diff --git "a/checkpoint-54664/trainer_state.json" "b/checkpoint-54664/trainer_state.json" new file mode 100644--- /dev/null +++ "b/checkpoint-54664/trainer_state.json" @@ -0,0 +1,328021 @@ +{ + "best_metric": 0.6931644082069397, + "best_model_checkpoint": "ai-generated-essay-detection-distilbert/checkpoint-54664", + "epoch": 2.0, + "eval_steps": 1, + "global_step": 54664, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0, + "learning_rate": 4e-09, + "loss": 0.6688, + "step": 1 + }, + { + "epoch": 0.0, + "learning_rate": 8e-09, + "loss": 0.7065, + "step": 2 + }, + { + "epoch": 0.0, + "learning_rate": 1.1999999999999998e-08, + "loss": 0.6933, + "step": 3 + }, + { + "epoch": 0.0, + "learning_rate": 1.6e-08, + "loss": 0.695, + "step": 4 + }, + { + "epoch": 0.0, + "learning_rate": 2e-08, + "loss": 0.6923, + "step": 5 + }, + { + "epoch": 0.0, + "learning_rate": 2.3999999999999997e-08, + "loss": 0.6937, + "step": 6 + }, + { + "epoch": 0.0, + "learning_rate": 2.8000000000000003e-08, + "loss": 0.6997, + "step": 7 + }, + { + "epoch": 0.0, + "learning_rate": 3.2e-08, + "loss": 0.7029, + "step": 8 + }, + { + "epoch": 0.0, + "learning_rate": 3.6e-08, + "loss": 0.7237, + "step": 9 + }, + { + "epoch": 0.0, + "learning_rate": 4e-08, + "loss": 0.6973, + "step": 10 + }, + { + "epoch": 0.0, + "learning_rate": 4.4e-08, + "loss": 0.6903, + "step": 11 + }, + { + "epoch": 0.0, + "learning_rate": 4.799999999999999e-08, + "loss": 0.6789, + "step": 12 + }, + { + "epoch": 0.0, + "learning_rate": 5.2e-08, + "loss": 0.6794, + "step": 13 + }, + { + "epoch": 0.0, + "learning_rate": 5.6000000000000005e-08, + "loss": 0.6639, + "step": 14 + }, + { + "epoch": 0.0, + "learning_rate": 6e-08, + "loss": 0.7066, + "step": 15 + }, + { + "epoch": 0.0, + "learning_rate": 6.4e-08, + "loss": 0.716, + "step": 16 + }, + { + "epoch": 0.0, + "learning_rate": 6.8e-08, + "loss": 0.6984, + "step": 17 + }, + { + "epoch": 0.0, + "learning_rate": 7.2e-08, + "loss": 0.7056, + "step": 18 + }, + { + "epoch": 0.0, + "learning_rate": 7.599999999999999e-08, + "loss": 0.699, + "step": 19 + }, + { + "epoch": 0.0, + "learning_rate": 8e-08, + "loss": 0.6885, + "step": 20 + }, + { + "epoch": 0.0, + "learning_rate": 8.4e-08, + "loss": 0.6909, + "step": 21 + }, + { + "epoch": 0.0, + "learning_rate": 8.8e-08, + "loss": 0.6674, + "step": 22 + }, + { + "epoch": 0.0, + "learning_rate": 9.2e-08, + "loss": 0.6828, + "step": 23 + }, + { + "epoch": 0.0, + "learning_rate": 9.599999999999999e-08, + "loss": 0.7223, + "step": 24 + }, + { + "epoch": 0.0, + "learning_rate": 1e-07, + "loss": 0.7241, + "step": 25 + }, + { + "epoch": 0.0, + "learning_rate": 1.04e-07, + "loss": 0.7019, + "step": 26 + }, + { + "epoch": 0.0, + "learning_rate": 1.08e-07, + "loss": 0.7126, + "step": 27 + }, + { + "epoch": 0.0, + "learning_rate": 1.1200000000000001e-07, + "loss": 0.6905, + "step": 28 + }, + { + "epoch": 0.0, + "learning_rate": 1.1599999999999999e-07, + "loss": 0.7259, + "step": 29 + }, + { + "epoch": 0.0, + "learning_rate": 1.2e-07, + "loss": 0.6671, + "step": 30 + }, + { + "epoch": 0.0, + "learning_rate": 1.24e-07, + "loss": 0.7156, + "step": 31 + }, + { + "epoch": 0.0, + "learning_rate": 1.28e-07, + "loss": 0.7185, + "step": 32 + }, + { + "epoch": 0.0, + "learning_rate": 1.32e-07, + "loss": 0.6903, + "step": 33 + }, + { + "epoch": 0.0, + "learning_rate": 1.36e-07, + "loss": 0.7013, + "step": 34 + }, + { + "epoch": 0.0, + "learning_rate": 1.3999999999999998e-07, + "loss": 0.6987, + "step": 35 + }, + { + "epoch": 0.0, + "learning_rate": 1.44e-07, + "loss": 0.7106, + "step": 36 + }, + { + "epoch": 0.0, + "learning_rate": 1.48e-07, + "loss": 0.7018, + "step": 37 + }, + { + "epoch": 0.0, + "learning_rate": 1.5199999999999998e-07, + "loss": 0.7087, + "step": 38 + }, + { + "epoch": 0.0, + "learning_rate": 1.56e-07, + "loss": 0.6921, + "step": 39 + }, + { + "epoch": 0.0, + "learning_rate": 1.6e-07, + "loss": 0.6954, + "step": 40 + }, + { + "epoch": 0.0, + "learning_rate": 1.6399999999999999e-07, + "loss": 0.7553, + "step": 41 + }, + { + "epoch": 0.0, + "learning_rate": 1.68e-07, + "loss": 0.6838, + "step": 42 + }, + { + "epoch": 0.0, + "learning_rate": 1.7199999999999998e-07, + "loss": 0.6962, + "step": 43 + }, + { + "epoch": 0.0, + "learning_rate": 1.76e-07, + "loss": 0.6996, + "step": 44 + }, + { + "epoch": 0.0, + "learning_rate": 1.8e-07, + "loss": 0.7276, + "step": 45 + }, + { + "epoch": 0.0, + "learning_rate": 1.84e-07, + "loss": 0.6757, + "step": 46 + }, + { + "epoch": 0.0, + "learning_rate": 1.88e-07, + "loss": 0.704, + "step": 47 + }, + { + "epoch": 0.0, + "learning_rate": 1.9199999999999997e-07, + "loss": 0.6948, + "step": 48 + }, + { + "epoch": 0.0, + "learning_rate": 1.9599999999999998e-07, + "loss": 0.7058, + "step": 49 + }, + { + "epoch": 0.0, + "learning_rate": 2e-07, + "loss": 0.6883, + "step": 50 + }, + { + "epoch": 0.0, + "learning_rate": 1.9999633793532792e-07, + "loss": 0.6378, + "step": 51 + }, + { + "epoch": 0.0, + "learning_rate": 1.9999267587065587e-07, + "loss": 0.6947, + "step": 52 + }, + { + "epoch": 0.0, + "learning_rate": 1.999890138059838e-07, + "loss": 0.6671, + "step": 53 + }, + { + "epoch": 0.0, + "learning_rate": 1.9998535174131175e-07, + "loss": 0.6923, + "step": 54 + }, + { + "epoch": 0.0, + "learning_rate": 1.9998168967663968e-07, + "loss": 0.6907, + "step": 55 + }, + { + "epoch": 0.0, + "learning_rate": 1.9997802761196763e-07, + "loss": 0.7284, + "step": 56 + }, + { + "epoch": 0.0, + "learning_rate": 1.9997436554729556e-07, + "loss": 0.7058, + "step": 57 + }, + { + "epoch": 0.0, + "learning_rate": 1.999707034826235e-07, + "loss": 0.6835, + "step": 58 + }, + { + "epoch": 0.0, + "learning_rate": 1.999670414179514e-07, + "loss": 0.6996, + "step": 59 + }, + { + "epoch": 0.0, + "learning_rate": 1.9996337935327936e-07, + "loss": 0.6791, + "step": 60 + }, + { + "epoch": 0.0, + "learning_rate": 1.999597172886073e-07, + "loss": 0.6963, + "step": 61 + }, + { + "epoch": 0.0, + "learning_rate": 1.9995605522393524e-07, + "loss": 0.7238, + "step": 62 + }, + { + "epoch": 0.0, + "learning_rate": 1.9995239315926317e-07, + "loss": 0.6843, + "step": 63 + }, + { + "epoch": 0.0, + "learning_rate": 1.9994873109459112e-07, + "loss": 0.6983, + "step": 64 + }, + { + "epoch": 0.0, + "learning_rate": 1.9994506902991905e-07, + "loss": 0.6965, + "step": 65 + }, + { + "epoch": 0.0, + "learning_rate": 1.99941406965247e-07, + "loss": 0.7119, + "step": 66 + }, + { + "epoch": 0.0, + "learning_rate": 1.9993774490057493e-07, + "loss": 0.7047, + "step": 67 + }, + { + "epoch": 0.0, + "learning_rate": 1.9993408283590288e-07, + "loss": 0.6912, + "step": 68 + }, + { + "epoch": 0.0, + "learning_rate": 1.999304207712308e-07, + "loss": 0.7027, + "step": 69 + }, + { + "epoch": 0.0, + "learning_rate": 1.9992675870655876e-07, + "loss": 0.7125, + "step": 70 + }, + { + "epoch": 0.0, + "learning_rate": 1.999230966418867e-07, + "loss": 0.7063, + "step": 71 + }, + { + "epoch": 0.0, + "learning_rate": 1.9991943457721462e-07, + "loss": 0.7368, + "step": 72 + }, + { + "epoch": 0.0, + "learning_rate": 1.9991577251254257e-07, + "loss": 0.6913, + "step": 73 + }, + { + "epoch": 0.0, + "learning_rate": 1.999121104478705e-07, + "loss": 0.6985, + "step": 74 + }, + { + "epoch": 0.0, + "learning_rate": 1.9990844838319845e-07, + "loss": 0.6935, + "step": 75 + }, + { + "epoch": 0.0, + "learning_rate": 1.9990478631852638e-07, + "loss": 0.7093, + "step": 76 + }, + { + "epoch": 0.0, + "learning_rate": 1.999011242538543e-07, + "loss": 0.7111, + "step": 77 + }, + { + "epoch": 0.0, + "learning_rate": 1.9989746218918223e-07, + "loss": 0.6686, + "step": 78 + }, + { + "epoch": 0.0, + "learning_rate": 1.9989380012451018e-07, + "loss": 0.6784, + "step": 79 + }, + { + "epoch": 0.0, + "learning_rate": 1.998901380598381e-07, + "loss": 0.6671, + "step": 80 + }, + { + "epoch": 0.0, + "learning_rate": 1.9988647599516606e-07, + "loss": 0.6857, + "step": 81 + }, + { + "epoch": 0.0, + "learning_rate": 1.99882813930494e-07, + "loss": 0.6892, + "step": 82 + }, + { + "epoch": 0.0, + "learning_rate": 1.9987915186582194e-07, + "loss": 0.7067, + "step": 83 + }, + { + "epoch": 0.0, + "learning_rate": 1.9987548980114987e-07, + "loss": 0.6948, + "step": 84 + }, + { + "epoch": 0.0, + "learning_rate": 1.9987182773647782e-07, + "loss": 0.705, + "step": 85 + }, + { + "epoch": 0.0, + "learning_rate": 1.9986816567180575e-07, + "loss": 0.6967, + "step": 86 + }, + { + "epoch": 0.0, + "learning_rate": 1.998645036071337e-07, + "loss": 0.7016, + "step": 87 + }, + { + "epoch": 0.0, + "learning_rate": 1.9986084154246163e-07, + "loss": 0.6545, + "step": 88 + }, + { + "epoch": 0.0, + "learning_rate": 1.9985717947778958e-07, + "loss": 0.7108, + "step": 89 + }, + { + "epoch": 0.0, + "learning_rate": 1.998535174131175e-07, + "loss": 0.6837, + "step": 90 + }, + { + "epoch": 0.0, + "learning_rate": 1.9984985534844546e-07, + "loss": 0.6837, + "step": 91 + }, + { + "epoch": 0.0, + "learning_rate": 1.998461932837734e-07, + "loss": 0.6943, + "step": 92 + }, + { + "epoch": 0.0, + "learning_rate": 1.9984253121910134e-07, + "loss": 0.6948, + "step": 93 + }, + { + "epoch": 0.0, + "learning_rate": 1.9983886915442927e-07, + "loss": 0.6961, + "step": 94 + }, + { + "epoch": 0.0, + "learning_rate": 1.998352070897572e-07, + "loss": 0.7099, + "step": 95 + }, + { + "epoch": 0.0, + "learning_rate": 1.9983154502508512e-07, + "loss": 0.7063, + "step": 96 + }, + { + "epoch": 0.0, + "learning_rate": 1.9982788296041307e-07, + "loss": 0.6861, + "step": 97 + }, + { + "epoch": 0.0, + "learning_rate": 1.99824220895741e-07, + "loss": 0.6889, + "step": 98 + }, + { + "epoch": 0.0, + "learning_rate": 1.9982055883106895e-07, + "loss": 0.7134, + "step": 99 + }, + { + "epoch": 0.0, + "learning_rate": 1.9981689676639688e-07, + "loss": 0.6916, + "step": 100 + }, + { + "epoch": 0.0, + "learning_rate": 1.998132347017248e-07, + "loss": 0.7365, + "step": 101 + }, + { + "epoch": 0.0, + "learning_rate": 1.9980957263705276e-07, + "loss": 0.7115, + "step": 102 + }, + { + "epoch": 0.0, + "learning_rate": 1.998059105723807e-07, + "loss": 0.7565, + "step": 103 + }, + { + "epoch": 0.0, + "learning_rate": 1.9980224850770864e-07, + "loss": 0.6782, + "step": 104 + }, + { + "epoch": 0.0, + "learning_rate": 1.9979858644303657e-07, + "loss": 0.6716, + "step": 105 + }, + { + "epoch": 0.0, + "learning_rate": 1.9979492437836452e-07, + "loss": 0.7011, + "step": 106 + }, + { + "epoch": 0.0, + "learning_rate": 1.9979126231369245e-07, + "loss": 0.6694, + "step": 107 + }, + { + "epoch": 0.0, + "learning_rate": 1.997876002490204e-07, + "loss": 0.689, + "step": 108 + }, + { + "epoch": 0.0, + "learning_rate": 1.9978393818434833e-07, + "loss": 0.6735, + "step": 109 + }, + { + "epoch": 0.0, + "learning_rate": 1.9978027611967628e-07, + "loss": 0.6773, + "step": 110 + }, + { + "epoch": 0.0, + "learning_rate": 1.997766140550042e-07, + "loss": 0.7096, + "step": 111 + }, + { + "epoch": 0.0, + "learning_rate": 1.9977295199033216e-07, + "loss": 0.6821, + "step": 112 + }, + { + "epoch": 0.0, + "learning_rate": 1.9976928992566006e-07, + "loss": 0.7095, + "step": 113 + }, + { + "epoch": 0.0, + "learning_rate": 1.9976562786098801e-07, + "loss": 0.6898, + "step": 114 + }, + { + "epoch": 0.0, + "learning_rate": 1.9976196579631594e-07, + "loss": 0.718, + "step": 115 + }, + { + "epoch": 0.0, + "learning_rate": 1.997583037316439e-07, + "loss": 0.6824, + "step": 116 + }, + { + "epoch": 0.0, + "learning_rate": 1.9975464166697182e-07, + "loss": 0.6815, + "step": 117 + }, + { + "epoch": 0.0, + "learning_rate": 1.9975097960229977e-07, + "loss": 0.7108, + "step": 118 + }, + { + "epoch": 0.0, + "learning_rate": 1.997473175376277e-07, + "loss": 0.6907, + "step": 119 + }, + { + "epoch": 0.0, + "learning_rate": 1.9974365547295565e-07, + "loss": 0.7102, + "step": 120 + }, + { + "epoch": 0.0, + "learning_rate": 1.9973999340828358e-07, + "loss": 0.7061, + "step": 121 + }, + { + "epoch": 0.0, + "learning_rate": 1.997363313436115e-07, + "loss": 0.6815, + "step": 122 + }, + { + "epoch": 0.0, + "learning_rate": 1.9973266927893946e-07, + "loss": 0.6976, + "step": 123 + }, + { + "epoch": 0.0, + "learning_rate": 1.997290072142674e-07, + "loss": 0.6953, + "step": 124 + }, + { + "epoch": 0.0, + "learning_rate": 1.9972534514959534e-07, + "loss": 0.6715, + "step": 125 + }, + { + "epoch": 0.0, + "learning_rate": 1.9972168308492327e-07, + "loss": 0.6852, + "step": 126 + }, + { + "epoch": 0.0, + "learning_rate": 1.9971802102025122e-07, + "loss": 0.6759, + "step": 127 + }, + { + "epoch": 0.0, + "learning_rate": 1.9971435895557915e-07, + "loss": 0.6822, + "step": 128 + }, + { + "epoch": 0.0, + "learning_rate": 1.997106968909071e-07, + "loss": 0.6862, + "step": 129 + }, + { + "epoch": 0.0, + "learning_rate": 1.9970703482623503e-07, + "loss": 0.6925, + "step": 130 + }, + { + "epoch": 0.0, + "learning_rate": 1.9970337276156295e-07, + "loss": 0.6833, + "step": 131 + }, + { + "epoch": 0.0, + "learning_rate": 1.9969971069689088e-07, + "loss": 0.6862, + "step": 132 + }, + { + "epoch": 0.0, + "learning_rate": 1.9969604863221883e-07, + "loss": 0.7003, + "step": 133 + }, + { + "epoch": 0.0, + "learning_rate": 1.9969238656754676e-07, + "loss": 0.6887, + "step": 134 + }, + { + "epoch": 0.0, + "learning_rate": 1.996887245028747e-07, + "loss": 0.7022, + "step": 135 + }, + { + "epoch": 0.0, + "learning_rate": 1.9968506243820264e-07, + "loss": 0.6983, + "step": 136 + }, + { + "epoch": 0.01, + "learning_rate": 1.996814003735306e-07, + "loss": 0.7253, + "step": 137 + }, + { + "epoch": 0.01, + "learning_rate": 1.9967773830885852e-07, + "loss": 0.688, + "step": 138 + }, + { + "epoch": 0.01, + "learning_rate": 1.9967407624418647e-07, + "loss": 0.6621, + "step": 139 + }, + { + "epoch": 0.01, + "learning_rate": 1.996704141795144e-07, + "loss": 0.7006, + "step": 140 + }, + { + "epoch": 0.01, + "learning_rate": 1.9966675211484235e-07, + "loss": 0.6876, + "step": 141 + }, + { + "epoch": 0.01, + "learning_rate": 1.9966309005017028e-07, + "loss": 0.6994, + "step": 142 + }, + { + "epoch": 0.01, + "learning_rate": 1.996594279854982e-07, + "loss": 0.6908, + "step": 143 + }, + { + "epoch": 0.01, + "learning_rate": 1.9965576592082616e-07, + "loss": 0.6966, + "step": 144 + }, + { + "epoch": 0.01, + "learning_rate": 1.9965210385615409e-07, + "loss": 0.6732, + "step": 145 + }, + { + "epoch": 0.01, + "learning_rate": 1.9964844179148204e-07, + "loss": 0.6807, + "step": 146 + }, + { + "epoch": 0.01, + "learning_rate": 1.9964477972680997e-07, + "loss": 0.6962, + "step": 147 + }, + { + "epoch": 0.01, + "learning_rate": 1.9964111766213792e-07, + "loss": 0.6754, + "step": 148 + }, + { + "epoch": 0.01, + "learning_rate": 1.9963745559746582e-07, + "loss": 0.6982, + "step": 149 + }, + { + "epoch": 0.01, + "learning_rate": 1.9963379353279377e-07, + "loss": 0.6682, + "step": 150 + }, + { + "epoch": 0.01, + "learning_rate": 1.996301314681217e-07, + "loss": 0.7063, + "step": 151 + }, + { + "epoch": 0.01, + "learning_rate": 1.9962646940344965e-07, + "loss": 0.6945, + "step": 152 + }, + { + "epoch": 0.01, + "learning_rate": 1.9962280733877758e-07, + "loss": 0.6649, + "step": 153 + }, + { + "epoch": 0.01, + "learning_rate": 1.9961914527410553e-07, + "loss": 0.6929, + "step": 154 + }, + { + "epoch": 0.01, + "learning_rate": 1.9961548320943346e-07, + "loss": 0.6894, + "step": 155 + }, + { + "epoch": 0.01, + "learning_rate": 1.996118211447614e-07, + "loss": 0.6989, + "step": 156 + }, + { + "epoch": 0.01, + "learning_rate": 1.9960815908008934e-07, + "loss": 0.727, + "step": 157 + }, + { + "epoch": 0.01, + "learning_rate": 1.996044970154173e-07, + "loss": 0.6724, + "step": 158 + }, + { + "epoch": 0.01, + "learning_rate": 1.9960083495074522e-07, + "loss": 0.7327, + "step": 159 + }, + { + "epoch": 0.01, + "learning_rate": 1.9959717288607317e-07, + "loss": 0.6938, + "step": 160 + }, + { + "epoch": 0.01, + "learning_rate": 1.995935108214011e-07, + "loss": 0.7087, + "step": 161 + }, + { + "epoch": 0.01, + "learning_rate": 1.9958984875672905e-07, + "loss": 0.682, + "step": 162 + }, + { + "epoch": 0.01, + "learning_rate": 1.9958618669205698e-07, + "loss": 0.6846, + "step": 163 + }, + { + "epoch": 0.01, + "learning_rate": 1.9958252462738493e-07, + "loss": 0.6918, + "step": 164 + }, + { + "epoch": 0.01, + "learning_rate": 1.9957886256271286e-07, + "loss": 0.6773, + "step": 165 + }, + { + "epoch": 0.01, + "learning_rate": 1.9957520049804078e-07, + "loss": 0.7106, + "step": 166 + }, + { + "epoch": 0.01, + "learning_rate": 1.995715384333687e-07, + "loss": 0.6693, + "step": 167 + }, + { + "epoch": 0.01, + "learning_rate": 1.9956787636869666e-07, + "loss": 0.6617, + "step": 168 + }, + { + "epoch": 0.01, + "learning_rate": 1.995642143040246e-07, + "loss": 0.6967, + "step": 169 + }, + { + "epoch": 0.01, + "learning_rate": 1.9956055223935252e-07, + "loss": 0.7265, + "step": 170 + }, + { + "epoch": 0.01, + "learning_rate": 1.9955689017468047e-07, + "loss": 0.7073, + "step": 171 + }, + { + "epoch": 0.01, + "learning_rate": 1.995532281100084e-07, + "loss": 0.7019, + "step": 172 + }, + { + "epoch": 0.01, + "learning_rate": 1.9954956604533635e-07, + "loss": 0.7267, + "step": 173 + }, + { + "epoch": 0.01, + "learning_rate": 1.9954590398066428e-07, + "loss": 0.7002, + "step": 174 + }, + { + "epoch": 0.01, + "learning_rate": 1.9954224191599223e-07, + "loss": 0.6909, + "step": 175 + }, + { + "epoch": 0.01, + "learning_rate": 1.9953857985132016e-07, + "loss": 0.6629, + "step": 176 + }, + { + "epoch": 0.01, + "learning_rate": 1.995349177866481e-07, + "loss": 0.6999, + "step": 177 + }, + { + "epoch": 0.01, + "learning_rate": 1.9953125572197604e-07, + "loss": 0.695, + "step": 178 + }, + { + "epoch": 0.01, + "learning_rate": 1.99527593657304e-07, + "loss": 0.6881, + "step": 179 + }, + { + "epoch": 0.01, + "learning_rate": 1.9952393159263192e-07, + "loss": 0.7258, + "step": 180 + }, + { + "epoch": 0.01, + "learning_rate": 1.9952026952795987e-07, + "loss": 0.6901, + "step": 181 + }, + { + "epoch": 0.01, + "learning_rate": 1.995166074632878e-07, + "loss": 0.6815, + "step": 182 + }, + { + "epoch": 0.01, + "learning_rate": 1.9951294539861575e-07, + "loss": 0.6778, + "step": 183 + }, + { + "epoch": 0.01, + "learning_rate": 1.9950928333394368e-07, + "loss": 0.6797, + "step": 184 + }, + { + "epoch": 0.01, + "learning_rate": 1.995056212692716e-07, + "loss": 0.6973, + "step": 185 + }, + { + "epoch": 0.01, + "learning_rate": 1.9950195920459953e-07, + "loss": 0.7049, + "step": 186 + }, + { + "epoch": 0.01, + "learning_rate": 1.9949829713992748e-07, + "loss": 0.6924, + "step": 187 + }, + { + "epoch": 0.01, + "learning_rate": 1.994946350752554e-07, + "loss": 0.7003, + "step": 188 + }, + { + "epoch": 0.01, + "learning_rate": 1.9949097301058336e-07, + "loss": 0.6904, + "step": 189 + }, + { + "epoch": 0.01, + "learning_rate": 1.994873109459113e-07, + "loss": 0.6847, + "step": 190 + }, + { + "epoch": 0.01, + "learning_rate": 1.9948364888123924e-07, + "loss": 0.7156, + "step": 191 + }, + { + "epoch": 0.01, + "learning_rate": 1.9947998681656717e-07, + "loss": 0.6822, + "step": 192 + }, + { + "epoch": 0.01, + "learning_rate": 1.994763247518951e-07, + "loss": 0.6749, + "step": 193 + }, + { + "epoch": 0.01, + "learning_rate": 1.9947266268722305e-07, + "loss": 0.6788, + "step": 194 + }, + { + "epoch": 0.01, + "learning_rate": 1.9946900062255098e-07, + "loss": 0.6977, + "step": 195 + }, + { + "epoch": 0.01, + "learning_rate": 1.9946533855787893e-07, + "loss": 0.6987, + "step": 196 + }, + { + "epoch": 0.01, + "learning_rate": 1.9946167649320686e-07, + "loss": 0.701, + "step": 197 + }, + { + "epoch": 0.01, + "learning_rate": 1.994580144285348e-07, + "loss": 0.6956, + "step": 198 + }, + { + "epoch": 0.01, + "learning_rate": 1.9945435236386274e-07, + "loss": 0.6837, + "step": 199 + }, + { + "epoch": 0.01, + "learning_rate": 1.994506902991907e-07, + "loss": 0.7023, + "step": 200 + }, + { + "epoch": 0.01, + "learning_rate": 1.9944702823451862e-07, + "loss": 0.6863, + "step": 201 + }, + { + "epoch": 0.01, + "learning_rate": 1.9944336616984657e-07, + "loss": 0.6856, + "step": 202 + }, + { + "epoch": 0.01, + "learning_rate": 1.9943970410517447e-07, + "loss": 0.6676, + "step": 203 + }, + { + "epoch": 0.01, + "learning_rate": 1.9943604204050242e-07, + "loss": 0.6825, + "step": 204 + }, + { + "epoch": 0.01, + "learning_rate": 1.9943237997583035e-07, + "loss": 0.6951, + "step": 205 + }, + { + "epoch": 0.01, + "learning_rate": 1.994287179111583e-07, + "loss": 0.7025, + "step": 206 + }, + { + "epoch": 0.01, + "learning_rate": 1.9942505584648623e-07, + "loss": 0.7095, + "step": 207 + }, + { + "epoch": 0.01, + "learning_rate": 1.9942139378181418e-07, + "loss": 0.7087, + "step": 208 + }, + { + "epoch": 0.01, + "learning_rate": 1.994177317171421e-07, + "loss": 0.6857, + "step": 209 + }, + { + "epoch": 0.01, + "learning_rate": 1.9941406965247006e-07, + "loss": 0.6975, + "step": 210 + }, + { + "epoch": 0.01, + "learning_rate": 1.99410407587798e-07, + "loss": 0.6865, + "step": 211 + }, + { + "epoch": 0.01, + "learning_rate": 1.9940674552312594e-07, + "loss": 0.7212, + "step": 212 + }, + { + "epoch": 0.01, + "learning_rate": 1.9940308345845387e-07, + "loss": 0.7024, + "step": 213 + }, + { + "epoch": 0.01, + "learning_rate": 1.993994213937818e-07, + "loss": 0.6938, + "step": 214 + }, + { + "epoch": 0.01, + "learning_rate": 1.9939575932910975e-07, + "loss": 0.6875, + "step": 215 + }, + { + "epoch": 0.01, + "learning_rate": 1.9939209726443767e-07, + "loss": 0.7281, + "step": 216 + }, + { + "epoch": 0.01, + "learning_rate": 1.9938843519976563e-07, + "loss": 0.6861, + "step": 217 + }, + { + "epoch": 0.01, + "learning_rate": 1.9938477313509355e-07, + "loss": 0.6909, + "step": 218 + }, + { + "epoch": 0.01, + "learning_rate": 1.993811110704215e-07, + "loss": 0.6981, + "step": 219 + }, + { + "epoch": 0.01, + "learning_rate": 1.9937744900574943e-07, + "loss": 0.6952, + "step": 220 + }, + { + "epoch": 0.01, + "learning_rate": 1.9937378694107736e-07, + "loss": 0.6772, + "step": 221 + }, + { + "epoch": 0.01, + "learning_rate": 1.993701248764053e-07, + "loss": 0.7337, + "step": 222 + }, + { + "epoch": 0.01, + "learning_rate": 1.9936646281173324e-07, + "loss": 0.6948, + "step": 223 + }, + { + "epoch": 0.01, + "learning_rate": 1.9936280074706117e-07, + "loss": 0.6945, + "step": 224 + }, + { + "epoch": 0.01, + "learning_rate": 1.9935913868238912e-07, + "loss": 0.6688, + "step": 225 + }, + { + "epoch": 0.01, + "learning_rate": 1.9935547661771705e-07, + "loss": 0.6839, + "step": 226 + }, + { + "epoch": 0.01, + "learning_rate": 1.99351814553045e-07, + "loss": 0.7039, + "step": 227 + }, + { + "epoch": 0.01, + "learning_rate": 1.9934815248837293e-07, + "loss": 0.693, + "step": 228 + }, + { + "epoch": 0.01, + "learning_rate": 1.9934449042370088e-07, + "loss": 0.6867, + "step": 229 + }, + { + "epoch": 0.01, + "learning_rate": 1.993408283590288e-07, + "loss": 0.689, + "step": 230 + }, + { + "epoch": 0.01, + "learning_rate": 1.9933716629435676e-07, + "loss": 0.6972, + "step": 231 + }, + { + "epoch": 0.01, + "learning_rate": 1.993335042296847e-07, + "loss": 0.6889, + "step": 232 + }, + { + "epoch": 0.01, + "learning_rate": 1.9932984216501264e-07, + "loss": 0.7066, + "step": 233 + }, + { + "epoch": 0.01, + "learning_rate": 1.9932618010034057e-07, + "loss": 0.7131, + "step": 234 + }, + { + "epoch": 0.01, + "learning_rate": 1.9932251803566852e-07, + "loss": 0.6792, + "step": 235 + }, + { + "epoch": 0.01, + "learning_rate": 1.9931885597099645e-07, + "loss": 0.7137, + "step": 236 + }, + { + "epoch": 0.01, + "learning_rate": 1.9931519390632437e-07, + "loss": 0.6844, + "step": 237 + }, + { + "epoch": 0.01, + "learning_rate": 1.9931153184165233e-07, + "loss": 0.696, + "step": 238 + }, + { + "epoch": 0.01, + "learning_rate": 1.9930786977698025e-07, + "loss": 0.698, + "step": 239 + }, + { + "epoch": 0.01, + "learning_rate": 1.9930420771230818e-07, + "loss": 0.694, + "step": 240 + }, + { + "epoch": 0.01, + "learning_rate": 1.993005456476361e-07, + "loss": 0.7072, + "step": 241 + }, + { + "epoch": 0.01, + "learning_rate": 1.9929688358296406e-07, + "loss": 0.6884, + "step": 242 + }, + { + "epoch": 0.01, + "learning_rate": 1.99293221518292e-07, + "loss": 0.6732, + "step": 243 + }, + { + "epoch": 0.01, + "learning_rate": 1.9928955945361994e-07, + "loss": 0.703, + "step": 244 + }, + { + "epoch": 0.01, + "learning_rate": 1.9928589738894787e-07, + "loss": 0.6977, + "step": 245 + }, + { + "epoch": 0.01, + "learning_rate": 1.9928223532427582e-07, + "loss": 0.7004, + "step": 246 + }, + { + "epoch": 0.01, + "learning_rate": 1.9927857325960375e-07, + "loss": 0.7022, + "step": 247 + }, + { + "epoch": 0.01, + "learning_rate": 1.992749111949317e-07, + "loss": 0.6899, + "step": 248 + }, + { + "epoch": 0.01, + "learning_rate": 1.9927124913025963e-07, + "loss": 0.6879, + "step": 249 + }, + { + "epoch": 0.01, + "learning_rate": 1.9926758706558758e-07, + "loss": 0.6849, + "step": 250 + }, + { + "epoch": 0.01, + "learning_rate": 1.992639250009155e-07, + "loss": 0.7004, + "step": 251 + }, + { + "epoch": 0.01, + "learning_rate": 1.9926026293624346e-07, + "loss": 0.7182, + "step": 252 + }, + { + "epoch": 0.01, + "learning_rate": 1.9925660087157139e-07, + "loss": 0.6807, + "step": 253 + }, + { + "epoch": 0.01, + "learning_rate": 1.9925293880689934e-07, + "loss": 0.7049, + "step": 254 + }, + { + "epoch": 0.01, + "learning_rate": 1.9924927674222727e-07, + "loss": 0.6939, + "step": 255 + }, + { + "epoch": 0.01, + "learning_rate": 1.9924561467755522e-07, + "loss": 0.7072, + "step": 256 + }, + { + "epoch": 0.01, + "learning_rate": 1.9924195261288312e-07, + "loss": 0.707, + "step": 257 + }, + { + "epoch": 0.01, + "learning_rate": 1.9923829054821107e-07, + "loss": 0.6755, + "step": 258 + }, + { + "epoch": 0.01, + "learning_rate": 1.99234628483539e-07, + "loss": 0.7193, + "step": 259 + }, + { + "epoch": 0.01, + "learning_rate": 1.9923096641886695e-07, + "loss": 0.6921, + "step": 260 + }, + { + "epoch": 0.01, + "learning_rate": 1.9922730435419488e-07, + "loss": 0.7055, + "step": 261 + }, + { + "epoch": 0.01, + "learning_rate": 1.9922364228952283e-07, + "loss": 0.6853, + "step": 262 + }, + { + "epoch": 0.01, + "learning_rate": 1.9921998022485076e-07, + "loss": 0.7158, + "step": 263 + }, + { + "epoch": 0.01, + "learning_rate": 1.9921631816017869e-07, + "loss": 0.6965, + "step": 264 + }, + { + "epoch": 0.01, + "learning_rate": 1.9921265609550664e-07, + "loss": 0.6748, + "step": 265 + }, + { + "epoch": 0.01, + "learning_rate": 1.9920899403083457e-07, + "loss": 0.7066, + "step": 266 + }, + { + "epoch": 0.01, + "learning_rate": 1.9920533196616252e-07, + "loss": 0.6937, + "step": 267 + }, + { + "epoch": 0.01, + "learning_rate": 1.9920166990149045e-07, + "loss": 0.7066, + "step": 268 + }, + { + "epoch": 0.01, + "learning_rate": 1.991980078368184e-07, + "loss": 0.6999, + "step": 269 + }, + { + "epoch": 0.01, + "learning_rate": 1.9919434577214633e-07, + "loss": 0.6785, + "step": 270 + }, + { + "epoch": 0.01, + "learning_rate": 1.9919068370747428e-07, + "loss": 0.6798, + "step": 271 + }, + { + "epoch": 0.01, + "learning_rate": 1.991870216428022e-07, + "loss": 0.7173, + "step": 272 + }, + { + "epoch": 0.01, + "learning_rate": 1.9918335957813016e-07, + "loss": 0.6915, + "step": 273 + }, + { + "epoch": 0.01, + "learning_rate": 1.9917969751345808e-07, + "loss": 0.6974, + "step": 274 + }, + { + "epoch": 0.01, + "learning_rate": 1.99176035448786e-07, + "loss": 0.6976, + "step": 275 + }, + { + "epoch": 0.01, + "learning_rate": 1.9917237338411394e-07, + "loss": 0.676, + "step": 276 + }, + { + "epoch": 0.01, + "learning_rate": 1.991687113194419e-07, + "loss": 0.6846, + "step": 277 + }, + { + "epoch": 0.01, + "learning_rate": 1.9916504925476982e-07, + "loss": 0.6983, + "step": 278 + }, + { + "epoch": 0.01, + "learning_rate": 1.9916138719009777e-07, + "loss": 0.6937, + "step": 279 + }, + { + "epoch": 0.01, + "learning_rate": 1.991577251254257e-07, + "loss": 0.6932, + "step": 280 + }, + { + "epoch": 0.01, + "learning_rate": 1.9915406306075365e-07, + "loss": 0.7051, + "step": 281 + }, + { + "epoch": 0.01, + "learning_rate": 1.9915040099608158e-07, + "loss": 0.6932, + "step": 282 + }, + { + "epoch": 0.01, + "learning_rate": 1.9914673893140953e-07, + "loss": 0.7016, + "step": 283 + }, + { + "epoch": 0.01, + "learning_rate": 1.9914307686673746e-07, + "loss": 0.7004, + "step": 284 + }, + { + "epoch": 0.01, + "learning_rate": 1.9913941480206538e-07, + "loss": 0.6679, + "step": 285 + }, + { + "epoch": 0.01, + "learning_rate": 1.9913575273739334e-07, + "loss": 0.7074, + "step": 286 + }, + { + "epoch": 0.01, + "learning_rate": 1.9913209067272126e-07, + "loss": 0.6972, + "step": 287 + }, + { + "epoch": 0.01, + "learning_rate": 1.9912842860804922e-07, + "loss": 0.6918, + "step": 288 + }, + { + "epoch": 0.01, + "learning_rate": 1.9912476654337714e-07, + "loss": 0.6872, + "step": 289 + }, + { + "epoch": 0.01, + "learning_rate": 1.991211044787051e-07, + "loss": 0.6847, + "step": 290 + }, + { + "epoch": 0.01, + "learning_rate": 1.9911744241403302e-07, + "loss": 0.69, + "step": 291 + }, + { + "epoch": 0.01, + "learning_rate": 1.9911378034936098e-07, + "loss": 0.6937, + "step": 292 + }, + { + "epoch": 0.01, + "learning_rate": 1.9911011828468888e-07, + "loss": 0.6858, + "step": 293 + }, + { + "epoch": 0.01, + "learning_rate": 1.9910645622001683e-07, + "loss": 0.6805, + "step": 294 + }, + { + "epoch": 0.01, + "learning_rate": 1.9910279415534476e-07, + "loss": 0.7215, + "step": 295 + }, + { + "epoch": 0.01, + "learning_rate": 1.990991320906727e-07, + "loss": 0.7081, + "step": 296 + }, + { + "epoch": 0.01, + "learning_rate": 1.9909547002600064e-07, + "loss": 0.6893, + "step": 297 + }, + { + "epoch": 0.01, + "learning_rate": 1.990918079613286e-07, + "loss": 0.6832, + "step": 298 + }, + { + "epoch": 0.01, + "learning_rate": 1.9908814589665652e-07, + "loss": 0.6965, + "step": 299 + }, + { + "epoch": 0.01, + "learning_rate": 1.9908448383198447e-07, + "loss": 0.6927, + "step": 300 + }, + { + "epoch": 0.01, + "learning_rate": 1.990808217673124e-07, + "loss": 0.7167, + "step": 301 + }, + { + "epoch": 0.01, + "learning_rate": 1.9907715970264035e-07, + "loss": 0.6642, + "step": 302 + }, + { + "epoch": 0.01, + "learning_rate": 1.9907349763796828e-07, + "loss": 0.7041, + "step": 303 + }, + { + "epoch": 0.01, + "learning_rate": 1.9906983557329623e-07, + "loss": 0.6975, + "step": 304 + }, + { + "epoch": 0.01, + "learning_rate": 1.9906617350862416e-07, + "loss": 0.7099, + "step": 305 + }, + { + "epoch": 0.01, + "learning_rate": 1.990625114439521e-07, + "loss": 0.6864, + "step": 306 + }, + { + "epoch": 0.01, + "learning_rate": 1.9905884937928004e-07, + "loss": 0.6854, + "step": 307 + }, + { + "epoch": 0.01, + "learning_rate": 1.9905518731460796e-07, + "loss": 0.6688, + "step": 308 + }, + { + "epoch": 0.01, + "learning_rate": 1.9905152524993592e-07, + "loss": 0.6879, + "step": 309 + }, + { + "epoch": 0.01, + "learning_rate": 1.9904786318526384e-07, + "loss": 0.693, + "step": 310 + }, + { + "epoch": 0.01, + "learning_rate": 1.9904420112059177e-07, + "loss": 0.7004, + "step": 311 + }, + { + "epoch": 0.01, + "learning_rate": 1.990405390559197e-07, + "loss": 0.6972, + "step": 312 + }, + { + "epoch": 0.01, + "learning_rate": 1.9903687699124765e-07, + "loss": 0.6867, + "step": 313 + }, + { + "epoch": 0.01, + "learning_rate": 1.9903321492657558e-07, + "loss": 0.6984, + "step": 314 + }, + { + "epoch": 0.01, + "learning_rate": 1.9902955286190353e-07, + "loss": 0.6612, + "step": 315 + }, + { + "epoch": 0.01, + "learning_rate": 1.9902589079723146e-07, + "loss": 0.6683, + "step": 316 + }, + { + "epoch": 0.01, + "learning_rate": 1.990222287325594e-07, + "loss": 0.6942, + "step": 317 + }, + { + "epoch": 0.01, + "learning_rate": 1.9901856666788734e-07, + "loss": 0.7138, + "step": 318 + }, + { + "epoch": 0.01, + "learning_rate": 1.990149046032153e-07, + "loss": 0.6985, + "step": 319 + }, + { + "epoch": 0.01, + "learning_rate": 1.9901124253854322e-07, + "loss": 0.6879, + "step": 320 + }, + { + "epoch": 0.01, + "learning_rate": 1.9900758047387117e-07, + "loss": 0.6874, + "step": 321 + }, + { + "epoch": 0.01, + "learning_rate": 1.990039184091991e-07, + "loss": 0.6879, + "step": 322 + }, + { + "epoch": 0.01, + "learning_rate": 1.9900025634452705e-07, + "loss": 0.6824, + "step": 323 + }, + { + "epoch": 0.01, + "learning_rate": 1.9899659427985498e-07, + "loss": 0.7075, + "step": 324 + }, + { + "epoch": 0.01, + "learning_rate": 1.9899293221518293e-07, + "loss": 0.703, + "step": 325 + }, + { + "epoch": 0.01, + "learning_rate": 1.9898927015051086e-07, + "loss": 0.6838, + "step": 326 + }, + { + "epoch": 0.01, + "learning_rate": 1.989856080858388e-07, + "loss": 0.7149, + "step": 327 + }, + { + "epoch": 0.01, + "learning_rate": 1.9898194602116674e-07, + "loss": 0.7046, + "step": 328 + }, + { + "epoch": 0.01, + "learning_rate": 1.9897828395649466e-07, + "loss": 0.7239, + "step": 329 + }, + { + "epoch": 0.01, + "learning_rate": 1.989746218918226e-07, + "loss": 0.7144, + "step": 330 + }, + { + "epoch": 0.01, + "learning_rate": 1.9897095982715054e-07, + "loss": 0.6866, + "step": 331 + }, + { + "epoch": 0.01, + "learning_rate": 1.9896729776247847e-07, + "loss": 0.7107, + "step": 332 + }, + { + "epoch": 0.01, + "learning_rate": 1.9896363569780642e-07, + "loss": 0.7021, + "step": 333 + }, + { + "epoch": 0.01, + "learning_rate": 1.9895997363313435e-07, + "loss": 0.6883, + "step": 334 + }, + { + "epoch": 0.01, + "learning_rate": 1.9895631156846228e-07, + "loss": 0.6744, + "step": 335 + }, + { + "epoch": 0.01, + "learning_rate": 1.9895264950379023e-07, + "loss": 0.7034, + "step": 336 + }, + { + "epoch": 0.01, + "learning_rate": 1.9894898743911815e-07, + "loss": 0.704, + "step": 337 + }, + { + "epoch": 0.01, + "learning_rate": 1.989453253744461e-07, + "loss": 0.7161, + "step": 338 + }, + { + "epoch": 0.01, + "learning_rate": 1.9894166330977403e-07, + "loss": 0.6966, + "step": 339 + }, + { + "epoch": 0.01, + "learning_rate": 1.98938001245102e-07, + "loss": 0.7071, + "step": 340 + }, + { + "epoch": 0.01, + "learning_rate": 1.9893433918042991e-07, + "loss": 0.6827, + "step": 341 + }, + { + "epoch": 0.01, + "learning_rate": 1.9893067711575787e-07, + "loss": 0.6978, + "step": 342 + }, + { + "epoch": 0.01, + "learning_rate": 1.989270150510858e-07, + "loss": 0.6896, + "step": 343 + }, + { + "epoch": 0.01, + "learning_rate": 1.9892335298641375e-07, + "loss": 0.6586, + "step": 344 + }, + { + "epoch": 0.01, + "learning_rate": 1.9891969092174167e-07, + "loss": 0.6909, + "step": 345 + }, + { + "epoch": 0.01, + "learning_rate": 1.9891602885706963e-07, + "loss": 0.6647, + "step": 346 + }, + { + "epoch": 0.01, + "learning_rate": 1.9891236679239753e-07, + "loss": 0.6758, + "step": 347 + }, + { + "epoch": 0.01, + "learning_rate": 1.9890870472772548e-07, + "loss": 0.7086, + "step": 348 + }, + { + "epoch": 0.01, + "learning_rate": 1.989050426630534e-07, + "loss": 0.6978, + "step": 349 + }, + { + "epoch": 0.01, + "learning_rate": 1.9890138059838136e-07, + "loss": 0.6863, + "step": 350 + }, + { + "epoch": 0.01, + "learning_rate": 1.988977185337093e-07, + "loss": 0.6898, + "step": 351 + }, + { + "epoch": 0.01, + "learning_rate": 1.9889405646903724e-07, + "loss": 0.7186, + "step": 352 + }, + { + "epoch": 0.01, + "learning_rate": 1.9889039440436517e-07, + "loss": 0.705, + "step": 353 + }, + { + "epoch": 0.01, + "learning_rate": 1.9888673233969312e-07, + "loss": 0.6864, + "step": 354 + }, + { + "epoch": 0.01, + "learning_rate": 1.9888307027502105e-07, + "loss": 0.6918, + "step": 355 + }, + { + "epoch": 0.01, + "learning_rate": 1.9887940821034897e-07, + "loss": 0.6874, + "step": 356 + }, + { + "epoch": 0.01, + "learning_rate": 1.9887574614567693e-07, + "loss": 0.6812, + "step": 357 + }, + { + "epoch": 0.01, + "learning_rate": 1.9887208408100485e-07, + "loss": 0.6943, + "step": 358 + }, + { + "epoch": 0.01, + "learning_rate": 1.988684220163328e-07, + "loss": 0.7206, + "step": 359 + }, + { + "epoch": 0.01, + "learning_rate": 1.9886475995166073e-07, + "loss": 0.6581, + "step": 360 + }, + { + "epoch": 0.01, + "learning_rate": 1.9886109788698869e-07, + "loss": 0.6924, + "step": 361 + }, + { + "epoch": 0.01, + "learning_rate": 1.9885743582231661e-07, + "loss": 0.7122, + "step": 362 + }, + { + "epoch": 0.01, + "learning_rate": 1.9885377375764457e-07, + "loss": 0.7125, + "step": 363 + }, + { + "epoch": 0.01, + "learning_rate": 1.988501116929725e-07, + "loss": 0.6983, + "step": 364 + }, + { + "epoch": 0.01, + "learning_rate": 1.9884644962830042e-07, + "loss": 0.7136, + "step": 365 + }, + { + "epoch": 0.01, + "learning_rate": 1.9884278756362835e-07, + "loss": 0.7201, + "step": 366 + }, + { + "epoch": 0.01, + "learning_rate": 1.988391254989563e-07, + "loss": 0.6849, + "step": 367 + }, + { + "epoch": 0.01, + "learning_rate": 1.9883546343428423e-07, + "loss": 0.6808, + "step": 368 + }, + { + "epoch": 0.01, + "learning_rate": 1.9883180136961218e-07, + "loss": 0.6753, + "step": 369 + }, + { + "epoch": 0.01, + "learning_rate": 1.988281393049401e-07, + "loss": 0.7143, + "step": 370 + }, + { + "epoch": 0.01, + "learning_rate": 1.9882447724026806e-07, + "loss": 0.6726, + "step": 371 + }, + { + "epoch": 0.01, + "learning_rate": 1.9882081517559599e-07, + "loss": 0.7116, + "step": 372 + }, + { + "epoch": 0.01, + "learning_rate": 1.9881715311092394e-07, + "loss": 0.6964, + "step": 373 + }, + { + "epoch": 0.01, + "learning_rate": 1.9881349104625187e-07, + "loss": 0.7191, + "step": 374 + }, + { + "epoch": 0.01, + "learning_rate": 1.9880982898157982e-07, + "loss": 0.7078, + "step": 375 + }, + { + "epoch": 0.01, + "learning_rate": 1.9880616691690775e-07, + "loss": 0.6744, + "step": 376 + }, + { + "epoch": 0.01, + "learning_rate": 1.988025048522357e-07, + "loss": 0.6864, + "step": 377 + }, + { + "epoch": 0.01, + "learning_rate": 1.9879884278756363e-07, + "loss": 0.6958, + "step": 378 + }, + { + "epoch": 0.01, + "learning_rate": 1.9879518072289155e-07, + "loss": 0.6969, + "step": 379 + }, + { + "epoch": 0.01, + "learning_rate": 1.987915186582195e-07, + "loss": 0.7095, + "step": 380 + }, + { + "epoch": 0.01, + "learning_rate": 1.9878785659354743e-07, + "loss": 0.695, + "step": 381 + }, + { + "epoch": 0.01, + "learning_rate": 1.9878419452887539e-07, + "loss": 0.7043, + "step": 382 + }, + { + "epoch": 0.01, + "learning_rate": 1.9878053246420329e-07, + "loss": 0.6863, + "step": 383 + }, + { + "epoch": 0.01, + "learning_rate": 1.9877687039953124e-07, + "loss": 0.6609, + "step": 384 + }, + { + "epoch": 0.01, + "learning_rate": 1.9877320833485917e-07, + "loss": 0.6984, + "step": 385 + }, + { + "epoch": 0.01, + "learning_rate": 1.9876954627018712e-07, + "loss": 0.7088, + "step": 386 + }, + { + "epoch": 0.01, + "learning_rate": 1.9876588420551505e-07, + "loss": 0.7055, + "step": 387 + }, + { + "epoch": 0.01, + "learning_rate": 1.98762222140843e-07, + "loss": 0.6868, + "step": 388 + }, + { + "epoch": 0.01, + "learning_rate": 1.9875856007617093e-07, + "loss": 0.7008, + "step": 389 + }, + { + "epoch": 0.01, + "learning_rate": 1.9875489801149888e-07, + "loss": 0.6952, + "step": 390 + }, + { + "epoch": 0.01, + "learning_rate": 1.987512359468268e-07, + "loss": 0.7014, + "step": 391 + }, + { + "epoch": 0.01, + "learning_rate": 1.9874757388215476e-07, + "loss": 0.6972, + "step": 392 + }, + { + "epoch": 0.01, + "learning_rate": 1.9874391181748269e-07, + "loss": 0.6788, + "step": 393 + }, + { + "epoch": 0.01, + "learning_rate": 1.9874024975281064e-07, + "loss": 0.6825, + "step": 394 + }, + { + "epoch": 0.01, + "learning_rate": 1.9873658768813856e-07, + "loss": 0.6828, + "step": 395 + }, + { + "epoch": 0.01, + "learning_rate": 1.9873292562346652e-07, + "loss": 0.6828, + "step": 396 + }, + { + "epoch": 0.01, + "learning_rate": 1.9872926355879444e-07, + "loss": 0.6819, + "step": 397 + }, + { + "epoch": 0.01, + "learning_rate": 1.987256014941224e-07, + "loss": 0.6805, + "step": 398 + }, + { + "epoch": 0.01, + "learning_rate": 1.9872193942945032e-07, + "loss": 0.7073, + "step": 399 + }, + { + "epoch": 0.01, + "learning_rate": 1.9871827736477825e-07, + "loss": 0.7216, + "step": 400 + }, + { + "epoch": 0.01, + "learning_rate": 1.9871461530010618e-07, + "loss": 0.6857, + "step": 401 + }, + { + "epoch": 0.01, + "learning_rate": 1.9871095323543413e-07, + "loss": 0.6794, + "step": 402 + }, + { + "epoch": 0.01, + "learning_rate": 1.9870729117076206e-07, + "loss": 0.7101, + "step": 403 + }, + { + "epoch": 0.01, + "learning_rate": 1.9870362910609e-07, + "loss": 0.6751, + "step": 404 + }, + { + "epoch": 0.01, + "learning_rate": 1.9869996704141794e-07, + "loss": 0.6788, + "step": 405 + }, + { + "epoch": 0.01, + "learning_rate": 1.9869630497674586e-07, + "loss": 0.6871, + "step": 406 + }, + { + "epoch": 0.01, + "learning_rate": 1.9869264291207382e-07, + "loss": 0.6675, + "step": 407 + }, + { + "epoch": 0.01, + "learning_rate": 1.9868898084740174e-07, + "loss": 0.6962, + "step": 408 + }, + { + "epoch": 0.01, + "learning_rate": 1.986853187827297e-07, + "loss": 0.7012, + "step": 409 + }, + { + "epoch": 0.02, + "learning_rate": 1.9868165671805762e-07, + "loss": 0.7063, + "step": 410 + }, + { + "epoch": 0.02, + "learning_rate": 1.9867799465338558e-07, + "loss": 0.6865, + "step": 411 + }, + { + "epoch": 0.02, + "learning_rate": 1.986743325887135e-07, + "loss": 0.7119, + "step": 412 + }, + { + "epoch": 0.02, + "learning_rate": 1.9867067052404146e-07, + "loss": 0.7062, + "step": 413 + }, + { + "epoch": 0.02, + "learning_rate": 1.9866700845936938e-07, + "loss": 0.7018, + "step": 414 + }, + { + "epoch": 0.02, + "learning_rate": 1.9866334639469734e-07, + "loss": 0.7052, + "step": 415 + }, + { + "epoch": 0.02, + "learning_rate": 1.9865968433002526e-07, + "loss": 0.7052, + "step": 416 + }, + { + "epoch": 0.02, + "learning_rate": 1.9865602226535322e-07, + "loss": 0.6799, + "step": 417 + }, + { + "epoch": 0.02, + "learning_rate": 1.9865236020068114e-07, + "loss": 0.6802, + "step": 418 + }, + { + "epoch": 0.02, + "learning_rate": 1.9864869813600907e-07, + "loss": 0.6844, + "step": 419 + }, + { + "epoch": 0.02, + "learning_rate": 1.98645036071337e-07, + "loss": 0.6848, + "step": 420 + }, + { + "epoch": 0.02, + "learning_rate": 1.9864137400666495e-07, + "loss": 0.7051, + "step": 421 + }, + { + "epoch": 0.02, + "learning_rate": 1.9863771194199288e-07, + "loss": 0.6601, + "step": 422 + }, + { + "epoch": 0.02, + "learning_rate": 1.9863404987732083e-07, + "loss": 0.6809, + "step": 423 + }, + { + "epoch": 0.02, + "learning_rate": 1.9863038781264876e-07, + "loss": 0.6996, + "step": 424 + }, + { + "epoch": 0.02, + "learning_rate": 1.986267257479767e-07, + "loss": 0.7226, + "step": 425 + }, + { + "epoch": 0.02, + "learning_rate": 1.9862306368330464e-07, + "loss": 0.6912, + "step": 426 + }, + { + "epoch": 0.02, + "learning_rate": 1.9861940161863256e-07, + "loss": 0.689, + "step": 427 + }, + { + "epoch": 0.02, + "learning_rate": 1.9861573955396052e-07, + "loss": 0.6888, + "step": 428 + }, + { + "epoch": 0.02, + "learning_rate": 1.9861207748928844e-07, + "loss": 0.7181, + "step": 429 + }, + { + "epoch": 0.02, + "learning_rate": 1.986084154246164e-07, + "loss": 0.7044, + "step": 430 + }, + { + "epoch": 0.02, + "learning_rate": 1.9860475335994432e-07, + "loss": 0.6941, + "step": 431 + }, + { + "epoch": 0.02, + "learning_rate": 1.9860109129527228e-07, + "loss": 0.6743, + "step": 432 + }, + { + "epoch": 0.02, + "learning_rate": 1.985974292306002e-07, + "loss": 0.6895, + "step": 433 + }, + { + "epoch": 0.02, + "learning_rate": 1.9859376716592816e-07, + "loss": 0.6966, + "step": 434 + }, + { + "epoch": 0.02, + "learning_rate": 1.9859010510125608e-07, + "loss": 0.7058, + "step": 435 + }, + { + "epoch": 0.02, + "learning_rate": 1.98586443036584e-07, + "loss": 0.6688, + "step": 436 + }, + { + "epoch": 0.02, + "learning_rate": 1.9858278097191194e-07, + "loss": 0.6952, + "step": 437 + }, + { + "epoch": 0.02, + "learning_rate": 1.985791189072399e-07, + "loss": 0.6719, + "step": 438 + }, + { + "epoch": 0.02, + "learning_rate": 1.9857545684256782e-07, + "loss": 0.6982, + "step": 439 + }, + { + "epoch": 0.02, + "learning_rate": 1.9857179477789577e-07, + "loss": 0.7214, + "step": 440 + }, + { + "epoch": 0.02, + "learning_rate": 1.985681327132237e-07, + "loss": 0.714, + "step": 441 + }, + { + "epoch": 0.02, + "learning_rate": 1.9856447064855165e-07, + "loss": 0.7016, + "step": 442 + }, + { + "epoch": 0.02, + "learning_rate": 1.9856080858387958e-07, + "loss": 0.694, + "step": 443 + }, + { + "epoch": 0.02, + "learning_rate": 1.9855714651920753e-07, + "loss": 0.6732, + "step": 444 + }, + { + "epoch": 0.02, + "learning_rate": 1.9855348445453546e-07, + "loss": 0.7068, + "step": 445 + }, + { + "epoch": 0.02, + "learning_rate": 1.985498223898634e-07, + "loss": 0.6893, + "step": 446 + }, + { + "epoch": 0.02, + "learning_rate": 1.9854616032519134e-07, + "loss": 0.6824, + "step": 447 + }, + { + "epoch": 0.02, + "learning_rate": 1.985424982605193e-07, + "loss": 0.7069, + "step": 448 + }, + { + "epoch": 0.02, + "learning_rate": 1.9853883619584722e-07, + "loss": 0.6912, + "step": 449 + }, + { + "epoch": 0.02, + "learning_rate": 1.9853517413117514e-07, + "loss": 0.693, + "step": 450 + }, + { + "epoch": 0.02, + "learning_rate": 1.985315120665031e-07, + "loss": 0.6854, + "step": 451 + }, + { + "epoch": 0.02, + "learning_rate": 1.9852785000183102e-07, + "loss": 0.7228, + "step": 452 + }, + { + "epoch": 0.02, + "learning_rate": 1.9852418793715897e-07, + "loss": 0.6972, + "step": 453 + }, + { + "epoch": 0.02, + "learning_rate": 1.9852052587248688e-07, + "loss": 0.7042, + "step": 454 + }, + { + "epoch": 0.02, + "learning_rate": 1.9851686380781483e-07, + "loss": 0.6623, + "step": 455 + }, + { + "epoch": 0.02, + "learning_rate": 1.9851320174314275e-07, + "loss": 0.6767, + "step": 456 + }, + { + "epoch": 0.02, + "learning_rate": 1.985095396784707e-07, + "loss": 0.6739, + "step": 457 + }, + { + "epoch": 0.02, + "learning_rate": 1.9850587761379863e-07, + "loss": 0.6868, + "step": 458 + }, + { + "epoch": 0.02, + "learning_rate": 1.985022155491266e-07, + "loss": 0.7133, + "step": 459 + }, + { + "epoch": 0.02, + "learning_rate": 1.9849855348445451e-07, + "loss": 0.6977, + "step": 460 + }, + { + "epoch": 0.02, + "learning_rate": 1.9849489141978247e-07, + "loss": 0.6859, + "step": 461 + }, + { + "epoch": 0.02, + "learning_rate": 1.984912293551104e-07, + "loss": 0.6973, + "step": 462 + }, + { + "epoch": 0.02, + "learning_rate": 1.9848756729043835e-07, + "loss": 0.7217, + "step": 463 + }, + { + "epoch": 0.02, + "learning_rate": 1.9848390522576627e-07, + "loss": 0.7274, + "step": 464 + }, + { + "epoch": 0.02, + "learning_rate": 1.9848024316109423e-07, + "loss": 0.6881, + "step": 465 + }, + { + "epoch": 0.02, + "learning_rate": 1.9847658109642215e-07, + "loss": 0.6955, + "step": 466 + }, + { + "epoch": 0.02, + "learning_rate": 1.984729190317501e-07, + "loss": 0.6688, + "step": 467 + }, + { + "epoch": 0.02, + "learning_rate": 1.9846925696707803e-07, + "loss": 0.7017, + "step": 468 + }, + { + "epoch": 0.02, + "learning_rate": 1.98465594902406e-07, + "loss": 0.7005, + "step": 469 + }, + { + "epoch": 0.02, + "learning_rate": 1.9846193283773391e-07, + "loss": 0.7188, + "step": 470 + }, + { + "epoch": 0.02, + "learning_rate": 1.9845827077306184e-07, + "loss": 0.733, + "step": 471 + }, + { + "epoch": 0.02, + "learning_rate": 1.9845460870838977e-07, + "loss": 0.7101, + "step": 472 + }, + { + "epoch": 0.02, + "learning_rate": 1.9845094664371772e-07, + "loss": 0.6896, + "step": 473 + }, + { + "epoch": 0.02, + "learning_rate": 1.9844728457904565e-07, + "loss": 0.6913, + "step": 474 + }, + { + "epoch": 0.02, + "learning_rate": 1.984436225143736e-07, + "loss": 0.6648, + "step": 475 + }, + { + "epoch": 0.02, + "learning_rate": 1.9843996044970153e-07, + "loss": 0.6855, + "step": 476 + }, + { + "epoch": 0.02, + "learning_rate": 1.9843629838502945e-07, + "loss": 0.7066, + "step": 477 + }, + { + "epoch": 0.02, + "learning_rate": 1.984326363203574e-07, + "loss": 0.6975, + "step": 478 + }, + { + "epoch": 0.02, + "learning_rate": 1.9842897425568533e-07, + "loss": 0.7014, + "step": 479 + }, + { + "epoch": 0.02, + "learning_rate": 1.9842531219101329e-07, + "loss": 0.7058, + "step": 480 + }, + { + "epoch": 0.02, + "learning_rate": 1.9842165012634121e-07, + "loss": 0.6709, + "step": 481 + }, + { + "epoch": 0.02, + "learning_rate": 1.9841798806166917e-07, + "loss": 0.7203, + "step": 482 + }, + { + "epoch": 0.02, + "learning_rate": 1.984143259969971e-07, + "loss": 0.6691, + "step": 483 + }, + { + "epoch": 0.02, + "learning_rate": 1.9841066393232505e-07, + "loss": 0.7022, + "step": 484 + }, + { + "epoch": 0.02, + "learning_rate": 1.9840700186765297e-07, + "loss": 0.68, + "step": 485 + }, + { + "epoch": 0.02, + "learning_rate": 1.9840333980298093e-07, + "loss": 0.7337, + "step": 486 + }, + { + "epoch": 0.02, + "learning_rate": 1.9839967773830885e-07, + "loss": 0.6911, + "step": 487 + }, + { + "epoch": 0.02, + "learning_rate": 1.983960156736368e-07, + "loss": 0.6925, + "step": 488 + }, + { + "epoch": 0.02, + "learning_rate": 1.9839235360896473e-07, + "loss": 0.6866, + "step": 489 + }, + { + "epoch": 0.02, + "learning_rate": 1.9838869154429266e-07, + "loss": 0.7249, + "step": 490 + }, + { + "epoch": 0.02, + "learning_rate": 1.9838502947962059e-07, + "loss": 0.6925, + "step": 491 + }, + { + "epoch": 0.02, + "learning_rate": 1.9838136741494854e-07, + "loss": 0.7065, + "step": 492 + }, + { + "epoch": 0.02, + "learning_rate": 1.9837770535027647e-07, + "loss": 0.6986, + "step": 493 + }, + { + "epoch": 0.02, + "learning_rate": 1.9837404328560442e-07, + "loss": 0.6932, + "step": 494 + }, + { + "epoch": 0.02, + "learning_rate": 1.9837038122093235e-07, + "loss": 0.6946, + "step": 495 + }, + { + "epoch": 0.02, + "learning_rate": 1.983667191562603e-07, + "loss": 0.6777, + "step": 496 + }, + { + "epoch": 0.02, + "learning_rate": 1.9836305709158823e-07, + "loss": 0.7176, + "step": 497 + }, + { + "epoch": 0.02, + "learning_rate": 1.9835939502691615e-07, + "loss": 0.6987, + "step": 498 + }, + { + "epoch": 0.02, + "learning_rate": 1.983557329622441e-07, + "loss": 0.7031, + "step": 499 + }, + { + "epoch": 0.02, + "learning_rate": 1.9835207089757203e-07, + "loss": 0.6928, + "step": 500 + }, + { + "epoch": 0.02, + "learning_rate": 1.9834840883289999e-07, + "loss": 0.6847, + "step": 501 + }, + { + "epoch": 0.02, + "learning_rate": 1.983447467682279e-07, + "loss": 0.6943, + "step": 502 + }, + { + "epoch": 0.02, + "learning_rate": 1.9834108470355587e-07, + "loss": 0.6721, + "step": 503 + }, + { + "epoch": 0.02, + "learning_rate": 1.983374226388838e-07, + "loss": 0.6947, + "step": 504 + }, + { + "epoch": 0.02, + "learning_rate": 1.9833376057421175e-07, + "loss": 0.6917, + "step": 505 + }, + { + "epoch": 0.02, + "learning_rate": 1.9833009850953967e-07, + "loss": 0.6693, + "step": 506 + }, + { + "epoch": 0.02, + "learning_rate": 1.9832643644486763e-07, + "loss": 0.6816, + "step": 507 + }, + { + "epoch": 0.02, + "learning_rate": 1.9832277438019553e-07, + "loss": 0.6987, + "step": 508 + }, + { + "epoch": 0.02, + "learning_rate": 1.9831911231552348e-07, + "loss": 0.6951, + "step": 509 + }, + { + "epoch": 0.02, + "learning_rate": 1.983154502508514e-07, + "loss": 0.6999, + "step": 510 + }, + { + "epoch": 0.02, + "learning_rate": 1.9831178818617936e-07, + "loss": 0.6785, + "step": 511 + }, + { + "epoch": 0.02, + "learning_rate": 1.9830812612150729e-07, + "loss": 0.698, + "step": 512 + }, + { + "epoch": 0.02, + "learning_rate": 1.9830446405683524e-07, + "loss": 0.6995, + "step": 513 + }, + { + "epoch": 0.02, + "learning_rate": 1.9830080199216316e-07, + "loss": 0.6688, + "step": 514 + }, + { + "epoch": 0.02, + "learning_rate": 1.9829713992749112e-07, + "loss": 0.7155, + "step": 515 + }, + { + "epoch": 0.02, + "learning_rate": 1.9829347786281904e-07, + "loss": 0.7039, + "step": 516 + }, + { + "epoch": 0.02, + "learning_rate": 1.98289815798147e-07, + "loss": 0.7316, + "step": 517 + }, + { + "epoch": 0.02, + "learning_rate": 1.9828615373347492e-07, + "loss": 0.6959, + "step": 518 + }, + { + "epoch": 0.02, + "learning_rate": 1.9828249166880288e-07, + "loss": 0.7029, + "step": 519 + }, + { + "epoch": 0.02, + "learning_rate": 1.982788296041308e-07, + "loss": 0.6964, + "step": 520 + }, + { + "epoch": 0.02, + "learning_rate": 1.9827516753945873e-07, + "loss": 0.7052, + "step": 521 + }, + { + "epoch": 0.02, + "learning_rate": 1.9827150547478668e-07, + "loss": 0.6733, + "step": 522 + }, + { + "epoch": 0.02, + "learning_rate": 1.982678434101146e-07, + "loss": 0.7073, + "step": 523 + }, + { + "epoch": 0.02, + "learning_rate": 1.9826418134544256e-07, + "loss": 0.6967, + "step": 524 + }, + { + "epoch": 0.02, + "learning_rate": 1.982605192807705e-07, + "loss": 0.6787, + "step": 525 + }, + { + "epoch": 0.02, + "learning_rate": 1.9825685721609842e-07, + "loss": 0.6887, + "step": 526 + }, + { + "epoch": 0.02, + "learning_rate": 1.9825319515142634e-07, + "loss": 0.6986, + "step": 527 + }, + { + "epoch": 0.02, + "learning_rate": 1.982495330867543e-07, + "loss": 0.6907, + "step": 528 + }, + { + "epoch": 0.02, + "learning_rate": 1.9824587102208222e-07, + "loss": 0.6928, + "step": 529 + }, + { + "epoch": 0.02, + "learning_rate": 1.9824220895741018e-07, + "loss": 0.7206, + "step": 530 + }, + { + "epoch": 0.02, + "learning_rate": 1.982385468927381e-07, + "loss": 0.7093, + "step": 531 + }, + { + "epoch": 0.02, + "learning_rate": 1.9823488482806606e-07, + "loss": 0.6996, + "step": 532 + }, + { + "epoch": 0.02, + "learning_rate": 1.9823122276339398e-07, + "loss": 0.6781, + "step": 533 + }, + { + "epoch": 0.02, + "learning_rate": 1.9822756069872194e-07, + "loss": 0.6969, + "step": 534 + }, + { + "epoch": 0.02, + "learning_rate": 1.9822389863404986e-07, + "loss": 0.6911, + "step": 535 + }, + { + "epoch": 0.02, + "learning_rate": 1.9822023656937782e-07, + "loss": 0.6913, + "step": 536 + }, + { + "epoch": 0.02, + "learning_rate": 1.9821657450470574e-07, + "loss": 0.7018, + "step": 537 + }, + { + "epoch": 0.02, + "learning_rate": 1.982129124400337e-07, + "loss": 0.7259, + "step": 538 + }, + { + "epoch": 0.02, + "learning_rate": 1.9820925037536162e-07, + "loss": 0.7029, + "step": 539 + }, + { + "epoch": 0.02, + "learning_rate": 1.9820558831068958e-07, + "loss": 0.6723, + "step": 540 + }, + { + "epoch": 0.02, + "learning_rate": 1.982019262460175e-07, + "loss": 0.6823, + "step": 541 + }, + { + "epoch": 0.02, + "learning_rate": 1.9819826418134543e-07, + "loss": 0.695, + "step": 542 + }, + { + "epoch": 0.02, + "learning_rate": 1.9819460211667338e-07, + "loss": 0.6744, + "step": 543 + }, + { + "epoch": 0.02, + "learning_rate": 1.981909400520013e-07, + "loss": 0.727, + "step": 544 + }, + { + "epoch": 0.02, + "learning_rate": 1.9818727798732924e-07, + "loss": 0.6759, + "step": 545 + }, + { + "epoch": 0.02, + "learning_rate": 1.981836159226572e-07, + "loss": 0.688, + "step": 546 + }, + { + "epoch": 0.02, + "learning_rate": 1.9817995385798512e-07, + "loss": 0.6858, + "step": 547 + }, + { + "epoch": 0.02, + "learning_rate": 1.9817629179331304e-07, + "loss": 0.6763, + "step": 548 + }, + { + "epoch": 0.02, + "learning_rate": 1.98172629728641e-07, + "loss": 0.6739, + "step": 549 + }, + { + "epoch": 0.02, + "learning_rate": 1.9816896766396892e-07, + "loss": 0.6918, + "step": 550 + }, + { + "epoch": 0.02, + "learning_rate": 1.9816530559929688e-07, + "loss": 0.6924, + "step": 551 + }, + { + "epoch": 0.02, + "learning_rate": 1.981616435346248e-07, + "loss": 0.6725, + "step": 552 + }, + { + "epoch": 0.02, + "learning_rate": 1.9815798146995276e-07, + "loss": 0.7067, + "step": 553 + }, + { + "epoch": 0.02, + "learning_rate": 1.9815431940528068e-07, + "loss": 0.6794, + "step": 554 + }, + { + "epoch": 0.02, + "learning_rate": 1.9815065734060864e-07, + "loss": 0.7078, + "step": 555 + }, + { + "epoch": 0.02, + "learning_rate": 1.9814699527593656e-07, + "loss": 0.6828, + "step": 556 + }, + { + "epoch": 0.02, + "learning_rate": 1.9814333321126452e-07, + "loss": 0.7037, + "step": 557 + }, + { + "epoch": 0.02, + "learning_rate": 1.9813967114659244e-07, + "loss": 0.683, + "step": 558 + }, + { + "epoch": 0.02, + "learning_rate": 1.981360090819204e-07, + "loss": 0.6971, + "step": 559 + }, + { + "epoch": 0.02, + "learning_rate": 1.9813234701724832e-07, + "loss": 0.6862, + "step": 560 + }, + { + "epoch": 0.02, + "learning_rate": 1.9812868495257628e-07, + "loss": 0.7098, + "step": 561 + }, + { + "epoch": 0.02, + "learning_rate": 1.9812502288790418e-07, + "loss": 0.693, + "step": 562 + }, + { + "epoch": 0.02, + "learning_rate": 1.9812136082323213e-07, + "loss": 0.6765, + "step": 563 + }, + { + "epoch": 0.02, + "learning_rate": 1.9811769875856006e-07, + "loss": 0.7107, + "step": 564 + }, + { + "epoch": 0.02, + "learning_rate": 1.98114036693888e-07, + "loss": 0.6815, + "step": 565 + }, + { + "epoch": 0.02, + "learning_rate": 1.9811037462921594e-07, + "loss": 0.6758, + "step": 566 + }, + { + "epoch": 0.02, + "learning_rate": 1.981067125645439e-07, + "loss": 0.7013, + "step": 567 + }, + { + "epoch": 0.02, + "learning_rate": 1.9810305049987182e-07, + "loss": 0.6964, + "step": 568 + }, + { + "epoch": 0.02, + "learning_rate": 1.9809938843519974e-07, + "loss": 0.6989, + "step": 569 + }, + { + "epoch": 0.02, + "learning_rate": 1.980957263705277e-07, + "loss": 0.6737, + "step": 570 + }, + { + "epoch": 0.02, + "learning_rate": 1.9809206430585562e-07, + "loss": 0.6809, + "step": 571 + }, + { + "epoch": 0.02, + "learning_rate": 1.9808840224118357e-07, + "loss": 0.7008, + "step": 572 + }, + { + "epoch": 0.02, + "learning_rate": 1.980847401765115e-07, + "loss": 0.7006, + "step": 573 + }, + { + "epoch": 0.02, + "learning_rate": 1.9808107811183945e-07, + "loss": 0.7103, + "step": 574 + }, + { + "epoch": 0.02, + "learning_rate": 1.9807741604716738e-07, + "loss": 0.6976, + "step": 575 + }, + { + "epoch": 0.02, + "learning_rate": 1.9807375398249533e-07, + "loss": 0.6952, + "step": 576 + }, + { + "epoch": 0.02, + "learning_rate": 1.9807009191782326e-07, + "loss": 0.6988, + "step": 577 + }, + { + "epoch": 0.02, + "learning_rate": 1.9806642985315121e-07, + "loss": 0.691, + "step": 578 + }, + { + "epoch": 0.02, + "learning_rate": 1.9806276778847914e-07, + "loss": 0.7055, + "step": 579 + }, + { + "epoch": 0.02, + "learning_rate": 1.9805910572380707e-07, + "loss": 0.6894, + "step": 580 + }, + { + "epoch": 0.02, + "learning_rate": 1.98055443659135e-07, + "loss": 0.6906, + "step": 581 + }, + { + "epoch": 0.02, + "learning_rate": 1.9805178159446295e-07, + "loss": 0.6744, + "step": 582 + }, + { + "epoch": 0.02, + "learning_rate": 1.9804811952979087e-07, + "loss": 0.6878, + "step": 583 + }, + { + "epoch": 0.02, + "learning_rate": 1.9804445746511883e-07, + "loss": 0.7161, + "step": 584 + }, + { + "epoch": 0.02, + "learning_rate": 1.9804079540044675e-07, + "loss": 0.7037, + "step": 585 + }, + { + "epoch": 0.02, + "learning_rate": 1.980371333357747e-07, + "loss": 0.6928, + "step": 586 + }, + { + "epoch": 0.02, + "learning_rate": 1.9803347127110263e-07, + "loss": 0.6847, + "step": 587 + }, + { + "epoch": 0.02, + "learning_rate": 1.980298092064306e-07, + "loss": 0.7093, + "step": 588 + }, + { + "epoch": 0.02, + "learning_rate": 1.9802614714175851e-07, + "loss": 0.7048, + "step": 589 + }, + { + "epoch": 0.02, + "learning_rate": 1.9802248507708647e-07, + "loss": 0.6907, + "step": 590 + }, + { + "epoch": 0.02, + "learning_rate": 1.980188230124144e-07, + "loss": 0.6664, + "step": 591 + }, + { + "epoch": 0.02, + "learning_rate": 1.9801516094774232e-07, + "loss": 0.7031, + "step": 592 + }, + { + "epoch": 0.02, + "learning_rate": 1.9801149888307027e-07, + "loss": 0.6821, + "step": 593 + }, + { + "epoch": 0.02, + "learning_rate": 1.980078368183982e-07, + "loss": 0.6987, + "step": 594 + }, + { + "epoch": 0.02, + "learning_rate": 1.9800417475372615e-07, + "loss": 0.6717, + "step": 595 + }, + { + "epoch": 0.02, + "learning_rate": 1.9800051268905408e-07, + "loss": 0.6994, + "step": 596 + }, + { + "epoch": 0.02, + "learning_rate": 1.9799685062438203e-07, + "loss": 0.6926, + "step": 597 + }, + { + "epoch": 0.02, + "learning_rate": 1.9799318855970993e-07, + "loss": 0.6784, + "step": 598 + }, + { + "epoch": 0.02, + "learning_rate": 1.979895264950379e-07, + "loss": 0.6707, + "step": 599 + }, + { + "epoch": 0.02, + "learning_rate": 1.9798586443036581e-07, + "loss": 0.6952, + "step": 600 + }, + { + "epoch": 0.02, + "learning_rate": 1.9798220236569377e-07, + "loss": 0.6942, + "step": 601 + }, + { + "epoch": 0.02, + "learning_rate": 1.979785403010217e-07, + "loss": 0.7028, + "step": 602 + }, + { + "epoch": 0.02, + "learning_rate": 1.9797487823634965e-07, + "loss": 0.6888, + "step": 603 + }, + { + "epoch": 0.02, + "learning_rate": 1.9797121617167757e-07, + "loss": 0.6717, + "step": 604 + }, + { + "epoch": 0.02, + "learning_rate": 1.9796755410700553e-07, + "loss": 0.7008, + "step": 605 + }, + { + "epoch": 0.02, + "learning_rate": 1.9796389204233345e-07, + "loss": 0.6844, + "step": 606 + }, + { + "epoch": 0.02, + "learning_rate": 1.979602299776614e-07, + "loss": 0.7136, + "step": 607 + }, + { + "epoch": 0.02, + "learning_rate": 1.9795656791298933e-07, + "loss": 0.6894, + "step": 608 + }, + { + "epoch": 0.02, + "learning_rate": 1.9795290584831729e-07, + "loss": 0.6948, + "step": 609 + }, + { + "epoch": 0.02, + "learning_rate": 1.979492437836452e-07, + "loss": 0.6979, + "step": 610 + }, + { + "epoch": 0.02, + "learning_rate": 1.9794558171897317e-07, + "loss": 0.6862, + "step": 611 + }, + { + "epoch": 0.02, + "learning_rate": 1.979419196543011e-07, + "loss": 0.6958, + "step": 612 + }, + { + "epoch": 0.02, + "learning_rate": 1.9793825758962902e-07, + "loss": 0.713, + "step": 613 + }, + { + "epoch": 0.02, + "learning_rate": 1.9793459552495697e-07, + "loss": 0.7036, + "step": 614 + }, + { + "epoch": 0.02, + "learning_rate": 1.979309334602849e-07, + "loss": 0.6864, + "step": 615 + }, + { + "epoch": 0.02, + "learning_rate": 1.9792727139561283e-07, + "loss": 0.6972, + "step": 616 + }, + { + "epoch": 0.02, + "learning_rate": 1.9792360933094078e-07, + "loss": 0.6823, + "step": 617 + }, + { + "epoch": 0.02, + "learning_rate": 1.979199472662687e-07, + "loss": 0.7129, + "step": 618 + }, + { + "epoch": 0.02, + "learning_rate": 1.9791628520159663e-07, + "loss": 0.6691, + "step": 619 + }, + { + "epoch": 0.02, + "learning_rate": 1.9791262313692459e-07, + "loss": 0.6843, + "step": 620 + }, + { + "epoch": 0.02, + "learning_rate": 1.979089610722525e-07, + "loss": 0.6874, + "step": 621 + }, + { + "epoch": 0.02, + "learning_rate": 1.9790529900758047e-07, + "loss": 0.7111, + "step": 622 + }, + { + "epoch": 0.02, + "learning_rate": 1.979016369429084e-07, + "loss": 0.6912, + "step": 623 + }, + { + "epoch": 0.02, + "learning_rate": 1.9789797487823635e-07, + "loss": 0.6941, + "step": 624 + }, + { + "epoch": 0.02, + "learning_rate": 1.9789431281356427e-07, + "loss": 0.6931, + "step": 625 + }, + { + "epoch": 0.02, + "learning_rate": 1.9789065074889223e-07, + "loss": 0.7189, + "step": 626 + }, + { + "epoch": 0.02, + "learning_rate": 1.9788698868422015e-07, + "loss": 0.7119, + "step": 627 + }, + { + "epoch": 0.02, + "learning_rate": 1.978833266195481e-07, + "loss": 0.6767, + "step": 628 + }, + { + "epoch": 0.02, + "learning_rate": 1.9787966455487603e-07, + "loss": 0.7081, + "step": 629 + }, + { + "epoch": 0.02, + "learning_rate": 1.9787600249020398e-07, + "loss": 0.706, + "step": 630 + }, + { + "epoch": 0.02, + "learning_rate": 1.978723404255319e-07, + "loss": 0.674, + "step": 631 + }, + { + "epoch": 0.02, + "learning_rate": 1.9786867836085986e-07, + "loss": 0.6988, + "step": 632 + }, + { + "epoch": 0.02, + "learning_rate": 1.978650162961878e-07, + "loss": 0.6886, + "step": 633 + }, + { + "epoch": 0.02, + "learning_rate": 1.9786135423151572e-07, + "loss": 0.689, + "step": 634 + }, + { + "epoch": 0.02, + "learning_rate": 1.9785769216684364e-07, + "loss": 0.714, + "step": 635 + }, + { + "epoch": 0.02, + "learning_rate": 1.978540301021716e-07, + "loss": 0.6889, + "step": 636 + }, + { + "epoch": 0.02, + "learning_rate": 1.9785036803749952e-07, + "loss": 0.6594, + "step": 637 + }, + { + "epoch": 0.02, + "learning_rate": 1.9784670597282748e-07, + "loss": 0.6965, + "step": 638 + }, + { + "epoch": 0.02, + "learning_rate": 1.978430439081554e-07, + "loss": 0.6888, + "step": 639 + }, + { + "epoch": 0.02, + "learning_rate": 1.9783938184348333e-07, + "loss": 0.7054, + "step": 640 + }, + { + "epoch": 0.02, + "learning_rate": 1.9783571977881128e-07, + "loss": 0.6983, + "step": 641 + }, + { + "epoch": 0.02, + "learning_rate": 1.978320577141392e-07, + "loss": 0.6886, + "step": 642 + }, + { + "epoch": 0.02, + "learning_rate": 1.9782839564946716e-07, + "loss": 0.6866, + "step": 643 + }, + { + "epoch": 0.02, + "learning_rate": 1.978247335847951e-07, + "loss": 0.6958, + "step": 644 + }, + { + "epoch": 0.02, + "learning_rate": 1.9782107152012304e-07, + "loss": 0.7, + "step": 645 + }, + { + "epoch": 0.02, + "learning_rate": 1.9781740945545097e-07, + "loss": 0.6985, + "step": 646 + }, + { + "epoch": 0.02, + "learning_rate": 1.9781374739077892e-07, + "loss": 0.6826, + "step": 647 + }, + { + "epoch": 0.02, + "learning_rate": 1.9781008532610685e-07, + "loss": 0.7037, + "step": 648 + }, + { + "epoch": 0.02, + "learning_rate": 1.978064232614348e-07, + "loss": 0.6831, + "step": 649 + }, + { + "epoch": 0.02, + "learning_rate": 1.9780276119676273e-07, + "loss": 0.6876, + "step": 650 + }, + { + "epoch": 0.02, + "learning_rate": 1.9779909913209068e-07, + "loss": 0.6949, + "step": 651 + }, + { + "epoch": 0.02, + "learning_rate": 1.9779543706741858e-07, + "loss": 0.6972, + "step": 652 + }, + { + "epoch": 0.02, + "learning_rate": 1.9779177500274654e-07, + "loss": 0.6696, + "step": 653 + }, + { + "epoch": 0.02, + "learning_rate": 1.9778811293807446e-07, + "loss": 0.6838, + "step": 654 + }, + { + "epoch": 0.02, + "learning_rate": 1.9778445087340242e-07, + "loss": 0.6801, + "step": 655 + }, + { + "epoch": 0.02, + "learning_rate": 1.9778078880873034e-07, + "loss": 0.679, + "step": 656 + }, + { + "epoch": 0.02, + "learning_rate": 1.977771267440583e-07, + "loss": 0.6958, + "step": 657 + }, + { + "epoch": 0.02, + "learning_rate": 1.9777346467938622e-07, + "loss": 0.7106, + "step": 658 + }, + { + "epoch": 0.02, + "learning_rate": 1.9776980261471418e-07, + "loss": 0.7128, + "step": 659 + }, + { + "epoch": 0.02, + "learning_rate": 1.977661405500421e-07, + "loss": 0.6879, + "step": 660 + }, + { + "epoch": 0.02, + "learning_rate": 1.9776247848537006e-07, + "loss": 0.6789, + "step": 661 + }, + { + "epoch": 0.02, + "learning_rate": 1.9775881642069798e-07, + "loss": 0.7027, + "step": 662 + }, + { + "epoch": 0.02, + "learning_rate": 1.977551543560259e-07, + "loss": 0.6525, + "step": 663 + }, + { + "epoch": 0.02, + "learning_rate": 1.9775149229135386e-07, + "loss": 0.681, + "step": 664 + }, + { + "epoch": 0.02, + "learning_rate": 1.977478302266818e-07, + "loss": 0.7028, + "step": 665 + }, + { + "epoch": 0.02, + "learning_rate": 1.9774416816200974e-07, + "loss": 0.7018, + "step": 666 + }, + { + "epoch": 0.02, + "learning_rate": 1.9774050609733767e-07, + "loss": 0.6923, + "step": 667 + }, + { + "epoch": 0.02, + "learning_rate": 1.9773684403266562e-07, + "loss": 0.7179, + "step": 668 + }, + { + "epoch": 0.02, + "learning_rate": 1.9773318196799355e-07, + "loss": 0.6762, + "step": 669 + }, + { + "epoch": 0.02, + "learning_rate": 1.9772951990332148e-07, + "loss": 0.7135, + "step": 670 + }, + { + "epoch": 0.02, + "learning_rate": 1.977258578386494e-07, + "loss": 0.6887, + "step": 671 + }, + { + "epoch": 0.02, + "learning_rate": 1.9772219577397736e-07, + "loss": 0.7065, + "step": 672 + }, + { + "epoch": 0.02, + "learning_rate": 1.9771853370930528e-07, + "loss": 0.7078, + "step": 673 + }, + { + "epoch": 0.02, + "learning_rate": 1.9771487164463324e-07, + "loss": 0.6967, + "step": 674 + }, + { + "epoch": 0.02, + "learning_rate": 1.9771120957996116e-07, + "loss": 0.6774, + "step": 675 + }, + { + "epoch": 0.02, + "learning_rate": 1.9770754751528912e-07, + "loss": 0.7014, + "step": 676 + }, + { + "epoch": 0.02, + "learning_rate": 1.9770388545061704e-07, + "loss": 0.7106, + "step": 677 + }, + { + "epoch": 0.02, + "learning_rate": 1.97700223385945e-07, + "loss": 0.6991, + "step": 678 + }, + { + "epoch": 0.02, + "learning_rate": 1.9769656132127292e-07, + "loss": 0.6734, + "step": 679 + }, + { + "epoch": 0.02, + "learning_rate": 1.9769289925660088e-07, + "loss": 0.6883, + "step": 680 + }, + { + "epoch": 0.02, + "learning_rate": 1.976892371919288e-07, + "loss": 0.7221, + "step": 681 + }, + { + "epoch": 0.02, + "learning_rate": 1.9768557512725676e-07, + "loss": 0.7123, + "step": 682 + }, + { + "epoch": 0.02, + "learning_rate": 1.9768191306258468e-07, + "loss": 0.6616, + "step": 683 + }, + { + "epoch": 0.03, + "learning_rate": 1.976782509979126e-07, + "loss": 0.715, + "step": 684 + }, + { + "epoch": 0.03, + "learning_rate": 1.9767458893324056e-07, + "loss": 0.686, + "step": 685 + }, + { + "epoch": 0.03, + "learning_rate": 1.976709268685685e-07, + "loss": 0.6762, + "step": 686 + }, + { + "epoch": 0.03, + "learning_rate": 1.9766726480389644e-07, + "loss": 0.7181, + "step": 687 + }, + { + "epoch": 0.03, + "learning_rate": 1.9766360273922437e-07, + "loss": 0.6954, + "step": 688 + }, + { + "epoch": 0.03, + "learning_rate": 1.976599406745523e-07, + "loss": 0.7062, + "step": 689 + }, + { + "epoch": 0.03, + "learning_rate": 1.9765627860988022e-07, + "loss": 0.7052, + "step": 690 + }, + { + "epoch": 0.03, + "learning_rate": 1.9765261654520817e-07, + "loss": 0.6942, + "step": 691 + }, + { + "epoch": 0.03, + "learning_rate": 1.976489544805361e-07, + "loss": 0.6848, + "step": 692 + }, + { + "epoch": 0.03, + "learning_rate": 1.9764529241586405e-07, + "loss": 0.7027, + "step": 693 + }, + { + "epoch": 0.03, + "learning_rate": 1.9764163035119198e-07, + "loss": 0.6426, + "step": 694 + }, + { + "epoch": 0.03, + "learning_rate": 1.9763796828651993e-07, + "loss": 0.6888, + "step": 695 + }, + { + "epoch": 0.03, + "learning_rate": 1.9763430622184786e-07, + "loss": 0.6896, + "step": 696 + }, + { + "epoch": 0.03, + "learning_rate": 1.9763064415717581e-07, + "loss": 0.6855, + "step": 697 + }, + { + "epoch": 0.03, + "learning_rate": 1.9762698209250374e-07, + "loss": 0.6966, + "step": 698 + }, + { + "epoch": 0.03, + "learning_rate": 1.976233200278317e-07, + "loss": 0.7022, + "step": 699 + }, + { + "epoch": 0.03, + "learning_rate": 1.9761965796315962e-07, + "loss": 0.704, + "step": 700 + }, + { + "epoch": 0.03, + "learning_rate": 1.9761599589848757e-07, + "loss": 0.6909, + "step": 701 + }, + { + "epoch": 0.03, + "learning_rate": 1.976123338338155e-07, + "loss": 0.6962, + "step": 702 + }, + { + "epoch": 0.03, + "learning_rate": 1.9760867176914345e-07, + "loss": 0.7009, + "step": 703 + }, + { + "epoch": 0.03, + "learning_rate": 1.9760500970447138e-07, + "loss": 0.6877, + "step": 704 + }, + { + "epoch": 0.03, + "learning_rate": 1.976013476397993e-07, + "loss": 0.6856, + "step": 705 + }, + { + "epoch": 0.03, + "learning_rate": 1.9759768557512723e-07, + "loss": 0.7127, + "step": 706 + }, + { + "epoch": 0.03, + "learning_rate": 1.975940235104552e-07, + "loss": 0.7058, + "step": 707 + }, + { + "epoch": 0.03, + "learning_rate": 1.9759036144578311e-07, + "loss": 0.6798, + "step": 708 + }, + { + "epoch": 0.03, + "learning_rate": 1.9758669938111107e-07, + "loss": 0.6901, + "step": 709 + }, + { + "epoch": 0.03, + "learning_rate": 1.97583037316439e-07, + "loss": 0.7005, + "step": 710 + }, + { + "epoch": 0.03, + "learning_rate": 1.9757937525176692e-07, + "loss": 0.6944, + "step": 711 + }, + { + "epoch": 0.03, + "learning_rate": 1.9757571318709487e-07, + "loss": 0.6935, + "step": 712 + }, + { + "epoch": 0.03, + "learning_rate": 1.975720511224228e-07, + "loss": 0.688, + "step": 713 + }, + { + "epoch": 0.03, + "learning_rate": 1.9756838905775075e-07, + "loss": 0.6936, + "step": 714 + }, + { + "epoch": 0.03, + "learning_rate": 1.9756472699307868e-07, + "loss": 0.6862, + "step": 715 + }, + { + "epoch": 0.03, + "learning_rate": 1.9756106492840663e-07, + "loss": 0.7142, + "step": 716 + }, + { + "epoch": 0.03, + "learning_rate": 1.9755740286373456e-07, + "loss": 0.7146, + "step": 717 + }, + { + "epoch": 0.03, + "learning_rate": 1.9755374079906251e-07, + "loss": 0.6781, + "step": 718 + }, + { + "epoch": 0.03, + "learning_rate": 1.9755007873439044e-07, + "loss": 0.6754, + "step": 719 + }, + { + "epoch": 0.03, + "learning_rate": 1.975464166697184e-07, + "loss": 0.6913, + "step": 720 + }, + { + "epoch": 0.03, + "learning_rate": 1.9754275460504632e-07, + "loss": 0.7071, + "step": 721 + }, + { + "epoch": 0.03, + "learning_rate": 1.9753909254037427e-07, + "loss": 0.7051, + "step": 722 + }, + { + "epoch": 0.03, + "learning_rate": 1.975354304757022e-07, + "loss": 0.7001, + "step": 723 + }, + { + "epoch": 0.03, + "learning_rate": 1.9753176841103013e-07, + "loss": 0.7122, + "step": 724 + }, + { + "epoch": 0.03, + "learning_rate": 1.9752810634635805e-07, + "loss": 0.6991, + "step": 725 + }, + { + "epoch": 0.03, + "learning_rate": 1.97524444281686e-07, + "loss": 0.7007, + "step": 726 + }, + { + "epoch": 0.03, + "learning_rate": 1.9752078221701393e-07, + "loss": 0.7166, + "step": 727 + }, + { + "epoch": 0.03, + "learning_rate": 1.9751712015234189e-07, + "loss": 0.6851, + "step": 728 + }, + { + "epoch": 0.03, + "learning_rate": 1.975134580876698e-07, + "loss": 0.7122, + "step": 729 + }, + { + "epoch": 0.03, + "learning_rate": 1.9750979602299777e-07, + "loss": 0.6995, + "step": 730 + }, + { + "epoch": 0.03, + "learning_rate": 1.975061339583257e-07, + "loss": 0.6968, + "step": 731 + }, + { + "epoch": 0.03, + "learning_rate": 1.9750247189365365e-07, + "loss": 0.7046, + "step": 732 + }, + { + "epoch": 0.03, + "learning_rate": 1.9749880982898157e-07, + "loss": 0.6974, + "step": 733 + }, + { + "epoch": 0.03, + "learning_rate": 1.974951477643095e-07, + "loss": 0.7122, + "step": 734 + }, + { + "epoch": 0.03, + "learning_rate": 1.9749148569963745e-07, + "loss": 0.7034, + "step": 735 + }, + { + "epoch": 0.03, + "learning_rate": 1.9748782363496538e-07, + "loss": 0.6994, + "step": 736 + }, + { + "epoch": 0.03, + "learning_rate": 1.9748416157029333e-07, + "loss": 0.7025, + "step": 737 + }, + { + "epoch": 0.03, + "learning_rate": 1.9748049950562126e-07, + "loss": 0.6954, + "step": 738 + }, + { + "epoch": 0.03, + "learning_rate": 1.974768374409492e-07, + "loss": 0.7089, + "step": 739 + }, + { + "epoch": 0.03, + "learning_rate": 1.9747317537627714e-07, + "loss": 0.6694, + "step": 740 + }, + { + "epoch": 0.03, + "learning_rate": 1.974695133116051e-07, + "loss": 0.6946, + "step": 741 + }, + { + "epoch": 0.03, + "learning_rate": 1.97465851246933e-07, + "loss": 0.6979, + "step": 742 + }, + { + "epoch": 0.03, + "learning_rate": 1.9746218918226095e-07, + "loss": 0.7105, + "step": 743 + }, + { + "epoch": 0.03, + "learning_rate": 1.9745852711758887e-07, + "loss": 0.6928, + "step": 744 + }, + { + "epoch": 0.03, + "learning_rate": 1.9745486505291683e-07, + "loss": 0.7022, + "step": 745 + }, + { + "epoch": 0.03, + "learning_rate": 1.9745120298824475e-07, + "loss": 0.7035, + "step": 746 + }, + { + "epoch": 0.03, + "learning_rate": 1.974475409235727e-07, + "loss": 0.6953, + "step": 747 + }, + { + "epoch": 0.03, + "learning_rate": 1.9744387885890063e-07, + "loss": 0.6967, + "step": 748 + }, + { + "epoch": 0.03, + "learning_rate": 1.9744021679422858e-07, + "loss": 0.7023, + "step": 749 + }, + { + "epoch": 0.03, + "learning_rate": 1.974365547295565e-07, + "loss": 0.6972, + "step": 750 + }, + { + "epoch": 0.03, + "learning_rate": 1.9743289266488446e-07, + "loss": 0.6911, + "step": 751 + }, + { + "epoch": 0.03, + "learning_rate": 1.974292306002124e-07, + "loss": 0.675, + "step": 752 + }, + { + "epoch": 0.03, + "learning_rate": 1.9742556853554034e-07, + "loss": 0.6989, + "step": 753 + }, + { + "epoch": 0.03, + "learning_rate": 1.9742190647086827e-07, + "loss": 0.6975, + "step": 754 + }, + { + "epoch": 0.03, + "learning_rate": 1.974182444061962e-07, + "loss": 0.6884, + "step": 755 + }, + { + "epoch": 0.03, + "learning_rate": 1.9741458234152415e-07, + "loss": 0.6943, + "step": 756 + }, + { + "epoch": 0.03, + "learning_rate": 1.9741092027685208e-07, + "loss": 0.6997, + "step": 757 + }, + { + "epoch": 0.03, + "learning_rate": 1.9740725821218003e-07, + "loss": 0.7079, + "step": 758 + }, + { + "epoch": 0.03, + "learning_rate": 1.9740359614750796e-07, + "loss": 0.707, + "step": 759 + }, + { + "epoch": 0.03, + "learning_rate": 1.9739993408283588e-07, + "loss": 0.723, + "step": 760 + }, + { + "epoch": 0.03, + "learning_rate": 1.973962720181638e-07, + "loss": 0.6885, + "step": 761 + }, + { + "epoch": 0.03, + "learning_rate": 1.9739260995349176e-07, + "loss": 0.7008, + "step": 762 + }, + { + "epoch": 0.03, + "learning_rate": 1.973889478888197e-07, + "loss": 0.7012, + "step": 763 + }, + { + "epoch": 0.03, + "learning_rate": 1.9738528582414764e-07, + "loss": 0.6948, + "step": 764 + }, + { + "epoch": 0.03, + "learning_rate": 1.9738162375947557e-07, + "loss": 0.7052, + "step": 765 + }, + { + "epoch": 0.03, + "learning_rate": 1.9737796169480352e-07, + "loss": 0.7277, + "step": 766 + }, + { + "epoch": 0.03, + "learning_rate": 1.9737429963013145e-07, + "loss": 0.6712, + "step": 767 + }, + { + "epoch": 0.03, + "learning_rate": 1.973706375654594e-07, + "loss": 0.6652, + "step": 768 + }, + { + "epoch": 0.03, + "learning_rate": 1.9736697550078733e-07, + "loss": 0.7104, + "step": 769 + }, + { + "epoch": 0.03, + "learning_rate": 1.9736331343611528e-07, + "loss": 0.687, + "step": 770 + }, + { + "epoch": 0.03, + "learning_rate": 1.973596513714432e-07, + "loss": 0.6896, + "step": 771 + }, + { + "epoch": 0.03, + "learning_rate": 1.9735598930677116e-07, + "loss": 0.7011, + "step": 772 + }, + { + "epoch": 0.03, + "learning_rate": 1.973523272420991e-07, + "loss": 0.7159, + "step": 773 + }, + { + "epoch": 0.03, + "learning_rate": 1.9734866517742704e-07, + "loss": 0.7093, + "step": 774 + }, + { + "epoch": 0.03, + "learning_rate": 1.9734500311275497e-07, + "loss": 0.692, + "step": 775 + }, + { + "epoch": 0.03, + "learning_rate": 1.973413410480829e-07, + "loss": 0.7042, + "step": 776 + }, + { + "epoch": 0.03, + "learning_rate": 1.9733767898341085e-07, + "loss": 0.7176, + "step": 777 + }, + { + "epoch": 0.03, + "learning_rate": 1.9733401691873878e-07, + "loss": 0.6805, + "step": 778 + }, + { + "epoch": 0.03, + "learning_rate": 1.973303548540667e-07, + "loss": 0.7086, + "step": 779 + }, + { + "epoch": 0.03, + "learning_rate": 1.9732669278939466e-07, + "loss": 0.7049, + "step": 780 + }, + { + "epoch": 0.03, + "learning_rate": 1.9732303072472258e-07, + "loss": 0.686, + "step": 781 + }, + { + "epoch": 0.03, + "learning_rate": 1.973193686600505e-07, + "loss": 0.6957, + "step": 782 + }, + { + "epoch": 0.03, + "learning_rate": 1.9731570659537846e-07, + "loss": 0.7156, + "step": 783 + }, + { + "epoch": 0.03, + "learning_rate": 1.973120445307064e-07, + "loss": 0.7022, + "step": 784 + }, + { + "epoch": 0.03, + "learning_rate": 1.9730838246603434e-07, + "loss": 0.6923, + "step": 785 + }, + { + "epoch": 0.03, + "learning_rate": 1.9730472040136227e-07, + "loss": 0.6939, + "step": 786 + }, + { + "epoch": 0.03, + "learning_rate": 1.9730105833669022e-07, + "loss": 0.7117, + "step": 787 + }, + { + "epoch": 0.03, + "learning_rate": 1.9729739627201815e-07, + "loss": 0.6952, + "step": 788 + }, + { + "epoch": 0.03, + "learning_rate": 1.972937342073461e-07, + "loss": 0.7107, + "step": 789 + }, + { + "epoch": 0.03, + "learning_rate": 1.9729007214267403e-07, + "loss": 0.6974, + "step": 790 + }, + { + "epoch": 0.03, + "learning_rate": 1.9728641007800198e-07, + "loss": 0.6907, + "step": 791 + }, + { + "epoch": 0.03, + "learning_rate": 1.972827480133299e-07, + "loss": 0.6965, + "step": 792 + }, + { + "epoch": 0.03, + "learning_rate": 1.9727908594865786e-07, + "loss": 0.6778, + "step": 793 + }, + { + "epoch": 0.03, + "learning_rate": 1.972754238839858e-07, + "loss": 0.6985, + "step": 794 + }, + { + "epoch": 0.03, + "learning_rate": 1.9727176181931374e-07, + "loss": 0.6986, + "step": 795 + }, + { + "epoch": 0.03, + "learning_rate": 1.9726809975464164e-07, + "loss": 0.6938, + "step": 796 + }, + { + "epoch": 0.03, + "learning_rate": 1.972644376899696e-07, + "loss": 0.6918, + "step": 797 + }, + { + "epoch": 0.03, + "learning_rate": 1.9726077562529752e-07, + "loss": 0.6903, + "step": 798 + }, + { + "epoch": 0.03, + "learning_rate": 1.9725711356062548e-07, + "loss": 0.6991, + "step": 799 + }, + { + "epoch": 0.03, + "learning_rate": 1.972534514959534e-07, + "loss": 0.691, + "step": 800 + }, + { + "epoch": 0.03, + "learning_rate": 1.9724978943128136e-07, + "loss": 0.6731, + "step": 801 + }, + { + "epoch": 0.03, + "learning_rate": 1.9724612736660928e-07, + "loss": 0.73, + "step": 802 + }, + { + "epoch": 0.03, + "learning_rate": 1.9724246530193724e-07, + "loss": 0.7013, + "step": 803 + }, + { + "epoch": 0.03, + "learning_rate": 1.9723880323726516e-07, + "loss": 0.692, + "step": 804 + }, + { + "epoch": 0.03, + "learning_rate": 1.972351411725931e-07, + "loss": 0.701, + "step": 805 + }, + { + "epoch": 0.03, + "learning_rate": 1.9723147910792104e-07, + "loss": 0.6762, + "step": 806 + }, + { + "epoch": 0.03, + "learning_rate": 1.9722781704324897e-07, + "loss": 0.7145, + "step": 807 + }, + { + "epoch": 0.03, + "learning_rate": 1.9722415497857692e-07, + "loss": 0.6878, + "step": 808 + }, + { + "epoch": 0.03, + "learning_rate": 1.9722049291390485e-07, + "loss": 0.6719, + "step": 809 + }, + { + "epoch": 0.03, + "learning_rate": 1.972168308492328e-07, + "loss": 0.6805, + "step": 810 + }, + { + "epoch": 0.03, + "learning_rate": 1.9721316878456073e-07, + "loss": 0.6961, + "step": 811 + }, + { + "epoch": 0.03, + "learning_rate": 1.9720950671988868e-07, + "loss": 0.6768, + "step": 812 + }, + { + "epoch": 0.03, + "learning_rate": 1.972058446552166e-07, + "loss": 0.693, + "step": 813 + }, + { + "epoch": 0.03, + "learning_rate": 1.9720218259054453e-07, + "loss": 0.7181, + "step": 814 + }, + { + "epoch": 0.03, + "learning_rate": 1.9719852052587246e-07, + "loss": 0.6966, + "step": 815 + }, + { + "epoch": 0.03, + "learning_rate": 1.9719485846120041e-07, + "loss": 0.6975, + "step": 816 + }, + { + "epoch": 0.03, + "learning_rate": 1.9719119639652834e-07, + "loss": 0.684, + "step": 817 + }, + { + "epoch": 0.03, + "learning_rate": 1.971875343318563e-07, + "loss": 0.6818, + "step": 818 + }, + { + "epoch": 0.03, + "learning_rate": 1.9718387226718422e-07, + "loss": 0.7059, + "step": 819 + }, + { + "epoch": 0.03, + "learning_rate": 1.9718021020251217e-07, + "loss": 0.6893, + "step": 820 + }, + { + "epoch": 0.03, + "learning_rate": 1.971765481378401e-07, + "loss": 0.7128, + "step": 821 + }, + { + "epoch": 0.03, + "learning_rate": 1.9717288607316805e-07, + "loss": 0.6928, + "step": 822 + }, + { + "epoch": 0.03, + "learning_rate": 1.9716922400849598e-07, + "loss": 0.7009, + "step": 823 + }, + { + "epoch": 0.03, + "learning_rate": 1.9716556194382393e-07, + "loss": 0.6917, + "step": 824 + }, + { + "epoch": 0.03, + "learning_rate": 1.9716189987915186e-07, + "loss": 0.6848, + "step": 825 + }, + { + "epoch": 0.03, + "learning_rate": 1.971582378144798e-07, + "loss": 0.6901, + "step": 826 + }, + { + "epoch": 0.03, + "learning_rate": 1.9715457574980774e-07, + "loss": 0.701, + "step": 827 + }, + { + "epoch": 0.03, + "learning_rate": 1.9715091368513567e-07, + "loss": 0.6815, + "step": 828 + }, + { + "epoch": 0.03, + "learning_rate": 1.9714725162046362e-07, + "loss": 0.6857, + "step": 829 + }, + { + "epoch": 0.03, + "learning_rate": 1.9714358955579155e-07, + "loss": 0.7113, + "step": 830 + }, + { + "epoch": 0.03, + "learning_rate": 1.9713992749111947e-07, + "loss": 0.6855, + "step": 831 + }, + { + "epoch": 0.03, + "learning_rate": 1.971362654264474e-07, + "loss": 0.6738, + "step": 832 + }, + { + "epoch": 0.03, + "learning_rate": 1.9713260336177535e-07, + "loss": 0.7093, + "step": 833 + }, + { + "epoch": 0.03, + "learning_rate": 1.9712894129710328e-07, + "loss": 0.6709, + "step": 834 + }, + { + "epoch": 0.03, + "learning_rate": 1.9712527923243123e-07, + "loss": 0.7017, + "step": 835 + }, + { + "epoch": 0.03, + "learning_rate": 1.9712161716775916e-07, + "loss": 0.6849, + "step": 836 + }, + { + "epoch": 0.03, + "learning_rate": 1.9711795510308711e-07, + "loss": 0.7024, + "step": 837 + }, + { + "epoch": 0.03, + "learning_rate": 1.9711429303841504e-07, + "loss": 0.6952, + "step": 838 + }, + { + "epoch": 0.03, + "learning_rate": 1.97110630973743e-07, + "loss": 0.6947, + "step": 839 + }, + { + "epoch": 0.03, + "learning_rate": 1.9710696890907092e-07, + "loss": 0.6768, + "step": 840 + }, + { + "epoch": 0.03, + "learning_rate": 1.9710330684439887e-07, + "loss": 0.692, + "step": 841 + }, + { + "epoch": 0.03, + "learning_rate": 1.970996447797268e-07, + "loss": 0.6569, + "step": 842 + }, + { + "epoch": 0.03, + "learning_rate": 1.9709598271505475e-07, + "loss": 0.6805, + "step": 843 + }, + { + "epoch": 0.03, + "learning_rate": 1.9709232065038268e-07, + "loss": 0.7049, + "step": 844 + }, + { + "epoch": 0.03, + "learning_rate": 1.9708865858571063e-07, + "loss": 0.7042, + "step": 845 + }, + { + "epoch": 0.03, + "learning_rate": 1.9708499652103856e-07, + "loss": 0.6922, + "step": 846 + }, + { + "epoch": 0.03, + "learning_rate": 1.9708133445636649e-07, + "loss": 0.6899, + "step": 847 + }, + { + "epoch": 0.03, + "learning_rate": 1.9707767239169444e-07, + "loss": 0.6892, + "step": 848 + }, + { + "epoch": 0.03, + "learning_rate": 1.9707401032702237e-07, + "loss": 0.7056, + "step": 849 + }, + { + "epoch": 0.03, + "learning_rate": 1.970703482623503e-07, + "loss": 0.7173, + "step": 850 + }, + { + "epoch": 0.03, + "learning_rate": 1.9706668619767825e-07, + "loss": 0.6961, + "step": 851 + }, + { + "epoch": 0.03, + "learning_rate": 1.9706302413300617e-07, + "loss": 0.7126, + "step": 852 + }, + { + "epoch": 0.03, + "learning_rate": 1.970593620683341e-07, + "loss": 0.7024, + "step": 853 + }, + { + "epoch": 0.03, + "learning_rate": 1.9705570000366205e-07, + "loss": 0.6665, + "step": 854 + }, + { + "epoch": 0.03, + "learning_rate": 1.9705203793898998e-07, + "loss": 0.6963, + "step": 855 + }, + { + "epoch": 0.03, + "learning_rate": 1.9704837587431793e-07, + "loss": 0.6795, + "step": 856 + }, + { + "epoch": 0.03, + "learning_rate": 1.9704471380964586e-07, + "loss": 0.6655, + "step": 857 + }, + { + "epoch": 0.03, + "learning_rate": 1.970410517449738e-07, + "loss": 0.6983, + "step": 858 + }, + { + "epoch": 0.03, + "learning_rate": 1.9703738968030174e-07, + "loss": 0.6903, + "step": 859 + }, + { + "epoch": 0.03, + "learning_rate": 1.970337276156297e-07, + "loss": 0.692, + "step": 860 + }, + { + "epoch": 0.03, + "learning_rate": 1.9703006555095762e-07, + "loss": 0.7077, + "step": 861 + }, + { + "epoch": 0.03, + "learning_rate": 1.9702640348628557e-07, + "loss": 0.6926, + "step": 862 + }, + { + "epoch": 0.03, + "learning_rate": 1.970227414216135e-07, + "loss": 0.6928, + "step": 863 + }, + { + "epoch": 0.03, + "learning_rate": 1.9701907935694145e-07, + "loss": 0.7069, + "step": 864 + }, + { + "epoch": 0.03, + "learning_rate": 1.9701541729226938e-07, + "loss": 0.6858, + "step": 865 + }, + { + "epoch": 0.03, + "learning_rate": 1.9701175522759733e-07, + "loss": 0.6855, + "step": 866 + }, + { + "epoch": 0.03, + "learning_rate": 1.9700809316292523e-07, + "loss": 0.7041, + "step": 867 + }, + { + "epoch": 0.03, + "learning_rate": 1.9700443109825318e-07, + "loss": 0.7017, + "step": 868 + }, + { + "epoch": 0.03, + "learning_rate": 1.970007690335811e-07, + "loss": 0.6784, + "step": 869 + }, + { + "epoch": 0.03, + "learning_rate": 1.9699710696890906e-07, + "loss": 0.7009, + "step": 870 + }, + { + "epoch": 0.03, + "learning_rate": 1.96993444904237e-07, + "loss": 0.6793, + "step": 871 + }, + { + "epoch": 0.03, + "learning_rate": 1.9698978283956494e-07, + "loss": 0.716, + "step": 872 + }, + { + "epoch": 0.03, + "learning_rate": 1.9698612077489287e-07, + "loss": 0.6873, + "step": 873 + }, + { + "epoch": 0.03, + "learning_rate": 1.969824587102208e-07, + "loss": 0.7119, + "step": 874 + }, + { + "epoch": 0.03, + "learning_rate": 1.9697879664554875e-07, + "loss": 0.6594, + "step": 875 + }, + { + "epoch": 0.03, + "learning_rate": 1.9697513458087668e-07, + "loss": 0.6691, + "step": 876 + }, + { + "epoch": 0.03, + "learning_rate": 1.9697147251620463e-07, + "loss": 0.7222, + "step": 877 + }, + { + "epoch": 0.03, + "learning_rate": 1.9696781045153256e-07, + "loss": 0.6867, + "step": 878 + }, + { + "epoch": 0.03, + "learning_rate": 1.969641483868605e-07, + "loss": 0.7033, + "step": 879 + }, + { + "epoch": 0.03, + "learning_rate": 1.9696048632218844e-07, + "loss": 0.7012, + "step": 880 + }, + { + "epoch": 0.03, + "learning_rate": 1.969568242575164e-07, + "loss": 0.7051, + "step": 881 + }, + { + "epoch": 0.03, + "learning_rate": 1.9695316219284432e-07, + "loss": 0.6924, + "step": 882 + }, + { + "epoch": 0.03, + "learning_rate": 1.9694950012817227e-07, + "loss": 0.689, + "step": 883 + }, + { + "epoch": 0.03, + "learning_rate": 1.969458380635002e-07, + "loss": 0.6899, + "step": 884 + }, + { + "epoch": 0.03, + "learning_rate": 1.9694217599882812e-07, + "loss": 0.7118, + "step": 885 + }, + { + "epoch": 0.03, + "learning_rate": 1.9693851393415605e-07, + "loss": 0.721, + "step": 886 + }, + { + "epoch": 0.03, + "learning_rate": 1.96934851869484e-07, + "loss": 0.6911, + "step": 887 + }, + { + "epoch": 0.03, + "learning_rate": 1.9693118980481193e-07, + "loss": 0.6915, + "step": 888 + }, + { + "epoch": 0.03, + "learning_rate": 1.9692752774013988e-07, + "loss": 0.6954, + "step": 889 + }, + { + "epoch": 0.03, + "learning_rate": 1.969238656754678e-07, + "loss": 0.6954, + "step": 890 + }, + { + "epoch": 0.03, + "learning_rate": 1.9692020361079576e-07, + "loss": 0.7031, + "step": 891 + }, + { + "epoch": 0.03, + "learning_rate": 1.969165415461237e-07, + "loss": 0.7026, + "step": 892 + }, + { + "epoch": 0.03, + "learning_rate": 1.9691287948145164e-07, + "loss": 0.6871, + "step": 893 + }, + { + "epoch": 0.03, + "learning_rate": 1.9690921741677957e-07, + "loss": 0.7032, + "step": 894 + }, + { + "epoch": 0.03, + "learning_rate": 1.9690555535210752e-07, + "loss": 0.7111, + "step": 895 + }, + { + "epoch": 0.03, + "learning_rate": 1.9690189328743545e-07, + "loss": 0.6963, + "step": 896 + }, + { + "epoch": 0.03, + "learning_rate": 1.9689823122276338e-07, + "loss": 0.6751, + "step": 897 + }, + { + "epoch": 0.03, + "learning_rate": 1.9689456915809133e-07, + "loss": 0.6919, + "step": 898 + }, + { + "epoch": 0.03, + "learning_rate": 1.9689090709341926e-07, + "loss": 0.6992, + "step": 899 + }, + { + "epoch": 0.03, + "learning_rate": 1.968872450287472e-07, + "loss": 0.707, + "step": 900 + }, + { + "epoch": 0.03, + "learning_rate": 1.9688358296407514e-07, + "loss": 0.6685, + "step": 901 + }, + { + "epoch": 0.03, + "learning_rate": 1.968799208994031e-07, + "loss": 0.6995, + "step": 902 + }, + { + "epoch": 0.03, + "learning_rate": 1.96876258834731e-07, + "loss": 0.6917, + "step": 903 + }, + { + "epoch": 0.03, + "learning_rate": 1.9687259677005894e-07, + "loss": 0.707, + "step": 904 + }, + { + "epoch": 0.03, + "learning_rate": 1.9686893470538687e-07, + "loss": 0.7055, + "step": 905 + }, + { + "epoch": 0.03, + "learning_rate": 1.9686527264071482e-07, + "loss": 0.7192, + "step": 906 + }, + { + "epoch": 0.03, + "learning_rate": 1.9686161057604275e-07, + "loss": 0.7094, + "step": 907 + }, + { + "epoch": 0.03, + "learning_rate": 1.968579485113707e-07, + "loss": 0.7279, + "step": 908 + }, + { + "epoch": 0.03, + "learning_rate": 1.9685428644669863e-07, + "loss": 0.6916, + "step": 909 + }, + { + "epoch": 0.03, + "learning_rate": 1.9685062438202658e-07, + "loss": 0.698, + "step": 910 + }, + { + "epoch": 0.03, + "learning_rate": 1.968469623173545e-07, + "loss": 0.6944, + "step": 911 + }, + { + "epoch": 0.03, + "learning_rate": 1.9684330025268246e-07, + "loss": 0.6907, + "step": 912 + }, + { + "epoch": 0.03, + "learning_rate": 1.968396381880104e-07, + "loss": 0.6919, + "step": 913 + }, + { + "epoch": 0.03, + "learning_rate": 1.9683597612333834e-07, + "loss": 0.7019, + "step": 914 + }, + { + "epoch": 0.03, + "learning_rate": 1.9683231405866627e-07, + "loss": 0.6847, + "step": 915 + }, + { + "epoch": 0.03, + "learning_rate": 1.9682865199399422e-07, + "loss": 0.734, + "step": 916 + }, + { + "epoch": 0.03, + "learning_rate": 1.9682498992932215e-07, + "loss": 0.6853, + "step": 917 + }, + { + "epoch": 0.03, + "learning_rate": 1.9682132786465008e-07, + "loss": 0.6646, + "step": 918 + }, + { + "epoch": 0.03, + "learning_rate": 1.9681766579997803e-07, + "loss": 0.6777, + "step": 919 + }, + { + "epoch": 0.03, + "learning_rate": 1.9681400373530596e-07, + "loss": 0.6917, + "step": 920 + }, + { + "epoch": 0.03, + "learning_rate": 1.9681034167063388e-07, + "loss": 0.7013, + "step": 921 + }, + { + "epoch": 0.03, + "learning_rate": 1.9680667960596184e-07, + "loss": 0.7125, + "step": 922 + }, + { + "epoch": 0.03, + "learning_rate": 1.9680301754128976e-07, + "loss": 0.6928, + "step": 923 + }, + { + "epoch": 0.03, + "learning_rate": 1.967993554766177e-07, + "loss": 0.6843, + "step": 924 + }, + { + "epoch": 0.03, + "learning_rate": 1.9679569341194564e-07, + "loss": 0.6812, + "step": 925 + }, + { + "epoch": 0.03, + "learning_rate": 1.9679203134727357e-07, + "loss": 0.7075, + "step": 926 + }, + { + "epoch": 0.03, + "learning_rate": 1.9678836928260152e-07, + "loss": 0.6858, + "step": 927 + }, + { + "epoch": 0.03, + "learning_rate": 1.9678470721792945e-07, + "loss": 0.7253, + "step": 928 + }, + { + "epoch": 0.03, + "learning_rate": 1.967810451532574e-07, + "loss": 0.6748, + "step": 929 + }, + { + "epoch": 0.03, + "learning_rate": 1.9677738308858533e-07, + "loss": 0.6943, + "step": 930 + }, + { + "epoch": 0.03, + "learning_rate": 1.9677372102391328e-07, + "loss": 0.6657, + "step": 931 + }, + { + "epoch": 0.03, + "learning_rate": 1.967700589592412e-07, + "loss": 0.6825, + "step": 932 + }, + { + "epoch": 0.03, + "learning_rate": 1.9676639689456916e-07, + "loss": 0.6932, + "step": 933 + }, + { + "epoch": 0.03, + "learning_rate": 1.967627348298971e-07, + "loss": 0.6833, + "step": 934 + }, + { + "epoch": 0.03, + "learning_rate": 1.9675907276522504e-07, + "loss": 0.6995, + "step": 935 + }, + { + "epoch": 0.03, + "learning_rate": 1.9675541070055297e-07, + "loss": 0.7142, + "step": 936 + }, + { + "epoch": 0.03, + "learning_rate": 1.9675174863588092e-07, + "loss": 0.7068, + "step": 937 + }, + { + "epoch": 0.03, + "learning_rate": 1.9674808657120885e-07, + "loss": 0.6998, + "step": 938 + }, + { + "epoch": 0.03, + "learning_rate": 1.9674442450653677e-07, + "loss": 0.6861, + "step": 939 + }, + { + "epoch": 0.03, + "learning_rate": 1.967407624418647e-07, + "loss": 0.6867, + "step": 940 + }, + { + "epoch": 0.03, + "learning_rate": 1.9673710037719265e-07, + "loss": 0.6966, + "step": 941 + }, + { + "epoch": 0.03, + "learning_rate": 1.9673343831252058e-07, + "loss": 0.7177, + "step": 942 + }, + { + "epoch": 0.03, + "learning_rate": 1.9672977624784853e-07, + "loss": 0.6979, + "step": 943 + }, + { + "epoch": 0.03, + "learning_rate": 1.9672611418317646e-07, + "loss": 0.7079, + "step": 944 + }, + { + "epoch": 0.03, + "learning_rate": 1.967224521185044e-07, + "loss": 0.6708, + "step": 945 + }, + { + "epoch": 0.03, + "learning_rate": 1.9671879005383234e-07, + "loss": 0.7005, + "step": 946 + }, + { + "epoch": 0.03, + "learning_rate": 1.9671512798916027e-07, + "loss": 0.6893, + "step": 947 + }, + { + "epoch": 0.03, + "learning_rate": 1.9671146592448822e-07, + "loss": 0.6928, + "step": 948 + }, + { + "epoch": 0.03, + "learning_rate": 1.9670780385981615e-07, + "loss": 0.6953, + "step": 949 + }, + { + "epoch": 0.03, + "learning_rate": 1.967041417951441e-07, + "loss": 0.7036, + "step": 950 + }, + { + "epoch": 0.03, + "learning_rate": 1.9670047973047203e-07, + "loss": 0.6926, + "step": 951 + }, + { + "epoch": 0.03, + "learning_rate": 1.9669681766579998e-07, + "loss": 0.6978, + "step": 952 + }, + { + "epoch": 0.03, + "learning_rate": 1.966931556011279e-07, + "loss": 0.7097, + "step": 953 + }, + { + "epoch": 0.03, + "learning_rate": 1.9668949353645586e-07, + "loss": 0.7118, + "step": 954 + }, + { + "epoch": 0.03, + "learning_rate": 1.9668583147178379e-07, + "loss": 0.6684, + "step": 955 + }, + { + "epoch": 0.03, + "learning_rate": 1.9668216940711174e-07, + "loss": 0.6844, + "step": 956 + }, + { + "epoch": 0.04, + "learning_rate": 1.9667850734243964e-07, + "loss": 0.67, + "step": 957 + }, + { + "epoch": 0.04, + "learning_rate": 1.966748452777676e-07, + "loss": 0.6906, + "step": 958 + }, + { + "epoch": 0.04, + "learning_rate": 1.9667118321309552e-07, + "loss": 0.7022, + "step": 959 + }, + { + "epoch": 0.04, + "learning_rate": 1.9666752114842347e-07, + "loss": 0.6902, + "step": 960 + }, + { + "epoch": 0.04, + "learning_rate": 1.966638590837514e-07, + "loss": 0.701, + "step": 961 + }, + { + "epoch": 0.04, + "learning_rate": 1.9666019701907935e-07, + "loss": 0.6998, + "step": 962 + }, + { + "epoch": 0.04, + "learning_rate": 1.9665653495440728e-07, + "loss": 0.715, + "step": 963 + }, + { + "epoch": 0.04, + "learning_rate": 1.9665287288973523e-07, + "loss": 0.6898, + "step": 964 + }, + { + "epoch": 0.04, + "learning_rate": 1.9664921082506316e-07, + "loss": 0.7143, + "step": 965 + }, + { + "epoch": 0.04, + "learning_rate": 1.966455487603911e-07, + "loss": 0.6817, + "step": 966 + }, + { + "epoch": 0.04, + "learning_rate": 1.9664188669571904e-07, + "loss": 0.6718, + "step": 967 + }, + { + "epoch": 0.04, + "learning_rate": 1.9663822463104697e-07, + "loss": 0.684, + "step": 968 + }, + { + "epoch": 0.04, + "learning_rate": 1.9663456256637492e-07, + "loss": 0.7029, + "step": 969 + }, + { + "epoch": 0.04, + "learning_rate": 1.9663090050170285e-07, + "loss": 0.7002, + "step": 970 + }, + { + "epoch": 0.04, + "learning_rate": 1.966272384370308e-07, + "loss": 0.6681, + "step": 971 + }, + { + "epoch": 0.04, + "learning_rate": 1.9662357637235873e-07, + "loss": 0.7039, + "step": 972 + }, + { + "epoch": 0.04, + "learning_rate": 1.9661991430768668e-07, + "loss": 0.6822, + "step": 973 + }, + { + "epoch": 0.04, + "learning_rate": 1.966162522430146e-07, + "loss": 0.6814, + "step": 974 + }, + { + "epoch": 0.04, + "learning_rate": 1.9661259017834253e-07, + "loss": 0.6833, + "step": 975 + }, + { + "epoch": 0.04, + "learning_rate": 1.9660892811367046e-07, + "loss": 0.6892, + "step": 976 + }, + { + "epoch": 0.04, + "learning_rate": 1.966052660489984e-07, + "loss": 0.6957, + "step": 977 + }, + { + "epoch": 0.04, + "learning_rate": 1.9660160398432634e-07, + "loss": 0.6809, + "step": 978 + }, + { + "epoch": 0.04, + "learning_rate": 1.965979419196543e-07, + "loss": 0.7015, + "step": 979 + }, + { + "epoch": 0.04, + "learning_rate": 1.9659427985498222e-07, + "loss": 0.6917, + "step": 980 + }, + { + "epoch": 0.04, + "learning_rate": 1.9659061779031017e-07, + "loss": 0.7093, + "step": 981 + }, + { + "epoch": 0.04, + "learning_rate": 1.965869557256381e-07, + "loss": 0.7077, + "step": 982 + }, + { + "epoch": 0.04, + "learning_rate": 1.9658329366096605e-07, + "loss": 0.6736, + "step": 983 + }, + { + "epoch": 0.04, + "learning_rate": 1.9657963159629398e-07, + "loss": 0.6912, + "step": 984 + }, + { + "epoch": 0.04, + "learning_rate": 1.9657596953162193e-07, + "loss": 0.6961, + "step": 985 + }, + { + "epoch": 0.04, + "learning_rate": 1.9657230746694986e-07, + "loss": 0.6789, + "step": 986 + }, + { + "epoch": 0.04, + "learning_rate": 1.965686454022778e-07, + "loss": 0.6941, + "step": 987 + }, + { + "epoch": 0.04, + "learning_rate": 1.9656498333760574e-07, + "loss": 0.6807, + "step": 988 + }, + { + "epoch": 0.04, + "learning_rate": 1.9656132127293366e-07, + "loss": 0.6904, + "step": 989 + }, + { + "epoch": 0.04, + "learning_rate": 1.9655765920826162e-07, + "loss": 0.7074, + "step": 990 + }, + { + "epoch": 0.04, + "learning_rate": 1.9655399714358954e-07, + "loss": 0.713, + "step": 991 + }, + { + "epoch": 0.04, + "learning_rate": 1.965503350789175e-07, + "loss": 0.6764, + "step": 992 + }, + { + "epoch": 0.04, + "learning_rate": 1.9654667301424542e-07, + "loss": 0.6918, + "step": 993 + }, + { + "epoch": 0.04, + "learning_rate": 1.9654301094957335e-07, + "loss": 0.6674, + "step": 994 + }, + { + "epoch": 0.04, + "learning_rate": 1.9653934888490128e-07, + "loss": 0.6916, + "step": 995 + }, + { + "epoch": 0.04, + "learning_rate": 1.9653568682022923e-07, + "loss": 0.6792, + "step": 996 + }, + { + "epoch": 0.04, + "learning_rate": 1.9653202475555716e-07, + "loss": 0.6798, + "step": 997 + }, + { + "epoch": 0.04, + "learning_rate": 1.965283626908851e-07, + "loss": 0.7085, + "step": 998 + }, + { + "epoch": 0.04, + "learning_rate": 1.9652470062621304e-07, + "loss": 0.681, + "step": 999 + }, + { + "epoch": 0.04, + "learning_rate": 1.96521038561541e-07, + "loss": 0.685, + "step": 1000 + }, + { + "epoch": 0.04, + "learning_rate": 1.9651737649686892e-07, + "loss": 0.6935, + "step": 1001 + }, + { + "epoch": 0.04, + "learning_rate": 1.9651371443219687e-07, + "loss": 0.6983, + "step": 1002 + }, + { + "epoch": 0.04, + "learning_rate": 1.965100523675248e-07, + "loss": 0.7134, + "step": 1003 + }, + { + "epoch": 0.04, + "learning_rate": 1.9650639030285275e-07, + "loss": 0.691, + "step": 1004 + }, + { + "epoch": 0.04, + "learning_rate": 1.9650272823818068e-07, + "loss": 0.7154, + "step": 1005 + }, + { + "epoch": 0.04, + "learning_rate": 1.9649906617350863e-07, + "loss": 0.7043, + "step": 1006 + }, + { + "epoch": 0.04, + "learning_rate": 1.9649540410883656e-07, + "loss": 0.7039, + "step": 1007 + }, + { + "epoch": 0.04, + "learning_rate": 1.964917420441645e-07, + "loss": 0.6869, + "step": 1008 + }, + { + "epoch": 0.04, + "learning_rate": 1.9648807997949244e-07, + "loss": 0.6847, + "step": 1009 + }, + { + "epoch": 0.04, + "learning_rate": 1.964844179148204e-07, + "loss": 0.7168, + "step": 1010 + }, + { + "epoch": 0.04, + "learning_rate": 1.964807558501483e-07, + "loss": 0.6655, + "step": 1011 + }, + { + "epoch": 0.04, + "learning_rate": 1.9647709378547624e-07, + "loss": 0.6754, + "step": 1012 + }, + { + "epoch": 0.04, + "learning_rate": 1.9647343172080417e-07, + "loss": 0.664, + "step": 1013 + }, + { + "epoch": 0.04, + "learning_rate": 1.9646976965613212e-07, + "loss": 0.6862, + "step": 1014 + }, + { + "epoch": 0.04, + "learning_rate": 1.9646610759146005e-07, + "loss": 0.6951, + "step": 1015 + }, + { + "epoch": 0.04, + "learning_rate": 1.9646244552678798e-07, + "loss": 0.6849, + "step": 1016 + }, + { + "epoch": 0.04, + "learning_rate": 1.9645878346211593e-07, + "loss": 0.7114, + "step": 1017 + }, + { + "epoch": 0.04, + "learning_rate": 1.9645512139744386e-07, + "loss": 0.7196, + "step": 1018 + }, + { + "epoch": 0.04, + "learning_rate": 1.964514593327718e-07, + "loss": 0.6905, + "step": 1019 + }, + { + "epoch": 0.04, + "learning_rate": 1.9644779726809974e-07, + "loss": 0.6829, + "step": 1020 + }, + { + "epoch": 0.04, + "learning_rate": 1.964441352034277e-07, + "loss": 0.6866, + "step": 1021 + }, + { + "epoch": 0.04, + "learning_rate": 1.9644047313875562e-07, + "loss": 0.7017, + "step": 1022 + }, + { + "epoch": 0.04, + "learning_rate": 1.9643681107408357e-07, + "loss": 0.6926, + "step": 1023 + }, + { + "epoch": 0.04, + "learning_rate": 1.964331490094115e-07, + "loss": 0.6767, + "step": 1024 + }, + { + "epoch": 0.04, + "learning_rate": 1.9642948694473945e-07, + "loss": 0.6846, + "step": 1025 + }, + { + "epoch": 0.04, + "learning_rate": 1.9642582488006738e-07, + "loss": 0.7096, + "step": 1026 + }, + { + "epoch": 0.04, + "learning_rate": 1.9642216281539533e-07, + "loss": 0.6853, + "step": 1027 + }, + { + "epoch": 0.04, + "learning_rate": 1.9641850075072326e-07, + "loss": 0.6919, + "step": 1028 + }, + { + "epoch": 0.04, + "learning_rate": 1.9641483868605118e-07, + "loss": 0.6857, + "step": 1029 + }, + { + "epoch": 0.04, + "learning_rate": 1.964111766213791e-07, + "loss": 0.6998, + "step": 1030 + }, + { + "epoch": 0.04, + "learning_rate": 1.9640751455670706e-07, + "loss": 0.704, + "step": 1031 + }, + { + "epoch": 0.04, + "learning_rate": 1.96403852492035e-07, + "loss": 0.6679, + "step": 1032 + }, + { + "epoch": 0.04, + "learning_rate": 1.9640019042736294e-07, + "loss": 0.6977, + "step": 1033 + }, + { + "epoch": 0.04, + "learning_rate": 1.9639652836269087e-07, + "loss": 0.6967, + "step": 1034 + }, + { + "epoch": 0.04, + "learning_rate": 1.9639286629801882e-07, + "loss": 0.6892, + "step": 1035 + }, + { + "epoch": 0.04, + "learning_rate": 1.9638920423334675e-07, + "loss": 0.6963, + "step": 1036 + }, + { + "epoch": 0.04, + "learning_rate": 1.963855421686747e-07, + "loss": 0.6994, + "step": 1037 + }, + { + "epoch": 0.04, + "learning_rate": 1.9638188010400263e-07, + "loss": 0.7255, + "step": 1038 + }, + { + "epoch": 0.04, + "learning_rate": 1.9637821803933056e-07, + "loss": 0.7, + "step": 1039 + }, + { + "epoch": 0.04, + "learning_rate": 1.963745559746585e-07, + "loss": 0.6924, + "step": 1040 + }, + { + "epoch": 0.04, + "learning_rate": 1.9637089390998644e-07, + "loss": 0.6902, + "step": 1041 + }, + { + "epoch": 0.04, + "learning_rate": 1.963672318453144e-07, + "loss": 0.6834, + "step": 1042 + }, + { + "epoch": 0.04, + "learning_rate": 1.9636356978064232e-07, + "loss": 0.6994, + "step": 1043 + }, + { + "epoch": 0.04, + "learning_rate": 1.9635990771597027e-07, + "loss": 0.7091, + "step": 1044 + }, + { + "epoch": 0.04, + "learning_rate": 1.963562456512982e-07, + "loss": 0.6935, + "step": 1045 + }, + { + "epoch": 0.04, + "learning_rate": 1.9635258358662615e-07, + "loss": 0.6925, + "step": 1046 + }, + { + "epoch": 0.04, + "learning_rate": 1.9634892152195405e-07, + "loss": 0.7061, + "step": 1047 + }, + { + "epoch": 0.04, + "learning_rate": 1.96345259457282e-07, + "loss": 0.6661, + "step": 1048 + }, + { + "epoch": 0.04, + "learning_rate": 1.9634159739260993e-07, + "loss": 0.7017, + "step": 1049 + }, + { + "epoch": 0.04, + "learning_rate": 1.9633793532793788e-07, + "loss": 0.6884, + "step": 1050 + }, + { + "epoch": 0.04, + "learning_rate": 1.963342732632658e-07, + "loss": 0.6933, + "step": 1051 + }, + { + "epoch": 0.04, + "learning_rate": 1.9633061119859376e-07, + "loss": 0.6912, + "step": 1052 + }, + { + "epoch": 0.04, + "learning_rate": 1.963269491339217e-07, + "loss": 0.697, + "step": 1053 + }, + { + "epoch": 0.04, + "learning_rate": 1.9632328706924964e-07, + "loss": 0.6864, + "step": 1054 + }, + { + "epoch": 0.04, + "learning_rate": 1.9631962500457757e-07, + "loss": 0.6984, + "step": 1055 + }, + { + "epoch": 0.04, + "learning_rate": 1.9631596293990552e-07, + "loss": 0.69, + "step": 1056 + }, + { + "epoch": 0.04, + "learning_rate": 1.9631230087523345e-07, + "loss": 0.6805, + "step": 1057 + }, + { + "epoch": 0.04, + "learning_rate": 1.963086388105614e-07, + "loss": 0.6915, + "step": 1058 + }, + { + "epoch": 0.04, + "learning_rate": 1.9630497674588933e-07, + "loss": 0.692, + "step": 1059 + }, + { + "epoch": 0.04, + "learning_rate": 1.9630131468121725e-07, + "loss": 0.7195, + "step": 1060 + }, + { + "epoch": 0.04, + "learning_rate": 1.962976526165452e-07, + "loss": 0.7066, + "step": 1061 + }, + { + "epoch": 0.04, + "learning_rate": 1.9629399055187313e-07, + "loss": 0.6942, + "step": 1062 + }, + { + "epoch": 0.04, + "learning_rate": 1.962903284872011e-07, + "loss": 0.6949, + "step": 1063 + }, + { + "epoch": 0.04, + "learning_rate": 1.9628666642252901e-07, + "loss": 0.6834, + "step": 1064 + }, + { + "epoch": 0.04, + "learning_rate": 1.9628300435785694e-07, + "loss": 0.7237, + "step": 1065 + }, + { + "epoch": 0.04, + "learning_rate": 1.9627934229318487e-07, + "loss": 0.6891, + "step": 1066 + }, + { + "epoch": 0.04, + "learning_rate": 1.9627568022851282e-07, + "loss": 0.6896, + "step": 1067 + }, + { + "epoch": 0.04, + "learning_rate": 1.9627201816384075e-07, + "loss": 0.6957, + "step": 1068 + }, + { + "epoch": 0.04, + "learning_rate": 1.962683560991687e-07, + "loss": 0.6984, + "step": 1069 + }, + { + "epoch": 0.04, + "learning_rate": 1.9626469403449663e-07, + "loss": 0.7021, + "step": 1070 + }, + { + "epoch": 0.04, + "learning_rate": 1.9626103196982458e-07, + "loss": 0.6896, + "step": 1071 + }, + { + "epoch": 0.04, + "learning_rate": 1.962573699051525e-07, + "loss": 0.7002, + "step": 1072 + }, + { + "epoch": 0.04, + "learning_rate": 1.9625370784048046e-07, + "loss": 0.6898, + "step": 1073 + }, + { + "epoch": 0.04, + "learning_rate": 1.9625004577580839e-07, + "loss": 0.7199, + "step": 1074 + }, + { + "epoch": 0.04, + "learning_rate": 1.9624638371113634e-07, + "loss": 0.6966, + "step": 1075 + }, + { + "epoch": 0.04, + "learning_rate": 1.9624272164646427e-07, + "loss": 0.6689, + "step": 1076 + }, + { + "epoch": 0.04, + "learning_rate": 1.9623905958179222e-07, + "loss": 0.6904, + "step": 1077 + }, + { + "epoch": 0.04, + "learning_rate": 1.9623539751712015e-07, + "loss": 0.7002, + "step": 1078 + }, + { + "epoch": 0.04, + "learning_rate": 1.962317354524481e-07, + "loss": 0.6896, + "step": 1079 + }, + { + "epoch": 0.04, + "learning_rate": 1.9622807338777603e-07, + "loss": 0.6999, + "step": 1080 + }, + { + "epoch": 0.04, + "learning_rate": 1.9622441132310398e-07, + "loss": 0.7231, + "step": 1081 + }, + { + "epoch": 0.04, + "learning_rate": 1.962207492584319e-07, + "loss": 0.6903, + "step": 1082 + }, + { + "epoch": 0.04, + "learning_rate": 1.9621708719375983e-07, + "loss": 0.6978, + "step": 1083 + }, + { + "epoch": 0.04, + "learning_rate": 1.9621342512908776e-07, + "loss": 0.6837, + "step": 1084 + }, + { + "epoch": 0.04, + "learning_rate": 1.962097630644157e-07, + "loss": 0.6922, + "step": 1085 + }, + { + "epoch": 0.04, + "learning_rate": 1.9620610099974364e-07, + "loss": 0.6732, + "step": 1086 + }, + { + "epoch": 0.04, + "learning_rate": 1.9620243893507157e-07, + "loss": 0.6804, + "step": 1087 + }, + { + "epoch": 0.04, + "learning_rate": 1.9619877687039952e-07, + "loss": 0.7133, + "step": 1088 + }, + { + "epoch": 0.04, + "learning_rate": 1.9619511480572745e-07, + "loss": 0.6842, + "step": 1089 + }, + { + "epoch": 0.04, + "learning_rate": 1.961914527410554e-07, + "loss": 0.6857, + "step": 1090 + }, + { + "epoch": 0.04, + "learning_rate": 1.9618779067638333e-07, + "loss": 0.7021, + "step": 1091 + }, + { + "epoch": 0.04, + "learning_rate": 1.9618412861171128e-07, + "loss": 0.7034, + "step": 1092 + }, + { + "epoch": 0.04, + "learning_rate": 1.961804665470392e-07, + "loss": 0.6846, + "step": 1093 + }, + { + "epoch": 0.04, + "learning_rate": 1.9617680448236716e-07, + "loss": 0.6792, + "step": 1094 + }, + { + "epoch": 0.04, + "learning_rate": 1.9617314241769509e-07, + "loss": 0.6981, + "step": 1095 + }, + { + "epoch": 0.04, + "learning_rate": 1.9616948035302304e-07, + "loss": 0.7017, + "step": 1096 + }, + { + "epoch": 0.04, + "learning_rate": 1.9616581828835097e-07, + "loss": 0.7227, + "step": 1097 + }, + { + "epoch": 0.04, + "learning_rate": 1.9616215622367892e-07, + "loss": 0.7074, + "step": 1098 + }, + { + "epoch": 0.04, + "learning_rate": 1.9615849415900685e-07, + "loss": 0.6975, + "step": 1099 + }, + { + "epoch": 0.04, + "learning_rate": 1.961548320943348e-07, + "loss": 0.7084, + "step": 1100 + }, + { + "epoch": 0.04, + "learning_rate": 1.961511700296627e-07, + "loss": 0.6858, + "step": 1101 + }, + { + "epoch": 0.04, + "learning_rate": 1.9614750796499065e-07, + "loss": 0.7012, + "step": 1102 + }, + { + "epoch": 0.04, + "learning_rate": 1.9614384590031858e-07, + "loss": 0.713, + "step": 1103 + }, + { + "epoch": 0.04, + "learning_rate": 1.9614018383564653e-07, + "loss": 0.6872, + "step": 1104 + }, + { + "epoch": 0.04, + "learning_rate": 1.9613652177097446e-07, + "loss": 0.6837, + "step": 1105 + }, + { + "epoch": 0.04, + "learning_rate": 1.961328597063024e-07, + "loss": 0.6668, + "step": 1106 + }, + { + "epoch": 0.04, + "learning_rate": 1.9612919764163034e-07, + "loss": 0.6969, + "step": 1107 + }, + { + "epoch": 0.04, + "learning_rate": 1.961255355769583e-07, + "loss": 0.7426, + "step": 1108 + }, + { + "epoch": 0.04, + "learning_rate": 1.9612187351228622e-07, + "loss": 0.6973, + "step": 1109 + }, + { + "epoch": 0.04, + "learning_rate": 1.9611821144761414e-07, + "loss": 0.6784, + "step": 1110 + }, + { + "epoch": 0.04, + "learning_rate": 1.961145493829421e-07, + "loss": 0.6968, + "step": 1111 + }, + { + "epoch": 0.04, + "learning_rate": 1.9611088731827002e-07, + "loss": 0.7014, + "step": 1112 + }, + { + "epoch": 0.04, + "learning_rate": 1.9610722525359798e-07, + "loss": 0.6803, + "step": 1113 + }, + { + "epoch": 0.04, + "learning_rate": 1.961035631889259e-07, + "loss": 0.7189, + "step": 1114 + }, + { + "epoch": 0.04, + "learning_rate": 1.9609990112425386e-07, + "loss": 0.6934, + "step": 1115 + }, + { + "epoch": 0.04, + "learning_rate": 1.9609623905958178e-07, + "loss": 0.6777, + "step": 1116 + }, + { + "epoch": 0.04, + "learning_rate": 1.9609257699490974e-07, + "loss": 0.7159, + "step": 1117 + }, + { + "epoch": 0.04, + "learning_rate": 1.9608891493023766e-07, + "loss": 0.6657, + "step": 1118 + }, + { + "epoch": 0.04, + "learning_rate": 1.960852528655656e-07, + "loss": 0.7201, + "step": 1119 + }, + { + "epoch": 0.04, + "learning_rate": 1.9608159080089352e-07, + "loss": 0.6908, + "step": 1120 + }, + { + "epoch": 0.04, + "learning_rate": 1.9607792873622147e-07, + "loss": 0.7035, + "step": 1121 + }, + { + "epoch": 0.04, + "learning_rate": 1.960742666715494e-07, + "loss": 0.6824, + "step": 1122 + }, + { + "epoch": 0.04, + "learning_rate": 1.9607060460687735e-07, + "loss": 0.7182, + "step": 1123 + }, + { + "epoch": 0.04, + "learning_rate": 1.9606694254220528e-07, + "loss": 0.6967, + "step": 1124 + }, + { + "epoch": 0.04, + "learning_rate": 1.9606328047753323e-07, + "loss": 0.6813, + "step": 1125 + }, + { + "epoch": 0.04, + "learning_rate": 1.9605961841286116e-07, + "loss": 0.7036, + "step": 1126 + }, + { + "epoch": 0.04, + "learning_rate": 1.960559563481891e-07, + "loss": 0.7181, + "step": 1127 + }, + { + "epoch": 0.04, + "learning_rate": 1.9605229428351704e-07, + "loss": 0.6969, + "step": 1128 + }, + { + "epoch": 0.04, + "learning_rate": 1.96048632218845e-07, + "loss": 0.6802, + "step": 1129 + }, + { + "epoch": 0.04, + "learning_rate": 1.9604497015417292e-07, + "loss": 0.7077, + "step": 1130 + }, + { + "epoch": 0.04, + "learning_rate": 1.9604130808950084e-07, + "loss": 0.6833, + "step": 1131 + }, + { + "epoch": 0.04, + "learning_rate": 1.960376460248288e-07, + "loss": 0.6869, + "step": 1132 + }, + { + "epoch": 0.04, + "learning_rate": 1.9603398396015672e-07, + "loss": 0.6974, + "step": 1133 + }, + { + "epoch": 0.04, + "learning_rate": 1.9603032189548468e-07, + "loss": 0.6829, + "step": 1134 + }, + { + "epoch": 0.04, + "learning_rate": 1.960266598308126e-07, + "loss": 0.6945, + "step": 1135 + }, + { + "epoch": 0.04, + "learning_rate": 1.9602299776614056e-07, + "loss": 0.704, + "step": 1136 + }, + { + "epoch": 0.04, + "learning_rate": 1.9601933570146846e-07, + "loss": 0.6938, + "step": 1137 + }, + { + "epoch": 0.04, + "learning_rate": 1.960156736367964e-07, + "loss": 0.6918, + "step": 1138 + }, + { + "epoch": 0.04, + "learning_rate": 1.9601201157212434e-07, + "loss": 0.6844, + "step": 1139 + }, + { + "epoch": 0.04, + "learning_rate": 1.960083495074523e-07, + "loss": 0.711, + "step": 1140 + }, + { + "epoch": 0.04, + "learning_rate": 1.9600468744278022e-07, + "loss": 0.6722, + "step": 1141 + }, + { + "epoch": 0.04, + "learning_rate": 1.9600102537810817e-07, + "loss": 0.6722, + "step": 1142 + }, + { + "epoch": 0.04, + "learning_rate": 1.959973633134361e-07, + "loss": 0.684, + "step": 1143 + }, + { + "epoch": 0.04, + "learning_rate": 1.9599370124876405e-07, + "loss": 0.7009, + "step": 1144 + }, + { + "epoch": 0.04, + "learning_rate": 1.9599003918409198e-07, + "loss": 0.7048, + "step": 1145 + }, + { + "epoch": 0.04, + "learning_rate": 1.9598637711941993e-07, + "loss": 0.69, + "step": 1146 + }, + { + "epoch": 0.04, + "learning_rate": 1.9598271505474786e-07, + "loss": 0.6914, + "step": 1147 + }, + { + "epoch": 0.04, + "learning_rate": 1.959790529900758e-07, + "loss": 0.6943, + "step": 1148 + }, + { + "epoch": 0.04, + "learning_rate": 1.9597539092540374e-07, + "loss": 0.6813, + "step": 1149 + }, + { + "epoch": 0.04, + "learning_rate": 1.959717288607317e-07, + "loss": 0.7086, + "step": 1150 + }, + { + "epoch": 0.04, + "learning_rate": 1.9596806679605962e-07, + "loss": 0.7129, + "step": 1151 + }, + { + "epoch": 0.04, + "learning_rate": 1.9596440473138757e-07, + "loss": 0.695, + "step": 1152 + }, + { + "epoch": 0.04, + "learning_rate": 1.959607426667155e-07, + "loss": 0.7035, + "step": 1153 + }, + { + "epoch": 0.04, + "learning_rate": 1.9595708060204342e-07, + "loss": 0.6845, + "step": 1154 + }, + { + "epoch": 0.04, + "learning_rate": 1.9595341853737135e-07, + "loss": 0.699, + "step": 1155 + }, + { + "epoch": 0.04, + "learning_rate": 1.959497564726993e-07, + "loss": 0.7145, + "step": 1156 + }, + { + "epoch": 0.04, + "learning_rate": 1.9594609440802723e-07, + "loss": 0.6838, + "step": 1157 + }, + { + "epoch": 0.04, + "learning_rate": 1.9594243234335516e-07, + "loss": 0.6727, + "step": 1158 + }, + { + "epoch": 0.04, + "learning_rate": 1.959387702786831e-07, + "loss": 0.6985, + "step": 1159 + }, + { + "epoch": 0.04, + "learning_rate": 1.9593510821401104e-07, + "loss": 0.6617, + "step": 1160 + }, + { + "epoch": 0.04, + "learning_rate": 1.95931446149339e-07, + "loss": 0.6986, + "step": 1161 + }, + { + "epoch": 0.04, + "learning_rate": 1.9592778408466692e-07, + "loss": 0.6695, + "step": 1162 + }, + { + "epoch": 0.04, + "learning_rate": 1.9592412201999487e-07, + "loss": 0.6994, + "step": 1163 + }, + { + "epoch": 0.04, + "learning_rate": 1.959204599553228e-07, + "loss": 0.6813, + "step": 1164 + }, + { + "epoch": 0.04, + "learning_rate": 1.9591679789065075e-07, + "loss": 0.684, + "step": 1165 + }, + { + "epoch": 0.04, + "learning_rate": 1.9591313582597867e-07, + "loss": 0.6886, + "step": 1166 + }, + { + "epoch": 0.04, + "learning_rate": 1.9590947376130663e-07, + "loss": 0.6964, + "step": 1167 + }, + { + "epoch": 0.04, + "learning_rate": 1.9590581169663455e-07, + "loss": 0.6875, + "step": 1168 + }, + { + "epoch": 0.04, + "learning_rate": 1.959021496319625e-07, + "loss": 0.7041, + "step": 1169 + }, + { + "epoch": 0.04, + "learning_rate": 1.9589848756729043e-07, + "loss": 0.7111, + "step": 1170 + }, + { + "epoch": 0.04, + "learning_rate": 1.958948255026184e-07, + "loss": 0.6932, + "step": 1171 + }, + { + "epoch": 0.04, + "learning_rate": 1.9589116343794631e-07, + "loss": 0.7029, + "step": 1172 + }, + { + "epoch": 0.04, + "learning_rate": 1.9588750137327424e-07, + "loss": 0.6909, + "step": 1173 + }, + { + "epoch": 0.04, + "learning_rate": 1.9588383930860217e-07, + "loss": 0.6893, + "step": 1174 + }, + { + "epoch": 0.04, + "learning_rate": 1.9588017724393012e-07, + "loss": 0.6892, + "step": 1175 + }, + { + "epoch": 0.04, + "learning_rate": 1.9587651517925805e-07, + "loss": 0.7034, + "step": 1176 + }, + { + "epoch": 0.04, + "learning_rate": 1.95872853114586e-07, + "loss": 0.6821, + "step": 1177 + }, + { + "epoch": 0.04, + "learning_rate": 1.9586919104991393e-07, + "loss": 0.6847, + "step": 1178 + }, + { + "epoch": 0.04, + "learning_rate": 1.9586552898524188e-07, + "loss": 0.6962, + "step": 1179 + }, + { + "epoch": 0.04, + "learning_rate": 1.958618669205698e-07, + "loss": 0.691, + "step": 1180 + }, + { + "epoch": 0.04, + "learning_rate": 1.9585820485589773e-07, + "loss": 0.6919, + "step": 1181 + }, + { + "epoch": 0.04, + "learning_rate": 1.958545427912257e-07, + "loss": 0.687, + "step": 1182 + }, + { + "epoch": 0.04, + "learning_rate": 1.9585088072655361e-07, + "loss": 0.7198, + "step": 1183 + }, + { + "epoch": 0.04, + "learning_rate": 1.9584721866188157e-07, + "loss": 0.7103, + "step": 1184 + }, + { + "epoch": 0.04, + "learning_rate": 1.958435565972095e-07, + "loss": 0.6897, + "step": 1185 + }, + { + "epoch": 0.04, + "learning_rate": 1.9583989453253745e-07, + "loss": 0.6765, + "step": 1186 + }, + { + "epoch": 0.04, + "learning_rate": 1.9583623246786537e-07, + "loss": 0.6858, + "step": 1187 + }, + { + "epoch": 0.04, + "learning_rate": 1.9583257040319333e-07, + "loss": 0.6883, + "step": 1188 + }, + { + "epoch": 0.04, + "learning_rate": 1.9582890833852125e-07, + "loss": 0.6873, + "step": 1189 + }, + { + "epoch": 0.04, + "learning_rate": 1.958252462738492e-07, + "loss": 0.7108, + "step": 1190 + }, + { + "epoch": 0.04, + "learning_rate": 1.958215842091771e-07, + "loss": 0.7038, + "step": 1191 + }, + { + "epoch": 0.04, + "learning_rate": 1.9581792214450506e-07, + "loss": 0.6766, + "step": 1192 + }, + { + "epoch": 0.04, + "learning_rate": 1.95814260079833e-07, + "loss": 0.717, + "step": 1193 + }, + { + "epoch": 0.04, + "learning_rate": 1.9581059801516094e-07, + "loss": 0.6763, + "step": 1194 + }, + { + "epoch": 0.04, + "learning_rate": 1.9580693595048887e-07, + "loss": 0.7105, + "step": 1195 + }, + { + "epoch": 0.04, + "learning_rate": 1.9580327388581682e-07, + "loss": 0.7001, + "step": 1196 + }, + { + "epoch": 0.04, + "learning_rate": 1.9579961182114475e-07, + "loss": 0.7095, + "step": 1197 + }, + { + "epoch": 0.04, + "learning_rate": 1.957959497564727e-07, + "loss": 0.6956, + "step": 1198 + }, + { + "epoch": 0.04, + "learning_rate": 1.9579228769180063e-07, + "loss": 0.7058, + "step": 1199 + }, + { + "epoch": 0.04, + "learning_rate": 1.9578862562712858e-07, + "loss": 0.685, + "step": 1200 + }, + { + "epoch": 0.04, + "learning_rate": 1.957849635624565e-07, + "loss": 0.6986, + "step": 1201 + }, + { + "epoch": 0.04, + "learning_rate": 1.9578130149778443e-07, + "loss": 0.6955, + "step": 1202 + }, + { + "epoch": 0.04, + "learning_rate": 1.9577763943311239e-07, + "loss": 0.7231, + "step": 1203 + }, + { + "epoch": 0.04, + "learning_rate": 1.957739773684403e-07, + "loss": 0.7087, + "step": 1204 + }, + { + "epoch": 0.04, + "learning_rate": 1.9577031530376827e-07, + "loss": 0.6685, + "step": 1205 + }, + { + "epoch": 0.04, + "learning_rate": 1.957666532390962e-07, + "loss": 0.6874, + "step": 1206 + }, + { + "epoch": 0.04, + "learning_rate": 1.9576299117442415e-07, + "loss": 0.6765, + "step": 1207 + }, + { + "epoch": 0.04, + "learning_rate": 1.9575932910975205e-07, + "loss": 0.7104, + "step": 1208 + }, + { + "epoch": 0.04, + "learning_rate": 1.9575566704508e-07, + "loss": 0.6732, + "step": 1209 + }, + { + "epoch": 0.04, + "learning_rate": 1.9575200498040793e-07, + "loss": 0.7085, + "step": 1210 + }, + { + "epoch": 0.04, + "learning_rate": 1.9574834291573588e-07, + "loss": 0.6712, + "step": 1211 + }, + { + "epoch": 0.04, + "learning_rate": 1.957446808510638e-07, + "loss": 0.6964, + "step": 1212 + }, + { + "epoch": 0.04, + "learning_rate": 1.9574101878639176e-07, + "loss": 0.709, + "step": 1213 + }, + { + "epoch": 0.04, + "learning_rate": 1.9573735672171969e-07, + "loss": 0.7004, + "step": 1214 + }, + { + "epoch": 0.04, + "learning_rate": 1.9573369465704764e-07, + "loss": 0.695, + "step": 1215 + }, + { + "epoch": 0.04, + "learning_rate": 1.9573003259237557e-07, + "loss": 0.6823, + "step": 1216 + }, + { + "epoch": 0.04, + "learning_rate": 1.9572637052770352e-07, + "loss": 0.7124, + "step": 1217 + }, + { + "epoch": 0.04, + "learning_rate": 1.9572270846303145e-07, + "loss": 0.6814, + "step": 1218 + }, + { + "epoch": 0.04, + "learning_rate": 1.957190463983594e-07, + "loss": 0.7052, + "step": 1219 + }, + { + "epoch": 0.04, + "learning_rate": 1.9571538433368733e-07, + "loss": 0.699, + "step": 1220 + }, + { + "epoch": 0.04, + "learning_rate": 1.9571172226901528e-07, + "loss": 0.6848, + "step": 1221 + }, + { + "epoch": 0.04, + "learning_rate": 1.957080602043432e-07, + "loss": 0.6957, + "step": 1222 + }, + { + "epoch": 0.04, + "learning_rate": 1.9570439813967116e-07, + "loss": 0.7007, + "step": 1223 + }, + { + "epoch": 0.04, + "learning_rate": 1.9570073607499908e-07, + "loss": 0.6941, + "step": 1224 + }, + { + "epoch": 0.04, + "learning_rate": 1.95697074010327e-07, + "loss": 0.6928, + "step": 1225 + }, + { + "epoch": 0.04, + "learning_rate": 1.9569341194565494e-07, + "loss": 0.6911, + "step": 1226 + }, + { + "epoch": 0.04, + "learning_rate": 1.956897498809829e-07, + "loss": 0.6961, + "step": 1227 + }, + { + "epoch": 0.04, + "learning_rate": 1.9568608781631082e-07, + "loss": 0.7016, + "step": 1228 + }, + { + "epoch": 0.04, + "learning_rate": 1.9568242575163874e-07, + "loss": 0.698, + "step": 1229 + }, + { + "epoch": 0.05, + "learning_rate": 1.956787636869667e-07, + "loss": 0.6664, + "step": 1230 + }, + { + "epoch": 0.05, + "learning_rate": 1.9567510162229462e-07, + "loss": 0.6699, + "step": 1231 + }, + { + "epoch": 0.05, + "learning_rate": 1.9567143955762258e-07, + "loss": 0.6984, + "step": 1232 + }, + { + "epoch": 0.05, + "learning_rate": 1.956677774929505e-07, + "loss": 0.6545, + "step": 1233 + }, + { + "epoch": 0.05, + "learning_rate": 1.9566411542827846e-07, + "loss": 0.697, + "step": 1234 + }, + { + "epoch": 0.05, + "learning_rate": 1.9566045336360638e-07, + "loss": 0.7021, + "step": 1235 + }, + { + "epoch": 0.05, + "learning_rate": 1.9565679129893434e-07, + "loss": 0.7111, + "step": 1236 + }, + { + "epoch": 0.05, + "learning_rate": 1.9565312923426226e-07, + "loss": 0.6937, + "step": 1237 + }, + { + "epoch": 0.05, + "learning_rate": 1.9564946716959022e-07, + "loss": 0.708, + "step": 1238 + }, + { + "epoch": 0.05, + "learning_rate": 1.9564580510491814e-07, + "loss": 0.6805, + "step": 1239 + }, + { + "epoch": 0.05, + "learning_rate": 1.956421430402461e-07, + "loss": 0.7072, + "step": 1240 + }, + { + "epoch": 0.05, + "learning_rate": 1.9563848097557402e-07, + "loss": 0.7121, + "step": 1241 + }, + { + "epoch": 0.05, + "learning_rate": 1.9563481891090198e-07, + "loss": 0.6915, + "step": 1242 + }, + { + "epoch": 0.05, + "learning_rate": 1.956311568462299e-07, + "loss": 0.7138, + "step": 1243 + }, + { + "epoch": 0.05, + "learning_rate": 1.9562749478155783e-07, + "loss": 0.7122, + "step": 1244 + }, + { + "epoch": 0.05, + "learning_rate": 1.9562383271688576e-07, + "loss": 0.7277, + "step": 1245 + }, + { + "epoch": 0.05, + "learning_rate": 1.956201706522137e-07, + "loss": 0.6862, + "step": 1246 + }, + { + "epoch": 0.05, + "learning_rate": 1.9561650858754164e-07, + "loss": 0.6705, + "step": 1247 + }, + { + "epoch": 0.05, + "learning_rate": 1.956128465228696e-07, + "loss": 0.6982, + "step": 1248 + }, + { + "epoch": 0.05, + "learning_rate": 1.9560918445819752e-07, + "loss": 0.6893, + "step": 1249 + }, + { + "epoch": 0.05, + "learning_rate": 1.9560552239352547e-07, + "loss": 0.6788, + "step": 1250 + }, + { + "epoch": 0.05, + "learning_rate": 1.956018603288534e-07, + "loss": 0.6924, + "step": 1251 + }, + { + "epoch": 0.05, + "learning_rate": 1.9559819826418132e-07, + "loss": 0.702, + "step": 1252 + }, + { + "epoch": 0.05, + "learning_rate": 1.9559453619950928e-07, + "loss": 0.6765, + "step": 1253 + }, + { + "epoch": 0.05, + "learning_rate": 1.955908741348372e-07, + "loss": 0.6952, + "step": 1254 + }, + { + "epoch": 0.05, + "learning_rate": 1.9558721207016516e-07, + "loss": 0.6973, + "step": 1255 + }, + { + "epoch": 0.05, + "learning_rate": 1.9558355000549308e-07, + "loss": 0.7004, + "step": 1256 + }, + { + "epoch": 0.05, + "learning_rate": 1.9557988794082104e-07, + "loss": 0.6996, + "step": 1257 + }, + { + "epoch": 0.05, + "learning_rate": 1.9557622587614896e-07, + "loss": 0.6856, + "step": 1258 + }, + { + "epoch": 0.05, + "learning_rate": 1.9557256381147692e-07, + "loss": 0.7034, + "step": 1259 + }, + { + "epoch": 0.05, + "learning_rate": 1.9556890174680484e-07, + "loss": 0.705, + "step": 1260 + }, + { + "epoch": 0.05, + "learning_rate": 1.955652396821328e-07, + "loss": 0.6895, + "step": 1261 + }, + { + "epoch": 0.05, + "learning_rate": 1.955615776174607e-07, + "loss": 0.7029, + "step": 1262 + }, + { + "epoch": 0.05, + "learning_rate": 1.9555791555278865e-07, + "loss": 0.7099, + "step": 1263 + }, + { + "epoch": 0.05, + "learning_rate": 1.9555425348811658e-07, + "loss": 0.6885, + "step": 1264 + }, + { + "epoch": 0.05, + "learning_rate": 1.9555059142344453e-07, + "loss": 0.6897, + "step": 1265 + }, + { + "epoch": 0.05, + "learning_rate": 1.9554692935877246e-07, + "loss": 0.7025, + "step": 1266 + }, + { + "epoch": 0.05, + "learning_rate": 1.955432672941004e-07, + "loss": 0.7038, + "step": 1267 + }, + { + "epoch": 0.05, + "learning_rate": 1.9553960522942834e-07, + "loss": 0.719, + "step": 1268 + }, + { + "epoch": 0.05, + "learning_rate": 1.955359431647563e-07, + "loss": 0.705, + "step": 1269 + }, + { + "epoch": 0.05, + "learning_rate": 1.9553228110008422e-07, + "loss": 0.6898, + "step": 1270 + }, + { + "epoch": 0.05, + "learning_rate": 1.9552861903541217e-07, + "loss": 0.7155, + "step": 1271 + }, + { + "epoch": 0.05, + "learning_rate": 1.955249569707401e-07, + "loss": 0.7134, + "step": 1272 + }, + { + "epoch": 0.05, + "learning_rate": 1.9552129490606802e-07, + "loss": 0.6983, + "step": 1273 + }, + { + "epoch": 0.05, + "learning_rate": 1.9551763284139598e-07, + "loss": 0.7189, + "step": 1274 + }, + { + "epoch": 0.05, + "learning_rate": 1.955139707767239e-07, + "loss": 0.7125, + "step": 1275 + }, + { + "epoch": 0.05, + "learning_rate": 1.9551030871205186e-07, + "loss": 0.7047, + "step": 1276 + }, + { + "epoch": 0.05, + "learning_rate": 1.9550664664737978e-07, + "loss": 0.7033, + "step": 1277 + }, + { + "epoch": 0.05, + "learning_rate": 1.9550298458270774e-07, + "loss": 0.7072, + "step": 1278 + }, + { + "epoch": 0.05, + "learning_rate": 1.9549932251803566e-07, + "loss": 0.7167, + "step": 1279 + }, + { + "epoch": 0.05, + "learning_rate": 1.954956604533636e-07, + "loss": 0.7235, + "step": 1280 + }, + { + "epoch": 0.05, + "learning_rate": 1.9549199838869152e-07, + "loss": 0.6817, + "step": 1281 + }, + { + "epoch": 0.05, + "learning_rate": 1.9548833632401947e-07, + "loss": 0.6829, + "step": 1282 + }, + { + "epoch": 0.05, + "learning_rate": 1.954846742593474e-07, + "loss": 0.6784, + "step": 1283 + }, + { + "epoch": 0.05, + "learning_rate": 1.9548101219467535e-07, + "loss": 0.696, + "step": 1284 + }, + { + "epoch": 0.05, + "learning_rate": 1.9547735013000327e-07, + "loss": 0.6945, + "step": 1285 + }, + { + "epoch": 0.05, + "learning_rate": 1.9547368806533123e-07, + "loss": 0.6825, + "step": 1286 + }, + { + "epoch": 0.05, + "learning_rate": 1.9547002600065915e-07, + "loss": 0.6898, + "step": 1287 + }, + { + "epoch": 0.05, + "learning_rate": 1.954663639359871e-07, + "loss": 0.7225, + "step": 1288 + }, + { + "epoch": 0.05, + "learning_rate": 1.9546270187131503e-07, + "loss": 0.7045, + "step": 1289 + }, + { + "epoch": 0.05, + "learning_rate": 1.95459039806643e-07, + "loss": 0.6986, + "step": 1290 + }, + { + "epoch": 0.05, + "learning_rate": 1.9545537774197091e-07, + "loss": 0.6972, + "step": 1291 + }, + { + "epoch": 0.05, + "learning_rate": 1.9545171567729887e-07, + "loss": 0.6817, + "step": 1292 + }, + { + "epoch": 0.05, + "learning_rate": 1.954480536126268e-07, + "loss": 0.7059, + "step": 1293 + }, + { + "epoch": 0.05, + "learning_rate": 1.9544439154795475e-07, + "loss": 0.6766, + "step": 1294 + }, + { + "epoch": 0.05, + "learning_rate": 1.9544072948328267e-07, + "loss": 0.6809, + "step": 1295 + }, + { + "epoch": 0.05, + "learning_rate": 1.954370674186106e-07, + "loss": 0.724, + "step": 1296 + }, + { + "epoch": 0.05, + "learning_rate": 1.9543340535393855e-07, + "loss": 0.7063, + "step": 1297 + }, + { + "epoch": 0.05, + "learning_rate": 1.9542974328926648e-07, + "loss": 0.7172, + "step": 1298 + }, + { + "epoch": 0.05, + "learning_rate": 1.954260812245944e-07, + "loss": 0.7064, + "step": 1299 + }, + { + "epoch": 0.05, + "learning_rate": 1.9542241915992233e-07, + "loss": 0.7029, + "step": 1300 + }, + { + "epoch": 0.05, + "learning_rate": 1.954187570952503e-07, + "loss": 0.6859, + "step": 1301 + }, + { + "epoch": 0.05, + "learning_rate": 1.9541509503057821e-07, + "loss": 0.6907, + "step": 1302 + }, + { + "epoch": 0.05, + "learning_rate": 1.9541143296590617e-07, + "loss": 0.685, + "step": 1303 + }, + { + "epoch": 0.05, + "learning_rate": 1.954077709012341e-07, + "loss": 0.6941, + "step": 1304 + }, + { + "epoch": 0.05, + "learning_rate": 1.9540410883656205e-07, + "loss": 0.7073, + "step": 1305 + }, + { + "epoch": 0.05, + "learning_rate": 1.9540044677188997e-07, + "loss": 0.6805, + "step": 1306 + }, + { + "epoch": 0.05, + "learning_rate": 1.9539678470721793e-07, + "loss": 0.6977, + "step": 1307 + }, + { + "epoch": 0.05, + "learning_rate": 1.9539312264254585e-07, + "loss": 0.6989, + "step": 1308 + }, + { + "epoch": 0.05, + "learning_rate": 1.953894605778738e-07, + "loss": 0.6952, + "step": 1309 + }, + { + "epoch": 0.05, + "learning_rate": 1.9538579851320173e-07, + "loss": 0.6891, + "step": 1310 + }, + { + "epoch": 0.05, + "learning_rate": 1.9538213644852969e-07, + "loss": 0.6778, + "step": 1311 + }, + { + "epoch": 0.05, + "learning_rate": 1.9537847438385761e-07, + "loss": 0.7074, + "step": 1312 + }, + { + "epoch": 0.05, + "learning_rate": 1.9537481231918557e-07, + "loss": 0.7039, + "step": 1313 + }, + { + "epoch": 0.05, + "learning_rate": 1.953711502545135e-07, + "loss": 0.69, + "step": 1314 + }, + { + "epoch": 0.05, + "learning_rate": 1.9536748818984145e-07, + "loss": 0.6834, + "step": 1315 + }, + { + "epoch": 0.05, + "learning_rate": 1.9536382612516935e-07, + "loss": 0.6944, + "step": 1316 + }, + { + "epoch": 0.05, + "learning_rate": 1.953601640604973e-07, + "loss": 0.6831, + "step": 1317 + }, + { + "epoch": 0.05, + "learning_rate": 1.9535650199582523e-07, + "loss": 0.6826, + "step": 1318 + }, + { + "epoch": 0.05, + "learning_rate": 1.9535283993115318e-07, + "loss": 0.7, + "step": 1319 + }, + { + "epoch": 0.05, + "learning_rate": 1.953491778664811e-07, + "loss": 0.6982, + "step": 1320 + }, + { + "epoch": 0.05, + "learning_rate": 1.9534551580180906e-07, + "loss": 0.6921, + "step": 1321 + }, + { + "epoch": 0.05, + "learning_rate": 1.9534185373713699e-07, + "loss": 0.7082, + "step": 1322 + }, + { + "epoch": 0.05, + "learning_rate": 1.953381916724649e-07, + "loss": 0.7001, + "step": 1323 + }, + { + "epoch": 0.05, + "learning_rate": 1.9533452960779287e-07, + "loss": 0.7013, + "step": 1324 + }, + { + "epoch": 0.05, + "learning_rate": 1.953308675431208e-07, + "loss": 0.7114, + "step": 1325 + }, + { + "epoch": 0.05, + "learning_rate": 1.9532720547844875e-07, + "loss": 0.6946, + "step": 1326 + }, + { + "epoch": 0.05, + "learning_rate": 1.9532354341377667e-07, + "loss": 0.686, + "step": 1327 + }, + { + "epoch": 0.05, + "learning_rate": 1.9531988134910463e-07, + "loss": 0.694, + "step": 1328 + }, + { + "epoch": 0.05, + "learning_rate": 1.9531621928443255e-07, + "loss": 0.7056, + "step": 1329 + }, + { + "epoch": 0.05, + "learning_rate": 1.953125572197605e-07, + "loss": 0.6939, + "step": 1330 + }, + { + "epoch": 0.05, + "learning_rate": 1.9530889515508843e-07, + "loss": 0.7065, + "step": 1331 + }, + { + "epoch": 0.05, + "learning_rate": 1.9530523309041639e-07, + "loss": 0.701, + "step": 1332 + }, + { + "epoch": 0.05, + "learning_rate": 1.953015710257443e-07, + "loss": 0.6929, + "step": 1333 + }, + { + "epoch": 0.05, + "learning_rate": 1.9529790896107224e-07, + "loss": 0.7036, + "step": 1334 + }, + { + "epoch": 0.05, + "learning_rate": 1.9529424689640017e-07, + "loss": 0.6902, + "step": 1335 + }, + { + "epoch": 0.05, + "learning_rate": 1.9529058483172812e-07, + "loss": 0.7005, + "step": 1336 + }, + { + "epoch": 0.05, + "learning_rate": 1.9528692276705605e-07, + "loss": 0.705, + "step": 1337 + }, + { + "epoch": 0.05, + "learning_rate": 1.95283260702384e-07, + "loss": 0.6939, + "step": 1338 + }, + { + "epoch": 0.05, + "learning_rate": 1.9527959863771193e-07, + "loss": 0.692, + "step": 1339 + }, + { + "epoch": 0.05, + "learning_rate": 1.9527593657303988e-07, + "loss": 0.7056, + "step": 1340 + }, + { + "epoch": 0.05, + "learning_rate": 1.952722745083678e-07, + "loss": 0.7037, + "step": 1341 + }, + { + "epoch": 0.05, + "learning_rate": 1.9526861244369576e-07, + "loss": 0.6961, + "step": 1342 + }, + { + "epoch": 0.05, + "learning_rate": 1.9526495037902368e-07, + "loss": 0.6977, + "step": 1343 + }, + { + "epoch": 0.05, + "learning_rate": 1.952612883143516e-07, + "loss": 0.7014, + "step": 1344 + }, + { + "epoch": 0.05, + "learning_rate": 1.9525762624967956e-07, + "loss": 0.6967, + "step": 1345 + }, + { + "epoch": 0.05, + "learning_rate": 1.952539641850075e-07, + "loss": 0.6916, + "step": 1346 + }, + { + "epoch": 0.05, + "learning_rate": 1.9525030212033544e-07, + "loss": 0.687, + "step": 1347 + }, + { + "epoch": 0.05, + "learning_rate": 1.9524664005566337e-07, + "loss": 0.7009, + "step": 1348 + }, + { + "epoch": 0.05, + "learning_rate": 1.9524297799099132e-07, + "loss": 0.6672, + "step": 1349 + }, + { + "epoch": 0.05, + "learning_rate": 1.9523931592631925e-07, + "loss": 0.6943, + "step": 1350 + }, + { + "epoch": 0.05, + "learning_rate": 1.952356538616472e-07, + "loss": 0.707, + "step": 1351 + }, + { + "epoch": 0.05, + "learning_rate": 1.952319917969751e-07, + "loss": 0.6936, + "step": 1352 + }, + { + "epoch": 0.05, + "learning_rate": 1.9522832973230306e-07, + "loss": 0.6982, + "step": 1353 + }, + { + "epoch": 0.05, + "learning_rate": 1.9522466766763098e-07, + "loss": 0.6981, + "step": 1354 + }, + { + "epoch": 0.05, + "learning_rate": 1.9522100560295894e-07, + "loss": 0.6808, + "step": 1355 + }, + { + "epoch": 0.05, + "learning_rate": 1.9521734353828686e-07, + "loss": 0.727, + "step": 1356 + }, + { + "epoch": 0.05, + "learning_rate": 1.9521368147361482e-07, + "loss": 0.6899, + "step": 1357 + }, + { + "epoch": 0.05, + "learning_rate": 1.9521001940894274e-07, + "loss": 0.6997, + "step": 1358 + }, + { + "epoch": 0.05, + "learning_rate": 1.952063573442707e-07, + "loss": 0.6999, + "step": 1359 + }, + { + "epoch": 0.05, + "learning_rate": 1.9520269527959862e-07, + "loss": 0.6907, + "step": 1360 + }, + { + "epoch": 0.05, + "learning_rate": 1.9519903321492658e-07, + "loss": 0.699, + "step": 1361 + }, + { + "epoch": 0.05, + "learning_rate": 1.951953711502545e-07, + "loss": 0.6881, + "step": 1362 + }, + { + "epoch": 0.05, + "learning_rate": 1.9519170908558246e-07, + "loss": 0.7016, + "step": 1363 + }, + { + "epoch": 0.05, + "learning_rate": 1.9518804702091038e-07, + "loss": 0.6863, + "step": 1364 + }, + { + "epoch": 0.05, + "learning_rate": 1.9518438495623834e-07, + "loss": 0.6868, + "step": 1365 + }, + { + "epoch": 0.05, + "learning_rate": 1.9518072289156626e-07, + "loss": 0.6819, + "step": 1366 + }, + { + "epoch": 0.05, + "learning_rate": 1.951770608268942e-07, + "loss": 0.6949, + "step": 1367 + }, + { + "epoch": 0.05, + "learning_rate": 1.9517339876222214e-07, + "loss": 0.7085, + "step": 1368 + }, + { + "epoch": 0.05, + "learning_rate": 1.9516973669755007e-07, + "loss": 0.706, + "step": 1369 + }, + { + "epoch": 0.05, + "learning_rate": 1.95166074632878e-07, + "loss": 0.6945, + "step": 1370 + }, + { + "epoch": 0.05, + "learning_rate": 1.9516241256820592e-07, + "loss": 0.673, + "step": 1371 + }, + { + "epoch": 0.05, + "learning_rate": 1.9515875050353388e-07, + "loss": 0.6619, + "step": 1372 + }, + { + "epoch": 0.05, + "learning_rate": 1.951550884388618e-07, + "loss": 0.6965, + "step": 1373 + }, + { + "epoch": 0.05, + "learning_rate": 1.9515142637418976e-07, + "loss": 0.7179, + "step": 1374 + }, + { + "epoch": 0.05, + "learning_rate": 1.9514776430951768e-07, + "loss": 0.708, + "step": 1375 + }, + { + "epoch": 0.05, + "learning_rate": 1.9514410224484564e-07, + "loss": 0.7059, + "step": 1376 + }, + { + "epoch": 0.05, + "learning_rate": 1.9514044018017356e-07, + "loss": 0.687, + "step": 1377 + }, + { + "epoch": 0.05, + "learning_rate": 1.9513677811550152e-07, + "loss": 0.6986, + "step": 1378 + }, + { + "epoch": 0.05, + "learning_rate": 1.9513311605082944e-07, + "loss": 0.688, + "step": 1379 + }, + { + "epoch": 0.05, + "learning_rate": 1.951294539861574e-07, + "loss": 0.7144, + "step": 1380 + }, + { + "epoch": 0.05, + "learning_rate": 1.9512579192148532e-07, + "loss": 0.6975, + "step": 1381 + }, + { + "epoch": 0.05, + "learning_rate": 1.9512212985681328e-07, + "loss": 0.6753, + "step": 1382 + }, + { + "epoch": 0.05, + "learning_rate": 1.951184677921412e-07, + "loss": 0.7012, + "step": 1383 + }, + { + "epoch": 0.05, + "learning_rate": 1.9511480572746916e-07, + "loss": 0.7256, + "step": 1384 + }, + { + "epoch": 0.05, + "learning_rate": 1.9511114366279708e-07, + "loss": 0.7123, + "step": 1385 + }, + { + "epoch": 0.05, + "learning_rate": 1.9510748159812504e-07, + "loss": 0.7142, + "step": 1386 + }, + { + "epoch": 0.05, + "learning_rate": 1.9510381953345296e-07, + "loss": 0.7011, + "step": 1387 + }, + { + "epoch": 0.05, + "learning_rate": 1.951001574687809e-07, + "loss": 0.7015, + "step": 1388 + }, + { + "epoch": 0.05, + "learning_rate": 1.9509649540410882e-07, + "loss": 0.6758, + "step": 1389 + }, + { + "epoch": 0.05, + "learning_rate": 1.9509283333943677e-07, + "loss": 0.6975, + "step": 1390 + }, + { + "epoch": 0.05, + "learning_rate": 1.950891712747647e-07, + "loss": 0.694, + "step": 1391 + }, + { + "epoch": 0.05, + "learning_rate": 1.9508550921009265e-07, + "loss": 0.6651, + "step": 1392 + }, + { + "epoch": 0.05, + "learning_rate": 1.9508184714542058e-07, + "loss": 0.6551, + "step": 1393 + }, + { + "epoch": 0.05, + "learning_rate": 1.950781850807485e-07, + "loss": 0.6727, + "step": 1394 + }, + { + "epoch": 0.05, + "learning_rate": 1.9507452301607646e-07, + "loss": 0.7042, + "step": 1395 + }, + { + "epoch": 0.05, + "learning_rate": 1.9507086095140438e-07, + "loss": 0.6802, + "step": 1396 + }, + { + "epoch": 0.05, + "learning_rate": 1.9506719888673234e-07, + "loss": 0.6914, + "step": 1397 + }, + { + "epoch": 0.05, + "learning_rate": 1.9506353682206026e-07, + "loss": 0.694, + "step": 1398 + }, + { + "epoch": 0.05, + "learning_rate": 1.9505987475738822e-07, + "loss": 0.7078, + "step": 1399 + }, + { + "epoch": 0.05, + "learning_rate": 1.9505621269271614e-07, + "loss": 0.6896, + "step": 1400 + }, + { + "epoch": 0.05, + "learning_rate": 1.950525506280441e-07, + "loss": 0.708, + "step": 1401 + }, + { + "epoch": 0.05, + "learning_rate": 1.9504888856337202e-07, + "loss": 0.7081, + "step": 1402 + }, + { + "epoch": 0.05, + "learning_rate": 1.9504522649869997e-07, + "loss": 0.6992, + "step": 1403 + }, + { + "epoch": 0.05, + "learning_rate": 1.950415644340279e-07, + "loss": 0.6962, + "step": 1404 + }, + { + "epoch": 0.05, + "learning_rate": 1.9503790236935585e-07, + "loss": 0.6883, + "step": 1405 + }, + { + "epoch": 0.05, + "learning_rate": 1.9503424030468375e-07, + "loss": 0.6907, + "step": 1406 + }, + { + "epoch": 0.05, + "learning_rate": 1.950305782400117e-07, + "loss": 0.6817, + "step": 1407 + }, + { + "epoch": 0.05, + "learning_rate": 1.9502691617533963e-07, + "loss": 0.6867, + "step": 1408 + }, + { + "epoch": 0.05, + "learning_rate": 1.950232541106676e-07, + "loss": 0.6814, + "step": 1409 + }, + { + "epoch": 0.05, + "learning_rate": 1.9501959204599551e-07, + "loss": 0.7039, + "step": 1410 + }, + { + "epoch": 0.05, + "learning_rate": 1.9501592998132347e-07, + "loss": 0.7095, + "step": 1411 + }, + { + "epoch": 0.05, + "learning_rate": 1.950122679166514e-07, + "loss": 0.6842, + "step": 1412 + }, + { + "epoch": 0.05, + "learning_rate": 1.9500860585197935e-07, + "loss": 0.6851, + "step": 1413 + }, + { + "epoch": 0.05, + "learning_rate": 1.9500494378730727e-07, + "loss": 0.6989, + "step": 1414 + }, + { + "epoch": 0.05, + "learning_rate": 1.950012817226352e-07, + "loss": 0.6931, + "step": 1415 + }, + { + "epoch": 0.05, + "learning_rate": 1.9499761965796315e-07, + "loss": 0.6965, + "step": 1416 + }, + { + "epoch": 0.05, + "learning_rate": 1.9499395759329108e-07, + "loss": 0.6754, + "step": 1417 + }, + { + "epoch": 0.05, + "learning_rate": 1.9499029552861903e-07, + "loss": 0.6849, + "step": 1418 + }, + { + "epoch": 0.05, + "learning_rate": 1.9498663346394696e-07, + "loss": 0.6898, + "step": 1419 + }, + { + "epoch": 0.05, + "learning_rate": 1.9498297139927491e-07, + "loss": 0.6828, + "step": 1420 + }, + { + "epoch": 0.05, + "learning_rate": 1.9497930933460284e-07, + "loss": 0.7154, + "step": 1421 + }, + { + "epoch": 0.05, + "learning_rate": 1.949756472699308e-07, + "loss": 0.7029, + "step": 1422 + }, + { + "epoch": 0.05, + "learning_rate": 1.9497198520525872e-07, + "loss": 0.6946, + "step": 1423 + }, + { + "epoch": 0.05, + "learning_rate": 1.9496832314058665e-07, + "loss": 0.7026, + "step": 1424 + }, + { + "epoch": 0.05, + "learning_rate": 1.9496466107591457e-07, + "loss": 0.6976, + "step": 1425 + }, + { + "epoch": 0.05, + "learning_rate": 1.9496099901124253e-07, + "loss": 0.6946, + "step": 1426 + }, + { + "epoch": 0.05, + "learning_rate": 1.9495733694657045e-07, + "loss": 0.6899, + "step": 1427 + }, + { + "epoch": 0.05, + "learning_rate": 1.949536748818984e-07, + "loss": 0.7296, + "step": 1428 + }, + { + "epoch": 0.05, + "learning_rate": 1.9495001281722633e-07, + "loss": 0.6726, + "step": 1429 + }, + { + "epoch": 0.05, + "learning_rate": 1.9494635075255429e-07, + "loss": 0.6997, + "step": 1430 + }, + { + "epoch": 0.05, + "learning_rate": 1.9494268868788221e-07, + "loss": 0.694, + "step": 1431 + }, + { + "epoch": 0.05, + "learning_rate": 1.9493902662321017e-07, + "loss": 0.693, + "step": 1432 + }, + { + "epoch": 0.05, + "learning_rate": 1.949353645585381e-07, + "loss": 0.6985, + "step": 1433 + }, + { + "epoch": 0.05, + "learning_rate": 1.9493170249386605e-07, + "loss": 0.6833, + "step": 1434 + }, + { + "epoch": 0.05, + "learning_rate": 1.9492804042919397e-07, + "loss": 0.6905, + "step": 1435 + }, + { + "epoch": 0.05, + "learning_rate": 1.9492437836452193e-07, + "loss": 0.6869, + "step": 1436 + }, + { + "epoch": 0.05, + "learning_rate": 1.9492071629984985e-07, + "loss": 0.678, + "step": 1437 + }, + { + "epoch": 0.05, + "learning_rate": 1.9491705423517778e-07, + "loss": 0.714, + "step": 1438 + }, + { + "epoch": 0.05, + "learning_rate": 1.9491339217050573e-07, + "loss": 0.6846, + "step": 1439 + }, + { + "epoch": 0.05, + "learning_rate": 1.9490973010583366e-07, + "loss": 0.7208, + "step": 1440 + }, + { + "epoch": 0.05, + "learning_rate": 1.949060680411616e-07, + "loss": 0.6844, + "step": 1441 + }, + { + "epoch": 0.05, + "learning_rate": 1.949024059764895e-07, + "loss": 0.7186, + "step": 1442 + }, + { + "epoch": 0.05, + "learning_rate": 1.9489874391181747e-07, + "loss": 0.7268, + "step": 1443 + }, + { + "epoch": 0.05, + "learning_rate": 1.948950818471454e-07, + "loss": 0.6879, + "step": 1444 + }, + { + "epoch": 0.05, + "learning_rate": 1.9489141978247335e-07, + "loss": 0.7246, + "step": 1445 + }, + { + "epoch": 0.05, + "learning_rate": 1.9488775771780127e-07, + "loss": 0.6894, + "step": 1446 + }, + { + "epoch": 0.05, + "learning_rate": 1.9488409565312923e-07, + "loss": 0.7081, + "step": 1447 + }, + { + "epoch": 0.05, + "learning_rate": 1.9488043358845715e-07, + "loss": 0.6978, + "step": 1448 + }, + { + "epoch": 0.05, + "learning_rate": 1.948767715237851e-07, + "loss": 0.6623, + "step": 1449 + }, + { + "epoch": 0.05, + "learning_rate": 1.9487310945911303e-07, + "loss": 0.6888, + "step": 1450 + }, + { + "epoch": 0.05, + "learning_rate": 1.9486944739444099e-07, + "loss": 0.6847, + "step": 1451 + }, + { + "epoch": 0.05, + "learning_rate": 1.948657853297689e-07, + "loss": 0.6844, + "step": 1452 + }, + { + "epoch": 0.05, + "learning_rate": 1.9486212326509687e-07, + "loss": 0.7084, + "step": 1453 + }, + { + "epoch": 0.05, + "learning_rate": 1.948584612004248e-07, + "loss": 0.6946, + "step": 1454 + }, + { + "epoch": 0.05, + "learning_rate": 1.9485479913575275e-07, + "loss": 0.6881, + "step": 1455 + }, + { + "epoch": 0.05, + "learning_rate": 1.9485113707108067e-07, + "loss": 0.7405, + "step": 1456 + }, + { + "epoch": 0.05, + "learning_rate": 1.9484747500640863e-07, + "loss": 0.7248, + "step": 1457 + }, + { + "epoch": 0.05, + "learning_rate": 1.9484381294173655e-07, + "loss": 0.7108, + "step": 1458 + }, + { + "epoch": 0.05, + "learning_rate": 1.9484015087706448e-07, + "loss": 0.6998, + "step": 1459 + }, + { + "epoch": 0.05, + "learning_rate": 1.948364888123924e-07, + "loss": 0.6933, + "step": 1460 + }, + { + "epoch": 0.05, + "learning_rate": 1.9483282674772036e-07, + "loss": 0.681, + "step": 1461 + }, + { + "epoch": 0.05, + "learning_rate": 1.9482916468304828e-07, + "loss": 0.713, + "step": 1462 + }, + { + "epoch": 0.05, + "learning_rate": 1.9482550261837624e-07, + "loss": 0.6776, + "step": 1463 + }, + { + "epoch": 0.05, + "learning_rate": 1.9482184055370416e-07, + "loss": 0.7063, + "step": 1464 + }, + { + "epoch": 0.05, + "learning_rate": 1.948181784890321e-07, + "loss": 0.6853, + "step": 1465 + }, + { + "epoch": 0.05, + "learning_rate": 1.9481451642436004e-07, + "loss": 0.6988, + "step": 1466 + }, + { + "epoch": 0.05, + "learning_rate": 1.9481085435968797e-07, + "loss": 0.6982, + "step": 1467 + }, + { + "epoch": 0.05, + "learning_rate": 1.9480719229501592e-07, + "loss": 0.6909, + "step": 1468 + }, + { + "epoch": 0.05, + "learning_rate": 1.9480353023034385e-07, + "loss": 0.6936, + "step": 1469 + }, + { + "epoch": 0.05, + "learning_rate": 1.947998681656718e-07, + "loss": 0.6987, + "step": 1470 + }, + { + "epoch": 0.05, + "learning_rate": 1.9479620610099973e-07, + "loss": 0.6771, + "step": 1471 + }, + { + "epoch": 0.05, + "learning_rate": 1.9479254403632768e-07, + "loss": 0.7026, + "step": 1472 + }, + { + "epoch": 0.05, + "learning_rate": 1.947888819716556e-07, + "loss": 0.6872, + "step": 1473 + }, + { + "epoch": 0.05, + "learning_rate": 1.9478521990698356e-07, + "loss": 0.7072, + "step": 1474 + }, + { + "epoch": 0.05, + "learning_rate": 1.947815578423115e-07, + "loss": 0.6965, + "step": 1475 + }, + { + "epoch": 0.05, + "learning_rate": 1.9477789577763944e-07, + "loss": 0.718, + "step": 1476 + }, + { + "epoch": 0.05, + "learning_rate": 1.9477423371296737e-07, + "loss": 0.6911, + "step": 1477 + }, + { + "epoch": 0.05, + "learning_rate": 1.947705716482953e-07, + "loss": 0.6725, + "step": 1478 + }, + { + "epoch": 0.05, + "learning_rate": 1.9476690958362322e-07, + "loss": 0.6803, + "step": 1479 + }, + { + "epoch": 0.05, + "learning_rate": 1.9476324751895118e-07, + "loss": 0.7111, + "step": 1480 + }, + { + "epoch": 0.05, + "learning_rate": 1.947595854542791e-07, + "loss": 0.7017, + "step": 1481 + }, + { + "epoch": 0.05, + "learning_rate": 1.9475592338960706e-07, + "loss": 0.6937, + "step": 1482 + }, + { + "epoch": 0.05, + "learning_rate": 1.9475226132493498e-07, + "loss": 0.6817, + "step": 1483 + }, + { + "epoch": 0.05, + "learning_rate": 1.9474859926026294e-07, + "loss": 0.6989, + "step": 1484 + }, + { + "epoch": 0.05, + "learning_rate": 1.9474493719559086e-07, + "loss": 0.6888, + "step": 1485 + }, + { + "epoch": 0.05, + "learning_rate": 1.947412751309188e-07, + "loss": 0.69, + "step": 1486 + }, + { + "epoch": 0.05, + "learning_rate": 1.9473761306624674e-07, + "loss": 0.677, + "step": 1487 + }, + { + "epoch": 0.05, + "learning_rate": 1.9473395100157467e-07, + "loss": 0.7057, + "step": 1488 + }, + { + "epoch": 0.05, + "learning_rate": 1.9473028893690262e-07, + "loss": 0.6999, + "step": 1489 + }, + { + "epoch": 0.05, + "learning_rate": 1.9472662687223055e-07, + "loss": 0.7039, + "step": 1490 + }, + { + "epoch": 0.05, + "learning_rate": 1.947229648075585e-07, + "loss": 0.6863, + "step": 1491 + }, + { + "epoch": 0.05, + "learning_rate": 1.9471930274288643e-07, + "loss": 0.6941, + "step": 1492 + }, + { + "epoch": 0.05, + "learning_rate": 1.9471564067821438e-07, + "loss": 0.7153, + "step": 1493 + }, + { + "epoch": 0.05, + "learning_rate": 1.947119786135423e-07, + "loss": 0.6804, + "step": 1494 + }, + { + "epoch": 0.05, + "learning_rate": 1.9470831654887026e-07, + "loss": 0.6868, + "step": 1495 + }, + { + "epoch": 0.05, + "learning_rate": 1.9470465448419816e-07, + "loss": 0.6963, + "step": 1496 + }, + { + "epoch": 0.05, + "learning_rate": 1.9470099241952612e-07, + "loss": 0.7026, + "step": 1497 + }, + { + "epoch": 0.05, + "learning_rate": 1.9469733035485404e-07, + "loss": 0.679, + "step": 1498 + }, + { + "epoch": 0.05, + "learning_rate": 1.94693668290182e-07, + "loss": 0.7025, + "step": 1499 + }, + { + "epoch": 0.05, + "learning_rate": 1.9469000622550992e-07, + "loss": 0.6945, + "step": 1500 + }, + { + "epoch": 0.05, + "learning_rate": 1.9468634416083788e-07, + "loss": 0.6876, + "step": 1501 + }, + { + "epoch": 0.05, + "learning_rate": 1.946826820961658e-07, + "loss": 0.6957, + "step": 1502 + }, + { + "epoch": 0.05, + "learning_rate": 1.9467902003149376e-07, + "loss": 0.6867, + "step": 1503 + }, + { + "epoch": 0.06, + "learning_rate": 1.9467535796682168e-07, + "loss": 0.6878, + "step": 1504 + }, + { + "epoch": 0.06, + "learning_rate": 1.9467169590214964e-07, + "loss": 0.7052, + "step": 1505 + }, + { + "epoch": 0.06, + "learning_rate": 1.9466803383747756e-07, + "loss": 0.6867, + "step": 1506 + }, + { + "epoch": 0.06, + "learning_rate": 1.9466437177280552e-07, + "loss": 0.6865, + "step": 1507 + }, + { + "epoch": 0.06, + "learning_rate": 1.9466070970813344e-07, + "loss": 0.6875, + "step": 1508 + }, + { + "epoch": 0.06, + "learning_rate": 1.9465704764346137e-07, + "loss": 0.7056, + "step": 1509 + }, + { + "epoch": 0.06, + "learning_rate": 1.9465338557878932e-07, + "loss": 0.6726, + "step": 1510 + }, + { + "epoch": 0.06, + "learning_rate": 1.9464972351411725e-07, + "loss": 0.7082, + "step": 1511 + }, + { + "epoch": 0.06, + "learning_rate": 1.946460614494452e-07, + "loss": 0.6755, + "step": 1512 + }, + { + "epoch": 0.06, + "learning_rate": 1.9464239938477313e-07, + "loss": 0.7076, + "step": 1513 + }, + { + "epoch": 0.06, + "learning_rate": 1.9463873732010106e-07, + "loss": 0.7141, + "step": 1514 + }, + { + "epoch": 0.06, + "learning_rate": 1.9463507525542898e-07, + "loss": 0.6606, + "step": 1515 + }, + { + "epoch": 0.06, + "learning_rate": 1.9463141319075694e-07, + "loss": 0.6939, + "step": 1516 + }, + { + "epoch": 0.06, + "learning_rate": 1.9462775112608486e-07, + "loss": 0.6979, + "step": 1517 + }, + { + "epoch": 0.06, + "learning_rate": 1.9462408906141282e-07, + "loss": 0.6867, + "step": 1518 + }, + { + "epoch": 0.06, + "learning_rate": 1.9462042699674074e-07, + "loss": 0.6991, + "step": 1519 + }, + { + "epoch": 0.06, + "learning_rate": 1.946167649320687e-07, + "loss": 0.6638, + "step": 1520 + }, + { + "epoch": 0.06, + "learning_rate": 1.9461310286739662e-07, + "loss": 0.6858, + "step": 1521 + }, + { + "epoch": 0.06, + "learning_rate": 1.9460944080272457e-07, + "loss": 0.6849, + "step": 1522 + }, + { + "epoch": 0.06, + "learning_rate": 1.946057787380525e-07, + "loss": 0.6989, + "step": 1523 + }, + { + "epoch": 0.06, + "learning_rate": 1.9460211667338045e-07, + "loss": 0.6877, + "step": 1524 + }, + { + "epoch": 0.06, + "learning_rate": 1.9459845460870838e-07, + "loss": 0.7114, + "step": 1525 + }, + { + "epoch": 0.06, + "learning_rate": 1.9459479254403633e-07, + "loss": 0.6995, + "step": 1526 + }, + { + "epoch": 0.06, + "learning_rate": 1.9459113047936426e-07, + "loss": 0.6852, + "step": 1527 + }, + { + "epoch": 0.06, + "learning_rate": 1.9458746841469221e-07, + "loss": 0.6878, + "step": 1528 + }, + { + "epoch": 0.06, + "learning_rate": 1.9458380635002014e-07, + "loss": 0.7143, + "step": 1529 + }, + { + "epoch": 0.06, + "learning_rate": 1.9458014428534807e-07, + "loss": 0.6893, + "step": 1530 + }, + { + "epoch": 0.06, + "learning_rate": 1.9457648222067602e-07, + "loss": 0.7077, + "step": 1531 + }, + { + "epoch": 0.06, + "learning_rate": 1.9457282015600395e-07, + "loss": 0.6909, + "step": 1532 + }, + { + "epoch": 0.06, + "learning_rate": 1.9456915809133187e-07, + "loss": 0.7065, + "step": 1533 + }, + { + "epoch": 0.06, + "learning_rate": 1.9456549602665983e-07, + "loss": 0.7053, + "step": 1534 + }, + { + "epoch": 0.06, + "learning_rate": 1.9456183396198775e-07, + "loss": 0.7335, + "step": 1535 + }, + { + "epoch": 0.06, + "learning_rate": 1.9455817189731568e-07, + "loss": 0.6912, + "step": 1536 + }, + { + "epoch": 0.06, + "learning_rate": 1.9455450983264363e-07, + "loss": 0.6963, + "step": 1537 + }, + { + "epoch": 0.06, + "learning_rate": 1.9455084776797156e-07, + "loss": 0.6879, + "step": 1538 + }, + { + "epoch": 0.06, + "learning_rate": 1.9454718570329951e-07, + "loss": 0.6783, + "step": 1539 + }, + { + "epoch": 0.06, + "learning_rate": 1.9454352363862744e-07, + "loss": 0.7024, + "step": 1540 + }, + { + "epoch": 0.06, + "learning_rate": 1.945398615739554e-07, + "loss": 0.6846, + "step": 1541 + }, + { + "epoch": 0.06, + "learning_rate": 1.9453619950928332e-07, + "loss": 0.6926, + "step": 1542 + }, + { + "epoch": 0.06, + "learning_rate": 1.9453253744461127e-07, + "loss": 0.7114, + "step": 1543 + }, + { + "epoch": 0.06, + "learning_rate": 1.945288753799392e-07, + "loss": 0.6778, + "step": 1544 + }, + { + "epoch": 0.06, + "learning_rate": 1.9452521331526715e-07, + "loss": 0.6979, + "step": 1545 + }, + { + "epoch": 0.06, + "learning_rate": 1.9452155125059508e-07, + "loss": 0.693, + "step": 1546 + }, + { + "epoch": 0.06, + "learning_rate": 1.9451788918592303e-07, + "loss": 0.7103, + "step": 1547 + }, + { + "epoch": 0.06, + "learning_rate": 1.9451422712125096e-07, + "loss": 0.6875, + "step": 1548 + }, + { + "epoch": 0.06, + "learning_rate": 1.945105650565789e-07, + "loss": 0.7239, + "step": 1549 + }, + { + "epoch": 0.06, + "learning_rate": 1.9450690299190681e-07, + "loss": 0.7079, + "step": 1550 + }, + { + "epoch": 0.06, + "learning_rate": 1.9450324092723477e-07, + "loss": 0.6874, + "step": 1551 + }, + { + "epoch": 0.06, + "learning_rate": 1.944995788625627e-07, + "loss": 0.6899, + "step": 1552 + }, + { + "epoch": 0.06, + "learning_rate": 1.9449591679789065e-07, + "loss": 0.6903, + "step": 1553 + }, + { + "epoch": 0.06, + "learning_rate": 1.9449225473321857e-07, + "loss": 0.7251, + "step": 1554 + }, + { + "epoch": 0.06, + "learning_rate": 1.9448859266854653e-07, + "loss": 0.7049, + "step": 1555 + }, + { + "epoch": 0.06, + "learning_rate": 1.9448493060387445e-07, + "loss": 0.7009, + "step": 1556 + }, + { + "epoch": 0.06, + "learning_rate": 1.9448126853920238e-07, + "loss": 0.7167, + "step": 1557 + }, + { + "epoch": 0.06, + "learning_rate": 1.9447760647453033e-07, + "loss": 0.6982, + "step": 1558 + }, + { + "epoch": 0.06, + "learning_rate": 1.9447394440985826e-07, + "loss": 0.685, + "step": 1559 + }, + { + "epoch": 0.06, + "learning_rate": 1.944702823451862e-07, + "loss": 0.693, + "step": 1560 + }, + { + "epoch": 0.06, + "learning_rate": 1.9446662028051414e-07, + "loss": 0.6847, + "step": 1561 + }, + { + "epoch": 0.06, + "learning_rate": 1.944629582158421e-07, + "loss": 0.6937, + "step": 1562 + }, + { + "epoch": 0.06, + "learning_rate": 1.9445929615117002e-07, + "loss": 0.7062, + "step": 1563 + }, + { + "epoch": 0.06, + "learning_rate": 1.9445563408649797e-07, + "loss": 0.6698, + "step": 1564 + }, + { + "epoch": 0.06, + "learning_rate": 1.944519720218259e-07, + "loss": 0.7062, + "step": 1565 + }, + { + "epoch": 0.06, + "learning_rate": 1.9444830995715385e-07, + "loss": 0.6997, + "step": 1566 + }, + { + "epoch": 0.06, + "learning_rate": 1.9444464789248178e-07, + "loss": 0.6907, + "step": 1567 + }, + { + "epoch": 0.06, + "learning_rate": 1.944409858278097e-07, + "loss": 0.7073, + "step": 1568 + }, + { + "epoch": 0.06, + "learning_rate": 1.9443732376313763e-07, + "loss": 0.6979, + "step": 1569 + }, + { + "epoch": 0.06, + "learning_rate": 1.9443366169846559e-07, + "loss": 0.6982, + "step": 1570 + }, + { + "epoch": 0.06, + "learning_rate": 1.944299996337935e-07, + "loss": 0.6619, + "step": 1571 + }, + { + "epoch": 0.06, + "learning_rate": 1.9442633756912147e-07, + "loss": 0.6925, + "step": 1572 + }, + { + "epoch": 0.06, + "learning_rate": 1.944226755044494e-07, + "loss": 0.7137, + "step": 1573 + }, + { + "epoch": 0.06, + "learning_rate": 1.9441901343977735e-07, + "loss": 0.6913, + "step": 1574 + }, + { + "epoch": 0.06, + "learning_rate": 1.9441535137510527e-07, + "loss": 0.7053, + "step": 1575 + }, + { + "epoch": 0.06, + "learning_rate": 1.9441168931043323e-07, + "loss": 0.6761, + "step": 1576 + }, + { + "epoch": 0.06, + "learning_rate": 1.9440802724576115e-07, + "loss": 0.7059, + "step": 1577 + }, + { + "epoch": 0.06, + "learning_rate": 1.9440436518108908e-07, + "loss": 0.7081, + "step": 1578 + }, + { + "epoch": 0.06, + "learning_rate": 1.9440070311641703e-07, + "loss": 0.6918, + "step": 1579 + }, + { + "epoch": 0.06, + "learning_rate": 1.9439704105174496e-07, + "loss": 0.6817, + "step": 1580 + }, + { + "epoch": 0.06, + "learning_rate": 1.943933789870729e-07, + "loss": 0.6953, + "step": 1581 + }, + { + "epoch": 0.06, + "learning_rate": 1.9438971692240084e-07, + "loss": 0.7, + "step": 1582 + }, + { + "epoch": 0.06, + "learning_rate": 1.943860548577288e-07, + "loss": 0.7037, + "step": 1583 + }, + { + "epoch": 0.06, + "learning_rate": 1.9438239279305672e-07, + "loss": 0.6954, + "step": 1584 + }, + { + "epoch": 0.06, + "learning_rate": 1.9437873072838467e-07, + "loss": 0.7152, + "step": 1585 + }, + { + "epoch": 0.06, + "learning_rate": 1.9437506866371257e-07, + "loss": 0.6936, + "step": 1586 + }, + { + "epoch": 0.06, + "learning_rate": 1.9437140659904052e-07, + "loss": 0.7075, + "step": 1587 + }, + { + "epoch": 0.06, + "learning_rate": 1.9436774453436845e-07, + "loss": 0.6829, + "step": 1588 + }, + { + "epoch": 0.06, + "learning_rate": 1.943640824696964e-07, + "loss": 0.6979, + "step": 1589 + }, + { + "epoch": 0.06, + "learning_rate": 1.9436042040502433e-07, + "loss": 0.6952, + "step": 1590 + }, + { + "epoch": 0.06, + "learning_rate": 1.9435675834035228e-07, + "loss": 0.714, + "step": 1591 + }, + { + "epoch": 0.06, + "learning_rate": 1.943530962756802e-07, + "loss": 0.6823, + "step": 1592 + }, + { + "epoch": 0.06, + "learning_rate": 1.9434943421100816e-07, + "loss": 0.6857, + "step": 1593 + }, + { + "epoch": 0.06, + "learning_rate": 1.943457721463361e-07, + "loss": 0.7003, + "step": 1594 + }, + { + "epoch": 0.06, + "learning_rate": 1.9434211008166404e-07, + "loss": 0.6635, + "step": 1595 + }, + { + "epoch": 0.06, + "learning_rate": 1.9433844801699197e-07, + "loss": 0.7093, + "step": 1596 + }, + { + "epoch": 0.06, + "learning_rate": 1.9433478595231992e-07, + "loss": 0.7071, + "step": 1597 + }, + { + "epoch": 0.06, + "learning_rate": 1.9433112388764785e-07, + "loss": 0.6954, + "step": 1598 + }, + { + "epoch": 0.06, + "learning_rate": 1.943274618229758e-07, + "loss": 0.7079, + "step": 1599 + }, + { + "epoch": 0.06, + "learning_rate": 1.9432379975830373e-07, + "loss": 0.7083, + "step": 1600 + }, + { + "epoch": 0.06, + "learning_rate": 1.9432013769363166e-07, + "loss": 0.7074, + "step": 1601 + }, + { + "epoch": 0.06, + "learning_rate": 1.943164756289596e-07, + "loss": 0.6844, + "step": 1602 + }, + { + "epoch": 0.06, + "learning_rate": 1.9431281356428754e-07, + "loss": 0.6853, + "step": 1603 + }, + { + "epoch": 0.06, + "learning_rate": 1.9430915149961546e-07, + "loss": 0.7313, + "step": 1604 + }, + { + "epoch": 0.06, + "learning_rate": 1.9430548943494342e-07, + "loss": 0.6822, + "step": 1605 + }, + { + "epoch": 0.06, + "learning_rate": 1.9430182737027134e-07, + "loss": 0.6823, + "step": 1606 + }, + { + "epoch": 0.06, + "learning_rate": 1.9429816530559927e-07, + "loss": 0.67, + "step": 1607 + }, + { + "epoch": 0.06, + "learning_rate": 1.9429450324092722e-07, + "loss": 0.7024, + "step": 1608 + }, + { + "epoch": 0.06, + "learning_rate": 1.9429084117625515e-07, + "loss": 0.7039, + "step": 1609 + }, + { + "epoch": 0.06, + "learning_rate": 1.942871791115831e-07, + "loss": 0.6906, + "step": 1610 + }, + { + "epoch": 0.06, + "learning_rate": 1.9428351704691103e-07, + "loss": 0.7155, + "step": 1611 + }, + { + "epoch": 0.06, + "learning_rate": 1.9427985498223898e-07, + "loss": 0.6777, + "step": 1612 + }, + { + "epoch": 0.06, + "learning_rate": 1.942761929175669e-07, + "loss": 0.6913, + "step": 1613 + }, + { + "epoch": 0.06, + "learning_rate": 1.9427253085289486e-07, + "loss": 0.6853, + "step": 1614 + }, + { + "epoch": 0.06, + "learning_rate": 1.942688687882228e-07, + "loss": 0.6905, + "step": 1615 + }, + { + "epoch": 0.06, + "learning_rate": 1.9426520672355074e-07, + "loss": 0.7044, + "step": 1616 + }, + { + "epoch": 0.06, + "learning_rate": 1.9426154465887867e-07, + "loss": 0.6861, + "step": 1617 + }, + { + "epoch": 0.06, + "learning_rate": 1.9425788259420662e-07, + "loss": 0.6796, + "step": 1618 + }, + { + "epoch": 0.06, + "learning_rate": 1.9425422052953455e-07, + "loss": 0.6965, + "step": 1619 + }, + { + "epoch": 0.06, + "learning_rate": 1.942505584648625e-07, + "loss": 0.6993, + "step": 1620 + }, + { + "epoch": 0.06, + "learning_rate": 1.942468964001904e-07, + "loss": 0.6934, + "step": 1621 + }, + { + "epoch": 0.06, + "learning_rate": 1.9424323433551836e-07, + "loss": 0.6685, + "step": 1622 + }, + { + "epoch": 0.06, + "learning_rate": 1.9423957227084628e-07, + "loss": 0.6879, + "step": 1623 + }, + { + "epoch": 0.06, + "learning_rate": 1.9423591020617424e-07, + "loss": 0.69, + "step": 1624 + }, + { + "epoch": 0.06, + "learning_rate": 1.9423224814150216e-07, + "loss": 0.6966, + "step": 1625 + }, + { + "epoch": 0.06, + "learning_rate": 1.9422858607683012e-07, + "loss": 0.6831, + "step": 1626 + }, + { + "epoch": 0.06, + "learning_rate": 1.9422492401215804e-07, + "loss": 0.721, + "step": 1627 + }, + { + "epoch": 0.06, + "learning_rate": 1.9422126194748597e-07, + "loss": 0.6853, + "step": 1628 + }, + { + "epoch": 0.06, + "learning_rate": 1.9421759988281392e-07, + "loss": 0.6822, + "step": 1629 + }, + { + "epoch": 0.06, + "learning_rate": 1.9421393781814185e-07, + "loss": 0.7065, + "step": 1630 + }, + { + "epoch": 0.06, + "learning_rate": 1.942102757534698e-07, + "loss": 0.7028, + "step": 1631 + }, + { + "epoch": 0.06, + "learning_rate": 1.9420661368879773e-07, + "loss": 0.6851, + "step": 1632 + }, + { + "epoch": 0.06, + "learning_rate": 1.9420295162412568e-07, + "loss": 0.6796, + "step": 1633 + }, + { + "epoch": 0.06, + "learning_rate": 1.941992895594536e-07, + "loss": 0.6993, + "step": 1634 + }, + { + "epoch": 0.06, + "learning_rate": 1.9419562749478156e-07, + "loss": 0.6954, + "step": 1635 + }, + { + "epoch": 0.06, + "learning_rate": 1.941919654301095e-07, + "loss": 0.6984, + "step": 1636 + }, + { + "epoch": 0.06, + "learning_rate": 1.9418830336543744e-07, + "loss": 0.7038, + "step": 1637 + }, + { + "epoch": 0.06, + "learning_rate": 1.9418464130076537e-07, + "loss": 0.6929, + "step": 1638 + }, + { + "epoch": 0.06, + "learning_rate": 1.941809792360933e-07, + "loss": 0.7045, + "step": 1639 + }, + { + "epoch": 0.06, + "learning_rate": 1.9417731717142122e-07, + "loss": 0.7268, + "step": 1640 + }, + { + "epoch": 0.06, + "learning_rate": 1.9417365510674917e-07, + "loss": 0.7141, + "step": 1641 + }, + { + "epoch": 0.06, + "learning_rate": 1.941699930420771e-07, + "loss": 0.706, + "step": 1642 + }, + { + "epoch": 0.06, + "learning_rate": 1.9416633097740505e-07, + "loss": 0.6965, + "step": 1643 + }, + { + "epoch": 0.06, + "learning_rate": 1.9416266891273298e-07, + "loss": 0.7187, + "step": 1644 + }, + { + "epoch": 0.06, + "learning_rate": 1.9415900684806093e-07, + "loss": 0.6906, + "step": 1645 + }, + { + "epoch": 0.06, + "learning_rate": 1.9415534478338886e-07, + "loss": 0.6892, + "step": 1646 + }, + { + "epoch": 0.06, + "learning_rate": 1.9415168271871681e-07, + "loss": 0.6898, + "step": 1647 + }, + { + "epoch": 0.06, + "learning_rate": 1.9414802065404474e-07, + "loss": 0.7043, + "step": 1648 + }, + { + "epoch": 0.06, + "learning_rate": 1.9414435858937267e-07, + "loss": 0.6974, + "step": 1649 + }, + { + "epoch": 0.06, + "learning_rate": 1.9414069652470062e-07, + "loss": 0.7356, + "step": 1650 + }, + { + "epoch": 0.06, + "learning_rate": 1.9413703446002855e-07, + "loss": 0.6844, + "step": 1651 + }, + { + "epoch": 0.06, + "learning_rate": 1.941333723953565e-07, + "loss": 0.683, + "step": 1652 + }, + { + "epoch": 0.06, + "learning_rate": 1.9412971033068443e-07, + "loss": 0.6948, + "step": 1653 + }, + { + "epoch": 0.06, + "learning_rate": 1.9412604826601238e-07, + "loss": 0.6871, + "step": 1654 + }, + { + "epoch": 0.06, + "learning_rate": 1.941223862013403e-07, + "loss": 0.6855, + "step": 1655 + }, + { + "epoch": 0.06, + "learning_rate": 1.9411872413666826e-07, + "loss": 0.6819, + "step": 1656 + }, + { + "epoch": 0.06, + "learning_rate": 1.9411506207199616e-07, + "loss": 0.6847, + "step": 1657 + }, + { + "epoch": 0.06, + "learning_rate": 1.9411140000732411e-07, + "loss": 0.7043, + "step": 1658 + }, + { + "epoch": 0.06, + "learning_rate": 1.9410773794265204e-07, + "loss": 0.6858, + "step": 1659 + }, + { + "epoch": 0.06, + "learning_rate": 1.9410407587798e-07, + "loss": 0.6873, + "step": 1660 + }, + { + "epoch": 0.06, + "learning_rate": 1.9410041381330792e-07, + "loss": 0.6867, + "step": 1661 + }, + { + "epoch": 0.06, + "learning_rate": 1.9409675174863587e-07, + "loss": 0.6954, + "step": 1662 + }, + { + "epoch": 0.06, + "learning_rate": 1.940930896839638e-07, + "loss": 0.6874, + "step": 1663 + }, + { + "epoch": 0.06, + "learning_rate": 1.9408942761929175e-07, + "loss": 0.6916, + "step": 1664 + }, + { + "epoch": 0.06, + "learning_rate": 1.9408576555461968e-07, + "loss": 0.6887, + "step": 1665 + }, + { + "epoch": 0.06, + "learning_rate": 1.9408210348994763e-07, + "loss": 0.6886, + "step": 1666 + }, + { + "epoch": 0.06, + "learning_rate": 1.9407844142527556e-07, + "loss": 0.6998, + "step": 1667 + }, + { + "epoch": 0.06, + "learning_rate": 1.940747793606035e-07, + "loss": 0.7002, + "step": 1668 + }, + { + "epoch": 0.06, + "learning_rate": 1.9407111729593144e-07, + "loss": 0.6836, + "step": 1669 + }, + { + "epoch": 0.06, + "learning_rate": 1.940674552312594e-07, + "loss": 0.7013, + "step": 1670 + }, + { + "epoch": 0.06, + "learning_rate": 1.9406379316658732e-07, + "loss": 0.6793, + "step": 1671 + }, + { + "epoch": 0.06, + "learning_rate": 1.9406013110191525e-07, + "loss": 0.6981, + "step": 1672 + }, + { + "epoch": 0.06, + "learning_rate": 1.940564690372432e-07, + "loss": 0.6842, + "step": 1673 + }, + { + "epoch": 0.06, + "learning_rate": 1.9405280697257113e-07, + "loss": 0.7042, + "step": 1674 + }, + { + "epoch": 0.06, + "learning_rate": 1.9404914490789905e-07, + "loss": 0.7164, + "step": 1675 + }, + { + "epoch": 0.06, + "learning_rate": 1.94045482843227e-07, + "loss": 0.7069, + "step": 1676 + }, + { + "epoch": 0.06, + "learning_rate": 1.9404182077855493e-07, + "loss": 0.7084, + "step": 1677 + }, + { + "epoch": 0.06, + "learning_rate": 1.9403815871388286e-07, + "loss": 0.7237, + "step": 1678 + }, + { + "epoch": 0.06, + "learning_rate": 1.940344966492108e-07, + "loss": 0.6931, + "step": 1679 + }, + { + "epoch": 0.06, + "learning_rate": 1.9403083458453874e-07, + "loss": 0.6899, + "step": 1680 + }, + { + "epoch": 0.06, + "learning_rate": 1.940271725198667e-07, + "loss": 0.691, + "step": 1681 + }, + { + "epoch": 0.06, + "learning_rate": 1.9402351045519462e-07, + "loss": 0.6842, + "step": 1682 + }, + { + "epoch": 0.06, + "learning_rate": 1.9401984839052257e-07, + "loss": 0.6805, + "step": 1683 + }, + { + "epoch": 0.06, + "learning_rate": 1.940161863258505e-07, + "loss": 0.7187, + "step": 1684 + }, + { + "epoch": 0.06, + "learning_rate": 1.9401252426117845e-07, + "loss": 0.6933, + "step": 1685 + }, + { + "epoch": 0.06, + "learning_rate": 1.9400886219650638e-07, + "loss": 0.6863, + "step": 1686 + }, + { + "epoch": 0.06, + "learning_rate": 1.9400520013183433e-07, + "loss": 0.6748, + "step": 1687 + }, + { + "epoch": 0.06, + "learning_rate": 1.9400153806716226e-07, + "loss": 0.6906, + "step": 1688 + }, + { + "epoch": 0.06, + "learning_rate": 1.939978760024902e-07, + "loss": 0.7092, + "step": 1689 + }, + { + "epoch": 0.06, + "learning_rate": 1.9399421393781814e-07, + "loss": 0.698, + "step": 1690 + }, + { + "epoch": 0.06, + "learning_rate": 1.939905518731461e-07, + "loss": 0.6905, + "step": 1691 + }, + { + "epoch": 0.06, + "learning_rate": 1.9398688980847402e-07, + "loss": 0.6912, + "step": 1692 + }, + { + "epoch": 0.06, + "learning_rate": 1.9398322774380195e-07, + "loss": 0.6928, + "step": 1693 + }, + { + "epoch": 0.06, + "learning_rate": 1.9397956567912987e-07, + "loss": 0.6897, + "step": 1694 + }, + { + "epoch": 0.06, + "learning_rate": 1.9397590361445783e-07, + "loss": 0.6887, + "step": 1695 + }, + { + "epoch": 0.06, + "learning_rate": 1.9397224154978575e-07, + "loss": 0.6863, + "step": 1696 + }, + { + "epoch": 0.06, + "learning_rate": 1.939685794851137e-07, + "loss": 0.6875, + "step": 1697 + }, + { + "epoch": 0.06, + "learning_rate": 1.9396491742044163e-07, + "loss": 0.7088, + "step": 1698 + }, + { + "epoch": 0.06, + "learning_rate": 1.9396125535576956e-07, + "loss": 0.688, + "step": 1699 + }, + { + "epoch": 0.06, + "learning_rate": 1.939575932910975e-07, + "loss": 0.6976, + "step": 1700 + }, + { + "epoch": 0.06, + "learning_rate": 1.9395393122642544e-07, + "loss": 0.6903, + "step": 1701 + }, + { + "epoch": 0.06, + "learning_rate": 1.939502691617534e-07, + "loss": 0.7075, + "step": 1702 + }, + { + "epoch": 0.06, + "learning_rate": 1.9394660709708132e-07, + "loss": 0.705, + "step": 1703 + }, + { + "epoch": 0.06, + "learning_rate": 1.9394294503240927e-07, + "loss": 0.707, + "step": 1704 + }, + { + "epoch": 0.06, + "learning_rate": 1.939392829677372e-07, + "loss": 0.7081, + "step": 1705 + }, + { + "epoch": 0.06, + "learning_rate": 1.9393562090306515e-07, + "loss": 0.6746, + "step": 1706 + }, + { + "epoch": 0.06, + "learning_rate": 1.9393195883839308e-07, + "loss": 0.6897, + "step": 1707 + }, + { + "epoch": 0.06, + "learning_rate": 1.9392829677372103e-07, + "loss": 0.664, + "step": 1708 + }, + { + "epoch": 0.06, + "learning_rate": 1.9392463470904896e-07, + "loss": 0.7077, + "step": 1709 + }, + { + "epoch": 0.06, + "learning_rate": 1.939209726443769e-07, + "loss": 0.6876, + "step": 1710 + }, + { + "epoch": 0.06, + "learning_rate": 1.939173105797048e-07, + "loss": 0.716, + "step": 1711 + }, + { + "epoch": 0.06, + "learning_rate": 1.9391364851503276e-07, + "loss": 0.6964, + "step": 1712 + }, + { + "epoch": 0.06, + "learning_rate": 1.939099864503607e-07, + "loss": 0.6772, + "step": 1713 + }, + { + "epoch": 0.06, + "learning_rate": 1.9390632438568864e-07, + "loss": 0.6984, + "step": 1714 + }, + { + "epoch": 0.06, + "learning_rate": 1.9390266232101657e-07, + "loss": 0.7015, + "step": 1715 + }, + { + "epoch": 0.06, + "learning_rate": 1.9389900025634452e-07, + "loss": 0.6702, + "step": 1716 + }, + { + "epoch": 0.06, + "learning_rate": 1.9389533819167245e-07, + "loss": 0.6789, + "step": 1717 + }, + { + "epoch": 0.06, + "learning_rate": 1.938916761270004e-07, + "loss": 0.6867, + "step": 1718 + }, + { + "epoch": 0.06, + "learning_rate": 1.9388801406232833e-07, + "loss": 0.7079, + "step": 1719 + }, + { + "epoch": 0.06, + "learning_rate": 1.9388435199765626e-07, + "loss": 0.6953, + "step": 1720 + }, + { + "epoch": 0.06, + "learning_rate": 1.938806899329842e-07, + "loss": 0.6888, + "step": 1721 + }, + { + "epoch": 0.06, + "learning_rate": 1.9387702786831214e-07, + "loss": 0.6968, + "step": 1722 + }, + { + "epoch": 0.06, + "learning_rate": 1.938733658036401e-07, + "loss": 0.6698, + "step": 1723 + }, + { + "epoch": 0.06, + "learning_rate": 1.9386970373896802e-07, + "loss": 0.6827, + "step": 1724 + }, + { + "epoch": 0.06, + "learning_rate": 1.9386604167429597e-07, + "loss": 0.674, + "step": 1725 + }, + { + "epoch": 0.06, + "learning_rate": 1.938623796096239e-07, + "loss": 0.701, + "step": 1726 + }, + { + "epoch": 0.06, + "learning_rate": 1.9385871754495185e-07, + "loss": 0.6816, + "step": 1727 + }, + { + "epoch": 0.06, + "learning_rate": 1.9385505548027978e-07, + "loss": 0.695, + "step": 1728 + }, + { + "epoch": 0.06, + "learning_rate": 1.938513934156077e-07, + "loss": 0.7009, + "step": 1729 + }, + { + "epoch": 0.06, + "learning_rate": 1.9384773135093563e-07, + "loss": 0.6895, + "step": 1730 + }, + { + "epoch": 0.06, + "learning_rate": 1.9384406928626358e-07, + "loss": 0.7155, + "step": 1731 + }, + { + "epoch": 0.06, + "learning_rate": 1.938404072215915e-07, + "loss": 0.6888, + "step": 1732 + }, + { + "epoch": 0.06, + "learning_rate": 1.9383674515691946e-07, + "loss": 0.6875, + "step": 1733 + }, + { + "epoch": 0.06, + "learning_rate": 1.938330830922474e-07, + "loss": 0.6933, + "step": 1734 + }, + { + "epoch": 0.06, + "learning_rate": 1.9382942102757534e-07, + "loss": 0.715, + "step": 1735 + }, + { + "epoch": 0.06, + "learning_rate": 1.9382575896290327e-07, + "loss": 0.7055, + "step": 1736 + }, + { + "epoch": 0.06, + "learning_rate": 1.9382209689823122e-07, + "loss": 0.7161, + "step": 1737 + }, + { + "epoch": 0.06, + "learning_rate": 1.9381843483355915e-07, + "loss": 0.7013, + "step": 1738 + }, + { + "epoch": 0.06, + "learning_rate": 1.938147727688871e-07, + "loss": 0.6851, + "step": 1739 + }, + { + "epoch": 0.06, + "learning_rate": 1.9381111070421503e-07, + "loss": 0.6977, + "step": 1740 + }, + { + "epoch": 0.06, + "learning_rate": 1.9380744863954298e-07, + "loss": 0.697, + "step": 1741 + }, + { + "epoch": 0.06, + "learning_rate": 1.938037865748709e-07, + "loss": 0.685, + "step": 1742 + }, + { + "epoch": 0.06, + "learning_rate": 1.9380012451019884e-07, + "loss": 0.7142, + "step": 1743 + }, + { + "epoch": 0.06, + "learning_rate": 1.937964624455268e-07, + "loss": 0.6905, + "step": 1744 + }, + { + "epoch": 0.06, + "learning_rate": 1.9379280038085472e-07, + "loss": 0.6883, + "step": 1745 + }, + { + "epoch": 0.06, + "learning_rate": 1.9378913831618267e-07, + "loss": 0.6951, + "step": 1746 + }, + { + "epoch": 0.06, + "learning_rate": 1.937854762515106e-07, + "loss": 0.6945, + "step": 1747 + }, + { + "epoch": 0.06, + "learning_rate": 1.9378181418683852e-07, + "loss": 0.6795, + "step": 1748 + }, + { + "epoch": 0.06, + "learning_rate": 1.9377815212216645e-07, + "loss": 0.7026, + "step": 1749 + }, + { + "epoch": 0.06, + "learning_rate": 1.937744900574944e-07, + "loss": 0.6671, + "step": 1750 + }, + { + "epoch": 0.06, + "learning_rate": 1.9377082799282233e-07, + "loss": 0.6847, + "step": 1751 + }, + { + "epoch": 0.06, + "learning_rate": 1.9376716592815028e-07, + "loss": 0.7102, + "step": 1752 + }, + { + "epoch": 0.06, + "learning_rate": 1.937635038634782e-07, + "loss": 0.6843, + "step": 1753 + }, + { + "epoch": 0.06, + "learning_rate": 1.9375984179880616e-07, + "loss": 0.6848, + "step": 1754 + }, + { + "epoch": 0.06, + "learning_rate": 1.937561797341341e-07, + "loss": 0.6897, + "step": 1755 + }, + { + "epoch": 0.06, + "learning_rate": 1.9375251766946204e-07, + "loss": 0.696, + "step": 1756 + }, + { + "epoch": 0.06, + "learning_rate": 1.9374885560478997e-07, + "loss": 0.7037, + "step": 1757 + }, + { + "epoch": 0.06, + "learning_rate": 1.9374519354011792e-07, + "loss": 0.6932, + "step": 1758 + }, + { + "epoch": 0.06, + "learning_rate": 1.9374153147544585e-07, + "loss": 0.701, + "step": 1759 + }, + { + "epoch": 0.06, + "learning_rate": 1.937378694107738e-07, + "loss": 0.6886, + "step": 1760 + }, + { + "epoch": 0.06, + "learning_rate": 1.9373420734610173e-07, + "loss": 0.7189, + "step": 1761 + }, + { + "epoch": 0.06, + "learning_rate": 1.9373054528142968e-07, + "loss": 0.6828, + "step": 1762 + }, + { + "epoch": 0.06, + "learning_rate": 1.937268832167576e-07, + "loss": 0.7202, + "step": 1763 + }, + { + "epoch": 0.06, + "learning_rate": 1.9372322115208553e-07, + "loss": 0.6766, + "step": 1764 + }, + { + "epoch": 0.06, + "learning_rate": 1.9371955908741346e-07, + "loss": 0.718, + "step": 1765 + }, + { + "epoch": 0.06, + "learning_rate": 1.9371589702274141e-07, + "loss": 0.6983, + "step": 1766 + }, + { + "epoch": 0.06, + "learning_rate": 1.9371223495806934e-07, + "loss": 0.7117, + "step": 1767 + }, + { + "epoch": 0.06, + "learning_rate": 1.937085728933973e-07, + "loss": 0.6893, + "step": 1768 + }, + { + "epoch": 0.06, + "learning_rate": 1.9370491082872522e-07, + "loss": 0.6892, + "step": 1769 + }, + { + "epoch": 0.06, + "learning_rate": 1.9370124876405315e-07, + "loss": 0.7155, + "step": 1770 + }, + { + "epoch": 0.06, + "learning_rate": 1.936975866993811e-07, + "loss": 0.6909, + "step": 1771 + }, + { + "epoch": 0.06, + "learning_rate": 1.9369392463470903e-07, + "loss": 0.6796, + "step": 1772 + }, + { + "epoch": 0.06, + "learning_rate": 1.9369026257003698e-07, + "loss": 0.7121, + "step": 1773 + }, + { + "epoch": 0.06, + "learning_rate": 1.936866005053649e-07, + "loss": 0.7056, + "step": 1774 + }, + { + "epoch": 0.06, + "learning_rate": 1.9368293844069286e-07, + "loss": 0.6962, + "step": 1775 + }, + { + "epoch": 0.06, + "learning_rate": 1.936792763760208e-07, + "loss": 0.6984, + "step": 1776 + }, + { + "epoch": 0.07, + "learning_rate": 1.9367561431134874e-07, + "loss": 0.6854, + "step": 1777 + }, + { + "epoch": 0.07, + "learning_rate": 1.9367195224667667e-07, + "loss": 0.6917, + "step": 1778 + }, + { + "epoch": 0.07, + "learning_rate": 1.9366829018200462e-07, + "loss": 0.714, + "step": 1779 + }, + { + "epoch": 0.07, + "learning_rate": 1.9366462811733255e-07, + "loss": 0.6867, + "step": 1780 + }, + { + "epoch": 0.07, + "learning_rate": 1.936609660526605e-07, + "loss": 0.69, + "step": 1781 + }, + { + "epoch": 0.07, + "learning_rate": 1.9365730398798843e-07, + "loss": 0.6921, + "step": 1782 + }, + { + "epoch": 0.07, + "learning_rate": 1.9365364192331635e-07, + "loss": 0.6922, + "step": 1783 + }, + { + "epoch": 0.07, + "learning_rate": 1.9364997985864428e-07, + "loss": 0.6887, + "step": 1784 + }, + { + "epoch": 0.07, + "learning_rate": 1.9364631779397223e-07, + "loss": 0.6757, + "step": 1785 + }, + { + "epoch": 0.07, + "learning_rate": 1.9364265572930016e-07, + "loss": 0.6834, + "step": 1786 + }, + { + "epoch": 0.07, + "learning_rate": 1.9363899366462811e-07, + "loss": 0.6859, + "step": 1787 + }, + { + "epoch": 0.07, + "learning_rate": 1.9363533159995604e-07, + "loss": 0.6895, + "step": 1788 + }, + { + "epoch": 0.07, + "learning_rate": 1.93631669535284e-07, + "loss": 0.6997, + "step": 1789 + }, + { + "epoch": 0.07, + "learning_rate": 1.9362800747061192e-07, + "loss": 0.7034, + "step": 1790 + }, + { + "epoch": 0.07, + "learning_rate": 1.9362434540593985e-07, + "loss": 0.6906, + "step": 1791 + }, + { + "epoch": 0.07, + "learning_rate": 1.936206833412678e-07, + "loss": 0.7036, + "step": 1792 + }, + { + "epoch": 0.07, + "learning_rate": 1.9361702127659573e-07, + "loss": 0.6841, + "step": 1793 + }, + { + "epoch": 0.07, + "learning_rate": 1.9361335921192368e-07, + "loss": 0.6977, + "step": 1794 + }, + { + "epoch": 0.07, + "learning_rate": 1.936096971472516e-07, + "loss": 0.7134, + "step": 1795 + }, + { + "epoch": 0.07, + "learning_rate": 1.9360603508257956e-07, + "loss": 0.6982, + "step": 1796 + }, + { + "epoch": 0.07, + "learning_rate": 1.9360237301790749e-07, + "loss": 0.7094, + "step": 1797 + }, + { + "epoch": 0.07, + "learning_rate": 1.9359871095323544e-07, + "loss": 0.6996, + "step": 1798 + }, + { + "epoch": 0.07, + "learning_rate": 1.9359504888856337e-07, + "loss": 0.6936, + "step": 1799 + }, + { + "epoch": 0.07, + "learning_rate": 1.9359138682389132e-07, + "loss": 0.6762, + "step": 1800 + }, + { + "epoch": 0.07, + "learning_rate": 1.9358772475921922e-07, + "loss": 0.6683, + "step": 1801 + }, + { + "epoch": 0.07, + "learning_rate": 1.9358406269454717e-07, + "loss": 0.7087, + "step": 1802 + }, + { + "epoch": 0.07, + "learning_rate": 1.935804006298751e-07, + "loss": 0.7006, + "step": 1803 + }, + { + "epoch": 0.07, + "learning_rate": 1.9357673856520305e-07, + "loss": 0.6815, + "step": 1804 + }, + { + "epoch": 0.07, + "learning_rate": 1.9357307650053098e-07, + "loss": 0.7025, + "step": 1805 + }, + { + "epoch": 0.07, + "learning_rate": 1.9356941443585893e-07, + "loss": 0.6763, + "step": 1806 + }, + { + "epoch": 0.07, + "learning_rate": 1.9356575237118686e-07, + "loss": 0.6788, + "step": 1807 + }, + { + "epoch": 0.07, + "learning_rate": 1.935620903065148e-07, + "loss": 0.7129, + "step": 1808 + }, + { + "epoch": 0.07, + "learning_rate": 1.9355842824184274e-07, + "loss": 0.6782, + "step": 1809 + }, + { + "epoch": 0.07, + "learning_rate": 1.935547661771707e-07, + "loss": 0.7053, + "step": 1810 + }, + { + "epoch": 0.07, + "learning_rate": 1.9355110411249862e-07, + "loss": 0.6995, + "step": 1811 + }, + { + "epoch": 0.07, + "learning_rate": 1.9354744204782657e-07, + "loss": 0.7004, + "step": 1812 + }, + { + "epoch": 0.07, + "learning_rate": 1.935437799831545e-07, + "loss": 0.6929, + "step": 1813 + }, + { + "epoch": 0.07, + "learning_rate": 1.9354011791848243e-07, + "loss": 0.6842, + "step": 1814 + }, + { + "epoch": 0.07, + "learning_rate": 1.9353645585381038e-07, + "loss": 0.6882, + "step": 1815 + }, + { + "epoch": 0.07, + "learning_rate": 1.935327937891383e-07, + "loss": 0.6923, + "step": 1816 + }, + { + "epoch": 0.07, + "learning_rate": 1.9352913172446626e-07, + "loss": 0.7029, + "step": 1817 + }, + { + "epoch": 0.07, + "learning_rate": 1.9352546965979418e-07, + "loss": 0.6862, + "step": 1818 + }, + { + "epoch": 0.07, + "learning_rate": 1.935218075951221e-07, + "loss": 0.6899, + "step": 1819 + }, + { + "epoch": 0.07, + "learning_rate": 1.9351814553045004e-07, + "loss": 0.686, + "step": 1820 + }, + { + "epoch": 0.07, + "learning_rate": 1.93514483465778e-07, + "loss": 0.6976, + "step": 1821 + }, + { + "epoch": 0.07, + "learning_rate": 1.9351082140110592e-07, + "loss": 0.6944, + "step": 1822 + }, + { + "epoch": 0.07, + "learning_rate": 1.9350715933643387e-07, + "loss": 0.7058, + "step": 1823 + }, + { + "epoch": 0.07, + "learning_rate": 1.935034972717618e-07, + "loss": 0.7055, + "step": 1824 + }, + { + "epoch": 0.07, + "learning_rate": 1.9349983520708975e-07, + "loss": 0.6705, + "step": 1825 + }, + { + "epoch": 0.07, + "learning_rate": 1.9349617314241768e-07, + "loss": 0.7055, + "step": 1826 + }, + { + "epoch": 0.07, + "learning_rate": 1.9349251107774563e-07, + "loss": 0.6895, + "step": 1827 + }, + { + "epoch": 0.07, + "learning_rate": 1.9348884901307356e-07, + "loss": 0.694, + "step": 1828 + }, + { + "epoch": 0.07, + "learning_rate": 1.934851869484015e-07, + "loss": 0.702, + "step": 1829 + }, + { + "epoch": 0.07, + "learning_rate": 1.9348152488372944e-07, + "loss": 0.709, + "step": 1830 + }, + { + "epoch": 0.07, + "learning_rate": 1.934778628190574e-07, + "loss": 0.6891, + "step": 1831 + }, + { + "epoch": 0.07, + "learning_rate": 1.9347420075438532e-07, + "loss": 0.6871, + "step": 1832 + }, + { + "epoch": 0.07, + "learning_rate": 1.9347053868971327e-07, + "loss": 0.7012, + "step": 1833 + }, + { + "epoch": 0.07, + "learning_rate": 1.934668766250412e-07, + "loss": 0.6896, + "step": 1834 + }, + { + "epoch": 0.07, + "learning_rate": 1.9346321456036912e-07, + "loss": 0.6902, + "step": 1835 + }, + { + "epoch": 0.07, + "learning_rate": 1.9345955249569708e-07, + "loss": 0.6837, + "step": 1836 + }, + { + "epoch": 0.07, + "learning_rate": 1.93455890431025e-07, + "loss": 0.6951, + "step": 1837 + }, + { + "epoch": 0.07, + "learning_rate": 1.9345222836635293e-07, + "loss": 0.7071, + "step": 1838 + }, + { + "epoch": 0.07, + "learning_rate": 1.9344856630168088e-07, + "loss": 0.6963, + "step": 1839 + }, + { + "epoch": 0.07, + "learning_rate": 1.934449042370088e-07, + "loss": 0.6919, + "step": 1840 + }, + { + "epoch": 0.07, + "learning_rate": 1.9344124217233674e-07, + "loss": 0.6982, + "step": 1841 + }, + { + "epoch": 0.07, + "learning_rate": 1.934375801076647e-07, + "loss": 0.685, + "step": 1842 + }, + { + "epoch": 0.07, + "learning_rate": 1.9343391804299262e-07, + "loss": 0.7083, + "step": 1843 + }, + { + "epoch": 0.07, + "learning_rate": 1.9343025597832057e-07, + "loss": 0.694, + "step": 1844 + }, + { + "epoch": 0.07, + "learning_rate": 1.934265939136485e-07, + "loss": 0.677, + "step": 1845 + }, + { + "epoch": 0.07, + "learning_rate": 1.9342293184897645e-07, + "loss": 0.6885, + "step": 1846 + }, + { + "epoch": 0.07, + "learning_rate": 1.9341926978430438e-07, + "loss": 0.7103, + "step": 1847 + }, + { + "epoch": 0.07, + "learning_rate": 1.9341560771963233e-07, + "loss": 0.7046, + "step": 1848 + }, + { + "epoch": 0.07, + "learning_rate": 1.9341194565496026e-07, + "loss": 0.6759, + "step": 1849 + }, + { + "epoch": 0.07, + "learning_rate": 1.934082835902882e-07, + "loss": 0.6915, + "step": 1850 + }, + { + "epoch": 0.07, + "learning_rate": 1.9340462152561614e-07, + "loss": 0.6827, + "step": 1851 + }, + { + "epoch": 0.07, + "learning_rate": 1.934009594609441e-07, + "loss": 0.6953, + "step": 1852 + }, + { + "epoch": 0.07, + "learning_rate": 1.9339729739627202e-07, + "loss": 0.71, + "step": 1853 + }, + { + "epoch": 0.07, + "learning_rate": 1.9339363533159997e-07, + "loss": 0.7093, + "step": 1854 + }, + { + "epoch": 0.07, + "learning_rate": 1.9338997326692787e-07, + "loss": 0.7056, + "step": 1855 + }, + { + "epoch": 0.07, + "learning_rate": 1.9338631120225582e-07, + "loss": 0.6864, + "step": 1856 + }, + { + "epoch": 0.07, + "learning_rate": 1.9338264913758375e-07, + "loss": 0.7033, + "step": 1857 + }, + { + "epoch": 0.07, + "learning_rate": 1.933789870729117e-07, + "loss": 0.6957, + "step": 1858 + }, + { + "epoch": 0.07, + "learning_rate": 1.9337532500823963e-07, + "loss": 0.6967, + "step": 1859 + }, + { + "epoch": 0.07, + "learning_rate": 1.9337166294356758e-07, + "loss": 0.7138, + "step": 1860 + }, + { + "epoch": 0.07, + "learning_rate": 1.933680008788955e-07, + "loss": 0.6876, + "step": 1861 + }, + { + "epoch": 0.07, + "learning_rate": 1.9336433881422344e-07, + "loss": 0.7056, + "step": 1862 + }, + { + "epoch": 0.07, + "learning_rate": 1.933606767495514e-07, + "loss": 0.6859, + "step": 1863 + }, + { + "epoch": 0.07, + "learning_rate": 1.9335701468487932e-07, + "loss": 0.6726, + "step": 1864 + }, + { + "epoch": 0.07, + "learning_rate": 1.9335335262020727e-07, + "loss": 0.6715, + "step": 1865 + }, + { + "epoch": 0.07, + "learning_rate": 1.933496905555352e-07, + "loss": 0.6956, + "step": 1866 + }, + { + "epoch": 0.07, + "learning_rate": 1.9334602849086315e-07, + "loss": 0.7021, + "step": 1867 + }, + { + "epoch": 0.07, + "learning_rate": 1.9334236642619108e-07, + "loss": 0.7149, + "step": 1868 + }, + { + "epoch": 0.07, + "learning_rate": 1.9333870436151903e-07, + "loss": 0.6801, + "step": 1869 + }, + { + "epoch": 0.07, + "learning_rate": 1.9333504229684696e-07, + "loss": 0.6741, + "step": 1870 + }, + { + "epoch": 0.07, + "learning_rate": 1.933313802321749e-07, + "loss": 0.7217, + "step": 1871 + }, + { + "epoch": 0.07, + "learning_rate": 1.9332771816750284e-07, + "loss": 0.7108, + "step": 1872 + }, + { + "epoch": 0.07, + "learning_rate": 1.9332405610283076e-07, + "loss": 0.6921, + "step": 1873 + }, + { + "epoch": 0.07, + "learning_rate": 1.933203940381587e-07, + "loss": 0.6967, + "step": 1874 + }, + { + "epoch": 0.07, + "learning_rate": 1.9331673197348664e-07, + "loss": 0.6865, + "step": 1875 + }, + { + "epoch": 0.07, + "learning_rate": 1.9331306990881457e-07, + "loss": 0.6997, + "step": 1876 + }, + { + "epoch": 0.07, + "learning_rate": 1.9330940784414252e-07, + "loss": 0.6593, + "step": 1877 + }, + { + "epoch": 0.07, + "learning_rate": 1.9330574577947045e-07, + "loss": 0.6871, + "step": 1878 + }, + { + "epoch": 0.07, + "learning_rate": 1.933020837147984e-07, + "loss": 0.6852, + "step": 1879 + }, + { + "epoch": 0.07, + "learning_rate": 1.9329842165012633e-07, + "loss": 0.6945, + "step": 1880 + }, + { + "epoch": 0.07, + "learning_rate": 1.9329475958545428e-07, + "loss": 0.6825, + "step": 1881 + }, + { + "epoch": 0.07, + "learning_rate": 1.932910975207822e-07, + "loss": 0.6791, + "step": 1882 + }, + { + "epoch": 0.07, + "learning_rate": 1.9328743545611016e-07, + "loss": 0.6879, + "step": 1883 + }, + { + "epoch": 0.07, + "learning_rate": 1.932837733914381e-07, + "loss": 0.7355, + "step": 1884 + }, + { + "epoch": 0.07, + "learning_rate": 1.9328011132676601e-07, + "loss": 0.6936, + "step": 1885 + }, + { + "epoch": 0.07, + "learning_rate": 1.9327644926209397e-07, + "loss": 0.7023, + "step": 1886 + }, + { + "epoch": 0.07, + "learning_rate": 1.932727871974219e-07, + "loss": 0.7002, + "step": 1887 + }, + { + "epoch": 0.07, + "learning_rate": 1.9326912513274985e-07, + "loss": 0.6991, + "step": 1888 + }, + { + "epoch": 0.07, + "learning_rate": 1.9326546306807777e-07, + "loss": 0.7011, + "step": 1889 + }, + { + "epoch": 0.07, + "learning_rate": 1.9326180100340573e-07, + "loss": 0.7078, + "step": 1890 + }, + { + "epoch": 0.07, + "learning_rate": 1.9325813893873363e-07, + "loss": 0.688, + "step": 1891 + }, + { + "epoch": 0.07, + "learning_rate": 1.9325447687406158e-07, + "loss": 0.6699, + "step": 1892 + }, + { + "epoch": 0.07, + "learning_rate": 1.932508148093895e-07, + "loss": 0.6874, + "step": 1893 + }, + { + "epoch": 0.07, + "learning_rate": 1.9324715274471746e-07, + "loss": 0.6693, + "step": 1894 + }, + { + "epoch": 0.07, + "learning_rate": 1.932434906800454e-07, + "loss": 0.6981, + "step": 1895 + }, + { + "epoch": 0.07, + "learning_rate": 1.9323982861537334e-07, + "loss": 0.6986, + "step": 1896 + }, + { + "epoch": 0.07, + "learning_rate": 1.9323616655070127e-07, + "loss": 0.699, + "step": 1897 + }, + { + "epoch": 0.07, + "learning_rate": 1.9323250448602922e-07, + "loss": 0.6861, + "step": 1898 + }, + { + "epoch": 0.07, + "learning_rate": 1.9322884242135715e-07, + "loss": 0.7171, + "step": 1899 + }, + { + "epoch": 0.07, + "learning_rate": 1.932251803566851e-07, + "loss": 0.6904, + "step": 1900 + }, + { + "epoch": 0.07, + "learning_rate": 1.9322151829201303e-07, + "loss": 0.6936, + "step": 1901 + }, + { + "epoch": 0.07, + "learning_rate": 1.9321785622734098e-07, + "loss": 0.6883, + "step": 1902 + }, + { + "epoch": 0.07, + "learning_rate": 1.932141941626689e-07, + "loss": 0.6891, + "step": 1903 + }, + { + "epoch": 0.07, + "learning_rate": 1.9321053209799686e-07, + "loss": 0.7032, + "step": 1904 + }, + { + "epoch": 0.07, + "learning_rate": 1.9320687003332479e-07, + "loss": 0.6832, + "step": 1905 + }, + { + "epoch": 0.07, + "learning_rate": 1.9320320796865271e-07, + "loss": 0.7016, + "step": 1906 + }, + { + "epoch": 0.07, + "learning_rate": 1.9319954590398067e-07, + "loss": 0.7151, + "step": 1907 + }, + { + "epoch": 0.07, + "learning_rate": 1.931958838393086e-07, + "loss": 0.6991, + "step": 1908 + }, + { + "epoch": 0.07, + "learning_rate": 1.9319222177463652e-07, + "loss": 0.7143, + "step": 1909 + }, + { + "epoch": 0.07, + "learning_rate": 1.9318855970996447e-07, + "loss": 0.6668, + "step": 1910 + }, + { + "epoch": 0.07, + "learning_rate": 1.931848976452924e-07, + "loss": 0.6839, + "step": 1911 + }, + { + "epoch": 0.07, + "learning_rate": 1.9318123558062033e-07, + "loss": 0.6979, + "step": 1912 + }, + { + "epoch": 0.07, + "learning_rate": 1.9317757351594828e-07, + "loss": 0.6953, + "step": 1913 + }, + { + "epoch": 0.07, + "learning_rate": 1.931739114512762e-07, + "loss": 0.6905, + "step": 1914 + }, + { + "epoch": 0.07, + "learning_rate": 1.9317024938660416e-07, + "loss": 0.6998, + "step": 1915 + }, + { + "epoch": 0.07, + "learning_rate": 1.9316658732193209e-07, + "loss": 0.6873, + "step": 1916 + }, + { + "epoch": 0.07, + "learning_rate": 1.9316292525726004e-07, + "loss": 0.6822, + "step": 1917 + }, + { + "epoch": 0.07, + "learning_rate": 1.9315926319258797e-07, + "loss": 0.6882, + "step": 1918 + }, + { + "epoch": 0.07, + "learning_rate": 1.9315560112791592e-07, + "loss": 0.6835, + "step": 1919 + }, + { + "epoch": 0.07, + "learning_rate": 1.9315193906324385e-07, + "loss": 0.699, + "step": 1920 + }, + { + "epoch": 0.07, + "learning_rate": 1.931482769985718e-07, + "loss": 0.6885, + "step": 1921 + }, + { + "epoch": 0.07, + "learning_rate": 1.9314461493389973e-07, + "loss": 0.7041, + "step": 1922 + }, + { + "epoch": 0.07, + "learning_rate": 1.9314095286922768e-07, + "loss": 0.713, + "step": 1923 + }, + { + "epoch": 0.07, + "learning_rate": 1.931372908045556e-07, + "loss": 0.6896, + "step": 1924 + }, + { + "epoch": 0.07, + "learning_rate": 1.9313362873988356e-07, + "loss": 0.6914, + "step": 1925 + }, + { + "epoch": 0.07, + "learning_rate": 1.9312996667521149e-07, + "loss": 0.6802, + "step": 1926 + }, + { + "epoch": 0.07, + "learning_rate": 1.931263046105394e-07, + "loss": 0.6869, + "step": 1927 + }, + { + "epoch": 0.07, + "learning_rate": 1.9312264254586734e-07, + "loss": 0.7113, + "step": 1928 + }, + { + "epoch": 0.07, + "learning_rate": 1.931189804811953e-07, + "loss": 0.7018, + "step": 1929 + }, + { + "epoch": 0.07, + "learning_rate": 1.9311531841652322e-07, + "loss": 0.6817, + "step": 1930 + }, + { + "epoch": 0.07, + "learning_rate": 1.9311165635185117e-07, + "loss": 0.6964, + "step": 1931 + }, + { + "epoch": 0.07, + "learning_rate": 1.931079942871791e-07, + "loss": 0.7177, + "step": 1932 + }, + { + "epoch": 0.07, + "learning_rate": 1.9310433222250703e-07, + "loss": 0.6719, + "step": 1933 + }, + { + "epoch": 0.07, + "learning_rate": 1.9310067015783498e-07, + "loss": 0.689, + "step": 1934 + }, + { + "epoch": 0.07, + "learning_rate": 1.930970080931629e-07, + "loss": 0.655, + "step": 1935 + }, + { + "epoch": 0.07, + "learning_rate": 1.9309334602849086e-07, + "loss": 0.7146, + "step": 1936 + }, + { + "epoch": 0.07, + "learning_rate": 1.9308968396381878e-07, + "loss": 0.6914, + "step": 1937 + }, + { + "epoch": 0.07, + "learning_rate": 1.9308602189914674e-07, + "loss": 0.6974, + "step": 1938 + }, + { + "epoch": 0.07, + "learning_rate": 1.9308235983447466e-07, + "loss": 0.7084, + "step": 1939 + }, + { + "epoch": 0.07, + "learning_rate": 1.9307869776980262e-07, + "loss": 0.6757, + "step": 1940 + }, + { + "epoch": 0.07, + "learning_rate": 1.9307503570513054e-07, + "loss": 0.6969, + "step": 1941 + }, + { + "epoch": 0.07, + "learning_rate": 1.930713736404585e-07, + "loss": 0.7298, + "step": 1942 + }, + { + "epoch": 0.07, + "learning_rate": 1.9306771157578642e-07, + "loss": 0.7284, + "step": 1943 + }, + { + "epoch": 0.07, + "learning_rate": 1.9306404951111438e-07, + "loss": 0.7159, + "step": 1944 + }, + { + "epoch": 0.07, + "learning_rate": 1.9306038744644228e-07, + "loss": 0.7077, + "step": 1945 + }, + { + "epoch": 0.07, + "learning_rate": 1.9305672538177023e-07, + "loss": 0.7177, + "step": 1946 + }, + { + "epoch": 0.07, + "learning_rate": 1.9305306331709816e-07, + "loss": 0.6773, + "step": 1947 + }, + { + "epoch": 0.07, + "learning_rate": 1.930494012524261e-07, + "loss": 0.694, + "step": 1948 + }, + { + "epoch": 0.07, + "learning_rate": 1.9304573918775404e-07, + "loss": 0.7089, + "step": 1949 + }, + { + "epoch": 0.07, + "learning_rate": 1.93042077123082e-07, + "loss": 0.6975, + "step": 1950 + }, + { + "epoch": 0.07, + "learning_rate": 1.9303841505840992e-07, + "loss": 0.6898, + "step": 1951 + }, + { + "epoch": 0.07, + "learning_rate": 1.9303475299373787e-07, + "loss": 0.7213, + "step": 1952 + }, + { + "epoch": 0.07, + "learning_rate": 1.930310909290658e-07, + "loss": 0.6919, + "step": 1953 + }, + { + "epoch": 0.07, + "learning_rate": 1.9302742886439375e-07, + "loss": 0.6991, + "step": 1954 + }, + { + "epoch": 0.07, + "learning_rate": 1.9302376679972168e-07, + "loss": 0.7031, + "step": 1955 + }, + { + "epoch": 0.07, + "learning_rate": 1.930201047350496e-07, + "loss": 0.7095, + "step": 1956 + }, + { + "epoch": 0.07, + "learning_rate": 1.9301644267037756e-07, + "loss": 0.6908, + "step": 1957 + }, + { + "epoch": 0.07, + "learning_rate": 1.9301278060570548e-07, + "loss": 0.6982, + "step": 1958 + }, + { + "epoch": 0.07, + "learning_rate": 1.9300911854103344e-07, + "loss": 0.6991, + "step": 1959 + }, + { + "epoch": 0.07, + "learning_rate": 1.9300545647636136e-07, + "loss": 0.7062, + "step": 1960 + }, + { + "epoch": 0.07, + "learning_rate": 1.9300179441168932e-07, + "loss": 0.7142, + "step": 1961 + }, + { + "epoch": 0.07, + "learning_rate": 1.9299813234701724e-07, + "loss": 0.683, + "step": 1962 + }, + { + "epoch": 0.07, + "learning_rate": 1.9299447028234517e-07, + "loss": 0.7117, + "step": 1963 + }, + { + "epoch": 0.07, + "learning_rate": 1.929908082176731e-07, + "loss": 0.6931, + "step": 1964 + }, + { + "epoch": 0.07, + "learning_rate": 1.9298714615300105e-07, + "loss": 0.6927, + "step": 1965 + }, + { + "epoch": 0.07, + "learning_rate": 1.9298348408832898e-07, + "loss": 0.7064, + "step": 1966 + }, + { + "epoch": 0.07, + "learning_rate": 1.9297982202365693e-07, + "loss": 0.6894, + "step": 1967 + }, + { + "epoch": 0.07, + "learning_rate": 1.9297615995898486e-07, + "loss": 0.6915, + "step": 1968 + }, + { + "epoch": 0.07, + "learning_rate": 1.929724978943128e-07, + "loss": 0.701, + "step": 1969 + }, + { + "epoch": 0.07, + "learning_rate": 1.9296883582964074e-07, + "loss": 0.7037, + "step": 1970 + }, + { + "epoch": 0.07, + "learning_rate": 1.929651737649687e-07, + "loss": 0.6972, + "step": 1971 + }, + { + "epoch": 0.07, + "learning_rate": 1.9296151170029662e-07, + "loss": 0.6857, + "step": 1972 + }, + { + "epoch": 0.07, + "learning_rate": 1.9295784963562457e-07, + "loss": 0.6818, + "step": 1973 + }, + { + "epoch": 0.07, + "learning_rate": 1.929541875709525e-07, + "loss": 0.7003, + "step": 1974 + }, + { + "epoch": 0.07, + "learning_rate": 1.9295052550628045e-07, + "loss": 0.6851, + "step": 1975 + }, + { + "epoch": 0.07, + "learning_rate": 1.9294686344160838e-07, + "loss": 0.7033, + "step": 1976 + }, + { + "epoch": 0.07, + "learning_rate": 1.929432013769363e-07, + "loss": 0.7094, + "step": 1977 + }, + { + "epoch": 0.07, + "learning_rate": 1.9293953931226426e-07, + "loss": 0.709, + "step": 1978 + }, + { + "epoch": 0.07, + "learning_rate": 1.9293587724759218e-07, + "loss": 0.6973, + "step": 1979 + }, + { + "epoch": 0.07, + "learning_rate": 1.929322151829201e-07, + "loss": 0.6991, + "step": 1980 + }, + { + "epoch": 0.07, + "learning_rate": 1.9292855311824806e-07, + "loss": 0.6944, + "step": 1981 + }, + { + "epoch": 0.07, + "learning_rate": 1.92924891053576e-07, + "loss": 0.7065, + "step": 1982 + }, + { + "epoch": 0.07, + "learning_rate": 1.9292122898890392e-07, + "loss": 0.688, + "step": 1983 + }, + { + "epoch": 0.07, + "learning_rate": 1.9291756692423187e-07, + "loss": 0.699, + "step": 1984 + }, + { + "epoch": 0.07, + "learning_rate": 1.929139048595598e-07, + "loss": 0.7006, + "step": 1985 + }, + { + "epoch": 0.07, + "learning_rate": 1.9291024279488775e-07, + "loss": 0.6992, + "step": 1986 + }, + { + "epoch": 0.07, + "learning_rate": 1.9290658073021568e-07, + "loss": 0.6619, + "step": 1987 + }, + { + "epoch": 0.07, + "learning_rate": 1.9290291866554363e-07, + "loss": 0.7022, + "step": 1988 + }, + { + "epoch": 0.07, + "learning_rate": 1.9289925660087156e-07, + "loss": 0.6816, + "step": 1989 + }, + { + "epoch": 0.07, + "learning_rate": 1.928955945361995e-07, + "loss": 0.7089, + "step": 1990 + }, + { + "epoch": 0.07, + "learning_rate": 1.9289193247152744e-07, + "loss": 0.697, + "step": 1991 + }, + { + "epoch": 0.07, + "learning_rate": 1.928882704068554e-07, + "loss": 0.7085, + "step": 1992 + }, + { + "epoch": 0.07, + "learning_rate": 1.9288460834218332e-07, + "loss": 0.6767, + "step": 1993 + }, + { + "epoch": 0.07, + "learning_rate": 1.9288094627751127e-07, + "loss": 0.7192, + "step": 1994 + }, + { + "epoch": 0.07, + "learning_rate": 1.928772842128392e-07, + "loss": 0.7014, + "step": 1995 + }, + { + "epoch": 0.07, + "learning_rate": 1.9287362214816715e-07, + "loss": 0.6939, + "step": 1996 + }, + { + "epoch": 0.07, + "learning_rate": 1.9286996008349507e-07, + "loss": 0.6737, + "step": 1997 + }, + { + "epoch": 0.07, + "learning_rate": 1.92866298018823e-07, + "loss": 0.6799, + "step": 1998 + }, + { + "epoch": 0.07, + "learning_rate": 1.9286263595415093e-07, + "loss": 0.6974, + "step": 1999 + }, + { + "epoch": 0.07, + "learning_rate": 1.9285897388947888e-07, + "loss": 0.7021, + "step": 2000 + }, + { + "epoch": 0.07, + "learning_rate": 1.928553118248068e-07, + "loss": 0.6966, + "step": 2001 + }, + { + "epoch": 0.07, + "learning_rate": 1.9285164976013476e-07, + "loss": 0.7112, + "step": 2002 + }, + { + "epoch": 0.07, + "learning_rate": 1.928479876954627e-07, + "loss": 0.6848, + "step": 2003 + }, + { + "epoch": 0.07, + "learning_rate": 1.9284432563079061e-07, + "loss": 0.6835, + "step": 2004 + }, + { + "epoch": 0.07, + "learning_rate": 1.9284066356611857e-07, + "loss": 0.6967, + "step": 2005 + }, + { + "epoch": 0.07, + "learning_rate": 1.928370015014465e-07, + "loss": 0.7047, + "step": 2006 + }, + { + "epoch": 0.07, + "learning_rate": 1.9283333943677445e-07, + "loss": 0.7137, + "step": 2007 + }, + { + "epoch": 0.07, + "learning_rate": 1.9282967737210237e-07, + "loss": 0.7116, + "step": 2008 + }, + { + "epoch": 0.07, + "learning_rate": 1.9282601530743033e-07, + "loss": 0.6902, + "step": 2009 + }, + { + "epoch": 0.07, + "learning_rate": 1.9282235324275825e-07, + "loss": 0.6806, + "step": 2010 + }, + { + "epoch": 0.07, + "learning_rate": 1.928186911780862e-07, + "loss": 0.7139, + "step": 2011 + }, + { + "epoch": 0.07, + "learning_rate": 1.9281502911341413e-07, + "loss": 0.6969, + "step": 2012 + }, + { + "epoch": 0.07, + "learning_rate": 1.928113670487421e-07, + "loss": 0.7005, + "step": 2013 + }, + { + "epoch": 0.07, + "learning_rate": 1.9280770498407001e-07, + "loss": 0.6906, + "step": 2014 + }, + { + "epoch": 0.07, + "learning_rate": 1.9280404291939797e-07, + "loss": 0.6975, + "step": 2015 + }, + { + "epoch": 0.07, + "learning_rate": 1.9280038085472587e-07, + "loss": 0.6899, + "step": 2016 + }, + { + "epoch": 0.07, + "learning_rate": 1.9279671879005382e-07, + "loss": 0.6926, + "step": 2017 + }, + { + "epoch": 0.07, + "learning_rate": 1.9279305672538175e-07, + "loss": 0.7003, + "step": 2018 + }, + { + "epoch": 0.07, + "learning_rate": 1.927893946607097e-07, + "loss": 0.6919, + "step": 2019 + }, + { + "epoch": 0.07, + "learning_rate": 1.9278573259603763e-07, + "loss": 0.7278, + "step": 2020 + }, + { + "epoch": 0.07, + "learning_rate": 1.9278207053136558e-07, + "loss": 0.6992, + "step": 2021 + }, + { + "epoch": 0.07, + "learning_rate": 1.927784084666935e-07, + "loss": 0.6923, + "step": 2022 + }, + { + "epoch": 0.07, + "learning_rate": 1.9277474640202146e-07, + "loss": 0.6891, + "step": 2023 + }, + { + "epoch": 0.07, + "learning_rate": 1.9277108433734939e-07, + "loss": 0.706, + "step": 2024 + }, + { + "epoch": 0.07, + "learning_rate": 1.9276742227267734e-07, + "loss": 0.6787, + "step": 2025 + }, + { + "epoch": 0.07, + "learning_rate": 1.9276376020800527e-07, + "loss": 0.6825, + "step": 2026 + }, + { + "epoch": 0.07, + "learning_rate": 1.927600981433332e-07, + "loss": 0.6998, + "step": 2027 + }, + { + "epoch": 0.07, + "learning_rate": 1.9275643607866115e-07, + "loss": 0.7091, + "step": 2028 + }, + { + "epoch": 0.07, + "learning_rate": 1.9275277401398907e-07, + "loss": 0.701, + "step": 2029 + }, + { + "epoch": 0.07, + "learning_rate": 1.9274911194931703e-07, + "loss": 0.6731, + "step": 2030 + }, + { + "epoch": 0.07, + "learning_rate": 1.9274544988464495e-07, + "loss": 0.7006, + "step": 2031 + }, + { + "epoch": 0.07, + "learning_rate": 1.927417878199729e-07, + "loss": 0.7009, + "step": 2032 + }, + { + "epoch": 0.07, + "learning_rate": 1.9273812575530083e-07, + "loss": 0.7072, + "step": 2033 + }, + { + "epoch": 0.07, + "learning_rate": 1.9273446369062876e-07, + "loss": 0.6972, + "step": 2034 + }, + { + "epoch": 0.07, + "learning_rate": 1.9273080162595669e-07, + "loss": 0.6945, + "step": 2035 + }, + { + "epoch": 0.07, + "learning_rate": 1.9272713956128464e-07, + "loss": 0.6897, + "step": 2036 + }, + { + "epoch": 0.07, + "learning_rate": 1.9272347749661257e-07, + "loss": 0.6604, + "step": 2037 + }, + { + "epoch": 0.07, + "learning_rate": 1.9271981543194052e-07, + "loss": 0.6961, + "step": 2038 + }, + { + "epoch": 0.07, + "learning_rate": 1.9271615336726845e-07, + "loss": 0.6948, + "step": 2039 + }, + { + "epoch": 0.07, + "learning_rate": 1.927124913025964e-07, + "loss": 0.6818, + "step": 2040 + }, + { + "epoch": 0.07, + "learning_rate": 1.9270882923792433e-07, + "loss": 0.6561, + "step": 2041 + }, + { + "epoch": 0.07, + "learning_rate": 1.9270516717325228e-07, + "loss": 0.698, + "step": 2042 + }, + { + "epoch": 0.07, + "learning_rate": 1.927015051085802e-07, + "loss": 0.6655, + "step": 2043 + }, + { + "epoch": 0.07, + "learning_rate": 1.9269784304390816e-07, + "loss": 0.6832, + "step": 2044 + }, + { + "epoch": 0.07, + "learning_rate": 1.9269418097923609e-07, + "loss": 0.7102, + "step": 2045 + }, + { + "epoch": 0.07, + "learning_rate": 1.9269051891456404e-07, + "loss": 0.7047, + "step": 2046 + }, + { + "epoch": 0.07, + "learning_rate": 1.9268685684989197e-07, + "loss": 0.7005, + "step": 2047 + }, + { + "epoch": 0.07, + "learning_rate": 1.926831947852199e-07, + "loss": 0.6882, + "step": 2048 + }, + { + "epoch": 0.07, + "learning_rate": 1.9267953272054785e-07, + "loss": 0.6817, + "step": 2049 + }, + { + "epoch": 0.08, + "learning_rate": 1.9267587065587577e-07, + "loss": 0.7015, + "step": 2050 + }, + { + "epoch": 0.08, + "learning_rate": 1.9267220859120373e-07, + "loss": 0.7001, + "step": 2051 + }, + { + "epoch": 0.08, + "learning_rate": 1.9266854652653165e-07, + "loss": 0.6731, + "step": 2052 + }, + { + "epoch": 0.08, + "learning_rate": 1.9266488446185958e-07, + "loss": 0.6829, + "step": 2053 + }, + { + "epoch": 0.08, + "learning_rate": 1.926612223971875e-07, + "loss": 0.6921, + "step": 2054 + }, + { + "epoch": 0.08, + "learning_rate": 1.9265756033251546e-07, + "loss": 0.6702, + "step": 2055 + }, + { + "epoch": 0.08, + "learning_rate": 1.9265389826784338e-07, + "loss": 0.7158, + "step": 2056 + }, + { + "epoch": 0.08, + "learning_rate": 1.9265023620317134e-07, + "loss": 0.6992, + "step": 2057 + }, + { + "epoch": 0.08, + "learning_rate": 1.9264657413849926e-07, + "loss": 0.6916, + "step": 2058 + }, + { + "epoch": 0.08, + "learning_rate": 1.9264291207382722e-07, + "loss": 0.6933, + "step": 2059 + }, + { + "epoch": 0.08, + "learning_rate": 1.9263925000915514e-07, + "loss": 0.7155, + "step": 2060 + }, + { + "epoch": 0.08, + "learning_rate": 1.926355879444831e-07, + "loss": 0.6767, + "step": 2061 + }, + { + "epoch": 0.08, + "learning_rate": 1.9263192587981102e-07, + "loss": 0.6815, + "step": 2062 + }, + { + "epoch": 0.08, + "learning_rate": 1.9262826381513898e-07, + "loss": 0.7103, + "step": 2063 + }, + { + "epoch": 0.08, + "learning_rate": 1.926246017504669e-07, + "loss": 0.7195, + "step": 2064 + }, + { + "epoch": 0.08, + "learning_rate": 1.9262093968579486e-07, + "loss": 0.7043, + "step": 2065 + }, + { + "epoch": 0.08, + "learning_rate": 1.9261727762112278e-07, + "loss": 0.692, + "step": 2066 + }, + { + "epoch": 0.08, + "learning_rate": 1.9261361555645074e-07, + "loss": 0.6869, + "step": 2067 + }, + { + "epoch": 0.08, + "learning_rate": 1.9260995349177866e-07, + "loss": 0.7055, + "step": 2068 + }, + { + "epoch": 0.08, + "learning_rate": 1.9260629142710662e-07, + "loss": 0.671, + "step": 2069 + }, + { + "epoch": 0.08, + "learning_rate": 1.9260262936243452e-07, + "loss": 0.7221, + "step": 2070 + }, + { + "epoch": 0.08, + "learning_rate": 1.9259896729776247e-07, + "loss": 0.6809, + "step": 2071 + }, + { + "epoch": 0.08, + "learning_rate": 1.925953052330904e-07, + "loss": 0.6868, + "step": 2072 + }, + { + "epoch": 0.08, + "learning_rate": 1.9259164316841835e-07, + "loss": 0.7175, + "step": 2073 + }, + { + "epoch": 0.08, + "learning_rate": 1.9258798110374628e-07, + "loss": 0.7085, + "step": 2074 + }, + { + "epoch": 0.08, + "learning_rate": 1.925843190390742e-07, + "loss": 0.6987, + "step": 2075 + }, + { + "epoch": 0.08, + "learning_rate": 1.9258065697440216e-07, + "loss": 0.7015, + "step": 2076 + }, + { + "epoch": 0.08, + "learning_rate": 1.9257699490973008e-07, + "loss": 0.7023, + "step": 2077 + }, + { + "epoch": 0.08, + "learning_rate": 1.9257333284505804e-07, + "loss": 0.6742, + "step": 2078 + }, + { + "epoch": 0.08, + "learning_rate": 1.9256967078038596e-07, + "loss": 0.7054, + "step": 2079 + }, + { + "epoch": 0.08, + "learning_rate": 1.9256600871571392e-07, + "loss": 0.6934, + "step": 2080 + }, + { + "epoch": 0.08, + "learning_rate": 1.9256234665104184e-07, + "loss": 0.6846, + "step": 2081 + }, + { + "epoch": 0.08, + "learning_rate": 1.925586845863698e-07, + "loss": 0.7114, + "step": 2082 + }, + { + "epoch": 0.08, + "learning_rate": 1.9255502252169772e-07, + "loss": 0.6822, + "step": 2083 + }, + { + "epoch": 0.08, + "learning_rate": 1.9255136045702568e-07, + "loss": 0.6726, + "step": 2084 + }, + { + "epoch": 0.08, + "learning_rate": 1.925476983923536e-07, + "loss": 0.6779, + "step": 2085 + }, + { + "epoch": 0.08, + "learning_rate": 1.9254403632768156e-07, + "loss": 0.7068, + "step": 2086 + }, + { + "epoch": 0.08, + "learning_rate": 1.9254037426300948e-07, + "loss": 0.691, + "step": 2087 + }, + { + "epoch": 0.08, + "learning_rate": 1.925367121983374e-07, + "loss": 0.7072, + "step": 2088 + }, + { + "epoch": 0.08, + "learning_rate": 1.9253305013366534e-07, + "loss": 0.6917, + "step": 2089 + }, + { + "epoch": 0.08, + "learning_rate": 1.925293880689933e-07, + "loss": 0.7089, + "step": 2090 + }, + { + "epoch": 0.08, + "learning_rate": 1.9252572600432122e-07, + "loss": 0.6917, + "step": 2091 + }, + { + "epoch": 0.08, + "learning_rate": 1.9252206393964917e-07, + "loss": 0.6652, + "step": 2092 + }, + { + "epoch": 0.08, + "learning_rate": 1.925184018749771e-07, + "loss": 0.6977, + "step": 2093 + }, + { + "epoch": 0.08, + "learning_rate": 1.9251473981030505e-07, + "loss": 0.7009, + "step": 2094 + }, + { + "epoch": 0.08, + "learning_rate": 1.9251107774563298e-07, + "loss": 0.6838, + "step": 2095 + }, + { + "epoch": 0.08, + "learning_rate": 1.9250741568096093e-07, + "loss": 0.7131, + "step": 2096 + }, + { + "epoch": 0.08, + "learning_rate": 1.9250375361628886e-07, + "loss": 0.6957, + "step": 2097 + }, + { + "epoch": 0.08, + "learning_rate": 1.9250009155161678e-07, + "loss": 0.6921, + "step": 2098 + }, + { + "epoch": 0.08, + "learning_rate": 1.9249642948694474e-07, + "loss": 0.6852, + "step": 2099 + }, + { + "epoch": 0.08, + "learning_rate": 1.9249276742227266e-07, + "loss": 0.6832, + "step": 2100 + }, + { + "epoch": 0.08, + "learning_rate": 1.9248910535760062e-07, + "loss": 0.6777, + "step": 2101 + }, + { + "epoch": 0.08, + "learning_rate": 1.9248544329292854e-07, + "loss": 0.6823, + "step": 2102 + }, + { + "epoch": 0.08, + "learning_rate": 1.924817812282565e-07, + "loss": 0.699, + "step": 2103 + }, + { + "epoch": 0.08, + "learning_rate": 1.9247811916358442e-07, + "loss": 0.6901, + "step": 2104 + }, + { + "epoch": 0.08, + "learning_rate": 1.9247445709891238e-07, + "loss": 0.6929, + "step": 2105 + }, + { + "epoch": 0.08, + "learning_rate": 1.9247079503424028e-07, + "loss": 0.6624, + "step": 2106 + }, + { + "epoch": 0.08, + "learning_rate": 1.9246713296956823e-07, + "loss": 0.6933, + "step": 2107 + }, + { + "epoch": 0.08, + "learning_rate": 1.9246347090489616e-07, + "loss": 0.7041, + "step": 2108 + }, + { + "epoch": 0.08, + "learning_rate": 1.924598088402241e-07, + "loss": 0.7115, + "step": 2109 + }, + { + "epoch": 0.08, + "learning_rate": 1.9245614677555204e-07, + "loss": 0.6704, + "step": 2110 + }, + { + "epoch": 0.08, + "learning_rate": 1.9245248471088e-07, + "loss": 0.6747, + "step": 2111 + }, + { + "epoch": 0.08, + "learning_rate": 1.9244882264620792e-07, + "loss": 0.6978, + "step": 2112 + }, + { + "epoch": 0.08, + "learning_rate": 1.9244516058153587e-07, + "loss": 0.6882, + "step": 2113 + }, + { + "epoch": 0.08, + "learning_rate": 1.924414985168638e-07, + "loss": 0.7001, + "step": 2114 + }, + { + "epoch": 0.08, + "learning_rate": 1.9243783645219175e-07, + "loss": 0.7003, + "step": 2115 + }, + { + "epoch": 0.08, + "learning_rate": 1.9243417438751967e-07, + "loss": 0.6911, + "step": 2116 + }, + { + "epoch": 0.08, + "learning_rate": 1.9243051232284763e-07, + "loss": 0.7088, + "step": 2117 + }, + { + "epoch": 0.08, + "learning_rate": 1.9242685025817555e-07, + "loss": 0.6907, + "step": 2118 + }, + { + "epoch": 0.08, + "learning_rate": 1.9242318819350348e-07, + "loss": 0.7039, + "step": 2119 + }, + { + "epoch": 0.08, + "learning_rate": 1.9241952612883143e-07, + "loss": 0.6823, + "step": 2120 + }, + { + "epoch": 0.08, + "learning_rate": 1.9241586406415936e-07, + "loss": 0.6951, + "step": 2121 + }, + { + "epoch": 0.08, + "learning_rate": 1.9241220199948731e-07, + "loss": 0.7003, + "step": 2122 + }, + { + "epoch": 0.08, + "learning_rate": 1.9240853993481524e-07, + "loss": 0.6699, + "step": 2123 + }, + { + "epoch": 0.08, + "learning_rate": 1.9240487787014317e-07, + "loss": 0.672, + "step": 2124 + }, + { + "epoch": 0.08, + "learning_rate": 1.924012158054711e-07, + "loss": 0.7015, + "step": 2125 + }, + { + "epoch": 0.08, + "learning_rate": 1.9239755374079905e-07, + "loss": 0.6837, + "step": 2126 + }, + { + "epoch": 0.08, + "learning_rate": 1.9239389167612697e-07, + "loss": 0.7037, + "step": 2127 + }, + { + "epoch": 0.08, + "learning_rate": 1.9239022961145493e-07, + "loss": 0.698, + "step": 2128 + }, + { + "epoch": 0.08, + "learning_rate": 1.9238656754678285e-07, + "loss": 0.687, + "step": 2129 + }, + { + "epoch": 0.08, + "learning_rate": 1.923829054821108e-07, + "loss": 0.694, + "step": 2130 + }, + { + "epoch": 0.08, + "learning_rate": 1.9237924341743873e-07, + "loss": 0.6984, + "step": 2131 + }, + { + "epoch": 0.08, + "learning_rate": 1.923755813527667e-07, + "loss": 0.6908, + "step": 2132 + }, + { + "epoch": 0.08, + "learning_rate": 1.9237191928809461e-07, + "loss": 0.6678, + "step": 2133 + }, + { + "epoch": 0.08, + "learning_rate": 1.9236825722342257e-07, + "loss": 0.692, + "step": 2134 + }, + { + "epoch": 0.08, + "learning_rate": 1.923645951587505e-07, + "loss": 0.6833, + "step": 2135 + }, + { + "epoch": 0.08, + "learning_rate": 1.9236093309407845e-07, + "loss": 0.6805, + "step": 2136 + }, + { + "epoch": 0.08, + "learning_rate": 1.9235727102940637e-07, + "loss": 0.699, + "step": 2137 + }, + { + "epoch": 0.08, + "learning_rate": 1.9235360896473433e-07, + "loss": 0.6868, + "step": 2138 + }, + { + "epoch": 0.08, + "learning_rate": 1.9234994690006225e-07, + "loss": 0.6871, + "step": 2139 + }, + { + "epoch": 0.08, + "learning_rate": 1.923462848353902e-07, + "loss": 0.6909, + "step": 2140 + }, + { + "epoch": 0.08, + "learning_rate": 1.9234262277071813e-07, + "loss": 0.7106, + "step": 2141 + }, + { + "epoch": 0.08, + "learning_rate": 1.9233896070604606e-07, + "loss": 0.7081, + "step": 2142 + }, + { + "epoch": 0.08, + "learning_rate": 1.9233529864137399e-07, + "loss": 0.6888, + "step": 2143 + }, + { + "epoch": 0.08, + "learning_rate": 1.9233163657670194e-07, + "loss": 0.683, + "step": 2144 + }, + { + "epoch": 0.08, + "learning_rate": 1.9232797451202987e-07, + "loss": 0.7214, + "step": 2145 + }, + { + "epoch": 0.08, + "learning_rate": 1.923243124473578e-07, + "loss": 0.7023, + "step": 2146 + }, + { + "epoch": 0.08, + "learning_rate": 1.9232065038268575e-07, + "loss": 0.7125, + "step": 2147 + }, + { + "epoch": 0.08, + "learning_rate": 1.9231698831801367e-07, + "loss": 0.6935, + "step": 2148 + }, + { + "epoch": 0.08, + "learning_rate": 1.9231332625334163e-07, + "loss": 0.6851, + "step": 2149 + }, + { + "epoch": 0.08, + "learning_rate": 1.9230966418866955e-07, + "loss": 0.6849, + "step": 2150 + }, + { + "epoch": 0.08, + "learning_rate": 1.923060021239975e-07, + "loss": 0.7102, + "step": 2151 + }, + { + "epoch": 0.08, + "learning_rate": 1.9230234005932543e-07, + "loss": 0.7021, + "step": 2152 + }, + { + "epoch": 0.08, + "learning_rate": 1.9229867799465339e-07, + "loss": 0.7017, + "step": 2153 + }, + { + "epoch": 0.08, + "learning_rate": 1.922950159299813e-07, + "loss": 0.6974, + "step": 2154 + }, + { + "epoch": 0.08, + "learning_rate": 1.9229135386530927e-07, + "loss": 0.6906, + "step": 2155 + }, + { + "epoch": 0.08, + "learning_rate": 1.922876918006372e-07, + "loss": 0.6922, + "step": 2156 + }, + { + "epoch": 0.08, + "learning_rate": 1.9228402973596515e-07, + "loss": 0.711, + "step": 2157 + }, + { + "epoch": 0.08, + "learning_rate": 1.9228036767129307e-07, + "loss": 0.6818, + "step": 2158 + }, + { + "epoch": 0.08, + "learning_rate": 1.9227670560662103e-07, + "loss": 0.6968, + "step": 2159 + }, + { + "epoch": 0.08, + "learning_rate": 1.9227304354194893e-07, + "loss": 0.7115, + "step": 2160 + }, + { + "epoch": 0.08, + "learning_rate": 1.9226938147727688e-07, + "loss": 0.6993, + "step": 2161 + }, + { + "epoch": 0.08, + "learning_rate": 1.922657194126048e-07, + "loss": 0.6978, + "step": 2162 + }, + { + "epoch": 0.08, + "learning_rate": 1.9226205734793276e-07, + "loss": 0.6981, + "step": 2163 + }, + { + "epoch": 0.08, + "learning_rate": 1.9225839528326069e-07, + "loss": 0.7058, + "step": 2164 + }, + { + "epoch": 0.08, + "learning_rate": 1.9225473321858864e-07, + "loss": 0.6898, + "step": 2165 + }, + { + "epoch": 0.08, + "learning_rate": 1.9225107115391657e-07, + "loss": 0.7024, + "step": 2166 + }, + { + "epoch": 0.08, + "learning_rate": 1.9224740908924452e-07, + "loss": 0.6815, + "step": 2167 + }, + { + "epoch": 0.08, + "learning_rate": 1.9224374702457245e-07, + "loss": 0.671, + "step": 2168 + }, + { + "epoch": 0.08, + "learning_rate": 1.9224008495990037e-07, + "loss": 0.6594, + "step": 2169 + }, + { + "epoch": 0.08, + "learning_rate": 1.9223642289522833e-07, + "loss": 0.6947, + "step": 2170 + }, + { + "epoch": 0.08, + "learning_rate": 1.9223276083055625e-07, + "loss": 0.7, + "step": 2171 + }, + { + "epoch": 0.08, + "learning_rate": 1.922290987658842e-07, + "loss": 0.6887, + "step": 2172 + }, + { + "epoch": 0.08, + "learning_rate": 1.9222543670121213e-07, + "loss": 0.6956, + "step": 2173 + }, + { + "epoch": 0.08, + "learning_rate": 1.9222177463654008e-07, + "loss": 0.6936, + "step": 2174 + }, + { + "epoch": 0.08, + "learning_rate": 1.92218112571868e-07, + "loss": 0.7081, + "step": 2175 + }, + { + "epoch": 0.08, + "learning_rate": 1.9221445050719596e-07, + "loss": 0.6732, + "step": 2176 + }, + { + "epoch": 0.08, + "learning_rate": 1.922107884425239e-07, + "loss": 0.7016, + "step": 2177 + }, + { + "epoch": 0.08, + "learning_rate": 1.9220712637785182e-07, + "loss": 0.6846, + "step": 2178 + }, + { + "epoch": 0.08, + "learning_rate": 1.9220346431317974e-07, + "loss": 0.6943, + "step": 2179 + }, + { + "epoch": 0.08, + "learning_rate": 1.921998022485077e-07, + "loss": 0.7152, + "step": 2180 + }, + { + "epoch": 0.08, + "learning_rate": 1.9219614018383562e-07, + "loss": 0.6907, + "step": 2181 + }, + { + "epoch": 0.08, + "learning_rate": 1.9219247811916358e-07, + "loss": 0.6865, + "step": 2182 + }, + { + "epoch": 0.08, + "learning_rate": 1.921888160544915e-07, + "loss": 0.6839, + "step": 2183 + }, + { + "epoch": 0.08, + "learning_rate": 1.9218515398981946e-07, + "loss": 0.6921, + "step": 2184 + }, + { + "epoch": 0.08, + "learning_rate": 1.9218149192514738e-07, + "loss": 0.712, + "step": 2185 + }, + { + "epoch": 0.08, + "learning_rate": 1.9217782986047534e-07, + "loss": 0.6786, + "step": 2186 + }, + { + "epoch": 0.08, + "learning_rate": 1.9217416779580326e-07, + "loss": 0.711, + "step": 2187 + }, + { + "epoch": 0.08, + "learning_rate": 1.9217050573113122e-07, + "loss": 0.6956, + "step": 2188 + }, + { + "epoch": 0.08, + "learning_rate": 1.9216684366645914e-07, + "loss": 0.7171, + "step": 2189 + }, + { + "epoch": 0.08, + "learning_rate": 1.9216318160178707e-07, + "loss": 0.7066, + "step": 2190 + }, + { + "epoch": 0.08, + "learning_rate": 1.9215951953711502e-07, + "loss": 0.6978, + "step": 2191 + }, + { + "epoch": 0.08, + "learning_rate": 1.9215585747244295e-07, + "loss": 0.6889, + "step": 2192 + }, + { + "epoch": 0.08, + "learning_rate": 1.921521954077709e-07, + "loss": 0.7073, + "step": 2193 + }, + { + "epoch": 0.08, + "learning_rate": 1.9214853334309883e-07, + "loss": 0.6915, + "step": 2194 + }, + { + "epoch": 0.08, + "learning_rate": 1.9214487127842678e-07, + "loss": 0.7001, + "step": 2195 + }, + { + "epoch": 0.08, + "learning_rate": 1.9214120921375468e-07, + "loss": 0.6845, + "step": 2196 + }, + { + "epoch": 0.08, + "learning_rate": 1.9213754714908264e-07, + "loss": 0.7076, + "step": 2197 + }, + { + "epoch": 0.08, + "learning_rate": 1.9213388508441056e-07, + "loss": 0.7093, + "step": 2198 + }, + { + "epoch": 0.08, + "learning_rate": 1.9213022301973852e-07, + "loss": 0.7029, + "step": 2199 + }, + { + "epoch": 0.08, + "learning_rate": 1.9212656095506644e-07, + "loss": 0.6983, + "step": 2200 + }, + { + "epoch": 0.08, + "learning_rate": 1.921228988903944e-07, + "loss": 0.7086, + "step": 2201 + }, + { + "epoch": 0.08, + "learning_rate": 1.9211923682572232e-07, + "loss": 0.6854, + "step": 2202 + }, + { + "epoch": 0.08, + "learning_rate": 1.9211557476105028e-07, + "loss": 0.6966, + "step": 2203 + }, + { + "epoch": 0.08, + "learning_rate": 1.921119126963782e-07, + "loss": 0.6958, + "step": 2204 + }, + { + "epoch": 0.08, + "learning_rate": 1.9210825063170616e-07, + "loss": 0.691, + "step": 2205 + }, + { + "epoch": 0.08, + "learning_rate": 1.9210458856703408e-07, + "loss": 0.7111, + "step": 2206 + }, + { + "epoch": 0.08, + "learning_rate": 1.9210092650236204e-07, + "loss": 0.7172, + "step": 2207 + }, + { + "epoch": 0.08, + "learning_rate": 1.9209726443768996e-07, + "loss": 0.6814, + "step": 2208 + }, + { + "epoch": 0.08, + "learning_rate": 1.9209360237301792e-07, + "loss": 0.6896, + "step": 2209 + }, + { + "epoch": 0.08, + "learning_rate": 1.9208994030834584e-07, + "loss": 0.6861, + "step": 2210 + }, + { + "epoch": 0.08, + "learning_rate": 1.920862782436738e-07, + "loss": 0.695, + "step": 2211 + }, + { + "epoch": 0.08, + "learning_rate": 1.9208261617900172e-07, + "loss": 0.6885, + "step": 2212 + }, + { + "epoch": 0.08, + "learning_rate": 1.9207895411432965e-07, + "loss": 0.7036, + "step": 2213 + }, + { + "epoch": 0.08, + "learning_rate": 1.9207529204965758e-07, + "loss": 0.6891, + "step": 2214 + }, + { + "epoch": 0.08, + "learning_rate": 1.9207162998498553e-07, + "loss": 0.6642, + "step": 2215 + }, + { + "epoch": 0.08, + "learning_rate": 1.9206796792031346e-07, + "loss": 0.6929, + "step": 2216 + }, + { + "epoch": 0.08, + "learning_rate": 1.9206430585564138e-07, + "loss": 0.6799, + "step": 2217 + }, + { + "epoch": 0.08, + "learning_rate": 1.9206064379096934e-07, + "loss": 0.7014, + "step": 2218 + }, + { + "epoch": 0.08, + "learning_rate": 1.9205698172629726e-07, + "loss": 0.6813, + "step": 2219 + }, + { + "epoch": 0.08, + "learning_rate": 1.9205331966162522e-07, + "loss": 0.6828, + "step": 2220 + }, + { + "epoch": 0.08, + "learning_rate": 1.9204965759695314e-07, + "loss": 0.6955, + "step": 2221 + }, + { + "epoch": 0.08, + "learning_rate": 1.920459955322811e-07, + "loss": 0.7052, + "step": 2222 + }, + { + "epoch": 0.08, + "learning_rate": 1.9204233346760902e-07, + "loss": 0.6844, + "step": 2223 + }, + { + "epoch": 0.08, + "learning_rate": 1.9203867140293698e-07, + "loss": 0.6841, + "step": 2224 + }, + { + "epoch": 0.08, + "learning_rate": 1.920350093382649e-07, + "loss": 0.6963, + "step": 2225 + }, + { + "epoch": 0.08, + "learning_rate": 1.9203134727359286e-07, + "loss": 0.6937, + "step": 2226 + }, + { + "epoch": 0.08, + "learning_rate": 1.9202768520892078e-07, + "loss": 0.693, + "step": 2227 + }, + { + "epoch": 0.08, + "learning_rate": 1.9202402314424874e-07, + "loss": 0.7027, + "step": 2228 + }, + { + "epoch": 0.08, + "learning_rate": 1.9202036107957666e-07, + "loss": 0.6947, + "step": 2229 + }, + { + "epoch": 0.08, + "learning_rate": 1.9201669901490461e-07, + "loss": 0.7114, + "step": 2230 + }, + { + "epoch": 0.08, + "learning_rate": 1.9201303695023254e-07, + "loss": 0.7101, + "step": 2231 + }, + { + "epoch": 0.08, + "learning_rate": 1.9200937488556047e-07, + "loss": 0.716, + "step": 2232 + }, + { + "epoch": 0.08, + "learning_rate": 1.920057128208884e-07, + "loss": 0.6951, + "step": 2233 + }, + { + "epoch": 0.08, + "learning_rate": 1.9200205075621635e-07, + "loss": 0.6977, + "step": 2234 + }, + { + "epoch": 0.08, + "learning_rate": 1.9199838869154427e-07, + "loss": 0.6954, + "step": 2235 + }, + { + "epoch": 0.08, + "learning_rate": 1.9199472662687223e-07, + "loss": 0.7184, + "step": 2236 + }, + { + "epoch": 0.08, + "learning_rate": 1.9199106456220015e-07, + "loss": 0.6965, + "step": 2237 + }, + { + "epoch": 0.08, + "learning_rate": 1.919874024975281e-07, + "loss": 0.6918, + "step": 2238 + }, + { + "epoch": 0.08, + "learning_rate": 1.9198374043285603e-07, + "loss": 0.7255, + "step": 2239 + }, + { + "epoch": 0.08, + "learning_rate": 1.9198007836818396e-07, + "loss": 0.6834, + "step": 2240 + }, + { + "epoch": 0.08, + "learning_rate": 1.9197641630351191e-07, + "loss": 0.6908, + "step": 2241 + }, + { + "epoch": 0.08, + "learning_rate": 1.9197275423883984e-07, + "loss": 0.6827, + "step": 2242 + }, + { + "epoch": 0.08, + "learning_rate": 1.919690921741678e-07, + "loss": 0.6789, + "step": 2243 + }, + { + "epoch": 0.08, + "learning_rate": 1.9196543010949572e-07, + "loss": 0.6981, + "step": 2244 + }, + { + "epoch": 0.08, + "learning_rate": 1.9196176804482367e-07, + "loss": 0.6895, + "step": 2245 + }, + { + "epoch": 0.08, + "learning_rate": 1.919581059801516e-07, + "loss": 0.6939, + "step": 2246 + }, + { + "epoch": 0.08, + "learning_rate": 1.9195444391547955e-07, + "loss": 0.6893, + "step": 2247 + }, + { + "epoch": 0.08, + "learning_rate": 1.9195078185080748e-07, + "loss": 0.6951, + "step": 2248 + }, + { + "epoch": 0.08, + "learning_rate": 1.9194711978613543e-07, + "loss": 0.6883, + "step": 2249 + }, + { + "epoch": 0.08, + "learning_rate": 1.9194345772146333e-07, + "loss": 0.6724, + "step": 2250 + }, + { + "epoch": 0.08, + "learning_rate": 1.919397956567913e-07, + "loss": 0.7141, + "step": 2251 + }, + { + "epoch": 0.08, + "learning_rate": 1.9193613359211921e-07, + "loss": 0.702, + "step": 2252 + }, + { + "epoch": 0.08, + "learning_rate": 1.9193247152744717e-07, + "loss": 0.6589, + "step": 2253 + }, + { + "epoch": 0.08, + "learning_rate": 1.919288094627751e-07, + "loss": 0.7122, + "step": 2254 + }, + { + "epoch": 0.08, + "learning_rate": 1.9192514739810305e-07, + "loss": 0.7264, + "step": 2255 + }, + { + "epoch": 0.08, + "learning_rate": 1.9192148533343097e-07, + "loss": 0.7165, + "step": 2256 + }, + { + "epoch": 0.08, + "learning_rate": 1.9191782326875893e-07, + "loss": 0.6852, + "step": 2257 + }, + { + "epoch": 0.08, + "learning_rate": 1.9191416120408685e-07, + "loss": 0.6692, + "step": 2258 + }, + { + "epoch": 0.08, + "learning_rate": 1.919104991394148e-07, + "loss": 0.6918, + "step": 2259 + }, + { + "epoch": 0.08, + "learning_rate": 1.9190683707474273e-07, + "loss": 0.6882, + "step": 2260 + }, + { + "epoch": 0.08, + "learning_rate": 1.9190317501007066e-07, + "loss": 0.6959, + "step": 2261 + }, + { + "epoch": 0.08, + "learning_rate": 1.918995129453986e-07, + "loss": 0.6912, + "step": 2262 + }, + { + "epoch": 0.08, + "learning_rate": 1.9189585088072654e-07, + "loss": 0.7143, + "step": 2263 + }, + { + "epoch": 0.08, + "learning_rate": 1.918921888160545e-07, + "loss": 0.7073, + "step": 2264 + }, + { + "epoch": 0.08, + "learning_rate": 1.9188852675138242e-07, + "loss": 0.692, + "step": 2265 + }, + { + "epoch": 0.08, + "learning_rate": 1.9188486468671037e-07, + "loss": 0.6867, + "step": 2266 + }, + { + "epoch": 0.08, + "learning_rate": 1.918812026220383e-07, + "loss": 0.6961, + "step": 2267 + }, + { + "epoch": 0.08, + "learning_rate": 1.9187754055736623e-07, + "loss": 0.6918, + "step": 2268 + }, + { + "epoch": 0.08, + "learning_rate": 1.9187387849269415e-07, + "loss": 0.7025, + "step": 2269 + }, + { + "epoch": 0.08, + "learning_rate": 1.918702164280221e-07, + "loss": 0.6958, + "step": 2270 + }, + { + "epoch": 0.08, + "learning_rate": 1.9186655436335003e-07, + "loss": 0.6958, + "step": 2271 + }, + { + "epoch": 0.08, + "learning_rate": 1.9186289229867799e-07, + "loss": 0.6654, + "step": 2272 + }, + { + "epoch": 0.08, + "learning_rate": 1.918592302340059e-07, + "loss": 0.6963, + "step": 2273 + }, + { + "epoch": 0.08, + "learning_rate": 1.9185556816933387e-07, + "loss": 0.6888, + "step": 2274 + }, + { + "epoch": 0.08, + "learning_rate": 1.918519061046618e-07, + "loss": 0.6932, + "step": 2275 + }, + { + "epoch": 0.08, + "learning_rate": 1.9184824403998975e-07, + "loss": 0.7096, + "step": 2276 + }, + { + "epoch": 0.08, + "learning_rate": 1.9184458197531767e-07, + "loss": 0.7057, + "step": 2277 + }, + { + "epoch": 0.08, + "learning_rate": 1.9184091991064563e-07, + "loss": 0.7032, + "step": 2278 + }, + { + "epoch": 0.08, + "learning_rate": 1.9183725784597355e-07, + "loss": 0.6972, + "step": 2279 + }, + { + "epoch": 0.08, + "learning_rate": 1.918335957813015e-07, + "loss": 0.7069, + "step": 2280 + }, + { + "epoch": 0.08, + "learning_rate": 1.9182993371662943e-07, + "loss": 0.6894, + "step": 2281 + }, + { + "epoch": 0.08, + "learning_rate": 1.9182627165195736e-07, + "loss": 0.6999, + "step": 2282 + }, + { + "epoch": 0.08, + "learning_rate": 1.918226095872853e-07, + "loss": 0.6848, + "step": 2283 + }, + { + "epoch": 0.08, + "learning_rate": 1.9181894752261324e-07, + "loss": 0.7042, + "step": 2284 + }, + { + "epoch": 0.08, + "learning_rate": 1.918152854579412e-07, + "loss": 0.6886, + "step": 2285 + }, + { + "epoch": 0.08, + "learning_rate": 1.9181162339326912e-07, + "loss": 0.6816, + "step": 2286 + }, + { + "epoch": 0.08, + "learning_rate": 1.9180796132859705e-07, + "loss": 0.6942, + "step": 2287 + }, + { + "epoch": 0.08, + "learning_rate": 1.9180429926392497e-07, + "loss": 0.7123, + "step": 2288 + }, + { + "epoch": 0.08, + "learning_rate": 1.9180063719925293e-07, + "loss": 0.7093, + "step": 2289 + }, + { + "epoch": 0.08, + "learning_rate": 1.9179697513458085e-07, + "loss": 0.7013, + "step": 2290 + }, + { + "epoch": 0.08, + "learning_rate": 1.917933130699088e-07, + "loss": 0.679, + "step": 2291 + }, + { + "epoch": 0.08, + "learning_rate": 1.9178965100523673e-07, + "loss": 0.7015, + "step": 2292 + }, + { + "epoch": 0.08, + "learning_rate": 1.9178598894056468e-07, + "loss": 0.7023, + "step": 2293 + }, + { + "epoch": 0.08, + "learning_rate": 1.917823268758926e-07, + "loss": 0.6797, + "step": 2294 + }, + { + "epoch": 0.08, + "learning_rate": 1.9177866481122056e-07, + "loss": 0.6913, + "step": 2295 + }, + { + "epoch": 0.08, + "learning_rate": 1.917750027465485e-07, + "loss": 0.6862, + "step": 2296 + }, + { + "epoch": 0.08, + "learning_rate": 1.9177134068187644e-07, + "loss": 0.6831, + "step": 2297 + }, + { + "epoch": 0.08, + "learning_rate": 1.9176767861720437e-07, + "loss": 0.6941, + "step": 2298 + }, + { + "epoch": 0.08, + "learning_rate": 1.9176401655253232e-07, + "loss": 0.6981, + "step": 2299 + }, + { + "epoch": 0.08, + "learning_rate": 1.9176035448786025e-07, + "loss": 0.7039, + "step": 2300 + }, + { + "epoch": 0.08, + "learning_rate": 1.917566924231882e-07, + "loss": 0.6884, + "step": 2301 + }, + { + "epoch": 0.08, + "learning_rate": 1.9175303035851613e-07, + "loss": 0.6856, + "step": 2302 + }, + { + "epoch": 0.08, + "learning_rate": 1.9174936829384408e-07, + "loss": 0.6901, + "step": 2303 + }, + { + "epoch": 0.08, + "learning_rate": 1.9174570622917198e-07, + "loss": 0.6845, + "step": 2304 + }, + { + "epoch": 0.08, + "learning_rate": 1.9174204416449994e-07, + "loss": 0.7228, + "step": 2305 + }, + { + "epoch": 0.08, + "learning_rate": 1.9173838209982786e-07, + "loss": 0.698, + "step": 2306 + }, + { + "epoch": 0.08, + "learning_rate": 1.9173472003515582e-07, + "loss": 0.6934, + "step": 2307 + }, + { + "epoch": 0.08, + "learning_rate": 1.9173105797048374e-07, + "loss": 0.6934, + "step": 2308 + }, + { + "epoch": 0.08, + "learning_rate": 1.917273959058117e-07, + "loss": 0.6825, + "step": 2309 + }, + { + "epoch": 0.08, + "learning_rate": 1.9172373384113962e-07, + "loss": 0.7037, + "step": 2310 + }, + { + "epoch": 0.08, + "learning_rate": 1.9172007177646755e-07, + "loss": 0.6848, + "step": 2311 + }, + { + "epoch": 0.08, + "learning_rate": 1.917164097117955e-07, + "loss": 0.7195, + "step": 2312 + }, + { + "epoch": 0.08, + "learning_rate": 1.9171274764712343e-07, + "loss": 0.6718, + "step": 2313 + }, + { + "epoch": 0.08, + "learning_rate": 1.9170908558245138e-07, + "loss": 0.6763, + "step": 2314 + }, + { + "epoch": 0.08, + "learning_rate": 1.917054235177793e-07, + "loss": 0.6951, + "step": 2315 + }, + { + "epoch": 0.08, + "learning_rate": 1.9170176145310726e-07, + "loss": 0.6789, + "step": 2316 + }, + { + "epoch": 0.08, + "learning_rate": 1.916980993884352e-07, + "loss": 0.702, + "step": 2317 + }, + { + "epoch": 0.08, + "learning_rate": 1.9169443732376314e-07, + "loss": 0.6992, + "step": 2318 + }, + { + "epoch": 0.08, + "learning_rate": 1.9169077525909107e-07, + "loss": 0.7006, + "step": 2319 + }, + { + "epoch": 0.08, + "learning_rate": 1.9168711319441902e-07, + "loss": 0.691, + "step": 2320 + }, + { + "epoch": 0.08, + "learning_rate": 1.9168345112974695e-07, + "loss": 0.6982, + "step": 2321 + }, + { + "epoch": 0.08, + "learning_rate": 1.9167978906507488e-07, + "loss": 0.7008, + "step": 2322 + }, + { + "epoch": 0.08, + "learning_rate": 1.916761270004028e-07, + "loss": 0.6709, + "step": 2323 + }, + { + "epoch": 0.09, + "learning_rate": 1.9167246493573076e-07, + "loss": 0.7114, + "step": 2324 + }, + { + "epoch": 0.09, + "learning_rate": 1.9166880287105868e-07, + "loss": 0.6704, + "step": 2325 + }, + { + "epoch": 0.09, + "learning_rate": 1.9166514080638664e-07, + "loss": 0.6879, + "step": 2326 + }, + { + "epoch": 0.09, + "learning_rate": 1.9166147874171456e-07, + "loss": 0.6801, + "step": 2327 + }, + { + "epoch": 0.09, + "learning_rate": 1.9165781667704252e-07, + "loss": 0.6749, + "step": 2328 + }, + { + "epoch": 0.09, + "learning_rate": 1.9165415461237044e-07, + "loss": 0.6825, + "step": 2329 + }, + { + "epoch": 0.09, + "learning_rate": 1.916504925476984e-07, + "loss": 0.6839, + "step": 2330 + }, + { + "epoch": 0.09, + "learning_rate": 1.9164683048302632e-07, + "loss": 0.6796, + "step": 2331 + }, + { + "epoch": 0.09, + "learning_rate": 1.9164316841835425e-07, + "loss": 0.7025, + "step": 2332 + }, + { + "epoch": 0.09, + "learning_rate": 1.916395063536822e-07, + "loss": 0.7057, + "step": 2333 + }, + { + "epoch": 0.09, + "learning_rate": 1.9163584428901013e-07, + "loss": 0.6991, + "step": 2334 + }, + { + "epoch": 0.09, + "learning_rate": 1.9163218222433808e-07, + "loss": 0.7041, + "step": 2335 + }, + { + "epoch": 0.09, + "learning_rate": 1.91628520159666e-07, + "loss": 0.6836, + "step": 2336 + }, + { + "epoch": 0.09, + "learning_rate": 1.9162485809499396e-07, + "loss": 0.7035, + "step": 2337 + }, + { + "epoch": 0.09, + "learning_rate": 1.916211960303219e-07, + "loss": 0.7031, + "step": 2338 + }, + { + "epoch": 0.09, + "learning_rate": 1.9161753396564984e-07, + "loss": 0.7307, + "step": 2339 + }, + { + "epoch": 0.09, + "learning_rate": 1.9161387190097774e-07, + "loss": 0.695, + "step": 2340 + }, + { + "epoch": 0.09, + "learning_rate": 1.916102098363057e-07, + "loss": 0.6833, + "step": 2341 + }, + { + "epoch": 0.09, + "learning_rate": 1.9160654777163362e-07, + "loss": 0.6788, + "step": 2342 + }, + { + "epoch": 0.09, + "learning_rate": 1.9160288570696158e-07, + "loss": 0.6908, + "step": 2343 + }, + { + "epoch": 0.09, + "learning_rate": 1.915992236422895e-07, + "loss": 0.6866, + "step": 2344 + }, + { + "epoch": 0.09, + "learning_rate": 1.9159556157761746e-07, + "loss": 0.6871, + "step": 2345 + }, + { + "epoch": 0.09, + "learning_rate": 1.9159189951294538e-07, + "loss": 0.7039, + "step": 2346 + }, + { + "epoch": 0.09, + "learning_rate": 1.9158823744827334e-07, + "loss": 0.6927, + "step": 2347 + }, + { + "epoch": 0.09, + "learning_rate": 1.9158457538360126e-07, + "loss": 0.7063, + "step": 2348 + }, + { + "epoch": 0.09, + "learning_rate": 1.9158091331892921e-07, + "loss": 0.7024, + "step": 2349 + }, + { + "epoch": 0.09, + "learning_rate": 1.9157725125425714e-07, + "loss": 0.7129, + "step": 2350 + }, + { + "epoch": 0.09, + "learning_rate": 1.915735891895851e-07, + "loss": 0.723, + "step": 2351 + }, + { + "epoch": 0.09, + "learning_rate": 1.9156992712491302e-07, + "loss": 0.7021, + "step": 2352 + }, + { + "epoch": 0.09, + "learning_rate": 1.9156626506024095e-07, + "loss": 0.7075, + "step": 2353 + }, + { + "epoch": 0.09, + "learning_rate": 1.915626029955689e-07, + "loss": 0.7179, + "step": 2354 + }, + { + "epoch": 0.09, + "learning_rate": 1.9155894093089683e-07, + "loss": 0.7239, + "step": 2355 + }, + { + "epoch": 0.09, + "learning_rate": 1.9155527886622478e-07, + "loss": 0.6658, + "step": 2356 + }, + { + "epoch": 0.09, + "learning_rate": 1.915516168015527e-07, + "loss": 0.6771, + "step": 2357 + }, + { + "epoch": 0.09, + "learning_rate": 1.9154795473688063e-07, + "loss": 0.6884, + "step": 2358 + }, + { + "epoch": 0.09, + "learning_rate": 1.9154429267220856e-07, + "loss": 0.692, + "step": 2359 + }, + { + "epoch": 0.09, + "learning_rate": 1.9154063060753651e-07, + "loss": 0.6823, + "step": 2360 + }, + { + "epoch": 0.09, + "learning_rate": 1.9153696854286444e-07, + "loss": 0.7087, + "step": 2361 + }, + { + "epoch": 0.09, + "learning_rate": 1.915333064781924e-07, + "loss": 0.6942, + "step": 2362 + }, + { + "epoch": 0.09, + "learning_rate": 1.9152964441352032e-07, + "loss": 0.7021, + "step": 2363 + }, + { + "epoch": 0.09, + "learning_rate": 1.9152598234884827e-07, + "loss": 0.6863, + "step": 2364 + }, + { + "epoch": 0.09, + "learning_rate": 1.915223202841762e-07, + "loss": 0.6662, + "step": 2365 + }, + { + "epoch": 0.09, + "learning_rate": 1.9151865821950415e-07, + "loss": 0.7066, + "step": 2366 + }, + { + "epoch": 0.09, + "learning_rate": 1.9151499615483208e-07, + "loss": 0.6975, + "step": 2367 + }, + { + "epoch": 0.09, + "learning_rate": 1.9151133409016003e-07, + "loss": 0.7066, + "step": 2368 + }, + { + "epoch": 0.09, + "learning_rate": 1.9150767202548796e-07, + "loss": 0.7306, + "step": 2369 + }, + { + "epoch": 0.09, + "learning_rate": 1.9150400996081591e-07, + "loss": 0.6813, + "step": 2370 + }, + { + "epoch": 0.09, + "learning_rate": 1.9150034789614384e-07, + "loss": 0.7057, + "step": 2371 + }, + { + "epoch": 0.09, + "learning_rate": 1.914966858314718e-07, + "loss": 0.6922, + "step": 2372 + }, + { + "epoch": 0.09, + "learning_rate": 1.9149302376679972e-07, + "loss": 0.7036, + "step": 2373 + }, + { + "epoch": 0.09, + "learning_rate": 1.9148936170212767e-07, + "loss": 0.6997, + "step": 2374 + }, + { + "epoch": 0.09, + "learning_rate": 1.9148569963745557e-07, + "loss": 0.7067, + "step": 2375 + }, + { + "epoch": 0.09, + "learning_rate": 1.9148203757278353e-07, + "loss": 0.691, + "step": 2376 + }, + { + "epoch": 0.09, + "learning_rate": 1.9147837550811145e-07, + "loss": 0.6874, + "step": 2377 + }, + { + "epoch": 0.09, + "learning_rate": 1.914747134434394e-07, + "loss": 0.6971, + "step": 2378 + }, + { + "epoch": 0.09, + "learning_rate": 1.9147105137876733e-07, + "loss": 0.6914, + "step": 2379 + }, + { + "epoch": 0.09, + "learning_rate": 1.9146738931409529e-07, + "loss": 0.6954, + "step": 2380 + }, + { + "epoch": 0.09, + "learning_rate": 1.914637272494232e-07, + "loss": 0.6838, + "step": 2381 + }, + { + "epoch": 0.09, + "learning_rate": 1.9146006518475114e-07, + "loss": 0.7062, + "step": 2382 + }, + { + "epoch": 0.09, + "learning_rate": 1.914564031200791e-07, + "loss": 0.6714, + "step": 2383 + }, + { + "epoch": 0.09, + "learning_rate": 1.9145274105540702e-07, + "loss": 0.6969, + "step": 2384 + }, + { + "epoch": 0.09, + "learning_rate": 1.9144907899073497e-07, + "loss": 0.7155, + "step": 2385 + }, + { + "epoch": 0.09, + "learning_rate": 1.914454169260629e-07, + "loss": 0.687, + "step": 2386 + }, + { + "epoch": 0.09, + "learning_rate": 1.9144175486139085e-07, + "loss": 0.6985, + "step": 2387 + }, + { + "epoch": 0.09, + "learning_rate": 1.9143809279671878e-07, + "loss": 0.6971, + "step": 2388 + }, + { + "epoch": 0.09, + "learning_rate": 1.9143443073204673e-07, + "loss": 0.6954, + "step": 2389 + }, + { + "epoch": 0.09, + "learning_rate": 1.9143076866737466e-07, + "loss": 0.6939, + "step": 2390 + }, + { + "epoch": 0.09, + "learning_rate": 1.914271066027026e-07, + "loss": 0.6884, + "step": 2391 + }, + { + "epoch": 0.09, + "learning_rate": 1.9142344453803054e-07, + "loss": 0.6942, + "step": 2392 + }, + { + "epoch": 0.09, + "learning_rate": 1.9141978247335847e-07, + "loss": 0.7253, + "step": 2393 + }, + { + "epoch": 0.09, + "learning_rate": 1.914161204086864e-07, + "loss": 0.6821, + "step": 2394 + }, + { + "epoch": 0.09, + "learning_rate": 1.9141245834401435e-07, + "loss": 0.6627, + "step": 2395 + }, + { + "epoch": 0.09, + "learning_rate": 1.9140879627934227e-07, + "loss": 0.6679, + "step": 2396 + }, + { + "epoch": 0.09, + "learning_rate": 1.9140513421467023e-07, + "loss": 0.6986, + "step": 2397 + }, + { + "epoch": 0.09, + "learning_rate": 1.9140147214999815e-07, + "loss": 0.6956, + "step": 2398 + }, + { + "epoch": 0.09, + "learning_rate": 1.913978100853261e-07, + "loss": 0.7005, + "step": 2399 + }, + { + "epoch": 0.09, + "learning_rate": 1.9139414802065403e-07, + "loss": 0.6933, + "step": 2400 + }, + { + "epoch": 0.09, + "learning_rate": 1.9139048595598199e-07, + "loss": 0.6841, + "step": 2401 + }, + { + "epoch": 0.09, + "learning_rate": 1.913868238913099e-07, + "loss": 0.7043, + "step": 2402 + }, + { + "epoch": 0.09, + "learning_rate": 1.9138316182663784e-07, + "loss": 0.6949, + "step": 2403 + }, + { + "epoch": 0.09, + "learning_rate": 1.913794997619658e-07, + "loss": 0.683, + "step": 2404 + }, + { + "epoch": 0.09, + "learning_rate": 1.9137583769729372e-07, + "loss": 0.6943, + "step": 2405 + }, + { + "epoch": 0.09, + "learning_rate": 1.9137217563262167e-07, + "loss": 0.6754, + "step": 2406 + }, + { + "epoch": 0.09, + "learning_rate": 1.913685135679496e-07, + "loss": 0.6943, + "step": 2407 + }, + { + "epoch": 0.09, + "learning_rate": 1.9136485150327755e-07, + "loss": 0.6896, + "step": 2408 + }, + { + "epoch": 0.09, + "learning_rate": 1.9136118943860548e-07, + "loss": 0.6793, + "step": 2409 + }, + { + "epoch": 0.09, + "learning_rate": 1.9135752737393343e-07, + "loss": 0.6665, + "step": 2410 + }, + { + "epoch": 0.09, + "learning_rate": 1.9135386530926133e-07, + "loss": 0.704, + "step": 2411 + }, + { + "epoch": 0.09, + "learning_rate": 1.9135020324458928e-07, + "loss": 0.703, + "step": 2412 + }, + { + "epoch": 0.09, + "learning_rate": 1.913465411799172e-07, + "loss": 0.6805, + "step": 2413 + }, + { + "epoch": 0.09, + "learning_rate": 1.9134287911524516e-07, + "loss": 0.7096, + "step": 2414 + }, + { + "epoch": 0.09, + "learning_rate": 1.913392170505731e-07, + "loss": 0.692, + "step": 2415 + }, + { + "epoch": 0.09, + "learning_rate": 1.9133555498590104e-07, + "loss": 0.7032, + "step": 2416 + }, + { + "epoch": 0.09, + "learning_rate": 1.9133189292122897e-07, + "loss": 0.6755, + "step": 2417 + }, + { + "epoch": 0.09, + "learning_rate": 1.9132823085655692e-07, + "loss": 0.6976, + "step": 2418 + }, + { + "epoch": 0.09, + "learning_rate": 1.9132456879188485e-07, + "loss": 0.6805, + "step": 2419 + }, + { + "epoch": 0.09, + "learning_rate": 1.913209067272128e-07, + "loss": 0.7024, + "step": 2420 + }, + { + "epoch": 0.09, + "learning_rate": 1.9131724466254073e-07, + "loss": 0.6979, + "step": 2421 + }, + { + "epoch": 0.09, + "learning_rate": 1.9131358259786868e-07, + "loss": 0.6881, + "step": 2422 + }, + { + "epoch": 0.09, + "learning_rate": 1.913099205331966e-07, + "loss": 0.707, + "step": 2423 + }, + { + "epoch": 0.09, + "learning_rate": 1.9130625846852454e-07, + "loss": 0.6951, + "step": 2424 + }, + { + "epoch": 0.09, + "learning_rate": 1.913025964038525e-07, + "loss": 0.6976, + "step": 2425 + }, + { + "epoch": 0.09, + "learning_rate": 1.9129893433918042e-07, + "loss": 0.692, + "step": 2426 + }, + { + "epoch": 0.09, + "learning_rate": 1.9129527227450837e-07, + "loss": 0.7066, + "step": 2427 + }, + { + "epoch": 0.09, + "learning_rate": 1.912916102098363e-07, + "loss": 0.6897, + "step": 2428 + }, + { + "epoch": 0.09, + "learning_rate": 1.9128794814516422e-07, + "loss": 0.7084, + "step": 2429 + }, + { + "epoch": 0.09, + "learning_rate": 1.9128428608049215e-07, + "loss": 0.68, + "step": 2430 + }, + { + "epoch": 0.09, + "learning_rate": 1.912806240158201e-07, + "loss": 0.7119, + "step": 2431 + }, + { + "epoch": 0.09, + "learning_rate": 1.9127696195114803e-07, + "loss": 0.7026, + "step": 2432 + }, + { + "epoch": 0.09, + "learning_rate": 1.9127329988647598e-07, + "loss": 0.698, + "step": 2433 + }, + { + "epoch": 0.09, + "learning_rate": 1.912696378218039e-07, + "loss": 0.6934, + "step": 2434 + }, + { + "epoch": 0.09, + "learning_rate": 1.9126597575713186e-07, + "loss": 0.7116, + "step": 2435 + }, + { + "epoch": 0.09, + "learning_rate": 1.912623136924598e-07, + "loss": 0.6925, + "step": 2436 + }, + { + "epoch": 0.09, + "learning_rate": 1.9125865162778774e-07, + "loss": 0.681, + "step": 2437 + }, + { + "epoch": 0.09, + "learning_rate": 1.9125498956311567e-07, + "loss": 0.6682, + "step": 2438 + }, + { + "epoch": 0.09, + "learning_rate": 1.9125132749844362e-07, + "loss": 0.6973, + "step": 2439 + }, + { + "epoch": 0.09, + "learning_rate": 1.9124766543377155e-07, + "loss": 0.7123, + "step": 2440 + }, + { + "epoch": 0.09, + "learning_rate": 1.912440033690995e-07, + "loss": 0.7063, + "step": 2441 + }, + { + "epoch": 0.09, + "learning_rate": 1.9124034130442743e-07, + "loss": 0.6951, + "step": 2442 + }, + { + "epoch": 0.09, + "learning_rate": 1.9123667923975538e-07, + "loss": 0.6872, + "step": 2443 + }, + { + "epoch": 0.09, + "learning_rate": 1.912330171750833e-07, + "loss": 0.6907, + "step": 2444 + }, + { + "epoch": 0.09, + "learning_rate": 1.9122935511041126e-07, + "loss": 0.6702, + "step": 2445 + }, + { + "epoch": 0.09, + "learning_rate": 1.912256930457392e-07, + "loss": 0.7027, + "step": 2446 + }, + { + "epoch": 0.09, + "learning_rate": 1.9122203098106712e-07, + "loss": 0.6976, + "step": 2447 + }, + { + "epoch": 0.09, + "learning_rate": 1.9121836891639504e-07, + "loss": 0.69, + "step": 2448 + }, + { + "epoch": 0.09, + "learning_rate": 1.91214706851723e-07, + "loss": 0.6901, + "step": 2449 + }, + { + "epoch": 0.09, + "learning_rate": 1.9121104478705092e-07, + "loss": 0.6803, + "step": 2450 + }, + { + "epoch": 0.09, + "learning_rate": 1.9120738272237888e-07, + "loss": 0.6903, + "step": 2451 + }, + { + "epoch": 0.09, + "learning_rate": 1.912037206577068e-07, + "loss": 0.7161, + "step": 2452 + }, + { + "epoch": 0.09, + "learning_rate": 1.9120005859303473e-07, + "loss": 0.6845, + "step": 2453 + }, + { + "epoch": 0.09, + "learning_rate": 1.9119639652836268e-07, + "loss": 0.6754, + "step": 2454 + }, + { + "epoch": 0.09, + "learning_rate": 1.911927344636906e-07, + "loss": 0.6994, + "step": 2455 + }, + { + "epoch": 0.09, + "learning_rate": 1.9118907239901856e-07, + "loss": 0.6879, + "step": 2456 + }, + { + "epoch": 0.09, + "learning_rate": 1.911854103343465e-07, + "loss": 0.7033, + "step": 2457 + }, + { + "epoch": 0.09, + "learning_rate": 1.9118174826967444e-07, + "loss": 0.7095, + "step": 2458 + }, + { + "epoch": 0.09, + "learning_rate": 1.9117808620500237e-07, + "loss": 0.7095, + "step": 2459 + }, + { + "epoch": 0.09, + "learning_rate": 1.9117442414033032e-07, + "loss": 0.69, + "step": 2460 + }, + { + "epoch": 0.09, + "learning_rate": 1.9117076207565825e-07, + "loss": 0.6866, + "step": 2461 + }, + { + "epoch": 0.09, + "learning_rate": 1.911671000109862e-07, + "loss": 0.6946, + "step": 2462 + }, + { + "epoch": 0.09, + "learning_rate": 1.9116343794631413e-07, + "loss": 0.692, + "step": 2463 + }, + { + "epoch": 0.09, + "learning_rate": 1.9115977588164208e-07, + "loss": 0.7078, + "step": 2464 + }, + { + "epoch": 0.09, + "learning_rate": 1.9115611381696998e-07, + "loss": 0.6841, + "step": 2465 + }, + { + "epoch": 0.09, + "learning_rate": 1.9115245175229794e-07, + "loss": 0.7113, + "step": 2466 + }, + { + "epoch": 0.09, + "learning_rate": 1.9114878968762586e-07, + "loss": 0.6886, + "step": 2467 + }, + { + "epoch": 0.09, + "learning_rate": 1.9114512762295381e-07, + "loss": 0.6935, + "step": 2468 + }, + { + "epoch": 0.09, + "learning_rate": 1.9114146555828174e-07, + "loss": 0.6955, + "step": 2469 + }, + { + "epoch": 0.09, + "learning_rate": 1.911378034936097e-07, + "loss": 0.6909, + "step": 2470 + }, + { + "epoch": 0.09, + "learning_rate": 1.9113414142893762e-07, + "loss": 0.6722, + "step": 2471 + }, + { + "epoch": 0.09, + "learning_rate": 1.9113047936426557e-07, + "loss": 0.6782, + "step": 2472 + }, + { + "epoch": 0.09, + "learning_rate": 1.911268172995935e-07, + "loss": 0.6736, + "step": 2473 + }, + { + "epoch": 0.09, + "learning_rate": 1.9112315523492143e-07, + "loss": 0.7072, + "step": 2474 + }, + { + "epoch": 0.09, + "learning_rate": 1.9111949317024938e-07, + "loss": 0.6833, + "step": 2475 + }, + { + "epoch": 0.09, + "learning_rate": 1.911158311055773e-07, + "loss": 0.6933, + "step": 2476 + }, + { + "epoch": 0.09, + "learning_rate": 1.9111216904090526e-07, + "loss": 0.7122, + "step": 2477 + }, + { + "epoch": 0.09, + "learning_rate": 1.911085069762332e-07, + "loss": 0.6949, + "step": 2478 + }, + { + "epoch": 0.09, + "learning_rate": 1.9110484491156114e-07, + "loss": 0.7081, + "step": 2479 + }, + { + "epoch": 0.09, + "learning_rate": 1.9110118284688907e-07, + "loss": 0.701, + "step": 2480 + }, + { + "epoch": 0.09, + "learning_rate": 1.9109752078221702e-07, + "loss": 0.7078, + "step": 2481 + }, + { + "epoch": 0.09, + "learning_rate": 1.9109385871754495e-07, + "loss": 0.6959, + "step": 2482 + }, + { + "epoch": 0.09, + "learning_rate": 1.9109019665287287e-07, + "loss": 0.7, + "step": 2483 + }, + { + "epoch": 0.09, + "learning_rate": 1.910865345882008e-07, + "loss": 0.6944, + "step": 2484 + }, + { + "epoch": 0.09, + "learning_rate": 1.9108287252352875e-07, + "loss": 0.6937, + "step": 2485 + }, + { + "epoch": 0.09, + "learning_rate": 1.9107921045885668e-07, + "loss": 0.6885, + "step": 2486 + }, + { + "epoch": 0.09, + "learning_rate": 1.9107554839418463e-07, + "loss": 0.6919, + "step": 2487 + }, + { + "epoch": 0.09, + "learning_rate": 1.9107188632951256e-07, + "loss": 0.6929, + "step": 2488 + }, + { + "epoch": 0.09, + "learning_rate": 1.9106822426484051e-07, + "loss": 0.7228, + "step": 2489 + }, + { + "epoch": 0.09, + "learning_rate": 1.9106456220016844e-07, + "loss": 0.702, + "step": 2490 + }, + { + "epoch": 0.09, + "learning_rate": 1.910609001354964e-07, + "loss": 0.698, + "step": 2491 + }, + { + "epoch": 0.09, + "learning_rate": 1.9105723807082432e-07, + "loss": 0.7098, + "step": 2492 + }, + { + "epoch": 0.09, + "learning_rate": 1.9105357600615227e-07, + "loss": 0.7309, + "step": 2493 + }, + { + "epoch": 0.09, + "learning_rate": 1.910499139414802e-07, + "loss": 0.7101, + "step": 2494 + }, + { + "epoch": 0.09, + "learning_rate": 1.9104625187680813e-07, + "loss": 0.6813, + "step": 2495 + }, + { + "epoch": 0.09, + "learning_rate": 1.9104258981213608e-07, + "loss": 0.686, + "step": 2496 + }, + { + "epoch": 0.09, + "learning_rate": 1.91038927747464e-07, + "loss": 0.7016, + "step": 2497 + }, + { + "epoch": 0.09, + "learning_rate": 1.9103526568279196e-07, + "loss": 0.6649, + "step": 2498 + }, + { + "epoch": 0.09, + "learning_rate": 1.9103160361811989e-07, + "loss": 0.6774, + "step": 2499 + }, + { + "epoch": 0.09, + "learning_rate": 1.9102794155344784e-07, + "loss": 0.6994, + "step": 2500 + }, + { + "epoch": 0.09, + "learning_rate": 1.9102427948877574e-07, + "loss": 0.6755, + "step": 2501 + }, + { + "epoch": 0.09, + "learning_rate": 1.910206174241037e-07, + "loss": 0.7222, + "step": 2502 + }, + { + "epoch": 0.09, + "learning_rate": 1.9101695535943162e-07, + "loss": 0.7104, + "step": 2503 + }, + { + "epoch": 0.09, + "learning_rate": 1.9101329329475957e-07, + "loss": 0.6834, + "step": 2504 + }, + { + "epoch": 0.09, + "learning_rate": 1.910096312300875e-07, + "loss": 0.6935, + "step": 2505 + }, + { + "epoch": 0.09, + "learning_rate": 1.9100596916541545e-07, + "loss": 0.6696, + "step": 2506 + }, + { + "epoch": 0.09, + "learning_rate": 1.9100230710074338e-07, + "loss": 0.7097, + "step": 2507 + }, + { + "epoch": 0.09, + "learning_rate": 1.9099864503607133e-07, + "loss": 0.6872, + "step": 2508 + }, + { + "epoch": 0.09, + "learning_rate": 1.9099498297139926e-07, + "loss": 0.6941, + "step": 2509 + }, + { + "epoch": 0.09, + "learning_rate": 1.909913209067272e-07, + "loss": 0.6866, + "step": 2510 + }, + { + "epoch": 0.09, + "learning_rate": 1.9098765884205514e-07, + "loss": 0.7122, + "step": 2511 + }, + { + "epoch": 0.09, + "learning_rate": 1.909839967773831e-07, + "loss": 0.6902, + "step": 2512 + }, + { + "epoch": 0.09, + "learning_rate": 1.9098033471271102e-07, + "loss": 0.6922, + "step": 2513 + }, + { + "epoch": 0.09, + "learning_rate": 1.9097667264803897e-07, + "loss": 0.6953, + "step": 2514 + }, + { + "epoch": 0.09, + "learning_rate": 1.909730105833669e-07, + "loss": 0.7034, + "step": 2515 + }, + { + "epoch": 0.09, + "learning_rate": 1.9096934851869485e-07, + "loss": 0.7071, + "step": 2516 + }, + { + "epoch": 0.09, + "learning_rate": 1.9096568645402278e-07, + "loss": 0.661, + "step": 2517 + }, + { + "epoch": 0.09, + "learning_rate": 1.909620243893507e-07, + "loss": 0.6897, + "step": 2518 + }, + { + "epoch": 0.09, + "learning_rate": 1.9095836232467863e-07, + "loss": 0.7096, + "step": 2519 + }, + { + "epoch": 0.09, + "learning_rate": 1.9095470026000659e-07, + "loss": 0.7074, + "step": 2520 + }, + { + "epoch": 0.09, + "learning_rate": 1.909510381953345e-07, + "loss": 0.6935, + "step": 2521 + }, + { + "epoch": 0.09, + "learning_rate": 1.9094737613066244e-07, + "loss": 0.6828, + "step": 2522 + }, + { + "epoch": 0.09, + "learning_rate": 1.909437140659904e-07, + "loss": 0.7091, + "step": 2523 + }, + { + "epoch": 0.09, + "learning_rate": 1.9094005200131832e-07, + "loss": 0.6976, + "step": 2524 + }, + { + "epoch": 0.09, + "learning_rate": 1.9093638993664627e-07, + "loss": 0.6995, + "step": 2525 + }, + { + "epoch": 0.09, + "learning_rate": 1.909327278719742e-07, + "loss": 0.699, + "step": 2526 + }, + { + "epoch": 0.09, + "learning_rate": 1.9092906580730215e-07, + "loss": 0.6788, + "step": 2527 + }, + { + "epoch": 0.09, + "learning_rate": 1.9092540374263008e-07, + "loss": 0.6984, + "step": 2528 + }, + { + "epoch": 0.09, + "learning_rate": 1.9092174167795803e-07, + "loss": 0.6804, + "step": 2529 + }, + { + "epoch": 0.09, + "learning_rate": 1.9091807961328596e-07, + "loss": 0.7078, + "step": 2530 + }, + { + "epoch": 0.09, + "learning_rate": 1.909144175486139e-07, + "loss": 0.681, + "step": 2531 + }, + { + "epoch": 0.09, + "learning_rate": 1.9091075548394184e-07, + "loss": 0.7076, + "step": 2532 + }, + { + "epoch": 0.09, + "learning_rate": 1.909070934192698e-07, + "loss": 0.7059, + "step": 2533 + }, + { + "epoch": 0.09, + "learning_rate": 1.9090343135459772e-07, + "loss": 0.7185, + "step": 2534 + }, + { + "epoch": 0.09, + "learning_rate": 1.9089976928992567e-07, + "loss": 0.7319, + "step": 2535 + }, + { + "epoch": 0.09, + "learning_rate": 1.908961072252536e-07, + "loss": 0.7015, + "step": 2536 + }, + { + "epoch": 0.09, + "learning_rate": 1.9089244516058152e-07, + "loss": 0.7077, + "step": 2537 + }, + { + "epoch": 0.09, + "learning_rate": 1.9088878309590945e-07, + "loss": 0.7208, + "step": 2538 + }, + { + "epoch": 0.09, + "learning_rate": 1.908851210312374e-07, + "loss": 0.6794, + "step": 2539 + }, + { + "epoch": 0.09, + "learning_rate": 1.9088145896656533e-07, + "loss": 0.6901, + "step": 2540 + }, + { + "epoch": 0.09, + "learning_rate": 1.9087779690189328e-07, + "loss": 0.7047, + "step": 2541 + }, + { + "epoch": 0.09, + "learning_rate": 1.908741348372212e-07, + "loss": 0.7044, + "step": 2542 + }, + { + "epoch": 0.09, + "learning_rate": 1.9087047277254916e-07, + "loss": 0.666, + "step": 2543 + }, + { + "epoch": 0.09, + "learning_rate": 1.908668107078771e-07, + "loss": 0.6945, + "step": 2544 + }, + { + "epoch": 0.09, + "learning_rate": 1.9086314864320502e-07, + "loss": 0.7025, + "step": 2545 + }, + { + "epoch": 0.09, + "learning_rate": 1.9085948657853297e-07, + "loss": 0.682, + "step": 2546 + }, + { + "epoch": 0.09, + "learning_rate": 1.908558245138609e-07, + "loss": 0.6873, + "step": 2547 + }, + { + "epoch": 0.09, + "learning_rate": 1.9085216244918885e-07, + "loss": 0.6946, + "step": 2548 + }, + { + "epoch": 0.09, + "learning_rate": 1.9084850038451678e-07, + "loss": 0.7245, + "step": 2549 + }, + { + "epoch": 0.09, + "learning_rate": 1.9084483831984473e-07, + "loss": 0.7021, + "step": 2550 + }, + { + "epoch": 0.09, + "learning_rate": 1.9084117625517266e-07, + "loss": 0.7122, + "step": 2551 + }, + { + "epoch": 0.09, + "learning_rate": 1.908375141905006e-07, + "loss": 0.713, + "step": 2552 + }, + { + "epoch": 0.09, + "learning_rate": 1.9083385212582854e-07, + "loss": 0.7104, + "step": 2553 + }, + { + "epoch": 0.09, + "learning_rate": 1.908301900611565e-07, + "loss": 0.6736, + "step": 2554 + }, + { + "epoch": 0.09, + "learning_rate": 1.908265279964844e-07, + "loss": 0.6982, + "step": 2555 + }, + { + "epoch": 0.09, + "learning_rate": 1.9082286593181234e-07, + "loss": 0.6858, + "step": 2556 + }, + { + "epoch": 0.09, + "learning_rate": 1.9081920386714027e-07, + "loss": 0.6862, + "step": 2557 + }, + { + "epoch": 0.09, + "learning_rate": 1.9081554180246822e-07, + "loss": 0.7064, + "step": 2558 + }, + { + "epoch": 0.09, + "learning_rate": 1.9081187973779615e-07, + "loss": 0.6832, + "step": 2559 + }, + { + "epoch": 0.09, + "learning_rate": 1.908082176731241e-07, + "loss": 0.6974, + "step": 2560 + }, + { + "epoch": 0.09, + "learning_rate": 1.9080455560845203e-07, + "loss": 0.7035, + "step": 2561 + }, + { + "epoch": 0.09, + "learning_rate": 1.9080089354377998e-07, + "loss": 0.6791, + "step": 2562 + }, + { + "epoch": 0.09, + "learning_rate": 1.907972314791079e-07, + "loss": 0.6859, + "step": 2563 + }, + { + "epoch": 0.09, + "learning_rate": 1.9079356941443586e-07, + "loss": 0.693, + "step": 2564 + }, + { + "epoch": 0.09, + "learning_rate": 1.907899073497638e-07, + "loss": 0.6904, + "step": 2565 + }, + { + "epoch": 0.09, + "learning_rate": 1.9078624528509172e-07, + "loss": 0.6913, + "step": 2566 + }, + { + "epoch": 0.09, + "learning_rate": 1.9078258322041967e-07, + "loss": 0.7079, + "step": 2567 + }, + { + "epoch": 0.09, + "learning_rate": 1.907789211557476e-07, + "loss": 0.6908, + "step": 2568 + }, + { + "epoch": 0.09, + "learning_rate": 1.9077525909107555e-07, + "loss": 0.7084, + "step": 2569 + }, + { + "epoch": 0.09, + "learning_rate": 1.9077159702640348e-07, + "loss": 0.7108, + "step": 2570 + }, + { + "epoch": 0.09, + "learning_rate": 1.9076793496173143e-07, + "loss": 0.6866, + "step": 2571 + }, + { + "epoch": 0.09, + "learning_rate": 1.9076427289705936e-07, + "loss": 0.7203, + "step": 2572 + }, + { + "epoch": 0.09, + "learning_rate": 1.9076061083238728e-07, + "loss": 0.6758, + "step": 2573 + }, + { + "epoch": 0.09, + "learning_rate": 1.907569487677152e-07, + "loss": 0.6876, + "step": 2574 + }, + { + "epoch": 0.09, + "learning_rate": 1.9075328670304316e-07, + "loss": 0.684, + "step": 2575 + }, + { + "epoch": 0.09, + "learning_rate": 1.907496246383711e-07, + "loss": 0.6939, + "step": 2576 + }, + { + "epoch": 0.09, + "learning_rate": 1.9074596257369904e-07, + "loss": 0.701, + "step": 2577 + }, + { + "epoch": 0.09, + "learning_rate": 1.9074230050902697e-07, + "loss": 0.6765, + "step": 2578 + }, + { + "epoch": 0.09, + "learning_rate": 1.9073863844435492e-07, + "loss": 0.7131, + "step": 2579 + }, + { + "epoch": 0.09, + "learning_rate": 1.9073497637968285e-07, + "loss": 0.709, + "step": 2580 + }, + { + "epoch": 0.09, + "learning_rate": 1.907313143150108e-07, + "loss": 0.6977, + "step": 2581 + }, + { + "epoch": 0.09, + "learning_rate": 1.9072765225033873e-07, + "loss": 0.6866, + "step": 2582 + }, + { + "epoch": 0.09, + "learning_rate": 1.9072399018566668e-07, + "loss": 0.7117, + "step": 2583 + }, + { + "epoch": 0.09, + "learning_rate": 1.907203281209946e-07, + "loss": 0.6852, + "step": 2584 + }, + { + "epoch": 0.09, + "learning_rate": 1.9071666605632256e-07, + "loss": 0.6844, + "step": 2585 + }, + { + "epoch": 0.09, + "learning_rate": 1.907130039916505e-07, + "loss": 0.6738, + "step": 2586 + }, + { + "epoch": 0.09, + "learning_rate": 1.9070934192697844e-07, + "loss": 0.6786, + "step": 2587 + }, + { + "epoch": 0.09, + "learning_rate": 1.9070567986230637e-07, + "loss": 0.6849, + "step": 2588 + }, + { + "epoch": 0.09, + "learning_rate": 1.907020177976343e-07, + "loss": 0.6855, + "step": 2589 + }, + { + "epoch": 0.09, + "learning_rate": 1.9069835573296225e-07, + "loss": 0.6956, + "step": 2590 + }, + { + "epoch": 0.09, + "learning_rate": 1.9069469366829017e-07, + "loss": 0.6762, + "step": 2591 + }, + { + "epoch": 0.09, + "learning_rate": 1.906910316036181e-07, + "loss": 0.6949, + "step": 2592 + }, + { + "epoch": 0.09, + "learning_rate": 1.9068736953894603e-07, + "loss": 0.6909, + "step": 2593 + }, + { + "epoch": 0.09, + "learning_rate": 1.9068370747427398e-07, + "loss": 0.7079, + "step": 2594 + }, + { + "epoch": 0.09, + "learning_rate": 1.906800454096019e-07, + "loss": 0.6632, + "step": 2595 + }, + { + "epoch": 0.09, + "learning_rate": 1.9067638334492986e-07, + "loss": 0.7125, + "step": 2596 + }, + { + "epoch": 0.1, + "learning_rate": 1.906727212802578e-07, + "loss": 0.6912, + "step": 2597 + }, + { + "epoch": 0.1, + "learning_rate": 1.9066905921558574e-07, + "loss": 0.6948, + "step": 2598 + }, + { + "epoch": 0.1, + "learning_rate": 1.9066539715091367e-07, + "loss": 0.6872, + "step": 2599 + }, + { + "epoch": 0.1, + "learning_rate": 1.9066173508624162e-07, + "loss": 0.6886, + "step": 2600 + }, + { + "epoch": 0.1, + "learning_rate": 1.9065807302156955e-07, + "loss": 0.7007, + "step": 2601 + }, + { + "epoch": 0.1, + "learning_rate": 1.906544109568975e-07, + "loss": 0.712, + "step": 2602 + }, + { + "epoch": 0.1, + "learning_rate": 1.9065074889222543e-07, + "loss": 0.6833, + "step": 2603 + }, + { + "epoch": 0.1, + "learning_rate": 1.9064708682755338e-07, + "loss": 0.6967, + "step": 2604 + }, + { + "epoch": 0.1, + "learning_rate": 1.906434247628813e-07, + "loss": 0.6932, + "step": 2605 + }, + { + "epoch": 0.1, + "learning_rate": 1.9063976269820926e-07, + "loss": 0.697, + "step": 2606 + }, + { + "epoch": 0.1, + "learning_rate": 1.906361006335372e-07, + "loss": 0.6952, + "step": 2607 + }, + { + "epoch": 0.1, + "learning_rate": 1.9063243856886514e-07, + "loss": 0.6922, + "step": 2608 + }, + { + "epoch": 0.1, + "learning_rate": 1.9062877650419304e-07, + "loss": 0.6778, + "step": 2609 + }, + { + "epoch": 0.1, + "learning_rate": 1.90625114439521e-07, + "loss": 0.6884, + "step": 2610 + }, + { + "epoch": 0.1, + "learning_rate": 1.9062145237484892e-07, + "loss": 0.6946, + "step": 2611 + }, + { + "epoch": 0.1, + "learning_rate": 1.9061779031017687e-07, + "loss": 0.6739, + "step": 2612 + }, + { + "epoch": 0.1, + "learning_rate": 1.906141282455048e-07, + "loss": 0.6986, + "step": 2613 + }, + { + "epoch": 0.1, + "learning_rate": 1.9061046618083275e-07, + "loss": 0.6978, + "step": 2614 + }, + { + "epoch": 0.1, + "learning_rate": 1.9060680411616068e-07, + "loss": 0.6893, + "step": 2615 + }, + { + "epoch": 0.1, + "learning_rate": 1.906031420514886e-07, + "loss": 0.6721, + "step": 2616 + }, + { + "epoch": 0.1, + "learning_rate": 1.9059947998681656e-07, + "loss": 0.6791, + "step": 2617 + }, + { + "epoch": 0.1, + "learning_rate": 1.9059581792214449e-07, + "loss": 0.7005, + "step": 2618 + }, + { + "epoch": 0.1, + "learning_rate": 1.9059215585747244e-07, + "loss": 0.6856, + "step": 2619 + }, + { + "epoch": 0.1, + "learning_rate": 1.9058849379280037e-07, + "loss": 0.6813, + "step": 2620 + }, + { + "epoch": 0.1, + "learning_rate": 1.9058483172812832e-07, + "loss": 0.7071, + "step": 2621 + }, + { + "epoch": 0.1, + "learning_rate": 1.9058116966345625e-07, + "loss": 0.6839, + "step": 2622 + }, + { + "epoch": 0.1, + "learning_rate": 1.905775075987842e-07, + "loss": 0.6681, + "step": 2623 + }, + { + "epoch": 0.1, + "learning_rate": 1.9057384553411213e-07, + "loss": 0.6964, + "step": 2624 + }, + { + "epoch": 0.1, + "learning_rate": 1.9057018346944008e-07, + "loss": 0.7053, + "step": 2625 + }, + { + "epoch": 0.1, + "learning_rate": 1.90566521404768e-07, + "loss": 0.7045, + "step": 2626 + }, + { + "epoch": 0.1, + "learning_rate": 1.9056285934009593e-07, + "loss": 0.6954, + "step": 2627 + }, + { + "epoch": 0.1, + "learning_rate": 1.9055919727542386e-07, + "loss": 0.6789, + "step": 2628 + }, + { + "epoch": 0.1, + "learning_rate": 1.905555352107518e-07, + "loss": 0.6877, + "step": 2629 + }, + { + "epoch": 0.1, + "learning_rate": 1.9055187314607974e-07, + "loss": 0.6866, + "step": 2630 + }, + { + "epoch": 0.1, + "learning_rate": 1.905482110814077e-07, + "loss": 0.6717, + "step": 2631 + }, + { + "epoch": 0.1, + "learning_rate": 1.9054454901673562e-07, + "loss": 0.7165, + "step": 2632 + }, + { + "epoch": 0.1, + "learning_rate": 1.9054088695206357e-07, + "loss": 0.6731, + "step": 2633 + }, + { + "epoch": 0.1, + "learning_rate": 1.905372248873915e-07, + "loss": 0.6895, + "step": 2634 + }, + { + "epoch": 0.1, + "learning_rate": 1.9053356282271945e-07, + "loss": 0.6927, + "step": 2635 + }, + { + "epoch": 0.1, + "learning_rate": 1.9052990075804738e-07, + "loss": 0.689, + "step": 2636 + }, + { + "epoch": 0.1, + "learning_rate": 1.905262386933753e-07, + "loss": 0.6902, + "step": 2637 + }, + { + "epoch": 0.1, + "learning_rate": 1.9052257662870326e-07, + "loss": 0.6981, + "step": 2638 + }, + { + "epoch": 0.1, + "learning_rate": 1.9051891456403119e-07, + "loss": 0.7042, + "step": 2639 + }, + { + "epoch": 0.1, + "learning_rate": 1.9051525249935914e-07, + "loss": 0.6875, + "step": 2640 + }, + { + "epoch": 0.1, + "learning_rate": 1.9051159043468707e-07, + "loss": 0.6941, + "step": 2641 + }, + { + "epoch": 0.1, + "learning_rate": 1.9050792837001502e-07, + "loss": 0.7032, + "step": 2642 + }, + { + "epoch": 0.1, + "learning_rate": 1.9050426630534295e-07, + "loss": 0.6974, + "step": 2643 + }, + { + "epoch": 0.1, + "learning_rate": 1.905006042406709e-07, + "loss": 0.6942, + "step": 2644 + }, + { + "epoch": 0.1, + "learning_rate": 1.904969421759988e-07, + "loss": 0.6795, + "step": 2645 + }, + { + "epoch": 0.1, + "learning_rate": 1.9049328011132675e-07, + "loss": 0.6761, + "step": 2646 + }, + { + "epoch": 0.1, + "learning_rate": 1.9048961804665468e-07, + "loss": 0.6893, + "step": 2647 + }, + { + "epoch": 0.1, + "learning_rate": 1.9048595598198263e-07, + "loss": 0.6896, + "step": 2648 + }, + { + "epoch": 0.1, + "learning_rate": 1.9048229391731056e-07, + "loss": 0.6952, + "step": 2649 + }, + { + "epoch": 0.1, + "learning_rate": 1.904786318526385e-07, + "loss": 0.6539, + "step": 2650 + }, + { + "epoch": 0.1, + "learning_rate": 1.9047496978796644e-07, + "loss": 0.6881, + "step": 2651 + }, + { + "epoch": 0.1, + "learning_rate": 1.904713077232944e-07, + "loss": 0.7029, + "step": 2652 + }, + { + "epoch": 0.1, + "learning_rate": 1.9046764565862232e-07, + "loss": 0.702, + "step": 2653 + }, + { + "epoch": 0.1, + "learning_rate": 1.9046398359395027e-07, + "loss": 0.7024, + "step": 2654 + }, + { + "epoch": 0.1, + "learning_rate": 1.904603215292782e-07, + "loss": 0.6836, + "step": 2655 + }, + { + "epoch": 0.1, + "learning_rate": 1.9045665946460615e-07, + "loss": 0.6978, + "step": 2656 + }, + { + "epoch": 0.1, + "learning_rate": 1.9045299739993408e-07, + "loss": 0.7032, + "step": 2657 + }, + { + "epoch": 0.1, + "learning_rate": 1.9044933533526203e-07, + "loss": 0.6853, + "step": 2658 + }, + { + "epoch": 0.1, + "learning_rate": 1.9044567327058996e-07, + "loss": 0.6919, + "step": 2659 + }, + { + "epoch": 0.1, + "learning_rate": 1.9044201120591788e-07, + "loss": 0.685, + "step": 2660 + }, + { + "epoch": 0.1, + "learning_rate": 1.9043834914124584e-07, + "loss": 0.6981, + "step": 2661 + }, + { + "epoch": 0.1, + "learning_rate": 1.9043468707657376e-07, + "loss": 0.7011, + "step": 2662 + }, + { + "epoch": 0.1, + "learning_rate": 1.904310250119017e-07, + "loss": 0.6756, + "step": 2663 + }, + { + "epoch": 0.1, + "learning_rate": 1.9042736294722962e-07, + "loss": 0.7128, + "step": 2664 + }, + { + "epoch": 0.1, + "learning_rate": 1.9042370088255757e-07, + "loss": 0.6992, + "step": 2665 + }, + { + "epoch": 0.1, + "learning_rate": 1.904200388178855e-07, + "loss": 0.6712, + "step": 2666 + }, + { + "epoch": 0.1, + "learning_rate": 1.9041637675321345e-07, + "loss": 0.6984, + "step": 2667 + }, + { + "epoch": 0.1, + "learning_rate": 1.9041271468854138e-07, + "loss": 0.724, + "step": 2668 + }, + { + "epoch": 0.1, + "learning_rate": 1.9040905262386933e-07, + "loss": 0.6957, + "step": 2669 + }, + { + "epoch": 0.1, + "learning_rate": 1.9040539055919726e-07, + "loss": 0.6924, + "step": 2670 + }, + { + "epoch": 0.1, + "learning_rate": 1.904017284945252e-07, + "loss": 0.6983, + "step": 2671 + }, + { + "epoch": 0.1, + "learning_rate": 1.9039806642985314e-07, + "loss": 0.69, + "step": 2672 + }, + { + "epoch": 0.1, + "learning_rate": 1.903944043651811e-07, + "loss": 0.6973, + "step": 2673 + }, + { + "epoch": 0.1, + "learning_rate": 1.9039074230050902e-07, + "loss": 0.7204, + "step": 2674 + }, + { + "epoch": 0.1, + "learning_rate": 1.9038708023583697e-07, + "loss": 0.6927, + "step": 2675 + }, + { + "epoch": 0.1, + "learning_rate": 1.903834181711649e-07, + "loss": 0.7058, + "step": 2676 + }, + { + "epoch": 0.1, + "learning_rate": 1.9037975610649285e-07, + "loss": 0.6828, + "step": 2677 + }, + { + "epoch": 0.1, + "learning_rate": 1.9037609404182078e-07, + "loss": 0.6897, + "step": 2678 + }, + { + "epoch": 0.1, + "learning_rate": 1.9037243197714873e-07, + "loss": 0.6709, + "step": 2679 + }, + { + "epoch": 0.1, + "learning_rate": 1.9036876991247666e-07, + "loss": 0.6956, + "step": 2680 + }, + { + "epoch": 0.1, + "learning_rate": 1.9036510784780458e-07, + "loss": 0.7041, + "step": 2681 + }, + { + "epoch": 0.1, + "learning_rate": 1.903614457831325e-07, + "loss": 0.703, + "step": 2682 + }, + { + "epoch": 0.1, + "learning_rate": 1.9035778371846046e-07, + "loss": 0.7052, + "step": 2683 + }, + { + "epoch": 0.1, + "learning_rate": 1.903541216537884e-07, + "loss": 0.6865, + "step": 2684 + }, + { + "epoch": 0.1, + "learning_rate": 1.9035045958911634e-07, + "loss": 0.6986, + "step": 2685 + }, + { + "epoch": 0.1, + "learning_rate": 1.9034679752444427e-07, + "loss": 0.6957, + "step": 2686 + }, + { + "epoch": 0.1, + "learning_rate": 1.903431354597722e-07, + "loss": 0.694, + "step": 2687 + }, + { + "epoch": 0.1, + "learning_rate": 1.9033947339510015e-07, + "loss": 0.6814, + "step": 2688 + }, + { + "epoch": 0.1, + "learning_rate": 1.9033581133042808e-07, + "loss": 0.7023, + "step": 2689 + }, + { + "epoch": 0.1, + "learning_rate": 1.9033214926575603e-07, + "loss": 0.6638, + "step": 2690 + }, + { + "epoch": 0.1, + "learning_rate": 1.9032848720108396e-07, + "loss": 0.6869, + "step": 2691 + }, + { + "epoch": 0.1, + "learning_rate": 1.903248251364119e-07, + "loss": 0.6945, + "step": 2692 + }, + { + "epoch": 0.1, + "learning_rate": 1.9032116307173984e-07, + "loss": 0.6961, + "step": 2693 + }, + { + "epoch": 0.1, + "learning_rate": 1.903175010070678e-07, + "loss": 0.6897, + "step": 2694 + }, + { + "epoch": 0.1, + "learning_rate": 1.9031383894239572e-07, + "loss": 0.6844, + "step": 2695 + }, + { + "epoch": 0.1, + "learning_rate": 1.9031017687772367e-07, + "loss": 0.6898, + "step": 2696 + }, + { + "epoch": 0.1, + "learning_rate": 1.903065148130516e-07, + "loss": 0.7012, + "step": 2697 + }, + { + "epoch": 0.1, + "learning_rate": 1.9030285274837955e-07, + "loss": 0.7308, + "step": 2698 + }, + { + "epoch": 0.1, + "learning_rate": 1.9029919068370745e-07, + "loss": 0.7173, + "step": 2699 + }, + { + "epoch": 0.1, + "learning_rate": 1.902955286190354e-07, + "loss": 0.7059, + "step": 2700 + }, + { + "epoch": 0.1, + "learning_rate": 1.9029186655436333e-07, + "loss": 0.6739, + "step": 2701 + }, + { + "epoch": 0.1, + "learning_rate": 1.9028820448969128e-07, + "loss": 0.6885, + "step": 2702 + }, + { + "epoch": 0.1, + "learning_rate": 1.902845424250192e-07, + "loss": 0.693, + "step": 2703 + }, + { + "epoch": 0.1, + "learning_rate": 1.9028088036034716e-07, + "loss": 0.6643, + "step": 2704 + }, + { + "epoch": 0.1, + "learning_rate": 1.902772182956751e-07, + "loss": 0.7005, + "step": 2705 + }, + { + "epoch": 0.1, + "learning_rate": 1.9027355623100304e-07, + "loss": 0.6983, + "step": 2706 + }, + { + "epoch": 0.1, + "learning_rate": 1.9026989416633097e-07, + "loss": 0.6813, + "step": 2707 + }, + { + "epoch": 0.1, + "learning_rate": 1.902662321016589e-07, + "loss": 0.7116, + "step": 2708 + }, + { + "epoch": 0.1, + "learning_rate": 1.9026257003698685e-07, + "loss": 0.6785, + "step": 2709 + }, + { + "epoch": 0.1, + "learning_rate": 1.9025890797231477e-07, + "loss": 0.7006, + "step": 2710 + }, + { + "epoch": 0.1, + "learning_rate": 1.9025524590764273e-07, + "loss": 0.6835, + "step": 2711 + }, + { + "epoch": 0.1, + "learning_rate": 1.9025158384297065e-07, + "loss": 0.6878, + "step": 2712 + }, + { + "epoch": 0.1, + "learning_rate": 1.902479217782986e-07, + "loss": 0.7195, + "step": 2713 + }, + { + "epoch": 0.1, + "learning_rate": 1.9024425971362653e-07, + "loss": 0.7025, + "step": 2714 + }, + { + "epoch": 0.1, + "learning_rate": 1.902405976489545e-07, + "loss": 0.6848, + "step": 2715 + }, + { + "epoch": 0.1, + "learning_rate": 1.9023693558428241e-07, + "loss": 0.6974, + "step": 2716 + }, + { + "epoch": 0.1, + "learning_rate": 1.9023327351961034e-07, + "loss": 0.6797, + "step": 2717 + }, + { + "epoch": 0.1, + "learning_rate": 1.9022961145493827e-07, + "loss": 0.7145, + "step": 2718 + }, + { + "epoch": 0.1, + "learning_rate": 1.9022594939026622e-07, + "loss": 0.6785, + "step": 2719 + }, + { + "epoch": 0.1, + "learning_rate": 1.9022228732559415e-07, + "loss": 0.7013, + "step": 2720 + }, + { + "epoch": 0.1, + "learning_rate": 1.902186252609221e-07, + "loss": 0.6886, + "step": 2721 + }, + { + "epoch": 0.1, + "learning_rate": 1.9021496319625003e-07, + "loss": 0.6842, + "step": 2722 + }, + { + "epoch": 0.1, + "learning_rate": 1.9021130113157798e-07, + "loss": 0.6856, + "step": 2723 + }, + { + "epoch": 0.1, + "learning_rate": 1.902076390669059e-07, + "loss": 0.6921, + "step": 2724 + }, + { + "epoch": 0.1, + "learning_rate": 1.9020397700223386e-07, + "loss": 0.713, + "step": 2725 + }, + { + "epoch": 0.1, + "learning_rate": 1.902003149375618e-07, + "loss": 0.7032, + "step": 2726 + }, + { + "epoch": 0.1, + "learning_rate": 1.9019665287288974e-07, + "loss": 0.7074, + "step": 2727 + }, + { + "epoch": 0.1, + "learning_rate": 1.9019299080821767e-07, + "loss": 0.6856, + "step": 2728 + }, + { + "epoch": 0.1, + "learning_rate": 1.9018932874354562e-07, + "loss": 0.6806, + "step": 2729 + }, + { + "epoch": 0.1, + "learning_rate": 1.9018566667887355e-07, + "loss": 0.6961, + "step": 2730 + }, + { + "epoch": 0.1, + "learning_rate": 1.9018200461420147e-07, + "loss": 0.6981, + "step": 2731 + }, + { + "epoch": 0.1, + "learning_rate": 1.9017834254952943e-07, + "loss": 0.6838, + "step": 2732 + }, + { + "epoch": 0.1, + "learning_rate": 1.9017468048485735e-07, + "loss": 0.705, + "step": 2733 + }, + { + "epoch": 0.1, + "learning_rate": 1.901710184201853e-07, + "loss": 0.6981, + "step": 2734 + }, + { + "epoch": 0.1, + "learning_rate": 1.901673563555132e-07, + "loss": 0.6584, + "step": 2735 + }, + { + "epoch": 0.1, + "learning_rate": 1.9016369429084116e-07, + "loss": 0.6946, + "step": 2736 + }, + { + "epoch": 0.1, + "learning_rate": 1.9016003222616909e-07, + "loss": 0.7206, + "step": 2737 + }, + { + "epoch": 0.1, + "learning_rate": 1.9015637016149704e-07, + "loss": 0.6937, + "step": 2738 + }, + { + "epoch": 0.1, + "learning_rate": 1.9015270809682497e-07, + "loss": 0.7011, + "step": 2739 + }, + { + "epoch": 0.1, + "learning_rate": 1.9014904603215292e-07, + "loss": 0.6875, + "step": 2740 + }, + { + "epoch": 0.1, + "learning_rate": 1.9014538396748085e-07, + "loss": 0.6952, + "step": 2741 + }, + { + "epoch": 0.1, + "learning_rate": 1.901417219028088e-07, + "loss": 0.6879, + "step": 2742 + }, + { + "epoch": 0.1, + "learning_rate": 1.9013805983813673e-07, + "loss": 0.6803, + "step": 2743 + }, + { + "epoch": 0.1, + "learning_rate": 1.9013439777346468e-07, + "loss": 0.683, + "step": 2744 + }, + { + "epoch": 0.1, + "learning_rate": 1.901307357087926e-07, + "loss": 0.6976, + "step": 2745 + }, + { + "epoch": 0.1, + "learning_rate": 1.9012707364412056e-07, + "loss": 0.6995, + "step": 2746 + }, + { + "epoch": 0.1, + "learning_rate": 1.9012341157944849e-07, + "loss": 0.6963, + "step": 2747 + }, + { + "epoch": 0.1, + "learning_rate": 1.9011974951477644e-07, + "loss": 0.724, + "step": 2748 + }, + { + "epoch": 0.1, + "learning_rate": 1.9011608745010437e-07, + "loss": 0.6855, + "step": 2749 + }, + { + "epoch": 0.1, + "learning_rate": 1.9011242538543232e-07, + "loss": 0.7007, + "step": 2750 + }, + { + "epoch": 0.1, + "learning_rate": 1.9010876332076025e-07, + "loss": 0.6969, + "step": 2751 + }, + { + "epoch": 0.1, + "learning_rate": 1.9010510125608817e-07, + "loss": 0.7016, + "step": 2752 + }, + { + "epoch": 0.1, + "learning_rate": 1.901014391914161e-07, + "loss": 0.6766, + "step": 2753 + }, + { + "epoch": 0.1, + "learning_rate": 1.9009777712674405e-07, + "loss": 0.6979, + "step": 2754 + }, + { + "epoch": 0.1, + "learning_rate": 1.9009411506207198e-07, + "loss": 0.6968, + "step": 2755 + }, + { + "epoch": 0.1, + "learning_rate": 1.9009045299739993e-07, + "loss": 0.6925, + "step": 2756 + }, + { + "epoch": 0.1, + "learning_rate": 1.9008679093272786e-07, + "loss": 0.6934, + "step": 2757 + }, + { + "epoch": 0.1, + "learning_rate": 1.9008312886805579e-07, + "loss": 0.6925, + "step": 2758 + }, + { + "epoch": 0.1, + "learning_rate": 1.9007946680338374e-07, + "loss": 0.7151, + "step": 2759 + }, + { + "epoch": 0.1, + "learning_rate": 1.9007580473871167e-07, + "loss": 0.6992, + "step": 2760 + }, + { + "epoch": 0.1, + "learning_rate": 1.9007214267403962e-07, + "loss": 0.7047, + "step": 2761 + }, + { + "epoch": 0.1, + "learning_rate": 1.9006848060936755e-07, + "loss": 0.6977, + "step": 2762 + }, + { + "epoch": 0.1, + "learning_rate": 1.900648185446955e-07, + "loss": 0.7031, + "step": 2763 + }, + { + "epoch": 0.1, + "learning_rate": 1.9006115648002343e-07, + "loss": 0.7021, + "step": 2764 + }, + { + "epoch": 0.1, + "learning_rate": 1.9005749441535138e-07, + "loss": 0.7079, + "step": 2765 + }, + { + "epoch": 0.1, + "learning_rate": 1.900538323506793e-07, + "loss": 0.6899, + "step": 2766 + }, + { + "epoch": 0.1, + "learning_rate": 1.9005017028600726e-07, + "loss": 0.6955, + "step": 2767 + }, + { + "epoch": 0.1, + "learning_rate": 1.9004650822133518e-07, + "loss": 0.6659, + "step": 2768 + }, + { + "epoch": 0.1, + "learning_rate": 1.9004284615666314e-07, + "loss": 0.6928, + "step": 2769 + }, + { + "epoch": 0.1, + "learning_rate": 1.9003918409199104e-07, + "loss": 0.6757, + "step": 2770 + }, + { + "epoch": 0.1, + "learning_rate": 1.90035522027319e-07, + "loss": 0.6799, + "step": 2771 + }, + { + "epoch": 0.1, + "learning_rate": 1.9003185996264692e-07, + "loss": 0.6927, + "step": 2772 + }, + { + "epoch": 0.1, + "learning_rate": 1.9002819789797487e-07, + "loss": 0.6988, + "step": 2773 + }, + { + "epoch": 0.1, + "learning_rate": 1.900245358333028e-07, + "loss": 0.7183, + "step": 2774 + }, + { + "epoch": 0.1, + "learning_rate": 1.9002087376863075e-07, + "loss": 0.6859, + "step": 2775 + }, + { + "epoch": 0.1, + "learning_rate": 1.9001721170395868e-07, + "loss": 0.675, + "step": 2776 + }, + { + "epoch": 0.1, + "learning_rate": 1.9001354963928663e-07, + "loss": 0.6722, + "step": 2777 + }, + { + "epoch": 0.1, + "learning_rate": 1.9000988757461456e-07, + "loss": 0.686, + "step": 2778 + }, + { + "epoch": 0.1, + "learning_rate": 1.9000622550994248e-07, + "loss": 0.6879, + "step": 2779 + }, + { + "epoch": 0.1, + "learning_rate": 1.9000256344527044e-07, + "loss": 0.6946, + "step": 2780 + }, + { + "epoch": 0.1, + "learning_rate": 1.8999890138059836e-07, + "loss": 0.6903, + "step": 2781 + }, + { + "epoch": 0.1, + "learning_rate": 1.8999523931592632e-07, + "loss": 0.6814, + "step": 2782 + }, + { + "epoch": 0.1, + "learning_rate": 1.8999157725125424e-07, + "loss": 0.6875, + "step": 2783 + }, + { + "epoch": 0.1, + "learning_rate": 1.899879151865822e-07, + "loss": 0.6922, + "step": 2784 + }, + { + "epoch": 0.1, + "learning_rate": 1.8998425312191012e-07, + "loss": 0.6708, + "step": 2785 + }, + { + "epoch": 0.1, + "learning_rate": 1.8998059105723808e-07, + "loss": 0.6952, + "step": 2786 + }, + { + "epoch": 0.1, + "learning_rate": 1.89976928992566e-07, + "loss": 0.6928, + "step": 2787 + }, + { + "epoch": 0.1, + "learning_rate": 1.8997326692789393e-07, + "loss": 0.7104, + "step": 2788 + }, + { + "epoch": 0.1, + "learning_rate": 1.8996960486322186e-07, + "loss": 0.7061, + "step": 2789 + }, + { + "epoch": 0.1, + "learning_rate": 1.899659427985498e-07, + "loss": 0.689, + "step": 2790 + }, + { + "epoch": 0.1, + "learning_rate": 1.8996228073387774e-07, + "loss": 0.685, + "step": 2791 + }, + { + "epoch": 0.1, + "learning_rate": 1.899586186692057e-07, + "loss": 0.7058, + "step": 2792 + }, + { + "epoch": 0.1, + "learning_rate": 1.8995495660453362e-07, + "loss": 0.6756, + "step": 2793 + }, + { + "epoch": 0.1, + "learning_rate": 1.8995129453986157e-07, + "loss": 0.6737, + "step": 2794 + }, + { + "epoch": 0.1, + "learning_rate": 1.899476324751895e-07, + "loss": 0.7054, + "step": 2795 + }, + { + "epoch": 0.1, + "learning_rate": 1.8994397041051745e-07, + "loss": 0.6765, + "step": 2796 + }, + { + "epoch": 0.1, + "learning_rate": 1.8994030834584538e-07, + "loss": 0.7122, + "step": 2797 + }, + { + "epoch": 0.1, + "learning_rate": 1.8993664628117333e-07, + "loss": 0.6674, + "step": 2798 + }, + { + "epoch": 0.1, + "learning_rate": 1.8993298421650126e-07, + "loss": 0.7075, + "step": 2799 + }, + { + "epoch": 0.1, + "learning_rate": 1.899293221518292e-07, + "loss": 0.6596, + "step": 2800 + }, + { + "epoch": 0.1, + "learning_rate": 1.8992566008715714e-07, + "loss": 0.6856, + "step": 2801 + }, + { + "epoch": 0.1, + "learning_rate": 1.8992199802248506e-07, + "loss": 0.6809, + "step": 2802 + }, + { + "epoch": 0.1, + "learning_rate": 1.8991833595781302e-07, + "loss": 0.6864, + "step": 2803 + }, + { + "epoch": 0.1, + "learning_rate": 1.8991467389314094e-07, + "loss": 0.6823, + "step": 2804 + }, + { + "epoch": 0.1, + "learning_rate": 1.899110118284689e-07, + "loss": 0.7023, + "step": 2805 + }, + { + "epoch": 0.1, + "learning_rate": 1.899073497637968e-07, + "loss": 0.6921, + "step": 2806 + }, + { + "epoch": 0.1, + "learning_rate": 1.8990368769912475e-07, + "loss": 0.6927, + "step": 2807 + }, + { + "epoch": 0.1, + "learning_rate": 1.8990002563445268e-07, + "loss": 0.6905, + "step": 2808 + }, + { + "epoch": 0.1, + "learning_rate": 1.8989636356978063e-07, + "loss": 0.7052, + "step": 2809 + }, + { + "epoch": 0.1, + "learning_rate": 1.8989270150510856e-07, + "loss": 0.6813, + "step": 2810 + }, + { + "epoch": 0.1, + "learning_rate": 1.898890394404365e-07, + "loss": 0.7096, + "step": 2811 + }, + { + "epoch": 0.1, + "learning_rate": 1.8988537737576444e-07, + "loss": 0.6809, + "step": 2812 + }, + { + "epoch": 0.1, + "learning_rate": 1.898817153110924e-07, + "loss": 0.7077, + "step": 2813 + }, + { + "epoch": 0.1, + "learning_rate": 1.8987805324642032e-07, + "loss": 0.6818, + "step": 2814 + }, + { + "epoch": 0.1, + "learning_rate": 1.8987439118174827e-07, + "loss": 0.7058, + "step": 2815 + }, + { + "epoch": 0.1, + "learning_rate": 1.898707291170762e-07, + "loss": 0.7049, + "step": 2816 + }, + { + "epoch": 0.1, + "learning_rate": 1.8986706705240415e-07, + "loss": 0.6828, + "step": 2817 + }, + { + "epoch": 0.1, + "learning_rate": 1.8986340498773208e-07, + "loss": 0.7183, + "step": 2818 + }, + { + "epoch": 0.1, + "learning_rate": 1.8985974292306003e-07, + "loss": 0.7341, + "step": 2819 + }, + { + "epoch": 0.1, + "learning_rate": 1.8985608085838796e-07, + "loss": 0.7012, + "step": 2820 + }, + { + "epoch": 0.1, + "learning_rate": 1.898524187937159e-07, + "loss": 0.7015, + "step": 2821 + }, + { + "epoch": 0.1, + "learning_rate": 1.8984875672904384e-07, + "loss": 0.6855, + "step": 2822 + }, + { + "epoch": 0.1, + "learning_rate": 1.8984509466437176e-07, + "loss": 0.6841, + "step": 2823 + }, + { + "epoch": 0.1, + "learning_rate": 1.898414325996997e-07, + "loss": 0.7076, + "step": 2824 + }, + { + "epoch": 0.1, + "learning_rate": 1.8983777053502764e-07, + "loss": 0.6886, + "step": 2825 + }, + { + "epoch": 0.1, + "learning_rate": 1.8983410847035557e-07, + "loss": 0.6888, + "step": 2826 + }, + { + "epoch": 0.1, + "learning_rate": 1.8983044640568352e-07, + "loss": 0.6895, + "step": 2827 + }, + { + "epoch": 0.1, + "learning_rate": 1.8982678434101145e-07, + "loss": 0.6883, + "step": 2828 + }, + { + "epoch": 0.1, + "learning_rate": 1.8982312227633937e-07, + "loss": 0.6959, + "step": 2829 + }, + { + "epoch": 0.1, + "learning_rate": 1.8981946021166733e-07, + "loss": 0.696, + "step": 2830 + }, + { + "epoch": 0.1, + "learning_rate": 1.8981579814699525e-07, + "loss": 0.6826, + "step": 2831 + }, + { + "epoch": 0.1, + "learning_rate": 1.898121360823232e-07, + "loss": 0.6938, + "step": 2832 + }, + { + "epoch": 0.1, + "learning_rate": 1.8980847401765113e-07, + "loss": 0.7236, + "step": 2833 + }, + { + "epoch": 0.1, + "learning_rate": 1.898048119529791e-07, + "loss": 0.6912, + "step": 2834 + }, + { + "epoch": 0.1, + "learning_rate": 1.8980114988830701e-07, + "loss": 0.6761, + "step": 2835 + }, + { + "epoch": 0.1, + "learning_rate": 1.8979748782363497e-07, + "loss": 0.6922, + "step": 2836 + }, + { + "epoch": 0.1, + "learning_rate": 1.897938257589629e-07, + "loss": 0.7015, + "step": 2837 + }, + { + "epoch": 0.1, + "learning_rate": 1.8979016369429085e-07, + "loss": 0.7062, + "step": 2838 + }, + { + "epoch": 0.1, + "learning_rate": 1.8978650162961877e-07, + "loss": 0.6885, + "step": 2839 + }, + { + "epoch": 0.1, + "learning_rate": 1.8978283956494673e-07, + "loss": 0.7079, + "step": 2840 + }, + { + "epoch": 0.1, + "learning_rate": 1.8977917750027465e-07, + "loss": 0.6804, + "step": 2841 + }, + { + "epoch": 0.1, + "learning_rate": 1.8977551543560258e-07, + "loss": 0.6983, + "step": 2842 + }, + { + "epoch": 0.1, + "learning_rate": 1.897718533709305e-07, + "loss": 0.7006, + "step": 2843 + }, + { + "epoch": 0.1, + "learning_rate": 1.8976819130625846e-07, + "loss": 0.6918, + "step": 2844 + }, + { + "epoch": 0.1, + "learning_rate": 1.897645292415864e-07, + "loss": 0.6718, + "step": 2845 + }, + { + "epoch": 0.1, + "learning_rate": 1.8976086717691434e-07, + "loss": 0.6871, + "step": 2846 + }, + { + "epoch": 0.1, + "learning_rate": 1.8975720511224227e-07, + "loss": 0.6868, + "step": 2847 + }, + { + "epoch": 0.1, + "learning_rate": 1.8975354304757022e-07, + "loss": 0.7121, + "step": 2848 + }, + { + "epoch": 0.1, + "learning_rate": 1.8974988098289815e-07, + "loss": 0.6958, + "step": 2849 + }, + { + "epoch": 0.1, + "learning_rate": 1.8974621891822607e-07, + "loss": 0.6661, + "step": 2850 + }, + { + "epoch": 0.1, + "learning_rate": 1.8974255685355403e-07, + "loss": 0.6962, + "step": 2851 + }, + { + "epoch": 0.1, + "learning_rate": 1.8973889478888195e-07, + "loss": 0.6787, + "step": 2852 + }, + { + "epoch": 0.1, + "learning_rate": 1.897352327242099e-07, + "loss": 0.7022, + "step": 2853 + }, + { + "epoch": 0.1, + "learning_rate": 1.8973157065953783e-07, + "loss": 0.6951, + "step": 2854 + }, + { + "epoch": 0.1, + "learning_rate": 1.8972790859486579e-07, + "loss": 0.7213, + "step": 2855 + }, + { + "epoch": 0.1, + "learning_rate": 1.897242465301937e-07, + "loss": 0.6995, + "step": 2856 + }, + { + "epoch": 0.1, + "learning_rate": 1.8972058446552167e-07, + "loss": 0.6951, + "step": 2857 + }, + { + "epoch": 0.1, + "learning_rate": 1.897169224008496e-07, + "loss": 0.7035, + "step": 2858 + }, + { + "epoch": 0.1, + "learning_rate": 1.8971326033617755e-07, + "loss": 0.6759, + "step": 2859 + }, + { + "epoch": 0.1, + "learning_rate": 1.8970959827150545e-07, + "loss": 0.6994, + "step": 2860 + }, + { + "epoch": 0.1, + "learning_rate": 1.897059362068334e-07, + "loss": 0.7075, + "step": 2861 + }, + { + "epoch": 0.1, + "learning_rate": 1.8970227414216133e-07, + "loss": 0.6766, + "step": 2862 + }, + { + "epoch": 0.1, + "learning_rate": 1.8969861207748928e-07, + "loss": 0.686, + "step": 2863 + }, + { + "epoch": 0.1, + "learning_rate": 1.896949500128172e-07, + "loss": 0.6943, + "step": 2864 + }, + { + "epoch": 0.1, + "learning_rate": 1.8969128794814516e-07, + "loss": 0.6802, + "step": 2865 + }, + { + "epoch": 0.1, + "learning_rate": 1.8968762588347309e-07, + "loss": 0.6863, + "step": 2866 + }, + { + "epoch": 0.1, + "learning_rate": 1.8968396381880104e-07, + "loss": 0.6892, + "step": 2867 + }, + { + "epoch": 0.1, + "learning_rate": 1.8968030175412897e-07, + "loss": 0.6814, + "step": 2868 + }, + { + "epoch": 0.1, + "learning_rate": 1.8967663968945692e-07, + "loss": 0.696, + "step": 2869 + }, + { + "epoch": 0.11, + "learning_rate": 1.8967297762478485e-07, + "loss": 0.6929, + "step": 2870 + }, + { + "epoch": 0.11, + "learning_rate": 1.896693155601128e-07, + "loss": 0.6881, + "step": 2871 + }, + { + "epoch": 0.11, + "learning_rate": 1.8966565349544073e-07, + "loss": 0.6736, + "step": 2872 + }, + { + "epoch": 0.11, + "learning_rate": 1.8966199143076865e-07, + "loss": 0.6982, + "step": 2873 + }, + { + "epoch": 0.11, + "learning_rate": 1.896583293660966e-07, + "loss": 0.7052, + "step": 2874 + }, + { + "epoch": 0.11, + "learning_rate": 1.8965466730142453e-07, + "loss": 0.707, + "step": 2875 + }, + { + "epoch": 0.11, + "learning_rate": 1.8965100523675249e-07, + "loss": 0.6791, + "step": 2876 + }, + { + "epoch": 0.11, + "learning_rate": 1.896473431720804e-07, + "loss": 0.7096, + "step": 2877 + }, + { + "epoch": 0.11, + "learning_rate": 1.8964368110740834e-07, + "loss": 0.6867, + "step": 2878 + }, + { + "epoch": 0.11, + "learning_rate": 1.8964001904273627e-07, + "loss": 0.7012, + "step": 2879 + }, + { + "epoch": 0.11, + "learning_rate": 1.8963635697806422e-07, + "loss": 0.6721, + "step": 2880 + }, + { + "epoch": 0.11, + "learning_rate": 1.8963269491339215e-07, + "loss": 0.71, + "step": 2881 + }, + { + "epoch": 0.11, + "learning_rate": 1.896290328487201e-07, + "loss": 0.6997, + "step": 2882 + }, + { + "epoch": 0.11, + "learning_rate": 1.8962537078404803e-07, + "loss": 0.6981, + "step": 2883 + }, + { + "epoch": 0.11, + "learning_rate": 1.8962170871937598e-07, + "loss": 0.6977, + "step": 2884 + }, + { + "epoch": 0.11, + "learning_rate": 1.896180466547039e-07, + "loss": 0.7012, + "step": 2885 + }, + { + "epoch": 0.11, + "learning_rate": 1.8961438459003186e-07, + "loss": 0.6845, + "step": 2886 + }, + { + "epoch": 0.11, + "learning_rate": 1.8961072252535978e-07, + "loss": 0.6874, + "step": 2887 + }, + { + "epoch": 0.11, + "learning_rate": 1.8960706046068774e-07, + "loss": 0.6737, + "step": 2888 + }, + { + "epoch": 0.11, + "learning_rate": 1.8960339839601566e-07, + "loss": 0.6818, + "step": 2889 + }, + { + "epoch": 0.11, + "learning_rate": 1.8959973633134362e-07, + "loss": 0.6929, + "step": 2890 + }, + { + "epoch": 0.11, + "learning_rate": 1.8959607426667154e-07, + "loss": 0.709, + "step": 2891 + }, + { + "epoch": 0.11, + "learning_rate": 1.895924122019995e-07, + "loss": 0.6945, + "step": 2892 + }, + { + "epoch": 0.11, + "learning_rate": 1.8958875013732742e-07, + "loss": 0.6714, + "step": 2893 + }, + { + "epoch": 0.11, + "learning_rate": 1.8958508807265535e-07, + "loss": 0.7038, + "step": 2894 + }, + { + "epoch": 0.11, + "learning_rate": 1.895814260079833e-07, + "loss": 0.6849, + "step": 2895 + }, + { + "epoch": 0.11, + "learning_rate": 1.8957776394331123e-07, + "loss": 0.6799, + "step": 2896 + }, + { + "epoch": 0.11, + "learning_rate": 1.8957410187863916e-07, + "loss": 0.6905, + "step": 2897 + }, + { + "epoch": 0.11, + "learning_rate": 1.895704398139671e-07, + "loss": 0.6902, + "step": 2898 + }, + { + "epoch": 0.11, + "learning_rate": 1.8956677774929504e-07, + "loss": 0.7032, + "step": 2899 + }, + { + "epoch": 0.11, + "learning_rate": 1.8956311568462296e-07, + "loss": 0.7159, + "step": 2900 + }, + { + "epoch": 0.11, + "learning_rate": 1.8955945361995092e-07, + "loss": 0.685, + "step": 2901 + }, + { + "epoch": 0.11, + "learning_rate": 1.8955579155527884e-07, + "loss": 0.6715, + "step": 2902 + }, + { + "epoch": 0.11, + "learning_rate": 1.895521294906068e-07, + "loss": 0.6737, + "step": 2903 + }, + { + "epoch": 0.11, + "learning_rate": 1.8954846742593472e-07, + "loss": 0.7047, + "step": 2904 + }, + { + "epoch": 0.11, + "learning_rate": 1.8954480536126268e-07, + "loss": 0.6886, + "step": 2905 + }, + { + "epoch": 0.11, + "learning_rate": 1.895411432965906e-07, + "loss": 0.6906, + "step": 2906 + }, + { + "epoch": 0.11, + "learning_rate": 1.8953748123191856e-07, + "loss": 0.6779, + "step": 2907 + }, + { + "epoch": 0.11, + "learning_rate": 1.8953381916724648e-07, + "loss": 0.6996, + "step": 2908 + }, + { + "epoch": 0.11, + "learning_rate": 1.8953015710257444e-07, + "loss": 0.6943, + "step": 2909 + }, + { + "epoch": 0.11, + "learning_rate": 1.8952649503790236e-07, + "loss": 0.6986, + "step": 2910 + }, + { + "epoch": 0.11, + "learning_rate": 1.8952283297323032e-07, + "loss": 0.6791, + "step": 2911 + }, + { + "epoch": 0.11, + "learning_rate": 1.8951917090855824e-07, + "loss": 0.7018, + "step": 2912 + }, + { + "epoch": 0.11, + "learning_rate": 1.895155088438862e-07, + "loss": 0.7088, + "step": 2913 + }, + { + "epoch": 0.11, + "learning_rate": 1.895118467792141e-07, + "loss": 0.6813, + "step": 2914 + }, + { + "epoch": 0.11, + "learning_rate": 1.8950818471454205e-07, + "loss": 0.6778, + "step": 2915 + }, + { + "epoch": 0.11, + "learning_rate": 1.8950452264986998e-07, + "loss": 0.6937, + "step": 2916 + }, + { + "epoch": 0.11, + "learning_rate": 1.8950086058519793e-07, + "loss": 0.6924, + "step": 2917 + }, + { + "epoch": 0.11, + "learning_rate": 1.8949719852052586e-07, + "loss": 0.6882, + "step": 2918 + }, + { + "epoch": 0.11, + "learning_rate": 1.894935364558538e-07, + "loss": 0.6823, + "step": 2919 + }, + { + "epoch": 0.11, + "learning_rate": 1.8948987439118174e-07, + "loss": 0.6958, + "step": 2920 + }, + { + "epoch": 0.11, + "learning_rate": 1.8948621232650966e-07, + "loss": 0.7103, + "step": 2921 + }, + { + "epoch": 0.11, + "learning_rate": 1.8948255026183762e-07, + "loss": 0.6824, + "step": 2922 + }, + { + "epoch": 0.11, + "learning_rate": 1.8947888819716554e-07, + "loss": 0.6973, + "step": 2923 + }, + { + "epoch": 0.11, + "learning_rate": 1.894752261324935e-07, + "loss": 0.6933, + "step": 2924 + }, + { + "epoch": 0.11, + "learning_rate": 1.8947156406782142e-07, + "loss": 0.6996, + "step": 2925 + }, + { + "epoch": 0.11, + "learning_rate": 1.8946790200314938e-07, + "loss": 0.6989, + "step": 2926 + }, + { + "epoch": 0.11, + "learning_rate": 1.894642399384773e-07, + "loss": 0.6809, + "step": 2927 + }, + { + "epoch": 0.11, + "learning_rate": 1.8946057787380526e-07, + "loss": 0.6923, + "step": 2928 + }, + { + "epoch": 0.11, + "learning_rate": 1.8945691580913318e-07, + "loss": 0.7026, + "step": 2929 + }, + { + "epoch": 0.11, + "learning_rate": 1.8945325374446114e-07, + "loss": 0.677, + "step": 2930 + }, + { + "epoch": 0.11, + "learning_rate": 1.8944959167978906e-07, + "loss": 0.7021, + "step": 2931 + }, + { + "epoch": 0.11, + "learning_rate": 1.89445929615117e-07, + "loss": 0.719, + "step": 2932 + }, + { + "epoch": 0.11, + "learning_rate": 1.8944226755044492e-07, + "loss": 0.6925, + "step": 2933 + }, + { + "epoch": 0.11, + "learning_rate": 1.8943860548577287e-07, + "loss": 0.7032, + "step": 2934 + }, + { + "epoch": 0.11, + "learning_rate": 1.894349434211008e-07, + "loss": 0.6837, + "step": 2935 + }, + { + "epoch": 0.11, + "learning_rate": 1.8943128135642875e-07, + "loss": 0.6794, + "step": 2936 + }, + { + "epoch": 0.11, + "learning_rate": 1.8942761929175668e-07, + "loss": 0.6851, + "step": 2937 + }, + { + "epoch": 0.11, + "learning_rate": 1.8942395722708463e-07, + "loss": 0.6923, + "step": 2938 + }, + { + "epoch": 0.11, + "learning_rate": 1.8942029516241256e-07, + "loss": 0.7041, + "step": 2939 + }, + { + "epoch": 0.11, + "learning_rate": 1.894166330977405e-07, + "loss": 0.6884, + "step": 2940 + }, + { + "epoch": 0.11, + "learning_rate": 1.8941297103306844e-07, + "loss": 0.6975, + "step": 2941 + }, + { + "epoch": 0.11, + "learning_rate": 1.894093089683964e-07, + "loss": 0.6937, + "step": 2942 + }, + { + "epoch": 0.11, + "learning_rate": 1.8940564690372431e-07, + "loss": 0.6807, + "step": 2943 + }, + { + "epoch": 0.11, + "learning_rate": 1.8940198483905224e-07, + "loss": 0.7121, + "step": 2944 + }, + { + "epoch": 0.11, + "learning_rate": 1.893983227743802e-07, + "loss": 0.7167, + "step": 2945 + }, + { + "epoch": 0.11, + "learning_rate": 1.8939466070970812e-07, + "loss": 0.6904, + "step": 2946 + }, + { + "epoch": 0.11, + "learning_rate": 1.8939099864503607e-07, + "loss": 0.6917, + "step": 2947 + }, + { + "epoch": 0.11, + "learning_rate": 1.89387336580364e-07, + "loss": 0.6786, + "step": 2948 + }, + { + "epoch": 0.11, + "learning_rate": 1.8938367451569195e-07, + "loss": 0.7001, + "step": 2949 + }, + { + "epoch": 0.11, + "learning_rate": 1.8938001245101985e-07, + "loss": 0.7051, + "step": 2950 + }, + { + "epoch": 0.11, + "learning_rate": 1.893763503863478e-07, + "loss": 0.6884, + "step": 2951 + }, + { + "epoch": 0.11, + "learning_rate": 1.8937268832167573e-07, + "loss": 0.7118, + "step": 2952 + }, + { + "epoch": 0.11, + "learning_rate": 1.893690262570037e-07, + "loss": 0.6864, + "step": 2953 + }, + { + "epoch": 0.11, + "learning_rate": 1.8936536419233161e-07, + "loss": 0.6827, + "step": 2954 + }, + { + "epoch": 0.11, + "learning_rate": 1.8936170212765957e-07, + "loss": 0.6769, + "step": 2955 + }, + { + "epoch": 0.11, + "learning_rate": 1.893580400629875e-07, + "loss": 0.6821, + "step": 2956 + }, + { + "epoch": 0.11, + "learning_rate": 1.8935437799831545e-07, + "loss": 0.6806, + "step": 2957 + }, + { + "epoch": 0.11, + "learning_rate": 1.8935071593364337e-07, + "loss": 0.7098, + "step": 2958 + }, + { + "epoch": 0.11, + "learning_rate": 1.8934705386897133e-07, + "loss": 0.682, + "step": 2959 + }, + { + "epoch": 0.11, + "learning_rate": 1.8934339180429925e-07, + "loss": 0.6908, + "step": 2960 + }, + { + "epoch": 0.11, + "learning_rate": 1.893397297396272e-07, + "loss": 0.6942, + "step": 2961 + }, + { + "epoch": 0.11, + "learning_rate": 1.8933606767495513e-07, + "loss": 0.692, + "step": 2962 + }, + { + "epoch": 0.11, + "learning_rate": 1.893324056102831e-07, + "loss": 0.6949, + "step": 2963 + }, + { + "epoch": 0.11, + "learning_rate": 1.8932874354561101e-07, + "loss": 0.704, + "step": 2964 + }, + { + "epoch": 0.11, + "learning_rate": 1.8932508148093894e-07, + "loss": 0.6982, + "step": 2965 + }, + { + "epoch": 0.11, + "learning_rate": 1.893214194162669e-07, + "loss": 0.6831, + "step": 2966 + }, + { + "epoch": 0.11, + "learning_rate": 1.8931775735159482e-07, + "loss": 0.704, + "step": 2967 + }, + { + "epoch": 0.11, + "learning_rate": 1.8931409528692275e-07, + "loss": 0.7012, + "step": 2968 + }, + { + "epoch": 0.11, + "learning_rate": 1.893104332222507e-07, + "loss": 0.7067, + "step": 2969 + }, + { + "epoch": 0.11, + "learning_rate": 1.8930677115757863e-07, + "loss": 0.6983, + "step": 2970 + }, + { + "epoch": 0.11, + "learning_rate": 1.8930310909290655e-07, + "loss": 0.6793, + "step": 2971 + }, + { + "epoch": 0.11, + "learning_rate": 1.892994470282345e-07, + "loss": 0.6805, + "step": 2972 + }, + { + "epoch": 0.11, + "learning_rate": 1.8929578496356243e-07, + "loss": 0.6922, + "step": 2973 + }, + { + "epoch": 0.11, + "learning_rate": 1.8929212289889039e-07, + "loss": 0.6843, + "step": 2974 + }, + { + "epoch": 0.11, + "learning_rate": 1.892884608342183e-07, + "loss": 0.7119, + "step": 2975 + }, + { + "epoch": 0.11, + "learning_rate": 1.8928479876954627e-07, + "loss": 0.7235, + "step": 2976 + }, + { + "epoch": 0.11, + "learning_rate": 1.892811367048742e-07, + "loss": 0.6947, + "step": 2977 + }, + { + "epoch": 0.11, + "learning_rate": 1.8927747464020215e-07, + "loss": 0.7075, + "step": 2978 + }, + { + "epoch": 0.11, + "learning_rate": 1.8927381257553007e-07, + "loss": 0.666, + "step": 2979 + }, + { + "epoch": 0.11, + "learning_rate": 1.8927015051085803e-07, + "loss": 0.6837, + "step": 2980 + }, + { + "epoch": 0.11, + "learning_rate": 1.8926648844618595e-07, + "loss": 0.6893, + "step": 2981 + }, + { + "epoch": 0.11, + "learning_rate": 1.892628263815139e-07, + "loss": 0.7049, + "step": 2982 + }, + { + "epoch": 0.11, + "learning_rate": 1.8925916431684183e-07, + "loss": 0.6886, + "step": 2983 + }, + { + "epoch": 0.11, + "learning_rate": 1.8925550225216979e-07, + "loss": 0.7016, + "step": 2984 + }, + { + "epoch": 0.11, + "learning_rate": 1.892518401874977e-07, + "loss": 0.678, + "step": 2985 + }, + { + "epoch": 0.11, + "learning_rate": 1.8924817812282564e-07, + "loss": 0.6933, + "step": 2986 + }, + { + "epoch": 0.11, + "learning_rate": 1.8924451605815357e-07, + "loss": 0.7009, + "step": 2987 + }, + { + "epoch": 0.11, + "learning_rate": 1.8924085399348152e-07, + "loss": 0.7201, + "step": 2988 + }, + { + "epoch": 0.11, + "learning_rate": 1.8923719192880945e-07, + "loss": 0.6952, + "step": 2989 + }, + { + "epoch": 0.11, + "learning_rate": 1.892335298641374e-07, + "loss": 0.6981, + "step": 2990 + }, + { + "epoch": 0.11, + "learning_rate": 1.8922986779946533e-07, + "loss": 0.6912, + "step": 2991 + }, + { + "epoch": 0.11, + "learning_rate": 1.8922620573479325e-07, + "loss": 0.6901, + "step": 2992 + }, + { + "epoch": 0.11, + "learning_rate": 1.892225436701212e-07, + "loss": 0.668, + "step": 2993 + }, + { + "epoch": 0.11, + "learning_rate": 1.8921888160544913e-07, + "loss": 0.6981, + "step": 2994 + }, + { + "epoch": 0.11, + "learning_rate": 1.8921521954077709e-07, + "loss": 0.704, + "step": 2995 + }, + { + "epoch": 0.11, + "learning_rate": 1.89211557476105e-07, + "loss": 0.7005, + "step": 2996 + }, + { + "epoch": 0.11, + "learning_rate": 1.8920789541143297e-07, + "loss": 0.7018, + "step": 2997 + }, + { + "epoch": 0.11, + "learning_rate": 1.892042333467609e-07, + "loss": 0.6933, + "step": 2998 + }, + { + "epoch": 0.11, + "learning_rate": 1.8920057128208885e-07, + "loss": 0.684, + "step": 2999 + }, + { + "epoch": 0.11, + "learning_rate": 1.8919690921741677e-07, + "loss": 0.6945, + "step": 3000 + }, + { + "epoch": 0.11, + "learning_rate": 1.8919324715274472e-07, + "loss": 0.6815, + "step": 3001 + }, + { + "epoch": 0.11, + "learning_rate": 1.8918958508807265e-07, + "loss": 0.7047, + "step": 3002 + }, + { + "epoch": 0.11, + "learning_rate": 1.891859230234006e-07, + "loss": 0.7131, + "step": 3003 + }, + { + "epoch": 0.11, + "learning_rate": 1.891822609587285e-07, + "loss": 0.6972, + "step": 3004 + }, + { + "epoch": 0.11, + "learning_rate": 1.8917859889405646e-07, + "loss": 0.6968, + "step": 3005 + }, + { + "epoch": 0.11, + "learning_rate": 1.8917493682938438e-07, + "loss": 0.7097, + "step": 3006 + }, + { + "epoch": 0.11, + "learning_rate": 1.8917127476471234e-07, + "loss": 0.7004, + "step": 3007 + }, + { + "epoch": 0.11, + "learning_rate": 1.8916761270004026e-07, + "loss": 0.6933, + "step": 3008 + }, + { + "epoch": 0.11, + "learning_rate": 1.8916395063536822e-07, + "loss": 0.6952, + "step": 3009 + }, + { + "epoch": 0.11, + "learning_rate": 1.8916028857069614e-07, + "loss": 0.7104, + "step": 3010 + }, + { + "epoch": 0.11, + "learning_rate": 1.891566265060241e-07, + "loss": 0.7046, + "step": 3011 + }, + { + "epoch": 0.11, + "learning_rate": 1.8915296444135202e-07, + "loss": 0.6809, + "step": 3012 + }, + { + "epoch": 0.11, + "learning_rate": 1.8914930237667998e-07, + "loss": 0.6978, + "step": 3013 + }, + { + "epoch": 0.11, + "learning_rate": 1.891456403120079e-07, + "loss": 0.677, + "step": 3014 + }, + { + "epoch": 0.11, + "learning_rate": 1.8914197824733583e-07, + "loss": 0.7057, + "step": 3015 + }, + { + "epoch": 0.11, + "learning_rate": 1.8913831618266378e-07, + "loss": 0.6935, + "step": 3016 + }, + { + "epoch": 0.11, + "learning_rate": 1.891346541179917e-07, + "loss": 0.6826, + "step": 3017 + }, + { + "epoch": 0.11, + "learning_rate": 1.8913099205331966e-07, + "loss": 0.7189, + "step": 3018 + }, + { + "epoch": 0.11, + "learning_rate": 1.891273299886476e-07, + "loss": 0.6995, + "step": 3019 + }, + { + "epoch": 0.11, + "learning_rate": 1.8912366792397554e-07, + "loss": 0.7049, + "step": 3020 + }, + { + "epoch": 0.11, + "learning_rate": 1.8912000585930347e-07, + "loss": 0.6924, + "step": 3021 + }, + { + "epoch": 0.11, + "learning_rate": 1.891163437946314e-07, + "loss": 0.6895, + "step": 3022 + }, + { + "epoch": 0.11, + "learning_rate": 1.8911268172995932e-07, + "loss": 0.7011, + "step": 3023 + }, + { + "epoch": 0.11, + "learning_rate": 1.8910901966528728e-07, + "loss": 0.6978, + "step": 3024 + }, + { + "epoch": 0.11, + "learning_rate": 1.891053576006152e-07, + "loss": 0.6935, + "step": 3025 + }, + { + "epoch": 0.11, + "learning_rate": 1.8910169553594316e-07, + "loss": 0.6876, + "step": 3026 + }, + { + "epoch": 0.11, + "learning_rate": 1.8909803347127108e-07, + "loss": 0.6858, + "step": 3027 + }, + { + "epoch": 0.11, + "learning_rate": 1.8909437140659904e-07, + "loss": 0.6975, + "step": 3028 + }, + { + "epoch": 0.11, + "learning_rate": 1.8909070934192696e-07, + "loss": 0.6979, + "step": 3029 + }, + { + "epoch": 0.11, + "learning_rate": 1.8908704727725492e-07, + "loss": 0.6862, + "step": 3030 + }, + { + "epoch": 0.11, + "learning_rate": 1.8908338521258284e-07, + "loss": 0.6758, + "step": 3031 + }, + { + "epoch": 0.11, + "learning_rate": 1.890797231479108e-07, + "loss": 0.7064, + "step": 3032 + }, + { + "epoch": 0.11, + "learning_rate": 1.8907606108323872e-07, + "loss": 0.7035, + "step": 3033 + }, + { + "epoch": 0.11, + "learning_rate": 1.8907239901856668e-07, + "loss": 0.6954, + "step": 3034 + }, + { + "epoch": 0.11, + "learning_rate": 1.890687369538946e-07, + "loss": 0.6904, + "step": 3035 + }, + { + "epoch": 0.11, + "learning_rate": 1.8906507488922253e-07, + "loss": 0.6862, + "step": 3036 + }, + { + "epoch": 0.11, + "learning_rate": 1.8906141282455048e-07, + "loss": 0.6894, + "step": 3037 + }, + { + "epoch": 0.11, + "learning_rate": 1.890577507598784e-07, + "loss": 0.6775, + "step": 3038 + }, + { + "epoch": 0.11, + "learning_rate": 1.8905408869520636e-07, + "loss": 0.6859, + "step": 3039 + }, + { + "epoch": 0.11, + "learning_rate": 1.890504266305343e-07, + "loss": 0.6783, + "step": 3040 + }, + { + "epoch": 0.11, + "learning_rate": 1.8904676456586222e-07, + "loss": 0.6957, + "step": 3041 + }, + { + "epoch": 0.11, + "learning_rate": 1.8904310250119014e-07, + "loss": 0.678, + "step": 3042 + }, + { + "epoch": 0.11, + "learning_rate": 1.890394404365181e-07, + "loss": 0.7054, + "step": 3043 + }, + { + "epoch": 0.11, + "learning_rate": 1.8903577837184602e-07, + "loss": 0.7037, + "step": 3044 + }, + { + "epoch": 0.11, + "learning_rate": 1.8903211630717398e-07, + "loss": 0.7125, + "step": 3045 + }, + { + "epoch": 0.11, + "learning_rate": 1.890284542425019e-07, + "loss": 0.6951, + "step": 3046 + }, + { + "epoch": 0.11, + "learning_rate": 1.8902479217782986e-07, + "loss": 0.7007, + "step": 3047 + }, + { + "epoch": 0.11, + "learning_rate": 1.8902113011315778e-07, + "loss": 0.6894, + "step": 3048 + }, + { + "epoch": 0.11, + "learning_rate": 1.8901746804848574e-07, + "loss": 0.7021, + "step": 3049 + }, + { + "epoch": 0.11, + "learning_rate": 1.8901380598381366e-07, + "loss": 0.6962, + "step": 3050 + }, + { + "epoch": 0.11, + "learning_rate": 1.8901014391914162e-07, + "loss": 0.696, + "step": 3051 + }, + { + "epoch": 0.11, + "learning_rate": 1.8900648185446954e-07, + "loss": 0.7049, + "step": 3052 + }, + { + "epoch": 0.11, + "learning_rate": 1.890028197897975e-07, + "loss": 0.7114, + "step": 3053 + }, + { + "epoch": 0.11, + "learning_rate": 1.8899915772512542e-07, + "loss": 0.6983, + "step": 3054 + }, + { + "epoch": 0.11, + "learning_rate": 1.8899549566045338e-07, + "loss": 0.6745, + "step": 3055 + }, + { + "epoch": 0.11, + "learning_rate": 1.889918335957813e-07, + "loss": 0.6864, + "step": 3056 + }, + { + "epoch": 0.11, + "learning_rate": 1.8898817153110923e-07, + "loss": 0.6962, + "step": 3057 + }, + { + "epoch": 0.11, + "learning_rate": 1.8898450946643716e-07, + "loss": 0.6891, + "step": 3058 + }, + { + "epoch": 0.11, + "learning_rate": 1.889808474017651e-07, + "loss": 0.6905, + "step": 3059 + }, + { + "epoch": 0.11, + "learning_rate": 1.8897718533709304e-07, + "loss": 0.7088, + "step": 3060 + }, + { + "epoch": 0.11, + "learning_rate": 1.88973523272421e-07, + "loss": 0.6815, + "step": 3061 + }, + { + "epoch": 0.11, + "learning_rate": 1.8896986120774891e-07, + "loss": 0.6968, + "step": 3062 + }, + { + "epoch": 0.11, + "learning_rate": 1.8896619914307684e-07, + "loss": 0.7014, + "step": 3063 + }, + { + "epoch": 0.11, + "learning_rate": 1.889625370784048e-07, + "loss": 0.6803, + "step": 3064 + }, + { + "epoch": 0.11, + "learning_rate": 1.8895887501373272e-07, + "loss": 0.7074, + "step": 3065 + }, + { + "epoch": 0.11, + "learning_rate": 1.8895521294906067e-07, + "loss": 0.6992, + "step": 3066 + }, + { + "epoch": 0.11, + "learning_rate": 1.889515508843886e-07, + "loss": 0.6807, + "step": 3067 + }, + { + "epoch": 0.11, + "learning_rate": 1.8894788881971655e-07, + "loss": 0.7017, + "step": 3068 + }, + { + "epoch": 0.11, + "learning_rate": 1.8894422675504448e-07, + "loss": 0.7092, + "step": 3069 + }, + { + "epoch": 0.11, + "learning_rate": 1.8894056469037243e-07, + "loss": 0.6844, + "step": 3070 + }, + { + "epoch": 0.11, + "learning_rate": 1.8893690262570036e-07, + "loss": 0.7036, + "step": 3071 + }, + { + "epoch": 0.11, + "learning_rate": 1.8893324056102831e-07, + "loss": 0.6893, + "step": 3072 + }, + { + "epoch": 0.11, + "learning_rate": 1.8892957849635624e-07, + "loss": 0.7112, + "step": 3073 + }, + { + "epoch": 0.11, + "learning_rate": 1.889259164316842e-07, + "loss": 0.7017, + "step": 3074 + }, + { + "epoch": 0.11, + "learning_rate": 1.8892225436701212e-07, + "loss": 0.7014, + "step": 3075 + }, + { + "epoch": 0.11, + "learning_rate": 1.8891859230234005e-07, + "loss": 0.6851, + "step": 3076 + }, + { + "epoch": 0.11, + "learning_rate": 1.8891493023766797e-07, + "loss": 0.7043, + "step": 3077 + }, + { + "epoch": 0.11, + "learning_rate": 1.8891126817299593e-07, + "loss": 0.6986, + "step": 3078 + }, + { + "epoch": 0.11, + "learning_rate": 1.8890760610832385e-07, + "loss": 0.6983, + "step": 3079 + }, + { + "epoch": 0.11, + "learning_rate": 1.889039440436518e-07, + "loss": 0.6956, + "step": 3080 + }, + { + "epoch": 0.11, + "learning_rate": 1.8890028197897973e-07, + "loss": 0.7119, + "step": 3081 + }, + { + "epoch": 0.11, + "learning_rate": 1.888966199143077e-07, + "loss": 0.7137, + "step": 3082 + }, + { + "epoch": 0.11, + "learning_rate": 1.8889295784963561e-07, + "loss": 0.6769, + "step": 3083 + }, + { + "epoch": 0.11, + "learning_rate": 1.8888929578496357e-07, + "loss": 0.6764, + "step": 3084 + }, + { + "epoch": 0.11, + "learning_rate": 1.888856337202915e-07, + "loss": 0.6847, + "step": 3085 + }, + { + "epoch": 0.11, + "learning_rate": 1.8888197165561942e-07, + "loss": 0.6999, + "step": 3086 + }, + { + "epoch": 0.11, + "learning_rate": 1.8887830959094737e-07, + "loss": 0.7134, + "step": 3087 + }, + { + "epoch": 0.11, + "learning_rate": 1.888746475262753e-07, + "loss": 0.7265, + "step": 3088 + }, + { + "epoch": 0.11, + "learning_rate": 1.8887098546160325e-07, + "loss": 0.7081, + "step": 3089 + }, + { + "epoch": 0.11, + "learning_rate": 1.8886732339693118e-07, + "loss": 0.6853, + "step": 3090 + }, + { + "epoch": 0.11, + "learning_rate": 1.8886366133225913e-07, + "loss": 0.693, + "step": 3091 + }, + { + "epoch": 0.11, + "learning_rate": 1.8885999926758706e-07, + "loss": 0.6829, + "step": 3092 + }, + { + "epoch": 0.11, + "learning_rate": 1.88856337202915e-07, + "loss": 0.6872, + "step": 3093 + }, + { + "epoch": 0.11, + "learning_rate": 1.8885267513824291e-07, + "loss": 0.7132, + "step": 3094 + }, + { + "epoch": 0.11, + "learning_rate": 1.8884901307357087e-07, + "loss": 0.7083, + "step": 3095 + }, + { + "epoch": 0.11, + "learning_rate": 1.888453510088988e-07, + "loss": 0.6533, + "step": 3096 + }, + { + "epoch": 0.11, + "learning_rate": 1.8884168894422675e-07, + "loss": 0.6669, + "step": 3097 + }, + { + "epoch": 0.11, + "learning_rate": 1.8883802687955467e-07, + "loss": 0.6901, + "step": 3098 + }, + { + "epoch": 0.11, + "learning_rate": 1.8883436481488263e-07, + "loss": 0.6819, + "step": 3099 + }, + { + "epoch": 0.11, + "learning_rate": 1.8883070275021055e-07, + "loss": 0.7046, + "step": 3100 + }, + { + "epoch": 0.11, + "learning_rate": 1.888270406855385e-07, + "loss": 0.6912, + "step": 3101 + }, + { + "epoch": 0.11, + "learning_rate": 1.8882337862086643e-07, + "loss": 0.7301, + "step": 3102 + }, + { + "epoch": 0.11, + "learning_rate": 1.8881971655619439e-07, + "loss": 0.6879, + "step": 3103 + }, + { + "epoch": 0.11, + "learning_rate": 1.888160544915223e-07, + "loss": 0.6755, + "step": 3104 + }, + { + "epoch": 0.11, + "learning_rate": 1.8881239242685027e-07, + "loss": 0.6724, + "step": 3105 + }, + { + "epoch": 0.11, + "learning_rate": 1.888087303621782e-07, + "loss": 0.6936, + "step": 3106 + }, + { + "epoch": 0.11, + "learning_rate": 1.8880506829750612e-07, + "loss": 0.707, + "step": 3107 + }, + { + "epoch": 0.11, + "learning_rate": 1.8880140623283407e-07, + "loss": 0.7008, + "step": 3108 + }, + { + "epoch": 0.11, + "learning_rate": 1.88797744168162e-07, + "loss": 0.7243, + "step": 3109 + }, + { + "epoch": 0.11, + "learning_rate": 1.8879408210348995e-07, + "loss": 0.7062, + "step": 3110 + }, + { + "epoch": 0.11, + "learning_rate": 1.8879042003881788e-07, + "loss": 0.6722, + "step": 3111 + }, + { + "epoch": 0.11, + "learning_rate": 1.887867579741458e-07, + "loss": 0.6889, + "step": 3112 + }, + { + "epoch": 0.11, + "learning_rate": 1.8878309590947373e-07, + "loss": 0.7068, + "step": 3113 + }, + { + "epoch": 0.11, + "learning_rate": 1.8877943384480169e-07, + "loss": 0.7268, + "step": 3114 + }, + { + "epoch": 0.11, + "learning_rate": 1.887757717801296e-07, + "loss": 0.6882, + "step": 3115 + }, + { + "epoch": 0.11, + "learning_rate": 1.8877210971545757e-07, + "loss": 0.7113, + "step": 3116 + }, + { + "epoch": 0.11, + "learning_rate": 1.887684476507855e-07, + "loss": 0.687, + "step": 3117 + }, + { + "epoch": 0.11, + "learning_rate": 1.8876478558611345e-07, + "loss": 0.691, + "step": 3118 + }, + { + "epoch": 0.11, + "learning_rate": 1.8876112352144137e-07, + "loss": 0.6982, + "step": 3119 + }, + { + "epoch": 0.11, + "learning_rate": 1.8875746145676932e-07, + "loss": 0.6715, + "step": 3120 + }, + { + "epoch": 0.11, + "learning_rate": 1.8875379939209725e-07, + "loss": 0.6782, + "step": 3121 + }, + { + "epoch": 0.11, + "learning_rate": 1.887501373274252e-07, + "loss": 0.7375, + "step": 3122 + }, + { + "epoch": 0.11, + "learning_rate": 1.8874647526275313e-07, + "loss": 0.6903, + "step": 3123 + }, + { + "epoch": 0.11, + "learning_rate": 1.8874281319808108e-07, + "loss": 0.6714, + "step": 3124 + }, + { + "epoch": 0.11, + "learning_rate": 1.88739151133409e-07, + "loss": 0.6735, + "step": 3125 + }, + { + "epoch": 0.11, + "learning_rate": 1.8873548906873696e-07, + "loss": 0.7288, + "step": 3126 + }, + { + "epoch": 0.11, + "learning_rate": 1.887318270040649e-07, + "loss": 0.6841, + "step": 3127 + }, + { + "epoch": 0.11, + "learning_rate": 1.8872816493939282e-07, + "loss": 0.6986, + "step": 3128 + }, + { + "epoch": 0.11, + "learning_rate": 1.8872450287472074e-07, + "loss": 0.701, + "step": 3129 + }, + { + "epoch": 0.11, + "learning_rate": 1.887208408100487e-07, + "loss": 0.7159, + "step": 3130 + }, + { + "epoch": 0.11, + "learning_rate": 1.8871717874537662e-07, + "loss": 0.6965, + "step": 3131 + }, + { + "epoch": 0.11, + "learning_rate": 1.8871351668070458e-07, + "loss": 0.6984, + "step": 3132 + }, + { + "epoch": 0.11, + "learning_rate": 1.887098546160325e-07, + "loss": 0.7002, + "step": 3133 + }, + { + "epoch": 0.11, + "learning_rate": 1.8870619255136043e-07, + "loss": 0.7062, + "step": 3134 + }, + { + "epoch": 0.11, + "learning_rate": 1.8870253048668838e-07, + "loss": 0.7121, + "step": 3135 + }, + { + "epoch": 0.11, + "learning_rate": 1.886988684220163e-07, + "loss": 0.7214, + "step": 3136 + }, + { + "epoch": 0.11, + "learning_rate": 1.8869520635734426e-07, + "loss": 0.6983, + "step": 3137 + }, + { + "epoch": 0.11, + "learning_rate": 1.886915442926722e-07, + "loss": 0.6773, + "step": 3138 + }, + { + "epoch": 0.11, + "learning_rate": 1.8868788222800014e-07, + "loss": 0.6928, + "step": 3139 + }, + { + "epoch": 0.11, + "learning_rate": 1.8868422016332807e-07, + "loss": 0.7046, + "step": 3140 + }, + { + "epoch": 0.11, + "learning_rate": 1.8868055809865602e-07, + "loss": 0.6734, + "step": 3141 + }, + { + "epoch": 0.11, + "learning_rate": 1.8867689603398395e-07, + "loss": 0.6786, + "step": 3142 + }, + { + "epoch": 0.11, + "learning_rate": 1.886732339693119e-07, + "loss": 0.6964, + "step": 3143 + }, + { + "epoch": 0.12, + "learning_rate": 1.8866957190463983e-07, + "loss": 0.7085, + "step": 3144 + }, + { + "epoch": 0.12, + "learning_rate": 1.8866590983996778e-07, + "loss": 0.7112, + "step": 3145 + }, + { + "epoch": 0.12, + "learning_rate": 1.886622477752957e-07, + "loss": 0.6959, + "step": 3146 + }, + { + "epoch": 0.12, + "learning_rate": 1.8865858571062364e-07, + "loss": 0.6996, + "step": 3147 + }, + { + "epoch": 0.12, + "learning_rate": 1.8865492364595156e-07, + "loss": 0.6845, + "step": 3148 + }, + { + "epoch": 0.12, + "learning_rate": 1.8865126158127952e-07, + "loss": 0.6798, + "step": 3149 + }, + { + "epoch": 0.12, + "learning_rate": 1.8864759951660744e-07, + "loss": 0.6947, + "step": 3150 + }, + { + "epoch": 0.12, + "learning_rate": 1.886439374519354e-07, + "loss": 0.7388, + "step": 3151 + }, + { + "epoch": 0.12, + "learning_rate": 1.8864027538726332e-07, + "loss": 0.7039, + "step": 3152 + }, + { + "epoch": 0.12, + "learning_rate": 1.8863661332259128e-07, + "loss": 0.7004, + "step": 3153 + }, + { + "epoch": 0.12, + "learning_rate": 1.886329512579192e-07, + "loss": 0.6923, + "step": 3154 + }, + { + "epoch": 0.12, + "learning_rate": 1.8862928919324716e-07, + "loss": 0.6747, + "step": 3155 + }, + { + "epoch": 0.12, + "learning_rate": 1.8862562712857508e-07, + "loss": 0.7001, + "step": 3156 + }, + { + "epoch": 0.12, + "learning_rate": 1.88621965063903e-07, + "loss": 0.6789, + "step": 3157 + }, + { + "epoch": 0.12, + "learning_rate": 1.8861830299923096e-07, + "loss": 0.6968, + "step": 3158 + }, + { + "epoch": 0.12, + "learning_rate": 1.886146409345589e-07, + "loss": 0.6986, + "step": 3159 + }, + { + "epoch": 0.12, + "learning_rate": 1.8861097886988684e-07, + "loss": 0.6808, + "step": 3160 + }, + { + "epoch": 0.12, + "learning_rate": 1.8860731680521477e-07, + "loss": 0.6925, + "step": 3161 + }, + { + "epoch": 0.12, + "learning_rate": 1.8860365474054272e-07, + "loss": 0.6928, + "step": 3162 + }, + { + "epoch": 0.12, + "learning_rate": 1.8859999267587065e-07, + "loss": 0.707, + "step": 3163 + }, + { + "epoch": 0.12, + "learning_rate": 1.885963306111986e-07, + "loss": 0.7046, + "step": 3164 + }, + { + "epoch": 0.12, + "learning_rate": 1.885926685465265e-07, + "loss": 0.677, + "step": 3165 + }, + { + "epoch": 0.12, + "learning_rate": 1.8858900648185446e-07, + "loss": 0.6723, + "step": 3166 + }, + { + "epoch": 0.12, + "learning_rate": 1.8858534441718238e-07, + "loss": 0.702, + "step": 3167 + }, + { + "epoch": 0.12, + "learning_rate": 1.8858168235251034e-07, + "loss": 0.7025, + "step": 3168 + }, + { + "epoch": 0.12, + "learning_rate": 1.8857802028783826e-07, + "loss": 0.6906, + "step": 3169 + }, + { + "epoch": 0.12, + "learning_rate": 1.8857435822316622e-07, + "loss": 0.7013, + "step": 3170 + }, + { + "epoch": 0.12, + "learning_rate": 1.8857069615849414e-07, + "loss": 0.6946, + "step": 3171 + }, + { + "epoch": 0.12, + "learning_rate": 1.885670340938221e-07, + "loss": 0.6985, + "step": 3172 + }, + { + "epoch": 0.12, + "learning_rate": 1.8856337202915002e-07, + "loss": 0.6812, + "step": 3173 + }, + { + "epoch": 0.12, + "learning_rate": 1.8855970996447798e-07, + "loss": 0.6856, + "step": 3174 + }, + { + "epoch": 0.12, + "learning_rate": 1.885560478998059e-07, + "loss": 0.6979, + "step": 3175 + }, + { + "epoch": 0.12, + "learning_rate": 1.8855238583513386e-07, + "loss": 0.7032, + "step": 3176 + }, + { + "epoch": 0.12, + "learning_rate": 1.8854872377046178e-07, + "loss": 0.6893, + "step": 3177 + }, + { + "epoch": 0.12, + "learning_rate": 1.885450617057897e-07, + "loss": 0.7013, + "step": 3178 + }, + { + "epoch": 0.12, + "learning_rate": 1.8854139964111766e-07, + "loss": 0.6893, + "step": 3179 + }, + { + "epoch": 0.12, + "learning_rate": 1.885377375764456e-07, + "loss": 0.6976, + "step": 3180 + }, + { + "epoch": 0.12, + "learning_rate": 1.8853407551177354e-07, + "loss": 0.6958, + "step": 3181 + }, + { + "epoch": 0.12, + "learning_rate": 1.8853041344710147e-07, + "loss": 0.6916, + "step": 3182 + }, + { + "epoch": 0.12, + "learning_rate": 1.885267513824294e-07, + "loss": 0.7068, + "step": 3183 + }, + { + "epoch": 0.12, + "learning_rate": 1.8852308931775732e-07, + "loss": 0.6866, + "step": 3184 + }, + { + "epoch": 0.12, + "learning_rate": 1.8851942725308527e-07, + "loss": 0.6734, + "step": 3185 + }, + { + "epoch": 0.12, + "learning_rate": 1.885157651884132e-07, + "loss": 0.6934, + "step": 3186 + }, + { + "epoch": 0.12, + "learning_rate": 1.8851210312374115e-07, + "loss": 0.6824, + "step": 3187 + }, + { + "epoch": 0.12, + "learning_rate": 1.8850844105906908e-07, + "loss": 0.7013, + "step": 3188 + }, + { + "epoch": 0.12, + "learning_rate": 1.8850477899439703e-07, + "loss": 0.7056, + "step": 3189 + }, + { + "epoch": 0.12, + "learning_rate": 1.8850111692972496e-07, + "loss": 0.6784, + "step": 3190 + }, + { + "epoch": 0.12, + "learning_rate": 1.8849745486505291e-07, + "loss": 0.6934, + "step": 3191 + }, + { + "epoch": 0.12, + "learning_rate": 1.8849379280038084e-07, + "loss": 0.6843, + "step": 3192 + }, + { + "epoch": 0.12, + "learning_rate": 1.884901307357088e-07, + "loss": 0.6821, + "step": 3193 + }, + { + "epoch": 0.12, + "learning_rate": 1.8848646867103672e-07, + "loss": 0.7079, + "step": 3194 + }, + { + "epoch": 0.12, + "learning_rate": 1.8848280660636467e-07, + "loss": 0.6821, + "step": 3195 + }, + { + "epoch": 0.12, + "learning_rate": 1.884791445416926e-07, + "loss": 0.6991, + "step": 3196 + }, + { + "epoch": 0.12, + "learning_rate": 1.8847548247702055e-07, + "loss": 0.6849, + "step": 3197 + }, + { + "epoch": 0.12, + "learning_rate": 1.8847182041234848e-07, + "loss": 0.6914, + "step": 3198 + }, + { + "epoch": 0.12, + "learning_rate": 1.884681583476764e-07, + "loss": 0.6924, + "step": 3199 + }, + { + "epoch": 0.12, + "learning_rate": 1.8846449628300436e-07, + "loss": 0.6732, + "step": 3200 + }, + { + "epoch": 0.12, + "learning_rate": 1.884608342183323e-07, + "loss": 0.7041, + "step": 3201 + }, + { + "epoch": 0.12, + "learning_rate": 1.8845717215366021e-07, + "loss": 0.7233, + "step": 3202 + }, + { + "epoch": 0.12, + "learning_rate": 1.8845351008898817e-07, + "loss": 0.6911, + "step": 3203 + }, + { + "epoch": 0.12, + "learning_rate": 1.884498480243161e-07, + "loss": 0.7084, + "step": 3204 + }, + { + "epoch": 0.12, + "learning_rate": 1.8844618595964402e-07, + "loss": 0.6899, + "step": 3205 + }, + { + "epoch": 0.12, + "learning_rate": 1.8844252389497197e-07, + "loss": 0.6832, + "step": 3206 + }, + { + "epoch": 0.12, + "learning_rate": 1.884388618302999e-07, + "loss": 0.6948, + "step": 3207 + }, + { + "epoch": 0.12, + "learning_rate": 1.8843519976562785e-07, + "loss": 0.6842, + "step": 3208 + }, + { + "epoch": 0.12, + "learning_rate": 1.8843153770095578e-07, + "loss": 0.7012, + "step": 3209 + }, + { + "epoch": 0.12, + "learning_rate": 1.8842787563628373e-07, + "loss": 0.6857, + "step": 3210 + }, + { + "epoch": 0.12, + "learning_rate": 1.8842421357161166e-07, + "loss": 0.7015, + "step": 3211 + }, + { + "epoch": 0.12, + "learning_rate": 1.884205515069396e-07, + "loss": 0.692, + "step": 3212 + }, + { + "epoch": 0.12, + "learning_rate": 1.8841688944226754e-07, + "loss": 0.7111, + "step": 3213 + }, + { + "epoch": 0.12, + "learning_rate": 1.884132273775955e-07, + "loss": 0.7037, + "step": 3214 + }, + { + "epoch": 0.12, + "learning_rate": 1.8840956531292342e-07, + "loss": 0.7003, + "step": 3215 + }, + { + "epoch": 0.12, + "learning_rate": 1.8840590324825137e-07, + "loss": 0.7076, + "step": 3216 + }, + { + "epoch": 0.12, + "learning_rate": 1.884022411835793e-07, + "loss": 0.6996, + "step": 3217 + }, + { + "epoch": 0.12, + "learning_rate": 1.8839857911890725e-07, + "loss": 0.6835, + "step": 3218 + }, + { + "epoch": 0.12, + "learning_rate": 1.8839491705423515e-07, + "loss": 0.7236, + "step": 3219 + }, + { + "epoch": 0.12, + "learning_rate": 1.883912549895631e-07, + "loss": 0.684, + "step": 3220 + }, + { + "epoch": 0.12, + "learning_rate": 1.8838759292489103e-07, + "loss": 0.6871, + "step": 3221 + }, + { + "epoch": 0.12, + "learning_rate": 1.8838393086021899e-07, + "loss": 0.6897, + "step": 3222 + }, + { + "epoch": 0.12, + "learning_rate": 1.883802687955469e-07, + "loss": 0.6959, + "step": 3223 + }, + { + "epoch": 0.12, + "learning_rate": 1.8837660673087487e-07, + "loss": 0.6918, + "step": 3224 + }, + { + "epoch": 0.12, + "learning_rate": 1.883729446662028e-07, + "loss": 0.6888, + "step": 3225 + }, + { + "epoch": 0.12, + "learning_rate": 1.8836928260153072e-07, + "loss": 0.6935, + "step": 3226 + }, + { + "epoch": 0.12, + "learning_rate": 1.8836562053685867e-07, + "loss": 0.6808, + "step": 3227 + }, + { + "epoch": 0.12, + "learning_rate": 1.883619584721866e-07, + "loss": 0.6966, + "step": 3228 + }, + { + "epoch": 0.12, + "learning_rate": 1.8835829640751455e-07, + "loss": 0.6973, + "step": 3229 + }, + { + "epoch": 0.12, + "learning_rate": 1.8835463434284248e-07, + "loss": 0.6939, + "step": 3230 + }, + { + "epoch": 0.12, + "learning_rate": 1.8835097227817043e-07, + "loss": 0.6981, + "step": 3231 + }, + { + "epoch": 0.12, + "learning_rate": 1.8834731021349836e-07, + "loss": 0.7037, + "step": 3232 + }, + { + "epoch": 0.12, + "learning_rate": 1.883436481488263e-07, + "loss": 0.6993, + "step": 3233 + }, + { + "epoch": 0.12, + "learning_rate": 1.8833998608415424e-07, + "loss": 0.6881, + "step": 3234 + }, + { + "epoch": 0.12, + "learning_rate": 1.883363240194822e-07, + "loss": 0.6983, + "step": 3235 + }, + { + "epoch": 0.12, + "learning_rate": 1.8833266195481012e-07, + "loss": 0.7081, + "step": 3236 + }, + { + "epoch": 0.12, + "learning_rate": 1.8832899989013805e-07, + "loss": 0.7034, + "step": 3237 + }, + { + "epoch": 0.12, + "learning_rate": 1.8832533782546597e-07, + "loss": 0.6821, + "step": 3238 + }, + { + "epoch": 0.12, + "learning_rate": 1.8832167576079392e-07, + "loss": 0.6972, + "step": 3239 + }, + { + "epoch": 0.12, + "learning_rate": 1.8831801369612185e-07, + "loss": 0.7015, + "step": 3240 + }, + { + "epoch": 0.12, + "learning_rate": 1.883143516314498e-07, + "loss": 0.6817, + "step": 3241 + }, + { + "epoch": 0.12, + "learning_rate": 1.8831068956677773e-07, + "loss": 0.6897, + "step": 3242 + }, + { + "epoch": 0.12, + "learning_rate": 1.8830702750210568e-07, + "loss": 0.6869, + "step": 3243 + }, + { + "epoch": 0.12, + "learning_rate": 1.883033654374336e-07, + "loss": 0.7044, + "step": 3244 + }, + { + "epoch": 0.12, + "learning_rate": 1.8829970337276156e-07, + "loss": 0.6983, + "step": 3245 + }, + { + "epoch": 0.12, + "learning_rate": 1.882960413080895e-07, + "loss": 0.694, + "step": 3246 + }, + { + "epoch": 0.12, + "learning_rate": 1.8829237924341744e-07, + "loss": 0.6927, + "step": 3247 + }, + { + "epoch": 0.12, + "learning_rate": 1.8828871717874537e-07, + "loss": 0.6932, + "step": 3248 + }, + { + "epoch": 0.12, + "learning_rate": 1.882850551140733e-07, + "loss": 0.6826, + "step": 3249 + }, + { + "epoch": 0.12, + "learning_rate": 1.8828139304940125e-07, + "loss": 0.6824, + "step": 3250 + }, + { + "epoch": 0.12, + "learning_rate": 1.8827773098472918e-07, + "loss": 0.6999, + "step": 3251 + }, + { + "epoch": 0.12, + "learning_rate": 1.8827406892005713e-07, + "loss": 0.6975, + "step": 3252 + }, + { + "epoch": 0.12, + "learning_rate": 1.8827040685538506e-07, + "loss": 0.696, + "step": 3253 + }, + { + "epoch": 0.12, + "learning_rate": 1.88266744790713e-07, + "loss": 0.6877, + "step": 3254 + }, + { + "epoch": 0.12, + "learning_rate": 1.882630827260409e-07, + "loss": 0.6803, + "step": 3255 + }, + { + "epoch": 0.12, + "learning_rate": 1.8825942066136886e-07, + "loss": 0.6649, + "step": 3256 + }, + { + "epoch": 0.12, + "learning_rate": 1.882557585966968e-07, + "loss": 0.6777, + "step": 3257 + }, + { + "epoch": 0.12, + "learning_rate": 1.8825209653202474e-07, + "loss": 0.6967, + "step": 3258 + }, + { + "epoch": 0.12, + "learning_rate": 1.8824843446735267e-07, + "loss": 0.6892, + "step": 3259 + }, + { + "epoch": 0.12, + "learning_rate": 1.8824477240268062e-07, + "loss": 0.6888, + "step": 3260 + }, + { + "epoch": 0.12, + "learning_rate": 1.8824111033800855e-07, + "loss": 0.6871, + "step": 3261 + }, + { + "epoch": 0.12, + "learning_rate": 1.882374482733365e-07, + "loss": 0.6761, + "step": 3262 + }, + { + "epoch": 0.12, + "learning_rate": 1.8823378620866443e-07, + "loss": 0.6916, + "step": 3263 + }, + { + "epoch": 0.12, + "learning_rate": 1.8823012414399238e-07, + "loss": 0.6947, + "step": 3264 + }, + { + "epoch": 0.12, + "learning_rate": 1.882264620793203e-07, + "loss": 0.6899, + "step": 3265 + }, + { + "epoch": 0.12, + "learning_rate": 1.8822280001464826e-07, + "loss": 0.6989, + "step": 3266 + }, + { + "epoch": 0.12, + "learning_rate": 1.882191379499762e-07, + "loss": 0.6999, + "step": 3267 + }, + { + "epoch": 0.12, + "learning_rate": 1.8821547588530414e-07, + "loss": 0.7122, + "step": 3268 + }, + { + "epoch": 0.12, + "learning_rate": 1.8821181382063207e-07, + "loss": 0.6926, + "step": 3269 + }, + { + "epoch": 0.12, + "learning_rate": 1.8820815175596e-07, + "loss": 0.7025, + "step": 3270 + }, + { + "epoch": 0.12, + "learning_rate": 1.8820448969128795e-07, + "loss": 0.7061, + "step": 3271 + }, + { + "epoch": 0.12, + "learning_rate": 1.8820082762661588e-07, + "loss": 0.6891, + "step": 3272 + }, + { + "epoch": 0.12, + "learning_rate": 1.881971655619438e-07, + "loss": 0.6735, + "step": 3273 + }, + { + "epoch": 0.12, + "learning_rate": 1.8819350349727176e-07, + "loss": 0.7114, + "step": 3274 + }, + { + "epoch": 0.12, + "learning_rate": 1.8818984143259968e-07, + "loss": 0.6907, + "step": 3275 + }, + { + "epoch": 0.12, + "learning_rate": 1.881861793679276e-07, + "loss": 0.6965, + "step": 3276 + }, + { + "epoch": 0.12, + "learning_rate": 1.8818251730325556e-07, + "loss": 0.706, + "step": 3277 + }, + { + "epoch": 0.12, + "learning_rate": 1.881788552385835e-07, + "loss": 0.6885, + "step": 3278 + }, + { + "epoch": 0.12, + "learning_rate": 1.8817519317391144e-07, + "loss": 0.6832, + "step": 3279 + }, + { + "epoch": 0.12, + "learning_rate": 1.8817153110923937e-07, + "loss": 0.6962, + "step": 3280 + }, + { + "epoch": 0.12, + "learning_rate": 1.8816786904456732e-07, + "loss": 0.699, + "step": 3281 + }, + { + "epoch": 0.12, + "learning_rate": 1.8816420697989525e-07, + "loss": 0.6958, + "step": 3282 + }, + { + "epoch": 0.12, + "learning_rate": 1.881605449152232e-07, + "loss": 0.7101, + "step": 3283 + }, + { + "epoch": 0.12, + "learning_rate": 1.8815688285055113e-07, + "loss": 0.6879, + "step": 3284 + }, + { + "epoch": 0.12, + "learning_rate": 1.8815322078587908e-07, + "loss": 0.6968, + "step": 3285 + }, + { + "epoch": 0.12, + "learning_rate": 1.88149558721207e-07, + "loss": 0.6975, + "step": 3286 + }, + { + "epoch": 0.12, + "learning_rate": 1.8814589665653496e-07, + "loss": 0.6932, + "step": 3287 + }, + { + "epoch": 0.12, + "learning_rate": 1.881422345918629e-07, + "loss": 0.6761, + "step": 3288 + }, + { + "epoch": 0.12, + "learning_rate": 1.8813857252719084e-07, + "loss": 0.6985, + "step": 3289 + }, + { + "epoch": 0.12, + "learning_rate": 1.8813491046251877e-07, + "loss": 0.6746, + "step": 3290 + }, + { + "epoch": 0.12, + "learning_rate": 1.881312483978467e-07, + "loss": 0.6857, + "step": 3291 + }, + { + "epoch": 0.12, + "learning_rate": 1.8812758633317462e-07, + "loss": 0.6736, + "step": 3292 + }, + { + "epoch": 0.12, + "learning_rate": 1.8812392426850258e-07, + "loss": 0.7101, + "step": 3293 + }, + { + "epoch": 0.12, + "learning_rate": 1.881202622038305e-07, + "loss": 0.7019, + "step": 3294 + }, + { + "epoch": 0.12, + "learning_rate": 1.8811660013915846e-07, + "loss": 0.6941, + "step": 3295 + }, + { + "epoch": 0.12, + "learning_rate": 1.8811293807448638e-07, + "loss": 0.6772, + "step": 3296 + }, + { + "epoch": 0.12, + "learning_rate": 1.881092760098143e-07, + "loss": 0.7125, + "step": 3297 + }, + { + "epoch": 0.12, + "learning_rate": 1.8810561394514226e-07, + "loss": 0.6949, + "step": 3298 + }, + { + "epoch": 0.12, + "learning_rate": 1.881019518804702e-07, + "loss": 0.6895, + "step": 3299 + }, + { + "epoch": 0.12, + "learning_rate": 1.8809828981579814e-07, + "loss": 0.6915, + "step": 3300 + }, + { + "epoch": 0.12, + "learning_rate": 1.8809462775112607e-07, + "loss": 0.7016, + "step": 3301 + }, + { + "epoch": 0.12, + "learning_rate": 1.8809096568645402e-07, + "loss": 0.6837, + "step": 3302 + }, + { + "epoch": 0.12, + "learning_rate": 1.8808730362178195e-07, + "loss": 0.6781, + "step": 3303 + }, + { + "epoch": 0.12, + "learning_rate": 1.880836415571099e-07, + "loss": 0.6894, + "step": 3304 + }, + { + "epoch": 0.12, + "learning_rate": 1.8807997949243783e-07, + "loss": 0.6865, + "step": 3305 + }, + { + "epoch": 0.12, + "learning_rate": 1.8807631742776578e-07, + "loss": 0.6909, + "step": 3306 + }, + { + "epoch": 0.12, + "learning_rate": 1.880726553630937e-07, + "loss": 0.6976, + "step": 3307 + }, + { + "epoch": 0.12, + "learning_rate": 1.8806899329842166e-07, + "loss": 0.691, + "step": 3308 + }, + { + "epoch": 0.12, + "learning_rate": 1.8806533123374956e-07, + "loss": 0.6796, + "step": 3309 + }, + { + "epoch": 0.12, + "learning_rate": 1.8806166916907751e-07, + "loss": 0.6889, + "step": 3310 + }, + { + "epoch": 0.12, + "learning_rate": 1.8805800710440544e-07, + "loss": 0.6868, + "step": 3311 + }, + { + "epoch": 0.12, + "learning_rate": 1.880543450397334e-07, + "loss": 0.7067, + "step": 3312 + }, + { + "epoch": 0.12, + "learning_rate": 1.8805068297506132e-07, + "loss": 0.6973, + "step": 3313 + }, + { + "epoch": 0.12, + "learning_rate": 1.8804702091038927e-07, + "loss": 0.6903, + "step": 3314 + }, + { + "epoch": 0.12, + "learning_rate": 1.880433588457172e-07, + "loss": 0.7064, + "step": 3315 + }, + { + "epoch": 0.12, + "learning_rate": 1.8803969678104515e-07, + "loss": 0.6796, + "step": 3316 + }, + { + "epoch": 0.12, + "learning_rate": 1.8803603471637308e-07, + "loss": 0.6796, + "step": 3317 + }, + { + "epoch": 0.12, + "learning_rate": 1.8803237265170103e-07, + "loss": 0.6924, + "step": 3318 + }, + { + "epoch": 0.12, + "learning_rate": 1.8802871058702896e-07, + "loss": 0.7103, + "step": 3319 + }, + { + "epoch": 0.12, + "learning_rate": 1.880250485223569e-07, + "loss": 0.7144, + "step": 3320 + }, + { + "epoch": 0.12, + "learning_rate": 1.8802138645768484e-07, + "loss": 0.6928, + "step": 3321 + }, + { + "epoch": 0.12, + "learning_rate": 1.8801772439301277e-07, + "loss": 0.6879, + "step": 3322 + }, + { + "epoch": 0.12, + "learning_rate": 1.8801406232834072e-07, + "loss": 0.6797, + "step": 3323 + }, + { + "epoch": 0.12, + "learning_rate": 1.8801040026366865e-07, + "loss": 0.7069, + "step": 3324 + }, + { + "epoch": 0.12, + "learning_rate": 1.880067381989966e-07, + "loss": 0.7019, + "step": 3325 + }, + { + "epoch": 0.12, + "learning_rate": 1.8800307613432453e-07, + "loss": 0.6913, + "step": 3326 + }, + { + "epoch": 0.12, + "learning_rate": 1.8799941406965245e-07, + "loss": 0.697, + "step": 3327 + }, + { + "epoch": 0.12, + "learning_rate": 1.8799575200498038e-07, + "loss": 0.6919, + "step": 3328 + }, + { + "epoch": 0.12, + "learning_rate": 1.8799208994030833e-07, + "loss": 0.7036, + "step": 3329 + }, + { + "epoch": 0.12, + "learning_rate": 1.8798842787563626e-07, + "loss": 0.6882, + "step": 3330 + }, + { + "epoch": 0.12, + "learning_rate": 1.879847658109642e-07, + "loss": 0.7006, + "step": 3331 + }, + { + "epoch": 0.12, + "learning_rate": 1.8798110374629214e-07, + "loss": 0.6987, + "step": 3332 + }, + { + "epoch": 0.12, + "learning_rate": 1.879774416816201e-07, + "loss": 0.7027, + "step": 3333 + }, + { + "epoch": 0.12, + "learning_rate": 1.8797377961694802e-07, + "loss": 0.6935, + "step": 3334 + }, + { + "epoch": 0.12, + "learning_rate": 1.8797011755227597e-07, + "loss": 0.68, + "step": 3335 + }, + { + "epoch": 0.12, + "learning_rate": 1.879664554876039e-07, + "loss": 0.6866, + "step": 3336 + }, + { + "epoch": 0.12, + "learning_rate": 1.8796279342293185e-07, + "loss": 0.6897, + "step": 3337 + }, + { + "epoch": 0.12, + "learning_rate": 1.8795913135825978e-07, + "loss": 0.6947, + "step": 3338 + }, + { + "epoch": 0.12, + "learning_rate": 1.8795546929358773e-07, + "loss": 0.6899, + "step": 3339 + }, + { + "epoch": 0.12, + "learning_rate": 1.8795180722891566e-07, + "loss": 0.7016, + "step": 3340 + }, + { + "epoch": 0.12, + "learning_rate": 1.8794814516424359e-07, + "loss": 0.7177, + "step": 3341 + }, + { + "epoch": 0.12, + "learning_rate": 1.8794448309957154e-07, + "loss": 0.6887, + "step": 3342 + }, + { + "epoch": 0.12, + "learning_rate": 1.8794082103489947e-07, + "loss": 0.6892, + "step": 3343 + }, + { + "epoch": 0.12, + "learning_rate": 1.8793715897022742e-07, + "loss": 0.6776, + "step": 3344 + }, + { + "epoch": 0.12, + "learning_rate": 1.8793349690555535e-07, + "loss": 0.6885, + "step": 3345 + }, + { + "epoch": 0.12, + "learning_rate": 1.8792983484088327e-07, + "loss": 0.6886, + "step": 3346 + }, + { + "epoch": 0.12, + "learning_rate": 1.879261727762112e-07, + "loss": 0.6945, + "step": 3347 + }, + { + "epoch": 0.12, + "learning_rate": 1.8792251071153915e-07, + "loss": 0.7144, + "step": 3348 + }, + { + "epoch": 0.12, + "learning_rate": 1.8791884864686708e-07, + "loss": 0.6959, + "step": 3349 + }, + { + "epoch": 0.12, + "learning_rate": 1.8791518658219503e-07, + "loss": 0.6932, + "step": 3350 + }, + { + "epoch": 0.12, + "learning_rate": 1.8791152451752296e-07, + "loss": 0.6885, + "step": 3351 + }, + { + "epoch": 0.12, + "learning_rate": 1.879078624528509e-07, + "loss": 0.7204, + "step": 3352 + }, + { + "epoch": 0.12, + "learning_rate": 1.8790420038817884e-07, + "loss": 0.6875, + "step": 3353 + }, + { + "epoch": 0.12, + "learning_rate": 1.879005383235068e-07, + "loss": 0.68, + "step": 3354 + }, + { + "epoch": 0.12, + "learning_rate": 1.8789687625883472e-07, + "loss": 0.6831, + "step": 3355 + }, + { + "epoch": 0.12, + "learning_rate": 1.8789321419416267e-07, + "loss": 0.7112, + "step": 3356 + }, + { + "epoch": 0.12, + "learning_rate": 1.878895521294906e-07, + "loss": 0.6899, + "step": 3357 + }, + { + "epoch": 0.12, + "learning_rate": 1.8788589006481855e-07, + "loss": 0.6901, + "step": 3358 + }, + { + "epoch": 0.12, + "learning_rate": 1.8788222800014648e-07, + "loss": 0.6869, + "step": 3359 + }, + { + "epoch": 0.12, + "learning_rate": 1.8787856593547443e-07, + "loss": 0.6849, + "step": 3360 + }, + { + "epoch": 0.12, + "learning_rate": 1.8787490387080236e-07, + "loss": 0.6801, + "step": 3361 + }, + { + "epoch": 0.12, + "learning_rate": 1.878712418061303e-07, + "loss": 0.6935, + "step": 3362 + }, + { + "epoch": 0.12, + "learning_rate": 1.878675797414582e-07, + "loss": 0.7263, + "step": 3363 + }, + { + "epoch": 0.12, + "learning_rate": 1.8786391767678616e-07, + "loss": 0.7041, + "step": 3364 + }, + { + "epoch": 0.12, + "learning_rate": 1.878602556121141e-07, + "loss": 0.688, + "step": 3365 + }, + { + "epoch": 0.12, + "learning_rate": 1.8785659354744204e-07, + "loss": 0.6977, + "step": 3366 + }, + { + "epoch": 0.12, + "learning_rate": 1.8785293148276997e-07, + "loss": 0.6868, + "step": 3367 + }, + { + "epoch": 0.12, + "learning_rate": 1.878492694180979e-07, + "loss": 0.694, + "step": 3368 + }, + { + "epoch": 0.12, + "learning_rate": 1.8784560735342585e-07, + "loss": 0.6932, + "step": 3369 + }, + { + "epoch": 0.12, + "learning_rate": 1.8784194528875378e-07, + "loss": 0.7147, + "step": 3370 + }, + { + "epoch": 0.12, + "learning_rate": 1.8783828322408173e-07, + "loss": 0.6977, + "step": 3371 + }, + { + "epoch": 0.12, + "learning_rate": 1.8783462115940966e-07, + "loss": 0.6993, + "step": 3372 + }, + { + "epoch": 0.12, + "learning_rate": 1.878309590947376e-07, + "loss": 0.6929, + "step": 3373 + }, + { + "epoch": 0.12, + "learning_rate": 1.8782729703006554e-07, + "loss": 0.698, + "step": 3374 + }, + { + "epoch": 0.12, + "learning_rate": 1.878236349653935e-07, + "loss": 0.7002, + "step": 3375 + }, + { + "epoch": 0.12, + "learning_rate": 1.8781997290072142e-07, + "loss": 0.718, + "step": 3376 + }, + { + "epoch": 0.12, + "learning_rate": 1.8781631083604937e-07, + "loss": 0.6979, + "step": 3377 + }, + { + "epoch": 0.12, + "learning_rate": 1.878126487713773e-07, + "loss": 0.6995, + "step": 3378 + }, + { + "epoch": 0.12, + "learning_rate": 1.8780898670670525e-07, + "loss": 0.683, + "step": 3379 + }, + { + "epoch": 0.12, + "learning_rate": 1.8780532464203318e-07, + "loss": 0.6828, + "step": 3380 + }, + { + "epoch": 0.12, + "learning_rate": 1.878016625773611e-07, + "loss": 0.7004, + "step": 3381 + }, + { + "epoch": 0.12, + "learning_rate": 1.8779800051268903e-07, + "loss": 0.6854, + "step": 3382 + }, + { + "epoch": 0.12, + "learning_rate": 1.8779433844801698e-07, + "loss": 0.6841, + "step": 3383 + }, + { + "epoch": 0.12, + "learning_rate": 1.877906763833449e-07, + "loss": 0.6839, + "step": 3384 + }, + { + "epoch": 0.12, + "learning_rate": 1.8778701431867286e-07, + "loss": 0.6988, + "step": 3385 + }, + { + "epoch": 0.12, + "learning_rate": 1.877833522540008e-07, + "loss": 0.6873, + "step": 3386 + }, + { + "epoch": 0.12, + "learning_rate": 1.8777969018932874e-07, + "loss": 0.6857, + "step": 3387 + }, + { + "epoch": 0.12, + "learning_rate": 1.8777602812465667e-07, + "loss": 0.6948, + "step": 3388 + }, + { + "epoch": 0.12, + "learning_rate": 1.8777236605998462e-07, + "loss": 0.6745, + "step": 3389 + }, + { + "epoch": 0.12, + "learning_rate": 1.8776870399531255e-07, + "loss": 0.6777, + "step": 3390 + }, + { + "epoch": 0.12, + "learning_rate": 1.8776504193064048e-07, + "loss": 0.704, + "step": 3391 + }, + { + "epoch": 0.12, + "learning_rate": 1.8776137986596843e-07, + "loss": 0.7, + "step": 3392 + }, + { + "epoch": 0.12, + "learning_rate": 1.8775771780129636e-07, + "loss": 0.723, + "step": 3393 + }, + { + "epoch": 0.12, + "learning_rate": 1.877540557366243e-07, + "loss": 0.6883, + "step": 3394 + }, + { + "epoch": 0.12, + "learning_rate": 1.8775039367195224e-07, + "loss": 0.6841, + "step": 3395 + }, + { + "epoch": 0.12, + "learning_rate": 1.877467316072802e-07, + "loss": 0.6732, + "step": 3396 + }, + { + "epoch": 0.12, + "learning_rate": 1.8774306954260812e-07, + "loss": 0.6857, + "step": 3397 + }, + { + "epoch": 0.12, + "learning_rate": 1.8773940747793607e-07, + "loss": 0.69, + "step": 3398 + }, + { + "epoch": 0.12, + "learning_rate": 1.8773574541326397e-07, + "loss": 0.6989, + "step": 3399 + }, + { + "epoch": 0.12, + "learning_rate": 1.8773208334859192e-07, + "loss": 0.705, + "step": 3400 + }, + { + "epoch": 0.12, + "learning_rate": 1.8772842128391985e-07, + "loss": 0.6902, + "step": 3401 + }, + { + "epoch": 0.12, + "learning_rate": 1.877247592192478e-07, + "loss": 0.6988, + "step": 3402 + }, + { + "epoch": 0.12, + "learning_rate": 1.8772109715457573e-07, + "loss": 0.6916, + "step": 3403 + }, + { + "epoch": 0.12, + "learning_rate": 1.8771743508990368e-07, + "loss": 0.6754, + "step": 3404 + }, + { + "epoch": 0.12, + "learning_rate": 1.877137730252316e-07, + "loss": 0.6825, + "step": 3405 + }, + { + "epoch": 0.12, + "learning_rate": 1.8771011096055956e-07, + "loss": 0.6924, + "step": 3406 + }, + { + "epoch": 0.12, + "learning_rate": 1.877064488958875e-07, + "loss": 0.6992, + "step": 3407 + }, + { + "epoch": 0.12, + "learning_rate": 1.8770278683121544e-07, + "loss": 0.6858, + "step": 3408 + }, + { + "epoch": 0.12, + "learning_rate": 1.8769912476654337e-07, + "loss": 0.6947, + "step": 3409 + }, + { + "epoch": 0.12, + "learning_rate": 1.8769546270187132e-07, + "loss": 0.6875, + "step": 3410 + }, + { + "epoch": 0.12, + "learning_rate": 1.8769180063719925e-07, + "loss": 0.6959, + "step": 3411 + }, + { + "epoch": 0.12, + "learning_rate": 1.8768813857252718e-07, + "loss": 0.6854, + "step": 3412 + }, + { + "epoch": 0.12, + "learning_rate": 1.8768447650785513e-07, + "loss": 0.6928, + "step": 3413 + }, + { + "epoch": 0.12, + "learning_rate": 1.8768081444318306e-07, + "loss": 0.7073, + "step": 3414 + }, + { + "epoch": 0.12, + "learning_rate": 1.87677152378511e-07, + "loss": 0.6926, + "step": 3415 + }, + { + "epoch": 0.12, + "learning_rate": 1.8767349031383894e-07, + "loss": 0.6787, + "step": 3416 + }, + { + "epoch": 0.13, + "learning_rate": 1.8766982824916686e-07, + "loss": 0.6933, + "step": 3417 + }, + { + "epoch": 0.13, + "learning_rate": 1.876661661844948e-07, + "loss": 0.6974, + "step": 3418 + }, + { + "epoch": 0.13, + "learning_rate": 1.8766250411982274e-07, + "loss": 0.7062, + "step": 3419 + }, + { + "epoch": 0.13, + "learning_rate": 1.8765884205515067e-07, + "loss": 0.7125, + "step": 3420 + }, + { + "epoch": 0.13, + "learning_rate": 1.8765517999047862e-07, + "loss": 0.7047, + "step": 3421 + }, + { + "epoch": 0.13, + "learning_rate": 1.8765151792580655e-07, + "loss": 0.6881, + "step": 3422 + }, + { + "epoch": 0.13, + "learning_rate": 1.876478558611345e-07, + "loss": 0.677, + "step": 3423 + }, + { + "epoch": 0.13, + "learning_rate": 1.8764419379646243e-07, + "loss": 0.6928, + "step": 3424 + }, + { + "epoch": 0.13, + "learning_rate": 1.8764053173179038e-07, + "loss": 0.6836, + "step": 3425 + }, + { + "epoch": 0.13, + "learning_rate": 1.876368696671183e-07, + "loss": 0.7085, + "step": 3426 + }, + { + "epoch": 0.13, + "learning_rate": 1.8763320760244626e-07, + "loss": 0.6795, + "step": 3427 + }, + { + "epoch": 0.13, + "learning_rate": 1.876295455377742e-07, + "loss": 0.7252, + "step": 3428 + }, + { + "epoch": 0.13, + "learning_rate": 1.8762588347310214e-07, + "loss": 0.7002, + "step": 3429 + }, + { + "epoch": 0.13, + "learning_rate": 1.8762222140843007e-07, + "loss": 0.6942, + "step": 3430 + }, + { + "epoch": 0.13, + "learning_rate": 1.8761855934375802e-07, + "loss": 0.6898, + "step": 3431 + }, + { + "epoch": 0.13, + "learning_rate": 1.8761489727908595e-07, + "loss": 0.7256, + "step": 3432 + }, + { + "epoch": 0.13, + "learning_rate": 1.876112352144139e-07, + "loss": 0.7069, + "step": 3433 + }, + { + "epoch": 0.13, + "learning_rate": 1.8760757314974183e-07, + "loss": 0.687, + "step": 3434 + }, + { + "epoch": 0.13, + "learning_rate": 1.8760391108506975e-07, + "loss": 0.706, + "step": 3435 + }, + { + "epoch": 0.13, + "learning_rate": 1.8760024902039768e-07, + "loss": 0.6839, + "step": 3436 + }, + { + "epoch": 0.13, + "learning_rate": 1.8759658695572563e-07, + "loss": 0.7053, + "step": 3437 + }, + { + "epoch": 0.13, + "learning_rate": 1.8759292489105356e-07, + "loss": 0.6926, + "step": 3438 + }, + { + "epoch": 0.13, + "learning_rate": 1.875892628263815e-07, + "loss": 0.687, + "step": 3439 + }, + { + "epoch": 0.13, + "learning_rate": 1.8758560076170944e-07, + "loss": 0.7075, + "step": 3440 + }, + { + "epoch": 0.13, + "learning_rate": 1.8758193869703737e-07, + "loss": 0.7071, + "step": 3441 + }, + { + "epoch": 0.13, + "learning_rate": 1.8757827663236532e-07, + "loss": 0.6919, + "step": 3442 + }, + { + "epoch": 0.13, + "learning_rate": 1.8757461456769325e-07, + "loss": 0.6926, + "step": 3443 + }, + { + "epoch": 0.13, + "learning_rate": 1.875709525030212e-07, + "loss": 0.7076, + "step": 3444 + }, + { + "epoch": 0.13, + "learning_rate": 1.8756729043834913e-07, + "loss": 0.692, + "step": 3445 + }, + { + "epoch": 0.13, + "learning_rate": 1.8756362837367708e-07, + "loss": 0.7002, + "step": 3446 + }, + { + "epoch": 0.13, + "learning_rate": 1.87559966309005e-07, + "loss": 0.7004, + "step": 3447 + }, + { + "epoch": 0.13, + "learning_rate": 1.8755630424433296e-07, + "loss": 0.6966, + "step": 3448 + }, + { + "epoch": 0.13, + "learning_rate": 1.8755264217966089e-07, + "loss": 0.7051, + "step": 3449 + }, + { + "epoch": 0.13, + "learning_rate": 1.8754898011498884e-07, + "loss": 0.6949, + "step": 3450 + }, + { + "epoch": 0.13, + "learning_rate": 1.8754531805031677e-07, + "loss": 0.6996, + "step": 3451 + }, + { + "epoch": 0.13, + "learning_rate": 1.8754165598564472e-07, + "loss": 0.6814, + "step": 3452 + }, + { + "epoch": 0.13, + "learning_rate": 1.8753799392097262e-07, + "loss": 0.6661, + "step": 3453 + }, + { + "epoch": 0.13, + "learning_rate": 1.8753433185630057e-07, + "loss": 0.6428, + "step": 3454 + }, + { + "epoch": 0.13, + "learning_rate": 1.875306697916285e-07, + "loss": 0.6881, + "step": 3455 + }, + { + "epoch": 0.13, + "learning_rate": 1.8752700772695645e-07, + "loss": 0.6748, + "step": 3456 + }, + { + "epoch": 0.13, + "learning_rate": 1.8752334566228438e-07, + "loss": 0.6755, + "step": 3457 + }, + { + "epoch": 0.13, + "learning_rate": 1.8751968359761233e-07, + "loss": 0.6848, + "step": 3458 + }, + { + "epoch": 0.13, + "learning_rate": 1.8751602153294026e-07, + "loss": 0.6901, + "step": 3459 + }, + { + "epoch": 0.13, + "learning_rate": 1.875123594682682e-07, + "loss": 0.695, + "step": 3460 + }, + { + "epoch": 0.13, + "learning_rate": 1.8750869740359614e-07, + "loss": 0.7082, + "step": 3461 + }, + { + "epoch": 0.13, + "learning_rate": 1.8750503533892407e-07, + "loss": 0.6793, + "step": 3462 + }, + { + "epoch": 0.13, + "learning_rate": 1.8750137327425202e-07, + "loss": 0.69, + "step": 3463 + }, + { + "epoch": 0.13, + "learning_rate": 1.8749771120957995e-07, + "loss": 0.6988, + "step": 3464 + }, + { + "epoch": 0.13, + "learning_rate": 1.874940491449079e-07, + "loss": 0.7199, + "step": 3465 + }, + { + "epoch": 0.13, + "learning_rate": 1.8749038708023583e-07, + "loss": 0.7118, + "step": 3466 + }, + { + "epoch": 0.13, + "learning_rate": 1.8748672501556378e-07, + "loss": 0.6942, + "step": 3467 + }, + { + "epoch": 0.13, + "learning_rate": 1.874830629508917e-07, + "loss": 0.6954, + "step": 3468 + }, + { + "epoch": 0.13, + "learning_rate": 1.8747940088621966e-07, + "loss": 0.6923, + "step": 3469 + }, + { + "epoch": 0.13, + "learning_rate": 1.8747573882154759e-07, + "loss": 0.6862, + "step": 3470 + }, + { + "epoch": 0.13, + "learning_rate": 1.874720767568755e-07, + "loss": 0.6975, + "step": 3471 + }, + { + "epoch": 0.13, + "learning_rate": 1.8746841469220344e-07, + "loss": 0.6847, + "step": 3472 + }, + { + "epoch": 0.13, + "learning_rate": 1.874647526275314e-07, + "loss": 0.6833, + "step": 3473 + }, + { + "epoch": 0.13, + "learning_rate": 1.8746109056285932e-07, + "loss": 0.698, + "step": 3474 + }, + { + "epoch": 0.13, + "learning_rate": 1.8745742849818727e-07, + "loss": 0.6712, + "step": 3475 + }, + { + "epoch": 0.13, + "learning_rate": 1.874537664335152e-07, + "loss": 0.6793, + "step": 3476 + }, + { + "epoch": 0.13, + "learning_rate": 1.8745010436884315e-07, + "loss": 0.6897, + "step": 3477 + }, + { + "epoch": 0.13, + "learning_rate": 1.8744644230417108e-07, + "loss": 0.7031, + "step": 3478 + }, + { + "epoch": 0.13, + "learning_rate": 1.8744278023949903e-07, + "loss": 0.6778, + "step": 3479 + }, + { + "epoch": 0.13, + "learning_rate": 1.8743911817482696e-07, + "loss": 0.6905, + "step": 3480 + }, + { + "epoch": 0.13, + "learning_rate": 1.874354561101549e-07, + "loss": 0.6826, + "step": 3481 + }, + { + "epoch": 0.13, + "learning_rate": 1.8743179404548284e-07, + "loss": 0.6924, + "step": 3482 + }, + { + "epoch": 0.13, + "learning_rate": 1.8742813198081076e-07, + "loss": 0.7025, + "step": 3483 + }, + { + "epoch": 0.13, + "learning_rate": 1.8742446991613872e-07, + "loss": 0.6787, + "step": 3484 + }, + { + "epoch": 0.13, + "learning_rate": 1.8742080785146664e-07, + "loss": 0.6938, + "step": 3485 + }, + { + "epoch": 0.13, + "learning_rate": 1.874171457867946e-07, + "loss": 0.6819, + "step": 3486 + }, + { + "epoch": 0.13, + "learning_rate": 1.8741348372212252e-07, + "loss": 0.6777, + "step": 3487 + }, + { + "epoch": 0.13, + "learning_rate": 1.8740982165745048e-07, + "loss": 0.6919, + "step": 3488 + }, + { + "epoch": 0.13, + "learning_rate": 1.8740615959277838e-07, + "loss": 0.7073, + "step": 3489 + }, + { + "epoch": 0.13, + "learning_rate": 1.8740249752810633e-07, + "loss": 0.6805, + "step": 3490 + }, + { + "epoch": 0.13, + "learning_rate": 1.8739883546343426e-07, + "loss": 0.6874, + "step": 3491 + }, + { + "epoch": 0.13, + "learning_rate": 1.873951733987622e-07, + "loss": 0.6812, + "step": 3492 + }, + { + "epoch": 0.13, + "learning_rate": 1.8739151133409014e-07, + "loss": 0.6894, + "step": 3493 + }, + { + "epoch": 0.13, + "learning_rate": 1.873878492694181e-07, + "loss": 0.7044, + "step": 3494 + }, + { + "epoch": 0.13, + "learning_rate": 1.8738418720474602e-07, + "loss": 0.6728, + "step": 3495 + }, + { + "epoch": 0.13, + "learning_rate": 1.8738052514007397e-07, + "loss": 0.694, + "step": 3496 + }, + { + "epoch": 0.13, + "learning_rate": 1.873768630754019e-07, + "loss": 0.6877, + "step": 3497 + }, + { + "epoch": 0.13, + "learning_rate": 1.8737320101072985e-07, + "loss": 0.6797, + "step": 3498 + }, + { + "epoch": 0.13, + "learning_rate": 1.8736953894605778e-07, + "loss": 0.7015, + "step": 3499 + }, + { + "epoch": 0.13, + "learning_rate": 1.8736587688138573e-07, + "loss": 0.7021, + "step": 3500 + }, + { + "epoch": 0.13, + "learning_rate": 1.8736221481671366e-07, + "loss": 0.7105, + "step": 3501 + }, + { + "epoch": 0.13, + "learning_rate": 1.873585527520416e-07, + "loss": 0.728, + "step": 3502 + }, + { + "epoch": 0.13, + "learning_rate": 1.8735489068736954e-07, + "loss": 0.6915, + "step": 3503 + }, + { + "epoch": 0.13, + "learning_rate": 1.873512286226975e-07, + "loss": 0.6903, + "step": 3504 + }, + { + "epoch": 0.13, + "learning_rate": 1.8734756655802542e-07, + "loss": 0.6958, + "step": 3505 + }, + { + "epoch": 0.13, + "learning_rate": 1.8734390449335334e-07, + "loss": 0.6884, + "step": 3506 + }, + { + "epoch": 0.13, + "learning_rate": 1.8734024242868127e-07, + "loss": 0.6796, + "step": 3507 + }, + { + "epoch": 0.13, + "learning_rate": 1.8733658036400922e-07, + "loss": 0.7116, + "step": 3508 + }, + { + "epoch": 0.13, + "learning_rate": 1.8733291829933715e-07, + "loss": 0.687, + "step": 3509 + }, + { + "epoch": 0.13, + "learning_rate": 1.8732925623466508e-07, + "loss": 0.6941, + "step": 3510 + }, + { + "epoch": 0.13, + "learning_rate": 1.8732559416999303e-07, + "loss": 0.6824, + "step": 3511 + }, + { + "epoch": 0.13, + "learning_rate": 1.8732193210532096e-07, + "loss": 0.6775, + "step": 3512 + }, + { + "epoch": 0.13, + "learning_rate": 1.873182700406489e-07, + "loss": 0.6822, + "step": 3513 + }, + { + "epoch": 0.13, + "learning_rate": 1.8731460797597684e-07, + "loss": 0.6745, + "step": 3514 + }, + { + "epoch": 0.13, + "learning_rate": 1.873109459113048e-07, + "loss": 0.7006, + "step": 3515 + }, + { + "epoch": 0.13, + "learning_rate": 1.8730728384663272e-07, + "loss": 0.6835, + "step": 3516 + }, + { + "epoch": 0.13, + "learning_rate": 1.8730362178196067e-07, + "loss": 0.6831, + "step": 3517 + }, + { + "epoch": 0.13, + "learning_rate": 1.872999597172886e-07, + "loss": 0.681, + "step": 3518 + }, + { + "epoch": 0.13, + "learning_rate": 1.8729629765261655e-07, + "loss": 0.6813, + "step": 3519 + }, + { + "epoch": 0.13, + "learning_rate": 1.8729263558794448e-07, + "loss": 0.6927, + "step": 3520 + }, + { + "epoch": 0.13, + "learning_rate": 1.8728897352327243e-07, + "loss": 0.7119, + "step": 3521 + }, + { + "epoch": 0.13, + "learning_rate": 1.8728531145860036e-07, + "loss": 0.6833, + "step": 3522 + }, + { + "epoch": 0.13, + "learning_rate": 1.872816493939283e-07, + "loss": 0.6747, + "step": 3523 + }, + { + "epoch": 0.13, + "learning_rate": 1.872779873292562e-07, + "loss": 0.678, + "step": 3524 + }, + { + "epoch": 0.13, + "learning_rate": 1.8727432526458416e-07, + "loss": 0.6965, + "step": 3525 + }, + { + "epoch": 0.13, + "learning_rate": 1.872706631999121e-07, + "loss": 0.7092, + "step": 3526 + }, + { + "epoch": 0.13, + "learning_rate": 1.8726700113524004e-07, + "loss": 0.6919, + "step": 3527 + }, + { + "epoch": 0.13, + "learning_rate": 1.8726333907056797e-07, + "loss": 0.6999, + "step": 3528 + }, + { + "epoch": 0.13, + "learning_rate": 1.8725967700589592e-07, + "loss": 0.6922, + "step": 3529 + }, + { + "epoch": 0.13, + "learning_rate": 1.8725601494122385e-07, + "loss": 0.6975, + "step": 3530 + }, + { + "epoch": 0.13, + "learning_rate": 1.872523528765518e-07, + "loss": 0.6981, + "step": 3531 + }, + { + "epoch": 0.13, + "learning_rate": 1.8724869081187973e-07, + "loss": 0.7045, + "step": 3532 + }, + { + "epoch": 0.13, + "learning_rate": 1.8724502874720766e-07, + "loss": 0.6896, + "step": 3533 + }, + { + "epoch": 0.13, + "learning_rate": 1.872413666825356e-07, + "loss": 0.6986, + "step": 3534 + }, + { + "epoch": 0.13, + "learning_rate": 1.8723770461786354e-07, + "loss": 0.6882, + "step": 3535 + }, + { + "epoch": 0.13, + "learning_rate": 1.872340425531915e-07, + "loss": 0.7062, + "step": 3536 + }, + { + "epoch": 0.13, + "learning_rate": 1.8723038048851941e-07, + "loss": 0.6888, + "step": 3537 + }, + { + "epoch": 0.13, + "learning_rate": 1.8722671842384737e-07, + "loss": 0.6853, + "step": 3538 + }, + { + "epoch": 0.13, + "learning_rate": 1.872230563591753e-07, + "loss": 0.6963, + "step": 3539 + }, + { + "epoch": 0.13, + "learning_rate": 1.8721939429450325e-07, + "loss": 0.6809, + "step": 3540 + }, + { + "epoch": 0.13, + "learning_rate": 1.8721573222983117e-07, + "loss": 0.6781, + "step": 3541 + }, + { + "epoch": 0.13, + "learning_rate": 1.872120701651591e-07, + "loss": 0.6996, + "step": 3542 + }, + { + "epoch": 0.13, + "learning_rate": 1.8720840810048703e-07, + "loss": 0.7039, + "step": 3543 + }, + { + "epoch": 0.13, + "learning_rate": 1.8720474603581498e-07, + "loss": 0.6933, + "step": 3544 + }, + { + "epoch": 0.13, + "learning_rate": 1.872010839711429e-07, + "loss": 0.6978, + "step": 3545 + }, + { + "epoch": 0.13, + "learning_rate": 1.8719742190647086e-07, + "loss": 0.6978, + "step": 3546 + }, + { + "epoch": 0.13, + "learning_rate": 1.871937598417988e-07, + "loss": 0.687, + "step": 3547 + }, + { + "epoch": 0.13, + "learning_rate": 1.8719009777712674e-07, + "loss": 0.6948, + "step": 3548 + }, + { + "epoch": 0.13, + "learning_rate": 1.8718643571245467e-07, + "loss": 0.6953, + "step": 3549 + }, + { + "epoch": 0.13, + "learning_rate": 1.8718277364778262e-07, + "loss": 0.6774, + "step": 3550 + }, + { + "epoch": 0.13, + "learning_rate": 1.8717911158311055e-07, + "loss": 0.6999, + "step": 3551 + }, + { + "epoch": 0.13, + "learning_rate": 1.871754495184385e-07, + "loss": 0.6996, + "step": 3552 + }, + { + "epoch": 0.13, + "learning_rate": 1.8717178745376643e-07, + "loss": 0.6979, + "step": 3553 + }, + { + "epoch": 0.13, + "learning_rate": 1.8716812538909435e-07, + "loss": 0.6929, + "step": 3554 + }, + { + "epoch": 0.13, + "learning_rate": 1.871644633244223e-07, + "loss": 0.6976, + "step": 3555 + }, + { + "epoch": 0.13, + "learning_rate": 1.8716080125975023e-07, + "loss": 0.6915, + "step": 3556 + }, + { + "epoch": 0.13, + "learning_rate": 1.871571391950782e-07, + "loss": 0.702, + "step": 3557 + }, + { + "epoch": 0.13, + "learning_rate": 1.8715347713040611e-07, + "loss": 0.7004, + "step": 3558 + }, + { + "epoch": 0.13, + "learning_rate": 1.8714981506573407e-07, + "loss": 0.7142, + "step": 3559 + }, + { + "epoch": 0.13, + "learning_rate": 1.8714615300106197e-07, + "loss": 0.6942, + "step": 3560 + }, + { + "epoch": 0.13, + "learning_rate": 1.8714249093638992e-07, + "loss": 0.6879, + "step": 3561 + }, + { + "epoch": 0.13, + "learning_rate": 1.8713882887171785e-07, + "loss": 0.7077, + "step": 3562 + }, + { + "epoch": 0.13, + "learning_rate": 1.871351668070458e-07, + "loss": 0.7036, + "step": 3563 + }, + { + "epoch": 0.13, + "learning_rate": 1.8713150474237373e-07, + "loss": 0.7239, + "step": 3564 + }, + { + "epoch": 0.13, + "learning_rate": 1.8712784267770168e-07, + "loss": 0.6944, + "step": 3565 + }, + { + "epoch": 0.13, + "learning_rate": 1.871241806130296e-07, + "loss": 0.6961, + "step": 3566 + }, + { + "epoch": 0.13, + "learning_rate": 1.8712051854835756e-07, + "loss": 0.6938, + "step": 3567 + }, + { + "epoch": 0.13, + "learning_rate": 1.8711685648368549e-07, + "loss": 0.69, + "step": 3568 + }, + { + "epoch": 0.13, + "learning_rate": 1.8711319441901344e-07, + "loss": 0.7037, + "step": 3569 + }, + { + "epoch": 0.13, + "learning_rate": 1.8710953235434137e-07, + "loss": 0.73, + "step": 3570 + }, + { + "epoch": 0.13, + "learning_rate": 1.8710587028966932e-07, + "loss": 0.7114, + "step": 3571 + }, + { + "epoch": 0.13, + "learning_rate": 1.8710220822499725e-07, + "loss": 0.7007, + "step": 3572 + }, + { + "epoch": 0.13, + "learning_rate": 1.870985461603252e-07, + "loss": 0.6857, + "step": 3573 + }, + { + "epoch": 0.13, + "learning_rate": 1.8709488409565313e-07, + "loss": 0.6938, + "step": 3574 + }, + { + "epoch": 0.13, + "learning_rate": 1.8709122203098108e-07, + "loss": 0.6957, + "step": 3575 + }, + { + "epoch": 0.13, + "learning_rate": 1.87087559966309e-07, + "loss": 0.6845, + "step": 3576 + }, + { + "epoch": 0.13, + "learning_rate": 1.8708389790163693e-07, + "loss": 0.7037, + "step": 3577 + }, + { + "epoch": 0.13, + "learning_rate": 1.8708023583696486e-07, + "loss": 0.6959, + "step": 3578 + }, + { + "epoch": 0.13, + "learning_rate": 1.870765737722928e-07, + "loss": 0.6773, + "step": 3579 + }, + { + "epoch": 0.13, + "learning_rate": 1.8707291170762074e-07, + "loss": 0.6979, + "step": 3580 + }, + { + "epoch": 0.13, + "learning_rate": 1.8706924964294867e-07, + "loss": 0.6964, + "step": 3581 + }, + { + "epoch": 0.13, + "learning_rate": 1.8706558757827662e-07, + "loss": 0.6835, + "step": 3582 + }, + { + "epoch": 0.13, + "learning_rate": 1.8706192551360455e-07, + "loss": 0.6899, + "step": 3583 + }, + { + "epoch": 0.13, + "learning_rate": 1.870582634489325e-07, + "loss": 0.7048, + "step": 3584 + }, + { + "epoch": 0.13, + "learning_rate": 1.8705460138426043e-07, + "loss": 0.6973, + "step": 3585 + }, + { + "epoch": 0.13, + "learning_rate": 1.8705093931958838e-07, + "loss": 0.6887, + "step": 3586 + }, + { + "epoch": 0.13, + "learning_rate": 1.870472772549163e-07, + "loss": 0.6865, + "step": 3587 + }, + { + "epoch": 0.13, + "learning_rate": 1.8704361519024426e-07, + "loss": 0.6863, + "step": 3588 + }, + { + "epoch": 0.13, + "learning_rate": 1.8703995312557219e-07, + "loss": 0.6896, + "step": 3589 + }, + { + "epoch": 0.13, + "learning_rate": 1.8703629106090014e-07, + "loss": 0.6883, + "step": 3590 + }, + { + "epoch": 0.13, + "learning_rate": 1.8703262899622807e-07, + "loss": 0.7077, + "step": 3591 + }, + { + "epoch": 0.13, + "learning_rate": 1.8702896693155602e-07, + "loss": 0.6988, + "step": 3592 + }, + { + "epoch": 0.13, + "learning_rate": 1.8702530486688395e-07, + "loss": 0.702, + "step": 3593 + }, + { + "epoch": 0.13, + "learning_rate": 1.870216428022119e-07, + "loss": 0.6848, + "step": 3594 + }, + { + "epoch": 0.13, + "learning_rate": 1.8701798073753982e-07, + "loss": 0.6866, + "step": 3595 + }, + { + "epoch": 0.13, + "learning_rate": 1.8701431867286775e-07, + "loss": 0.6986, + "step": 3596 + }, + { + "epoch": 0.13, + "learning_rate": 1.8701065660819568e-07, + "loss": 0.6804, + "step": 3597 + }, + { + "epoch": 0.13, + "learning_rate": 1.8700699454352363e-07, + "loss": 0.6947, + "step": 3598 + }, + { + "epoch": 0.13, + "learning_rate": 1.8700333247885156e-07, + "loss": 0.7006, + "step": 3599 + }, + { + "epoch": 0.13, + "learning_rate": 1.869996704141795e-07, + "loss": 0.6982, + "step": 3600 + }, + { + "epoch": 0.13, + "learning_rate": 1.8699600834950744e-07, + "loss": 0.6887, + "step": 3601 + }, + { + "epoch": 0.13, + "learning_rate": 1.869923462848354e-07, + "loss": 0.6907, + "step": 3602 + }, + { + "epoch": 0.13, + "learning_rate": 1.8698868422016332e-07, + "loss": 0.6982, + "step": 3603 + }, + { + "epoch": 0.13, + "learning_rate": 1.8698502215549124e-07, + "loss": 0.6788, + "step": 3604 + }, + { + "epoch": 0.13, + "learning_rate": 1.869813600908192e-07, + "loss": 0.6916, + "step": 3605 + }, + { + "epoch": 0.13, + "learning_rate": 1.8697769802614712e-07, + "loss": 0.7019, + "step": 3606 + }, + { + "epoch": 0.13, + "learning_rate": 1.8697403596147508e-07, + "loss": 0.6951, + "step": 3607 + }, + { + "epoch": 0.13, + "learning_rate": 1.86970373896803e-07, + "loss": 0.6977, + "step": 3608 + }, + { + "epoch": 0.13, + "learning_rate": 1.8696671183213096e-07, + "loss": 0.7031, + "step": 3609 + }, + { + "epoch": 0.13, + "learning_rate": 1.8696304976745888e-07, + "loss": 0.6696, + "step": 3610 + }, + { + "epoch": 0.13, + "learning_rate": 1.8695938770278684e-07, + "loss": 0.6848, + "step": 3611 + }, + { + "epoch": 0.13, + "learning_rate": 1.8695572563811476e-07, + "loss": 0.6739, + "step": 3612 + }, + { + "epoch": 0.13, + "learning_rate": 1.8695206357344272e-07, + "loss": 0.6872, + "step": 3613 + }, + { + "epoch": 0.13, + "learning_rate": 1.8694840150877062e-07, + "loss": 0.6993, + "step": 3614 + }, + { + "epoch": 0.13, + "learning_rate": 1.8694473944409857e-07, + "loss": 0.6978, + "step": 3615 + }, + { + "epoch": 0.13, + "learning_rate": 1.869410773794265e-07, + "loss": 0.6723, + "step": 3616 + }, + { + "epoch": 0.13, + "learning_rate": 1.8693741531475445e-07, + "loss": 0.7051, + "step": 3617 + }, + { + "epoch": 0.13, + "learning_rate": 1.8693375325008238e-07, + "loss": 0.6999, + "step": 3618 + }, + { + "epoch": 0.13, + "learning_rate": 1.8693009118541033e-07, + "loss": 0.6889, + "step": 3619 + }, + { + "epoch": 0.13, + "learning_rate": 1.8692642912073826e-07, + "loss": 0.7007, + "step": 3620 + }, + { + "epoch": 0.13, + "learning_rate": 1.869227670560662e-07, + "loss": 0.7009, + "step": 3621 + }, + { + "epoch": 0.13, + "learning_rate": 1.8691910499139414e-07, + "loss": 0.6733, + "step": 3622 + }, + { + "epoch": 0.13, + "learning_rate": 1.869154429267221e-07, + "loss": 0.7028, + "step": 3623 + }, + { + "epoch": 0.13, + "learning_rate": 1.8691178086205002e-07, + "loss": 0.7018, + "step": 3624 + }, + { + "epoch": 0.13, + "learning_rate": 1.8690811879737794e-07, + "loss": 0.6985, + "step": 3625 + }, + { + "epoch": 0.13, + "learning_rate": 1.869044567327059e-07, + "loss": 0.6884, + "step": 3626 + }, + { + "epoch": 0.13, + "learning_rate": 1.8690079466803382e-07, + "loss": 0.6816, + "step": 3627 + }, + { + "epoch": 0.13, + "learning_rate": 1.8689713260336178e-07, + "loss": 0.6959, + "step": 3628 + }, + { + "epoch": 0.13, + "learning_rate": 1.868934705386897e-07, + "loss": 0.7075, + "step": 3629 + }, + { + "epoch": 0.13, + "learning_rate": 1.8688980847401766e-07, + "loss": 0.7077, + "step": 3630 + }, + { + "epoch": 0.13, + "learning_rate": 1.8688614640934558e-07, + "loss": 0.6812, + "step": 3631 + }, + { + "epoch": 0.13, + "learning_rate": 1.868824843446735e-07, + "loss": 0.6959, + "step": 3632 + }, + { + "epoch": 0.13, + "learning_rate": 1.8687882228000144e-07, + "loss": 0.7173, + "step": 3633 + }, + { + "epoch": 0.13, + "learning_rate": 1.868751602153294e-07, + "loss": 0.6929, + "step": 3634 + }, + { + "epoch": 0.13, + "learning_rate": 1.8687149815065732e-07, + "loss": 0.7123, + "step": 3635 + }, + { + "epoch": 0.13, + "learning_rate": 1.8686783608598527e-07, + "loss": 0.6963, + "step": 3636 + }, + { + "epoch": 0.13, + "learning_rate": 1.868641740213132e-07, + "loss": 0.7083, + "step": 3637 + }, + { + "epoch": 0.13, + "learning_rate": 1.8686051195664115e-07, + "loss": 0.7086, + "step": 3638 + }, + { + "epoch": 0.13, + "learning_rate": 1.8685684989196908e-07, + "loss": 0.7061, + "step": 3639 + }, + { + "epoch": 0.13, + "learning_rate": 1.8685318782729703e-07, + "loss": 0.6855, + "step": 3640 + }, + { + "epoch": 0.13, + "learning_rate": 1.8684952576262496e-07, + "loss": 0.7038, + "step": 3641 + }, + { + "epoch": 0.13, + "learning_rate": 1.868458636979529e-07, + "loss": 0.6981, + "step": 3642 + }, + { + "epoch": 0.13, + "learning_rate": 1.8684220163328084e-07, + "loss": 0.7099, + "step": 3643 + }, + { + "epoch": 0.13, + "learning_rate": 1.868385395686088e-07, + "loss": 0.699, + "step": 3644 + }, + { + "epoch": 0.13, + "learning_rate": 1.8683487750393672e-07, + "loss": 0.6953, + "step": 3645 + }, + { + "epoch": 0.13, + "learning_rate": 1.8683121543926467e-07, + "loss": 0.699, + "step": 3646 + }, + { + "epoch": 0.13, + "learning_rate": 1.868275533745926e-07, + "loss": 0.7155, + "step": 3647 + }, + { + "epoch": 0.13, + "learning_rate": 1.8682389130992052e-07, + "loss": 0.711, + "step": 3648 + }, + { + "epoch": 0.13, + "learning_rate": 1.8682022924524848e-07, + "loss": 0.6863, + "step": 3649 + }, + { + "epoch": 0.13, + "learning_rate": 1.868165671805764e-07, + "loss": 0.6827, + "step": 3650 + }, + { + "epoch": 0.13, + "learning_rate": 1.8681290511590433e-07, + "loss": 0.6818, + "step": 3651 + }, + { + "epoch": 0.13, + "learning_rate": 1.8680924305123226e-07, + "loss": 0.68, + "step": 3652 + }, + { + "epoch": 0.13, + "learning_rate": 1.868055809865602e-07, + "loss": 0.6842, + "step": 3653 + }, + { + "epoch": 0.13, + "learning_rate": 1.8680191892188814e-07, + "loss": 0.6985, + "step": 3654 + }, + { + "epoch": 0.13, + "learning_rate": 1.867982568572161e-07, + "loss": 0.6928, + "step": 3655 + }, + { + "epoch": 0.13, + "learning_rate": 1.8679459479254401e-07, + "loss": 0.682, + "step": 3656 + }, + { + "epoch": 0.13, + "learning_rate": 1.8679093272787197e-07, + "loss": 0.698, + "step": 3657 + }, + { + "epoch": 0.13, + "learning_rate": 1.867872706631999e-07, + "loss": 0.7047, + "step": 3658 + }, + { + "epoch": 0.13, + "learning_rate": 1.8678360859852785e-07, + "loss": 0.6959, + "step": 3659 + }, + { + "epoch": 0.13, + "learning_rate": 1.8677994653385577e-07, + "loss": 0.6939, + "step": 3660 + }, + { + "epoch": 0.13, + "learning_rate": 1.8677628446918373e-07, + "loss": 0.6877, + "step": 3661 + }, + { + "epoch": 0.13, + "learning_rate": 1.8677262240451165e-07, + "loss": 0.7058, + "step": 3662 + }, + { + "epoch": 0.13, + "learning_rate": 1.867689603398396e-07, + "loss": 0.7066, + "step": 3663 + }, + { + "epoch": 0.13, + "learning_rate": 1.8676529827516753e-07, + "loss": 0.6979, + "step": 3664 + }, + { + "epoch": 0.13, + "learning_rate": 1.867616362104955e-07, + "loss": 0.6878, + "step": 3665 + }, + { + "epoch": 0.13, + "learning_rate": 1.8675797414582341e-07, + "loss": 0.6907, + "step": 3666 + }, + { + "epoch": 0.13, + "learning_rate": 1.8675431208115137e-07, + "loss": 0.6969, + "step": 3667 + }, + { + "epoch": 0.13, + "learning_rate": 1.8675065001647927e-07, + "loss": 0.6866, + "step": 3668 + }, + { + "epoch": 0.13, + "learning_rate": 1.8674698795180722e-07, + "loss": 0.683, + "step": 3669 + }, + { + "epoch": 0.13, + "learning_rate": 1.8674332588713515e-07, + "loss": 0.6842, + "step": 3670 + }, + { + "epoch": 0.13, + "learning_rate": 1.867396638224631e-07, + "loss": 0.6926, + "step": 3671 + }, + { + "epoch": 0.13, + "learning_rate": 1.8673600175779103e-07, + "loss": 0.69, + "step": 3672 + }, + { + "epoch": 0.13, + "learning_rate": 1.8673233969311898e-07, + "loss": 0.6781, + "step": 3673 + }, + { + "epoch": 0.13, + "learning_rate": 1.867286776284469e-07, + "loss": 0.6805, + "step": 3674 + }, + { + "epoch": 0.13, + "learning_rate": 1.8672501556377483e-07, + "loss": 0.6926, + "step": 3675 + }, + { + "epoch": 0.13, + "learning_rate": 1.867213534991028e-07, + "loss": 0.7032, + "step": 3676 + }, + { + "epoch": 0.13, + "learning_rate": 1.8671769143443071e-07, + "loss": 0.6962, + "step": 3677 + }, + { + "epoch": 0.13, + "learning_rate": 1.8671402936975867e-07, + "loss": 0.7072, + "step": 3678 + }, + { + "epoch": 0.13, + "learning_rate": 1.867103673050866e-07, + "loss": 0.7037, + "step": 3679 + }, + { + "epoch": 0.13, + "learning_rate": 1.8670670524041455e-07, + "loss": 0.6983, + "step": 3680 + }, + { + "epoch": 0.13, + "learning_rate": 1.8670304317574247e-07, + "loss": 0.7091, + "step": 3681 + }, + { + "epoch": 0.13, + "learning_rate": 1.8669938111107043e-07, + "loss": 0.6851, + "step": 3682 + }, + { + "epoch": 0.13, + "learning_rate": 1.8669571904639835e-07, + "loss": 0.6895, + "step": 3683 + }, + { + "epoch": 0.13, + "learning_rate": 1.866920569817263e-07, + "loss": 0.6936, + "step": 3684 + }, + { + "epoch": 0.13, + "learning_rate": 1.8668839491705423e-07, + "loss": 0.7092, + "step": 3685 + }, + { + "epoch": 0.13, + "learning_rate": 1.8668473285238216e-07, + "loss": 0.6949, + "step": 3686 + }, + { + "epoch": 0.13, + "learning_rate": 1.8668107078771009e-07, + "loss": 0.693, + "step": 3687 + }, + { + "epoch": 0.13, + "learning_rate": 1.8667740872303804e-07, + "loss": 0.7005, + "step": 3688 + }, + { + "epoch": 0.13, + "learning_rate": 1.8667374665836597e-07, + "loss": 0.6768, + "step": 3689 + }, + { + "epoch": 0.14, + "learning_rate": 1.8667008459369392e-07, + "loss": 0.7007, + "step": 3690 + }, + { + "epoch": 0.14, + "learning_rate": 1.8666642252902185e-07, + "loss": 0.6589, + "step": 3691 + }, + { + "epoch": 0.14, + "learning_rate": 1.866627604643498e-07, + "loss": 0.6999, + "step": 3692 + }, + { + "epoch": 0.14, + "learning_rate": 1.8665909839967773e-07, + "loss": 0.6995, + "step": 3693 + }, + { + "epoch": 0.14, + "learning_rate": 1.8665543633500568e-07, + "loss": 0.6877, + "step": 3694 + }, + { + "epoch": 0.14, + "learning_rate": 1.866517742703336e-07, + "loss": 0.6956, + "step": 3695 + }, + { + "epoch": 0.14, + "learning_rate": 1.8664811220566153e-07, + "loss": 0.6786, + "step": 3696 + }, + { + "epoch": 0.14, + "learning_rate": 1.8664445014098949e-07, + "loss": 0.6973, + "step": 3697 + }, + { + "epoch": 0.14, + "learning_rate": 1.866407880763174e-07, + "loss": 0.6725, + "step": 3698 + }, + { + "epoch": 0.14, + "learning_rate": 1.8663712601164537e-07, + "loss": 0.6721, + "step": 3699 + }, + { + "epoch": 0.14, + "learning_rate": 1.866334639469733e-07, + "loss": 0.6964, + "step": 3700 + }, + { + "epoch": 0.14, + "learning_rate": 1.8662980188230125e-07, + "loss": 0.7083, + "step": 3701 + }, + { + "epoch": 0.14, + "learning_rate": 1.8662613981762917e-07, + "loss": 0.6858, + "step": 3702 + }, + { + "epoch": 0.14, + "learning_rate": 1.8662247775295713e-07, + "loss": 0.7054, + "step": 3703 + }, + { + "epoch": 0.14, + "learning_rate": 1.8661881568828503e-07, + "loss": 0.6803, + "step": 3704 + }, + { + "epoch": 0.14, + "learning_rate": 1.8661515362361298e-07, + "loss": 0.7269, + "step": 3705 + }, + { + "epoch": 0.14, + "learning_rate": 1.866114915589409e-07, + "loss": 0.7093, + "step": 3706 + }, + { + "epoch": 0.14, + "learning_rate": 1.8660782949426886e-07, + "loss": 0.6875, + "step": 3707 + }, + { + "epoch": 0.14, + "learning_rate": 1.8660416742959679e-07, + "loss": 0.6957, + "step": 3708 + }, + { + "epoch": 0.14, + "learning_rate": 1.8660050536492474e-07, + "loss": 0.6809, + "step": 3709 + }, + { + "epoch": 0.14, + "learning_rate": 1.8659684330025267e-07, + "loss": 0.6839, + "step": 3710 + }, + { + "epoch": 0.14, + "learning_rate": 1.8659318123558062e-07, + "loss": 0.7068, + "step": 3711 + }, + { + "epoch": 0.14, + "learning_rate": 1.8658951917090855e-07, + "loss": 0.7121, + "step": 3712 + }, + { + "epoch": 0.14, + "learning_rate": 1.865858571062365e-07, + "loss": 0.6975, + "step": 3713 + }, + { + "epoch": 0.14, + "learning_rate": 1.8658219504156442e-07, + "loss": 0.6867, + "step": 3714 + }, + { + "epoch": 0.14, + "learning_rate": 1.8657853297689238e-07, + "loss": 0.6812, + "step": 3715 + }, + { + "epoch": 0.14, + "learning_rate": 1.865748709122203e-07, + "loss": 0.7038, + "step": 3716 + }, + { + "epoch": 0.14, + "learning_rate": 1.8657120884754826e-07, + "loss": 0.678, + "step": 3717 + }, + { + "epoch": 0.14, + "learning_rate": 1.8656754678287618e-07, + "loss": 0.6999, + "step": 3718 + }, + { + "epoch": 0.14, + "learning_rate": 1.865638847182041e-07, + "loss": 0.6891, + "step": 3719 + }, + { + "epoch": 0.14, + "learning_rate": 1.8656022265353206e-07, + "loss": 0.6821, + "step": 3720 + }, + { + "epoch": 0.14, + "learning_rate": 1.8655656058886e-07, + "loss": 0.7044, + "step": 3721 + }, + { + "epoch": 0.14, + "learning_rate": 1.8655289852418792e-07, + "loss": 0.6866, + "step": 3722 + }, + { + "epoch": 0.14, + "learning_rate": 1.8654923645951584e-07, + "loss": 0.7025, + "step": 3723 + }, + { + "epoch": 0.14, + "learning_rate": 1.865455743948438e-07, + "loss": 0.6771, + "step": 3724 + }, + { + "epoch": 0.14, + "learning_rate": 1.8654191233017172e-07, + "loss": 0.7063, + "step": 3725 + }, + { + "epoch": 0.14, + "learning_rate": 1.8653825026549968e-07, + "loss": 0.6806, + "step": 3726 + }, + { + "epoch": 0.14, + "learning_rate": 1.865345882008276e-07, + "loss": 0.6933, + "step": 3727 + }, + { + "epoch": 0.14, + "learning_rate": 1.8653092613615556e-07, + "loss": 0.6881, + "step": 3728 + }, + { + "epoch": 0.14, + "learning_rate": 1.8652726407148348e-07, + "loss": 0.7101, + "step": 3729 + }, + { + "epoch": 0.14, + "learning_rate": 1.8652360200681144e-07, + "loss": 0.6842, + "step": 3730 + }, + { + "epoch": 0.14, + "learning_rate": 1.8651993994213936e-07, + "loss": 0.6917, + "step": 3731 + }, + { + "epoch": 0.14, + "learning_rate": 1.8651627787746732e-07, + "loss": 0.6824, + "step": 3732 + }, + { + "epoch": 0.14, + "learning_rate": 1.8651261581279524e-07, + "loss": 0.6908, + "step": 3733 + }, + { + "epoch": 0.14, + "learning_rate": 1.865089537481232e-07, + "loss": 0.6851, + "step": 3734 + }, + { + "epoch": 0.14, + "learning_rate": 1.8650529168345112e-07, + "loss": 0.6779, + "step": 3735 + }, + { + "epoch": 0.14, + "learning_rate": 1.8650162961877908e-07, + "loss": 0.6956, + "step": 3736 + }, + { + "epoch": 0.14, + "learning_rate": 1.86497967554107e-07, + "loss": 0.696, + "step": 3737 + }, + { + "epoch": 0.14, + "learning_rate": 1.8649430548943496e-07, + "loss": 0.7044, + "step": 3738 + }, + { + "epoch": 0.14, + "learning_rate": 1.8649064342476288e-07, + "loss": 0.6881, + "step": 3739 + }, + { + "epoch": 0.14, + "learning_rate": 1.864869813600908e-07, + "loss": 0.6883, + "step": 3740 + }, + { + "epoch": 0.14, + "learning_rate": 1.8648331929541874e-07, + "loss": 0.6934, + "step": 3741 + }, + { + "epoch": 0.14, + "learning_rate": 1.864796572307467e-07, + "loss": 0.6986, + "step": 3742 + }, + { + "epoch": 0.14, + "learning_rate": 1.8647599516607462e-07, + "loss": 0.6824, + "step": 3743 + }, + { + "epoch": 0.14, + "learning_rate": 1.8647233310140257e-07, + "loss": 0.7036, + "step": 3744 + }, + { + "epoch": 0.14, + "learning_rate": 1.864686710367305e-07, + "loss": 0.6836, + "step": 3745 + }, + { + "epoch": 0.14, + "learning_rate": 1.8646500897205842e-07, + "loss": 0.6804, + "step": 3746 + }, + { + "epoch": 0.14, + "learning_rate": 1.8646134690738638e-07, + "loss": 0.6919, + "step": 3747 + }, + { + "epoch": 0.14, + "learning_rate": 1.864576848427143e-07, + "loss": 0.6988, + "step": 3748 + }, + { + "epoch": 0.14, + "learning_rate": 1.8645402277804226e-07, + "loss": 0.6746, + "step": 3749 + }, + { + "epoch": 0.14, + "learning_rate": 1.8645036071337018e-07, + "loss": 0.7266, + "step": 3750 + }, + { + "epoch": 0.14, + "learning_rate": 1.8644669864869814e-07, + "loss": 0.6904, + "step": 3751 + }, + { + "epoch": 0.14, + "learning_rate": 1.8644303658402606e-07, + "loss": 0.6964, + "step": 3752 + }, + { + "epoch": 0.14, + "learning_rate": 1.8643937451935402e-07, + "loss": 0.6755, + "step": 3753 + }, + { + "epoch": 0.14, + "learning_rate": 1.8643571245468194e-07, + "loss": 0.7252, + "step": 3754 + }, + { + "epoch": 0.14, + "learning_rate": 1.864320503900099e-07, + "loss": 0.7003, + "step": 3755 + }, + { + "epoch": 0.14, + "learning_rate": 1.8642838832533782e-07, + "loss": 0.6811, + "step": 3756 + }, + { + "epoch": 0.14, + "learning_rate": 1.8642472626066578e-07, + "loss": 0.6678, + "step": 3757 + }, + { + "epoch": 0.14, + "learning_rate": 1.8642106419599368e-07, + "loss": 0.719, + "step": 3758 + }, + { + "epoch": 0.14, + "learning_rate": 1.8641740213132163e-07, + "loss": 0.6994, + "step": 3759 + }, + { + "epoch": 0.14, + "learning_rate": 1.8641374006664956e-07, + "loss": 0.6927, + "step": 3760 + }, + { + "epoch": 0.14, + "learning_rate": 1.864100780019775e-07, + "loss": 0.7162, + "step": 3761 + }, + { + "epoch": 0.14, + "learning_rate": 1.8640641593730544e-07, + "loss": 0.7122, + "step": 3762 + }, + { + "epoch": 0.14, + "learning_rate": 1.864027538726334e-07, + "loss": 0.6953, + "step": 3763 + }, + { + "epoch": 0.14, + "learning_rate": 1.8639909180796132e-07, + "loss": 0.6825, + "step": 3764 + }, + { + "epoch": 0.14, + "learning_rate": 1.8639542974328927e-07, + "loss": 0.6736, + "step": 3765 + }, + { + "epoch": 0.14, + "learning_rate": 1.863917676786172e-07, + "loss": 0.698, + "step": 3766 + }, + { + "epoch": 0.14, + "learning_rate": 1.8638810561394512e-07, + "loss": 0.7187, + "step": 3767 + }, + { + "epoch": 0.14, + "learning_rate": 1.8638444354927308e-07, + "loss": 0.6853, + "step": 3768 + }, + { + "epoch": 0.14, + "learning_rate": 1.86380781484601e-07, + "loss": 0.7009, + "step": 3769 + }, + { + "epoch": 0.14, + "learning_rate": 1.8637711941992896e-07, + "loss": 0.6834, + "step": 3770 + }, + { + "epoch": 0.14, + "learning_rate": 1.8637345735525688e-07, + "loss": 0.6784, + "step": 3771 + }, + { + "epoch": 0.14, + "learning_rate": 1.8636979529058483e-07, + "loss": 0.7044, + "step": 3772 + }, + { + "epoch": 0.14, + "learning_rate": 1.8636613322591276e-07, + "loss": 0.6776, + "step": 3773 + }, + { + "epoch": 0.14, + "learning_rate": 1.8636247116124071e-07, + "loss": 0.6781, + "step": 3774 + }, + { + "epoch": 0.14, + "learning_rate": 1.8635880909656864e-07, + "loss": 0.6883, + "step": 3775 + }, + { + "epoch": 0.14, + "learning_rate": 1.8635514703189657e-07, + "loss": 0.7069, + "step": 3776 + }, + { + "epoch": 0.14, + "learning_rate": 1.863514849672245e-07, + "loss": 0.6857, + "step": 3777 + }, + { + "epoch": 0.14, + "learning_rate": 1.8634782290255245e-07, + "loss": 0.6921, + "step": 3778 + }, + { + "epoch": 0.14, + "learning_rate": 1.8634416083788037e-07, + "loss": 0.6788, + "step": 3779 + }, + { + "epoch": 0.14, + "learning_rate": 1.8634049877320833e-07, + "loss": 0.6999, + "step": 3780 + }, + { + "epoch": 0.14, + "learning_rate": 1.8633683670853625e-07, + "loss": 0.6675, + "step": 3781 + }, + { + "epoch": 0.14, + "learning_rate": 1.863331746438642e-07, + "loss": 0.7002, + "step": 3782 + }, + { + "epoch": 0.14, + "learning_rate": 1.8632951257919213e-07, + "loss": 0.7061, + "step": 3783 + }, + { + "epoch": 0.14, + "learning_rate": 1.863258505145201e-07, + "loss": 0.6833, + "step": 3784 + }, + { + "epoch": 0.14, + "learning_rate": 1.8632218844984801e-07, + "loss": 0.7037, + "step": 3785 + }, + { + "epoch": 0.14, + "learning_rate": 1.8631852638517597e-07, + "loss": 0.6984, + "step": 3786 + }, + { + "epoch": 0.14, + "learning_rate": 1.863148643205039e-07, + "loss": 0.7332, + "step": 3787 + }, + { + "epoch": 0.14, + "learning_rate": 1.8631120225583185e-07, + "loss": 0.7321, + "step": 3788 + }, + { + "epoch": 0.14, + "learning_rate": 1.8630754019115977e-07, + "loss": 0.7, + "step": 3789 + }, + { + "epoch": 0.14, + "learning_rate": 1.863038781264877e-07, + "loss": 0.6905, + "step": 3790 + }, + { + "epoch": 0.14, + "learning_rate": 1.8630021606181565e-07, + "loss": 0.6981, + "step": 3791 + }, + { + "epoch": 0.14, + "learning_rate": 1.8629655399714358e-07, + "loss": 0.7097, + "step": 3792 + }, + { + "epoch": 0.14, + "learning_rate": 1.8629289193247153e-07, + "loss": 0.6974, + "step": 3793 + }, + { + "epoch": 0.14, + "learning_rate": 1.8628922986779943e-07, + "loss": 0.6934, + "step": 3794 + }, + { + "epoch": 0.14, + "learning_rate": 1.862855678031274e-07, + "loss": 0.7038, + "step": 3795 + }, + { + "epoch": 0.14, + "learning_rate": 1.8628190573845531e-07, + "loss": 0.6857, + "step": 3796 + }, + { + "epoch": 0.14, + "learning_rate": 1.8627824367378327e-07, + "loss": 0.6886, + "step": 3797 + }, + { + "epoch": 0.14, + "learning_rate": 1.862745816091112e-07, + "loss": 0.6818, + "step": 3798 + }, + { + "epoch": 0.14, + "learning_rate": 1.8627091954443915e-07, + "loss": 0.6972, + "step": 3799 + }, + { + "epoch": 0.14, + "learning_rate": 1.8626725747976707e-07, + "loss": 0.7003, + "step": 3800 + }, + { + "epoch": 0.14, + "learning_rate": 1.8626359541509503e-07, + "loss": 0.6841, + "step": 3801 + }, + { + "epoch": 0.14, + "learning_rate": 1.8625993335042295e-07, + "loss": 0.67, + "step": 3802 + }, + { + "epoch": 0.14, + "learning_rate": 1.862562712857509e-07, + "loss": 0.6956, + "step": 3803 + }, + { + "epoch": 0.14, + "learning_rate": 1.8625260922107883e-07, + "loss": 0.7042, + "step": 3804 + }, + { + "epoch": 0.14, + "learning_rate": 1.8624894715640679e-07, + "loss": 0.6787, + "step": 3805 + }, + { + "epoch": 0.14, + "learning_rate": 1.862452850917347e-07, + "loss": 0.6845, + "step": 3806 + }, + { + "epoch": 0.14, + "learning_rate": 1.8624162302706267e-07, + "loss": 0.701, + "step": 3807 + }, + { + "epoch": 0.14, + "learning_rate": 1.862379609623906e-07, + "loss": 0.7161, + "step": 3808 + }, + { + "epoch": 0.14, + "learning_rate": 1.8623429889771855e-07, + "loss": 0.7015, + "step": 3809 + }, + { + "epoch": 0.14, + "learning_rate": 1.8623063683304647e-07, + "loss": 0.6872, + "step": 3810 + }, + { + "epoch": 0.14, + "learning_rate": 1.862269747683744e-07, + "loss": 0.6962, + "step": 3811 + }, + { + "epoch": 0.14, + "learning_rate": 1.8622331270370233e-07, + "loss": 0.7084, + "step": 3812 + }, + { + "epoch": 0.14, + "learning_rate": 1.8621965063903028e-07, + "loss": 0.7172, + "step": 3813 + }, + { + "epoch": 0.14, + "learning_rate": 1.862159885743582e-07, + "loss": 0.6865, + "step": 3814 + }, + { + "epoch": 0.14, + "learning_rate": 1.8621232650968616e-07, + "loss": 0.6906, + "step": 3815 + }, + { + "epoch": 0.14, + "learning_rate": 1.8620866444501409e-07, + "loss": 0.7001, + "step": 3816 + }, + { + "epoch": 0.14, + "learning_rate": 1.86205002380342e-07, + "loss": 0.6901, + "step": 3817 + }, + { + "epoch": 0.14, + "learning_rate": 1.8620134031566997e-07, + "loss": 0.6602, + "step": 3818 + }, + { + "epoch": 0.14, + "learning_rate": 1.861976782509979e-07, + "loss": 0.7172, + "step": 3819 + }, + { + "epoch": 0.14, + "learning_rate": 1.8619401618632585e-07, + "loss": 0.7097, + "step": 3820 + }, + { + "epoch": 0.14, + "learning_rate": 1.8619035412165377e-07, + "loss": 0.6901, + "step": 3821 + }, + { + "epoch": 0.14, + "learning_rate": 1.8618669205698173e-07, + "loss": 0.6821, + "step": 3822 + }, + { + "epoch": 0.14, + "learning_rate": 1.8618302999230965e-07, + "loss": 0.7012, + "step": 3823 + }, + { + "epoch": 0.14, + "learning_rate": 1.861793679276376e-07, + "loss": 0.6949, + "step": 3824 + }, + { + "epoch": 0.14, + "learning_rate": 1.8617570586296553e-07, + "loss": 0.6936, + "step": 3825 + }, + { + "epoch": 0.14, + "learning_rate": 1.8617204379829349e-07, + "loss": 0.694, + "step": 3826 + }, + { + "epoch": 0.14, + "learning_rate": 1.861683817336214e-07, + "loss": 0.6984, + "step": 3827 + }, + { + "epoch": 0.14, + "learning_rate": 1.8616471966894937e-07, + "loss": 0.6951, + "step": 3828 + }, + { + "epoch": 0.14, + "learning_rate": 1.861610576042773e-07, + "loss": 0.7053, + "step": 3829 + }, + { + "epoch": 0.14, + "learning_rate": 1.8615739553960522e-07, + "loss": 0.6818, + "step": 3830 + }, + { + "epoch": 0.14, + "learning_rate": 1.8615373347493315e-07, + "loss": 0.6949, + "step": 3831 + }, + { + "epoch": 0.14, + "learning_rate": 1.861500714102611e-07, + "loss": 0.7027, + "step": 3832 + }, + { + "epoch": 0.14, + "learning_rate": 1.8614640934558902e-07, + "loss": 0.704, + "step": 3833 + }, + { + "epoch": 0.14, + "learning_rate": 1.8614274728091698e-07, + "loss": 0.6855, + "step": 3834 + }, + { + "epoch": 0.14, + "learning_rate": 1.861390852162449e-07, + "loss": 0.6806, + "step": 3835 + }, + { + "epoch": 0.14, + "learning_rate": 1.8613542315157286e-07, + "loss": 0.6859, + "step": 3836 + }, + { + "epoch": 0.14, + "learning_rate": 1.8613176108690078e-07, + "loss": 0.7062, + "step": 3837 + }, + { + "epoch": 0.14, + "learning_rate": 1.861280990222287e-07, + "loss": 0.71, + "step": 3838 + }, + { + "epoch": 0.14, + "learning_rate": 1.8612443695755666e-07, + "loss": 0.7013, + "step": 3839 + }, + { + "epoch": 0.14, + "learning_rate": 1.861207748928846e-07, + "loss": 0.7023, + "step": 3840 + }, + { + "epoch": 0.14, + "learning_rate": 1.8611711282821254e-07, + "loss": 0.6962, + "step": 3841 + }, + { + "epoch": 0.14, + "learning_rate": 1.8611345076354047e-07, + "loss": 0.6871, + "step": 3842 + }, + { + "epoch": 0.14, + "learning_rate": 1.8610978869886842e-07, + "loss": 0.6923, + "step": 3843 + }, + { + "epoch": 0.14, + "learning_rate": 1.8610612663419635e-07, + "loss": 0.6854, + "step": 3844 + }, + { + "epoch": 0.14, + "learning_rate": 1.861024645695243e-07, + "loss": 0.6989, + "step": 3845 + }, + { + "epoch": 0.14, + "learning_rate": 1.8609880250485223e-07, + "loss": 0.6895, + "step": 3846 + }, + { + "epoch": 0.14, + "learning_rate": 1.8609514044018018e-07, + "loss": 0.6863, + "step": 3847 + }, + { + "epoch": 0.14, + "learning_rate": 1.8609147837550808e-07, + "loss": 0.6899, + "step": 3848 + }, + { + "epoch": 0.14, + "learning_rate": 1.8608781631083604e-07, + "loss": 0.6988, + "step": 3849 + }, + { + "epoch": 0.14, + "learning_rate": 1.8608415424616396e-07, + "loss": 0.6827, + "step": 3850 + }, + { + "epoch": 0.14, + "learning_rate": 1.8608049218149192e-07, + "loss": 0.6899, + "step": 3851 + }, + { + "epoch": 0.14, + "learning_rate": 1.8607683011681984e-07, + "loss": 0.6697, + "step": 3852 + }, + { + "epoch": 0.14, + "learning_rate": 1.860731680521478e-07, + "loss": 0.6933, + "step": 3853 + }, + { + "epoch": 0.14, + "learning_rate": 1.8606950598747572e-07, + "loss": 0.705, + "step": 3854 + }, + { + "epoch": 0.14, + "learning_rate": 1.8606584392280368e-07, + "loss": 0.6803, + "step": 3855 + }, + { + "epoch": 0.14, + "learning_rate": 1.860621818581316e-07, + "loss": 0.7025, + "step": 3856 + }, + { + "epoch": 0.14, + "learning_rate": 1.8605851979345956e-07, + "loss": 0.6982, + "step": 3857 + }, + { + "epoch": 0.14, + "learning_rate": 1.8605485772878748e-07, + "loss": 0.6998, + "step": 3858 + }, + { + "epoch": 0.14, + "learning_rate": 1.8605119566411544e-07, + "loss": 0.6905, + "step": 3859 + }, + { + "epoch": 0.14, + "learning_rate": 1.8604753359944336e-07, + "loss": 0.6794, + "step": 3860 + }, + { + "epoch": 0.14, + "learning_rate": 1.860438715347713e-07, + "loss": 0.708, + "step": 3861 + }, + { + "epoch": 0.14, + "learning_rate": 1.8604020947009924e-07, + "loss": 0.7121, + "step": 3862 + }, + { + "epoch": 0.14, + "learning_rate": 1.8603654740542717e-07, + "loss": 0.6936, + "step": 3863 + }, + { + "epoch": 0.14, + "learning_rate": 1.8603288534075512e-07, + "loss": 0.6971, + "step": 3864 + }, + { + "epoch": 0.14, + "learning_rate": 1.8602922327608305e-07, + "loss": 0.6843, + "step": 3865 + }, + { + "epoch": 0.14, + "learning_rate": 1.8602556121141098e-07, + "loss": 0.6892, + "step": 3866 + }, + { + "epoch": 0.14, + "learning_rate": 1.860218991467389e-07, + "loss": 0.7044, + "step": 3867 + }, + { + "epoch": 0.14, + "learning_rate": 1.8601823708206686e-07, + "loss": 0.7004, + "step": 3868 + }, + { + "epoch": 0.14, + "learning_rate": 1.8601457501739478e-07, + "loss": 0.6929, + "step": 3869 + }, + { + "epoch": 0.14, + "learning_rate": 1.8601091295272274e-07, + "loss": 0.6821, + "step": 3870 + }, + { + "epoch": 0.14, + "learning_rate": 1.8600725088805066e-07, + "loss": 0.6687, + "step": 3871 + }, + { + "epoch": 0.14, + "learning_rate": 1.8600358882337862e-07, + "loss": 0.6899, + "step": 3872 + }, + { + "epoch": 0.14, + "learning_rate": 1.8599992675870654e-07, + "loss": 0.6931, + "step": 3873 + }, + { + "epoch": 0.14, + "learning_rate": 1.859962646940345e-07, + "loss": 0.6762, + "step": 3874 + }, + { + "epoch": 0.14, + "learning_rate": 1.8599260262936242e-07, + "loss": 0.7036, + "step": 3875 + }, + { + "epoch": 0.14, + "learning_rate": 1.8598894056469038e-07, + "loss": 0.6908, + "step": 3876 + }, + { + "epoch": 0.14, + "learning_rate": 1.859852785000183e-07, + "loss": 0.7111, + "step": 3877 + }, + { + "epoch": 0.14, + "learning_rate": 1.8598161643534626e-07, + "loss": 0.7132, + "step": 3878 + }, + { + "epoch": 0.14, + "learning_rate": 1.8597795437067418e-07, + "loss": 0.6966, + "step": 3879 + }, + { + "epoch": 0.14, + "learning_rate": 1.8597429230600214e-07, + "loss": 0.6904, + "step": 3880 + }, + { + "epoch": 0.14, + "learning_rate": 1.8597063024133006e-07, + "loss": 0.6967, + "step": 3881 + }, + { + "epoch": 0.14, + "learning_rate": 1.85966968176658e-07, + "loss": 0.6873, + "step": 3882 + }, + { + "epoch": 0.14, + "learning_rate": 1.8596330611198594e-07, + "loss": 0.7041, + "step": 3883 + }, + { + "epoch": 0.14, + "learning_rate": 1.8595964404731387e-07, + "loss": 0.695, + "step": 3884 + }, + { + "epoch": 0.14, + "learning_rate": 1.859559819826418e-07, + "loss": 0.6883, + "step": 3885 + }, + { + "epoch": 0.14, + "learning_rate": 1.8595231991796975e-07, + "loss": 0.6844, + "step": 3886 + }, + { + "epoch": 0.14, + "learning_rate": 1.8594865785329768e-07, + "loss": 0.6913, + "step": 3887 + }, + { + "epoch": 0.14, + "learning_rate": 1.859449957886256e-07, + "loss": 0.6903, + "step": 3888 + }, + { + "epoch": 0.14, + "learning_rate": 1.8594133372395356e-07, + "loss": 0.6899, + "step": 3889 + }, + { + "epoch": 0.14, + "learning_rate": 1.8593767165928148e-07, + "loss": 0.7118, + "step": 3890 + }, + { + "epoch": 0.14, + "learning_rate": 1.8593400959460943e-07, + "loss": 0.7025, + "step": 3891 + }, + { + "epoch": 0.14, + "learning_rate": 1.8593034752993736e-07, + "loss": 0.6839, + "step": 3892 + }, + { + "epoch": 0.14, + "learning_rate": 1.8592668546526531e-07, + "loss": 0.6977, + "step": 3893 + }, + { + "epoch": 0.14, + "learning_rate": 1.8592302340059324e-07, + "loss": 0.7025, + "step": 3894 + }, + { + "epoch": 0.14, + "learning_rate": 1.859193613359212e-07, + "loss": 0.6924, + "step": 3895 + }, + { + "epoch": 0.14, + "learning_rate": 1.8591569927124912e-07, + "loss": 0.7114, + "step": 3896 + }, + { + "epoch": 0.14, + "learning_rate": 1.8591203720657707e-07, + "loss": 0.7149, + "step": 3897 + }, + { + "epoch": 0.14, + "learning_rate": 1.85908375141905e-07, + "loss": 0.7245, + "step": 3898 + }, + { + "epoch": 0.14, + "learning_rate": 1.8590471307723295e-07, + "loss": 0.7019, + "step": 3899 + }, + { + "epoch": 0.14, + "learning_rate": 1.8590105101256088e-07, + "loss": 0.7021, + "step": 3900 + }, + { + "epoch": 0.14, + "learning_rate": 1.858973889478888e-07, + "loss": 0.6894, + "step": 3901 + }, + { + "epoch": 0.14, + "learning_rate": 1.8589372688321673e-07, + "loss": 0.6907, + "step": 3902 + }, + { + "epoch": 0.14, + "learning_rate": 1.858900648185447e-07, + "loss": 0.7069, + "step": 3903 + }, + { + "epoch": 0.14, + "learning_rate": 1.8588640275387261e-07, + "loss": 0.6829, + "step": 3904 + }, + { + "epoch": 0.14, + "learning_rate": 1.8588274068920057e-07, + "loss": 0.6907, + "step": 3905 + }, + { + "epoch": 0.14, + "learning_rate": 1.858790786245285e-07, + "loss": 0.7019, + "step": 3906 + }, + { + "epoch": 0.14, + "learning_rate": 1.8587541655985645e-07, + "loss": 0.7012, + "step": 3907 + }, + { + "epoch": 0.14, + "learning_rate": 1.8587175449518437e-07, + "loss": 0.688, + "step": 3908 + }, + { + "epoch": 0.14, + "learning_rate": 1.858680924305123e-07, + "loss": 0.7074, + "step": 3909 + }, + { + "epoch": 0.14, + "learning_rate": 1.8586443036584025e-07, + "loss": 0.7031, + "step": 3910 + }, + { + "epoch": 0.14, + "learning_rate": 1.8586076830116818e-07, + "loss": 0.6958, + "step": 3911 + }, + { + "epoch": 0.14, + "learning_rate": 1.8585710623649613e-07, + "loss": 0.6856, + "step": 3912 + }, + { + "epoch": 0.14, + "learning_rate": 1.8585344417182406e-07, + "loss": 0.7056, + "step": 3913 + }, + { + "epoch": 0.14, + "learning_rate": 1.8584978210715201e-07, + "loss": 0.6809, + "step": 3914 + }, + { + "epoch": 0.14, + "learning_rate": 1.8584612004247994e-07, + "loss": 0.6852, + "step": 3915 + }, + { + "epoch": 0.14, + "learning_rate": 1.858424579778079e-07, + "loss": 0.7061, + "step": 3916 + }, + { + "epoch": 0.14, + "learning_rate": 1.8583879591313582e-07, + "loss": 0.6971, + "step": 3917 + }, + { + "epoch": 0.14, + "learning_rate": 1.8583513384846377e-07, + "loss": 0.6829, + "step": 3918 + }, + { + "epoch": 0.14, + "learning_rate": 1.8583147178379167e-07, + "loss": 0.7025, + "step": 3919 + }, + { + "epoch": 0.14, + "learning_rate": 1.8582780971911963e-07, + "loss": 0.7139, + "step": 3920 + }, + { + "epoch": 0.14, + "learning_rate": 1.8582414765444755e-07, + "loss": 0.692, + "step": 3921 + }, + { + "epoch": 0.14, + "learning_rate": 1.858204855897755e-07, + "loss": 0.7057, + "step": 3922 + }, + { + "epoch": 0.14, + "learning_rate": 1.8581682352510343e-07, + "loss": 0.6959, + "step": 3923 + }, + { + "epoch": 0.14, + "learning_rate": 1.8581316146043139e-07, + "loss": 0.682, + "step": 3924 + }, + { + "epoch": 0.14, + "learning_rate": 1.858094993957593e-07, + "loss": 0.693, + "step": 3925 + }, + { + "epoch": 0.14, + "learning_rate": 1.8580583733108727e-07, + "loss": 0.6994, + "step": 3926 + }, + { + "epoch": 0.14, + "learning_rate": 1.858021752664152e-07, + "loss": 0.6791, + "step": 3927 + }, + { + "epoch": 0.14, + "learning_rate": 1.8579851320174315e-07, + "loss": 0.6887, + "step": 3928 + }, + { + "epoch": 0.14, + "learning_rate": 1.8579485113707107e-07, + "loss": 0.6847, + "step": 3929 + }, + { + "epoch": 0.14, + "learning_rate": 1.85791189072399e-07, + "loss": 0.6997, + "step": 3930 + }, + { + "epoch": 0.14, + "learning_rate": 1.8578752700772695e-07, + "loss": 0.6971, + "step": 3931 + }, + { + "epoch": 0.14, + "learning_rate": 1.8578386494305488e-07, + "loss": 0.6818, + "step": 3932 + }, + { + "epoch": 0.14, + "learning_rate": 1.8578020287838283e-07, + "loss": 0.6806, + "step": 3933 + }, + { + "epoch": 0.14, + "learning_rate": 1.8577654081371076e-07, + "loss": 0.7041, + "step": 3934 + }, + { + "epoch": 0.14, + "learning_rate": 1.857728787490387e-07, + "loss": 0.7003, + "step": 3935 + }, + { + "epoch": 0.14, + "learning_rate": 1.8576921668436664e-07, + "loss": 0.6935, + "step": 3936 + }, + { + "epoch": 0.14, + "learning_rate": 1.8576555461969457e-07, + "loss": 0.6991, + "step": 3937 + }, + { + "epoch": 0.14, + "learning_rate": 1.857618925550225e-07, + "loss": 0.6973, + "step": 3938 + }, + { + "epoch": 0.14, + "learning_rate": 1.8575823049035045e-07, + "loss": 0.6921, + "step": 3939 + }, + { + "epoch": 0.14, + "learning_rate": 1.8575456842567837e-07, + "loss": 0.6997, + "step": 3940 + }, + { + "epoch": 0.14, + "learning_rate": 1.8575090636100633e-07, + "loss": 0.6917, + "step": 3941 + }, + { + "epoch": 0.14, + "learning_rate": 1.8574724429633425e-07, + "loss": 0.707, + "step": 3942 + }, + { + "epoch": 0.14, + "learning_rate": 1.857435822316622e-07, + "loss": 0.6995, + "step": 3943 + }, + { + "epoch": 0.14, + "learning_rate": 1.8573992016699013e-07, + "loss": 0.6758, + "step": 3944 + }, + { + "epoch": 0.14, + "learning_rate": 1.8573625810231809e-07, + "loss": 0.6893, + "step": 3945 + }, + { + "epoch": 0.14, + "learning_rate": 1.85732596037646e-07, + "loss": 0.6799, + "step": 3946 + }, + { + "epoch": 0.14, + "learning_rate": 1.8572893397297397e-07, + "loss": 0.6895, + "step": 3947 + }, + { + "epoch": 0.14, + "learning_rate": 1.857252719083019e-07, + "loss": 0.6832, + "step": 3948 + }, + { + "epoch": 0.14, + "learning_rate": 1.8572160984362984e-07, + "loss": 0.6957, + "step": 3949 + }, + { + "epoch": 0.14, + "learning_rate": 1.8571794777895777e-07, + "loss": 0.7193, + "step": 3950 + }, + { + "epoch": 0.14, + "learning_rate": 1.8571428571428572e-07, + "loss": 0.6788, + "step": 3951 + }, + { + "epoch": 0.14, + "learning_rate": 1.8571062364961365e-07, + "loss": 0.6913, + "step": 3952 + }, + { + "epoch": 0.14, + "learning_rate": 1.8570696158494158e-07, + "loss": 0.6865, + "step": 3953 + }, + { + "epoch": 0.14, + "learning_rate": 1.8570329952026953e-07, + "loss": 0.7015, + "step": 3954 + }, + { + "epoch": 0.14, + "learning_rate": 1.8569963745559746e-07, + "loss": 0.6905, + "step": 3955 + }, + { + "epoch": 0.14, + "learning_rate": 1.8569597539092538e-07, + "loss": 0.6965, + "step": 3956 + }, + { + "epoch": 0.14, + "learning_rate": 1.8569231332625334e-07, + "loss": 0.7168, + "step": 3957 + }, + { + "epoch": 0.14, + "learning_rate": 1.8568865126158126e-07, + "loss": 0.6919, + "step": 3958 + }, + { + "epoch": 0.14, + "learning_rate": 1.856849891969092e-07, + "loss": 0.6748, + "step": 3959 + }, + { + "epoch": 0.14, + "learning_rate": 1.8568132713223714e-07, + "loss": 0.6964, + "step": 3960 + }, + { + "epoch": 0.14, + "learning_rate": 1.8567766506756507e-07, + "loss": 0.7226, + "step": 3961 + }, + { + "epoch": 0.14, + "learning_rate": 1.8567400300289302e-07, + "loss": 0.6993, + "step": 3962 + }, + { + "epoch": 0.14, + "learning_rate": 1.8567034093822095e-07, + "loss": 0.7059, + "step": 3963 + }, + { + "epoch": 0.15, + "learning_rate": 1.856666788735489e-07, + "loss": 0.6982, + "step": 3964 + }, + { + "epoch": 0.15, + "learning_rate": 1.8566301680887683e-07, + "loss": 0.6859, + "step": 3965 + }, + { + "epoch": 0.15, + "learning_rate": 1.8565935474420478e-07, + "loss": 0.7, + "step": 3966 + }, + { + "epoch": 0.15, + "learning_rate": 1.856556926795327e-07, + "loss": 0.6926, + "step": 3967 + }, + { + "epoch": 0.15, + "learning_rate": 1.8565203061486066e-07, + "loss": 0.7039, + "step": 3968 + }, + { + "epoch": 0.15, + "learning_rate": 1.856483685501886e-07, + "loss": 0.699, + "step": 3969 + }, + { + "epoch": 0.15, + "learning_rate": 1.8564470648551654e-07, + "loss": 0.6792, + "step": 3970 + }, + { + "epoch": 0.15, + "learning_rate": 1.8564104442084447e-07, + "loss": 0.6934, + "step": 3971 + }, + { + "epoch": 0.15, + "learning_rate": 1.8563738235617242e-07, + "loss": 0.7017, + "step": 3972 + }, + { + "epoch": 0.15, + "learning_rate": 1.8563372029150032e-07, + "loss": 0.6829, + "step": 3973 + }, + { + "epoch": 0.15, + "learning_rate": 1.8563005822682828e-07, + "loss": 0.6761, + "step": 3974 + }, + { + "epoch": 0.15, + "learning_rate": 1.856263961621562e-07, + "loss": 0.6652, + "step": 3975 + }, + { + "epoch": 0.15, + "learning_rate": 1.8562273409748416e-07, + "loss": 0.6888, + "step": 3976 + }, + { + "epoch": 0.15, + "learning_rate": 1.8561907203281208e-07, + "loss": 0.6759, + "step": 3977 + }, + { + "epoch": 0.15, + "learning_rate": 1.8561540996814004e-07, + "loss": 0.7063, + "step": 3978 + }, + { + "epoch": 0.15, + "learning_rate": 1.8561174790346796e-07, + "loss": 0.6852, + "step": 3979 + }, + { + "epoch": 0.15, + "learning_rate": 1.856080858387959e-07, + "loss": 0.6901, + "step": 3980 + }, + { + "epoch": 0.15, + "learning_rate": 1.8560442377412384e-07, + "loss": 0.6849, + "step": 3981 + }, + { + "epoch": 0.15, + "learning_rate": 1.8560076170945177e-07, + "loss": 0.6902, + "step": 3982 + }, + { + "epoch": 0.15, + "learning_rate": 1.8559709964477972e-07, + "loss": 0.6867, + "step": 3983 + }, + { + "epoch": 0.15, + "learning_rate": 1.8559343758010765e-07, + "loss": 0.6897, + "step": 3984 + }, + { + "epoch": 0.15, + "learning_rate": 1.855897755154356e-07, + "loss": 0.6787, + "step": 3985 + }, + { + "epoch": 0.15, + "learning_rate": 1.8558611345076353e-07, + "loss": 0.6928, + "step": 3986 + }, + { + "epoch": 0.15, + "learning_rate": 1.8558245138609148e-07, + "loss": 0.7053, + "step": 3987 + }, + { + "epoch": 0.15, + "learning_rate": 1.855787893214194e-07, + "loss": 0.6989, + "step": 3988 + }, + { + "epoch": 0.15, + "learning_rate": 1.8557512725674736e-07, + "loss": 0.6953, + "step": 3989 + }, + { + "epoch": 0.15, + "learning_rate": 1.855714651920753e-07, + "loss": 0.7142, + "step": 3990 + }, + { + "epoch": 0.15, + "learning_rate": 1.8556780312740322e-07, + "loss": 0.7028, + "step": 3991 + }, + { + "epoch": 0.15, + "learning_rate": 1.8556414106273114e-07, + "loss": 0.7079, + "step": 3992 + }, + { + "epoch": 0.15, + "learning_rate": 1.855604789980591e-07, + "loss": 0.7013, + "step": 3993 + }, + { + "epoch": 0.15, + "learning_rate": 1.8555681693338702e-07, + "loss": 0.6881, + "step": 3994 + }, + { + "epoch": 0.15, + "learning_rate": 1.8555315486871498e-07, + "loss": 0.6847, + "step": 3995 + }, + { + "epoch": 0.15, + "learning_rate": 1.855494928040429e-07, + "loss": 0.7213, + "step": 3996 + }, + { + "epoch": 0.15, + "learning_rate": 1.8554583073937086e-07, + "loss": 0.697, + "step": 3997 + }, + { + "epoch": 0.15, + "learning_rate": 1.8554216867469878e-07, + "loss": 0.6925, + "step": 3998 + }, + { + "epoch": 0.15, + "learning_rate": 1.8553850661002674e-07, + "loss": 0.6753, + "step": 3999 + }, + { + "epoch": 0.15, + "learning_rate": 1.8553484454535466e-07, + "loss": 0.6813, + "step": 4000 + }, + { + "epoch": 0.15, + "learning_rate": 1.855311824806826e-07, + "loss": 0.7054, + "step": 4001 + }, + { + "epoch": 0.15, + "learning_rate": 1.8552752041601054e-07, + "loss": 0.6942, + "step": 4002 + }, + { + "epoch": 0.15, + "learning_rate": 1.8552385835133847e-07, + "loss": 0.6746, + "step": 4003 + }, + { + "epoch": 0.15, + "learning_rate": 1.8552019628666642e-07, + "loss": 0.6896, + "step": 4004 + }, + { + "epoch": 0.15, + "learning_rate": 1.8551653422199435e-07, + "loss": 0.6688, + "step": 4005 + }, + { + "epoch": 0.15, + "learning_rate": 1.855128721573223e-07, + "loss": 0.6994, + "step": 4006 + }, + { + "epoch": 0.15, + "learning_rate": 1.8550921009265023e-07, + "loss": 0.7023, + "step": 4007 + }, + { + "epoch": 0.15, + "learning_rate": 1.8550554802797818e-07, + "loss": 0.7016, + "step": 4008 + }, + { + "epoch": 0.15, + "learning_rate": 1.8550188596330608e-07, + "loss": 0.6779, + "step": 4009 + }, + { + "epoch": 0.15, + "learning_rate": 1.8549822389863403e-07, + "loss": 0.6952, + "step": 4010 + }, + { + "epoch": 0.15, + "learning_rate": 1.8549456183396196e-07, + "loss": 0.7024, + "step": 4011 + }, + { + "epoch": 0.15, + "learning_rate": 1.8549089976928991e-07, + "loss": 0.7017, + "step": 4012 + }, + { + "epoch": 0.15, + "learning_rate": 1.8548723770461784e-07, + "loss": 0.7074, + "step": 4013 + }, + { + "epoch": 0.15, + "learning_rate": 1.854835756399458e-07, + "loss": 0.6977, + "step": 4014 + }, + { + "epoch": 0.15, + "learning_rate": 1.8547991357527372e-07, + "loss": 0.7123, + "step": 4015 + }, + { + "epoch": 0.15, + "learning_rate": 1.8547625151060167e-07, + "loss": 0.7021, + "step": 4016 + }, + { + "epoch": 0.15, + "learning_rate": 1.854725894459296e-07, + "loss": 0.7081, + "step": 4017 + }, + { + "epoch": 0.15, + "learning_rate": 1.8546892738125755e-07, + "loss": 0.7051, + "step": 4018 + }, + { + "epoch": 0.15, + "learning_rate": 1.8546526531658548e-07, + "loss": 0.6849, + "step": 4019 + }, + { + "epoch": 0.15, + "learning_rate": 1.8546160325191343e-07, + "loss": 0.6903, + "step": 4020 + }, + { + "epoch": 0.15, + "learning_rate": 1.8545794118724136e-07, + "loss": 0.7149, + "step": 4021 + }, + { + "epoch": 0.15, + "learning_rate": 1.8545427912256931e-07, + "loss": 0.7065, + "step": 4022 + }, + { + "epoch": 0.15, + "learning_rate": 1.8545061705789724e-07, + "loss": 0.7011, + "step": 4023 + }, + { + "epoch": 0.15, + "learning_rate": 1.8544695499322517e-07, + "loss": 0.6769, + "step": 4024 + }, + { + "epoch": 0.15, + "learning_rate": 1.8544329292855312e-07, + "loss": 0.7009, + "step": 4025 + }, + { + "epoch": 0.15, + "learning_rate": 1.8543963086388105e-07, + "loss": 0.6988, + "step": 4026 + }, + { + "epoch": 0.15, + "learning_rate": 1.8543596879920897e-07, + "loss": 0.7074, + "step": 4027 + }, + { + "epoch": 0.15, + "learning_rate": 1.8543230673453693e-07, + "loss": 0.6831, + "step": 4028 + }, + { + "epoch": 0.15, + "learning_rate": 1.8542864466986485e-07, + "loss": 0.7039, + "step": 4029 + }, + { + "epoch": 0.15, + "learning_rate": 1.8542498260519278e-07, + "loss": 0.694, + "step": 4030 + }, + { + "epoch": 0.15, + "learning_rate": 1.8542132054052073e-07, + "loss": 0.7005, + "step": 4031 + }, + { + "epoch": 0.15, + "learning_rate": 1.8541765847584866e-07, + "loss": 0.675, + "step": 4032 + }, + { + "epoch": 0.15, + "learning_rate": 1.8541399641117661e-07, + "loss": 0.6905, + "step": 4033 + }, + { + "epoch": 0.15, + "learning_rate": 1.8541033434650454e-07, + "loss": 0.685, + "step": 4034 + }, + { + "epoch": 0.15, + "learning_rate": 1.854066722818325e-07, + "loss": 0.6817, + "step": 4035 + }, + { + "epoch": 0.15, + "learning_rate": 1.8540301021716042e-07, + "loss": 0.695, + "step": 4036 + }, + { + "epoch": 0.15, + "learning_rate": 1.8539934815248837e-07, + "loss": 0.6708, + "step": 4037 + }, + { + "epoch": 0.15, + "learning_rate": 1.853956860878163e-07, + "loss": 0.7193, + "step": 4038 + }, + { + "epoch": 0.15, + "learning_rate": 1.8539202402314425e-07, + "loss": 0.69, + "step": 4039 + }, + { + "epoch": 0.15, + "learning_rate": 1.8538836195847218e-07, + "loss": 0.7156, + "step": 4040 + }, + { + "epoch": 0.15, + "learning_rate": 1.8538469989380013e-07, + "loss": 0.7049, + "step": 4041 + }, + { + "epoch": 0.15, + "learning_rate": 1.8538103782912806e-07, + "loss": 0.7085, + "step": 4042 + }, + { + "epoch": 0.15, + "learning_rate": 1.85377375764456e-07, + "loss": 0.6976, + "step": 4043 + }, + { + "epoch": 0.15, + "learning_rate": 1.8537371369978394e-07, + "loss": 0.6861, + "step": 4044 + }, + { + "epoch": 0.15, + "learning_rate": 1.8537005163511187e-07, + "loss": 0.6924, + "step": 4045 + }, + { + "epoch": 0.15, + "learning_rate": 1.853663895704398e-07, + "loss": 0.6989, + "step": 4046 + }, + { + "epoch": 0.15, + "learning_rate": 1.8536272750576775e-07, + "loss": 0.6896, + "step": 4047 + }, + { + "epoch": 0.15, + "learning_rate": 1.8535906544109567e-07, + "loss": 0.6898, + "step": 4048 + }, + { + "epoch": 0.15, + "learning_rate": 1.8535540337642363e-07, + "loss": 0.7093, + "step": 4049 + }, + { + "epoch": 0.15, + "learning_rate": 1.8535174131175155e-07, + "loss": 0.7021, + "step": 4050 + }, + { + "epoch": 0.15, + "learning_rate": 1.8534807924707948e-07, + "loss": 0.7005, + "step": 4051 + }, + { + "epoch": 0.15, + "learning_rate": 1.8534441718240743e-07, + "loss": 0.6873, + "step": 4052 + }, + { + "epoch": 0.15, + "learning_rate": 1.8534075511773536e-07, + "loss": 0.6848, + "step": 4053 + }, + { + "epoch": 0.15, + "learning_rate": 1.853370930530633e-07, + "loss": 0.6919, + "step": 4054 + }, + { + "epoch": 0.15, + "learning_rate": 1.8533343098839124e-07, + "loss": 0.6933, + "step": 4055 + }, + { + "epoch": 0.15, + "learning_rate": 1.853297689237192e-07, + "loss": 0.7042, + "step": 4056 + }, + { + "epoch": 0.15, + "learning_rate": 1.8532610685904712e-07, + "loss": 0.6891, + "step": 4057 + }, + { + "epoch": 0.15, + "learning_rate": 1.8532244479437507e-07, + "loss": 0.689, + "step": 4058 + }, + { + "epoch": 0.15, + "learning_rate": 1.85318782729703e-07, + "loss": 0.7119, + "step": 4059 + }, + { + "epoch": 0.15, + "learning_rate": 1.8531512066503095e-07, + "loss": 0.688, + "step": 4060 + }, + { + "epoch": 0.15, + "learning_rate": 1.8531145860035888e-07, + "loss": 0.6962, + "step": 4061 + }, + { + "epoch": 0.15, + "learning_rate": 1.8530779653568683e-07, + "loss": 0.6799, + "step": 4062 + }, + { + "epoch": 0.15, + "learning_rate": 1.8530413447101473e-07, + "loss": 0.6994, + "step": 4063 + }, + { + "epoch": 0.15, + "learning_rate": 1.8530047240634269e-07, + "loss": 0.6728, + "step": 4064 + }, + { + "epoch": 0.15, + "learning_rate": 1.852968103416706e-07, + "loss": 0.696, + "step": 4065 + }, + { + "epoch": 0.15, + "learning_rate": 1.8529314827699857e-07, + "loss": 0.6698, + "step": 4066 + }, + { + "epoch": 0.15, + "learning_rate": 1.852894862123265e-07, + "loss": 0.6716, + "step": 4067 + }, + { + "epoch": 0.15, + "learning_rate": 1.8528582414765444e-07, + "loss": 0.7056, + "step": 4068 + }, + { + "epoch": 0.15, + "learning_rate": 1.8528216208298237e-07, + "loss": 0.6963, + "step": 4069 + }, + { + "epoch": 0.15, + "learning_rate": 1.8527850001831032e-07, + "loss": 0.6979, + "step": 4070 + }, + { + "epoch": 0.15, + "learning_rate": 1.8527483795363825e-07, + "loss": 0.6913, + "step": 4071 + }, + { + "epoch": 0.15, + "learning_rate": 1.8527117588896618e-07, + "loss": 0.6926, + "step": 4072 + }, + { + "epoch": 0.15, + "learning_rate": 1.8526751382429413e-07, + "loss": 0.7048, + "step": 4073 + }, + { + "epoch": 0.15, + "learning_rate": 1.8526385175962206e-07, + "loss": 0.7013, + "step": 4074 + }, + { + "epoch": 0.15, + "learning_rate": 1.8526018969495e-07, + "loss": 0.6949, + "step": 4075 + }, + { + "epoch": 0.15, + "learning_rate": 1.8525652763027794e-07, + "loss": 0.6817, + "step": 4076 + }, + { + "epoch": 0.15, + "learning_rate": 1.852528655656059e-07, + "loss": 0.6954, + "step": 4077 + }, + { + "epoch": 0.15, + "learning_rate": 1.8524920350093382e-07, + "loss": 0.6909, + "step": 4078 + }, + { + "epoch": 0.15, + "learning_rate": 1.8524554143626177e-07, + "loss": 0.6917, + "step": 4079 + }, + { + "epoch": 0.15, + "learning_rate": 1.852418793715897e-07, + "loss": 0.6849, + "step": 4080 + }, + { + "epoch": 0.15, + "learning_rate": 1.8523821730691762e-07, + "loss": 0.6672, + "step": 4081 + }, + { + "epoch": 0.15, + "learning_rate": 1.8523455524224555e-07, + "loss": 0.6905, + "step": 4082 + }, + { + "epoch": 0.15, + "learning_rate": 1.852308931775735e-07, + "loss": 0.6916, + "step": 4083 + }, + { + "epoch": 0.15, + "learning_rate": 1.8522723111290143e-07, + "loss": 0.6708, + "step": 4084 + }, + { + "epoch": 0.15, + "learning_rate": 1.8522356904822938e-07, + "loss": 0.6753, + "step": 4085 + }, + { + "epoch": 0.15, + "learning_rate": 1.852199069835573e-07, + "loss": 0.6963, + "step": 4086 + }, + { + "epoch": 0.15, + "learning_rate": 1.8521624491888526e-07, + "loss": 0.6868, + "step": 4087 + }, + { + "epoch": 0.15, + "learning_rate": 1.852125828542132e-07, + "loss": 0.6936, + "step": 4088 + }, + { + "epoch": 0.15, + "learning_rate": 1.8520892078954114e-07, + "loss": 0.6749, + "step": 4089 + }, + { + "epoch": 0.15, + "learning_rate": 1.8520525872486907e-07, + "loss": 0.6784, + "step": 4090 + }, + { + "epoch": 0.15, + "learning_rate": 1.8520159666019702e-07, + "loss": 0.6869, + "step": 4091 + }, + { + "epoch": 0.15, + "learning_rate": 1.8519793459552495e-07, + "loss": 0.6772, + "step": 4092 + }, + { + "epoch": 0.15, + "learning_rate": 1.851942725308529e-07, + "loss": 0.7067, + "step": 4093 + }, + { + "epoch": 0.15, + "learning_rate": 1.8519061046618083e-07, + "loss": 0.686, + "step": 4094 + }, + { + "epoch": 0.15, + "learning_rate": 1.8518694840150876e-07, + "loss": 0.6923, + "step": 4095 + }, + { + "epoch": 0.15, + "learning_rate": 1.851832863368367e-07, + "loss": 0.6885, + "step": 4096 + }, + { + "epoch": 0.15, + "learning_rate": 1.8517962427216464e-07, + "loss": 0.6942, + "step": 4097 + }, + { + "epoch": 0.15, + "learning_rate": 1.851759622074926e-07, + "loss": 0.7154, + "step": 4098 + }, + { + "epoch": 0.15, + "learning_rate": 1.8517230014282052e-07, + "loss": 0.6911, + "step": 4099 + }, + { + "epoch": 0.15, + "learning_rate": 1.8516863807814844e-07, + "loss": 0.6754, + "step": 4100 + }, + { + "epoch": 0.15, + "learning_rate": 1.8516497601347637e-07, + "loss": 0.6859, + "step": 4101 + }, + { + "epoch": 0.15, + "learning_rate": 1.8516131394880432e-07, + "loss": 0.6836, + "step": 4102 + }, + { + "epoch": 0.15, + "learning_rate": 1.8515765188413225e-07, + "loss": 0.7054, + "step": 4103 + }, + { + "epoch": 0.15, + "learning_rate": 1.851539898194602e-07, + "loss": 0.674, + "step": 4104 + }, + { + "epoch": 0.15, + "learning_rate": 1.8515032775478813e-07, + "loss": 0.7058, + "step": 4105 + }, + { + "epoch": 0.15, + "learning_rate": 1.8514666569011608e-07, + "loss": 0.7043, + "step": 4106 + }, + { + "epoch": 0.15, + "learning_rate": 1.85143003625444e-07, + "loss": 0.7045, + "step": 4107 + }, + { + "epoch": 0.15, + "learning_rate": 1.8513934156077196e-07, + "loss": 0.7022, + "step": 4108 + }, + { + "epoch": 0.15, + "learning_rate": 1.851356794960999e-07, + "loss": 0.6972, + "step": 4109 + }, + { + "epoch": 0.15, + "learning_rate": 1.8513201743142784e-07, + "loss": 0.7035, + "step": 4110 + }, + { + "epoch": 0.15, + "learning_rate": 1.8512835536675577e-07, + "loss": 0.6992, + "step": 4111 + }, + { + "epoch": 0.15, + "learning_rate": 1.8512469330208372e-07, + "loss": 0.6698, + "step": 4112 + }, + { + "epoch": 0.15, + "learning_rate": 1.8512103123741165e-07, + "loss": 0.7084, + "step": 4113 + }, + { + "epoch": 0.15, + "learning_rate": 1.851173691727396e-07, + "loss": 0.6599, + "step": 4114 + }, + { + "epoch": 0.15, + "learning_rate": 1.8511370710806753e-07, + "loss": 0.6947, + "step": 4115 + }, + { + "epoch": 0.15, + "learning_rate": 1.8511004504339546e-07, + "loss": 0.6872, + "step": 4116 + }, + { + "epoch": 0.15, + "learning_rate": 1.8510638297872338e-07, + "loss": 0.681, + "step": 4117 + }, + { + "epoch": 0.15, + "learning_rate": 1.8510272091405134e-07, + "loss": 0.7197, + "step": 4118 + }, + { + "epoch": 0.15, + "learning_rate": 1.8509905884937926e-07, + "loss": 0.7103, + "step": 4119 + }, + { + "epoch": 0.15, + "learning_rate": 1.8509539678470722e-07, + "loss": 0.6896, + "step": 4120 + }, + { + "epoch": 0.15, + "learning_rate": 1.8509173472003514e-07, + "loss": 0.6831, + "step": 4121 + }, + { + "epoch": 0.15, + "learning_rate": 1.8508807265536307e-07, + "loss": 0.6971, + "step": 4122 + }, + { + "epoch": 0.15, + "learning_rate": 1.8508441059069102e-07, + "loss": 0.7074, + "step": 4123 + }, + { + "epoch": 0.15, + "learning_rate": 1.8508074852601895e-07, + "loss": 0.7004, + "step": 4124 + }, + { + "epoch": 0.15, + "learning_rate": 1.850770864613469e-07, + "loss": 0.6867, + "step": 4125 + }, + { + "epoch": 0.15, + "learning_rate": 1.8507342439667483e-07, + "loss": 0.6905, + "step": 4126 + }, + { + "epoch": 0.15, + "learning_rate": 1.8506976233200278e-07, + "loss": 0.6678, + "step": 4127 + }, + { + "epoch": 0.15, + "learning_rate": 1.850661002673307e-07, + "loss": 0.6838, + "step": 4128 + }, + { + "epoch": 0.15, + "learning_rate": 1.8506243820265866e-07, + "loss": 0.7084, + "step": 4129 + }, + { + "epoch": 0.15, + "learning_rate": 1.850587761379866e-07, + "loss": 0.6989, + "step": 4130 + }, + { + "epoch": 0.15, + "learning_rate": 1.8505511407331454e-07, + "loss": 0.7166, + "step": 4131 + }, + { + "epoch": 0.15, + "learning_rate": 1.8505145200864247e-07, + "loss": 0.7045, + "step": 4132 + }, + { + "epoch": 0.15, + "learning_rate": 1.8504778994397042e-07, + "loss": 0.6761, + "step": 4133 + }, + { + "epoch": 0.15, + "learning_rate": 1.8504412787929835e-07, + "loss": 0.6777, + "step": 4134 + }, + { + "epoch": 0.15, + "learning_rate": 1.8504046581462627e-07, + "loss": 0.7058, + "step": 4135 + }, + { + "epoch": 0.15, + "learning_rate": 1.850368037499542e-07, + "loss": 0.6949, + "step": 4136 + }, + { + "epoch": 0.15, + "learning_rate": 1.8503314168528215e-07, + "loss": 0.7053, + "step": 4137 + }, + { + "epoch": 0.15, + "learning_rate": 1.8502947962061008e-07, + "loss": 0.7217, + "step": 4138 + }, + { + "epoch": 0.15, + "learning_rate": 1.8502581755593803e-07, + "loss": 0.6971, + "step": 4139 + }, + { + "epoch": 0.15, + "learning_rate": 1.8502215549126596e-07, + "loss": 0.7082, + "step": 4140 + }, + { + "epoch": 0.15, + "learning_rate": 1.8501849342659391e-07, + "loss": 0.718, + "step": 4141 + }, + { + "epoch": 0.15, + "learning_rate": 1.8501483136192184e-07, + "loss": 0.7048, + "step": 4142 + }, + { + "epoch": 0.15, + "learning_rate": 1.8501116929724977e-07, + "loss": 0.6801, + "step": 4143 + }, + { + "epoch": 0.15, + "learning_rate": 1.8500750723257772e-07, + "loss": 0.6954, + "step": 4144 + }, + { + "epoch": 0.15, + "learning_rate": 1.8500384516790565e-07, + "loss": 0.6949, + "step": 4145 + }, + { + "epoch": 0.15, + "learning_rate": 1.850001831032336e-07, + "loss": 0.6883, + "step": 4146 + }, + { + "epoch": 0.15, + "learning_rate": 1.8499652103856153e-07, + "loss": 0.6891, + "step": 4147 + }, + { + "epoch": 0.15, + "learning_rate": 1.8499285897388948e-07, + "loss": 0.6848, + "step": 4148 + }, + { + "epoch": 0.15, + "learning_rate": 1.849891969092174e-07, + "loss": 0.6969, + "step": 4149 + }, + { + "epoch": 0.15, + "learning_rate": 1.8498553484454536e-07, + "loss": 0.7305, + "step": 4150 + }, + { + "epoch": 0.15, + "learning_rate": 1.849818727798733e-07, + "loss": 0.6875, + "step": 4151 + }, + { + "epoch": 0.15, + "learning_rate": 1.8497821071520124e-07, + "loss": 0.6895, + "step": 4152 + }, + { + "epoch": 0.15, + "learning_rate": 1.8497454865052914e-07, + "loss": 0.6855, + "step": 4153 + }, + { + "epoch": 0.15, + "learning_rate": 1.849708865858571e-07, + "loss": 0.6879, + "step": 4154 + }, + { + "epoch": 0.15, + "learning_rate": 1.8496722452118502e-07, + "loss": 0.7267, + "step": 4155 + }, + { + "epoch": 0.15, + "learning_rate": 1.8496356245651297e-07, + "loss": 0.6896, + "step": 4156 + }, + { + "epoch": 0.15, + "learning_rate": 1.849599003918409e-07, + "loss": 0.697, + "step": 4157 + }, + { + "epoch": 0.15, + "learning_rate": 1.8495623832716885e-07, + "loss": 0.6984, + "step": 4158 + }, + { + "epoch": 0.15, + "learning_rate": 1.8495257626249678e-07, + "loss": 0.6834, + "step": 4159 + }, + { + "epoch": 0.15, + "learning_rate": 1.8494891419782473e-07, + "loss": 0.7052, + "step": 4160 + }, + { + "epoch": 0.15, + "learning_rate": 1.8494525213315266e-07, + "loss": 0.7293, + "step": 4161 + }, + { + "epoch": 0.15, + "learning_rate": 1.849415900684806e-07, + "loss": 0.6869, + "step": 4162 + }, + { + "epoch": 0.15, + "learning_rate": 1.8493792800380854e-07, + "loss": 0.6949, + "step": 4163 + }, + { + "epoch": 0.15, + "learning_rate": 1.849342659391365e-07, + "loss": 0.6984, + "step": 4164 + }, + { + "epoch": 0.15, + "learning_rate": 1.8493060387446442e-07, + "loss": 0.6875, + "step": 4165 + }, + { + "epoch": 0.15, + "learning_rate": 1.8492694180979235e-07, + "loss": 0.7089, + "step": 4166 + }, + { + "epoch": 0.15, + "learning_rate": 1.849232797451203e-07, + "loss": 0.6839, + "step": 4167 + }, + { + "epoch": 0.15, + "learning_rate": 1.8491961768044823e-07, + "loss": 0.6812, + "step": 4168 + }, + { + "epoch": 0.15, + "learning_rate": 1.8491595561577618e-07, + "loss": 0.6878, + "step": 4169 + }, + { + "epoch": 0.15, + "learning_rate": 1.849122935511041e-07, + "loss": 0.7027, + "step": 4170 + }, + { + "epoch": 0.15, + "learning_rate": 1.8490863148643203e-07, + "loss": 0.6931, + "step": 4171 + }, + { + "epoch": 0.15, + "learning_rate": 1.8490496942175996e-07, + "loss": 0.7141, + "step": 4172 + }, + { + "epoch": 0.15, + "learning_rate": 1.849013073570879e-07, + "loss": 0.6978, + "step": 4173 + }, + { + "epoch": 0.15, + "learning_rate": 1.8489764529241584e-07, + "loss": 0.6793, + "step": 4174 + }, + { + "epoch": 0.15, + "learning_rate": 1.848939832277438e-07, + "loss": 0.6948, + "step": 4175 + }, + { + "epoch": 0.15, + "learning_rate": 1.8489032116307172e-07, + "loss": 0.6797, + "step": 4176 + }, + { + "epoch": 0.15, + "learning_rate": 1.8488665909839967e-07, + "loss": 0.6885, + "step": 4177 + }, + { + "epoch": 0.15, + "learning_rate": 1.848829970337276e-07, + "loss": 0.6809, + "step": 4178 + }, + { + "epoch": 0.15, + "learning_rate": 1.8487933496905555e-07, + "loss": 0.7042, + "step": 4179 + }, + { + "epoch": 0.15, + "learning_rate": 1.8487567290438348e-07, + "loss": 0.6982, + "step": 4180 + }, + { + "epoch": 0.15, + "learning_rate": 1.8487201083971143e-07, + "loss": 0.6961, + "step": 4181 + }, + { + "epoch": 0.15, + "learning_rate": 1.8486834877503936e-07, + "loss": 0.6949, + "step": 4182 + }, + { + "epoch": 0.15, + "learning_rate": 1.848646867103673e-07, + "loss": 0.6896, + "step": 4183 + }, + { + "epoch": 0.15, + "learning_rate": 1.8486102464569524e-07, + "loss": 0.7056, + "step": 4184 + }, + { + "epoch": 0.15, + "learning_rate": 1.848573625810232e-07, + "loss": 0.6856, + "step": 4185 + }, + { + "epoch": 0.15, + "learning_rate": 1.8485370051635112e-07, + "loss": 0.699, + "step": 4186 + }, + { + "epoch": 0.15, + "learning_rate": 1.8485003845167905e-07, + "loss": 0.6949, + "step": 4187 + }, + { + "epoch": 0.15, + "learning_rate": 1.84846376387007e-07, + "loss": 0.6847, + "step": 4188 + }, + { + "epoch": 0.15, + "learning_rate": 1.8484271432233492e-07, + "loss": 0.6972, + "step": 4189 + }, + { + "epoch": 0.15, + "learning_rate": 1.8483905225766285e-07, + "loss": 0.6797, + "step": 4190 + }, + { + "epoch": 0.15, + "learning_rate": 1.848353901929908e-07, + "loss": 0.6928, + "step": 4191 + }, + { + "epoch": 0.15, + "learning_rate": 1.8483172812831873e-07, + "loss": 0.6854, + "step": 4192 + }, + { + "epoch": 0.15, + "learning_rate": 1.8482806606364666e-07, + "loss": 0.6903, + "step": 4193 + }, + { + "epoch": 0.15, + "learning_rate": 1.848244039989746e-07, + "loss": 0.6977, + "step": 4194 + }, + { + "epoch": 0.15, + "learning_rate": 1.8482074193430254e-07, + "loss": 0.6765, + "step": 4195 + }, + { + "epoch": 0.15, + "learning_rate": 1.848170798696305e-07, + "loss": 0.69, + "step": 4196 + }, + { + "epoch": 0.15, + "learning_rate": 1.8481341780495842e-07, + "loss": 0.7053, + "step": 4197 + }, + { + "epoch": 0.15, + "learning_rate": 1.8480975574028637e-07, + "loss": 0.6925, + "step": 4198 + }, + { + "epoch": 0.15, + "learning_rate": 1.848060936756143e-07, + "loss": 0.708, + "step": 4199 + }, + { + "epoch": 0.15, + "learning_rate": 1.8480243161094225e-07, + "loss": 0.6923, + "step": 4200 + }, + { + "epoch": 0.15, + "learning_rate": 1.8479876954627018e-07, + "loss": 0.7216, + "step": 4201 + }, + { + "epoch": 0.15, + "learning_rate": 1.8479510748159813e-07, + "loss": 0.6864, + "step": 4202 + }, + { + "epoch": 0.15, + "learning_rate": 1.8479144541692606e-07, + "loss": 0.6691, + "step": 4203 + }, + { + "epoch": 0.15, + "learning_rate": 1.84787783352254e-07, + "loss": 0.6966, + "step": 4204 + }, + { + "epoch": 0.15, + "learning_rate": 1.8478412128758194e-07, + "loss": 0.7193, + "step": 4205 + }, + { + "epoch": 0.15, + "learning_rate": 1.847804592229099e-07, + "loss": 0.7165, + "step": 4206 + }, + { + "epoch": 0.15, + "learning_rate": 1.847767971582378e-07, + "loss": 0.6973, + "step": 4207 + }, + { + "epoch": 0.15, + "learning_rate": 1.8477313509356574e-07, + "loss": 0.7037, + "step": 4208 + }, + { + "epoch": 0.15, + "learning_rate": 1.8476947302889367e-07, + "loss": 0.6975, + "step": 4209 + }, + { + "epoch": 0.15, + "learning_rate": 1.8476581096422162e-07, + "loss": 0.7048, + "step": 4210 + }, + { + "epoch": 0.15, + "learning_rate": 1.8476214889954955e-07, + "loss": 0.6755, + "step": 4211 + }, + { + "epoch": 0.15, + "learning_rate": 1.847584868348775e-07, + "loss": 0.73, + "step": 4212 + }, + { + "epoch": 0.15, + "learning_rate": 1.8475482477020543e-07, + "loss": 0.6922, + "step": 4213 + }, + { + "epoch": 0.15, + "learning_rate": 1.8475116270553336e-07, + "loss": 0.6911, + "step": 4214 + }, + { + "epoch": 0.15, + "learning_rate": 1.847475006408613e-07, + "loss": 0.7029, + "step": 4215 + }, + { + "epoch": 0.15, + "learning_rate": 1.8474383857618924e-07, + "loss": 0.6801, + "step": 4216 + }, + { + "epoch": 0.15, + "learning_rate": 1.847401765115172e-07, + "loss": 0.6982, + "step": 4217 + }, + { + "epoch": 0.15, + "learning_rate": 1.8473651444684512e-07, + "loss": 0.691, + "step": 4218 + }, + { + "epoch": 0.15, + "learning_rate": 1.8473285238217307e-07, + "loss": 0.6744, + "step": 4219 + }, + { + "epoch": 0.15, + "learning_rate": 1.84729190317501e-07, + "loss": 0.7079, + "step": 4220 + }, + { + "epoch": 0.15, + "learning_rate": 1.8472552825282895e-07, + "loss": 0.7036, + "step": 4221 + }, + { + "epoch": 0.15, + "learning_rate": 1.8472186618815688e-07, + "loss": 0.7199, + "step": 4222 + }, + { + "epoch": 0.15, + "learning_rate": 1.8471820412348483e-07, + "loss": 0.7032, + "step": 4223 + }, + { + "epoch": 0.15, + "learning_rate": 1.8471454205881276e-07, + "loss": 0.6885, + "step": 4224 + }, + { + "epoch": 0.15, + "learning_rate": 1.8471087999414068e-07, + "loss": 0.696, + "step": 4225 + }, + { + "epoch": 0.15, + "learning_rate": 1.847072179294686e-07, + "loss": 0.6877, + "step": 4226 + }, + { + "epoch": 0.15, + "learning_rate": 1.8470355586479656e-07, + "loss": 0.7062, + "step": 4227 + }, + { + "epoch": 0.15, + "learning_rate": 1.846998938001245e-07, + "loss": 0.6827, + "step": 4228 + }, + { + "epoch": 0.15, + "learning_rate": 1.8469623173545244e-07, + "loss": 0.6989, + "step": 4229 + }, + { + "epoch": 0.15, + "learning_rate": 1.8469256967078037e-07, + "loss": 0.6979, + "step": 4230 + }, + { + "epoch": 0.15, + "learning_rate": 1.8468890760610832e-07, + "loss": 0.6946, + "step": 4231 + }, + { + "epoch": 0.15, + "learning_rate": 1.8468524554143625e-07, + "loss": 0.7056, + "step": 4232 + }, + { + "epoch": 0.15, + "learning_rate": 1.846815834767642e-07, + "loss": 0.6935, + "step": 4233 + }, + { + "epoch": 0.15, + "learning_rate": 1.8467792141209213e-07, + "loss": 0.7135, + "step": 4234 + }, + { + "epoch": 0.15, + "learning_rate": 1.8467425934742008e-07, + "loss": 0.702, + "step": 4235 + }, + { + "epoch": 0.15, + "learning_rate": 1.84670597282748e-07, + "loss": 0.7026, + "step": 4236 + }, + { + "epoch": 0.16, + "learning_rate": 1.8466693521807594e-07, + "loss": 0.6808, + "step": 4237 + }, + { + "epoch": 0.16, + "learning_rate": 1.846632731534039e-07, + "loss": 0.6728, + "step": 4238 + }, + { + "epoch": 0.16, + "learning_rate": 1.8465961108873182e-07, + "loss": 0.6887, + "step": 4239 + }, + { + "epoch": 0.16, + "learning_rate": 1.8465594902405977e-07, + "loss": 0.7013, + "step": 4240 + }, + { + "epoch": 0.16, + "learning_rate": 1.846522869593877e-07, + "loss": 0.6943, + "step": 4241 + }, + { + "epoch": 0.16, + "learning_rate": 1.8464862489471565e-07, + "loss": 0.675, + "step": 4242 + }, + { + "epoch": 0.16, + "learning_rate": 1.8464496283004355e-07, + "loss": 0.7022, + "step": 4243 + }, + { + "epoch": 0.16, + "learning_rate": 1.846413007653715e-07, + "loss": 0.6955, + "step": 4244 + }, + { + "epoch": 0.16, + "learning_rate": 1.8463763870069943e-07, + "loss": 0.7148, + "step": 4245 + }, + { + "epoch": 0.16, + "learning_rate": 1.8463397663602738e-07, + "loss": 0.6925, + "step": 4246 + }, + { + "epoch": 0.16, + "learning_rate": 1.846303145713553e-07, + "loss": 0.7028, + "step": 4247 + }, + { + "epoch": 0.16, + "learning_rate": 1.8462665250668326e-07, + "loss": 0.6967, + "step": 4248 + }, + { + "epoch": 0.16, + "learning_rate": 1.846229904420112e-07, + "loss": 0.6878, + "step": 4249 + }, + { + "epoch": 0.16, + "learning_rate": 1.8461932837733914e-07, + "loss": 0.6836, + "step": 4250 + }, + { + "epoch": 0.16, + "learning_rate": 1.8461566631266707e-07, + "loss": 0.6988, + "step": 4251 + }, + { + "epoch": 0.16, + "learning_rate": 1.8461200424799502e-07, + "loss": 0.7039, + "step": 4252 + }, + { + "epoch": 0.16, + "learning_rate": 1.8460834218332295e-07, + "loss": 0.7065, + "step": 4253 + }, + { + "epoch": 0.16, + "learning_rate": 1.846046801186509e-07, + "loss": 0.6985, + "step": 4254 + }, + { + "epoch": 0.16, + "learning_rate": 1.8460101805397883e-07, + "loss": 0.6912, + "step": 4255 + }, + { + "epoch": 0.16, + "learning_rate": 1.8459735598930678e-07, + "loss": 0.6842, + "step": 4256 + }, + { + "epoch": 0.16, + "learning_rate": 1.845936939246347e-07, + "loss": 0.6987, + "step": 4257 + }, + { + "epoch": 0.16, + "learning_rate": 1.8459003185996263e-07, + "loss": 0.7068, + "step": 4258 + }, + { + "epoch": 0.16, + "learning_rate": 1.845863697952906e-07, + "loss": 0.6837, + "step": 4259 + }, + { + "epoch": 0.16, + "learning_rate": 1.8458270773061851e-07, + "loss": 0.6787, + "step": 4260 + }, + { + "epoch": 0.16, + "learning_rate": 1.8457904566594644e-07, + "loss": 0.6795, + "step": 4261 + }, + { + "epoch": 0.16, + "learning_rate": 1.845753836012744e-07, + "loss": 0.683, + "step": 4262 + }, + { + "epoch": 0.16, + "learning_rate": 1.8457172153660232e-07, + "loss": 0.6792, + "step": 4263 + }, + { + "epoch": 0.16, + "learning_rate": 1.8456805947193025e-07, + "loss": 0.6861, + "step": 4264 + }, + { + "epoch": 0.16, + "learning_rate": 1.845643974072582e-07, + "loss": 0.6862, + "step": 4265 + }, + { + "epoch": 0.16, + "learning_rate": 1.8456073534258613e-07, + "loss": 0.7097, + "step": 4266 + }, + { + "epoch": 0.16, + "learning_rate": 1.8455707327791408e-07, + "loss": 0.6872, + "step": 4267 + }, + { + "epoch": 0.16, + "learning_rate": 1.84553411213242e-07, + "loss": 0.6832, + "step": 4268 + }, + { + "epoch": 0.16, + "learning_rate": 1.8454974914856996e-07, + "loss": 0.6947, + "step": 4269 + }, + { + "epoch": 0.16, + "learning_rate": 1.845460870838979e-07, + "loss": 0.6913, + "step": 4270 + }, + { + "epoch": 0.16, + "learning_rate": 1.8454242501922584e-07, + "loss": 0.706, + "step": 4271 + }, + { + "epoch": 0.16, + "learning_rate": 1.8453876295455377e-07, + "loss": 0.6975, + "step": 4272 + }, + { + "epoch": 0.16, + "learning_rate": 1.8453510088988172e-07, + "loss": 0.6746, + "step": 4273 + }, + { + "epoch": 0.16, + "learning_rate": 1.8453143882520965e-07, + "loss": 0.6968, + "step": 4274 + }, + { + "epoch": 0.16, + "learning_rate": 1.845277767605376e-07, + "loss": 0.6874, + "step": 4275 + }, + { + "epoch": 0.16, + "learning_rate": 1.8452411469586553e-07, + "loss": 0.7228, + "step": 4276 + }, + { + "epoch": 0.16, + "learning_rate": 1.8452045263119348e-07, + "loss": 0.7049, + "step": 4277 + }, + { + "epoch": 0.16, + "learning_rate": 1.845167905665214e-07, + "loss": 0.7209, + "step": 4278 + }, + { + "epoch": 0.16, + "learning_rate": 1.8451312850184933e-07, + "loss": 0.7016, + "step": 4279 + }, + { + "epoch": 0.16, + "learning_rate": 1.8450946643717726e-07, + "loss": 0.6861, + "step": 4280 + }, + { + "epoch": 0.16, + "learning_rate": 1.845058043725052e-07, + "loss": 0.7068, + "step": 4281 + }, + { + "epoch": 0.16, + "learning_rate": 1.8450214230783314e-07, + "loss": 0.6998, + "step": 4282 + }, + { + "epoch": 0.16, + "learning_rate": 1.844984802431611e-07, + "loss": 0.688, + "step": 4283 + }, + { + "epoch": 0.16, + "learning_rate": 1.8449481817848902e-07, + "loss": 0.6958, + "step": 4284 + }, + { + "epoch": 0.16, + "learning_rate": 1.8449115611381695e-07, + "loss": 0.7209, + "step": 4285 + }, + { + "epoch": 0.16, + "learning_rate": 1.844874940491449e-07, + "loss": 0.6884, + "step": 4286 + }, + { + "epoch": 0.16, + "learning_rate": 1.8448383198447283e-07, + "loss": 0.698, + "step": 4287 + }, + { + "epoch": 0.16, + "learning_rate": 1.8448016991980078e-07, + "loss": 0.7064, + "step": 4288 + }, + { + "epoch": 0.16, + "learning_rate": 1.844765078551287e-07, + "loss": 0.6756, + "step": 4289 + }, + { + "epoch": 0.16, + "learning_rate": 1.8447284579045666e-07, + "loss": 0.6872, + "step": 4290 + }, + { + "epoch": 0.16, + "learning_rate": 1.8446918372578459e-07, + "loss": 0.6867, + "step": 4291 + }, + { + "epoch": 0.16, + "learning_rate": 1.8446552166111254e-07, + "loss": 0.7055, + "step": 4292 + }, + { + "epoch": 0.16, + "learning_rate": 1.8446185959644047e-07, + "loss": 0.691, + "step": 4293 + }, + { + "epoch": 0.16, + "learning_rate": 1.8445819753176842e-07, + "loss": 0.6747, + "step": 4294 + }, + { + "epoch": 0.16, + "learning_rate": 1.8445453546709635e-07, + "loss": 0.6983, + "step": 4295 + }, + { + "epoch": 0.16, + "learning_rate": 1.8445087340242427e-07, + "loss": 0.677, + "step": 4296 + }, + { + "epoch": 0.16, + "learning_rate": 1.844472113377522e-07, + "loss": 0.7071, + "step": 4297 + }, + { + "epoch": 0.16, + "learning_rate": 1.8444354927308015e-07, + "loss": 0.6804, + "step": 4298 + }, + { + "epoch": 0.16, + "learning_rate": 1.8443988720840808e-07, + "loss": 0.6953, + "step": 4299 + }, + { + "epoch": 0.16, + "learning_rate": 1.8443622514373603e-07, + "loss": 0.6931, + "step": 4300 + }, + { + "epoch": 0.16, + "learning_rate": 1.8443256307906396e-07, + "loss": 0.7117, + "step": 4301 + }, + { + "epoch": 0.16, + "learning_rate": 1.844289010143919e-07, + "loss": 0.6907, + "step": 4302 + }, + { + "epoch": 0.16, + "learning_rate": 1.8442523894971984e-07, + "loss": 0.7005, + "step": 4303 + }, + { + "epoch": 0.16, + "learning_rate": 1.844215768850478e-07, + "loss": 0.698, + "step": 4304 + }, + { + "epoch": 0.16, + "learning_rate": 1.8441791482037572e-07, + "loss": 0.6629, + "step": 4305 + }, + { + "epoch": 0.16, + "learning_rate": 1.8441425275570367e-07, + "loss": 0.6755, + "step": 4306 + }, + { + "epoch": 0.16, + "learning_rate": 1.844105906910316e-07, + "loss": 0.7009, + "step": 4307 + }, + { + "epoch": 0.16, + "learning_rate": 1.8440692862635952e-07, + "loss": 0.6896, + "step": 4308 + }, + { + "epoch": 0.16, + "learning_rate": 1.8440326656168748e-07, + "loss": 0.711, + "step": 4309 + }, + { + "epoch": 0.16, + "learning_rate": 1.843996044970154e-07, + "loss": 0.6868, + "step": 4310 + }, + { + "epoch": 0.16, + "learning_rate": 1.8439594243234336e-07, + "loss": 0.7004, + "step": 4311 + }, + { + "epoch": 0.16, + "learning_rate": 1.8439228036767128e-07, + "loss": 0.6865, + "step": 4312 + }, + { + "epoch": 0.16, + "learning_rate": 1.8438861830299924e-07, + "loss": 0.6934, + "step": 4313 + }, + { + "epoch": 0.16, + "learning_rate": 1.8438495623832714e-07, + "loss": 0.6985, + "step": 4314 + }, + { + "epoch": 0.16, + "learning_rate": 1.843812941736551e-07, + "loss": 0.6748, + "step": 4315 + }, + { + "epoch": 0.16, + "learning_rate": 1.8437763210898302e-07, + "loss": 0.6714, + "step": 4316 + }, + { + "epoch": 0.16, + "learning_rate": 1.8437397004431097e-07, + "loss": 0.6927, + "step": 4317 + }, + { + "epoch": 0.16, + "learning_rate": 1.843703079796389e-07, + "loss": 0.6956, + "step": 4318 + }, + { + "epoch": 0.16, + "learning_rate": 1.8436664591496685e-07, + "loss": 0.6915, + "step": 4319 + }, + { + "epoch": 0.16, + "learning_rate": 1.8436298385029478e-07, + "loss": 0.686, + "step": 4320 + }, + { + "epoch": 0.16, + "learning_rate": 1.8435932178562273e-07, + "loss": 0.703, + "step": 4321 + }, + { + "epoch": 0.16, + "learning_rate": 1.8435565972095066e-07, + "loss": 0.719, + "step": 4322 + }, + { + "epoch": 0.16, + "learning_rate": 1.843519976562786e-07, + "loss": 0.6989, + "step": 4323 + }, + { + "epoch": 0.16, + "learning_rate": 1.8434833559160654e-07, + "loss": 0.6786, + "step": 4324 + }, + { + "epoch": 0.16, + "learning_rate": 1.843446735269345e-07, + "loss": 0.7031, + "step": 4325 + }, + { + "epoch": 0.16, + "learning_rate": 1.8434101146226242e-07, + "loss": 0.6783, + "step": 4326 + }, + { + "epoch": 0.16, + "learning_rate": 1.8433734939759037e-07, + "loss": 0.7086, + "step": 4327 + }, + { + "epoch": 0.16, + "learning_rate": 1.843336873329183e-07, + "loss": 0.7089, + "step": 4328 + }, + { + "epoch": 0.16, + "learning_rate": 1.8433002526824622e-07, + "loss": 0.6903, + "step": 4329 + }, + { + "epoch": 0.16, + "learning_rate": 1.8432636320357418e-07, + "loss": 0.6875, + "step": 4330 + }, + { + "epoch": 0.16, + "learning_rate": 1.843227011389021e-07, + "loss": 0.7037, + "step": 4331 + }, + { + "epoch": 0.16, + "learning_rate": 1.8431903907423003e-07, + "loss": 0.6831, + "step": 4332 + }, + { + "epoch": 0.16, + "learning_rate": 1.8431537700955798e-07, + "loss": 0.7073, + "step": 4333 + }, + { + "epoch": 0.16, + "learning_rate": 1.843117149448859e-07, + "loss": 0.6758, + "step": 4334 + }, + { + "epoch": 0.16, + "learning_rate": 1.8430805288021384e-07, + "loss": 0.7119, + "step": 4335 + }, + { + "epoch": 0.16, + "learning_rate": 1.843043908155418e-07, + "loss": 0.6961, + "step": 4336 + }, + { + "epoch": 0.16, + "learning_rate": 1.8430072875086972e-07, + "loss": 0.7056, + "step": 4337 + }, + { + "epoch": 0.16, + "learning_rate": 1.8429706668619767e-07, + "loss": 0.6811, + "step": 4338 + }, + { + "epoch": 0.16, + "learning_rate": 1.842934046215256e-07, + "loss": 0.7036, + "step": 4339 + }, + { + "epoch": 0.16, + "learning_rate": 1.8428974255685355e-07, + "loss": 0.6959, + "step": 4340 + }, + { + "epoch": 0.16, + "learning_rate": 1.8428608049218148e-07, + "loss": 0.7024, + "step": 4341 + }, + { + "epoch": 0.16, + "learning_rate": 1.8428241842750943e-07, + "loss": 0.6793, + "step": 4342 + }, + { + "epoch": 0.16, + "learning_rate": 1.8427875636283736e-07, + "loss": 0.6822, + "step": 4343 + }, + { + "epoch": 0.16, + "learning_rate": 1.842750942981653e-07, + "loss": 0.709, + "step": 4344 + }, + { + "epoch": 0.16, + "learning_rate": 1.8427143223349324e-07, + "loss": 0.7032, + "step": 4345 + }, + { + "epoch": 0.16, + "learning_rate": 1.842677701688212e-07, + "loss": 0.6913, + "step": 4346 + }, + { + "epoch": 0.16, + "learning_rate": 1.8426410810414912e-07, + "loss": 0.6962, + "step": 4347 + }, + { + "epoch": 0.16, + "learning_rate": 1.8426044603947707e-07, + "loss": 0.679, + "step": 4348 + }, + { + "epoch": 0.16, + "learning_rate": 1.84256783974805e-07, + "loss": 0.7079, + "step": 4349 + }, + { + "epoch": 0.16, + "learning_rate": 1.8425312191013292e-07, + "loss": 0.7075, + "step": 4350 + }, + { + "epoch": 0.16, + "learning_rate": 1.8424945984546085e-07, + "loss": 0.663, + "step": 4351 + }, + { + "epoch": 0.16, + "learning_rate": 1.842457977807888e-07, + "loss": 0.7107, + "step": 4352 + }, + { + "epoch": 0.16, + "learning_rate": 1.8424213571611673e-07, + "loss": 0.6978, + "step": 4353 + }, + { + "epoch": 0.16, + "learning_rate": 1.8423847365144468e-07, + "loss": 0.7022, + "step": 4354 + }, + { + "epoch": 0.16, + "learning_rate": 1.842348115867726e-07, + "loss": 0.6927, + "step": 4355 + }, + { + "epoch": 0.16, + "learning_rate": 1.8423114952210054e-07, + "loss": 0.6945, + "step": 4356 + }, + { + "epoch": 0.16, + "learning_rate": 1.842274874574285e-07, + "loss": 0.6853, + "step": 4357 + }, + { + "epoch": 0.16, + "learning_rate": 1.8422382539275642e-07, + "loss": 0.7056, + "step": 4358 + }, + { + "epoch": 0.16, + "learning_rate": 1.8422016332808437e-07, + "loss": 0.6903, + "step": 4359 + }, + { + "epoch": 0.16, + "learning_rate": 1.842165012634123e-07, + "loss": 0.6923, + "step": 4360 + }, + { + "epoch": 0.16, + "learning_rate": 1.8421283919874025e-07, + "loss": 0.6989, + "step": 4361 + }, + { + "epoch": 0.16, + "learning_rate": 1.8420917713406818e-07, + "loss": 0.7158, + "step": 4362 + }, + { + "epoch": 0.16, + "learning_rate": 1.8420551506939613e-07, + "loss": 0.7045, + "step": 4363 + }, + { + "epoch": 0.16, + "learning_rate": 1.8420185300472406e-07, + "loss": 0.6882, + "step": 4364 + }, + { + "epoch": 0.16, + "learning_rate": 1.84198190940052e-07, + "loss": 0.6616, + "step": 4365 + }, + { + "epoch": 0.16, + "learning_rate": 1.8419452887537993e-07, + "loss": 0.6912, + "step": 4366 + }, + { + "epoch": 0.16, + "learning_rate": 1.841908668107079e-07, + "loss": 0.6908, + "step": 4367 + }, + { + "epoch": 0.16, + "learning_rate": 1.841872047460358e-07, + "loss": 0.6998, + "step": 4368 + }, + { + "epoch": 0.16, + "learning_rate": 1.8418354268136374e-07, + "loss": 0.7156, + "step": 4369 + }, + { + "epoch": 0.16, + "learning_rate": 1.8417988061669167e-07, + "loss": 0.7112, + "step": 4370 + }, + { + "epoch": 0.16, + "learning_rate": 1.8417621855201962e-07, + "loss": 0.7123, + "step": 4371 + }, + { + "epoch": 0.16, + "learning_rate": 1.8417255648734755e-07, + "loss": 0.6967, + "step": 4372 + }, + { + "epoch": 0.16, + "learning_rate": 1.841688944226755e-07, + "loss": 0.6973, + "step": 4373 + }, + { + "epoch": 0.16, + "learning_rate": 1.8416523235800343e-07, + "loss": 0.7131, + "step": 4374 + }, + { + "epoch": 0.16, + "learning_rate": 1.8416157029333138e-07, + "loss": 0.7022, + "step": 4375 + }, + { + "epoch": 0.16, + "learning_rate": 1.841579082286593e-07, + "loss": 0.6917, + "step": 4376 + }, + { + "epoch": 0.16, + "learning_rate": 1.8415424616398726e-07, + "loss": 0.6849, + "step": 4377 + }, + { + "epoch": 0.16, + "learning_rate": 1.841505840993152e-07, + "loss": 0.697, + "step": 4378 + }, + { + "epoch": 0.16, + "learning_rate": 1.8414692203464311e-07, + "loss": 0.6744, + "step": 4379 + }, + { + "epoch": 0.16, + "learning_rate": 1.8414325996997107e-07, + "loss": 0.6911, + "step": 4380 + }, + { + "epoch": 0.16, + "learning_rate": 1.84139597905299e-07, + "loss": 0.7052, + "step": 4381 + }, + { + "epoch": 0.16, + "learning_rate": 1.8413593584062695e-07, + "loss": 0.6831, + "step": 4382 + }, + { + "epoch": 0.16, + "learning_rate": 1.8413227377595487e-07, + "loss": 0.6909, + "step": 4383 + }, + { + "epoch": 0.16, + "learning_rate": 1.8412861171128283e-07, + "loss": 0.7236, + "step": 4384 + }, + { + "epoch": 0.16, + "learning_rate": 1.8412494964661075e-07, + "loss": 0.702, + "step": 4385 + }, + { + "epoch": 0.16, + "learning_rate": 1.8412128758193868e-07, + "loss": 0.6968, + "step": 4386 + }, + { + "epoch": 0.16, + "learning_rate": 1.841176255172666e-07, + "loss": 0.7248, + "step": 4387 + }, + { + "epoch": 0.16, + "learning_rate": 1.8411396345259456e-07, + "loss": 0.7008, + "step": 4388 + }, + { + "epoch": 0.16, + "learning_rate": 1.841103013879225e-07, + "loss": 0.6819, + "step": 4389 + }, + { + "epoch": 0.16, + "learning_rate": 1.8410663932325044e-07, + "loss": 0.6791, + "step": 4390 + }, + { + "epoch": 0.16, + "learning_rate": 1.8410297725857837e-07, + "loss": 0.6945, + "step": 4391 + }, + { + "epoch": 0.16, + "learning_rate": 1.8409931519390632e-07, + "loss": 0.6938, + "step": 4392 + }, + { + "epoch": 0.16, + "learning_rate": 1.8409565312923425e-07, + "loss": 0.7002, + "step": 4393 + }, + { + "epoch": 0.16, + "learning_rate": 1.840919910645622e-07, + "loss": 0.6955, + "step": 4394 + }, + { + "epoch": 0.16, + "learning_rate": 1.8408832899989013e-07, + "loss": 0.6791, + "step": 4395 + }, + { + "epoch": 0.16, + "learning_rate": 1.8408466693521808e-07, + "loss": 0.6893, + "step": 4396 + }, + { + "epoch": 0.16, + "learning_rate": 1.84081004870546e-07, + "loss": 0.687, + "step": 4397 + }, + { + "epoch": 0.16, + "learning_rate": 1.8407734280587396e-07, + "loss": 0.6626, + "step": 4398 + }, + { + "epoch": 0.16, + "learning_rate": 1.8407368074120189e-07, + "loss": 0.6962, + "step": 4399 + }, + { + "epoch": 0.16, + "learning_rate": 1.840700186765298e-07, + "loss": 0.6829, + "step": 4400 + }, + { + "epoch": 0.16, + "learning_rate": 1.8406635661185777e-07, + "loss": 0.6838, + "step": 4401 + }, + { + "epoch": 0.16, + "learning_rate": 1.840626945471857e-07, + "loss": 0.7001, + "step": 4402 + }, + { + "epoch": 0.16, + "learning_rate": 1.8405903248251365e-07, + "loss": 0.6753, + "step": 4403 + }, + { + "epoch": 0.16, + "learning_rate": 1.8405537041784157e-07, + "loss": 0.6959, + "step": 4404 + }, + { + "epoch": 0.16, + "learning_rate": 1.840517083531695e-07, + "loss": 0.7003, + "step": 4405 + }, + { + "epoch": 0.16, + "learning_rate": 1.8404804628849743e-07, + "loss": 0.6968, + "step": 4406 + }, + { + "epoch": 0.16, + "learning_rate": 1.8404438422382538e-07, + "loss": 0.6926, + "step": 4407 + }, + { + "epoch": 0.16, + "learning_rate": 1.840407221591533e-07, + "loss": 0.6978, + "step": 4408 + }, + { + "epoch": 0.16, + "learning_rate": 1.8403706009448126e-07, + "loss": 0.6945, + "step": 4409 + }, + { + "epoch": 0.16, + "learning_rate": 1.8403339802980919e-07, + "loss": 0.6919, + "step": 4410 + }, + { + "epoch": 0.16, + "learning_rate": 1.8402973596513714e-07, + "loss": 0.6883, + "step": 4411 + }, + { + "epoch": 0.16, + "learning_rate": 1.8402607390046507e-07, + "loss": 0.6879, + "step": 4412 + }, + { + "epoch": 0.16, + "learning_rate": 1.8402241183579302e-07, + "loss": 0.6968, + "step": 4413 + }, + { + "epoch": 0.16, + "learning_rate": 1.8401874977112095e-07, + "loss": 0.7239, + "step": 4414 + }, + { + "epoch": 0.16, + "learning_rate": 1.840150877064489e-07, + "loss": 0.7116, + "step": 4415 + }, + { + "epoch": 0.16, + "learning_rate": 1.8401142564177683e-07, + "loss": 0.6754, + "step": 4416 + }, + { + "epoch": 0.16, + "learning_rate": 1.8400776357710478e-07, + "loss": 0.6971, + "step": 4417 + }, + { + "epoch": 0.16, + "learning_rate": 1.840041015124327e-07, + "loss": 0.6974, + "step": 4418 + }, + { + "epoch": 0.16, + "learning_rate": 1.8400043944776066e-07, + "loss": 0.689, + "step": 4419 + }, + { + "epoch": 0.16, + "learning_rate": 1.8399677738308859e-07, + "loss": 0.6741, + "step": 4420 + }, + { + "epoch": 0.16, + "learning_rate": 1.8399311531841654e-07, + "loss": 0.7013, + "step": 4421 + }, + { + "epoch": 0.16, + "learning_rate": 1.8398945325374444e-07, + "loss": 0.7038, + "step": 4422 + }, + { + "epoch": 0.16, + "learning_rate": 1.839857911890724e-07, + "loss": 0.6968, + "step": 4423 + }, + { + "epoch": 0.16, + "learning_rate": 1.8398212912440032e-07, + "loss": 0.6932, + "step": 4424 + }, + { + "epoch": 0.16, + "learning_rate": 1.8397846705972827e-07, + "loss": 0.6863, + "step": 4425 + }, + { + "epoch": 0.16, + "learning_rate": 1.839748049950562e-07, + "loss": 0.7097, + "step": 4426 + }, + { + "epoch": 0.16, + "learning_rate": 1.8397114293038412e-07, + "loss": 0.7116, + "step": 4427 + }, + { + "epoch": 0.16, + "learning_rate": 1.8396748086571208e-07, + "loss": 0.7204, + "step": 4428 + }, + { + "epoch": 0.16, + "learning_rate": 1.8396381880104e-07, + "loss": 0.7099, + "step": 4429 + }, + { + "epoch": 0.16, + "learning_rate": 1.8396015673636796e-07, + "loss": 0.6941, + "step": 4430 + }, + { + "epoch": 0.16, + "learning_rate": 1.8395649467169588e-07, + "loss": 0.6774, + "step": 4431 + }, + { + "epoch": 0.16, + "learning_rate": 1.8395283260702384e-07, + "loss": 0.6995, + "step": 4432 + }, + { + "epoch": 0.16, + "learning_rate": 1.8394917054235176e-07, + "loss": 0.6889, + "step": 4433 + }, + { + "epoch": 0.16, + "learning_rate": 1.8394550847767972e-07, + "loss": 0.6927, + "step": 4434 + }, + { + "epoch": 0.16, + "learning_rate": 1.8394184641300764e-07, + "loss": 0.7044, + "step": 4435 + }, + { + "epoch": 0.16, + "learning_rate": 1.839381843483356e-07, + "loss": 0.6976, + "step": 4436 + }, + { + "epoch": 0.16, + "learning_rate": 1.8393452228366352e-07, + "loss": 0.6918, + "step": 4437 + }, + { + "epoch": 0.16, + "learning_rate": 1.8393086021899148e-07, + "loss": 0.6854, + "step": 4438 + }, + { + "epoch": 0.16, + "learning_rate": 1.839271981543194e-07, + "loss": 0.7168, + "step": 4439 + }, + { + "epoch": 0.16, + "learning_rate": 1.8392353608964733e-07, + "loss": 0.6905, + "step": 4440 + }, + { + "epoch": 0.16, + "learning_rate": 1.8391987402497526e-07, + "loss": 0.7094, + "step": 4441 + }, + { + "epoch": 0.16, + "learning_rate": 1.839162119603032e-07, + "loss": 0.7107, + "step": 4442 + }, + { + "epoch": 0.16, + "learning_rate": 1.8391254989563114e-07, + "loss": 0.686, + "step": 4443 + }, + { + "epoch": 0.16, + "learning_rate": 1.839088878309591e-07, + "loss": 0.7212, + "step": 4444 + }, + { + "epoch": 0.16, + "learning_rate": 1.8390522576628702e-07, + "loss": 0.6809, + "step": 4445 + }, + { + "epoch": 0.16, + "learning_rate": 1.8390156370161497e-07, + "loss": 0.7158, + "step": 4446 + }, + { + "epoch": 0.16, + "learning_rate": 1.838979016369429e-07, + "loss": 0.6986, + "step": 4447 + }, + { + "epoch": 0.16, + "learning_rate": 1.8389423957227085e-07, + "loss": 0.7032, + "step": 4448 + }, + { + "epoch": 0.16, + "learning_rate": 1.8389057750759878e-07, + "loss": 0.6888, + "step": 4449 + }, + { + "epoch": 0.16, + "learning_rate": 1.838869154429267e-07, + "loss": 0.6928, + "step": 4450 + }, + { + "epoch": 0.16, + "learning_rate": 1.8388325337825466e-07, + "loss": 0.6793, + "step": 4451 + }, + { + "epoch": 0.16, + "learning_rate": 1.8387959131358258e-07, + "loss": 0.6958, + "step": 4452 + }, + { + "epoch": 0.16, + "learning_rate": 1.8387592924891054e-07, + "loss": 0.695, + "step": 4453 + }, + { + "epoch": 0.16, + "learning_rate": 1.8387226718423846e-07, + "loss": 0.6904, + "step": 4454 + }, + { + "epoch": 0.16, + "learning_rate": 1.8386860511956642e-07, + "loss": 0.6905, + "step": 4455 + }, + { + "epoch": 0.16, + "learning_rate": 1.8386494305489434e-07, + "loss": 0.6947, + "step": 4456 + }, + { + "epoch": 0.16, + "learning_rate": 1.838612809902223e-07, + "loss": 0.698, + "step": 4457 + }, + { + "epoch": 0.16, + "learning_rate": 1.838576189255502e-07, + "loss": 0.6994, + "step": 4458 + }, + { + "epoch": 0.16, + "learning_rate": 1.8385395686087815e-07, + "loss": 0.6898, + "step": 4459 + }, + { + "epoch": 0.16, + "learning_rate": 1.8385029479620608e-07, + "loss": 0.69, + "step": 4460 + }, + { + "epoch": 0.16, + "learning_rate": 1.8384663273153403e-07, + "loss": 0.6905, + "step": 4461 + }, + { + "epoch": 0.16, + "learning_rate": 1.8384297066686196e-07, + "loss": 0.6947, + "step": 4462 + }, + { + "epoch": 0.16, + "learning_rate": 1.838393086021899e-07, + "loss": 0.7049, + "step": 4463 + }, + { + "epoch": 0.16, + "learning_rate": 1.8383564653751784e-07, + "loss": 0.7022, + "step": 4464 + }, + { + "epoch": 0.16, + "learning_rate": 1.838319844728458e-07, + "loss": 0.6937, + "step": 4465 + }, + { + "epoch": 0.16, + "learning_rate": 1.8382832240817372e-07, + "loss": 0.6895, + "step": 4466 + }, + { + "epoch": 0.16, + "learning_rate": 1.8382466034350167e-07, + "loss": 0.6778, + "step": 4467 + }, + { + "epoch": 0.16, + "learning_rate": 1.838209982788296e-07, + "loss": 0.6793, + "step": 4468 + }, + { + "epoch": 0.16, + "learning_rate": 1.8381733621415755e-07, + "loss": 0.6975, + "step": 4469 + }, + { + "epoch": 0.16, + "learning_rate": 1.8381367414948548e-07, + "loss": 0.7037, + "step": 4470 + }, + { + "epoch": 0.16, + "learning_rate": 1.838100120848134e-07, + "loss": 0.6808, + "step": 4471 + }, + { + "epoch": 0.16, + "learning_rate": 1.8380635002014136e-07, + "loss": 0.7008, + "step": 4472 + }, + { + "epoch": 0.16, + "learning_rate": 1.8380268795546928e-07, + "loss": 0.7161, + "step": 4473 + }, + { + "epoch": 0.16, + "learning_rate": 1.8379902589079724e-07, + "loss": 0.715, + "step": 4474 + }, + { + "epoch": 0.16, + "learning_rate": 1.8379536382612516e-07, + "loss": 0.6854, + "step": 4475 + }, + { + "epoch": 0.16, + "learning_rate": 1.837917017614531e-07, + "loss": 0.6961, + "step": 4476 + }, + { + "epoch": 0.16, + "learning_rate": 1.8378803969678102e-07, + "loss": 0.6958, + "step": 4477 + }, + { + "epoch": 0.16, + "learning_rate": 1.8378437763210897e-07, + "loss": 0.6747, + "step": 4478 + }, + { + "epoch": 0.16, + "learning_rate": 1.837807155674369e-07, + "loss": 0.6794, + "step": 4479 + }, + { + "epoch": 0.16, + "learning_rate": 1.8377705350276485e-07, + "loss": 0.6954, + "step": 4480 + }, + { + "epoch": 0.16, + "learning_rate": 1.8377339143809278e-07, + "loss": 0.7054, + "step": 4481 + }, + { + "epoch": 0.16, + "learning_rate": 1.8376972937342073e-07, + "loss": 0.6887, + "step": 4482 + }, + { + "epoch": 0.16, + "learning_rate": 1.8376606730874866e-07, + "loss": 0.6956, + "step": 4483 + }, + { + "epoch": 0.16, + "learning_rate": 1.837624052440766e-07, + "loss": 0.6964, + "step": 4484 + }, + { + "epoch": 0.16, + "learning_rate": 1.8375874317940453e-07, + "loss": 0.6838, + "step": 4485 + }, + { + "epoch": 0.16, + "learning_rate": 1.837550811147325e-07, + "loss": 0.6981, + "step": 4486 + }, + { + "epoch": 0.16, + "learning_rate": 1.8375141905006041e-07, + "loss": 0.688, + "step": 4487 + }, + { + "epoch": 0.16, + "learning_rate": 1.8374775698538837e-07, + "loss": 0.7003, + "step": 4488 + }, + { + "epoch": 0.16, + "learning_rate": 1.837440949207163e-07, + "loss": 0.6893, + "step": 4489 + }, + { + "epoch": 0.16, + "learning_rate": 1.8374043285604425e-07, + "loss": 0.7165, + "step": 4490 + }, + { + "epoch": 0.16, + "learning_rate": 1.8373677079137217e-07, + "loss": 0.6913, + "step": 4491 + }, + { + "epoch": 0.16, + "learning_rate": 1.8373310872670013e-07, + "loss": 0.697, + "step": 4492 + }, + { + "epoch": 0.16, + "learning_rate": 1.8372944666202805e-07, + "loss": 0.7066, + "step": 4493 + }, + { + "epoch": 0.16, + "learning_rate": 1.8372578459735598e-07, + "loss": 0.7101, + "step": 4494 + }, + { + "epoch": 0.16, + "learning_rate": 1.837221225326839e-07, + "loss": 0.6811, + "step": 4495 + }, + { + "epoch": 0.16, + "learning_rate": 1.8371846046801186e-07, + "loss": 0.6954, + "step": 4496 + }, + { + "epoch": 0.16, + "learning_rate": 1.837147984033398e-07, + "loss": 0.7024, + "step": 4497 + }, + { + "epoch": 0.16, + "learning_rate": 1.8371113633866771e-07, + "loss": 0.7103, + "step": 4498 + }, + { + "epoch": 0.16, + "learning_rate": 1.8370747427399567e-07, + "loss": 0.7172, + "step": 4499 + }, + { + "epoch": 0.16, + "learning_rate": 1.837038122093236e-07, + "loss": 0.6869, + "step": 4500 + }, + { + "epoch": 0.16, + "learning_rate": 1.8370015014465155e-07, + "loss": 0.6677, + "step": 4501 + }, + { + "epoch": 0.16, + "learning_rate": 1.8369648807997947e-07, + "loss": 0.6982, + "step": 4502 + }, + { + "epoch": 0.16, + "learning_rate": 1.8369282601530743e-07, + "loss": 0.6957, + "step": 4503 + }, + { + "epoch": 0.16, + "learning_rate": 1.8368916395063535e-07, + "loss": 0.7006, + "step": 4504 + }, + { + "epoch": 0.16, + "learning_rate": 1.836855018859633e-07, + "loss": 0.6955, + "step": 4505 + }, + { + "epoch": 0.16, + "learning_rate": 1.8368183982129123e-07, + "loss": 0.6877, + "step": 4506 + }, + { + "epoch": 0.16, + "learning_rate": 1.836781777566192e-07, + "loss": 0.6912, + "step": 4507 + }, + { + "epoch": 0.16, + "learning_rate": 1.8367451569194711e-07, + "loss": 0.6881, + "step": 4508 + }, + { + "epoch": 0.16, + "learning_rate": 1.8367085362727507e-07, + "loss": 0.6899, + "step": 4509 + }, + { + "epoch": 0.17, + "learning_rate": 1.83667191562603e-07, + "loss": 0.713, + "step": 4510 + }, + { + "epoch": 0.17, + "learning_rate": 1.8366352949793095e-07, + "loss": 0.6815, + "step": 4511 + }, + { + "epoch": 0.17, + "learning_rate": 1.8365986743325885e-07, + "loss": 0.6985, + "step": 4512 + }, + { + "epoch": 0.17, + "learning_rate": 1.836562053685868e-07, + "loss": 0.6777, + "step": 4513 + }, + { + "epoch": 0.17, + "learning_rate": 1.8365254330391473e-07, + "loss": 0.7006, + "step": 4514 + }, + { + "epoch": 0.17, + "learning_rate": 1.8364888123924268e-07, + "loss": 0.6967, + "step": 4515 + }, + { + "epoch": 0.17, + "learning_rate": 1.836452191745706e-07, + "loss": 0.7024, + "step": 4516 + }, + { + "epoch": 0.17, + "learning_rate": 1.8364155710989856e-07, + "loss": 0.6978, + "step": 4517 + }, + { + "epoch": 0.17, + "learning_rate": 1.8363789504522649e-07, + "loss": 0.6797, + "step": 4518 + }, + { + "epoch": 0.17, + "learning_rate": 1.8363423298055444e-07, + "loss": 0.6777, + "step": 4519 + }, + { + "epoch": 0.17, + "learning_rate": 1.8363057091588237e-07, + "loss": 0.6677, + "step": 4520 + }, + { + "epoch": 0.17, + "learning_rate": 1.836269088512103e-07, + "loss": 0.7236, + "step": 4521 + }, + { + "epoch": 0.17, + "learning_rate": 1.8362324678653825e-07, + "loss": 0.7051, + "step": 4522 + }, + { + "epoch": 0.17, + "learning_rate": 1.8361958472186617e-07, + "loss": 0.6934, + "step": 4523 + }, + { + "epoch": 0.17, + "learning_rate": 1.8361592265719413e-07, + "loss": 0.7046, + "step": 4524 + }, + { + "epoch": 0.17, + "learning_rate": 1.8361226059252205e-07, + "loss": 0.6925, + "step": 4525 + }, + { + "epoch": 0.17, + "learning_rate": 1.8360859852785e-07, + "loss": 0.7063, + "step": 4526 + }, + { + "epoch": 0.17, + "learning_rate": 1.8360493646317793e-07, + "loss": 0.6825, + "step": 4527 + }, + { + "epoch": 0.17, + "learning_rate": 1.8360127439850589e-07, + "loss": 0.6743, + "step": 4528 + }, + { + "epoch": 0.17, + "learning_rate": 1.835976123338338e-07, + "loss": 0.7081, + "step": 4529 + }, + { + "epoch": 0.17, + "learning_rate": 1.8359395026916174e-07, + "loss": 0.7055, + "step": 4530 + }, + { + "epoch": 0.17, + "learning_rate": 1.8359028820448967e-07, + "loss": 0.7269, + "step": 4531 + }, + { + "epoch": 0.17, + "learning_rate": 1.8358662613981762e-07, + "loss": 0.7001, + "step": 4532 + }, + { + "epoch": 0.17, + "learning_rate": 1.8358296407514555e-07, + "loss": 0.6903, + "step": 4533 + }, + { + "epoch": 0.17, + "learning_rate": 1.835793020104735e-07, + "loss": 0.6918, + "step": 4534 + }, + { + "epoch": 0.17, + "learning_rate": 1.8357563994580143e-07, + "loss": 0.6971, + "step": 4535 + }, + { + "epoch": 0.17, + "learning_rate": 1.8357197788112938e-07, + "loss": 0.6876, + "step": 4536 + }, + { + "epoch": 0.17, + "learning_rate": 1.835683158164573e-07, + "loss": 0.689, + "step": 4537 + }, + { + "epoch": 0.17, + "learning_rate": 1.8356465375178526e-07, + "loss": 0.6987, + "step": 4538 + }, + { + "epoch": 0.17, + "learning_rate": 1.8356099168711319e-07, + "loss": 0.6951, + "step": 4539 + }, + { + "epoch": 0.17, + "learning_rate": 1.8355732962244114e-07, + "loss": 0.6772, + "step": 4540 + }, + { + "epoch": 0.17, + "learning_rate": 1.8355366755776907e-07, + "loss": 0.6922, + "step": 4541 + }, + { + "epoch": 0.17, + "learning_rate": 1.83550005493097e-07, + "loss": 0.7011, + "step": 4542 + }, + { + "epoch": 0.17, + "learning_rate": 1.8354634342842494e-07, + "loss": 0.6899, + "step": 4543 + }, + { + "epoch": 0.17, + "learning_rate": 1.8354268136375287e-07, + "loss": 0.6848, + "step": 4544 + }, + { + "epoch": 0.17, + "learning_rate": 1.8353901929908082e-07, + "loss": 0.69, + "step": 4545 + }, + { + "epoch": 0.17, + "learning_rate": 1.8353535723440875e-07, + "loss": 0.684, + "step": 4546 + }, + { + "epoch": 0.17, + "learning_rate": 1.835316951697367e-07, + "loss": 0.6998, + "step": 4547 + }, + { + "epoch": 0.17, + "learning_rate": 1.835280331050646e-07, + "loss": 0.708, + "step": 4548 + }, + { + "epoch": 0.17, + "learning_rate": 1.8352437104039256e-07, + "loss": 0.6878, + "step": 4549 + }, + { + "epoch": 0.17, + "learning_rate": 1.8352070897572048e-07, + "loss": 0.6767, + "step": 4550 + }, + { + "epoch": 0.17, + "learning_rate": 1.8351704691104844e-07, + "loss": 0.6875, + "step": 4551 + }, + { + "epoch": 0.17, + "learning_rate": 1.8351338484637636e-07, + "loss": 0.6711, + "step": 4552 + }, + { + "epoch": 0.17, + "learning_rate": 1.8350972278170432e-07, + "loss": 0.6875, + "step": 4553 + }, + { + "epoch": 0.17, + "learning_rate": 1.8350606071703224e-07, + "loss": 0.6955, + "step": 4554 + }, + { + "epoch": 0.17, + "learning_rate": 1.835023986523602e-07, + "loss": 0.6789, + "step": 4555 + }, + { + "epoch": 0.17, + "learning_rate": 1.8349873658768812e-07, + "loss": 0.6923, + "step": 4556 + }, + { + "epoch": 0.17, + "learning_rate": 1.8349507452301608e-07, + "loss": 0.6967, + "step": 4557 + }, + { + "epoch": 0.17, + "learning_rate": 1.83491412458344e-07, + "loss": 0.6947, + "step": 4558 + }, + { + "epoch": 0.17, + "learning_rate": 1.8348775039367196e-07, + "loss": 0.704, + "step": 4559 + }, + { + "epoch": 0.17, + "learning_rate": 1.8348408832899988e-07, + "loss": 0.6822, + "step": 4560 + }, + { + "epoch": 0.17, + "learning_rate": 1.8348042626432784e-07, + "loss": 0.6889, + "step": 4561 + }, + { + "epoch": 0.17, + "learning_rate": 1.8347676419965576e-07, + "loss": 0.6996, + "step": 4562 + }, + { + "epoch": 0.17, + "learning_rate": 1.8347310213498372e-07, + "loss": 0.6853, + "step": 4563 + }, + { + "epoch": 0.17, + "learning_rate": 1.8346944007031164e-07, + "loss": 0.6928, + "step": 4564 + }, + { + "epoch": 0.17, + "learning_rate": 1.8346577800563957e-07, + "loss": 0.6895, + "step": 4565 + }, + { + "epoch": 0.17, + "learning_rate": 1.834621159409675e-07, + "loss": 0.6907, + "step": 4566 + }, + { + "epoch": 0.17, + "learning_rate": 1.8345845387629545e-07, + "loss": 0.7107, + "step": 4567 + }, + { + "epoch": 0.17, + "learning_rate": 1.8345479181162338e-07, + "loss": 0.7032, + "step": 4568 + }, + { + "epoch": 0.17, + "learning_rate": 1.834511297469513e-07, + "loss": 0.6921, + "step": 4569 + }, + { + "epoch": 0.17, + "learning_rate": 1.8344746768227926e-07, + "loss": 0.7015, + "step": 4570 + }, + { + "epoch": 0.17, + "learning_rate": 1.8344380561760718e-07, + "loss": 0.7065, + "step": 4571 + }, + { + "epoch": 0.17, + "learning_rate": 1.8344014355293514e-07, + "loss": 0.6964, + "step": 4572 + }, + { + "epoch": 0.17, + "learning_rate": 1.8343648148826306e-07, + "loss": 0.6773, + "step": 4573 + }, + { + "epoch": 0.17, + "learning_rate": 1.8343281942359102e-07, + "loss": 0.7031, + "step": 4574 + }, + { + "epoch": 0.17, + "learning_rate": 1.8342915735891894e-07, + "loss": 0.6939, + "step": 4575 + }, + { + "epoch": 0.17, + "learning_rate": 1.834254952942469e-07, + "loss": 0.7087, + "step": 4576 + }, + { + "epoch": 0.17, + "learning_rate": 1.8342183322957482e-07, + "loss": 0.6964, + "step": 4577 + }, + { + "epoch": 0.17, + "learning_rate": 1.8341817116490278e-07, + "loss": 0.6933, + "step": 4578 + }, + { + "epoch": 0.17, + "learning_rate": 1.834145091002307e-07, + "loss": 0.7122, + "step": 4579 + }, + { + "epoch": 0.17, + "learning_rate": 1.8341084703555866e-07, + "loss": 0.7079, + "step": 4580 + }, + { + "epoch": 0.17, + "learning_rate": 1.8340718497088658e-07, + "loss": 0.7019, + "step": 4581 + }, + { + "epoch": 0.17, + "learning_rate": 1.8340352290621454e-07, + "loss": 0.7128, + "step": 4582 + }, + { + "epoch": 0.17, + "learning_rate": 1.8339986084154246e-07, + "loss": 0.6855, + "step": 4583 + }, + { + "epoch": 0.17, + "learning_rate": 1.833961987768704e-07, + "loss": 0.6896, + "step": 4584 + }, + { + "epoch": 0.17, + "learning_rate": 1.8339253671219832e-07, + "loss": 0.693, + "step": 4585 + }, + { + "epoch": 0.17, + "learning_rate": 1.8338887464752627e-07, + "loss": 0.7159, + "step": 4586 + }, + { + "epoch": 0.17, + "learning_rate": 1.833852125828542e-07, + "loss": 0.7082, + "step": 4587 + }, + { + "epoch": 0.17, + "learning_rate": 1.8338155051818215e-07, + "loss": 0.6886, + "step": 4588 + }, + { + "epoch": 0.17, + "learning_rate": 1.8337788845351008e-07, + "loss": 0.6771, + "step": 4589 + }, + { + "epoch": 0.17, + "learning_rate": 1.8337422638883803e-07, + "loss": 0.6931, + "step": 4590 + }, + { + "epoch": 0.17, + "learning_rate": 1.8337056432416596e-07, + "loss": 0.6851, + "step": 4591 + }, + { + "epoch": 0.17, + "learning_rate": 1.8336690225949388e-07, + "loss": 0.6858, + "step": 4592 + }, + { + "epoch": 0.17, + "learning_rate": 1.8336324019482184e-07, + "loss": 0.6896, + "step": 4593 + }, + { + "epoch": 0.17, + "learning_rate": 1.8335957813014976e-07, + "loss": 0.6966, + "step": 4594 + }, + { + "epoch": 0.17, + "learning_rate": 1.8335591606547772e-07, + "loss": 0.6931, + "step": 4595 + }, + { + "epoch": 0.17, + "learning_rate": 1.8335225400080564e-07, + "loss": 0.7034, + "step": 4596 + }, + { + "epoch": 0.17, + "learning_rate": 1.833485919361336e-07, + "loss": 0.674, + "step": 4597 + }, + { + "epoch": 0.17, + "learning_rate": 1.8334492987146152e-07, + "loss": 0.7029, + "step": 4598 + }, + { + "epoch": 0.17, + "learning_rate": 1.8334126780678948e-07, + "loss": 0.6808, + "step": 4599 + }, + { + "epoch": 0.17, + "learning_rate": 1.833376057421174e-07, + "loss": 0.697, + "step": 4600 + }, + { + "epoch": 0.17, + "learning_rate": 1.8333394367744535e-07, + "loss": 0.6909, + "step": 4601 + }, + { + "epoch": 0.17, + "learning_rate": 1.8333028161277326e-07, + "loss": 0.6773, + "step": 4602 + }, + { + "epoch": 0.17, + "learning_rate": 1.833266195481012e-07, + "loss": 0.673, + "step": 4603 + }, + { + "epoch": 0.17, + "learning_rate": 1.8332295748342913e-07, + "loss": 0.6944, + "step": 4604 + }, + { + "epoch": 0.17, + "learning_rate": 1.833192954187571e-07, + "loss": 0.6904, + "step": 4605 + }, + { + "epoch": 0.17, + "learning_rate": 1.8331563335408501e-07, + "loss": 0.6927, + "step": 4606 + }, + { + "epoch": 0.17, + "learning_rate": 1.8331197128941297e-07, + "loss": 0.6813, + "step": 4607 + }, + { + "epoch": 0.17, + "learning_rate": 1.833083092247409e-07, + "loss": 0.6788, + "step": 4608 + }, + { + "epoch": 0.17, + "learning_rate": 1.8330464716006885e-07, + "loss": 0.7335, + "step": 4609 + }, + { + "epoch": 0.17, + "learning_rate": 1.8330098509539677e-07, + "loss": 0.6919, + "step": 4610 + }, + { + "epoch": 0.17, + "learning_rate": 1.8329732303072473e-07, + "loss": 0.6932, + "step": 4611 + }, + { + "epoch": 0.17, + "learning_rate": 1.8329366096605265e-07, + "loss": 0.6816, + "step": 4612 + }, + { + "epoch": 0.17, + "learning_rate": 1.8328999890138058e-07, + "loss": 0.6807, + "step": 4613 + }, + { + "epoch": 0.17, + "learning_rate": 1.8328633683670853e-07, + "loss": 0.7023, + "step": 4614 + }, + { + "epoch": 0.17, + "learning_rate": 1.8328267477203646e-07, + "loss": 0.668, + "step": 4615 + }, + { + "epoch": 0.17, + "learning_rate": 1.8327901270736441e-07, + "loss": 0.7144, + "step": 4616 + }, + { + "epoch": 0.17, + "learning_rate": 1.8327535064269234e-07, + "loss": 0.7203, + "step": 4617 + }, + { + "epoch": 0.17, + "learning_rate": 1.832716885780203e-07, + "loss": 0.6956, + "step": 4618 + }, + { + "epoch": 0.17, + "learning_rate": 1.8326802651334822e-07, + "loss": 0.6958, + "step": 4619 + }, + { + "epoch": 0.17, + "learning_rate": 1.8326436444867615e-07, + "loss": 0.6955, + "step": 4620 + }, + { + "epoch": 0.17, + "learning_rate": 1.8326070238400407e-07, + "loss": 0.6991, + "step": 4621 + }, + { + "epoch": 0.17, + "learning_rate": 1.8325704031933203e-07, + "loss": 0.6894, + "step": 4622 + }, + { + "epoch": 0.17, + "learning_rate": 1.8325337825465995e-07, + "loss": 0.692, + "step": 4623 + }, + { + "epoch": 0.17, + "learning_rate": 1.832497161899879e-07, + "loss": 0.6911, + "step": 4624 + }, + { + "epoch": 0.17, + "learning_rate": 1.8324605412531583e-07, + "loss": 0.6905, + "step": 4625 + }, + { + "epoch": 0.17, + "learning_rate": 1.832423920606438e-07, + "loss": 0.6822, + "step": 4626 + }, + { + "epoch": 0.17, + "learning_rate": 1.8323872999597171e-07, + "loss": 0.6814, + "step": 4627 + }, + { + "epoch": 0.17, + "learning_rate": 1.8323506793129967e-07, + "loss": 0.7021, + "step": 4628 + }, + { + "epoch": 0.17, + "learning_rate": 1.832314058666276e-07, + "loss": 0.6948, + "step": 4629 + }, + { + "epoch": 0.17, + "learning_rate": 1.8322774380195555e-07, + "loss": 0.6885, + "step": 4630 + }, + { + "epoch": 0.17, + "learning_rate": 1.8322408173728347e-07, + "loss": 0.7075, + "step": 4631 + }, + { + "epoch": 0.17, + "learning_rate": 1.8322041967261143e-07, + "loss": 0.7126, + "step": 4632 + }, + { + "epoch": 0.17, + "learning_rate": 1.8321675760793935e-07, + "loss": 0.7033, + "step": 4633 + }, + { + "epoch": 0.17, + "learning_rate": 1.8321309554326728e-07, + "loss": 0.6781, + "step": 4634 + }, + { + "epoch": 0.17, + "learning_rate": 1.8320943347859523e-07, + "loss": 0.6858, + "step": 4635 + }, + { + "epoch": 0.17, + "learning_rate": 1.8320577141392316e-07, + "loss": 0.7006, + "step": 4636 + }, + { + "epoch": 0.17, + "learning_rate": 1.832021093492511e-07, + "loss": 0.6772, + "step": 4637 + }, + { + "epoch": 0.17, + "learning_rate": 1.8319844728457904e-07, + "loss": 0.72, + "step": 4638 + }, + { + "epoch": 0.17, + "learning_rate": 1.8319478521990697e-07, + "loss": 0.6951, + "step": 4639 + }, + { + "epoch": 0.17, + "learning_rate": 1.831911231552349e-07, + "loss": 0.6989, + "step": 4640 + }, + { + "epoch": 0.17, + "learning_rate": 1.8318746109056285e-07, + "loss": 0.691, + "step": 4641 + }, + { + "epoch": 0.17, + "learning_rate": 1.8318379902589077e-07, + "loss": 0.6875, + "step": 4642 + }, + { + "epoch": 0.17, + "learning_rate": 1.8318013696121873e-07, + "loss": 0.6807, + "step": 4643 + }, + { + "epoch": 0.17, + "learning_rate": 1.8317647489654665e-07, + "loss": 0.6975, + "step": 4644 + }, + { + "epoch": 0.17, + "learning_rate": 1.831728128318746e-07, + "loss": 0.6808, + "step": 4645 + }, + { + "epoch": 0.17, + "learning_rate": 1.8316915076720253e-07, + "loss": 0.6783, + "step": 4646 + }, + { + "epoch": 0.17, + "learning_rate": 1.8316548870253049e-07, + "loss": 0.7119, + "step": 4647 + }, + { + "epoch": 0.17, + "learning_rate": 1.831618266378584e-07, + "loss": 0.6663, + "step": 4648 + }, + { + "epoch": 0.17, + "learning_rate": 1.8315816457318637e-07, + "loss": 0.6905, + "step": 4649 + }, + { + "epoch": 0.17, + "learning_rate": 1.831545025085143e-07, + "loss": 0.6996, + "step": 4650 + }, + { + "epoch": 0.17, + "learning_rate": 1.8315084044384225e-07, + "loss": 0.6883, + "step": 4651 + }, + { + "epoch": 0.17, + "learning_rate": 1.8314717837917017e-07, + "loss": 0.6925, + "step": 4652 + }, + { + "epoch": 0.17, + "learning_rate": 1.8314351631449813e-07, + "loss": 0.7067, + "step": 4653 + }, + { + "epoch": 0.17, + "learning_rate": 1.8313985424982605e-07, + "loss": 0.6958, + "step": 4654 + }, + { + "epoch": 0.17, + "learning_rate": 1.83136192185154e-07, + "loss": 0.7054, + "step": 4655 + }, + { + "epoch": 0.17, + "learning_rate": 1.831325301204819e-07, + "loss": 0.7065, + "step": 4656 + }, + { + "epoch": 0.17, + "learning_rate": 1.8312886805580986e-07, + "loss": 0.6824, + "step": 4657 + }, + { + "epoch": 0.17, + "learning_rate": 1.8312520599113779e-07, + "loss": 0.6931, + "step": 4658 + }, + { + "epoch": 0.17, + "learning_rate": 1.8312154392646574e-07, + "loss": 0.6909, + "step": 4659 + }, + { + "epoch": 0.17, + "learning_rate": 1.8311788186179367e-07, + "loss": 0.6867, + "step": 4660 + }, + { + "epoch": 0.17, + "learning_rate": 1.8311421979712162e-07, + "loss": 0.6942, + "step": 4661 + }, + { + "epoch": 0.17, + "learning_rate": 1.8311055773244954e-07, + "loss": 0.6898, + "step": 4662 + }, + { + "epoch": 0.17, + "learning_rate": 1.8310689566777747e-07, + "loss": 0.6921, + "step": 4663 + }, + { + "epoch": 0.17, + "learning_rate": 1.8310323360310542e-07, + "loss": 0.6958, + "step": 4664 + }, + { + "epoch": 0.17, + "learning_rate": 1.8309957153843335e-07, + "loss": 0.7079, + "step": 4665 + }, + { + "epoch": 0.17, + "learning_rate": 1.830959094737613e-07, + "loss": 0.6904, + "step": 4666 + }, + { + "epoch": 0.17, + "learning_rate": 1.8309224740908923e-07, + "loss": 0.705, + "step": 4667 + }, + { + "epoch": 0.17, + "learning_rate": 1.8308858534441718e-07, + "loss": 0.6809, + "step": 4668 + }, + { + "epoch": 0.17, + "learning_rate": 1.830849232797451e-07, + "loss": 0.7014, + "step": 4669 + }, + { + "epoch": 0.17, + "learning_rate": 1.8308126121507306e-07, + "loss": 0.7026, + "step": 4670 + }, + { + "epoch": 0.17, + "learning_rate": 1.83077599150401e-07, + "loss": 0.6953, + "step": 4671 + }, + { + "epoch": 0.17, + "learning_rate": 1.8307393708572894e-07, + "loss": 0.6725, + "step": 4672 + }, + { + "epoch": 0.17, + "learning_rate": 1.8307027502105684e-07, + "loss": 0.6905, + "step": 4673 + }, + { + "epoch": 0.17, + "learning_rate": 1.830666129563848e-07, + "loss": 0.7084, + "step": 4674 + }, + { + "epoch": 0.17, + "learning_rate": 1.8306295089171272e-07, + "loss": 0.6923, + "step": 4675 + }, + { + "epoch": 0.17, + "learning_rate": 1.8305928882704068e-07, + "loss": 0.6923, + "step": 4676 + }, + { + "epoch": 0.17, + "learning_rate": 1.830556267623686e-07, + "loss": 0.6967, + "step": 4677 + }, + { + "epoch": 0.17, + "learning_rate": 1.8305196469769656e-07, + "loss": 0.6999, + "step": 4678 + }, + { + "epoch": 0.17, + "learning_rate": 1.8304830263302448e-07, + "loss": 0.7034, + "step": 4679 + }, + { + "epoch": 0.17, + "learning_rate": 1.8304464056835244e-07, + "loss": 0.6774, + "step": 4680 + }, + { + "epoch": 0.17, + "learning_rate": 1.8304097850368036e-07, + "loss": 0.6967, + "step": 4681 + }, + { + "epoch": 0.17, + "learning_rate": 1.8303731643900832e-07, + "loss": 0.6833, + "step": 4682 + }, + { + "epoch": 0.17, + "learning_rate": 1.8303365437433624e-07, + "loss": 0.6878, + "step": 4683 + }, + { + "epoch": 0.17, + "learning_rate": 1.8302999230966417e-07, + "loss": 0.6955, + "step": 4684 + }, + { + "epoch": 0.17, + "learning_rate": 1.8302633024499212e-07, + "loss": 0.6747, + "step": 4685 + }, + { + "epoch": 0.17, + "learning_rate": 1.8302266818032005e-07, + "loss": 0.6898, + "step": 4686 + }, + { + "epoch": 0.17, + "learning_rate": 1.83019006115648e-07, + "loss": 0.706, + "step": 4687 + }, + { + "epoch": 0.17, + "learning_rate": 1.8301534405097593e-07, + "loss": 0.6968, + "step": 4688 + }, + { + "epoch": 0.17, + "learning_rate": 1.8301168198630388e-07, + "loss": 0.7157, + "step": 4689 + }, + { + "epoch": 0.17, + "learning_rate": 1.830080199216318e-07, + "loss": 0.6834, + "step": 4690 + }, + { + "epoch": 0.17, + "learning_rate": 1.8300435785695974e-07, + "loss": 0.6968, + "step": 4691 + }, + { + "epoch": 0.17, + "learning_rate": 1.8300069579228766e-07, + "loss": 0.6755, + "step": 4692 + }, + { + "epoch": 0.17, + "learning_rate": 1.8299703372761562e-07, + "loss": 0.7187, + "step": 4693 + }, + { + "epoch": 0.17, + "learning_rate": 1.8299337166294354e-07, + "loss": 0.6836, + "step": 4694 + }, + { + "epoch": 0.17, + "learning_rate": 1.829897095982715e-07, + "loss": 0.6816, + "step": 4695 + }, + { + "epoch": 0.17, + "learning_rate": 1.8298604753359942e-07, + "loss": 0.703, + "step": 4696 + }, + { + "epoch": 0.17, + "learning_rate": 1.8298238546892738e-07, + "loss": 0.7037, + "step": 4697 + }, + { + "epoch": 0.17, + "learning_rate": 1.829787234042553e-07, + "loss": 0.6949, + "step": 4698 + }, + { + "epoch": 0.17, + "learning_rate": 1.8297506133958326e-07, + "loss": 0.6971, + "step": 4699 + }, + { + "epoch": 0.17, + "learning_rate": 1.8297139927491118e-07, + "loss": 0.6821, + "step": 4700 + }, + { + "epoch": 0.17, + "learning_rate": 1.8296773721023914e-07, + "loss": 0.6982, + "step": 4701 + }, + { + "epoch": 0.17, + "learning_rate": 1.8296407514556706e-07, + "loss": 0.7109, + "step": 4702 + }, + { + "epoch": 0.17, + "learning_rate": 1.8296041308089502e-07, + "loss": 0.6773, + "step": 4703 + }, + { + "epoch": 0.17, + "learning_rate": 1.8295675101622294e-07, + "loss": 0.6893, + "step": 4704 + }, + { + "epoch": 0.17, + "learning_rate": 1.8295308895155087e-07, + "loss": 0.7183, + "step": 4705 + }, + { + "epoch": 0.17, + "learning_rate": 1.8294942688687882e-07, + "loss": 0.7145, + "step": 4706 + }, + { + "epoch": 0.17, + "learning_rate": 1.8294576482220675e-07, + "loss": 0.6993, + "step": 4707 + }, + { + "epoch": 0.17, + "learning_rate": 1.829421027575347e-07, + "loss": 0.6838, + "step": 4708 + }, + { + "epoch": 0.17, + "learning_rate": 1.8293844069286263e-07, + "loss": 0.6803, + "step": 4709 + }, + { + "epoch": 0.17, + "learning_rate": 1.8293477862819056e-07, + "loss": 0.6626, + "step": 4710 + }, + { + "epoch": 0.17, + "learning_rate": 1.8293111656351848e-07, + "loss": 0.709, + "step": 4711 + }, + { + "epoch": 0.17, + "learning_rate": 1.8292745449884644e-07, + "loss": 0.6718, + "step": 4712 + }, + { + "epoch": 0.17, + "learning_rate": 1.8292379243417436e-07, + "loss": 0.6873, + "step": 4713 + }, + { + "epoch": 0.17, + "learning_rate": 1.8292013036950232e-07, + "loss": 0.6984, + "step": 4714 + }, + { + "epoch": 0.17, + "learning_rate": 1.8291646830483024e-07, + "loss": 0.6874, + "step": 4715 + }, + { + "epoch": 0.17, + "learning_rate": 1.829128062401582e-07, + "loss": 0.6908, + "step": 4716 + }, + { + "epoch": 0.17, + "learning_rate": 1.8290914417548612e-07, + "loss": 0.72, + "step": 4717 + }, + { + "epoch": 0.17, + "learning_rate": 1.8290548211081408e-07, + "loss": 0.6924, + "step": 4718 + }, + { + "epoch": 0.17, + "learning_rate": 1.82901820046142e-07, + "loss": 0.703, + "step": 4719 + }, + { + "epoch": 0.17, + "learning_rate": 1.8289815798146995e-07, + "loss": 0.6914, + "step": 4720 + }, + { + "epoch": 0.17, + "learning_rate": 1.8289449591679788e-07, + "loss": 0.6768, + "step": 4721 + }, + { + "epoch": 0.17, + "learning_rate": 1.8289083385212583e-07, + "loss": 0.6904, + "step": 4722 + }, + { + "epoch": 0.17, + "learning_rate": 1.8288717178745376e-07, + "loss": 0.7039, + "step": 4723 + }, + { + "epoch": 0.17, + "learning_rate": 1.8288350972278171e-07, + "loss": 0.6837, + "step": 4724 + }, + { + "epoch": 0.17, + "learning_rate": 1.8287984765810964e-07, + "loss": 0.6996, + "step": 4725 + }, + { + "epoch": 0.17, + "learning_rate": 1.828761855934376e-07, + "loss": 0.7014, + "step": 4726 + }, + { + "epoch": 0.17, + "learning_rate": 1.828725235287655e-07, + "loss": 0.6842, + "step": 4727 + }, + { + "epoch": 0.17, + "learning_rate": 1.8286886146409345e-07, + "loss": 0.6855, + "step": 4728 + }, + { + "epoch": 0.17, + "learning_rate": 1.8286519939942137e-07, + "loss": 0.6918, + "step": 4729 + }, + { + "epoch": 0.17, + "learning_rate": 1.8286153733474933e-07, + "loss": 0.6871, + "step": 4730 + }, + { + "epoch": 0.17, + "learning_rate": 1.8285787527007725e-07, + "loss": 0.6859, + "step": 4731 + }, + { + "epoch": 0.17, + "learning_rate": 1.828542132054052e-07, + "loss": 0.6916, + "step": 4732 + }, + { + "epoch": 0.17, + "learning_rate": 1.8285055114073313e-07, + "loss": 0.7019, + "step": 4733 + }, + { + "epoch": 0.17, + "learning_rate": 1.8284688907606106e-07, + "loss": 0.7234, + "step": 4734 + }, + { + "epoch": 0.17, + "learning_rate": 1.8284322701138901e-07, + "loss": 0.6847, + "step": 4735 + }, + { + "epoch": 0.17, + "learning_rate": 1.8283956494671694e-07, + "loss": 0.6799, + "step": 4736 + }, + { + "epoch": 0.17, + "learning_rate": 1.828359028820449e-07, + "loss": 0.7136, + "step": 4737 + }, + { + "epoch": 0.17, + "learning_rate": 1.8283224081737282e-07, + "loss": 0.6826, + "step": 4738 + }, + { + "epoch": 0.17, + "learning_rate": 1.8282857875270077e-07, + "loss": 0.6918, + "step": 4739 + }, + { + "epoch": 0.17, + "learning_rate": 1.828249166880287e-07, + "loss": 0.6767, + "step": 4740 + }, + { + "epoch": 0.17, + "learning_rate": 1.8282125462335665e-07, + "loss": 0.7084, + "step": 4741 + }, + { + "epoch": 0.17, + "learning_rate": 1.8281759255868458e-07, + "loss": 0.7074, + "step": 4742 + }, + { + "epoch": 0.17, + "learning_rate": 1.8281393049401253e-07, + "loss": 0.6711, + "step": 4743 + }, + { + "epoch": 0.17, + "learning_rate": 1.8281026842934046e-07, + "loss": 0.6848, + "step": 4744 + }, + { + "epoch": 0.17, + "learning_rate": 1.828066063646684e-07, + "loss": 0.6583, + "step": 4745 + }, + { + "epoch": 0.17, + "learning_rate": 1.8280294429999631e-07, + "loss": 0.6832, + "step": 4746 + }, + { + "epoch": 0.17, + "learning_rate": 1.8279928223532427e-07, + "loss": 0.6925, + "step": 4747 + }, + { + "epoch": 0.17, + "learning_rate": 1.827956201706522e-07, + "loss": 0.714, + "step": 4748 + }, + { + "epoch": 0.17, + "learning_rate": 1.8279195810598015e-07, + "loss": 0.7121, + "step": 4749 + }, + { + "epoch": 0.17, + "learning_rate": 1.8278829604130807e-07, + "loss": 0.6842, + "step": 4750 + }, + { + "epoch": 0.17, + "learning_rate": 1.8278463397663603e-07, + "loss": 0.687, + "step": 4751 + }, + { + "epoch": 0.17, + "learning_rate": 1.8278097191196395e-07, + "loss": 0.706, + "step": 4752 + }, + { + "epoch": 0.17, + "learning_rate": 1.827773098472919e-07, + "loss": 0.6947, + "step": 4753 + }, + { + "epoch": 0.17, + "learning_rate": 1.8277364778261983e-07, + "loss": 0.698, + "step": 4754 + }, + { + "epoch": 0.17, + "learning_rate": 1.8276998571794776e-07, + "loss": 0.699, + "step": 4755 + }, + { + "epoch": 0.17, + "learning_rate": 1.827663236532757e-07, + "loss": 0.6898, + "step": 4756 + }, + { + "epoch": 0.17, + "learning_rate": 1.8276266158860364e-07, + "loss": 0.7115, + "step": 4757 + }, + { + "epoch": 0.17, + "learning_rate": 1.827589995239316e-07, + "loss": 0.7059, + "step": 4758 + }, + { + "epoch": 0.17, + "learning_rate": 1.8275533745925952e-07, + "loss": 0.6955, + "step": 4759 + }, + { + "epoch": 0.17, + "learning_rate": 1.8275167539458747e-07, + "loss": 0.6814, + "step": 4760 + }, + { + "epoch": 0.17, + "learning_rate": 1.827480133299154e-07, + "loss": 0.6859, + "step": 4761 + }, + { + "epoch": 0.17, + "learning_rate": 1.8274435126524335e-07, + "loss": 0.702, + "step": 4762 + }, + { + "epoch": 0.17, + "learning_rate": 1.8274068920057125e-07, + "loss": 0.6853, + "step": 4763 + }, + { + "epoch": 0.17, + "learning_rate": 1.827370271358992e-07, + "loss": 0.679, + "step": 4764 + }, + { + "epoch": 0.17, + "learning_rate": 1.8273336507122713e-07, + "loss": 0.69, + "step": 4765 + }, + { + "epoch": 0.17, + "learning_rate": 1.8272970300655509e-07, + "loss": 0.7065, + "step": 4766 + }, + { + "epoch": 0.17, + "learning_rate": 1.82726040941883e-07, + "loss": 0.6824, + "step": 4767 + }, + { + "epoch": 0.17, + "learning_rate": 1.8272237887721097e-07, + "loss": 0.6912, + "step": 4768 + }, + { + "epoch": 0.17, + "learning_rate": 1.827187168125389e-07, + "loss": 0.7058, + "step": 4769 + }, + { + "epoch": 0.17, + "learning_rate": 1.8271505474786685e-07, + "loss": 0.6919, + "step": 4770 + }, + { + "epoch": 0.17, + "learning_rate": 1.8271139268319477e-07, + "loss": 0.7004, + "step": 4771 + }, + { + "epoch": 0.17, + "learning_rate": 1.8270773061852273e-07, + "loss": 0.7068, + "step": 4772 + }, + { + "epoch": 0.17, + "learning_rate": 1.8270406855385065e-07, + "loss": 0.712, + "step": 4773 + }, + { + "epoch": 0.17, + "learning_rate": 1.827004064891786e-07, + "loss": 0.7001, + "step": 4774 + }, + { + "epoch": 0.17, + "learning_rate": 1.8269674442450653e-07, + "loss": 0.6846, + "step": 4775 + }, + { + "epoch": 0.17, + "learning_rate": 1.8269308235983446e-07, + "loss": 0.7074, + "step": 4776 + }, + { + "epoch": 0.17, + "learning_rate": 1.826894202951624e-07, + "loss": 0.6924, + "step": 4777 + }, + { + "epoch": 0.17, + "learning_rate": 1.8268575823049034e-07, + "loss": 0.6936, + "step": 4778 + }, + { + "epoch": 0.17, + "learning_rate": 1.826820961658183e-07, + "loss": 0.6981, + "step": 4779 + }, + { + "epoch": 0.17, + "learning_rate": 1.8267843410114622e-07, + "loss": 0.7059, + "step": 4780 + }, + { + "epoch": 0.17, + "learning_rate": 1.8267477203647415e-07, + "loss": 0.6888, + "step": 4781 + }, + { + "epoch": 0.17, + "learning_rate": 1.8267110997180207e-07, + "loss": 0.6871, + "step": 4782 + }, + { + "epoch": 0.17, + "learning_rate": 1.8266744790713002e-07, + "loss": 0.7184, + "step": 4783 + }, + { + "epoch": 0.18, + "learning_rate": 1.8266378584245795e-07, + "loss": 0.6727, + "step": 4784 + }, + { + "epoch": 0.18, + "learning_rate": 1.826601237777859e-07, + "loss": 0.7038, + "step": 4785 + }, + { + "epoch": 0.18, + "learning_rate": 1.8265646171311383e-07, + "loss": 0.6959, + "step": 4786 + }, + { + "epoch": 0.18, + "learning_rate": 1.8265279964844178e-07, + "loss": 0.6978, + "step": 4787 + }, + { + "epoch": 0.18, + "learning_rate": 1.826491375837697e-07, + "loss": 0.6859, + "step": 4788 + }, + { + "epoch": 0.18, + "learning_rate": 1.8264547551909766e-07, + "loss": 0.7079, + "step": 4789 + }, + { + "epoch": 0.18, + "learning_rate": 1.826418134544256e-07, + "loss": 0.7002, + "step": 4790 + }, + { + "epoch": 0.18, + "learning_rate": 1.8263815138975354e-07, + "loss": 0.6993, + "step": 4791 + }, + { + "epoch": 0.18, + "learning_rate": 1.8263448932508147e-07, + "loss": 0.7159, + "step": 4792 + }, + { + "epoch": 0.18, + "learning_rate": 1.8263082726040942e-07, + "loss": 0.7007, + "step": 4793 + }, + { + "epoch": 0.18, + "learning_rate": 1.8262716519573735e-07, + "loss": 0.6829, + "step": 4794 + }, + { + "epoch": 0.18, + "learning_rate": 1.826235031310653e-07, + "loss": 0.6926, + "step": 4795 + }, + { + "epoch": 0.18, + "learning_rate": 1.8261984106639323e-07, + "loss": 0.6831, + "step": 4796 + }, + { + "epoch": 0.18, + "learning_rate": 1.8261617900172118e-07, + "loss": 0.7025, + "step": 4797 + }, + { + "epoch": 0.18, + "learning_rate": 1.826125169370491e-07, + "loss": 0.6885, + "step": 4798 + }, + { + "epoch": 0.18, + "learning_rate": 1.8260885487237704e-07, + "loss": 0.7089, + "step": 4799 + }, + { + "epoch": 0.18, + "learning_rate": 1.8260519280770496e-07, + "loss": 0.6963, + "step": 4800 + }, + { + "epoch": 0.18, + "learning_rate": 1.8260153074303292e-07, + "loss": 0.6787, + "step": 4801 + }, + { + "epoch": 0.18, + "learning_rate": 1.8259786867836084e-07, + "loss": 0.6947, + "step": 4802 + }, + { + "epoch": 0.18, + "learning_rate": 1.825942066136888e-07, + "loss": 0.6935, + "step": 4803 + }, + { + "epoch": 0.18, + "learning_rate": 1.8259054454901672e-07, + "loss": 0.6847, + "step": 4804 + }, + { + "epoch": 0.18, + "learning_rate": 1.8258688248434465e-07, + "loss": 0.6776, + "step": 4805 + }, + { + "epoch": 0.18, + "learning_rate": 1.825832204196726e-07, + "loss": 0.7098, + "step": 4806 + }, + { + "epoch": 0.18, + "learning_rate": 1.8257955835500053e-07, + "loss": 0.687, + "step": 4807 + }, + { + "epoch": 0.18, + "learning_rate": 1.8257589629032848e-07, + "loss": 0.7146, + "step": 4808 + }, + { + "epoch": 0.18, + "learning_rate": 1.825722342256564e-07, + "loss": 0.6954, + "step": 4809 + }, + { + "epoch": 0.18, + "learning_rate": 1.8256857216098436e-07, + "loss": 0.6951, + "step": 4810 + }, + { + "epoch": 0.18, + "learning_rate": 1.825649100963123e-07, + "loss": 0.6966, + "step": 4811 + }, + { + "epoch": 0.18, + "learning_rate": 1.8256124803164024e-07, + "loss": 0.6961, + "step": 4812 + }, + { + "epoch": 0.18, + "learning_rate": 1.8255758596696817e-07, + "loss": 0.7016, + "step": 4813 + }, + { + "epoch": 0.18, + "learning_rate": 1.8255392390229612e-07, + "loss": 0.6997, + "step": 4814 + }, + { + "epoch": 0.18, + "learning_rate": 1.8255026183762405e-07, + "loss": 0.716, + "step": 4815 + }, + { + "epoch": 0.18, + "learning_rate": 1.82546599772952e-07, + "loss": 0.6873, + "step": 4816 + }, + { + "epoch": 0.18, + "learning_rate": 1.825429377082799e-07, + "loss": 0.6769, + "step": 4817 + }, + { + "epoch": 0.18, + "learning_rate": 1.8253927564360786e-07, + "loss": 0.6784, + "step": 4818 + }, + { + "epoch": 0.18, + "learning_rate": 1.8253561357893578e-07, + "loss": 0.6938, + "step": 4819 + }, + { + "epoch": 0.18, + "learning_rate": 1.8253195151426374e-07, + "loss": 0.7018, + "step": 4820 + }, + { + "epoch": 0.18, + "learning_rate": 1.8252828944959166e-07, + "loss": 0.6819, + "step": 4821 + }, + { + "epoch": 0.18, + "learning_rate": 1.8252462738491962e-07, + "loss": 0.6981, + "step": 4822 + }, + { + "epoch": 0.18, + "learning_rate": 1.8252096532024754e-07, + "loss": 0.6835, + "step": 4823 + }, + { + "epoch": 0.18, + "learning_rate": 1.825173032555755e-07, + "loss": 0.6918, + "step": 4824 + }, + { + "epoch": 0.18, + "learning_rate": 1.8251364119090342e-07, + "loss": 0.6689, + "step": 4825 + }, + { + "epoch": 0.18, + "learning_rate": 1.8250997912623135e-07, + "loss": 0.6508, + "step": 4826 + }, + { + "epoch": 0.18, + "learning_rate": 1.825063170615593e-07, + "loss": 0.6758, + "step": 4827 + }, + { + "epoch": 0.18, + "learning_rate": 1.8250265499688723e-07, + "loss": 0.6891, + "step": 4828 + }, + { + "epoch": 0.18, + "learning_rate": 1.8249899293221518e-07, + "loss": 0.7095, + "step": 4829 + }, + { + "epoch": 0.18, + "learning_rate": 1.824953308675431e-07, + "loss": 0.7149, + "step": 4830 + }, + { + "epoch": 0.18, + "learning_rate": 1.8249166880287106e-07, + "loss": 0.7131, + "step": 4831 + }, + { + "epoch": 0.18, + "learning_rate": 1.82488006738199e-07, + "loss": 0.6757, + "step": 4832 + }, + { + "epoch": 0.18, + "learning_rate": 1.8248434467352694e-07, + "loss": 0.6937, + "step": 4833 + }, + { + "epoch": 0.18, + "learning_rate": 1.8248068260885487e-07, + "loss": 0.6897, + "step": 4834 + }, + { + "epoch": 0.18, + "learning_rate": 1.824770205441828e-07, + "loss": 0.718, + "step": 4835 + }, + { + "epoch": 0.18, + "learning_rate": 1.8247335847951072e-07, + "loss": 0.6831, + "step": 4836 + }, + { + "epoch": 0.18, + "learning_rate": 1.8246969641483868e-07, + "loss": 0.6905, + "step": 4837 + }, + { + "epoch": 0.18, + "learning_rate": 1.824660343501666e-07, + "loss": 0.7108, + "step": 4838 + }, + { + "epoch": 0.18, + "learning_rate": 1.8246237228549455e-07, + "loss": 0.696, + "step": 4839 + }, + { + "epoch": 0.18, + "learning_rate": 1.8245871022082248e-07, + "loss": 0.6964, + "step": 4840 + }, + { + "epoch": 0.18, + "learning_rate": 1.8245504815615043e-07, + "loss": 0.7052, + "step": 4841 + }, + { + "epoch": 0.18, + "learning_rate": 1.8245138609147836e-07, + "loss": 0.6922, + "step": 4842 + }, + { + "epoch": 0.18, + "learning_rate": 1.8244772402680631e-07, + "loss": 0.6987, + "step": 4843 + }, + { + "epoch": 0.18, + "learning_rate": 1.8244406196213424e-07, + "loss": 0.7097, + "step": 4844 + }, + { + "epoch": 0.18, + "learning_rate": 1.824403998974622e-07, + "loss": 0.6958, + "step": 4845 + }, + { + "epoch": 0.18, + "learning_rate": 1.8243673783279012e-07, + "loss": 0.6936, + "step": 4846 + }, + { + "epoch": 0.18, + "learning_rate": 1.8243307576811805e-07, + "loss": 0.6903, + "step": 4847 + }, + { + "epoch": 0.18, + "learning_rate": 1.82429413703446e-07, + "loss": 0.6758, + "step": 4848 + }, + { + "epoch": 0.18, + "learning_rate": 1.8242575163877393e-07, + "loss": 0.6817, + "step": 4849 + }, + { + "epoch": 0.18, + "learning_rate": 1.8242208957410188e-07, + "loss": 0.6848, + "step": 4850 + }, + { + "epoch": 0.18, + "learning_rate": 1.824184275094298e-07, + "loss": 0.6899, + "step": 4851 + }, + { + "epoch": 0.18, + "learning_rate": 1.8241476544475776e-07, + "loss": 0.6997, + "step": 4852 + }, + { + "epoch": 0.18, + "learning_rate": 1.8241110338008566e-07, + "loss": 0.6813, + "step": 4853 + }, + { + "epoch": 0.18, + "learning_rate": 1.8240744131541361e-07, + "loss": 0.6824, + "step": 4854 + }, + { + "epoch": 0.18, + "learning_rate": 1.8240377925074154e-07, + "loss": 0.7094, + "step": 4855 + }, + { + "epoch": 0.18, + "learning_rate": 1.824001171860695e-07, + "loss": 0.6871, + "step": 4856 + }, + { + "epoch": 0.18, + "learning_rate": 1.8239645512139742e-07, + "loss": 0.702, + "step": 4857 + }, + { + "epoch": 0.18, + "learning_rate": 1.8239279305672537e-07, + "loss": 0.7055, + "step": 4858 + }, + { + "epoch": 0.18, + "learning_rate": 1.823891309920533e-07, + "loss": 0.699, + "step": 4859 + }, + { + "epoch": 0.18, + "learning_rate": 1.8238546892738125e-07, + "loss": 0.7146, + "step": 4860 + }, + { + "epoch": 0.18, + "learning_rate": 1.8238180686270918e-07, + "loss": 0.7053, + "step": 4861 + }, + { + "epoch": 0.18, + "learning_rate": 1.8237814479803713e-07, + "loss": 0.6964, + "step": 4862 + }, + { + "epoch": 0.18, + "learning_rate": 1.8237448273336506e-07, + "loss": 0.6909, + "step": 4863 + }, + { + "epoch": 0.18, + "learning_rate": 1.8237082066869301e-07, + "loss": 0.6884, + "step": 4864 + }, + { + "epoch": 0.18, + "learning_rate": 1.8236715860402094e-07, + "loss": 0.7185, + "step": 4865 + }, + { + "epoch": 0.18, + "learning_rate": 1.823634965393489e-07, + "loss": 0.6961, + "step": 4866 + }, + { + "epoch": 0.18, + "learning_rate": 1.8235983447467682e-07, + "loss": 0.675, + "step": 4867 + }, + { + "epoch": 0.18, + "learning_rate": 1.8235617241000477e-07, + "loss": 0.6873, + "step": 4868 + }, + { + "epoch": 0.18, + "learning_rate": 1.823525103453327e-07, + "loss": 0.7043, + "step": 4869 + }, + { + "epoch": 0.18, + "learning_rate": 1.8234884828066063e-07, + "loss": 0.6701, + "step": 4870 + }, + { + "epoch": 0.18, + "learning_rate": 1.8234518621598855e-07, + "loss": 0.6868, + "step": 4871 + }, + { + "epoch": 0.18, + "learning_rate": 1.823415241513165e-07, + "loss": 0.678, + "step": 4872 + }, + { + "epoch": 0.18, + "learning_rate": 1.8233786208664443e-07, + "loss": 0.6847, + "step": 4873 + }, + { + "epoch": 0.18, + "learning_rate": 1.8233420002197236e-07, + "loss": 0.6803, + "step": 4874 + }, + { + "epoch": 0.18, + "learning_rate": 1.823305379573003e-07, + "loss": 0.7033, + "step": 4875 + }, + { + "epoch": 0.18, + "learning_rate": 1.8232687589262824e-07, + "loss": 0.6666, + "step": 4876 + }, + { + "epoch": 0.18, + "learning_rate": 1.823232138279562e-07, + "loss": 0.7004, + "step": 4877 + }, + { + "epoch": 0.18, + "learning_rate": 1.8231955176328412e-07, + "loss": 0.7063, + "step": 4878 + }, + { + "epoch": 0.18, + "learning_rate": 1.8231588969861207e-07, + "loss": 0.6994, + "step": 4879 + }, + { + "epoch": 0.18, + "learning_rate": 1.8231222763394e-07, + "loss": 0.6872, + "step": 4880 + }, + { + "epoch": 0.18, + "learning_rate": 1.8230856556926795e-07, + "loss": 0.6834, + "step": 4881 + }, + { + "epoch": 0.18, + "learning_rate": 1.8230490350459588e-07, + "loss": 0.7186, + "step": 4882 + }, + { + "epoch": 0.18, + "learning_rate": 1.8230124143992383e-07, + "loss": 0.698, + "step": 4883 + }, + { + "epoch": 0.18, + "learning_rate": 1.8229757937525176e-07, + "loss": 0.6962, + "step": 4884 + }, + { + "epoch": 0.18, + "learning_rate": 1.822939173105797e-07, + "loss": 0.7193, + "step": 4885 + }, + { + "epoch": 0.18, + "learning_rate": 1.8229025524590764e-07, + "loss": 0.7067, + "step": 4886 + }, + { + "epoch": 0.18, + "learning_rate": 1.822865931812356e-07, + "loss": 0.6761, + "step": 4887 + }, + { + "epoch": 0.18, + "learning_rate": 1.8228293111656352e-07, + "loss": 0.6831, + "step": 4888 + }, + { + "epoch": 0.18, + "learning_rate": 1.8227926905189145e-07, + "loss": 0.6854, + "step": 4889 + }, + { + "epoch": 0.18, + "learning_rate": 1.8227560698721937e-07, + "loss": 0.6801, + "step": 4890 + }, + { + "epoch": 0.18, + "learning_rate": 1.8227194492254733e-07, + "loss": 0.699, + "step": 4891 + }, + { + "epoch": 0.18, + "learning_rate": 1.8226828285787525e-07, + "loss": 0.6907, + "step": 4892 + }, + { + "epoch": 0.18, + "learning_rate": 1.822646207932032e-07, + "loss": 0.7088, + "step": 4893 + }, + { + "epoch": 0.18, + "learning_rate": 1.8226095872853113e-07, + "loss": 0.6879, + "step": 4894 + }, + { + "epoch": 0.18, + "learning_rate": 1.8225729666385909e-07, + "loss": 0.7053, + "step": 4895 + }, + { + "epoch": 0.18, + "learning_rate": 1.82253634599187e-07, + "loss": 0.6879, + "step": 4896 + }, + { + "epoch": 0.18, + "learning_rate": 1.8224997253451494e-07, + "loss": 0.6776, + "step": 4897 + }, + { + "epoch": 0.18, + "learning_rate": 1.822463104698429e-07, + "loss": 0.6855, + "step": 4898 + }, + { + "epoch": 0.18, + "learning_rate": 1.8224264840517082e-07, + "loss": 0.6893, + "step": 4899 + }, + { + "epoch": 0.18, + "learning_rate": 1.8223898634049877e-07, + "loss": 0.7104, + "step": 4900 + }, + { + "epoch": 0.18, + "learning_rate": 1.822353242758267e-07, + "loss": 0.6965, + "step": 4901 + }, + { + "epoch": 0.18, + "learning_rate": 1.8223166221115465e-07, + "loss": 0.6804, + "step": 4902 + }, + { + "epoch": 0.18, + "learning_rate": 1.8222800014648258e-07, + "loss": 0.6881, + "step": 4903 + }, + { + "epoch": 0.18, + "learning_rate": 1.8222433808181053e-07, + "loss": 0.6984, + "step": 4904 + }, + { + "epoch": 0.18, + "learning_rate": 1.8222067601713846e-07, + "loss": 0.7131, + "step": 4905 + }, + { + "epoch": 0.18, + "learning_rate": 1.822170139524664e-07, + "loss": 0.7218, + "step": 4906 + }, + { + "epoch": 0.18, + "learning_rate": 1.822133518877943e-07, + "loss": 0.6945, + "step": 4907 + }, + { + "epoch": 0.18, + "learning_rate": 1.8220968982312226e-07, + "loss": 0.6997, + "step": 4908 + }, + { + "epoch": 0.18, + "learning_rate": 1.822060277584502e-07, + "loss": 0.7215, + "step": 4909 + }, + { + "epoch": 0.18, + "learning_rate": 1.8220236569377814e-07, + "loss": 0.7069, + "step": 4910 + }, + { + "epoch": 0.18, + "learning_rate": 1.8219870362910607e-07, + "loss": 0.6885, + "step": 4911 + }, + { + "epoch": 0.18, + "learning_rate": 1.8219504156443402e-07, + "loss": 0.7037, + "step": 4912 + }, + { + "epoch": 0.18, + "learning_rate": 1.8219137949976195e-07, + "loss": 0.7086, + "step": 4913 + }, + { + "epoch": 0.18, + "learning_rate": 1.821877174350899e-07, + "loss": 0.6972, + "step": 4914 + }, + { + "epoch": 0.18, + "learning_rate": 1.8218405537041783e-07, + "loss": 0.7026, + "step": 4915 + }, + { + "epoch": 0.18, + "learning_rate": 1.8218039330574578e-07, + "loss": 0.6903, + "step": 4916 + }, + { + "epoch": 0.18, + "learning_rate": 1.821767312410737e-07, + "loss": 0.6991, + "step": 4917 + }, + { + "epoch": 0.18, + "learning_rate": 1.8217306917640164e-07, + "loss": 0.6885, + "step": 4918 + }, + { + "epoch": 0.18, + "learning_rate": 1.821694071117296e-07, + "loss": 0.7082, + "step": 4919 + }, + { + "epoch": 0.18, + "learning_rate": 1.8216574504705752e-07, + "loss": 0.6726, + "step": 4920 + }, + { + "epoch": 0.18, + "learning_rate": 1.8216208298238547e-07, + "loss": 0.6889, + "step": 4921 + }, + { + "epoch": 0.18, + "learning_rate": 1.821584209177134e-07, + "loss": 0.6979, + "step": 4922 + }, + { + "epoch": 0.18, + "learning_rate": 1.8215475885304135e-07, + "loss": 0.6884, + "step": 4923 + }, + { + "epoch": 0.18, + "learning_rate": 1.8215109678836928e-07, + "loss": 0.681, + "step": 4924 + }, + { + "epoch": 0.18, + "learning_rate": 1.821474347236972e-07, + "loss": 0.6925, + "step": 4925 + }, + { + "epoch": 0.18, + "learning_rate": 1.8214377265902513e-07, + "loss": 0.6796, + "step": 4926 + }, + { + "epoch": 0.18, + "learning_rate": 1.8214011059435308e-07, + "loss": 0.6928, + "step": 4927 + }, + { + "epoch": 0.18, + "learning_rate": 1.82136448529681e-07, + "loss": 0.6918, + "step": 4928 + }, + { + "epoch": 0.18, + "learning_rate": 1.8213278646500896e-07, + "loss": 0.7079, + "step": 4929 + }, + { + "epoch": 0.18, + "learning_rate": 1.821291244003369e-07, + "loss": 0.6879, + "step": 4930 + }, + { + "epoch": 0.18, + "learning_rate": 1.8212546233566484e-07, + "loss": 0.6872, + "step": 4931 + }, + { + "epoch": 0.18, + "learning_rate": 1.8212180027099277e-07, + "loss": 0.6881, + "step": 4932 + }, + { + "epoch": 0.18, + "learning_rate": 1.8211813820632072e-07, + "loss": 0.7052, + "step": 4933 + }, + { + "epoch": 0.18, + "learning_rate": 1.8211447614164865e-07, + "loss": 0.6835, + "step": 4934 + }, + { + "epoch": 0.18, + "learning_rate": 1.821108140769766e-07, + "loss": 0.6843, + "step": 4935 + }, + { + "epoch": 0.18, + "learning_rate": 1.8210715201230453e-07, + "loss": 0.6873, + "step": 4936 + }, + { + "epoch": 0.18, + "learning_rate": 1.8210348994763248e-07, + "loss": 0.7016, + "step": 4937 + }, + { + "epoch": 0.18, + "learning_rate": 1.820998278829604e-07, + "loss": 0.7056, + "step": 4938 + }, + { + "epoch": 0.18, + "learning_rate": 1.8209616581828836e-07, + "loss": 0.6986, + "step": 4939 + }, + { + "epoch": 0.18, + "learning_rate": 1.820925037536163e-07, + "loss": 0.7134, + "step": 4940 + }, + { + "epoch": 0.18, + "learning_rate": 1.8208884168894422e-07, + "loss": 0.6896, + "step": 4941 + }, + { + "epoch": 0.18, + "learning_rate": 1.8208517962427217e-07, + "loss": 0.7044, + "step": 4942 + }, + { + "epoch": 0.18, + "learning_rate": 1.820815175596001e-07, + "loss": 0.6979, + "step": 4943 + }, + { + "epoch": 0.18, + "learning_rate": 1.8207785549492802e-07, + "loss": 0.7178, + "step": 4944 + }, + { + "epoch": 0.18, + "learning_rate": 1.8207419343025595e-07, + "loss": 0.6866, + "step": 4945 + }, + { + "epoch": 0.18, + "learning_rate": 1.820705313655839e-07, + "loss": 0.6949, + "step": 4946 + }, + { + "epoch": 0.18, + "learning_rate": 1.8206686930091183e-07, + "loss": 0.6619, + "step": 4947 + }, + { + "epoch": 0.18, + "learning_rate": 1.8206320723623978e-07, + "loss": 0.6749, + "step": 4948 + }, + { + "epoch": 0.18, + "learning_rate": 1.820595451715677e-07, + "loss": 0.7136, + "step": 4949 + }, + { + "epoch": 0.18, + "learning_rate": 1.8205588310689566e-07, + "loss": 0.7088, + "step": 4950 + }, + { + "epoch": 0.18, + "learning_rate": 1.820522210422236e-07, + "loss": 0.6967, + "step": 4951 + }, + { + "epoch": 0.18, + "learning_rate": 1.8204855897755154e-07, + "loss": 0.6867, + "step": 4952 + }, + { + "epoch": 0.18, + "learning_rate": 1.8204489691287947e-07, + "loss": 0.6942, + "step": 4953 + }, + { + "epoch": 0.18, + "learning_rate": 1.8204123484820742e-07, + "loss": 0.6834, + "step": 4954 + }, + { + "epoch": 0.18, + "learning_rate": 1.8203757278353535e-07, + "loss": 0.6914, + "step": 4955 + }, + { + "epoch": 0.18, + "learning_rate": 1.820339107188633e-07, + "loss": 0.6942, + "step": 4956 + }, + { + "epoch": 0.18, + "learning_rate": 1.8203024865419123e-07, + "loss": 0.6834, + "step": 4957 + }, + { + "epoch": 0.18, + "learning_rate": 1.8202658658951918e-07, + "loss": 0.6931, + "step": 4958 + }, + { + "epoch": 0.18, + "learning_rate": 1.820229245248471e-07, + "loss": 0.6917, + "step": 4959 + }, + { + "epoch": 0.18, + "learning_rate": 1.8201926246017506e-07, + "loss": 0.6761, + "step": 4960 + }, + { + "epoch": 0.18, + "learning_rate": 1.8201560039550296e-07, + "loss": 0.6925, + "step": 4961 + }, + { + "epoch": 0.18, + "learning_rate": 1.8201193833083091e-07, + "loss": 0.6874, + "step": 4962 + }, + { + "epoch": 0.18, + "learning_rate": 1.8200827626615884e-07, + "loss": 0.6932, + "step": 4963 + }, + { + "epoch": 0.18, + "learning_rate": 1.820046142014868e-07, + "loss": 0.7086, + "step": 4964 + }, + { + "epoch": 0.18, + "learning_rate": 1.8200095213681472e-07, + "loss": 0.6981, + "step": 4965 + }, + { + "epoch": 0.18, + "learning_rate": 1.8199729007214267e-07, + "loss": 0.7048, + "step": 4966 + }, + { + "epoch": 0.18, + "learning_rate": 1.819936280074706e-07, + "loss": 0.7008, + "step": 4967 + }, + { + "epoch": 0.18, + "learning_rate": 1.8198996594279853e-07, + "loss": 0.6749, + "step": 4968 + }, + { + "epoch": 0.18, + "learning_rate": 1.8198630387812648e-07, + "loss": 0.708, + "step": 4969 + }, + { + "epoch": 0.18, + "learning_rate": 1.819826418134544e-07, + "loss": 0.6861, + "step": 4970 + }, + { + "epoch": 0.18, + "learning_rate": 1.8197897974878236e-07, + "loss": 0.6925, + "step": 4971 + }, + { + "epoch": 0.18, + "learning_rate": 1.819753176841103e-07, + "loss": 0.7103, + "step": 4972 + }, + { + "epoch": 0.18, + "learning_rate": 1.8197165561943824e-07, + "loss": 0.691, + "step": 4973 + }, + { + "epoch": 0.18, + "learning_rate": 1.8196799355476617e-07, + "loss": 0.698, + "step": 4974 + }, + { + "epoch": 0.18, + "learning_rate": 1.8196433149009412e-07, + "loss": 0.6661, + "step": 4975 + }, + { + "epoch": 0.18, + "learning_rate": 1.8196066942542205e-07, + "loss": 0.701, + "step": 4976 + }, + { + "epoch": 0.18, + "learning_rate": 1.8195700736075e-07, + "loss": 0.7038, + "step": 4977 + }, + { + "epoch": 0.18, + "learning_rate": 1.8195334529607793e-07, + "loss": 0.6825, + "step": 4978 + }, + { + "epoch": 0.18, + "learning_rate": 1.8194968323140585e-07, + "loss": 0.678, + "step": 4979 + }, + { + "epoch": 0.18, + "learning_rate": 1.8194602116673378e-07, + "loss": 0.6937, + "step": 4980 + }, + { + "epoch": 0.18, + "learning_rate": 1.8194235910206173e-07, + "loss": 0.6795, + "step": 4981 + }, + { + "epoch": 0.18, + "learning_rate": 1.8193869703738966e-07, + "loss": 0.7061, + "step": 4982 + }, + { + "epoch": 0.18, + "learning_rate": 1.8193503497271761e-07, + "loss": 0.6836, + "step": 4983 + }, + { + "epoch": 0.18, + "learning_rate": 1.8193137290804554e-07, + "loss": 0.703, + "step": 4984 + }, + { + "epoch": 0.18, + "learning_rate": 1.819277108433735e-07, + "loss": 0.7119, + "step": 4985 + }, + { + "epoch": 0.18, + "learning_rate": 1.8192404877870142e-07, + "loss": 0.7162, + "step": 4986 + }, + { + "epoch": 0.18, + "learning_rate": 1.8192038671402937e-07, + "loss": 0.6966, + "step": 4987 + }, + { + "epoch": 0.18, + "learning_rate": 1.819167246493573e-07, + "loss": 0.6988, + "step": 4988 + }, + { + "epoch": 0.18, + "learning_rate": 1.8191306258468523e-07, + "loss": 0.7014, + "step": 4989 + }, + { + "epoch": 0.18, + "learning_rate": 1.8190940052001318e-07, + "loss": 0.6555, + "step": 4990 + }, + { + "epoch": 0.18, + "learning_rate": 1.819057384553411e-07, + "loss": 0.7326, + "step": 4991 + }, + { + "epoch": 0.18, + "learning_rate": 1.8190207639066906e-07, + "loss": 0.7256, + "step": 4992 + }, + { + "epoch": 0.18, + "learning_rate": 1.8189841432599699e-07, + "loss": 0.6994, + "step": 4993 + }, + { + "epoch": 0.18, + "learning_rate": 1.8189475226132494e-07, + "loss": 0.6735, + "step": 4994 + }, + { + "epoch": 0.18, + "learning_rate": 1.8189109019665287e-07, + "loss": 0.7009, + "step": 4995 + }, + { + "epoch": 0.18, + "learning_rate": 1.8188742813198082e-07, + "loss": 0.6832, + "step": 4996 + }, + { + "epoch": 0.18, + "learning_rate": 1.8188376606730872e-07, + "loss": 0.6851, + "step": 4997 + }, + { + "epoch": 0.18, + "learning_rate": 1.8188010400263667e-07, + "loss": 0.6916, + "step": 4998 + }, + { + "epoch": 0.18, + "learning_rate": 1.818764419379646e-07, + "loss": 0.7, + "step": 4999 + }, + { + "epoch": 0.18, + "learning_rate": 1.8187277987329255e-07, + "loss": 0.718, + "step": 5000 + }, + { + "epoch": 0.18, + "learning_rate": 1.8186911780862048e-07, + "loss": 0.69, + "step": 5001 + }, + { + "epoch": 0.18, + "learning_rate": 1.8186545574394843e-07, + "loss": 0.6823, + "step": 5002 + }, + { + "epoch": 0.18, + "learning_rate": 1.8186179367927636e-07, + "loss": 0.7155, + "step": 5003 + }, + { + "epoch": 0.18, + "learning_rate": 1.818581316146043e-07, + "loss": 0.711, + "step": 5004 + }, + { + "epoch": 0.18, + "learning_rate": 1.8185446954993224e-07, + "loss": 0.6918, + "step": 5005 + }, + { + "epoch": 0.18, + "learning_rate": 1.818508074852602e-07, + "loss": 0.6964, + "step": 5006 + }, + { + "epoch": 0.18, + "learning_rate": 1.8184714542058812e-07, + "loss": 0.7112, + "step": 5007 + }, + { + "epoch": 0.18, + "learning_rate": 1.8184348335591607e-07, + "loss": 0.7068, + "step": 5008 + }, + { + "epoch": 0.18, + "learning_rate": 1.81839821291244e-07, + "loss": 0.6895, + "step": 5009 + }, + { + "epoch": 0.18, + "learning_rate": 1.8183615922657195e-07, + "loss": 0.6921, + "step": 5010 + }, + { + "epoch": 0.18, + "learning_rate": 1.8183249716189988e-07, + "loss": 0.6874, + "step": 5011 + }, + { + "epoch": 0.18, + "learning_rate": 1.818288350972278e-07, + "loss": 0.6776, + "step": 5012 + }, + { + "epoch": 0.18, + "learning_rate": 1.8182517303255576e-07, + "loss": 0.6934, + "step": 5013 + }, + { + "epoch": 0.18, + "learning_rate": 1.8182151096788369e-07, + "loss": 0.6894, + "step": 5014 + }, + { + "epoch": 0.18, + "learning_rate": 1.818178489032116e-07, + "loss": 0.7056, + "step": 5015 + }, + { + "epoch": 0.18, + "learning_rate": 1.8181418683853954e-07, + "loss": 0.6995, + "step": 5016 + }, + { + "epoch": 0.18, + "learning_rate": 1.818105247738675e-07, + "loss": 0.6796, + "step": 5017 + }, + { + "epoch": 0.18, + "learning_rate": 1.8180686270919542e-07, + "loss": 0.6887, + "step": 5018 + }, + { + "epoch": 0.18, + "learning_rate": 1.8180320064452337e-07, + "loss": 0.6781, + "step": 5019 + }, + { + "epoch": 0.18, + "learning_rate": 1.817995385798513e-07, + "loss": 0.6852, + "step": 5020 + }, + { + "epoch": 0.18, + "learning_rate": 1.8179587651517925e-07, + "loss": 0.7082, + "step": 5021 + }, + { + "epoch": 0.18, + "learning_rate": 1.8179221445050718e-07, + "loss": 0.6786, + "step": 5022 + }, + { + "epoch": 0.18, + "learning_rate": 1.8178855238583513e-07, + "loss": 0.6824, + "step": 5023 + }, + { + "epoch": 0.18, + "learning_rate": 1.8178489032116306e-07, + "loss": 0.6957, + "step": 5024 + }, + { + "epoch": 0.18, + "learning_rate": 1.81781228256491e-07, + "loss": 0.6891, + "step": 5025 + }, + { + "epoch": 0.18, + "learning_rate": 1.8177756619181894e-07, + "loss": 0.704, + "step": 5026 + }, + { + "epoch": 0.18, + "learning_rate": 1.817739041271469e-07, + "loss": 0.705, + "step": 5027 + }, + { + "epoch": 0.18, + "learning_rate": 1.8177024206247482e-07, + "loss": 0.676, + "step": 5028 + }, + { + "epoch": 0.18, + "learning_rate": 1.8176657999780277e-07, + "loss": 0.6999, + "step": 5029 + }, + { + "epoch": 0.18, + "learning_rate": 1.817629179331307e-07, + "loss": 0.7184, + "step": 5030 + }, + { + "epoch": 0.18, + "learning_rate": 1.8175925586845865e-07, + "loss": 0.6669, + "step": 5031 + }, + { + "epoch": 0.18, + "learning_rate": 1.8175559380378658e-07, + "loss": 0.707, + "step": 5032 + }, + { + "epoch": 0.18, + "learning_rate": 1.817519317391145e-07, + "loss": 0.6833, + "step": 5033 + }, + { + "epoch": 0.18, + "learning_rate": 1.8174826967444243e-07, + "loss": 0.6933, + "step": 5034 + }, + { + "epoch": 0.18, + "learning_rate": 1.8174460760977038e-07, + "loss": 0.7029, + "step": 5035 + }, + { + "epoch": 0.18, + "learning_rate": 1.817409455450983e-07, + "loss": 0.6789, + "step": 5036 + }, + { + "epoch": 0.18, + "learning_rate": 1.8173728348042626e-07, + "loss": 0.688, + "step": 5037 + }, + { + "epoch": 0.18, + "learning_rate": 1.817336214157542e-07, + "loss": 0.7027, + "step": 5038 + }, + { + "epoch": 0.18, + "learning_rate": 1.8172995935108212e-07, + "loss": 0.7028, + "step": 5039 + }, + { + "epoch": 0.18, + "learning_rate": 1.8172629728641007e-07, + "loss": 0.6733, + "step": 5040 + }, + { + "epoch": 0.18, + "learning_rate": 1.81722635221738e-07, + "loss": 0.7008, + "step": 5041 + }, + { + "epoch": 0.18, + "learning_rate": 1.8171897315706595e-07, + "loss": 0.7014, + "step": 5042 + }, + { + "epoch": 0.18, + "learning_rate": 1.8171531109239388e-07, + "loss": 0.7386, + "step": 5043 + }, + { + "epoch": 0.18, + "learning_rate": 1.8171164902772183e-07, + "loss": 0.6789, + "step": 5044 + }, + { + "epoch": 0.18, + "learning_rate": 1.8170798696304976e-07, + "loss": 0.6682, + "step": 5045 + }, + { + "epoch": 0.18, + "learning_rate": 1.817043248983777e-07, + "loss": 0.6998, + "step": 5046 + }, + { + "epoch": 0.18, + "learning_rate": 1.8170066283370564e-07, + "loss": 0.6961, + "step": 5047 + }, + { + "epoch": 0.18, + "learning_rate": 1.816970007690336e-07, + "loss": 0.6992, + "step": 5048 + }, + { + "epoch": 0.18, + "learning_rate": 1.8169333870436152e-07, + "loss": 0.6941, + "step": 5049 + }, + { + "epoch": 0.18, + "learning_rate": 1.8168967663968944e-07, + "loss": 0.6963, + "step": 5050 + }, + { + "epoch": 0.18, + "learning_rate": 1.8168601457501737e-07, + "loss": 0.6978, + "step": 5051 + }, + { + "epoch": 0.18, + "learning_rate": 1.8168235251034532e-07, + "loss": 0.7042, + "step": 5052 + }, + { + "epoch": 0.18, + "learning_rate": 1.8167869044567325e-07, + "loss": 0.7142, + "step": 5053 + }, + { + "epoch": 0.18, + "learning_rate": 1.816750283810012e-07, + "loss": 0.6895, + "step": 5054 + }, + { + "epoch": 0.18, + "learning_rate": 1.8167136631632913e-07, + "loss": 0.6953, + "step": 5055 + }, + { + "epoch": 0.18, + "learning_rate": 1.8166770425165708e-07, + "loss": 0.7073, + "step": 5056 + }, + { + "epoch": 0.19, + "learning_rate": 1.81664042186985e-07, + "loss": 0.6928, + "step": 5057 + }, + { + "epoch": 0.19, + "learning_rate": 1.8166038012231296e-07, + "loss": 0.6923, + "step": 5058 + }, + { + "epoch": 0.19, + "learning_rate": 1.816567180576409e-07, + "loss": 0.7305, + "step": 5059 + }, + { + "epoch": 0.19, + "learning_rate": 1.8165305599296882e-07, + "loss": 0.6973, + "step": 5060 + }, + { + "epoch": 0.19, + "learning_rate": 1.8164939392829677e-07, + "loss": 0.6852, + "step": 5061 + }, + { + "epoch": 0.19, + "learning_rate": 1.816457318636247e-07, + "loss": 0.6982, + "step": 5062 + }, + { + "epoch": 0.19, + "learning_rate": 1.8164206979895265e-07, + "loss": 0.6934, + "step": 5063 + }, + { + "epoch": 0.19, + "learning_rate": 1.8163840773428058e-07, + "loss": 0.6919, + "step": 5064 + }, + { + "epoch": 0.19, + "learning_rate": 1.8163474566960853e-07, + "loss": 0.7044, + "step": 5065 + }, + { + "epoch": 0.19, + "learning_rate": 1.8163108360493646e-07, + "loss": 0.7159, + "step": 5066 + }, + { + "epoch": 0.19, + "learning_rate": 1.816274215402644e-07, + "loss": 0.6798, + "step": 5067 + }, + { + "epoch": 0.19, + "learning_rate": 1.816237594755923e-07, + "loss": 0.6756, + "step": 5068 + }, + { + "epoch": 0.19, + "learning_rate": 1.8162009741092026e-07, + "loss": 0.7071, + "step": 5069 + }, + { + "epoch": 0.19, + "learning_rate": 1.816164353462482e-07, + "loss": 0.6965, + "step": 5070 + }, + { + "epoch": 0.19, + "learning_rate": 1.8161277328157614e-07, + "loss": 0.681, + "step": 5071 + }, + { + "epoch": 0.19, + "learning_rate": 1.8160911121690407e-07, + "loss": 0.6961, + "step": 5072 + }, + { + "epoch": 0.19, + "learning_rate": 1.8160544915223202e-07, + "loss": 0.704, + "step": 5073 + }, + { + "epoch": 0.19, + "learning_rate": 1.8160178708755995e-07, + "loss": 0.6897, + "step": 5074 + }, + { + "epoch": 0.19, + "learning_rate": 1.815981250228879e-07, + "loss": 0.6976, + "step": 5075 + }, + { + "epoch": 0.19, + "learning_rate": 1.8159446295821583e-07, + "loss": 0.6959, + "step": 5076 + }, + { + "epoch": 0.19, + "learning_rate": 1.8159080089354378e-07, + "loss": 0.7058, + "step": 5077 + }, + { + "epoch": 0.19, + "learning_rate": 1.815871388288717e-07, + "loss": 0.7075, + "step": 5078 + }, + { + "epoch": 0.19, + "learning_rate": 1.8158347676419966e-07, + "loss": 0.6889, + "step": 5079 + }, + { + "epoch": 0.19, + "learning_rate": 1.815798146995276e-07, + "loss": 0.6969, + "step": 5080 + }, + { + "epoch": 0.19, + "learning_rate": 1.8157615263485554e-07, + "loss": 0.6863, + "step": 5081 + }, + { + "epoch": 0.19, + "learning_rate": 1.8157249057018347e-07, + "loss": 0.6792, + "step": 5082 + }, + { + "epoch": 0.19, + "learning_rate": 1.815688285055114e-07, + "loss": 0.7032, + "step": 5083 + }, + { + "epoch": 0.19, + "learning_rate": 1.8156516644083935e-07, + "loss": 0.691, + "step": 5084 + }, + { + "epoch": 0.19, + "learning_rate": 1.8156150437616727e-07, + "loss": 0.6795, + "step": 5085 + }, + { + "epoch": 0.19, + "learning_rate": 1.815578423114952e-07, + "loss": 0.6955, + "step": 5086 + }, + { + "epoch": 0.19, + "learning_rate": 1.8155418024682313e-07, + "loss": 0.6839, + "step": 5087 + }, + { + "epoch": 0.19, + "learning_rate": 1.8155051818215108e-07, + "loss": 0.7214, + "step": 5088 + }, + { + "epoch": 0.19, + "learning_rate": 1.81546856117479e-07, + "loss": 0.6886, + "step": 5089 + }, + { + "epoch": 0.19, + "learning_rate": 1.8154319405280696e-07, + "loss": 0.7177, + "step": 5090 + }, + { + "epoch": 0.19, + "learning_rate": 1.815395319881349e-07, + "loss": 0.7007, + "step": 5091 + }, + { + "epoch": 0.19, + "learning_rate": 1.8153586992346284e-07, + "loss": 0.6888, + "step": 5092 + }, + { + "epoch": 0.19, + "learning_rate": 1.8153220785879077e-07, + "loss": 0.6752, + "step": 5093 + }, + { + "epoch": 0.19, + "learning_rate": 1.8152854579411872e-07, + "loss": 0.7186, + "step": 5094 + }, + { + "epoch": 0.19, + "learning_rate": 1.8152488372944665e-07, + "loss": 0.6933, + "step": 5095 + }, + { + "epoch": 0.19, + "learning_rate": 1.815212216647746e-07, + "loss": 0.705, + "step": 5096 + }, + { + "epoch": 0.19, + "learning_rate": 1.8151755960010253e-07, + "loss": 0.6982, + "step": 5097 + }, + { + "epoch": 0.19, + "learning_rate": 1.8151389753543048e-07, + "loss": 0.6574, + "step": 5098 + }, + { + "epoch": 0.19, + "learning_rate": 1.815102354707584e-07, + "loss": 0.7082, + "step": 5099 + }, + { + "epoch": 0.19, + "learning_rate": 1.8150657340608636e-07, + "loss": 0.6938, + "step": 5100 + }, + { + "epoch": 0.19, + "learning_rate": 1.815029113414143e-07, + "loss": 0.703, + "step": 5101 + }, + { + "epoch": 0.19, + "learning_rate": 1.8149924927674224e-07, + "loss": 0.6917, + "step": 5102 + }, + { + "epoch": 0.19, + "learning_rate": 1.8149558721207017e-07, + "loss": 0.7267, + "step": 5103 + }, + { + "epoch": 0.19, + "learning_rate": 1.814919251473981e-07, + "loss": 0.7161, + "step": 5104 + }, + { + "epoch": 0.19, + "learning_rate": 1.8148826308272602e-07, + "loss": 0.7278, + "step": 5105 + }, + { + "epoch": 0.19, + "learning_rate": 1.8148460101805397e-07, + "loss": 0.7033, + "step": 5106 + }, + { + "epoch": 0.19, + "learning_rate": 1.814809389533819e-07, + "loss": 0.6812, + "step": 5107 + }, + { + "epoch": 0.19, + "learning_rate": 1.8147727688870985e-07, + "loss": 0.6832, + "step": 5108 + }, + { + "epoch": 0.19, + "learning_rate": 1.8147361482403778e-07, + "loss": 0.6819, + "step": 5109 + }, + { + "epoch": 0.19, + "learning_rate": 1.814699527593657e-07, + "loss": 0.7007, + "step": 5110 + }, + { + "epoch": 0.19, + "learning_rate": 1.8146629069469366e-07, + "loss": 0.6637, + "step": 5111 + }, + { + "epoch": 0.19, + "learning_rate": 1.8146262863002159e-07, + "loss": 0.7038, + "step": 5112 + }, + { + "epoch": 0.19, + "learning_rate": 1.8145896656534954e-07, + "loss": 0.7044, + "step": 5113 + }, + { + "epoch": 0.19, + "learning_rate": 1.8145530450067747e-07, + "loss": 0.6815, + "step": 5114 + }, + { + "epoch": 0.19, + "learning_rate": 1.8145164243600542e-07, + "loss": 0.6864, + "step": 5115 + }, + { + "epoch": 0.19, + "learning_rate": 1.8144798037133335e-07, + "loss": 0.6947, + "step": 5116 + }, + { + "epoch": 0.19, + "learning_rate": 1.814443183066613e-07, + "loss": 0.6765, + "step": 5117 + }, + { + "epoch": 0.19, + "learning_rate": 1.8144065624198923e-07, + "loss": 0.6941, + "step": 5118 + }, + { + "epoch": 0.19, + "learning_rate": 1.8143699417731718e-07, + "loss": 0.6968, + "step": 5119 + }, + { + "epoch": 0.19, + "learning_rate": 1.814333321126451e-07, + "loss": 0.6882, + "step": 5120 + }, + { + "epoch": 0.19, + "learning_rate": 1.8142967004797306e-07, + "loss": 0.6859, + "step": 5121 + }, + { + "epoch": 0.19, + "learning_rate": 1.8142600798330096e-07, + "loss": 0.6899, + "step": 5122 + }, + { + "epoch": 0.19, + "learning_rate": 1.814223459186289e-07, + "loss": 0.6818, + "step": 5123 + }, + { + "epoch": 0.19, + "learning_rate": 1.8141868385395684e-07, + "loss": 0.6947, + "step": 5124 + }, + { + "epoch": 0.19, + "learning_rate": 1.814150217892848e-07, + "loss": 0.7147, + "step": 5125 + }, + { + "epoch": 0.19, + "learning_rate": 1.8141135972461272e-07, + "loss": 0.7003, + "step": 5126 + }, + { + "epoch": 0.19, + "learning_rate": 1.8140769765994067e-07, + "loss": 0.6814, + "step": 5127 + }, + { + "epoch": 0.19, + "learning_rate": 1.814040355952686e-07, + "loss": 0.699, + "step": 5128 + }, + { + "epoch": 0.19, + "learning_rate": 1.8140037353059655e-07, + "loss": 0.6916, + "step": 5129 + }, + { + "epoch": 0.19, + "learning_rate": 1.8139671146592448e-07, + "loss": 0.6843, + "step": 5130 + }, + { + "epoch": 0.19, + "learning_rate": 1.813930494012524e-07, + "loss": 0.7121, + "step": 5131 + }, + { + "epoch": 0.19, + "learning_rate": 1.8138938733658036e-07, + "loss": 0.7024, + "step": 5132 + }, + { + "epoch": 0.19, + "learning_rate": 1.8138572527190829e-07, + "loss": 0.7052, + "step": 5133 + }, + { + "epoch": 0.19, + "learning_rate": 1.8138206320723624e-07, + "loss": 0.6657, + "step": 5134 + }, + { + "epoch": 0.19, + "learning_rate": 1.8137840114256417e-07, + "loss": 0.6796, + "step": 5135 + }, + { + "epoch": 0.19, + "learning_rate": 1.8137473907789212e-07, + "loss": 0.7043, + "step": 5136 + }, + { + "epoch": 0.19, + "learning_rate": 1.8137107701322004e-07, + "loss": 0.7026, + "step": 5137 + }, + { + "epoch": 0.19, + "learning_rate": 1.81367414948548e-07, + "loss": 0.712, + "step": 5138 + }, + { + "epoch": 0.19, + "learning_rate": 1.8136375288387592e-07, + "loss": 0.6868, + "step": 5139 + }, + { + "epoch": 0.19, + "learning_rate": 1.8136009081920385e-07, + "loss": 0.6915, + "step": 5140 + }, + { + "epoch": 0.19, + "learning_rate": 1.8135642875453178e-07, + "loss": 0.6882, + "step": 5141 + }, + { + "epoch": 0.19, + "learning_rate": 1.8135276668985973e-07, + "loss": 0.7032, + "step": 5142 + }, + { + "epoch": 0.19, + "learning_rate": 1.8134910462518766e-07, + "loss": 0.6763, + "step": 5143 + }, + { + "epoch": 0.19, + "learning_rate": 1.813454425605156e-07, + "loss": 0.6878, + "step": 5144 + }, + { + "epoch": 0.19, + "learning_rate": 1.8134178049584354e-07, + "loss": 0.6813, + "step": 5145 + }, + { + "epoch": 0.19, + "learning_rate": 1.813381184311715e-07, + "loss": 0.6737, + "step": 5146 + }, + { + "epoch": 0.19, + "learning_rate": 1.8133445636649942e-07, + "loss": 0.7016, + "step": 5147 + }, + { + "epoch": 0.19, + "learning_rate": 1.8133079430182737e-07, + "loss": 0.6916, + "step": 5148 + }, + { + "epoch": 0.19, + "learning_rate": 1.813271322371553e-07, + "loss": 0.6935, + "step": 5149 + }, + { + "epoch": 0.19, + "learning_rate": 1.8132347017248325e-07, + "loss": 0.705, + "step": 5150 + }, + { + "epoch": 0.19, + "learning_rate": 1.8131980810781118e-07, + "loss": 0.6894, + "step": 5151 + }, + { + "epoch": 0.19, + "learning_rate": 1.8131614604313913e-07, + "loss": 0.7028, + "step": 5152 + }, + { + "epoch": 0.19, + "learning_rate": 1.8131248397846706e-07, + "loss": 0.7079, + "step": 5153 + }, + { + "epoch": 0.19, + "learning_rate": 1.8130882191379498e-07, + "loss": 0.6983, + "step": 5154 + }, + { + "epoch": 0.19, + "learning_rate": 1.8130515984912294e-07, + "loss": 0.6876, + "step": 5155 + }, + { + "epoch": 0.19, + "learning_rate": 1.8130149778445086e-07, + "loss": 0.6832, + "step": 5156 + }, + { + "epoch": 0.19, + "learning_rate": 1.8129783571977882e-07, + "loss": 0.6835, + "step": 5157 + }, + { + "epoch": 0.19, + "learning_rate": 1.8129417365510672e-07, + "loss": 0.692, + "step": 5158 + }, + { + "epoch": 0.19, + "learning_rate": 1.8129051159043467e-07, + "loss": 0.7014, + "step": 5159 + }, + { + "epoch": 0.19, + "learning_rate": 1.812868495257626e-07, + "loss": 0.6907, + "step": 5160 + }, + { + "epoch": 0.19, + "learning_rate": 1.8128318746109055e-07, + "loss": 0.6873, + "step": 5161 + }, + { + "epoch": 0.19, + "learning_rate": 1.8127952539641848e-07, + "loss": 0.701, + "step": 5162 + }, + { + "epoch": 0.19, + "learning_rate": 1.8127586333174643e-07, + "loss": 0.6839, + "step": 5163 + }, + { + "epoch": 0.19, + "learning_rate": 1.8127220126707436e-07, + "loss": 0.7044, + "step": 5164 + }, + { + "epoch": 0.19, + "learning_rate": 1.812685392024023e-07, + "loss": 0.6853, + "step": 5165 + }, + { + "epoch": 0.19, + "learning_rate": 1.8126487713773024e-07, + "loss": 0.6921, + "step": 5166 + }, + { + "epoch": 0.19, + "learning_rate": 1.812612150730582e-07, + "loss": 0.6849, + "step": 5167 + }, + { + "epoch": 0.19, + "learning_rate": 1.8125755300838612e-07, + "loss": 0.7084, + "step": 5168 + }, + { + "epoch": 0.19, + "learning_rate": 1.8125389094371407e-07, + "loss": 0.6993, + "step": 5169 + }, + { + "epoch": 0.19, + "learning_rate": 1.81250228879042e-07, + "loss": 0.7128, + "step": 5170 + }, + { + "epoch": 0.19, + "learning_rate": 1.8124656681436995e-07, + "loss": 0.7002, + "step": 5171 + }, + { + "epoch": 0.19, + "learning_rate": 1.8124290474969788e-07, + "loss": 0.6982, + "step": 5172 + }, + { + "epoch": 0.19, + "learning_rate": 1.8123924268502583e-07, + "loss": 0.7046, + "step": 5173 + }, + { + "epoch": 0.19, + "learning_rate": 1.8123558062035376e-07, + "loss": 0.6929, + "step": 5174 + }, + { + "epoch": 0.19, + "learning_rate": 1.8123191855568168e-07, + "loss": 0.6701, + "step": 5175 + }, + { + "epoch": 0.19, + "learning_rate": 1.812282564910096e-07, + "loss": 0.6923, + "step": 5176 + }, + { + "epoch": 0.19, + "learning_rate": 1.8122459442633756e-07, + "loss": 0.691, + "step": 5177 + }, + { + "epoch": 0.19, + "learning_rate": 1.812209323616655e-07, + "loss": 0.6978, + "step": 5178 + }, + { + "epoch": 0.19, + "learning_rate": 1.8121727029699344e-07, + "loss": 0.6897, + "step": 5179 + }, + { + "epoch": 0.19, + "learning_rate": 1.8121360823232137e-07, + "loss": 0.7078, + "step": 5180 + }, + { + "epoch": 0.19, + "learning_rate": 1.812099461676493e-07, + "loss": 0.6936, + "step": 5181 + }, + { + "epoch": 0.19, + "learning_rate": 1.8120628410297725e-07, + "loss": 0.7039, + "step": 5182 + }, + { + "epoch": 0.19, + "learning_rate": 1.8120262203830518e-07, + "loss": 0.6992, + "step": 5183 + }, + { + "epoch": 0.19, + "learning_rate": 1.8119895997363313e-07, + "loss": 0.7045, + "step": 5184 + }, + { + "epoch": 0.19, + "learning_rate": 1.8119529790896106e-07, + "loss": 0.6865, + "step": 5185 + }, + { + "epoch": 0.19, + "learning_rate": 1.81191635844289e-07, + "loss": 0.676, + "step": 5186 + }, + { + "epoch": 0.19, + "learning_rate": 1.8118797377961694e-07, + "loss": 0.6839, + "step": 5187 + }, + { + "epoch": 0.19, + "learning_rate": 1.811843117149449e-07, + "loss": 0.6737, + "step": 5188 + }, + { + "epoch": 0.19, + "learning_rate": 1.8118064965027282e-07, + "loss": 0.6966, + "step": 5189 + }, + { + "epoch": 0.19, + "learning_rate": 1.8117698758560077e-07, + "loss": 0.6908, + "step": 5190 + }, + { + "epoch": 0.19, + "learning_rate": 1.811733255209287e-07, + "loss": 0.7174, + "step": 5191 + }, + { + "epoch": 0.19, + "learning_rate": 1.8116966345625665e-07, + "loss": 0.6959, + "step": 5192 + }, + { + "epoch": 0.19, + "learning_rate": 1.8116600139158458e-07, + "loss": 0.714, + "step": 5193 + }, + { + "epoch": 0.19, + "learning_rate": 1.811623393269125e-07, + "loss": 0.6827, + "step": 5194 + }, + { + "epoch": 0.19, + "learning_rate": 1.8115867726224043e-07, + "loss": 0.7048, + "step": 5195 + }, + { + "epoch": 0.19, + "learning_rate": 1.8115501519756838e-07, + "loss": 0.7051, + "step": 5196 + }, + { + "epoch": 0.19, + "learning_rate": 1.811513531328963e-07, + "loss": 0.6947, + "step": 5197 + }, + { + "epoch": 0.19, + "learning_rate": 1.8114769106822426e-07, + "loss": 0.7058, + "step": 5198 + }, + { + "epoch": 0.19, + "learning_rate": 1.811440290035522e-07, + "loss": 0.6886, + "step": 5199 + }, + { + "epoch": 0.19, + "learning_rate": 1.8114036693888014e-07, + "loss": 0.68, + "step": 5200 + }, + { + "epoch": 0.19, + "learning_rate": 1.8113670487420807e-07, + "loss": 0.7049, + "step": 5201 + }, + { + "epoch": 0.19, + "learning_rate": 1.81133042809536e-07, + "loss": 0.7024, + "step": 5202 + }, + { + "epoch": 0.19, + "learning_rate": 1.8112938074486395e-07, + "loss": 0.6979, + "step": 5203 + }, + { + "epoch": 0.19, + "learning_rate": 1.8112571868019187e-07, + "loss": 0.6912, + "step": 5204 + }, + { + "epoch": 0.19, + "learning_rate": 1.8112205661551983e-07, + "loss": 0.6803, + "step": 5205 + }, + { + "epoch": 0.19, + "learning_rate": 1.8111839455084775e-07, + "loss": 0.6926, + "step": 5206 + }, + { + "epoch": 0.19, + "learning_rate": 1.811147324861757e-07, + "loss": 0.6987, + "step": 5207 + }, + { + "epoch": 0.19, + "learning_rate": 1.8111107042150363e-07, + "loss": 0.7106, + "step": 5208 + }, + { + "epoch": 0.19, + "learning_rate": 1.811074083568316e-07, + "loss": 0.6956, + "step": 5209 + }, + { + "epoch": 0.19, + "learning_rate": 1.8110374629215951e-07, + "loss": 0.7065, + "step": 5210 + }, + { + "epoch": 0.19, + "learning_rate": 1.8110008422748747e-07, + "loss": 0.6887, + "step": 5211 + }, + { + "epoch": 0.19, + "learning_rate": 1.8109642216281537e-07, + "loss": 0.6972, + "step": 5212 + }, + { + "epoch": 0.19, + "learning_rate": 1.8109276009814332e-07, + "loss": 0.686, + "step": 5213 + }, + { + "epoch": 0.19, + "learning_rate": 1.8108909803347125e-07, + "loss": 0.687, + "step": 5214 + }, + { + "epoch": 0.19, + "learning_rate": 1.810854359687992e-07, + "loss": 0.688, + "step": 5215 + }, + { + "epoch": 0.19, + "learning_rate": 1.8108177390412713e-07, + "loss": 0.6728, + "step": 5216 + }, + { + "epoch": 0.19, + "learning_rate": 1.8107811183945508e-07, + "loss": 0.6799, + "step": 5217 + }, + { + "epoch": 0.19, + "learning_rate": 1.81074449774783e-07, + "loss": 0.7019, + "step": 5218 + }, + { + "epoch": 0.19, + "learning_rate": 1.8107078771011096e-07, + "loss": 0.6971, + "step": 5219 + }, + { + "epoch": 0.19, + "learning_rate": 1.810671256454389e-07, + "loss": 0.703, + "step": 5220 + }, + { + "epoch": 0.19, + "learning_rate": 1.8106346358076684e-07, + "loss": 0.692, + "step": 5221 + }, + { + "epoch": 0.19, + "learning_rate": 1.8105980151609477e-07, + "loss": 0.6861, + "step": 5222 + }, + { + "epoch": 0.19, + "learning_rate": 1.8105613945142272e-07, + "loss": 0.6881, + "step": 5223 + }, + { + "epoch": 0.19, + "learning_rate": 1.8105247738675065e-07, + "loss": 0.7023, + "step": 5224 + }, + { + "epoch": 0.19, + "learning_rate": 1.8104881532207857e-07, + "loss": 0.6994, + "step": 5225 + }, + { + "epoch": 0.19, + "learning_rate": 1.8104515325740653e-07, + "loss": 0.6957, + "step": 5226 + }, + { + "epoch": 0.19, + "learning_rate": 1.8104149119273445e-07, + "loss": 0.6958, + "step": 5227 + }, + { + "epoch": 0.19, + "learning_rate": 1.810378291280624e-07, + "loss": 0.692, + "step": 5228 + }, + { + "epoch": 0.19, + "learning_rate": 1.8103416706339033e-07, + "loss": 0.706, + "step": 5229 + }, + { + "epoch": 0.19, + "learning_rate": 1.8103050499871826e-07, + "loss": 0.701, + "step": 5230 + }, + { + "epoch": 0.19, + "learning_rate": 1.8102684293404619e-07, + "loss": 0.6763, + "step": 5231 + }, + { + "epoch": 0.19, + "learning_rate": 1.8102318086937414e-07, + "loss": 0.6997, + "step": 5232 + }, + { + "epoch": 0.19, + "learning_rate": 1.8101951880470207e-07, + "loss": 0.6916, + "step": 5233 + }, + { + "epoch": 0.19, + "learning_rate": 1.8101585674003002e-07, + "loss": 0.6737, + "step": 5234 + }, + { + "epoch": 0.19, + "learning_rate": 1.8101219467535795e-07, + "loss": 0.6967, + "step": 5235 + }, + { + "epoch": 0.19, + "learning_rate": 1.810085326106859e-07, + "loss": 0.6926, + "step": 5236 + }, + { + "epoch": 0.19, + "learning_rate": 1.8100487054601383e-07, + "loss": 0.7072, + "step": 5237 + }, + { + "epoch": 0.19, + "learning_rate": 1.8100120848134178e-07, + "loss": 0.6848, + "step": 5238 + }, + { + "epoch": 0.19, + "learning_rate": 1.809975464166697e-07, + "loss": 0.683, + "step": 5239 + }, + { + "epoch": 0.19, + "learning_rate": 1.8099388435199766e-07, + "loss": 0.6998, + "step": 5240 + }, + { + "epoch": 0.19, + "learning_rate": 1.8099022228732559e-07, + "loss": 0.7214, + "step": 5241 + }, + { + "epoch": 0.19, + "learning_rate": 1.8098656022265354e-07, + "loss": 0.6962, + "step": 5242 + }, + { + "epoch": 0.19, + "learning_rate": 1.8098289815798147e-07, + "loss": 0.6854, + "step": 5243 + }, + { + "epoch": 0.19, + "learning_rate": 1.8097923609330942e-07, + "loss": 0.6837, + "step": 5244 + }, + { + "epoch": 0.19, + "learning_rate": 1.8097557402863735e-07, + "loss": 0.708, + "step": 5245 + }, + { + "epoch": 0.19, + "learning_rate": 1.8097191196396527e-07, + "loss": 0.6877, + "step": 5246 + }, + { + "epoch": 0.19, + "learning_rate": 1.8096824989929323e-07, + "loss": 0.7081, + "step": 5247 + }, + { + "epoch": 0.19, + "learning_rate": 1.8096458783462115e-07, + "loss": 0.6906, + "step": 5248 + }, + { + "epoch": 0.19, + "learning_rate": 1.8096092576994908e-07, + "loss": 0.6776, + "step": 5249 + }, + { + "epoch": 0.19, + "learning_rate": 1.8095726370527703e-07, + "loss": 0.6911, + "step": 5250 + }, + { + "epoch": 0.19, + "learning_rate": 1.8095360164060496e-07, + "loss": 0.6969, + "step": 5251 + }, + { + "epoch": 0.19, + "learning_rate": 1.8094993957593289e-07, + "loss": 0.691, + "step": 5252 + }, + { + "epoch": 0.19, + "learning_rate": 1.8094627751126084e-07, + "loss": 0.6915, + "step": 5253 + }, + { + "epoch": 0.19, + "learning_rate": 1.8094261544658877e-07, + "loss": 0.7007, + "step": 5254 + }, + { + "epoch": 0.19, + "learning_rate": 1.8093895338191672e-07, + "loss": 0.6836, + "step": 5255 + }, + { + "epoch": 0.19, + "learning_rate": 1.8093529131724464e-07, + "loss": 0.6786, + "step": 5256 + }, + { + "epoch": 0.19, + "learning_rate": 1.809316292525726e-07, + "loss": 0.6918, + "step": 5257 + }, + { + "epoch": 0.19, + "learning_rate": 1.8092796718790052e-07, + "loss": 0.7123, + "step": 5258 + }, + { + "epoch": 0.19, + "learning_rate": 1.8092430512322848e-07, + "loss": 0.6904, + "step": 5259 + }, + { + "epoch": 0.19, + "learning_rate": 1.809206430585564e-07, + "loss": 0.6895, + "step": 5260 + }, + { + "epoch": 0.19, + "learning_rate": 1.8091698099388436e-07, + "loss": 0.6754, + "step": 5261 + }, + { + "epoch": 0.19, + "learning_rate": 1.8091331892921228e-07, + "loss": 0.6936, + "step": 5262 + }, + { + "epoch": 0.19, + "learning_rate": 1.8090965686454024e-07, + "loss": 0.691, + "step": 5263 + }, + { + "epoch": 0.19, + "learning_rate": 1.8090599479986816e-07, + "loss": 0.6907, + "step": 5264 + }, + { + "epoch": 0.19, + "learning_rate": 1.8090233273519612e-07, + "loss": 0.6772, + "step": 5265 + }, + { + "epoch": 0.19, + "learning_rate": 1.8089867067052402e-07, + "loss": 0.683, + "step": 5266 + }, + { + "epoch": 0.19, + "learning_rate": 1.8089500860585197e-07, + "loss": 0.6892, + "step": 5267 + }, + { + "epoch": 0.19, + "learning_rate": 1.808913465411799e-07, + "loss": 0.6916, + "step": 5268 + }, + { + "epoch": 0.19, + "learning_rate": 1.8088768447650785e-07, + "loss": 0.6999, + "step": 5269 + }, + { + "epoch": 0.19, + "learning_rate": 1.8088402241183578e-07, + "loss": 0.6815, + "step": 5270 + }, + { + "epoch": 0.19, + "learning_rate": 1.8088036034716373e-07, + "loss": 0.6933, + "step": 5271 + }, + { + "epoch": 0.19, + "learning_rate": 1.8087669828249166e-07, + "loss": 0.6792, + "step": 5272 + }, + { + "epoch": 0.19, + "learning_rate": 1.8087303621781958e-07, + "loss": 0.6951, + "step": 5273 + }, + { + "epoch": 0.19, + "learning_rate": 1.8086937415314754e-07, + "loss": 0.6985, + "step": 5274 + }, + { + "epoch": 0.19, + "learning_rate": 1.8086571208847546e-07, + "loss": 0.6985, + "step": 5275 + }, + { + "epoch": 0.19, + "learning_rate": 1.8086205002380342e-07, + "loss": 0.7031, + "step": 5276 + }, + { + "epoch": 0.19, + "learning_rate": 1.8085838795913134e-07, + "loss": 0.6882, + "step": 5277 + }, + { + "epoch": 0.19, + "learning_rate": 1.808547258944593e-07, + "loss": 0.7025, + "step": 5278 + }, + { + "epoch": 0.19, + "learning_rate": 1.8085106382978722e-07, + "loss": 0.6706, + "step": 5279 + }, + { + "epoch": 0.19, + "learning_rate": 1.8084740176511518e-07, + "loss": 0.6816, + "step": 5280 + }, + { + "epoch": 0.19, + "learning_rate": 1.808437397004431e-07, + "loss": 0.7195, + "step": 5281 + }, + { + "epoch": 0.19, + "learning_rate": 1.8084007763577106e-07, + "loss": 0.6794, + "step": 5282 + }, + { + "epoch": 0.19, + "learning_rate": 1.8083641557109898e-07, + "loss": 0.7047, + "step": 5283 + }, + { + "epoch": 0.19, + "learning_rate": 1.808327535064269e-07, + "loss": 0.6861, + "step": 5284 + }, + { + "epoch": 0.19, + "learning_rate": 1.8082909144175484e-07, + "loss": 0.687, + "step": 5285 + }, + { + "epoch": 0.19, + "learning_rate": 1.808254293770828e-07, + "loss": 0.6888, + "step": 5286 + }, + { + "epoch": 0.19, + "learning_rate": 1.8082176731241072e-07, + "loss": 0.6774, + "step": 5287 + }, + { + "epoch": 0.19, + "learning_rate": 1.8081810524773867e-07, + "loss": 0.6935, + "step": 5288 + }, + { + "epoch": 0.19, + "learning_rate": 1.808144431830666e-07, + "loss": 0.6896, + "step": 5289 + }, + { + "epoch": 0.19, + "learning_rate": 1.8081078111839455e-07, + "loss": 0.6913, + "step": 5290 + }, + { + "epoch": 0.19, + "learning_rate": 1.8080711905372248e-07, + "loss": 0.7066, + "step": 5291 + }, + { + "epoch": 0.19, + "learning_rate": 1.8080345698905043e-07, + "loss": 0.6857, + "step": 5292 + }, + { + "epoch": 0.19, + "learning_rate": 1.8079979492437836e-07, + "loss": 0.6759, + "step": 5293 + }, + { + "epoch": 0.19, + "learning_rate": 1.807961328597063e-07, + "loss": 0.6802, + "step": 5294 + }, + { + "epoch": 0.19, + "learning_rate": 1.8079247079503424e-07, + "loss": 0.6958, + "step": 5295 + }, + { + "epoch": 0.19, + "learning_rate": 1.8078880873036216e-07, + "loss": 0.7099, + "step": 5296 + }, + { + "epoch": 0.19, + "learning_rate": 1.8078514666569012e-07, + "loss": 0.6984, + "step": 5297 + }, + { + "epoch": 0.19, + "learning_rate": 1.8078148460101804e-07, + "loss": 0.6806, + "step": 5298 + }, + { + "epoch": 0.19, + "learning_rate": 1.80777822536346e-07, + "loss": 0.6843, + "step": 5299 + }, + { + "epoch": 0.19, + "learning_rate": 1.8077416047167392e-07, + "loss": 0.7028, + "step": 5300 + }, + { + "epoch": 0.19, + "learning_rate": 1.8077049840700188e-07, + "loss": 0.7108, + "step": 5301 + }, + { + "epoch": 0.19, + "learning_rate": 1.8076683634232978e-07, + "loss": 0.6764, + "step": 5302 + }, + { + "epoch": 0.19, + "learning_rate": 1.8076317427765773e-07, + "loss": 0.6814, + "step": 5303 + }, + { + "epoch": 0.19, + "learning_rate": 1.8075951221298566e-07, + "loss": 0.6812, + "step": 5304 + }, + { + "epoch": 0.19, + "learning_rate": 1.807558501483136e-07, + "loss": 0.6876, + "step": 5305 + }, + { + "epoch": 0.19, + "learning_rate": 1.8075218808364154e-07, + "loss": 0.6883, + "step": 5306 + }, + { + "epoch": 0.19, + "learning_rate": 1.807485260189695e-07, + "loss": 0.6913, + "step": 5307 + }, + { + "epoch": 0.19, + "learning_rate": 1.8074486395429742e-07, + "loss": 0.6906, + "step": 5308 + }, + { + "epoch": 0.19, + "learning_rate": 1.8074120188962537e-07, + "loss": 0.7148, + "step": 5309 + }, + { + "epoch": 0.19, + "learning_rate": 1.807375398249533e-07, + "loss": 0.6912, + "step": 5310 + }, + { + "epoch": 0.19, + "learning_rate": 1.8073387776028125e-07, + "loss": 0.6842, + "step": 5311 + }, + { + "epoch": 0.19, + "learning_rate": 1.8073021569560918e-07, + "loss": 0.6877, + "step": 5312 + }, + { + "epoch": 0.19, + "learning_rate": 1.8072655363093713e-07, + "loss": 0.7097, + "step": 5313 + }, + { + "epoch": 0.19, + "learning_rate": 1.8072289156626505e-07, + "loss": 0.7227, + "step": 5314 + }, + { + "epoch": 0.19, + "learning_rate": 1.80719229501593e-07, + "loss": 0.7088, + "step": 5315 + }, + { + "epoch": 0.19, + "learning_rate": 1.8071556743692093e-07, + "loss": 0.7004, + "step": 5316 + }, + { + "epoch": 0.19, + "learning_rate": 1.8071190537224886e-07, + "loss": 0.6745, + "step": 5317 + }, + { + "epoch": 0.19, + "learning_rate": 1.8070824330757681e-07, + "loss": 0.7161, + "step": 5318 + }, + { + "epoch": 0.19, + "learning_rate": 1.8070458124290474e-07, + "loss": 0.689, + "step": 5319 + }, + { + "epoch": 0.19, + "learning_rate": 1.8070091917823267e-07, + "loss": 0.6752, + "step": 5320 + }, + { + "epoch": 0.19, + "learning_rate": 1.8069725711356062e-07, + "loss": 0.6832, + "step": 5321 + }, + { + "epoch": 0.19, + "learning_rate": 1.8069359504888855e-07, + "loss": 0.6782, + "step": 5322 + }, + { + "epoch": 0.19, + "learning_rate": 1.8068993298421647e-07, + "loss": 0.7103, + "step": 5323 + }, + { + "epoch": 0.19, + "learning_rate": 1.8068627091954443e-07, + "loss": 0.6822, + "step": 5324 + }, + { + "epoch": 0.19, + "learning_rate": 1.8068260885487235e-07, + "loss": 0.6905, + "step": 5325 + }, + { + "epoch": 0.19, + "learning_rate": 1.806789467902003e-07, + "loss": 0.6854, + "step": 5326 + }, + { + "epoch": 0.19, + "learning_rate": 1.8067528472552823e-07, + "loss": 0.6934, + "step": 5327 + }, + { + "epoch": 0.19, + "learning_rate": 1.806716226608562e-07, + "loss": 0.7179, + "step": 5328 + }, + { + "epoch": 0.19, + "learning_rate": 1.8066796059618411e-07, + "loss": 0.6798, + "step": 5329 + }, + { + "epoch": 0.2, + "learning_rate": 1.8066429853151207e-07, + "loss": 0.6841, + "step": 5330 + }, + { + "epoch": 0.2, + "learning_rate": 1.8066063646684e-07, + "loss": 0.6971, + "step": 5331 + }, + { + "epoch": 0.2, + "learning_rate": 1.8065697440216795e-07, + "loss": 0.6981, + "step": 5332 + }, + { + "epoch": 0.2, + "learning_rate": 1.8065331233749587e-07, + "loss": 0.6897, + "step": 5333 + }, + { + "epoch": 0.2, + "learning_rate": 1.8064965027282383e-07, + "loss": 0.7025, + "step": 5334 + }, + { + "epoch": 0.2, + "learning_rate": 1.8064598820815175e-07, + "loss": 0.6932, + "step": 5335 + }, + { + "epoch": 0.2, + "learning_rate": 1.806423261434797e-07, + "loss": 0.6993, + "step": 5336 + }, + { + "epoch": 0.2, + "learning_rate": 1.8063866407880763e-07, + "loss": 0.6939, + "step": 5337 + }, + { + "epoch": 0.2, + "learning_rate": 1.8063500201413556e-07, + "loss": 0.6942, + "step": 5338 + }, + { + "epoch": 0.2, + "learning_rate": 1.806313399494635e-07, + "loss": 0.7069, + "step": 5339 + }, + { + "epoch": 0.2, + "learning_rate": 1.8062767788479144e-07, + "loss": 0.6653, + "step": 5340 + }, + { + "epoch": 0.2, + "learning_rate": 1.8062401582011937e-07, + "loss": 0.688, + "step": 5341 + }, + { + "epoch": 0.2, + "learning_rate": 1.8062035375544732e-07, + "loss": 0.6837, + "step": 5342 + }, + { + "epoch": 0.2, + "learning_rate": 1.8061669169077525e-07, + "loss": 0.6903, + "step": 5343 + }, + { + "epoch": 0.2, + "learning_rate": 1.8061302962610317e-07, + "loss": 0.7021, + "step": 5344 + }, + { + "epoch": 0.2, + "learning_rate": 1.8060936756143113e-07, + "loss": 0.7071, + "step": 5345 + }, + { + "epoch": 0.2, + "learning_rate": 1.8060570549675905e-07, + "loss": 0.6967, + "step": 5346 + }, + { + "epoch": 0.2, + "learning_rate": 1.80602043432087e-07, + "loss": 0.7018, + "step": 5347 + }, + { + "epoch": 0.2, + "learning_rate": 1.8059838136741493e-07, + "loss": 0.6918, + "step": 5348 + }, + { + "epoch": 0.2, + "learning_rate": 1.8059471930274289e-07, + "loss": 0.689, + "step": 5349 + }, + { + "epoch": 0.2, + "learning_rate": 1.805910572380708e-07, + "loss": 0.6754, + "step": 5350 + }, + { + "epoch": 0.2, + "learning_rate": 1.8058739517339877e-07, + "loss": 0.6837, + "step": 5351 + }, + { + "epoch": 0.2, + "learning_rate": 1.805837331087267e-07, + "loss": 0.6984, + "step": 5352 + }, + { + "epoch": 0.2, + "learning_rate": 1.8058007104405465e-07, + "loss": 0.6895, + "step": 5353 + }, + { + "epoch": 0.2, + "learning_rate": 1.8057640897938257e-07, + "loss": 0.6601, + "step": 5354 + }, + { + "epoch": 0.2, + "learning_rate": 1.8057274691471053e-07, + "loss": 0.6843, + "step": 5355 + }, + { + "epoch": 0.2, + "learning_rate": 1.8056908485003843e-07, + "loss": 0.6998, + "step": 5356 + }, + { + "epoch": 0.2, + "learning_rate": 1.8056542278536638e-07, + "loss": 0.6872, + "step": 5357 + }, + { + "epoch": 0.2, + "learning_rate": 1.805617607206943e-07, + "loss": 0.6883, + "step": 5358 + }, + { + "epoch": 0.2, + "learning_rate": 1.8055809865602226e-07, + "loss": 0.6958, + "step": 5359 + }, + { + "epoch": 0.2, + "learning_rate": 1.8055443659135019e-07, + "loss": 0.7037, + "step": 5360 + }, + { + "epoch": 0.2, + "learning_rate": 1.8055077452667814e-07, + "loss": 0.7095, + "step": 5361 + }, + { + "epoch": 0.2, + "learning_rate": 1.8054711246200607e-07, + "loss": 0.694, + "step": 5362 + }, + { + "epoch": 0.2, + "learning_rate": 1.8054345039733402e-07, + "loss": 0.7075, + "step": 5363 + }, + { + "epoch": 0.2, + "learning_rate": 1.8053978833266195e-07, + "loss": 0.6894, + "step": 5364 + }, + { + "epoch": 0.2, + "learning_rate": 1.805361262679899e-07, + "loss": 0.6929, + "step": 5365 + }, + { + "epoch": 0.2, + "learning_rate": 1.8053246420331783e-07, + "loss": 0.7015, + "step": 5366 + }, + { + "epoch": 0.2, + "learning_rate": 1.8052880213864575e-07, + "loss": 0.7038, + "step": 5367 + }, + { + "epoch": 0.2, + "learning_rate": 1.805251400739737e-07, + "loss": 0.6779, + "step": 5368 + }, + { + "epoch": 0.2, + "learning_rate": 1.8052147800930163e-07, + "loss": 0.6764, + "step": 5369 + }, + { + "epoch": 0.2, + "learning_rate": 1.8051781594462959e-07, + "loss": 0.7036, + "step": 5370 + }, + { + "epoch": 0.2, + "learning_rate": 1.805141538799575e-07, + "loss": 0.705, + "step": 5371 + }, + { + "epoch": 0.2, + "learning_rate": 1.8051049181528546e-07, + "loss": 0.6754, + "step": 5372 + }, + { + "epoch": 0.2, + "learning_rate": 1.805068297506134e-07, + "loss": 0.6912, + "step": 5373 + }, + { + "epoch": 0.2, + "learning_rate": 1.8050316768594132e-07, + "loss": 0.7213, + "step": 5374 + }, + { + "epoch": 0.2, + "learning_rate": 1.8049950562126924e-07, + "loss": 0.6968, + "step": 5375 + }, + { + "epoch": 0.2, + "learning_rate": 1.804958435565972e-07, + "loss": 0.6839, + "step": 5376 + }, + { + "epoch": 0.2, + "learning_rate": 1.8049218149192512e-07, + "loss": 0.6586, + "step": 5377 + }, + { + "epoch": 0.2, + "learning_rate": 1.8048851942725308e-07, + "loss": 0.6883, + "step": 5378 + }, + { + "epoch": 0.2, + "learning_rate": 1.80484857362581e-07, + "loss": 0.6956, + "step": 5379 + }, + { + "epoch": 0.2, + "learning_rate": 1.8048119529790896e-07, + "loss": 0.6944, + "step": 5380 + }, + { + "epoch": 0.2, + "learning_rate": 1.8047753323323688e-07, + "loss": 0.6985, + "step": 5381 + }, + { + "epoch": 0.2, + "learning_rate": 1.8047387116856484e-07, + "loss": 0.698, + "step": 5382 + }, + { + "epoch": 0.2, + "learning_rate": 1.8047020910389276e-07, + "loss": 0.6976, + "step": 5383 + }, + { + "epoch": 0.2, + "learning_rate": 1.8046654703922072e-07, + "loss": 0.6925, + "step": 5384 + }, + { + "epoch": 0.2, + "learning_rate": 1.8046288497454864e-07, + "loss": 0.7035, + "step": 5385 + }, + { + "epoch": 0.2, + "learning_rate": 1.804592229098766e-07, + "loss": 0.7041, + "step": 5386 + }, + { + "epoch": 0.2, + "learning_rate": 1.8045556084520452e-07, + "loss": 0.6825, + "step": 5387 + }, + { + "epoch": 0.2, + "learning_rate": 1.8045189878053245e-07, + "loss": 0.6889, + "step": 5388 + }, + { + "epoch": 0.2, + "learning_rate": 1.804482367158604e-07, + "loss": 0.7112, + "step": 5389 + }, + { + "epoch": 0.2, + "learning_rate": 1.8044457465118833e-07, + "loss": 0.6951, + "step": 5390 + }, + { + "epoch": 0.2, + "learning_rate": 1.8044091258651628e-07, + "loss": 0.7019, + "step": 5391 + }, + { + "epoch": 0.2, + "learning_rate": 1.804372505218442e-07, + "loss": 0.6909, + "step": 5392 + }, + { + "epoch": 0.2, + "learning_rate": 1.8043358845717214e-07, + "loss": 0.6925, + "step": 5393 + }, + { + "epoch": 0.2, + "learning_rate": 1.8042992639250006e-07, + "loss": 0.695, + "step": 5394 + }, + { + "epoch": 0.2, + "learning_rate": 1.8042626432782802e-07, + "loss": 0.6995, + "step": 5395 + }, + { + "epoch": 0.2, + "learning_rate": 1.8042260226315594e-07, + "loss": 0.6854, + "step": 5396 + }, + { + "epoch": 0.2, + "learning_rate": 1.804189401984839e-07, + "loss": 0.6969, + "step": 5397 + }, + { + "epoch": 0.2, + "learning_rate": 1.8041527813381182e-07, + "loss": 0.6945, + "step": 5398 + }, + { + "epoch": 0.2, + "learning_rate": 1.8041161606913978e-07, + "loss": 0.6906, + "step": 5399 + }, + { + "epoch": 0.2, + "learning_rate": 1.804079540044677e-07, + "loss": 0.7066, + "step": 5400 + }, + { + "epoch": 0.2, + "learning_rate": 1.8040429193979566e-07, + "loss": 0.6909, + "step": 5401 + }, + { + "epoch": 0.2, + "learning_rate": 1.8040062987512358e-07, + "loss": 0.6779, + "step": 5402 + }, + { + "epoch": 0.2, + "learning_rate": 1.8039696781045154e-07, + "loss": 0.6945, + "step": 5403 + }, + { + "epoch": 0.2, + "learning_rate": 1.8039330574577946e-07, + "loss": 0.6976, + "step": 5404 + }, + { + "epoch": 0.2, + "learning_rate": 1.8038964368110742e-07, + "loss": 0.7082, + "step": 5405 + }, + { + "epoch": 0.2, + "learning_rate": 1.8038598161643534e-07, + "loss": 0.6891, + "step": 5406 + }, + { + "epoch": 0.2, + "learning_rate": 1.803823195517633e-07, + "loss": 0.6981, + "step": 5407 + }, + { + "epoch": 0.2, + "learning_rate": 1.8037865748709122e-07, + "loss": 0.685, + "step": 5408 + }, + { + "epoch": 0.2, + "learning_rate": 1.8037499542241915e-07, + "loss": 0.6871, + "step": 5409 + }, + { + "epoch": 0.2, + "learning_rate": 1.8037133335774708e-07, + "loss": 0.6971, + "step": 5410 + }, + { + "epoch": 0.2, + "learning_rate": 1.8036767129307503e-07, + "loss": 0.6981, + "step": 5411 + }, + { + "epoch": 0.2, + "learning_rate": 1.8036400922840296e-07, + "loss": 0.7047, + "step": 5412 + }, + { + "epoch": 0.2, + "learning_rate": 1.803603471637309e-07, + "loss": 0.6963, + "step": 5413 + }, + { + "epoch": 0.2, + "learning_rate": 1.8035668509905884e-07, + "loss": 0.6789, + "step": 5414 + }, + { + "epoch": 0.2, + "learning_rate": 1.8035302303438676e-07, + "loss": 0.6833, + "step": 5415 + }, + { + "epoch": 0.2, + "learning_rate": 1.8034936096971472e-07, + "loss": 0.6855, + "step": 5416 + }, + { + "epoch": 0.2, + "learning_rate": 1.8034569890504264e-07, + "loss": 0.7085, + "step": 5417 + }, + { + "epoch": 0.2, + "learning_rate": 1.803420368403706e-07, + "loss": 0.7019, + "step": 5418 + }, + { + "epoch": 0.2, + "learning_rate": 1.8033837477569852e-07, + "loss": 0.6951, + "step": 5419 + }, + { + "epoch": 0.2, + "learning_rate": 1.8033471271102648e-07, + "loss": 0.7085, + "step": 5420 + }, + { + "epoch": 0.2, + "learning_rate": 1.803310506463544e-07, + "loss": 0.6795, + "step": 5421 + }, + { + "epoch": 0.2, + "learning_rate": 1.8032738858168236e-07, + "loss": 0.681, + "step": 5422 + }, + { + "epoch": 0.2, + "learning_rate": 1.8032372651701028e-07, + "loss": 0.6923, + "step": 5423 + }, + { + "epoch": 0.2, + "learning_rate": 1.8032006445233824e-07, + "loss": 0.6854, + "step": 5424 + }, + { + "epoch": 0.2, + "learning_rate": 1.8031640238766616e-07, + "loss": 0.7122, + "step": 5425 + }, + { + "epoch": 0.2, + "learning_rate": 1.8031274032299412e-07, + "loss": 0.6924, + "step": 5426 + }, + { + "epoch": 0.2, + "learning_rate": 1.8030907825832204e-07, + "loss": 0.7021, + "step": 5427 + }, + { + "epoch": 0.2, + "learning_rate": 1.8030541619364997e-07, + "loss": 0.7056, + "step": 5428 + }, + { + "epoch": 0.2, + "learning_rate": 1.803017541289779e-07, + "loss": 0.7068, + "step": 5429 + }, + { + "epoch": 0.2, + "learning_rate": 1.8029809206430585e-07, + "loss": 0.6998, + "step": 5430 + }, + { + "epoch": 0.2, + "learning_rate": 1.8029442999963378e-07, + "loss": 0.6939, + "step": 5431 + }, + { + "epoch": 0.2, + "learning_rate": 1.8029076793496173e-07, + "loss": 0.6716, + "step": 5432 + }, + { + "epoch": 0.2, + "learning_rate": 1.8028710587028965e-07, + "loss": 0.699, + "step": 5433 + }, + { + "epoch": 0.2, + "learning_rate": 1.802834438056176e-07, + "loss": 0.7136, + "step": 5434 + }, + { + "epoch": 0.2, + "learning_rate": 1.8027978174094553e-07, + "loss": 0.7136, + "step": 5435 + }, + { + "epoch": 0.2, + "learning_rate": 1.802761196762735e-07, + "loss": 0.7111, + "step": 5436 + }, + { + "epoch": 0.2, + "learning_rate": 1.8027245761160141e-07, + "loss": 0.6931, + "step": 5437 + }, + { + "epoch": 0.2, + "learning_rate": 1.8026879554692934e-07, + "loss": 0.7018, + "step": 5438 + }, + { + "epoch": 0.2, + "learning_rate": 1.802651334822573e-07, + "loss": 0.7004, + "step": 5439 + }, + { + "epoch": 0.2, + "learning_rate": 1.8026147141758522e-07, + "loss": 0.695, + "step": 5440 + }, + { + "epoch": 0.2, + "learning_rate": 1.8025780935291317e-07, + "loss": 0.6899, + "step": 5441 + }, + { + "epoch": 0.2, + "learning_rate": 1.802541472882411e-07, + "loss": 0.7059, + "step": 5442 + }, + { + "epoch": 0.2, + "learning_rate": 1.8025048522356905e-07, + "loss": 0.6966, + "step": 5443 + }, + { + "epoch": 0.2, + "learning_rate": 1.8024682315889698e-07, + "loss": 0.6898, + "step": 5444 + }, + { + "epoch": 0.2, + "learning_rate": 1.802431610942249e-07, + "loss": 0.6977, + "step": 5445 + }, + { + "epoch": 0.2, + "learning_rate": 1.8023949902955283e-07, + "loss": 0.7084, + "step": 5446 + }, + { + "epoch": 0.2, + "learning_rate": 1.802358369648808e-07, + "loss": 0.7058, + "step": 5447 + }, + { + "epoch": 0.2, + "learning_rate": 1.8023217490020871e-07, + "loss": 0.6942, + "step": 5448 + }, + { + "epoch": 0.2, + "learning_rate": 1.8022851283553667e-07, + "loss": 0.6927, + "step": 5449 + }, + { + "epoch": 0.2, + "learning_rate": 1.802248507708646e-07, + "loss": 0.6959, + "step": 5450 + }, + { + "epoch": 0.2, + "learning_rate": 1.8022118870619255e-07, + "loss": 0.696, + "step": 5451 + }, + { + "epoch": 0.2, + "learning_rate": 1.8021752664152047e-07, + "loss": 0.6942, + "step": 5452 + }, + { + "epoch": 0.2, + "learning_rate": 1.8021386457684843e-07, + "loss": 0.6866, + "step": 5453 + }, + { + "epoch": 0.2, + "learning_rate": 1.8021020251217635e-07, + "loss": 0.7098, + "step": 5454 + }, + { + "epoch": 0.2, + "learning_rate": 1.802065404475043e-07, + "loss": 0.6947, + "step": 5455 + }, + { + "epoch": 0.2, + "learning_rate": 1.8020287838283223e-07, + "loss": 0.7245, + "step": 5456 + }, + { + "epoch": 0.2, + "learning_rate": 1.801992163181602e-07, + "loss": 0.6976, + "step": 5457 + }, + { + "epoch": 0.2, + "learning_rate": 1.8019555425348811e-07, + "loss": 0.6743, + "step": 5458 + }, + { + "epoch": 0.2, + "learning_rate": 1.8019189218881604e-07, + "loss": 0.6927, + "step": 5459 + }, + { + "epoch": 0.2, + "learning_rate": 1.80188230124144e-07, + "loss": 0.7083, + "step": 5460 + }, + { + "epoch": 0.2, + "learning_rate": 1.8018456805947192e-07, + "loss": 0.6851, + "step": 5461 + }, + { + "epoch": 0.2, + "learning_rate": 1.8018090599479987e-07, + "loss": 0.6942, + "step": 5462 + }, + { + "epoch": 0.2, + "learning_rate": 1.801772439301278e-07, + "loss": 0.7077, + "step": 5463 + }, + { + "epoch": 0.2, + "learning_rate": 1.8017358186545573e-07, + "loss": 0.6818, + "step": 5464 + }, + { + "epoch": 0.2, + "learning_rate": 1.8016991980078365e-07, + "loss": 0.6789, + "step": 5465 + }, + { + "epoch": 0.2, + "learning_rate": 1.801662577361116e-07, + "loss": 0.6911, + "step": 5466 + }, + { + "epoch": 0.2, + "learning_rate": 1.8016259567143953e-07, + "loss": 0.6959, + "step": 5467 + }, + { + "epoch": 0.2, + "learning_rate": 1.8015893360676749e-07, + "loss": 0.7, + "step": 5468 + }, + { + "epoch": 0.2, + "learning_rate": 1.801552715420954e-07, + "loss": 0.7017, + "step": 5469 + }, + { + "epoch": 0.2, + "learning_rate": 1.8015160947742337e-07, + "loss": 0.6864, + "step": 5470 + }, + { + "epoch": 0.2, + "learning_rate": 1.801479474127513e-07, + "loss": 0.7096, + "step": 5471 + }, + { + "epoch": 0.2, + "learning_rate": 1.8014428534807925e-07, + "loss": 0.6987, + "step": 5472 + }, + { + "epoch": 0.2, + "learning_rate": 1.8014062328340717e-07, + "loss": 0.6655, + "step": 5473 + }, + { + "epoch": 0.2, + "learning_rate": 1.8013696121873513e-07, + "loss": 0.6887, + "step": 5474 + }, + { + "epoch": 0.2, + "learning_rate": 1.8013329915406305e-07, + "loss": 0.7017, + "step": 5475 + }, + { + "epoch": 0.2, + "learning_rate": 1.80129637089391e-07, + "loss": 0.7061, + "step": 5476 + }, + { + "epoch": 0.2, + "learning_rate": 1.8012597502471893e-07, + "loss": 0.6905, + "step": 5477 + }, + { + "epoch": 0.2, + "learning_rate": 1.8012231296004689e-07, + "loss": 0.6996, + "step": 5478 + }, + { + "epoch": 0.2, + "learning_rate": 1.801186508953748e-07, + "loss": 0.6987, + "step": 5479 + }, + { + "epoch": 0.2, + "learning_rate": 1.8011498883070274e-07, + "loss": 0.6997, + "step": 5480 + }, + { + "epoch": 0.2, + "learning_rate": 1.8011132676603067e-07, + "loss": 0.6891, + "step": 5481 + }, + { + "epoch": 0.2, + "learning_rate": 1.8010766470135862e-07, + "loss": 0.702, + "step": 5482 + }, + { + "epoch": 0.2, + "learning_rate": 1.8010400263668655e-07, + "loss": 0.7173, + "step": 5483 + }, + { + "epoch": 0.2, + "learning_rate": 1.801003405720145e-07, + "loss": 0.6911, + "step": 5484 + }, + { + "epoch": 0.2, + "learning_rate": 1.8009667850734243e-07, + "loss": 0.7057, + "step": 5485 + }, + { + "epoch": 0.2, + "learning_rate": 1.8009301644267035e-07, + "loss": 0.697, + "step": 5486 + }, + { + "epoch": 0.2, + "learning_rate": 1.800893543779983e-07, + "loss": 0.6923, + "step": 5487 + }, + { + "epoch": 0.2, + "learning_rate": 1.8008569231332623e-07, + "loss": 0.7111, + "step": 5488 + }, + { + "epoch": 0.2, + "learning_rate": 1.8008203024865419e-07, + "loss": 0.6995, + "step": 5489 + }, + { + "epoch": 0.2, + "learning_rate": 1.800783681839821e-07, + "loss": 0.7139, + "step": 5490 + }, + { + "epoch": 0.2, + "learning_rate": 1.8007470611931006e-07, + "loss": 0.6956, + "step": 5491 + }, + { + "epoch": 0.2, + "learning_rate": 1.80071044054638e-07, + "loss": 0.7022, + "step": 5492 + }, + { + "epoch": 0.2, + "learning_rate": 1.8006738198996594e-07, + "loss": 0.6954, + "step": 5493 + }, + { + "epoch": 0.2, + "learning_rate": 1.8006371992529387e-07, + "loss": 0.6837, + "step": 5494 + }, + { + "epoch": 0.2, + "learning_rate": 1.8006005786062182e-07, + "loss": 0.6999, + "step": 5495 + }, + { + "epoch": 0.2, + "learning_rate": 1.8005639579594975e-07, + "loss": 0.6935, + "step": 5496 + }, + { + "epoch": 0.2, + "learning_rate": 1.800527337312777e-07, + "loss": 0.6998, + "step": 5497 + }, + { + "epoch": 0.2, + "learning_rate": 1.8004907166660563e-07, + "loss": 0.6739, + "step": 5498 + }, + { + "epoch": 0.2, + "learning_rate": 1.8004540960193356e-07, + "loss": 0.6788, + "step": 5499 + }, + { + "epoch": 0.2, + "learning_rate": 1.8004174753726148e-07, + "loss": 0.6843, + "step": 5500 + }, + { + "epoch": 0.2, + "learning_rate": 1.8003808547258944e-07, + "loss": 0.6978, + "step": 5501 + }, + { + "epoch": 0.2, + "learning_rate": 1.8003442340791736e-07, + "loss": 0.7063, + "step": 5502 + }, + { + "epoch": 0.2, + "learning_rate": 1.8003076134324532e-07, + "loss": 0.6984, + "step": 5503 + }, + { + "epoch": 0.2, + "learning_rate": 1.8002709927857324e-07, + "loss": 0.6896, + "step": 5504 + }, + { + "epoch": 0.2, + "learning_rate": 1.800234372139012e-07, + "loss": 0.7059, + "step": 5505 + }, + { + "epoch": 0.2, + "learning_rate": 1.8001977514922912e-07, + "loss": 0.6844, + "step": 5506 + }, + { + "epoch": 0.2, + "learning_rate": 1.8001611308455708e-07, + "loss": 0.6986, + "step": 5507 + }, + { + "epoch": 0.2, + "learning_rate": 1.80012451019885e-07, + "loss": 0.6776, + "step": 5508 + }, + { + "epoch": 0.2, + "learning_rate": 1.8000878895521293e-07, + "loss": 0.6852, + "step": 5509 + }, + { + "epoch": 0.2, + "learning_rate": 1.8000512689054088e-07, + "loss": 0.6959, + "step": 5510 + }, + { + "epoch": 0.2, + "learning_rate": 1.800014648258688e-07, + "loss": 0.7104, + "step": 5511 + }, + { + "epoch": 0.2, + "learning_rate": 1.7999780276119676e-07, + "loss": 0.694, + "step": 5512 + }, + { + "epoch": 0.2, + "learning_rate": 1.799941406965247e-07, + "loss": 0.7012, + "step": 5513 + }, + { + "epoch": 0.2, + "learning_rate": 1.7999047863185264e-07, + "loss": 0.7033, + "step": 5514 + }, + { + "epoch": 0.2, + "learning_rate": 1.7998681656718057e-07, + "loss": 0.6887, + "step": 5515 + }, + { + "epoch": 0.2, + "learning_rate": 1.7998315450250852e-07, + "loss": 0.7183, + "step": 5516 + }, + { + "epoch": 0.2, + "learning_rate": 1.7997949243783642e-07, + "loss": 0.7069, + "step": 5517 + }, + { + "epoch": 0.2, + "learning_rate": 1.7997583037316438e-07, + "loss": 0.6918, + "step": 5518 + }, + { + "epoch": 0.2, + "learning_rate": 1.799721683084923e-07, + "loss": 0.6856, + "step": 5519 + }, + { + "epoch": 0.2, + "learning_rate": 1.7996850624382026e-07, + "loss": 0.6944, + "step": 5520 + }, + { + "epoch": 0.2, + "learning_rate": 1.7996484417914818e-07, + "loss": 0.6914, + "step": 5521 + }, + { + "epoch": 0.2, + "learning_rate": 1.7996118211447614e-07, + "loss": 0.6856, + "step": 5522 + }, + { + "epoch": 0.2, + "learning_rate": 1.7995752004980406e-07, + "loss": 0.7028, + "step": 5523 + }, + { + "epoch": 0.2, + "learning_rate": 1.7995385798513202e-07, + "loss": 0.6989, + "step": 5524 + }, + { + "epoch": 0.2, + "learning_rate": 1.7995019592045994e-07, + "loss": 0.6994, + "step": 5525 + }, + { + "epoch": 0.2, + "learning_rate": 1.799465338557879e-07, + "loss": 0.6775, + "step": 5526 + }, + { + "epoch": 0.2, + "learning_rate": 1.7994287179111582e-07, + "loss": 0.7032, + "step": 5527 + }, + { + "epoch": 0.2, + "learning_rate": 1.7993920972644378e-07, + "loss": 0.6897, + "step": 5528 + }, + { + "epoch": 0.2, + "learning_rate": 1.799355476617717e-07, + "loss": 0.6929, + "step": 5529 + }, + { + "epoch": 0.2, + "learning_rate": 1.7993188559709963e-07, + "loss": 0.6886, + "step": 5530 + }, + { + "epoch": 0.2, + "learning_rate": 1.7992822353242758e-07, + "loss": 0.6791, + "step": 5531 + }, + { + "epoch": 0.2, + "learning_rate": 1.799245614677555e-07, + "loss": 0.6922, + "step": 5532 + }, + { + "epoch": 0.2, + "learning_rate": 1.7992089940308346e-07, + "loss": 0.6887, + "step": 5533 + }, + { + "epoch": 0.2, + "learning_rate": 1.799172373384114e-07, + "loss": 0.6979, + "step": 5534 + }, + { + "epoch": 0.2, + "learning_rate": 1.7991357527373932e-07, + "loss": 0.677, + "step": 5535 + }, + { + "epoch": 0.2, + "learning_rate": 1.7990991320906724e-07, + "loss": 0.6901, + "step": 5536 + }, + { + "epoch": 0.2, + "learning_rate": 1.799062511443952e-07, + "loss": 0.6816, + "step": 5537 + }, + { + "epoch": 0.2, + "learning_rate": 1.7990258907972312e-07, + "loss": 0.7061, + "step": 5538 + }, + { + "epoch": 0.2, + "learning_rate": 1.7989892701505108e-07, + "loss": 0.6966, + "step": 5539 + }, + { + "epoch": 0.2, + "learning_rate": 1.79895264950379e-07, + "loss": 0.7061, + "step": 5540 + }, + { + "epoch": 0.2, + "learning_rate": 1.7989160288570696e-07, + "loss": 0.7057, + "step": 5541 + }, + { + "epoch": 0.2, + "learning_rate": 1.7988794082103488e-07, + "loss": 0.7003, + "step": 5542 + }, + { + "epoch": 0.2, + "learning_rate": 1.7988427875636284e-07, + "loss": 0.7117, + "step": 5543 + }, + { + "epoch": 0.2, + "learning_rate": 1.7988061669169076e-07, + "loss": 0.6906, + "step": 5544 + }, + { + "epoch": 0.2, + "learning_rate": 1.7987695462701872e-07, + "loss": 0.6879, + "step": 5545 + }, + { + "epoch": 0.2, + "learning_rate": 1.7987329256234664e-07, + "loss": 0.6879, + "step": 5546 + }, + { + "epoch": 0.2, + "learning_rate": 1.798696304976746e-07, + "loss": 0.6911, + "step": 5547 + }, + { + "epoch": 0.2, + "learning_rate": 1.7986596843300252e-07, + "loss": 0.695, + "step": 5548 + }, + { + "epoch": 0.2, + "learning_rate": 1.7986230636833047e-07, + "loss": 0.6878, + "step": 5549 + }, + { + "epoch": 0.2, + "learning_rate": 1.798586443036584e-07, + "loss": 0.6857, + "step": 5550 + }, + { + "epoch": 0.2, + "learning_rate": 1.7985498223898633e-07, + "loss": 0.6985, + "step": 5551 + }, + { + "epoch": 0.2, + "learning_rate": 1.7985132017431428e-07, + "loss": 0.7065, + "step": 5552 + }, + { + "epoch": 0.2, + "learning_rate": 1.798476581096422e-07, + "loss": 0.7187, + "step": 5553 + }, + { + "epoch": 0.2, + "learning_rate": 1.7984399604497013e-07, + "loss": 0.6921, + "step": 5554 + }, + { + "epoch": 0.2, + "learning_rate": 1.798403339802981e-07, + "loss": 0.6899, + "step": 5555 + }, + { + "epoch": 0.2, + "learning_rate": 1.7983667191562601e-07, + "loss": 0.6996, + "step": 5556 + }, + { + "epoch": 0.2, + "learning_rate": 1.7983300985095394e-07, + "loss": 0.7053, + "step": 5557 + }, + { + "epoch": 0.2, + "learning_rate": 1.798293477862819e-07, + "loss": 0.6903, + "step": 5558 + }, + { + "epoch": 0.2, + "learning_rate": 1.7982568572160982e-07, + "loss": 0.7041, + "step": 5559 + }, + { + "epoch": 0.2, + "learning_rate": 1.7982202365693777e-07, + "loss": 0.6931, + "step": 5560 + }, + { + "epoch": 0.2, + "learning_rate": 1.798183615922657e-07, + "loss": 0.6898, + "step": 5561 + }, + { + "epoch": 0.2, + "learning_rate": 1.7981469952759365e-07, + "loss": 0.6872, + "step": 5562 + }, + { + "epoch": 0.2, + "learning_rate": 1.7981103746292158e-07, + "loss": 0.6928, + "step": 5563 + }, + { + "epoch": 0.2, + "learning_rate": 1.7980737539824953e-07, + "loss": 0.6796, + "step": 5564 + }, + { + "epoch": 0.2, + "learning_rate": 1.7980371333357746e-07, + "loss": 0.7023, + "step": 5565 + }, + { + "epoch": 0.2, + "learning_rate": 1.7980005126890541e-07, + "loss": 0.7027, + "step": 5566 + }, + { + "epoch": 0.2, + "learning_rate": 1.7979638920423334e-07, + "loss": 0.6934, + "step": 5567 + }, + { + "epoch": 0.2, + "learning_rate": 1.797927271395613e-07, + "loss": 0.7031, + "step": 5568 + }, + { + "epoch": 0.2, + "learning_rate": 1.7978906507488922e-07, + "loss": 0.6937, + "step": 5569 + }, + { + "epoch": 0.2, + "learning_rate": 1.7978540301021717e-07, + "loss": 0.6963, + "step": 5570 + }, + { + "epoch": 0.2, + "learning_rate": 1.7978174094554507e-07, + "loss": 0.6872, + "step": 5571 + }, + { + "epoch": 0.2, + "learning_rate": 1.7977807888087303e-07, + "loss": 0.7, + "step": 5572 + }, + { + "epoch": 0.2, + "learning_rate": 1.7977441681620095e-07, + "loss": 0.6884, + "step": 5573 + }, + { + "epoch": 0.2, + "learning_rate": 1.797707547515289e-07, + "loss": 0.6897, + "step": 5574 + }, + { + "epoch": 0.2, + "learning_rate": 1.7976709268685683e-07, + "loss": 0.68, + "step": 5575 + }, + { + "epoch": 0.2, + "learning_rate": 1.797634306221848e-07, + "loss": 0.6865, + "step": 5576 + }, + { + "epoch": 0.2, + "learning_rate": 1.7975976855751271e-07, + "loss": 0.6673, + "step": 5577 + }, + { + "epoch": 0.2, + "learning_rate": 1.7975610649284064e-07, + "loss": 0.6963, + "step": 5578 + }, + { + "epoch": 0.2, + "learning_rate": 1.797524444281686e-07, + "loss": 0.7049, + "step": 5579 + }, + { + "epoch": 0.2, + "learning_rate": 1.7974878236349652e-07, + "loss": 0.6957, + "step": 5580 + }, + { + "epoch": 0.2, + "learning_rate": 1.7974512029882447e-07, + "loss": 0.7082, + "step": 5581 + }, + { + "epoch": 0.2, + "learning_rate": 1.797414582341524e-07, + "loss": 0.6899, + "step": 5582 + }, + { + "epoch": 0.2, + "learning_rate": 1.7973779616948035e-07, + "loss": 0.7108, + "step": 5583 + }, + { + "epoch": 0.2, + "learning_rate": 1.7973413410480828e-07, + "loss": 0.709, + "step": 5584 + }, + { + "epoch": 0.2, + "learning_rate": 1.7973047204013623e-07, + "loss": 0.6716, + "step": 5585 + }, + { + "epoch": 0.2, + "learning_rate": 1.7972680997546416e-07, + "loss": 0.6996, + "step": 5586 + }, + { + "epoch": 0.2, + "learning_rate": 1.797231479107921e-07, + "loss": 0.6892, + "step": 5587 + }, + { + "epoch": 0.2, + "learning_rate": 1.7971948584612004e-07, + "loss": 0.7232, + "step": 5588 + }, + { + "epoch": 0.2, + "learning_rate": 1.7971582378144797e-07, + "loss": 0.7137, + "step": 5589 + }, + { + "epoch": 0.2, + "learning_rate": 1.797121617167759e-07, + "loss": 0.6889, + "step": 5590 + }, + { + "epoch": 0.2, + "learning_rate": 1.7970849965210385e-07, + "loss": 0.6859, + "step": 5591 + }, + { + "epoch": 0.2, + "learning_rate": 1.7970483758743177e-07, + "loss": 0.6923, + "step": 5592 + }, + { + "epoch": 0.2, + "learning_rate": 1.7970117552275973e-07, + "loss": 0.6851, + "step": 5593 + }, + { + "epoch": 0.2, + "learning_rate": 1.7969751345808765e-07, + "loss": 0.7009, + "step": 5594 + }, + { + "epoch": 0.2, + "learning_rate": 1.796938513934156e-07, + "loss": 0.6896, + "step": 5595 + }, + { + "epoch": 0.2, + "learning_rate": 1.7969018932874353e-07, + "loss": 0.7015, + "step": 5596 + }, + { + "epoch": 0.2, + "learning_rate": 1.7968652726407149e-07, + "loss": 0.7225, + "step": 5597 + }, + { + "epoch": 0.2, + "learning_rate": 1.796828651993994e-07, + "loss": 0.7039, + "step": 5598 + }, + { + "epoch": 0.2, + "learning_rate": 1.7967920313472737e-07, + "loss": 0.6911, + "step": 5599 + }, + { + "epoch": 0.2, + "learning_rate": 1.796755410700553e-07, + "loss": 0.6809, + "step": 5600 + }, + { + "epoch": 0.2, + "learning_rate": 1.7967187900538322e-07, + "loss": 0.7053, + "step": 5601 + }, + { + "epoch": 0.2, + "learning_rate": 1.7966821694071117e-07, + "loss": 0.7045, + "step": 5602 + }, + { + "epoch": 0.2, + "learning_rate": 1.796645548760391e-07, + "loss": 0.6913, + "step": 5603 + }, + { + "epoch": 0.21, + "learning_rate": 1.7966089281136705e-07, + "loss": 0.7088, + "step": 5604 + }, + { + "epoch": 0.21, + "learning_rate": 1.7965723074669498e-07, + "loss": 0.695, + "step": 5605 + }, + { + "epoch": 0.21, + "learning_rate": 1.7965356868202293e-07, + "loss": 0.7071, + "step": 5606 + }, + { + "epoch": 0.21, + "learning_rate": 1.7964990661735083e-07, + "loss": 0.685, + "step": 5607 + }, + { + "epoch": 0.21, + "learning_rate": 1.7964624455267879e-07, + "loss": 0.7166, + "step": 5608 + }, + { + "epoch": 0.21, + "learning_rate": 1.796425824880067e-07, + "loss": 0.6891, + "step": 5609 + }, + { + "epoch": 0.21, + "learning_rate": 1.7963892042333466e-07, + "loss": 0.695, + "step": 5610 + }, + { + "epoch": 0.21, + "learning_rate": 1.796352583586626e-07, + "loss": 0.6924, + "step": 5611 + }, + { + "epoch": 0.21, + "learning_rate": 1.7963159629399054e-07, + "loss": 0.6827, + "step": 5612 + }, + { + "epoch": 0.21, + "learning_rate": 1.7962793422931847e-07, + "loss": 0.7048, + "step": 5613 + }, + { + "epoch": 0.21, + "learning_rate": 1.7962427216464642e-07, + "loss": 0.6934, + "step": 5614 + }, + { + "epoch": 0.21, + "learning_rate": 1.7962061009997435e-07, + "loss": 0.6969, + "step": 5615 + }, + { + "epoch": 0.21, + "learning_rate": 1.796169480353023e-07, + "loss": 0.6801, + "step": 5616 + }, + { + "epoch": 0.21, + "learning_rate": 1.7961328597063023e-07, + "loss": 0.6911, + "step": 5617 + }, + { + "epoch": 0.21, + "learning_rate": 1.7960962390595818e-07, + "loss": 0.715, + "step": 5618 + }, + { + "epoch": 0.21, + "learning_rate": 1.796059618412861e-07, + "loss": 0.6942, + "step": 5619 + }, + { + "epoch": 0.21, + "learning_rate": 1.7960229977661406e-07, + "loss": 0.6765, + "step": 5620 + }, + { + "epoch": 0.21, + "learning_rate": 1.79598637711942e-07, + "loss": 0.6839, + "step": 5621 + }, + { + "epoch": 0.21, + "learning_rate": 1.7959497564726992e-07, + "loss": 0.6713, + "step": 5622 + }, + { + "epoch": 0.21, + "learning_rate": 1.7959131358259787e-07, + "loss": 0.6879, + "step": 5623 + }, + { + "epoch": 0.21, + "learning_rate": 1.795876515179258e-07, + "loss": 0.7, + "step": 5624 + }, + { + "epoch": 0.21, + "learning_rate": 1.7958398945325372e-07, + "loss": 0.6959, + "step": 5625 + }, + { + "epoch": 0.21, + "learning_rate": 1.7958032738858168e-07, + "loss": 0.6949, + "step": 5626 + }, + { + "epoch": 0.21, + "learning_rate": 1.795766653239096e-07, + "loss": 0.6969, + "step": 5627 + }, + { + "epoch": 0.21, + "learning_rate": 1.7957300325923753e-07, + "loss": 0.6675, + "step": 5628 + }, + { + "epoch": 0.21, + "learning_rate": 1.7956934119456548e-07, + "loss": 0.6839, + "step": 5629 + }, + { + "epoch": 0.21, + "learning_rate": 1.795656791298934e-07, + "loss": 0.6908, + "step": 5630 + }, + { + "epoch": 0.21, + "learning_rate": 1.7956201706522136e-07, + "loss": 0.6971, + "step": 5631 + }, + { + "epoch": 0.21, + "learning_rate": 1.795583550005493e-07, + "loss": 0.6947, + "step": 5632 + }, + { + "epoch": 0.21, + "learning_rate": 1.7955469293587724e-07, + "loss": 0.6773, + "step": 5633 + }, + { + "epoch": 0.21, + "learning_rate": 1.7955103087120517e-07, + "loss": 0.7112, + "step": 5634 + }, + { + "epoch": 0.21, + "learning_rate": 1.7954736880653312e-07, + "loss": 0.68, + "step": 5635 + }, + { + "epoch": 0.21, + "learning_rate": 1.7954370674186105e-07, + "loss": 0.6984, + "step": 5636 + }, + { + "epoch": 0.21, + "learning_rate": 1.79540044677189e-07, + "loss": 0.6852, + "step": 5637 + }, + { + "epoch": 0.21, + "learning_rate": 1.7953638261251693e-07, + "loss": 0.6935, + "step": 5638 + }, + { + "epoch": 0.21, + "learning_rate": 1.7953272054784488e-07, + "loss": 0.6881, + "step": 5639 + }, + { + "epoch": 0.21, + "learning_rate": 1.795290584831728e-07, + "loss": 0.686, + "step": 5640 + }, + { + "epoch": 0.21, + "learning_rate": 1.7952539641850076e-07, + "loss": 0.6924, + "step": 5641 + }, + { + "epoch": 0.21, + "learning_rate": 1.795217343538287e-07, + "loss": 0.6722, + "step": 5642 + }, + { + "epoch": 0.21, + "learning_rate": 1.7951807228915662e-07, + "loss": 0.7068, + "step": 5643 + }, + { + "epoch": 0.21, + "learning_rate": 1.7951441022448454e-07, + "loss": 0.6793, + "step": 5644 + }, + { + "epoch": 0.21, + "learning_rate": 1.795107481598125e-07, + "loss": 0.693, + "step": 5645 + }, + { + "epoch": 0.21, + "learning_rate": 1.7950708609514042e-07, + "loss": 0.7032, + "step": 5646 + }, + { + "epoch": 0.21, + "learning_rate": 1.7950342403046838e-07, + "loss": 0.6826, + "step": 5647 + }, + { + "epoch": 0.21, + "learning_rate": 1.794997619657963e-07, + "loss": 0.691, + "step": 5648 + }, + { + "epoch": 0.21, + "learning_rate": 1.7949609990112423e-07, + "loss": 0.6971, + "step": 5649 + }, + { + "epoch": 0.21, + "learning_rate": 1.7949243783645218e-07, + "loss": 0.7003, + "step": 5650 + }, + { + "epoch": 0.21, + "learning_rate": 1.794887757717801e-07, + "loss": 0.7133, + "step": 5651 + }, + { + "epoch": 0.21, + "learning_rate": 1.7948511370710806e-07, + "loss": 0.7011, + "step": 5652 + }, + { + "epoch": 0.21, + "learning_rate": 1.79481451642436e-07, + "loss": 0.6911, + "step": 5653 + }, + { + "epoch": 0.21, + "learning_rate": 1.7947778957776394e-07, + "loss": 0.6867, + "step": 5654 + }, + { + "epoch": 0.21, + "learning_rate": 1.7947412751309187e-07, + "loss": 0.7072, + "step": 5655 + }, + { + "epoch": 0.21, + "learning_rate": 1.7947046544841982e-07, + "loss": 0.7006, + "step": 5656 + }, + { + "epoch": 0.21, + "learning_rate": 1.7946680338374775e-07, + "loss": 0.6653, + "step": 5657 + }, + { + "epoch": 0.21, + "learning_rate": 1.794631413190757e-07, + "loss": 0.6972, + "step": 5658 + }, + { + "epoch": 0.21, + "learning_rate": 1.7945947925440363e-07, + "loss": 0.6923, + "step": 5659 + }, + { + "epoch": 0.21, + "learning_rate": 1.7945581718973158e-07, + "loss": 0.6905, + "step": 5660 + }, + { + "epoch": 0.21, + "learning_rate": 1.7945215512505948e-07, + "loss": 0.7095, + "step": 5661 + }, + { + "epoch": 0.21, + "learning_rate": 1.7944849306038744e-07, + "loss": 0.6948, + "step": 5662 + }, + { + "epoch": 0.21, + "learning_rate": 1.7944483099571536e-07, + "loss": 0.6964, + "step": 5663 + }, + { + "epoch": 0.21, + "learning_rate": 1.7944116893104332e-07, + "loss": 0.6951, + "step": 5664 + }, + { + "epoch": 0.21, + "learning_rate": 1.7943750686637124e-07, + "loss": 0.686, + "step": 5665 + }, + { + "epoch": 0.21, + "learning_rate": 1.794338448016992e-07, + "loss": 0.7207, + "step": 5666 + }, + { + "epoch": 0.21, + "learning_rate": 1.7943018273702712e-07, + "loss": 0.7059, + "step": 5667 + }, + { + "epoch": 0.21, + "learning_rate": 1.7942652067235507e-07, + "loss": 0.7297, + "step": 5668 + }, + { + "epoch": 0.21, + "learning_rate": 1.79422858607683e-07, + "loss": 0.6601, + "step": 5669 + }, + { + "epoch": 0.21, + "learning_rate": 1.7941919654301095e-07, + "loss": 0.6869, + "step": 5670 + }, + { + "epoch": 0.21, + "learning_rate": 1.7941553447833888e-07, + "loss": 0.7, + "step": 5671 + }, + { + "epoch": 0.21, + "learning_rate": 1.794118724136668e-07, + "loss": 0.7049, + "step": 5672 + }, + { + "epoch": 0.21, + "learning_rate": 1.7940821034899476e-07, + "loss": 0.6819, + "step": 5673 + }, + { + "epoch": 0.21, + "learning_rate": 1.794045482843227e-07, + "loss": 0.6984, + "step": 5674 + }, + { + "epoch": 0.21, + "learning_rate": 1.7940088621965064e-07, + "loss": 0.696, + "step": 5675 + }, + { + "epoch": 0.21, + "learning_rate": 1.7939722415497857e-07, + "loss": 0.6888, + "step": 5676 + }, + { + "epoch": 0.21, + "learning_rate": 1.7939356209030652e-07, + "loss": 0.6964, + "step": 5677 + }, + { + "epoch": 0.21, + "learning_rate": 1.7938990002563445e-07, + "loss": 0.6978, + "step": 5678 + }, + { + "epoch": 0.21, + "learning_rate": 1.7938623796096237e-07, + "loss": 0.683, + "step": 5679 + }, + { + "epoch": 0.21, + "learning_rate": 1.793825758962903e-07, + "loss": 0.6923, + "step": 5680 + }, + { + "epoch": 0.21, + "learning_rate": 1.7937891383161825e-07, + "loss": 0.6864, + "step": 5681 + }, + { + "epoch": 0.21, + "learning_rate": 1.7937525176694618e-07, + "loss": 0.677, + "step": 5682 + }, + { + "epoch": 0.21, + "learning_rate": 1.7937158970227413e-07, + "loss": 0.6801, + "step": 5683 + }, + { + "epoch": 0.21, + "learning_rate": 1.7936792763760206e-07, + "loss": 0.6983, + "step": 5684 + }, + { + "epoch": 0.21, + "learning_rate": 1.7936426557293001e-07, + "loss": 0.6942, + "step": 5685 + }, + { + "epoch": 0.21, + "learning_rate": 1.7936060350825794e-07, + "loss": 0.684, + "step": 5686 + }, + { + "epoch": 0.21, + "learning_rate": 1.793569414435859e-07, + "loss": 0.6843, + "step": 5687 + }, + { + "epoch": 0.21, + "learning_rate": 1.7935327937891382e-07, + "loss": 0.6711, + "step": 5688 + }, + { + "epoch": 0.21, + "learning_rate": 1.7934961731424177e-07, + "loss": 0.6855, + "step": 5689 + }, + { + "epoch": 0.21, + "learning_rate": 1.793459552495697e-07, + "loss": 0.7051, + "step": 5690 + }, + { + "epoch": 0.21, + "learning_rate": 1.7934229318489765e-07, + "loss": 0.6996, + "step": 5691 + }, + { + "epoch": 0.21, + "learning_rate": 1.7933863112022558e-07, + "loss": 0.7083, + "step": 5692 + }, + { + "epoch": 0.21, + "learning_rate": 1.793349690555535e-07, + "loss": 0.6816, + "step": 5693 + }, + { + "epoch": 0.21, + "learning_rate": 1.7933130699088146e-07, + "loss": 0.6936, + "step": 5694 + }, + { + "epoch": 0.21, + "learning_rate": 1.793276449262094e-07, + "loss": 0.6902, + "step": 5695 + }, + { + "epoch": 0.21, + "learning_rate": 1.7932398286153734e-07, + "loss": 0.702, + "step": 5696 + }, + { + "epoch": 0.21, + "learning_rate": 1.7932032079686527e-07, + "loss": 0.7018, + "step": 5697 + }, + { + "epoch": 0.21, + "learning_rate": 1.793166587321932e-07, + "loss": 0.6712, + "step": 5698 + }, + { + "epoch": 0.21, + "learning_rate": 1.7931299666752112e-07, + "loss": 0.6975, + "step": 5699 + }, + { + "epoch": 0.21, + "learning_rate": 1.7930933460284907e-07, + "loss": 0.685, + "step": 5700 + }, + { + "epoch": 0.21, + "learning_rate": 1.79305672538177e-07, + "loss": 0.6885, + "step": 5701 + }, + { + "epoch": 0.21, + "learning_rate": 1.7930201047350495e-07, + "loss": 0.6938, + "step": 5702 + }, + { + "epoch": 0.21, + "learning_rate": 1.7929834840883288e-07, + "loss": 0.6897, + "step": 5703 + }, + { + "epoch": 0.21, + "learning_rate": 1.7929468634416083e-07, + "loss": 0.6993, + "step": 5704 + }, + { + "epoch": 0.21, + "learning_rate": 1.7929102427948876e-07, + "loss": 0.6898, + "step": 5705 + }, + { + "epoch": 0.21, + "learning_rate": 1.792873622148167e-07, + "loss": 0.7104, + "step": 5706 + }, + { + "epoch": 0.21, + "learning_rate": 1.7928370015014464e-07, + "loss": 0.701, + "step": 5707 + }, + { + "epoch": 0.21, + "learning_rate": 1.792800380854726e-07, + "loss": 0.7089, + "step": 5708 + }, + { + "epoch": 0.21, + "learning_rate": 1.7927637602080052e-07, + "loss": 0.6783, + "step": 5709 + }, + { + "epoch": 0.21, + "learning_rate": 1.7927271395612847e-07, + "loss": 0.7002, + "step": 5710 + }, + { + "epoch": 0.21, + "learning_rate": 1.792690518914564e-07, + "loss": 0.7046, + "step": 5711 + }, + { + "epoch": 0.21, + "learning_rate": 1.7926538982678435e-07, + "loss": 0.6773, + "step": 5712 + }, + { + "epoch": 0.21, + "learning_rate": 1.7926172776211228e-07, + "loss": 0.7035, + "step": 5713 + }, + { + "epoch": 0.21, + "learning_rate": 1.7925806569744023e-07, + "loss": 0.6868, + "step": 5714 + }, + { + "epoch": 0.21, + "learning_rate": 1.7925440363276813e-07, + "loss": 0.6948, + "step": 5715 + }, + { + "epoch": 0.21, + "learning_rate": 1.7925074156809609e-07, + "loss": 0.7019, + "step": 5716 + }, + { + "epoch": 0.21, + "learning_rate": 1.79247079503424e-07, + "loss": 0.694, + "step": 5717 + }, + { + "epoch": 0.21, + "learning_rate": 1.7924341743875197e-07, + "loss": 0.6907, + "step": 5718 + }, + { + "epoch": 0.21, + "learning_rate": 1.792397553740799e-07, + "loss": 0.7141, + "step": 5719 + }, + { + "epoch": 0.21, + "learning_rate": 1.7923609330940782e-07, + "loss": 0.6969, + "step": 5720 + }, + { + "epoch": 0.21, + "learning_rate": 1.7923243124473577e-07, + "loss": 0.6907, + "step": 5721 + }, + { + "epoch": 0.21, + "learning_rate": 1.792287691800637e-07, + "loss": 0.6815, + "step": 5722 + }, + { + "epoch": 0.21, + "learning_rate": 1.7922510711539165e-07, + "loss": 0.7027, + "step": 5723 + }, + { + "epoch": 0.21, + "learning_rate": 1.7922144505071958e-07, + "loss": 0.7027, + "step": 5724 + }, + { + "epoch": 0.21, + "learning_rate": 1.7921778298604753e-07, + "loss": 0.6887, + "step": 5725 + }, + { + "epoch": 0.21, + "learning_rate": 1.7921412092137546e-07, + "loss": 0.6877, + "step": 5726 + }, + { + "epoch": 0.21, + "learning_rate": 1.792104588567034e-07, + "loss": 0.7098, + "step": 5727 + }, + { + "epoch": 0.21, + "learning_rate": 1.7920679679203134e-07, + "loss": 0.6837, + "step": 5728 + }, + { + "epoch": 0.21, + "learning_rate": 1.792031347273593e-07, + "loss": 0.6966, + "step": 5729 + }, + { + "epoch": 0.21, + "learning_rate": 1.7919947266268722e-07, + "loss": 0.6996, + "step": 5730 + }, + { + "epoch": 0.21, + "learning_rate": 1.7919581059801517e-07, + "loss": 0.688, + "step": 5731 + }, + { + "epoch": 0.21, + "learning_rate": 1.791921485333431e-07, + "loss": 0.6833, + "step": 5732 + }, + { + "epoch": 0.21, + "learning_rate": 1.7918848646867102e-07, + "loss": 0.668, + "step": 5733 + }, + { + "epoch": 0.21, + "learning_rate": 1.7918482440399895e-07, + "loss": 0.6792, + "step": 5734 + }, + { + "epoch": 0.21, + "learning_rate": 1.791811623393269e-07, + "loss": 0.7085, + "step": 5735 + }, + { + "epoch": 0.21, + "learning_rate": 1.7917750027465483e-07, + "loss": 0.7041, + "step": 5736 + }, + { + "epoch": 0.21, + "learning_rate": 1.7917383820998278e-07, + "loss": 0.6947, + "step": 5737 + }, + { + "epoch": 0.21, + "learning_rate": 1.791701761453107e-07, + "loss": 0.696, + "step": 5738 + }, + { + "epoch": 0.21, + "learning_rate": 1.7916651408063866e-07, + "loss": 0.692, + "step": 5739 + }, + { + "epoch": 0.21, + "learning_rate": 1.791628520159666e-07, + "loss": 0.7061, + "step": 5740 + }, + { + "epoch": 0.21, + "learning_rate": 1.7915918995129454e-07, + "loss": 0.6888, + "step": 5741 + }, + { + "epoch": 0.21, + "learning_rate": 1.7915552788662247e-07, + "loss": 0.684, + "step": 5742 + }, + { + "epoch": 0.21, + "learning_rate": 1.791518658219504e-07, + "loss": 0.7007, + "step": 5743 + }, + { + "epoch": 0.21, + "learning_rate": 1.7914820375727835e-07, + "loss": 0.6898, + "step": 5744 + }, + { + "epoch": 0.21, + "learning_rate": 1.7914454169260628e-07, + "loss": 0.6935, + "step": 5745 + }, + { + "epoch": 0.21, + "learning_rate": 1.7914087962793423e-07, + "loss": 0.6965, + "step": 5746 + }, + { + "epoch": 0.21, + "learning_rate": 1.7913721756326216e-07, + "loss": 0.6896, + "step": 5747 + }, + { + "epoch": 0.21, + "learning_rate": 1.791335554985901e-07, + "loss": 0.7047, + "step": 5748 + }, + { + "epoch": 0.21, + "learning_rate": 1.7912989343391804e-07, + "loss": 0.6848, + "step": 5749 + }, + { + "epoch": 0.21, + "learning_rate": 1.79126231369246e-07, + "loss": 0.6971, + "step": 5750 + }, + { + "epoch": 0.21, + "learning_rate": 1.791225693045739e-07, + "loss": 0.6886, + "step": 5751 + }, + { + "epoch": 0.21, + "learning_rate": 1.7911890723990184e-07, + "loss": 0.6908, + "step": 5752 + }, + { + "epoch": 0.21, + "learning_rate": 1.7911524517522977e-07, + "loss": 0.6765, + "step": 5753 + }, + { + "epoch": 0.21, + "learning_rate": 1.7911158311055772e-07, + "loss": 0.7029, + "step": 5754 + }, + { + "epoch": 0.21, + "learning_rate": 1.7910792104588565e-07, + "loss": 0.6876, + "step": 5755 + }, + { + "epoch": 0.21, + "learning_rate": 1.791042589812136e-07, + "loss": 0.6919, + "step": 5756 + }, + { + "epoch": 0.21, + "learning_rate": 1.7910059691654153e-07, + "loss": 0.6909, + "step": 5757 + }, + { + "epoch": 0.21, + "learning_rate": 1.7909693485186948e-07, + "loss": 0.6948, + "step": 5758 + }, + { + "epoch": 0.21, + "learning_rate": 1.790932727871974e-07, + "loss": 0.6879, + "step": 5759 + }, + { + "epoch": 0.21, + "learning_rate": 1.7908961072252536e-07, + "loss": 0.694, + "step": 5760 + }, + { + "epoch": 0.21, + "learning_rate": 1.790859486578533e-07, + "loss": 0.6981, + "step": 5761 + }, + { + "epoch": 0.21, + "learning_rate": 1.7908228659318124e-07, + "loss": 0.7137, + "step": 5762 + }, + { + "epoch": 0.21, + "learning_rate": 1.7907862452850917e-07, + "loss": 0.694, + "step": 5763 + }, + { + "epoch": 0.21, + "learning_rate": 1.790749624638371e-07, + "loss": 0.6951, + "step": 5764 + }, + { + "epoch": 0.21, + "learning_rate": 1.7907130039916505e-07, + "loss": 0.6997, + "step": 5765 + }, + { + "epoch": 0.21, + "learning_rate": 1.7906763833449298e-07, + "loss": 0.6927, + "step": 5766 + }, + { + "epoch": 0.21, + "learning_rate": 1.7906397626982093e-07, + "loss": 0.6951, + "step": 5767 + }, + { + "epoch": 0.21, + "learning_rate": 1.7906031420514886e-07, + "loss": 0.6972, + "step": 5768 + }, + { + "epoch": 0.21, + "learning_rate": 1.7905665214047678e-07, + "loss": 0.6651, + "step": 5769 + }, + { + "epoch": 0.21, + "learning_rate": 1.790529900758047e-07, + "loss": 0.6621, + "step": 5770 + }, + { + "epoch": 0.21, + "learning_rate": 1.7904932801113266e-07, + "loss": 0.7004, + "step": 5771 + }, + { + "epoch": 0.21, + "learning_rate": 1.790456659464606e-07, + "loss": 0.6892, + "step": 5772 + }, + { + "epoch": 0.21, + "learning_rate": 1.7904200388178854e-07, + "loss": 0.6822, + "step": 5773 + }, + { + "epoch": 0.21, + "learning_rate": 1.7903834181711647e-07, + "loss": 0.6987, + "step": 5774 + }, + { + "epoch": 0.21, + "learning_rate": 1.7903467975244442e-07, + "loss": 0.6896, + "step": 5775 + }, + { + "epoch": 0.21, + "learning_rate": 1.7903101768777235e-07, + "loss": 0.6901, + "step": 5776 + }, + { + "epoch": 0.21, + "learning_rate": 1.790273556231003e-07, + "loss": 0.6931, + "step": 5777 + }, + { + "epoch": 0.21, + "learning_rate": 1.7902369355842823e-07, + "loss": 0.6929, + "step": 5778 + }, + { + "epoch": 0.21, + "learning_rate": 1.7902003149375618e-07, + "loss": 0.6957, + "step": 5779 + }, + { + "epoch": 0.21, + "learning_rate": 1.790163694290841e-07, + "loss": 0.7009, + "step": 5780 + }, + { + "epoch": 0.21, + "learning_rate": 1.7901270736441206e-07, + "loss": 0.6999, + "step": 5781 + }, + { + "epoch": 0.21, + "learning_rate": 1.7900904529974e-07, + "loss": 0.6991, + "step": 5782 + }, + { + "epoch": 0.21, + "learning_rate": 1.7900538323506794e-07, + "loss": 0.6962, + "step": 5783 + }, + { + "epoch": 0.21, + "learning_rate": 1.7900172117039587e-07, + "loss": 0.6584, + "step": 5784 + }, + { + "epoch": 0.21, + "learning_rate": 1.7899805910572382e-07, + "loss": 0.6907, + "step": 5785 + }, + { + "epoch": 0.21, + "learning_rate": 1.7899439704105175e-07, + "loss": 0.6958, + "step": 5786 + }, + { + "epoch": 0.21, + "learning_rate": 1.7899073497637968e-07, + "loss": 0.6942, + "step": 5787 + }, + { + "epoch": 0.21, + "learning_rate": 1.789870729117076e-07, + "loss": 0.6944, + "step": 5788 + }, + { + "epoch": 0.21, + "learning_rate": 1.7898341084703555e-07, + "loss": 0.694, + "step": 5789 + }, + { + "epoch": 0.21, + "learning_rate": 1.7897974878236348e-07, + "loss": 0.6965, + "step": 5790 + }, + { + "epoch": 0.21, + "learning_rate": 1.789760867176914e-07, + "loss": 0.6942, + "step": 5791 + }, + { + "epoch": 0.21, + "learning_rate": 1.7897242465301936e-07, + "loss": 0.6806, + "step": 5792 + }, + { + "epoch": 0.21, + "learning_rate": 1.789687625883473e-07, + "loss": 0.6952, + "step": 5793 + }, + { + "epoch": 0.21, + "learning_rate": 1.7896510052367524e-07, + "loss": 0.6956, + "step": 5794 + }, + { + "epoch": 0.21, + "learning_rate": 1.7896143845900317e-07, + "loss": 0.7082, + "step": 5795 + }, + { + "epoch": 0.21, + "learning_rate": 1.7895777639433112e-07, + "loss": 0.6784, + "step": 5796 + }, + { + "epoch": 0.21, + "learning_rate": 1.7895411432965905e-07, + "loss": 0.6982, + "step": 5797 + }, + { + "epoch": 0.21, + "learning_rate": 1.78950452264987e-07, + "loss": 0.686, + "step": 5798 + }, + { + "epoch": 0.21, + "learning_rate": 1.7894679020031493e-07, + "loss": 0.6877, + "step": 5799 + }, + { + "epoch": 0.21, + "learning_rate": 1.7894312813564288e-07, + "loss": 0.6898, + "step": 5800 + }, + { + "epoch": 0.21, + "learning_rate": 1.789394660709708e-07, + "loss": 0.7001, + "step": 5801 + }, + { + "epoch": 0.21, + "learning_rate": 1.7893580400629876e-07, + "loss": 0.6844, + "step": 5802 + }, + { + "epoch": 0.21, + "learning_rate": 1.789321419416267e-07, + "loss": 0.6963, + "step": 5803 + }, + { + "epoch": 0.21, + "learning_rate": 1.7892847987695464e-07, + "loss": 0.6972, + "step": 5804 + }, + { + "epoch": 0.21, + "learning_rate": 1.7892481781228254e-07, + "loss": 0.7011, + "step": 5805 + }, + { + "epoch": 0.21, + "learning_rate": 1.789211557476105e-07, + "loss": 0.7093, + "step": 5806 + }, + { + "epoch": 0.21, + "learning_rate": 1.7891749368293842e-07, + "loss": 0.6662, + "step": 5807 + }, + { + "epoch": 0.21, + "learning_rate": 1.7891383161826637e-07, + "loss": 0.7121, + "step": 5808 + }, + { + "epoch": 0.21, + "learning_rate": 1.789101695535943e-07, + "loss": 0.6897, + "step": 5809 + }, + { + "epoch": 0.21, + "learning_rate": 1.7890650748892225e-07, + "loss": 0.6888, + "step": 5810 + }, + { + "epoch": 0.21, + "learning_rate": 1.7890284542425018e-07, + "loss": 0.717, + "step": 5811 + }, + { + "epoch": 0.21, + "learning_rate": 1.7889918335957813e-07, + "loss": 0.6743, + "step": 5812 + }, + { + "epoch": 0.21, + "learning_rate": 1.7889552129490606e-07, + "loss": 0.6919, + "step": 5813 + }, + { + "epoch": 0.21, + "learning_rate": 1.78891859230234e-07, + "loss": 0.669, + "step": 5814 + }, + { + "epoch": 0.21, + "learning_rate": 1.7888819716556194e-07, + "loss": 0.6885, + "step": 5815 + }, + { + "epoch": 0.21, + "learning_rate": 1.7888453510088987e-07, + "loss": 0.7054, + "step": 5816 + }, + { + "epoch": 0.21, + "learning_rate": 1.7888087303621782e-07, + "loss": 0.696, + "step": 5817 + }, + { + "epoch": 0.21, + "learning_rate": 1.7887721097154575e-07, + "loss": 0.6897, + "step": 5818 + }, + { + "epoch": 0.21, + "learning_rate": 1.788735489068737e-07, + "loss": 0.7033, + "step": 5819 + }, + { + "epoch": 0.21, + "learning_rate": 1.7886988684220163e-07, + "loss": 0.6831, + "step": 5820 + }, + { + "epoch": 0.21, + "learning_rate": 1.7886622477752958e-07, + "loss": 0.7037, + "step": 5821 + }, + { + "epoch": 0.21, + "learning_rate": 1.7886256271285748e-07, + "loss": 0.6914, + "step": 5822 + }, + { + "epoch": 0.21, + "learning_rate": 1.7885890064818543e-07, + "loss": 0.6929, + "step": 5823 + }, + { + "epoch": 0.21, + "learning_rate": 1.7885523858351336e-07, + "loss": 0.6952, + "step": 5824 + }, + { + "epoch": 0.21, + "learning_rate": 1.788515765188413e-07, + "loss": 0.6904, + "step": 5825 + }, + { + "epoch": 0.21, + "learning_rate": 1.7884791445416924e-07, + "loss": 0.6831, + "step": 5826 + }, + { + "epoch": 0.21, + "learning_rate": 1.788442523894972e-07, + "loss": 0.6952, + "step": 5827 + }, + { + "epoch": 0.21, + "learning_rate": 1.7884059032482512e-07, + "loss": 0.6877, + "step": 5828 + }, + { + "epoch": 0.21, + "learning_rate": 1.7883692826015307e-07, + "loss": 0.6809, + "step": 5829 + }, + { + "epoch": 0.21, + "learning_rate": 1.78833266195481e-07, + "loss": 0.7007, + "step": 5830 + }, + { + "epoch": 0.21, + "learning_rate": 1.7882960413080895e-07, + "loss": 0.6984, + "step": 5831 + }, + { + "epoch": 0.21, + "learning_rate": 1.7882594206613688e-07, + "loss": 0.6986, + "step": 5832 + }, + { + "epoch": 0.21, + "learning_rate": 1.7882228000146483e-07, + "loss": 0.6857, + "step": 5833 + }, + { + "epoch": 0.21, + "learning_rate": 1.7881861793679276e-07, + "loss": 0.6985, + "step": 5834 + }, + { + "epoch": 0.21, + "learning_rate": 1.7881495587212069e-07, + "loss": 0.6909, + "step": 5835 + }, + { + "epoch": 0.21, + "learning_rate": 1.7881129380744864e-07, + "loss": 0.6948, + "step": 5836 + }, + { + "epoch": 0.21, + "learning_rate": 1.7880763174277657e-07, + "loss": 0.6883, + "step": 5837 + }, + { + "epoch": 0.21, + "learning_rate": 1.7880396967810452e-07, + "loss": 0.7031, + "step": 5838 + }, + { + "epoch": 0.21, + "learning_rate": 1.7880030761343245e-07, + "loss": 0.6847, + "step": 5839 + }, + { + "epoch": 0.21, + "learning_rate": 1.7879664554876037e-07, + "loss": 0.6746, + "step": 5840 + }, + { + "epoch": 0.21, + "learning_rate": 1.787929834840883e-07, + "loss": 0.6962, + "step": 5841 + }, + { + "epoch": 0.21, + "learning_rate": 1.7878932141941625e-07, + "loss": 0.6988, + "step": 5842 + }, + { + "epoch": 0.21, + "learning_rate": 1.7878565935474418e-07, + "loss": 0.7015, + "step": 5843 + }, + { + "epoch": 0.21, + "learning_rate": 1.7878199729007213e-07, + "loss": 0.681, + "step": 5844 + }, + { + "epoch": 0.21, + "learning_rate": 1.7877833522540006e-07, + "loss": 0.7022, + "step": 5845 + }, + { + "epoch": 0.21, + "learning_rate": 1.78774673160728e-07, + "loss": 0.6819, + "step": 5846 + }, + { + "epoch": 0.21, + "learning_rate": 1.7877101109605594e-07, + "loss": 0.6903, + "step": 5847 + }, + { + "epoch": 0.21, + "learning_rate": 1.787673490313839e-07, + "loss": 0.6909, + "step": 5848 + }, + { + "epoch": 0.21, + "learning_rate": 1.7876368696671182e-07, + "loss": 0.679, + "step": 5849 + }, + { + "epoch": 0.21, + "learning_rate": 1.7876002490203977e-07, + "loss": 0.6902, + "step": 5850 + }, + { + "epoch": 0.21, + "learning_rate": 1.787563628373677e-07, + "loss": 0.6789, + "step": 5851 + }, + { + "epoch": 0.21, + "learning_rate": 1.7875270077269565e-07, + "loss": 0.6921, + "step": 5852 + }, + { + "epoch": 0.21, + "learning_rate": 1.7874903870802358e-07, + "loss": 0.7091, + "step": 5853 + }, + { + "epoch": 0.21, + "learning_rate": 1.7874537664335153e-07, + "loss": 0.7007, + "step": 5854 + }, + { + "epoch": 0.21, + "learning_rate": 1.7874171457867946e-07, + "loss": 0.6846, + "step": 5855 + }, + { + "epoch": 0.21, + "learning_rate": 1.787380525140074e-07, + "loss": 0.6971, + "step": 5856 + }, + { + "epoch": 0.21, + "learning_rate": 1.7873439044933534e-07, + "loss": 0.6814, + "step": 5857 + }, + { + "epoch": 0.21, + "learning_rate": 1.7873072838466326e-07, + "loss": 0.6749, + "step": 5858 + }, + { + "epoch": 0.21, + "learning_rate": 1.787270663199912e-07, + "loss": 0.6932, + "step": 5859 + }, + { + "epoch": 0.21, + "learning_rate": 1.7872340425531914e-07, + "loss": 0.6844, + "step": 5860 + }, + { + "epoch": 0.21, + "learning_rate": 1.7871974219064707e-07, + "loss": 0.7025, + "step": 5861 + }, + { + "epoch": 0.21, + "learning_rate": 1.78716080125975e-07, + "loss": 0.6944, + "step": 5862 + }, + { + "epoch": 0.21, + "learning_rate": 1.7871241806130295e-07, + "loss": 0.6998, + "step": 5863 + }, + { + "epoch": 0.21, + "learning_rate": 1.7870875599663088e-07, + "loss": 0.6901, + "step": 5864 + }, + { + "epoch": 0.21, + "learning_rate": 1.7870509393195883e-07, + "loss": 0.6673, + "step": 5865 + }, + { + "epoch": 0.21, + "learning_rate": 1.7870143186728676e-07, + "loss": 0.6984, + "step": 5866 + }, + { + "epoch": 0.21, + "learning_rate": 1.786977698026147e-07, + "loss": 0.6881, + "step": 5867 + }, + { + "epoch": 0.21, + "learning_rate": 1.7869410773794264e-07, + "loss": 0.7033, + "step": 5868 + }, + { + "epoch": 0.21, + "learning_rate": 1.786904456732706e-07, + "loss": 0.6953, + "step": 5869 + }, + { + "epoch": 0.21, + "learning_rate": 1.7868678360859852e-07, + "loss": 0.6726, + "step": 5870 + }, + { + "epoch": 0.21, + "learning_rate": 1.7868312154392647e-07, + "loss": 0.6766, + "step": 5871 + }, + { + "epoch": 0.21, + "learning_rate": 1.786794594792544e-07, + "loss": 0.6812, + "step": 5872 + }, + { + "epoch": 0.21, + "learning_rate": 1.7867579741458235e-07, + "loss": 0.6701, + "step": 5873 + }, + { + "epoch": 0.21, + "learning_rate": 1.7867213534991028e-07, + "loss": 0.6949, + "step": 5874 + }, + { + "epoch": 0.21, + "learning_rate": 1.7866847328523823e-07, + "loss": 0.6975, + "step": 5875 + }, + { + "epoch": 0.21, + "learning_rate": 1.7866481122056613e-07, + "loss": 0.7046, + "step": 5876 + }, + { + "epoch": 0.22, + "learning_rate": 1.7866114915589408e-07, + "loss": 0.6921, + "step": 5877 + }, + { + "epoch": 0.22, + "learning_rate": 1.78657487091222e-07, + "loss": 0.7044, + "step": 5878 + }, + { + "epoch": 0.22, + "learning_rate": 1.7865382502654996e-07, + "loss": 0.6943, + "step": 5879 + }, + { + "epoch": 0.22, + "learning_rate": 1.786501629618779e-07, + "loss": 0.6864, + "step": 5880 + }, + { + "epoch": 0.22, + "learning_rate": 1.7864650089720584e-07, + "loss": 0.6903, + "step": 5881 + }, + { + "epoch": 0.22, + "learning_rate": 1.7864283883253377e-07, + "loss": 0.6787, + "step": 5882 + }, + { + "epoch": 0.22, + "learning_rate": 1.7863917676786172e-07, + "loss": 0.6871, + "step": 5883 + }, + { + "epoch": 0.22, + "learning_rate": 1.7863551470318965e-07, + "loss": 0.6964, + "step": 5884 + }, + { + "epoch": 0.22, + "learning_rate": 1.7863185263851758e-07, + "loss": 0.6987, + "step": 5885 + }, + { + "epoch": 0.22, + "learning_rate": 1.7862819057384553e-07, + "loss": 0.7001, + "step": 5886 + }, + { + "epoch": 0.22, + "learning_rate": 1.7862452850917346e-07, + "loss": 0.6929, + "step": 5887 + }, + { + "epoch": 0.22, + "learning_rate": 1.786208664445014e-07, + "loss": 0.6828, + "step": 5888 + }, + { + "epoch": 0.22, + "learning_rate": 1.7861720437982934e-07, + "loss": 0.6977, + "step": 5889 + }, + { + "epoch": 0.22, + "learning_rate": 1.786135423151573e-07, + "loss": 0.7023, + "step": 5890 + }, + { + "epoch": 0.22, + "learning_rate": 1.7860988025048522e-07, + "loss": 0.7062, + "step": 5891 + }, + { + "epoch": 0.22, + "learning_rate": 1.7860621818581317e-07, + "loss": 0.7023, + "step": 5892 + }, + { + "epoch": 0.22, + "learning_rate": 1.786025561211411e-07, + "loss": 0.6751, + "step": 5893 + }, + { + "epoch": 0.22, + "learning_rate": 1.7859889405646902e-07, + "loss": 0.6958, + "step": 5894 + }, + { + "epoch": 0.22, + "learning_rate": 1.7859523199179695e-07, + "loss": 0.6781, + "step": 5895 + }, + { + "epoch": 0.22, + "learning_rate": 1.785915699271249e-07, + "loss": 0.6931, + "step": 5896 + }, + { + "epoch": 0.22, + "learning_rate": 1.7858790786245283e-07, + "loss": 0.6904, + "step": 5897 + }, + { + "epoch": 0.22, + "learning_rate": 1.7858424579778078e-07, + "loss": 0.6912, + "step": 5898 + }, + { + "epoch": 0.22, + "learning_rate": 1.785805837331087e-07, + "loss": 0.6976, + "step": 5899 + }, + { + "epoch": 0.22, + "learning_rate": 1.7857692166843666e-07, + "loss": 0.7057, + "step": 5900 + }, + { + "epoch": 0.22, + "learning_rate": 1.785732596037646e-07, + "loss": 0.7102, + "step": 5901 + }, + { + "epoch": 0.22, + "learning_rate": 1.7856959753909254e-07, + "loss": 0.6872, + "step": 5902 + }, + { + "epoch": 0.22, + "learning_rate": 1.7856593547442047e-07, + "loss": 0.6976, + "step": 5903 + }, + { + "epoch": 0.22, + "learning_rate": 1.7856227340974842e-07, + "loss": 0.6949, + "step": 5904 + }, + { + "epoch": 0.22, + "learning_rate": 1.7855861134507635e-07, + "loss": 0.6976, + "step": 5905 + }, + { + "epoch": 0.22, + "learning_rate": 1.7855494928040428e-07, + "loss": 0.6963, + "step": 5906 + }, + { + "epoch": 0.22, + "learning_rate": 1.7855128721573223e-07, + "loss": 0.6784, + "step": 5907 + }, + { + "epoch": 0.22, + "learning_rate": 1.7854762515106015e-07, + "loss": 0.6961, + "step": 5908 + }, + { + "epoch": 0.22, + "learning_rate": 1.785439630863881e-07, + "loss": 0.6962, + "step": 5909 + }, + { + "epoch": 0.22, + "learning_rate": 1.7854030102171603e-07, + "loss": 0.6913, + "step": 5910 + }, + { + "epoch": 0.22, + "learning_rate": 1.78536638957044e-07, + "loss": 0.6897, + "step": 5911 + }, + { + "epoch": 0.22, + "learning_rate": 1.785329768923719e-07, + "loss": 0.6993, + "step": 5912 + }, + { + "epoch": 0.22, + "learning_rate": 1.7852931482769984e-07, + "loss": 0.6986, + "step": 5913 + }, + { + "epoch": 0.22, + "learning_rate": 1.7852565276302777e-07, + "loss": 0.6935, + "step": 5914 + }, + { + "epoch": 0.22, + "learning_rate": 1.7852199069835572e-07, + "loss": 0.6988, + "step": 5915 + }, + { + "epoch": 0.22, + "learning_rate": 1.7851832863368365e-07, + "loss": 0.6991, + "step": 5916 + }, + { + "epoch": 0.22, + "learning_rate": 1.785146665690116e-07, + "loss": 0.6848, + "step": 5917 + }, + { + "epoch": 0.22, + "learning_rate": 1.7851100450433953e-07, + "loss": 0.6794, + "step": 5918 + }, + { + "epoch": 0.22, + "learning_rate": 1.7850734243966748e-07, + "loss": 0.6962, + "step": 5919 + }, + { + "epoch": 0.22, + "learning_rate": 1.785036803749954e-07, + "loss": 0.7035, + "step": 5920 + }, + { + "epoch": 0.22, + "learning_rate": 1.7850001831032336e-07, + "loss": 0.6853, + "step": 5921 + }, + { + "epoch": 0.22, + "learning_rate": 1.784963562456513e-07, + "loss": 0.6909, + "step": 5922 + }, + { + "epoch": 0.22, + "learning_rate": 1.7849269418097924e-07, + "loss": 0.6905, + "step": 5923 + }, + { + "epoch": 0.22, + "learning_rate": 1.7848903211630717e-07, + "loss": 0.6999, + "step": 5924 + }, + { + "epoch": 0.22, + "learning_rate": 1.7848537005163512e-07, + "loss": 0.6919, + "step": 5925 + }, + { + "epoch": 0.22, + "learning_rate": 1.7848170798696305e-07, + "loss": 0.7015, + "step": 5926 + }, + { + "epoch": 0.22, + "learning_rate": 1.78478045922291e-07, + "loss": 0.6927, + "step": 5927 + }, + { + "epoch": 0.22, + "learning_rate": 1.7847438385761893e-07, + "loss": 0.669, + "step": 5928 + }, + { + "epoch": 0.22, + "learning_rate": 1.7847072179294685e-07, + "loss": 0.6983, + "step": 5929 + }, + { + "epoch": 0.22, + "learning_rate": 1.7846705972827478e-07, + "loss": 0.7023, + "step": 5930 + }, + { + "epoch": 0.22, + "learning_rate": 1.7846339766360273e-07, + "loss": 0.7001, + "step": 5931 + }, + { + "epoch": 0.22, + "learning_rate": 1.7845973559893066e-07, + "loss": 0.7066, + "step": 5932 + }, + { + "epoch": 0.22, + "learning_rate": 1.784560735342586e-07, + "loss": 0.7248, + "step": 5933 + }, + { + "epoch": 0.22, + "learning_rate": 1.7845241146958654e-07, + "loss": 0.6872, + "step": 5934 + }, + { + "epoch": 0.22, + "learning_rate": 1.7844874940491447e-07, + "loss": 0.7126, + "step": 5935 + }, + { + "epoch": 0.22, + "learning_rate": 1.7844508734024242e-07, + "loss": 0.719, + "step": 5936 + }, + { + "epoch": 0.22, + "learning_rate": 1.7844142527557035e-07, + "loss": 0.7162, + "step": 5937 + }, + { + "epoch": 0.22, + "learning_rate": 1.784377632108983e-07, + "loss": 0.6964, + "step": 5938 + }, + { + "epoch": 0.22, + "learning_rate": 1.7843410114622623e-07, + "loss": 0.6651, + "step": 5939 + }, + { + "epoch": 0.22, + "learning_rate": 1.7843043908155418e-07, + "loss": 0.6835, + "step": 5940 + }, + { + "epoch": 0.22, + "learning_rate": 1.784267770168821e-07, + "loss": 0.6921, + "step": 5941 + }, + { + "epoch": 0.22, + "learning_rate": 1.7842311495221006e-07, + "loss": 0.7193, + "step": 5942 + }, + { + "epoch": 0.22, + "learning_rate": 1.7841945288753799e-07, + "loss": 0.6985, + "step": 5943 + }, + { + "epoch": 0.22, + "learning_rate": 1.7841579082286594e-07, + "loss": 0.6948, + "step": 5944 + }, + { + "epoch": 0.22, + "learning_rate": 1.7841212875819387e-07, + "loss": 0.6957, + "step": 5945 + }, + { + "epoch": 0.22, + "learning_rate": 1.7840846669352182e-07, + "loss": 0.6942, + "step": 5946 + }, + { + "epoch": 0.22, + "learning_rate": 1.7840480462884975e-07, + "loss": 0.6927, + "step": 5947 + }, + { + "epoch": 0.22, + "learning_rate": 1.7840114256417767e-07, + "loss": 0.6958, + "step": 5948 + }, + { + "epoch": 0.22, + "learning_rate": 1.783974804995056e-07, + "loss": 0.6917, + "step": 5949 + }, + { + "epoch": 0.22, + "learning_rate": 1.7839381843483355e-07, + "loss": 0.6978, + "step": 5950 + }, + { + "epoch": 0.22, + "learning_rate": 1.7839015637016148e-07, + "loss": 0.6832, + "step": 5951 + }, + { + "epoch": 0.22, + "learning_rate": 1.7838649430548943e-07, + "loss": 0.7057, + "step": 5952 + }, + { + "epoch": 0.22, + "learning_rate": 1.7838283224081736e-07, + "loss": 0.6817, + "step": 5953 + }, + { + "epoch": 0.22, + "learning_rate": 1.783791701761453e-07, + "loss": 0.6956, + "step": 5954 + }, + { + "epoch": 0.22, + "learning_rate": 1.7837550811147324e-07, + "loss": 0.6995, + "step": 5955 + }, + { + "epoch": 0.22, + "learning_rate": 1.7837184604680117e-07, + "loss": 0.6857, + "step": 5956 + }, + { + "epoch": 0.22, + "learning_rate": 1.7836818398212912e-07, + "loss": 0.7046, + "step": 5957 + }, + { + "epoch": 0.22, + "learning_rate": 1.7836452191745705e-07, + "loss": 0.6992, + "step": 5958 + }, + { + "epoch": 0.22, + "learning_rate": 1.78360859852785e-07, + "loss": 0.6996, + "step": 5959 + }, + { + "epoch": 0.22, + "learning_rate": 1.7835719778811293e-07, + "loss": 0.7123, + "step": 5960 + }, + { + "epoch": 0.22, + "learning_rate": 1.7835353572344088e-07, + "loss": 0.6908, + "step": 5961 + }, + { + "epoch": 0.22, + "learning_rate": 1.783498736587688e-07, + "loss": 0.6811, + "step": 5962 + }, + { + "epoch": 0.22, + "learning_rate": 1.7834621159409676e-07, + "loss": 0.7031, + "step": 5963 + }, + { + "epoch": 0.22, + "learning_rate": 1.7834254952942469e-07, + "loss": 0.7075, + "step": 5964 + }, + { + "epoch": 0.22, + "learning_rate": 1.7833888746475264e-07, + "loss": 0.7064, + "step": 5965 + }, + { + "epoch": 0.22, + "learning_rate": 1.7833522540008054e-07, + "loss": 0.6987, + "step": 5966 + }, + { + "epoch": 0.22, + "learning_rate": 1.783315633354085e-07, + "loss": 0.7088, + "step": 5967 + }, + { + "epoch": 0.22, + "learning_rate": 1.7832790127073642e-07, + "loss": 0.6975, + "step": 5968 + }, + { + "epoch": 0.22, + "learning_rate": 1.7832423920606437e-07, + "loss": 0.7127, + "step": 5969 + }, + { + "epoch": 0.22, + "learning_rate": 1.783205771413923e-07, + "loss": 0.6899, + "step": 5970 + }, + { + "epoch": 0.22, + "learning_rate": 1.7831691507672025e-07, + "loss": 0.6946, + "step": 5971 + }, + { + "epoch": 0.22, + "learning_rate": 1.7831325301204818e-07, + "loss": 0.7051, + "step": 5972 + }, + { + "epoch": 0.22, + "learning_rate": 1.7830959094737613e-07, + "loss": 0.7004, + "step": 5973 + }, + { + "epoch": 0.22, + "learning_rate": 1.7830592888270406e-07, + "loss": 0.6986, + "step": 5974 + }, + { + "epoch": 0.22, + "learning_rate": 1.78302266818032e-07, + "loss": 0.7061, + "step": 5975 + }, + { + "epoch": 0.22, + "learning_rate": 1.7829860475335994e-07, + "loss": 0.6926, + "step": 5976 + }, + { + "epoch": 0.22, + "learning_rate": 1.7829494268868786e-07, + "loss": 0.6725, + "step": 5977 + }, + { + "epoch": 0.22, + "learning_rate": 1.7829128062401582e-07, + "loss": 0.6845, + "step": 5978 + }, + { + "epoch": 0.22, + "learning_rate": 1.7828761855934374e-07, + "loss": 0.6813, + "step": 5979 + }, + { + "epoch": 0.22, + "learning_rate": 1.782839564946717e-07, + "loss": 0.6823, + "step": 5980 + }, + { + "epoch": 0.22, + "learning_rate": 1.7828029442999962e-07, + "loss": 0.7074, + "step": 5981 + }, + { + "epoch": 0.22, + "learning_rate": 1.7827663236532758e-07, + "loss": 0.6745, + "step": 5982 + }, + { + "epoch": 0.22, + "learning_rate": 1.782729703006555e-07, + "loss": 0.698, + "step": 5983 + }, + { + "epoch": 0.22, + "learning_rate": 1.7826930823598343e-07, + "loss": 0.6979, + "step": 5984 + }, + { + "epoch": 0.22, + "learning_rate": 1.7826564617131136e-07, + "loss": 0.6988, + "step": 5985 + }, + { + "epoch": 0.22, + "learning_rate": 1.782619841066393e-07, + "loss": 0.6953, + "step": 5986 + }, + { + "epoch": 0.22, + "learning_rate": 1.7825832204196724e-07, + "loss": 0.7157, + "step": 5987 + }, + { + "epoch": 0.22, + "learning_rate": 1.782546599772952e-07, + "loss": 0.7077, + "step": 5988 + }, + { + "epoch": 0.22, + "learning_rate": 1.7825099791262312e-07, + "loss": 0.7001, + "step": 5989 + }, + { + "epoch": 0.22, + "learning_rate": 1.7824733584795107e-07, + "loss": 0.6931, + "step": 5990 + }, + { + "epoch": 0.22, + "learning_rate": 1.78243673783279e-07, + "loss": 0.6925, + "step": 5991 + }, + { + "epoch": 0.22, + "learning_rate": 1.7824001171860695e-07, + "loss": 0.6861, + "step": 5992 + }, + { + "epoch": 0.22, + "learning_rate": 1.7823634965393488e-07, + "loss": 0.7036, + "step": 5993 + }, + { + "epoch": 0.22, + "learning_rate": 1.7823268758926283e-07, + "loss": 0.6964, + "step": 5994 + }, + { + "epoch": 0.22, + "learning_rate": 1.7822902552459076e-07, + "loss": 0.7023, + "step": 5995 + }, + { + "epoch": 0.22, + "learning_rate": 1.782253634599187e-07, + "loss": 0.7034, + "step": 5996 + }, + { + "epoch": 0.22, + "learning_rate": 1.7822170139524664e-07, + "loss": 0.6887, + "step": 5997 + }, + { + "epoch": 0.22, + "learning_rate": 1.782180393305746e-07, + "loss": 0.6886, + "step": 5998 + }, + { + "epoch": 0.22, + "learning_rate": 1.7821437726590252e-07, + "loss": 0.6949, + "step": 5999 + }, + { + "epoch": 0.22, + "learning_rate": 1.7821071520123044e-07, + "loss": 0.6665, + "step": 6000 + }, + { + "epoch": 0.22, + "learning_rate": 1.782070531365584e-07, + "loss": 0.686, + "step": 6001 + }, + { + "epoch": 0.22, + "learning_rate": 1.7820339107188632e-07, + "loss": 0.6964, + "step": 6002 + }, + { + "epoch": 0.22, + "learning_rate": 1.7819972900721425e-07, + "loss": 0.6728, + "step": 6003 + }, + { + "epoch": 0.22, + "learning_rate": 1.7819606694254218e-07, + "loss": 0.6999, + "step": 6004 + }, + { + "epoch": 0.22, + "learning_rate": 1.7819240487787013e-07, + "loss": 0.6868, + "step": 6005 + }, + { + "epoch": 0.22, + "learning_rate": 1.7818874281319806e-07, + "loss": 0.7021, + "step": 6006 + }, + { + "epoch": 0.22, + "learning_rate": 1.78185080748526e-07, + "loss": 0.6832, + "step": 6007 + }, + { + "epoch": 0.22, + "learning_rate": 1.7818141868385394e-07, + "loss": 0.6922, + "step": 6008 + }, + { + "epoch": 0.22, + "learning_rate": 1.781777566191819e-07, + "loss": 0.6866, + "step": 6009 + }, + { + "epoch": 0.22, + "learning_rate": 1.7817409455450982e-07, + "loss": 0.6837, + "step": 6010 + }, + { + "epoch": 0.22, + "learning_rate": 1.7817043248983777e-07, + "loss": 0.6827, + "step": 6011 + }, + { + "epoch": 0.22, + "learning_rate": 1.781667704251657e-07, + "loss": 0.6898, + "step": 6012 + }, + { + "epoch": 0.22, + "learning_rate": 1.7816310836049365e-07, + "loss": 0.7208, + "step": 6013 + }, + { + "epoch": 0.22, + "learning_rate": 1.7815944629582158e-07, + "loss": 0.7026, + "step": 6014 + }, + { + "epoch": 0.22, + "learning_rate": 1.7815578423114953e-07, + "loss": 0.6921, + "step": 6015 + }, + { + "epoch": 0.22, + "learning_rate": 1.7815212216647746e-07, + "loss": 0.6816, + "step": 6016 + }, + { + "epoch": 0.22, + "learning_rate": 1.781484601018054e-07, + "loss": 0.6788, + "step": 6017 + }, + { + "epoch": 0.22, + "learning_rate": 1.7814479803713334e-07, + "loss": 0.6898, + "step": 6018 + }, + { + "epoch": 0.22, + "learning_rate": 1.781411359724613e-07, + "loss": 0.6959, + "step": 6019 + }, + { + "epoch": 0.22, + "learning_rate": 1.781374739077892e-07, + "loss": 0.6997, + "step": 6020 + }, + { + "epoch": 0.22, + "learning_rate": 1.7813381184311714e-07, + "loss": 0.6914, + "step": 6021 + }, + { + "epoch": 0.22, + "learning_rate": 1.7813014977844507e-07, + "loss": 0.6878, + "step": 6022 + }, + { + "epoch": 0.22, + "learning_rate": 1.7812648771377302e-07, + "loss": 0.6859, + "step": 6023 + }, + { + "epoch": 0.22, + "learning_rate": 1.7812282564910095e-07, + "loss": 0.7029, + "step": 6024 + }, + { + "epoch": 0.22, + "learning_rate": 1.781191635844289e-07, + "loss": 0.6878, + "step": 6025 + }, + { + "epoch": 0.22, + "learning_rate": 1.7811550151975683e-07, + "loss": 0.6855, + "step": 6026 + }, + { + "epoch": 0.22, + "learning_rate": 1.7811183945508475e-07, + "loss": 0.6929, + "step": 6027 + }, + { + "epoch": 0.22, + "learning_rate": 1.781081773904127e-07, + "loss": 0.6949, + "step": 6028 + }, + { + "epoch": 0.22, + "learning_rate": 1.7810451532574063e-07, + "loss": 0.6793, + "step": 6029 + }, + { + "epoch": 0.22, + "learning_rate": 1.781008532610686e-07, + "loss": 0.6899, + "step": 6030 + }, + { + "epoch": 0.22, + "learning_rate": 1.7809719119639651e-07, + "loss": 0.6658, + "step": 6031 + }, + { + "epoch": 0.22, + "learning_rate": 1.7809352913172447e-07, + "loss": 0.6925, + "step": 6032 + }, + { + "epoch": 0.22, + "learning_rate": 1.780898670670524e-07, + "loss": 0.6931, + "step": 6033 + }, + { + "epoch": 0.22, + "learning_rate": 1.7808620500238035e-07, + "loss": 0.7153, + "step": 6034 + }, + { + "epoch": 0.22, + "learning_rate": 1.7808254293770827e-07, + "loss": 0.6826, + "step": 6035 + }, + { + "epoch": 0.22, + "learning_rate": 1.7807888087303623e-07, + "loss": 0.6951, + "step": 6036 + }, + { + "epoch": 0.22, + "learning_rate": 1.7807521880836415e-07, + "loss": 0.7054, + "step": 6037 + }, + { + "epoch": 0.22, + "learning_rate": 1.7807155674369208e-07, + "loss": 0.6981, + "step": 6038 + }, + { + "epoch": 0.22, + "learning_rate": 1.7806789467902e-07, + "loss": 0.6878, + "step": 6039 + }, + { + "epoch": 0.22, + "learning_rate": 1.7806423261434796e-07, + "loss": 0.7019, + "step": 6040 + }, + { + "epoch": 0.22, + "learning_rate": 1.780605705496759e-07, + "loss": 0.7007, + "step": 6041 + }, + { + "epoch": 0.22, + "learning_rate": 1.7805690848500384e-07, + "loss": 0.6871, + "step": 6042 + }, + { + "epoch": 0.22, + "learning_rate": 1.7805324642033177e-07, + "loss": 0.6971, + "step": 6043 + }, + { + "epoch": 0.22, + "learning_rate": 1.7804958435565972e-07, + "loss": 0.6813, + "step": 6044 + }, + { + "epoch": 0.22, + "learning_rate": 1.7804592229098765e-07, + "loss": 0.6911, + "step": 6045 + }, + { + "epoch": 0.22, + "learning_rate": 1.780422602263156e-07, + "loss": 0.659, + "step": 6046 + }, + { + "epoch": 0.22, + "learning_rate": 1.7803859816164353e-07, + "loss": 0.6995, + "step": 6047 + }, + { + "epoch": 0.22, + "learning_rate": 1.7803493609697145e-07, + "loss": 0.6879, + "step": 6048 + }, + { + "epoch": 0.22, + "learning_rate": 1.780312740322994e-07, + "loss": 0.6873, + "step": 6049 + }, + { + "epoch": 0.22, + "learning_rate": 1.7802761196762733e-07, + "loss": 0.691, + "step": 6050 + }, + { + "epoch": 0.22, + "learning_rate": 1.780239499029553e-07, + "loss": 0.7056, + "step": 6051 + }, + { + "epoch": 0.22, + "learning_rate": 1.7802028783828321e-07, + "loss": 0.6991, + "step": 6052 + }, + { + "epoch": 0.22, + "learning_rate": 1.7801662577361117e-07, + "loss": 0.7016, + "step": 6053 + }, + { + "epoch": 0.22, + "learning_rate": 1.780129637089391e-07, + "loss": 0.699, + "step": 6054 + }, + { + "epoch": 0.22, + "learning_rate": 1.7800930164426705e-07, + "loss": 0.6957, + "step": 6055 + }, + { + "epoch": 0.22, + "learning_rate": 1.7800563957959495e-07, + "loss": 0.6985, + "step": 6056 + }, + { + "epoch": 0.22, + "learning_rate": 1.780019775149229e-07, + "loss": 0.6938, + "step": 6057 + }, + { + "epoch": 0.22, + "learning_rate": 1.7799831545025083e-07, + "loss": 0.6831, + "step": 6058 + }, + { + "epoch": 0.22, + "learning_rate": 1.7799465338557878e-07, + "loss": 0.6792, + "step": 6059 + }, + { + "epoch": 0.22, + "learning_rate": 1.779909913209067e-07, + "loss": 0.7077, + "step": 6060 + }, + { + "epoch": 0.22, + "learning_rate": 1.7798732925623466e-07, + "loss": 0.6999, + "step": 6061 + }, + { + "epoch": 0.22, + "learning_rate": 1.7798366719156259e-07, + "loss": 0.6812, + "step": 6062 + }, + { + "epoch": 0.22, + "learning_rate": 1.7798000512689054e-07, + "loss": 0.7045, + "step": 6063 + }, + { + "epoch": 0.22, + "learning_rate": 1.7797634306221847e-07, + "loss": 0.7017, + "step": 6064 + }, + { + "epoch": 0.22, + "learning_rate": 1.7797268099754642e-07, + "loss": 0.6937, + "step": 6065 + }, + { + "epoch": 0.22, + "learning_rate": 1.7796901893287435e-07, + "loss": 0.6887, + "step": 6066 + }, + { + "epoch": 0.22, + "learning_rate": 1.779653568682023e-07, + "loss": 0.7075, + "step": 6067 + }, + { + "epoch": 0.22, + "learning_rate": 1.7796169480353023e-07, + "loss": 0.7077, + "step": 6068 + }, + { + "epoch": 0.22, + "learning_rate": 1.7795803273885818e-07, + "loss": 0.6935, + "step": 6069 + }, + { + "epoch": 0.22, + "learning_rate": 1.779543706741861e-07, + "loss": 0.6857, + "step": 6070 + }, + { + "epoch": 0.22, + "learning_rate": 1.7795070860951403e-07, + "loss": 0.7074, + "step": 6071 + }, + { + "epoch": 0.22, + "learning_rate": 1.7794704654484199e-07, + "loss": 0.6957, + "step": 6072 + }, + { + "epoch": 0.22, + "learning_rate": 1.779433844801699e-07, + "loss": 0.6745, + "step": 6073 + }, + { + "epoch": 0.22, + "learning_rate": 1.7793972241549784e-07, + "loss": 0.7251, + "step": 6074 + }, + { + "epoch": 0.22, + "learning_rate": 1.7793606035082577e-07, + "loss": 0.7049, + "step": 6075 + }, + { + "epoch": 0.22, + "learning_rate": 1.7793239828615372e-07, + "loss": 0.6963, + "step": 6076 + }, + { + "epoch": 0.22, + "learning_rate": 1.7792873622148165e-07, + "loss": 0.7037, + "step": 6077 + }, + { + "epoch": 0.22, + "learning_rate": 1.779250741568096e-07, + "loss": 0.6964, + "step": 6078 + }, + { + "epoch": 0.22, + "learning_rate": 1.7792141209213753e-07, + "loss": 0.6859, + "step": 6079 + }, + { + "epoch": 0.22, + "learning_rate": 1.7791775002746548e-07, + "loss": 0.6908, + "step": 6080 + }, + { + "epoch": 0.22, + "learning_rate": 1.779140879627934e-07, + "loss": 0.6921, + "step": 6081 + }, + { + "epoch": 0.22, + "learning_rate": 1.7791042589812136e-07, + "loss": 0.6806, + "step": 6082 + }, + { + "epoch": 0.22, + "learning_rate": 1.7790676383344929e-07, + "loss": 0.6826, + "step": 6083 + }, + { + "epoch": 0.22, + "learning_rate": 1.7790310176877724e-07, + "loss": 0.7011, + "step": 6084 + }, + { + "epoch": 0.22, + "learning_rate": 1.7789943970410516e-07, + "loss": 0.6905, + "step": 6085 + }, + { + "epoch": 0.22, + "learning_rate": 1.7789577763943312e-07, + "loss": 0.6885, + "step": 6086 + }, + { + "epoch": 0.22, + "learning_rate": 1.7789211557476104e-07, + "loss": 0.703, + "step": 6087 + }, + { + "epoch": 0.22, + "learning_rate": 1.77888453510089e-07, + "loss": 0.6924, + "step": 6088 + }, + { + "epoch": 0.22, + "learning_rate": 1.7788479144541692e-07, + "loss": 0.7025, + "step": 6089 + }, + { + "epoch": 0.22, + "learning_rate": 1.7788112938074488e-07, + "loss": 0.6926, + "step": 6090 + }, + { + "epoch": 0.22, + "learning_rate": 1.778774673160728e-07, + "loss": 0.7118, + "step": 6091 + }, + { + "epoch": 0.22, + "learning_rate": 1.7787380525140073e-07, + "loss": 0.6875, + "step": 6092 + }, + { + "epoch": 0.22, + "learning_rate": 1.7787014318672866e-07, + "loss": 0.6976, + "step": 6093 + }, + { + "epoch": 0.22, + "learning_rate": 1.778664811220566e-07, + "loss": 0.688, + "step": 6094 + }, + { + "epoch": 0.22, + "learning_rate": 1.7786281905738454e-07, + "loss": 0.6917, + "step": 6095 + }, + { + "epoch": 0.22, + "learning_rate": 1.778591569927125e-07, + "loss": 0.6883, + "step": 6096 + }, + { + "epoch": 0.22, + "learning_rate": 1.7785549492804042e-07, + "loss": 0.6917, + "step": 6097 + }, + { + "epoch": 0.22, + "learning_rate": 1.7785183286336834e-07, + "loss": 0.6933, + "step": 6098 + }, + { + "epoch": 0.22, + "learning_rate": 1.778481707986963e-07, + "loss": 0.6979, + "step": 6099 + }, + { + "epoch": 0.22, + "learning_rate": 1.7784450873402422e-07, + "loss": 0.6809, + "step": 6100 + }, + { + "epoch": 0.22, + "learning_rate": 1.7784084666935218e-07, + "loss": 0.6984, + "step": 6101 + }, + { + "epoch": 0.22, + "learning_rate": 1.778371846046801e-07, + "loss": 0.6907, + "step": 6102 + }, + { + "epoch": 0.22, + "learning_rate": 1.7783352254000806e-07, + "loss": 0.7046, + "step": 6103 + }, + { + "epoch": 0.22, + "learning_rate": 1.7782986047533598e-07, + "loss": 0.6843, + "step": 6104 + }, + { + "epoch": 0.22, + "learning_rate": 1.7782619841066394e-07, + "loss": 0.684, + "step": 6105 + }, + { + "epoch": 0.22, + "learning_rate": 1.7782253634599186e-07, + "loss": 0.7074, + "step": 6106 + }, + { + "epoch": 0.22, + "learning_rate": 1.7781887428131982e-07, + "loss": 0.6833, + "step": 6107 + }, + { + "epoch": 0.22, + "learning_rate": 1.7781521221664774e-07, + "loss": 0.6884, + "step": 6108 + }, + { + "epoch": 0.22, + "learning_rate": 1.778115501519757e-07, + "loss": 0.6994, + "step": 6109 + }, + { + "epoch": 0.22, + "learning_rate": 1.778078880873036e-07, + "loss": 0.707, + "step": 6110 + }, + { + "epoch": 0.22, + "learning_rate": 1.7780422602263155e-07, + "loss": 0.7035, + "step": 6111 + }, + { + "epoch": 0.22, + "learning_rate": 1.7780056395795948e-07, + "loss": 0.6927, + "step": 6112 + }, + { + "epoch": 0.22, + "learning_rate": 1.7779690189328743e-07, + "loss": 0.703, + "step": 6113 + }, + { + "epoch": 0.22, + "learning_rate": 1.7779323982861536e-07, + "loss": 0.6821, + "step": 6114 + }, + { + "epoch": 0.22, + "learning_rate": 1.777895777639433e-07, + "loss": 0.6886, + "step": 6115 + }, + { + "epoch": 0.22, + "learning_rate": 1.7778591569927124e-07, + "loss": 0.6807, + "step": 6116 + }, + { + "epoch": 0.22, + "learning_rate": 1.777822536345992e-07, + "loss": 0.6962, + "step": 6117 + }, + { + "epoch": 0.22, + "learning_rate": 1.7777859156992712e-07, + "loss": 0.6849, + "step": 6118 + }, + { + "epoch": 0.22, + "learning_rate": 1.7777492950525504e-07, + "loss": 0.6873, + "step": 6119 + }, + { + "epoch": 0.22, + "learning_rate": 1.77771267440583e-07, + "loss": 0.7115, + "step": 6120 + }, + { + "epoch": 0.22, + "learning_rate": 1.7776760537591092e-07, + "loss": 0.6996, + "step": 6121 + }, + { + "epoch": 0.22, + "learning_rate": 1.7776394331123888e-07, + "loss": 0.7114, + "step": 6122 + }, + { + "epoch": 0.22, + "learning_rate": 1.777602812465668e-07, + "loss": 0.6911, + "step": 6123 + }, + { + "epoch": 0.22, + "learning_rate": 1.7775661918189476e-07, + "loss": 0.6812, + "step": 6124 + }, + { + "epoch": 0.22, + "learning_rate": 1.7775295711722268e-07, + "loss": 0.6866, + "step": 6125 + }, + { + "epoch": 0.22, + "learning_rate": 1.7774929505255064e-07, + "loss": 0.6935, + "step": 6126 + }, + { + "epoch": 0.22, + "learning_rate": 1.7774563298787856e-07, + "loss": 0.7076, + "step": 6127 + }, + { + "epoch": 0.22, + "learning_rate": 1.777419709232065e-07, + "loss": 0.7011, + "step": 6128 + }, + { + "epoch": 0.22, + "learning_rate": 1.7773830885853442e-07, + "loss": 0.6947, + "step": 6129 + }, + { + "epoch": 0.22, + "learning_rate": 1.7773464679386237e-07, + "loss": 0.6927, + "step": 6130 + }, + { + "epoch": 0.22, + "learning_rate": 1.777309847291903e-07, + "loss": 0.6859, + "step": 6131 + }, + { + "epoch": 0.22, + "learning_rate": 1.7772732266451825e-07, + "loss": 0.6905, + "step": 6132 + }, + { + "epoch": 0.22, + "learning_rate": 1.7772366059984618e-07, + "loss": 0.6938, + "step": 6133 + }, + { + "epoch": 0.22, + "learning_rate": 1.7771999853517413e-07, + "loss": 0.699, + "step": 6134 + }, + { + "epoch": 0.22, + "learning_rate": 1.7771633647050206e-07, + "loss": 0.6802, + "step": 6135 + }, + { + "epoch": 0.22, + "learning_rate": 1.7771267440583e-07, + "loss": 0.684, + "step": 6136 + }, + { + "epoch": 0.22, + "learning_rate": 1.7770901234115794e-07, + "loss": 0.7148, + "step": 6137 + }, + { + "epoch": 0.22, + "learning_rate": 1.777053502764859e-07, + "loss": 0.6993, + "step": 6138 + }, + { + "epoch": 0.22, + "learning_rate": 1.7770168821181382e-07, + "loss": 0.6935, + "step": 6139 + }, + { + "epoch": 0.22, + "learning_rate": 1.7769802614714177e-07, + "loss": 0.6999, + "step": 6140 + }, + { + "epoch": 0.22, + "learning_rate": 1.776943640824697e-07, + "loss": 0.7139, + "step": 6141 + }, + { + "epoch": 0.22, + "learning_rate": 1.7769070201779762e-07, + "loss": 0.6926, + "step": 6142 + }, + { + "epoch": 0.22, + "learning_rate": 1.7768703995312557e-07, + "loss": 0.6941, + "step": 6143 + }, + { + "epoch": 0.22, + "learning_rate": 1.776833778884535e-07, + "loss": 0.6896, + "step": 6144 + }, + { + "epoch": 0.22, + "learning_rate": 1.7767971582378145e-07, + "loss": 0.6871, + "step": 6145 + }, + { + "epoch": 0.22, + "learning_rate": 1.7767605375910936e-07, + "loss": 0.6937, + "step": 6146 + }, + { + "epoch": 0.22, + "learning_rate": 1.776723916944373e-07, + "loss": 0.6827, + "step": 6147 + }, + { + "epoch": 0.22, + "learning_rate": 1.7766872962976523e-07, + "loss": 0.6906, + "step": 6148 + }, + { + "epoch": 0.22, + "learning_rate": 1.776650675650932e-07, + "loss": 0.6917, + "step": 6149 + }, + { + "epoch": 0.23, + "learning_rate": 1.7766140550042111e-07, + "loss": 0.6788, + "step": 6150 + }, + { + "epoch": 0.23, + "learning_rate": 1.7765774343574907e-07, + "loss": 0.7055, + "step": 6151 + }, + { + "epoch": 0.23, + "learning_rate": 1.77654081371077e-07, + "loss": 0.6884, + "step": 6152 + }, + { + "epoch": 0.23, + "learning_rate": 1.7765041930640495e-07, + "loss": 0.6941, + "step": 6153 + }, + { + "epoch": 0.23, + "learning_rate": 1.7764675724173287e-07, + "loss": 0.6948, + "step": 6154 + }, + { + "epoch": 0.23, + "learning_rate": 1.7764309517706083e-07, + "loss": 0.7058, + "step": 6155 + }, + { + "epoch": 0.23, + "learning_rate": 1.7763943311238875e-07, + "loss": 0.694, + "step": 6156 + }, + { + "epoch": 0.23, + "learning_rate": 1.776357710477167e-07, + "loss": 0.6985, + "step": 6157 + }, + { + "epoch": 0.23, + "learning_rate": 1.7763210898304463e-07, + "loss": 0.7063, + "step": 6158 + }, + { + "epoch": 0.23, + "learning_rate": 1.776284469183726e-07, + "loss": 0.7063, + "step": 6159 + }, + { + "epoch": 0.23, + "learning_rate": 1.7762478485370051e-07, + "loss": 0.6875, + "step": 6160 + }, + { + "epoch": 0.23, + "learning_rate": 1.7762112278902847e-07, + "loss": 0.6999, + "step": 6161 + }, + { + "epoch": 0.23, + "learning_rate": 1.776174607243564e-07, + "loss": 0.6783, + "step": 6162 + }, + { + "epoch": 0.23, + "learning_rate": 1.7761379865968432e-07, + "loss": 0.6992, + "step": 6163 + }, + { + "epoch": 0.23, + "learning_rate": 1.7761013659501225e-07, + "loss": 0.6951, + "step": 6164 + }, + { + "epoch": 0.23, + "learning_rate": 1.776064745303402e-07, + "loss": 0.7074, + "step": 6165 + }, + { + "epoch": 0.23, + "learning_rate": 1.7760281246566813e-07, + "loss": 0.6874, + "step": 6166 + }, + { + "epoch": 0.23, + "learning_rate": 1.7759915040099608e-07, + "loss": 0.7138, + "step": 6167 + }, + { + "epoch": 0.23, + "learning_rate": 1.77595488336324e-07, + "loss": 0.6885, + "step": 6168 + }, + { + "epoch": 0.23, + "learning_rate": 1.7759182627165193e-07, + "loss": 0.7067, + "step": 6169 + }, + { + "epoch": 0.23, + "learning_rate": 1.775881642069799e-07, + "loss": 0.6939, + "step": 6170 + }, + { + "epoch": 0.23, + "learning_rate": 1.7758450214230781e-07, + "loss": 0.7075, + "step": 6171 + }, + { + "epoch": 0.23, + "learning_rate": 1.7758084007763577e-07, + "loss": 0.6851, + "step": 6172 + }, + { + "epoch": 0.23, + "learning_rate": 1.775771780129637e-07, + "loss": 0.7035, + "step": 6173 + }, + { + "epoch": 0.23, + "learning_rate": 1.7757351594829165e-07, + "loss": 0.711, + "step": 6174 + }, + { + "epoch": 0.23, + "learning_rate": 1.7756985388361957e-07, + "loss": 0.7108, + "step": 6175 + }, + { + "epoch": 0.23, + "learning_rate": 1.7756619181894753e-07, + "loss": 0.7068, + "step": 6176 + }, + { + "epoch": 0.23, + "learning_rate": 1.7756252975427545e-07, + "loss": 0.6862, + "step": 6177 + }, + { + "epoch": 0.23, + "learning_rate": 1.775588676896034e-07, + "loss": 0.6868, + "step": 6178 + }, + { + "epoch": 0.23, + "learning_rate": 1.7755520562493133e-07, + "loss": 0.6983, + "step": 6179 + }, + { + "epoch": 0.23, + "learning_rate": 1.7755154356025929e-07, + "loss": 0.7119, + "step": 6180 + }, + { + "epoch": 0.23, + "learning_rate": 1.775478814955872e-07, + "loss": 0.6874, + "step": 6181 + }, + { + "epoch": 0.23, + "learning_rate": 1.7754421943091514e-07, + "loss": 0.694, + "step": 6182 + }, + { + "epoch": 0.23, + "learning_rate": 1.7754055736624307e-07, + "loss": 0.7022, + "step": 6183 + }, + { + "epoch": 0.23, + "learning_rate": 1.7753689530157102e-07, + "loss": 0.7073, + "step": 6184 + }, + { + "epoch": 0.23, + "learning_rate": 1.7753323323689895e-07, + "loss": 0.6893, + "step": 6185 + }, + { + "epoch": 0.23, + "learning_rate": 1.775295711722269e-07, + "loss": 0.7096, + "step": 6186 + }, + { + "epoch": 0.23, + "learning_rate": 1.7752590910755483e-07, + "loss": 0.6909, + "step": 6187 + }, + { + "epoch": 0.23, + "learning_rate": 1.7752224704288278e-07, + "loss": 0.7017, + "step": 6188 + }, + { + "epoch": 0.23, + "learning_rate": 1.775185849782107e-07, + "loss": 0.6918, + "step": 6189 + }, + { + "epoch": 0.23, + "learning_rate": 1.7751492291353863e-07, + "loss": 0.6802, + "step": 6190 + }, + { + "epoch": 0.23, + "learning_rate": 1.7751126084886659e-07, + "loss": 0.6964, + "step": 6191 + }, + { + "epoch": 0.23, + "learning_rate": 1.775075987841945e-07, + "loss": 0.7057, + "step": 6192 + }, + { + "epoch": 0.23, + "learning_rate": 1.7750393671952247e-07, + "loss": 0.6938, + "step": 6193 + }, + { + "epoch": 0.23, + "learning_rate": 1.775002746548504e-07, + "loss": 0.6876, + "step": 6194 + }, + { + "epoch": 0.23, + "learning_rate": 1.7749661259017835e-07, + "loss": 0.7042, + "step": 6195 + }, + { + "epoch": 0.23, + "learning_rate": 1.7749295052550627e-07, + "loss": 0.6891, + "step": 6196 + }, + { + "epoch": 0.23, + "learning_rate": 1.7748928846083423e-07, + "loss": 0.7019, + "step": 6197 + }, + { + "epoch": 0.23, + "learning_rate": 1.7748562639616215e-07, + "loss": 0.6899, + "step": 6198 + }, + { + "epoch": 0.23, + "learning_rate": 1.774819643314901e-07, + "loss": 0.6907, + "step": 6199 + }, + { + "epoch": 0.23, + "learning_rate": 1.77478302266818e-07, + "loss": 0.6698, + "step": 6200 + }, + { + "epoch": 0.23, + "learning_rate": 1.7747464020214596e-07, + "loss": 0.6949, + "step": 6201 + }, + { + "epoch": 0.23, + "learning_rate": 1.7747097813747389e-07, + "loss": 0.6908, + "step": 6202 + }, + { + "epoch": 0.23, + "learning_rate": 1.7746731607280184e-07, + "loss": 0.7019, + "step": 6203 + }, + { + "epoch": 0.23, + "learning_rate": 1.7746365400812976e-07, + "loss": 0.6965, + "step": 6204 + }, + { + "epoch": 0.23, + "learning_rate": 1.7745999194345772e-07, + "loss": 0.6996, + "step": 6205 + }, + { + "epoch": 0.23, + "learning_rate": 1.7745632987878564e-07, + "loss": 0.7057, + "step": 6206 + }, + { + "epoch": 0.23, + "learning_rate": 1.774526678141136e-07, + "loss": 0.6776, + "step": 6207 + }, + { + "epoch": 0.23, + "learning_rate": 1.7744900574944152e-07, + "loss": 0.676, + "step": 6208 + }, + { + "epoch": 0.23, + "learning_rate": 1.7744534368476948e-07, + "loss": 0.6759, + "step": 6209 + }, + { + "epoch": 0.23, + "learning_rate": 1.774416816200974e-07, + "loss": 0.6852, + "step": 6210 + }, + { + "epoch": 0.23, + "learning_rate": 1.7743801955542536e-07, + "loss": 0.6919, + "step": 6211 + }, + { + "epoch": 0.23, + "learning_rate": 1.7743435749075328e-07, + "loss": 0.6663, + "step": 6212 + }, + { + "epoch": 0.23, + "learning_rate": 1.774306954260812e-07, + "loss": 0.7058, + "step": 6213 + }, + { + "epoch": 0.23, + "learning_rate": 1.7742703336140916e-07, + "loss": 0.6919, + "step": 6214 + }, + { + "epoch": 0.23, + "learning_rate": 1.774233712967371e-07, + "loss": 0.7006, + "step": 6215 + }, + { + "epoch": 0.23, + "learning_rate": 1.7741970923206504e-07, + "loss": 0.6853, + "step": 6216 + }, + { + "epoch": 0.23, + "learning_rate": 1.7741604716739294e-07, + "loss": 0.7002, + "step": 6217 + }, + { + "epoch": 0.23, + "learning_rate": 1.774123851027209e-07, + "loss": 0.7008, + "step": 6218 + }, + { + "epoch": 0.23, + "learning_rate": 1.7740872303804882e-07, + "loss": 0.6807, + "step": 6219 + }, + { + "epoch": 0.23, + "learning_rate": 1.7740506097337678e-07, + "loss": 0.6937, + "step": 6220 + }, + { + "epoch": 0.23, + "learning_rate": 1.774013989087047e-07, + "loss": 0.6889, + "step": 6221 + }, + { + "epoch": 0.23, + "learning_rate": 1.7739773684403266e-07, + "loss": 0.698, + "step": 6222 + }, + { + "epoch": 0.23, + "learning_rate": 1.7739407477936058e-07, + "loss": 0.693, + "step": 6223 + }, + { + "epoch": 0.23, + "learning_rate": 1.7739041271468854e-07, + "loss": 0.6905, + "step": 6224 + }, + { + "epoch": 0.23, + "learning_rate": 1.7738675065001646e-07, + "loss": 0.7023, + "step": 6225 + }, + { + "epoch": 0.23, + "learning_rate": 1.7738308858534442e-07, + "loss": 0.6817, + "step": 6226 + }, + { + "epoch": 0.23, + "learning_rate": 1.7737942652067234e-07, + "loss": 0.6815, + "step": 6227 + }, + { + "epoch": 0.23, + "learning_rate": 1.773757644560003e-07, + "loss": 0.7037, + "step": 6228 + }, + { + "epoch": 0.23, + "learning_rate": 1.7737210239132822e-07, + "loss": 0.7012, + "step": 6229 + }, + { + "epoch": 0.23, + "learning_rate": 1.7736844032665618e-07, + "loss": 0.6827, + "step": 6230 + }, + { + "epoch": 0.23, + "learning_rate": 1.773647782619841e-07, + "loss": 0.6795, + "step": 6231 + }, + { + "epoch": 0.23, + "learning_rate": 1.7736111619731206e-07, + "loss": 0.6874, + "step": 6232 + }, + { + "epoch": 0.23, + "learning_rate": 1.7735745413263998e-07, + "loss": 0.679, + "step": 6233 + }, + { + "epoch": 0.23, + "learning_rate": 1.773537920679679e-07, + "loss": 0.7006, + "step": 6234 + }, + { + "epoch": 0.23, + "learning_rate": 1.7735013000329584e-07, + "loss": 0.6767, + "step": 6235 + }, + { + "epoch": 0.23, + "learning_rate": 1.773464679386238e-07, + "loss": 0.6988, + "step": 6236 + }, + { + "epoch": 0.23, + "learning_rate": 1.7734280587395172e-07, + "loss": 0.6918, + "step": 6237 + }, + { + "epoch": 0.23, + "learning_rate": 1.7733914380927967e-07, + "loss": 0.6732, + "step": 6238 + }, + { + "epoch": 0.23, + "learning_rate": 1.773354817446076e-07, + "loss": 0.686, + "step": 6239 + }, + { + "epoch": 0.23, + "learning_rate": 1.7733181967993552e-07, + "loss": 0.691, + "step": 6240 + }, + { + "epoch": 0.23, + "learning_rate": 1.7732815761526348e-07, + "loss": 0.6987, + "step": 6241 + }, + { + "epoch": 0.23, + "learning_rate": 1.773244955505914e-07, + "loss": 0.7073, + "step": 6242 + }, + { + "epoch": 0.23, + "learning_rate": 1.7732083348591936e-07, + "loss": 0.6957, + "step": 6243 + }, + { + "epoch": 0.23, + "learning_rate": 1.7731717142124728e-07, + "loss": 0.6632, + "step": 6244 + }, + { + "epoch": 0.23, + "learning_rate": 1.7731350935657524e-07, + "loss": 0.6884, + "step": 6245 + }, + { + "epoch": 0.23, + "learning_rate": 1.7730984729190316e-07, + "loss": 0.6852, + "step": 6246 + }, + { + "epoch": 0.23, + "learning_rate": 1.7730618522723112e-07, + "loss": 0.6944, + "step": 6247 + }, + { + "epoch": 0.23, + "learning_rate": 1.7730252316255904e-07, + "loss": 0.7026, + "step": 6248 + }, + { + "epoch": 0.23, + "learning_rate": 1.77298861097887e-07, + "loss": 0.6888, + "step": 6249 + }, + { + "epoch": 0.23, + "learning_rate": 1.7729519903321492e-07, + "loss": 0.6882, + "step": 6250 + }, + { + "epoch": 0.23, + "learning_rate": 1.7729153696854288e-07, + "loss": 0.6943, + "step": 6251 + }, + { + "epoch": 0.23, + "learning_rate": 1.772878749038708e-07, + "loss": 0.7007, + "step": 6252 + }, + { + "epoch": 0.23, + "learning_rate": 1.7728421283919873e-07, + "loss": 0.7009, + "step": 6253 + }, + { + "epoch": 0.23, + "learning_rate": 1.7728055077452666e-07, + "loss": 0.7015, + "step": 6254 + }, + { + "epoch": 0.23, + "learning_rate": 1.772768887098546e-07, + "loss": 0.7016, + "step": 6255 + }, + { + "epoch": 0.23, + "learning_rate": 1.7727322664518254e-07, + "loss": 0.6947, + "step": 6256 + }, + { + "epoch": 0.23, + "learning_rate": 1.772695645805105e-07, + "loss": 0.6912, + "step": 6257 + }, + { + "epoch": 0.23, + "learning_rate": 1.7726590251583842e-07, + "loss": 0.6934, + "step": 6258 + }, + { + "epoch": 0.23, + "learning_rate": 1.7726224045116637e-07, + "loss": 0.677, + "step": 6259 + }, + { + "epoch": 0.23, + "learning_rate": 1.772585783864943e-07, + "loss": 0.6857, + "step": 6260 + }, + { + "epoch": 0.23, + "learning_rate": 1.7725491632182222e-07, + "loss": 0.6937, + "step": 6261 + }, + { + "epoch": 0.23, + "learning_rate": 1.7725125425715017e-07, + "loss": 0.6843, + "step": 6262 + }, + { + "epoch": 0.23, + "learning_rate": 1.772475921924781e-07, + "loss": 0.6836, + "step": 6263 + }, + { + "epoch": 0.23, + "learning_rate": 1.7724393012780605e-07, + "loss": 0.6773, + "step": 6264 + }, + { + "epoch": 0.23, + "learning_rate": 1.7724026806313398e-07, + "loss": 0.6662, + "step": 6265 + }, + { + "epoch": 0.23, + "learning_rate": 1.7723660599846193e-07, + "loss": 0.6951, + "step": 6266 + }, + { + "epoch": 0.23, + "learning_rate": 1.7723294393378986e-07, + "loss": 0.7064, + "step": 6267 + }, + { + "epoch": 0.23, + "learning_rate": 1.7722928186911781e-07, + "loss": 0.6922, + "step": 6268 + }, + { + "epoch": 0.23, + "learning_rate": 1.7722561980444574e-07, + "loss": 0.6868, + "step": 6269 + }, + { + "epoch": 0.23, + "learning_rate": 1.772219577397737e-07, + "loss": 0.6653, + "step": 6270 + }, + { + "epoch": 0.23, + "learning_rate": 1.772182956751016e-07, + "loss": 0.6867, + "step": 6271 + }, + { + "epoch": 0.23, + "learning_rate": 1.7721463361042955e-07, + "loss": 0.699, + "step": 6272 + }, + { + "epoch": 0.23, + "learning_rate": 1.7721097154575747e-07, + "loss": 0.7012, + "step": 6273 + }, + { + "epoch": 0.23, + "learning_rate": 1.7720730948108543e-07, + "loss": 0.6945, + "step": 6274 + }, + { + "epoch": 0.23, + "learning_rate": 1.7720364741641335e-07, + "loss": 0.6836, + "step": 6275 + }, + { + "epoch": 0.23, + "learning_rate": 1.771999853517413e-07, + "loss": 0.6929, + "step": 6276 + }, + { + "epoch": 0.23, + "learning_rate": 1.7719632328706923e-07, + "loss": 0.6901, + "step": 6277 + }, + { + "epoch": 0.23, + "learning_rate": 1.771926612223972e-07, + "loss": 0.6958, + "step": 6278 + }, + { + "epoch": 0.23, + "learning_rate": 1.7718899915772511e-07, + "loss": 0.6957, + "step": 6279 + }, + { + "epoch": 0.23, + "learning_rate": 1.7718533709305307e-07, + "loss": 0.7174, + "step": 6280 + }, + { + "epoch": 0.23, + "learning_rate": 1.77181675028381e-07, + "loss": 0.6945, + "step": 6281 + }, + { + "epoch": 0.23, + "learning_rate": 1.7717801296370892e-07, + "loss": 0.6814, + "step": 6282 + }, + { + "epoch": 0.23, + "learning_rate": 1.7717435089903687e-07, + "loss": 0.6919, + "step": 6283 + }, + { + "epoch": 0.23, + "learning_rate": 1.771706888343648e-07, + "loss": 0.6842, + "step": 6284 + }, + { + "epoch": 0.23, + "learning_rate": 1.7716702676969275e-07, + "loss": 0.6969, + "step": 6285 + }, + { + "epoch": 0.23, + "learning_rate": 1.7716336470502068e-07, + "loss": 0.7082, + "step": 6286 + }, + { + "epoch": 0.23, + "learning_rate": 1.7715970264034863e-07, + "loss": 0.6875, + "step": 6287 + }, + { + "epoch": 0.23, + "learning_rate": 1.7715604057567656e-07, + "loss": 0.6932, + "step": 6288 + }, + { + "epoch": 0.23, + "learning_rate": 1.771523785110045e-07, + "loss": 0.6994, + "step": 6289 + }, + { + "epoch": 0.23, + "learning_rate": 1.7714871644633241e-07, + "loss": 0.7015, + "step": 6290 + }, + { + "epoch": 0.23, + "learning_rate": 1.7714505438166037e-07, + "loss": 0.6859, + "step": 6291 + }, + { + "epoch": 0.23, + "learning_rate": 1.771413923169883e-07, + "loss": 0.6912, + "step": 6292 + }, + { + "epoch": 0.23, + "learning_rate": 1.7713773025231625e-07, + "loss": 0.6747, + "step": 6293 + }, + { + "epoch": 0.23, + "learning_rate": 1.7713406818764417e-07, + "loss": 0.6971, + "step": 6294 + }, + { + "epoch": 0.23, + "learning_rate": 1.7713040612297213e-07, + "loss": 0.6862, + "step": 6295 + }, + { + "epoch": 0.23, + "learning_rate": 1.7712674405830005e-07, + "loss": 0.6921, + "step": 6296 + }, + { + "epoch": 0.23, + "learning_rate": 1.77123081993628e-07, + "loss": 0.6896, + "step": 6297 + }, + { + "epoch": 0.23, + "learning_rate": 1.7711941992895593e-07, + "loss": 0.6892, + "step": 6298 + }, + { + "epoch": 0.23, + "learning_rate": 1.7711575786428389e-07, + "loss": 0.6787, + "step": 6299 + }, + { + "epoch": 0.23, + "learning_rate": 1.771120957996118e-07, + "loss": 0.6983, + "step": 6300 + }, + { + "epoch": 0.23, + "learning_rate": 1.7710843373493977e-07, + "loss": 0.7115, + "step": 6301 + }, + { + "epoch": 0.23, + "learning_rate": 1.771047716702677e-07, + "loss": 0.6778, + "step": 6302 + }, + { + "epoch": 0.23, + "learning_rate": 1.7710110960559565e-07, + "loss": 0.6809, + "step": 6303 + }, + { + "epoch": 0.23, + "learning_rate": 1.7709744754092357e-07, + "loss": 0.6808, + "step": 6304 + }, + { + "epoch": 0.23, + "learning_rate": 1.770937854762515e-07, + "loss": 0.6851, + "step": 6305 + }, + { + "epoch": 0.23, + "learning_rate": 1.7709012341157945e-07, + "loss": 0.6822, + "step": 6306 + }, + { + "epoch": 0.23, + "learning_rate": 1.7708646134690738e-07, + "loss": 0.7079, + "step": 6307 + }, + { + "epoch": 0.23, + "learning_rate": 1.770827992822353e-07, + "loss": 0.6962, + "step": 6308 + }, + { + "epoch": 0.23, + "learning_rate": 1.7707913721756326e-07, + "loss": 0.7153, + "step": 6309 + }, + { + "epoch": 0.23, + "learning_rate": 1.7707547515289119e-07, + "loss": 0.7107, + "step": 6310 + }, + { + "epoch": 0.23, + "learning_rate": 1.770718130882191e-07, + "loss": 0.6923, + "step": 6311 + }, + { + "epoch": 0.23, + "learning_rate": 1.7706815102354707e-07, + "loss": 0.7267, + "step": 6312 + }, + { + "epoch": 0.23, + "learning_rate": 1.77064488958875e-07, + "loss": 0.6985, + "step": 6313 + }, + { + "epoch": 0.23, + "learning_rate": 1.7706082689420295e-07, + "loss": 0.7001, + "step": 6314 + }, + { + "epoch": 0.23, + "learning_rate": 1.7705716482953087e-07, + "loss": 0.693, + "step": 6315 + }, + { + "epoch": 0.23, + "learning_rate": 1.7705350276485883e-07, + "loss": 0.707, + "step": 6316 + }, + { + "epoch": 0.23, + "learning_rate": 1.7704984070018675e-07, + "loss": 0.697, + "step": 6317 + }, + { + "epoch": 0.23, + "learning_rate": 1.770461786355147e-07, + "loss": 0.7087, + "step": 6318 + }, + { + "epoch": 0.23, + "learning_rate": 1.7704251657084263e-07, + "loss": 0.7077, + "step": 6319 + }, + { + "epoch": 0.23, + "learning_rate": 1.7703885450617058e-07, + "loss": 0.689, + "step": 6320 + }, + { + "epoch": 0.23, + "learning_rate": 1.770351924414985e-07, + "loss": 0.701, + "step": 6321 + }, + { + "epoch": 0.23, + "learning_rate": 1.7703153037682646e-07, + "loss": 0.6864, + "step": 6322 + }, + { + "epoch": 0.23, + "learning_rate": 1.770278683121544e-07, + "loss": 0.6829, + "step": 6323 + }, + { + "epoch": 0.23, + "learning_rate": 1.7702420624748234e-07, + "loss": 0.6928, + "step": 6324 + }, + { + "epoch": 0.23, + "learning_rate": 1.7702054418281024e-07, + "loss": 0.7082, + "step": 6325 + }, + { + "epoch": 0.23, + "learning_rate": 1.770168821181382e-07, + "loss": 0.6877, + "step": 6326 + }, + { + "epoch": 0.23, + "learning_rate": 1.7701322005346612e-07, + "loss": 0.7051, + "step": 6327 + }, + { + "epoch": 0.23, + "learning_rate": 1.7700955798879408e-07, + "loss": 0.7015, + "step": 6328 + }, + { + "epoch": 0.23, + "learning_rate": 1.77005895924122e-07, + "loss": 0.6962, + "step": 6329 + }, + { + "epoch": 0.23, + "learning_rate": 1.7700223385944996e-07, + "loss": 0.698, + "step": 6330 + }, + { + "epoch": 0.23, + "learning_rate": 1.7699857179477788e-07, + "loss": 0.7091, + "step": 6331 + }, + { + "epoch": 0.23, + "learning_rate": 1.769949097301058e-07, + "loss": 0.6904, + "step": 6332 + }, + { + "epoch": 0.23, + "learning_rate": 1.7699124766543376e-07, + "loss": 0.6765, + "step": 6333 + }, + { + "epoch": 0.23, + "learning_rate": 1.769875856007617e-07, + "loss": 0.6909, + "step": 6334 + }, + { + "epoch": 0.23, + "learning_rate": 1.7698392353608964e-07, + "loss": 0.6918, + "step": 6335 + }, + { + "epoch": 0.23, + "learning_rate": 1.7698026147141757e-07, + "loss": 0.7154, + "step": 6336 + }, + { + "epoch": 0.23, + "learning_rate": 1.7697659940674552e-07, + "loss": 0.6959, + "step": 6337 + }, + { + "epoch": 0.23, + "learning_rate": 1.7697293734207345e-07, + "loss": 0.7009, + "step": 6338 + }, + { + "epoch": 0.23, + "learning_rate": 1.769692752774014e-07, + "loss": 0.7005, + "step": 6339 + }, + { + "epoch": 0.23, + "learning_rate": 1.7696561321272933e-07, + "loss": 0.6979, + "step": 6340 + }, + { + "epoch": 0.23, + "learning_rate": 1.7696195114805728e-07, + "loss": 0.7115, + "step": 6341 + }, + { + "epoch": 0.23, + "learning_rate": 1.769582890833852e-07, + "loss": 0.6885, + "step": 6342 + }, + { + "epoch": 0.23, + "learning_rate": 1.7695462701871314e-07, + "loss": 0.7033, + "step": 6343 + }, + { + "epoch": 0.23, + "learning_rate": 1.7695096495404106e-07, + "loss": 0.7014, + "step": 6344 + }, + { + "epoch": 0.23, + "learning_rate": 1.7694730288936902e-07, + "loss": 0.6774, + "step": 6345 + }, + { + "epoch": 0.23, + "learning_rate": 1.7694364082469694e-07, + "loss": 0.6966, + "step": 6346 + }, + { + "epoch": 0.23, + "learning_rate": 1.769399787600249e-07, + "loss": 0.7089, + "step": 6347 + }, + { + "epoch": 0.23, + "learning_rate": 1.7693631669535282e-07, + "loss": 0.7107, + "step": 6348 + }, + { + "epoch": 0.23, + "learning_rate": 1.7693265463068078e-07, + "loss": 0.6964, + "step": 6349 + }, + { + "epoch": 0.23, + "learning_rate": 1.769289925660087e-07, + "loss": 0.7177, + "step": 6350 + }, + { + "epoch": 0.23, + "learning_rate": 1.7692533050133666e-07, + "loss": 0.6845, + "step": 6351 + }, + { + "epoch": 0.23, + "learning_rate": 1.7692166843666458e-07, + "loss": 0.6924, + "step": 6352 + }, + { + "epoch": 0.23, + "learning_rate": 1.769180063719925e-07, + "loss": 0.6732, + "step": 6353 + }, + { + "epoch": 0.23, + "learning_rate": 1.7691434430732046e-07, + "loss": 0.7029, + "step": 6354 + }, + { + "epoch": 0.23, + "learning_rate": 1.769106822426484e-07, + "loss": 0.6832, + "step": 6355 + }, + { + "epoch": 0.23, + "learning_rate": 1.7690702017797634e-07, + "loss": 0.6837, + "step": 6356 + }, + { + "epoch": 0.23, + "learning_rate": 1.7690335811330427e-07, + "loss": 0.6923, + "step": 6357 + }, + { + "epoch": 0.23, + "learning_rate": 1.7689969604863222e-07, + "loss": 0.6992, + "step": 6358 + }, + { + "epoch": 0.23, + "learning_rate": 1.7689603398396015e-07, + "loss": 0.7071, + "step": 6359 + }, + { + "epoch": 0.23, + "learning_rate": 1.768923719192881e-07, + "loss": 0.6875, + "step": 6360 + }, + { + "epoch": 0.23, + "learning_rate": 1.76888709854616e-07, + "loss": 0.6909, + "step": 6361 + }, + { + "epoch": 0.23, + "learning_rate": 1.7688504778994396e-07, + "loss": 0.7044, + "step": 6362 + }, + { + "epoch": 0.23, + "learning_rate": 1.7688138572527188e-07, + "loss": 0.7005, + "step": 6363 + }, + { + "epoch": 0.23, + "learning_rate": 1.7687772366059984e-07, + "loss": 0.6917, + "step": 6364 + }, + { + "epoch": 0.23, + "learning_rate": 1.7687406159592776e-07, + "loss": 0.704, + "step": 6365 + }, + { + "epoch": 0.23, + "learning_rate": 1.7687039953125572e-07, + "loss": 0.6897, + "step": 6366 + }, + { + "epoch": 0.23, + "learning_rate": 1.7686673746658364e-07, + "loss": 0.6986, + "step": 6367 + }, + { + "epoch": 0.23, + "learning_rate": 1.768630754019116e-07, + "loss": 0.6727, + "step": 6368 + }, + { + "epoch": 0.23, + "learning_rate": 1.7685941333723952e-07, + "loss": 0.694, + "step": 6369 + }, + { + "epoch": 0.23, + "learning_rate": 1.7685575127256748e-07, + "loss": 0.6832, + "step": 6370 + }, + { + "epoch": 0.23, + "learning_rate": 1.768520892078954e-07, + "loss": 0.6899, + "step": 6371 + }, + { + "epoch": 0.23, + "learning_rate": 1.7684842714322336e-07, + "loss": 0.6937, + "step": 6372 + }, + { + "epoch": 0.23, + "learning_rate": 1.7684476507855128e-07, + "loss": 0.6857, + "step": 6373 + }, + { + "epoch": 0.23, + "learning_rate": 1.7684110301387924e-07, + "loss": 0.6925, + "step": 6374 + }, + { + "epoch": 0.23, + "learning_rate": 1.7683744094920716e-07, + "loss": 0.6962, + "step": 6375 + }, + { + "epoch": 0.23, + "learning_rate": 1.768337788845351e-07, + "loss": 0.6956, + "step": 6376 + }, + { + "epoch": 0.23, + "learning_rate": 1.7683011681986304e-07, + "loss": 0.6917, + "step": 6377 + }, + { + "epoch": 0.23, + "learning_rate": 1.7682645475519097e-07, + "loss": 0.7051, + "step": 6378 + }, + { + "epoch": 0.23, + "learning_rate": 1.768227926905189e-07, + "loss": 0.6968, + "step": 6379 + }, + { + "epoch": 0.23, + "learning_rate": 1.7681913062584685e-07, + "loss": 0.6967, + "step": 6380 + }, + { + "epoch": 0.23, + "learning_rate": 1.7681546856117478e-07, + "loss": 0.7053, + "step": 6381 + }, + { + "epoch": 0.23, + "learning_rate": 1.768118064965027e-07, + "loss": 0.6835, + "step": 6382 + }, + { + "epoch": 0.23, + "learning_rate": 1.7680814443183065e-07, + "loss": 0.6875, + "step": 6383 + }, + { + "epoch": 0.23, + "learning_rate": 1.7680448236715858e-07, + "loss": 0.6952, + "step": 6384 + }, + { + "epoch": 0.23, + "learning_rate": 1.7680082030248653e-07, + "loss": 0.7027, + "step": 6385 + }, + { + "epoch": 0.23, + "learning_rate": 1.7679715823781446e-07, + "loss": 0.6992, + "step": 6386 + }, + { + "epoch": 0.23, + "learning_rate": 1.7679349617314241e-07, + "loss": 0.7065, + "step": 6387 + }, + { + "epoch": 0.23, + "learning_rate": 1.7678983410847034e-07, + "loss": 0.6782, + "step": 6388 + }, + { + "epoch": 0.23, + "learning_rate": 1.767861720437983e-07, + "loss": 0.699, + "step": 6389 + }, + { + "epoch": 0.23, + "learning_rate": 1.7678250997912622e-07, + "loss": 0.7043, + "step": 6390 + }, + { + "epoch": 0.23, + "learning_rate": 1.7677884791445417e-07, + "loss": 0.6884, + "step": 6391 + }, + { + "epoch": 0.23, + "learning_rate": 1.767751858497821e-07, + "loss": 0.7078, + "step": 6392 + }, + { + "epoch": 0.23, + "learning_rate": 1.7677152378511005e-07, + "loss": 0.6931, + "step": 6393 + }, + { + "epoch": 0.23, + "learning_rate": 1.7676786172043798e-07, + "loss": 0.688, + "step": 6394 + }, + { + "epoch": 0.23, + "learning_rate": 1.7676419965576593e-07, + "loss": 0.6985, + "step": 6395 + }, + { + "epoch": 0.23, + "learning_rate": 1.7676053759109386e-07, + "loss": 0.7022, + "step": 6396 + }, + { + "epoch": 0.23, + "learning_rate": 1.767568755264218e-07, + "loss": 0.6909, + "step": 6397 + }, + { + "epoch": 0.23, + "learning_rate": 1.7675321346174971e-07, + "loss": 0.7027, + "step": 6398 + }, + { + "epoch": 0.23, + "learning_rate": 1.7674955139707767e-07, + "loss": 0.6934, + "step": 6399 + }, + { + "epoch": 0.23, + "learning_rate": 1.767458893324056e-07, + "loss": 0.718, + "step": 6400 + }, + { + "epoch": 0.23, + "learning_rate": 1.7674222726773355e-07, + "loss": 0.7051, + "step": 6401 + }, + { + "epoch": 0.23, + "learning_rate": 1.7673856520306147e-07, + "loss": 0.6764, + "step": 6402 + }, + { + "epoch": 0.23, + "learning_rate": 1.767349031383894e-07, + "loss": 0.71, + "step": 6403 + }, + { + "epoch": 0.23, + "learning_rate": 1.7673124107371735e-07, + "loss": 0.7028, + "step": 6404 + }, + { + "epoch": 0.23, + "learning_rate": 1.7672757900904528e-07, + "loss": 0.7005, + "step": 6405 + }, + { + "epoch": 0.23, + "learning_rate": 1.7672391694437323e-07, + "loss": 0.6891, + "step": 6406 + }, + { + "epoch": 0.23, + "learning_rate": 1.7672025487970116e-07, + "loss": 0.6913, + "step": 6407 + }, + { + "epoch": 0.23, + "learning_rate": 1.7671659281502911e-07, + "loss": 0.6863, + "step": 6408 + }, + { + "epoch": 0.23, + "learning_rate": 1.7671293075035704e-07, + "loss": 0.6949, + "step": 6409 + }, + { + "epoch": 0.23, + "learning_rate": 1.76709268685685e-07, + "loss": 0.716, + "step": 6410 + }, + { + "epoch": 0.23, + "learning_rate": 1.7670560662101292e-07, + "loss": 0.6859, + "step": 6411 + }, + { + "epoch": 0.23, + "learning_rate": 1.7670194455634087e-07, + "loss": 0.6964, + "step": 6412 + }, + { + "epoch": 0.23, + "learning_rate": 1.766982824916688e-07, + "loss": 0.6727, + "step": 6413 + }, + { + "epoch": 0.23, + "learning_rate": 1.7669462042699675e-07, + "loss": 0.7017, + "step": 6414 + }, + { + "epoch": 0.23, + "learning_rate": 1.7669095836232465e-07, + "loss": 0.6923, + "step": 6415 + }, + { + "epoch": 0.23, + "learning_rate": 1.766872962976526e-07, + "loss": 0.6984, + "step": 6416 + }, + { + "epoch": 0.23, + "learning_rate": 1.7668363423298053e-07, + "loss": 0.6756, + "step": 6417 + }, + { + "epoch": 0.23, + "learning_rate": 1.7667997216830849e-07, + "loss": 0.7076, + "step": 6418 + }, + { + "epoch": 0.23, + "learning_rate": 1.766763101036364e-07, + "loss": 0.6942, + "step": 6419 + }, + { + "epoch": 0.23, + "learning_rate": 1.7667264803896437e-07, + "loss": 0.6791, + "step": 6420 + }, + { + "epoch": 0.23, + "learning_rate": 1.766689859742923e-07, + "loss": 0.7047, + "step": 6421 + }, + { + "epoch": 0.23, + "learning_rate": 1.7666532390962025e-07, + "loss": 0.7048, + "step": 6422 + }, + { + "epoch": 0.23, + "learning_rate": 1.7666166184494817e-07, + "loss": 0.7048, + "step": 6423 + }, + { + "epoch": 0.24, + "learning_rate": 1.766579997802761e-07, + "loss": 0.6985, + "step": 6424 + }, + { + "epoch": 0.24, + "learning_rate": 1.7665433771560405e-07, + "loss": 0.6761, + "step": 6425 + }, + { + "epoch": 0.24, + "learning_rate": 1.7665067565093198e-07, + "loss": 0.6878, + "step": 6426 + }, + { + "epoch": 0.24, + "learning_rate": 1.7664701358625993e-07, + "loss": 0.7019, + "step": 6427 + }, + { + "epoch": 0.24, + "learning_rate": 1.7664335152158786e-07, + "loss": 0.6754, + "step": 6428 + }, + { + "epoch": 0.24, + "learning_rate": 1.766396894569158e-07, + "loss": 0.6943, + "step": 6429 + }, + { + "epoch": 0.24, + "learning_rate": 1.7663602739224374e-07, + "loss": 0.6882, + "step": 6430 + }, + { + "epoch": 0.24, + "learning_rate": 1.766323653275717e-07, + "loss": 0.6843, + "step": 6431 + }, + { + "epoch": 0.24, + "learning_rate": 1.7662870326289962e-07, + "loss": 0.7024, + "step": 6432 + }, + { + "epoch": 0.24, + "learning_rate": 1.7662504119822755e-07, + "loss": 0.7157, + "step": 6433 + }, + { + "epoch": 0.24, + "learning_rate": 1.7662137913355547e-07, + "loss": 0.6931, + "step": 6434 + }, + { + "epoch": 0.24, + "learning_rate": 1.7661771706888343e-07, + "loss": 0.6921, + "step": 6435 + }, + { + "epoch": 0.24, + "learning_rate": 1.7661405500421135e-07, + "loss": 0.6881, + "step": 6436 + }, + { + "epoch": 0.24, + "learning_rate": 1.766103929395393e-07, + "loss": 0.6859, + "step": 6437 + }, + { + "epoch": 0.24, + "learning_rate": 1.7660673087486723e-07, + "loss": 0.6958, + "step": 6438 + }, + { + "epoch": 0.24, + "learning_rate": 1.7660306881019518e-07, + "loss": 0.695, + "step": 6439 + }, + { + "epoch": 0.24, + "learning_rate": 1.765994067455231e-07, + "loss": 0.6905, + "step": 6440 + }, + { + "epoch": 0.24, + "learning_rate": 1.7659574468085106e-07, + "loss": 0.6815, + "step": 6441 + }, + { + "epoch": 0.24, + "learning_rate": 1.76592082616179e-07, + "loss": 0.6903, + "step": 6442 + }, + { + "epoch": 0.24, + "learning_rate": 1.7658842055150694e-07, + "loss": 0.6996, + "step": 6443 + }, + { + "epoch": 0.24, + "learning_rate": 1.7658475848683487e-07, + "loss": 0.6798, + "step": 6444 + }, + { + "epoch": 0.24, + "learning_rate": 1.7658109642216282e-07, + "loss": 0.7036, + "step": 6445 + }, + { + "epoch": 0.24, + "learning_rate": 1.7657743435749075e-07, + "loss": 0.6893, + "step": 6446 + }, + { + "epoch": 0.24, + "learning_rate": 1.7657377229281868e-07, + "loss": 0.6936, + "step": 6447 + }, + { + "epoch": 0.24, + "learning_rate": 1.7657011022814663e-07, + "loss": 0.6845, + "step": 6448 + }, + { + "epoch": 0.24, + "learning_rate": 1.7656644816347456e-07, + "loss": 0.6862, + "step": 6449 + }, + { + "epoch": 0.24, + "learning_rate": 1.765627860988025e-07, + "loss": 0.7005, + "step": 6450 + }, + { + "epoch": 0.24, + "learning_rate": 1.7655912403413044e-07, + "loss": 0.6757, + "step": 6451 + }, + { + "epoch": 0.24, + "learning_rate": 1.7655546196945836e-07, + "loss": 0.6974, + "step": 6452 + }, + { + "epoch": 0.24, + "learning_rate": 1.765517999047863e-07, + "loss": 0.6963, + "step": 6453 + }, + { + "epoch": 0.24, + "learning_rate": 1.7654813784011424e-07, + "loss": 0.6902, + "step": 6454 + }, + { + "epoch": 0.24, + "learning_rate": 1.7654447577544217e-07, + "loss": 0.6999, + "step": 6455 + }, + { + "epoch": 0.24, + "learning_rate": 1.7654081371077012e-07, + "loss": 0.6753, + "step": 6456 + }, + { + "epoch": 0.24, + "learning_rate": 1.7653715164609805e-07, + "loss": 0.6995, + "step": 6457 + }, + { + "epoch": 0.24, + "learning_rate": 1.76533489581426e-07, + "loss": 0.6835, + "step": 6458 + }, + { + "epoch": 0.24, + "learning_rate": 1.7652982751675393e-07, + "loss": 0.6973, + "step": 6459 + }, + { + "epoch": 0.24, + "learning_rate": 1.7652616545208188e-07, + "loss": 0.7054, + "step": 6460 + }, + { + "epoch": 0.24, + "learning_rate": 1.765225033874098e-07, + "loss": 0.7054, + "step": 6461 + }, + { + "epoch": 0.24, + "learning_rate": 1.7651884132273776e-07, + "loss": 0.6954, + "step": 6462 + }, + { + "epoch": 0.24, + "learning_rate": 1.765151792580657e-07, + "loss": 0.6884, + "step": 6463 + }, + { + "epoch": 0.24, + "learning_rate": 1.7651151719339364e-07, + "loss": 0.7046, + "step": 6464 + }, + { + "epoch": 0.24, + "learning_rate": 1.7650785512872157e-07, + "loss": 0.7, + "step": 6465 + }, + { + "epoch": 0.24, + "learning_rate": 1.7650419306404952e-07, + "loss": 0.6942, + "step": 6466 + }, + { + "epoch": 0.24, + "learning_rate": 1.7650053099937745e-07, + "loss": 0.6996, + "step": 6467 + }, + { + "epoch": 0.24, + "learning_rate": 1.7649686893470538e-07, + "loss": 0.6921, + "step": 6468 + }, + { + "epoch": 0.24, + "learning_rate": 1.764932068700333e-07, + "loss": 0.6747, + "step": 6469 + }, + { + "epoch": 0.24, + "learning_rate": 1.7648954480536126e-07, + "loss": 0.6913, + "step": 6470 + }, + { + "epoch": 0.24, + "learning_rate": 1.7648588274068918e-07, + "loss": 0.7075, + "step": 6471 + }, + { + "epoch": 0.24, + "learning_rate": 1.7648222067601714e-07, + "loss": 0.7069, + "step": 6472 + }, + { + "epoch": 0.24, + "learning_rate": 1.7647855861134506e-07, + "loss": 0.7013, + "step": 6473 + }, + { + "epoch": 0.24, + "learning_rate": 1.76474896546673e-07, + "loss": 0.7057, + "step": 6474 + }, + { + "epoch": 0.24, + "learning_rate": 1.7647123448200094e-07, + "loss": 0.6946, + "step": 6475 + }, + { + "epoch": 0.24, + "learning_rate": 1.7646757241732887e-07, + "loss": 0.7088, + "step": 6476 + }, + { + "epoch": 0.24, + "learning_rate": 1.7646391035265682e-07, + "loss": 0.6767, + "step": 6477 + }, + { + "epoch": 0.24, + "learning_rate": 1.7646024828798475e-07, + "loss": 0.7037, + "step": 6478 + }, + { + "epoch": 0.24, + "learning_rate": 1.764565862233127e-07, + "loss": 0.7016, + "step": 6479 + }, + { + "epoch": 0.24, + "learning_rate": 1.7645292415864063e-07, + "loss": 0.6925, + "step": 6480 + }, + { + "epoch": 0.24, + "learning_rate": 1.7644926209396858e-07, + "loss": 0.6879, + "step": 6481 + }, + { + "epoch": 0.24, + "learning_rate": 1.764456000292965e-07, + "loss": 0.6928, + "step": 6482 + }, + { + "epoch": 0.24, + "learning_rate": 1.7644193796462446e-07, + "loss": 0.7075, + "step": 6483 + }, + { + "epoch": 0.24, + "learning_rate": 1.764382758999524e-07, + "loss": 0.6912, + "step": 6484 + }, + { + "epoch": 0.24, + "learning_rate": 1.7643461383528034e-07, + "loss": 0.7029, + "step": 6485 + }, + { + "epoch": 0.24, + "learning_rate": 1.7643095177060827e-07, + "loss": 0.6945, + "step": 6486 + }, + { + "epoch": 0.24, + "learning_rate": 1.764272897059362e-07, + "loss": 0.7133, + "step": 6487 + }, + { + "epoch": 0.24, + "learning_rate": 1.7642362764126412e-07, + "loss": 0.6918, + "step": 6488 + }, + { + "epoch": 0.24, + "learning_rate": 1.7641996557659208e-07, + "loss": 0.6888, + "step": 6489 + }, + { + "epoch": 0.24, + "learning_rate": 1.7641630351192e-07, + "loss": 0.7092, + "step": 6490 + }, + { + "epoch": 0.24, + "learning_rate": 1.7641264144724796e-07, + "loss": 0.6842, + "step": 6491 + }, + { + "epoch": 0.24, + "learning_rate": 1.7640897938257588e-07, + "loss": 0.709, + "step": 6492 + }, + { + "epoch": 0.24, + "learning_rate": 1.7640531731790384e-07, + "loss": 0.6935, + "step": 6493 + }, + { + "epoch": 0.24, + "learning_rate": 1.7640165525323176e-07, + "loss": 0.6746, + "step": 6494 + }, + { + "epoch": 0.24, + "learning_rate": 1.763979931885597e-07, + "loss": 0.6865, + "step": 6495 + }, + { + "epoch": 0.24, + "learning_rate": 1.7639433112388764e-07, + "loss": 0.6863, + "step": 6496 + }, + { + "epoch": 0.24, + "learning_rate": 1.7639066905921557e-07, + "loss": 0.6975, + "step": 6497 + }, + { + "epoch": 0.24, + "learning_rate": 1.7638700699454352e-07, + "loss": 0.6755, + "step": 6498 + }, + { + "epoch": 0.24, + "learning_rate": 1.7638334492987145e-07, + "loss": 0.6895, + "step": 6499 + }, + { + "epoch": 0.24, + "learning_rate": 1.763796828651994e-07, + "loss": 0.6925, + "step": 6500 + }, + { + "epoch": 0.24, + "learning_rate": 1.7637602080052733e-07, + "loss": 0.6839, + "step": 6501 + }, + { + "epoch": 0.24, + "learning_rate": 1.7637235873585528e-07, + "loss": 0.7106, + "step": 6502 + }, + { + "epoch": 0.24, + "learning_rate": 1.763686966711832e-07, + "loss": 0.692, + "step": 6503 + }, + { + "epoch": 0.24, + "learning_rate": 1.7636503460651116e-07, + "loss": 0.6838, + "step": 6504 + }, + { + "epoch": 0.24, + "learning_rate": 1.7636137254183906e-07, + "loss": 0.7049, + "step": 6505 + }, + { + "epoch": 0.24, + "learning_rate": 1.7635771047716701e-07, + "loss": 0.6912, + "step": 6506 + }, + { + "epoch": 0.24, + "learning_rate": 1.7635404841249494e-07, + "loss": 0.6972, + "step": 6507 + }, + { + "epoch": 0.24, + "learning_rate": 1.763503863478229e-07, + "loss": 0.6946, + "step": 6508 + }, + { + "epoch": 0.24, + "learning_rate": 1.7634672428315082e-07, + "loss": 0.7132, + "step": 6509 + }, + { + "epoch": 0.24, + "learning_rate": 1.7634306221847877e-07, + "loss": 0.6932, + "step": 6510 + }, + { + "epoch": 0.24, + "learning_rate": 1.763394001538067e-07, + "loss": 0.7026, + "step": 6511 + }, + { + "epoch": 0.24, + "learning_rate": 1.7633573808913465e-07, + "loss": 0.6919, + "step": 6512 + }, + { + "epoch": 0.24, + "learning_rate": 1.7633207602446258e-07, + "loss": 0.7069, + "step": 6513 + }, + { + "epoch": 0.24, + "learning_rate": 1.7632841395979053e-07, + "loss": 0.6915, + "step": 6514 + }, + { + "epoch": 0.24, + "learning_rate": 1.7632475189511846e-07, + "loss": 0.6963, + "step": 6515 + }, + { + "epoch": 0.24, + "learning_rate": 1.7632108983044641e-07, + "loss": 0.6974, + "step": 6516 + }, + { + "epoch": 0.24, + "learning_rate": 1.7631742776577434e-07, + "loss": 0.6913, + "step": 6517 + }, + { + "epoch": 0.24, + "learning_rate": 1.7631376570110227e-07, + "loss": 0.7042, + "step": 6518 + }, + { + "epoch": 0.24, + "learning_rate": 1.7631010363643022e-07, + "loss": 0.6881, + "step": 6519 + }, + { + "epoch": 0.24, + "learning_rate": 1.7630644157175815e-07, + "loss": 0.6955, + "step": 6520 + }, + { + "epoch": 0.24, + "learning_rate": 1.763027795070861e-07, + "loss": 0.6866, + "step": 6521 + }, + { + "epoch": 0.24, + "learning_rate": 1.7629911744241403e-07, + "loss": 0.696, + "step": 6522 + }, + { + "epoch": 0.24, + "learning_rate": 1.7629545537774195e-07, + "loss": 0.6826, + "step": 6523 + }, + { + "epoch": 0.24, + "learning_rate": 1.7629179331306988e-07, + "loss": 0.6863, + "step": 6524 + }, + { + "epoch": 0.24, + "learning_rate": 1.7628813124839783e-07, + "loss": 0.6934, + "step": 6525 + }, + { + "epoch": 0.24, + "learning_rate": 1.7628446918372576e-07, + "loss": 0.6818, + "step": 6526 + }, + { + "epoch": 0.24, + "learning_rate": 1.7628080711905371e-07, + "loss": 0.6864, + "step": 6527 + }, + { + "epoch": 0.24, + "learning_rate": 1.7627714505438164e-07, + "loss": 0.7047, + "step": 6528 + }, + { + "epoch": 0.24, + "learning_rate": 1.762734829897096e-07, + "loss": 0.6832, + "step": 6529 + }, + { + "epoch": 0.24, + "learning_rate": 1.7626982092503752e-07, + "loss": 0.6977, + "step": 6530 + }, + { + "epoch": 0.24, + "learning_rate": 1.7626615886036547e-07, + "loss": 0.686, + "step": 6531 + }, + { + "epoch": 0.24, + "learning_rate": 1.762624967956934e-07, + "loss": 0.6943, + "step": 6532 + }, + { + "epoch": 0.24, + "learning_rate": 1.7625883473102135e-07, + "loss": 0.6922, + "step": 6533 + }, + { + "epoch": 0.24, + "learning_rate": 1.7625517266634928e-07, + "loss": 0.701, + "step": 6534 + }, + { + "epoch": 0.24, + "learning_rate": 1.7625151060167723e-07, + "loss": 0.7025, + "step": 6535 + }, + { + "epoch": 0.24, + "learning_rate": 1.7624784853700516e-07, + "loss": 0.6908, + "step": 6536 + }, + { + "epoch": 0.24, + "learning_rate": 1.762441864723331e-07, + "loss": 0.6897, + "step": 6537 + }, + { + "epoch": 0.24, + "learning_rate": 1.7624052440766104e-07, + "loss": 0.7051, + "step": 6538 + }, + { + "epoch": 0.24, + "learning_rate": 1.7623686234298897e-07, + "loss": 0.6967, + "step": 6539 + }, + { + "epoch": 0.24, + "learning_rate": 1.7623320027831692e-07, + "loss": 0.6996, + "step": 6540 + }, + { + "epoch": 0.24, + "learning_rate": 1.7622953821364485e-07, + "loss": 0.6897, + "step": 6541 + }, + { + "epoch": 0.24, + "learning_rate": 1.7622587614897277e-07, + "loss": 0.6868, + "step": 6542 + }, + { + "epoch": 0.24, + "learning_rate": 1.7622221408430073e-07, + "loss": 0.6931, + "step": 6543 + }, + { + "epoch": 0.24, + "learning_rate": 1.7621855201962865e-07, + "loss": 0.6921, + "step": 6544 + }, + { + "epoch": 0.24, + "learning_rate": 1.7621488995495658e-07, + "loss": 0.6946, + "step": 6545 + }, + { + "epoch": 0.24, + "learning_rate": 1.7621122789028453e-07, + "loss": 0.6989, + "step": 6546 + }, + { + "epoch": 0.24, + "learning_rate": 1.7620756582561246e-07, + "loss": 0.6961, + "step": 6547 + }, + { + "epoch": 0.24, + "learning_rate": 1.762039037609404e-07, + "loss": 0.6704, + "step": 6548 + }, + { + "epoch": 0.24, + "learning_rate": 1.7620024169626834e-07, + "loss": 0.6923, + "step": 6549 + }, + { + "epoch": 0.24, + "learning_rate": 1.761965796315963e-07, + "loss": 0.7042, + "step": 6550 + }, + { + "epoch": 0.24, + "learning_rate": 1.7619291756692422e-07, + "loss": 0.6879, + "step": 6551 + }, + { + "epoch": 0.24, + "learning_rate": 1.7618925550225217e-07, + "loss": 0.6948, + "step": 6552 + }, + { + "epoch": 0.24, + "learning_rate": 1.761855934375801e-07, + "loss": 0.6824, + "step": 6553 + }, + { + "epoch": 0.24, + "learning_rate": 1.7618193137290805e-07, + "loss": 0.6936, + "step": 6554 + }, + { + "epoch": 0.24, + "learning_rate": 1.7617826930823598e-07, + "loss": 0.6859, + "step": 6555 + }, + { + "epoch": 0.24, + "learning_rate": 1.7617460724356393e-07, + "loss": 0.687, + "step": 6556 + }, + { + "epoch": 0.24, + "learning_rate": 1.7617094517889186e-07, + "loss": 0.6973, + "step": 6557 + }, + { + "epoch": 0.24, + "learning_rate": 1.761672831142198e-07, + "loss": 0.6796, + "step": 6558 + }, + { + "epoch": 0.24, + "learning_rate": 1.761636210495477e-07, + "loss": 0.686, + "step": 6559 + }, + { + "epoch": 0.24, + "learning_rate": 1.7615995898487566e-07, + "loss": 0.6921, + "step": 6560 + }, + { + "epoch": 0.24, + "learning_rate": 1.761562969202036e-07, + "loss": 0.6991, + "step": 6561 + }, + { + "epoch": 0.24, + "learning_rate": 1.7615263485553154e-07, + "loss": 0.6786, + "step": 6562 + }, + { + "epoch": 0.24, + "learning_rate": 1.7614897279085947e-07, + "loss": 0.6976, + "step": 6563 + }, + { + "epoch": 0.24, + "learning_rate": 1.7614531072618742e-07, + "loss": 0.6939, + "step": 6564 + }, + { + "epoch": 0.24, + "learning_rate": 1.7614164866151535e-07, + "loss": 0.6814, + "step": 6565 + }, + { + "epoch": 0.24, + "learning_rate": 1.7613798659684328e-07, + "loss": 0.6803, + "step": 6566 + }, + { + "epoch": 0.24, + "learning_rate": 1.7613432453217123e-07, + "loss": 0.6989, + "step": 6567 + }, + { + "epoch": 0.24, + "learning_rate": 1.7613066246749916e-07, + "loss": 0.6904, + "step": 6568 + }, + { + "epoch": 0.24, + "learning_rate": 1.761270004028271e-07, + "loss": 0.6837, + "step": 6569 + }, + { + "epoch": 0.24, + "learning_rate": 1.7612333833815504e-07, + "loss": 0.6982, + "step": 6570 + }, + { + "epoch": 0.24, + "learning_rate": 1.76119676273483e-07, + "loss": 0.6963, + "step": 6571 + }, + { + "epoch": 0.24, + "learning_rate": 1.7611601420881092e-07, + "loss": 0.7084, + "step": 6572 + }, + { + "epoch": 0.24, + "learning_rate": 1.7611235214413887e-07, + "loss": 0.7044, + "step": 6573 + }, + { + "epoch": 0.24, + "learning_rate": 1.761086900794668e-07, + "loss": 0.7027, + "step": 6574 + }, + { + "epoch": 0.24, + "learning_rate": 1.7610502801479475e-07, + "loss": 0.6884, + "step": 6575 + }, + { + "epoch": 0.24, + "learning_rate": 1.7610136595012268e-07, + "loss": 0.7081, + "step": 6576 + }, + { + "epoch": 0.24, + "learning_rate": 1.760977038854506e-07, + "loss": 0.6965, + "step": 6577 + }, + { + "epoch": 0.24, + "learning_rate": 1.7609404182077853e-07, + "loss": 0.6804, + "step": 6578 + }, + { + "epoch": 0.24, + "learning_rate": 1.7609037975610648e-07, + "loss": 0.7049, + "step": 6579 + }, + { + "epoch": 0.24, + "learning_rate": 1.760867176914344e-07, + "loss": 0.679, + "step": 6580 + }, + { + "epoch": 0.24, + "learning_rate": 1.7608305562676236e-07, + "loss": 0.6903, + "step": 6581 + }, + { + "epoch": 0.24, + "learning_rate": 1.760793935620903e-07, + "loss": 0.6899, + "step": 6582 + }, + { + "epoch": 0.24, + "learning_rate": 1.7607573149741824e-07, + "loss": 0.6953, + "step": 6583 + }, + { + "epoch": 0.24, + "learning_rate": 1.7607206943274617e-07, + "loss": 0.7074, + "step": 6584 + }, + { + "epoch": 0.24, + "learning_rate": 1.7606840736807412e-07, + "loss": 0.6941, + "step": 6585 + }, + { + "epoch": 0.24, + "learning_rate": 1.7606474530340205e-07, + "loss": 0.715, + "step": 6586 + }, + { + "epoch": 0.24, + "learning_rate": 1.7606108323873e-07, + "loss": 0.705, + "step": 6587 + }, + { + "epoch": 0.24, + "learning_rate": 1.7605742117405793e-07, + "loss": 0.679, + "step": 6588 + }, + { + "epoch": 0.24, + "learning_rate": 1.7605375910938586e-07, + "loss": 0.6861, + "step": 6589 + }, + { + "epoch": 0.24, + "learning_rate": 1.760500970447138e-07, + "loss": 0.7033, + "step": 6590 + }, + { + "epoch": 0.24, + "learning_rate": 1.7604643498004174e-07, + "loss": 0.7201, + "step": 6591 + }, + { + "epoch": 0.24, + "learning_rate": 1.760427729153697e-07, + "loss": 0.6843, + "step": 6592 + }, + { + "epoch": 0.24, + "learning_rate": 1.7603911085069762e-07, + "loss": 0.6787, + "step": 6593 + }, + { + "epoch": 0.24, + "learning_rate": 1.7603544878602554e-07, + "loss": 0.6903, + "step": 6594 + }, + { + "epoch": 0.24, + "learning_rate": 1.7603178672135347e-07, + "loss": 0.6943, + "step": 6595 + }, + { + "epoch": 0.24, + "learning_rate": 1.7602812465668142e-07, + "loss": 0.6805, + "step": 6596 + }, + { + "epoch": 0.24, + "learning_rate": 1.7602446259200935e-07, + "loss": 0.7011, + "step": 6597 + }, + { + "epoch": 0.24, + "learning_rate": 1.760208005273373e-07, + "loss": 0.703, + "step": 6598 + }, + { + "epoch": 0.24, + "learning_rate": 1.7601713846266523e-07, + "loss": 0.6835, + "step": 6599 + }, + { + "epoch": 0.24, + "learning_rate": 1.7601347639799318e-07, + "loss": 0.6902, + "step": 6600 + }, + { + "epoch": 0.24, + "learning_rate": 1.760098143333211e-07, + "loss": 0.7027, + "step": 6601 + }, + { + "epoch": 0.24, + "learning_rate": 1.7600615226864906e-07, + "loss": 0.7082, + "step": 6602 + }, + { + "epoch": 0.24, + "learning_rate": 1.76002490203977e-07, + "loss": 0.6885, + "step": 6603 + }, + { + "epoch": 0.24, + "learning_rate": 1.7599882813930494e-07, + "loss": 0.698, + "step": 6604 + }, + { + "epoch": 0.24, + "learning_rate": 1.7599516607463287e-07, + "loss": 0.7039, + "step": 6605 + }, + { + "epoch": 0.24, + "learning_rate": 1.7599150400996082e-07, + "loss": 0.6982, + "step": 6606 + }, + { + "epoch": 0.24, + "learning_rate": 1.7598784194528875e-07, + "loss": 0.7105, + "step": 6607 + }, + { + "epoch": 0.24, + "learning_rate": 1.759841798806167e-07, + "loss": 0.6894, + "step": 6608 + }, + { + "epoch": 0.24, + "learning_rate": 1.7598051781594463e-07, + "loss": 0.6919, + "step": 6609 + }, + { + "epoch": 0.24, + "learning_rate": 1.7597685575127256e-07, + "loss": 0.6808, + "step": 6610 + }, + { + "epoch": 0.24, + "learning_rate": 1.759731936866005e-07, + "loss": 0.686, + "step": 6611 + }, + { + "epoch": 0.24, + "learning_rate": 1.7596953162192844e-07, + "loss": 0.6878, + "step": 6612 + }, + { + "epoch": 0.24, + "learning_rate": 1.7596586955725636e-07, + "loss": 0.6938, + "step": 6613 + }, + { + "epoch": 0.24, + "learning_rate": 1.7596220749258432e-07, + "loss": 0.6838, + "step": 6614 + }, + { + "epoch": 0.24, + "learning_rate": 1.7595854542791224e-07, + "loss": 0.7055, + "step": 6615 + }, + { + "epoch": 0.24, + "learning_rate": 1.7595488336324017e-07, + "loss": 0.7017, + "step": 6616 + }, + { + "epoch": 0.24, + "learning_rate": 1.7595122129856812e-07, + "loss": 0.6995, + "step": 6617 + }, + { + "epoch": 0.24, + "learning_rate": 1.7594755923389605e-07, + "loss": 0.6903, + "step": 6618 + }, + { + "epoch": 0.24, + "learning_rate": 1.75943897169224e-07, + "loss": 0.6946, + "step": 6619 + }, + { + "epoch": 0.24, + "learning_rate": 1.7594023510455193e-07, + "loss": 0.7008, + "step": 6620 + }, + { + "epoch": 0.24, + "learning_rate": 1.7593657303987988e-07, + "loss": 0.6957, + "step": 6621 + }, + { + "epoch": 0.24, + "learning_rate": 1.759329109752078e-07, + "loss": 0.7097, + "step": 6622 + }, + { + "epoch": 0.24, + "learning_rate": 1.7592924891053576e-07, + "loss": 0.6977, + "step": 6623 + }, + { + "epoch": 0.24, + "learning_rate": 1.759255868458637e-07, + "loss": 0.7038, + "step": 6624 + }, + { + "epoch": 0.24, + "learning_rate": 1.7592192478119164e-07, + "loss": 0.6911, + "step": 6625 + }, + { + "epoch": 0.24, + "learning_rate": 1.7591826271651957e-07, + "loss": 0.6966, + "step": 6626 + }, + { + "epoch": 0.24, + "learning_rate": 1.7591460065184752e-07, + "loss": 0.6955, + "step": 6627 + }, + { + "epoch": 0.24, + "learning_rate": 1.7591093858717545e-07, + "loss": 0.6932, + "step": 6628 + }, + { + "epoch": 0.24, + "learning_rate": 1.759072765225034e-07, + "loss": 0.6736, + "step": 6629 + }, + { + "epoch": 0.24, + "learning_rate": 1.759036144578313e-07, + "loss": 0.7025, + "step": 6630 + }, + { + "epoch": 0.24, + "learning_rate": 1.7589995239315925e-07, + "loss": 0.6923, + "step": 6631 + }, + { + "epoch": 0.24, + "learning_rate": 1.7589629032848718e-07, + "loss": 0.6943, + "step": 6632 + }, + { + "epoch": 0.24, + "learning_rate": 1.7589262826381513e-07, + "loss": 0.6964, + "step": 6633 + }, + { + "epoch": 0.24, + "learning_rate": 1.7588896619914306e-07, + "loss": 0.6775, + "step": 6634 + }, + { + "epoch": 0.24, + "learning_rate": 1.7588530413447101e-07, + "loss": 0.6898, + "step": 6635 + }, + { + "epoch": 0.24, + "learning_rate": 1.7588164206979894e-07, + "loss": 0.6933, + "step": 6636 + }, + { + "epoch": 0.24, + "learning_rate": 1.7587798000512687e-07, + "loss": 0.687, + "step": 6637 + }, + { + "epoch": 0.24, + "learning_rate": 1.7587431794045482e-07, + "loss": 0.6883, + "step": 6638 + }, + { + "epoch": 0.24, + "learning_rate": 1.7587065587578275e-07, + "loss": 0.7011, + "step": 6639 + }, + { + "epoch": 0.24, + "learning_rate": 1.758669938111107e-07, + "loss": 0.6776, + "step": 6640 + }, + { + "epoch": 0.24, + "learning_rate": 1.7586333174643863e-07, + "loss": 0.6862, + "step": 6641 + }, + { + "epoch": 0.24, + "learning_rate": 1.7585966968176658e-07, + "loss": 0.6814, + "step": 6642 + }, + { + "epoch": 0.24, + "learning_rate": 1.758560076170945e-07, + "loss": 0.6825, + "step": 6643 + }, + { + "epoch": 0.24, + "learning_rate": 1.7585234555242246e-07, + "loss": 0.6942, + "step": 6644 + }, + { + "epoch": 0.24, + "learning_rate": 1.758486834877504e-07, + "loss": 0.6956, + "step": 6645 + }, + { + "epoch": 0.24, + "learning_rate": 1.7584502142307834e-07, + "loss": 0.6882, + "step": 6646 + }, + { + "epoch": 0.24, + "learning_rate": 1.7584135935840627e-07, + "loss": 0.6796, + "step": 6647 + }, + { + "epoch": 0.24, + "learning_rate": 1.758376972937342e-07, + "loss": 0.6927, + "step": 6648 + }, + { + "epoch": 0.24, + "learning_rate": 1.7583403522906212e-07, + "loss": 0.6736, + "step": 6649 + }, + { + "epoch": 0.24, + "learning_rate": 1.7583037316439007e-07, + "loss": 0.7201, + "step": 6650 + }, + { + "epoch": 0.24, + "learning_rate": 1.75826711099718e-07, + "loss": 0.7005, + "step": 6651 + }, + { + "epoch": 0.24, + "learning_rate": 1.7582304903504595e-07, + "loss": 0.7047, + "step": 6652 + }, + { + "epoch": 0.24, + "learning_rate": 1.7581938697037388e-07, + "loss": 0.6984, + "step": 6653 + }, + { + "epoch": 0.24, + "learning_rate": 1.7581572490570183e-07, + "loss": 0.6819, + "step": 6654 + }, + { + "epoch": 0.24, + "learning_rate": 1.7581206284102976e-07, + "loss": 0.6859, + "step": 6655 + }, + { + "epoch": 0.24, + "learning_rate": 1.758084007763577e-07, + "loss": 0.7133, + "step": 6656 + }, + { + "epoch": 0.24, + "learning_rate": 1.7580473871168564e-07, + "loss": 0.693, + "step": 6657 + }, + { + "epoch": 0.24, + "learning_rate": 1.758010766470136e-07, + "loss": 0.6926, + "step": 6658 + }, + { + "epoch": 0.24, + "learning_rate": 1.7579741458234152e-07, + "loss": 0.6986, + "step": 6659 + }, + { + "epoch": 0.24, + "learning_rate": 1.7579375251766945e-07, + "loss": 0.6937, + "step": 6660 + }, + { + "epoch": 0.24, + "learning_rate": 1.757900904529974e-07, + "loss": 0.7041, + "step": 6661 + }, + { + "epoch": 0.24, + "learning_rate": 1.7578642838832533e-07, + "loss": 0.6949, + "step": 6662 + }, + { + "epoch": 0.24, + "learning_rate": 1.7578276632365328e-07, + "loss": 0.691, + "step": 6663 + }, + { + "epoch": 0.24, + "learning_rate": 1.757791042589812e-07, + "loss": 0.7053, + "step": 6664 + }, + { + "epoch": 0.24, + "learning_rate": 1.7577544219430916e-07, + "loss": 0.6854, + "step": 6665 + }, + { + "epoch": 0.24, + "learning_rate": 1.7577178012963706e-07, + "loss": 0.6965, + "step": 6666 + }, + { + "epoch": 0.24, + "learning_rate": 1.75768118064965e-07, + "loss": 0.7133, + "step": 6667 + }, + { + "epoch": 0.24, + "learning_rate": 1.7576445600029294e-07, + "loss": 0.6991, + "step": 6668 + }, + { + "epoch": 0.24, + "learning_rate": 1.757607939356209e-07, + "loss": 0.67, + "step": 6669 + }, + { + "epoch": 0.24, + "learning_rate": 1.7575713187094882e-07, + "loss": 0.6822, + "step": 6670 + }, + { + "epoch": 0.24, + "learning_rate": 1.7575346980627677e-07, + "loss": 0.6913, + "step": 6671 + }, + { + "epoch": 0.24, + "learning_rate": 1.757498077416047e-07, + "loss": 0.6824, + "step": 6672 + }, + { + "epoch": 0.24, + "learning_rate": 1.7574614567693265e-07, + "loss": 0.684, + "step": 6673 + }, + { + "epoch": 0.24, + "learning_rate": 1.7574248361226058e-07, + "loss": 0.6858, + "step": 6674 + }, + { + "epoch": 0.24, + "learning_rate": 1.7573882154758853e-07, + "loss": 0.6946, + "step": 6675 + }, + { + "epoch": 0.24, + "learning_rate": 1.7573515948291646e-07, + "loss": 0.6847, + "step": 6676 + }, + { + "epoch": 0.24, + "learning_rate": 1.757314974182444e-07, + "loss": 0.6737, + "step": 6677 + }, + { + "epoch": 0.24, + "learning_rate": 1.7572783535357234e-07, + "loss": 0.6926, + "step": 6678 + }, + { + "epoch": 0.24, + "learning_rate": 1.757241732889003e-07, + "loss": 0.6839, + "step": 6679 + }, + { + "epoch": 0.24, + "learning_rate": 1.7572051122422822e-07, + "loss": 0.6854, + "step": 6680 + }, + { + "epoch": 0.24, + "learning_rate": 1.7571684915955614e-07, + "loss": 0.6893, + "step": 6681 + }, + { + "epoch": 0.24, + "learning_rate": 1.757131870948841e-07, + "loss": 0.6774, + "step": 6682 + }, + { + "epoch": 0.24, + "learning_rate": 1.7570952503021202e-07, + "loss": 0.6795, + "step": 6683 + }, + { + "epoch": 0.24, + "learning_rate": 1.7570586296553995e-07, + "loss": 0.7306, + "step": 6684 + }, + { + "epoch": 0.24, + "learning_rate": 1.757022009008679e-07, + "loss": 0.6897, + "step": 6685 + }, + { + "epoch": 0.24, + "learning_rate": 1.7569853883619583e-07, + "loss": 0.6917, + "step": 6686 + }, + { + "epoch": 0.24, + "learning_rate": 1.7569487677152376e-07, + "loss": 0.6935, + "step": 6687 + }, + { + "epoch": 0.24, + "learning_rate": 1.756912147068517e-07, + "loss": 0.7, + "step": 6688 + }, + { + "epoch": 0.24, + "learning_rate": 1.7568755264217964e-07, + "loss": 0.6764, + "step": 6689 + }, + { + "epoch": 0.24, + "learning_rate": 1.756838905775076e-07, + "loss": 0.6763, + "step": 6690 + }, + { + "epoch": 0.24, + "learning_rate": 1.7568022851283552e-07, + "loss": 0.6869, + "step": 6691 + }, + { + "epoch": 0.24, + "learning_rate": 1.7567656644816347e-07, + "loss": 0.6963, + "step": 6692 + }, + { + "epoch": 0.24, + "learning_rate": 1.756729043834914e-07, + "loss": 0.6777, + "step": 6693 + }, + { + "epoch": 0.24, + "learning_rate": 1.7566924231881935e-07, + "loss": 0.6716, + "step": 6694 + }, + { + "epoch": 0.24, + "learning_rate": 1.7566558025414728e-07, + "loss": 0.6945, + "step": 6695 + }, + { + "epoch": 0.24, + "learning_rate": 1.7566191818947523e-07, + "loss": 0.6928, + "step": 6696 + }, + { + "epoch": 0.25, + "learning_rate": 1.7565825612480316e-07, + "loss": 0.6727, + "step": 6697 + }, + { + "epoch": 0.25, + "learning_rate": 1.756545940601311e-07, + "loss": 0.6963, + "step": 6698 + }, + { + "epoch": 0.25, + "learning_rate": 1.7565093199545904e-07, + "loss": 0.7019, + "step": 6699 + }, + { + "epoch": 0.25, + "learning_rate": 1.75647269930787e-07, + "loss": 0.7003, + "step": 6700 + }, + { + "epoch": 0.25, + "learning_rate": 1.7564360786611492e-07, + "loss": 0.6897, + "step": 6701 + }, + { + "epoch": 0.25, + "learning_rate": 1.7563994580144284e-07, + "loss": 0.7097, + "step": 6702 + }, + { + "epoch": 0.25, + "learning_rate": 1.7563628373677077e-07, + "loss": 0.6856, + "step": 6703 + }, + { + "epoch": 0.25, + "learning_rate": 1.7563262167209872e-07, + "loss": 0.7038, + "step": 6704 + }, + { + "epoch": 0.25, + "learning_rate": 1.7562895960742665e-07, + "loss": 0.6984, + "step": 6705 + }, + { + "epoch": 0.25, + "learning_rate": 1.756252975427546e-07, + "loss": 0.698, + "step": 6706 + }, + { + "epoch": 0.25, + "learning_rate": 1.7562163547808253e-07, + "loss": 0.6767, + "step": 6707 + }, + { + "epoch": 0.25, + "learning_rate": 1.7561797341341046e-07, + "loss": 0.7033, + "step": 6708 + }, + { + "epoch": 0.25, + "learning_rate": 1.756143113487384e-07, + "loss": 0.7033, + "step": 6709 + }, + { + "epoch": 0.25, + "learning_rate": 1.7561064928406634e-07, + "loss": 0.6935, + "step": 6710 + }, + { + "epoch": 0.25, + "learning_rate": 1.756069872193943e-07, + "loss": 0.69, + "step": 6711 + }, + { + "epoch": 0.25, + "learning_rate": 1.7560332515472222e-07, + "loss": 0.6999, + "step": 6712 + }, + { + "epoch": 0.25, + "learning_rate": 1.7559966309005017e-07, + "loss": 0.6903, + "step": 6713 + }, + { + "epoch": 0.25, + "learning_rate": 1.755960010253781e-07, + "loss": 0.7038, + "step": 6714 + }, + { + "epoch": 0.25, + "learning_rate": 1.7559233896070605e-07, + "loss": 0.7127, + "step": 6715 + }, + { + "epoch": 0.25, + "learning_rate": 1.7558867689603398e-07, + "loss": 0.6839, + "step": 6716 + }, + { + "epoch": 0.25, + "learning_rate": 1.7558501483136193e-07, + "loss": 0.7086, + "step": 6717 + }, + { + "epoch": 0.25, + "learning_rate": 1.7558135276668986e-07, + "loss": 0.6968, + "step": 6718 + }, + { + "epoch": 0.25, + "learning_rate": 1.755776907020178e-07, + "loss": 0.6947, + "step": 6719 + }, + { + "epoch": 0.25, + "learning_rate": 1.755740286373457e-07, + "loss": 0.6975, + "step": 6720 + }, + { + "epoch": 0.25, + "learning_rate": 1.7557036657267366e-07, + "loss": 0.6928, + "step": 6721 + }, + { + "epoch": 0.25, + "learning_rate": 1.755667045080016e-07, + "loss": 0.7032, + "step": 6722 + }, + { + "epoch": 0.25, + "learning_rate": 1.7556304244332954e-07, + "loss": 0.682, + "step": 6723 + }, + { + "epoch": 0.25, + "learning_rate": 1.7555938037865747e-07, + "loss": 0.6993, + "step": 6724 + }, + { + "epoch": 0.25, + "learning_rate": 1.7555571831398542e-07, + "loss": 0.6821, + "step": 6725 + }, + { + "epoch": 0.25, + "learning_rate": 1.7555205624931335e-07, + "loss": 0.6955, + "step": 6726 + }, + { + "epoch": 0.25, + "learning_rate": 1.755483941846413e-07, + "loss": 0.6972, + "step": 6727 + }, + { + "epoch": 0.25, + "learning_rate": 1.7554473211996923e-07, + "loss": 0.6921, + "step": 6728 + }, + { + "epoch": 0.25, + "learning_rate": 1.7554107005529718e-07, + "loss": 0.6971, + "step": 6729 + }, + { + "epoch": 0.25, + "learning_rate": 1.755374079906251e-07, + "loss": 0.7028, + "step": 6730 + }, + { + "epoch": 0.25, + "learning_rate": 1.7553374592595304e-07, + "loss": 0.7102, + "step": 6731 + }, + { + "epoch": 0.25, + "learning_rate": 1.75530083861281e-07, + "loss": 0.6997, + "step": 6732 + }, + { + "epoch": 0.25, + "learning_rate": 1.7552642179660892e-07, + "loss": 0.7106, + "step": 6733 + }, + { + "epoch": 0.25, + "learning_rate": 1.7552275973193687e-07, + "loss": 0.6913, + "step": 6734 + }, + { + "epoch": 0.25, + "learning_rate": 1.755190976672648e-07, + "loss": 0.717, + "step": 6735 + }, + { + "epoch": 0.25, + "learning_rate": 1.7551543560259275e-07, + "loss": 0.6872, + "step": 6736 + }, + { + "epoch": 0.25, + "learning_rate": 1.7551177353792067e-07, + "loss": 0.7203, + "step": 6737 + }, + { + "epoch": 0.25, + "learning_rate": 1.755081114732486e-07, + "loss": 0.687, + "step": 6738 + }, + { + "epoch": 0.25, + "learning_rate": 1.7550444940857653e-07, + "loss": 0.6886, + "step": 6739 + }, + { + "epoch": 0.25, + "learning_rate": 1.7550078734390448e-07, + "loss": 0.7139, + "step": 6740 + }, + { + "epoch": 0.25, + "learning_rate": 1.754971252792324e-07, + "loss": 0.7106, + "step": 6741 + }, + { + "epoch": 0.25, + "learning_rate": 1.7549346321456036e-07, + "loss": 0.6896, + "step": 6742 + }, + { + "epoch": 0.25, + "learning_rate": 1.754898011498883e-07, + "loss": 0.7108, + "step": 6743 + }, + { + "epoch": 0.25, + "learning_rate": 1.7548613908521624e-07, + "loss": 0.6925, + "step": 6744 + }, + { + "epoch": 0.25, + "learning_rate": 1.7548247702054417e-07, + "loss": 0.6948, + "step": 6745 + }, + { + "epoch": 0.25, + "learning_rate": 1.7547881495587212e-07, + "loss": 0.7066, + "step": 6746 + }, + { + "epoch": 0.25, + "learning_rate": 1.7547515289120005e-07, + "loss": 0.7042, + "step": 6747 + }, + { + "epoch": 0.25, + "learning_rate": 1.75471490826528e-07, + "loss": 0.6744, + "step": 6748 + }, + { + "epoch": 0.25, + "learning_rate": 1.7546782876185593e-07, + "loss": 0.6893, + "step": 6749 + }, + { + "epoch": 0.25, + "learning_rate": 1.7546416669718388e-07, + "loss": 0.6995, + "step": 6750 + }, + { + "epoch": 0.25, + "learning_rate": 1.754605046325118e-07, + "loss": 0.7029, + "step": 6751 + }, + { + "epoch": 0.25, + "learning_rate": 1.7545684256783973e-07, + "loss": 0.6955, + "step": 6752 + }, + { + "epoch": 0.25, + "learning_rate": 1.754531805031677e-07, + "loss": 0.6933, + "step": 6753 + }, + { + "epoch": 0.25, + "learning_rate": 1.7544951843849561e-07, + "loss": 0.7101, + "step": 6754 + }, + { + "epoch": 0.25, + "learning_rate": 1.7544585637382357e-07, + "loss": 0.6621, + "step": 6755 + }, + { + "epoch": 0.25, + "learning_rate": 1.754421943091515e-07, + "loss": 0.6799, + "step": 6756 + }, + { + "epoch": 0.25, + "learning_rate": 1.7543853224447942e-07, + "loss": 0.6948, + "step": 6757 + }, + { + "epoch": 0.25, + "learning_rate": 1.7543487017980735e-07, + "loss": 0.7061, + "step": 6758 + }, + { + "epoch": 0.25, + "learning_rate": 1.754312081151353e-07, + "loss": 0.6902, + "step": 6759 + }, + { + "epoch": 0.25, + "learning_rate": 1.7542754605046323e-07, + "loss": 0.7225, + "step": 6760 + }, + { + "epoch": 0.25, + "learning_rate": 1.7542388398579118e-07, + "loss": 0.6726, + "step": 6761 + }, + { + "epoch": 0.25, + "learning_rate": 1.754202219211191e-07, + "loss": 0.7001, + "step": 6762 + }, + { + "epoch": 0.25, + "learning_rate": 1.7541655985644706e-07, + "loss": 0.6896, + "step": 6763 + }, + { + "epoch": 0.25, + "learning_rate": 1.75412897791775e-07, + "loss": 0.6853, + "step": 6764 + }, + { + "epoch": 0.25, + "learning_rate": 1.7540923572710294e-07, + "loss": 0.6975, + "step": 6765 + }, + { + "epoch": 0.25, + "learning_rate": 1.7540557366243087e-07, + "loss": 0.6914, + "step": 6766 + }, + { + "epoch": 0.25, + "learning_rate": 1.7540191159775882e-07, + "loss": 0.6969, + "step": 6767 + }, + { + "epoch": 0.25, + "learning_rate": 1.7539824953308675e-07, + "loss": 0.6902, + "step": 6768 + }, + { + "epoch": 0.25, + "learning_rate": 1.753945874684147e-07, + "loss": 0.7021, + "step": 6769 + }, + { + "epoch": 0.25, + "learning_rate": 1.7539092540374263e-07, + "loss": 0.6899, + "step": 6770 + }, + { + "epoch": 0.25, + "learning_rate": 1.7538726333907058e-07, + "loss": 0.6959, + "step": 6771 + }, + { + "epoch": 0.25, + "learning_rate": 1.753836012743985e-07, + "loss": 0.6932, + "step": 6772 + }, + { + "epoch": 0.25, + "learning_rate": 1.7537993920972646e-07, + "loss": 0.6913, + "step": 6773 + }, + { + "epoch": 0.25, + "learning_rate": 1.7537627714505436e-07, + "loss": 0.6883, + "step": 6774 + }, + { + "epoch": 0.25, + "learning_rate": 1.753726150803823e-07, + "loss": 0.6785, + "step": 6775 + }, + { + "epoch": 0.25, + "learning_rate": 1.7536895301571024e-07, + "loss": 0.6995, + "step": 6776 + }, + { + "epoch": 0.25, + "learning_rate": 1.753652909510382e-07, + "loss": 0.7059, + "step": 6777 + }, + { + "epoch": 0.25, + "learning_rate": 1.7536162888636612e-07, + "loss": 0.6886, + "step": 6778 + }, + { + "epoch": 0.25, + "learning_rate": 1.7535796682169405e-07, + "loss": 0.7008, + "step": 6779 + }, + { + "epoch": 0.25, + "learning_rate": 1.75354304757022e-07, + "loss": 0.6801, + "step": 6780 + }, + { + "epoch": 0.25, + "learning_rate": 1.7535064269234993e-07, + "loss": 0.7082, + "step": 6781 + }, + { + "epoch": 0.25, + "learning_rate": 1.7534698062767788e-07, + "loss": 0.6941, + "step": 6782 + }, + { + "epoch": 0.25, + "learning_rate": 1.753433185630058e-07, + "loss": 0.6974, + "step": 6783 + }, + { + "epoch": 0.25, + "learning_rate": 1.7533965649833376e-07, + "loss": 0.6996, + "step": 6784 + }, + { + "epoch": 0.25, + "learning_rate": 1.7533599443366169e-07, + "loss": 0.6852, + "step": 6785 + }, + { + "epoch": 0.25, + "learning_rate": 1.7533233236898964e-07, + "loss": 0.6971, + "step": 6786 + }, + { + "epoch": 0.25, + "learning_rate": 1.7532867030431757e-07, + "loss": 0.705, + "step": 6787 + }, + { + "epoch": 0.25, + "learning_rate": 1.7532500823964552e-07, + "loss": 0.6997, + "step": 6788 + }, + { + "epoch": 0.25, + "learning_rate": 1.7532134617497345e-07, + "loss": 0.6799, + "step": 6789 + }, + { + "epoch": 0.25, + "learning_rate": 1.753176841103014e-07, + "loss": 0.7034, + "step": 6790 + }, + { + "epoch": 0.25, + "learning_rate": 1.7531402204562933e-07, + "loss": 0.6938, + "step": 6791 + }, + { + "epoch": 0.25, + "learning_rate": 1.7531035998095725e-07, + "loss": 0.7013, + "step": 6792 + }, + { + "epoch": 0.25, + "learning_rate": 1.7530669791628518e-07, + "loss": 0.6898, + "step": 6793 + }, + { + "epoch": 0.25, + "learning_rate": 1.7530303585161313e-07, + "loss": 0.6961, + "step": 6794 + }, + { + "epoch": 0.25, + "learning_rate": 1.7529937378694106e-07, + "loss": 0.7052, + "step": 6795 + }, + { + "epoch": 0.25, + "learning_rate": 1.75295711722269e-07, + "loss": 0.6978, + "step": 6796 + }, + { + "epoch": 0.25, + "learning_rate": 1.7529204965759694e-07, + "loss": 0.6843, + "step": 6797 + }, + { + "epoch": 0.25, + "learning_rate": 1.752883875929249e-07, + "loss": 0.6927, + "step": 6798 + }, + { + "epoch": 0.25, + "learning_rate": 1.7528472552825282e-07, + "loss": 0.6899, + "step": 6799 + }, + { + "epoch": 0.25, + "learning_rate": 1.7528106346358077e-07, + "loss": 0.6946, + "step": 6800 + }, + { + "epoch": 0.25, + "learning_rate": 1.752774013989087e-07, + "loss": 0.671, + "step": 6801 + }, + { + "epoch": 0.25, + "learning_rate": 1.7527373933423662e-07, + "loss": 0.7054, + "step": 6802 + }, + { + "epoch": 0.25, + "learning_rate": 1.7527007726956458e-07, + "loss": 0.6922, + "step": 6803 + }, + { + "epoch": 0.25, + "learning_rate": 1.752664152048925e-07, + "loss": 0.7073, + "step": 6804 + }, + { + "epoch": 0.25, + "learning_rate": 1.7526275314022046e-07, + "loss": 0.7168, + "step": 6805 + }, + { + "epoch": 0.25, + "learning_rate": 1.7525909107554838e-07, + "loss": 0.6886, + "step": 6806 + }, + { + "epoch": 0.25, + "learning_rate": 1.7525542901087634e-07, + "loss": 0.6948, + "step": 6807 + }, + { + "epoch": 0.25, + "learning_rate": 1.7525176694620426e-07, + "loss": 0.697, + "step": 6808 + }, + { + "epoch": 0.25, + "learning_rate": 1.7524810488153222e-07, + "loss": 0.6883, + "step": 6809 + }, + { + "epoch": 0.25, + "learning_rate": 1.7524444281686012e-07, + "loss": 0.6931, + "step": 6810 + }, + { + "epoch": 0.25, + "learning_rate": 1.7524078075218807e-07, + "loss": 0.6935, + "step": 6811 + }, + { + "epoch": 0.25, + "learning_rate": 1.75237118687516e-07, + "loss": 0.7073, + "step": 6812 + }, + { + "epoch": 0.25, + "learning_rate": 1.7523345662284395e-07, + "loss": 0.6913, + "step": 6813 + }, + { + "epoch": 0.25, + "learning_rate": 1.7522979455817188e-07, + "loss": 0.687, + "step": 6814 + }, + { + "epoch": 0.25, + "learning_rate": 1.7522613249349983e-07, + "loss": 0.6987, + "step": 6815 + }, + { + "epoch": 0.25, + "learning_rate": 1.7522247042882776e-07, + "loss": 0.7014, + "step": 6816 + }, + { + "epoch": 0.25, + "learning_rate": 1.752188083641557e-07, + "loss": 0.6976, + "step": 6817 + }, + { + "epoch": 0.25, + "learning_rate": 1.7521514629948364e-07, + "loss": 0.6726, + "step": 6818 + }, + { + "epoch": 0.25, + "learning_rate": 1.752114842348116e-07, + "loss": 0.7015, + "step": 6819 + }, + { + "epoch": 0.25, + "learning_rate": 1.7520782217013952e-07, + "loss": 0.6939, + "step": 6820 + }, + { + "epoch": 0.25, + "learning_rate": 1.7520416010546747e-07, + "loss": 0.7195, + "step": 6821 + }, + { + "epoch": 0.25, + "learning_rate": 1.752004980407954e-07, + "loss": 0.6891, + "step": 6822 + }, + { + "epoch": 0.25, + "learning_rate": 1.7519683597612332e-07, + "loss": 0.6876, + "step": 6823 + }, + { + "epoch": 0.25, + "learning_rate": 1.7519317391145128e-07, + "loss": 0.6905, + "step": 6824 + }, + { + "epoch": 0.25, + "learning_rate": 1.751895118467792e-07, + "loss": 0.7037, + "step": 6825 + }, + { + "epoch": 0.25, + "learning_rate": 1.7518584978210716e-07, + "loss": 0.6981, + "step": 6826 + }, + { + "epoch": 0.25, + "learning_rate": 1.7518218771743508e-07, + "loss": 0.6965, + "step": 6827 + }, + { + "epoch": 0.25, + "learning_rate": 1.75178525652763e-07, + "loss": 0.6974, + "step": 6828 + }, + { + "epoch": 0.25, + "learning_rate": 1.7517486358809094e-07, + "loss": 0.7024, + "step": 6829 + }, + { + "epoch": 0.25, + "learning_rate": 1.751712015234189e-07, + "loss": 0.6915, + "step": 6830 + }, + { + "epoch": 0.25, + "learning_rate": 1.7516753945874682e-07, + "loss": 0.6896, + "step": 6831 + }, + { + "epoch": 0.25, + "learning_rate": 1.7516387739407477e-07, + "loss": 0.689, + "step": 6832 + }, + { + "epoch": 0.25, + "learning_rate": 1.751602153294027e-07, + "loss": 0.7116, + "step": 6833 + }, + { + "epoch": 0.25, + "learning_rate": 1.7515655326473065e-07, + "loss": 0.6797, + "step": 6834 + }, + { + "epoch": 0.25, + "learning_rate": 1.7515289120005858e-07, + "loss": 0.6809, + "step": 6835 + }, + { + "epoch": 0.25, + "learning_rate": 1.7514922913538653e-07, + "loss": 0.6951, + "step": 6836 + }, + { + "epoch": 0.25, + "learning_rate": 1.7514556707071446e-07, + "loss": 0.7078, + "step": 6837 + }, + { + "epoch": 0.25, + "learning_rate": 1.751419050060424e-07, + "loss": 0.6862, + "step": 6838 + }, + { + "epoch": 0.25, + "learning_rate": 1.7513824294137034e-07, + "loss": 0.6919, + "step": 6839 + }, + { + "epoch": 0.25, + "learning_rate": 1.751345808766983e-07, + "loss": 0.6983, + "step": 6840 + }, + { + "epoch": 0.25, + "learning_rate": 1.7513091881202622e-07, + "loss": 0.6916, + "step": 6841 + }, + { + "epoch": 0.25, + "learning_rate": 1.7512725674735417e-07, + "loss": 0.6974, + "step": 6842 + }, + { + "epoch": 0.25, + "learning_rate": 1.751235946826821e-07, + "loss": 0.6924, + "step": 6843 + }, + { + "epoch": 0.25, + "learning_rate": 1.7511993261801005e-07, + "loss": 0.6877, + "step": 6844 + }, + { + "epoch": 0.25, + "learning_rate": 1.7511627055333798e-07, + "loss": 0.675, + "step": 6845 + }, + { + "epoch": 0.25, + "learning_rate": 1.751126084886659e-07, + "loss": 0.6994, + "step": 6846 + }, + { + "epoch": 0.25, + "learning_rate": 1.7510894642399383e-07, + "loss": 0.7066, + "step": 6847 + }, + { + "epoch": 0.25, + "learning_rate": 1.7510528435932178e-07, + "loss": 0.7225, + "step": 6848 + }, + { + "epoch": 0.25, + "learning_rate": 1.751016222946497e-07, + "loss": 0.6799, + "step": 6849 + }, + { + "epoch": 0.25, + "learning_rate": 1.7509796022997764e-07, + "loss": 0.685, + "step": 6850 + }, + { + "epoch": 0.25, + "learning_rate": 1.750942981653056e-07, + "loss": 0.691, + "step": 6851 + }, + { + "epoch": 0.25, + "learning_rate": 1.7509063610063352e-07, + "loss": 0.6918, + "step": 6852 + }, + { + "epoch": 0.25, + "learning_rate": 1.7508697403596147e-07, + "loss": 0.6964, + "step": 6853 + }, + { + "epoch": 0.25, + "learning_rate": 1.750833119712894e-07, + "loss": 0.7089, + "step": 6854 + }, + { + "epoch": 0.25, + "learning_rate": 1.7507964990661735e-07, + "loss": 0.6871, + "step": 6855 + }, + { + "epoch": 0.25, + "learning_rate": 1.7507598784194527e-07, + "loss": 0.679, + "step": 6856 + }, + { + "epoch": 0.25, + "learning_rate": 1.7507232577727323e-07, + "loss": 0.6769, + "step": 6857 + }, + { + "epoch": 0.25, + "learning_rate": 1.7506866371260115e-07, + "loss": 0.7, + "step": 6858 + }, + { + "epoch": 0.25, + "learning_rate": 1.750650016479291e-07, + "loss": 0.6828, + "step": 6859 + }, + { + "epoch": 0.25, + "learning_rate": 1.7506133958325703e-07, + "loss": 0.6947, + "step": 6860 + }, + { + "epoch": 0.25, + "learning_rate": 1.75057677518585e-07, + "loss": 0.6894, + "step": 6861 + }, + { + "epoch": 0.25, + "learning_rate": 1.7505401545391291e-07, + "loss": 0.6997, + "step": 6862 + }, + { + "epoch": 0.25, + "learning_rate": 1.7505035338924087e-07, + "loss": 0.7129, + "step": 6863 + }, + { + "epoch": 0.25, + "learning_rate": 1.7504669132456877e-07, + "loss": 0.7041, + "step": 6864 + }, + { + "epoch": 0.25, + "learning_rate": 1.7504302925989672e-07, + "loss": 0.6944, + "step": 6865 + }, + { + "epoch": 0.25, + "learning_rate": 1.7503936719522465e-07, + "loss": 0.688, + "step": 6866 + }, + { + "epoch": 0.25, + "learning_rate": 1.750357051305526e-07, + "loss": 0.6763, + "step": 6867 + }, + { + "epoch": 0.25, + "learning_rate": 1.7503204306588053e-07, + "loss": 0.6779, + "step": 6868 + }, + { + "epoch": 0.25, + "learning_rate": 1.7502838100120848e-07, + "loss": 0.6899, + "step": 6869 + }, + { + "epoch": 0.25, + "learning_rate": 1.750247189365364e-07, + "loss": 0.6851, + "step": 6870 + }, + { + "epoch": 0.25, + "learning_rate": 1.7502105687186436e-07, + "loss": 0.6938, + "step": 6871 + }, + { + "epoch": 0.25, + "learning_rate": 1.750173948071923e-07, + "loss": 0.7097, + "step": 6872 + }, + { + "epoch": 0.25, + "learning_rate": 1.7501373274252021e-07, + "loss": 0.6977, + "step": 6873 + }, + { + "epoch": 0.25, + "learning_rate": 1.7501007067784817e-07, + "loss": 0.7019, + "step": 6874 + }, + { + "epoch": 0.25, + "learning_rate": 1.750064086131761e-07, + "loss": 0.6857, + "step": 6875 + }, + { + "epoch": 0.25, + "learning_rate": 1.7500274654850405e-07, + "loss": 0.6899, + "step": 6876 + }, + { + "epoch": 0.25, + "learning_rate": 1.7499908448383197e-07, + "loss": 0.6795, + "step": 6877 + }, + { + "epoch": 0.25, + "learning_rate": 1.7499542241915993e-07, + "loss": 0.7004, + "step": 6878 + }, + { + "epoch": 0.25, + "learning_rate": 1.7499176035448785e-07, + "loss": 0.7144, + "step": 6879 + }, + { + "epoch": 0.25, + "learning_rate": 1.749880982898158e-07, + "loss": 0.6903, + "step": 6880 + }, + { + "epoch": 0.25, + "learning_rate": 1.7498443622514373e-07, + "loss": 0.6957, + "step": 6881 + }, + { + "epoch": 0.25, + "learning_rate": 1.7498077416047166e-07, + "loss": 0.6887, + "step": 6882 + }, + { + "epoch": 0.25, + "learning_rate": 1.749771120957996e-07, + "loss": 0.6991, + "step": 6883 + }, + { + "epoch": 0.25, + "learning_rate": 1.7497345003112754e-07, + "loss": 0.7041, + "step": 6884 + }, + { + "epoch": 0.25, + "learning_rate": 1.7496978796645547e-07, + "loss": 0.6954, + "step": 6885 + }, + { + "epoch": 0.25, + "learning_rate": 1.7496612590178342e-07, + "loss": 0.7041, + "step": 6886 + }, + { + "epoch": 0.25, + "learning_rate": 1.7496246383711135e-07, + "loss": 0.6939, + "step": 6887 + }, + { + "epoch": 0.25, + "learning_rate": 1.749588017724393e-07, + "loss": 0.6912, + "step": 6888 + }, + { + "epoch": 0.25, + "learning_rate": 1.7495513970776723e-07, + "loss": 0.706, + "step": 6889 + }, + { + "epoch": 0.25, + "learning_rate": 1.7495147764309518e-07, + "loss": 0.6752, + "step": 6890 + }, + { + "epoch": 0.25, + "learning_rate": 1.749478155784231e-07, + "loss": 0.6872, + "step": 6891 + }, + { + "epoch": 0.25, + "learning_rate": 1.7494415351375106e-07, + "loss": 0.7056, + "step": 6892 + }, + { + "epoch": 0.25, + "learning_rate": 1.7494049144907899e-07, + "loss": 0.7052, + "step": 6893 + }, + { + "epoch": 0.25, + "learning_rate": 1.749368293844069e-07, + "loss": 0.6845, + "step": 6894 + }, + { + "epoch": 0.25, + "learning_rate": 1.7493316731973487e-07, + "loss": 0.6965, + "step": 6895 + }, + { + "epoch": 0.25, + "learning_rate": 1.749295052550628e-07, + "loss": 0.6919, + "step": 6896 + }, + { + "epoch": 0.25, + "learning_rate": 1.7492584319039075e-07, + "loss": 0.6841, + "step": 6897 + }, + { + "epoch": 0.25, + "learning_rate": 1.7492218112571867e-07, + "loss": 0.6907, + "step": 6898 + }, + { + "epoch": 0.25, + "learning_rate": 1.7491851906104663e-07, + "loss": 0.6835, + "step": 6899 + }, + { + "epoch": 0.25, + "learning_rate": 1.7491485699637453e-07, + "loss": 0.6808, + "step": 6900 + }, + { + "epoch": 0.25, + "learning_rate": 1.7491119493170248e-07, + "loss": 0.6918, + "step": 6901 + }, + { + "epoch": 0.25, + "learning_rate": 1.749075328670304e-07, + "loss": 0.685, + "step": 6902 + }, + { + "epoch": 0.25, + "learning_rate": 1.7490387080235836e-07, + "loss": 0.6941, + "step": 6903 + }, + { + "epoch": 0.25, + "learning_rate": 1.7490020873768629e-07, + "loss": 0.6856, + "step": 6904 + }, + { + "epoch": 0.25, + "learning_rate": 1.7489654667301424e-07, + "loss": 0.6784, + "step": 6905 + }, + { + "epoch": 0.25, + "learning_rate": 1.7489288460834217e-07, + "loss": 0.6878, + "step": 6906 + }, + { + "epoch": 0.25, + "learning_rate": 1.7488922254367012e-07, + "loss": 0.6927, + "step": 6907 + }, + { + "epoch": 0.25, + "learning_rate": 1.7488556047899805e-07, + "loss": 0.6826, + "step": 6908 + }, + { + "epoch": 0.25, + "learning_rate": 1.74881898414326e-07, + "loss": 0.7041, + "step": 6909 + }, + { + "epoch": 0.25, + "learning_rate": 1.7487823634965393e-07, + "loss": 0.6993, + "step": 6910 + }, + { + "epoch": 0.25, + "learning_rate": 1.7487457428498188e-07, + "loss": 0.691, + "step": 6911 + }, + { + "epoch": 0.25, + "learning_rate": 1.748709122203098e-07, + "loss": 0.6956, + "step": 6912 + }, + { + "epoch": 0.25, + "learning_rate": 1.7486725015563776e-07, + "loss": 0.6785, + "step": 6913 + }, + { + "epoch": 0.25, + "learning_rate": 1.7486358809096568e-07, + "loss": 0.6874, + "step": 6914 + }, + { + "epoch": 0.25, + "learning_rate": 1.7485992602629364e-07, + "loss": 0.6905, + "step": 6915 + }, + { + "epoch": 0.25, + "learning_rate": 1.7485626396162156e-07, + "loss": 0.7069, + "step": 6916 + }, + { + "epoch": 0.25, + "learning_rate": 1.748526018969495e-07, + "loss": 0.6793, + "step": 6917 + }, + { + "epoch": 0.25, + "learning_rate": 1.7484893983227742e-07, + "loss": 0.7092, + "step": 6918 + }, + { + "epoch": 0.25, + "learning_rate": 1.7484527776760537e-07, + "loss": 0.6882, + "step": 6919 + }, + { + "epoch": 0.25, + "learning_rate": 1.748416157029333e-07, + "loss": 0.6971, + "step": 6920 + }, + { + "epoch": 0.25, + "learning_rate": 1.7483795363826122e-07, + "loss": 0.6876, + "step": 6921 + }, + { + "epoch": 0.25, + "learning_rate": 1.7483429157358918e-07, + "loss": 0.723, + "step": 6922 + }, + { + "epoch": 0.25, + "learning_rate": 1.748306295089171e-07, + "loss": 0.6923, + "step": 6923 + }, + { + "epoch": 0.25, + "learning_rate": 1.7482696744424506e-07, + "loss": 0.7049, + "step": 6924 + }, + { + "epoch": 0.25, + "learning_rate": 1.7482330537957298e-07, + "loss": 0.7068, + "step": 6925 + }, + { + "epoch": 0.25, + "learning_rate": 1.7481964331490094e-07, + "loss": 0.6859, + "step": 6926 + }, + { + "epoch": 0.25, + "learning_rate": 1.7481598125022886e-07, + "loss": 0.7007, + "step": 6927 + }, + { + "epoch": 0.25, + "learning_rate": 1.7481231918555682e-07, + "loss": 0.6887, + "step": 6928 + }, + { + "epoch": 0.25, + "learning_rate": 1.7480865712088474e-07, + "loss": 0.684, + "step": 6929 + }, + { + "epoch": 0.25, + "learning_rate": 1.748049950562127e-07, + "loss": 0.6922, + "step": 6930 + }, + { + "epoch": 0.25, + "learning_rate": 1.7480133299154062e-07, + "loss": 0.6928, + "step": 6931 + }, + { + "epoch": 0.25, + "learning_rate": 1.7479767092686858e-07, + "loss": 0.6909, + "step": 6932 + }, + { + "epoch": 0.25, + "learning_rate": 1.747940088621965e-07, + "loss": 0.695, + "step": 6933 + }, + { + "epoch": 0.25, + "learning_rate": 1.7479034679752446e-07, + "loss": 0.687, + "step": 6934 + }, + { + "epoch": 0.25, + "learning_rate": 1.7478668473285238e-07, + "loss": 0.6941, + "step": 6935 + }, + { + "epoch": 0.25, + "learning_rate": 1.747830226681803e-07, + "loss": 0.6886, + "step": 6936 + }, + { + "epoch": 0.25, + "learning_rate": 1.7477936060350824e-07, + "loss": 0.6989, + "step": 6937 + }, + { + "epoch": 0.25, + "learning_rate": 1.747756985388362e-07, + "loss": 0.676, + "step": 6938 + }, + { + "epoch": 0.25, + "learning_rate": 1.7477203647416412e-07, + "loss": 0.6834, + "step": 6939 + }, + { + "epoch": 0.25, + "learning_rate": 1.7476837440949207e-07, + "loss": 0.7058, + "step": 6940 + }, + { + "epoch": 0.25, + "learning_rate": 1.7476471234482e-07, + "loss": 0.7036, + "step": 6941 + }, + { + "epoch": 0.25, + "learning_rate": 1.7476105028014795e-07, + "loss": 0.6891, + "step": 6942 + }, + { + "epoch": 0.25, + "learning_rate": 1.7475738821547588e-07, + "loss": 0.6932, + "step": 6943 + }, + { + "epoch": 0.25, + "learning_rate": 1.747537261508038e-07, + "loss": 0.708, + "step": 6944 + }, + { + "epoch": 0.25, + "learning_rate": 1.7475006408613176e-07, + "loss": 0.6935, + "step": 6945 + }, + { + "epoch": 0.25, + "learning_rate": 1.7474640202145968e-07, + "loss": 0.6928, + "step": 6946 + }, + { + "epoch": 0.25, + "learning_rate": 1.7474273995678764e-07, + "loss": 0.6854, + "step": 6947 + }, + { + "epoch": 0.25, + "learning_rate": 1.7473907789211556e-07, + "loss": 0.6916, + "step": 6948 + }, + { + "epoch": 0.25, + "learning_rate": 1.7473541582744352e-07, + "loss": 0.709, + "step": 6949 + }, + { + "epoch": 0.25, + "learning_rate": 1.7473175376277144e-07, + "loss": 0.6876, + "step": 6950 + }, + { + "epoch": 0.25, + "learning_rate": 1.747280916980994e-07, + "loss": 0.6991, + "step": 6951 + }, + { + "epoch": 0.25, + "learning_rate": 1.7472442963342732e-07, + "loss": 0.705, + "step": 6952 + }, + { + "epoch": 0.25, + "learning_rate": 1.7472076756875528e-07, + "loss": 0.7041, + "step": 6953 + }, + { + "epoch": 0.25, + "learning_rate": 1.7471710550408318e-07, + "loss": 0.6943, + "step": 6954 + }, + { + "epoch": 0.25, + "learning_rate": 1.7471344343941113e-07, + "loss": 0.687, + "step": 6955 + }, + { + "epoch": 0.25, + "learning_rate": 1.7470978137473906e-07, + "loss": 0.6956, + "step": 6956 + }, + { + "epoch": 0.25, + "learning_rate": 1.74706119310067e-07, + "loss": 0.6955, + "step": 6957 + }, + { + "epoch": 0.25, + "learning_rate": 1.7470245724539494e-07, + "loss": 0.68, + "step": 6958 + }, + { + "epoch": 0.25, + "learning_rate": 1.746987951807229e-07, + "loss": 0.681, + "step": 6959 + }, + { + "epoch": 0.25, + "learning_rate": 1.7469513311605082e-07, + "loss": 0.6824, + "step": 6960 + }, + { + "epoch": 0.25, + "learning_rate": 1.7469147105137877e-07, + "loss": 0.6984, + "step": 6961 + }, + { + "epoch": 0.25, + "learning_rate": 1.746878089867067e-07, + "loss": 0.6914, + "step": 6962 + }, + { + "epoch": 0.25, + "learning_rate": 1.7468414692203465e-07, + "loss": 0.6835, + "step": 6963 + }, + { + "epoch": 0.25, + "learning_rate": 1.7468048485736258e-07, + "loss": 0.6876, + "step": 6964 + }, + { + "epoch": 0.25, + "learning_rate": 1.746768227926905e-07, + "loss": 0.6986, + "step": 6965 + }, + { + "epoch": 0.25, + "learning_rate": 1.7467316072801846e-07, + "loss": 0.6734, + "step": 6966 + }, + { + "epoch": 0.25, + "learning_rate": 1.7466949866334638e-07, + "loss": 0.7042, + "step": 6967 + }, + { + "epoch": 0.25, + "learning_rate": 1.7466583659867434e-07, + "loss": 0.6985, + "step": 6968 + }, + { + "epoch": 0.25, + "learning_rate": 1.7466217453400226e-07, + "loss": 0.686, + "step": 6969 + }, + { + "epoch": 0.26, + "learning_rate": 1.7465851246933022e-07, + "loss": 0.6989, + "step": 6970 + }, + { + "epoch": 0.26, + "learning_rate": 1.7465485040465814e-07, + "loss": 0.6999, + "step": 6971 + }, + { + "epoch": 0.26, + "learning_rate": 1.7465118833998607e-07, + "loss": 0.6734, + "step": 6972 + }, + { + "epoch": 0.26, + "learning_rate": 1.74647526275314e-07, + "loss": 0.6921, + "step": 6973 + }, + { + "epoch": 0.26, + "learning_rate": 1.7464386421064195e-07, + "loss": 0.706, + "step": 6974 + }, + { + "epoch": 0.26, + "learning_rate": 1.7464020214596988e-07, + "loss": 0.6867, + "step": 6975 + }, + { + "epoch": 0.26, + "learning_rate": 1.7463654008129783e-07, + "loss": 0.6968, + "step": 6976 + }, + { + "epoch": 0.26, + "learning_rate": 1.7463287801662575e-07, + "loss": 0.6932, + "step": 6977 + }, + { + "epoch": 0.26, + "learning_rate": 1.746292159519537e-07, + "loss": 0.704, + "step": 6978 + }, + { + "epoch": 0.26, + "learning_rate": 1.7462555388728163e-07, + "loss": 0.7101, + "step": 6979 + }, + { + "epoch": 0.26, + "learning_rate": 1.746218918226096e-07, + "loss": 0.6816, + "step": 6980 + }, + { + "epoch": 0.26, + "learning_rate": 1.7461822975793751e-07, + "loss": 0.6931, + "step": 6981 + }, + { + "epoch": 0.26, + "learning_rate": 1.7461456769326547e-07, + "loss": 0.7023, + "step": 6982 + }, + { + "epoch": 0.26, + "learning_rate": 1.746109056285934e-07, + "loss": 0.6914, + "step": 6983 + }, + { + "epoch": 0.26, + "learning_rate": 1.7460724356392135e-07, + "loss": 0.6912, + "step": 6984 + }, + { + "epoch": 0.26, + "learning_rate": 1.7460358149924927e-07, + "loss": 0.6945, + "step": 6985 + }, + { + "epoch": 0.26, + "learning_rate": 1.745999194345772e-07, + "loss": 0.6786, + "step": 6986 + }, + { + "epoch": 0.26, + "learning_rate": 1.7459625736990515e-07, + "loss": 0.6997, + "step": 6987 + }, + { + "epoch": 0.26, + "learning_rate": 1.7459259530523308e-07, + "loss": 0.6951, + "step": 6988 + }, + { + "epoch": 0.26, + "learning_rate": 1.74588933240561e-07, + "loss": 0.7013, + "step": 6989 + }, + { + "epoch": 0.26, + "learning_rate": 1.7458527117588896e-07, + "loss": 0.691, + "step": 6990 + }, + { + "epoch": 0.26, + "learning_rate": 1.745816091112169e-07, + "loss": 0.6999, + "step": 6991 + }, + { + "epoch": 0.26, + "learning_rate": 1.7457794704654481e-07, + "loss": 0.6751, + "step": 6992 + }, + { + "epoch": 0.26, + "learning_rate": 1.7457428498187277e-07, + "loss": 0.6876, + "step": 6993 + }, + { + "epoch": 0.26, + "learning_rate": 1.745706229172007e-07, + "loss": 0.6817, + "step": 6994 + }, + { + "epoch": 0.26, + "learning_rate": 1.7456696085252865e-07, + "loss": 0.6854, + "step": 6995 + }, + { + "epoch": 0.26, + "learning_rate": 1.7456329878785657e-07, + "loss": 0.6995, + "step": 6996 + }, + { + "epoch": 0.26, + "learning_rate": 1.7455963672318453e-07, + "loss": 0.7073, + "step": 6997 + }, + { + "epoch": 0.26, + "learning_rate": 1.7455597465851245e-07, + "loss": 0.6994, + "step": 6998 + }, + { + "epoch": 0.26, + "learning_rate": 1.745523125938404e-07, + "loss": 0.6972, + "step": 6999 + }, + { + "epoch": 0.26, + "learning_rate": 1.7454865052916833e-07, + "loss": 0.7022, + "step": 7000 + }, + { + "epoch": 0.26, + "learning_rate": 1.7454498846449629e-07, + "loss": 0.6968, + "step": 7001 + }, + { + "epoch": 0.26, + "learning_rate": 1.7454132639982421e-07, + "loss": 0.696, + "step": 7002 + }, + { + "epoch": 0.26, + "learning_rate": 1.7453766433515217e-07, + "loss": 0.7082, + "step": 7003 + }, + { + "epoch": 0.26, + "learning_rate": 1.745340022704801e-07, + "loss": 0.6918, + "step": 7004 + }, + { + "epoch": 0.26, + "learning_rate": 1.7453034020580805e-07, + "loss": 0.6834, + "step": 7005 + }, + { + "epoch": 0.26, + "learning_rate": 1.7452667814113597e-07, + "loss": 0.6731, + "step": 7006 + }, + { + "epoch": 0.26, + "learning_rate": 1.745230160764639e-07, + "loss": 0.7099, + "step": 7007 + }, + { + "epoch": 0.26, + "learning_rate": 1.7451935401179183e-07, + "loss": 0.6987, + "step": 7008 + }, + { + "epoch": 0.26, + "learning_rate": 1.7451569194711978e-07, + "loss": 0.6886, + "step": 7009 + }, + { + "epoch": 0.26, + "learning_rate": 1.745120298824477e-07, + "loss": 0.7152, + "step": 7010 + }, + { + "epoch": 0.26, + "learning_rate": 1.7450836781777566e-07, + "loss": 0.6937, + "step": 7011 + }, + { + "epoch": 0.26, + "learning_rate": 1.7450470575310359e-07, + "loss": 0.7032, + "step": 7012 + }, + { + "epoch": 0.26, + "learning_rate": 1.7450104368843154e-07, + "loss": 0.681, + "step": 7013 + }, + { + "epoch": 0.26, + "learning_rate": 1.7449738162375947e-07, + "loss": 0.6972, + "step": 7014 + }, + { + "epoch": 0.26, + "learning_rate": 1.744937195590874e-07, + "loss": 0.6902, + "step": 7015 + }, + { + "epoch": 0.26, + "learning_rate": 1.7449005749441535e-07, + "loss": 0.6902, + "step": 7016 + }, + { + "epoch": 0.26, + "learning_rate": 1.7448639542974327e-07, + "loss": 0.6887, + "step": 7017 + }, + { + "epoch": 0.26, + "learning_rate": 1.7448273336507123e-07, + "loss": 0.6886, + "step": 7018 + }, + { + "epoch": 0.26, + "learning_rate": 1.7447907130039915e-07, + "loss": 0.6861, + "step": 7019 + }, + { + "epoch": 0.26, + "learning_rate": 1.744754092357271e-07, + "loss": 0.6802, + "step": 7020 + }, + { + "epoch": 0.26, + "learning_rate": 1.7447174717105503e-07, + "loss": 0.6924, + "step": 7021 + }, + { + "epoch": 0.26, + "learning_rate": 1.7446808510638299e-07, + "loss": 0.6807, + "step": 7022 + }, + { + "epoch": 0.26, + "learning_rate": 1.744644230417109e-07, + "loss": 0.6775, + "step": 7023 + }, + { + "epoch": 0.26, + "learning_rate": 1.7446076097703887e-07, + "loss": 0.6982, + "step": 7024 + }, + { + "epoch": 0.26, + "learning_rate": 1.7445709891236677e-07, + "loss": 0.6883, + "step": 7025 + }, + { + "epoch": 0.26, + "learning_rate": 1.7445343684769472e-07, + "loss": 0.7054, + "step": 7026 + }, + { + "epoch": 0.26, + "learning_rate": 1.7444977478302265e-07, + "loss": 0.6993, + "step": 7027 + }, + { + "epoch": 0.26, + "learning_rate": 1.744461127183506e-07, + "loss": 0.6848, + "step": 7028 + }, + { + "epoch": 0.26, + "learning_rate": 1.7444245065367853e-07, + "loss": 0.6992, + "step": 7029 + }, + { + "epoch": 0.26, + "learning_rate": 1.7443878858900648e-07, + "loss": 0.7155, + "step": 7030 + }, + { + "epoch": 0.26, + "learning_rate": 1.744351265243344e-07, + "loss": 0.6758, + "step": 7031 + }, + { + "epoch": 0.26, + "learning_rate": 1.7443146445966236e-07, + "loss": 0.6991, + "step": 7032 + }, + { + "epoch": 0.26, + "learning_rate": 1.7442780239499028e-07, + "loss": 0.6863, + "step": 7033 + }, + { + "epoch": 0.26, + "learning_rate": 1.7442414033031824e-07, + "loss": 0.6966, + "step": 7034 + }, + { + "epoch": 0.26, + "learning_rate": 1.7442047826564616e-07, + "loss": 0.7043, + "step": 7035 + }, + { + "epoch": 0.26, + "learning_rate": 1.744168162009741e-07, + "loss": 0.6765, + "step": 7036 + }, + { + "epoch": 0.26, + "learning_rate": 1.7441315413630204e-07, + "loss": 0.705, + "step": 7037 + }, + { + "epoch": 0.26, + "learning_rate": 1.7440949207162997e-07, + "loss": 0.6891, + "step": 7038 + }, + { + "epoch": 0.26, + "learning_rate": 1.7440583000695792e-07, + "loss": 0.694, + "step": 7039 + }, + { + "epoch": 0.26, + "learning_rate": 1.7440216794228585e-07, + "loss": 0.6838, + "step": 7040 + }, + { + "epoch": 0.26, + "learning_rate": 1.743985058776138e-07, + "loss": 0.6782, + "step": 7041 + }, + { + "epoch": 0.26, + "learning_rate": 1.7439484381294173e-07, + "loss": 0.6854, + "step": 7042 + }, + { + "epoch": 0.26, + "learning_rate": 1.7439118174826966e-07, + "loss": 0.6785, + "step": 7043 + }, + { + "epoch": 0.26, + "learning_rate": 1.7438751968359758e-07, + "loss": 0.6933, + "step": 7044 + }, + { + "epoch": 0.26, + "learning_rate": 1.7438385761892554e-07, + "loss": 0.6974, + "step": 7045 + }, + { + "epoch": 0.26, + "learning_rate": 1.7438019555425346e-07, + "loss": 0.7075, + "step": 7046 + }, + { + "epoch": 0.26, + "learning_rate": 1.7437653348958142e-07, + "loss": 0.6715, + "step": 7047 + }, + { + "epoch": 0.26, + "learning_rate": 1.7437287142490934e-07, + "loss": 0.6919, + "step": 7048 + }, + { + "epoch": 0.26, + "learning_rate": 1.743692093602373e-07, + "loss": 0.6933, + "step": 7049 + }, + { + "epoch": 0.26, + "learning_rate": 1.7436554729556522e-07, + "loss": 0.6927, + "step": 7050 + }, + { + "epoch": 0.26, + "learning_rate": 1.7436188523089318e-07, + "loss": 0.7116, + "step": 7051 + }, + { + "epoch": 0.26, + "learning_rate": 1.743582231662211e-07, + "loss": 0.7038, + "step": 7052 + }, + { + "epoch": 0.26, + "learning_rate": 1.7435456110154906e-07, + "loss": 0.7081, + "step": 7053 + }, + { + "epoch": 0.26, + "learning_rate": 1.7435089903687698e-07, + "loss": 0.689, + "step": 7054 + }, + { + "epoch": 0.26, + "learning_rate": 1.7434723697220494e-07, + "loss": 0.6894, + "step": 7055 + }, + { + "epoch": 0.26, + "learning_rate": 1.7434357490753286e-07, + "loss": 0.6886, + "step": 7056 + }, + { + "epoch": 0.26, + "learning_rate": 1.743399128428608e-07, + "loss": 0.6964, + "step": 7057 + }, + { + "epoch": 0.26, + "learning_rate": 1.7433625077818874e-07, + "loss": 0.6787, + "step": 7058 + }, + { + "epoch": 0.26, + "learning_rate": 1.7433258871351667e-07, + "loss": 0.6832, + "step": 7059 + }, + { + "epoch": 0.26, + "learning_rate": 1.7432892664884462e-07, + "loss": 0.6916, + "step": 7060 + }, + { + "epoch": 0.26, + "learning_rate": 1.7432526458417255e-07, + "loss": 0.6763, + "step": 7061 + }, + { + "epoch": 0.26, + "learning_rate": 1.7432160251950048e-07, + "loss": 0.7045, + "step": 7062 + }, + { + "epoch": 0.26, + "learning_rate": 1.743179404548284e-07, + "loss": 0.6969, + "step": 7063 + }, + { + "epoch": 0.26, + "learning_rate": 1.7431427839015636e-07, + "loss": 0.7022, + "step": 7064 + }, + { + "epoch": 0.26, + "learning_rate": 1.7431061632548428e-07, + "loss": 0.6842, + "step": 7065 + }, + { + "epoch": 0.26, + "learning_rate": 1.7430695426081224e-07, + "loss": 0.7075, + "step": 7066 + }, + { + "epoch": 0.26, + "learning_rate": 1.7430329219614016e-07, + "loss": 0.6992, + "step": 7067 + }, + { + "epoch": 0.26, + "learning_rate": 1.7429963013146812e-07, + "loss": 0.689, + "step": 7068 + }, + { + "epoch": 0.26, + "learning_rate": 1.7429596806679604e-07, + "loss": 0.6899, + "step": 7069 + }, + { + "epoch": 0.26, + "learning_rate": 1.74292306002124e-07, + "loss": 0.7054, + "step": 7070 + }, + { + "epoch": 0.26, + "learning_rate": 1.7428864393745192e-07, + "loss": 0.6942, + "step": 7071 + }, + { + "epoch": 0.26, + "learning_rate": 1.7428498187277988e-07, + "loss": 0.6921, + "step": 7072 + }, + { + "epoch": 0.26, + "learning_rate": 1.742813198081078e-07, + "loss": 0.7104, + "step": 7073 + }, + { + "epoch": 0.26, + "learning_rate": 1.7427765774343576e-07, + "loss": 0.708, + "step": 7074 + }, + { + "epoch": 0.26, + "learning_rate": 1.7427399567876368e-07, + "loss": 0.6821, + "step": 7075 + }, + { + "epoch": 0.26, + "learning_rate": 1.7427033361409164e-07, + "loss": 0.6958, + "step": 7076 + }, + { + "epoch": 0.26, + "learning_rate": 1.7426667154941956e-07, + "loss": 0.6998, + "step": 7077 + }, + { + "epoch": 0.26, + "learning_rate": 1.7426300948474752e-07, + "loss": 0.6826, + "step": 7078 + }, + { + "epoch": 0.26, + "learning_rate": 1.7425934742007542e-07, + "loss": 0.6989, + "step": 7079 + }, + { + "epoch": 0.26, + "learning_rate": 1.7425568535540337e-07, + "loss": 0.6877, + "step": 7080 + }, + { + "epoch": 0.26, + "learning_rate": 1.742520232907313e-07, + "loss": 0.6975, + "step": 7081 + }, + { + "epoch": 0.26, + "learning_rate": 1.7424836122605925e-07, + "loss": 0.6954, + "step": 7082 + }, + { + "epoch": 0.26, + "learning_rate": 1.7424469916138718e-07, + "loss": 0.6995, + "step": 7083 + }, + { + "epoch": 0.26, + "learning_rate": 1.7424103709671513e-07, + "loss": 0.6875, + "step": 7084 + }, + { + "epoch": 0.26, + "learning_rate": 1.7423737503204306e-07, + "loss": 0.6854, + "step": 7085 + }, + { + "epoch": 0.26, + "learning_rate": 1.7423371296737098e-07, + "loss": 0.6961, + "step": 7086 + }, + { + "epoch": 0.26, + "learning_rate": 1.7423005090269894e-07, + "loss": 0.693, + "step": 7087 + }, + { + "epoch": 0.26, + "learning_rate": 1.7422638883802686e-07, + "loss": 0.6978, + "step": 7088 + }, + { + "epoch": 0.26, + "learning_rate": 1.7422272677335482e-07, + "loss": 0.6863, + "step": 7089 + }, + { + "epoch": 0.26, + "learning_rate": 1.7421906470868274e-07, + "loss": 0.7073, + "step": 7090 + }, + { + "epoch": 0.26, + "learning_rate": 1.742154026440107e-07, + "loss": 0.6965, + "step": 7091 + }, + { + "epoch": 0.26, + "learning_rate": 1.7421174057933862e-07, + "loss": 0.7122, + "step": 7092 + }, + { + "epoch": 0.26, + "learning_rate": 1.7420807851466657e-07, + "loss": 0.7016, + "step": 7093 + }, + { + "epoch": 0.26, + "learning_rate": 1.742044164499945e-07, + "loss": 0.675, + "step": 7094 + }, + { + "epoch": 0.26, + "learning_rate": 1.7420075438532245e-07, + "loss": 0.7033, + "step": 7095 + }, + { + "epoch": 0.26, + "learning_rate": 1.7419709232065038e-07, + "loss": 0.6923, + "step": 7096 + }, + { + "epoch": 0.26, + "learning_rate": 1.741934302559783e-07, + "loss": 0.6895, + "step": 7097 + }, + { + "epoch": 0.26, + "learning_rate": 1.7418976819130623e-07, + "loss": 0.7039, + "step": 7098 + }, + { + "epoch": 0.26, + "learning_rate": 1.741861061266342e-07, + "loss": 0.6898, + "step": 7099 + }, + { + "epoch": 0.26, + "learning_rate": 1.7418244406196211e-07, + "loss": 0.7046, + "step": 7100 + }, + { + "epoch": 0.26, + "learning_rate": 1.7417878199729007e-07, + "loss": 0.7059, + "step": 7101 + }, + { + "epoch": 0.26, + "learning_rate": 1.74175119932618e-07, + "loss": 0.6806, + "step": 7102 + }, + { + "epoch": 0.26, + "learning_rate": 1.7417145786794595e-07, + "loss": 0.6991, + "step": 7103 + }, + { + "epoch": 0.26, + "learning_rate": 1.7416779580327387e-07, + "loss": 0.7065, + "step": 7104 + }, + { + "epoch": 0.26, + "learning_rate": 1.7416413373860183e-07, + "loss": 0.6807, + "step": 7105 + }, + { + "epoch": 0.26, + "learning_rate": 1.7416047167392975e-07, + "loss": 0.6929, + "step": 7106 + }, + { + "epoch": 0.26, + "learning_rate": 1.7415680960925768e-07, + "loss": 0.708, + "step": 7107 + }, + { + "epoch": 0.26, + "learning_rate": 1.7415314754458563e-07, + "loss": 0.6789, + "step": 7108 + }, + { + "epoch": 0.26, + "learning_rate": 1.7414948547991356e-07, + "loss": 0.6878, + "step": 7109 + }, + { + "epoch": 0.26, + "learning_rate": 1.7414582341524151e-07, + "loss": 0.7024, + "step": 7110 + }, + { + "epoch": 0.26, + "learning_rate": 1.7414216135056944e-07, + "loss": 0.7112, + "step": 7111 + }, + { + "epoch": 0.26, + "learning_rate": 1.741384992858974e-07, + "loss": 0.6982, + "step": 7112 + }, + { + "epoch": 0.26, + "learning_rate": 1.7413483722122532e-07, + "loss": 0.6887, + "step": 7113 + }, + { + "epoch": 0.26, + "learning_rate": 1.7413117515655327e-07, + "loss": 0.6851, + "step": 7114 + }, + { + "epoch": 0.26, + "learning_rate": 1.7412751309188117e-07, + "loss": 0.704, + "step": 7115 + }, + { + "epoch": 0.26, + "learning_rate": 1.7412385102720913e-07, + "loss": 0.6937, + "step": 7116 + }, + { + "epoch": 0.26, + "learning_rate": 1.7412018896253705e-07, + "loss": 0.6928, + "step": 7117 + }, + { + "epoch": 0.26, + "learning_rate": 1.74116526897865e-07, + "loss": 0.6827, + "step": 7118 + }, + { + "epoch": 0.26, + "learning_rate": 1.7411286483319293e-07, + "loss": 0.7013, + "step": 7119 + }, + { + "epoch": 0.26, + "learning_rate": 1.7410920276852089e-07, + "loss": 0.6821, + "step": 7120 + }, + { + "epoch": 0.26, + "learning_rate": 1.7410554070384881e-07, + "loss": 0.6971, + "step": 7121 + }, + { + "epoch": 0.26, + "learning_rate": 1.7410187863917677e-07, + "loss": 0.6939, + "step": 7122 + }, + { + "epoch": 0.26, + "learning_rate": 1.740982165745047e-07, + "loss": 0.7001, + "step": 7123 + }, + { + "epoch": 0.26, + "learning_rate": 1.7409455450983265e-07, + "loss": 0.6979, + "step": 7124 + }, + { + "epoch": 0.26, + "learning_rate": 1.7409089244516057e-07, + "loss": 0.6938, + "step": 7125 + }, + { + "epoch": 0.26, + "learning_rate": 1.7408723038048853e-07, + "loss": 0.6771, + "step": 7126 + }, + { + "epoch": 0.26, + "learning_rate": 1.7408356831581645e-07, + "loss": 0.6837, + "step": 7127 + }, + { + "epoch": 0.26, + "learning_rate": 1.7407990625114438e-07, + "loss": 0.6889, + "step": 7128 + }, + { + "epoch": 0.26, + "learning_rate": 1.7407624418647233e-07, + "loss": 0.702, + "step": 7129 + }, + { + "epoch": 0.26, + "learning_rate": 1.7407258212180026e-07, + "loss": 0.6981, + "step": 7130 + }, + { + "epoch": 0.26, + "learning_rate": 1.740689200571282e-07, + "loss": 0.7005, + "step": 7131 + }, + { + "epoch": 0.26, + "learning_rate": 1.7406525799245614e-07, + "loss": 0.6999, + "step": 7132 + }, + { + "epoch": 0.26, + "learning_rate": 1.7406159592778407e-07, + "loss": 0.6758, + "step": 7133 + }, + { + "epoch": 0.26, + "learning_rate": 1.74057933863112e-07, + "loss": 0.6914, + "step": 7134 + }, + { + "epoch": 0.26, + "learning_rate": 1.7405427179843995e-07, + "loss": 0.7087, + "step": 7135 + }, + { + "epoch": 0.26, + "learning_rate": 1.7405060973376787e-07, + "loss": 0.7172, + "step": 7136 + }, + { + "epoch": 0.26, + "learning_rate": 1.7404694766909583e-07, + "loss": 0.6934, + "step": 7137 + }, + { + "epoch": 0.26, + "learning_rate": 1.7404328560442375e-07, + "loss": 0.6912, + "step": 7138 + }, + { + "epoch": 0.26, + "learning_rate": 1.740396235397517e-07, + "loss": 0.704, + "step": 7139 + }, + { + "epoch": 0.26, + "learning_rate": 1.7403596147507963e-07, + "loss": 0.704, + "step": 7140 + }, + { + "epoch": 0.26, + "learning_rate": 1.7403229941040759e-07, + "loss": 0.6866, + "step": 7141 + }, + { + "epoch": 0.26, + "learning_rate": 1.740286373457355e-07, + "loss": 0.6861, + "step": 7142 + }, + { + "epoch": 0.26, + "learning_rate": 1.7402497528106347e-07, + "loss": 0.6871, + "step": 7143 + }, + { + "epoch": 0.26, + "learning_rate": 1.740213132163914e-07, + "loss": 0.6924, + "step": 7144 + }, + { + "epoch": 0.26, + "learning_rate": 1.7401765115171935e-07, + "loss": 0.7012, + "step": 7145 + }, + { + "epoch": 0.26, + "learning_rate": 1.7401398908704727e-07, + "loss": 0.6966, + "step": 7146 + }, + { + "epoch": 0.26, + "learning_rate": 1.7401032702237523e-07, + "loss": 0.6806, + "step": 7147 + }, + { + "epoch": 0.26, + "learning_rate": 1.7400666495770315e-07, + "loss": 0.7017, + "step": 7148 + }, + { + "epoch": 0.26, + "learning_rate": 1.740030028930311e-07, + "loss": 0.7024, + "step": 7149 + }, + { + "epoch": 0.26, + "learning_rate": 1.7399934082835903e-07, + "loss": 0.6908, + "step": 7150 + }, + { + "epoch": 0.26, + "learning_rate": 1.7399567876368696e-07, + "loss": 0.694, + "step": 7151 + }, + { + "epoch": 0.26, + "learning_rate": 1.7399201669901489e-07, + "loss": 0.7113, + "step": 7152 + }, + { + "epoch": 0.26, + "learning_rate": 1.7398835463434284e-07, + "loss": 0.6871, + "step": 7153 + }, + { + "epoch": 0.26, + "learning_rate": 1.7398469256967076e-07, + "loss": 0.6873, + "step": 7154 + }, + { + "epoch": 0.26, + "learning_rate": 1.7398103050499872e-07, + "loss": 0.6929, + "step": 7155 + }, + { + "epoch": 0.26, + "learning_rate": 1.7397736844032664e-07, + "loss": 0.695, + "step": 7156 + }, + { + "epoch": 0.26, + "learning_rate": 1.7397370637565457e-07, + "loss": 0.6876, + "step": 7157 + }, + { + "epoch": 0.26, + "learning_rate": 1.7397004431098252e-07, + "loss": 0.7085, + "step": 7158 + }, + { + "epoch": 0.26, + "learning_rate": 1.7396638224631045e-07, + "loss": 0.6933, + "step": 7159 + }, + { + "epoch": 0.26, + "learning_rate": 1.739627201816384e-07, + "loss": 0.6923, + "step": 7160 + }, + { + "epoch": 0.26, + "learning_rate": 1.7395905811696633e-07, + "loss": 0.6842, + "step": 7161 + }, + { + "epoch": 0.26, + "learning_rate": 1.7395539605229428e-07, + "loss": 0.7089, + "step": 7162 + }, + { + "epoch": 0.26, + "learning_rate": 1.739517339876222e-07, + "loss": 0.7061, + "step": 7163 + }, + { + "epoch": 0.26, + "learning_rate": 1.7394807192295016e-07, + "loss": 0.7014, + "step": 7164 + }, + { + "epoch": 0.26, + "learning_rate": 1.739444098582781e-07, + "loss": 0.6796, + "step": 7165 + }, + { + "epoch": 0.26, + "learning_rate": 1.7394074779360604e-07, + "loss": 0.6964, + "step": 7166 + }, + { + "epoch": 0.26, + "learning_rate": 1.7393708572893397e-07, + "loss": 0.701, + "step": 7167 + }, + { + "epoch": 0.26, + "learning_rate": 1.7393342366426192e-07, + "loss": 0.6996, + "step": 7168 + }, + { + "epoch": 0.26, + "learning_rate": 1.7392976159958982e-07, + "loss": 0.6915, + "step": 7169 + }, + { + "epoch": 0.26, + "learning_rate": 1.7392609953491778e-07, + "loss": 0.6802, + "step": 7170 + }, + { + "epoch": 0.26, + "learning_rate": 1.739224374702457e-07, + "loss": 0.6997, + "step": 7171 + }, + { + "epoch": 0.26, + "learning_rate": 1.7391877540557366e-07, + "loss": 0.6763, + "step": 7172 + }, + { + "epoch": 0.26, + "learning_rate": 1.7391511334090158e-07, + "loss": 0.6981, + "step": 7173 + }, + { + "epoch": 0.26, + "learning_rate": 1.7391145127622954e-07, + "loss": 0.6929, + "step": 7174 + }, + { + "epoch": 0.26, + "learning_rate": 1.7390778921155746e-07, + "loss": 0.6992, + "step": 7175 + }, + { + "epoch": 0.26, + "learning_rate": 1.7390412714688542e-07, + "loss": 0.6794, + "step": 7176 + }, + { + "epoch": 0.26, + "learning_rate": 1.7390046508221334e-07, + "loss": 0.6736, + "step": 7177 + }, + { + "epoch": 0.26, + "learning_rate": 1.7389680301754127e-07, + "loss": 0.6909, + "step": 7178 + }, + { + "epoch": 0.26, + "learning_rate": 1.7389314095286922e-07, + "loss": 0.6968, + "step": 7179 + }, + { + "epoch": 0.26, + "learning_rate": 1.7388947888819715e-07, + "loss": 0.6904, + "step": 7180 + }, + { + "epoch": 0.26, + "learning_rate": 1.738858168235251e-07, + "loss": 0.6975, + "step": 7181 + }, + { + "epoch": 0.26, + "learning_rate": 1.7388215475885303e-07, + "loss": 0.6832, + "step": 7182 + }, + { + "epoch": 0.26, + "learning_rate": 1.7387849269418098e-07, + "loss": 0.6949, + "step": 7183 + }, + { + "epoch": 0.26, + "learning_rate": 1.738748306295089e-07, + "loss": 0.6856, + "step": 7184 + }, + { + "epoch": 0.26, + "learning_rate": 1.7387116856483686e-07, + "loss": 0.692, + "step": 7185 + }, + { + "epoch": 0.26, + "learning_rate": 1.738675065001648e-07, + "loss": 0.686, + "step": 7186 + }, + { + "epoch": 0.26, + "learning_rate": 1.7386384443549272e-07, + "loss": 0.6886, + "step": 7187 + }, + { + "epoch": 0.26, + "learning_rate": 1.7386018237082064e-07, + "loss": 0.6916, + "step": 7188 + }, + { + "epoch": 0.26, + "learning_rate": 1.738565203061486e-07, + "loss": 0.7064, + "step": 7189 + }, + { + "epoch": 0.26, + "learning_rate": 1.7385285824147652e-07, + "loss": 0.7116, + "step": 7190 + }, + { + "epoch": 0.26, + "learning_rate": 1.7384919617680448e-07, + "loss": 0.6922, + "step": 7191 + }, + { + "epoch": 0.26, + "learning_rate": 1.738455341121324e-07, + "loss": 0.7112, + "step": 7192 + }, + { + "epoch": 0.26, + "learning_rate": 1.7384187204746036e-07, + "loss": 0.7061, + "step": 7193 + }, + { + "epoch": 0.26, + "learning_rate": 1.7383820998278828e-07, + "loss": 0.6755, + "step": 7194 + }, + { + "epoch": 0.26, + "learning_rate": 1.7383454791811624e-07, + "loss": 0.6888, + "step": 7195 + }, + { + "epoch": 0.26, + "learning_rate": 1.7383088585344416e-07, + "loss": 0.7184, + "step": 7196 + }, + { + "epoch": 0.26, + "learning_rate": 1.7382722378877212e-07, + "loss": 0.7089, + "step": 7197 + }, + { + "epoch": 0.26, + "learning_rate": 1.7382356172410004e-07, + "loss": 0.6903, + "step": 7198 + }, + { + "epoch": 0.26, + "learning_rate": 1.7381989965942797e-07, + "loss": 0.6918, + "step": 7199 + }, + { + "epoch": 0.26, + "learning_rate": 1.7381623759475592e-07, + "loss": 0.697, + "step": 7200 + }, + { + "epoch": 0.26, + "learning_rate": 1.7381257553008385e-07, + "loss": 0.7092, + "step": 7201 + }, + { + "epoch": 0.26, + "learning_rate": 1.738089134654118e-07, + "loss": 0.6885, + "step": 7202 + }, + { + "epoch": 0.26, + "learning_rate": 1.7380525140073973e-07, + "loss": 0.6948, + "step": 7203 + }, + { + "epoch": 0.26, + "learning_rate": 1.7380158933606768e-07, + "loss": 0.6915, + "step": 7204 + }, + { + "epoch": 0.26, + "learning_rate": 1.7379792727139558e-07, + "loss": 0.6971, + "step": 7205 + }, + { + "epoch": 0.26, + "learning_rate": 1.7379426520672354e-07, + "loss": 0.6874, + "step": 7206 + }, + { + "epoch": 0.26, + "learning_rate": 1.7379060314205146e-07, + "loss": 0.6808, + "step": 7207 + }, + { + "epoch": 0.26, + "learning_rate": 1.7378694107737942e-07, + "loss": 0.7116, + "step": 7208 + }, + { + "epoch": 0.26, + "learning_rate": 1.7378327901270734e-07, + "loss": 0.69, + "step": 7209 + }, + { + "epoch": 0.26, + "learning_rate": 1.737796169480353e-07, + "loss": 0.6918, + "step": 7210 + }, + { + "epoch": 0.26, + "learning_rate": 1.7377595488336322e-07, + "loss": 0.6822, + "step": 7211 + }, + { + "epoch": 0.26, + "learning_rate": 1.7377229281869117e-07, + "loss": 0.6787, + "step": 7212 + }, + { + "epoch": 0.26, + "learning_rate": 1.737686307540191e-07, + "loss": 0.6835, + "step": 7213 + }, + { + "epoch": 0.26, + "learning_rate": 1.7376496868934705e-07, + "loss": 0.702, + "step": 7214 + }, + { + "epoch": 0.26, + "learning_rate": 1.7376130662467498e-07, + "loss": 0.6976, + "step": 7215 + }, + { + "epoch": 0.26, + "learning_rate": 1.7375764456000293e-07, + "loss": 0.6973, + "step": 7216 + }, + { + "epoch": 0.26, + "learning_rate": 1.7375398249533086e-07, + "loss": 0.702, + "step": 7217 + }, + { + "epoch": 0.26, + "learning_rate": 1.7375032043065881e-07, + "loss": 0.7084, + "step": 7218 + }, + { + "epoch": 0.26, + "learning_rate": 1.7374665836598674e-07, + "loss": 0.6957, + "step": 7219 + }, + { + "epoch": 0.26, + "learning_rate": 1.737429963013147e-07, + "loss": 0.6915, + "step": 7220 + }, + { + "epoch": 0.26, + "learning_rate": 1.7373933423664262e-07, + "loss": 0.6891, + "step": 7221 + }, + { + "epoch": 0.26, + "learning_rate": 1.7373567217197055e-07, + "loss": 0.71, + "step": 7222 + }, + { + "epoch": 0.26, + "learning_rate": 1.7373201010729847e-07, + "loss": 0.6837, + "step": 7223 + }, + { + "epoch": 0.26, + "learning_rate": 1.7372834804262643e-07, + "loss": 0.6923, + "step": 7224 + }, + { + "epoch": 0.26, + "learning_rate": 1.7372468597795435e-07, + "loss": 0.6929, + "step": 7225 + }, + { + "epoch": 0.26, + "learning_rate": 1.7372102391328228e-07, + "loss": 0.7016, + "step": 7226 + }, + { + "epoch": 0.26, + "learning_rate": 1.7371736184861023e-07, + "loss": 0.6949, + "step": 7227 + }, + { + "epoch": 0.26, + "learning_rate": 1.7371369978393816e-07, + "loss": 0.701, + "step": 7228 + }, + { + "epoch": 0.26, + "learning_rate": 1.7371003771926611e-07, + "loss": 0.689, + "step": 7229 + }, + { + "epoch": 0.26, + "learning_rate": 1.7370637565459404e-07, + "loss": 0.6955, + "step": 7230 + }, + { + "epoch": 0.26, + "learning_rate": 1.73702713589922e-07, + "loss": 0.7002, + "step": 7231 + }, + { + "epoch": 0.26, + "learning_rate": 1.7369905152524992e-07, + "loss": 0.6991, + "step": 7232 + }, + { + "epoch": 0.26, + "learning_rate": 1.7369538946057787e-07, + "loss": 0.6902, + "step": 7233 + }, + { + "epoch": 0.26, + "learning_rate": 1.736917273959058e-07, + "loss": 0.6894, + "step": 7234 + }, + { + "epoch": 0.26, + "learning_rate": 1.7368806533123375e-07, + "loss": 0.6939, + "step": 7235 + }, + { + "epoch": 0.26, + "learning_rate": 1.7368440326656168e-07, + "loss": 0.6907, + "step": 7236 + }, + { + "epoch": 0.26, + "learning_rate": 1.7368074120188963e-07, + "loss": 0.7003, + "step": 7237 + }, + { + "epoch": 0.26, + "learning_rate": 1.7367707913721756e-07, + "loss": 0.6936, + "step": 7238 + }, + { + "epoch": 0.26, + "learning_rate": 1.736734170725455e-07, + "loss": 0.6847, + "step": 7239 + }, + { + "epoch": 0.26, + "learning_rate": 1.7366975500787344e-07, + "loss": 0.6771, + "step": 7240 + }, + { + "epoch": 0.26, + "learning_rate": 1.7366609294320137e-07, + "loss": 0.7055, + "step": 7241 + }, + { + "epoch": 0.26, + "learning_rate": 1.736624308785293e-07, + "loss": 0.6872, + "step": 7242 + }, + { + "epoch": 0.27, + "learning_rate": 1.7365876881385725e-07, + "loss": 0.6845, + "step": 7243 + }, + { + "epoch": 0.27, + "learning_rate": 1.7365510674918517e-07, + "loss": 0.7006, + "step": 7244 + }, + { + "epoch": 0.27, + "learning_rate": 1.7365144468451313e-07, + "loss": 0.6947, + "step": 7245 + }, + { + "epoch": 0.27, + "learning_rate": 1.7364778261984105e-07, + "loss": 0.7108, + "step": 7246 + }, + { + "epoch": 0.27, + "learning_rate": 1.73644120555169e-07, + "loss": 0.6822, + "step": 7247 + }, + { + "epoch": 0.27, + "learning_rate": 1.7364045849049693e-07, + "loss": 0.6902, + "step": 7248 + }, + { + "epoch": 0.27, + "learning_rate": 1.7363679642582486e-07, + "loss": 0.7055, + "step": 7249 + }, + { + "epoch": 0.27, + "learning_rate": 1.736331343611528e-07, + "loss": 0.6862, + "step": 7250 + }, + { + "epoch": 0.27, + "learning_rate": 1.7362947229648074e-07, + "loss": 0.7002, + "step": 7251 + }, + { + "epoch": 0.27, + "learning_rate": 1.736258102318087e-07, + "loss": 0.6923, + "step": 7252 + }, + { + "epoch": 0.27, + "learning_rate": 1.7362214816713662e-07, + "loss": 0.6918, + "step": 7253 + }, + { + "epoch": 0.27, + "learning_rate": 1.7361848610246457e-07, + "loss": 0.6983, + "step": 7254 + }, + { + "epoch": 0.27, + "learning_rate": 1.736148240377925e-07, + "loss": 0.689, + "step": 7255 + }, + { + "epoch": 0.27, + "learning_rate": 1.7361116197312045e-07, + "loss": 0.7024, + "step": 7256 + }, + { + "epoch": 0.27, + "learning_rate": 1.7360749990844838e-07, + "loss": 0.6922, + "step": 7257 + }, + { + "epoch": 0.27, + "learning_rate": 1.7360383784377633e-07, + "loss": 0.6914, + "step": 7258 + }, + { + "epoch": 0.27, + "learning_rate": 1.7360017577910423e-07, + "loss": 0.6904, + "step": 7259 + }, + { + "epoch": 0.27, + "learning_rate": 1.7359651371443219e-07, + "loss": 0.693, + "step": 7260 + }, + { + "epoch": 0.27, + "learning_rate": 1.735928516497601e-07, + "loss": 0.6896, + "step": 7261 + }, + { + "epoch": 0.27, + "learning_rate": 1.7358918958508807e-07, + "loss": 0.6917, + "step": 7262 + }, + { + "epoch": 0.27, + "learning_rate": 1.73585527520416e-07, + "loss": 0.6957, + "step": 7263 + }, + { + "epoch": 0.27, + "learning_rate": 1.7358186545574395e-07, + "loss": 0.6893, + "step": 7264 + }, + { + "epoch": 0.27, + "learning_rate": 1.7357820339107187e-07, + "loss": 0.6737, + "step": 7265 + }, + { + "epoch": 0.27, + "learning_rate": 1.7357454132639983e-07, + "loss": 0.6832, + "step": 7266 + }, + { + "epoch": 0.27, + "learning_rate": 1.7357087926172775e-07, + "loss": 0.6941, + "step": 7267 + }, + { + "epoch": 0.27, + "learning_rate": 1.735672171970557e-07, + "loss": 0.6903, + "step": 7268 + }, + { + "epoch": 0.27, + "learning_rate": 1.7356355513238363e-07, + "loss": 0.6883, + "step": 7269 + }, + { + "epoch": 0.27, + "learning_rate": 1.7355989306771156e-07, + "loss": 0.6987, + "step": 7270 + }, + { + "epoch": 0.27, + "learning_rate": 1.735562310030395e-07, + "loss": 0.694, + "step": 7271 + }, + { + "epoch": 0.27, + "learning_rate": 1.7355256893836744e-07, + "loss": 0.6868, + "step": 7272 + }, + { + "epoch": 0.27, + "learning_rate": 1.735489068736954e-07, + "loss": 0.7015, + "step": 7273 + }, + { + "epoch": 0.27, + "learning_rate": 1.7354524480902332e-07, + "loss": 0.6968, + "step": 7274 + }, + { + "epoch": 0.27, + "learning_rate": 1.7354158274435127e-07, + "loss": 0.6772, + "step": 7275 + }, + { + "epoch": 0.27, + "learning_rate": 1.735379206796792e-07, + "loss": 0.686, + "step": 7276 + }, + { + "epoch": 0.27, + "learning_rate": 1.7353425861500712e-07, + "loss": 0.6958, + "step": 7277 + }, + { + "epoch": 0.27, + "learning_rate": 1.7353059655033505e-07, + "loss": 0.6945, + "step": 7278 + }, + { + "epoch": 0.27, + "learning_rate": 1.73526934485663e-07, + "loss": 0.7002, + "step": 7279 + }, + { + "epoch": 0.27, + "learning_rate": 1.7352327242099093e-07, + "loss": 0.6821, + "step": 7280 + }, + { + "epoch": 0.27, + "learning_rate": 1.7351961035631888e-07, + "loss": 0.6955, + "step": 7281 + }, + { + "epoch": 0.27, + "learning_rate": 1.735159482916468e-07, + "loss": 0.6879, + "step": 7282 + }, + { + "epoch": 0.27, + "learning_rate": 1.7351228622697476e-07, + "loss": 0.6973, + "step": 7283 + }, + { + "epoch": 0.27, + "learning_rate": 1.735086241623027e-07, + "loss": 0.6992, + "step": 7284 + }, + { + "epoch": 0.27, + "learning_rate": 1.7350496209763064e-07, + "loss": 0.6921, + "step": 7285 + }, + { + "epoch": 0.27, + "learning_rate": 1.7350130003295857e-07, + "loss": 0.6952, + "step": 7286 + }, + { + "epoch": 0.27, + "learning_rate": 1.7349763796828652e-07, + "loss": 0.7072, + "step": 7287 + }, + { + "epoch": 0.27, + "learning_rate": 1.7349397590361445e-07, + "loss": 0.7053, + "step": 7288 + }, + { + "epoch": 0.27, + "learning_rate": 1.734903138389424e-07, + "loss": 0.6824, + "step": 7289 + }, + { + "epoch": 0.27, + "learning_rate": 1.7348665177427033e-07, + "loss": 0.7156, + "step": 7290 + }, + { + "epoch": 0.27, + "learning_rate": 1.7348298970959828e-07, + "loss": 0.7168, + "step": 7291 + }, + { + "epoch": 0.27, + "learning_rate": 1.734793276449262e-07, + "loss": 0.7079, + "step": 7292 + }, + { + "epoch": 0.27, + "learning_rate": 1.7347566558025414e-07, + "loss": 0.6983, + "step": 7293 + }, + { + "epoch": 0.27, + "learning_rate": 1.734720035155821e-07, + "loss": 0.7008, + "step": 7294 + }, + { + "epoch": 0.27, + "learning_rate": 1.7346834145091002e-07, + "loss": 0.6956, + "step": 7295 + }, + { + "epoch": 0.27, + "learning_rate": 1.7346467938623794e-07, + "loss": 0.7082, + "step": 7296 + }, + { + "epoch": 0.27, + "learning_rate": 1.7346101732156587e-07, + "loss": 0.6921, + "step": 7297 + }, + { + "epoch": 0.27, + "learning_rate": 1.7345735525689382e-07, + "loss": 0.6931, + "step": 7298 + }, + { + "epoch": 0.27, + "learning_rate": 1.7345369319222175e-07, + "loss": 0.6935, + "step": 7299 + }, + { + "epoch": 0.27, + "learning_rate": 1.734500311275497e-07, + "loss": 0.6952, + "step": 7300 + }, + { + "epoch": 0.27, + "learning_rate": 1.7344636906287763e-07, + "loss": 0.7014, + "step": 7301 + }, + { + "epoch": 0.27, + "learning_rate": 1.7344270699820558e-07, + "loss": 0.707, + "step": 7302 + }, + { + "epoch": 0.27, + "learning_rate": 1.734390449335335e-07, + "loss": 0.7002, + "step": 7303 + }, + { + "epoch": 0.27, + "learning_rate": 1.7343538286886146e-07, + "loss": 0.7138, + "step": 7304 + }, + { + "epoch": 0.27, + "learning_rate": 1.734317208041894e-07, + "loss": 0.6967, + "step": 7305 + }, + { + "epoch": 0.27, + "learning_rate": 1.7342805873951734e-07, + "loss": 0.7043, + "step": 7306 + }, + { + "epoch": 0.27, + "learning_rate": 1.7342439667484527e-07, + "loss": 0.6971, + "step": 7307 + }, + { + "epoch": 0.27, + "learning_rate": 1.7342073461017322e-07, + "loss": 0.7108, + "step": 7308 + }, + { + "epoch": 0.27, + "learning_rate": 1.7341707254550115e-07, + "loss": 0.699, + "step": 7309 + }, + { + "epoch": 0.27, + "learning_rate": 1.734134104808291e-07, + "loss": 0.6999, + "step": 7310 + }, + { + "epoch": 0.27, + "learning_rate": 1.7340974841615703e-07, + "loss": 0.6867, + "step": 7311 + }, + { + "epoch": 0.27, + "learning_rate": 1.7340608635148498e-07, + "loss": 0.7002, + "step": 7312 + }, + { + "epoch": 0.27, + "learning_rate": 1.7340242428681288e-07, + "loss": 0.6785, + "step": 7313 + }, + { + "epoch": 0.27, + "learning_rate": 1.7339876222214084e-07, + "loss": 0.6926, + "step": 7314 + }, + { + "epoch": 0.27, + "learning_rate": 1.7339510015746876e-07, + "loss": 0.6928, + "step": 7315 + }, + { + "epoch": 0.27, + "learning_rate": 1.7339143809279672e-07, + "loss": 0.6912, + "step": 7316 + }, + { + "epoch": 0.27, + "learning_rate": 1.7338777602812464e-07, + "loss": 0.7048, + "step": 7317 + }, + { + "epoch": 0.27, + "learning_rate": 1.733841139634526e-07, + "loss": 0.69, + "step": 7318 + }, + { + "epoch": 0.27, + "learning_rate": 1.7338045189878052e-07, + "loss": 0.6907, + "step": 7319 + }, + { + "epoch": 0.27, + "learning_rate": 1.7337678983410845e-07, + "loss": 0.6924, + "step": 7320 + }, + { + "epoch": 0.27, + "learning_rate": 1.733731277694364e-07, + "loss": 0.694, + "step": 7321 + }, + { + "epoch": 0.27, + "learning_rate": 1.7336946570476433e-07, + "loss": 0.6916, + "step": 7322 + }, + { + "epoch": 0.27, + "learning_rate": 1.7336580364009228e-07, + "loss": 0.7075, + "step": 7323 + }, + { + "epoch": 0.27, + "learning_rate": 1.733621415754202e-07, + "loss": 0.6874, + "step": 7324 + }, + { + "epoch": 0.27, + "learning_rate": 1.7335847951074816e-07, + "loss": 0.687, + "step": 7325 + }, + { + "epoch": 0.27, + "learning_rate": 1.733548174460761e-07, + "loss": 0.6892, + "step": 7326 + }, + { + "epoch": 0.27, + "learning_rate": 1.7335115538140404e-07, + "loss": 0.6917, + "step": 7327 + }, + { + "epoch": 0.27, + "learning_rate": 1.7334749331673197e-07, + "loss": 0.7069, + "step": 7328 + }, + { + "epoch": 0.27, + "learning_rate": 1.7334383125205992e-07, + "loss": 0.6944, + "step": 7329 + }, + { + "epoch": 0.27, + "learning_rate": 1.7334016918738785e-07, + "loss": 0.6878, + "step": 7330 + }, + { + "epoch": 0.27, + "learning_rate": 1.7333650712271577e-07, + "loss": 0.7009, + "step": 7331 + }, + { + "epoch": 0.27, + "learning_rate": 1.733328450580437e-07, + "loss": 0.7042, + "step": 7332 + }, + { + "epoch": 0.27, + "learning_rate": 1.7332918299337165e-07, + "loss": 0.6953, + "step": 7333 + }, + { + "epoch": 0.27, + "learning_rate": 1.7332552092869958e-07, + "loss": 0.6888, + "step": 7334 + }, + { + "epoch": 0.27, + "learning_rate": 1.7332185886402753e-07, + "loss": 0.6897, + "step": 7335 + }, + { + "epoch": 0.27, + "learning_rate": 1.7331819679935546e-07, + "loss": 0.6897, + "step": 7336 + }, + { + "epoch": 0.27, + "learning_rate": 1.7331453473468341e-07, + "loss": 0.6802, + "step": 7337 + }, + { + "epoch": 0.27, + "learning_rate": 1.7331087267001134e-07, + "loss": 0.6824, + "step": 7338 + }, + { + "epoch": 0.27, + "learning_rate": 1.733072106053393e-07, + "loss": 0.697, + "step": 7339 + }, + { + "epoch": 0.27, + "learning_rate": 1.7330354854066722e-07, + "loss": 0.6901, + "step": 7340 + }, + { + "epoch": 0.27, + "learning_rate": 1.7329988647599515e-07, + "loss": 0.6886, + "step": 7341 + }, + { + "epoch": 0.27, + "learning_rate": 1.732962244113231e-07, + "loss": 0.6977, + "step": 7342 + }, + { + "epoch": 0.27, + "learning_rate": 1.7329256234665103e-07, + "loss": 0.7077, + "step": 7343 + }, + { + "epoch": 0.27, + "learning_rate": 1.7328890028197898e-07, + "loss": 0.6964, + "step": 7344 + }, + { + "epoch": 0.27, + "learning_rate": 1.732852382173069e-07, + "loss": 0.7108, + "step": 7345 + }, + { + "epoch": 0.27, + "learning_rate": 1.7328157615263486e-07, + "loss": 0.685, + "step": 7346 + }, + { + "epoch": 0.27, + "learning_rate": 1.732779140879628e-07, + "loss": 0.6911, + "step": 7347 + }, + { + "epoch": 0.27, + "learning_rate": 1.7327425202329074e-07, + "loss": 0.7014, + "step": 7348 + }, + { + "epoch": 0.27, + "learning_rate": 1.7327058995861864e-07, + "loss": 0.6913, + "step": 7349 + }, + { + "epoch": 0.27, + "learning_rate": 1.732669278939466e-07, + "loss": 0.7012, + "step": 7350 + }, + { + "epoch": 0.27, + "learning_rate": 1.7326326582927452e-07, + "loss": 0.6821, + "step": 7351 + }, + { + "epoch": 0.27, + "learning_rate": 1.7325960376460247e-07, + "loss": 0.6971, + "step": 7352 + }, + { + "epoch": 0.27, + "learning_rate": 1.732559416999304e-07, + "loss": 0.6997, + "step": 7353 + }, + { + "epoch": 0.27, + "learning_rate": 1.7325227963525835e-07, + "loss": 0.6898, + "step": 7354 + }, + { + "epoch": 0.27, + "learning_rate": 1.7324861757058628e-07, + "loss": 0.6895, + "step": 7355 + }, + { + "epoch": 0.27, + "learning_rate": 1.7324495550591423e-07, + "loss": 0.7051, + "step": 7356 + }, + { + "epoch": 0.27, + "learning_rate": 1.7324129344124216e-07, + "loss": 0.688, + "step": 7357 + }, + { + "epoch": 0.27, + "learning_rate": 1.7323763137657011e-07, + "loss": 0.7131, + "step": 7358 + }, + { + "epoch": 0.27, + "learning_rate": 1.7323396931189804e-07, + "loss": 0.6926, + "step": 7359 + }, + { + "epoch": 0.27, + "learning_rate": 1.73230307247226e-07, + "loss": 0.6911, + "step": 7360 + }, + { + "epoch": 0.27, + "learning_rate": 1.7322664518255392e-07, + "loss": 0.6954, + "step": 7361 + }, + { + "epoch": 0.27, + "learning_rate": 1.7322298311788187e-07, + "loss": 0.68, + "step": 7362 + }, + { + "epoch": 0.27, + "learning_rate": 1.732193210532098e-07, + "loss": 0.6937, + "step": 7363 + }, + { + "epoch": 0.27, + "learning_rate": 1.7321565898853773e-07, + "loss": 0.7048, + "step": 7364 + }, + { + "epoch": 0.27, + "learning_rate": 1.7321199692386568e-07, + "loss": 0.6885, + "step": 7365 + }, + { + "epoch": 0.27, + "learning_rate": 1.732083348591936e-07, + "loss": 0.6997, + "step": 7366 + }, + { + "epoch": 0.27, + "learning_rate": 1.7320467279452153e-07, + "loss": 0.6852, + "step": 7367 + }, + { + "epoch": 0.27, + "learning_rate": 1.7320101072984946e-07, + "loss": 0.6925, + "step": 7368 + }, + { + "epoch": 0.27, + "learning_rate": 1.731973486651774e-07, + "loss": 0.6979, + "step": 7369 + }, + { + "epoch": 0.27, + "learning_rate": 1.7319368660050534e-07, + "loss": 0.6782, + "step": 7370 + }, + { + "epoch": 0.27, + "learning_rate": 1.731900245358333e-07, + "loss": 0.6928, + "step": 7371 + }, + { + "epoch": 0.27, + "learning_rate": 1.7318636247116122e-07, + "loss": 0.6848, + "step": 7372 + }, + { + "epoch": 0.27, + "learning_rate": 1.7318270040648917e-07, + "loss": 0.6797, + "step": 7373 + }, + { + "epoch": 0.27, + "learning_rate": 1.731790383418171e-07, + "loss": 0.7098, + "step": 7374 + }, + { + "epoch": 0.27, + "learning_rate": 1.7317537627714505e-07, + "loss": 0.6976, + "step": 7375 + }, + { + "epoch": 0.27, + "learning_rate": 1.7317171421247298e-07, + "loss": 0.6961, + "step": 7376 + }, + { + "epoch": 0.27, + "learning_rate": 1.7316805214780093e-07, + "loss": 0.6804, + "step": 7377 + }, + { + "epoch": 0.27, + "learning_rate": 1.7316439008312886e-07, + "loss": 0.6945, + "step": 7378 + }, + { + "epoch": 0.27, + "learning_rate": 1.731607280184568e-07, + "loss": 0.6911, + "step": 7379 + }, + { + "epoch": 0.27, + "learning_rate": 1.7315706595378474e-07, + "loss": 0.6838, + "step": 7380 + }, + { + "epoch": 0.27, + "learning_rate": 1.731534038891127e-07, + "loss": 0.6965, + "step": 7381 + }, + { + "epoch": 0.27, + "learning_rate": 1.7314974182444062e-07, + "loss": 0.698, + "step": 7382 + }, + { + "epoch": 0.27, + "learning_rate": 1.7314607975976857e-07, + "loss": 0.6914, + "step": 7383 + }, + { + "epoch": 0.27, + "learning_rate": 1.7314241769509647e-07, + "loss": 0.6889, + "step": 7384 + }, + { + "epoch": 0.27, + "learning_rate": 1.7313875563042443e-07, + "loss": 0.6987, + "step": 7385 + }, + { + "epoch": 0.27, + "learning_rate": 1.7313509356575235e-07, + "loss": 0.6887, + "step": 7386 + }, + { + "epoch": 0.27, + "learning_rate": 1.731314315010803e-07, + "loss": 0.6875, + "step": 7387 + }, + { + "epoch": 0.27, + "learning_rate": 1.7312776943640823e-07, + "loss": 0.7021, + "step": 7388 + }, + { + "epoch": 0.27, + "learning_rate": 1.7312410737173618e-07, + "loss": 0.6979, + "step": 7389 + }, + { + "epoch": 0.27, + "learning_rate": 1.731204453070641e-07, + "loss": 0.6993, + "step": 7390 + }, + { + "epoch": 0.27, + "learning_rate": 1.7311678324239204e-07, + "loss": 0.6968, + "step": 7391 + }, + { + "epoch": 0.27, + "learning_rate": 1.7311312117772e-07, + "loss": 0.6974, + "step": 7392 + }, + { + "epoch": 0.27, + "learning_rate": 1.7310945911304792e-07, + "loss": 0.6986, + "step": 7393 + }, + { + "epoch": 0.27, + "learning_rate": 1.7310579704837587e-07, + "loss": 0.6929, + "step": 7394 + }, + { + "epoch": 0.27, + "learning_rate": 1.731021349837038e-07, + "loss": 0.6992, + "step": 7395 + }, + { + "epoch": 0.27, + "learning_rate": 1.7309847291903175e-07, + "loss": 0.69, + "step": 7396 + }, + { + "epoch": 0.27, + "learning_rate": 1.7309481085435968e-07, + "loss": 0.6968, + "step": 7397 + }, + { + "epoch": 0.27, + "learning_rate": 1.7309114878968763e-07, + "loss": 0.7125, + "step": 7398 + }, + { + "epoch": 0.27, + "learning_rate": 1.7308748672501556e-07, + "loss": 0.6844, + "step": 7399 + }, + { + "epoch": 0.27, + "learning_rate": 1.730838246603435e-07, + "loss": 0.6733, + "step": 7400 + }, + { + "epoch": 0.27, + "learning_rate": 1.7308016259567144e-07, + "loss": 0.7062, + "step": 7401 + }, + { + "epoch": 0.27, + "learning_rate": 1.7307650053099936e-07, + "loss": 0.6759, + "step": 7402 + }, + { + "epoch": 0.27, + "learning_rate": 1.730728384663273e-07, + "loss": 0.7071, + "step": 7403 + }, + { + "epoch": 0.27, + "learning_rate": 1.7306917640165524e-07, + "loss": 0.6792, + "step": 7404 + }, + { + "epoch": 0.27, + "learning_rate": 1.7306551433698317e-07, + "loss": 0.705, + "step": 7405 + }, + { + "epoch": 0.27, + "learning_rate": 1.7306185227231112e-07, + "loss": 0.6786, + "step": 7406 + }, + { + "epoch": 0.27, + "learning_rate": 1.7305819020763905e-07, + "loss": 0.6943, + "step": 7407 + }, + { + "epoch": 0.27, + "learning_rate": 1.73054528142967e-07, + "loss": 0.6939, + "step": 7408 + }, + { + "epoch": 0.27, + "learning_rate": 1.7305086607829493e-07, + "loss": 0.6856, + "step": 7409 + }, + { + "epoch": 0.27, + "learning_rate": 1.7304720401362288e-07, + "loss": 0.6921, + "step": 7410 + }, + { + "epoch": 0.27, + "learning_rate": 1.730435419489508e-07, + "loss": 0.6909, + "step": 7411 + }, + { + "epoch": 0.27, + "learning_rate": 1.7303987988427874e-07, + "loss": 0.7156, + "step": 7412 + }, + { + "epoch": 0.27, + "learning_rate": 1.730362178196067e-07, + "loss": 0.6853, + "step": 7413 + }, + { + "epoch": 0.27, + "learning_rate": 1.7303255575493462e-07, + "loss": 0.6832, + "step": 7414 + }, + { + "epoch": 0.27, + "learning_rate": 1.7302889369026257e-07, + "loss": 0.6858, + "step": 7415 + }, + { + "epoch": 0.27, + "learning_rate": 1.730252316255905e-07, + "loss": 0.6853, + "step": 7416 + }, + { + "epoch": 0.27, + "learning_rate": 1.7302156956091845e-07, + "loss": 0.6878, + "step": 7417 + }, + { + "epoch": 0.27, + "learning_rate": 1.7301790749624638e-07, + "loss": 0.6957, + "step": 7418 + }, + { + "epoch": 0.27, + "learning_rate": 1.7301424543157433e-07, + "loss": 0.6844, + "step": 7419 + }, + { + "epoch": 0.27, + "learning_rate": 1.7301058336690223e-07, + "loss": 0.7004, + "step": 7420 + }, + { + "epoch": 0.27, + "learning_rate": 1.7300692130223018e-07, + "loss": 0.7007, + "step": 7421 + }, + { + "epoch": 0.27, + "learning_rate": 1.730032592375581e-07, + "loss": 0.6922, + "step": 7422 + }, + { + "epoch": 0.27, + "learning_rate": 1.7299959717288606e-07, + "loss": 0.6894, + "step": 7423 + }, + { + "epoch": 0.27, + "learning_rate": 1.72995935108214e-07, + "loss": 0.6995, + "step": 7424 + }, + { + "epoch": 0.27, + "learning_rate": 1.7299227304354194e-07, + "loss": 0.6971, + "step": 7425 + }, + { + "epoch": 0.27, + "learning_rate": 1.7298861097886987e-07, + "loss": 0.6979, + "step": 7426 + }, + { + "epoch": 0.27, + "learning_rate": 1.7298494891419782e-07, + "loss": 0.6833, + "step": 7427 + }, + { + "epoch": 0.27, + "learning_rate": 1.7298128684952575e-07, + "loss": 0.6831, + "step": 7428 + }, + { + "epoch": 0.27, + "learning_rate": 1.729776247848537e-07, + "loss": 0.709, + "step": 7429 + }, + { + "epoch": 0.27, + "learning_rate": 1.7297396272018163e-07, + "loss": 0.7123, + "step": 7430 + }, + { + "epoch": 0.27, + "learning_rate": 1.7297030065550958e-07, + "loss": 0.7049, + "step": 7431 + }, + { + "epoch": 0.27, + "learning_rate": 1.729666385908375e-07, + "loss": 0.6952, + "step": 7432 + }, + { + "epoch": 0.27, + "learning_rate": 1.7296297652616546e-07, + "loss": 0.7197, + "step": 7433 + }, + { + "epoch": 0.27, + "learning_rate": 1.729593144614934e-07, + "loss": 0.6922, + "step": 7434 + }, + { + "epoch": 0.27, + "learning_rate": 1.7295565239682132e-07, + "loss": 0.693, + "step": 7435 + }, + { + "epoch": 0.27, + "learning_rate": 1.7295199033214927e-07, + "loss": 0.678, + "step": 7436 + }, + { + "epoch": 0.27, + "learning_rate": 1.729483282674772e-07, + "loss": 0.6936, + "step": 7437 + }, + { + "epoch": 0.27, + "learning_rate": 1.7294466620280512e-07, + "loss": 0.7004, + "step": 7438 + }, + { + "epoch": 0.27, + "learning_rate": 1.7294100413813305e-07, + "loss": 0.7032, + "step": 7439 + }, + { + "epoch": 0.27, + "learning_rate": 1.72937342073461e-07, + "loss": 0.7022, + "step": 7440 + }, + { + "epoch": 0.27, + "learning_rate": 1.7293368000878893e-07, + "loss": 0.6892, + "step": 7441 + }, + { + "epoch": 0.27, + "learning_rate": 1.7293001794411688e-07, + "loss": 0.6889, + "step": 7442 + }, + { + "epoch": 0.27, + "learning_rate": 1.729263558794448e-07, + "loss": 0.6901, + "step": 7443 + }, + { + "epoch": 0.27, + "learning_rate": 1.7292269381477276e-07, + "loss": 0.6945, + "step": 7444 + }, + { + "epoch": 0.27, + "learning_rate": 1.729190317501007e-07, + "loss": 0.6953, + "step": 7445 + }, + { + "epoch": 0.27, + "learning_rate": 1.7291536968542864e-07, + "loss": 0.6834, + "step": 7446 + }, + { + "epoch": 0.27, + "learning_rate": 1.7291170762075657e-07, + "loss": 0.6884, + "step": 7447 + }, + { + "epoch": 0.27, + "learning_rate": 1.7290804555608452e-07, + "loss": 0.7122, + "step": 7448 + }, + { + "epoch": 0.27, + "learning_rate": 1.7290438349141245e-07, + "loss": 0.7004, + "step": 7449 + }, + { + "epoch": 0.27, + "learning_rate": 1.729007214267404e-07, + "loss": 0.6779, + "step": 7450 + }, + { + "epoch": 0.27, + "learning_rate": 1.7289705936206833e-07, + "loss": 0.6924, + "step": 7451 + }, + { + "epoch": 0.27, + "learning_rate": 1.7289339729739628e-07, + "loss": 0.7157, + "step": 7452 + }, + { + "epoch": 0.27, + "learning_rate": 1.728897352327242e-07, + "loss": 0.6915, + "step": 7453 + }, + { + "epoch": 0.27, + "learning_rate": 1.7288607316805216e-07, + "loss": 0.6954, + "step": 7454 + }, + { + "epoch": 0.27, + "learning_rate": 1.728824111033801e-07, + "loss": 0.6877, + "step": 7455 + }, + { + "epoch": 0.27, + "learning_rate": 1.7287874903870801e-07, + "loss": 0.712, + "step": 7456 + }, + { + "epoch": 0.27, + "learning_rate": 1.7287508697403594e-07, + "loss": 0.6979, + "step": 7457 + }, + { + "epoch": 0.27, + "learning_rate": 1.728714249093639e-07, + "loss": 0.7017, + "step": 7458 + }, + { + "epoch": 0.27, + "learning_rate": 1.7286776284469182e-07, + "loss": 0.7001, + "step": 7459 + }, + { + "epoch": 0.27, + "learning_rate": 1.7286410078001977e-07, + "loss": 0.7039, + "step": 7460 + }, + { + "epoch": 0.27, + "learning_rate": 1.728604387153477e-07, + "loss": 0.6954, + "step": 7461 + }, + { + "epoch": 0.27, + "learning_rate": 1.7285677665067563e-07, + "loss": 0.7071, + "step": 7462 + }, + { + "epoch": 0.27, + "learning_rate": 1.7285311458600358e-07, + "loss": 0.6974, + "step": 7463 + }, + { + "epoch": 0.27, + "learning_rate": 1.728494525213315e-07, + "loss": 0.6989, + "step": 7464 + }, + { + "epoch": 0.27, + "learning_rate": 1.7284579045665946e-07, + "loss": 0.6887, + "step": 7465 + }, + { + "epoch": 0.27, + "learning_rate": 1.728421283919874e-07, + "loss": 0.6799, + "step": 7466 + }, + { + "epoch": 0.27, + "learning_rate": 1.7283846632731534e-07, + "loss": 0.7003, + "step": 7467 + }, + { + "epoch": 0.27, + "learning_rate": 1.7283480426264327e-07, + "loss": 0.6945, + "step": 7468 + }, + { + "epoch": 0.27, + "learning_rate": 1.7283114219797122e-07, + "loss": 0.6985, + "step": 7469 + }, + { + "epoch": 0.27, + "learning_rate": 1.7282748013329915e-07, + "loss": 0.6962, + "step": 7470 + }, + { + "epoch": 0.27, + "learning_rate": 1.728238180686271e-07, + "loss": 0.6863, + "step": 7471 + }, + { + "epoch": 0.27, + "learning_rate": 1.7282015600395503e-07, + "loss": 0.6901, + "step": 7472 + }, + { + "epoch": 0.27, + "learning_rate": 1.7281649393928298e-07, + "loss": 0.6892, + "step": 7473 + }, + { + "epoch": 0.27, + "learning_rate": 1.7281283187461088e-07, + "loss": 0.6987, + "step": 7474 + }, + { + "epoch": 0.27, + "learning_rate": 1.7280916980993883e-07, + "loss": 0.6918, + "step": 7475 + }, + { + "epoch": 0.27, + "learning_rate": 1.7280550774526676e-07, + "loss": 0.6908, + "step": 7476 + }, + { + "epoch": 0.27, + "learning_rate": 1.7280184568059471e-07, + "loss": 0.6943, + "step": 7477 + }, + { + "epoch": 0.27, + "learning_rate": 1.7279818361592264e-07, + "loss": 0.6997, + "step": 7478 + }, + { + "epoch": 0.27, + "learning_rate": 1.727945215512506e-07, + "loss": 0.6981, + "step": 7479 + }, + { + "epoch": 0.27, + "learning_rate": 1.7279085948657852e-07, + "loss": 0.6977, + "step": 7480 + }, + { + "epoch": 0.27, + "learning_rate": 1.7278719742190647e-07, + "loss": 0.6904, + "step": 7481 + }, + { + "epoch": 0.27, + "learning_rate": 1.727835353572344e-07, + "loss": 0.6968, + "step": 7482 + }, + { + "epoch": 0.27, + "learning_rate": 1.7277987329256233e-07, + "loss": 0.6968, + "step": 7483 + }, + { + "epoch": 0.27, + "learning_rate": 1.7277621122789028e-07, + "loss": 0.6953, + "step": 7484 + }, + { + "epoch": 0.27, + "learning_rate": 1.727725491632182e-07, + "loss": 0.6761, + "step": 7485 + }, + { + "epoch": 0.27, + "learning_rate": 1.7276888709854616e-07, + "loss": 0.6904, + "step": 7486 + }, + { + "epoch": 0.27, + "learning_rate": 1.7276522503387409e-07, + "loss": 0.6889, + "step": 7487 + }, + { + "epoch": 0.27, + "learning_rate": 1.7276156296920204e-07, + "loss": 0.6945, + "step": 7488 + }, + { + "epoch": 0.27, + "learning_rate": 1.7275790090452997e-07, + "loss": 0.6967, + "step": 7489 + }, + { + "epoch": 0.27, + "learning_rate": 1.7275423883985792e-07, + "loss": 0.6908, + "step": 7490 + }, + { + "epoch": 0.27, + "learning_rate": 1.7275057677518585e-07, + "loss": 0.7049, + "step": 7491 + }, + { + "epoch": 0.27, + "learning_rate": 1.7274691471051377e-07, + "loss": 0.6932, + "step": 7492 + }, + { + "epoch": 0.27, + "learning_rate": 1.727432526458417e-07, + "loss": 0.6917, + "step": 7493 + }, + { + "epoch": 0.27, + "learning_rate": 1.7273959058116965e-07, + "loss": 0.6905, + "step": 7494 + }, + { + "epoch": 0.27, + "learning_rate": 1.7273592851649758e-07, + "loss": 0.7013, + "step": 7495 + }, + { + "epoch": 0.27, + "learning_rate": 1.7273226645182553e-07, + "loss": 0.6993, + "step": 7496 + }, + { + "epoch": 0.27, + "learning_rate": 1.7272860438715346e-07, + "loss": 0.6868, + "step": 7497 + }, + { + "epoch": 0.27, + "learning_rate": 1.727249423224814e-07, + "loss": 0.7014, + "step": 7498 + }, + { + "epoch": 0.27, + "learning_rate": 1.7272128025780934e-07, + "loss": 0.6998, + "step": 7499 + }, + { + "epoch": 0.27, + "learning_rate": 1.727176181931373e-07, + "loss": 0.6971, + "step": 7500 + }, + { + "epoch": 0.27, + "learning_rate": 1.7271395612846522e-07, + "loss": 0.6874, + "step": 7501 + }, + { + "epoch": 0.27, + "learning_rate": 1.7271029406379317e-07, + "loss": 0.7046, + "step": 7502 + }, + { + "epoch": 0.27, + "learning_rate": 1.727066319991211e-07, + "loss": 0.6873, + "step": 7503 + }, + { + "epoch": 0.27, + "learning_rate": 1.7270296993444905e-07, + "loss": 0.6928, + "step": 7504 + }, + { + "epoch": 0.27, + "learning_rate": 1.7269930786977698e-07, + "loss": 0.6879, + "step": 7505 + }, + { + "epoch": 0.27, + "learning_rate": 1.726956458051049e-07, + "loss": 0.6904, + "step": 7506 + }, + { + "epoch": 0.27, + "learning_rate": 1.7269198374043286e-07, + "loss": 0.6942, + "step": 7507 + }, + { + "epoch": 0.27, + "learning_rate": 1.7268832167576078e-07, + "loss": 0.6976, + "step": 7508 + }, + { + "epoch": 0.27, + "learning_rate": 1.7268465961108874e-07, + "loss": 0.6941, + "step": 7509 + }, + { + "epoch": 0.27, + "learning_rate": 1.7268099754641664e-07, + "loss": 0.6885, + "step": 7510 + }, + { + "epoch": 0.27, + "learning_rate": 1.726773354817446e-07, + "loss": 0.7023, + "step": 7511 + }, + { + "epoch": 0.27, + "learning_rate": 1.7267367341707252e-07, + "loss": 0.6961, + "step": 7512 + }, + { + "epoch": 0.27, + "learning_rate": 1.7267001135240047e-07, + "loss": 0.689, + "step": 7513 + }, + { + "epoch": 0.27, + "learning_rate": 1.726663492877284e-07, + "loss": 0.6909, + "step": 7514 + }, + { + "epoch": 0.27, + "learning_rate": 1.7266268722305635e-07, + "loss": 0.6884, + "step": 7515 + }, + { + "epoch": 0.27, + "learning_rate": 1.7265902515838428e-07, + "loss": 0.7079, + "step": 7516 + }, + { + "epoch": 0.28, + "learning_rate": 1.7265536309371223e-07, + "loss": 0.6949, + "step": 7517 + }, + { + "epoch": 0.28, + "learning_rate": 1.7265170102904016e-07, + "loss": 0.6859, + "step": 7518 + }, + { + "epoch": 0.28, + "learning_rate": 1.726480389643681e-07, + "loss": 0.7084, + "step": 7519 + }, + { + "epoch": 0.28, + "learning_rate": 1.7264437689969604e-07, + "loss": 0.7225, + "step": 7520 + }, + { + "epoch": 0.28, + "learning_rate": 1.72640714835024e-07, + "loss": 0.6879, + "step": 7521 + }, + { + "epoch": 0.28, + "learning_rate": 1.7263705277035192e-07, + "loss": 0.6958, + "step": 7522 + }, + { + "epoch": 0.28, + "learning_rate": 1.7263339070567987e-07, + "loss": 0.7025, + "step": 7523 + }, + { + "epoch": 0.28, + "learning_rate": 1.726297286410078e-07, + "loss": 0.684, + "step": 7524 + }, + { + "epoch": 0.28, + "learning_rate": 1.7262606657633575e-07, + "loss": 0.6916, + "step": 7525 + }, + { + "epoch": 0.28, + "learning_rate": 1.7262240451166368e-07, + "loss": 0.701, + "step": 7526 + }, + { + "epoch": 0.28, + "learning_rate": 1.726187424469916e-07, + "loss": 0.7038, + "step": 7527 + }, + { + "epoch": 0.28, + "learning_rate": 1.7261508038231953e-07, + "loss": 0.7028, + "step": 7528 + }, + { + "epoch": 0.28, + "learning_rate": 1.7261141831764748e-07, + "loss": 0.6897, + "step": 7529 + }, + { + "epoch": 0.28, + "learning_rate": 1.726077562529754e-07, + "loss": 0.6874, + "step": 7530 + }, + { + "epoch": 0.28, + "learning_rate": 1.7260409418830336e-07, + "loss": 0.6817, + "step": 7531 + }, + { + "epoch": 0.28, + "learning_rate": 1.726004321236313e-07, + "loss": 0.7033, + "step": 7532 + }, + { + "epoch": 0.28, + "learning_rate": 1.7259677005895922e-07, + "loss": 0.6991, + "step": 7533 + }, + { + "epoch": 0.28, + "learning_rate": 1.7259310799428717e-07, + "loss": 0.7032, + "step": 7534 + }, + { + "epoch": 0.28, + "learning_rate": 1.725894459296151e-07, + "loss": 0.6958, + "step": 7535 + }, + { + "epoch": 0.28, + "learning_rate": 1.7258578386494305e-07, + "loss": 0.6866, + "step": 7536 + }, + { + "epoch": 0.28, + "learning_rate": 1.7258212180027098e-07, + "loss": 0.702, + "step": 7537 + }, + { + "epoch": 0.28, + "learning_rate": 1.7257845973559893e-07, + "loss": 0.7033, + "step": 7538 + }, + { + "epoch": 0.28, + "learning_rate": 1.7257479767092686e-07, + "loss": 0.7055, + "step": 7539 + }, + { + "epoch": 0.28, + "learning_rate": 1.725711356062548e-07, + "loss": 0.7115, + "step": 7540 + }, + { + "epoch": 0.28, + "learning_rate": 1.7256747354158274e-07, + "loss": 0.6971, + "step": 7541 + }, + { + "epoch": 0.28, + "learning_rate": 1.725638114769107e-07, + "loss": 0.6932, + "step": 7542 + }, + { + "epoch": 0.28, + "learning_rate": 1.7256014941223862e-07, + "loss": 0.6956, + "step": 7543 + }, + { + "epoch": 0.28, + "learning_rate": 1.7255648734756657e-07, + "loss": 0.6999, + "step": 7544 + }, + { + "epoch": 0.28, + "learning_rate": 1.725528252828945e-07, + "loss": 0.6965, + "step": 7545 + }, + { + "epoch": 0.28, + "learning_rate": 1.7254916321822242e-07, + "loss": 0.6931, + "step": 7546 + }, + { + "epoch": 0.28, + "learning_rate": 1.7254550115355035e-07, + "loss": 0.6845, + "step": 7547 + }, + { + "epoch": 0.28, + "learning_rate": 1.725418390888783e-07, + "loss": 0.6741, + "step": 7548 + }, + { + "epoch": 0.28, + "learning_rate": 1.7253817702420623e-07, + "loss": 0.6813, + "step": 7549 + }, + { + "epoch": 0.28, + "learning_rate": 1.7253451495953418e-07, + "loss": 0.6998, + "step": 7550 + }, + { + "epoch": 0.28, + "learning_rate": 1.725308528948621e-07, + "loss": 0.6881, + "step": 7551 + }, + { + "epoch": 0.28, + "learning_rate": 1.7252719083019006e-07, + "loss": 0.6995, + "step": 7552 + }, + { + "epoch": 0.28, + "learning_rate": 1.72523528765518e-07, + "loss": 0.7001, + "step": 7553 + }, + { + "epoch": 0.28, + "learning_rate": 1.7251986670084592e-07, + "loss": 0.7029, + "step": 7554 + }, + { + "epoch": 0.28, + "learning_rate": 1.7251620463617387e-07, + "loss": 0.6932, + "step": 7555 + }, + { + "epoch": 0.28, + "learning_rate": 1.725125425715018e-07, + "loss": 0.7085, + "step": 7556 + }, + { + "epoch": 0.28, + "learning_rate": 1.7250888050682975e-07, + "loss": 0.6904, + "step": 7557 + }, + { + "epoch": 0.28, + "learning_rate": 1.7250521844215768e-07, + "loss": 0.6976, + "step": 7558 + }, + { + "epoch": 0.28, + "learning_rate": 1.7250155637748563e-07, + "loss": 0.6847, + "step": 7559 + }, + { + "epoch": 0.28, + "learning_rate": 1.7249789431281356e-07, + "loss": 0.6945, + "step": 7560 + }, + { + "epoch": 0.28, + "learning_rate": 1.724942322481415e-07, + "loss": 0.6922, + "step": 7561 + }, + { + "epoch": 0.28, + "learning_rate": 1.7249057018346944e-07, + "loss": 0.6981, + "step": 7562 + }, + { + "epoch": 0.28, + "learning_rate": 1.724869081187974e-07, + "loss": 0.6979, + "step": 7563 + }, + { + "epoch": 0.28, + "learning_rate": 1.724832460541253e-07, + "loss": 0.676, + "step": 7564 + }, + { + "epoch": 0.28, + "learning_rate": 1.7247958398945324e-07, + "loss": 0.6832, + "step": 7565 + }, + { + "epoch": 0.28, + "learning_rate": 1.7247592192478117e-07, + "loss": 0.69, + "step": 7566 + }, + { + "epoch": 0.28, + "learning_rate": 1.7247225986010912e-07, + "loss": 0.6716, + "step": 7567 + }, + { + "epoch": 0.28, + "learning_rate": 1.7246859779543705e-07, + "loss": 0.7056, + "step": 7568 + }, + { + "epoch": 0.28, + "learning_rate": 1.72464935730765e-07, + "loss": 0.6929, + "step": 7569 + }, + { + "epoch": 0.28, + "learning_rate": 1.7246127366609293e-07, + "loss": 0.6923, + "step": 7570 + }, + { + "epoch": 0.28, + "learning_rate": 1.7245761160142088e-07, + "loss": 0.7265, + "step": 7571 + }, + { + "epoch": 0.28, + "learning_rate": 1.724539495367488e-07, + "loss": 0.6943, + "step": 7572 + }, + { + "epoch": 0.28, + "learning_rate": 1.7245028747207676e-07, + "loss": 0.6861, + "step": 7573 + }, + { + "epoch": 0.28, + "learning_rate": 1.724466254074047e-07, + "loss": 0.6918, + "step": 7574 + }, + { + "epoch": 0.28, + "learning_rate": 1.7244296334273264e-07, + "loss": 0.6817, + "step": 7575 + }, + { + "epoch": 0.28, + "learning_rate": 1.7243930127806057e-07, + "loss": 0.6911, + "step": 7576 + }, + { + "epoch": 0.28, + "learning_rate": 1.724356392133885e-07, + "loss": 0.6872, + "step": 7577 + }, + { + "epoch": 0.28, + "learning_rate": 1.7243197714871645e-07, + "loss": 0.677, + "step": 7578 + }, + { + "epoch": 0.28, + "learning_rate": 1.7242831508404437e-07, + "loss": 0.7095, + "step": 7579 + }, + { + "epoch": 0.28, + "learning_rate": 1.7242465301937233e-07, + "loss": 0.6806, + "step": 7580 + }, + { + "epoch": 0.28, + "learning_rate": 1.7242099095470025e-07, + "loss": 0.6999, + "step": 7581 + }, + { + "epoch": 0.28, + "learning_rate": 1.7241732889002818e-07, + "loss": 0.6894, + "step": 7582 + }, + { + "epoch": 0.28, + "learning_rate": 1.724136668253561e-07, + "loss": 0.6911, + "step": 7583 + }, + { + "epoch": 0.28, + "learning_rate": 1.7241000476068406e-07, + "loss": 0.6976, + "step": 7584 + }, + { + "epoch": 0.28, + "learning_rate": 1.72406342696012e-07, + "loss": 0.6993, + "step": 7585 + }, + { + "epoch": 0.28, + "learning_rate": 1.7240268063133994e-07, + "loss": 0.7205, + "step": 7586 + }, + { + "epoch": 0.28, + "learning_rate": 1.7239901856666787e-07, + "loss": 0.6889, + "step": 7587 + }, + { + "epoch": 0.28, + "learning_rate": 1.7239535650199582e-07, + "loss": 0.6839, + "step": 7588 + }, + { + "epoch": 0.28, + "learning_rate": 1.7239169443732375e-07, + "loss": 0.7217, + "step": 7589 + }, + { + "epoch": 0.28, + "learning_rate": 1.723880323726517e-07, + "loss": 0.6962, + "step": 7590 + }, + { + "epoch": 0.28, + "learning_rate": 1.7238437030797963e-07, + "loss": 0.7021, + "step": 7591 + }, + { + "epoch": 0.28, + "learning_rate": 1.7238070824330758e-07, + "loss": 0.6907, + "step": 7592 + }, + { + "epoch": 0.28, + "learning_rate": 1.723770461786355e-07, + "loss": 0.6779, + "step": 7593 + }, + { + "epoch": 0.28, + "learning_rate": 1.7237338411396346e-07, + "loss": 0.6808, + "step": 7594 + }, + { + "epoch": 0.28, + "learning_rate": 1.7236972204929139e-07, + "loss": 0.6962, + "step": 7595 + }, + { + "epoch": 0.28, + "learning_rate": 1.7236605998461934e-07, + "loss": 0.6851, + "step": 7596 + }, + { + "epoch": 0.28, + "learning_rate": 1.7236239791994727e-07, + "loss": 0.714, + "step": 7597 + }, + { + "epoch": 0.28, + "learning_rate": 1.723587358552752e-07, + "loss": 0.6945, + "step": 7598 + }, + { + "epoch": 0.28, + "learning_rate": 1.7235507379060315e-07, + "loss": 0.6847, + "step": 7599 + }, + { + "epoch": 0.28, + "learning_rate": 1.7235141172593107e-07, + "loss": 0.6843, + "step": 7600 + }, + { + "epoch": 0.28, + "learning_rate": 1.72347749661259e-07, + "loss": 0.6964, + "step": 7601 + }, + { + "epoch": 0.28, + "learning_rate": 1.7234408759658695e-07, + "loss": 0.6863, + "step": 7602 + }, + { + "epoch": 0.28, + "learning_rate": 1.7234042553191488e-07, + "loss": 0.6828, + "step": 7603 + }, + { + "epoch": 0.28, + "learning_rate": 1.723367634672428e-07, + "loss": 0.6833, + "step": 7604 + }, + { + "epoch": 0.28, + "learning_rate": 1.7233310140257076e-07, + "loss": 0.7024, + "step": 7605 + }, + { + "epoch": 0.28, + "learning_rate": 1.7232943933789869e-07, + "loss": 0.7005, + "step": 7606 + }, + { + "epoch": 0.28, + "learning_rate": 1.7232577727322664e-07, + "loss": 0.6911, + "step": 7607 + }, + { + "epoch": 0.28, + "learning_rate": 1.7232211520855457e-07, + "loss": 0.6879, + "step": 7608 + }, + { + "epoch": 0.28, + "learning_rate": 1.7231845314388252e-07, + "loss": 0.6916, + "step": 7609 + }, + { + "epoch": 0.28, + "learning_rate": 1.7231479107921045e-07, + "loss": 0.6829, + "step": 7610 + }, + { + "epoch": 0.28, + "learning_rate": 1.723111290145384e-07, + "loss": 0.6729, + "step": 7611 + }, + { + "epoch": 0.28, + "learning_rate": 1.7230746694986633e-07, + "loss": 0.7053, + "step": 7612 + }, + { + "epoch": 0.28, + "learning_rate": 1.7230380488519428e-07, + "loss": 0.6968, + "step": 7613 + }, + { + "epoch": 0.28, + "learning_rate": 1.723001428205222e-07, + "loss": 0.6865, + "step": 7614 + }, + { + "epoch": 0.28, + "learning_rate": 1.7229648075585016e-07, + "loss": 0.6887, + "step": 7615 + }, + { + "epoch": 0.28, + "learning_rate": 1.7229281869117809e-07, + "loss": 0.7047, + "step": 7616 + }, + { + "epoch": 0.28, + "learning_rate": 1.7228915662650604e-07, + "loss": 0.6948, + "step": 7617 + }, + { + "epoch": 0.28, + "learning_rate": 1.7228549456183394e-07, + "loss": 0.6903, + "step": 7618 + }, + { + "epoch": 0.28, + "learning_rate": 1.722818324971619e-07, + "loss": 0.6834, + "step": 7619 + }, + { + "epoch": 0.28, + "learning_rate": 1.7227817043248982e-07, + "loss": 0.7057, + "step": 7620 + }, + { + "epoch": 0.28, + "learning_rate": 1.7227450836781777e-07, + "loss": 0.6902, + "step": 7621 + }, + { + "epoch": 0.28, + "learning_rate": 1.722708463031457e-07, + "loss": 0.6859, + "step": 7622 + }, + { + "epoch": 0.28, + "learning_rate": 1.7226718423847365e-07, + "loss": 0.68, + "step": 7623 + }, + { + "epoch": 0.28, + "learning_rate": 1.7226352217380158e-07, + "loss": 0.6712, + "step": 7624 + }, + { + "epoch": 0.28, + "learning_rate": 1.722598601091295e-07, + "loss": 0.7104, + "step": 7625 + }, + { + "epoch": 0.28, + "learning_rate": 1.7225619804445746e-07, + "loss": 0.7024, + "step": 7626 + }, + { + "epoch": 0.28, + "learning_rate": 1.7225253597978538e-07, + "loss": 0.6918, + "step": 7627 + }, + { + "epoch": 0.28, + "learning_rate": 1.7224887391511334e-07, + "loss": 0.7027, + "step": 7628 + }, + { + "epoch": 0.28, + "learning_rate": 1.7224521185044126e-07, + "loss": 0.7118, + "step": 7629 + }, + { + "epoch": 0.28, + "learning_rate": 1.7224154978576922e-07, + "loss": 0.6934, + "step": 7630 + }, + { + "epoch": 0.28, + "learning_rate": 1.7223788772109714e-07, + "loss": 0.6832, + "step": 7631 + }, + { + "epoch": 0.28, + "learning_rate": 1.722342256564251e-07, + "loss": 0.6954, + "step": 7632 + }, + { + "epoch": 0.28, + "learning_rate": 1.7223056359175302e-07, + "loss": 0.715, + "step": 7633 + }, + { + "epoch": 0.28, + "learning_rate": 1.7222690152708098e-07, + "loss": 0.6879, + "step": 7634 + }, + { + "epoch": 0.28, + "learning_rate": 1.722232394624089e-07, + "loss": 0.71, + "step": 7635 + }, + { + "epoch": 0.28, + "learning_rate": 1.7221957739773683e-07, + "loss": 0.6938, + "step": 7636 + }, + { + "epoch": 0.28, + "learning_rate": 1.7221591533306476e-07, + "loss": 0.6876, + "step": 7637 + }, + { + "epoch": 0.28, + "learning_rate": 1.722122532683927e-07, + "loss": 0.6869, + "step": 7638 + }, + { + "epoch": 0.28, + "learning_rate": 1.7220859120372064e-07, + "loss": 0.6968, + "step": 7639 + }, + { + "epoch": 0.28, + "learning_rate": 1.722049291390486e-07, + "loss": 0.6911, + "step": 7640 + }, + { + "epoch": 0.28, + "learning_rate": 1.7220126707437652e-07, + "loss": 0.6925, + "step": 7641 + }, + { + "epoch": 0.28, + "learning_rate": 1.7219760500970447e-07, + "loss": 0.6806, + "step": 7642 + }, + { + "epoch": 0.28, + "learning_rate": 1.721939429450324e-07, + "loss": 0.6937, + "step": 7643 + }, + { + "epoch": 0.28, + "learning_rate": 1.7219028088036035e-07, + "loss": 0.6766, + "step": 7644 + }, + { + "epoch": 0.28, + "learning_rate": 1.7218661881568828e-07, + "loss": 0.7146, + "step": 7645 + }, + { + "epoch": 0.28, + "learning_rate": 1.7218295675101623e-07, + "loss": 0.7007, + "step": 7646 + }, + { + "epoch": 0.28, + "learning_rate": 1.7217929468634416e-07, + "loss": 0.6872, + "step": 7647 + }, + { + "epoch": 0.28, + "learning_rate": 1.7217563262167208e-07, + "loss": 0.7103, + "step": 7648 + }, + { + "epoch": 0.28, + "learning_rate": 1.7217197055700004e-07, + "loss": 0.7102, + "step": 7649 + }, + { + "epoch": 0.28, + "learning_rate": 1.7216830849232796e-07, + "loss": 0.6838, + "step": 7650 + }, + { + "epoch": 0.28, + "learning_rate": 1.7216464642765592e-07, + "loss": 0.6908, + "step": 7651 + }, + { + "epoch": 0.28, + "learning_rate": 1.7216098436298384e-07, + "loss": 0.6893, + "step": 7652 + }, + { + "epoch": 0.28, + "learning_rate": 1.721573222983118e-07, + "loss": 0.6924, + "step": 7653 + }, + { + "epoch": 0.28, + "learning_rate": 1.721536602336397e-07, + "loss": 0.7113, + "step": 7654 + }, + { + "epoch": 0.28, + "learning_rate": 1.7214999816896765e-07, + "loss": 0.6822, + "step": 7655 + }, + { + "epoch": 0.28, + "learning_rate": 1.7214633610429558e-07, + "loss": 0.6935, + "step": 7656 + }, + { + "epoch": 0.28, + "learning_rate": 1.7214267403962353e-07, + "loss": 0.694, + "step": 7657 + }, + { + "epoch": 0.28, + "learning_rate": 1.7213901197495146e-07, + "loss": 0.6929, + "step": 7658 + }, + { + "epoch": 0.28, + "learning_rate": 1.721353499102794e-07, + "loss": 0.6723, + "step": 7659 + }, + { + "epoch": 0.28, + "learning_rate": 1.7213168784560734e-07, + "loss": 0.6967, + "step": 7660 + }, + { + "epoch": 0.28, + "learning_rate": 1.721280257809353e-07, + "loss": 0.6846, + "step": 7661 + }, + { + "epoch": 0.28, + "learning_rate": 1.7212436371626322e-07, + "loss": 0.6902, + "step": 7662 + }, + { + "epoch": 0.28, + "learning_rate": 1.7212070165159117e-07, + "loss": 0.701, + "step": 7663 + }, + { + "epoch": 0.28, + "learning_rate": 1.721170395869191e-07, + "loss": 0.7125, + "step": 7664 + }, + { + "epoch": 0.28, + "learning_rate": 1.7211337752224705e-07, + "loss": 0.7019, + "step": 7665 + }, + { + "epoch": 0.28, + "learning_rate": 1.7210971545757498e-07, + "loss": 0.6945, + "step": 7666 + }, + { + "epoch": 0.28, + "learning_rate": 1.7210605339290293e-07, + "loss": 0.6954, + "step": 7667 + }, + { + "epoch": 0.28, + "learning_rate": 1.7210239132823086e-07, + "loss": 0.6979, + "step": 7668 + }, + { + "epoch": 0.28, + "learning_rate": 1.7209872926355878e-07, + "loss": 0.704, + "step": 7669 + }, + { + "epoch": 0.28, + "learning_rate": 1.7209506719888674e-07, + "loss": 0.6997, + "step": 7670 + }, + { + "epoch": 0.28, + "learning_rate": 1.7209140513421466e-07, + "loss": 0.6963, + "step": 7671 + }, + { + "epoch": 0.28, + "learning_rate": 1.720877430695426e-07, + "loss": 0.7046, + "step": 7672 + }, + { + "epoch": 0.28, + "learning_rate": 1.7208408100487054e-07, + "loss": 0.7029, + "step": 7673 + }, + { + "epoch": 0.28, + "learning_rate": 1.7208041894019847e-07, + "loss": 0.7031, + "step": 7674 + }, + { + "epoch": 0.28, + "learning_rate": 1.720767568755264e-07, + "loss": 0.6991, + "step": 7675 + }, + { + "epoch": 0.28, + "learning_rate": 1.7207309481085435e-07, + "loss": 0.6913, + "step": 7676 + }, + { + "epoch": 0.28, + "learning_rate": 1.7206943274618228e-07, + "loss": 0.6912, + "step": 7677 + }, + { + "epoch": 0.28, + "learning_rate": 1.7206577068151023e-07, + "loss": 0.6923, + "step": 7678 + }, + { + "epoch": 0.28, + "learning_rate": 1.7206210861683816e-07, + "loss": 0.691, + "step": 7679 + }, + { + "epoch": 0.28, + "learning_rate": 1.720584465521661e-07, + "loss": 0.6947, + "step": 7680 + }, + { + "epoch": 0.28, + "learning_rate": 1.7205478448749404e-07, + "loss": 0.6969, + "step": 7681 + }, + { + "epoch": 0.28, + "learning_rate": 1.72051122422822e-07, + "loss": 0.6919, + "step": 7682 + }, + { + "epoch": 0.28, + "learning_rate": 1.7204746035814992e-07, + "loss": 0.6964, + "step": 7683 + }, + { + "epoch": 0.28, + "learning_rate": 1.7204379829347787e-07, + "loss": 0.672, + "step": 7684 + }, + { + "epoch": 0.28, + "learning_rate": 1.720401362288058e-07, + "loss": 0.693, + "step": 7685 + }, + { + "epoch": 0.28, + "learning_rate": 1.7203647416413375e-07, + "loss": 0.7099, + "step": 7686 + }, + { + "epoch": 0.28, + "learning_rate": 1.7203281209946167e-07, + "loss": 0.6891, + "step": 7687 + }, + { + "epoch": 0.28, + "learning_rate": 1.7202915003478963e-07, + "loss": 0.7039, + "step": 7688 + }, + { + "epoch": 0.28, + "learning_rate": 1.7202548797011755e-07, + "loss": 0.697, + "step": 7689 + }, + { + "epoch": 0.28, + "learning_rate": 1.7202182590544548e-07, + "loss": 0.7046, + "step": 7690 + }, + { + "epoch": 0.28, + "learning_rate": 1.720181638407734e-07, + "loss": 0.6898, + "step": 7691 + }, + { + "epoch": 0.28, + "learning_rate": 1.7201450177610136e-07, + "loss": 0.6879, + "step": 7692 + }, + { + "epoch": 0.28, + "learning_rate": 1.720108397114293e-07, + "loss": 0.6936, + "step": 7693 + }, + { + "epoch": 0.28, + "learning_rate": 1.7200717764675724e-07, + "loss": 0.6952, + "step": 7694 + }, + { + "epoch": 0.28, + "learning_rate": 1.7200351558208517e-07, + "loss": 0.6857, + "step": 7695 + }, + { + "epoch": 0.28, + "learning_rate": 1.719998535174131e-07, + "loss": 0.6936, + "step": 7696 + }, + { + "epoch": 0.28, + "learning_rate": 1.7199619145274105e-07, + "loss": 0.6731, + "step": 7697 + }, + { + "epoch": 0.28, + "learning_rate": 1.7199252938806897e-07, + "loss": 0.6889, + "step": 7698 + }, + { + "epoch": 0.28, + "learning_rate": 1.7198886732339693e-07, + "loss": 0.7014, + "step": 7699 + }, + { + "epoch": 0.28, + "learning_rate": 1.7198520525872485e-07, + "loss": 0.6905, + "step": 7700 + }, + { + "epoch": 0.28, + "learning_rate": 1.719815431940528e-07, + "loss": 0.6892, + "step": 7701 + }, + { + "epoch": 0.28, + "learning_rate": 1.7197788112938073e-07, + "loss": 0.6869, + "step": 7702 + }, + { + "epoch": 0.28, + "learning_rate": 1.719742190647087e-07, + "loss": 0.6884, + "step": 7703 + }, + { + "epoch": 0.28, + "learning_rate": 1.7197055700003661e-07, + "loss": 0.691, + "step": 7704 + }, + { + "epoch": 0.28, + "learning_rate": 1.7196689493536457e-07, + "loss": 0.6747, + "step": 7705 + }, + { + "epoch": 0.28, + "learning_rate": 1.719632328706925e-07, + "loss": 0.6848, + "step": 7706 + }, + { + "epoch": 0.28, + "learning_rate": 1.7195957080602045e-07, + "loss": 0.6953, + "step": 7707 + }, + { + "epoch": 0.28, + "learning_rate": 1.7195590874134835e-07, + "loss": 0.7108, + "step": 7708 + }, + { + "epoch": 0.28, + "learning_rate": 1.719522466766763e-07, + "loss": 0.7019, + "step": 7709 + }, + { + "epoch": 0.28, + "learning_rate": 1.7194858461200423e-07, + "loss": 0.6872, + "step": 7710 + }, + { + "epoch": 0.28, + "learning_rate": 1.7194492254733218e-07, + "loss": 0.6918, + "step": 7711 + }, + { + "epoch": 0.28, + "learning_rate": 1.719412604826601e-07, + "loss": 0.6976, + "step": 7712 + }, + { + "epoch": 0.28, + "learning_rate": 1.7193759841798806e-07, + "loss": 0.6927, + "step": 7713 + }, + { + "epoch": 0.28, + "learning_rate": 1.7193393635331599e-07, + "loss": 0.705, + "step": 7714 + }, + { + "epoch": 0.28, + "learning_rate": 1.7193027428864394e-07, + "loss": 0.693, + "step": 7715 + }, + { + "epoch": 0.28, + "learning_rate": 1.7192661222397187e-07, + "loss": 0.6845, + "step": 7716 + }, + { + "epoch": 0.28, + "learning_rate": 1.7192295015929982e-07, + "loss": 0.6946, + "step": 7717 + }, + { + "epoch": 0.28, + "learning_rate": 1.7191928809462775e-07, + "loss": 0.6996, + "step": 7718 + }, + { + "epoch": 0.28, + "learning_rate": 1.7191562602995567e-07, + "loss": 0.6937, + "step": 7719 + }, + { + "epoch": 0.28, + "learning_rate": 1.7191196396528363e-07, + "loss": 0.6833, + "step": 7720 + }, + { + "epoch": 0.28, + "learning_rate": 1.7190830190061155e-07, + "loss": 0.6852, + "step": 7721 + }, + { + "epoch": 0.28, + "learning_rate": 1.719046398359395e-07, + "loss": 0.7005, + "step": 7722 + }, + { + "epoch": 0.28, + "learning_rate": 1.7190097777126743e-07, + "loss": 0.7004, + "step": 7723 + }, + { + "epoch": 0.28, + "learning_rate": 1.7189731570659539e-07, + "loss": 0.6982, + "step": 7724 + }, + { + "epoch": 0.28, + "learning_rate": 1.718936536419233e-07, + "loss": 0.6795, + "step": 7725 + }, + { + "epoch": 0.28, + "learning_rate": 1.7188999157725124e-07, + "loss": 0.6892, + "step": 7726 + }, + { + "epoch": 0.28, + "learning_rate": 1.7188632951257917e-07, + "loss": 0.6967, + "step": 7727 + }, + { + "epoch": 0.28, + "learning_rate": 1.7188266744790712e-07, + "loss": 0.6999, + "step": 7728 + }, + { + "epoch": 0.28, + "learning_rate": 1.7187900538323505e-07, + "loss": 0.6992, + "step": 7729 + }, + { + "epoch": 0.28, + "learning_rate": 1.71875343318563e-07, + "loss": 0.6949, + "step": 7730 + }, + { + "epoch": 0.28, + "learning_rate": 1.7187168125389093e-07, + "loss": 0.6835, + "step": 7731 + }, + { + "epoch": 0.28, + "learning_rate": 1.7186801918921888e-07, + "loss": 0.6931, + "step": 7732 + }, + { + "epoch": 0.28, + "learning_rate": 1.718643571245468e-07, + "loss": 0.6929, + "step": 7733 + }, + { + "epoch": 0.28, + "learning_rate": 1.7186069505987476e-07, + "loss": 0.7, + "step": 7734 + }, + { + "epoch": 0.28, + "learning_rate": 1.7185703299520269e-07, + "loss": 0.6932, + "step": 7735 + }, + { + "epoch": 0.28, + "learning_rate": 1.7185337093053064e-07, + "loss": 0.6831, + "step": 7736 + }, + { + "epoch": 0.28, + "learning_rate": 1.7184970886585857e-07, + "loss": 0.6987, + "step": 7737 + }, + { + "epoch": 0.28, + "learning_rate": 1.7184604680118652e-07, + "loss": 0.6903, + "step": 7738 + }, + { + "epoch": 0.28, + "learning_rate": 1.7184238473651445e-07, + "loss": 0.6967, + "step": 7739 + }, + { + "epoch": 0.28, + "learning_rate": 1.7183872267184237e-07, + "loss": 0.6998, + "step": 7740 + }, + { + "epoch": 0.28, + "learning_rate": 1.7183506060717033e-07, + "loss": 0.6934, + "step": 7741 + }, + { + "epoch": 0.28, + "learning_rate": 1.7183139854249825e-07, + "loss": 0.6949, + "step": 7742 + }, + { + "epoch": 0.28, + "learning_rate": 1.7182773647782618e-07, + "loss": 0.6976, + "step": 7743 + }, + { + "epoch": 0.28, + "learning_rate": 1.7182407441315413e-07, + "loss": 0.6866, + "step": 7744 + }, + { + "epoch": 0.28, + "learning_rate": 1.7182041234848206e-07, + "loss": 0.6757, + "step": 7745 + }, + { + "epoch": 0.28, + "learning_rate": 1.7181675028380999e-07, + "loss": 0.6931, + "step": 7746 + }, + { + "epoch": 0.28, + "learning_rate": 1.7181308821913794e-07, + "loss": 0.6953, + "step": 7747 + }, + { + "epoch": 0.28, + "learning_rate": 1.7180942615446586e-07, + "loss": 0.6904, + "step": 7748 + }, + { + "epoch": 0.28, + "learning_rate": 1.7180576408979382e-07, + "loss": 0.6853, + "step": 7749 + }, + { + "epoch": 0.28, + "learning_rate": 1.7180210202512174e-07, + "loss": 0.6828, + "step": 7750 + }, + { + "epoch": 0.28, + "learning_rate": 1.717984399604497e-07, + "loss": 0.6917, + "step": 7751 + }, + { + "epoch": 0.28, + "learning_rate": 1.7179477789577762e-07, + "loss": 0.7004, + "step": 7752 + }, + { + "epoch": 0.28, + "learning_rate": 1.7179111583110558e-07, + "loss": 0.7009, + "step": 7753 + }, + { + "epoch": 0.28, + "learning_rate": 1.717874537664335e-07, + "loss": 0.6786, + "step": 7754 + }, + { + "epoch": 0.28, + "learning_rate": 1.7178379170176146e-07, + "loss": 0.6879, + "step": 7755 + }, + { + "epoch": 0.28, + "learning_rate": 1.7178012963708938e-07, + "loss": 0.6918, + "step": 7756 + }, + { + "epoch": 0.28, + "learning_rate": 1.7177646757241734e-07, + "loss": 0.6952, + "step": 7757 + }, + { + "epoch": 0.28, + "learning_rate": 1.7177280550774526e-07, + "loss": 0.7019, + "step": 7758 + }, + { + "epoch": 0.28, + "learning_rate": 1.7176914344307322e-07, + "loss": 0.7016, + "step": 7759 + }, + { + "epoch": 0.28, + "learning_rate": 1.7176548137840114e-07, + "loss": 0.684, + "step": 7760 + }, + { + "epoch": 0.28, + "learning_rate": 1.7176181931372907e-07, + "loss": 0.6972, + "step": 7761 + }, + { + "epoch": 0.28, + "learning_rate": 1.71758157249057e-07, + "loss": 0.6962, + "step": 7762 + }, + { + "epoch": 0.28, + "learning_rate": 1.7175449518438495e-07, + "loss": 0.6978, + "step": 7763 + }, + { + "epoch": 0.28, + "learning_rate": 1.7175083311971288e-07, + "loss": 0.6939, + "step": 7764 + }, + { + "epoch": 0.28, + "learning_rate": 1.7174717105504083e-07, + "loss": 0.6842, + "step": 7765 + }, + { + "epoch": 0.28, + "learning_rate": 1.7174350899036876e-07, + "loss": 0.6882, + "step": 7766 + }, + { + "epoch": 0.28, + "learning_rate": 1.7173984692569668e-07, + "loss": 0.6943, + "step": 7767 + }, + { + "epoch": 0.28, + "learning_rate": 1.7173618486102464e-07, + "loss": 0.7002, + "step": 7768 + }, + { + "epoch": 0.28, + "learning_rate": 1.7173252279635256e-07, + "loss": 0.6796, + "step": 7769 + }, + { + "epoch": 0.28, + "learning_rate": 1.7172886073168052e-07, + "loss": 0.7017, + "step": 7770 + }, + { + "epoch": 0.28, + "learning_rate": 1.7172519866700844e-07, + "loss": 0.6935, + "step": 7771 + }, + { + "epoch": 0.28, + "learning_rate": 1.717215366023364e-07, + "loss": 0.6929, + "step": 7772 + }, + { + "epoch": 0.28, + "learning_rate": 1.7171787453766432e-07, + "loss": 0.6984, + "step": 7773 + }, + { + "epoch": 0.28, + "learning_rate": 1.7171421247299228e-07, + "loss": 0.6996, + "step": 7774 + }, + { + "epoch": 0.28, + "learning_rate": 1.717105504083202e-07, + "loss": 0.6878, + "step": 7775 + }, + { + "epoch": 0.28, + "learning_rate": 1.7170688834364816e-07, + "loss": 0.7091, + "step": 7776 + }, + { + "epoch": 0.28, + "learning_rate": 1.7170322627897608e-07, + "loss": 0.6914, + "step": 7777 + }, + { + "epoch": 0.28, + "learning_rate": 1.7169956421430404e-07, + "loss": 0.691, + "step": 7778 + }, + { + "epoch": 0.28, + "learning_rate": 1.7169590214963194e-07, + "loss": 0.6675, + "step": 7779 + }, + { + "epoch": 0.28, + "learning_rate": 1.716922400849599e-07, + "loss": 0.6975, + "step": 7780 + }, + { + "epoch": 0.28, + "learning_rate": 1.7168857802028782e-07, + "loss": 0.7025, + "step": 7781 + }, + { + "epoch": 0.28, + "learning_rate": 1.7168491595561577e-07, + "loss": 0.6956, + "step": 7782 + }, + { + "epoch": 0.28, + "learning_rate": 1.716812538909437e-07, + "loss": 0.6847, + "step": 7783 + }, + { + "epoch": 0.28, + "learning_rate": 1.7167759182627165e-07, + "loss": 0.6888, + "step": 7784 + }, + { + "epoch": 0.28, + "learning_rate": 1.7167392976159958e-07, + "loss": 0.7196, + "step": 7785 + }, + { + "epoch": 0.28, + "learning_rate": 1.7167026769692753e-07, + "loss": 0.7043, + "step": 7786 + }, + { + "epoch": 0.28, + "learning_rate": 1.7166660563225546e-07, + "loss": 0.7018, + "step": 7787 + }, + { + "epoch": 0.28, + "learning_rate": 1.716629435675834e-07, + "loss": 0.6944, + "step": 7788 + }, + { + "epoch": 0.28, + "learning_rate": 1.7165928150291134e-07, + "loss": 0.7016, + "step": 7789 + }, + { + "epoch": 0.29, + "learning_rate": 1.7165561943823926e-07, + "loss": 0.6752, + "step": 7790 + }, + { + "epoch": 0.29, + "learning_rate": 1.7165195737356722e-07, + "loss": 0.6984, + "step": 7791 + }, + { + "epoch": 0.29, + "learning_rate": 1.7164829530889514e-07, + "loss": 0.7039, + "step": 7792 + }, + { + "epoch": 0.29, + "learning_rate": 1.716446332442231e-07, + "loss": 0.7044, + "step": 7793 + }, + { + "epoch": 0.29, + "learning_rate": 1.7164097117955102e-07, + "loss": 0.6928, + "step": 7794 + }, + { + "epoch": 0.29, + "learning_rate": 1.7163730911487898e-07, + "loss": 0.7077, + "step": 7795 + }, + { + "epoch": 0.29, + "learning_rate": 1.716336470502069e-07, + "loss": 0.6953, + "step": 7796 + }, + { + "epoch": 0.29, + "learning_rate": 1.7162998498553483e-07, + "loss": 0.6838, + "step": 7797 + }, + { + "epoch": 0.29, + "learning_rate": 1.7162632292086276e-07, + "loss": 0.7047, + "step": 7798 + }, + { + "epoch": 0.29, + "learning_rate": 1.716226608561907e-07, + "loss": 0.6913, + "step": 7799 + }, + { + "epoch": 0.29, + "learning_rate": 1.7161899879151864e-07, + "loss": 0.7045, + "step": 7800 + }, + { + "epoch": 0.29, + "learning_rate": 1.716153367268466e-07, + "loss": 0.6902, + "step": 7801 + }, + { + "epoch": 0.29, + "learning_rate": 1.7161167466217452e-07, + "loss": 0.6837, + "step": 7802 + }, + { + "epoch": 0.29, + "learning_rate": 1.7160801259750247e-07, + "loss": 0.6917, + "step": 7803 + }, + { + "epoch": 0.29, + "learning_rate": 1.716043505328304e-07, + "loss": 0.6909, + "step": 7804 + }, + { + "epoch": 0.29, + "learning_rate": 1.7160068846815835e-07, + "loss": 0.693, + "step": 7805 + }, + { + "epoch": 0.29, + "learning_rate": 1.7159702640348627e-07, + "loss": 0.7012, + "step": 7806 + }, + { + "epoch": 0.29, + "learning_rate": 1.7159336433881423e-07, + "loss": 0.6803, + "step": 7807 + }, + { + "epoch": 0.29, + "learning_rate": 1.7158970227414215e-07, + "loss": 0.6943, + "step": 7808 + }, + { + "epoch": 0.29, + "learning_rate": 1.715860402094701e-07, + "loss": 0.6947, + "step": 7809 + }, + { + "epoch": 0.29, + "learning_rate": 1.7158237814479803e-07, + "loss": 0.6921, + "step": 7810 + }, + { + "epoch": 0.29, + "learning_rate": 1.7157871608012596e-07, + "loss": 0.6932, + "step": 7811 + }, + { + "epoch": 0.29, + "learning_rate": 1.7157505401545391e-07, + "loss": 0.6994, + "step": 7812 + }, + { + "epoch": 0.29, + "learning_rate": 1.7157139195078184e-07, + "loss": 0.6875, + "step": 7813 + }, + { + "epoch": 0.29, + "learning_rate": 1.715677298861098e-07, + "loss": 0.7301, + "step": 7814 + }, + { + "epoch": 0.29, + "learning_rate": 1.7156406782143772e-07, + "loss": 0.692, + "step": 7815 + }, + { + "epoch": 0.29, + "learning_rate": 1.7156040575676565e-07, + "loss": 0.7022, + "step": 7816 + }, + { + "epoch": 0.29, + "learning_rate": 1.7155674369209357e-07, + "loss": 0.6909, + "step": 7817 + }, + { + "epoch": 0.29, + "learning_rate": 1.7155308162742153e-07, + "loss": 0.6939, + "step": 7818 + }, + { + "epoch": 0.29, + "learning_rate": 1.7154941956274945e-07, + "loss": 0.7016, + "step": 7819 + }, + { + "epoch": 0.29, + "learning_rate": 1.715457574980774e-07, + "loss": 0.6966, + "step": 7820 + }, + { + "epoch": 0.29, + "learning_rate": 1.7154209543340533e-07, + "loss": 0.6917, + "step": 7821 + }, + { + "epoch": 0.29, + "learning_rate": 1.715384333687333e-07, + "loss": 0.6939, + "step": 7822 + }, + { + "epoch": 0.29, + "learning_rate": 1.7153477130406121e-07, + "loss": 0.705, + "step": 7823 + }, + { + "epoch": 0.29, + "learning_rate": 1.7153110923938917e-07, + "loss": 0.7008, + "step": 7824 + }, + { + "epoch": 0.29, + "learning_rate": 1.715274471747171e-07, + "loss": 0.7116, + "step": 7825 + }, + { + "epoch": 0.29, + "learning_rate": 1.7152378511004505e-07, + "loss": 0.6976, + "step": 7826 + }, + { + "epoch": 0.29, + "learning_rate": 1.7152012304537297e-07, + "loss": 0.6913, + "step": 7827 + }, + { + "epoch": 0.29, + "learning_rate": 1.7151646098070093e-07, + "loss": 0.6866, + "step": 7828 + }, + { + "epoch": 0.29, + "learning_rate": 1.7151279891602885e-07, + "loss": 0.6816, + "step": 7829 + }, + { + "epoch": 0.29, + "learning_rate": 1.715091368513568e-07, + "loss": 0.6913, + "step": 7830 + }, + { + "epoch": 0.29, + "learning_rate": 1.7150547478668473e-07, + "loss": 0.6967, + "step": 7831 + }, + { + "epoch": 0.29, + "learning_rate": 1.7150181272201266e-07, + "loss": 0.6806, + "step": 7832 + }, + { + "epoch": 0.29, + "learning_rate": 1.7149815065734059e-07, + "loss": 0.7176, + "step": 7833 + }, + { + "epoch": 0.29, + "learning_rate": 1.7149448859266854e-07, + "loss": 0.7131, + "step": 7834 + }, + { + "epoch": 0.29, + "learning_rate": 1.7149082652799647e-07, + "loss": 0.6888, + "step": 7835 + }, + { + "epoch": 0.29, + "learning_rate": 1.7148716446332442e-07, + "loss": 0.6986, + "step": 7836 + }, + { + "epoch": 0.29, + "learning_rate": 1.7148350239865235e-07, + "loss": 0.7014, + "step": 7837 + }, + { + "epoch": 0.29, + "learning_rate": 1.7147984033398027e-07, + "loss": 0.6811, + "step": 7838 + }, + { + "epoch": 0.29, + "learning_rate": 1.7147617826930823e-07, + "loss": 0.704, + "step": 7839 + }, + { + "epoch": 0.29, + "learning_rate": 1.7147251620463615e-07, + "loss": 0.7131, + "step": 7840 + }, + { + "epoch": 0.29, + "learning_rate": 1.714688541399641e-07, + "loss": 0.6937, + "step": 7841 + }, + { + "epoch": 0.29, + "learning_rate": 1.7146519207529203e-07, + "loss": 0.7023, + "step": 7842 + }, + { + "epoch": 0.29, + "learning_rate": 1.7146153001061999e-07, + "loss": 0.6978, + "step": 7843 + }, + { + "epoch": 0.29, + "learning_rate": 1.714578679459479e-07, + "loss": 0.693, + "step": 7844 + }, + { + "epoch": 0.29, + "learning_rate": 1.7145420588127587e-07, + "loss": 0.6932, + "step": 7845 + }, + { + "epoch": 0.29, + "learning_rate": 1.714505438166038e-07, + "loss": 0.7012, + "step": 7846 + }, + { + "epoch": 0.29, + "learning_rate": 1.7144688175193175e-07, + "loss": 0.6895, + "step": 7847 + }, + { + "epoch": 0.29, + "learning_rate": 1.7144321968725967e-07, + "loss": 0.7052, + "step": 7848 + }, + { + "epoch": 0.29, + "learning_rate": 1.7143955762258763e-07, + "loss": 0.6951, + "step": 7849 + }, + { + "epoch": 0.29, + "learning_rate": 1.7143589555791555e-07, + "loss": 0.6959, + "step": 7850 + }, + { + "epoch": 0.29, + "learning_rate": 1.7143223349324348e-07, + "loss": 0.6963, + "step": 7851 + }, + { + "epoch": 0.29, + "learning_rate": 1.714285714285714e-07, + "loss": 0.6932, + "step": 7852 + }, + { + "epoch": 0.29, + "learning_rate": 1.7142490936389936e-07, + "loss": 0.6988, + "step": 7853 + }, + { + "epoch": 0.29, + "learning_rate": 1.7142124729922729e-07, + "loss": 0.6893, + "step": 7854 + }, + { + "epoch": 0.29, + "learning_rate": 1.7141758523455524e-07, + "loss": 0.7041, + "step": 7855 + }, + { + "epoch": 0.29, + "learning_rate": 1.7141392316988317e-07, + "loss": 0.6871, + "step": 7856 + }, + { + "epoch": 0.29, + "learning_rate": 1.7141026110521112e-07, + "loss": 0.69, + "step": 7857 + }, + { + "epoch": 0.29, + "learning_rate": 1.7140659904053905e-07, + "loss": 0.6876, + "step": 7858 + }, + { + "epoch": 0.29, + "learning_rate": 1.71402936975867e-07, + "loss": 0.6945, + "step": 7859 + }, + { + "epoch": 0.29, + "learning_rate": 1.7139927491119493e-07, + "loss": 0.6818, + "step": 7860 + }, + { + "epoch": 0.29, + "learning_rate": 1.7139561284652285e-07, + "loss": 0.6917, + "step": 7861 + }, + { + "epoch": 0.29, + "learning_rate": 1.713919507818508e-07, + "loss": 0.6738, + "step": 7862 + }, + { + "epoch": 0.29, + "learning_rate": 1.7138828871717873e-07, + "loss": 0.68, + "step": 7863 + }, + { + "epoch": 0.29, + "learning_rate": 1.7138462665250668e-07, + "loss": 0.6777, + "step": 7864 + }, + { + "epoch": 0.29, + "learning_rate": 1.713809645878346e-07, + "loss": 0.6805, + "step": 7865 + }, + { + "epoch": 0.29, + "learning_rate": 1.7137730252316256e-07, + "loss": 0.6924, + "step": 7866 + }, + { + "epoch": 0.29, + "learning_rate": 1.713736404584905e-07, + "loss": 0.7015, + "step": 7867 + }, + { + "epoch": 0.29, + "learning_rate": 1.7136997839381844e-07, + "loss": 0.6963, + "step": 7868 + }, + { + "epoch": 0.29, + "learning_rate": 1.7136631632914634e-07, + "loss": 0.69, + "step": 7869 + }, + { + "epoch": 0.29, + "learning_rate": 1.713626542644743e-07, + "loss": 0.6903, + "step": 7870 + }, + { + "epoch": 0.29, + "learning_rate": 1.7135899219980222e-07, + "loss": 0.6943, + "step": 7871 + }, + { + "epoch": 0.29, + "learning_rate": 1.7135533013513018e-07, + "loss": 0.673, + "step": 7872 + }, + { + "epoch": 0.29, + "learning_rate": 1.713516680704581e-07, + "loss": 0.6905, + "step": 7873 + }, + { + "epoch": 0.29, + "learning_rate": 1.7134800600578606e-07, + "loss": 0.6809, + "step": 7874 + }, + { + "epoch": 0.29, + "learning_rate": 1.7134434394111398e-07, + "loss": 0.6939, + "step": 7875 + }, + { + "epoch": 0.29, + "learning_rate": 1.7134068187644194e-07, + "loss": 0.7036, + "step": 7876 + }, + { + "epoch": 0.29, + "learning_rate": 1.7133701981176986e-07, + "loss": 0.6861, + "step": 7877 + }, + { + "epoch": 0.29, + "learning_rate": 1.7133335774709782e-07, + "loss": 0.6811, + "step": 7878 + }, + { + "epoch": 0.29, + "learning_rate": 1.7132969568242574e-07, + "loss": 0.6951, + "step": 7879 + }, + { + "epoch": 0.29, + "learning_rate": 1.713260336177537e-07, + "loss": 0.6967, + "step": 7880 + }, + { + "epoch": 0.29, + "learning_rate": 1.7132237155308162e-07, + "loss": 0.6929, + "step": 7881 + }, + { + "epoch": 0.29, + "learning_rate": 1.7131870948840955e-07, + "loss": 0.6976, + "step": 7882 + }, + { + "epoch": 0.29, + "learning_rate": 1.713150474237375e-07, + "loss": 0.6863, + "step": 7883 + }, + { + "epoch": 0.29, + "learning_rate": 1.7131138535906543e-07, + "loss": 0.6989, + "step": 7884 + }, + { + "epoch": 0.29, + "learning_rate": 1.7130772329439338e-07, + "loss": 0.6737, + "step": 7885 + }, + { + "epoch": 0.29, + "learning_rate": 1.713040612297213e-07, + "loss": 0.7057, + "step": 7886 + }, + { + "epoch": 0.29, + "learning_rate": 1.7130039916504924e-07, + "loss": 0.7014, + "step": 7887 + }, + { + "epoch": 0.29, + "learning_rate": 1.7129673710037716e-07, + "loss": 0.6882, + "step": 7888 + }, + { + "epoch": 0.29, + "learning_rate": 1.7129307503570512e-07, + "loss": 0.6993, + "step": 7889 + }, + { + "epoch": 0.29, + "learning_rate": 1.7128941297103304e-07, + "loss": 0.6942, + "step": 7890 + }, + { + "epoch": 0.29, + "learning_rate": 1.71285750906361e-07, + "loss": 0.6945, + "step": 7891 + }, + { + "epoch": 0.29, + "learning_rate": 1.7128208884168892e-07, + "loss": 0.6908, + "step": 7892 + }, + { + "epoch": 0.29, + "learning_rate": 1.7127842677701688e-07, + "loss": 0.7167, + "step": 7893 + }, + { + "epoch": 0.29, + "learning_rate": 1.712747647123448e-07, + "loss": 0.6908, + "step": 7894 + }, + { + "epoch": 0.29, + "learning_rate": 1.7127110264767276e-07, + "loss": 0.6819, + "step": 7895 + }, + { + "epoch": 0.29, + "learning_rate": 1.7126744058300068e-07, + "loss": 0.6998, + "step": 7896 + }, + { + "epoch": 0.29, + "learning_rate": 1.7126377851832864e-07, + "loss": 0.7044, + "step": 7897 + }, + { + "epoch": 0.29, + "learning_rate": 1.7126011645365656e-07, + "loss": 0.7101, + "step": 7898 + }, + { + "epoch": 0.29, + "learning_rate": 1.7125645438898452e-07, + "loss": 0.6639, + "step": 7899 + }, + { + "epoch": 0.29, + "learning_rate": 1.7125279232431244e-07, + "loss": 0.6839, + "step": 7900 + }, + { + "epoch": 0.29, + "learning_rate": 1.712491302596404e-07, + "loss": 0.6706, + "step": 7901 + }, + { + "epoch": 0.29, + "learning_rate": 1.7124546819496832e-07, + "loss": 0.7004, + "step": 7902 + }, + { + "epoch": 0.29, + "learning_rate": 1.7124180613029625e-07, + "loss": 0.6883, + "step": 7903 + }, + { + "epoch": 0.29, + "learning_rate": 1.712381440656242e-07, + "loss": 0.6919, + "step": 7904 + }, + { + "epoch": 0.29, + "learning_rate": 1.7123448200095213e-07, + "loss": 0.7009, + "step": 7905 + }, + { + "epoch": 0.29, + "learning_rate": 1.7123081993628006e-07, + "loss": 0.6863, + "step": 7906 + }, + { + "epoch": 0.29, + "learning_rate": 1.71227157871608e-07, + "loss": 0.6887, + "step": 7907 + }, + { + "epoch": 0.29, + "learning_rate": 1.7122349580693594e-07, + "loss": 0.7005, + "step": 7908 + }, + { + "epoch": 0.29, + "learning_rate": 1.7121983374226386e-07, + "loss": 0.6978, + "step": 7909 + }, + { + "epoch": 0.29, + "learning_rate": 1.7121617167759182e-07, + "loss": 0.6915, + "step": 7910 + }, + { + "epoch": 0.29, + "learning_rate": 1.7121250961291974e-07, + "loss": 0.7045, + "step": 7911 + }, + { + "epoch": 0.29, + "learning_rate": 1.712088475482477e-07, + "loss": 0.6978, + "step": 7912 + }, + { + "epoch": 0.29, + "learning_rate": 1.7120518548357562e-07, + "loss": 0.6879, + "step": 7913 + }, + { + "epoch": 0.29, + "learning_rate": 1.7120152341890358e-07, + "loss": 0.6794, + "step": 7914 + }, + { + "epoch": 0.29, + "learning_rate": 1.711978613542315e-07, + "loss": 0.6937, + "step": 7915 + }, + { + "epoch": 0.29, + "learning_rate": 1.7119419928955946e-07, + "loss": 0.6975, + "step": 7916 + }, + { + "epoch": 0.29, + "learning_rate": 1.7119053722488738e-07, + "loss": 0.7041, + "step": 7917 + }, + { + "epoch": 0.29, + "learning_rate": 1.7118687516021534e-07, + "loss": 0.6966, + "step": 7918 + }, + { + "epoch": 0.29, + "learning_rate": 1.7118321309554326e-07, + "loss": 0.684, + "step": 7919 + }, + { + "epoch": 0.29, + "learning_rate": 1.7117955103087121e-07, + "loss": 0.6864, + "step": 7920 + }, + { + "epoch": 0.29, + "learning_rate": 1.7117588896619914e-07, + "loss": 0.6998, + "step": 7921 + }, + { + "epoch": 0.29, + "learning_rate": 1.711722269015271e-07, + "loss": 0.7044, + "step": 7922 + }, + { + "epoch": 0.29, + "learning_rate": 1.71168564836855e-07, + "loss": 0.6874, + "step": 7923 + }, + { + "epoch": 0.29, + "learning_rate": 1.7116490277218295e-07, + "loss": 0.6795, + "step": 7924 + }, + { + "epoch": 0.29, + "learning_rate": 1.7116124070751087e-07, + "loss": 0.698, + "step": 7925 + }, + { + "epoch": 0.29, + "learning_rate": 1.7115757864283883e-07, + "loss": 0.6856, + "step": 7926 + }, + { + "epoch": 0.29, + "learning_rate": 1.7115391657816675e-07, + "loss": 0.6677, + "step": 7927 + }, + { + "epoch": 0.29, + "learning_rate": 1.711502545134947e-07, + "loss": 0.6814, + "step": 7928 + }, + { + "epoch": 0.29, + "learning_rate": 1.7114659244882263e-07, + "loss": 0.6929, + "step": 7929 + }, + { + "epoch": 0.29, + "learning_rate": 1.7114293038415056e-07, + "loss": 0.6957, + "step": 7930 + }, + { + "epoch": 0.29, + "learning_rate": 1.7113926831947851e-07, + "loss": 0.7022, + "step": 7931 + }, + { + "epoch": 0.29, + "learning_rate": 1.7113560625480644e-07, + "loss": 0.6917, + "step": 7932 + }, + { + "epoch": 0.29, + "learning_rate": 1.711319441901344e-07, + "loss": 0.7023, + "step": 7933 + }, + { + "epoch": 0.29, + "learning_rate": 1.7112828212546232e-07, + "loss": 0.6933, + "step": 7934 + }, + { + "epoch": 0.29, + "learning_rate": 1.7112462006079027e-07, + "loss": 0.699, + "step": 7935 + }, + { + "epoch": 0.29, + "learning_rate": 1.711209579961182e-07, + "loss": 0.709, + "step": 7936 + }, + { + "epoch": 0.29, + "learning_rate": 1.7111729593144615e-07, + "loss": 0.7032, + "step": 7937 + }, + { + "epoch": 0.29, + "learning_rate": 1.7111363386677408e-07, + "loss": 0.702, + "step": 7938 + }, + { + "epoch": 0.29, + "learning_rate": 1.7110997180210203e-07, + "loss": 0.6982, + "step": 7939 + }, + { + "epoch": 0.29, + "learning_rate": 1.7110630973742996e-07, + "loss": 0.6878, + "step": 7940 + }, + { + "epoch": 0.29, + "learning_rate": 1.711026476727579e-07, + "loss": 0.6945, + "step": 7941 + }, + { + "epoch": 0.29, + "learning_rate": 1.7109898560808581e-07, + "loss": 0.6973, + "step": 7942 + }, + { + "epoch": 0.29, + "learning_rate": 1.7109532354341377e-07, + "loss": 0.7027, + "step": 7943 + }, + { + "epoch": 0.29, + "learning_rate": 1.710916614787417e-07, + "loss": 0.692, + "step": 7944 + }, + { + "epoch": 0.29, + "learning_rate": 1.7108799941406965e-07, + "loss": 0.695, + "step": 7945 + }, + { + "epoch": 0.29, + "learning_rate": 1.7108433734939757e-07, + "loss": 0.6963, + "step": 7946 + }, + { + "epoch": 0.29, + "learning_rate": 1.7108067528472553e-07, + "loss": 0.6858, + "step": 7947 + }, + { + "epoch": 0.29, + "learning_rate": 1.7107701322005345e-07, + "loss": 0.6888, + "step": 7948 + }, + { + "epoch": 0.29, + "learning_rate": 1.710733511553814e-07, + "loss": 0.6864, + "step": 7949 + }, + { + "epoch": 0.29, + "learning_rate": 1.7106968909070933e-07, + "loss": 0.6883, + "step": 7950 + }, + { + "epoch": 0.29, + "learning_rate": 1.7106602702603729e-07, + "loss": 0.704, + "step": 7951 + }, + { + "epoch": 0.29, + "learning_rate": 1.7106236496136521e-07, + "loss": 0.699, + "step": 7952 + }, + { + "epoch": 0.29, + "learning_rate": 1.7105870289669314e-07, + "loss": 0.6891, + "step": 7953 + }, + { + "epoch": 0.29, + "learning_rate": 1.710550408320211e-07, + "loss": 0.6952, + "step": 7954 + }, + { + "epoch": 0.29, + "learning_rate": 1.7105137876734902e-07, + "loss": 0.7075, + "step": 7955 + }, + { + "epoch": 0.29, + "learning_rate": 1.7104771670267697e-07, + "loss": 0.6922, + "step": 7956 + }, + { + "epoch": 0.29, + "learning_rate": 1.710440546380049e-07, + "loss": 0.7006, + "step": 7957 + }, + { + "epoch": 0.29, + "learning_rate": 1.7104039257333285e-07, + "loss": 0.6907, + "step": 7958 + }, + { + "epoch": 0.29, + "learning_rate": 1.7103673050866075e-07, + "loss": 0.6986, + "step": 7959 + }, + { + "epoch": 0.29, + "learning_rate": 1.710330684439887e-07, + "loss": 0.6744, + "step": 7960 + }, + { + "epoch": 0.29, + "learning_rate": 1.7102940637931663e-07, + "loss": 0.7039, + "step": 7961 + }, + { + "epoch": 0.29, + "learning_rate": 1.7102574431464459e-07, + "loss": 0.6955, + "step": 7962 + }, + { + "epoch": 0.29, + "learning_rate": 1.710220822499725e-07, + "loss": 0.6851, + "step": 7963 + }, + { + "epoch": 0.29, + "learning_rate": 1.7101842018530047e-07, + "loss": 0.6922, + "step": 7964 + }, + { + "epoch": 0.29, + "learning_rate": 1.710147581206284e-07, + "loss": 0.6956, + "step": 7965 + }, + { + "epoch": 0.29, + "learning_rate": 1.7101109605595635e-07, + "loss": 0.6955, + "step": 7966 + }, + { + "epoch": 0.29, + "learning_rate": 1.7100743399128427e-07, + "loss": 0.6877, + "step": 7967 + }, + { + "epoch": 0.29, + "learning_rate": 1.7100377192661223e-07, + "loss": 0.7044, + "step": 7968 + }, + { + "epoch": 0.29, + "learning_rate": 1.7100010986194015e-07, + "loss": 0.68, + "step": 7969 + }, + { + "epoch": 0.29, + "learning_rate": 1.709964477972681e-07, + "loss": 0.7044, + "step": 7970 + }, + { + "epoch": 0.29, + "learning_rate": 1.7099278573259603e-07, + "loss": 0.7058, + "step": 7971 + }, + { + "epoch": 0.29, + "learning_rate": 1.7098912366792399e-07, + "loss": 0.674, + "step": 7972 + }, + { + "epoch": 0.29, + "learning_rate": 1.709854616032519e-07, + "loss": 0.6834, + "step": 7973 + }, + { + "epoch": 0.29, + "learning_rate": 1.7098179953857984e-07, + "loss": 0.704, + "step": 7974 + }, + { + "epoch": 0.29, + "learning_rate": 1.709781374739078e-07, + "loss": 0.6991, + "step": 7975 + }, + { + "epoch": 0.29, + "learning_rate": 1.7097447540923572e-07, + "loss": 0.7084, + "step": 7976 + }, + { + "epoch": 0.29, + "learning_rate": 1.7097081334456365e-07, + "loss": 0.6956, + "step": 7977 + }, + { + "epoch": 0.29, + "learning_rate": 1.709671512798916e-07, + "loss": 0.7013, + "step": 7978 + }, + { + "epoch": 0.29, + "learning_rate": 1.7096348921521953e-07, + "loss": 0.6702, + "step": 7979 + }, + { + "epoch": 0.29, + "learning_rate": 1.7095982715054745e-07, + "loss": 0.6788, + "step": 7980 + }, + { + "epoch": 0.29, + "learning_rate": 1.709561650858754e-07, + "loss": 0.666, + "step": 7981 + }, + { + "epoch": 0.29, + "learning_rate": 1.7095250302120333e-07, + "loss": 0.705, + "step": 7982 + }, + { + "epoch": 0.29, + "learning_rate": 1.7094884095653128e-07, + "loss": 0.684, + "step": 7983 + }, + { + "epoch": 0.29, + "learning_rate": 1.709451788918592e-07, + "loss": 0.6925, + "step": 7984 + }, + { + "epoch": 0.29, + "learning_rate": 1.7094151682718716e-07, + "loss": 0.6839, + "step": 7985 + }, + { + "epoch": 0.29, + "learning_rate": 1.709378547625151e-07, + "loss": 0.6974, + "step": 7986 + }, + { + "epoch": 0.29, + "learning_rate": 1.7093419269784304e-07, + "loss": 0.7026, + "step": 7987 + }, + { + "epoch": 0.29, + "learning_rate": 1.7093053063317097e-07, + "loss": 0.701, + "step": 7988 + }, + { + "epoch": 0.29, + "learning_rate": 1.7092686856849892e-07, + "loss": 0.7004, + "step": 7989 + }, + { + "epoch": 0.29, + "learning_rate": 1.7092320650382685e-07, + "loss": 0.6985, + "step": 7990 + }, + { + "epoch": 0.29, + "learning_rate": 1.709195444391548e-07, + "loss": 0.6999, + "step": 7991 + }, + { + "epoch": 0.29, + "learning_rate": 1.7091588237448273e-07, + "loss": 0.7063, + "step": 7992 + }, + { + "epoch": 0.29, + "learning_rate": 1.7091222030981068e-07, + "loss": 0.6968, + "step": 7993 + }, + { + "epoch": 0.29, + "learning_rate": 1.709085582451386e-07, + "loss": 0.6763, + "step": 7994 + }, + { + "epoch": 0.29, + "learning_rate": 1.7090489618046654e-07, + "loss": 0.6917, + "step": 7995 + }, + { + "epoch": 0.29, + "learning_rate": 1.7090123411579446e-07, + "loss": 0.7009, + "step": 7996 + }, + { + "epoch": 0.29, + "learning_rate": 1.7089757205112242e-07, + "loss": 0.6854, + "step": 7997 + }, + { + "epoch": 0.29, + "learning_rate": 1.7089390998645034e-07, + "loss": 0.6881, + "step": 7998 + }, + { + "epoch": 0.29, + "learning_rate": 1.708902479217783e-07, + "loss": 0.7007, + "step": 7999 + }, + { + "epoch": 0.29, + "learning_rate": 1.7088658585710622e-07, + "loss": 0.7008, + "step": 8000 + }, + { + "epoch": 0.29, + "learning_rate": 1.7088292379243415e-07, + "loss": 0.6835, + "step": 8001 + }, + { + "epoch": 0.29, + "learning_rate": 1.708792617277621e-07, + "loss": 0.6999, + "step": 8002 + }, + { + "epoch": 0.29, + "learning_rate": 1.7087559966309003e-07, + "loss": 0.6907, + "step": 8003 + }, + { + "epoch": 0.29, + "learning_rate": 1.7087193759841798e-07, + "loss": 0.6932, + "step": 8004 + }, + { + "epoch": 0.29, + "learning_rate": 1.708682755337459e-07, + "loss": 0.6812, + "step": 8005 + }, + { + "epoch": 0.29, + "learning_rate": 1.7086461346907386e-07, + "loss": 0.7022, + "step": 8006 + }, + { + "epoch": 0.29, + "learning_rate": 1.708609514044018e-07, + "loss": 0.6999, + "step": 8007 + }, + { + "epoch": 0.29, + "learning_rate": 1.7085728933972974e-07, + "loss": 0.6909, + "step": 8008 + }, + { + "epoch": 0.29, + "learning_rate": 1.7085362727505767e-07, + "loss": 0.7018, + "step": 8009 + }, + { + "epoch": 0.29, + "learning_rate": 1.7084996521038562e-07, + "loss": 0.6734, + "step": 8010 + }, + { + "epoch": 0.29, + "learning_rate": 1.7084630314571355e-07, + "loss": 0.6864, + "step": 8011 + }, + { + "epoch": 0.29, + "learning_rate": 1.708426410810415e-07, + "loss": 0.6885, + "step": 8012 + }, + { + "epoch": 0.29, + "learning_rate": 1.708389790163694e-07, + "loss": 0.703, + "step": 8013 + }, + { + "epoch": 0.29, + "learning_rate": 1.7083531695169736e-07, + "loss": 0.7151, + "step": 8014 + }, + { + "epoch": 0.29, + "learning_rate": 1.7083165488702528e-07, + "loss": 0.6993, + "step": 8015 + }, + { + "epoch": 0.29, + "learning_rate": 1.7082799282235324e-07, + "loss": 0.6871, + "step": 8016 + }, + { + "epoch": 0.29, + "learning_rate": 1.7082433075768116e-07, + "loss": 0.6873, + "step": 8017 + }, + { + "epoch": 0.29, + "learning_rate": 1.7082066869300912e-07, + "loss": 0.69, + "step": 8018 + }, + { + "epoch": 0.29, + "learning_rate": 1.7081700662833704e-07, + "loss": 0.6923, + "step": 8019 + }, + { + "epoch": 0.29, + "learning_rate": 1.70813344563665e-07, + "loss": 0.6979, + "step": 8020 + }, + { + "epoch": 0.29, + "learning_rate": 1.7080968249899292e-07, + "loss": 0.6839, + "step": 8021 + }, + { + "epoch": 0.29, + "learning_rate": 1.7080602043432088e-07, + "loss": 0.6801, + "step": 8022 + }, + { + "epoch": 0.29, + "learning_rate": 1.708023583696488e-07, + "loss": 0.6802, + "step": 8023 + }, + { + "epoch": 0.29, + "learning_rate": 1.7079869630497673e-07, + "loss": 0.6997, + "step": 8024 + }, + { + "epoch": 0.29, + "learning_rate": 1.7079503424030468e-07, + "loss": 0.6841, + "step": 8025 + }, + { + "epoch": 0.29, + "learning_rate": 1.707913721756326e-07, + "loss": 0.7069, + "step": 8026 + }, + { + "epoch": 0.29, + "learning_rate": 1.7078771011096056e-07, + "loss": 0.6858, + "step": 8027 + }, + { + "epoch": 0.29, + "learning_rate": 1.707840480462885e-07, + "loss": 0.7003, + "step": 8028 + }, + { + "epoch": 0.29, + "learning_rate": 1.7078038598161644e-07, + "loss": 0.6893, + "step": 8029 + }, + { + "epoch": 0.29, + "learning_rate": 1.7077672391694437e-07, + "loss": 0.6901, + "step": 8030 + }, + { + "epoch": 0.29, + "learning_rate": 1.707730618522723e-07, + "loss": 0.7029, + "step": 8031 + }, + { + "epoch": 0.29, + "learning_rate": 1.7076939978760022e-07, + "loss": 0.7012, + "step": 8032 + }, + { + "epoch": 0.29, + "learning_rate": 1.7076573772292818e-07, + "loss": 0.6847, + "step": 8033 + }, + { + "epoch": 0.29, + "learning_rate": 1.707620756582561e-07, + "loss": 0.6953, + "step": 8034 + }, + { + "epoch": 0.29, + "learning_rate": 1.7075841359358406e-07, + "loss": 0.6956, + "step": 8035 + }, + { + "epoch": 0.29, + "learning_rate": 1.7075475152891198e-07, + "loss": 0.7009, + "step": 8036 + }, + { + "epoch": 0.29, + "learning_rate": 1.7075108946423994e-07, + "loss": 0.703, + "step": 8037 + }, + { + "epoch": 0.29, + "learning_rate": 1.7074742739956786e-07, + "loss": 0.6696, + "step": 8038 + }, + { + "epoch": 0.29, + "learning_rate": 1.7074376533489581e-07, + "loss": 0.6792, + "step": 8039 + }, + { + "epoch": 0.29, + "learning_rate": 1.7074010327022374e-07, + "loss": 0.6892, + "step": 8040 + }, + { + "epoch": 0.29, + "learning_rate": 1.707364412055517e-07, + "loss": 0.7042, + "step": 8041 + }, + { + "epoch": 0.29, + "learning_rate": 1.7073277914087962e-07, + "loss": 0.6936, + "step": 8042 + }, + { + "epoch": 0.29, + "learning_rate": 1.7072911707620757e-07, + "loss": 0.7152, + "step": 8043 + }, + { + "epoch": 0.29, + "learning_rate": 1.707254550115355e-07, + "loss": 0.7036, + "step": 8044 + }, + { + "epoch": 0.29, + "learning_rate": 1.7072179294686343e-07, + "loss": 0.6951, + "step": 8045 + }, + { + "epoch": 0.29, + "learning_rate": 1.7071813088219138e-07, + "loss": 0.6919, + "step": 8046 + }, + { + "epoch": 0.29, + "learning_rate": 1.707144688175193e-07, + "loss": 0.6831, + "step": 8047 + }, + { + "epoch": 0.29, + "learning_rate": 1.7071080675284726e-07, + "loss": 0.6956, + "step": 8048 + }, + { + "epoch": 0.29, + "learning_rate": 1.707071446881752e-07, + "loss": 0.6796, + "step": 8049 + }, + { + "epoch": 0.29, + "learning_rate": 1.7070348262350311e-07, + "loss": 0.6699, + "step": 8050 + }, + { + "epoch": 0.29, + "learning_rate": 1.7069982055883104e-07, + "loss": 0.6814, + "step": 8051 + }, + { + "epoch": 0.29, + "learning_rate": 1.70696158494159e-07, + "loss": 0.7019, + "step": 8052 + }, + { + "epoch": 0.29, + "learning_rate": 1.7069249642948692e-07, + "loss": 0.6984, + "step": 8053 + }, + { + "epoch": 0.29, + "learning_rate": 1.7068883436481487e-07, + "loss": 0.6802, + "step": 8054 + }, + { + "epoch": 0.29, + "learning_rate": 1.706851723001428e-07, + "loss": 0.6864, + "step": 8055 + }, + { + "epoch": 0.29, + "learning_rate": 1.7068151023547075e-07, + "loss": 0.6948, + "step": 8056 + }, + { + "epoch": 0.29, + "learning_rate": 1.7067784817079868e-07, + "loss": 0.6938, + "step": 8057 + }, + { + "epoch": 0.29, + "learning_rate": 1.7067418610612663e-07, + "loss": 0.6888, + "step": 8058 + }, + { + "epoch": 0.29, + "learning_rate": 1.7067052404145456e-07, + "loss": 0.6821, + "step": 8059 + }, + { + "epoch": 0.29, + "learning_rate": 1.7066686197678251e-07, + "loss": 0.6844, + "step": 8060 + }, + { + "epoch": 0.29, + "learning_rate": 1.7066319991211044e-07, + "loss": 0.6975, + "step": 8061 + }, + { + "epoch": 0.29, + "learning_rate": 1.706595378474384e-07, + "loss": 0.6875, + "step": 8062 + }, + { + "epoch": 0.3, + "learning_rate": 1.7065587578276632e-07, + "loss": 0.6952, + "step": 8063 + }, + { + "epoch": 0.3, + "learning_rate": 1.7065221371809427e-07, + "loss": 0.6767, + "step": 8064 + }, + { + "epoch": 0.3, + "learning_rate": 1.706485516534222e-07, + "loss": 0.698, + "step": 8065 + }, + { + "epoch": 0.3, + "learning_rate": 1.7064488958875015e-07, + "loss": 0.7015, + "step": 8066 + }, + { + "epoch": 0.3, + "learning_rate": 1.7064122752407805e-07, + "loss": 0.6943, + "step": 8067 + }, + { + "epoch": 0.3, + "learning_rate": 1.70637565459406e-07, + "loss": 0.705, + "step": 8068 + }, + { + "epoch": 0.3, + "learning_rate": 1.7063390339473393e-07, + "loss": 0.6927, + "step": 8069 + }, + { + "epoch": 0.3, + "learning_rate": 1.7063024133006189e-07, + "loss": 0.691, + "step": 8070 + }, + { + "epoch": 0.3, + "learning_rate": 1.7062657926538981e-07, + "loss": 0.7017, + "step": 8071 + }, + { + "epoch": 0.3, + "learning_rate": 1.7062291720071774e-07, + "loss": 0.6938, + "step": 8072 + }, + { + "epoch": 0.3, + "learning_rate": 1.706192551360457e-07, + "loss": 0.7048, + "step": 8073 + }, + { + "epoch": 0.3, + "learning_rate": 1.7061559307137362e-07, + "loss": 0.6856, + "step": 8074 + }, + { + "epoch": 0.3, + "learning_rate": 1.7061193100670157e-07, + "loss": 0.6928, + "step": 8075 + }, + { + "epoch": 0.3, + "learning_rate": 1.706082689420295e-07, + "loss": 0.7005, + "step": 8076 + }, + { + "epoch": 0.3, + "learning_rate": 1.7060460687735745e-07, + "loss": 0.6782, + "step": 8077 + }, + { + "epoch": 0.3, + "learning_rate": 1.7060094481268538e-07, + "loss": 0.6934, + "step": 8078 + }, + { + "epoch": 0.3, + "learning_rate": 1.7059728274801333e-07, + "loss": 0.6946, + "step": 8079 + }, + { + "epoch": 0.3, + "learning_rate": 1.7059362068334126e-07, + "loss": 0.6853, + "step": 8080 + }, + { + "epoch": 0.3, + "learning_rate": 1.705899586186692e-07, + "loss": 0.7008, + "step": 8081 + }, + { + "epoch": 0.3, + "learning_rate": 1.7058629655399714e-07, + "loss": 0.6997, + "step": 8082 + }, + { + "epoch": 0.3, + "learning_rate": 1.705826344893251e-07, + "loss": 0.698, + "step": 8083 + }, + { + "epoch": 0.3, + "learning_rate": 1.7057897242465302e-07, + "loss": 0.6983, + "step": 8084 + }, + { + "epoch": 0.3, + "learning_rate": 1.7057531035998095e-07, + "loss": 0.6943, + "step": 8085 + }, + { + "epoch": 0.3, + "learning_rate": 1.7057164829530887e-07, + "loss": 0.68, + "step": 8086 + }, + { + "epoch": 0.3, + "learning_rate": 1.7056798623063683e-07, + "loss": 0.6686, + "step": 8087 + }, + { + "epoch": 0.3, + "learning_rate": 1.7056432416596475e-07, + "loss": 0.6909, + "step": 8088 + }, + { + "epoch": 0.3, + "learning_rate": 1.705606621012927e-07, + "loss": 0.6904, + "step": 8089 + }, + { + "epoch": 0.3, + "learning_rate": 1.7055700003662063e-07, + "loss": 0.7007, + "step": 8090 + }, + { + "epoch": 0.3, + "learning_rate": 1.7055333797194859e-07, + "loss": 0.6858, + "step": 8091 + }, + { + "epoch": 0.3, + "learning_rate": 1.705496759072765e-07, + "loss": 0.6944, + "step": 8092 + }, + { + "epoch": 0.3, + "learning_rate": 1.7054601384260447e-07, + "loss": 0.6867, + "step": 8093 + }, + { + "epoch": 0.3, + "learning_rate": 1.705423517779324e-07, + "loss": 0.6905, + "step": 8094 + }, + { + "epoch": 0.3, + "learning_rate": 1.7053868971326032e-07, + "loss": 0.6931, + "step": 8095 + }, + { + "epoch": 0.3, + "learning_rate": 1.7053502764858827e-07, + "loss": 0.6942, + "step": 8096 + }, + { + "epoch": 0.3, + "learning_rate": 1.705313655839162e-07, + "loss": 0.7001, + "step": 8097 + }, + { + "epoch": 0.3, + "learning_rate": 1.7052770351924415e-07, + "loss": 0.6961, + "step": 8098 + }, + { + "epoch": 0.3, + "learning_rate": 1.7052404145457208e-07, + "loss": 0.6849, + "step": 8099 + }, + { + "epoch": 0.3, + "learning_rate": 1.7052037938990003e-07, + "loss": 0.6924, + "step": 8100 + }, + { + "epoch": 0.3, + "learning_rate": 1.7051671732522796e-07, + "loss": 0.6968, + "step": 8101 + }, + { + "epoch": 0.3, + "learning_rate": 1.705130552605559e-07, + "loss": 0.7015, + "step": 8102 + }, + { + "epoch": 0.3, + "learning_rate": 1.705093931958838e-07, + "loss": 0.6953, + "step": 8103 + }, + { + "epoch": 0.3, + "learning_rate": 1.7050573113121176e-07, + "loss": 0.6939, + "step": 8104 + }, + { + "epoch": 0.3, + "learning_rate": 1.705020690665397e-07, + "loss": 0.6916, + "step": 8105 + }, + { + "epoch": 0.3, + "learning_rate": 1.7049840700186764e-07, + "loss": 0.6915, + "step": 8106 + }, + { + "epoch": 0.3, + "learning_rate": 1.7049474493719557e-07, + "loss": 0.6904, + "step": 8107 + }, + { + "epoch": 0.3, + "learning_rate": 1.7049108287252352e-07, + "loss": 0.6942, + "step": 8108 + }, + { + "epoch": 0.3, + "learning_rate": 1.7048742080785145e-07, + "loss": 0.69, + "step": 8109 + }, + { + "epoch": 0.3, + "learning_rate": 1.704837587431794e-07, + "loss": 0.6857, + "step": 8110 + }, + { + "epoch": 0.3, + "learning_rate": 1.7048009667850733e-07, + "loss": 0.6941, + "step": 8111 + }, + { + "epoch": 0.3, + "learning_rate": 1.7047643461383528e-07, + "loss": 0.6786, + "step": 8112 + }, + { + "epoch": 0.3, + "learning_rate": 1.704727725491632e-07, + "loss": 0.6848, + "step": 8113 + }, + { + "epoch": 0.3, + "learning_rate": 1.7046911048449116e-07, + "loss": 0.6853, + "step": 8114 + }, + { + "epoch": 0.3, + "learning_rate": 1.704654484198191e-07, + "loss": 0.6954, + "step": 8115 + }, + { + "epoch": 0.3, + "learning_rate": 1.7046178635514702e-07, + "loss": 0.6826, + "step": 8116 + }, + { + "epoch": 0.3, + "learning_rate": 1.7045812429047497e-07, + "loss": 0.6902, + "step": 8117 + }, + { + "epoch": 0.3, + "learning_rate": 1.704544622258029e-07, + "loss": 0.7118, + "step": 8118 + }, + { + "epoch": 0.3, + "learning_rate": 1.7045080016113085e-07, + "loss": 0.697, + "step": 8119 + }, + { + "epoch": 0.3, + "learning_rate": 1.7044713809645878e-07, + "loss": 0.6882, + "step": 8120 + }, + { + "epoch": 0.3, + "learning_rate": 1.704434760317867e-07, + "loss": 0.6922, + "step": 8121 + }, + { + "epoch": 0.3, + "learning_rate": 1.7043981396711463e-07, + "loss": 0.6856, + "step": 8122 + }, + { + "epoch": 0.3, + "learning_rate": 1.7043615190244258e-07, + "loss": 0.7048, + "step": 8123 + }, + { + "epoch": 0.3, + "learning_rate": 1.704324898377705e-07, + "loss": 0.6906, + "step": 8124 + }, + { + "epoch": 0.3, + "learning_rate": 1.7042882777309846e-07, + "loss": 0.694, + "step": 8125 + }, + { + "epoch": 0.3, + "learning_rate": 1.704251657084264e-07, + "loss": 0.7042, + "step": 8126 + }, + { + "epoch": 0.3, + "learning_rate": 1.7042150364375434e-07, + "loss": 0.6857, + "step": 8127 + }, + { + "epoch": 0.3, + "learning_rate": 1.7041784157908227e-07, + "loss": 0.6902, + "step": 8128 + }, + { + "epoch": 0.3, + "learning_rate": 1.7041417951441022e-07, + "loss": 0.6969, + "step": 8129 + }, + { + "epoch": 0.3, + "learning_rate": 1.7041051744973815e-07, + "loss": 0.6855, + "step": 8130 + }, + { + "epoch": 0.3, + "learning_rate": 1.704068553850661e-07, + "loss": 0.6931, + "step": 8131 + }, + { + "epoch": 0.3, + "learning_rate": 1.7040319332039403e-07, + "loss": 0.6947, + "step": 8132 + }, + { + "epoch": 0.3, + "learning_rate": 1.7039953125572198e-07, + "loss": 0.6839, + "step": 8133 + }, + { + "epoch": 0.3, + "learning_rate": 1.703958691910499e-07, + "loss": 0.7032, + "step": 8134 + }, + { + "epoch": 0.3, + "learning_rate": 1.7039220712637786e-07, + "loss": 0.6781, + "step": 8135 + }, + { + "epoch": 0.3, + "learning_rate": 1.703885450617058e-07, + "loss": 0.6926, + "step": 8136 + }, + { + "epoch": 0.3, + "learning_rate": 1.7038488299703374e-07, + "loss": 0.6909, + "step": 8137 + }, + { + "epoch": 0.3, + "learning_rate": 1.7038122093236164e-07, + "loss": 0.687, + "step": 8138 + }, + { + "epoch": 0.3, + "learning_rate": 1.703775588676896e-07, + "loss": 0.694, + "step": 8139 + }, + { + "epoch": 0.3, + "learning_rate": 1.7037389680301752e-07, + "loss": 0.6823, + "step": 8140 + }, + { + "epoch": 0.3, + "learning_rate": 1.7037023473834548e-07, + "loss": 0.6936, + "step": 8141 + }, + { + "epoch": 0.3, + "learning_rate": 1.703665726736734e-07, + "loss": 0.6998, + "step": 8142 + }, + { + "epoch": 0.3, + "learning_rate": 1.7036291060900133e-07, + "loss": 0.6802, + "step": 8143 + }, + { + "epoch": 0.3, + "learning_rate": 1.7035924854432928e-07, + "loss": 0.671, + "step": 8144 + }, + { + "epoch": 0.3, + "learning_rate": 1.703555864796572e-07, + "loss": 0.6978, + "step": 8145 + }, + { + "epoch": 0.3, + "learning_rate": 1.7035192441498516e-07, + "loss": 0.6851, + "step": 8146 + }, + { + "epoch": 0.3, + "learning_rate": 1.703482623503131e-07, + "loss": 0.6849, + "step": 8147 + }, + { + "epoch": 0.3, + "learning_rate": 1.7034460028564104e-07, + "loss": 0.7265, + "step": 8148 + }, + { + "epoch": 0.3, + "learning_rate": 1.7034093822096897e-07, + "loss": 0.6817, + "step": 8149 + }, + { + "epoch": 0.3, + "learning_rate": 1.7033727615629692e-07, + "loss": 0.6912, + "step": 8150 + }, + { + "epoch": 0.3, + "learning_rate": 1.7033361409162485e-07, + "loss": 0.7028, + "step": 8151 + }, + { + "epoch": 0.3, + "learning_rate": 1.703299520269528e-07, + "loss": 0.6916, + "step": 8152 + }, + { + "epoch": 0.3, + "learning_rate": 1.7032628996228073e-07, + "loss": 0.6831, + "step": 8153 + }, + { + "epoch": 0.3, + "learning_rate": 1.7032262789760868e-07, + "loss": 0.696, + "step": 8154 + }, + { + "epoch": 0.3, + "learning_rate": 1.703189658329366e-07, + "loss": 0.6864, + "step": 8155 + }, + { + "epoch": 0.3, + "learning_rate": 1.7031530376826454e-07, + "loss": 0.6965, + "step": 8156 + }, + { + "epoch": 0.3, + "learning_rate": 1.7031164170359246e-07, + "loss": 0.6816, + "step": 8157 + }, + { + "epoch": 0.3, + "learning_rate": 1.7030797963892042e-07, + "loss": 0.6967, + "step": 8158 + }, + { + "epoch": 0.3, + "learning_rate": 1.7030431757424834e-07, + "loss": 0.6914, + "step": 8159 + }, + { + "epoch": 0.3, + "learning_rate": 1.703006555095763e-07, + "loss": 0.6976, + "step": 8160 + }, + { + "epoch": 0.3, + "learning_rate": 1.7029699344490422e-07, + "loss": 0.6993, + "step": 8161 + }, + { + "epoch": 0.3, + "learning_rate": 1.7029333138023217e-07, + "loss": 0.694, + "step": 8162 + }, + { + "epoch": 0.3, + "learning_rate": 1.702896693155601e-07, + "loss": 0.695, + "step": 8163 + }, + { + "epoch": 0.3, + "learning_rate": 1.7028600725088805e-07, + "loss": 0.7101, + "step": 8164 + }, + { + "epoch": 0.3, + "learning_rate": 1.7028234518621598e-07, + "loss": 0.6942, + "step": 8165 + }, + { + "epoch": 0.3, + "learning_rate": 1.702786831215439e-07, + "loss": 0.6933, + "step": 8166 + }, + { + "epoch": 0.3, + "learning_rate": 1.7027502105687186e-07, + "loss": 0.6975, + "step": 8167 + }, + { + "epoch": 0.3, + "learning_rate": 1.702713589921998e-07, + "loss": 0.6861, + "step": 8168 + }, + { + "epoch": 0.3, + "learning_rate": 1.7026769692752774e-07, + "loss": 0.7007, + "step": 8169 + }, + { + "epoch": 0.3, + "learning_rate": 1.7026403486285567e-07, + "loss": 0.6888, + "step": 8170 + }, + { + "epoch": 0.3, + "learning_rate": 1.7026037279818362e-07, + "loss": 0.6816, + "step": 8171 + }, + { + "epoch": 0.3, + "learning_rate": 1.7025671073351155e-07, + "loss": 0.689, + "step": 8172 + }, + { + "epoch": 0.3, + "learning_rate": 1.702530486688395e-07, + "loss": 0.6967, + "step": 8173 + }, + { + "epoch": 0.3, + "learning_rate": 1.702493866041674e-07, + "loss": 0.7045, + "step": 8174 + }, + { + "epoch": 0.3, + "learning_rate": 1.7024572453949535e-07, + "loss": 0.6827, + "step": 8175 + }, + { + "epoch": 0.3, + "learning_rate": 1.7024206247482328e-07, + "loss": 0.6875, + "step": 8176 + }, + { + "epoch": 0.3, + "learning_rate": 1.7023840041015123e-07, + "loss": 0.6846, + "step": 8177 + }, + { + "epoch": 0.3, + "learning_rate": 1.7023473834547916e-07, + "loss": 0.6892, + "step": 8178 + }, + { + "epoch": 0.3, + "learning_rate": 1.7023107628080711e-07, + "loss": 0.6999, + "step": 8179 + }, + { + "epoch": 0.3, + "learning_rate": 1.7022741421613504e-07, + "loss": 0.6914, + "step": 8180 + }, + { + "epoch": 0.3, + "learning_rate": 1.70223752151463e-07, + "loss": 0.6971, + "step": 8181 + }, + { + "epoch": 0.3, + "learning_rate": 1.7022009008679092e-07, + "loss": 0.7169, + "step": 8182 + }, + { + "epoch": 0.3, + "learning_rate": 1.7021642802211887e-07, + "loss": 0.6952, + "step": 8183 + }, + { + "epoch": 0.3, + "learning_rate": 1.702127659574468e-07, + "loss": 0.704, + "step": 8184 + }, + { + "epoch": 0.3, + "learning_rate": 1.7020910389277475e-07, + "loss": 0.6879, + "step": 8185 + }, + { + "epoch": 0.3, + "learning_rate": 1.7020544182810268e-07, + "loss": 0.6985, + "step": 8186 + }, + { + "epoch": 0.3, + "learning_rate": 1.702017797634306e-07, + "loss": 0.6946, + "step": 8187 + }, + { + "epoch": 0.3, + "learning_rate": 1.7019811769875856e-07, + "loss": 0.6984, + "step": 8188 + }, + { + "epoch": 0.3, + "learning_rate": 1.7019445563408649e-07, + "loss": 0.7058, + "step": 8189 + }, + { + "epoch": 0.3, + "learning_rate": 1.7019079356941444e-07, + "loss": 0.6959, + "step": 8190 + }, + { + "epoch": 0.3, + "learning_rate": 1.7018713150474237e-07, + "loss": 0.7074, + "step": 8191 + }, + { + "epoch": 0.3, + "learning_rate": 1.701834694400703e-07, + "loss": 0.6914, + "step": 8192 + }, + { + "epoch": 0.3, + "learning_rate": 1.7017980737539822e-07, + "loss": 0.7006, + "step": 8193 + }, + { + "epoch": 0.3, + "learning_rate": 1.7017614531072617e-07, + "loss": 0.6932, + "step": 8194 + }, + { + "epoch": 0.3, + "learning_rate": 1.701724832460541e-07, + "loss": 0.6778, + "step": 8195 + }, + { + "epoch": 0.3, + "learning_rate": 1.7016882118138205e-07, + "loss": 0.6906, + "step": 8196 + }, + { + "epoch": 0.3, + "learning_rate": 1.7016515911670998e-07, + "loss": 0.6933, + "step": 8197 + }, + { + "epoch": 0.3, + "learning_rate": 1.7016149705203793e-07, + "loss": 0.6763, + "step": 8198 + }, + { + "epoch": 0.3, + "learning_rate": 1.7015783498736586e-07, + "loss": 0.7025, + "step": 8199 + }, + { + "epoch": 0.3, + "learning_rate": 1.701541729226938e-07, + "loss": 0.686, + "step": 8200 + }, + { + "epoch": 0.3, + "learning_rate": 1.7015051085802174e-07, + "loss": 0.6882, + "step": 8201 + }, + { + "epoch": 0.3, + "learning_rate": 1.701468487933497e-07, + "loss": 0.7037, + "step": 8202 + }, + { + "epoch": 0.3, + "learning_rate": 1.7014318672867762e-07, + "loss": 0.7007, + "step": 8203 + }, + { + "epoch": 0.3, + "learning_rate": 1.7013952466400557e-07, + "loss": 0.7069, + "step": 8204 + }, + { + "epoch": 0.3, + "learning_rate": 1.701358625993335e-07, + "loss": 0.7129, + "step": 8205 + }, + { + "epoch": 0.3, + "learning_rate": 1.7013220053466145e-07, + "loss": 0.6953, + "step": 8206 + }, + { + "epoch": 0.3, + "learning_rate": 1.7012853846998938e-07, + "loss": 0.6991, + "step": 8207 + }, + { + "epoch": 0.3, + "learning_rate": 1.7012487640531733e-07, + "loss": 0.6977, + "step": 8208 + }, + { + "epoch": 0.3, + "learning_rate": 1.7012121434064526e-07, + "loss": 0.6883, + "step": 8209 + }, + { + "epoch": 0.3, + "learning_rate": 1.7011755227597319e-07, + "loss": 0.6975, + "step": 8210 + }, + { + "epoch": 0.3, + "learning_rate": 1.701138902113011e-07, + "loss": 0.6989, + "step": 8211 + }, + { + "epoch": 0.3, + "learning_rate": 1.7011022814662907e-07, + "loss": 0.7095, + "step": 8212 + }, + { + "epoch": 0.3, + "learning_rate": 1.70106566081957e-07, + "loss": 0.694, + "step": 8213 + }, + { + "epoch": 0.3, + "learning_rate": 1.7010290401728492e-07, + "loss": 0.6994, + "step": 8214 + }, + { + "epoch": 0.3, + "learning_rate": 1.7009924195261287e-07, + "loss": 0.6964, + "step": 8215 + }, + { + "epoch": 0.3, + "learning_rate": 1.700955798879408e-07, + "loss": 0.7054, + "step": 8216 + }, + { + "epoch": 0.3, + "learning_rate": 1.7009191782326875e-07, + "loss": 0.7004, + "step": 8217 + }, + { + "epoch": 0.3, + "learning_rate": 1.7008825575859668e-07, + "loss": 0.6928, + "step": 8218 + }, + { + "epoch": 0.3, + "learning_rate": 1.7008459369392463e-07, + "loss": 0.6949, + "step": 8219 + }, + { + "epoch": 0.3, + "learning_rate": 1.7008093162925256e-07, + "loss": 0.6916, + "step": 8220 + }, + { + "epoch": 0.3, + "learning_rate": 1.700772695645805e-07, + "loss": 0.6955, + "step": 8221 + }, + { + "epoch": 0.3, + "learning_rate": 1.7007360749990844e-07, + "loss": 0.7043, + "step": 8222 + }, + { + "epoch": 0.3, + "learning_rate": 1.700699454352364e-07, + "loss": 0.6927, + "step": 8223 + }, + { + "epoch": 0.3, + "learning_rate": 1.7006628337056432e-07, + "loss": 0.6957, + "step": 8224 + }, + { + "epoch": 0.3, + "learning_rate": 1.7006262130589227e-07, + "loss": 0.6977, + "step": 8225 + }, + { + "epoch": 0.3, + "learning_rate": 1.700589592412202e-07, + "loss": 0.6842, + "step": 8226 + }, + { + "epoch": 0.3, + "learning_rate": 1.7005529717654815e-07, + "loss": 0.6916, + "step": 8227 + }, + { + "epoch": 0.3, + "learning_rate": 1.7005163511187605e-07, + "loss": 0.7077, + "step": 8228 + }, + { + "epoch": 0.3, + "learning_rate": 1.70047973047204e-07, + "loss": 0.7051, + "step": 8229 + }, + { + "epoch": 0.3, + "learning_rate": 1.7004431098253193e-07, + "loss": 0.7072, + "step": 8230 + }, + { + "epoch": 0.3, + "learning_rate": 1.7004064891785988e-07, + "loss": 0.6905, + "step": 8231 + }, + { + "epoch": 0.3, + "learning_rate": 1.700369868531878e-07, + "loss": 0.6853, + "step": 8232 + }, + { + "epoch": 0.3, + "learning_rate": 1.7003332478851576e-07, + "loss": 0.6931, + "step": 8233 + }, + { + "epoch": 0.3, + "learning_rate": 1.700296627238437e-07, + "loss": 0.6787, + "step": 8234 + }, + { + "epoch": 0.3, + "learning_rate": 1.7002600065917164e-07, + "loss": 0.6978, + "step": 8235 + }, + { + "epoch": 0.3, + "learning_rate": 1.7002233859449957e-07, + "loss": 0.6924, + "step": 8236 + }, + { + "epoch": 0.3, + "learning_rate": 1.700186765298275e-07, + "loss": 0.7195, + "step": 8237 + }, + { + "epoch": 0.3, + "learning_rate": 1.7001501446515545e-07, + "loss": 0.6995, + "step": 8238 + }, + { + "epoch": 0.3, + "learning_rate": 1.7001135240048338e-07, + "loss": 0.6867, + "step": 8239 + }, + { + "epoch": 0.3, + "learning_rate": 1.7000769033581133e-07, + "loss": 0.7042, + "step": 8240 + }, + { + "epoch": 0.3, + "learning_rate": 1.7000402827113926e-07, + "loss": 0.6984, + "step": 8241 + }, + { + "epoch": 0.3, + "learning_rate": 1.700003662064672e-07, + "loss": 0.6998, + "step": 8242 + }, + { + "epoch": 0.3, + "learning_rate": 1.6999670414179514e-07, + "loss": 0.676, + "step": 8243 + }, + { + "epoch": 0.3, + "learning_rate": 1.699930420771231e-07, + "loss": 0.7154, + "step": 8244 + }, + { + "epoch": 0.3, + "learning_rate": 1.6998938001245102e-07, + "loss": 0.6991, + "step": 8245 + }, + { + "epoch": 0.3, + "learning_rate": 1.6998571794777894e-07, + "loss": 0.6917, + "step": 8246 + }, + { + "epoch": 0.3, + "learning_rate": 1.6998205588310687e-07, + "loss": 0.6982, + "step": 8247 + }, + { + "epoch": 0.3, + "learning_rate": 1.6997839381843482e-07, + "loss": 0.696, + "step": 8248 + }, + { + "epoch": 0.3, + "learning_rate": 1.6997473175376275e-07, + "loss": 0.699, + "step": 8249 + }, + { + "epoch": 0.3, + "learning_rate": 1.699710696890907e-07, + "loss": 0.7012, + "step": 8250 + }, + { + "epoch": 0.3, + "learning_rate": 1.6996740762441863e-07, + "loss": 0.6924, + "step": 8251 + }, + { + "epoch": 0.3, + "learning_rate": 1.6996374555974658e-07, + "loss": 0.6936, + "step": 8252 + }, + { + "epoch": 0.3, + "learning_rate": 1.699600834950745e-07, + "loss": 0.6982, + "step": 8253 + }, + { + "epoch": 0.3, + "learning_rate": 1.6995642143040246e-07, + "loss": 0.6985, + "step": 8254 + }, + { + "epoch": 0.3, + "learning_rate": 1.699527593657304e-07, + "loss": 0.6995, + "step": 8255 + }, + { + "epoch": 0.3, + "learning_rate": 1.6994909730105834e-07, + "loss": 0.705, + "step": 8256 + }, + { + "epoch": 0.3, + "learning_rate": 1.6994543523638627e-07, + "loss": 0.6755, + "step": 8257 + }, + { + "epoch": 0.3, + "learning_rate": 1.699417731717142e-07, + "loss": 0.699, + "step": 8258 + }, + { + "epoch": 0.3, + "learning_rate": 1.6993811110704215e-07, + "loss": 0.7046, + "step": 8259 + }, + { + "epoch": 0.3, + "learning_rate": 1.6993444904237008e-07, + "loss": 0.7043, + "step": 8260 + }, + { + "epoch": 0.3, + "learning_rate": 1.6993078697769803e-07, + "loss": 0.6998, + "step": 8261 + }, + { + "epoch": 0.3, + "learning_rate": 1.6992712491302596e-07, + "loss": 0.6829, + "step": 8262 + }, + { + "epoch": 0.3, + "learning_rate": 1.699234628483539e-07, + "loss": 0.6922, + "step": 8263 + }, + { + "epoch": 0.3, + "learning_rate": 1.699198007836818e-07, + "loss": 0.6916, + "step": 8264 + }, + { + "epoch": 0.3, + "learning_rate": 1.6991613871900976e-07, + "loss": 0.684, + "step": 8265 + }, + { + "epoch": 0.3, + "learning_rate": 1.699124766543377e-07, + "loss": 0.7059, + "step": 8266 + }, + { + "epoch": 0.3, + "learning_rate": 1.6990881458966564e-07, + "loss": 0.6885, + "step": 8267 + }, + { + "epoch": 0.3, + "learning_rate": 1.6990515252499357e-07, + "loss": 0.6927, + "step": 8268 + }, + { + "epoch": 0.3, + "learning_rate": 1.6990149046032152e-07, + "loss": 0.7013, + "step": 8269 + }, + { + "epoch": 0.3, + "learning_rate": 1.6989782839564945e-07, + "loss": 0.6943, + "step": 8270 + }, + { + "epoch": 0.3, + "learning_rate": 1.698941663309774e-07, + "loss": 0.6943, + "step": 8271 + }, + { + "epoch": 0.3, + "learning_rate": 1.6989050426630533e-07, + "loss": 0.7028, + "step": 8272 + }, + { + "epoch": 0.3, + "learning_rate": 1.6988684220163328e-07, + "loss": 0.6749, + "step": 8273 + }, + { + "epoch": 0.3, + "learning_rate": 1.698831801369612e-07, + "loss": 0.6782, + "step": 8274 + }, + { + "epoch": 0.3, + "learning_rate": 1.6987951807228916e-07, + "loss": 0.7116, + "step": 8275 + }, + { + "epoch": 0.3, + "learning_rate": 1.698758560076171e-07, + "loss": 0.6948, + "step": 8276 + }, + { + "epoch": 0.3, + "learning_rate": 1.6987219394294504e-07, + "loss": 0.6885, + "step": 8277 + }, + { + "epoch": 0.3, + "learning_rate": 1.6986853187827297e-07, + "loss": 0.688, + "step": 8278 + }, + { + "epoch": 0.3, + "learning_rate": 1.6986486981360092e-07, + "loss": 0.6959, + "step": 8279 + }, + { + "epoch": 0.3, + "learning_rate": 1.6986120774892885e-07, + "loss": 0.6851, + "step": 8280 + }, + { + "epoch": 0.3, + "learning_rate": 1.6985754568425677e-07, + "loss": 0.6883, + "step": 8281 + }, + { + "epoch": 0.3, + "learning_rate": 1.698538836195847e-07, + "loss": 0.7126, + "step": 8282 + }, + { + "epoch": 0.3, + "learning_rate": 1.6985022155491265e-07, + "loss": 0.7018, + "step": 8283 + }, + { + "epoch": 0.3, + "learning_rate": 1.6984655949024058e-07, + "loss": 0.6921, + "step": 8284 + }, + { + "epoch": 0.3, + "learning_rate": 1.698428974255685e-07, + "loss": 0.6876, + "step": 8285 + }, + { + "epoch": 0.3, + "learning_rate": 1.6983923536089646e-07, + "loss": 0.696, + "step": 8286 + }, + { + "epoch": 0.3, + "learning_rate": 1.698355732962244e-07, + "loss": 0.6919, + "step": 8287 + }, + { + "epoch": 0.3, + "learning_rate": 1.6983191123155234e-07, + "loss": 0.6944, + "step": 8288 + }, + { + "epoch": 0.3, + "learning_rate": 1.6982824916688027e-07, + "loss": 0.6849, + "step": 8289 + }, + { + "epoch": 0.3, + "learning_rate": 1.6982458710220822e-07, + "loss": 0.6936, + "step": 8290 + }, + { + "epoch": 0.3, + "learning_rate": 1.6982092503753615e-07, + "loss": 0.7039, + "step": 8291 + }, + { + "epoch": 0.3, + "learning_rate": 1.698172629728641e-07, + "loss": 0.6768, + "step": 8292 + }, + { + "epoch": 0.3, + "learning_rate": 1.6981360090819203e-07, + "loss": 0.6834, + "step": 8293 + }, + { + "epoch": 0.3, + "learning_rate": 1.6980993884351998e-07, + "loss": 0.7104, + "step": 8294 + }, + { + "epoch": 0.3, + "learning_rate": 1.698062767788479e-07, + "loss": 0.6837, + "step": 8295 + }, + { + "epoch": 0.3, + "learning_rate": 1.6980261471417586e-07, + "loss": 0.6979, + "step": 8296 + }, + { + "epoch": 0.3, + "learning_rate": 1.697989526495038e-07, + "loss": 0.6918, + "step": 8297 + }, + { + "epoch": 0.3, + "learning_rate": 1.6979529058483174e-07, + "loss": 0.6804, + "step": 8298 + }, + { + "epoch": 0.3, + "learning_rate": 1.6979162852015967e-07, + "loss": 0.6872, + "step": 8299 + }, + { + "epoch": 0.3, + "learning_rate": 1.697879664554876e-07, + "loss": 0.6937, + "step": 8300 + }, + { + "epoch": 0.3, + "learning_rate": 1.6978430439081552e-07, + "loss": 0.6967, + "step": 8301 + }, + { + "epoch": 0.3, + "learning_rate": 1.6978064232614347e-07, + "loss": 0.6838, + "step": 8302 + }, + { + "epoch": 0.3, + "learning_rate": 1.697769802614714e-07, + "loss": 0.6897, + "step": 8303 + }, + { + "epoch": 0.3, + "learning_rate": 1.6977331819679935e-07, + "loss": 0.6885, + "step": 8304 + }, + { + "epoch": 0.3, + "learning_rate": 1.6976965613212728e-07, + "loss": 0.6973, + "step": 8305 + }, + { + "epoch": 0.3, + "learning_rate": 1.6976599406745523e-07, + "loss": 0.6871, + "step": 8306 + }, + { + "epoch": 0.3, + "learning_rate": 1.6976233200278316e-07, + "loss": 0.6972, + "step": 8307 + }, + { + "epoch": 0.3, + "learning_rate": 1.6975866993811109e-07, + "loss": 0.6976, + "step": 8308 + }, + { + "epoch": 0.3, + "learning_rate": 1.6975500787343904e-07, + "loss": 0.6915, + "step": 8309 + }, + { + "epoch": 0.3, + "learning_rate": 1.6975134580876697e-07, + "loss": 0.6929, + "step": 8310 + }, + { + "epoch": 0.3, + "learning_rate": 1.6974768374409492e-07, + "loss": 0.697, + "step": 8311 + }, + { + "epoch": 0.3, + "learning_rate": 1.6974402167942285e-07, + "loss": 0.7125, + "step": 8312 + }, + { + "epoch": 0.3, + "learning_rate": 1.697403596147508e-07, + "loss": 0.6824, + "step": 8313 + }, + { + "epoch": 0.3, + "learning_rate": 1.6973669755007873e-07, + "loss": 0.692, + "step": 8314 + }, + { + "epoch": 0.3, + "learning_rate": 1.6973303548540668e-07, + "loss": 0.6963, + "step": 8315 + }, + { + "epoch": 0.3, + "learning_rate": 1.697293734207346e-07, + "loss": 0.689, + "step": 8316 + }, + { + "epoch": 0.3, + "learning_rate": 1.6972571135606256e-07, + "loss": 0.7013, + "step": 8317 + }, + { + "epoch": 0.3, + "learning_rate": 1.6972204929139046e-07, + "loss": 0.6932, + "step": 8318 + }, + { + "epoch": 0.3, + "learning_rate": 1.697183872267184e-07, + "loss": 0.705, + "step": 8319 + }, + { + "epoch": 0.3, + "learning_rate": 1.6971472516204634e-07, + "loss": 0.684, + "step": 8320 + }, + { + "epoch": 0.3, + "learning_rate": 1.697110630973743e-07, + "loss": 0.6991, + "step": 8321 + }, + { + "epoch": 0.3, + "learning_rate": 1.6970740103270222e-07, + "loss": 0.7099, + "step": 8322 + }, + { + "epoch": 0.3, + "learning_rate": 1.6970373896803017e-07, + "loss": 0.6824, + "step": 8323 + }, + { + "epoch": 0.3, + "learning_rate": 1.697000769033581e-07, + "loss": 0.7061, + "step": 8324 + }, + { + "epoch": 0.3, + "learning_rate": 1.6969641483868605e-07, + "loss": 0.7017, + "step": 8325 + }, + { + "epoch": 0.3, + "learning_rate": 1.6969275277401398e-07, + "loss": 0.6909, + "step": 8326 + }, + { + "epoch": 0.3, + "learning_rate": 1.6968909070934193e-07, + "loss": 0.6996, + "step": 8327 + }, + { + "epoch": 0.3, + "learning_rate": 1.6968542864466986e-07, + "loss": 0.6885, + "step": 8328 + }, + { + "epoch": 0.3, + "learning_rate": 1.6968176657999779e-07, + "loss": 0.6982, + "step": 8329 + }, + { + "epoch": 0.3, + "learning_rate": 1.6967810451532574e-07, + "loss": 0.7072, + "step": 8330 + }, + { + "epoch": 0.3, + "learning_rate": 1.6967444245065367e-07, + "loss": 0.6849, + "step": 8331 + }, + { + "epoch": 0.3, + "learning_rate": 1.6967078038598162e-07, + "loss": 0.6948, + "step": 8332 + }, + { + "epoch": 0.3, + "learning_rate": 1.6966711832130955e-07, + "loss": 0.6837, + "step": 8333 + }, + { + "epoch": 0.3, + "learning_rate": 1.696634562566375e-07, + "loss": 0.6877, + "step": 8334 + }, + { + "epoch": 0.3, + "learning_rate": 1.6965979419196543e-07, + "loss": 0.6925, + "step": 8335 + }, + { + "epoch": 0.3, + "learning_rate": 1.6965613212729335e-07, + "loss": 0.6842, + "step": 8336 + }, + { + "epoch": 0.31, + "learning_rate": 1.6965247006262128e-07, + "loss": 0.698, + "step": 8337 + }, + { + "epoch": 0.31, + "learning_rate": 1.6964880799794923e-07, + "loss": 0.6818, + "step": 8338 + }, + { + "epoch": 0.31, + "learning_rate": 1.6964514593327716e-07, + "loss": 0.6838, + "step": 8339 + }, + { + "epoch": 0.31, + "learning_rate": 1.696414838686051e-07, + "loss": 0.699, + "step": 8340 + }, + { + "epoch": 0.31, + "learning_rate": 1.6963782180393304e-07, + "loss": 0.7044, + "step": 8341 + }, + { + "epoch": 0.31, + "learning_rate": 1.69634159739261e-07, + "loss": 0.6833, + "step": 8342 + }, + { + "epoch": 0.31, + "learning_rate": 1.6963049767458892e-07, + "loss": 0.6931, + "step": 8343 + }, + { + "epoch": 0.31, + "learning_rate": 1.6962683560991687e-07, + "loss": 0.7106, + "step": 8344 + }, + { + "epoch": 0.31, + "learning_rate": 1.696231735452448e-07, + "loss": 0.7002, + "step": 8345 + }, + { + "epoch": 0.31, + "learning_rate": 1.6961951148057275e-07, + "loss": 0.6945, + "step": 8346 + }, + { + "epoch": 0.31, + "learning_rate": 1.6961584941590068e-07, + "loss": 0.7052, + "step": 8347 + }, + { + "epoch": 0.31, + "learning_rate": 1.6961218735122863e-07, + "loss": 0.6881, + "step": 8348 + }, + { + "epoch": 0.31, + "learning_rate": 1.6960852528655656e-07, + "loss": 0.693, + "step": 8349 + }, + { + "epoch": 0.31, + "learning_rate": 1.696048632218845e-07, + "loss": 0.6904, + "step": 8350 + }, + { + "epoch": 0.31, + "learning_rate": 1.6960120115721244e-07, + "loss": 0.6796, + "step": 8351 + }, + { + "epoch": 0.31, + "learning_rate": 1.6959753909254036e-07, + "loss": 0.6904, + "step": 8352 + }, + { + "epoch": 0.31, + "learning_rate": 1.6959387702786832e-07, + "loss": 0.6875, + "step": 8353 + }, + { + "epoch": 0.31, + "learning_rate": 1.6959021496319624e-07, + "loss": 0.7057, + "step": 8354 + }, + { + "epoch": 0.31, + "learning_rate": 1.6958655289852417e-07, + "loss": 0.6993, + "step": 8355 + }, + { + "epoch": 0.31, + "learning_rate": 1.695828908338521e-07, + "loss": 0.6917, + "step": 8356 + }, + { + "epoch": 0.31, + "learning_rate": 1.6957922876918005e-07, + "loss": 0.7057, + "step": 8357 + }, + { + "epoch": 0.31, + "learning_rate": 1.6957556670450798e-07, + "loss": 0.708, + "step": 8358 + }, + { + "epoch": 0.31, + "learning_rate": 1.6957190463983593e-07, + "loss": 0.6928, + "step": 8359 + }, + { + "epoch": 0.31, + "learning_rate": 1.6956824257516386e-07, + "loss": 0.6948, + "step": 8360 + }, + { + "epoch": 0.31, + "learning_rate": 1.695645805104918e-07, + "loss": 0.688, + "step": 8361 + }, + { + "epoch": 0.31, + "learning_rate": 1.6956091844581974e-07, + "loss": 0.6958, + "step": 8362 + }, + { + "epoch": 0.31, + "learning_rate": 1.695572563811477e-07, + "loss": 0.685, + "step": 8363 + }, + { + "epoch": 0.31, + "learning_rate": 1.6955359431647562e-07, + "loss": 0.6921, + "step": 8364 + }, + { + "epoch": 0.31, + "learning_rate": 1.6954993225180357e-07, + "loss": 0.6991, + "step": 8365 + }, + { + "epoch": 0.31, + "learning_rate": 1.695462701871315e-07, + "loss": 0.6908, + "step": 8366 + }, + { + "epoch": 0.31, + "learning_rate": 1.6954260812245945e-07, + "loss": 0.6911, + "step": 8367 + }, + { + "epoch": 0.31, + "learning_rate": 1.6953894605778738e-07, + "loss": 0.7012, + "step": 8368 + }, + { + "epoch": 0.31, + "learning_rate": 1.6953528399311533e-07, + "loss": 0.6814, + "step": 8369 + }, + { + "epoch": 0.31, + "learning_rate": 1.6953162192844326e-07, + "loss": 0.7053, + "step": 8370 + }, + { + "epoch": 0.31, + "learning_rate": 1.695279598637712e-07, + "loss": 0.6853, + "step": 8371 + }, + { + "epoch": 0.31, + "learning_rate": 1.695242977990991e-07, + "loss": 0.6918, + "step": 8372 + }, + { + "epoch": 0.31, + "learning_rate": 1.6952063573442706e-07, + "loss": 0.69, + "step": 8373 + }, + { + "epoch": 0.31, + "learning_rate": 1.69516973669755e-07, + "loss": 0.6963, + "step": 8374 + }, + { + "epoch": 0.31, + "learning_rate": 1.6951331160508294e-07, + "loss": 0.6997, + "step": 8375 + }, + { + "epoch": 0.31, + "learning_rate": 1.6950964954041087e-07, + "loss": 0.6832, + "step": 8376 + }, + { + "epoch": 0.31, + "learning_rate": 1.6950598747573882e-07, + "loss": 0.6695, + "step": 8377 + }, + { + "epoch": 0.31, + "learning_rate": 1.6950232541106675e-07, + "loss": 0.6907, + "step": 8378 + }, + { + "epoch": 0.31, + "learning_rate": 1.6949866334639468e-07, + "loss": 0.6918, + "step": 8379 + }, + { + "epoch": 0.31, + "learning_rate": 1.6949500128172263e-07, + "loss": 0.6824, + "step": 8380 + }, + { + "epoch": 0.31, + "learning_rate": 1.6949133921705056e-07, + "loss": 0.6868, + "step": 8381 + }, + { + "epoch": 0.31, + "learning_rate": 1.694876771523785e-07, + "loss": 0.6895, + "step": 8382 + }, + { + "epoch": 0.31, + "learning_rate": 1.6948401508770644e-07, + "loss": 0.6941, + "step": 8383 + }, + { + "epoch": 0.31, + "learning_rate": 1.694803530230344e-07, + "loss": 0.6831, + "step": 8384 + }, + { + "epoch": 0.31, + "learning_rate": 1.6947669095836232e-07, + "loss": 0.6851, + "step": 8385 + }, + { + "epoch": 0.31, + "learning_rate": 1.6947302889369027e-07, + "loss": 0.6792, + "step": 8386 + }, + { + "epoch": 0.31, + "learning_rate": 1.694693668290182e-07, + "loss": 0.6905, + "step": 8387 + }, + { + "epoch": 0.31, + "learning_rate": 1.6946570476434615e-07, + "loss": 0.6985, + "step": 8388 + }, + { + "epoch": 0.31, + "learning_rate": 1.6946204269967408e-07, + "loss": 0.6839, + "step": 8389 + }, + { + "epoch": 0.31, + "learning_rate": 1.69458380635002e-07, + "loss": 0.6892, + "step": 8390 + }, + { + "epoch": 0.31, + "learning_rate": 1.6945471857032993e-07, + "loss": 0.6984, + "step": 8391 + }, + { + "epoch": 0.31, + "learning_rate": 1.6945105650565788e-07, + "loss": 0.7012, + "step": 8392 + }, + { + "epoch": 0.31, + "learning_rate": 1.694473944409858e-07, + "loss": 0.7004, + "step": 8393 + }, + { + "epoch": 0.31, + "learning_rate": 1.6944373237631376e-07, + "loss": 0.6885, + "step": 8394 + }, + { + "epoch": 0.31, + "learning_rate": 1.694400703116417e-07, + "loss": 0.6988, + "step": 8395 + }, + { + "epoch": 0.31, + "learning_rate": 1.6943640824696964e-07, + "loss": 0.7001, + "step": 8396 + }, + { + "epoch": 0.31, + "learning_rate": 1.6943274618229757e-07, + "loss": 0.6939, + "step": 8397 + }, + { + "epoch": 0.31, + "learning_rate": 1.6942908411762552e-07, + "loss": 0.7225, + "step": 8398 + }, + { + "epoch": 0.31, + "learning_rate": 1.6942542205295345e-07, + "loss": 0.6885, + "step": 8399 + }, + { + "epoch": 0.31, + "learning_rate": 1.6942175998828137e-07, + "loss": 0.7222, + "step": 8400 + }, + { + "epoch": 0.31, + "learning_rate": 1.6941809792360933e-07, + "loss": 0.6832, + "step": 8401 + }, + { + "epoch": 0.31, + "learning_rate": 1.6941443585893725e-07, + "loss": 0.672, + "step": 8402 + }, + { + "epoch": 0.31, + "learning_rate": 1.694107737942652e-07, + "loss": 0.6835, + "step": 8403 + }, + { + "epoch": 0.31, + "learning_rate": 1.6940711172959313e-07, + "loss": 0.6989, + "step": 8404 + }, + { + "epoch": 0.31, + "learning_rate": 1.694034496649211e-07, + "loss": 0.7034, + "step": 8405 + }, + { + "epoch": 0.31, + "learning_rate": 1.6939978760024901e-07, + "loss": 0.679, + "step": 8406 + }, + { + "epoch": 0.31, + "learning_rate": 1.6939612553557697e-07, + "loss": 0.6824, + "step": 8407 + }, + { + "epoch": 0.31, + "learning_rate": 1.6939246347090487e-07, + "loss": 0.6972, + "step": 8408 + }, + { + "epoch": 0.31, + "learning_rate": 1.6938880140623282e-07, + "loss": 0.6787, + "step": 8409 + }, + { + "epoch": 0.31, + "learning_rate": 1.6938513934156075e-07, + "loss": 0.7129, + "step": 8410 + }, + { + "epoch": 0.31, + "learning_rate": 1.693814772768887e-07, + "loss": 0.6805, + "step": 8411 + }, + { + "epoch": 0.31, + "learning_rate": 1.6937781521221663e-07, + "loss": 0.6964, + "step": 8412 + }, + { + "epoch": 0.31, + "learning_rate": 1.6937415314754458e-07, + "loss": 0.6978, + "step": 8413 + }, + { + "epoch": 0.31, + "learning_rate": 1.693704910828725e-07, + "loss": 0.6859, + "step": 8414 + }, + { + "epoch": 0.31, + "learning_rate": 1.6936682901820046e-07, + "loss": 0.6921, + "step": 8415 + }, + { + "epoch": 0.31, + "learning_rate": 1.693631669535284e-07, + "loss": 0.7005, + "step": 8416 + }, + { + "epoch": 0.31, + "learning_rate": 1.6935950488885634e-07, + "loss": 0.69, + "step": 8417 + }, + { + "epoch": 0.31, + "learning_rate": 1.6935584282418427e-07, + "loss": 0.7014, + "step": 8418 + }, + { + "epoch": 0.31, + "learning_rate": 1.6935218075951222e-07, + "loss": 0.7047, + "step": 8419 + }, + { + "epoch": 0.31, + "learning_rate": 1.6934851869484015e-07, + "loss": 0.7055, + "step": 8420 + }, + { + "epoch": 0.31, + "learning_rate": 1.693448566301681e-07, + "loss": 0.6906, + "step": 8421 + }, + { + "epoch": 0.31, + "learning_rate": 1.6934119456549603e-07, + "loss": 0.6969, + "step": 8422 + }, + { + "epoch": 0.31, + "learning_rate": 1.6933753250082395e-07, + "loss": 0.6911, + "step": 8423 + }, + { + "epoch": 0.31, + "learning_rate": 1.693338704361519e-07, + "loss": 0.6839, + "step": 8424 + }, + { + "epoch": 0.31, + "learning_rate": 1.6933020837147983e-07, + "loss": 0.7124, + "step": 8425 + }, + { + "epoch": 0.31, + "learning_rate": 1.6932654630680776e-07, + "loss": 0.6947, + "step": 8426 + }, + { + "epoch": 0.31, + "learning_rate": 1.6932288424213569e-07, + "loss": 0.6875, + "step": 8427 + }, + { + "epoch": 0.31, + "learning_rate": 1.6931922217746364e-07, + "loss": 0.6946, + "step": 8428 + }, + { + "epoch": 0.31, + "learning_rate": 1.6931556011279157e-07, + "loss": 0.6951, + "step": 8429 + }, + { + "epoch": 0.31, + "learning_rate": 1.6931189804811952e-07, + "loss": 0.689, + "step": 8430 + }, + { + "epoch": 0.31, + "learning_rate": 1.6930823598344745e-07, + "loss": 0.6904, + "step": 8431 + }, + { + "epoch": 0.31, + "learning_rate": 1.693045739187754e-07, + "loss": 0.7027, + "step": 8432 + }, + { + "epoch": 0.31, + "learning_rate": 1.6930091185410333e-07, + "loss": 0.6948, + "step": 8433 + }, + { + "epoch": 0.31, + "learning_rate": 1.6929724978943128e-07, + "loss": 0.692, + "step": 8434 + }, + { + "epoch": 0.31, + "learning_rate": 1.692935877247592e-07, + "loss": 0.6673, + "step": 8435 + }, + { + "epoch": 0.31, + "learning_rate": 1.6928992566008716e-07, + "loss": 0.7003, + "step": 8436 + }, + { + "epoch": 0.31, + "learning_rate": 1.6928626359541509e-07, + "loss": 0.7006, + "step": 8437 + }, + { + "epoch": 0.31, + "learning_rate": 1.6928260153074304e-07, + "loss": 0.6882, + "step": 8438 + }, + { + "epoch": 0.31, + "learning_rate": 1.6927893946607097e-07, + "loss": 0.6965, + "step": 8439 + }, + { + "epoch": 0.31, + "learning_rate": 1.6927527740139892e-07, + "loss": 0.7098, + "step": 8440 + }, + { + "epoch": 0.31, + "learning_rate": 1.6927161533672685e-07, + "loss": 0.7048, + "step": 8441 + }, + { + "epoch": 0.31, + "learning_rate": 1.692679532720548e-07, + "loss": 0.7002, + "step": 8442 + }, + { + "epoch": 0.31, + "learning_rate": 1.6926429120738273e-07, + "loss": 0.6958, + "step": 8443 + }, + { + "epoch": 0.31, + "learning_rate": 1.6926062914271065e-07, + "loss": 0.6914, + "step": 8444 + }, + { + "epoch": 0.31, + "learning_rate": 1.6925696707803858e-07, + "loss": 0.6818, + "step": 8445 + }, + { + "epoch": 0.31, + "learning_rate": 1.6925330501336653e-07, + "loss": 0.7063, + "step": 8446 + }, + { + "epoch": 0.31, + "learning_rate": 1.6924964294869446e-07, + "loss": 0.7086, + "step": 8447 + }, + { + "epoch": 0.31, + "learning_rate": 1.692459808840224e-07, + "loss": 0.6965, + "step": 8448 + }, + { + "epoch": 0.31, + "learning_rate": 1.6924231881935034e-07, + "loss": 0.6881, + "step": 8449 + }, + { + "epoch": 0.31, + "learning_rate": 1.6923865675467827e-07, + "loss": 0.6825, + "step": 8450 + }, + { + "epoch": 0.31, + "learning_rate": 1.6923499469000622e-07, + "loss": 0.6974, + "step": 8451 + }, + { + "epoch": 0.31, + "learning_rate": 1.6923133262533415e-07, + "loss": 0.6899, + "step": 8452 + }, + { + "epoch": 0.31, + "learning_rate": 1.692276705606621e-07, + "loss": 0.7029, + "step": 8453 + }, + { + "epoch": 0.31, + "learning_rate": 1.6922400849599003e-07, + "loss": 0.697, + "step": 8454 + }, + { + "epoch": 0.31, + "learning_rate": 1.6922034643131798e-07, + "loss": 0.6857, + "step": 8455 + }, + { + "epoch": 0.31, + "learning_rate": 1.692166843666459e-07, + "loss": 0.6943, + "step": 8456 + }, + { + "epoch": 0.31, + "learning_rate": 1.6921302230197386e-07, + "loss": 0.6996, + "step": 8457 + }, + { + "epoch": 0.31, + "learning_rate": 1.6920936023730178e-07, + "loss": 0.7063, + "step": 8458 + }, + { + "epoch": 0.31, + "learning_rate": 1.6920569817262974e-07, + "loss": 0.6974, + "step": 8459 + }, + { + "epoch": 0.31, + "learning_rate": 1.6920203610795766e-07, + "loss": 0.6991, + "step": 8460 + }, + { + "epoch": 0.31, + "learning_rate": 1.6919837404328562e-07, + "loss": 0.683, + "step": 8461 + }, + { + "epoch": 0.31, + "learning_rate": 1.6919471197861352e-07, + "loss": 0.6978, + "step": 8462 + }, + { + "epoch": 0.31, + "learning_rate": 1.6919104991394147e-07, + "loss": 0.6868, + "step": 8463 + }, + { + "epoch": 0.31, + "learning_rate": 1.691873878492694e-07, + "loss": 0.7004, + "step": 8464 + }, + { + "epoch": 0.31, + "learning_rate": 1.6918372578459735e-07, + "loss": 0.709, + "step": 8465 + }, + { + "epoch": 0.31, + "learning_rate": 1.6918006371992528e-07, + "loss": 0.7013, + "step": 8466 + }, + { + "epoch": 0.31, + "learning_rate": 1.6917640165525323e-07, + "loss": 0.721, + "step": 8467 + }, + { + "epoch": 0.31, + "learning_rate": 1.6917273959058116e-07, + "loss": 0.7103, + "step": 8468 + }, + { + "epoch": 0.31, + "learning_rate": 1.691690775259091e-07, + "loss": 0.7052, + "step": 8469 + }, + { + "epoch": 0.31, + "learning_rate": 1.6916541546123704e-07, + "loss": 0.6997, + "step": 8470 + }, + { + "epoch": 0.31, + "learning_rate": 1.6916175339656496e-07, + "loss": 0.6888, + "step": 8471 + }, + { + "epoch": 0.31, + "learning_rate": 1.6915809133189292e-07, + "loss": 0.6975, + "step": 8472 + }, + { + "epoch": 0.31, + "learning_rate": 1.6915442926722084e-07, + "loss": 0.684, + "step": 8473 + }, + { + "epoch": 0.31, + "learning_rate": 1.691507672025488e-07, + "loss": 0.6921, + "step": 8474 + }, + { + "epoch": 0.31, + "learning_rate": 1.6914710513787672e-07, + "loss": 0.7106, + "step": 8475 + }, + { + "epoch": 0.31, + "learning_rate": 1.6914344307320468e-07, + "loss": 0.704, + "step": 8476 + }, + { + "epoch": 0.31, + "learning_rate": 1.691397810085326e-07, + "loss": 0.6882, + "step": 8477 + }, + { + "epoch": 0.31, + "learning_rate": 1.6913611894386056e-07, + "loss": 0.6792, + "step": 8478 + }, + { + "epoch": 0.31, + "learning_rate": 1.6913245687918848e-07, + "loss": 0.6979, + "step": 8479 + }, + { + "epoch": 0.31, + "learning_rate": 1.691287948145164e-07, + "loss": 0.6921, + "step": 8480 + }, + { + "epoch": 0.31, + "learning_rate": 1.6912513274984434e-07, + "loss": 0.7014, + "step": 8481 + }, + { + "epoch": 0.31, + "learning_rate": 1.691214706851723e-07, + "loss": 0.6964, + "step": 8482 + }, + { + "epoch": 0.31, + "learning_rate": 1.6911780862050022e-07, + "loss": 0.6933, + "step": 8483 + }, + { + "epoch": 0.31, + "learning_rate": 1.6911414655582817e-07, + "loss": 0.6903, + "step": 8484 + }, + { + "epoch": 0.31, + "learning_rate": 1.691104844911561e-07, + "loss": 0.688, + "step": 8485 + }, + { + "epoch": 0.31, + "learning_rate": 1.6910682242648405e-07, + "loss": 0.6997, + "step": 8486 + }, + { + "epoch": 0.31, + "learning_rate": 1.6910316036181198e-07, + "loss": 0.692, + "step": 8487 + }, + { + "epoch": 0.31, + "learning_rate": 1.6909949829713993e-07, + "loss": 0.693, + "step": 8488 + }, + { + "epoch": 0.31, + "learning_rate": 1.6909583623246786e-07, + "loss": 0.6873, + "step": 8489 + }, + { + "epoch": 0.31, + "learning_rate": 1.690921741677958e-07, + "loss": 0.6905, + "step": 8490 + }, + { + "epoch": 0.31, + "learning_rate": 1.6908851210312374e-07, + "loss": 0.6987, + "step": 8491 + }, + { + "epoch": 0.31, + "learning_rate": 1.690848500384517e-07, + "loss": 0.6784, + "step": 8492 + }, + { + "epoch": 0.31, + "learning_rate": 1.6908118797377962e-07, + "loss": 0.7002, + "step": 8493 + }, + { + "epoch": 0.31, + "learning_rate": 1.6907752590910754e-07, + "loss": 0.6887, + "step": 8494 + }, + { + "epoch": 0.31, + "learning_rate": 1.690738638444355e-07, + "loss": 0.6878, + "step": 8495 + }, + { + "epoch": 0.31, + "learning_rate": 1.6907020177976342e-07, + "loss": 0.6838, + "step": 8496 + }, + { + "epoch": 0.31, + "learning_rate": 1.6906653971509138e-07, + "loss": 0.6832, + "step": 8497 + }, + { + "epoch": 0.31, + "learning_rate": 1.6906287765041928e-07, + "loss": 0.7146, + "step": 8498 + }, + { + "epoch": 0.31, + "learning_rate": 1.6905921558574723e-07, + "loss": 0.6995, + "step": 8499 + }, + { + "epoch": 0.31, + "learning_rate": 1.6905555352107516e-07, + "loss": 0.6917, + "step": 8500 + }, + { + "epoch": 0.31, + "learning_rate": 1.690518914564031e-07, + "loss": 0.68, + "step": 8501 + }, + { + "epoch": 0.31, + "learning_rate": 1.6904822939173104e-07, + "loss": 0.6825, + "step": 8502 + }, + { + "epoch": 0.31, + "learning_rate": 1.69044567327059e-07, + "loss": 0.6812, + "step": 8503 + }, + { + "epoch": 0.31, + "learning_rate": 1.6904090526238692e-07, + "loss": 0.6864, + "step": 8504 + }, + { + "epoch": 0.31, + "learning_rate": 1.6903724319771487e-07, + "loss": 0.6983, + "step": 8505 + }, + { + "epoch": 0.31, + "learning_rate": 1.690335811330428e-07, + "loss": 0.7083, + "step": 8506 + }, + { + "epoch": 0.31, + "learning_rate": 1.6902991906837075e-07, + "loss": 0.684, + "step": 8507 + }, + { + "epoch": 0.31, + "learning_rate": 1.6902625700369868e-07, + "loss": 0.6963, + "step": 8508 + }, + { + "epoch": 0.31, + "learning_rate": 1.6902259493902663e-07, + "loss": 0.6841, + "step": 8509 + }, + { + "epoch": 0.31, + "learning_rate": 1.6901893287435456e-07, + "loss": 0.6806, + "step": 8510 + }, + { + "epoch": 0.31, + "learning_rate": 1.690152708096825e-07, + "loss": 0.6947, + "step": 8511 + }, + { + "epoch": 0.31, + "learning_rate": 1.6901160874501044e-07, + "loss": 0.6878, + "step": 8512 + }, + { + "epoch": 0.31, + "learning_rate": 1.690079466803384e-07, + "loss": 0.7056, + "step": 8513 + }, + { + "epoch": 0.31, + "learning_rate": 1.6900428461566631e-07, + "loss": 0.6934, + "step": 8514 + }, + { + "epoch": 0.31, + "learning_rate": 1.6900062255099424e-07, + "loss": 0.6963, + "step": 8515 + }, + { + "epoch": 0.31, + "learning_rate": 1.6899696048632217e-07, + "loss": 0.6962, + "step": 8516 + }, + { + "epoch": 0.31, + "learning_rate": 1.6899329842165012e-07, + "loss": 0.704, + "step": 8517 + }, + { + "epoch": 0.31, + "learning_rate": 1.6898963635697805e-07, + "loss": 0.7039, + "step": 8518 + }, + { + "epoch": 0.31, + "learning_rate": 1.68985974292306e-07, + "loss": 0.6942, + "step": 8519 + }, + { + "epoch": 0.31, + "learning_rate": 1.6898231222763393e-07, + "loss": 0.6939, + "step": 8520 + }, + { + "epoch": 0.31, + "learning_rate": 1.6897865016296185e-07, + "loss": 0.6989, + "step": 8521 + }, + { + "epoch": 0.31, + "learning_rate": 1.689749880982898e-07, + "loss": 0.6859, + "step": 8522 + }, + { + "epoch": 0.31, + "learning_rate": 1.6897132603361773e-07, + "loss": 0.6801, + "step": 8523 + }, + { + "epoch": 0.31, + "learning_rate": 1.689676639689457e-07, + "loss": 0.6872, + "step": 8524 + }, + { + "epoch": 0.31, + "learning_rate": 1.6896400190427361e-07, + "loss": 0.6823, + "step": 8525 + }, + { + "epoch": 0.31, + "learning_rate": 1.6896033983960157e-07, + "loss": 0.6937, + "step": 8526 + }, + { + "epoch": 0.31, + "learning_rate": 1.689566777749295e-07, + "loss": 0.6931, + "step": 8527 + }, + { + "epoch": 0.31, + "learning_rate": 1.6895301571025745e-07, + "loss": 0.6893, + "step": 8528 + }, + { + "epoch": 0.31, + "learning_rate": 1.6894935364558537e-07, + "loss": 0.682, + "step": 8529 + }, + { + "epoch": 0.31, + "learning_rate": 1.6894569158091333e-07, + "loss": 0.6967, + "step": 8530 + }, + { + "epoch": 0.31, + "learning_rate": 1.6894202951624125e-07, + "loss": 0.6894, + "step": 8531 + }, + { + "epoch": 0.31, + "learning_rate": 1.689383674515692e-07, + "loss": 0.6887, + "step": 8532 + }, + { + "epoch": 0.31, + "learning_rate": 1.689347053868971e-07, + "loss": 0.6938, + "step": 8533 + }, + { + "epoch": 0.31, + "learning_rate": 1.6893104332222506e-07, + "loss": 0.7087, + "step": 8534 + }, + { + "epoch": 0.31, + "learning_rate": 1.68927381257553e-07, + "loss": 0.6908, + "step": 8535 + }, + { + "epoch": 0.31, + "learning_rate": 1.6892371919288094e-07, + "loss": 0.6715, + "step": 8536 + }, + { + "epoch": 0.31, + "learning_rate": 1.6892005712820887e-07, + "loss": 0.6782, + "step": 8537 + }, + { + "epoch": 0.31, + "learning_rate": 1.6891639506353682e-07, + "loss": 0.6975, + "step": 8538 + }, + { + "epoch": 0.31, + "learning_rate": 1.6891273299886475e-07, + "loss": 0.7056, + "step": 8539 + }, + { + "epoch": 0.31, + "learning_rate": 1.689090709341927e-07, + "loss": 0.7204, + "step": 8540 + }, + { + "epoch": 0.31, + "learning_rate": 1.6890540886952063e-07, + "loss": 0.7005, + "step": 8541 + }, + { + "epoch": 0.31, + "learning_rate": 1.6890174680484855e-07, + "loss": 0.7005, + "step": 8542 + }, + { + "epoch": 0.31, + "learning_rate": 1.688980847401765e-07, + "loss": 0.6858, + "step": 8543 + }, + { + "epoch": 0.31, + "learning_rate": 1.6889442267550443e-07, + "loss": 0.7003, + "step": 8544 + }, + { + "epoch": 0.31, + "learning_rate": 1.6889076061083239e-07, + "loss": 0.678, + "step": 8545 + }, + { + "epoch": 0.31, + "learning_rate": 1.6888709854616031e-07, + "loss": 0.7056, + "step": 8546 + }, + { + "epoch": 0.31, + "learning_rate": 1.6888343648148827e-07, + "loss": 0.6876, + "step": 8547 + }, + { + "epoch": 0.31, + "learning_rate": 1.688797744168162e-07, + "loss": 0.6966, + "step": 8548 + }, + { + "epoch": 0.31, + "learning_rate": 1.6887611235214415e-07, + "loss": 0.7028, + "step": 8549 + }, + { + "epoch": 0.31, + "learning_rate": 1.6887245028747207e-07, + "loss": 0.6899, + "step": 8550 + }, + { + "epoch": 0.31, + "learning_rate": 1.688687882228e-07, + "loss": 0.693, + "step": 8551 + }, + { + "epoch": 0.31, + "learning_rate": 1.6886512615812793e-07, + "loss": 0.7057, + "step": 8552 + }, + { + "epoch": 0.31, + "learning_rate": 1.6886146409345588e-07, + "loss": 0.6999, + "step": 8553 + }, + { + "epoch": 0.31, + "learning_rate": 1.688578020287838e-07, + "loss": 0.6921, + "step": 8554 + }, + { + "epoch": 0.31, + "learning_rate": 1.6885413996411176e-07, + "loss": 0.7005, + "step": 8555 + }, + { + "epoch": 0.31, + "learning_rate": 1.6885047789943969e-07, + "loss": 0.7062, + "step": 8556 + }, + { + "epoch": 0.31, + "learning_rate": 1.6884681583476764e-07, + "loss": 0.6917, + "step": 8557 + }, + { + "epoch": 0.31, + "learning_rate": 1.6884315377009557e-07, + "loss": 0.6867, + "step": 8558 + }, + { + "epoch": 0.31, + "learning_rate": 1.6883949170542352e-07, + "loss": 0.6901, + "step": 8559 + }, + { + "epoch": 0.31, + "learning_rate": 1.6883582964075145e-07, + "loss": 0.687, + "step": 8560 + }, + { + "epoch": 0.31, + "learning_rate": 1.688321675760794e-07, + "loss": 0.7035, + "step": 8561 + }, + { + "epoch": 0.31, + "learning_rate": 1.6882850551140733e-07, + "loss": 0.6946, + "step": 8562 + }, + { + "epoch": 0.31, + "learning_rate": 1.6882484344673528e-07, + "loss": 0.6843, + "step": 8563 + }, + { + "epoch": 0.31, + "learning_rate": 1.688211813820632e-07, + "loss": 0.6834, + "step": 8564 + }, + { + "epoch": 0.31, + "learning_rate": 1.6881751931739113e-07, + "loss": 0.7063, + "step": 8565 + }, + { + "epoch": 0.31, + "learning_rate": 1.6881385725271909e-07, + "loss": 0.7036, + "step": 8566 + }, + { + "epoch": 0.31, + "learning_rate": 1.68810195188047e-07, + "loss": 0.6753, + "step": 8567 + }, + { + "epoch": 0.31, + "learning_rate": 1.6880653312337497e-07, + "loss": 0.6785, + "step": 8568 + }, + { + "epoch": 0.31, + "learning_rate": 1.6880287105870287e-07, + "loss": 0.692, + "step": 8569 + }, + { + "epoch": 0.31, + "learning_rate": 1.6879920899403082e-07, + "loss": 0.6919, + "step": 8570 + }, + { + "epoch": 0.31, + "learning_rate": 1.6879554692935875e-07, + "loss": 0.7157, + "step": 8571 + }, + { + "epoch": 0.31, + "learning_rate": 1.687918848646867e-07, + "loss": 0.7079, + "step": 8572 + }, + { + "epoch": 0.31, + "learning_rate": 1.6878822280001463e-07, + "loss": 0.6973, + "step": 8573 + }, + { + "epoch": 0.31, + "learning_rate": 1.6878456073534258e-07, + "loss": 0.686, + "step": 8574 + }, + { + "epoch": 0.31, + "learning_rate": 1.687808986706705e-07, + "loss": 0.694, + "step": 8575 + }, + { + "epoch": 0.31, + "learning_rate": 1.6877723660599846e-07, + "loss": 0.6941, + "step": 8576 + }, + { + "epoch": 0.31, + "learning_rate": 1.6877357454132638e-07, + "loss": 0.6879, + "step": 8577 + }, + { + "epoch": 0.31, + "learning_rate": 1.6876991247665434e-07, + "loss": 0.696, + "step": 8578 + }, + { + "epoch": 0.31, + "learning_rate": 1.6876625041198226e-07, + "loss": 0.6936, + "step": 8579 + }, + { + "epoch": 0.31, + "learning_rate": 1.6876258834731022e-07, + "loss": 0.6917, + "step": 8580 + }, + { + "epoch": 0.31, + "learning_rate": 1.6875892628263814e-07, + "loss": 0.6911, + "step": 8581 + }, + { + "epoch": 0.31, + "learning_rate": 1.687552642179661e-07, + "loss": 0.6979, + "step": 8582 + }, + { + "epoch": 0.31, + "learning_rate": 1.6875160215329402e-07, + "loss": 0.7005, + "step": 8583 + }, + { + "epoch": 0.31, + "learning_rate": 1.6874794008862198e-07, + "loss": 0.6906, + "step": 8584 + }, + { + "epoch": 0.31, + "learning_rate": 1.687442780239499e-07, + "loss": 0.6786, + "step": 8585 + }, + { + "epoch": 0.31, + "learning_rate": 1.6874061595927783e-07, + "loss": 0.6902, + "step": 8586 + }, + { + "epoch": 0.31, + "learning_rate": 1.6873695389460576e-07, + "loss": 0.6844, + "step": 8587 + }, + { + "epoch": 0.31, + "learning_rate": 1.687332918299337e-07, + "loss": 0.7064, + "step": 8588 + }, + { + "epoch": 0.31, + "learning_rate": 1.6872962976526164e-07, + "loss": 0.6891, + "step": 8589 + }, + { + "epoch": 0.31, + "learning_rate": 1.687259677005896e-07, + "loss": 0.6761, + "step": 8590 + }, + { + "epoch": 0.31, + "learning_rate": 1.6872230563591752e-07, + "loss": 0.6765, + "step": 8591 + }, + { + "epoch": 0.31, + "learning_rate": 1.6871864357124544e-07, + "loss": 0.6813, + "step": 8592 + }, + { + "epoch": 0.31, + "learning_rate": 1.687149815065734e-07, + "loss": 0.7193, + "step": 8593 + }, + { + "epoch": 0.31, + "learning_rate": 1.6871131944190132e-07, + "loss": 0.6921, + "step": 8594 + }, + { + "epoch": 0.31, + "learning_rate": 1.6870765737722928e-07, + "loss": 0.6818, + "step": 8595 + }, + { + "epoch": 0.31, + "learning_rate": 1.687039953125572e-07, + "loss": 0.6912, + "step": 8596 + }, + { + "epoch": 0.31, + "learning_rate": 1.6870033324788516e-07, + "loss": 0.6974, + "step": 8597 + }, + { + "epoch": 0.31, + "learning_rate": 1.6869667118321308e-07, + "loss": 0.6892, + "step": 8598 + }, + { + "epoch": 0.31, + "learning_rate": 1.6869300911854104e-07, + "loss": 0.687, + "step": 8599 + }, + { + "epoch": 0.31, + "learning_rate": 1.6868934705386896e-07, + "loss": 0.6961, + "step": 8600 + }, + { + "epoch": 0.31, + "learning_rate": 1.6868568498919692e-07, + "loss": 0.7024, + "step": 8601 + }, + { + "epoch": 0.31, + "learning_rate": 1.6868202292452484e-07, + "loss": 0.6843, + "step": 8602 + }, + { + "epoch": 0.31, + "learning_rate": 1.686783608598528e-07, + "loss": 0.6903, + "step": 8603 + }, + { + "epoch": 0.31, + "learning_rate": 1.6867469879518072e-07, + "loss": 0.6927, + "step": 8604 + }, + { + "epoch": 0.31, + "learning_rate": 1.6867103673050865e-07, + "loss": 0.6928, + "step": 8605 + }, + { + "epoch": 0.31, + "learning_rate": 1.6866737466583658e-07, + "loss": 0.6891, + "step": 8606 + }, + { + "epoch": 0.31, + "learning_rate": 1.6866371260116453e-07, + "loss": 0.7016, + "step": 8607 + }, + { + "epoch": 0.31, + "learning_rate": 1.6866005053649246e-07, + "loss": 0.7029, + "step": 8608 + }, + { + "epoch": 0.31, + "learning_rate": 1.686563884718204e-07, + "loss": 0.6936, + "step": 8609 + }, + { + "epoch": 0.32, + "learning_rate": 1.6865272640714834e-07, + "loss": 0.6944, + "step": 8610 + }, + { + "epoch": 0.32, + "learning_rate": 1.686490643424763e-07, + "loss": 0.6837, + "step": 8611 + }, + { + "epoch": 0.32, + "learning_rate": 1.6864540227780422e-07, + "loss": 0.6941, + "step": 8612 + }, + { + "epoch": 0.32, + "learning_rate": 1.6864174021313214e-07, + "loss": 0.6915, + "step": 8613 + }, + { + "epoch": 0.32, + "learning_rate": 1.686380781484601e-07, + "loss": 0.6834, + "step": 8614 + }, + { + "epoch": 0.32, + "learning_rate": 1.6863441608378802e-07, + "loss": 0.6998, + "step": 8615 + }, + { + "epoch": 0.32, + "learning_rate": 1.6863075401911598e-07, + "loss": 0.6947, + "step": 8616 + }, + { + "epoch": 0.32, + "learning_rate": 1.686270919544439e-07, + "loss": 0.6905, + "step": 8617 + }, + { + "epoch": 0.32, + "learning_rate": 1.6862342988977186e-07, + "loss": 0.7099, + "step": 8618 + }, + { + "epoch": 0.32, + "learning_rate": 1.6861976782509978e-07, + "loss": 0.6991, + "step": 8619 + }, + { + "epoch": 0.32, + "learning_rate": 1.6861610576042774e-07, + "loss": 0.7098, + "step": 8620 + }, + { + "epoch": 0.32, + "learning_rate": 1.6861244369575566e-07, + "loss": 0.6926, + "step": 8621 + }, + { + "epoch": 0.32, + "learning_rate": 1.6860878163108362e-07, + "loss": 0.6899, + "step": 8622 + }, + { + "epoch": 0.32, + "learning_rate": 1.6860511956641152e-07, + "loss": 0.6913, + "step": 8623 + }, + { + "epoch": 0.32, + "learning_rate": 1.6860145750173947e-07, + "loss": 0.7007, + "step": 8624 + }, + { + "epoch": 0.32, + "learning_rate": 1.685977954370674e-07, + "loss": 0.6801, + "step": 8625 + }, + { + "epoch": 0.32, + "learning_rate": 1.6859413337239535e-07, + "loss": 0.7136, + "step": 8626 + }, + { + "epoch": 0.32, + "learning_rate": 1.6859047130772328e-07, + "loss": 0.6976, + "step": 8627 + }, + { + "epoch": 0.32, + "learning_rate": 1.6858680924305123e-07, + "loss": 0.6954, + "step": 8628 + }, + { + "epoch": 0.32, + "learning_rate": 1.6858314717837916e-07, + "loss": 0.6913, + "step": 8629 + }, + { + "epoch": 0.32, + "learning_rate": 1.685794851137071e-07, + "loss": 0.6895, + "step": 8630 + }, + { + "epoch": 0.32, + "learning_rate": 1.6857582304903504e-07, + "loss": 0.6993, + "step": 8631 + }, + { + "epoch": 0.32, + "learning_rate": 1.68572160984363e-07, + "loss": 0.6984, + "step": 8632 + }, + { + "epoch": 0.32, + "learning_rate": 1.6856849891969091e-07, + "loss": 0.6891, + "step": 8633 + }, + { + "epoch": 0.32, + "learning_rate": 1.6856483685501884e-07, + "loss": 0.6965, + "step": 8634 + }, + { + "epoch": 0.32, + "learning_rate": 1.685611747903468e-07, + "loss": 0.6881, + "step": 8635 + }, + { + "epoch": 0.32, + "learning_rate": 1.6855751272567472e-07, + "loss": 0.6857, + "step": 8636 + }, + { + "epoch": 0.32, + "learning_rate": 1.6855385066100267e-07, + "loss": 0.6918, + "step": 8637 + }, + { + "epoch": 0.32, + "learning_rate": 1.685501885963306e-07, + "loss": 0.6893, + "step": 8638 + }, + { + "epoch": 0.32, + "learning_rate": 1.6854652653165855e-07, + "loss": 0.6843, + "step": 8639 + }, + { + "epoch": 0.32, + "learning_rate": 1.6854286446698648e-07, + "loss": 0.6955, + "step": 8640 + }, + { + "epoch": 0.32, + "learning_rate": 1.685392024023144e-07, + "loss": 0.6975, + "step": 8641 + }, + { + "epoch": 0.32, + "learning_rate": 1.6853554033764233e-07, + "loss": 0.6929, + "step": 8642 + }, + { + "epoch": 0.32, + "learning_rate": 1.685318782729703e-07, + "loss": 0.6779, + "step": 8643 + }, + { + "epoch": 0.32, + "learning_rate": 1.6852821620829821e-07, + "loss": 0.7048, + "step": 8644 + }, + { + "epoch": 0.32, + "learning_rate": 1.6852455414362617e-07, + "loss": 0.6782, + "step": 8645 + }, + { + "epoch": 0.32, + "learning_rate": 1.685208920789541e-07, + "loss": 0.6988, + "step": 8646 + }, + { + "epoch": 0.32, + "learning_rate": 1.6851723001428205e-07, + "loss": 0.6919, + "step": 8647 + }, + { + "epoch": 0.32, + "learning_rate": 1.6851356794960997e-07, + "loss": 0.6832, + "step": 8648 + }, + { + "epoch": 0.32, + "learning_rate": 1.6850990588493793e-07, + "loss": 0.6793, + "step": 8649 + }, + { + "epoch": 0.32, + "learning_rate": 1.6850624382026585e-07, + "loss": 0.6812, + "step": 8650 + }, + { + "epoch": 0.32, + "learning_rate": 1.685025817555938e-07, + "loss": 0.6916, + "step": 8651 + }, + { + "epoch": 0.32, + "learning_rate": 1.6849891969092173e-07, + "loss": 0.704, + "step": 8652 + }, + { + "epoch": 0.32, + "learning_rate": 1.684952576262497e-07, + "loss": 0.6805, + "step": 8653 + }, + { + "epoch": 0.32, + "learning_rate": 1.6849159556157761e-07, + "loss": 0.6869, + "step": 8654 + }, + { + "epoch": 0.32, + "learning_rate": 1.6848793349690557e-07, + "loss": 0.6898, + "step": 8655 + }, + { + "epoch": 0.32, + "learning_rate": 1.684842714322335e-07, + "loss": 0.6966, + "step": 8656 + }, + { + "epoch": 0.32, + "learning_rate": 1.6848060936756142e-07, + "loss": 0.6891, + "step": 8657 + }, + { + "epoch": 0.32, + "learning_rate": 1.6847694730288937e-07, + "loss": 0.7074, + "step": 8658 + }, + { + "epoch": 0.32, + "learning_rate": 1.684732852382173e-07, + "loss": 0.6821, + "step": 8659 + }, + { + "epoch": 0.32, + "learning_rate": 1.6846962317354523e-07, + "loss": 0.7047, + "step": 8660 + }, + { + "epoch": 0.32, + "learning_rate": 1.6846596110887318e-07, + "loss": 0.6878, + "step": 8661 + }, + { + "epoch": 0.32, + "learning_rate": 1.684622990442011e-07, + "loss": 0.6943, + "step": 8662 + }, + { + "epoch": 0.32, + "learning_rate": 1.6845863697952903e-07, + "loss": 0.7136, + "step": 8663 + }, + { + "epoch": 0.32, + "learning_rate": 1.6845497491485699e-07, + "loss": 0.6831, + "step": 8664 + }, + { + "epoch": 0.32, + "learning_rate": 1.6845131285018491e-07, + "loss": 0.7039, + "step": 8665 + }, + { + "epoch": 0.32, + "learning_rate": 1.6844765078551287e-07, + "loss": 0.6893, + "step": 8666 + }, + { + "epoch": 0.32, + "learning_rate": 1.684439887208408e-07, + "loss": 0.6889, + "step": 8667 + }, + { + "epoch": 0.32, + "learning_rate": 1.6844032665616875e-07, + "loss": 0.6813, + "step": 8668 + }, + { + "epoch": 0.32, + "learning_rate": 1.6843666459149667e-07, + "loss": 0.6827, + "step": 8669 + }, + { + "epoch": 0.32, + "learning_rate": 1.6843300252682463e-07, + "loss": 0.7124, + "step": 8670 + }, + { + "epoch": 0.32, + "learning_rate": 1.6842934046215255e-07, + "loss": 0.6942, + "step": 8671 + }, + { + "epoch": 0.32, + "learning_rate": 1.684256783974805e-07, + "loss": 0.7092, + "step": 8672 + }, + { + "epoch": 0.32, + "learning_rate": 1.6842201633280843e-07, + "loss": 0.7074, + "step": 8673 + }, + { + "epoch": 0.32, + "learning_rate": 1.6841835426813639e-07, + "loss": 0.6863, + "step": 8674 + }, + { + "epoch": 0.32, + "learning_rate": 1.684146922034643e-07, + "loss": 0.6856, + "step": 8675 + }, + { + "epoch": 0.32, + "learning_rate": 1.6841103013879227e-07, + "loss": 0.6929, + "step": 8676 + }, + { + "epoch": 0.32, + "learning_rate": 1.6840736807412017e-07, + "loss": 0.6952, + "step": 8677 + }, + { + "epoch": 0.32, + "learning_rate": 1.6840370600944812e-07, + "loss": 0.6905, + "step": 8678 + }, + { + "epoch": 0.32, + "learning_rate": 1.6840004394477605e-07, + "loss": 0.6944, + "step": 8679 + }, + { + "epoch": 0.32, + "learning_rate": 1.68396381880104e-07, + "loss": 0.6888, + "step": 8680 + }, + { + "epoch": 0.32, + "learning_rate": 1.6839271981543193e-07, + "loss": 0.6958, + "step": 8681 + }, + { + "epoch": 0.32, + "learning_rate": 1.6838905775075988e-07, + "loss": 0.6997, + "step": 8682 + }, + { + "epoch": 0.32, + "learning_rate": 1.683853956860878e-07, + "loss": 0.6973, + "step": 8683 + }, + { + "epoch": 0.32, + "learning_rate": 1.6838173362141573e-07, + "loss": 0.7101, + "step": 8684 + }, + { + "epoch": 0.32, + "learning_rate": 1.6837807155674369e-07, + "loss": 0.6763, + "step": 8685 + }, + { + "epoch": 0.32, + "learning_rate": 1.683744094920716e-07, + "loss": 0.6802, + "step": 8686 + }, + { + "epoch": 0.32, + "learning_rate": 1.6837074742739957e-07, + "loss": 0.7012, + "step": 8687 + }, + { + "epoch": 0.32, + "learning_rate": 1.683670853627275e-07, + "loss": 0.702, + "step": 8688 + }, + { + "epoch": 0.32, + "learning_rate": 1.6836342329805545e-07, + "loss": 0.6916, + "step": 8689 + }, + { + "epoch": 0.32, + "learning_rate": 1.6835976123338337e-07, + "loss": 0.6882, + "step": 8690 + }, + { + "epoch": 0.32, + "learning_rate": 1.6835609916871132e-07, + "loss": 0.6817, + "step": 8691 + }, + { + "epoch": 0.32, + "learning_rate": 1.6835243710403925e-07, + "loss": 0.6965, + "step": 8692 + }, + { + "epoch": 0.32, + "learning_rate": 1.683487750393672e-07, + "loss": 0.7027, + "step": 8693 + }, + { + "epoch": 0.32, + "learning_rate": 1.6834511297469513e-07, + "loss": 0.6924, + "step": 8694 + }, + { + "epoch": 0.32, + "learning_rate": 1.6834145091002306e-07, + "loss": 0.6935, + "step": 8695 + }, + { + "epoch": 0.32, + "learning_rate": 1.6833778884535098e-07, + "loss": 0.6937, + "step": 8696 + }, + { + "epoch": 0.32, + "learning_rate": 1.6833412678067894e-07, + "loss": 0.6743, + "step": 8697 + }, + { + "epoch": 0.32, + "learning_rate": 1.6833046471600686e-07, + "loss": 0.6926, + "step": 8698 + }, + { + "epoch": 0.32, + "learning_rate": 1.6832680265133482e-07, + "loss": 0.6882, + "step": 8699 + }, + { + "epoch": 0.32, + "learning_rate": 1.6832314058666274e-07, + "loss": 0.6883, + "step": 8700 + }, + { + "epoch": 0.32, + "learning_rate": 1.683194785219907e-07, + "loss": 0.6941, + "step": 8701 + }, + { + "epoch": 0.32, + "learning_rate": 1.6831581645731862e-07, + "loss": 0.6911, + "step": 8702 + }, + { + "epoch": 0.32, + "learning_rate": 1.6831215439264658e-07, + "loss": 0.7106, + "step": 8703 + }, + { + "epoch": 0.32, + "learning_rate": 1.683084923279745e-07, + "loss": 0.672, + "step": 8704 + }, + { + "epoch": 0.32, + "learning_rate": 1.6830483026330243e-07, + "loss": 0.7004, + "step": 8705 + }, + { + "epoch": 0.32, + "learning_rate": 1.6830116819863038e-07, + "loss": 0.699, + "step": 8706 + }, + { + "epoch": 0.32, + "learning_rate": 1.682975061339583e-07, + "loss": 0.6884, + "step": 8707 + }, + { + "epoch": 0.32, + "learning_rate": 1.6829384406928626e-07, + "loss": 0.7069, + "step": 8708 + }, + { + "epoch": 0.32, + "learning_rate": 1.682901820046142e-07, + "loss": 0.6919, + "step": 8709 + }, + { + "epoch": 0.32, + "learning_rate": 1.6828651993994214e-07, + "loss": 0.6904, + "step": 8710 + }, + { + "epoch": 0.32, + "learning_rate": 1.6828285787527007e-07, + "loss": 0.6671, + "step": 8711 + }, + { + "epoch": 0.32, + "learning_rate": 1.6827919581059802e-07, + "loss": 0.7065, + "step": 8712 + }, + { + "epoch": 0.32, + "learning_rate": 1.6827553374592592e-07, + "loss": 0.6901, + "step": 8713 + }, + { + "epoch": 0.32, + "learning_rate": 1.6827187168125388e-07, + "loss": 0.6775, + "step": 8714 + }, + { + "epoch": 0.32, + "learning_rate": 1.682682096165818e-07, + "loss": 0.6896, + "step": 8715 + }, + { + "epoch": 0.32, + "learning_rate": 1.6826454755190976e-07, + "loss": 0.6948, + "step": 8716 + }, + { + "epoch": 0.32, + "learning_rate": 1.6826088548723768e-07, + "loss": 0.6817, + "step": 8717 + }, + { + "epoch": 0.32, + "learning_rate": 1.6825722342256564e-07, + "loss": 0.6802, + "step": 8718 + }, + { + "epoch": 0.32, + "learning_rate": 1.6825356135789356e-07, + "loss": 0.7023, + "step": 8719 + }, + { + "epoch": 0.32, + "learning_rate": 1.6824989929322152e-07, + "loss": 0.6971, + "step": 8720 + }, + { + "epoch": 0.32, + "learning_rate": 1.6824623722854944e-07, + "loss": 0.6956, + "step": 8721 + }, + { + "epoch": 0.32, + "learning_rate": 1.682425751638774e-07, + "loss": 0.6823, + "step": 8722 + }, + { + "epoch": 0.32, + "learning_rate": 1.6823891309920532e-07, + "loss": 0.6978, + "step": 8723 + }, + { + "epoch": 0.32, + "learning_rate": 1.6823525103453328e-07, + "loss": 0.7082, + "step": 8724 + }, + { + "epoch": 0.32, + "learning_rate": 1.682315889698612e-07, + "loss": 0.7018, + "step": 8725 + }, + { + "epoch": 0.32, + "learning_rate": 1.6822792690518916e-07, + "loss": 0.6892, + "step": 8726 + }, + { + "epoch": 0.32, + "learning_rate": 1.6822426484051708e-07, + "loss": 0.6899, + "step": 8727 + }, + { + "epoch": 0.32, + "learning_rate": 1.68220602775845e-07, + "loss": 0.6882, + "step": 8728 + }, + { + "epoch": 0.32, + "learning_rate": 1.6821694071117296e-07, + "loss": 0.7111, + "step": 8729 + }, + { + "epoch": 0.32, + "learning_rate": 1.682132786465009e-07, + "loss": 0.6807, + "step": 8730 + }, + { + "epoch": 0.32, + "learning_rate": 1.6820961658182882e-07, + "loss": 0.6828, + "step": 8731 + }, + { + "epoch": 0.32, + "learning_rate": 1.6820595451715677e-07, + "loss": 0.6949, + "step": 8732 + }, + { + "epoch": 0.32, + "learning_rate": 1.682022924524847e-07, + "loss": 0.6663, + "step": 8733 + }, + { + "epoch": 0.32, + "learning_rate": 1.6819863038781262e-07, + "loss": 0.6979, + "step": 8734 + }, + { + "epoch": 0.32, + "learning_rate": 1.6819496832314058e-07, + "loss": 0.7009, + "step": 8735 + }, + { + "epoch": 0.32, + "learning_rate": 1.681913062584685e-07, + "loss": 0.6972, + "step": 8736 + }, + { + "epoch": 0.32, + "learning_rate": 1.6818764419379646e-07, + "loss": 0.6952, + "step": 8737 + }, + { + "epoch": 0.32, + "learning_rate": 1.6818398212912438e-07, + "loss": 0.7065, + "step": 8738 + }, + { + "epoch": 0.32, + "learning_rate": 1.6818032006445234e-07, + "loss": 0.6919, + "step": 8739 + }, + { + "epoch": 0.32, + "learning_rate": 1.6817665799978026e-07, + "loss": 0.6856, + "step": 8740 + }, + { + "epoch": 0.32, + "learning_rate": 1.6817299593510822e-07, + "loss": 0.6918, + "step": 8741 + }, + { + "epoch": 0.32, + "learning_rate": 1.6816933387043614e-07, + "loss": 0.693, + "step": 8742 + }, + { + "epoch": 0.32, + "learning_rate": 1.681656718057641e-07, + "loss": 0.6889, + "step": 8743 + }, + { + "epoch": 0.32, + "learning_rate": 1.6816200974109202e-07, + "loss": 0.6982, + "step": 8744 + }, + { + "epoch": 0.32, + "learning_rate": 1.6815834767641998e-07, + "loss": 0.6902, + "step": 8745 + }, + { + "epoch": 0.32, + "learning_rate": 1.681546856117479e-07, + "loss": 0.6955, + "step": 8746 + }, + { + "epoch": 0.32, + "learning_rate": 1.6815102354707586e-07, + "loss": 0.6786, + "step": 8747 + }, + { + "epoch": 0.32, + "learning_rate": 1.6814736148240378e-07, + "loss": 0.68, + "step": 8748 + }, + { + "epoch": 0.32, + "learning_rate": 1.681436994177317e-07, + "loss": 0.7084, + "step": 8749 + }, + { + "epoch": 0.32, + "learning_rate": 1.6814003735305964e-07, + "loss": 0.6964, + "step": 8750 + }, + { + "epoch": 0.32, + "learning_rate": 1.681363752883876e-07, + "loss": 0.6911, + "step": 8751 + }, + { + "epoch": 0.32, + "learning_rate": 1.6813271322371552e-07, + "loss": 0.71, + "step": 8752 + }, + { + "epoch": 0.32, + "learning_rate": 1.6812905115904347e-07, + "loss": 0.6826, + "step": 8753 + }, + { + "epoch": 0.32, + "learning_rate": 1.681253890943714e-07, + "loss": 0.7184, + "step": 8754 + }, + { + "epoch": 0.32, + "learning_rate": 1.6812172702969932e-07, + "loss": 0.6942, + "step": 8755 + }, + { + "epoch": 0.32, + "learning_rate": 1.6811806496502727e-07, + "loss": 0.6993, + "step": 8756 + }, + { + "epoch": 0.32, + "learning_rate": 1.681144029003552e-07, + "loss": 0.7011, + "step": 8757 + }, + { + "epoch": 0.32, + "learning_rate": 1.6811074083568315e-07, + "loss": 0.6956, + "step": 8758 + }, + { + "epoch": 0.32, + "learning_rate": 1.6810707877101108e-07, + "loss": 0.701, + "step": 8759 + }, + { + "epoch": 0.32, + "learning_rate": 1.6810341670633903e-07, + "loss": 0.6925, + "step": 8760 + }, + { + "epoch": 0.32, + "learning_rate": 1.6809975464166696e-07, + "loss": 0.6862, + "step": 8761 + }, + { + "epoch": 0.32, + "learning_rate": 1.6809609257699491e-07, + "loss": 0.6925, + "step": 8762 + }, + { + "epoch": 0.32, + "learning_rate": 1.6809243051232284e-07, + "loss": 0.6702, + "step": 8763 + }, + { + "epoch": 0.32, + "learning_rate": 1.680887684476508e-07, + "loss": 0.7003, + "step": 8764 + }, + { + "epoch": 0.32, + "learning_rate": 1.6808510638297872e-07, + "loss": 0.6967, + "step": 8765 + }, + { + "epoch": 0.32, + "learning_rate": 1.6808144431830667e-07, + "loss": 0.6986, + "step": 8766 + }, + { + "epoch": 0.32, + "learning_rate": 1.6807778225363457e-07, + "loss": 0.688, + "step": 8767 + }, + { + "epoch": 0.32, + "learning_rate": 1.6807412018896253e-07, + "loss": 0.6757, + "step": 8768 + }, + { + "epoch": 0.32, + "learning_rate": 1.6807045812429045e-07, + "loss": 0.6989, + "step": 8769 + }, + { + "epoch": 0.32, + "learning_rate": 1.680667960596184e-07, + "loss": 0.6984, + "step": 8770 + }, + { + "epoch": 0.32, + "learning_rate": 1.6806313399494633e-07, + "loss": 0.71, + "step": 8771 + }, + { + "epoch": 0.32, + "learning_rate": 1.680594719302743e-07, + "loss": 0.6815, + "step": 8772 + }, + { + "epoch": 0.32, + "learning_rate": 1.6805580986560221e-07, + "loss": 0.6971, + "step": 8773 + }, + { + "epoch": 0.32, + "learning_rate": 1.6805214780093017e-07, + "loss": 0.7092, + "step": 8774 + }, + { + "epoch": 0.32, + "learning_rate": 1.680484857362581e-07, + "loss": 0.7002, + "step": 8775 + }, + { + "epoch": 0.32, + "learning_rate": 1.6804482367158602e-07, + "loss": 0.6943, + "step": 8776 + }, + { + "epoch": 0.32, + "learning_rate": 1.6804116160691397e-07, + "loss": 0.6957, + "step": 8777 + }, + { + "epoch": 0.32, + "learning_rate": 1.680374995422419e-07, + "loss": 0.7035, + "step": 8778 + }, + { + "epoch": 0.32, + "learning_rate": 1.6803383747756985e-07, + "loss": 0.6979, + "step": 8779 + }, + { + "epoch": 0.32, + "learning_rate": 1.6803017541289778e-07, + "loss": 0.6983, + "step": 8780 + }, + { + "epoch": 0.32, + "learning_rate": 1.6802651334822573e-07, + "loss": 0.6848, + "step": 8781 + }, + { + "epoch": 0.32, + "learning_rate": 1.6802285128355366e-07, + "loss": 0.7003, + "step": 8782 + }, + { + "epoch": 0.32, + "learning_rate": 1.680191892188816e-07, + "loss": 0.6968, + "step": 8783 + }, + { + "epoch": 0.32, + "learning_rate": 1.6801552715420954e-07, + "loss": 0.6926, + "step": 8784 + }, + { + "epoch": 0.32, + "learning_rate": 1.6801186508953747e-07, + "loss": 0.6916, + "step": 8785 + }, + { + "epoch": 0.32, + "learning_rate": 1.680082030248654e-07, + "loss": 0.6945, + "step": 8786 + }, + { + "epoch": 0.32, + "learning_rate": 1.6800454096019335e-07, + "loss": 0.7014, + "step": 8787 + }, + { + "epoch": 0.32, + "learning_rate": 1.6800087889552127e-07, + "loss": 0.6866, + "step": 8788 + }, + { + "epoch": 0.32, + "learning_rate": 1.6799721683084923e-07, + "loss": 0.698, + "step": 8789 + }, + { + "epoch": 0.32, + "learning_rate": 1.6799355476617715e-07, + "loss": 0.7019, + "step": 8790 + }, + { + "epoch": 0.32, + "learning_rate": 1.679898927015051e-07, + "loss": 0.7102, + "step": 8791 + }, + { + "epoch": 0.32, + "learning_rate": 1.6798623063683303e-07, + "loss": 0.7039, + "step": 8792 + }, + { + "epoch": 0.32, + "learning_rate": 1.6798256857216099e-07, + "loss": 0.6951, + "step": 8793 + }, + { + "epoch": 0.32, + "learning_rate": 1.679789065074889e-07, + "loss": 0.7041, + "step": 8794 + }, + { + "epoch": 0.32, + "learning_rate": 1.6797524444281687e-07, + "loss": 0.697, + "step": 8795 + }, + { + "epoch": 0.32, + "learning_rate": 1.679715823781448e-07, + "loss": 0.703, + "step": 8796 + }, + { + "epoch": 0.32, + "learning_rate": 1.6796792031347275e-07, + "loss": 0.7062, + "step": 8797 + }, + { + "epoch": 0.32, + "learning_rate": 1.6796425824880067e-07, + "loss": 0.6875, + "step": 8798 + }, + { + "epoch": 0.32, + "learning_rate": 1.679605961841286e-07, + "loss": 0.6912, + "step": 8799 + }, + { + "epoch": 0.32, + "learning_rate": 1.6795693411945655e-07, + "loss": 0.7001, + "step": 8800 + }, + { + "epoch": 0.32, + "learning_rate": 1.6795327205478448e-07, + "loss": 0.6942, + "step": 8801 + }, + { + "epoch": 0.32, + "learning_rate": 1.6794960999011243e-07, + "loss": 0.6863, + "step": 8802 + }, + { + "epoch": 0.32, + "learning_rate": 1.6794594792544036e-07, + "loss": 0.68, + "step": 8803 + }, + { + "epoch": 0.32, + "learning_rate": 1.6794228586076829e-07, + "loss": 0.6921, + "step": 8804 + }, + { + "epoch": 0.32, + "learning_rate": 1.679386237960962e-07, + "loss": 0.698, + "step": 8805 + }, + { + "epoch": 0.32, + "learning_rate": 1.6793496173142417e-07, + "loss": 0.6888, + "step": 8806 + }, + { + "epoch": 0.32, + "learning_rate": 1.679312996667521e-07, + "loss": 0.6976, + "step": 8807 + }, + { + "epoch": 0.32, + "learning_rate": 1.6792763760208005e-07, + "loss": 0.6739, + "step": 8808 + }, + { + "epoch": 0.32, + "learning_rate": 1.6792397553740797e-07, + "loss": 0.7231, + "step": 8809 + }, + { + "epoch": 0.32, + "learning_rate": 1.6792031347273593e-07, + "loss": 0.6844, + "step": 8810 + }, + { + "epoch": 0.32, + "learning_rate": 1.6791665140806385e-07, + "loss": 0.7122, + "step": 8811 + }, + { + "epoch": 0.32, + "learning_rate": 1.679129893433918e-07, + "loss": 0.7085, + "step": 8812 + }, + { + "epoch": 0.32, + "learning_rate": 1.6790932727871973e-07, + "loss": 0.7067, + "step": 8813 + }, + { + "epoch": 0.32, + "learning_rate": 1.6790566521404768e-07, + "loss": 0.6973, + "step": 8814 + }, + { + "epoch": 0.32, + "learning_rate": 1.679020031493756e-07, + "loss": 0.699, + "step": 8815 + }, + { + "epoch": 0.32, + "learning_rate": 1.6789834108470356e-07, + "loss": 0.6829, + "step": 8816 + }, + { + "epoch": 0.32, + "learning_rate": 1.678946790200315e-07, + "loss": 0.6947, + "step": 8817 + }, + { + "epoch": 0.32, + "learning_rate": 1.6789101695535944e-07, + "loss": 0.689, + "step": 8818 + }, + { + "epoch": 0.32, + "learning_rate": 1.6788735489068737e-07, + "loss": 0.6918, + "step": 8819 + }, + { + "epoch": 0.32, + "learning_rate": 1.678836928260153e-07, + "loss": 0.6806, + "step": 8820 + }, + { + "epoch": 0.32, + "learning_rate": 1.6788003076134322e-07, + "loss": 0.6866, + "step": 8821 + }, + { + "epoch": 0.32, + "learning_rate": 1.6787636869667118e-07, + "loss": 0.6813, + "step": 8822 + }, + { + "epoch": 0.32, + "learning_rate": 1.678727066319991e-07, + "loss": 0.6956, + "step": 8823 + }, + { + "epoch": 0.32, + "learning_rate": 1.6786904456732706e-07, + "loss": 0.7033, + "step": 8824 + }, + { + "epoch": 0.32, + "learning_rate": 1.6786538250265498e-07, + "loss": 0.697, + "step": 8825 + }, + { + "epoch": 0.32, + "learning_rate": 1.678617204379829e-07, + "loss": 0.6861, + "step": 8826 + }, + { + "epoch": 0.32, + "learning_rate": 1.6785805837331086e-07, + "loss": 0.6933, + "step": 8827 + }, + { + "epoch": 0.32, + "learning_rate": 1.678543963086388e-07, + "loss": 0.6916, + "step": 8828 + }, + { + "epoch": 0.32, + "learning_rate": 1.6785073424396674e-07, + "loss": 0.673, + "step": 8829 + }, + { + "epoch": 0.32, + "learning_rate": 1.6784707217929467e-07, + "loss": 0.7022, + "step": 8830 + }, + { + "epoch": 0.32, + "learning_rate": 1.6784341011462262e-07, + "loss": 0.704, + "step": 8831 + }, + { + "epoch": 0.32, + "learning_rate": 1.6783974804995055e-07, + "loss": 0.6791, + "step": 8832 + }, + { + "epoch": 0.32, + "learning_rate": 1.678360859852785e-07, + "loss": 0.6943, + "step": 8833 + }, + { + "epoch": 0.32, + "learning_rate": 1.6783242392060643e-07, + "loss": 0.6985, + "step": 8834 + }, + { + "epoch": 0.32, + "learning_rate": 1.6782876185593438e-07, + "loss": 0.6971, + "step": 8835 + }, + { + "epoch": 0.32, + "learning_rate": 1.678250997912623e-07, + "loss": 0.6882, + "step": 8836 + }, + { + "epoch": 0.32, + "learning_rate": 1.6782143772659026e-07, + "loss": 0.6784, + "step": 8837 + }, + { + "epoch": 0.32, + "learning_rate": 1.678177756619182e-07, + "loss": 0.703, + "step": 8838 + }, + { + "epoch": 0.32, + "learning_rate": 1.6781411359724612e-07, + "loss": 0.6802, + "step": 8839 + }, + { + "epoch": 0.32, + "learning_rate": 1.6781045153257404e-07, + "loss": 0.6976, + "step": 8840 + }, + { + "epoch": 0.32, + "learning_rate": 1.67806789467902e-07, + "loss": 0.6945, + "step": 8841 + }, + { + "epoch": 0.32, + "learning_rate": 1.6780312740322992e-07, + "loss": 0.6983, + "step": 8842 + }, + { + "epoch": 0.32, + "learning_rate": 1.6779946533855788e-07, + "loss": 0.7052, + "step": 8843 + }, + { + "epoch": 0.32, + "learning_rate": 1.677958032738858e-07, + "loss": 0.6889, + "step": 8844 + }, + { + "epoch": 0.32, + "learning_rate": 1.6779214120921376e-07, + "loss": 0.6809, + "step": 8845 + }, + { + "epoch": 0.32, + "learning_rate": 1.6778847914454168e-07, + "loss": 0.6949, + "step": 8846 + }, + { + "epoch": 0.32, + "learning_rate": 1.677848170798696e-07, + "loss": 0.6692, + "step": 8847 + }, + { + "epoch": 0.32, + "learning_rate": 1.6778115501519756e-07, + "loss": 0.6919, + "step": 8848 + }, + { + "epoch": 0.32, + "learning_rate": 1.677774929505255e-07, + "loss": 0.7192, + "step": 8849 + }, + { + "epoch": 0.32, + "learning_rate": 1.6777383088585344e-07, + "loss": 0.6943, + "step": 8850 + }, + { + "epoch": 0.32, + "learning_rate": 1.6777016882118137e-07, + "loss": 0.695, + "step": 8851 + }, + { + "epoch": 0.32, + "learning_rate": 1.6776650675650932e-07, + "loss": 0.71, + "step": 8852 + }, + { + "epoch": 0.32, + "learning_rate": 1.6776284469183725e-07, + "loss": 0.6902, + "step": 8853 + }, + { + "epoch": 0.32, + "learning_rate": 1.677591826271652e-07, + "loss": 0.685, + "step": 8854 + }, + { + "epoch": 0.32, + "learning_rate": 1.6775552056249313e-07, + "loss": 0.6808, + "step": 8855 + }, + { + "epoch": 0.32, + "learning_rate": 1.6775185849782108e-07, + "loss": 0.6997, + "step": 8856 + }, + { + "epoch": 0.32, + "learning_rate": 1.6774819643314898e-07, + "loss": 0.7032, + "step": 8857 + }, + { + "epoch": 0.32, + "learning_rate": 1.6774453436847694e-07, + "loss": 0.7091, + "step": 8858 + }, + { + "epoch": 0.32, + "learning_rate": 1.6774087230380486e-07, + "loss": 0.6799, + "step": 8859 + }, + { + "epoch": 0.32, + "learning_rate": 1.6773721023913282e-07, + "loss": 0.6928, + "step": 8860 + }, + { + "epoch": 0.32, + "learning_rate": 1.6773354817446074e-07, + "loss": 0.7, + "step": 8861 + }, + { + "epoch": 0.32, + "learning_rate": 1.677298861097887e-07, + "loss": 0.7014, + "step": 8862 + }, + { + "epoch": 0.32, + "learning_rate": 1.6772622404511662e-07, + "loss": 0.6878, + "step": 8863 + }, + { + "epoch": 0.32, + "learning_rate": 1.6772256198044458e-07, + "loss": 0.6894, + "step": 8864 + }, + { + "epoch": 0.32, + "learning_rate": 1.677188999157725e-07, + "loss": 0.691, + "step": 8865 + }, + { + "epoch": 0.32, + "learning_rate": 1.6771523785110046e-07, + "loss": 0.6979, + "step": 8866 + }, + { + "epoch": 0.32, + "learning_rate": 1.6771157578642838e-07, + "loss": 0.6797, + "step": 8867 + }, + { + "epoch": 0.32, + "learning_rate": 1.6770791372175633e-07, + "loss": 0.68, + "step": 8868 + }, + { + "epoch": 0.32, + "learning_rate": 1.6770425165708426e-07, + "loss": 0.7085, + "step": 8869 + }, + { + "epoch": 0.32, + "learning_rate": 1.677005895924122e-07, + "loss": 0.7081, + "step": 8870 + }, + { + "epoch": 0.32, + "learning_rate": 1.6769692752774014e-07, + "loss": 0.7195, + "step": 8871 + }, + { + "epoch": 0.32, + "learning_rate": 1.6769326546306807e-07, + "loss": 0.6882, + "step": 8872 + }, + { + "epoch": 0.32, + "learning_rate": 1.6768960339839602e-07, + "loss": 0.6764, + "step": 8873 + }, + { + "epoch": 0.32, + "learning_rate": 1.6768594133372395e-07, + "loss": 0.6901, + "step": 8874 + }, + { + "epoch": 0.32, + "learning_rate": 1.6768227926905187e-07, + "loss": 0.6863, + "step": 8875 + }, + { + "epoch": 0.32, + "learning_rate": 1.676786172043798e-07, + "loss": 0.6854, + "step": 8876 + }, + { + "epoch": 0.32, + "learning_rate": 1.6767495513970775e-07, + "loss": 0.6952, + "step": 8877 + }, + { + "epoch": 0.32, + "learning_rate": 1.6767129307503568e-07, + "loss": 0.6841, + "step": 8878 + }, + { + "epoch": 0.32, + "learning_rate": 1.6766763101036363e-07, + "loss": 0.6946, + "step": 8879 + }, + { + "epoch": 0.32, + "learning_rate": 1.6766396894569156e-07, + "loss": 0.6937, + "step": 8880 + }, + { + "epoch": 0.32, + "learning_rate": 1.6766030688101951e-07, + "loss": 0.6949, + "step": 8881 + }, + { + "epoch": 0.32, + "learning_rate": 1.6765664481634744e-07, + "loss": 0.687, + "step": 8882 + }, + { + "epoch": 0.33, + "learning_rate": 1.676529827516754e-07, + "loss": 0.6715, + "step": 8883 + }, + { + "epoch": 0.33, + "learning_rate": 1.6764932068700332e-07, + "loss": 0.69, + "step": 8884 + }, + { + "epoch": 0.33, + "learning_rate": 1.6764565862233127e-07, + "loss": 0.6868, + "step": 8885 + }, + { + "epoch": 0.33, + "learning_rate": 1.676419965576592e-07, + "loss": 0.6824, + "step": 8886 + }, + { + "epoch": 0.33, + "learning_rate": 1.6763833449298715e-07, + "loss": 0.6989, + "step": 8887 + }, + { + "epoch": 0.33, + "learning_rate": 1.6763467242831508e-07, + "loss": 0.6974, + "step": 8888 + }, + { + "epoch": 0.33, + "learning_rate": 1.6763101036364303e-07, + "loss": 0.7007, + "step": 8889 + }, + { + "epoch": 0.33, + "learning_rate": 1.6762734829897096e-07, + "loss": 0.6957, + "step": 8890 + }, + { + "epoch": 0.33, + "learning_rate": 1.676236862342989e-07, + "loss": 0.686, + "step": 8891 + }, + { + "epoch": 0.33, + "learning_rate": 1.6762002416962684e-07, + "loss": 0.6871, + "step": 8892 + }, + { + "epoch": 0.33, + "learning_rate": 1.6761636210495477e-07, + "loss": 0.6948, + "step": 8893 + }, + { + "epoch": 0.33, + "learning_rate": 1.676127000402827e-07, + "loss": 0.679, + "step": 8894 + }, + { + "epoch": 0.33, + "learning_rate": 1.6760903797561065e-07, + "loss": 0.697, + "step": 8895 + }, + { + "epoch": 0.33, + "learning_rate": 1.6760537591093857e-07, + "loss": 0.6899, + "step": 8896 + }, + { + "epoch": 0.33, + "learning_rate": 1.676017138462665e-07, + "loss": 0.6878, + "step": 8897 + }, + { + "epoch": 0.33, + "learning_rate": 1.6759805178159445e-07, + "loss": 0.7037, + "step": 8898 + }, + { + "epoch": 0.33, + "learning_rate": 1.6759438971692238e-07, + "loss": 0.7091, + "step": 8899 + }, + { + "epoch": 0.33, + "learning_rate": 1.6759072765225033e-07, + "loss": 0.6976, + "step": 8900 + }, + { + "epoch": 0.33, + "learning_rate": 1.6758706558757826e-07, + "loss": 0.6949, + "step": 8901 + }, + { + "epoch": 0.33, + "learning_rate": 1.675834035229062e-07, + "loss": 0.6882, + "step": 8902 + }, + { + "epoch": 0.33, + "learning_rate": 1.6757974145823414e-07, + "loss": 0.6831, + "step": 8903 + }, + { + "epoch": 0.33, + "learning_rate": 1.675760793935621e-07, + "loss": 0.6968, + "step": 8904 + }, + { + "epoch": 0.33, + "learning_rate": 1.6757241732889002e-07, + "loss": 0.686, + "step": 8905 + }, + { + "epoch": 0.33, + "learning_rate": 1.6756875526421797e-07, + "loss": 0.6846, + "step": 8906 + }, + { + "epoch": 0.33, + "learning_rate": 1.675650931995459e-07, + "loss": 0.7034, + "step": 8907 + }, + { + "epoch": 0.33, + "learning_rate": 1.6756143113487385e-07, + "loss": 0.6784, + "step": 8908 + }, + { + "epoch": 0.33, + "learning_rate": 1.6755776907020178e-07, + "loss": 0.6806, + "step": 8909 + }, + { + "epoch": 0.33, + "learning_rate": 1.675541070055297e-07, + "loss": 0.6918, + "step": 8910 + }, + { + "epoch": 0.33, + "learning_rate": 1.6755044494085763e-07, + "loss": 0.6954, + "step": 8911 + }, + { + "epoch": 0.33, + "learning_rate": 1.6754678287618559e-07, + "loss": 0.6942, + "step": 8912 + }, + { + "epoch": 0.33, + "learning_rate": 1.675431208115135e-07, + "loss": 0.6988, + "step": 8913 + }, + { + "epoch": 0.33, + "learning_rate": 1.6753945874684147e-07, + "loss": 0.7001, + "step": 8914 + }, + { + "epoch": 0.33, + "learning_rate": 1.675357966821694e-07, + "loss": 0.6955, + "step": 8915 + }, + { + "epoch": 0.33, + "learning_rate": 1.6753213461749735e-07, + "loss": 0.6806, + "step": 8916 + }, + { + "epoch": 0.33, + "learning_rate": 1.6752847255282527e-07, + "loss": 0.7033, + "step": 8917 + }, + { + "epoch": 0.33, + "learning_rate": 1.675248104881532e-07, + "loss": 0.6919, + "step": 8918 + }, + { + "epoch": 0.33, + "learning_rate": 1.6752114842348115e-07, + "loss": 0.6986, + "step": 8919 + }, + { + "epoch": 0.33, + "learning_rate": 1.6751748635880908e-07, + "loss": 0.6876, + "step": 8920 + }, + { + "epoch": 0.33, + "learning_rate": 1.6751382429413703e-07, + "loss": 0.7015, + "step": 8921 + }, + { + "epoch": 0.33, + "learning_rate": 1.6751016222946496e-07, + "loss": 0.6719, + "step": 8922 + }, + { + "epoch": 0.33, + "learning_rate": 1.675065001647929e-07, + "loss": 0.6799, + "step": 8923 + }, + { + "epoch": 0.33, + "learning_rate": 1.6750283810012084e-07, + "loss": 0.6877, + "step": 8924 + }, + { + "epoch": 0.33, + "learning_rate": 1.674991760354488e-07, + "loss": 0.7019, + "step": 8925 + }, + { + "epoch": 0.33, + "learning_rate": 1.6749551397077672e-07, + "loss": 0.6964, + "step": 8926 + }, + { + "epoch": 0.33, + "learning_rate": 1.6749185190610467e-07, + "loss": 0.6986, + "step": 8927 + }, + { + "epoch": 0.33, + "learning_rate": 1.6748818984143257e-07, + "loss": 0.676, + "step": 8928 + }, + { + "epoch": 0.33, + "learning_rate": 1.6748452777676053e-07, + "loss": 0.6875, + "step": 8929 + }, + { + "epoch": 0.33, + "learning_rate": 1.6748086571208845e-07, + "loss": 0.6872, + "step": 8930 + }, + { + "epoch": 0.33, + "learning_rate": 1.674772036474164e-07, + "loss": 0.6819, + "step": 8931 + }, + { + "epoch": 0.33, + "learning_rate": 1.6747354158274433e-07, + "loss": 0.6943, + "step": 8932 + }, + { + "epoch": 0.33, + "learning_rate": 1.6746987951807228e-07, + "loss": 0.7056, + "step": 8933 + }, + { + "epoch": 0.33, + "learning_rate": 1.674662174534002e-07, + "loss": 0.6983, + "step": 8934 + }, + { + "epoch": 0.33, + "learning_rate": 1.6746255538872816e-07, + "loss": 0.6672, + "step": 8935 + }, + { + "epoch": 0.33, + "learning_rate": 1.674588933240561e-07, + "loss": 0.6966, + "step": 8936 + }, + { + "epoch": 0.33, + "learning_rate": 1.6745523125938404e-07, + "loss": 0.704, + "step": 8937 + }, + { + "epoch": 0.33, + "learning_rate": 1.6745156919471197e-07, + "loss": 0.7081, + "step": 8938 + }, + { + "epoch": 0.33, + "learning_rate": 1.6744790713003992e-07, + "loss": 0.7206, + "step": 8939 + }, + { + "epoch": 0.33, + "learning_rate": 1.6744424506536785e-07, + "loss": 0.7023, + "step": 8940 + }, + { + "epoch": 0.33, + "learning_rate": 1.6744058300069578e-07, + "loss": 0.6817, + "step": 8941 + }, + { + "epoch": 0.33, + "learning_rate": 1.6743692093602373e-07, + "loss": 0.7107, + "step": 8942 + }, + { + "epoch": 0.33, + "learning_rate": 1.6743325887135166e-07, + "loss": 0.7125, + "step": 8943 + }, + { + "epoch": 0.33, + "learning_rate": 1.674295968066796e-07, + "loss": 0.6962, + "step": 8944 + }, + { + "epoch": 0.33, + "learning_rate": 1.6742593474200754e-07, + "loss": 0.7081, + "step": 8945 + }, + { + "epoch": 0.33, + "learning_rate": 1.6742227267733546e-07, + "loss": 0.702, + "step": 8946 + }, + { + "epoch": 0.33, + "learning_rate": 1.674186106126634e-07, + "loss": 0.7043, + "step": 8947 + }, + { + "epoch": 0.33, + "learning_rate": 1.6741494854799134e-07, + "loss": 0.7043, + "step": 8948 + }, + { + "epoch": 0.33, + "learning_rate": 1.6741128648331927e-07, + "loss": 0.6966, + "step": 8949 + }, + { + "epoch": 0.33, + "learning_rate": 1.6740762441864722e-07, + "loss": 0.6987, + "step": 8950 + }, + { + "epoch": 0.33, + "learning_rate": 1.6740396235397515e-07, + "loss": 0.6945, + "step": 8951 + }, + { + "epoch": 0.33, + "learning_rate": 1.674003002893031e-07, + "loss": 0.6918, + "step": 8952 + }, + { + "epoch": 0.33, + "learning_rate": 1.6739663822463103e-07, + "loss": 0.7022, + "step": 8953 + }, + { + "epoch": 0.33, + "learning_rate": 1.6739297615995898e-07, + "loss": 0.6895, + "step": 8954 + }, + { + "epoch": 0.33, + "learning_rate": 1.673893140952869e-07, + "loss": 0.6938, + "step": 8955 + }, + { + "epoch": 0.33, + "learning_rate": 1.6738565203061486e-07, + "loss": 0.6881, + "step": 8956 + }, + { + "epoch": 0.33, + "learning_rate": 1.673819899659428e-07, + "loss": 0.6752, + "step": 8957 + }, + { + "epoch": 0.33, + "learning_rate": 1.6737832790127074e-07, + "loss": 0.6994, + "step": 8958 + }, + { + "epoch": 0.33, + "learning_rate": 1.6737466583659867e-07, + "loss": 0.6857, + "step": 8959 + }, + { + "epoch": 0.33, + "learning_rate": 1.6737100377192662e-07, + "loss": 0.6848, + "step": 8960 + }, + { + "epoch": 0.33, + "learning_rate": 1.6736734170725455e-07, + "loss": 0.7086, + "step": 8961 + }, + { + "epoch": 0.33, + "learning_rate": 1.6736367964258248e-07, + "loss": 0.6958, + "step": 8962 + }, + { + "epoch": 0.33, + "learning_rate": 1.6736001757791043e-07, + "loss": 0.7044, + "step": 8963 + }, + { + "epoch": 0.33, + "learning_rate": 1.6735635551323836e-07, + "loss": 0.6788, + "step": 8964 + }, + { + "epoch": 0.33, + "learning_rate": 1.6735269344856628e-07, + "loss": 0.7159, + "step": 8965 + }, + { + "epoch": 0.33, + "learning_rate": 1.6734903138389424e-07, + "loss": 0.7069, + "step": 8966 + }, + { + "epoch": 0.33, + "learning_rate": 1.6734536931922216e-07, + "loss": 0.7095, + "step": 8967 + }, + { + "epoch": 0.33, + "learning_rate": 1.673417072545501e-07, + "loss": 0.6947, + "step": 8968 + }, + { + "epoch": 0.33, + "learning_rate": 1.6733804518987804e-07, + "loss": 0.6831, + "step": 8969 + }, + { + "epoch": 0.33, + "learning_rate": 1.6733438312520597e-07, + "loss": 0.6797, + "step": 8970 + }, + { + "epoch": 0.33, + "learning_rate": 1.6733072106053392e-07, + "loss": 0.684, + "step": 8971 + }, + { + "epoch": 0.33, + "learning_rate": 1.6732705899586185e-07, + "loss": 0.6948, + "step": 8972 + }, + { + "epoch": 0.33, + "learning_rate": 1.673233969311898e-07, + "loss": 0.6934, + "step": 8973 + }, + { + "epoch": 0.33, + "learning_rate": 1.6731973486651773e-07, + "loss": 0.6901, + "step": 8974 + }, + { + "epoch": 0.33, + "learning_rate": 1.6731607280184568e-07, + "loss": 0.6878, + "step": 8975 + }, + { + "epoch": 0.33, + "learning_rate": 1.673124107371736e-07, + "loss": 0.6993, + "step": 8976 + }, + { + "epoch": 0.33, + "learning_rate": 1.6730874867250156e-07, + "loss": 0.7004, + "step": 8977 + }, + { + "epoch": 0.33, + "learning_rate": 1.673050866078295e-07, + "loss": 0.6911, + "step": 8978 + }, + { + "epoch": 0.33, + "learning_rate": 1.6730142454315744e-07, + "loss": 0.6914, + "step": 8979 + }, + { + "epoch": 0.33, + "learning_rate": 1.6729776247848537e-07, + "loss": 0.6892, + "step": 8980 + }, + { + "epoch": 0.33, + "learning_rate": 1.6729410041381332e-07, + "loss": 0.6865, + "step": 8981 + }, + { + "epoch": 0.33, + "learning_rate": 1.6729043834914122e-07, + "loss": 0.7065, + "step": 8982 + }, + { + "epoch": 0.33, + "learning_rate": 1.6728677628446918e-07, + "loss": 0.7194, + "step": 8983 + }, + { + "epoch": 0.33, + "learning_rate": 1.672831142197971e-07, + "loss": 0.6932, + "step": 8984 + }, + { + "epoch": 0.33, + "learning_rate": 1.6727945215512506e-07, + "loss": 0.6992, + "step": 8985 + }, + { + "epoch": 0.33, + "learning_rate": 1.6727579009045298e-07, + "loss": 0.6915, + "step": 8986 + }, + { + "epoch": 0.33, + "learning_rate": 1.6727212802578094e-07, + "loss": 0.6834, + "step": 8987 + }, + { + "epoch": 0.33, + "learning_rate": 1.6726846596110886e-07, + "loss": 0.6969, + "step": 8988 + }, + { + "epoch": 0.33, + "learning_rate": 1.672648038964368e-07, + "loss": 0.7159, + "step": 8989 + }, + { + "epoch": 0.33, + "learning_rate": 1.6726114183176474e-07, + "loss": 0.6847, + "step": 8990 + }, + { + "epoch": 0.33, + "learning_rate": 1.6725747976709267e-07, + "loss": 0.6938, + "step": 8991 + }, + { + "epoch": 0.33, + "learning_rate": 1.6725381770242062e-07, + "loss": 0.6915, + "step": 8992 + }, + { + "epoch": 0.33, + "learning_rate": 1.6725015563774855e-07, + "loss": 0.6918, + "step": 8993 + }, + { + "epoch": 0.33, + "learning_rate": 1.672464935730765e-07, + "loss": 0.6906, + "step": 8994 + }, + { + "epoch": 0.33, + "learning_rate": 1.6724283150840443e-07, + "loss": 0.7077, + "step": 8995 + }, + { + "epoch": 0.33, + "learning_rate": 1.6723916944373238e-07, + "loss": 0.6904, + "step": 8996 + }, + { + "epoch": 0.33, + "learning_rate": 1.672355073790603e-07, + "loss": 0.6772, + "step": 8997 + }, + { + "epoch": 0.33, + "learning_rate": 1.6723184531438826e-07, + "loss": 0.6806, + "step": 8998 + }, + { + "epoch": 0.33, + "learning_rate": 1.672281832497162e-07, + "loss": 0.6887, + "step": 8999 + }, + { + "epoch": 0.33, + "learning_rate": 1.6722452118504411e-07, + "loss": 0.6813, + "step": 9000 + }, + { + "epoch": 0.33, + "learning_rate": 1.6722085912037204e-07, + "loss": 0.6912, + "step": 9001 + }, + { + "epoch": 0.33, + "learning_rate": 1.672171970557e-07, + "loss": 0.7141, + "step": 9002 + }, + { + "epoch": 0.33, + "learning_rate": 1.6721353499102792e-07, + "loss": 0.6837, + "step": 9003 + }, + { + "epoch": 0.33, + "learning_rate": 1.6720987292635587e-07, + "loss": 0.6924, + "step": 9004 + }, + { + "epoch": 0.33, + "learning_rate": 1.672062108616838e-07, + "loss": 0.6976, + "step": 9005 + }, + { + "epoch": 0.33, + "learning_rate": 1.6720254879701175e-07, + "loss": 0.68, + "step": 9006 + }, + { + "epoch": 0.33, + "learning_rate": 1.6719888673233968e-07, + "loss": 0.6982, + "step": 9007 + }, + { + "epoch": 0.33, + "learning_rate": 1.6719522466766763e-07, + "loss": 0.6934, + "step": 9008 + }, + { + "epoch": 0.33, + "learning_rate": 1.6719156260299556e-07, + "loss": 0.7024, + "step": 9009 + }, + { + "epoch": 0.33, + "learning_rate": 1.6718790053832351e-07, + "loss": 0.6869, + "step": 9010 + }, + { + "epoch": 0.33, + "learning_rate": 1.6718423847365144e-07, + "loss": 0.6897, + "step": 9011 + }, + { + "epoch": 0.33, + "learning_rate": 1.6718057640897937e-07, + "loss": 0.7088, + "step": 9012 + }, + { + "epoch": 0.33, + "learning_rate": 1.6717691434430732e-07, + "loss": 0.6902, + "step": 9013 + }, + { + "epoch": 0.33, + "learning_rate": 1.6717325227963525e-07, + "loss": 0.697, + "step": 9014 + }, + { + "epoch": 0.33, + "learning_rate": 1.671695902149632e-07, + "loss": 0.6958, + "step": 9015 + }, + { + "epoch": 0.33, + "learning_rate": 1.6716592815029113e-07, + "loss": 0.6758, + "step": 9016 + }, + { + "epoch": 0.33, + "learning_rate": 1.6716226608561908e-07, + "loss": 0.6798, + "step": 9017 + }, + { + "epoch": 0.33, + "learning_rate": 1.6715860402094698e-07, + "loss": 0.6889, + "step": 9018 + }, + { + "epoch": 0.33, + "learning_rate": 1.6715494195627493e-07, + "loss": 0.6969, + "step": 9019 + }, + { + "epoch": 0.33, + "learning_rate": 1.6715127989160286e-07, + "loss": 0.7195, + "step": 9020 + }, + { + "epoch": 0.33, + "learning_rate": 1.671476178269308e-07, + "loss": 0.6894, + "step": 9021 + }, + { + "epoch": 0.33, + "learning_rate": 1.6714395576225874e-07, + "loss": 0.6965, + "step": 9022 + }, + { + "epoch": 0.33, + "learning_rate": 1.671402936975867e-07, + "loss": 0.693, + "step": 9023 + }, + { + "epoch": 0.33, + "learning_rate": 1.6713663163291462e-07, + "loss": 0.6866, + "step": 9024 + }, + { + "epoch": 0.33, + "learning_rate": 1.6713296956824257e-07, + "loss": 0.6886, + "step": 9025 + }, + { + "epoch": 0.33, + "learning_rate": 1.671293075035705e-07, + "loss": 0.7, + "step": 9026 + }, + { + "epoch": 0.33, + "learning_rate": 1.6712564543889845e-07, + "loss": 0.7013, + "step": 9027 + }, + { + "epoch": 0.33, + "learning_rate": 1.6712198337422638e-07, + "loss": 0.695, + "step": 9028 + }, + { + "epoch": 0.33, + "learning_rate": 1.6711832130955433e-07, + "loss": 0.6965, + "step": 9029 + }, + { + "epoch": 0.33, + "learning_rate": 1.6711465924488226e-07, + "loss": 0.6914, + "step": 9030 + }, + { + "epoch": 0.33, + "learning_rate": 1.671109971802102e-07, + "loss": 0.6918, + "step": 9031 + }, + { + "epoch": 0.33, + "learning_rate": 1.6710733511553814e-07, + "loss": 0.6991, + "step": 9032 + }, + { + "epoch": 0.33, + "learning_rate": 1.6710367305086607e-07, + "loss": 0.7041, + "step": 9033 + }, + { + "epoch": 0.33, + "learning_rate": 1.6710001098619402e-07, + "loss": 0.6955, + "step": 9034 + }, + { + "epoch": 0.33, + "learning_rate": 1.6709634892152195e-07, + "loss": 0.6965, + "step": 9035 + }, + { + "epoch": 0.33, + "learning_rate": 1.6709268685684987e-07, + "loss": 0.6954, + "step": 9036 + }, + { + "epoch": 0.33, + "learning_rate": 1.6708902479217783e-07, + "loss": 0.6914, + "step": 9037 + }, + { + "epoch": 0.33, + "learning_rate": 1.6708536272750575e-07, + "loss": 0.7098, + "step": 9038 + }, + { + "epoch": 0.33, + "learning_rate": 1.6708170066283368e-07, + "loss": 0.7106, + "step": 9039 + }, + { + "epoch": 0.33, + "learning_rate": 1.6707803859816163e-07, + "loss": 0.6897, + "step": 9040 + }, + { + "epoch": 0.33, + "learning_rate": 1.6707437653348956e-07, + "loss": 0.6932, + "step": 9041 + }, + { + "epoch": 0.33, + "learning_rate": 1.670707144688175e-07, + "loss": 0.6741, + "step": 9042 + }, + { + "epoch": 0.33, + "learning_rate": 1.6706705240414544e-07, + "loss": 0.6745, + "step": 9043 + }, + { + "epoch": 0.33, + "learning_rate": 1.670633903394734e-07, + "loss": 0.6804, + "step": 9044 + }, + { + "epoch": 0.33, + "learning_rate": 1.6705972827480132e-07, + "loss": 0.6931, + "step": 9045 + }, + { + "epoch": 0.33, + "learning_rate": 1.6705606621012927e-07, + "loss": 0.6904, + "step": 9046 + }, + { + "epoch": 0.33, + "learning_rate": 1.670524041454572e-07, + "loss": 0.6983, + "step": 9047 + }, + { + "epoch": 0.33, + "learning_rate": 1.6704874208078515e-07, + "loss": 0.6881, + "step": 9048 + }, + { + "epoch": 0.33, + "learning_rate": 1.6704508001611308e-07, + "loss": 0.6896, + "step": 9049 + }, + { + "epoch": 0.33, + "learning_rate": 1.6704141795144103e-07, + "loss": 0.6838, + "step": 9050 + }, + { + "epoch": 0.33, + "learning_rate": 1.6703775588676896e-07, + "loss": 0.7055, + "step": 9051 + }, + { + "epoch": 0.33, + "learning_rate": 1.670340938220969e-07, + "loss": 0.6918, + "step": 9052 + }, + { + "epoch": 0.33, + "learning_rate": 1.6703043175742484e-07, + "loss": 0.6896, + "step": 9053 + }, + { + "epoch": 0.33, + "learning_rate": 1.6702676969275276e-07, + "loss": 0.6985, + "step": 9054 + }, + { + "epoch": 0.33, + "learning_rate": 1.670231076280807e-07, + "loss": 0.6955, + "step": 9055 + }, + { + "epoch": 0.33, + "learning_rate": 1.6701944556340864e-07, + "loss": 0.6895, + "step": 9056 + }, + { + "epoch": 0.33, + "learning_rate": 1.6701578349873657e-07, + "loss": 0.6913, + "step": 9057 + }, + { + "epoch": 0.33, + "learning_rate": 1.6701212143406452e-07, + "loss": 0.7037, + "step": 9058 + }, + { + "epoch": 0.33, + "learning_rate": 1.6700845936939245e-07, + "loss": 0.6965, + "step": 9059 + }, + { + "epoch": 0.33, + "learning_rate": 1.6700479730472038e-07, + "loss": 0.6946, + "step": 9060 + }, + { + "epoch": 0.33, + "learning_rate": 1.6700113524004833e-07, + "loss": 0.691, + "step": 9061 + }, + { + "epoch": 0.33, + "learning_rate": 1.6699747317537626e-07, + "loss": 0.6991, + "step": 9062 + }, + { + "epoch": 0.33, + "learning_rate": 1.669938111107042e-07, + "loss": 0.7004, + "step": 9063 + }, + { + "epoch": 0.33, + "learning_rate": 1.6699014904603214e-07, + "loss": 0.6892, + "step": 9064 + }, + { + "epoch": 0.33, + "learning_rate": 1.669864869813601e-07, + "loss": 0.6908, + "step": 9065 + }, + { + "epoch": 0.33, + "learning_rate": 1.6698282491668802e-07, + "loss": 0.6895, + "step": 9066 + }, + { + "epoch": 0.33, + "learning_rate": 1.6697916285201597e-07, + "loss": 0.6834, + "step": 9067 + }, + { + "epoch": 0.33, + "learning_rate": 1.669755007873439e-07, + "loss": 0.6825, + "step": 9068 + }, + { + "epoch": 0.33, + "learning_rate": 1.6697183872267185e-07, + "loss": 0.6865, + "step": 9069 + }, + { + "epoch": 0.33, + "learning_rate": 1.6696817665799978e-07, + "loss": 0.6821, + "step": 9070 + }, + { + "epoch": 0.33, + "learning_rate": 1.6696451459332773e-07, + "loss": 0.6958, + "step": 9071 + }, + { + "epoch": 0.33, + "learning_rate": 1.6696085252865563e-07, + "loss": 0.6986, + "step": 9072 + }, + { + "epoch": 0.33, + "learning_rate": 1.6695719046398358e-07, + "loss": 0.6975, + "step": 9073 + }, + { + "epoch": 0.33, + "learning_rate": 1.669535283993115e-07, + "loss": 0.6765, + "step": 9074 + }, + { + "epoch": 0.33, + "learning_rate": 1.6694986633463946e-07, + "loss": 0.6916, + "step": 9075 + }, + { + "epoch": 0.33, + "learning_rate": 1.669462042699674e-07, + "loss": 0.6803, + "step": 9076 + }, + { + "epoch": 0.33, + "learning_rate": 1.6694254220529534e-07, + "loss": 0.6975, + "step": 9077 + }, + { + "epoch": 0.33, + "learning_rate": 1.6693888014062327e-07, + "loss": 0.6956, + "step": 9078 + }, + { + "epoch": 0.33, + "learning_rate": 1.6693521807595122e-07, + "loss": 0.6806, + "step": 9079 + }, + { + "epoch": 0.33, + "learning_rate": 1.6693155601127915e-07, + "loss": 0.6903, + "step": 9080 + }, + { + "epoch": 0.33, + "learning_rate": 1.669278939466071e-07, + "loss": 0.6924, + "step": 9081 + }, + { + "epoch": 0.33, + "learning_rate": 1.6692423188193503e-07, + "loss": 0.701, + "step": 9082 + }, + { + "epoch": 0.33, + "learning_rate": 1.6692056981726296e-07, + "loss": 0.6897, + "step": 9083 + }, + { + "epoch": 0.33, + "learning_rate": 1.669169077525909e-07, + "loss": 0.6872, + "step": 9084 + }, + { + "epoch": 0.33, + "learning_rate": 1.6691324568791884e-07, + "loss": 0.6976, + "step": 9085 + }, + { + "epoch": 0.33, + "learning_rate": 1.669095836232468e-07, + "loss": 0.6895, + "step": 9086 + }, + { + "epoch": 0.33, + "learning_rate": 1.6690592155857472e-07, + "loss": 0.689, + "step": 9087 + }, + { + "epoch": 0.33, + "learning_rate": 1.6690225949390267e-07, + "loss": 0.7071, + "step": 9088 + }, + { + "epoch": 0.33, + "learning_rate": 1.668985974292306e-07, + "loss": 0.6964, + "step": 9089 + }, + { + "epoch": 0.33, + "learning_rate": 1.6689493536455852e-07, + "loss": 0.6817, + "step": 9090 + }, + { + "epoch": 0.33, + "learning_rate": 1.6689127329988645e-07, + "loss": 0.6917, + "step": 9091 + }, + { + "epoch": 0.33, + "learning_rate": 1.668876112352144e-07, + "loss": 0.6775, + "step": 9092 + }, + { + "epoch": 0.33, + "learning_rate": 1.6688394917054233e-07, + "loss": 0.6779, + "step": 9093 + }, + { + "epoch": 0.33, + "learning_rate": 1.6688028710587028e-07, + "loss": 0.6988, + "step": 9094 + }, + { + "epoch": 0.33, + "learning_rate": 1.668766250411982e-07, + "loss": 0.6809, + "step": 9095 + }, + { + "epoch": 0.33, + "learning_rate": 1.6687296297652616e-07, + "loss": 0.6987, + "step": 9096 + }, + { + "epoch": 0.33, + "learning_rate": 1.668693009118541e-07, + "loss": 0.7, + "step": 9097 + }, + { + "epoch": 0.33, + "learning_rate": 1.6686563884718204e-07, + "loss": 0.6978, + "step": 9098 + }, + { + "epoch": 0.33, + "learning_rate": 1.6686197678250997e-07, + "loss": 0.6772, + "step": 9099 + }, + { + "epoch": 0.33, + "learning_rate": 1.6685831471783792e-07, + "loss": 0.6991, + "step": 9100 + }, + { + "epoch": 0.33, + "learning_rate": 1.6685465265316585e-07, + "loss": 0.6952, + "step": 9101 + }, + { + "epoch": 0.33, + "learning_rate": 1.668509905884938e-07, + "loss": 0.6995, + "step": 9102 + }, + { + "epoch": 0.33, + "learning_rate": 1.6684732852382173e-07, + "loss": 0.7066, + "step": 9103 + }, + { + "epoch": 0.33, + "learning_rate": 1.6684366645914966e-07, + "loss": 0.7052, + "step": 9104 + }, + { + "epoch": 0.33, + "learning_rate": 1.668400043944776e-07, + "loss": 0.7053, + "step": 9105 + }, + { + "epoch": 0.33, + "learning_rate": 1.6683634232980554e-07, + "loss": 0.7235, + "step": 9106 + }, + { + "epoch": 0.33, + "learning_rate": 1.668326802651335e-07, + "loss": 0.6971, + "step": 9107 + }, + { + "epoch": 0.33, + "learning_rate": 1.6682901820046141e-07, + "loss": 0.7024, + "step": 9108 + }, + { + "epoch": 0.33, + "learning_rate": 1.6682535613578934e-07, + "loss": 0.6948, + "step": 9109 + }, + { + "epoch": 0.33, + "learning_rate": 1.6682169407111727e-07, + "loss": 0.6988, + "step": 9110 + }, + { + "epoch": 0.33, + "learning_rate": 1.6681803200644522e-07, + "loss": 0.7025, + "step": 9111 + }, + { + "epoch": 0.33, + "learning_rate": 1.6681436994177315e-07, + "loss": 0.6876, + "step": 9112 + }, + { + "epoch": 0.33, + "learning_rate": 1.668107078771011e-07, + "loss": 0.6955, + "step": 9113 + }, + { + "epoch": 0.33, + "learning_rate": 1.6680704581242903e-07, + "loss": 0.6921, + "step": 9114 + }, + { + "epoch": 0.33, + "learning_rate": 1.6680338374775698e-07, + "loss": 0.6996, + "step": 9115 + }, + { + "epoch": 0.33, + "learning_rate": 1.667997216830849e-07, + "loss": 0.6785, + "step": 9116 + }, + { + "epoch": 0.33, + "learning_rate": 1.6679605961841286e-07, + "loss": 0.6906, + "step": 9117 + }, + { + "epoch": 0.33, + "learning_rate": 1.667923975537408e-07, + "loss": 0.693, + "step": 9118 + }, + { + "epoch": 0.33, + "learning_rate": 1.6678873548906874e-07, + "loss": 0.6901, + "step": 9119 + }, + { + "epoch": 0.33, + "learning_rate": 1.6678507342439667e-07, + "loss": 0.707, + "step": 9120 + }, + { + "epoch": 0.33, + "learning_rate": 1.6678141135972462e-07, + "loss": 0.691, + "step": 9121 + }, + { + "epoch": 0.33, + "learning_rate": 1.6677774929505255e-07, + "loss": 0.6925, + "step": 9122 + }, + { + "epoch": 0.33, + "learning_rate": 1.667740872303805e-07, + "loss": 0.6936, + "step": 9123 + }, + { + "epoch": 0.33, + "learning_rate": 1.6677042516570843e-07, + "loss": 0.7114, + "step": 9124 + }, + { + "epoch": 0.33, + "learning_rate": 1.6676676310103638e-07, + "loss": 0.6907, + "step": 9125 + }, + { + "epoch": 0.33, + "learning_rate": 1.6676310103636428e-07, + "loss": 0.6708, + "step": 9126 + }, + { + "epoch": 0.33, + "learning_rate": 1.6675943897169223e-07, + "loss": 0.6873, + "step": 9127 + }, + { + "epoch": 0.33, + "learning_rate": 1.6675577690702016e-07, + "loss": 0.6894, + "step": 9128 + }, + { + "epoch": 0.33, + "learning_rate": 1.6675211484234811e-07, + "loss": 0.6931, + "step": 9129 + }, + { + "epoch": 0.33, + "learning_rate": 1.6674845277767604e-07, + "loss": 0.6962, + "step": 9130 + }, + { + "epoch": 0.33, + "learning_rate": 1.6674479071300397e-07, + "loss": 0.6706, + "step": 9131 + }, + { + "epoch": 0.33, + "learning_rate": 1.6674112864833192e-07, + "loss": 0.671, + "step": 9132 + }, + { + "epoch": 0.33, + "learning_rate": 1.6673746658365985e-07, + "loss": 0.7013, + "step": 9133 + }, + { + "epoch": 0.33, + "learning_rate": 1.667338045189878e-07, + "loss": 0.683, + "step": 9134 + }, + { + "epoch": 0.33, + "learning_rate": 1.6673014245431573e-07, + "loss": 0.6887, + "step": 9135 + }, + { + "epoch": 0.33, + "learning_rate": 1.6672648038964368e-07, + "loss": 0.6982, + "step": 9136 + }, + { + "epoch": 0.33, + "learning_rate": 1.667228183249716e-07, + "loss": 0.6854, + "step": 9137 + }, + { + "epoch": 0.33, + "learning_rate": 1.6671915626029956e-07, + "loss": 0.7012, + "step": 9138 + }, + { + "epoch": 0.33, + "learning_rate": 1.6671549419562749e-07, + "loss": 0.6818, + "step": 9139 + }, + { + "epoch": 0.33, + "learning_rate": 1.6671183213095544e-07, + "loss": 0.6913, + "step": 9140 + }, + { + "epoch": 0.33, + "learning_rate": 1.6670817006628337e-07, + "loss": 0.6953, + "step": 9141 + }, + { + "epoch": 0.33, + "learning_rate": 1.6670450800161132e-07, + "loss": 0.6818, + "step": 9142 + }, + { + "epoch": 0.33, + "learning_rate": 1.6670084593693925e-07, + "loss": 0.6958, + "step": 9143 + }, + { + "epoch": 0.33, + "learning_rate": 1.6669718387226717e-07, + "loss": 0.7106, + "step": 9144 + }, + { + "epoch": 0.33, + "learning_rate": 1.666935218075951e-07, + "loss": 0.7025, + "step": 9145 + }, + { + "epoch": 0.33, + "learning_rate": 1.6668985974292305e-07, + "loss": 0.6948, + "step": 9146 + }, + { + "epoch": 0.33, + "learning_rate": 1.6668619767825098e-07, + "loss": 0.6999, + "step": 9147 + }, + { + "epoch": 0.33, + "learning_rate": 1.6668253561357893e-07, + "loss": 0.6952, + "step": 9148 + }, + { + "epoch": 0.33, + "learning_rate": 1.6667887354890686e-07, + "loss": 0.6893, + "step": 9149 + }, + { + "epoch": 0.33, + "learning_rate": 1.666752114842348e-07, + "loss": 0.6883, + "step": 9150 + }, + { + "epoch": 0.33, + "learning_rate": 1.6667154941956274e-07, + "loss": 0.6963, + "step": 9151 + }, + { + "epoch": 0.33, + "learning_rate": 1.666678873548907e-07, + "loss": 0.7028, + "step": 9152 + }, + { + "epoch": 0.33, + "learning_rate": 1.6666422529021862e-07, + "loss": 0.6967, + "step": 9153 + }, + { + "epoch": 0.33, + "learning_rate": 1.6666056322554655e-07, + "loss": 0.6909, + "step": 9154 + }, + { + "epoch": 0.33, + "learning_rate": 1.666569011608745e-07, + "loss": 0.6919, + "step": 9155 + }, + { + "epoch": 0.33, + "learning_rate": 1.6665323909620243e-07, + "loss": 0.6942, + "step": 9156 + }, + { + "epoch": 0.34, + "learning_rate": 1.6664957703153038e-07, + "loss": 0.6891, + "step": 9157 + }, + { + "epoch": 0.34, + "learning_rate": 1.666459149668583e-07, + "loss": 0.6997, + "step": 9158 + }, + { + "epoch": 0.34, + "learning_rate": 1.6664225290218626e-07, + "loss": 0.696, + "step": 9159 + }, + { + "epoch": 0.34, + "learning_rate": 1.6663859083751419e-07, + "loss": 0.6856, + "step": 9160 + }, + { + "epoch": 0.34, + "learning_rate": 1.6663492877284214e-07, + "loss": 0.692, + "step": 9161 + }, + { + "epoch": 0.34, + "learning_rate": 1.6663126670817004e-07, + "loss": 0.6944, + "step": 9162 + }, + { + "epoch": 0.34, + "learning_rate": 1.66627604643498e-07, + "loss": 0.6903, + "step": 9163 + }, + { + "epoch": 0.34, + "learning_rate": 1.6662394257882592e-07, + "loss": 0.6978, + "step": 9164 + }, + { + "epoch": 0.34, + "learning_rate": 1.6662028051415387e-07, + "loss": 0.6816, + "step": 9165 + }, + { + "epoch": 0.34, + "learning_rate": 1.666166184494818e-07, + "loss": 0.6961, + "step": 9166 + }, + { + "epoch": 0.34, + "learning_rate": 1.6661295638480975e-07, + "loss": 0.6846, + "step": 9167 + }, + { + "epoch": 0.34, + "learning_rate": 1.6660929432013768e-07, + "loss": 0.6859, + "step": 9168 + }, + { + "epoch": 0.34, + "learning_rate": 1.6660563225546563e-07, + "loss": 0.6864, + "step": 9169 + }, + { + "epoch": 0.34, + "learning_rate": 1.6660197019079356e-07, + "loss": 0.6913, + "step": 9170 + }, + { + "epoch": 0.34, + "learning_rate": 1.665983081261215e-07, + "loss": 0.6951, + "step": 9171 + }, + { + "epoch": 0.34, + "learning_rate": 1.6659464606144944e-07, + "loss": 0.698, + "step": 9172 + }, + { + "epoch": 0.34, + "learning_rate": 1.665909839967774e-07, + "loss": 0.6841, + "step": 9173 + }, + { + "epoch": 0.34, + "learning_rate": 1.6658732193210532e-07, + "loss": 0.6925, + "step": 9174 + }, + { + "epoch": 0.34, + "learning_rate": 1.6658365986743324e-07, + "loss": 0.6926, + "step": 9175 + }, + { + "epoch": 0.34, + "learning_rate": 1.665799978027612e-07, + "loss": 0.7013, + "step": 9176 + }, + { + "epoch": 0.34, + "learning_rate": 1.6657633573808912e-07, + "loss": 0.6983, + "step": 9177 + }, + { + "epoch": 0.34, + "learning_rate": 1.6657267367341708e-07, + "loss": 0.7011, + "step": 9178 + }, + { + "epoch": 0.34, + "learning_rate": 1.66569011608745e-07, + "loss": 0.6898, + "step": 9179 + }, + { + "epoch": 0.34, + "learning_rate": 1.6656534954407293e-07, + "loss": 0.6914, + "step": 9180 + }, + { + "epoch": 0.34, + "learning_rate": 1.6656168747940086e-07, + "loss": 0.6972, + "step": 9181 + }, + { + "epoch": 0.34, + "learning_rate": 1.665580254147288e-07, + "loss": 0.6898, + "step": 9182 + }, + { + "epoch": 0.34, + "learning_rate": 1.6655436335005674e-07, + "loss": 0.6954, + "step": 9183 + }, + { + "epoch": 0.34, + "learning_rate": 1.665507012853847e-07, + "loss": 0.694, + "step": 9184 + }, + { + "epoch": 0.34, + "learning_rate": 1.6654703922071262e-07, + "loss": 0.7058, + "step": 9185 + }, + { + "epoch": 0.34, + "learning_rate": 1.6654337715604057e-07, + "loss": 0.6724, + "step": 9186 + }, + { + "epoch": 0.34, + "learning_rate": 1.665397150913685e-07, + "loss": 0.6881, + "step": 9187 + }, + { + "epoch": 0.34, + "learning_rate": 1.6653605302669645e-07, + "loss": 0.711, + "step": 9188 + }, + { + "epoch": 0.34, + "learning_rate": 1.6653239096202438e-07, + "loss": 0.6985, + "step": 9189 + }, + { + "epoch": 0.34, + "learning_rate": 1.6652872889735233e-07, + "loss": 0.6921, + "step": 9190 + }, + { + "epoch": 0.34, + "learning_rate": 1.6652506683268026e-07, + "loss": 0.6972, + "step": 9191 + }, + { + "epoch": 0.34, + "learning_rate": 1.665214047680082e-07, + "loss": 0.6822, + "step": 9192 + }, + { + "epoch": 0.34, + "learning_rate": 1.6651774270333614e-07, + "loss": 0.6817, + "step": 9193 + }, + { + "epoch": 0.34, + "learning_rate": 1.665140806386641e-07, + "loss": 0.6921, + "step": 9194 + }, + { + "epoch": 0.34, + "learning_rate": 1.6651041857399202e-07, + "loss": 0.6905, + "step": 9195 + }, + { + "epoch": 0.34, + "learning_rate": 1.6650675650931997e-07, + "loss": 0.7019, + "step": 9196 + }, + { + "epoch": 0.34, + "learning_rate": 1.665030944446479e-07, + "loss": 0.6814, + "step": 9197 + }, + { + "epoch": 0.34, + "learning_rate": 1.6649943237997582e-07, + "loss": 0.7041, + "step": 9198 + }, + { + "epoch": 0.34, + "learning_rate": 1.6649577031530375e-07, + "loss": 0.68, + "step": 9199 + }, + { + "epoch": 0.34, + "learning_rate": 1.664921082506317e-07, + "loss": 0.6983, + "step": 9200 + }, + { + "epoch": 0.34, + "learning_rate": 1.6648844618595963e-07, + "loss": 0.6836, + "step": 9201 + }, + { + "epoch": 0.34, + "learning_rate": 1.6648478412128756e-07, + "loss": 0.6947, + "step": 9202 + }, + { + "epoch": 0.34, + "learning_rate": 1.664811220566155e-07, + "loss": 0.6806, + "step": 9203 + }, + { + "epoch": 0.34, + "learning_rate": 1.6647745999194344e-07, + "loss": 0.6895, + "step": 9204 + }, + { + "epoch": 0.34, + "learning_rate": 1.664737979272714e-07, + "loss": 0.6902, + "step": 9205 + }, + { + "epoch": 0.34, + "learning_rate": 1.6647013586259932e-07, + "loss": 0.685, + "step": 9206 + }, + { + "epoch": 0.34, + "learning_rate": 1.6646647379792727e-07, + "loss": 0.6936, + "step": 9207 + }, + { + "epoch": 0.34, + "learning_rate": 1.664628117332552e-07, + "loss": 0.6973, + "step": 9208 + }, + { + "epoch": 0.34, + "learning_rate": 1.6645914966858315e-07, + "loss": 0.6862, + "step": 9209 + }, + { + "epoch": 0.34, + "learning_rate": 1.6645548760391108e-07, + "loss": 0.6887, + "step": 9210 + }, + { + "epoch": 0.34, + "learning_rate": 1.6645182553923903e-07, + "loss": 0.7116, + "step": 9211 + }, + { + "epoch": 0.34, + "learning_rate": 1.6644816347456696e-07, + "loss": 0.6894, + "step": 9212 + }, + { + "epoch": 0.34, + "learning_rate": 1.664445014098949e-07, + "loss": 0.7078, + "step": 9213 + }, + { + "epoch": 0.34, + "learning_rate": 1.6644083934522284e-07, + "loss": 0.7, + "step": 9214 + }, + { + "epoch": 0.34, + "learning_rate": 1.664371772805508e-07, + "loss": 0.6954, + "step": 9215 + }, + { + "epoch": 0.34, + "learning_rate": 1.664335152158787e-07, + "loss": 0.6941, + "step": 9216 + }, + { + "epoch": 0.34, + "learning_rate": 1.6642985315120664e-07, + "loss": 0.695, + "step": 9217 + }, + { + "epoch": 0.34, + "learning_rate": 1.6642619108653457e-07, + "loss": 0.6835, + "step": 9218 + }, + { + "epoch": 0.34, + "learning_rate": 1.6642252902186252e-07, + "loss": 0.6839, + "step": 9219 + }, + { + "epoch": 0.34, + "learning_rate": 1.6641886695719045e-07, + "loss": 0.6992, + "step": 9220 + }, + { + "epoch": 0.34, + "learning_rate": 1.664152048925184e-07, + "loss": 0.694, + "step": 9221 + }, + { + "epoch": 0.34, + "learning_rate": 1.6641154282784633e-07, + "loss": 0.7048, + "step": 9222 + }, + { + "epoch": 0.34, + "learning_rate": 1.6640788076317428e-07, + "loss": 0.6875, + "step": 9223 + }, + { + "epoch": 0.34, + "learning_rate": 1.664042186985022e-07, + "loss": 0.6953, + "step": 9224 + }, + { + "epoch": 0.34, + "learning_rate": 1.6640055663383014e-07, + "loss": 0.6942, + "step": 9225 + }, + { + "epoch": 0.34, + "learning_rate": 1.663968945691581e-07, + "loss": 0.7108, + "step": 9226 + }, + { + "epoch": 0.34, + "learning_rate": 1.6639323250448601e-07, + "loss": 0.7114, + "step": 9227 + }, + { + "epoch": 0.34, + "learning_rate": 1.6638957043981397e-07, + "loss": 0.686, + "step": 9228 + }, + { + "epoch": 0.34, + "learning_rate": 1.663859083751419e-07, + "loss": 0.6918, + "step": 9229 + }, + { + "epoch": 0.34, + "learning_rate": 1.6638224631046985e-07, + "loss": 0.6989, + "step": 9230 + }, + { + "epoch": 0.34, + "learning_rate": 1.6637858424579777e-07, + "loss": 0.699, + "step": 9231 + }, + { + "epoch": 0.34, + "learning_rate": 1.6637492218112573e-07, + "loss": 0.6987, + "step": 9232 + }, + { + "epoch": 0.34, + "learning_rate": 1.6637126011645365e-07, + "loss": 0.6952, + "step": 9233 + }, + { + "epoch": 0.34, + "learning_rate": 1.6636759805178158e-07, + "loss": 0.69, + "step": 9234 + }, + { + "epoch": 0.34, + "learning_rate": 1.663639359871095e-07, + "loss": 0.6808, + "step": 9235 + }, + { + "epoch": 0.34, + "learning_rate": 1.6636027392243746e-07, + "loss": 0.7013, + "step": 9236 + }, + { + "epoch": 0.34, + "learning_rate": 1.663566118577654e-07, + "loss": 0.7023, + "step": 9237 + }, + { + "epoch": 0.34, + "learning_rate": 1.6635294979309334e-07, + "loss": 0.6994, + "step": 9238 + }, + { + "epoch": 0.34, + "learning_rate": 1.6634928772842127e-07, + "loss": 0.6744, + "step": 9239 + }, + { + "epoch": 0.34, + "learning_rate": 1.6634562566374922e-07, + "loss": 0.6883, + "step": 9240 + }, + { + "epoch": 0.34, + "learning_rate": 1.6634196359907715e-07, + "loss": 0.7047, + "step": 9241 + }, + { + "epoch": 0.34, + "learning_rate": 1.663383015344051e-07, + "loss": 0.6996, + "step": 9242 + }, + { + "epoch": 0.34, + "learning_rate": 1.6633463946973303e-07, + "loss": 0.6679, + "step": 9243 + }, + { + "epoch": 0.34, + "learning_rate": 1.6633097740506098e-07, + "loss": 0.7047, + "step": 9244 + }, + { + "epoch": 0.34, + "learning_rate": 1.663273153403889e-07, + "loss": 0.6953, + "step": 9245 + }, + { + "epoch": 0.34, + "learning_rate": 1.6632365327571683e-07, + "loss": 0.6988, + "step": 9246 + }, + { + "epoch": 0.34, + "learning_rate": 1.663199912110448e-07, + "loss": 0.6899, + "step": 9247 + }, + { + "epoch": 0.34, + "learning_rate": 1.6631632914637271e-07, + "loss": 0.707, + "step": 9248 + }, + { + "epoch": 0.34, + "learning_rate": 1.6631266708170067e-07, + "loss": 0.6927, + "step": 9249 + }, + { + "epoch": 0.34, + "learning_rate": 1.663090050170286e-07, + "loss": 0.6953, + "step": 9250 + }, + { + "epoch": 0.34, + "learning_rate": 1.6630534295235655e-07, + "loss": 0.6749, + "step": 9251 + }, + { + "epoch": 0.34, + "learning_rate": 1.6630168088768445e-07, + "loss": 0.6999, + "step": 9252 + }, + { + "epoch": 0.34, + "learning_rate": 1.662980188230124e-07, + "loss": 0.6794, + "step": 9253 + }, + { + "epoch": 0.34, + "learning_rate": 1.6629435675834033e-07, + "loss": 0.6916, + "step": 9254 + }, + { + "epoch": 0.34, + "learning_rate": 1.6629069469366828e-07, + "loss": 0.6853, + "step": 9255 + }, + { + "epoch": 0.34, + "learning_rate": 1.662870326289962e-07, + "loss": 0.6895, + "step": 9256 + }, + { + "epoch": 0.34, + "learning_rate": 1.6628337056432416e-07, + "loss": 0.688, + "step": 9257 + }, + { + "epoch": 0.34, + "learning_rate": 1.6627970849965209e-07, + "loss": 0.6881, + "step": 9258 + }, + { + "epoch": 0.34, + "learning_rate": 1.6627604643498004e-07, + "loss": 0.6949, + "step": 9259 + }, + { + "epoch": 0.34, + "learning_rate": 1.6627238437030797e-07, + "loss": 0.6704, + "step": 9260 + }, + { + "epoch": 0.34, + "learning_rate": 1.6626872230563592e-07, + "loss": 0.6909, + "step": 9261 + }, + { + "epoch": 0.34, + "learning_rate": 1.6626506024096385e-07, + "loss": 0.6858, + "step": 9262 + }, + { + "epoch": 0.34, + "learning_rate": 1.662613981762918e-07, + "loss": 0.6908, + "step": 9263 + }, + { + "epoch": 0.34, + "learning_rate": 1.6625773611161973e-07, + "loss": 0.6935, + "step": 9264 + }, + { + "epoch": 0.34, + "learning_rate": 1.6625407404694768e-07, + "loss": 0.6926, + "step": 9265 + }, + { + "epoch": 0.34, + "learning_rate": 1.662504119822756e-07, + "loss": 0.6838, + "step": 9266 + }, + { + "epoch": 0.34, + "learning_rate": 1.6624674991760356e-07, + "loss": 0.6917, + "step": 9267 + }, + { + "epoch": 0.34, + "learning_rate": 1.6624308785293149e-07, + "loss": 0.7023, + "step": 9268 + }, + { + "epoch": 0.34, + "learning_rate": 1.662394257882594e-07, + "loss": 0.6967, + "step": 9269 + }, + { + "epoch": 0.34, + "learning_rate": 1.6623576372358734e-07, + "loss": 0.6863, + "step": 9270 + }, + { + "epoch": 0.34, + "learning_rate": 1.662321016589153e-07, + "loss": 0.6941, + "step": 9271 + }, + { + "epoch": 0.34, + "learning_rate": 1.6622843959424322e-07, + "loss": 0.7073, + "step": 9272 + }, + { + "epoch": 0.34, + "learning_rate": 1.6622477752957115e-07, + "loss": 0.6935, + "step": 9273 + }, + { + "epoch": 0.34, + "learning_rate": 1.662211154648991e-07, + "loss": 0.6951, + "step": 9274 + }, + { + "epoch": 0.34, + "learning_rate": 1.6621745340022703e-07, + "loss": 0.7077, + "step": 9275 + }, + { + "epoch": 0.34, + "learning_rate": 1.6621379133555498e-07, + "loss": 0.6999, + "step": 9276 + }, + { + "epoch": 0.34, + "learning_rate": 1.662101292708829e-07, + "loss": 0.7219, + "step": 9277 + }, + { + "epoch": 0.34, + "learning_rate": 1.6620646720621086e-07, + "loss": 0.6942, + "step": 9278 + }, + { + "epoch": 0.34, + "learning_rate": 1.6620280514153879e-07, + "loss": 0.6938, + "step": 9279 + }, + { + "epoch": 0.34, + "learning_rate": 1.6619914307686674e-07, + "loss": 0.6782, + "step": 9280 + }, + { + "epoch": 0.34, + "learning_rate": 1.6619548101219467e-07, + "loss": 0.6977, + "step": 9281 + }, + { + "epoch": 0.34, + "learning_rate": 1.6619181894752262e-07, + "loss": 0.7081, + "step": 9282 + }, + { + "epoch": 0.34, + "learning_rate": 1.6618815688285055e-07, + "loss": 0.6916, + "step": 9283 + }, + { + "epoch": 0.34, + "learning_rate": 1.661844948181785e-07, + "loss": 0.7096, + "step": 9284 + }, + { + "epoch": 0.34, + "learning_rate": 1.6618083275350642e-07, + "loss": 0.6954, + "step": 9285 + }, + { + "epoch": 0.34, + "learning_rate": 1.6617717068883438e-07, + "loss": 0.6839, + "step": 9286 + }, + { + "epoch": 0.34, + "learning_rate": 1.6617350862416228e-07, + "loss": 0.7066, + "step": 9287 + }, + { + "epoch": 0.34, + "learning_rate": 1.6616984655949023e-07, + "loss": 0.6994, + "step": 9288 + }, + { + "epoch": 0.34, + "learning_rate": 1.6616618449481816e-07, + "loss": 0.6881, + "step": 9289 + }, + { + "epoch": 0.34, + "learning_rate": 1.661625224301461e-07, + "loss": 0.6933, + "step": 9290 + }, + { + "epoch": 0.34, + "learning_rate": 1.6615886036547404e-07, + "loss": 0.6912, + "step": 9291 + }, + { + "epoch": 0.34, + "learning_rate": 1.66155198300802e-07, + "loss": 0.6868, + "step": 9292 + }, + { + "epoch": 0.34, + "learning_rate": 1.6615153623612992e-07, + "loss": 0.6898, + "step": 9293 + }, + { + "epoch": 0.34, + "learning_rate": 1.6614787417145787e-07, + "loss": 0.6913, + "step": 9294 + }, + { + "epoch": 0.34, + "learning_rate": 1.661442121067858e-07, + "loss": 0.7071, + "step": 9295 + }, + { + "epoch": 0.34, + "learning_rate": 1.6614055004211372e-07, + "loss": 0.7003, + "step": 9296 + }, + { + "epoch": 0.34, + "learning_rate": 1.6613688797744168e-07, + "loss": 0.6863, + "step": 9297 + }, + { + "epoch": 0.34, + "learning_rate": 1.661332259127696e-07, + "loss": 0.6824, + "step": 9298 + }, + { + "epoch": 0.34, + "learning_rate": 1.6612956384809756e-07, + "loss": 0.683, + "step": 9299 + }, + { + "epoch": 0.34, + "learning_rate": 1.6612590178342548e-07, + "loss": 0.681, + "step": 9300 + }, + { + "epoch": 0.34, + "learning_rate": 1.6612223971875344e-07, + "loss": 0.6872, + "step": 9301 + }, + { + "epoch": 0.34, + "learning_rate": 1.6611857765408136e-07, + "loss": 0.7064, + "step": 9302 + }, + { + "epoch": 0.34, + "learning_rate": 1.6611491558940932e-07, + "loss": 0.7052, + "step": 9303 + }, + { + "epoch": 0.34, + "learning_rate": 1.6611125352473724e-07, + "loss": 0.709, + "step": 9304 + }, + { + "epoch": 0.34, + "learning_rate": 1.6610759146006517e-07, + "loss": 0.6936, + "step": 9305 + }, + { + "epoch": 0.34, + "learning_rate": 1.661039293953931e-07, + "loss": 0.6886, + "step": 9306 + }, + { + "epoch": 0.34, + "learning_rate": 1.6610026733072105e-07, + "loss": 0.6883, + "step": 9307 + }, + { + "epoch": 0.34, + "learning_rate": 1.6609660526604898e-07, + "loss": 0.6733, + "step": 9308 + }, + { + "epoch": 0.34, + "learning_rate": 1.6609294320137693e-07, + "loss": 0.6811, + "step": 9309 + }, + { + "epoch": 0.34, + "learning_rate": 1.6608928113670486e-07, + "loss": 0.6981, + "step": 9310 + }, + { + "epoch": 0.34, + "learning_rate": 1.660856190720328e-07, + "loss": 0.7126, + "step": 9311 + }, + { + "epoch": 0.34, + "learning_rate": 1.6608195700736074e-07, + "loss": 0.6966, + "step": 9312 + }, + { + "epoch": 0.34, + "learning_rate": 1.660782949426887e-07, + "loss": 0.6806, + "step": 9313 + }, + { + "epoch": 0.34, + "learning_rate": 1.6607463287801662e-07, + "loss": 0.7049, + "step": 9314 + }, + { + "epoch": 0.34, + "learning_rate": 1.6607097081334457e-07, + "loss": 0.6897, + "step": 9315 + }, + { + "epoch": 0.34, + "learning_rate": 1.660673087486725e-07, + "loss": 0.687, + "step": 9316 + }, + { + "epoch": 0.34, + "learning_rate": 1.6606364668400042e-07, + "loss": 0.6953, + "step": 9317 + }, + { + "epoch": 0.34, + "learning_rate": 1.6605998461932838e-07, + "loss": 0.6891, + "step": 9318 + }, + { + "epoch": 0.34, + "learning_rate": 1.660563225546563e-07, + "loss": 0.6801, + "step": 9319 + }, + { + "epoch": 0.34, + "learning_rate": 1.6605266048998426e-07, + "loss": 0.685, + "step": 9320 + }, + { + "epoch": 0.34, + "learning_rate": 1.6604899842531218e-07, + "loss": 0.7085, + "step": 9321 + }, + { + "epoch": 0.34, + "learning_rate": 1.6604533636064014e-07, + "loss": 0.6722, + "step": 9322 + }, + { + "epoch": 0.34, + "learning_rate": 1.6604167429596804e-07, + "loss": 0.6957, + "step": 9323 + }, + { + "epoch": 0.34, + "learning_rate": 1.66038012231296e-07, + "loss": 0.6968, + "step": 9324 + }, + { + "epoch": 0.34, + "learning_rate": 1.6603435016662392e-07, + "loss": 0.6909, + "step": 9325 + }, + { + "epoch": 0.34, + "learning_rate": 1.6603068810195187e-07, + "loss": 0.695, + "step": 9326 + }, + { + "epoch": 0.34, + "learning_rate": 1.660270260372798e-07, + "loss": 0.6839, + "step": 9327 + }, + { + "epoch": 0.34, + "learning_rate": 1.6602336397260775e-07, + "loss": 0.7104, + "step": 9328 + }, + { + "epoch": 0.34, + "learning_rate": 1.6601970190793568e-07, + "loss": 0.7017, + "step": 9329 + }, + { + "epoch": 0.34, + "learning_rate": 1.6601603984326363e-07, + "loss": 0.6909, + "step": 9330 + }, + { + "epoch": 0.34, + "learning_rate": 1.6601237777859156e-07, + "loss": 0.7003, + "step": 9331 + }, + { + "epoch": 0.34, + "learning_rate": 1.660087157139195e-07, + "loss": 0.7022, + "step": 9332 + }, + { + "epoch": 0.34, + "learning_rate": 1.6600505364924744e-07, + "loss": 0.6942, + "step": 9333 + }, + { + "epoch": 0.34, + "learning_rate": 1.660013915845754e-07, + "loss": 0.7036, + "step": 9334 + }, + { + "epoch": 0.34, + "learning_rate": 1.6599772951990332e-07, + "loss": 0.6839, + "step": 9335 + }, + { + "epoch": 0.34, + "learning_rate": 1.6599406745523127e-07, + "loss": 0.6933, + "step": 9336 + }, + { + "epoch": 0.34, + "learning_rate": 1.659904053905592e-07, + "loss": 0.6898, + "step": 9337 + }, + { + "epoch": 0.34, + "learning_rate": 1.6598674332588712e-07, + "loss": 0.6887, + "step": 9338 + }, + { + "epoch": 0.34, + "learning_rate": 1.6598308126121508e-07, + "loss": 0.6954, + "step": 9339 + }, + { + "epoch": 0.34, + "learning_rate": 1.65979419196543e-07, + "loss": 0.6965, + "step": 9340 + }, + { + "epoch": 0.34, + "learning_rate": 1.6597575713187093e-07, + "loss": 0.694, + "step": 9341 + }, + { + "epoch": 0.34, + "learning_rate": 1.6597209506719888e-07, + "loss": 0.698, + "step": 9342 + }, + { + "epoch": 0.34, + "learning_rate": 1.659684330025268e-07, + "loss": 0.6742, + "step": 9343 + }, + { + "epoch": 0.34, + "learning_rate": 1.6596477093785474e-07, + "loss": 0.701, + "step": 9344 + }, + { + "epoch": 0.34, + "learning_rate": 1.659611088731827e-07, + "loss": 0.7054, + "step": 9345 + }, + { + "epoch": 0.34, + "learning_rate": 1.6595744680851062e-07, + "loss": 0.6953, + "step": 9346 + }, + { + "epoch": 0.34, + "learning_rate": 1.6595378474383857e-07, + "loss": 0.6911, + "step": 9347 + }, + { + "epoch": 0.34, + "learning_rate": 1.659501226791665e-07, + "loss": 0.6736, + "step": 9348 + }, + { + "epoch": 0.34, + "learning_rate": 1.6594646061449445e-07, + "loss": 0.7104, + "step": 9349 + }, + { + "epoch": 0.34, + "learning_rate": 1.6594279854982237e-07, + "loss": 0.6783, + "step": 9350 + }, + { + "epoch": 0.34, + "learning_rate": 1.6593913648515033e-07, + "loss": 0.6762, + "step": 9351 + }, + { + "epoch": 0.34, + "learning_rate": 1.6593547442047825e-07, + "loss": 0.6917, + "step": 9352 + }, + { + "epoch": 0.34, + "learning_rate": 1.659318123558062e-07, + "loss": 0.6913, + "step": 9353 + }, + { + "epoch": 0.34, + "learning_rate": 1.6592815029113413e-07, + "loss": 0.6934, + "step": 9354 + }, + { + "epoch": 0.34, + "learning_rate": 1.659244882264621e-07, + "loss": 0.6896, + "step": 9355 + }, + { + "epoch": 0.34, + "learning_rate": 1.6592082616179001e-07, + "loss": 0.683, + "step": 9356 + }, + { + "epoch": 0.34, + "learning_rate": 1.6591716409711797e-07, + "loss": 0.6917, + "step": 9357 + }, + { + "epoch": 0.34, + "learning_rate": 1.659135020324459e-07, + "loss": 0.6912, + "step": 9358 + }, + { + "epoch": 0.34, + "learning_rate": 1.6590983996777382e-07, + "loss": 0.6906, + "step": 9359 + }, + { + "epoch": 0.34, + "learning_rate": 1.6590617790310175e-07, + "loss": 0.6933, + "step": 9360 + }, + { + "epoch": 0.34, + "learning_rate": 1.659025158384297e-07, + "loss": 0.6946, + "step": 9361 + }, + { + "epoch": 0.34, + "learning_rate": 1.6589885377375763e-07, + "loss": 0.7027, + "step": 9362 + }, + { + "epoch": 0.34, + "learning_rate": 1.6589519170908558e-07, + "loss": 0.6879, + "step": 9363 + }, + { + "epoch": 0.34, + "learning_rate": 1.658915296444135e-07, + "loss": 0.6935, + "step": 9364 + }, + { + "epoch": 0.34, + "learning_rate": 1.6588786757974146e-07, + "loss": 0.7023, + "step": 9365 + }, + { + "epoch": 0.34, + "learning_rate": 1.658842055150694e-07, + "loss": 0.6925, + "step": 9366 + }, + { + "epoch": 0.34, + "learning_rate": 1.6588054345039731e-07, + "loss": 0.693, + "step": 9367 + }, + { + "epoch": 0.34, + "learning_rate": 1.6587688138572527e-07, + "loss": 0.697, + "step": 9368 + }, + { + "epoch": 0.34, + "learning_rate": 1.658732193210532e-07, + "loss": 0.6973, + "step": 9369 + }, + { + "epoch": 0.34, + "learning_rate": 1.6586955725638115e-07, + "loss": 0.7126, + "step": 9370 + }, + { + "epoch": 0.34, + "learning_rate": 1.6586589519170907e-07, + "loss": 0.7046, + "step": 9371 + }, + { + "epoch": 0.34, + "learning_rate": 1.6586223312703703e-07, + "loss": 0.6931, + "step": 9372 + }, + { + "epoch": 0.34, + "learning_rate": 1.6585857106236495e-07, + "loss": 0.7031, + "step": 9373 + }, + { + "epoch": 0.34, + "learning_rate": 1.658549089976929e-07, + "loss": 0.6926, + "step": 9374 + }, + { + "epoch": 0.34, + "learning_rate": 1.6585124693302083e-07, + "loss": 0.6919, + "step": 9375 + }, + { + "epoch": 0.34, + "learning_rate": 1.6584758486834879e-07, + "loss": 0.6718, + "step": 9376 + }, + { + "epoch": 0.34, + "learning_rate": 1.6584392280367669e-07, + "loss": 0.7031, + "step": 9377 + }, + { + "epoch": 0.34, + "learning_rate": 1.6584026073900464e-07, + "loss": 0.6968, + "step": 9378 + }, + { + "epoch": 0.34, + "learning_rate": 1.6583659867433257e-07, + "loss": 0.7069, + "step": 9379 + }, + { + "epoch": 0.34, + "learning_rate": 1.6583293660966052e-07, + "loss": 0.6798, + "step": 9380 + }, + { + "epoch": 0.34, + "learning_rate": 1.6582927454498845e-07, + "loss": 0.6992, + "step": 9381 + }, + { + "epoch": 0.34, + "learning_rate": 1.658256124803164e-07, + "loss": 0.7018, + "step": 9382 + }, + { + "epoch": 0.34, + "learning_rate": 1.6582195041564433e-07, + "loss": 0.6852, + "step": 9383 + }, + { + "epoch": 0.34, + "learning_rate": 1.6581828835097228e-07, + "loss": 0.6826, + "step": 9384 + }, + { + "epoch": 0.34, + "learning_rate": 1.658146262863002e-07, + "loss": 0.6877, + "step": 9385 + }, + { + "epoch": 0.34, + "learning_rate": 1.6581096422162816e-07, + "loss": 0.6909, + "step": 9386 + }, + { + "epoch": 0.34, + "learning_rate": 1.6580730215695609e-07, + "loss": 0.6962, + "step": 9387 + }, + { + "epoch": 0.34, + "learning_rate": 1.65803640092284e-07, + "loss": 0.7063, + "step": 9388 + }, + { + "epoch": 0.34, + "learning_rate": 1.6579997802761197e-07, + "loss": 0.6869, + "step": 9389 + }, + { + "epoch": 0.34, + "learning_rate": 1.657963159629399e-07, + "loss": 0.7028, + "step": 9390 + }, + { + "epoch": 0.34, + "learning_rate": 1.6579265389826785e-07, + "loss": 0.6942, + "step": 9391 + }, + { + "epoch": 0.34, + "learning_rate": 1.6578899183359577e-07, + "loss": 0.696, + "step": 9392 + }, + { + "epoch": 0.34, + "learning_rate": 1.6578532976892373e-07, + "loss": 0.6834, + "step": 9393 + }, + { + "epoch": 0.34, + "learning_rate": 1.6578166770425165e-07, + "loss": 0.6956, + "step": 9394 + }, + { + "epoch": 0.34, + "learning_rate": 1.6577800563957958e-07, + "loss": 0.6779, + "step": 9395 + }, + { + "epoch": 0.34, + "learning_rate": 1.657743435749075e-07, + "loss": 0.6955, + "step": 9396 + }, + { + "epoch": 0.34, + "learning_rate": 1.6577068151023546e-07, + "loss": 0.6848, + "step": 9397 + }, + { + "epoch": 0.34, + "learning_rate": 1.6576701944556339e-07, + "loss": 0.6957, + "step": 9398 + }, + { + "epoch": 0.34, + "learning_rate": 1.6576335738089134e-07, + "loss": 0.7011, + "step": 9399 + }, + { + "epoch": 0.34, + "learning_rate": 1.6575969531621927e-07, + "loss": 0.6903, + "step": 9400 + }, + { + "epoch": 0.34, + "learning_rate": 1.6575603325154722e-07, + "loss": 0.7133, + "step": 9401 + }, + { + "epoch": 0.34, + "learning_rate": 1.6575237118687515e-07, + "loss": 0.7028, + "step": 9402 + }, + { + "epoch": 0.34, + "learning_rate": 1.657487091222031e-07, + "loss": 0.7032, + "step": 9403 + }, + { + "epoch": 0.34, + "learning_rate": 1.6574504705753102e-07, + "loss": 0.6856, + "step": 9404 + }, + { + "epoch": 0.34, + "learning_rate": 1.6574138499285898e-07, + "loss": 0.6899, + "step": 9405 + }, + { + "epoch": 0.34, + "learning_rate": 1.657377229281869e-07, + "loss": 0.6891, + "step": 9406 + }, + { + "epoch": 0.34, + "learning_rate": 1.6573406086351486e-07, + "loss": 0.7119, + "step": 9407 + }, + { + "epoch": 0.34, + "learning_rate": 1.6573039879884278e-07, + "loss": 0.7066, + "step": 9408 + }, + { + "epoch": 0.34, + "learning_rate": 1.657267367341707e-07, + "loss": 0.6934, + "step": 9409 + }, + { + "epoch": 0.34, + "learning_rate": 1.6572307466949866e-07, + "loss": 0.6879, + "step": 9410 + }, + { + "epoch": 0.34, + "learning_rate": 1.657194126048266e-07, + "loss": 0.6867, + "step": 9411 + }, + { + "epoch": 0.34, + "learning_rate": 1.6571575054015454e-07, + "loss": 0.6847, + "step": 9412 + }, + { + "epoch": 0.34, + "learning_rate": 1.6571208847548247e-07, + "loss": 0.6831, + "step": 9413 + }, + { + "epoch": 0.34, + "learning_rate": 1.657084264108104e-07, + "loss": 0.6893, + "step": 9414 + }, + { + "epoch": 0.34, + "learning_rate": 1.6570476434613832e-07, + "loss": 0.6913, + "step": 9415 + }, + { + "epoch": 0.34, + "learning_rate": 1.6570110228146628e-07, + "loss": 0.6902, + "step": 9416 + }, + { + "epoch": 0.34, + "learning_rate": 1.656974402167942e-07, + "loss": 0.7052, + "step": 9417 + }, + { + "epoch": 0.34, + "learning_rate": 1.6569377815212216e-07, + "loss": 0.6859, + "step": 9418 + }, + { + "epoch": 0.34, + "learning_rate": 1.6569011608745008e-07, + "loss": 0.683, + "step": 9419 + }, + { + "epoch": 0.34, + "learning_rate": 1.6568645402277804e-07, + "loss": 0.6963, + "step": 9420 + }, + { + "epoch": 0.34, + "learning_rate": 1.6568279195810596e-07, + "loss": 0.6953, + "step": 9421 + }, + { + "epoch": 0.34, + "learning_rate": 1.6567912989343392e-07, + "loss": 0.6963, + "step": 9422 + }, + { + "epoch": 0.34, + "learning_rate": 1.6567546782876184e-07, + "loss": 0.6934, + "step": 9423 + }, + { + "epoch": 0.34, + "learning_rate": 1.656718057640898e-07, + "loss": 0.6803, + "step": 9424 + }, + { + "epoch": 0.34, + "learning_rate": 1.6566814369941772e-07, + "loss": 0.6917, + "step": 9425 + }, + { + "epoch": 0.34, + "learning_rate": 1.6566448163474568e-07, + "loss": 0.6881, + "step": 9426 + }, + { + "epoch": 0.34, + "learning_rate": 1.656608195700736e-07, + "loss": 0.711, + "step": 9427 + }, + { + "epoch": 0.34, + "learning_rate": 1.6565715750540156e-07, + "loss": 0.6997, + "step": 9428 + }, + { + "epoch": 0.34, + "learning_rate": 1.6565349544072948e-07, + "loss": 0.6896, + "step": 9429 + }, + { + "epoch": 0.35, + "learning_rate": 1.6564983337605744e-07, + "loss": 0.71, + "step": 9430 + }, + { + "epoch": 0.35, + "learning_rate": 1.6564617131138534e-07, + "loss": 0.6842, + "step": 9431 + }, + { + "epoch": 0.35, + "learning_rate": 1.656425092467133e-07, + "loss": 0.6962, + "step": 9432 + }, + { + "epoch": 0.35, + "learning_rate": 1.6563884718204122e-07, + "loss": 0.6869, + "step": 9433 + }, + { + "epoch": 0.35, + "learning_rate": 1.6563518511736917e-07, + "loss": 0.6974, + "step": 9434 + }, + { + "epoch": 0.35, + "learning_rate": 1.656315230526971e-07, + "loss": 0.6964, + "step": 9435 + }, + { + "epoch": 0.35, + "learning_rate": 1.6562786098802505e-07, + "loss": 0.689, + "step": 9436 + }, + { + "epoch": 0.35, + "learning_rate": 1.6562419892335298e-07, + "loss": 0.689, + "step": 9437 + }, + { + "epoch": 0.35, + "learning_rate": 1.656205368586809e-07, + "loss": 0.7047, + "step": 9438 + }, + { + "epoch": 0.35, + "learning_rate": 1.6561687479400886e-07, + "loss": 0.6855, + "step": 9439 + }, + { + "epoch": 0.35, + "learning_rate": 1.6561321272933678e-07, + "loss": 0.6867, + "step": 9440 + }, + { + "epoch": 0.35, + "learning_rate": 1.6560955066466474e-07, + "loss": 0.6944, + "step": 9441 + }, + { + "epoch": 0.35, + "learning_rate": 1.6560588859999266e-07, + "loss": 0.6915, + "step": 9442 + }, + { + "epoch": 0.35, + "learning_rate": 1.6560222653532062e-07, + "loss": 0.7034, + "step": 9443 + }, + { + "epoch": 0.35, + "learning_rate": 1.6559856447064854e-07, + "loss": 0.6983, + "step": 9444 + }, + { + "epoch": 0.35, + "learning_rate": 1.655949024059765e-07, + "loss": 0.6679, + "step": 9445 + }, + { + "epoch": 0.35, + "learning_rate": 1.6559124034130442e-07, + "loss": 0.6929, + "step": 9446 + }, + { + "epoch": 0.35, + "learning_rate": 1.6558757827663238e-07, + "loss": 0.692, + "step": 9447 + }, + { + "epoch": 0.35, + "learning_rate": 1.655839162119603e-07, + "loss": 0.6878, + "step": 9448 + }, + { + "epoch": 0.35, + "learning_rate": 1.6558025414728823e-07, + "loss": 0.7074, + "step": 9449 + }, + { + "epoch": 0.35, + "learning_rate": 1.6557659208261616e-07, + "loss": 0.6997, + "step": 9450 + }, + { + "epoch": 0.35, + "learning_rate": 1.655729300179441e-07, + "loss": 0.6893, + "step": 9451 + }, + { + "epoch": 0.35, + "learning_rate": 1.6556926795327204e-07, + "loss": 0.6898, + "step": 9452 + }, + { + "epoch": 0.35, + "learning_rate": 1.655656058886e-07, + "loss": 0.69, + "step": 9453 + }, + { + "epoch": 0.35, + "learning_rate": 1.6556194382392792e-07, + "loss": 0.6716, + "step": 9454 + }, + { + "epoch": 0.35, + "learning_rate": 1.6555828175925587e-07, + "loss": 0.6956, + "step": 9455 + }, + { + "epoch": 0.35, + "learning_rate": 1.655546196945838e-07, + "loss": 0.6752, + "step": 9456 + }, + { + "epoch": 0.35, + "learning_rate": 1.6555095762991175e-07, + "loss": 0.6797, + "step": 9457 + }, + { + "epoch": 0.35, + "learning_rate": 1.6554729556523968e-07, + "loss": 0.693, + "step": 9458 + }, + { + "epoch": 0.35, + "learning_rate": 1.655436335005676e-07, + "loss": 0.6906, + "step": 9459 + }, + { + "epoch": 0.35, + "learning_rate": 1.6553997143589556e-07, + "loss": 0.6963, + "step": 9460 + }, + { + "epoch": 0.35, + "learning_rate": 1.6553630937122348e-07, + "loss": 0.6965, + "step": 9461 + }, + { + "epoch": 0.35, + "learning_rate": 1.6553264730655143e-07, + "loss": 0.6794, + "step": 9462 + }, + { + "epoch": 0.35, + "learning_rate": 1.6552898524187936e-07, + "loss": 0.6952, + "step": 9463 + }, + { + "epoch": 0.35, + "learning_rate": 1.6552532317720731e-07, + "loss": 0.6972, + "step": 9464 + }, + { + "epoch": 0.35, + "learning_rate": 1.6552166111253524e-07, + "loss": 0.6954, + "step": 9465 + }, + { + "epoch": 0.35, + "learning_rate": 1.655179990478632e-07, + "loss": 0.6988, + "step": 9466 + }, + { + "epoch": 0.35, + "learning_rate": 1.655143369831911e-07, + "loss": 0.7024, + "step": 9467 + }, + { + "epoch": 0.35, + "learning_rate": 1.6551067491851905e-07, + "loss": 0.7138, + "step": 9468 + }, + { + "epoch": 0.35, + "learning_rate": 1.6550701285384697e-07, + "loss": 0.7065, + "step": 9469 + }, + { + "epoch": 0.35, + "learning_rate": 1.6550335078917493e-07, + "loss": 0.7063, + "step": 9470 + }, + { + "epoch": 0.35, + "learning_rate": 1.6549968872450285e-07, + "loss": 0.7046, + "step": 9471 + }, + { + "epoch": 0.35, + "learning_rate": 1.654960266598308e-07, + "loss": 0.7019, + "step": 9472 + }, + { + "epoch": 0.35, + "learning_rate": 1.6549236459515873e-07, + "loss": 0.698, + "step": 9473 + }, + { + "epoch": 0.35, + "learning_rate": 1.654887025304867e-07, + "loss": 0.6948, + "step": 9474 + }, + { + "epoch": 0.35, + "learning_rate": 1.6548504046581461e-07, + "loss": 0.693, + "step": 9475 + }, + { + "epoch": 0.35, + "learning_rate": 1.6548137840114257e-07, + "loss": 0.6921, + "step": 9476 + }, + { + "epoch": 0.35, + "learning_rate": 1.654777163364705e-07, + "loss": 0.6946, + "step": 9477 + }, + { + "epoch": 0.35, + "learning_rate": 1.6547405427179845e-07, + "loss": 0.694, + "step": 9478 + }, + { + "epoch": 0.35, + "learning_rate": 1.6547039220712637e-07, + "loss": 0.6912, + "step": 9479 + }, + { + "epoch": 0.35, + "learning_rate": 1.654667301424543e-07, + "loss": 0.6924, + "step": 9480 + }, + { + "epoch": 0.35, + "learning_rate": 1.6546306807778225e-07, + "loss": 0.6934, + "step": 9481 + }, + { + "epoch": 0.35, + "learning_rate": 1.6545940601311018e-07, + "loss": 0.6987, + "step": 9482 + }, + { + "epoch": 0.35, + "learning_rate": 1.6545574394843813e-07, + "loss": 0.6942, + "step": 9483 + }, + { + "epoch": 0.35, + "learning_rate": 1.6545208188376606e-07, + "loss": 0.6783, + "step": 9484 + }, + { + "epoch": 0.35, + "learning_rate": 1.65448419819094e-07, + "loss": 0.7102, + "step": 9485 + }, + { + "epoch": 0.35, + "learning_rate": 1.6544475775442191e-07, + "loss": 0.7096, + "step": 9486 + }, + { + "epoch": 0.35, + "learning_rate": 1.6544109568974987e-07, + "loss": 0.6989, + "step": 9487 + }, + { + "epoch": 0.35, + "learning_rate": 1.654374336250778e-07, + "loss": 0.6961, + "step": 9488 + }, + { + "epoch": 0.35, + "learning_rate": 1.6543377156040575e-07, + "loss": 0.7142, + "step": 9489 + }, + { + "epoch": 0.35, + "learning_rate": 1.6543010949573367e-07, + "loss": 0.6895, + "step": 9490 + }, + { + "epoch": 0.35, + "learning_rate": 1.6542644743106163e-07, + "loss": 0.6809, + "step": 9491 + }, + { + "epoch": 0.35, + "learning_rate": 1.6542278536638955e-07, + "loss": 0.6913, + "step": 9492 + }, + { + "epoch": 0.35, + "learning_rate": 1.654191233017175e-07, + "loss": 0.702, + "step": 9493 + }, + { + "epoch": 0.35, + "learning_rate": 1.6541546123704543e-07, + "loss": 0.6915, + "step": 9494 + }, + { + "epoch": 0.35, + "learning_rate": 1.6541179917237339e-07, + "loss": 0.6923, + "step": 9495 + }, + { + "epoch": 0.35, + "learning_rate": 1.654081371077013e-07, + "loss": 0.6854, + "step": 9496 + }, + { + "epoch": 0.35, + "learning_rate": 1.6540447504302927e-07, + "loss": 0.6933, + "step": 9497 + }, + { + "epoch": 0.35, + "learning_rate": 1.654008129783572e-07, + "loss": 0.6948, + "step": 9498 + }, + { + "epoch": 0.35, + "learning_rate": 1.6539715091368515e-07, + "loss": 0.6893, + "step": 9499 + }, + { + "epoch": 0.35, + "learning_rate": 1.6539348884901307e-07, + "loss": 0.7047, + "step": 9500 + }, + { + "epoch": 0.35, + "learning_rate": 1.6538982678434103e-07, + "loss": 0.696, + "step": 9501 + }, + { + "epoch": 0.35, + "learning_rate": 1.6538616471966895e-07, + "loss": 0.6881, + "step": 9502 + }, + { + "epoch": 0.35, + "learning_rate": 1.6538250265499688e-07, + "loss": 0.696, + "step": 9503 + }, + { + "epoch": 0.35, + "learning_rate": 1.653788405903248e-07, + "loss": 0.6923, + "step": 9504 + }, + { + "epoch": 0.35, + "learning_rate": 1.6537517852565276e-07, + "loss": 0.6908, + "step": 9505 + }, + { + "epoch": 0.35, + "learning_rate": 1.6537151646098069e-07, + "loss": 0.6932, + "step": 9506 + }, + { + "epoch": 0.35, + "learning_rate": 1.653678543963086e-07, + "loss": 0.6957, + "step": 9507 + }, + { + "epoch": 0.35, + "learning_rate": 1.6536419233163657e-07, + "loss": 0.7008, + "step": 9508 + }, + { + "epoch": 0.35, + "learning_rate": 1.653605302669645e-07, + "loss": 0.7057, + "step": 9509 + }, + { + "epoch": 0.35, + "learning_rate": 1.6535686820229245e-07, + "loss": 0.6998, + "step": 9510 + }, + { + "epoch": 0.35, + "learning_rate": 1.6535320613762037e-07, + "loss": 0.6927, + "step": 9511 + }, + { + "epoch": 0.35, + "learning_rate": 1.6534954407294833e-07, + "loss": 0.6935, + "step": 9512 + }, + { + "epoch": 0.35, + "learning_rate": 1.6534588200827625e-07, + "loss": 0.6848, + "step": 9513 + }, + { + "epoch": 0.35, + "learning_rate": 1.653422199436042e-07, + "loss": 0.6822, + "step": 9514 + }, + { + "epoch": 0.35, + "learning_rate": 1.6533855787893213e-07, + "loss": 0.6836, + "step": 9515 + }, + { + "epoch": 0.35, + "learning_rate": 1.6533489581426009e-07, + "loss": 0.7136, + "step": 9516 + }, + { + "epoch": 0.35, + "learning_rate": 1.65331233749588e-07, + "loss": 0.6874, + "step": 9517 + }, + { + "epoch": 0.35, + "learning_rate": 1.6532757168491597e-07, + "loss": 0.7008, + "step": 9518 + }, + { + "epoch": 0.35, + "learning_rate": 1.653239096202439e-07, + "loss": 0.7167, + "step": 9519 + }, + { + "epoch": 0.35, + "learning_rate": 1.6532024755557184e-07, + "loss": 0.706, + "step": 9520 + }, + { + "epoch": 0.35, + "learning_rate": 1.6531658549089975e-07, + "loss": 0.6994, + "step": 9521 + }, + { + "epoch": 0.35, + "learning_rate": 1.653129234262277e-07, + "loss": 0.6964, + "step": 9522 + }, + { + "epoch": 0.35, + "learning_rate": 1.6530926136155563e-07, + "loss": 0.6964, + "step": 9523 + }, + { + "epoch": 0.35, + "learning_rate": 1.6530559929688358e-07, + "loss": 0.7059, + "step": 9524 + }, + { + "epoch": 0.35, + "learning_rate": 1.653019372322115e-07, + "loss": 0.7039, + "step": 9525 + }, + { + "epoch": 0.35, + "learning_rate": 1.6529827516753946e-07, + "loss": 0.6894, + "step": 9526 + }, + { + "epoch": 0.35, + "learning_rate": 1.6529461310286738e-07, + "loss": 0.7055, + "step": 9527 + }, + { + "epoch": 0.35, + "learning_rate": 1.6529095103819534e-07, + "loss": 0.6871, + "step": 9528 + }, + { + "epoch": 0.35, + "learning_rate": 1.6528728897352326e-07, + "loss": 0.696, + "step": 9529 + }, + { + "epoch": 0.35, + "learning_rate": 1.652836269088512e-07, + "loss": 0.6959, + "step": 9530 + }, + { + "epoch": 0.35, + "learning_rate": 1.6527996484417914e-07, + "loss": 0.6901, + "step": 9531 + }, + { + "epoch": 0.35, + "learning_rate": 1.6527630277950707e-07, + "loss": 0.6909, + "step": 9532 + }, + { + "epoch": 0.35, + "learning_rate": 1.6527264071483502e-07, + "loss": 0.6988, + "step": 9533 + }, + { + "epoch": 0.35, + "learning_rate": 1.6526897865016295e-07, + "loss": 0.6952, + "step": 9534 + }, + { + "epoch": 0.35, + "learning_rate": 1.652653165854909e-07, + "loss": 0.6801, + "step": 9535 + }, + { + "epoch": 0.35, + "learning_rate": 1.6526165452081883e-07, + "loss": 0.707, + "step": 9536 + }, + { + "epoch": 0.35, + "learning_rate": 1.6525799245614678e-07, + "loss": 0.6947, + "step": 9537 + }, + { + "epoch": 0.35, + "learning_rate": 1.652543303914747e-07, + "loss": 0.6836, + "step": 9538 + }, + { + "epoch": 0.35, + "learning_rate": 1.6525066832680264e-07, + "loss": 0.7066, + "step": 9539 + }, + { + "epoch": 0.35, + "learning_rate": 1.6524700626213056e-07, + "loss": 0.6934, + "step": 9540 + }, + { + "epoch": 0.35, + "learning_rate": 1.6524334419745852e-07, + "loss": 0.6989, + "step": 9541 + }, + { + "epoch": 0.35, + "learning_rate": 1.6523968213278644e-07, + "loss": 0.6935, + "step": 9542 + }, + { + "epoch": 0.35, + "learning_rate": 1.652360200681144e-07, + "loss": 0.6842, + "step": 9543 + }, + { + "epoch": 0.35, + "learning_rate": 1.6523235800344232e-07, + "loss": 0.69, + "step": 9544 + }, + { + "epoch": 0.35, + "learning_rate": 1.6522869593877028e-07, + "loss": 0.6766, + "step": 9545 + }, + { + "epoch": 0.35, + "learning_rate": 1.652250338740982e-07, + "loss": 0.6955, + "step": 9546 + }, + { + "epoch": 0.35, + "learning_rate": 1.6522137180942616e-07, + "loss": 0.6977, + "step": 9547 + }, + { + "epoch": 0.35, + "learning_rate": 1.6521770974475408e-07, + "loss": 0.71, + "step": 9548 + }, + { + "epoch": 0.35, + "learning_rate": 1.6521404768008204e-07, + "loss": 0.6941, + "step": 9549 + }, + { + "epoch": 0.35, + "learning_rate": 1.6521038561540996e-07, + "loss": 0.6966, + "step": 9550 + }, + { + "epoch": 0.35, + "learning_rate": 1.652067235507379e-07, + "loss": 0.7048, + "step": 9551 + }, + { + "epoch": 0.35, + "learning_rate": 1.6520306148606584e-07, + "loss": 0.7059, + "step": 9552 + }, + { + "epoch": 0.35, + "learning_rate": 1.6519939942139377e-07, + "loss": 0.7069, + "step": 9553 + }, + { + "epoch": 0.35, + "learning_rate": 1.6519573735672172e-07, + "loss": 0.6852, + "step": 9554 + }, + { + "epoch": 0.35, + "learning_rate": 1.6519207529204965e-07, + "loss": 0.691, + "step": 9555 + }, + { + "epoch": 0.35, + "learning_rate": 1.651884132273776e-07, + "loss": 0.6953, + "step": 9556 + }, + { + "epoch": 0.35, + "learning_rate": 1.651847511627055e-07, + "loss": 0.7107, + "step": 9557 + }, + { + "epoch": 0.35, + "learning_rate": 1.6518108909803346e-07, + "loss": 0.6952, + "step": 9558 + }, + { + "epoch": 0.35, + "learning_rate": 1.6517742703336138e-07, + "loss": 0.7057, + "step": 9559 + }, + { + "epoch": 0.35, + "learning_rate": 1.6517376496868934e-07, + "loss": 0.6818, + "step": 9560 + }, + { + "epoch": 0.35, + "learning_rate": 1.6517010290401726e-07, + "loss": 0.6952, + "step": 9561 + }, + { + "epoch": 0.35, + "learning_rate": 1.6516644083934522e-07, + "loss": 0.7133, + "step": 9562 + }, + { + "epoch": 0.35, + "learning_rate": 1.6516277877467314e-07, + "loss": 0.6822, + "step": 9563 + }, + { + "epoch": 0.35, + "learning_rate": 1.651591167100011e-07, + "loss": 0.6758, + "step": 9564 + }, + { + "epoch": 0.35, + "learning_rate": 1.6515545464532902e-07, + "loss": 0.6939, + "step": 9565 + }, + { + "epoch": 0.35, + "learning_rate": 1.6515179258065698e-07, + "loss": 0.6924, + "step": 9566 + }, + { + "epoch": 0.35, + "learning_rate": 1.651481305159849e-07, + "loss": 0.714, + "step": 9567 + }, + { + "epoch": 0.35, + "learning_rate": 1.6514446845131286e-07, + "loss": 0.7118, + "step": 9568 + }, + { + "epoch": 0.35, + "learning_rate": 1.6514080638664078e-07, + "loss": 0.6884, + "step": 9569 + }, + { + "epoch": 0.35, + "learning_rate": 1.6513714432196874e-07, + "loss": 0.7074, + "step": 9570 + }, + { + "epoch": 0.35, + "learning_rate": 1.6513348225729666e-07, + "loss": 0.6929, + "step": 9571 + }, + { + "epoch": 0.35, + "learning_rate": 1.6512982019262462e-07, + "loss": 0.7074, + "step": 9572 + }, + { + "epoch": 0.35, + "learning_rate": 1.6512615812795254e-07, + "loss": 0.696, + "step": 9573 + }, + { + "epoch": 0.35, + "learning_rate": 1.6512249606328047e-07, + "loss": 0.6971, + "step": 9574 + }, + { + "epoch": 0.35, + "learning_rate": 1.651188339986084e-07, + "loss": 0.6843, + "step": 9575 + }, + { + "epoch": 0.35, + "learning_rate": 1.6511517193393635e-07, + "loss": 0.682, + "step": 9576 + }, + { + "epoch": 0.35, + "learning_rate": 1.6511150986926428e-07, + "loss": 0.6744, + "step": 9577 + }, + { + "epoch": 0.35, + "learning_rate": 1.651078478045922e-07, + "loss": 0.6911, + "step": 9578 + }, + { + "epoch": 0.35, + "learning_rate": 1.6510418573992016e-07, + "loss": 0.6897, + "step": 9579 + }, + { + "epoch": 0.35, + "learning_rate": 1.6510052367524808e-07, + "loss": 0.6915, + "step": 9580 + }, + { + "epoch": 0.35, + "learning_rate": 1.6509686161057604e-07, + "loss": 0.6837, + "step": 9581 + }, + { + "epoch": 0.35, + "learning_rate": 1.6509319954590396e-07, + "loss": 0.6793, + "step": 9582 + }, + { + "epoch": 0.35, + "learning_rate": 1.6508953748123191e-07, + "loss": 0.6811, + "step": 9583 + }, + { + "epoch": 0.35, + "learning_rate": 1.6508587541655984e-07, + "loss": 0.7032, + "step": 9584 + }, + { + "epoch": 0.35, + "learning_rate": 1.650822133518878e-07, + "loss": 0.6929, + "step": 9585 + }, + { + "epoch": 0.35, + "learning_rate": 1.6507855128721572e-07, + "loss": 0.6929, + "step": 9586 + }, + { + "epoch": 0.35, + "learning_rate": 1.6507488922254367e-07, + "loss": 0.698, + "step": 9587 + }, + { + "epoch": 0.35, + "learning_rate": 1.650712271578716e-07, + "loss": 0.707, + "step": 9588 + }, + { + "epoch": 0.35, + "learning_rate": 1.6506756509319955e-07, + "loss": 0.6844, + "step": 9589 + }, + { + "epoch": 0.35, + "learning_rate": 1.6506390302852748e-07, + "loss": 0.7217, + "step": 9590 + }, + { + "epoch": 0.35, + "learning_rate": 1.6506024096385543e-07, + "loss": 0.679, + "step": 9591 + }, + { + "epoch": 0.35, + "learning_rate": 1.6505657889918336e-07, + "loss": 0.6997, + "step": 9592 + }, + { + "epoch": 0.35, + "learning_rate": 1.650529168345113e-07, + "loss": 0.6908, + "step": 9593 + }, + { + "epoch": 0.35, + "learning_rate": 1.6504925476983921e-07, + "loss": 0.6761, + "step": 9594 + }, + { + "epoch": 0.35, + "learning_rate": 1.6504559270516717e-07, + "loss": 0.6953, + "step": 9595 + }, + { + "epoch": 0.35, + "learning_rate": 1.650419306404951e-07, + "loss": 0.6902, + "step": 9596 + }, + { + "epoch": 0.35, + "learning_rate": 1.6503826857582305e-07, + "loss": 0.6863, + "step": 9597 + }, + { + "epoch": 0.35, + "learning_rate": 1.6503460651115097e-07, + "loss": 0.6911, + "step": 9598 + }, + { + "epoch": 0.35, + "learning_rate": 1.6503094444647893e-07, + "loss": 0.6919, + "step": 9599 + }, + { + "epoch": 0.35, + "learning_rate": 1.6502728238180685e-07, + "loss": 0.7098, + "step": 9600 + }, + { + "epoch": 0.35, + "learning_rate": 1.6502362031713478e-07, + "loss": 0.6994, + "step": 9601 + }, + { + "epoch": 0.35, + "learning_rate": 1.6501995825246273e-07, + "loss": 0.6872, + "step": 9602 + }, + { + "epoch": 0.35, + "learning_rate": 1.6501629618779066e-07, + "loss": 0.6971, + "step": 9603 + }, + { + "epoch": 0.35, + "learning_rate": 1.6501263412311861e-07, + "loss": 0.6825, + "step": 9604 + }, + { + "epoch": 0.35, + "learning_rate": 1.6500897205844654e-07, + "loss": 0.691, + "step": 9605 + }, + { + "epoch": 0.35, + "learning_rate": 1.650053099937745e-07, + "loss": 0.6911, + "step": 9606 + }, + { + "epoch": 0.35, + "learning_rate": 1.6500164792910242e-07, + "loss": 0.6878, + "step": 9607 + }, + { + "epoch": 0.35, + "learning_rate": 1.6499798586443037e-07, + "loss": 0.6981, + "step": 9608 + }, + { + "epoch": 0.35, + "learning_rate": 1.649943237997583e-07, + "loss": 0.6916, + "step": 9609 + }, + { + "epoch": 0.35, + "learning_rate": 1.6499066173508625e-07, + "loss": 0.689, + "step": 9610 + }, + { + "epoch": 0.35, + "learning_rate": 1.6498699967041415e-07, + "loss": 0.6831, + "step": 9611 + }, + { + "epoch": 0.35, + "learning_rate": 1.649833376057421e-07, + "loss": 0.7036, + "step": 9612 + }, + { + "epoch": 0.35, + "learning_rate": 1.6497967554107003e-07, + "loss": 0.6831, + "step": 9613 + }, + { + "epoch": 0.35, + "learning_rate": 1.6497601347639799e-07, + "loss": 0.7005, + "step": 9614 + }, + { + "epoch": 0.35, + "learning_rate": 1.649723514117259e-07, + "loss": 0.6858, + "step": 9615 + }, + { + "epoch": 0.35, + "learning_rate": 1.6496868934705387e-07, + "loss": 0.7066, + "step": 9616 + }, + { + "epoch": 0.35, + "learning_rate": 1.649650272823818e-07, + "loss": 0.6855, + "step": 9617 + }, + { + "epoch": 0.35, + "learning_rate": 1.6496136521770975e-07, + "loss": 0.6798, + "step": 9618 + }, + { + "epoch": 0.35, + "learning_rate": 1.6495770315303767e-07, + "loss": 0.7032, + "step": 9619 + }, + { + "epoch": 0.35, + "learning_rate": 1.6495404108836563e-07, + "loss": 0.7064, + "step": 9620 + }, + { + "epoch": 0.35, + "learning_rate": 1.6495037902369355e-07, + "loss": 0.7069, + "step": 9621 + }, + { + "epoch": 0.35, + "learning_rate": 1.6494671695902148e-07, + "loss": 0.7078, + "step": 9622 + }, + { + "epoch": 0.35, + "learning_rate": 1.6494305489434943e-07, + "loss": 0.681, + "step": 9623 + }, + { + "epoch": 0.35, + "learning_rate": 1.6493939282967736e-07, + "loss": 0.6728, + "step": 9624 + }, + { + "epoch": 0.35, + "learning_rate": 1.649357307650053e-07, + "loss": 0.6906, + "step": 9625 + }, + { + "epoch": 0.35, + "learning_rate": 1.6493206870033324e-07, + "loss": 0.6966, + "step": 9626 + }, + { + "epoch": 0.35, + "learning_rate": 1.649284066356612e-07, + "loss": 0.7016, + "step": 9627 + }, + { + "epoch": 0.35, + "learning_rate": 1.6492474457098912e-07, + "loss": 0.6924, + "step": 9628 + }, + { + "epoch": 0.35, + "learning_rate": 1.6492108250631705e-07, + "loss": 0.6988, + "step": 9629 + }, + { + "epoch": 0.35, + "learning_rate": 1.6491742044164497e-07, + "loss": 0.6891, + "step": 9630 + }, + { + "epoch": 0.35, + "learning_rate": 1.6491375837697293e-07, + "loss": 0.7029, + "step": 9631 + }, + { + "epoch": 0.35, + "learning_rate": 1.6491009631230085e-07, + "loss": 0.7083, + "step": 9632 + }, + { + "epoch": 0.35, + "learning_rate": 1.649064342476288e-07, + "loss": 0.7008, + "step": 9633 + }, + { + "epoch": 0.35, + "learning_rate": 1.6490277218295673e-07, + "loss": 0.6858, + "step": 9634 + }, + { + "epoch": 0.35, + "learning_rate": 1.6489911011828469e-07, + "loss": 0.6996, + "step": 9635 + }, + { + "epoch": 0.35, + "learning_rate": 1.648954480536126e-07, + "loss": 0.6831, + "step": 9636 + }, + { + "epoch": 0.35, + "learning_rate": 1.6489178598894057e-07, + "loss": 0.6833, + "step": 9637 + }, + { + "epoch": 0.35, + "learning_rate": 1.648881239242685e-07, + "loss": 0.6911, + "step": 9638 + }, + { + "epoch": 0.35, + "learning_rate": 1.6488446185959645e-07, + "loss": 0.6916, + "step": 9639 + }, + { + "epoch": 0.35, + "learning_rate": 1.6488079979492437e-07, + "loss": 0.7062, + "step": 9640 + }, + { + "epoch": 0.35, + "learning_rate": 1.6487713773025232e-07, + "loss": 0.6899, + "step": 9641 + }, + { + "epoch": 0.35, + "learning_rate": 1.6487347566558025e-07, + "loss": 0.6921, + "step": 9642 + }, + { + "epoch": 0.35, + "learning_rate": 1.648698136009082e-07, + "loss": 0.6862, + "step": 9643 + }, + { + "epoch": 0.35, + "learning_rate": 1.6486615153623613e-07, + "loss": 0.6945, + "step": 9644 + }, + { + "epoch": 0.35, + "learning_rate": 1.6486248947156406e-07, + "loss": 0.7072, + "step": 9645 + }, + { + "epoch": 0.35, + "learning_rate": 1.64858827406892e-07, + "loss": 0.6859, + "step": 9646 + }, + { + "epoch": 0.35, + "learning_rate": 1.6485516534221994e-07, + "loss": 0.7109, + "step": 9647 + }, + { + "epoch": 0.35, + "learning_rate": 1.6485150327754786e-07, + "loss": 0.6952, + "step": 9648 + }, + { + "epoch": 0.35, + "learning_rate": 1.648478412128758e-07, + "loss": 0.6992, + "step": 9649 + }, + { + "epoch": 0.35, + "learning_rate": 1.6484417914820374e-07, + "loss": 0.7022, + "step": 9650 + }, + { + "epoch": 0.35, + "learning_rate": 1.6484051708353167e-07, + "loss": 0.6876, + "step": 9651 + }, + { + "epoch": 0.35, + "learning_rate": 1.6483685501885962e-07, + "loss": 0.6764, + "step": 9652 + }, + { + "epoch": 0.35, + "learning_rate": 1.6483319295418755e-07, + "loss": 0.6948, + "step": 9653 + }, + { + "epoch": 0.35, + "learning_rate": 1.648295308895155e-07, + "loss": 0.6873, + "step": 9654 + }, + { + "epoch": 0.35, + "learning_rate": 1.6482586882484343e-07, + "loss": 0.7038, + "step": 9655 + }, + { + "epoch": 0.35, + "learning_rate": 1.6482220676017138e-07, + "loss": 0.6961, + "step": 9656 + }, + { + "epoch": 0.35, + "learning_rate": 1.648185446954993e-07, + "loss": 0.6843, + "step": 9657 + }, + { + "epoch": 0.35, + "learning_rate": 1.6481488263082726e-07, + "loss": 0.7006, + "step": 9658 + }, + { + "epoch": 0.35, + "learning_rate": 1.648112205661552e-07, + "loss": 0.679, + "step": 9659 + }, + { + "epoch": 0.35, + "learning_rate": 1.6480755850148314e-07, + "loss": 0.7016, + "step": 9660 + }, + { + "epoch": 0.35, + "learning_rate": 1.6480389643681107e-07, + "loss": 0.6967, + "step": 9661 + }, + { + "epoch": 0.35, + "learning_rate": 1.6480023437213902e-07, + "loss": 0.6845, + "step": 9662 + }, + { + "epoch": 0.35, + "learning_rate": 1.6479657230746695e-07, + "loss": 0.6818, + "step": 9663 + }, + { + "epoch": 0.35, + "learning_rate": 1.647929102427949e-07, + "loss": 0.6932, + "step": 9664 + }, + { + "epoch": 0.35, + "learning_rate": 1.647892481781228e-07, + "loss": 0.6848, + "step": 9665 + }, + { + "epoch": 0.35, + "learning_rate": 1.6478558611345076e-07, + "loss": 0.696, + "step": 9666 + }, + { + "epoch": 0.35, + "learning_rate": 1.6478192404877868e-07, + "loss": 0.6952, + "step": 9667 + }, + { + "epoch": 0.35, + "learning_rate": 1.6477826198410664e-07, + "loss": 0.7148, + "step": 9668 + }, + { + "epoch": 0.35, + "learning_rate": 1.6477459991943456e-07, + "loss": 0.687, + "step": 9669 + }, + { + "epoch": 0.35, + "learning_rate": 1.6477093785476252e-07, + "loss": 0.6929, + "step": 9670 + }, + { + "epoch": 0.35, + "learning_rate": 1.6476727579009044e-07, + "loss": 0.695, + "step": 9671 + }, + { + "epoch": 0.35, + "learning_rate": 1.6476361372541837e-07, + "loss": 0.691, + "step": 9672 + }, + { + "epoch": 0.35, + "learning_rate": 1.6475995166074632e-07, + "loss": 0.6953, + "step": 9673 + }, + { + "epoch": 0.35, + "learning_rate": 1.6475628959607425e-07, + "loss": 0.6776, + "step": 9674 + }, + { + "epoch": 0.35, + "learning_rate": 1.647526275314022e-07, + "loss": 0.7022, + "step": 9675 + }, + { + "epoch": 0.35, + "learning_rate": 1.6474896546673013e-07, + "loss": 0.6992, + "step": 9676 + }, + { + "epoch": 0.35, + "learning_rate": 1.6474530340205808e-07, + "loss": 0.6894, + "step": 9677 + }, + { + "epoch": 0.35, + "learning_rate": 1.64741641337386e-07, + "loss": 0.6929, + "step": 9678 + }, + { + "epoch": 0.35, + "learning_rate": 1.6473797927271396e-07, + "loss": 0.6901, + "step": 9679 + }, + { + "epoch": 0.35, + "learning_rate": 1.647343172080419e-07, + "loss": 0.7036, + "step": 9680 + }, + { + "epoch": 0.35, + "learning_rate": 1.6473065514336984e-07, + "loss": 0.6881, + "step": 9681 + }, + { + "epoch": 0.35, + "learning_rate": 1.6472699307869774e-07, + "loss": 0.6971, + "step": 9682 + }, + { + "epoch": 0.35, + "learning_rate": 1.647233310140257e-07, + "loss": 0.699, + "step": 9683 + }, + { + "epoch": 0.35, + "learning_rate": 1.6471966894935362e-07, + "loss": 0.6919, + "step": 9684 + }, + { + "epoch": 0.35, + "learning_rate": 1.6471600688468158e-07, + "loss": 0.6978, + "step": 9685 + }, + { + "epoch": 0.35, + "learning_rate": 1.647123448200095e-07, + "loss": 0.7049, + "step": 9686 + }, + { + "epoch": 0.35, + "learning_rate": 1.6470868275533746e-07, + "loss": 0.6998, + "step": 9687 + }, + { + "epoch": 0.35, + "learning_rate": 1.6470502069066538e-07, + "loss": 0.7055, + "step": 9688 + }, + { + "epoch": 0.35, + "learning_rate": 1.6470135862599334e-07, + "loss": 0.7, + "step": 9689 + }, + { + "epoch": 0.35, + "learning_rate": 1.6469769656132126e-07, + "loss": 0.6844, + "step": 9690 + }, + { + "epoch": 0.35, + "learning_rate": 1.6469403449664922e-07, + "loss": 0.6846, + "step": 9691 + }, + { + "epoch": 0.35, + "learning_rate": 1.6469037243197714e-07, + "loss": 0.6984, + "step": 9692 + }, + { + "epoch": 0.35, + "learning_rate": 1.6468671036730507e-07, + "loss": 0.6955, + "step": 9693 + }, + { + "epoch": 0.35, + "learning_rate": 1.6468304830263302e-07, + "loss": 0.691, + "step": 9694 + }, + { + "epoch": 0.35, + "learning_rate": 1.6467938623796095e-07, + "loss": 0.6986, + "step": 9695 + }, + { + "epoch": 0.35, + "learning_rate": 1.646757241732889e-07, + "loss": 0.6923, + "step": 9696 + }, + { + "epoch": 0.35, + "learning_rate": 1.6467206210861683e-07, + "loss": 0.6946, + "step": 9697 + }, + { + "epoch": 0.35, + "learning_rate": 1.6466840004394478e-07, + "loss": 0.7026, + "step": 9698 + }, + { + "epoch": 0.35, + "learning_rate": 1.646647379792727e-07, + "loss": 0.6757, + "step": 9699 + }, + { + "epoch": 0.35, + "learning_rate": 1.6466107591460064e-07, + "loss": 0.7103, + "step": 9700 + }, + { + "epoch": 0.35, + "learning_rate": 1.6465741384992856e-07, + "loss": 0.6868, + "step": 9701 + }, + { + "epoch": 0.35, + "learning_rate": 1.6465375178525651e-07, + "loss": 0.7169, + "step": 9702 + }, + { + "epoch": 0.36, + "learning_rate": 1.6465008972058444e-07, + "loss": 0.6886, + "step": 9703 + }, + { + "epoch": 0.36, + "learning_rate": 1.646464276559124e-07, + "loss": 0.6841, + "step": 9704 + }, + { + "epoch": 0.36, + "learning_rate": 1.6464276559124032e-07, + "loss": 0.7061, + "step": 9705 + }, + { + "epoch": 0.36, + "learning_rate": 1.6463910352656827e-07, + "loss": 0.6962, + "step": 9706 + }, + { + "epoch": 0.36, + "learning_rate": 1.646354414618962e-07, + "loss": 0.6939, + "step": 9707 + }, + { + "epoch": 0.36, + "learning_rate": 1.6463177939722415e-07, + "loss": 0.6914, + "step": 9708 + }, + { + "epoch": 0.36, + "learning_rate": 1.6462811733255208e-07, + "loss": 0.6823, + "step": 9709 + }, + { + "epoch": 0.36, + "learning_rate": 1.6462445526788003e-07, + "loss": 0.7046, + "step": 9710 + }, + { + "epoch": 0.36, + "learning_rate": 1.6462079320320796e-07, + "loss": 0.7051, + "step": 9711 + }, + { + "epoch": 0.36, + "learning_rate": 1.6461713113853591e-07, + "loss": 0.683, + "step": 9712 + }, + { + "epoch": 0.36, + "learning_rate": 1.6461346907386384e-07, + "loss": 0.6926, + "step": 9713 + }, + { + "epoch": 0.36, + "learning_rate": 1.646098070091918e-07, + "loss": 0.7039, + "step": 9714 + }, + { + "epoch": 0.36, + "learning_rate": 1.6460614494451972e-07, + "loss": 0.6967, + "step": 9715 + }, + { + "epoch": 0.36, + "learning_rate": 1.6460248287984765e-07, + "loss": 0.6973, + "step": 9716 + }, + { + "epoch": 0.36, + "learning_rate": 1.645988208151756e-07, + "loss": 0.6936, + "step": 9717 + }, + { + "epoch": 0.36, + "learning_rate": 1.6459515875050353e-07, + "loss": 0.6969, + "step": 9718 + }, + { + "epoch": 0.36, + "learning_rate": 1.6459149668583145e-07, + "loss": 0.6923, + "step": 9719 + }, + { + "epoch": 0.36, + "learning_rate": 1.6458783462115938e-07, + "loss": 0.6795, + "step": 9720 + }, + { + "epoch": 0.36, + "learning_rate": 1.6458417255648733e-07, + "loss": 0.7054, + "step": 9721 + }, + { + "epoch": 0.36, + "learning_rate": 1.6458051049181526e-07, + "loss": 0.6871, + "step": 9722 + }, + { + "epoch": 0.36, + "learning_rate": 1.6457684842714321e-07, + "loss": 0.6914, + "step": 9723 + }, + { + "epoch": 0.36, + "learning_rate": 1.6457318636247114e-07, + "loss": 0.6917, + "step": 9724 + }, + { + "epoch": 0.36, + "learning_rate": 1.645695242977991e-07, + "loss": 0.6943, + "step": 9725 + }, + { + "epoch": 0.36, + "learning_rate": 1.6456586223312702e-07, + "loss": 0.7006, + "step": 9726 + }, + { + "epoch": 0.36, + "learning_rate": 1.6456220016845497e-07, + "loss": 0.6936, + "step": 9727 + }, + { + "epoch": 0.36, + "learning_rate": 1.645585381037829e-07, + "loss": 0.7029, + "step": 9728 + }, + { + "epoch": 0.36, + "learning_rate": 1.6455487603911085e-07, + "loss": 0.693, + "step": 9729 + }, + { + "epoch": 0.36, + "learning_rate": 1.6455121397443878e-07, + "loss": 0.6992, + "step": 9730 + }, + { + "epoch": 0.36, + "learning_rate": 1.6454755190976673e-07, + "loss": 0.6964, + "step": 9731 + }, + { + "epoch": 0.36, + "learning_rate": 1.6454388984509466e-07, + "loss": 0.7046, + "step": 9732 + }, + { + "epoch": 0.36, + "learning_rate": 1.645402277804226e-07, + "loss": 0.6885, + "step": 9733 + }, + { + "epoch": 0.36, + "learning_rate": 1.6453656571575054e-07, + "loss": 0.696, + "step": 9734 + }, + { + "epoch": 0.36, + "learning_rate": 1.645329036510785e-07, + "loss": 0.6875, + "step": 9735 + }, + { + "epoch": 0.36, + "learning_rate": 1.645292415864064e-07, + "loss": 0.7001, + "step": 9736 + }, + { + "epoch": 0.36, + "learning_rate": 1.6452557952173435e-07, + "loss": 0.6939, + "step": 9737 + }, + { + "epoch": 0.36, + "learning_rate": 1.6452191745706227e-07, + "loss": 0.6963, + "step": 9738 + }, + { + "epoch": 0.36, + "learning_rate": 1.6451825539239023e-07, + "loss": 0.6947, + "step": 9739 + }, + { + "epoch": 0.36, + "learning_rate": 1.6451459332771815e-07, + "loss": 0.6882, + "step": 9740 + }, + { + "epoch": 0.36, + "learning_rate": 1.645109312630461e-07, + "loss": 0.6968, + "step": 9741 + }, + { + "epoch": 0.36, + "learning_rate": 1.6450726919837403e-07, + "loss": 0.7059, + "step": 9742 + }, + { + "epoch": 0.36, + "learning_rate": 1.6450360713370196e-07, + "loss": 0.6861, + "step": 9743 + }, + { + "epoch": 0.36, + "learning_rate": 1.644999450690299e-07, + "loss": 0.6807, + "step": 9744 + }, + { + "epoch": 0.36, + "learning_rate": 1.6449628300435784e-07, + "loss": 0.7137, + "step": 9745 + }, + { + "epoch": 0.36, + "learning_rate": 1.644926209396858e-07, + "loss": 0.6958, + "step": 9746 + }, + { + "epoch": 0.36, + "learning_rate": 1.6448895887501372e-07, + "loss": 0.6993, + "step": 9747 + }, + { + "epoch": 0.36, + "learning_rate": 1.6448529681034167e-07, + "loss": 0.6965, + "step": 9748 + }, + { + "epoch": 0.36, + "learning_rate": 1.644816347456696e-07, + "loss": 0.6939, + "step": 9749 + }, + { + "epoch": 0.36, + "learning_rate": 1.6447797268099755e-07, + "loss": 0.7046, + "step": 9750 + }, + { + "epoch": 0.36, + "learning_rate": 1.6447431061632548e-07, + "loss": 0.6971, + "step": 9751 + }, + { + "epoch": 0.36, + "learning_rate": 1.6447064855165343e-07, + "loss": 0.6945, + "step": 9752 + }, + { + "epoch": 0.36, + "learning_rate": 1.6446698648698136e-07, + "loss": 0.706, + "step": 9753 + }, + { + "epoch": 0.36, + "learning_rate": 1.6446332442230929e-07, + "loss": 0.6716, + "step": 9754 + }, + { + "epoch": 0.36, + "learning_rate": 1.644596623576372e-07, + "loss": 0.6844, + "step": 9755 + }, + { + "epoch": 0.36, + "learning_rate": 1.6445600029296517e-07, + "loss": 0.6891, + "step": 9756 + }, + { + "epoch": 0.36, + "learning_rate": 1.644523382282931e-07, + "loss": 0.7018, + "step": 9757 + }, + { + "epoch": 0.36, + "learning_rate": 1.6444867616362105e-07, + "loss": 0.6918, + "step": 9758 + }, + { + "epoch": 0.36, + "learning_rate": 1.6444501409894897e-07, + "loss": 0.7049, + "step": 9759 + }, + { + "epoch": 0.36, + "learning_rate": 1.6444135203427692e-07, + "loss": 0.6891, + "step": 9760 + }, + { + "epoch": 0.36, + "learning_rate": 1.6443768996960485e-07, + "loss": 0.6833, + "step": 9761 + }, + { + "epoch": 0.36, + "learning_rate": 1.644340279049328e-07, + "loss": 0.6938, + "step": 9762 + }, + { + "epoch": 0.36, + "learning_rate": 1.6443036584026073e-07, + "loss": 0.6986, + "step": 9763 + }, + { + "epoch": 0.36, + "learning_rate": 1.6442670377558866e-07, + "loss": 0.691, + "step": 9764 + }, + { + "epoch": 0.36, + "learning_rate": 1.644230417109166e-07, + "loss": 0.6886, + "step": 9765 + }, + { + "epoch": 0.36, + "learning_rate": 1.6441937964624454e-07, + "loss": 0.683, + "step": 9766 + }, + { + "epoch": 0.36, + "learning_rate": 1.644157175815725e-07, + "loss": 0.6944, + "step": 9767 + }, + { + "epoch": 0.36, + "learning_rate": 1.6441205551690042e-07, + "loss": 0.6985, + "step": 9768 + }, + { + "epoch": 0.36, + "learning_rate": 1.6440839345222837e-07, + "loss": 0.6802, + "step": 9769 + }, + { + "epoch": 0.36, + "learning_rate": 1.644047313875563e-07, + "loss": 0.6829, + "step": 9770 + }, + { + "epoch": 0.36, + "learning_rate": 1.6440106932288425e-07, + "loss": 0.6903, + "step": 9771 + }, + { + "epoch": 0.36, + "learning_rate": 1.6439740725821215e-07, + "loss": 0.6893, + "step": 9772 + }, + { + "epoch": 0.36, + "learning_rate": 1.643937451935401e-07, + "loss": 0.6984, + "step": 9773 + }, + { + "epoch": 0.36, + "learning_rate": 1.6439008312886803e-07, + "loss": 0.6872, + "step": 9774 + }, + { + "epoch": 0.36, + "learning_rate": 1.6438642106419598e-07, + "loss": 0.6922, + "step": 9775 + }, + { + "epoch": 0.36, + "learning_rate": 1.643827589995239e-07, + "loss": 0.685, + "step": 9776 + }, + { + "epoch": 0.36, + "learning_rate": 1.6437909693485186e-07, + "loss": 0.6833, + "step": 9777 + }, + { + "epoch": 0.36, + "learning_rate": 1.643754348701798e-07, + "loss": 0.6993, + "step": 9778 + }, + { + "epoch": 0.36, + "learning_rate": 1.6437177280550774e-07, + "loss": 0.6937, + "step": 9779 + }, + { + "epoch": 0.36, + "learning_rate": 1.6436811074083567e-07, + "loss": 0.6997, + "step": 9780 + }, + { + "epoch": 0.36, + "learning_rate": 1.6436444867616362e-07, + "loss": 0.7026, + "step": 9781 + }, + { + "epoch": 0.36, + "learning_rate": 1.6436078661149155e-07, + "loss": 0.6948, + "step": 9782 + }, + { + "epoch": 0.36, + "learning_rate": 1.643571245468195e-07, + "loss": 0.6984, + "step": 9783 + }, + { + "epoch": 0.36, + "learning_rate": 1.6435346248214743e-07, + "loss": 0.6893, + "step": 9784 + }, + { + "epoch": 0.36, + "learning_rate": 1.6434980041747538e-07, + "loss": 0.6925, + "step": 9785 + }, + { + "epoch": 0.36, + "learning_rate": 1.643461383528033e-07, + "loss": 0.6826, + "step": 9786 + }, + { + "epoch": 0.36, + "learning_rate": 1.6434247628813124e-07, + "loss": 0.6861, + "step": 9787 + }, + { + "epoch": 0.36, + "learning_rate": 1.643388142234592e-07, + "loss": 0.6912, + "step": 9788 + }, + { + "epoch": 0.36, + "learning_rate": 1.6433515215878712e-07, + "loss": 0.6891, + "step": 9789 + }, + { + "epoch": 0.36, + "learning_rate": 1.6433149009411504e-07, + "loss": 0.6975, + "step": 9790 + }, + { + "epoch": 0.36, + "learning_rate": 1.6432782802944297e-07, + "loss": 0.6925, + "step": 9791 + }, + { + "epoch": 0.36, + "learning_rate": 1.6432416596477092e-07, + "loss": 0.6852, + "step": 9792 + }, + { + "epoch": 0.36, + "learning_rate": 1.6432050390009885e-07, + "loss": 0.7009, + "step": 9793 + }, + { + "epoch": 0.36, + "learning_rate": 1.643168418354268e-07, + "loss": 0.6942, + "step": 9794 + }, + { + "epoch": 0.36, + "learning_rate": 1.6431317977075473e-07, + "loss": 0.6859, + "step": 9795 + }, + { + "epoch": 0.36, + "learning_rate": 1.6430951770608268e-07, + "loss": 0.709, + "step": 9796 + }, + { + "epoch": 0.36, + "learning_rate": 1.643058556414106e-07, + "loss": 0.6838, + "step": 9797 + }, + { + "epoch": 0.36, + "learning_rate": 1.6430219357673856e-07, + "loss": 0.691, + "step": 9798 + }, + { + "epoch": 0.36, + "learning_rate": 1.642985315120665e-07, + "loss": 0.6957, + "step": 9799 + }, + { + "epoch": 0.36, + "learning_rate": 1.6429486944739444e-07, + "loss": 0.6989, + "step": 9800 + }, + { + "epoch": 0.36, + "learning_rate": 1.6429120738272237e-07, + "loss": 0.6897, + "step": 9801 + }, + { + "epoch": 0.36, + "learning_rate": 1.6428754531805032e-07, + "loss": 0.6953, + "step": 9802 + }, + { + "epoch": 0.36, + "learning_rate": 1.6428388325337825e-07, + "loss": 0.6976, + "step": 9803 + }, + { + "epoch": 0.36, + "learning_rate": 1.642802211887062e-07, + "loss": 0.6896, + "step": 9804 + }, + { + "epoch": 0.36, + "learning_rate": 1.6427655912403413e-07, + "loss": 0.6982, + "step": 9805 + }, + { + "epoch": 0.36, + "learning_rate": 1.6427289705936208e-07, + "loss": 0.6903, + "step": 9806 + }, + { + "epoch": 0.36, + "learning_rate": 1.6426923499469e-07, + "loss": 0.7052, + "step": 9807 + }, + { + "epoch": 0.36, + "learning_rate": 1.6426557293001794e-07, + "loss": 0.6946, + "step": 9808 + }, + { + "epoch": 0.36, + "learning_rate": 1.6426191086534586e-07, + "loss": 0.6966, + "step": 9809 + }, + { + "epoch": 0.36, + "learning_rate": 1.6425824880067382e-07, + "loss": 0.6865, + "step": 9810 + }, + { + "epoch": 0.36, + "learning_rate": 1.6425458673600174e-07, + "loss": 0.6888, + "step": 9811 + }, + { + "epoch": 0.36, + "learning_rate": 1.642509246713297e-07, + "loss": 0.7019, + "step": 9812 + }, + { + "epoch": 0.36, + "learning_rate": 1.6424726260665762e-07, + "loss": 0.7022, + "step": 9813 + }, + { + "epoch": 0.36, + "learning_rate": 1.6424360054198555e-07, + "loss": 0.6998, + "step": 9814 + }, + { + "epoch": 0.36, + "learning_rate": 1.642399384773135e-07, + "loss": 0.7008, + "step": 9815 + }, + { + "epoch": 0.36, + "learning_rate": 1.6423627641264143e-07, + "loss": 0.6923, + "step": 9816 + }, + { + "epoch": 0.36, + "learning_rate": 1.6423261434796938e-07, + "loss": 0.6918, + "step": 9817 + }, + { + "epoch": 0.36, + "learning_rate": 1.642289522832973e-07, + "loss": 0.685, + "step": 9818 + }, + { + "epoch": 0.36, + "learning_rate": 1.6422529021862526e-07, + "loss": 0.6785, + "step": 9819 + }, + { + "epoch": 0.36, + "learning_rate": 1.642216281539532e-07, + "loss": 0.7018, + "step": 9820 + }, + { + "epoch": 0.36, + "learning_rate": 1.6421796608928114e-07, + "loss": 0.6801, + "step": 9821 + }, + { + "epoch": 0.36, + "learning_rate": 1.6421430402460907e-07, + "loss": 0.6801, + "step": 9822 + }, + { + "epoch": 0.36, + "learning_rate": 1.6421064195993702e-07, + "loss": 0.693, + "step": 9823 + }, + { + "epoch": 0.36, + "learning_rate": 1.6420697989526495e-07, + "loss": 0.7072, + "step": 9824 + }, + { + "epoch": 0.36, + "learning_rate": 1.642033178305929e-07, + "loss": 0.688, + "step": 9825 + }, + { + "epoch": 0.36, + "learning_rate": 1.641996557659208e-07, + "loss": 0.6891, + "step": 9826 + }, + { + "epoch": 0.36, + "learning_rate": 1.6419599370124875e-07, + "loss": 0.6891, + "step": 9827 + }, + { + "epoch": 0.36, + "learning_rate": 1.6419233163657668e-07, + "loss": 0.6946, + "step": 9828 + }, + { + "epoch": 0.36, + "learning_rate": 1.6418866957190463e-07, + "loss": 0.6921, + "step": 9829 + }, + { + "epoch": 0.36, + "learning_rate": 1.6418500750723256e-07, + "loss": 0.6899, + "step": 9830 + }, + { + "epoch": 0.36, + "learning_rate": 1.6418134544256051e-07, + "loss": 0.6903, + "step": 9831 + }, + { + "epoch": 0.36, + "learning_rate": 1.6417768337788844e-07, + "loss": 0.6885, + "step": 9832 + }, + { + "epoch": 0.36, + "learning_rate": 1.641740213132164e-07, + "loss": 0.7022, + "step": 9833 + }, + { + "epoch": 0.36, + "learning_rate": 1.6417035924854432e-07, + "loss": 0.7066, + "step": 9834 + }, + { + "epoch": 0.36, + "learning_rate": 1.6416669718387225e-07, + "loss": 0.694, + "step": 9835 + }, + { + "epoch": 0.36, + "learning_rate": 1.641630351192002e-07, + "loss": 0.6862, + "step": 9836 + }, + { + "epoch": 0.36, + "learning_rate": 1.6415937305452813e-07, + "loss": 0.687, + "step": 9837 + }, + { + "epoch": 0.36, + "learning_rate": 1.6415571098985608e-07, + "loss": 0.6893, + "step": 9838 + }, + { + "epoch": 0.36, + "learning_rate": 1.64152048925184e-07, + "loss": 0.7093, + "step": 9839 + }, + { + "epoch": 0.36, + "learning_rate": 1.6414838686051196e-07, + "loss": 0.6932, + "step": 9840 + }, + { + "epoch": 0.36, + "learning_rate": 1.641447247958399e-07, + "loss": 0.6972, + "step": 9841 + }, + { + "epoch": 0.36, + "learning_rate": 1.6414106273116784e-07, + "loss": 0.6917, + "step": 9842 + }, + { + "epoch": 0.36, + "learning_rate": 1.6413740066649577e-07, + "loss": 0.684, + "step": 9843 + }, + { + "epoch": 0.36, + "learning_rate": 1.641337386018237e-07, + "loss": 0.6866, + "step": 9844 + }, + { + "epoch": 0.36, + "learning_rate": 1.6413007653715162e-07, + "loss": 0.6986, + "step": 9845 + }, + { + "epoch": 0.36, + "learning_rate": 1.6412641447247957e-07, + "loss": 0.6906, + "step": 9846 + }, + { + "epoch": 0.36, + "learning_rate": 1.641227524078075e-07, + "loss": 0.7045, + "step": 9847 + }, + { + "epoch": 0.36, + "learning_rate": 1.6411909034313545e-07, + "loss": 0.6895, + "step": 9848 + }, + { + "epoch": 0.36, + "learning_rate": 1.6411542827846338e-07, + "loss": 0.7001, + "step": 9849 + }, + { + "epoch": 0.36, + "learning_rate": 1.6411176621379133e-07, + "loss": 0.6969, + "step": 9850 + }, + { + "epoch": 0.36, + "learning_rate": 1.6410810414911926e-07, + "loss": 0.6977, + "step": 9851 + }, + { + "epoch": 0.36, + "learning_rate": 1.641044420844472e-07, + "loss": 0.7124, + "step": 9852 + }, + { + "epoch": 0.36, + "learning_rate": 1.6410078001977514e-07, + "loss": 0.6725, + "step": 9853 + }, + { + "epoch": 0.36, + "learning_rate": 1.640971179551031e-07, + "loss": 0.6763, + "step": 9854 + }, + { + "epoch": 0.36, + "learning_rate": 1.6409345589043102e-07, + "loss": 0.6942, + "step": 9855 + }, + { + "epoch": 0.36, + "learning_rate": 1.6408979382575897e-07, + "loss": 0.6806, + "step": 9856 + }, + { + "epoch": 0.36, + "learning_rate": 1.640861317610869e-07, + "loss": 0.6873, + "step": 9857 + }, + { + "epoch": 0.36, + "learning_rate": 1.6408246969641483e-07, + "loss": 0.6961, + "step": 9858 + }, + { + "epoch": 0.36, + "learning_rate": 1.6407880763174278e-07, + "loss": 0.6958, + "step": 9859 + }, + { + "epoch": 0.36, + "learning_rate": 1.640751455670707e-07, + "loss": 0.6881, + "step": 9860 + }, + { + "epoch": 0.36, + "learning_rate": 1.6407148350239866e-07, + "loss": 0.7016, + "step": 9861 + }, + { + "epoch": 0.36, + "learning_rate": 1.6406782143772656e-07, + "loss": 0.6922, + "step": 9862 + }, + { + "epoch": 0.36, + "learning_rate": 1.640641593730545e-07, + "loss": 0.6918, + "step": 9863 + }, + { + "epoch": 0.36, + "learning_rate": 1.6406049730838244e-07, + "loss": 0.6861, + "step": 9864 + }, + { + "epoch": 0.36, + "learning_rate": 1.640568352437104e-07, + "loss": 0.6939, + "step": 9865 + }, + { + "epoch": 0.36, + "learning_rate": 1.6405317317903832e-07, + "loss": 0.6905, + "step": 9866 + }, + { + "epoch": 0.36, + "learning_rate": 1.6404951111436627e-07, + "loss": 0.6994, + "step": 9867 + }, + { + "epoch": 0.36, + "learning_rate": 1.640458490496942e-07, + "loss": 0.6916, + "step": 9868 + }, + { + "epoch": 0.36, + "learning_rate": 1.6404218698502215e-07, + "loss": 0.6956, + "step": 9869 + }, + { + "epoch": 0.36, + "learning_rate": 1.6403852492035008e-07, + "loss": 0.6916, + "step": 9870 + }, + { + "epoch": 0.36, + "learning_rate": 1.6403486285567803e-07, + "loss": 0.692, + "step": 9871 + }, + { + "epoch": 0.36, + "learning_rate": 1.6403120079100596e-07, + "loss": 0.6877, + "step": 9872 + }, + { + "epoch": 0.36, + "learning_rate": 1.640275387263339e-07, + "loss": 0.6884, + "step": 9873 + }, + { + "epoch": 0.36, + "learning_rate": 1.6402387666166184e-07, + "loss": 0.6982, + "step": 9874 + }, + { + "epoch": 0.36, + "learning_rate": 1.640202145969898e-07, + "loss": 0.6789, + "step": 9875 + }, + { + "epoch": 0.36, + "learning_rate": 1.6401655253231772e-07, + "loss": 0.6958, + "step": 9876 + }, + { + "epoch": 0.36, + "learning_rate": 1.6401289046764567e-07, + "loss": 0.6969, + "step": 9877 + }, + { + "epoch": 0.36, + "learning_rate": 1.640092284029736e-07, + "loss": 0.6974, + "step": 9878 + }, + { + "epoch": 0.36, + "learning_rate": 1.6400556633830152e-07, + "loss": 0.6835, + "step": 9879 + }, + { + "epoch": 0.36, + "learning_rate": 1.6400190427362945e-07, + "loss": 0.7091, + "step": 9880 + }, + { + "epoch": 0.36, + "learning_rate": 1.639982422089574e-07, + "loss": 0.704, + "step": 9881 + }, + { + "epoch": 0.36, + "learning_rate": 1.6399458014428533e-07, + "loss": 0.6874, + "step": 9882 + }, + { + "epoch": 0.36, + "learning_rate": 1.6399091807961328e-07, + "loss": 0.7034, + "step": 9883 + }, + { + "epoch": 0.36, + "learning_rate": 1.639872560149412e-07, + "loss": 0.7022, + "step": 9884 + }, + { + "epoch": 0.36, + "learning_rate": 1.6398359395026914e-07, + "loss": 0.7017, + "step": 9885 + }, + { + "epoch": 0.36, + "learning_rate": 1.639799318855971e-07, + "loss": 0.7244, + "step": 9886 + }, + { + "epoch": 0.36, + "learning_rate": 1.6397626982092502e-07, + "loss": 0.6967, + "step": 9887 + }, + { + "epoch": 0.36, + "learning_rate": 1.6397260775625297e-07, + "loss": 0.6785, + "step": 9888 + }, + { + "epoch": 0.36, + "learning_rate": 1.639689456915809e-07, + "loss": 0.6858, + "step": 9889 + }, + { + "epoch": 0.36, + "learning_rate": 1.6396528362690885e-07, + "loss": 0.6815, + "step": 9890 + }, + { + "epoch": 0.36, + "learning_rate": 1.6396162156223678e-07, + "loss": 0.6965, + "step": 9891 + }, + { + "epoch": 0.36, + "learning_rate": 1.6395795949756473e-07, + "loss": 0.6844, + "step": 9892 + }, + { + "epoch": 0.36, + "learning_rate": 1.6395429743289266e-07, + "loss": 0.7031, + "step": 9893 + }, + { + "epoch": 0.36, + "learning_rate": 1.639506353682206e-07, + "loss": 0.6823, + "step": 9894 + }, + { + "epoch": 0.36, + "learning_rate": 1.6394697330354854e-07, + "loss": 0.7067, + "step": 9895 + }, + { + "epoch": 0.36, + "learning_rate": 1.639433112388765e-07, + "loss": 0.7073, + "step": 9896 + }, + { + "epoch": 0.36, + "learning_rate": 1.6393964917420442e-07, + "loss": 0.6878, + "step": 9897 + }, + { + "epoch": 0.36, + "learning_rate": 1.6393598710953234e-07, + "loss": 0.7119, + "step": 9898 + }, + { + "epoch": 0.36, + "learning_rate": 1.6393232504486027e-07, + "loss": 0.7064, + "step": 9899 + }, + { + "epoch": 0.36, + "learning_rate": 1.6392866298018822e-07, + "loss": 0.6876, + "step": 9900 + }, + { + "epoch": 0.36, + "learning_rate": 1.6392500091551615e-07, + "loss": 0.7038, + "step": 9901 + }, + { + "epoch": 0.36, + "learning_rate": 1.639213388508441e-07, + "loss": 0.7086, + "step": 9902 + }, + { + "epoch": 0.36, + "learning_rate": 1.6391767678617203e-07, + "loss": 0.6844, + "step": 9903 + }, + { + "epoch": 0.36, + "learning_rate": 1.6391401472149998e-07, + "loss": 0.6822, + "step": 9904 + }, + { + "epoch": 0.36, + "learning_rate": 1.639103526568279e-07, + "loss": 0.6933, + "step": 9905 + }, + { + "epoch": 0.36, + "learning_rate": 1.6390669059215584e-07, + "loss": 0.6938, + "step": 9906 + }, + { + "epoch": 0.36, + "learning_rate": 1.639030285274838e-07, + "loss": 0.6845, + "step": 9907 + }, + { + "epoch": 0.36, + "learning_rate": 1.6389936646281172e-07, + "loss": 0.6979, + "step": 9908 + }, + { + "epoch": 0.36, + "learning_rate": 1.6389570439813967e-07, + "loss": 0.687, + "step": 9909 + }, + { + "epoch": 0.36, + "learning_rate": 1.638920423334676e-07, + "loss": 0.7038, + "step": 9910 + }, + { + "epoch": 0.36, + "learning_rate": 1.6388838026879555e-07, + "loss": 0.6799, + "step": 9911 + }, + { + "epoch": 0.36, + "learning_rate": 1.6388471820412348e-07, + "loss": 0.6952, + "step": 9912 + }, + { + "epoch": 0.36, + "learning_rate": 1.6388105613945143e-07, + "loss": 0.6944, + "step": 9913 + }, + { + "epoch": 0.36, + "learning_rate": 1.6387739407477936e-07, + "loss": 0.6964, + "step": 9914 + }, + { + "epoch": 0.36, + "learning_rate": 1.638737320101073e-07, + "loss": 0.6983, + "step": 9915 + }, + { + "epoch": 0.36, + "learning_rate": 1.638700699454352e-07, + "loss": 0.6914, + "step": 9916 + }, + { + "epoch": 0.36, + "learning_rate": 1.6386640788076316e-07, + "loss": 0.689, + "step": 9917 + }, + { + "epoch": 0.36, + "learning_rate": 1.638627458160911e-07, + "loss": 0.6926, + "step": 9918 + }, + { + "epoch": 0.36, + "learning_rate": 1.6385908375141904e-07, + "loss": 0.703, + "step": 9919 + }, + { + "epoch": 0.36, + "learning_rate": 1.6385542168674697e-07, + "loss": 0.7065, + "step": 9920 + }, + { + "epoch": 0.36, + "learning_rate": 1.6385175962207492e-07, + "loss": 0.6796, + "step": 9921 + }, + { + "epoch": 0.36, + "learning_rate": 1.6384809755740285e-07, + "loss": 0.6749, + "step": 9922 + }, + { + "epoch": 0.36, + "learning_rate": 1.638444354927308e-07, + "loss": 0.6982, + "step": 9923 + }, + { + "epoch": 0.36, + "learning_rate": 1.6384077342805873e-07, + "loss": 0.7025, + "step": 9924 + }, + { + "epoch": 0.36, + "learning_rate": 1.6383711136338668e-07, + "loss": 0.6839, + "step": 9925 + }, + { + "epoch": 0.36, + "learning_rate": 1.638334492987146e-07, + "loss": 0.6868, + "step": 9926 + }, + { + "epoch": 0.36, + "learning_rate": 1.6382978723404256e-07, + "loss": 0.6813, + "step": 9927 + }, + { + "epoch": 0.36, + "learning_rate": 1.638261251693705e-07, + "loss": 0.6845, + "step": 9928 + }, + { + "epoch": 0.36, + "learning_rate": 1.6382246310469842e-07, + "loss": 0.6969, + "step": 9929 + }, + { + "epoch": 0.36, + "learning_rate": 1.6381880104002637e-07, + "loss": 0.6858, + "step": 9930 + }, + { + "epoch": 0.36, + "learning_rate": 1.638151389753543e-07, + "loss": 0.6829, + "step": 9931 + }, + { + "epoch": 0.36, + "learning_rate": 1.6381147691068225e-07, + "loss": 0.7015, + "step": 9932 + }, + { + "epoch": 0.36, + "learning_rate": 1.6380781484601018e-07, + "loss": 0.6874, + "step": 9933 + }, + { + "epoch": 0.36, + "learning_rate": 1.638041527813381e-07, + "loss": 0.6996, + "step": 9934 + }, + { + "epoch": 0.36, + "learning_rate": 1.6380049071666603e-07, + "loss": 0.6945, + "step": 9935 + }, + { + "epoch": 0.36, + "learning_rate": 1.6379682865199398e-07, + "loss": 0.6823, + "step": 9936 + }, + { + "epoch": 0.36, + "learning_rate": 1.637931665873219e-07, + "loss": 0.6869, + "step": 9937 + }, + { + "epoch": 0.36, + "learning_rate": 1.6378950452264986e-07, + "loss": 0.6902, + "step": 9938 + }, + { + "epoch": 0.36, + "learning_rate": 1.637858424579778e-07, + "loss": 0.6937, + "step": 9939 + }, + { + "epoch": 0.36, + "learning_rate": 1.6378218039330574e-07, + "loss": 0.6868, + "step": 9940 + }, + { + "epoch": 0.36, + "learning_rate": 1.6377851832863367e-07, + "loss": 0.6975, + "step": 9941 + }, + { + "epoch": 0.36, + "learning_rate": 1.6377485626396162e-07, + "loss": 0.6957, + "step": 9942 + }, + { + "epoch": 0.36, + "learning_rate": 1.6377119419928955e-07, + "loss": 0.7019, + "step": 9943 + }, + { + "epoch": 0.36, + "learning_rate": 1.637675321346175e-07, + "loss": 0.6965, + "step": 9944 + }, + { + "epoch": 0.36, + "learning_rate": 1.6376387006994543e-07, + "loss": 0.6923, + "step": 9945 + }, + { + "epoch": 0.36, + "learning_rate": 1.6376020800527338e-07, + "loss": 0.6765, + "step": 9946 + }, + { + "epoch": 0.36, + "learning_rate": 1.637565459406013e-07, + "loss": 0.6918, + "step": 9947 + }, + { + "epoch": 0.36, + "learning_rate": 1.6375288387592926e-07, + "loss": 0.7024, + "step": 9948 + }, + { + "epoch": 0.36, + "learning_rate": 1.637492218112572e-07, + "loss": 0.6857, + "step": 9949 + }, + { + "epoch": 0.36, + "learning_rate": 1.6374555974658511e-07, + "loss": 0.6872, + "step": 9950 + }, + { + "epoch": 0.36, + "learning_rate": 1.6374189768191307e-07, + "loss": 0.6827, + "step": 9951 + }, + { + "epoch": 0.36, + "learning_rate": 1.63738235617241e-07, + "loss": 0.6779, + "step": 9952 + }, + { + "epoch": 0.36, + "learning_rate": 1.6373457355256892e-07, + "loss": 0.6887, + "step": 9953 + }, + { + "epoch": 0.36, + "learning_rate": 1.6373091148789687e-07, + "loss": 0.69, + "step": 9954 + }, + { + "epoch": 0.36, + "learning_rate": 1.637272494232248e-07, + "loss": 0.6985, + "step": 9955 + }, + { + "epoch": 0.36, + "learning_rate": 1.6372358735855273e-07, + "loss": 0.6987, + "step": 9956 + }, + { + "epoch": 0.36, + "learning_rate": 1.6371992529388068e-07, + "loss": 0.6813, + "step": 9957 + }, + { + "epoch": 0.36, + "learning_rate": 1.637162632292086e-07, + "loss": 0.7011, + "step": 9958 + }, + { + "epoch": 0.36, + "learning_rate": 1.6371260116453656e-07, + "loss": 0.6935, + "step": 9959 + }, + { + "epoch": 0.36, + "learning_rate": 1.637089390998645e-07, + "loss": 0.6933, + "step": 9960 + }, + { + "epoch": 0.36, + "learning_rate": 1.6370527703519244e-07, + "loss": 0.6817, + "step": 9961 + }, + { + "epoch": 0.36, + "learning_rate": 1.6370161497052037e-07, + "loss": 0.6833, + "step": 9962 + }, + { + "epoch": 0.36, + "learning_rate": 1.6369795290584832e-07, + "loss": 0.6985, + "step": 9963 + }, + { + "epoch": 0.36, + "learning_rate": 1.6369429084117625e-07, + "loss": 0.7157, + "step": 9964 + }, + { + "epoch": 0.36, + "learning_rate": 1.636906287765042e-07, + "loss": 0.7144, + "step": 9965 + }, + { + "epoch": 0.36, + "learning_rate": 1.6368696671183213e-07, + "loss": 0.6907, + "step": 9966 + }, + { + "epoch": 0.36, + "learning_rate": 1.6368330464716008e-07, + "loss": 0.6809, + "step": 9967 + }, + { + "epoch": 0.36, + "learning_rate": 1.63679642582488e-07, + "loss": 0.6849, + "step": 9968 + }, + { + "epoch": 0.36, + "learning_rate": 1.6367598051781596e-07, + "loss": 0.6843, + "step": 9969 + }, + { + "epoch": 0.36, + "learning_rate": 1.6367231845314386e-07, + "loss": 0.6877, + "step": 9970 + }, + { + "epoch": 0.36, + "learning_rate": 1.636686563884718e-07, + "loss": 0.6988, + "step": 9971 + }, + { + "epoch": 0.36, + "learning_rate": 1.6366499432379974e-07, + "loss": 0.6984, + "step": 9972 + }, + { + "epoch": 0.36, + "learning_rate": 1.636613322591277e-07, + "loss": 0.682, + "step": 9973 + }, + { + "epoch": 0.36, + "learning_rate": 1.6365767019445562e-07, + "loss": 0.697, + "step": 9974 + }, + { + "epoch": 0.36, + "learning_rate": 1.6365400812978357e-07, + "loss": 0.6927, + "step": 9975 + }, + { + "epoch": 0.36, + "learning_rate": 1.636503460651115e-07, + "loss": 0.6932, + "step": 9976 + }, + { + "epoch": 0.37, + "learning_rate": 1.6364668400043943e-07, + "loss": 0.6969, + "step": 9977 + }, + { + "epoch": 0.37, + "learning_rate": 1.6364302193576738e-07, + "loss": 0.6948, + "step": 9978 + }, + { + "epoch": 0.37, + "learning_rate": 1.636393598710953e-07, + "loss": 0.7029, + "step": 9979 + }, + { + "epoch": 0.37, + "learning_rate": 1.6363569780642326e-07, + "loss": 0.693, + "step": 9980 + }, + { + "epoch": 0.37, + "learning_rate": 1.6363203574175119e-07, + "loss": 0.6869, + "step": 9981 + }, + { + "epoch": 0.37, + "learning_rate": 1.6362837367707914e-07, + "loss": 0.694, + "step": 9982 + }, + { + "epoch": 0.37, + "learning_rate": 1.6362471161240707e-07, + "loss": 0.6878, + "step": 9983 + }, + { + "epoch": 0.37, + "learning_rate": 1.6362104954773502e-07, + "loss": 0.694, + "step": 9984 + }, + { + "epoch": 0.37, + "learning_rate": 1.6361738748306295e-07, + "loss": 0.6872, + "step": 9985 + }, + { + "epoch": 0.37, + "learning_rate": 1.636137254183909e-07, + "loss": 0.6941, + "step": 9986 + }, + { + "epoch": 0.37, + "learning_rate": 1.6361006335371883e-07, + "loss": 0.6957, + "step": 9987 + }, + { + "epoch": 0.37, + "learning_rate": 1.6360640128904675e-07, + "loss": 0.6907, + "step": 9988 + }, + { + "epoch": 0.37, + "learning_rate": 1.6360273922437468e-07, + "loss": 0.6907, + "step": 9989 + }, + { + "epoch": 0.37, + "learning_rate": 1.6359907715970263e-07, + "loss": 0.7019, + "step": 9990 + }, + { + "epoch": 0.37, + "learning_rate": 1.6359541509503056e-07, + "loss": 0.6917, + "step": 9991 + }, + { + "epoch": 0.37, + "learning_rate": 1.635917530303585e-07, + "loss": 0.696, + "step": 9992 + }, + { + "epoch": 0.37, + "learning_rate": 1.6358809096568644e-07, + "loss": 0.6747, + "step": 9993 + }, + { + "epoch": 0.37, + "learning_rate": 1.635844289010144e-07, + "loss": 0.6923, + "step": 9994 + }, + { + "epoch": 0.37, + "learning_rate": 1.6358076683634232e-07, + "loss": 0.6828, + "step": 9995 + }, + { + "epoch": 0.37, + "learning_rate": 1.6357710477167027e-07, + "loss": 0.6991, + "step": 9996 + }, + { + "epoch": 0.37, + "learning_rate": 1.635734427069982e-07, + "loss": 0.696, + "step": 9997 + }, + { + "epoch": 0.37, + "learning_rate": 1.6356978064232615e-07, + "loss": 0.704, + "step": 9998 + }, + { + "epoch": 0.37, + "learning_rate": 1.6356611857765408e-07, + "loss": 0.6969, + "step": 9999 + }, + { + "epoch": 0.37, + "learning_rate": 1.63562456512982e-07, + "loss": 0.7035, + "step": 10000 + }, + { + "epoch": 0.37, + "learning_rate": 1.6355879444830996e-07, + "loss": 0.6891, + "step": 10001 + }, + { + "epoch": 0.37, + "learning_rate": 1.6355513238363788e-07, + "loss": 0.6919, + "step": 10002 + }, + { + "epoch": 0.37, + "learning_rate": 1.6355147031896584e-07, + "loss": 0.6835, + "step": 10003 + }, + { + "epoch": 0.37, + "learning_rate": 1.6354780825429376e-07, + "loss": 0.6935, + "step": 10004 + }, + { + "epoch": 0.37, + "learning_rate": 1.6354414618962172e-07, + "loss": 0.6832, + "step": 10005 + }, + { + "epoch": 0.37, + "learning_rate": 1.6354048412494962e-07, + "loss": 0.6872, + "step": 10006 + }, + { + "epoch": 0.37, + "learning_rate": 1.6353682206027757e-07, + "loss": 0.7002, + "step": 10007 + }, + { + "epoch": 0.37, + "learning_rate": 1.635331599956055e-07, + "loss": 0.6967, + "step": 10008 + }, + { + "epoch": 0.37, + "learning_rate": 1.6352949793093345e-07, + "loss": 0.6961, + "step": 10009 + }, + { + "epoch": 0.37, + "learning_rate": 1.6352583586626138e-07, + "loss": 0.6749, + "step": 10010 + }, + { + "epoch": 0.37, + "learning_rate": 1.6352217380158933e-07, + "loss": 0.7063, + "step": 10011 + }, + { + "epoch": 0.37, + "learning_rate": 1.6351851173691726e-07, + "loss": 0.7126, + "step": 10012 + }, + { + "epoch": 0.37, + "learning_rate": 1.635148496722452e-07, + "loss": 0.7002, + "step": 10013 + }, + { + "epoch": 0.37, + "learning_rate": 1.6351118760757314e-07, + "loss": 0.6954, + "step": 10014 + }, + { + "epoch": 0.37, + "learning_rate": 1.635075255429011e-07, + "loss": 0.7041, + "step": 10015 + }, + { + "epoch": 0.37, + "learning_rate": 1.6350386347822902e-07, + "loss": 0.6972, + "step": 10016 + }, + { + "epoch": 0.37, + "learning_rate": 1.6350020141355697e-07, + "loss": 0.6944, + "step": 10017 + }, + { + "epoch": 0.37, + "learning_rate": 1.634965393488849e-07, + "loss": 0.6985, + "step": 10018 + }, + { + "epoch": 0.37, + "learning_rate": 1.6349287728421285e-07, + "loss": 0.6769, + "step": 10019 + }, + { + "epoch": 0.37, + "learning_rate": 1.6348921521954078e-07, + "loss": 0.7039, + "step": 10020 + }, + { + "epoch": 0.37, + "learning_rate": 1.634855531548687e-07, + "loss": 0.7104, + "step": 10021 + }, + { + "epoch": 0.37, + "learning_rate": 1.6348189109019666e-07, + "loss": 0.7028, + "step": 10022 + }, + { + "epoch": 0.37, + "learning_rate": 1.6347822902552458e-07, + "loss": 0.685, + "step": 10023 + }, + { + "epoch": 0.37, + "learning_rate": 1.634745669608525e-07, + "loss": 0.6984, + "step": 10024 + }, + { + "epoch": 0.37, + "learning_rate": 1.6347090489618046e-07, + "loss": 0.6962, + "step": 10025 + }, + { + "epoch": 0.37, + "learning_rate": 1.634672428315084e-07, + "loss": 0.7039, + "step": 10026 + }, + { + "epoch": 0.37, + "learning_rate": 1.6346358076683632e-07, + "loss": 0.6852, + "step": 10027 + }, + { + "epoch": 0.37, + "learning_rate": 1.6345991870216427e-07, + "loss": 0.6921, + "step": 10028 + }, + { + "epoch": 0.37, + "learning_rate": 1.634562566374922e-07, + "loss": 0.6886, + "step": 10029 + }, + { + "epoch": 0.37, + "learning_rate": 1.6345259457282015e-07, + "loss": 0.7025, + "step": 10030 + }, + { + "epoch": 0.37, + "learning_rate": 1.6344893250814808e-07, + "loss": 0.6851, + "step": 10031 + }, + { + "epoch": 0.37, + "learning_rate": 1.6344527044347603e-07, + "loss": 0.6906, + "step": 10032 + }, + { + "epoch": 0.37, + "learning_rate": 1.6344160837880396e-07, + "loss": 0.6966, + "step": 10033 + }, + { + "epoch": 0.37, + "learning_rate": 1.634379463141319e-07, + "loss": 0.6939, + "step": 10034 + }, + { + "epoch": 0.37, + "learning_rate": 1.6343428424945984e-07, + "loss": 0.6924, + "step": 10035 + }, + { + "epoch": 0.37, + "learning_rate": 1.634306221847878e-07, + "loss": 0.6999, + "step": 10036 + }, + { + "epoch": 0.37, + "learning_rate": 1.6342696012011572e-07, + "loss": 0.6953, + "step": 10037 + }, + { + "epoch": 0.37, + "learning_rate": 1.6342329805544367e-07, + "loss": 0.6987, + "step": 10038 + }, + { + "epoch": 0.37, + "learning_rate": 1.634196359907716e-07, + "loss": 0.6996, + "step": 10039 + }, + { + "epoch": 0.37, + "learning_rate": 1.6341597392609955e-07, + "loss": 0.6883, + "step": 10040 + }, + { + "epoch": 0.37, + "learning_rate": 1.6341231186142748e-07, + "loss": 0.709, + "step": 10041 + }, + { + "epoch": 0.37, + "learning_rate": 1.634086497967554e-07, + "loss": 0.6896, + "step": 10042 + }, + { + "epoch": 0.37, + "learning_rate": 1.6340498773208333e-07, + "loss": 0.6951, + "step": 10043 + }, + { + "epoch": 0.37, + "learning_rate": 1.6340132566741128e-07, + "loss": 0.6887, + "step": 10044 + }, + { + "epoch": 0.37, + "learning_rate": 1.633976636027392e-07, + "loss": 0.685, + "step": 10045 + }, + { + "epoch": 0.37, + "learning_rate": 1.6339400153806716e-07, + "loss": 0.6979, + "step": 10046 + }, + { + "epoch": 0.37, + "learning_rate": 1.633903394733951e-07, + "loss": 0.6986, + "step": 10047 + }, + { + "epoch": 0.37, + "learning_rate": 1.6338667740872302e-07, + "loss": 0.6987, + "step": 10048 + }, + { + "epoch": 0.37, + "learning_rate": 1.6338301534405097e-07, + "loss": 0.7049, + "step": 10049 + }, + { + "epoch": 0.37, + "learning_rate": 1.633793532793789e-07, + "loss": 0.7025, + "step": 10050 + }, + { + "epoch": 0.37, + "learning_rate": 1.6337569121470685e-07, + "loss": 0.6974, + "step": 10051 + }, + { + "epoch": 0.37, + "learning_rate": 1.6337202915003478e-07, + "loss": 0.7097, + "step": 10052 + }, + { + "epoch": 0.37, + "learning_rate": 1.6336836708536273e-07, + "loss": 0.6699, + "step": 10053 + }, + { + "epoch": 0.37, + "learning_rate": 1.6336470502069066e-07, + "loss": 0.6934, + "step": 10054 + }, + { + "epoch": 0.37, + "learning_rate": 1.633610429560186e-07, + "loss": 0.692, + "step": 10055 + }, + { + "epoch": 0.37, + "learning_rate": 1.6335738089134653e-07, + "loss": 0.6898, + "step": 10056 + }, + { + "epoch": 0.37, + "learning_rate": 1.633537188266745e-07, + "loss": 0.6938, + "step": 10057 + }, + { + "epoch": 0.37, + "learning_rate": 1.6335005676200241e-07, + "loss": 0.6934, + "step": 10058 + }, + { + "epoch": 0.37, + "learning_rate": 1.6334639469733034e-07, + "loss": 0.6881, + "step": 10059 + }, + { + "epoch": 0.37, + "learning_rate": 1.6334273263265827e-07, + "loss": 0.6875, + "step": 10060 + }, + { + "epoch": 0.37, + "learning_rate": 1.6333907056798622e-07, + "loss": 0.7054, + "step": 10061 + }, + { + "epoch": 0.37, + "learning_rate": 1.6333540850331415e-07, + "loss": 0.7011, + "step": 10062 + }, + { + "epoch": 0.37, + "learning_rate": 1.633317464386421e-07, + "loss": 0.6922, + "step": 10063 + }, + { + "epoch": 0.37, + "learning_rate": 1.6332808437397003e-07, + "loss": 0.6887, + "step": 10064 + }, + { + "epoch": 0.37, + "learning_rate": 1.6332442230929798e-07, + "loss": 0.6997, + "step": 10065 + }, + { + "epoch": 0.37, + "learning_rate": 1.633207602446259e-07, + "loss": 0.6925, + "step": 10066 + }, + { + "epoch": 0.37, + "learning_rate": 1.6331709817995386e-07, + "loss": 0.6931, + "step": 10067 + }, + { + "epoch": 0.37, + "learning_rate": 1.633134361152818e-07, + "loss": 0.7028, + "step": 10068 + }, + { + "epoch": 0.37, + "learning_rate": 1.6330977405060974e-07, + "loss": 0.6973, + "step": 10069 + }, + { + "epoch": 0.37, + "learning_rate": 1.6330611198593767e-07, + "loss": 0.6892, + "step": 10070 + }, + { + "epoch": 0.37, + "learning_rate": 1.633024499212656e-07, + "loss": 0.694, + "step": 10071 + }, + { + "epoch": 0.37, + "learning_rate": 1.6329878785659355e-07, + "loss": 0.6827, + "step": 10072 + }, + { + "epoch": 0.37, + "learning_rate": 1.6329512579192147e-07, + "loss": 0.6928, + "step": 10073 + }, + { + "epoch": 0.37, + "learning_rate": 1.6329146372724943e-07, + "loss": 0.6928, + "step": 10074 + }, + { + "epoch": 0.37, + "learning_rate": 1.6328780166257735e-07, + "loss": 0.6944, + "step": 10075 + }, + { + "epoch": 0.37, + "learning_rate": 1.632841395979053e-07, + "loss": 0.6733, + "step": 10076 + }, + { + "epoch": 0.37, + "learning_rate": 1.632804775332332e-07, + "loss": 0.6939, + "step": 10077 + }, + { + "epoch": 0.37, + "learning_rate": 1.6327681546856116e-07, + "loss": 0.6961, + "step": 10078 + }, + { + "epoch": 0.37, + "learning_rate": 1.632731534038891e-07, + "loss": 0.6859, + "step": 10079 + }, + { + "epoch": 0.37, + "learning_rate": 1.6326949133921704e-07, + "loss": 0.6936, + "step": 10080 + }, + { + "epoch": 0.37, + "learning_rate": 1.6326582927454497e-07, + "loss": 0.7101, + "step": 10081 + }, + { + "epoch": 0.37, + "learning_rate": 1.6326216720987292e-07, + "loss": 0.7086, + "step": 10082 + }, + { + "epoch": 0.37, + "learning_rate": 1.6325850514520085e-07, + "loss": 0.6918, + "step": 10083 + }, + { + "epoch": 0.37, + "learning_rate": 1.632548430805288e-07, + "loss": 0.6866, + "step": 10084 + }, + { + "epoch": 0.37, + "learning_rate": 1.6325118101585673e-07, + "loss": 0.6904, + "step": 10085 + }, + { + "epoch": 0.37, + "learning_rate": 1.6324751895118468e-07, + "loss": 0.6975, + "step": 10086 + }, + { + "epoch": 0.37, + "learning_rate": 1.632438568865126e-07, + "loss": 0.6927, + "step": 10087 + }, + { + "epoch": 0.37, + "learning_rate": 1.6324019482184056e-07, + "loss": 0.702, + "step": 10088 + }, + { + "epoch": 0.37, + "learning_rate": 1.6323653275716849e-07, + "loss": 0.699, + "step": 10089 + }, + { + "epoch": 0.37, + "learning_rate": 1.6323287069249644e-07, + "loss": 0.6979, + "step": 10090 + }, + { + "epoch": 0.37, + "learning_rate": 1.6322920862782437e-07, + "loss": 0.6865, + "step": 10091 + }, + { + "epoch": 0.37, + "learning_rate": 1.632255465631523e-07, + "loss": 0.6911, + "step": 10092 + }, + { + "epoch": 0.37, + "learning_rate": 1.6322188449848025e-07, + "loss": 0.6976, + "step": 10093 + }, + { + "epoch": 0.37, + "learning_rate": 1.6321822243380817e-07, + "loss": 0.6959, + "step": 10094 + }, + { + "epoch": 0.37, + "learning_rate": 1.632145603691361e-07, + "loss": 0.7056, + "step": 10095 + }, + { + "epoch": 0.37, + "learning_rate": 1.6321089830446405e-07, + "loss": 0.6967, + "step": 10096 + }, + { + "epoch": 0.37, + "learning_rate": 1.6320723623979198e-07, + "loss": 0.6847, + "step": 10097 + }, + { + "epoch": 0.37, + "learning_rate": 1.632035741751199e-07, + "loss": 0.6897, + "step": 10098 + }, + { + "epoch": 0.37, + "learning_rate": 1.6319991211044786e-07, + "loss": 0.6954, + "step": 10099 + }, + { + "epoch": 0.37, + "learning_rate": 1.6319625004577579e-07, + "loss": 0.6802, + "step": 10100 + }, + { + "epoch": 0.37, + "learning_rate": 1.6319258798110374e-07, + "loss": 0.7014, + "step": 10101 + }, + { + "epoch": 0.37, + "learning_rate": 1.6318892591643167e-07, + "loss": 0.7009, + "step": 10102 + }, + { + "epoch": 0.37, + "learning_rate": 1.6318526385175962e-07, + "loss": 0.7016, + "step": 10103 + }, + { + "epoch": 0.37, + "learning_rate": 1.6318160178708755e-07, + "loss": 0.6861, + "step": 10104 + }, + { + "epoch": 0.37, + "learning_rate": 1.631779397224155e-07, + "loss": 0.6977, + "step": 10105 + }, + { + "epoch": 0.37, + "learning_rate": 1.6317427765774343e-07, + "loss": 0.6834, + "step": 10106 + }, + { + "epoch": 0.37, + "learning_rate": 1.6317061559307138e-07, + "loss": 0.721, + "step": 10107 + }, + { + "epoch": 0.37, + "learning_rate": 1.631669535283993e-07, + "loss": 0.6932, + "step": 10108 + }, + { + "epoch": 0.37, + "learning_rate": 1.6316329146372726e-07, + "loss": 0.7, + "step": 10109 + }, + { + "epoch": 0.37, + "learning_rate": 1.6315962939905519e-07, + "loss": 0.6786, + "step": 10110 + }, + { + "epoch": 0.37, + "learning_rate": 1.6315596733438314e-07, + "loss": 0.7014, + "step": 10111 + }, + { + "epoch": 0.37, + "learning_rate": 1.6315230526971107e-07, + "loss": 0.6987, + "step": 10112 + }, + { + "epoch": 0.37, + "learning_rate": 1.63148643205039e-07, + "loss": 0.6962, + "step": 10113 + }, + { + "epoch": 0.37, + "learning_rate": 1.6314498114036692e-07, + "loss": 0.6917, + "step": 10114 + }, + { + "epoch": 0.37, + "learning_rate": 1.6314131907569487e-07, + "loss": 0.6865, + "step": 10115 + }, + { + "epoch": 0.37, + "learning_rate": 1.631376570110228e-07, + "loss": 0.6898, + "step": 10116 + }, + { + "epoch": 0.37, + "learning_rate": 1.6313399494635075e-07, + "loss": 0.7027, + "step": 10117 + }, + { + "epoch": 0.37, + "learning_rate": 1.6313033288167868e-07, + "loss": 0.6806, + "step": 10118 + }, + { + "epoch": 0.37, + "learning_rate": 1.631266708170066e-07, + "loss": 0.6944, + "step": 10119 + }, + { + "epoch": 0.37, + "learning_rate": 1.6312300875233456e-07, + "loss": 0.6772, + "step": 10120 + }, + { + "epoch": 0.37, + "learning_rate": 1.6311934668766248e-07, + "loss": 0.7046, + "step": 10121 + }, + { + "epoch": 0.37, + "learning_rate": 1.6311568462299044e-07, + "loss": 0.6826, + "step": 10122 + }, + { + "epoch": 0.37, + "learning_rate": 1.6311202255831836e-07, + "loss": 0.7026, + "step": 10123 + }, + { + "epoch": 0.37, + "learning_rate": 1.6310836049364632e-07, + "loss": 0.6939, + "step": 10124 + }, + { + "epoch": 0.37, + "learning_rate": 1.6310469842897424e-07, + "loss": 0.6752, + "step": 10125 + }, + { + "epoch": 0.37, + "learning_rate": 1.631010363643022e-07, + "loss": 0.696, + "step": 10126 + }, + { + "epoch": 0.37, + "learning_rate": 1.6309737429963012e-07, + "loss": 0.6783, + "step": 10127 + }, + { + "epoch": 0.37, + "learning_rate": 1.6309371223495808e-07, + "loss": 0.6853, + "step": 10128 + }, + { + "epoch": 0.37, + "learning_rate": 1.63090050170286e-07, + "loss": 0.6812, + "step": 10129 + }, + { + "epoch": 0.37, + "learning_rate": 1.6308638810561396e-07, + "loss": 0.6898, + "step": 10130 + }, + { + "epoch": 0.37, + "learning_rate": 1.6308272604094186e-07, + "loss": 0.6981, + "step": 10131 + }, + { + "epoch": 0.37, + "learning_rate": 1.630790639762698e-07, + "loss": 0.6808, + "step": 10132 + }, + { + "epoch": 0.37, + "learning_rate": 1.6307540191159774e-07, + "loss": 0.7065, + "step": 10133 + }, + { + "epoch": 0.37, + "learning_rate": 1.630717398469257e-07, + "loss": 0.6837, + "step": 10134 + }, + { + "epoch": 0.37, + "learning_rate": 1.6306807778225362e-07, + "loss": 0.7116, + "step": 10135 + }, + { + "epoch": 0.37, + "learning_rate": 1.6306441571758157e-07, + "loss": 0.7093, + "step": 10136 + }, + { + "epoch": 0.37, + "learning_rate": 1.630607536529095e-07, + "loss": 0.6953, + "step": 10137 + }, + { + "epoch": 0.37, + "learning_rate": 1.6305709158823745e-07, + "loss": 0.6923, + "step": 10138 + }, + { + "epoch": 0.37, + "learning_rate": 1.6305342952356538e-07, + "loss": 0.7006, + "step": 10139 + }, + { + "epoch": 0.37, + "learning_rate": 1.6304976745889333e-07, + "loss": 0.6833, + "step": 10140 + }, + { + "epoch": 0.37, + "learning_rate": 1.6304610539422126e-07, + "loss": 0.702, + "step": 10141 + }, + { + "epoch": 0.37, + "learning_rate": 1.6304244332954918e-07, + "loss": 0.6806, + "step": 10142 + }, + { + "epoch": 0.37, + "learning_rate": 1.6303878126487714e-07, + "loss": 0.6887, + "step": 10143 + }, + { + "epoch": 0.37, + "learning_rate": 1.6303511920020506e-07, + "loss": 0.6886, + "step": 10144 + }, + { + "epoch": 0.37, + "learning_rate": 1.6303145713553302e-07, + "loss": 0.6781, + "step": 10145 + }, + { + "epoch": 0.37, + "learning_rate": 1.6302779507086094e-07, + "loss": 0.7026, + "step": 10146 + }, + { + "epoch": 0.37, + "learning_rate": 1.630241330061889e-07, + "loss": 0.6994, + "step": 10147 + }, + { + "epoch": 0.37, + "learning_rate": 1.6302047094151682e-07, + "loss": 0.6847, + "step": 10148 + }, + { + "epoch": 0.37, + "learning_rate": 1.6301680887684475e-07, + "loss": 0.694, + "step": 10149 + }, + { + "epoch": 0.37, + "learning_rate": 1.6301314681217268e-07, + "loss": 0.696, + "step": 10150 + }, + { + "epoch": 0.37, + "learning_rate": 1.6300948474750063e-07, + "loss": 0.6702, + "step": 10151 + }, + { + "epoch": 0.37, + "learning_rate": 1.6300582268282856e-07, + "loss": 0.7041, + "step": 10152 + }, + { + "epoch": 0.37, + "learning_rate": 1.630021606181565e-07, + "loss": 0.6933, + "step": 10153 + }, + { + "epoch": 0.37, + "learning_rate": 1.6299849855348444e-07, + "loss": 0.6791, + "step": 10154 + }, + { + "epoch": 0.37, + "learning_rate": 1.629948364888124e-07, + "loss": 0.686, + "step": 10155 + }, + { + "epoch": 0.37, + "learning_rate": 1.6299117442414032e-07, + "loss": 0.6986, + "step": 10156 + }, + { + "epoch": 0.37, + "learning_rate": 1.6298751235946827e-07, + "loss": 0.6958, + "step": 10157 + }, + { + "epoch": 0.37, + "learning_rate": 1.629838502947962e-07, + "loss": 0.7005, + "step": 10158 + }, + { + "epoch": 0.37, + "learning_rate": 1.6298018823012415e-07, + "loss": 0.6906, + "step": 10159 + }, + { + "epoch": 0.37, + "learning_rate": 1.6297652616545208e-07, + "loss": 0.6894, + "step": 10160 + }, + { + "epoch": 0.37, + "learning_rate": 1.6297286410078003e-07, + "loss": 0.6849, + "step": 10161 + }, + { + "epoch": 0.37, + "learning_rate": 1.6296920203610796e-07, + "loss": 0.6947, + "step": 10162 + }, + { + "epoch": 0.37, + "learning_rate": 1.6296553997143588e-07, + "loss": 0.6934, + "step": 10163 + }, + { + "epoch": 0.37, + "learning_rate": 1.6296187790676384e-07, + "loss": 0.6978, + "step": 10164 + }, + { + "epoch": 0.37, + "learning_rate": 1.6295821584209176e-07, + "loss": 0.6892, + "step": 10165 + }, + { + "epoch": 0.37, + "learning_rate": 1.6295455377741972e-07, + "loss": 0.6669, + "step": 10166 + }, + { + "epoch": 0.37, + "learning_rate": 1.6295089171274764e-07, + "loss": 0.6951, + "step": 10167 + }, + { + "epoch": 0.37, + "learning_rate": 1.6294722964807557e-07, + "loss": 0.6878, + "step": 10168 + }, + { + "epoch": 0.37, + "learning_rate": 1.629435675834035e-07, + "loss": 0.7027, + "step": 10169 + }, + { + "epoch": 0.37, + "learning_rate": 1.6293990551873145e-07, + "loss": 0.694, + "step": 10170 + }, + { + "epoch": 0.37, + "learning_rate": 1.6293624345405938e-07, + "loss": 0.6807, + "step": 10171 + }, + { + "epoch": 0.37, + "learning_rate": 1.6293258138938733e-07, + "loss": 0.7065, + "step": 10172 + }, + { + "epoch": 0.37, + "learning_rate": 1.6292891932471526e-07, + "loss": 0.695, + "step": 10173 + }, + { + "epoch": 0.37, + "learning_rate": 1.629252572600432e-07, + "loss": 0.7102, + "step": 10174 + }, + { + "epoch": 0.37, + "learning_rate": 1.6292159519537114e-07, + "loss": 0.6935, + "step": 10175 + }, + { + "epoch": 0.37, + "learning_rate": 1.629179331306991e-07, + "loss": 0.6877, + "step": 10176 + }, + { + "epoch": 0.37, + "learning_rate": 1.6291427106602701e-07, + "loss": 0.6937, + "step": 10177 + }, + { + "epoch": 0.37, + "learning_rate": 1.6291060900135497e-07, + "loss": 0.6844, + "step": 10178 + }, + { + "epoch": 0.37, + "learning_rate": 1.629069469366829e-07, + "loss": 0.6971, + "step": 10179 + }, + { + "epoch": 0.37, + "learning_rate": 1.6290328487201085e-07, + "loss": 0.7138, + "step": 10180 + }, + { + "epoch": 0.37, + "learning_rate": 1.6289962280733877e-07, + "loss": 0.6773, + "step": 10181 + }, + { + "epoch": 0.37, + "learning_rate": 1.6289596074266673e-07, + "loss": 0.6761, + "step": 10182 + }, + { + "epoch": 0.37, + "learning_rate": 1.6289229867799465e-07, + "loss": 0.6954, + "step": 10183 + }, + { + "epoch": 0.37, + "learning_rate": 1.6288863661332258e-07, + "loss": 0.6947, + "step": 10184 + }, + { + "epoch": 0.37, + "learning_rate": 1.628849745486505e-07, + "loss": 0.7014, + "step": 10185 + }, + { + "epoch": 0.37, + "learning_rate": 1.6288131248397846e-07, + "loss": 0.6905, + "step": 10186 + }, + { + "epoch": 0.37, + "learning_rate": 1.628776504193064e-07, + "loss": 0.6878, + "step": 10187 + }, + { + "epoch": 0.37, + "learning_rate": 1.6287398835463434e-07, + "loss": 0.6999, + "step": 10188 + }, + { + "epoch": 0.37, + "learning_rate": 1.6287032628996227e-07, + "loss": 0.6957, + "step": 10189 + }, + { + "epoch": 0.37, + "learning_rate": 1.628666642252902e-07, + "loss": 0.6915, + "step": 10190 + }, + { + "epoch": 0.37, + "learning_rate": 1.6286300216061815e-07, + "loss": 0.6864, + "step": 10191 + }, + { + "epoch": 0.37, + "learning_rate": 1.6285934009594607e-07, + "loss": 0.6872, + "step": 10192 + }, + { + "epoch": 0.37, + "learning_rate": 1.6285567803127403e-07, + "loss": 0.688, + "step": 10193 + }, + { + "epoch": 0.37, + "learning_rate": 1.6285201596660195e-07, + "loss": 0.6914, + "step": 10194 + }, + { + "epoch": 0.37, + "learning_rate": 1.628483539019299e-07, + "loss": 0.6882, + "step": 10195 + }, + { + "epoch": 0.37, + "learning_rate": 1.6284469183725783e-07, + "loss": 0.6958, + "step": 10196 + }, + { + "epoch": 0.37, + "learning_rate": 1.628410297725858e-07, + "loss": 0.6974, + "step": 10197 + }, + { + "epoch": 0.37, + "learning_rate": 1.6283736770791371e-07, + "loss": 0.7015, + "step": 10198 + }, + { + "epoch": 0.37, + "learning_rate": 1.6283370564324167e-07, + "loss": 0.6973, + "step": 10199 + }, + { + "epoch": 0.37, + "learning_rate": 1.628300435785696e-07, + "loss": 0.6903, + "step": 10200 + }, + { + "epoch": 0.37, + "learning_rate": 1.6282638151389755e-07, + "loss": 0.6819, + "step": 10201 + }, + { + "epoch": 0.37, + "learning_rate": 1.6282271944922547e-07, + "loss": 0.6996, + "step": 10202 + }, + { + "epoch": 0.37, + "learning_rate": 1.628190573845534e-07, + "loss": 0.6868, + "step": 10203 + }, + { + "epoch": 0.37, + "learning_rate": 1.6281539531988133e-07, + "loss": 0.6862, + "step": 10204 + }, + { + "epoch": 0.37, + "learning_rate": 1.6281173325520928e-07, + "loss": 0.6874, + "step": 10205 + }, + { + "epoch": 0.37, + "learning_rate": 1.628080711905372e-07, + "loss": 0.7015, + "step": 10206 + }, + { + "epoch": 0.37, + "learning_rate": 1.6280440912586516e-07, + "loss": 0.7005, + "step": 10207 + }, + { + "epoch": 0.37, + "learning_rate": 1.6280074706119309e-07, + "loss": 0.7048, + "step": 10208 + }, + { + "epoch": 0.37, + "learning_rate": 1.6279708499652104e-07, + "loss": 0.6947, + "step": 10209 + }, + { + "epoch": 0.37, + "learning_rate": 1.6279342293184897e-07, + "loss": 0.6831, + "step": 10210 + }, + { + "epoch": 0.37, + "learning_rate": 1.627897608671769e-07, + "loss": 0.6949, + "step": 10211 + }, + { + "epoch": 0.37, + "learning_rate": 1.6278609880250485e-07, + "loss": 0.6808, + "step": 10212 + }, + { + "epoch": 0.37, + "learning_rate": 1.6278243673783277e-07, + "loss": 0.7065, + "step": 10213 + }, + { + "epoch": 0.37, + "learning_rate": 1.6277877467316073e-07, + "loss": 0.6859, + "step": 10214 + }, + { + "epoch": 0.37, + "learning_rate": 1.6277511260848865e-07, + "loss": 0.7045, + "step": 10215 + }, + { + "epoch": 0.37, + "learning_rate": 1.627714505438166e-07, + "loss": 0.6964, + "step": 10216 + }, + { + "epoch": 0.37, + "learning_rate": 1.6276778847914453e-07, + "loss": 0.6828, + "step": 10217 + }, + { + "epoch": 0.37, + "learning_rate": 1.6276412641447249e-07, + "loss": 0.6952, + "step": 10218 + }, + { + "epoch": 0.37, + "learning_rate": 1.627604643498004e-07, + "loss": 0.6998, + "step": 10219 + }, + { + "epoch": 0.37, + "learning_rate": 1.6275680228512837e-07, + "loss": 0.7006, + "step": 10220 + }, + { + "epoch": 0.37, + "learning_rate": 1.6275314022045627e-07, + "loss": 0.7045, + "step": 10221 + }, + { + "epoch": 0.37, + "learning_rate": 1.6274947815578422e-07, + "loss": 0.6901, + "step": 10222 + }, + { + "epoch": 0.37, + "learning_rate": 1.6274581609111215e-07, + "loss": 0.6896, + "step": 10223 + }, + { + "epoch": 0.37, + "learning_rate": 1.627421540264401e-07, + "loss": 0.6765, + "step": 10224 + }, + { + "epoch": 0.37, + "learning_rate": 1.6273849196176803e-07, + "loss": 0.6961, + "step": 10225 + }, + { + "epoch": 0.37, + "learning_rate": 1.6273482989709598e-07, + "loss": 0.6933, + "step": 10226 + }, + { + "epoch": 0.37, + "learning_rate": 1.627311678324239e-07, + "loss": 0.6944, + "step": 10227 + }, + { + "epoch": 0.37, + "learning_rate": 1.6272750576775186e-07, + "loss": 0.6922, + "step": 10228 + }, + { + "epoch": 0.37, + "learning_rate": 1.6272384370307979e-07, + "loss": 0.6876, + "step": 10229 + }, + { + "epoch": 0.37, + "learning_rate": 1.6272018163840774e-07, + "loss": 0.6819, + "step": 10230 + }, + { + "epoch": 0.37, + "learning_rate": 1.6271651957373567e-07, + "loss": 0.7096, + "step": 10231 + }, + { + "epoch": 0.37, + "learning_rate": 1.6271285750906362e-07, + "loss": 0.7076, + "step": 10232 + }, + { + "epoch": 0.37, + "learning_rate": 1.6270919544439154e-07, + "loss": 0.697, + "step": 10233 + }, + { + "epoch": 0.37, + "learning_rate": 1.6270553337971947e-07, + "loss": 0.6954, + "step": 10234 + }, + { + "epoch": 0.37, + "learning_rate": 1.6270187131504742e-07, + "loss": 0.7113, + "step": 10235 + }, + { + "epoch": 0.37, + "learning_rate": 1.6269820925037535e-07, + "loss": 0.7031, + "step": 10236 + }, + { + "epoch": 0.37, + "learning_rate": 1.626945471857033e-07, + "loss": 0.6935, + "step": 10237 + }, + { + "epoch": 0.37, + "learning_rate": 1.6269088512103123e-07, + "loss": 0.6824, + "step": 10238 + }, + { + "epoch": 0.37, + "learning_rate": 1.6268722305635916e-07, + "loss": 0.6773, + "step": 10239 + }, + { + "epoch": 0.37, + "learning_rate": 1.6268356099168708e-07, + "loss": 0.6855, + "step": 10240 + }, + { + "epoch": 0.37, + "learning_rate": 1.6267989892701504e-07, + "loss": 0.697, + "step": 10241 + }, + { + "epoch": 0.37, + "learning_rate": 1.6267623686234296e-07, + "loss": 0.6962, + "step": 10242 + }, + { + "epoch": 0.37, + "learning_rate": 1.6267257479767092e-07, + "loss": 0.6887, + "step": 10243 + }, + { + "epoch": 0.37, + "learning_rate": 1.6266891273299884e-07, + "loss": 0.726, + "step": 10244 + }, + { + "epoch": 0.37, + "learning_rate": 1.626652506683268e-07, + "loss": 0.6818, + "step": 10245 + }, + { + "epoch": 0.37, + "learning_rate": 1.6266158860365472e-07, + "loss": 0.6917, + "step": 10246 + }, + { + "epoch": 0.37, + "learning_rate": 1.6265792653898268e-07, + "loss": 0.6948, + "step": 10247 + }, + { + "epoch": 0.37, + "learning_rate": 1.626542644743106e-07, + "loss": 0.7014, + "step": 10248 + }, + { + "epoch": 0.37, + "learning_rate": 1.6265060240963856e-07, + "loss": 0.6904, + "step": 10249 + }, + { + "epoch": 0.38, + "learning_rate": 1.6264694034496648e-07, + "loss": 0.6878, + "step": 10250 + }, + { + "epoch": 0.38, + "learning_rate": 1.6264327828029444e-07, + "loss": 0.7073, + "step": 10251 + }, + { + "epoch": 0.38, + "learning_rate": 1.6263961621562236e-07, + "loss": 0.6976, + "step": 10252 + }, + { + "epoch": 0.38, + "learning_rate": 1.6263595415095032e-07, + "loss": 0.6955, + "step": 10253 + }, + { + "epoch": 0.38, + "learning_rate": 1.6263229208627824e-07, + "loss": 0.692, + "step": 10254 + }, + { + "epoch": 0.38, + "learning_rate": 1.6262863002160617e-07, + "loss": 0.6939, + "step": 10255 + }, + { + "epoch": 0.38, + "learning_rate": 1.6262496795693412e-07, + "loss": 0.7064, + "step": 10256 + }, + { + "epoch": 0.38, + "learning_rate": 1.6262130589226205e-07, + "loss": 0.6936, + "step": 10257 + }, + { + "epoch": 0.38, + "learning_rate": 1.6261764382758998e-07, + "loss": 0.7093, + "step": 10258 + }, + { + "epoch": 0.38, + "learning_rate": 1.6261398176291793e-07, + "loss": 0.6929, + "step": 10259 + }, + { + "epoch": 0.38, + "learning_rate": 1.6261031969824586e-07, + "loss": 0.7092, + "step": 10260 + }, + { + "epoch": 0.38, + "learning_rate": 1.6260665763357378e-07, + "loss": 0.6858, + "step": 10261 + }, + { + "epoch": 0.38, + "learning_rate": 1.6260299556890174e-07, + "loss": 0.6985, + "step": 10262 + }, + { + "epoch": 0.38, + "learning_rate": 1.6259933350422966e-07, + "loss": 0.6932, + "step": 10263 + }, + { + "epoch": 0.38, + "learning_rate": 1.6259567143955762e-07, + "loss": 0.7019, + "step": 10264 + }, + { + "epoch": 0.38, + "learning_rate": 1.6259200937488554e-07, + "loss": 0.6872, + "step": 10265 + }, + { + "epoch": 0.38, + "learning_rate": 1.625883473102135e-07, + "loss": 0.6923, + "step": 10266 + }, + { + "epoch": 0.38, + "learning_rate": 1.6258468524554142e-07, + "loss": 0.6948, + "step": 10267 + }, + { + "epoch": 0.38, + "learning_rate": 1.6258102318086938e-07, + "loss": 0.6996, + "step": 10268 + }, + { + "epoch": 0.38, + "learning_rate": 1.625773611161973e-07, + "loss": 0.6844, + "step": 10269 + }, + { + "epoch": 0.38, + "learning_rate": 1.6257369905152526e-07, + "loss": 0.7006, + "step": 10270 + }, + { + "epoch": 0.38, + "learning_rate": 1.6257003698685318e-07, + "loss": 0.6973, + "step": 10271 + }, + { + "epoch": 0.38, + "learning_rate": 1.6256637492218114e-07, + "loss": 0.6894, + "step": 10272 + }, + { + "epoch": 0.38, + "learning_rate": 1.6256271285750906e-07, + "loss": 0.682, + "step": 10273 + }, + { + "epoch": 0.38, + "learning_rate": 1.6255905079283702e-07, + "loss": 0.6819, + "step": 10274 + }, + { + "epoch": 0.38, + "learning_rate": 1.6255538872816492e-07, + "loss": 0.7104, + "step": 10275 + }, + { + "epoch": 0.38, + "learning_rate": 1.6255172666349287e-07, + "loss": 0.6957, + "step": 10276 + }, + { + "epoch": 0.38, + "learning_rate": 1.625480645988208e-07, + "loss": 0.6934, + "step": 10277 + }, + { + "epoch": 0.38, + "learning_rate": 1.6254440253414875e-07, + "loss": 0.6951, + "step": 10278 + }, + { + "epoch": 0.38, + "learning_rate": 1.6254074046947668e-07, + "loss": 0.6913, + "step": 10279 + }, + { + "epoch": 0.38, + "learning_rate": 1.6253707840480463e-07, + "loss": 0.7009, + "step": 10280 + }, + { + "epoch": 0.38, + "learning_rate": 1.6253341634013256e-07, + "loss": 0.7021, + "step": 10281 + }, + { + "epoch": 0.38, + "learning_rate": 1.6252975427546048e-07, + "loss": 0.6951, + "step": 10282 + }, + { + "epoch": 0.38, + "learning_rate": 1.6252609221078844e-07, + "loss": 0.7029, + "step": 10283 + }, + { + "epoch": 0.38, + "learning_rate": 1.6252243014611636e-07, + "loss": 0.6883, + "step": 10284 + }, + { + "epoch": 0.38, + "learning_rate": 1.6251876808144432e-07, + "loss": 0.681, + "step": 10285 + }, + { + "epoch": 0.38, + "learning_rate": 1.6251510601677224e-07, + "loss": 0.709, + "step": 10286 + }, + { + "epoch": 0.38, + "learning_rate": 1.625114439521002e-07, + "loss": 0.6985, + "step": 10287 + }, + { + "epoch": 0.38, + "learning_rate": 1.6250778188742812e-07, + "loss": 0.6866, + "step": 10288 + }, + { + "epoch": 0.38, + "learning_rate": 1.6250411982275608e-07, + "loss": 0.6717, + "step": 10289 + }, + { + "epoch": 0.38, + "learning_rate": 1.62500457758084e-07, + "loss": 0.686, + "step": 10290 + }, + { + "epoch": 0.38, + "learning_rate": 1.6249679569341195e-07, + "loss": 0.6889, + "step": 10291 + }, + { + "epoch": 0.38, + "learning_rate": 1.6249313362873988e-07, + "loss": 0.6934, + "step": 10292 + }, + { + "epoch": 0.38, + "learning_rate": 1.624894715640678e-07, + "loss": 0.7171, + "step": 10293 + }, + { + "epoch": 0.38, + "learning_rate": 1.6248580949939574e-07, + "loss": 0.7073, + "step": 10294 + }, + { + "epoch": 0.38, + "learning_rate": 1.624821474347237e-07, + "loss": 0.6793, + "step": 10295 + }, + { + "epoch": 0.38, + "learning_rate": 1.6247848537005161e-07, + "loss": 0.6992, + "step": 10296 + }, + { + "epoch": 0.38, + "learning_rate": 1.6247482330537957e-07, + "loss": 0.6935, + "step": 10297 + }, + { + "epoch": 0.38, + "learning_rate": 1.624711612407075e-07, + "loss": 0.6896, + "step": 10298 + }, + { + "epoch": 0.38, + "learning_rate": 1.6246749917603545e-07, + "loss": 0.6923, + "step": 10299 + }, + { + "epoch": 0.38, + "learning_rate": 1.6246383711136337e-07, + "loss": 0.6946, + "step": 10300 + }, + { + "epoch": 0.38, + "learning_rate": 1.6246017504669133e-07, + "loss": 0.6985, + "step": 10301 + }, + { + "epoch": 0.38, + "learning_rate": 1.6245651298201925e-07, + "loss": 0.6875, + "step": 10302 + }, + { + "epoch": 0.38, + "learning_rate": 1.624528509173472e-07, + "loss": 0.7138, + "step": 10303 + }, + { + "epoch": 0.38, + "learning_rate": 1.6244918885267513e-07, + "loss": 0.6995, + "step": 10304 + }, + { + "epoch": 0.38, + "learning_rate": 1.6244552678800306e-07, + "loss": 0.679, + "step": 10305 + }, + { + "epoch": 0.38, + "learning_rate": 1.6244186472333101e-07, + "loss": 0.7073, + "step": 10306 + }, + { + "epoch": 0.38, + "learning_rate": 1.6243820265865894e-07, + "loss": 0.7173, + "step": 10307 + }, + { + "epoch": 0.38, + "learning_rate": 1.624345405939869e-07, + "loss": 0.6884, + "step": 10308 + }, + { + "epoch": 0.38, + "learning_rate": 1.6243087852931482e-07, + "loss": 0.6996, + "step": 10309 + }, + { + "epoch": 0.38, + "learning_rate": 1.6242721646464277e-07, + "loss": 0.6861, + "step": 10310 + }, + { + "epoch": 0.38, + "learning_rate": 1.6242355439997067e-07, + "loss": 0.7079, + "step": 10311 + }, + { + "epoch": 0.38, + "learning_rate": 1.6241989233529863e-07, + "loss": 0.7033, + "step": 10312 + }, + { + "epoch": 0.38, + "learning_rate": 1.6241623027062655e-07, + "loss": 0.6934, + "step": 10313 + }, + { + "epoch": 0.38, + "learning_rate": 1.624125682059545e-07, + "loss": 0.6853, + "step": 10314 + }, + { + "epoch": 0.38, + "learning_rate": 1.6240890614128243e-07, + "loss": 0.6891, + "step": 10315 + }, + { + "epoch": 0.38, + "learning_rate": 1.624052440766104e-07, + "loss": 0.6891, + "step": 10316 + }, + { + "epoch": 0.38, + "learning_rate": 1.6240158201193831e-07, + "loss": 0.7002, + "step": 10317 + }, + { + "epoch": 0.38, + "learning_rate": 1.6239791994726627e-07, + "loss": 0.6941, + "step": 10318 + }, + { + "epoch": 0.38, + "learning_rate": 1.623942578825942e-07, + "loss": 0.6909, + "step": 10319 + }, + { + "epoch": 0.38, + "learning_rate": 1.6239059581792215e-07, + "loss": 0.6899, + "step": 10320 + }, + { + "epoch": 0.38, + "learning_rate": 1.6238693375325007e-07, + "loss": 0.6891, + "step": 10321 + }, + { + "epoch": 0.38, + "learning_rate": 1.6238327168857803e-07, + "loss": 0.6908, + "step": 10322 + }, + { + "epoch": 0.38, + "learning_rate": 1.6237960962390595e-07, + "loss": 0.6987, + "step": 10323 + }, + { + "epoch": 0.38, + "learning_rate": 1.623759475592339e-07, + "loss": 0.7251, + "step": 10324 + }, + { + "epoch": 0.38, + "learning_rate": 1.6237228549456183e-07, + "loss": 0.6723, + "step": 10325 + }, + { + "epoch": 0.38, + "learning_rate": 1.6236862342988976e-07, + "loss": 0.71, + "step": 10326 + }, + { + "epoch": 0.38, + "learning_rate": 1.623649613652177e-07, + "loss": 0.6998, + "step": 10327 + }, + { + "epoch": 0.38, + "learning_rate": 1.6236129930054564e-07, + "loss": 0.6956, + "step": 10328 + }, + { + "epoch": 0.38, + "learning_rate": 1.6235763723587357e-07, + "loss": 0.6982, + "step": 10329 + }, + { + "epoch": 0.38, + "learning_rate": 1.6235397517120152e-07, + "loss": 0.6936, + "step": 10330 + }, + { + "epoch": 0.38, + "learning_rate": 1.6235031310652945e-07, + "loss": 0.6806, + "step": 10331 + }, + { + "epoch": 0.38, + "learning_rate": 1.6234665104185737e-07, + "loss": 0.6938, + "step": 10332 + }, + { + "epoch": 0.38, + "learning_rate": 1.6234298897718533e-07, + "loss": 0.6854, + "step": 10333 + }, + { + "epoch": 0.38, + "learning_rate": 1.6233932691251325e-07, + "loss": 0.6911, + "step": 10334 + }, + { + "epoch": 0.38, + "learning_rate": 1.623356648478412e-07, + "loss": 0.704, + "step": 10335 + }, + { + "epoch": 0.38, + "learning_rate": 1.6233200278316913e-07, + "loss": 0.6894, + "step": 10336 + }, + { + "epoch": 0.38, + "learning_rate": 1.6232834071849709e-07, + "loss": 0.6885, + "step": 10337 + }, + { + "epoch": 0.38, + "learning_rate": 1.62324678653825e-07, + "loss": 0.7112, + "step": 10338 + }, + { + "epoch": 0.38, + "learning_rate": 1.6232101658915297e-07, + "loss": 0.7048, + "step": 10339 + }, + { + "epoch": 0.38, + "learning_rate": 1.623173545244809e-07, + "loss": 0.6945, + "step": 10340 + }, + { + "epoch": 0.38, + "learning_rate": 1.6231369245980885e-07, + "loss": 0.6942, + "step": 10341 + }, + { + "epoch": 0.38, + "learning_rate": 1.6231003039513677e-07, + "loss": 0.6971, + "step": 10342 + }, + { + "epoch": 0.38, + "learning_rate": 1.6230636833046473e-07, + "loss": 0.6978, + "step": 10343 + }, + { + "epoch": 0.38, + "learning_rate": 1.6230270626579265e-07, + "loss": 0.6939, + "step": 10344 + }, + { + "epoch": 0.38, + "learning_rate": 1.622990442011206e-07, + "loss": 0.7017, + "step": 10345 + }, + { + "epoch": 0.38, + "learning_rate": 1.6229538213644853e-07, + "loss": 0.6932, + "step": 10346 + }, + { + "epoch": 0.38, + "learning_rate": 1.6229172007177646e-07, + "loss": 0.691, + "step": 10347 + }, + { + "epoch": 0.38, + "learning_rate": 1.6228805800710439e-07, + "loss": 0.708, + "step": 10348 + }, + { + "epoch": 0.38, + "learning_rate": 1.6228439594243234e-07, + "loss": 0.702, + "step": 10349 + }, + { + "epoch": 0.38, + "learning_rate": 1.6228073387776027e-07, + "loss": 0.7088, + "step": 10350 + }, + { + "epoch": 0.38, + "learning_rate": 1.6227707181308822e-07, + "loss": 0.6792, + "step": 10351 + }, + { + "epoch": 0.38, + "learning_rate": 1.6227340974841615e-07, + "loss": 0.6916, + "step": 10352 + }, + { + "epoch": 0.38, + "learning_rate": 1.6226974768374407e-07, + "loss": 0.6931, + "step": 10353 + }, + { + "epoch": 0.38, + "learning_rate": 1.6226608561907202e-07, + "loss": 0.6992, + "step": 10354 + }, + { + "epoch": 0.38, + "learning_rate": 1.6226242355439995e-07, + "loss": 0.6926, + "step": 10355 + }, + { + "epoch": 0.38, + "learning_rate": 1.622587614897279e-07, + "loss": 0.6821, + "step": 10356 + }, + { + "epoch": 0.38, + "learning_rate": 1.6225509942505583e-07, + "loss": 0.6937, + "step": 10357 + }, + { + "epoch": 0.38, + "learning_rate": 1.6225143736038378e-07, + "loss": 0.686, + "step": 10358 + }, + { + "epoch": 0.38, + "learning_rate": 1.622477752957117e-07, + "loss": 0.6743, + "step": 10359 + }, + { + "epoch": 0.38, + "learning_rate": 1.6224411323103966e-07, + "loss": 0.7006, + "step": 10360 + }, + { + "epoch": 0.38, + "learning_rate": 1.622404511663676e-07, + "loss": 0.6949, + "step": 10361 + }, + { + "epoch": 0.38, + "learning_rate": 1.6223678910169554e-07, + "loss": 0.6926, + "step": 10362 + }, + { + "epoch": 0.38, + "learning_rate": 1.6223312703702347e-07, + "loss": 0.6882, + "step": 10363 + }, + { + "epoch": 0.38, + "learning_rate": 1.6222946497235142e-07, + "loss": 0.6814, + "step": 10364 + }, + { + "epoch": 0.38, + "learning_rate": 1.6222580290767932e-07, + "loss": 0.6984, + "step": 10365 + }, + { + "epoch": 0.38, + "learning_rate": 1.6222214084300728e-07, + "loss": 0.6789, + "step": 10366 + }, + { + "epoch": 0.38, + "learning_rate": 1.622184787783352e-07, + "loss": 0.7008, + "step": 10367 + }, + { + "epoch": 0.38, + "learning_rate": 1.6221481671366316e-07, + "loss": 0.7012, + "step": 10368 + }, + { + "epoch": 0.38, + "learning_rate": 1.6221115464899108e-07, + "loss": 0.6997, + "step": 10369 + }, + { + "epoch": 0.38, + "learning_rate": 1.6220749258431904e-07, + "loss": 0.7004, + "step": 10370 + }, + { + "epoch": 0.38, + "learning_rate": 1.6220383051964696e-07, + "loss": 0.6928, + "step": 10371 + }, + { + "epoch": 0.38, + "learning_rate": 1.6220016845497492e-07, + "loss": 0.6876, + "step": 10372 + }, + { + "epoch": 0.38, + "learning_rate": 1.6219650639030284e-07, + "loss": 0.6806, + "step": 10373 + }, + { + "epoch": 0.38, + "learning_rate": 1.621928443256308e-07, + "loss": 0.6923, + "step": 10374 + }, + { + "epoch": 0.38, + "learning_rate": 1.6218918226095872e-07, + "loss": 0.7011, + "step": 10375 + }, + { + "epoch": 0.38, + "learning_rate": 1.6218552019628665e-07, + "loss": 0.6889, + "step": 10376 + }, + { + "epoch": 0.38, + "learning_rate": 1.621818581316146e-07, + "loss": 0.6917, + "step": 10377 + }, + { + "epoch": 0.38, + "learning_rate": 1.6217819606694253e-07, + "loss": 0.7004, + "step": 10378 + }, + { + "epoch": 0.38, + "learning_rate": 1.6217453400227048e-07, + "loss": 0.7042, + "step": 10379 + }, + { + "epoch": 0.38, + "learning_rate": 1.621708719375984e-07, + "loss": 0.7012, + "step": 10380 + }, + { + "epoch": 0.38, + "learning_rate": 1.6216720987292636e-07, + "loss": 0.6963, + "step": 10381 + }, + { + "epoch": 0.38, + "learning_rate": 1.621635478082543e-07, + "loss": 0.692, + "step": 10382 + }, + { + "epoch": 0.38, + "learning_rate": 1.6215988574358222e-07, + "loss": 0.6922, + "step": 10383 + }, + { + "epoch": 0.38, + "learning_rate": 1.6215622367891014e-07, + "loss": 0.6922, + "step": 10384 + }, + { + "epoch": 0.38, + "learning_rate": 1.621525616142381e-07, + "loss": 0.6932, + "step": 10385 + }, + { + "epoch": 0.38, + "learning_rate": 1.6214889954956602e-07, + "loss": 0.6978, + "step": 10386 + }, + { + "epoch": 0.38, + "learning_rate": 1.6214523748489398e-07, + "loss": 0.6874, + "step": 10387 + }, + { + "epoch": 0.38, + "learning_rate": 1.621415754202219e-07, + "loss": 0.6915, + "step": 10388 + }, + { + "epoch": 0.38, + "learning_rate": 1.6213791335554986e-07, + "loss": 0.7103, + "step": 10389 + }, + { + "epoch": 0.38, + "learning_rate": 1.6213425129087778e-07, + "loss": 0.687, + "step": 10390 + }, + { + "epoch": 0.38, + "learning_rate": 1.6213058922620574e-07, + "loss": 0.7087, + "step": 10391 + }, + { + "epoch": 0.38, + "learning_rate": 1.6212692716153366e-07, + "loss": 0.684, + "step": 10392 + }, + { + "epoch": 0.38, + "learning_rate": 1.6212326509686162e-07, + "loss": 0.7012, + "step": 10393 + }, + { + "epoch": 0.38, + "learning_rate": 1.6211960303218954e-07, + "loss": 0.6899, + "step": 10394 + }, + { + "epoch": 0.38, + "learning_rate": 1.621159409675175e-07, + "loss": 0.6784, + "step": 10395 + }, + { + "epoch": 0.38, + "learning_rate": 1.6211227890284542e-07, + "loss": 0.7079, + "step": 10396 + }, + { + "epoch": 0.38, + "learning_rate": 1.6210861683817335e-07, + "loss": 0.6932, + "step": 10397 + }, + { + "epoch": 0.38, + "learning_rate": 1.621049547735013e-07, + "loss": 0.68, + "step": 10398 + }, + { + "epoch": 0.38, + "learning_rate": 1.6210129270882923e-07, + "loss": 0.6931, + "step": 10399 + }, + { + "epoch": 0.38, + "learning_rate": 1.6209763064415718e-07, + "loss": 0.698, + "step": 10400 + }, + { + "epoch": 0.38, + "learning_rate": 1.620939685794851e-07, + "loss": 0.7029, + "step": 10401 + }, + { + "epoch": 0.38, + "learning_rate": 1.6209030651481304e-07, + "loss": 0.7014, + "step": 10402 + }, + { + "epoch": 0.38, + "learning_rate": 1.6208664445014096e-07, + "loss": 0.69, + "step": 10403 + }, + { + "epoch": 0.38, + "learning_rate": 1.6208298238546892e-07, + "loss": 0.6996, + "step": 10404 + }, + { + "epoch": 0.38, + "learning_rate": 1.6207932032079684e-07, + "loss": 0.6965, + "step": 10405 + }, + { + "epoch": 0.38, + "learning_rate": 1.620756582561248e-07, + "loss": 0.7071, + "step": 10406 + }, + { + "epoch": 0.38, + "learning_rate": 1.6207199619145272e-07, + "loss": 0.6973, + "step": 10407 + }, + { + "epoch": 0.38, + "learning_rate": 1.6206833412678068e-07, + "loss": 0.6884, + "step": 10408 + }, + { + "epoch": 0.38, + "learning_rate": 1.620646720621086e-07, + "loss": 0.6919, + "step": 10409 + }, + { + "epoch": 0.38, + "learning_rate": 1.6206100999743656e-07, + "loss": 0.6948, + "step": 10410 + }, + { + "epoch": 0.38, + "learning_rate": 1.6205734793276448e-07, + "loss": 0.7031, + "step": 10411 + }, + { + "epoch": 0.38, + "learning_rate": 1.6205368586809243e-07, + "loss": 0.6888, + "step": 10412 + }, + { + "epoch": 0.38, + "learning_rate": 1.6205002380342036e-07, + "loss": 0.6862, + "step": 10413 + }, + { + "epoch": 0.38, + "learning_rate": 1.6204636173874831e-07, + "loss": 0.694, + "step": 10414 + }, + { + "epoch": 0.38, + "learning_rate": 1.6204269967407624e-07, + "loss": 0.6804, + "step": 10415 + }, + { + "epoch": 0.38, + "learning_rate": 1.620390376094042e-07, + "loss": 0.6971, + "step": 10416 + }, + { + "epoch": 0.38, + "learning_rate": 1.6203537554473212e-07, + "loss": 0.6982, + "step": 10417 + }, + { + "epoch": 0.38, + "learning_rate": 1.6203171348006007e-07, + "loss": 0.6851, + "step": 10418 + }, + { + "epoch": 0.38, + "learning_rate": 1.6202805141538797e-07, + "loss": 0.6768, + "step": 10419 + }, + { + "epoch": 0.38, + "learning_rate": 1.6202438935071593e-07, + "loss": 0.6938, + "step": 10420 + }, + { + "epoch": 0.38, + "learning_rate": 1.6202072728604385e-07, + "loss": 0.7, + "step": 10421 + }, + { + "epoch": 0.38, + "learning_rate": 1.620170652213718e-07, + "loss": 0.69, + "step": 10422 + }, + { + "epoch": 0.38, + "learning_rate": 1.6201340315669973e-07, + "loss": 0.6924, + "step": 10423 + }, + { + "epoch": 0.38, + "learning_rate": 1.6200974109202766e-07, + "loss": 0.6879, + "step": 10424 + }, + { + "epoch": 0.38, + "learning_rate": 1.6200607902735561e-07, + "loss": 0.6952, + "step": 10425 + }, + { + "epoch": 0.38, + "learning_rate": 1.6200241696268354e-07, + "loss": 0.6822, + "step": 10426 + }, + { + "epoch": 0.38, + "learning_rate": 1.619987548980115e-07, + "loss": 0.6957, + "step": 10427 + }, + { + "epoch": 0.38, + "learning_rate": 1.6199509283333942e-07, + "loss": 0.7179, + "step": 10428 + }, + { + "epoch": 0.38, + "learning_rate": 1.6199143076866737e-07, + "loss": 0.6786, + "step": 10429 + }, + { + "epoch": 0.38, + "learning_rate": 1.619877687039953e-07, + "loss": 0.7137, + "step": 10430 + }, + { + "epoch": 0.38, + "learning_rate": 1.6198410663932325e-07, + "loss": 0.6882, + "step": 10431 + }, + { + "epoch": 0.38, + "learning_rate": 1.6198044457465118e-07, + "loss": 0.697, + "step": 10432 + }, + { + "epoch": 0.38, + "learning_rate": 1.6197678250997913e-07, + "loss": 0.7088, + "step": 10433 + }, + { + "epoch": 0.38, + "learning_rate": 1.6197312044530706e-07, + "loss": 0.6938, + "step": 10434 + }, + { + "epoch": 0.38, + "learning_rate": 1.6196945838063501e-07, + "loss": 0.7076, + "step": 10435 + }, + { + "epoch": 0.38, + "learning_rate": 1.6196579631596291e-07, + "loss": 0.6872, + "step": 10436 + }, + { + "epoch": 0.38, + "learning_rate": 1.6196213425129087e-07, + "loss": 0.6807, + "step": 10437 + }, + { + "epoch": 0.38, + "learning_rate": 1.619584721866188e-07, + "loss": 0.7029, + "step": 10438 + }, + { + "epoch": 0.38, + "learning_rate": 1.6195481012194675e-07, + "loss": 0.6994, + "step": 10439 + }, + { + "epoch": 0.38, + "learning_rate": 1.6195114805727467e-07, + "loss": 0.7141, + "step": 10440 + }, + { + "epoch": 0.38, + "learning_rate": 1.6194748599260263e-07, + "loss": 0.6952, + "step": 10441 + }, + { + "epoch": 0.38, + "learning_rate": 1.6194382392793055e-07, + "loss": 0.6979, + "step": 10442 + }, + { + "epoch": 0.38, + "learning_rate": 1.619401618632585e-07, + "loss": 0.7004, + "step": 10443 + }, + { + "epoch": 0.38, + "learning_rate": 1.6193649979858643e-07, + "loss": 0.6916, + "step": 10444 + }, + { + "epoch": 0.38, + "learning_rate": 1.6193283773391439e-07, + "loss": 0.7149, + "step": 10445 + }, + { + "epoch": 0.38, + "learning_rate": 1.619291756692423e-07, + "loss": 0.7076, + "step": 10446 + }, + { + "epoch": 0.38, + "learning_rate": 1.6192551360457024e-07, + "loss": 0.6951, + "step": 10447 + }, + { + "epoch": 0.38, + "learning_rate": 1.619218515398982e-07, + "loss": 0.7096, + "step": 10448 + }, + { + "epoch": 0.38, + "learning_rate": 1.6191818947522612e-07, + "loss": 0.6914, + "step": 10449 + }, + { + "epoch": 0.38, + "learning_rate": 1.6191452741055407e-07, + "loss": 0.6943, + "step": 10450 + }, + { + "epoch": 0.38, + "learning_rate": 1.61910865345882e-07, + "loss": 0.6734, + "step": 10451 + }, + { + "epoch": 0.38, + "learning_rate": 1.6190720328120995e-07, + "loss": 0.6892, + "step": 10452 + }, + { + "epoch": 0.38, + "learning_rate": 1.6190354121653788e-07, + "loss": 0.7126, + "step": 10453 + }, + { + "epoch": 0.38, + "learning_rate": 1.618998791518658e-07, + "loss": 0.6985, + "step": 10454 + }, + { + "epoch": 0.38, + "learning_rate": 1.6189621708719373e-07, + "loss": 0.7101, + "step": 10455 + }, + { + "epoch": 0.38, + "learning_rate": 1.6189255502252169e-07, + "loss": 0.6935, + "step": 10456 + }, + { + "epoch": 0.38, + "learning_rate": 1.618888929578496e-07, + "loss": 0.6956, + "step": 10457 + }, + { + "epoch": 0.38, + "learning_rate": 1.6188523089317757e-07, + "loss": 0.695, + "step": 10458 + }, + { + "epoch": 0.38, + "learning_rate": 1.618815688285055e-07, + "loss": 0.6887, + "step": 10459 + }, + { + "epoch": 0.38, + "learning_rate": 1.6187790676383345e-07, + "loss": 0.6979, + "step": 10460 + }, + { + "epoch": 0.38, + "learning_rate": 1.6187424469916137e-07, + "loss": 0.6939, + "step": 10461 + }, + { + "epoch": 0.38, + "learning_rate": 1.6187058263448933e-07, + "loss": 0.6982, + "step": 10462 + }, + { + "epoch": 0.38, + "learning_rate": 1.6186692056981725e-07, + "loss": 0.6984, + "step": 10463 + }, + { + "epoch": 0.38, + "learning_rate": 1.618632585051452e-07, + "loss": 0.692, + "step": 10464 + }, + { + "epoch": 0.38, + "learning_rate": 1.6185959644047313e-07, + "loss": 0.6881, + "step": 10465 + }, + { + "epoch": 0.38, + "learning_rate": 1.6185593437580109e-07, + "loss": 0.6906, + "step": 10466 + }, + { + "epoch": 0.38, + "learning_rate": 1.61852272311129e-07, + "loss": 0.699, + "step": 10467 + }, + { + "epoch": 0.38, + "learning_rate": 1.6184861024645694e-07, + "loss": 0.6887, + "step": 10468 + }, + { + "epoch": 0.38, + "learning_rate": 1.618449481817849e-07, + "loss": 0.6849, + "step": 10469 + }, + { + "epoch": 0.38, + "learning_rate": 1.6184128611711282e-07, + "loss": 0.6765, + "step": 10470 + }, + { + "epoch": 0.38, + "learning_rate": 1.6183762405244077e-07, + "loss": 0.7157, + "step": 10471 + }, + { + "epoch": 0.38, + "learning_rate": 1.618339619877687e-07, + "loss": 0.6789, + "step": 10472 + }, + { + "epoch": 0.38, + "learning_rate": 1.6183029992309662e-07, + "loss": 0.6865, + "step": 10473 + }, + { + "epoch": 0.38, + "learning_rate": 1.6182663785842455e-07, + "loss": 0.7055, + "step": 10474 + }, + { + "epoch": 0.38, + "learning_rate": 1.618229757937525e-07, + "loss": 0.6966, + "step": 10475 + }, + { + "epoch": 0.38, + "learning_rate": 1.6181931372908043e-07, + "loss": 0.6948, + "step": 10476 + }, + { + "epoch": 0.38, + "learning_rate": 1.6181565166440838e-07, + "loss": 0.6807, + "step": 10477 + }, + { + "epoch": 0.38, + "learning_rate": 1.618119895997363e-07, + "loss": 0.6843, + "step": 10478 + }, + { + "epoch": 0.38, + "learning_rate": 1.6180832753506426e-07, + "loss": 0.6988, + "step": 10479 + }, + { + "epoch": 0.38, + "learning_rate": 1.618046654703922e-07, + "loss": 0.6953, + "step": 10480 + }, + { + "epoch": 0.38, + "learning_rate": 1.6180100340572014e-07, + "loss": 0.6987, + "step": 10481 + }, + { + "epoch": 0.38, + "learning_rate": 1.6179734134104807e-07, + "loss": 0.7091, + "step": 10482 + }, + { + "epoch": 0.38, + "learning_rate": 1.6179367927637602e-07, + "loss": 0.7089, + "step": 10483 + }, + { + "epoch": 0.38, + "learning_rate": 1.6179001721170395e-07, + "loss": 0.6849, + "step": 10484 + }, + { + "epoch": 0.38, + "learning_rate": 1.617863551470319e-07, + "loss": 0.7051, + "step": 10485 + }, + { + "epoch": 0.38, + "learning_rate": 1.6178269308235983e-07, + "loss": 0.6845, + "step": 10486 + }, + { + "epoch": 0.38, + "learning_rate": 1.6177903101768778e-07, + "loss": 0.6928, + "step": 10487 + }, + { + "epoch": 0.38, + "learning_rate": 1.617753689530157e-07, + "loss": 0.683, + "step": 10488 + }, + { + "epoch": 0.38, + "learning_rate": 1.6177170688834366e-07, + "loss": 0.7125, + "step": 10489 + }, + { + "epoch": 0.38, + "learning_rate": 1.6176804482367156e-07, + "loss": 0.6987, + "step": 10490 + }, + { + "epoch": 0.38, + "learning_rate": 1.6176438275899952e-07, + "loss": 0.7013, + "step": 10491 + }, + { + "epoch": 0.38, + "learning_rate": 1.6176072069432744e-07, + "loss": 0.6807, + "step": 10492 + }, + { + "epoch": 0.38, + "learning_rate": 1.617570586296554e-07, + "loss": 0.69, + "step": 10493 + }, + { + "epoch": 0.38, + "learning_rate": 1.6175339656498332e-07, + "loss": 0.6899, + "step": 10494 + }, + { + "epoch": 0.38, + "learning_rate": 1.6174973450031125e-07, + "loss": 0.7047, + "step": 10495 + }, + { + "epoch": 0.38, + "learning_rate": 1.617460724356392e-07, + "loss": 0.7152, + "step": 10496 + }, + { + "epoch": 0.38, + "learning_rate": 1.6174241037096713e-07, + "loss": 0.6929, + "step": 10497 + }, + { + "epoch": 0.38, + "learning_rate": 1.6173874830629508e-07, + "loss": 0.7018, + "step": 10498 + }, + { + "epoch": 0.38, + "learning_rate": 1.61735086241623e-07, + "loss": 0.6872, + "step": 10499 + }, + { + "epoch": 0.38, + "learning_rate": 1.6173142417695096e-07, + "loss": 0.6926, + "step": 10500 + }, + { + "epoch": 0.38, + "learning_rate": 1.617277621122789e-07, + "loss": 0.6839, + "step": 10501 + }, + { + "epoch": 0.38, + "learning_rate": 1.6172410004760684e-07, + "loss": 0.6915, + "step": 10502 + }, + { + "epoch": 0.38, + "learning_rate": 1.6172043798293477e-07, + "loss": 0.6916, + "step": 10503 + }, + { + "epoch": 0.38, + "learning_rate": 1.6171677591826272e-07, + "loss": 0.695, + "step": 10504 + }, + { + "epoch": 0.38, + "learning_rate": 1.6171311385359065e-07, + "loss": 0.7089, + "step": 10505 + }, + { + "epoch": 0.38, + "learning_rate": 1.617094517889186e-07, + "loss": 0.6931, + "step": 10506 + }, + { + "epoch": 0.38, + "learning_rate": 1.6170578972424653e-07, + "loss": 0.6905, + "step": 10507 + }, + { + "epoch": 0.38, + "learning_rate": 1.6170212765957446e-07, + "loss": 0.6957, + "step": 10508 + }, + { + "epoch": 0.38, + "learning_rate": 1.6169846559490238e-07, + "loss": 0.6967, + "step": 10509 + }, + { + "epoch": 0.38, + "learning_rate": 1.6169480353023034e-07, + "loss": 0.6901, + "step": 10510 + }, + { + "epoch": 0.38, + "learning_rate": 1.6169114146555826e-07, + "loss": 0.692, + "step": 10511 + }, + { + "epoch": 0.38, + "learning_rate": 1.6168747940088622e-07, + "loss": 0.6956, + "step": 10512 + }, + { + "epoch": 0.38, + "learning_rate": 1.6168381733621414e-07, + "loss": 0.6966, + "step": 10513 + }, + { + "epoch": 0.38, + "learning_rate": 1.616801552715421e-07, + "loss": 0.6856, + "step": 10514 + }, + { + "epoch": 0.38, + "learning_rate": 1.6167649320687002e-07, + "loss": 0.7074, + "step": 10515 + }, + { + "epoch": 0.38, + "learning_rate": 1.6167283114219798e-07, + "loss": 0.6937, + "step": 10516 + }, + { + "epoch": 0.38, + "learning_rate": 1.616691690775259e-07, + "loss": 0.6919, + "step": 10517 + }, + { + "epoch": 0.38, + "learning_rate": 1.6166550701285383e-07, + "loss": 0.7043, + "step": 10518 + }, + { + "epoch": 0.38, + "learning_rate": 1.6166184494818178e-07, + "loss": 0.689, + "step": 10519 + }, + { + "epoch": 0.38, + "learning_rate": 1.616581828835097e-07, + "loss": 0.7048, + "step": 10520 + }, + { + "epoch": 0.38, + "learning_rate": 1.6165452081883766e-07, + "loss": 0.7022, + "step": 10521 + }, + { + "epoch": 0.38, + "learning_rate": 1.616508587541656e-07, + "loss": 0.6953, + "step": 10522 + }, + { + "epoch": 0.39, + "learning_rate": 1.6164719668949354e-07, + "loss": 0.6923, + "step": 10523 + }, + { + "epoch": 0.39, + "learning_rate": 1.6164353462482147e-07, + "loss": 0.6938, + "step": 10524 + }, + { + "epoch": 0.39, + "learning_rate": 1.6163987256014942e-07, + "loss": 0.6991, + "step": 10525 + }, + { + "epoch": 0.39, + "learning_rate": 1.6163621049547732e-07, + "loss": 0.7, + "step": 10526 + }, + { + "epoch": 0.39, + "learning_rate": 1.6163254843080528e-07, + "loss": 0.6889, + "step": 10527 + }, + { + "epoch": 0.39, + "learning_rate": 1.616288863661332e-07, + "loss": 0.6973, + "step": 10528 + }, + { + "epoch": 0.39, + "learning_rate": 1.6162522430146116e-07, + "loss": 0.6894, + "step": 10529 + }, + { + "epoch": 0.39, + "learning_rate": 1.6162156223678908e-07, + "loss": 0.7001, + "step": 10530 + }, + { + "epoch": 0.39, + "learning_rate": 1.6161790017211703e-07, + "loss": 0.6946, + "step": 10531 + }, + { + "epoch": 0.39, + "learning_rate": 1.6161423810744496e-07, + "loss": 0.6916, + "step": 10532 + }, + { + "epoch": 0.39, + "learning_rate": 1.6161057604277291e-07, + "loss": 0.6955, + "step": 10533 + }, + { + "epoch": 0.39, + "learning_rate": 1.6160691397810084e-07, + "loss": 0.6828, + "step": 10534 + }, + { + "epoch": 0.39, + "learning_rate": 1.616032519134288e-07, + "loss": 0.6898, + "step": 10535 + }, + { + "epoch": 0.39, + "learning_rate": 1.6159958984875672e-07, + "loss": 0.6923, + "step": 10536 + }, + { + "epoch": 0.39, + "learning_rate": 1.6159592778408467e-07, + "loss": 0.7106, + "step": 10537 + }, + { + "epoch": 0.39, + "learning_rate": 1.615922657194126e-07, + "loss": 0.6857, + "step": 10538 + }, + { + "epoch": 0.39, + "learning_rate": 1.6158860365474053e-07, + "loss": 0.6723, + "step": 10539 + }, + { + "epoch": 0.39, + "learning_rate": 1.6158494159006848e-07, + "loss": 0.6891, + "step": 10540 + }, + { + "epoch": 0.39, + "learning_rate": 1.615812795253964e-07, + "loss": 0.6954, + "step": 10541 + }, + { + "epoch": 0.39, + "learning_rate": 1.6157761746072436e-07, + "loss": 0.693, + "step": 10542 + }, + { + "epoch": 0.39, + "learning_rate": 1.615739553960523e-07, + "loss": 0.6893, + "step": 10543 + }, + { + "epoch": 0.39, + "learning_rate": 1.6157029333138021e-07, + "loss": 0.6824, + "step": 10544 + }, + { + "epoch": 0.39, + "learning_rate": 1.6156663126670814e-07, + "loss": 0.707, + "step": 10545 + }, + { + "epoch": 0.39, + "learning_rate": 1.615629692020361e-07, + "loss": 0.7007, + "step": 10546 + }, + { + "epoch": 0.39, + "learning_rate": 1.6155930713736402e-07, + "loss": 0.6949, + "step": 10547 + }, + { + "epoch": 0.39, + "learning_rate": 1.6155564507269197e-07, + "loss": 0.6878, + "step": 10548 + }, + { + "epoch": 0.39, + "learning_rate": 1.615519830080199e-07, + "loss": 0.6919, + "step": 10549 + }, + { + "epoch": 0.39, + "learning_rate": 1.6154832094334785e-07, + "loss": 0.6935, + "step": 10550 + }, + { + "epoch": 0.39, + "learning_rate": 1.6154465887867578e-07, + "loss": 0.687, + "step": 10551 + }, + { + "epoch": 0.39, + "learning_rate": 1.6154099681400373e-07, + "loss": 0.6828, + "step": 10552 + }, + { + "epoch": 0.39, + "learning_rate": 1.6153733474933166e-07, + "loss": 0.6717, + "step": 10553 + }, + { + "epoch": 0.39, + "learning_rate": 1.6153367268465961e-07, + "loss": 0.6963, + "step": 10554 + }, + { + "epoch": 0.39, + "learning_rate": 1.6153001061998754e-07, + "loss": 0.6896, + "step": 10555 + }, + { + "epoch": 0.39, + "learning_rate": 1.615263485553155e-07, + "loss": 0.6933, + "step": 10556 + }, + { + "epoch": 0.39, + "learning_rate": 1.6152268649064342e-07, + "loss": 0.7019, + "step": 10557 + }, + { + "epoch": 0.39, + "learning_rate": 1.6151902442597137e-07, + "loss": 0.7001, + "step": 10558 + }, + { + "epoch": 0.39, + "learning_rate": 1.615153623612993e-07, + "loss": 0.6919, + "step": 10559 + }, + { + "epoch": 0.39, + "learning_rate": 1.6151170029662725e-07, + "loss": 0.6836, + "step": 10560 + }, + { + "epoch": 0.39, + "learning_rate": 1.6150803823195518e-07, + "loss": 0.7014, + "step": 10561 + }, + { + "epoch": 0.39, + "learning_rate": 1.615043761672831e-07, + "loss": 0.6866, + "step": 10562 + }, + { + "epoch": 0.39, + "learning_rate": 1.6150071410261103e-07, + "loss": 0.6995, + "step": 10563 + }, + { + "epoch": 0.39, + "learning_rate": 1.6149705203793899e-07, + "loss": 0.6958, + "step": 10564 + }, + { + "epoch": 0.39, + "learning_rate": 1.614933899732669e-07, + "loss": 0.6964, + "step": 10565 + }, + { + "epoch": 0.39, + "learning_rate": 1.6148972790859484e-07, + "loss": 0.6905, + "step": 10566 + }, + { + "epoch": 0.39, + "learning_rate": 1.614860658439228e-07, + "loss": 0.6833, + "step": 10567 + }, + { + "epoch": 0.39, + "learning_rate": 1.6148240377925072e-07, + "loss": 0.6876, + "step": 10568 + }, + { + "epoch": 0.39, + "learning_rate": 1.6147874171457867e-07, + "loss": 0.7039, + "step": 10569 + }, + { + "epoch": 0.39, + "learning_rate": 1.614750796499066e-07, + "loss": 0.685, + "step": 10570 + }, + { + "epoch": 0.39, + "learning_rate": 1.6147141758523455e-07, + "loss": 0.6903, + "step": 10571 + }, + { + "epoch": 0.39, + "learning_rate": 1.6146775552056248e-07, + "loss": 0.6771, + "step": 10572 + }, + { + "epoch": 0.39, + "learning_rate": 1.6146409345589043e-07, + "loss": 0.6879, + "step": 10573 + }, + { + "epoch": 0.39, + "learning_rate": 1.6146043139121836e-07, + "loss": 0.686, + "step": 10574 + }, + { + "epoch": 0.39, + "learning_rate": 1.614567693265463e-07, + "loss": 0.6925, + "step": 10575 + }, + { + "epoch": 0.39, + "learning_rate": 1.6145310726187424e-07, + "loss": 0.7065, + "step": 10576 + }, + { + "epoch": 0.39, + "learning_rate": 1.614494451972022e-07, + "loss": 0.6943, + "step": 10577 + }, + { + "epoch": 0.39, + "learning_rate": 1.6144578313253012e-07, + "loss": 0.6856, + "step": 10578 + }, + { + "epoch": 0.39, + "learning_rate": 1.6144212106785807e-07, + "loss": 0.6919, + "step": 10579 + }, + { + "epoch": 0.39, + "learning_rate": 1.6143845900318597e-07, + "loss": 0.6883, + "step": 10580 + }, + { + "epoch": 0.39, + "learning_rate": 1.6143479693851393e-07, + "loss": 0.6948, + "step": 10581 + }, + { + "epoch": 0.39, + "learning_rate": 1.6143113487384185e-07, + "loss": 0.6954, + "step": 10582 + }, + { + "epoch": 0.39, + "learning_rate": 1.614274728091698e-07, + "loss": 0.6968, + "step": 10583 + }, + { + "epoch": 0.39, + "learning_rate": 1.6142381074449773e-07, + "loss": 0.6961, + "step": 10584 + }, + { + "epoch": 0.39, + "learning_rate": 1.6142014867982569e-07, + "loss": 0.7065, + "step": 10585 + }, + { + "epoch": 0.39, + "learning_rate": 1.614164866151536e-07, + "loss": 0.6965, + "step": 10586 + }, + { + "epoch": 0.39, + "learning_rate": 1.6141282455048157e-07, + "loss": 0.681, + "step": 10587 + }, + { + "epoch": 0.39, + "learning_rate": 1.614091624858095e-07, + "loss": 0.6934, + "step": 10588 + }, + { + "epoch": 0.39, + "learning_rate": 1.6140550042113742e-07, + "loss": 0.6989, + "step": 10589 + }, + { + "epoch": 0.39, + "learning_rate": 1.6140183835646537e-07, + "loss": 0.6796, + "step": 10590 + }, + { + "epoch": 0.39, + "learning_rate": 1.613981762917933e-07, + "loss": 0.6975, + "step": 10591 + }, + { + "epoch": 0.39, + "learning_rate": 1.6139451422712125e-07, + "loss": 0.6852, + "step": 10592 + }, + { + "epoch": 0.39, + "learning_rate": 1.6139085216244918e-07, + "loss": 0.7087, + "step": 10593 + }, + { + "epoch": 0.39, + "learning_rate": 1.6138719009777713e-07, + "loss": 0.6906, + "step": 10594 + }, + { + "epoch": 0.39, + "learning_rate": 1.6138352803310506e-07, + "loss": 0.7141, + "step": 10595 + }, + { + "epoch": 0.39, + "learning_rate": 1.61379865968433e-07, + "loss": 0.702, + "step": 10596 + }, + { + "epoch": 0.39, + "learning_rate": 1.6137620390376094e-07, + "loss": 0.7071, + "step": 10597 + }, + { + "epoch": 0.39, + "learning_rate": 1.6137254183908886e-07, + "loss": 0.6857, + "step": 10598 + }, + { + "epoch": 0.39, + "learning_rate": 1.613688797744168e-07, + "loss": 0.7026, + "step": 10599 + }, + { + "epoch": 0.39, + "learning_rate": 1.6136521770974474e-07, + "loss": 0.6929, + "step": 10600 + }, + { + "epoch": 0.39, + "learning_rate": 1.6136155564507267e-07, + "loss": 0.6849, + "step": 10601 + }, + { + "epoch": 0.39, + "learning_rate": 1.6135789358040062e-07, + "loss": 0.718, + "step": 10602 + }, + { + "epoch": 0.39, + "learning_rate": 1.6135423151572855e-07, + "loss": 0.6862, + "step": 10603 + }, + { + "epoch": 0.39, + "learning_rate": 1.613505694510565e-07, + "loss": 0.6974, + "step": 10604 + }, + { + "epoch": 0.39, + "learning_rate": 1.6134690738638443e-07, + "loss": 0.679, + "step": 10605 + }, + { + "epoch": 0.39, + "learning_rate": 1.6134324532171238e-07, + "loss": 0.6812, + "step": 10606 + }, + { + "epoch": 0.39, + "learning_rate": 1.613395832570403e-07, + "loss": 0.6832, + "step": 10607 + }, + { + "epoch": 0.39, + "learning_rate": 1.6133592119236826e-07, + "loss": 0.6751, + "step": 10608 + }, + { + "epoch": 0.39, + "learning_rate": 1.613322591276962e-07, + "loss": 0.6822, + "step": 10609 + }, + { + "epoch": 0.39, + "learning_rate": 1.6132859706302412e-07, + "loss": 0.6977, + "step": 10610 + }, + { + "epoch": 0.39, + "learning_rate": 1.6132493499835207e-07, + "loss": 0.6855, + "step": 10611 + }, + { + "epoch": 0.39, + "learning_rate": 1.6132127293368e-07, + "loss": 0.6808, + "step": 10612 + }, + { + "epoch": 0.39, + "learning_rate": 1.6131761086900795e-07, + "loss": 0.7095, + "step": 10613 + }, + { + "epoch": 0.39, + "learning_rate": 1.6131394880433588e-07, + "loss": 0.6949, + "step": 10614 + }, + { + "epoch": 0.39, + "learning_rate": 1.6131028673966383e-07, + "loss": 0.6701, + "step": 10615 + }, + { + "epoch": 0.39, + "learning_rate": 1.6130662467499173e-07, + "loss": 0.6891, + "step": 10616 + }, + { + "epoch": 0.39, + "learning_rate": 1.6130296261031968e-07, + "loss": 0.6762, + "step": 10617 + }, + { + "epoch": 0.39, + "learning_rate": 1.612993005456476e-07, + "loss": 0.7209, + "step": 10618 + }, + { + "epoch": 0.39, + "learning_rate": 1.6129563848097556e-07, + "loss": 0.6784, + "step": 10619 + }, + { + "epoch": 0.39, + "learning_rate": 1.612919764163035e-07, + "loss": 0.6946, + "step": 10620 + }, + { + "epoch": 0.39, + "learning_rate": 1.6128831435163144e-07, + "loss": 0.6907, + "step": 10621 + }, + { + "epoch": 0.39, + "learning_rate": 1.6128465228695937e-07, + "loss": 0.669, + "step": 10622 + }, + { + "epoch": 0.39, + "learning_rate": 1.6128099022228732e-07, + "loss": 0.6897, + "step": 10623 + }, + { + "epoch": 0.39, + "learning_rate": 1.6127732815761525e-07, + "loss": 0.7006, + "step": 10624 + }, + { + "epoch": 0.39, + "learning_rate": 1.612736660929432e-07, + "loss": 0.6864, + "step": 10625 + }, + { + "epoch": 0.39, + "learning_rate": 1.6127000402827113e-07, + "loss": 0.6931, + "step": 10626 + }, + { + "epoch": 0.39, + "learning_rate": 1.6126634196359908e-07, + "loss": 0.6899, + "step": 10627 + }, + { + "epoch": 0.39, + "learning_rate": 1.61262679898927e-07, + "loss": 0.6942, + "step": 10628 + }, + { + "epoch": 0.39, + "learning_rate": 1.6125901783425496e-07, + "loss": 0.6827, + "step": 10629 + }, + { + "epoch": 0.39, + "learning_rate": 1.612553557695829e-07, + "loss": 0.6903, + "step": 10630 + }, + { + "epoch": 0.39, + "learning_rate": 1.6125169370491084e-07, + "loss": 0.6873, + "step": 10631 + }, + { + "epoch": 0.39, + "learning_rate": 1.6124803164023877e-07, + "loss": 0.6884, + "step": 10632 + }, + { + "epoch": 0.39, + "learning_rate": 1.612443695755667e-07, + "loss": 0.688, + "step": 10633 + }, + { + "epoch": 0.39, + "learning_rate": 1.6124070751089462e-07, + "loss": 0.6926, + "step": 10634 + }, + { + "epoch": 0.39, + "learning_rate": 1.6123704544622258e-07, + "loss": 0.6903, + "step": 10635 + }, + { + "epoch": 0.39, + "learning_rate": 1.612333833815505e-07, + "loss": 0.7004, + "step": 10636 + }, + { + "epoch": 0.39, + "learning_rate": 1.6122972131687843e-07, + "loss": 0.68, + "step": 10637 + }, + { + "epoch": 0.39, + "learning_rate": 1.6122605925220638e-07, + "loss": 0.6912, + "step": 10638 + }, + { + "epoch": 0.39, + "learning_rate": 1.612223971875343e-07, + "loss": 0.7012, + "step": 10639 + }, + { + "epoch": 0.39, + "learning_rate": 1.6121873512286226e-07, + "loss": 0.7016, + "step": 10640 + }, + { + "epoch": 0.39, + "learning_rate": 1.612150730581902e-07, + "loss": 0.6867, + "step": 10641 + }, + { + "epoch": 0.39, + "learning_rate": 1.6121141099351814e-07, + "loss": 0.6887, + "step": 10642 + }, + { + "epoch": 0.39, + "learning_rate": 1.6120774892884607e-07, + "loss": 0.6972, + "step": 10643 + }, + { + "epoch": 0.39, + "learning_rate": 1.6120408686417402e-07, + "loss": 0.6846, + "step": 10644 + }, + { + "epoch": 0.39, + "learning_rate": 1.6120042479950195e-07, + "loss": 0.6939, + "step": 10645 + }, + { + "epoch": 0.39, + "learning_rate": 1.611967627348299e-07, + "loss": 0.7013, + "step": 10646 + }, + { + "epoch": 0.39, + "learning_rate": 1.6119310067015783e-07, + "loss": 0.677, + "step": 10647 + }, + { + "epoch": 0.39, + "learning_rate": 1.6118943860548578e-07, + "loss": 0.6984, + "step": 10648 + }, + { + "epoch": 0.39, + "learning_rate": 1.611857765408137e-07, + "loss": 0.6817, + "step": 10649 + }, + { + "epoch": 0.39, + "learning_rate": 1.6118211447614166e-07, + "loss": 0.6999, + "step": 10650 + }, + { + "epoch": 0.39, + "learning_rate": 1.611784524114696e-07, + "loss": 0.6957, + "step": 10651 + }, + { + "epoch": 0.39, + "learning_rate": 1.6117479034679751e-07, + "loss": 0.6816, + "step": 10652 + }, + { + "epoch": 0.39, + "learning_rate": 1.6117112828212544e-07, + "loss": 0.6796, + "step": 10653 + }, + { + "epoch": 0.39, + "learning_rate": 1.611674662174534e-07, + "loss": 0.694, + "step": 10654 + }, + { + "epoch": 0.39, + "learning_rate": 1.6116380415278132e-07, + "loss": 0.6898, + "step": 10655 + }, + { + "epoch": 0.39, + "learning_rate": 1.6116014208810927e-07, + "loss": 0.6759, + "step": 10656 + }, + { + "epoch": 0.39, + "learning_rate": 1.611564800234372e-07, + "loss": 0.6843, + "step": 10657 + }, + { + "epoch": 0.39, + "learning_rate": 1.6115281795876515e-07, + "loss": 0.6997, + "step": 10658 + }, + { + "epoch": 0.39, + "learning_rate": 1.6114915589409308e-07, + "loss": 0.6848, + "step": 10659 + }, + { + "epoch": 0.39, + "learning_rate": 1.61145493829421e-07, + "loss": 0.6888, + "step": 10660 + }, + { + "epoch": 0.39, + "learning_rate": 1.6114183176474896e-07, + "loss": 0.6876, + "step": 10661 + }, + { + "epoch": 0.39, + "learning_rate": 1.611381697000769e-07, + "loss": 0.6964, + "step": 10662 + }, + { + "epoch": 0.39, + "learning_rate": 1.6113450763540484e-07, + "loss": 0.6938, + "step": 10663 + }, + { + "epoch": 0.39, + "learning_rate": 1.6113084557073277e-07, + "loss": 0.6799, + "step": 10664 + }, + { + "epoch": 0.39, + "learning_rate": 1.6112718350606072e-07, + "loss": 0.6964, + "step": 10665 + }, + { + "epoch": 0.39, + "learning_rate": 1.6112352144138865e-07, + "loss": 0.689, + "step": 10666 + }, + { + "epoch": 0.39, + "learning_rate": 1.611198593767166e-07, + "loss": 0.6921, + "step": 10667 + }, + { + "epoch": 0.39, + "learning_rate": 1.6111619731204453e-07, + "loss": 0.6899, + "step": 10668 + }, + { + "epoch": 0.39, + "learning_rate": 1.6111253524737248e-07, + "loss": 0.6863, + "step": 10669 + }, + { + "epoch": 0.39, + "learning_rate": 1.6110887318270038e-07, + "loss": 0.6853, + "step": 10670 + }, + { + "epoch": 0.39, + "learning_rate": 1.6110521111802833e-07, + "loss": 0.7071, + "step": 10671 + }, + { + "epoch": 0.39, + "learning_rate": 1.6110154905335626e-07, + "loss": 0.6827, + "step": 10672 + }, + { + "epoch": 0.39, + "learning_rate": 1.6109788698868421e-07, + "loss": 0.6928, + "step": 10673 + }, + { + "epoch": 0.39, + "learning_rate": 1.6109422492401214e-07, + "loss": 0.6918, + "step": 10674 + }, + { + "epoch": 0.39, + "learning_rate": 1.610905628593401e-07, + "loss": 0.6844, + "step": 10675 + }, + { + "epoch": 0.39, + "learning_rate": 1.6108690079466802e-07, + "loss": 0.6836, + "step": 10676 + }, + { + "epoch": 0.39, + "learning_rate": 1.6108323872999597e-07, + "loss": 0.6889, + "step": 10677 + }, + { + "epoch": 0.39, + "learning_rate": 1.610795766653239e-07, + "loss": 0.698, + "step": 10678 + }, + { + "epoch": 0.39, + "learning_rate": 1.6107591460065185e-07, + "loss": 0.69, + "step": 10679 + }, + { + "epoch": 0.39, + "learning_rate": 1.6107225253597978e-07, + "loss": 0.6919, + "step": 10680 + }, + { + "epoch": 0.39, + "learning_rate": 1.610685904713077e-07, + "loss": 0.7026, + "step": 10681 + }, + { + "epoch": 0.39, + "learning_rate": 1.6106492840663566e-07, + "loss": 0.6952, + "step": 10682 + }, + { + "epoch": 0.39, + "learning_rate": 1.6106126634196359e-07, + "loss": 0.6978, + "step": 10683 + }, + { + "epoch": 0.39, + "learning_rate": 1.6105760427729154e-07, + "loss": 0.6799, + "step": 10684 + }, + { + "epoch": 0.39, + "learning_rate": 1.6105394221261947e-07, + "loss": 0.6969, + "step": 10685 + }, + { + "epoch": 0.39, + "learning_rate": 1.6105028014794742e-07, + "loss": 0.687, + "step": 10686 + }, + { + "epoch": 0.39, + "learning_rate": 1.6104661808327535e-07, + "loss": 0.6879, + "step": 10687 + }, + { + "epoch": 0.39, + "learning_rate": 1.6104295601860327e-07, + "loss": 0.6845, + "step": 10688 + }, + { + "epoch": 0.39, + "learning_rate": 1.610392939539312e-07, + "loss": 0.6997, + "step": 10689 + }, + { + "epoch": 0.39, + "learning_rate": 1.6103563188925915e-07, + "loss": 0.7059, + "step": 10690 + }, + { + "epoch": 0.39, + "learning_rate": 1.6103196982458708e-07, + "loss": 0.6895, + "step": 10691 + }, + { + "epoch": 0.39, + "learning_rate": 1.6102830775991503e-07, + "loss": 0.6904, + "step": 10692 + }, + { + "epoch": 0.39, + "learning_rate": 1.6102464569524296e-07, + "loss": 0.683, + "step": 10693 + }, + { + "epoch": 0.39, + "learning_rate": 1.610209836305709e-07, + "loss": 0.6989, + "step": 10694 + }, + { + "epoch": 0.39, + "learning_rate": 1.6101732156589884e-07, + "loss": 0.6891, + "step": 10695 + }, + { + "epoch": 0.39, + "learning_rate": 1.610136595012268e-07, + "loss": 0.6879, + "step": 10696 + }, + { + "epoch": 0.39, + "learning_rate": 1.6100999743655472e-07, + "loss": 0.6863, + "step": 10697 + }, + { + "epoch": 0.39, + "learning_rate": 1.6100633537188267e-07, + "loss": 0.6884, + "step": 10698 + }, + { + "epoch": 0.39, + "learning_rate": 1.610026733072106e-07, + "loss": 0.6957, + "step": 10699 + }, + { + "epoch": 0.39, + "learning_rate": 1.6099901124253855e-07, + "loss": 0.6853, + "step": 10700 + }, + { + "epoch": 0.39, + "learning_rate": 1.6099534917786648e-07, + "loss": 0.702, + "step": 10701 + }, + { + "epoch": 0.39, + "learning_rate": 1.6099168711319443e-07, + "loss": 0.6987, + "step": 10702 + }, + { + "epoch": 0.39, + "learning_rate": 1.6098802504852236e-07, + "loss": 0.6936, + "step": 10703 + }, + { + "epoch": 0.39, + "learning_rate": 1.6098436298385029e-07, + "loss": 0.7015, + "step": 10704 + }, + { + "epoch": 0.39, + "learning_rate": 1.6098070091917824e-07, + "loss": 0.6909, + "step": 10705 + }, + { + "epoch": 0.39, + "learning_rate": 1.6097703885450617e-07, + "loss": 0.6772, + "step": 10706 + }, + { + "epoch": 0.39, + "learning_rate": 1.609733767898341e-07, + "loss": 0.6907, + "step": 10707 + }, + { + "epoch": 0.39, + "learning_rate": 1.6096971472516202e-07, + "loss": 0.7021, + "step": 10708 + }, + { + "epoch": 0.39, + "learning_rate": 1.6096605266048997e-07, + "loss": 0.6837, + "step": 10709 + }, + { + "epoch": 0.39, + "learning_rate": 1.609623905958179e-07, + "loss": 0.6774, + "step": 10710 + }, + { + "epoch": 0.39, + "learning_rate": 1.6095872853114585e-07, + "loss": 0.6986, + "step": 10711 + }, + { + "epoch": 0.39, + "learning_rate": 1.6095506646647378e-07, + "loss": 0.6838, + "step": 10712 + }, + { + "epoch": 0.39, + "learning_rate": 1.6095140440180173e-07, + "loss": 0.6826, + "step": 10713 + }, + { + "epoch": 0.39, + "learning_rate": 1.6094774233712966e-07, + "loss": 0.7036, + "step": 10714 + }, + { + "epoch": 0.39, + "learning_rate": 1.609440802724576e-07, + "loss": 0.694, + "step": 10715 + }, + { + "epoch": 0.39, + "learning_rate": 1.6094041820778554e-07, + "loss": 0.6954, + "step": 10716 + }, + { + "epoch": 0.39, + "learning_rate": 1.609367561431135e-07, + "loss": 0.6937, + "step": 10717 + }, + { + "epoch": 0.39, + "learning_rate": 1.6093309407844142e-07, + "loss": 0.7017, + "step": 10718 + }, + { + "epoch": 0.39, + "learning_rate": 1.6092943201376937e-07, + "loss": 0.6876, + "step": 10719 + }, + { + "epoch": 0.39, + "learning_rate": 1.609257699490973e-07, + "loss": 0.7023, + "step": 10720 + }, + { + "epoch": 0.39, + "learning_rate": 1.6092210788442525e-07, + "loss": 0.6932, + "step": 10721 + }, + { + "epoch": 0.39, + "learning_rate": 1.6091844581975318e-07, + "loss": 0.7075, + "step": 10722 + }, + { + "epoch": 0.39, + "learning_rate": 1.6091478375508113e-07, + "loss": 0.6888, + "step": 10723 + }, + { + "epoch": 0.39, + "learning_rate": 1.6091112169040903e-07, + "loss": 0.6907, + "step": 10724 + }, + { + "epoch": 0.39, + "learning_rate": 1.6090745962573698e-07, + "loss": 0.6892, + "step": 10725 + }, + { + "epoch": 0.39, + "learning_rate": 1.609037975610649e-07, + "loss": 0.6811, + "step": 10726 + }, + { + "epoch": 0.39, + "learning_rate": 1.6090013549639286e-07, + "loss": 0.7038, + "step": 10727 + }, + { + "epoch": 0.39, + "learning_rate": 1.608964734317208e-07, + "loss": 0.6908, + "step": 10728 + }, + { + "epoch": 0.39, + "learning_rate": 1.6089281136704874e-07, + "loss": 0.7059, + "step": 10729 + }, + { + "epoch": 0.39, + "learning_rate": 1.6088914930237667e-07, + "loss": 0.6776, + "step": 10730 + }, + { + "epoch": 0.39, + "learning_rate": 1.608854872377046e-07, + "loss": 0.6918, + "step": 10731 + }, + { + "epoch": 0.39, + "learning_rate": 1.6088182517303255e-07, + "loss": 0.7011, + "step": 10732 + }, + { + "epoch": 0.39, + "learning_rate": 1.6087816310836048e-07, + "loss": 0.6935, + "step": 10733 + }, + { + "epoch": 0.39, + "learning_rate": 1.6087450104368843e-07, + "loss": 0.684, + "step": 10734 + }, + { + "epoch": 0.39, + "learning_rate": 1.6087083897901636e-07, + "loss": 0.6935, + "step": 10735 + }, + { + "epoch": 0.39, + "learning_rate": 1.608671769143443e-07, + "loss": 0.7011, + "step": 10736 + }, + { + "epoch": 0.39, + "learning_rate": 1.6086351484967224e-07, + "loss": 0.6836, + "step": 10737 + }, + { + "epoch": 0.39, + "learning_rate": 1.608598527850002e-07, + "loss": 0.6872, + "step": 10738 + }, + { + "epoch": 0.39, + "learning_rate": 1.6085619072032812e-07, + "loss": 0.7051, + "step": 10739 + }, + { + "epoch": 0.39, + "learning_rate": 1.6085252865565607e-07, + "loss": 0.6772, + "step": 10740 + }, + { + "epoch": 0.39, + "learning_rate": 1.60848866590984e-07, + "loss": 0.6931, + "step": 10741 + }, + { + "epoch": 0.39, + "learning_rate": 1.6084520452631192e-07, + "loss": 0.6716, + "step": 10742 + }, + { + "epoch": 0.39, + "learning_rate": 1.6084154246163985e-07, + "loss": 0.6992, + "step": 10743 + }, + { + "epoch": 0.39, + "learning_rate": 1.608378803969678e-07, + "loss": 0.6763, + "step": 10744 + }, + { + "epoch": 0.39, + "learning_rate": 1.6083421833229573e-07, + "loss": 0.68, + "step": 10745 + }, + { + "epoch": 0.39, + "learning_rate": 1.6083055626762368e-07, + "loss": 0.6907, + "step": 10746 + }, + { + "epoch": 0.39, + "learning_rate": 1.608268942029516e-07, + "loss": 0.7002, + "step": 10747 + }, + { + "epoch": 0.39, + "learning_rate": 1.6082323213827956e-07, + "loss": 0.6966, + "step": 10748 + }, + { + "epoch": 0.39, + "learning_rate": 1.608195700736075e-07, + "loss": 0.6992, + "step": 10749 + }, + { + "epoch": 0.39, + "learning_rate": 1.6081590800893544e-07, + "loss": 0.7003, + "step": 10750 + }, + { + "epoch": 0.39, + "learning_rate": 1.6081224594426337e-07, + "loss": 0.6985, + "step": 10751 + }, + { + "epoch": 0.39, + "learning_rate": 1.608085838795913e-07, + "loss": 0.6846, + "step": 10752 + }, + { + "epoch": 0.39, + "learning_rate": 1.6080492181491925e-07, + "loss": 0.6914, + "step": 10753 + }, + { + "epoch": 0.39, + "learning_rate": 1.6080125975024718e-07, + "loss": 0.6917, + "step": 10754 + }, + { + "epoch": 0.39, + "learning_rate": 1.6079759768557513e-07, + "loss": 0.6877, + "step": 10755 + }, + { + "epoch": 0.39, + "learning_rate": 1.6079393562090306e-07, + "loss": 0.6945, + "step": 10756 + }, + { + "epoch": 0.39, + "learning_rate": 1.60790273556231e-07, + "loss": 0.6862, + "step": 10757 + }, + { + "epoch": 0.39, + "learning_rate": 1.6078661149155894e-07, + "loss": 0.704, + "step": 10758 + }, + { + "epoch": 0.39, + "learning_rate": 1.607829494268869e-07, + "loss": 0.6952, + "step": 10759 + }, + { + "epoch": 0.39, + "learning_rate": 1.607792873622148e-07, + "loss": 0.698, + "step": 10760 + }, + { + "epoch": 0.39, + "learning_rate": 1.6077562529754274e-07, + "loss": 0.6935, + "step": 10761 + }, + { + "epoch": 0.39, + "learning_rate": 1.6077196323287067e-07, + "loss": 0.6969, + "step": 10762 + }, + { + "epoch": 0.39, + "learning_rate": 1.6076830116819862e-07, + "loss": 0.6895, + "step": 10763 + }, + { + "epoch": 0.39, + "learning_rate": 1.6076463910352655e-07, + "loss": 0.6982, + "step": 10764 + }, + { + "epoch": 0.39, + "learning_rate": 1.607609770388545e-07, + "loss": 0.7156, + "step": 10765 + }, + { + "epoch": 0.39, + "learning_rate": 1.6075731497418243e-07, + "loss": 0.7032, + "step": 10766 + }, + { + "epoch": 0.39, + "learning_rate": 1.6075365290951038e-07, + "loss": 0.7033, + "step": 10767 + }, + { + "epoch": 0.39, + "learning_rate": 1.607499908448383e-07, + "loss": 0.6988, + "step": 10768 + }, + { + "epoch": 0.39, + "learning_rate": 1.6074632878016626e-07, + "loss": 0.697, + "step": 10769 + }, + { + "epoch": 0.39, + "learning_rate": 1.607426667154942e-07, + "loss": 0.7, + "step": 10770 + }, + { + "epoch": 0.39, + "learning_rate": 1.6073900465082214e-07, + "loss": 0.6962, + "step": 10771 + }, + { + "epoch": 0.39, + "learning_rate": 1.6073534258615007e-07, + "loss": 0.7018, + "step": 10772 + }, + { + "epoch": 0.39, + "learning_rate": 1.6073168052147802e-07, + "loss": 0.6887, + "step": 10773 + }, + { + "epoch": 0.39, + "learning_rate": 1.6072801845680595e-07, + "loss": 0.6802, + "step": 10774 + }, + { + "epoch": 0.39, + "learning_rate": 1.6072435639213387e-07, + "loss": 0.683, + "step": 10775 + }, + { + "epoch": 0.39, + "learning_rate": 1.6072069432746183e-07, + "loss": 0.7069, + "step": 10776 + }, + { + "epoch": 0.39, + "learning_rate": 1.6071703226278975e-07, + "loss": 0.6964, + "step": 10777 + }, + { + "epoch": 0.39, + "learning_rate": 1.6071337019811768e-07, + "loss": 0.6849, + "step": 10778 + }, + { + "epoch": 0.39, + "learning_rate": 1.607097081334456e-07, + "loss": 0.6939, + "step": 10779 + }, + { + "epoch": 0.39, + "learning_rate": 1.6070604606877356e-07, + "loss": 0.6789, + "step": 10780 + }, + { + "epoch": 0.39, + "learning_rate": 1.607023840041015e-07, + "loss": 0.7107, + "step": 10781 + }, + { + "epoch": 0.39, + "learning_rate": 1.6069872193942944e-07, + "loss": 0.7043, + "step": 10782 + }, + { + "epoch": 0.39, + "learning_rate": 1.6069505987475737e-07, + "loss": 0.7025, + "step": 10783 + }, + { + "epoch": 0.39, + "learning_rate": 1.6069139781008532e-07, + "loss": 0.689, + "step": 10784 + }, + { + "epoch": 0.39, + "learning_rate": 1.6068773574541325e-07, + "loss": 0.6815, + "step": 10785 + }, + { + "epoch": 0.39, + "learning_rate": 1.606840736807412e-07, + "loss": 0.6995, + "step": 10786 + }, + { + "epoch": 0.39, + "learning_rate": 1.6068041161606913e-07, + "loss": 0.6865, + "step": 10787 + }, + { + "epoch": 0.39, + "learning_rate": 1.6067674955139708e-07, + "loss": 0.7157, + "step": 10788 + }, + { + "epoch": 0.39, + "learning_rate": 1.60673087486725e-07, + "loss": 0.6971, + "step": 10789 + }, + { + "epoch": 0.39, + "learning_rate": 1.6066942542205296e-07, + "loss": 0.6817, + "step": 10790 + }, + { + "epoch": 0.39, + "learning_rate": 1.606657633573809e-07, + "loss": 0.7079, + "step": 10791 + }, + { + "epoch": 0.39, + "learning_rate": 1.6066210129270884e-07, + "loss": 0.706, + "step": 10792 + }, + { + "epoch": 0.39, + "learning_rate": 1.6065843922803677e-07, + "loss": 0.6931, + "step": 10793 + }, + { + "epoch": 0.39, + "learning_rate": 1.6065477716336472e-07, + "loss": 0.6811, + "step": 10794 + }, + { + "epoch": 0.39, + "learning_rate": 1.6065111509869265e-07, + "loss": 0.7105, + "step": 10795 + }, + { + "epoch": 0.39, + "learning_rate": 1.6064745303402057e-07, + "loss": 0.6849, + "step": 10796 + }, + { + "epoch": 0.4, + "learning_rate": 1.606437909693485e-07, + "loss": 0.7037, + "step": 10797 + }, + { + "epoch": 0.4, + "learning_rate": 1.6064012890467645e-07, + "loss": 0.6862, + "step": 10798 + }, + { + "epoch": 0.4, + "learning_rate": 1.6063646684000438e-07, + "loss": 0.6949, + "step": 10799 + }, + { + "epoch": 0.4, + "learning_rate": 1.6063280477533233e-07, + "loss": 0.7052, + "step": 10800 + }, + { + "epoch": 0.4, + "learning_rate": 1.6062914271066026e-07, + "loss": 0.6902, + "step": 10801 + }, + { + "epoch": 0.4, + "learning_rate": 1.6062548064598819e-07, + "loss": 0.6994, + "step": 10802 + }, + { + "epoch": 0.4, + "learning_rate": 1.6062181858131614e-07, + "loss": 0.7108, + "step": 10803 + }, + { + "epoch": 0.4, + "learning_rate": 1.6061815651664407e-07, + "loss": 0.6889, + "step": 10804 + }, + { + "epoch": 0.4, + "learning_rate": 1.6061449445197202e-07, + "loss": 0.6999, + "step": 10805 + }, + { + "epoch": 0.4, + "learning_rate": 1.6061083238729995e-07, + "loss": 0.6842, + "step": 10806 + }, + { + "epoch": 0.4, + "learning_rate": 1.606071703226279e-07, + "loss": 0.6885, + "step": 10807 + }, + { + "epoch": 0.4, + "learning_rate": 1.6060350825795583e-07, + "loss": 0.6995, + "step": 10808 + }, + { + "epoch": 0.4, + "learning_rate": 1.6059984619328378e-07, + "loss": 0.6884, + "step": 10809 + }, + { + "epoch": 0.4, + "learning_rate": 1.605961841286117e-07, + "loss": 0.7007, + "step": 10810 + }, + { + "epoch": 0.4, + "learning_rate": 1.6059252206393966e-07, + "loss": 0.7304, + "step": 10811 + }, + { + "epoch": 0.4, + "learning_rate": 1.6058885999926759e-07, + "loss": 0.6772, + "step": 10812 + }, + { + "epoch": 0.4, + "learning_rate": 1.6058519793459554e-07, + "loss": 0.7021, + "step": 10813 + }, + { + "epoch": 0.4, + "learning_rate": 1.6058153586992344e-07, + "loss": 0.6775, + "step": 10814 + }, + { + "epoch": 0.4, + "learning_rate": 1.605778738052514e-07, + "loss": 0.6984, + "step": 10815 + }, + { + "epoch": 0.4, + "learning_rate": 1.6057421174057932e-07, + "loss": 0.685, + "step": 10816 + }, + { + "epoch": 0.4, + "learning_rate": 1.6057054967590727e-07, + "loss": 0.695, + "step": 10817 + }, + { + "epoch": 0.4, + "learning_rate": 1.605668876112352e-07, + "loss": 0.694, + "step": 10818 + }, + { + "epoch": 0.4, + "learning_rate": 1.6056322554656315e-07, + "loss": 0.6978, + "step": 10819 + }, + { + "epoch": 0.4, + "learning_rate": 1.6055956348189108e-07, + "loss": 0.6976, + "step": 10820 + }, + { + "epoch": 0.4, + "learning_rate": 1.6055590141721903e-07, + "loss": 0.6892, + "step": 10821 + }, + { + "epoch": 0.4, + "learning_rate": 1.6055223935254696e-07, + "loss": 0.6806, + "step": 10822 + }, + { + "epoch": 0.4, + "learning_rate": 1.6054857728787489e-07, + "loss": 0.7, + "step": 10823 + }, + { + "epoch": 0.4, + "learning_rate": 1.6054491522320284e-07, + "loss": 0.6963, + "step": 10824 + }, + { + "epoch": 0.4, + "learning_rate": 1.6054125315853077e-07, + "loss": 0.6767, + "step": 10825 + }, + { + "epoch": 0.4, + "learning_rate": 1.6053759109385872e-07, + "loss": 0.6853, + "step": 10826 + }, + { + "epoch": 0.4, + "learning_rate": 1.6053392902918664e-07, + "loss": 0.6975, + "step": 10827 + }, + { + "epoch": 0.4, + "learning_rate": 1.605302669645146e-07, + "loss": 0.7129, + "step": 10828 + }, + { + "epoch": 0.4, + "learning_rate": 1.6052660489984252e-07, + "loss": 0.7024, + "step": 10829 + }, + { + "epoch": 0.4, + "learning_rate": 1.6052294283517048e-07, + "loss": 0.6984, + "step": 10830 + }, + { + "epoch": 0.4, + "learning_rate": 1.6051928077049838e-07, + "loss": 0.7021, + "step": 10831 + }, + { + "epoch": 0.4, + "learning_rate": 1.6051561870582633e-07, + "loss": 0.6999, + "step": 10832 + }, + { + "epoch": 0.4, + "learning_rate": 1.6051195664115426e-07, + "loss": 0.7, + "step": 10833 + }, + { + "epoch": 0.4, + "learning_rate": 1.605082945764822e-07, + "loss": 0.6863, + "step": 10834 + }, + { + "epoch": 0.4, + "learning_rate": 1.6050463251181014e-07, + "loss": 0.7019, + "step": 10835 + }, + { + "epoch": 0.4, + "learning_rate": 1.605009704471381e-07, + "loss": 0.6831, + "step": 10836 + }, + { + "epoch": 0.4, + "learning_rate": 1.6049730838246602e-07, + "loss": 0.6859, + "step": 10837 + }, + { + "epoch": 0.4, + "learning_rate": 1.6049364631779397e-07, + "loss": 0.6832, + "step": 10838 + }, + { + "epoch": 0.4, + "learning_rate": 1.604899842531219e-07, + "loss": 0.6876, + "step": 10839 + }, + { + "epoch": 0.4, + "learning_rate": 1.6048632218844985e-07, + "loss": 0.6952, + "step": 10840 + }, + { + "epoch": 0.4, + "learning_rate": 1.6048266012377778e-07, + "loss": 0.689, + "step": 10841 + }, + { + "epoch": 0.4, + "learning_rate": 1.6047899805910573e-07, + "loss": 0.6983, + "step": 10842 + }, + { + "epoch": 0.4, + "learning_rate": 1.6047533599443366e-07, + "loss": 0.6925, + "step": 10843 + }, + { + "epoch": 0.4, + "learning_rate": 1.604716739297616e-07, + "loss": 0.6994, + "step": 10844 + }, + { + "epoch": 0.4, + "learning_rate": 1.6046801186508954e-07, + "loss": 0.6862, + "step": 10845 + }, + { + "epoch": 0.4, + "learning_rate": 1.6046434980041746e-07, + "loss": 0.6975, + "step": 10846 + }, + { + "epoch": 0.4, + "learning_rate": 1.6046068773574542e-07, + "loss": 0.6882, + "step": 10847 + }, + { + "epoch": 0.4, + "learning_rate": 1.6045702567107334e-07, + "loss": 0.7052, + "step": 10848 + }, + { + "epoch": 0.4, + "learning_rate": 1.6045336360640127e-07, + "loss": 0.6988, + "step": 10849 + }, + { + "epoch": 0.4, + "learning_rate": 1.604497015417292e-07, + "loss": 0.6899, + "step": 10850 + }, + { + "epoch": 0.4, + "learning_rate": 1.6044603947705715e-07, + "loss": 0.7035, + "step": 10851 + }, + { + "epoch": 0.4, + "learning_rate": 1.6044237741238508e-07, + "loss": 0.6983, + "step": 10852 + }, + { + "epoch": 0.4, + "learning_rate": 1.6043871534771303e-07, + "loss": 0.6887, + "step": 10853 + }, + { + "epoch": 0.4, + "learning_rate": 1.6043505328304096e-07, + "loss": 0.6942, + "step": 10854 + }, + { + "epoch": 0.4, + "learning_rate": 1.604313912183689e-07, + "loss": 0.6996, + "step": 10855 + }, + { + "epoch": 0.4, + "learning_rate": 1.6042772915369684e-07, + "loss": 0.6906, + "step": 10856 + }, + { + "epoch": 0.4, + "learning_rate": 1.604240670890248e-07, + "loss": 0.694, + "step": 10857 + }, + { + "epoch": 0.4, + "learning_rate": 1.6042040502435272e-07, + "loss": 0.6932, + "step": 10858 + }, + { + "epoch": 0.4, + "learning_rate": 1.6041674295968067e-07, + "loss": 0.6871, + "step": 10859 + }, + { + "epoch": 0.4, + "learning_rate": 1.604130808950086e-07, + "loss": 0.6976, + "step": 10860 + }, + { + "epoch": 0.4, + "learning_rate": 1.6040941883033655e-07, + "loss": 0.6963, + "step": 10861 + }, + { + "epoch": 0.4, + "learning_rate": 1.6040575676566448e-07, + "loss": 0.7051, + "step": 10862 + }, + { + "epoch": 0.4, + "learning_rate": 1.6040209470099243e-07, + "loss": 0.6837, + "step": 10863 + }, + { + "epoch": 0.4, + "learning_rate": 1.6039843263632036e-07, + "loss": 0.6876, + "step": 10864 + }, + { + "epoch": 0.4, + "learning_rate": 1.603947705716483e-07, + "loss": 0.6859, + "step": 10865 + }, + { + "epoch": 0.4, + "learning_rate": 1.6039110850697624e-07, + "loss": 0.692, + "step": 10866 + }, + { + "epoch": 0.4, + "learning_rate": 1.6038744644230416e-07, + "loss": 0.6904, + "step": 10867 + }, + { + "epoch": 0.4, + "learning_rate": 1.603837843776321e-07, + "loss": 0.6977, + "step": 10868 + }, + { + "epoch": 0.4, + "learning_rate": 1.6038012231296004e-07, + "loss": 0.7, + "step": 10869 + }, + { + "epoch": 0.4, + "learning_rate": 1.6037646024828797e-07, + "loss": 0.7001, + "step": 10870 + }, + { + "epoch": 0.4, + "learning_rate": 1.6037279818361592e-07, + "loss": 0.6946, + "step": 10871 + }, + { + "epoch": 0.4, + "learning_rate": 1.6036913611894385e-07, + "loss": 0.6818, + "step": 10872 + }, + { + "epoch": 0.4, + "learning_rate": 1.6036547405427178e-07, + "loss": 0.7013, + "step": 10873 + }, + { + "epoch": 0.4, + "learning_rate": 1.6036181198959973e-07, + "loss": 0.6797, + "step": 10874 + }, + { + "epoch": 0.4, + "learning_rate": 1.6035814992492766e-07, + "loss": 0.694, + "step": 10875 + }, + { + "epoch": 0.4, + "learning_rate": 1.603544878602556e-07, + "loss": 0.6949, + "step": 10876 + }, + { + "epoch": 0.4, + "learning_rate": 1.6035082579558354e-07, + "loss": 0.6961, + "step": 10877 + }, + { + "epoch": 0.4, + "learning_rate": 1.603471637309115e-07, + "loss": 0.6896, + "step": 10878 + }, + { + "epoch": 0.4, + "learning_rate": 1.6034350166623942e-07, + "loss": 0.7008, + "step": 10879 + }, + { + "epoch": 0.4, + "learning_rate": 1.6033983960156737e-07, + "loss": 0.6947, + "step": 10880 + }, + { + "epoch": 0.4, + "learning_rate": 1.603361775368953e-07, + "loss": 0.6908, + "step": 10881 + }, + { + "epoch": 0.4, + "learning_rate": 1.6033251547222325e-07, + "loss": 0.6741, + "step": 10882 + }, + { + "epoch": 0.4, + "learning_rate": 1.6032885340755118e-07, + "loss": 0.6918, + "step": 10883 + }, + { + "epoch": 0.4, + "learning_rate": 1.6032519134287913e-07, + "loss": 0.6915, + "step": 10884 + }, + { + "epoch": 0.4, + "learning_rate": 1.6032152927820703e-07, + "loss": 0.6951, + "step": 10885 + }, + { + "epoch": 0.4, + "learning_rate": 1.6031786721353498e-07, + "loss": 0.701, + "step": 10886 + }, + { + "epoch": 0.4, + "learning_rate": 1.603142051488629e-07, + "loss": 0.6903, + "step": 10887 + }, + { + "epoch": 0.4, + "learning_rate": 1.6031054308419086e-07, + "loss": 0.6959, + "step": 10888 + }, + { + "epoch": 0.4, + "learning_rate": 1.603068810195188e-07, + "loss": 0.6988, + "step": 10889 + }, + { + "epoch": 0.4, + "learning_rate": 1.6030321895484674e-07, + "loss": 0.6937, + "step": 10890 + }, + { + "epoch": 0.4, + "learning_rate": 1.6029955689017467e-07, + "loss": 0.7037, + "step": 10891 + }, + { + "epoch": 0.4, + "learning_rate": 1.6029589482550262e-07, + "loss": 0.7037, + "step": 10892 + }, + { + "epoch": 0.4, + "learning_rate": 1.6029223276083055e-07, + "loss": 0.6937, + "step": 10893 + }, + { + "epoch": 0.4, + "learning_rate": 1.6028857069615847e-07, + "loss": 0.6946, + "step": 10894 + }, + { + "epoch": 0.4, + "learning_rate": 1.6028490863148643e-07, + "loss": 0.6665, + "step": 10895 + }, + { + "epoch": 0.4, + "learning_rate": 1.6028124656681435e-07, + "loss": 0.6832, + "step": 10896 + }, + { + "epoch": 0.4, + "learning_rate": 1.602775845021423e-07, + "loss": 0.6885, + "step": 10897 + }, + { + "epoch": 0.4, + "learning_rate": 1.6027392243747023e-07, + "loss": 0.6699, + "step": 10898 + }, + { + "epoch": 0.4, + "learning_rate": 1.602702603727982e-07, + "loss": 0.7016, + "step": 10899 + }, + { + "epoch": 0.4, + "learning_rate": 1.6026659830812611e-07, + "loss": 0.7033, + "step": 10900 + }, + { + "epoch": 0.4, + "learning_rate": 1.6026293624345407e-07, + "loss": 0.6929, + "step": 10901 + }, + { + "epoch": 0.4, + "learning_rate": 1.60259274178782e-07, + "loss": 0.6954, + "step": 10902 + }, + { + "epoch": 0.4, + "learning_rate": 1.6025561211410992e-07, + "loss": 0.6822, + "step": 10903 + }, + { + "epoch": 0.4, + "learning_rate": 1.6025195004943785e-07, + "loss": 0.6837, + "step": 10904 + }, + { + "epoch": 0.4, + "learning_rate": 1.602482879847658e-07, + "loss": 0.7004, + "step": 10905 + }, + { + "epoch": 0.4, + "learning_rate": 1.6024462592009373e-07, + "loss": 0.6967, + "step": 10906 + }, + { + "epoch": 0.4, + "learning_rate": 1.6024096385542168e-07, + "loss": 0.6944, + "step": 10907 + }, + { + "epoch": 0.4, + "learning_rate": 1.602373017907496e-07, + "loss": 0.6984, + "step": 10908 + }, + { + "epoch": 0.4, + "learning_rate": 1.6023363972607756e-07, + "loss": 0.69, + "step": 10909 + }, + { + "epoch": 0.4, + "learning_rate": 1.602299776614055e-07, + "loss": 0.686, + "step": 10910 + }, + { + "epoch": 0.4, + "learning_rate": 1.6022631559673344e-07, + "loss": 0.6977, + "step": 10911 + }, + { + "epoch": 0.4, + "learning_rate": 1.6022265353206137e-07, + "loss": 0.6861, + "step": 10912 + }, + { + "epoch": 0.4, + "learning_rate": 1.6021899146738932e-07, + "loss": 0.6946, + "step": 10913 + }, + { + "epoch": 0.4, + "learning_rate": 1.6021532940271725e-07, + "loss": 0.6976, + "step": 10914 + }, + { + "epoch": 0.4, + "learning_rate": 1.6021166733804517e-07, + "loss": 0.6864, + "step": 10915 + }, + { + "epoch": 0.4, + "learning_rate": 1.6020800527337313e-07, + "loss": 0.7088, + "step": 10916 + }, + { + "epoch": 0.4, + "learning_rate": 1.6020434320870105e-07, + "loss": 0.7001, + "step": 10917 + }, + { + "epoch": 0.4, + "learning_rate": 1.60200681144029e-07, + "loss": 0.7093, + "step": 10918 + }, + { + "epoch": 0.4, + "learning_rate": 1.6019701907935693e-07, + "loss": 0.6883, + "step": 10919 + }, + { + "epoch": 0.4, + "learning_rate": 1.6019335701468489e-07, + "loss": 0.7169, + "step": 10920 + }, + { + "epoch": 0.4, + "learning_rate": 1.6018969495001279e-07, + "loss": 0.6986, + "step": 10921 + }, + { + "epoch": 0.4, + "learning_rate": 1.6018603288534074e-07, + "loss": 0.7011, + "step": 10922 + }, + { + "epoch": 0.4, + "learning_rate": 1.6018237082066867e-07, + "loss": 0.6953, + "step": 10923 + }, + { + "epoch": 0.4, + "learning_rate": 1.6017870875599662e-07, + "loss": 0.6904, + "step": 10924 + }, + { + "epoch": 0.4, + "learning_rate": 1.6017504669132455e-07, + "loss": 0.6937, + "step": 10925 + }, + { + "epoch": 0.4, + "learning_rate": 1.601713846266525e-07, + "loss": 0.6968, + "step": 10926 + }, + { + "epoch": 0.4, + "learning_rate": 1.6016772256198043e-07, + "loss": 0.7011, + "step": 10927 + }, + { + "epoch": 0.4, + "learning_rate": 1.6016406049730838e-07, + "loss": 0.7011, + "step": 10928 + }, + { + "epoch": 0.4, + "learning_rate": 1.601603984326363e-07, + "loss": 0.6799, + "step": 10929 + }, + { + "epoch": 0.4, + "learning_rate": 1.6015673636796426e-07, + "loss": 0.6836, + "step": 10930 + }, + { + "epoch": 0.4, + "learning_rate": 1.6015307430329219e-07, + "loss": 0.689, + "step": 10931 + }, + { + "epoch": 0.4, + "learning_rate": 1.6014941223862014e-07, + "loss": 0.703, + "step": 10932 + }, + { + "epoch": 0.4, + "learning_rate": 1.6014575017394807e-07, + "loss": 0.6921, + "step": 10933 + }, + { + "epoch": 0.4, + "learning_rate": 1.6014208810927602e-07, + "loss": 0.7061, + "step": 10934 + }, + { + "epoch": 0.4, + "learning_rate": 1.6013842604460395e-07, + "loss": 0.6865, + "step": 10935 + }, + { + "epoch": 0.4, + "learning_rate": 1.601347639799319e-07, + "loss": 0.6874, + "step": 10936 + }, + { + "epoch": 0.4, + "learning_rate": 1.6013110191525983e-07, + "loss": 0.6839, + "step": 10937 + }, + { + "epoch": 0.4, + "learning_rate": 1.6012743985058775e-07, + "loss": 0.7035, + "step": 10938 + }, + { + "epoch": 0.4, + "learning_rate": 1.6012377778591568e-07, + "loss": 0.6916, + "step": 10939 + }, + { + "epoch": 0.4, + "learning_rate": 1.6012011572124363e-07, + "loss": 0.6923, + "step": 10940 + }, + { + "epoch": 0.4, + "learning_rate": 1.6011645365657156e-07, + "loss": 0.7107, + "step": 10941 + }, + { + "epoch": 0.4, + "learning_rate": 1.601127915918995e-07, + "loss": 0.6902, + "step": 10942 + }, + { + "epoch": 0.4, + "learning_rate": 1.6010912952722744e-07, + "loss": 0.6991, + "step": 10943 + }, + { + "epoch": 0.4, + "learning_rate": 1.6010546746255537e-07, + "loss": 0.7098, + "step": 10944 + }, + { + "epoch": 0.4, + "learning_rate": 1.6010180539788332e-07, + "loss": 0.6879, + "step": 10945 + }, + { + "epoch": 0.4, + "learning_rate": 1.6009814333321125e-07, + "loss": 0.6873, + "step": 10946 + }, + { + "epoch": 0.4, + "learning_rate": 1.600944812685392e-07, + "loss": 0.6964, + "step": 10947 + }, + { + "epoch": 0.4, + "learning_rate": 1.6009081920386712e-07, + "loss": 0.6973, + "step": 10948 + }, + { + "epoch": 0.4, + "learning_rate": 1.6008715713919508e-07, + "loss": 0.6973, + "step": 10949 + }, + { + "epoch": 0.4, + "learning_rate": 1.60083495074523e-07, + "loss": 0.6958, + "step": 10950 + }, + { + "epoch": 0.4, + "learning_rate": 1.6007983300985096e-07, + "loss": 0.7074, + "step": 10951 + }, + { + "epoch": 0.4, + "learning_rate": 1.6007617094517888e-07, + "loss": 0.6945, + "step": 10952 + }, + { + "epoch": 0.4, + "learning_rate": 1.6007250888050684e-07, + "loss": 0.6896, + "step": 10953 + }, + { + "epoch": 0.4, + "learning_rate": 1.6006884681583476e-07, + "loss": 0.6963, + "step": 10954 + }, + { + "epoch": 0.4, + "learning_rate": 1.6006518475116272e-07, + "loss": 0.6995, + "step": 10955 + }, + { + "epoch": 0.4, + "learning_rate": 1.6006152268649064e-07, + "loss": 0.7002, + "step": 10956 + }, + { + "epoch": 0.4, + "learning_rate": 1.6005786062181857e-07, + "loss": 0.7088, + "step": 10957 + }, + { + "epoch": 0.4, + "learning_rate": 1.600541985571465e-07, + "loss": 0.6905, + "step": 10958 + }, + { + "epoch": 0.4, + "learning_rate": 1.6005053649247445e-07, + "loss": 0.6846, + "step": 10959 + }, + { + "epoch": 0.4, + "learning_rate": 1.6004687442780238e-07, + "loss": 0.6781, + "step": 10960 + }, + { + "epoch": 0.4, + "learning_rate": 1.6004321236313033e-07, + "loss": 0.6892, + "step": 10961 + }, + { + "epoch": 0.4, + "learning_rate": 1.6003955029845826e-07, + "loss": 0.7021, + "step": 10962 + }, + { + "epoch": 0.4, + "learning_rate": 1.600358882337862e-07, + "loss": 0.6959, + "step": 10963 + }, + { + "epoch": 0.4, + "learning_rate": 1.6003222616911414e-07, + "loss": 0.7041, + "step": 10964 + }, + { + "epoch": 0.4, + "learning_rate": 1.6002856410444206e-07, + "loss": 0.6849, + "step": 10965 + }, + { + "epoch": 0.4, + "learning_rate": 1.6002490203977002e-07, + "loss": 0.6884, + "step": 10966 + }, + { + "epoch": 0.4, + "learning_rate": 1.6002123997509794e-07, + "loss": 0.7016, + "step": 10967 + }, + { + "epoch": 0.4, + "learning_rate": 1.600175779104259e-07, + "loss": 0.7032, + "step": 10968 + }, + { + "epoch": 0.4, + "learning_rate": 1.6001391584575382e-07, + "loss": 0.7012, + "step": 10969 + }, + { + "epoch": 0.4, + "learning_rate": 1.6001025378108178e-07, + "loss": 0.698, + "step": 10970 + }, + { + "epoch": 0.4, + "learning_rate": 1.600065917164097e-07, + "loss": 0.688, + "step": 10971 + }, + { + "epoch": 0.4, + "learning_rate": 1.6000292965173766e-07, + "loss": 0.7076, + "step": 10972 + }, + { + "epoch": 0.4, + "learning_rate": 1.5999926758706558e-07, + "loss": 0.6792, + "step": 10973 + }, + { + "epoch": 0.4, + "learning_rate": 1.5999560552239354e-07, + "loss": 0.6867, + "step": 10974 + }, + { + "epoch": 0.4, + "learning_rate": 1.5999194345772144e-07, + "loss": 0.6955, + "step": 10975 + }, + { + "epoch": 0.4, + "learning_rate": 1.599882813930494e-07, + "loss": 0.6987, + "step": 10976 + }, + { + "epoch": 0.4, + "learning_rate": 1.5998461932837732e-07, + "loss": 0.6907, + "step": 10977 + }, + { + "epoch": 0.4, + "learning_rate": 1.5998095726370527e-07, + "loss": 0.6887, + "step": 10978 + }, + { + "epoch": 0.4, + "learning_rate": 1.599772951990332e-07, + "loss": 0.7001, + "step": 10979 + }, + { + "epoch": 0.4, + "learning_rate": 1.5997363313436115e-07, + "loss": 0.6877, + "step": 10980 + }, + { + "epoch": 0.4, + "learning_rate": 1.5996997106968908e-07, + "loss": 0.6999, + "step": 10981 + }, + { + "epoch": 0.4, + "learning_rate": 1.5996630900501703e-07, + "loss": 0.6828, + "step": 10982 + }, + { + "epoch": 0.4, + "learning_rate": 1.5996264694034496e-07, + "loss": 0.7045, + "step": 10983 + }, + { + "epoch": 0.4, + "learning_rate": 1.599589848756729e-07, + "loss": 0.686, + "step": 10984 + }, + { + "epoch": 0.4, + "learning_rate": 1.5995532281100084e-07, + "loss": 0.6893, + "step": 10985 + }, + { + "epoch": 0.4, + "learning_rate": 1.5995166074632876e-07, + "loss": 0.6882, + "step": 10986 + }, + { + "epoch": 0.4, + "learning_rate": 1.5994799868165672e-07, + "loss": 0.6882, + "step": 10987 + }, + { + "epoch": 0.4, + "learning_rate": 1.5994433661698464e-07, + "loss": 0.6778, + "step": 10988 + }, + { + "epoch": 0.4, + "learning_rate": 1.599406745523126e-07, + "loss": 0.6872, + "step": 10989 + }, + { + "epoch": 0.4, + "learning_rate": 1.5993701248764052e-07, + "loss": 0.701, + "step": 10990 + }, + { + "epoch": 0.4, + "learning_rate": 1.5993335042296848e-07, + "loss": 0.69, + "step": 10991 + }, + { + "epoch": 0.4, + "learning_rate": 1.599296883582964e-07, + "loss": 0.6933, + "step": 10992 + }, + { + "epoch": 0.4, + "learning_rate": 1.5992602629362433e-07, + "loss": 0.7072, + "step": 10993 + }, + { + "epoch": 0.4, + "learning_rate": 1.5992236422895226e-07, + "loss": 0.6951, + "step": 10994 + }, + { + "epoch": 0.4, + "learning_rate": 1.599187021642802e-07, + "loss": 0.7022, + "step": 10995 + }, + { + "epoch": 0.4, + "learning_rate": 1.5991504009960814e-07, + "loss": 0.7134, + "step": 10996 + }, + { + "epoch": 0.4, + "learning_rate": 1.599113780349361e-07, + "loss": 0.6951, + "step": 10997 + }, + { + "epoch": 0.4, + "learning_rate": 1.5990771597026402e-07, + "loss": 0.6938, + "step": 10998 + }, + { + "epoch": 0.4, + "learning_rate": 1.5990405390559197e-07, + "loss": 0.6967, + "step": 10999 + }, + { + "epoch": 0.4, + "learning_rate": 1.599003918409199e-07, + "loss": 0.7006, + "step": 11000 + }, + { + "epoch": 0.4, + "learning_rate": 1.5989672977624785e-07, + "loss": 0.6972, + "step": 11001 + }, + { + "epoch": 0.4, + "learning_rate": 1.5989306771157578e-07, + "loss": 0.707, + "step": 11002 + }, + { + "epoch": 0.4, + "learning_rate": 1.5988940564690373e-07, + "loss": 0.6758, + "step": 11003 + }, + { + "epoch": 0.4, + "learning_rate": 1.5988574358223166e-07, + "loss": 0.6903, + "step": 11004 + }, + { + "epoch": 0.4, + "learning_rate": 1.598820815175596e-07, + "loss": 0.6994, + "step": 11005 + }, + { + "epoch": 0.4, + "learning_rate": 1.5987841945288753e-07, + "loss": 0.6948, + "step": 11006 + }, + { + "epoch": 0.4, + "learning_rate": 1.598747573882155e-07, + "loss": 0.6873, + "step": 11007 + }, + { + "epoch": 0.4, + "learning_rate": 1.5987109532354341e-07, + "loss": 0.6991, + "step": 11008 + }, + { + "epoch": 0.4, + "learning_rate": 1.5986743325887134e-07, + "loss": 0.6893, + "step": 11009 + }, + { + "epoch": 0.4, + "learning_rate": 1.598637711941993e-07, + "loss": 0.687, + "step": 11010 + }, + { + "epoch": 0.4, + "learning_rate": 1.5986010912952722e-07, + "loss": 0.687, + "step": 11011 + }, + { + "epoch": 0.4, + "learning_rate": 1.5985644706485515e-07, + "loss": 0.7173, + "step": 11012 + }, + { + "epoch": 0.4, + "learning_rate": 1.598527850001831e-07, + "loss": 0.6792, + "step": 11013 + }, + { + "epoch": 0.4, + "learning_rate": 1.5984912293551103e-07, + "loss": 0.6854, + "step": 11014 + }, + { + "epoch": 0.4, + "learning_rate": 1.5984546087083895e-07, + "loss": 0.6958, + "step": 11015 + }, + { + "epoch": 0.4, + "learning_rate": 1.598417988061669e-07, + "loss": 0.6911, + "step": 11016 + }, + { + "epoch": 0.4, + "learning_rate": 1.5983813674149483e-07, + "loss": 0.6898, + "step": 11017 + }, + { + "epoch": 0.4, + "learning_rate": 1.598344746768228e-07, + "loss": 0.6857, + "step": 11018 + }, + { + "epoch": 0.4, + "learning_rate": 1.5983081261215071e-07, + "loss": 0.7001, + "step": 11019 + }, + { + "epoch": 0.4, + "learning_rate": 1.5982715054747867e-07, + "loss": 0.6913, + "step": 11020 + }, + { + "epoch": 0.4, + "learning_rate": 1.598234884828066e-07, + "loss": 0.6866, + "step": 11021 + }, + { + "epoch": 0.4, + "learning_rate": 1.5981982641813455e-07, + "loss": 0.6889, + "step": 11022 + }, + { + "epoch": 0.4, + "learning_rate": 1.5981616435346247e-07, + "loss": 0.6959, + "step": 11023 + }, + { + "epoch": 0.4, + "learning_rate": 1.5981250228879043e-07, + "loss": 0.7048, + "step": 11024 + }, + { + "epoch": 0.4, + "learning_rate": 1.5980884022411835e-07, + "loss": 0.7104, + "step": 11025 + }, + { + "epoch": 0.4, + "learning_rate": 1.598051781594463e-07, + "loss": 0.6842, + "step": 11026 + }, + { + "epoch": 0.4, + "learning_rate": 1.5980151609477423e-07, + "loss": 0.6995, + "step": 11027 + }, + { + "epoch": 0.4, + "learning_rate": 1.597978540301022e-07, + "loss": 0.6863, + "step": 11028 + }, + { + "epoch": 0.4, + "learning_rate": 1.597941919654301e-07, + "loss": 0.6797, + "step": 11029 + }, + { + "epoch": 0.4, + "learning_rate": 1.5979052990075804e-07, + "loss": 0.6982, + "step": 11030 + }, + { + "epoch": 0.4, + "learning_rate": 1.5978686783608597e-07, + "loss": 0.6875, + "step": 11031 + }, + { + "epoch": 0.4, + "learning_rate": 1.5978320577141392e-07, + "loss": 0.6883, + "step": 11032 + }, + { + "epoch": 0.4, + "learning_rate": 1.5977954370674185e-07, + "loss": 0.704, + "step": 11033 + }, + { + "epoch": 0.4, + "learning_rate": 1.597758816420698e-07, + "loss": 0.6914, + "step": 11034 + }, + { + "epoch": 0.4, + "learning_rate": 1.5977221957739773e-07, + "loss": 0.6887, + "step": 11035 + }, + { + "epoch": 0.4, + "learning_rate": 1.5976855751272565e-07, + "loss": 0.689, + "step": 11036 + }, + { + "epoch": 0.4, + "learning_rate": 1.597648954480536e-07, + "loss": 0.6996, + "step": 11037 + }, + { + "epoch": 0.4, + "learning_rate": 1.5976123338338153e-07, + "loss": 0.7026, + "step": 11038 + }, + { + "epoch": 0.4, + "learning_rate": 1.5975757131870949e-07, + "loss": 0.6848, + "step": 11039 + }, + { + "epoch": 0.4, + "learning_rate": 1.597539092540374e-07, + "loss": 0.7018, + "step": 11040 + }, + { + "epoch": 0.4, + "learning_rate": 1.5975024718936537e-07, + "loss": 0.697, + "step": 11041 + }, + { + "epoch": 0.4, + "learning_rate": 1.597465851246933e-07, + "loss": 0.6908, + "step": 11042 + }, + { + "epoch": 0.4, + "learning_rate": 1.5974292306002125e-07, + "loss": 0.6975, + "step": 11043 + }, + { + "epoch": 0.4, + "learning_rate": 1.5973926099534917e-07, + "loss": 0.6863, + "step": 11044 + }, + { + "epoch": 0.4, + "learning_rate": 1.5973559893067713e-07, + "loss": 0.6872, + "step": 11045 + }, + { + "epoch": 0.4, + "learning_rate": 1.5973193686600505e-07, + "loss": 0.6925, + "step": 11046 + }, + { + "epoch": 0.4, + "learning_rate": 1.5972827480133298e-07, + "loss": 0.6881, + "step": 11047 + }, + { + "epoch": 0.4, + "learning_rate": 1.597246127366609e-07, + "loss": 0.7005, + "step": 11048 + }, + { + "epoch": 0.4, + "learning_rate": 1.5972095067198886e-07, + "loss": 0.7006, + "step": 11049 + }, + { + "epoch": 0.4, + "learning_rate": 1.5971728860731679e-07, + "loss": 0.6951, + "step": 11050 + }, + { + "epoch": 0.4, + "learning_rate": 1.5971362654264474e-07, + "loss": 0.6938, + "step": 11051 + }, + { + "epoch": 0.4, + "learning_rate": 1.5970996447797267e-07, + "loss": 0.6957, + "step": 11052 + }, + { + "epoch": 0.4, + "learning_rate": 1.5970630241330062e-07, + "loss": 0.6923, + "step": 11053 + }, + { + "epoch": 0.4, + "learning_rate": 1.5970264034862855e-07, + "loss": 0.6913, + "step": 11054 + }, + { + "epoch": 0.4, + "learning_rate": 1.596989782839565e-07, + "loss": 0.6864, + "step": 11055 + }, + { + "epoch": 0.4, + "learning_rate": 1.5969531621928443e-07, + "loss": 0.6972, + "step": 11056 + }, + { + "epoch": 0.4, + "learning_rate": 1.5969165415461235e-07, + "loss": 0.6767, + "step": 11057 + }, + { + "epoch": 0.4, + "learning_rate": 1.596879920899403e-07, + "loss": 0.6959, + "step": 11058 + }, + { + "epoch": 0.4, + "learning_rate": 1.5968433002526823e-07, + "loss": 0.6923, + "step": 11059 + }, + { + "epoch": 0.4, + "learning_rate": 1.5968066796059619e-07, + "loss": 0.7049, + "step": 11060 + }, + { + "epoch": 0.4, + "learning_rate": 1.596770058959241e-07, + "loss": 0.6843, + "step": 11061 + }, + { + "epoch": 0.4, + "learning_rate": 1.5967334383125206e-07, + "loss": 0.697, + "step": 11062 + }, + { + "epoch": 0.4, + "learning_rate": 1.5966968176658e-07, + "loss": 0.6856, + "step": 11063 + }, + { + "epoch": 0.4, + "learning_rate": 1.5966601970190794e-07, + "loss": 0.6997, + "step": 11064 + }, + { + "epoch": 0.4, + "learning_rate": 1.5966235763723585e-07, + "loss": 0.6816, + "step": 11065 + }, + { + "epoch": 0.4, + "learning_rate": 1.596586955725638e-07, + "loss": 0.6893, + "step": 11066 + }, + { + "epoch": 0.4, + "learning_rate": 1.5965503350789172e-07, + "loss": 0.6883, + "step": 11067 + }, + { + "epoch": 0.4, + "learning_rate": 1.5965137144321968e-07, + "loss": 0.6881, + "step": 11068 + }, + { + "epoch": 0.4, + "learning_rate": 1.596477093785476e-07, + "loss": 0.6936, + "step": 11069 + }, + { + "epoch": 0.41, + "learning_rate": 1.5964404731387556e-07, + "loss": 0.7087, + "step": 11070 + }, + { + "epoch": 0.41, + "learning_rate": 1.5964038524920348e-07, + "loss": 0.6939, + "step": 11071 + }, + { + "epoch": 0.41, + "learning_rate": 1.5963672318453144e-07, + "loss": 0.6853, + "step": 11072 + }, + { + "epoch": 0.41, + "learning_rate": 1.5963306111985936e-07, + "loss": 0.6845, + "step": 11073 + }, + { + "epoch": 0.41, + "learning_rate": 1.5962939905518732e-07, + "loss": 0.6844, + "step": 11074 + }, + { + "epoch": 0.41, + "learning_rate": 1.5962573699051524e-07, + "loss": 0.6828, + "step": 11075 + }, + { + "epoch": 0.41, + "learning_rate": 1.596220749258432e-07, + "loss": 0.6853, + "step": 11076 + }, + { + "epoch": 0.41, + "learning_rate": 1.5961841286117112e-07, + "loss": 0.6874, + "step": 11077 + }, + { + "epoch": 0.41, + "learning_rate": 1.5961475079649908e-07, + "loss": 0.7034, + "step": 11078 + }, + { + "epoch": 0.41, + "learning_rate": 1.59611088731827e-07, + "loss": 0.6918, + "step": 11079 + }, + { + "epoch": 0.41, + "learning_rate": 1.5960742666715493e-07, + "loss": 0.6927, + "step": 11080 + }, + { + "epoch": 0.41, + "learning_rate": 1.5960376460248288e-07, + "loss": 0.691, + "step": 11081 + }, + { + "epoch": 0.41, + "learning_rate": 1.596001025378108e-07, + "loss": 0.6891, + "step": 11082 + }, + { + "epoch": 0.41, + "learning_rate": 1.5959644047313874e-07, + "loss": 0.6983, + "step": 11083 + }, + { + "epoch": 0.41, + "learning_rate": 1.595927784084667e-07, + "loss": 0.6834, + "step": 11084 + }, + { + "epoch": 0.41, + "learning_rate": 1.5958911634379462e-07, + "loss": 0.6886, + "step": 11085 + }, + { + "epoch": 0.41, + "learning_rate": 1.5958545427912254e-07, + "loss": 0.6963, + "step": 11086 + }, + { + "epoch": 0.41, + "learning_rate": 1.595817922144505e-07, + "loss": 0.7044, + "step": 11087 + }, + { + "epoch": 0.41, + "learning_rate": 1.5957813014977842e-07, + "loss": 0.7041, + "step": 11088 + }, + { + "epoch": 0.41, + "learning_rate": 1.5957446808510638e-07, + "loss": 0.7116, + "step": 11089 + }, + { + "epoch": 0.41, + "learning_rate": 1.595708060204343e-07, + "loss": 0.6955, + "step": 11090 + }, + { + "epoch": 0.41, + "learning_rate": 1.5956714395576226e-07, + "loss": 0.6719, + "step": 11091 + }, + { + "epoch": 0.41, + "learning_rate": 1.5956348189109018e-07, + "loss": 0.692, + "step": 11092 + }, + { + "epoch": 0.41, + "learning_rate": 1.5955981982641814e-07, + "loss": 0.7043, + "step": 11093 + }, + { + "epoch": 0.41, + "learning_rate": 1.5955615776174606e-07, + "loss": 0.6758, + "step": 11094 + }, + { + "epoch": 0.41, + "learning_rate": 1.5955249569707402e-07, + "loss": 0.6992, + "step": 11095 + }, + { + "epoch": 0.41, + "learning_rate": 1.5954883363240194e-07, + "loss": 0.6818, + "step": 11096 + }, + { + "epoch": 0.41, + "learning_rate": 1.595451715677299e-07, + "loss": 0.7003, + "step": 11097 + }, + { + "epoch": 0.41, + "learning_rate": 1.5954150950305782e-07, + "loss": 0.6852, + "step": 11098 + }, + { + "epoch": 0.41, + "learning_rate": 1.5953784743838578e-07, + "loss": 0.6873, + "step": 11099 + }, + { + "epoch": 0.41, + "learning_rate": 1.595341853737137e-07, + "loss": 0.7001, + "step": 11100 + }, + { + "epoch": 0.41, + "learning_rate": 1.5953052330904163e-07, + "loss": 0.7066, + "step": 11101 + }, + { + "epoch": 0.41, + "learning_rate": 1.5952686124436956e-07, + "loss": 0.699, + "step": 11102 + }, + { + "epoch": 0.41, + "learning_rate": 1.595231991796975e-07, + "loss": 0.7119, + "step": 11103 + }, + { + "epoch": 0.41, + "learning_rate": 1.5951953711502544e-07, + "loss": 0.6924, + "step": 11104 + }, + { + "epoch": 0.41, + "learning_rate": 1.595158750503534e-07, + "loss": 0.7019, + "step": 11105 + }, + { + "epoch": 0.41, + "learning_rate": 1.5951221298568132e-07, + "loss": 0.6886, + "step": 11106 + }, + { + "epoch": 0.41, + "learning_rate": 1.5950855092100924e-07, + "loss": 0.6903, + "step": 11107 + }, + { + "epoch": 0.41, + "learning_rate": 1.595048888563372e-07, + "loss": 0.6814, + "step": 11108 + }, + { + "epoch": 0.41, + "learning_rate": 1.5950122679166512e-07, + "loss": 0.6968, + "step": 11109 + }, + { + "epoch": 0.41, + "learning_rate": 1.5949756472699308e-07, + "loss": 0.6988, + "step": 11110 + }, + { + "epoch": 0.41, + "learning_rate": 1.59493902662321e-07, + "loss": 0.681, + "step": 11111 + }, + { + "epoch": 0.41, + "learning_rate": 1.5949024059764896e-07, + "loss": 0.6982, + "step": 11112 + }, + { + "epoch": 0.41, + "learning_rate": 1.5948657853297688e-07, + "loss": 0.6932, + "step": 11113 + }, + { + "epoch": 0.41, + "learning_rate": 1.5948291646830484e-07, + "loss": 0.7029, + "step": 11114 + }, + { + "epoch": 0.41, + "learning_rate": 1.5947925440363276e-07, + "loss": 0.6965, + "step": 11115 + }, + { + "epoch": 0.41, + "learning_rate": 1.5947559233896072e-07, + "loss": 0.6874, + "step": 11116 + }, + { + "epoch": 0.41, + "learning_rate": 1.5947193027428864e-07, + "loss": 0.6916, + "step": 11117 + }, + { + "epoch": 0.41, + "learning_rate": 1.594682682096166e-07, + "loss": 0.6875, + "step": 11118 + }, + { + "epoch": 0.41, + "learning_rate": 1.594646061449445e-07, + "loss": 0.6954, + "step": 11119 + }, + { + "epoch": 0.41, + "learning_rate": 1.5946094408027245e-07, + "loss": 0.696, + "step": 11120 + }, + { + "epoch": 0.41, + "learning_rate": 1.5945728201560038e-07, + "loss": 0.7039, + "step": 11121 + }, + { + "epoch": 0.41, + "learning_rate": 1.5945361995092833e-07, + "loss": 0.6811, + "step": 11122 + }, + { + "epoch": 0.41, + "learning_rate": 1.5944995788625626e-07, + "loss": 0.7018, + "step": 11123 + }, + { + "epoch": 0.41, + "learning_rate": 1.594462958215842e-07, + "loss": 0.6819, + "step": 11124 + }, + { + "epoch": 0.41, + "learning_rate": 1.5944263375691213e-07, + "loss": 0.6972, + "step": 11125 + }, + { + "epoch": 0.41, + "learning_rate": 1.594389716922401e-07, + "loss": 0.6989, + "step": 11126 + }, + { + "epoch": 0.41, + "learning_rate": 1.5943530962756801e-07, + "loss": 0.6869, + "step": 11127 + }, + { + "epoch": 0.41, + "learning_rate": 1.5943164756289594e-07, + "loss": 0.6905, + "step": 11128 + }, + { + "epoch": 0.41, + "learning_rate": 1.594279854982239e-07, + "loss": 0.6942, + "step": 11129 + }, + { + "epoch": 0.41, + "learning_rate": 1.5942432343355182e-07, + "loss": 0.6954, + "step": 11130 + }, + { + "epoch": 0.41, + "learning_rate": 1.5942066136887977e-07, + "loss": 0.6974, + "step": 11131 + }, + { + "epoch": 0.41, + "learning_rate": 1.594169993042077e-07, + "loss": 0.6988, + "step": 11132 + }, + { + "epoch": 0.41, + "learning_rate": 1.5941333723953565e-07, + "loss": 0.6963, + "step": 11133 + }, + { + "epoch": 0.41, + "learning_rate": 1.5940967517486358e-07, + "loss": 0.6894, + "step": 11134 + }, + { + "epoch": 0.41, + "learning_rate": 1.5940601311019153e-07, + "loss": 0.6783, + "step": 11135 + }, + { + "epoch": 0.41, + "learning_rate": 1.5940235104551946e-07, + "loss": 0.6955, + "step": 11136 + }, + { + "epoch": 0.41, + "learning_rate": 1.593986889808474e-07, + "loss": 0.6883, + "step": 11137 + }, + { + "epoch": 0.41, + "learning_rate": 1.5939502691617531e-07, + "loss": 0.6952, + "step": 11138 + }, + { + "epoch": 0.41, + "learning_rate": 1.5939136485150327e-07, + "loss": 0.6938, + "step": 11139 + }, + { + "epoch": 0.41, + "learning_rate": 1.593877027868312e-07, + "loss": 0.6772, + "step": 11140 + }, + { + "epoch": 0.41, + "learning_rate": 1.5938404072215915e-07, + "loss": 0.6812, + "step": 11141 + }, + { + "epoch": 0.41, + "learning_rate": 1.5938037865748707e-07, + "loss": 0.6929, + "step": 11142 + }, + { + "epoch": 0.41, + "learning_rate": 1.5937671659281503e-07, + "loss": 0.6829, + "step": 11143 + }, + { + "epoch": 0.41, + "learning_rate": 1.5937305452814295e-07, + "loss": 0.6972, + "step": 11144 + }, + { + "epoch": 0.41, + "learning_rate": 1.593693924634709e-07, + "loss": 0.6874, + "step": 11145 + }, + { + "epoch": 0.41, + "learning_rate": 1.5936573039879883e-07, + "loss": 0.6889, + "step": 11146 + }, + { + "epoch": 0.41, + "learning_rate": 1.593620683341268e-07, + "loss": 0.6835, + "step": 11147 + }, + { + "epoch": 0.41, + "learning_rate": 1.5935840626945471e-07, + "loss": 0.7011, + "step": 11148 + }, + { + "epoch": 0.41, + "learning_rate": 1.5935474420478267e-07, + "loss": 0.6981, + "step": 11149 + }, + { + "epoch": 0.41, + "learning_rate": 1.593510821401106e-07, + "loss": 0.6998, + "step": 11150 + }, + { + "epoch": 0.41, + "learning_rate": 1.5934742007543852e-07, + "loss": 0.6696, + "step": 11151 + }, + { + "epoch": 0.41, + "learning_rate": 1.5934375801076647e-07, + "loss": 0.6973, + "step": 11152 + }, + { + "epoch": 0.41, + "learning_rate": 1.593400959460944e-07, + "loss": 0.6836, + "step": 11153 + }, + { + "epoch": 0.41, + "learning_rate": 1.5933643388142235e-07, + "loss": 0.6987, + "step": 11154 + }, + { + "epoch": 0.41, + "learning_rate": 1.5933277181675025e-07, + "loss": 0.6999, + "step": 11155 + }, + { + "epoch": 0.41, + "learning_rate": 1.593291097520782e-07, + "loss": 0.7012, + "step": 11156 + }, + { + "epoch": 0.41, + "learning_rate": 1.5932544768740613e-07, + "loss": 0.691, + "step": 11157 + }, + { + "epoch": 0.41, + "learning_rate": 1.5932178562273409e-07, + "loss": 0.6965, + "step": 11158 + }, + { + "epoch": 0.41, + "learning_rate": 1.59318123558062e-07, + "loss": 0.6934, + "step": 11159 + }, + { + "epoch": 0.41, + "learning_rate": 1.5931446149338997e-07, + "loss": 0.688, + "step": 11160 + }, + { + "epoch": 0.41, + "learning_rate": 1.593107994287179e-07, + "loss": 0.7013, + "step": 11161 + }, + { + "epoch": 0.41, + "learning_rate": 1.5930713736404585e-07, + "loss": 0.6866, + "step": 11162 + }, + { + "epoch": 0.41, + "learning_rate": 1.5930347529937377e-07, + "loss": 0.6967, + "step": 11163 + }, + { + "epoch": 0.41, + "learning_rate": 1.5929981323470173e-07, + "loss": 0.7065, + "step": 11164 + }, + { + "epoch": 0.41, + "learning_rate": 1.5929615117002965e-07, + "loss": 0.6852, + "step": 11165 + }, + { + "epoch": 0.41, + "learning_rate": 1.592924891053576e-07, + "loss": 0.683, + "step": 11166 + }, + { + "epoch": 0.41, + "learning_rate": 1.5928882704068553e-07, + "loss": 0.7003, + "step": 11167 + }, + { + "epoch": 0.41, + "learning_rate": 1.5928516497601349e-07, + "loss": 0.6916, + "step": 11168 + }, + { + "epoch": 0.41, + "learning_rate": 1.592815029113414e-07, + "loss": 0.698, + "step": 11169 + }, + { + "epoch": 0.41, + "learning_rate": 1.5927784084666937e-07, + "loss": 0.6921, + "step": 11170 + }, + { + "epoch": 0.41, + "learning_rate": 1.592741787819973e-07, + "loss": 0.6908, + "step": 11171 + }, + { + "epoch": 0.41, + "learning_rate": 1.5927051671732522e-07, + "loss": 0.701, + "step": 11172 + }, + { + "epoch": 0.41, + "learning_rate": 1.5926685465265315e-07, + "loss": 0.6949, + "step": 11173 + }, + { + "epoch": 0.41, + "learning_rate": 1.592631925879811e-07, + "loss": 0.6813, + "step": 11174 + }, + { + "epoch": 0.41, + "learning_rate": 1.5925953052330903e-07, + "loss": 0.6917, + "step": 11175 + }, + { + "epoch": 0.41, + "learning_rate": 1.5925586845863698e-07, + "loss": 0.6878, + "step": 11176 + }, + { + "epoch": 0.41, + "learning_rate": 1.592522063939649e-07, + "loss": 0.6922, + "step": 11177 + }, + { + "epoch": 0.41, + "learning_rate": 1.5924854432929283e-07, + "loss": 0.6992, + "step": 11178 + }, + { + "epoch": 0.41, + "learning_rate": 1.5924488226462079e-07, + "loss": 0.6844, + "step": 11179 + }, + { + "epoch": 0.41, + "learning_rate": 1.592412201999487e-07, + "loss": 0.6883, + "step": 11180 + }, + { + "epoch": 0.41, + "learning_rate": 1.5923755813527667e-07, + "loss": 0.6955, + "step": 11181 + }, + { + "epoch": 0.41, + "learning_rate": 1.592338960706046e-07, + "loss": 0.7082, + "step": 11182 + }, + { + "epoch": 0.41, + "learning_rate": 1.5923023400593254e-07, + "loss": 0.6957, + "step": 11183 + }, + { + "epoch": 0.41, + "learning_rate": 1.5922657194126047e-07, + "loss": 0.6832, + "step": 11184 + }, + { + "epoch": 0.41, + "learning_rate": 1.5922290987658842e-07, + "loss": 0.6795, + "step": 11185 + }, + { + "epoch": 0.41, + "learning_rate": 1.5921924781191635e-07, + "loss": 0.6969, + "step": 11186 + }, + { + "epoch": 0.41, + "learning_rate": 1.592155857472443e-07, + "loss": 0.692, + "step": 11187 + }, + { + "epoch": 0.41, + "learning_rate": 1.5921192368257223e-07, + "loss": 0.7107, + "step": 11188 + }, + { + "epoch": 0.41, + "learning_rate": 1.5920826161790018e-07, + "loss": 0.6819, + "step": 11189 + }, + { + "epoch": 0.41, + "learning_rate": 1.592045995532281e-07, + "loss": 0.7209, + "step": 11190 + }, + { + "epoch": 0.41, + "learning_rate": 1.5920093748855604e-07, + "loss": 0.6936, + "step": 11191 + }, + { + "epoch": 0.41, + "learning_rate": 1.5919727542388396e-07, + "loss": 0.7005, + "step": 11192 + }, + { + "epoch": 0.41, + "learning_rate": 1.5919361335921192e-07, + "loss": 0.6914, + "step": 11193 + }, + { + "epoch": 0.41, + "learning_rate": 1.5918995129453984e-07, + "loss": 0.7036, + "step": 11194 + }, + { + "epoch": 0.41, + "learning_rate": 1.591862892298678e-07, + "loss": 0.7017, + "step": 11195 + }, + { + "epoch": 0.41, + "learning_rate": 1.5918262716519572e-07, + "loss": 0.6852, + "step": 11196 + }, + { + "epoch": 0.41, + "learning_rate": 1.5917896510052368e-07, + "loss": 0.6802, + "step": 11197 + }, + { + "epoch": 0.41, + "learning_rate": 1.591753030358516e-07, + "loss": 0.6921, + "step": 11198 + }, + { + "epoch": 0.41, + "learning_rate": 1.5917164097117953e-07, + "loss": 0.694, + "step": 11199 + }, + { + "epoch": 0.41, + "learning_rate": 1.5916797890650748e-07, + "loss": 0.6972, + "step": 11200 + }, + { + "epoch": 0.41, + "learning_rate": 1.591643168418354e-07, + "loss": 0.6783, + "step": 11201 + }, + { + "epoch": 0.41, + "learning_rate": 1.5916065477716336e-07, + "loss": 0.6974, + "step": 11202 + }, + { + "epoch": 0.41, + "learning_rate": 1.591569927124913e-07, + "loss": 0.7067, + "step": 11203 + }, + { + "epoch": 0.41, + "learning_rate": 1.5915333064781924e-07, + "loss": 0.6796, + "step": 11204 + }, + { + "epoch": 0.41, + "learning_rate": 1.5914966858314717e-07, + "loss": 0.6964, + "step": 11205 + }, + { + "epoch": 0.41, + "learning_rate": 1.5914600651847512e-07, + "loss": 0.6895, + "step": 11206 + }, + { + "epoch": 0.41, + "learning_rate": 1.5914234445380305e-07, + "loss": 0.6915, + "step": 11207 + }, + { + "epoch": 0.41, + "learning_rate": 1.5913868238913098e-07, + "loss": 0.696, + "step": 11208 + }, + { + "epoch": 0.41, + "learning_rate": 1.591350203244589e-07, + "loss": 0.6778, + "step": 11209 + }, + { + "epoch": 0.41, + "learning_rate": 1.5913135825978686e-07, + "loss": 0.6957, + "step": 11210 + }, + { + "epoch": 0.41, + "learning_rate": 1.5912769619511478e-07, + "loss": 0.6993, + "step": 11211 + }, + { + "epoch": 0.41, + "learning_rate": 1.5912403413044274e-07, + "loss": 0.6999, + "step": 11212 + }, + { + "epoch": 0.41, + "learning_rate": 1.5912037206577066e-07, + "loss": 0.6874, + "step": 11213 + }, + { + "epoch": 0.41, + "learning_rate": 1.5911671000109862e-07, + "loss": 0.7055, + "step": 11214 + }, + { + "epoch": 0.41, + "learning_rate": 1.5911304793642654e-07, + "loss": 0.6738, + "step": 11215 + }, + { + "epoch": 0.41, + "learning_rate": 1.591093858717545e-07, + "loss": 0.6806, + "step": 11216 + }, + { + "epoch": 0.41, + "learning_rate": 1.5910572380708242e-07, + "loss": 0.6879, + "step": 11217 + }, + { + "epoch": 0.41, + "learning_rate": 1.5910206174241038e-07, + "loss": 0.6876, + "step": 11218 + }, + { + "epoch": 0.41, + "learning_rate": 1.590983996777383e-07, + "loss": 0.7023, + "step": 11219 + }, + { + "epoch": 0.41, + "learning_rate": 1.5909473761306626e-07, + "loss": 0.695, + "step": 11220 + }, + { + "epoch": 0.41, + "learning_rate": 1.5909107554839418e-07, + "loss": 0.6984, + "step": 11221 + }, + { + "epoch": 0.41, + "learning_rate": 1.590874134837221e-07, + "loss": 0.6932, + "step": 11222 + }, + { + "epoch": 0.41, + "learning_rate": 1.5908375141905006e-07, + "loss": 0.6939, + "step": 11223 + }, + { + "epoch": 0.41, + "learning_rate": 1.59080089354378e-07, + "loss": 0.6949, + "step": 11224 + }, + { + "epoch": 0.41, + "learning_rate": 1.5907642728970594e-07, + "loss": 0.7001, + "step": 11225 + }, + { + "epoch": 0.41, + "learning_rate": 1.5907276522503384e-07, + "loss": 0.7037, + "step": 11226 + }, + { + "epoch": 0.41, + "learning_rate": 1.590691031603618e-07, + "loss": 0.6936, + "step": 11227 + }, + { + "epoch": 0.41, + "learning_rate": 1.5906544109568972e-07, + "loss": 0.6941, + "step": 11228 + }, + { + "epoch": 0.41, + "learning_rate": 1.5906177903101768e-07, + "loss": 0.6944, + "step": 11229 + }, + { + "epoch": 0.41, + "learning_rate": 1.590581169663456e-07, + "loss": 0.6761, + "step": 11230 + }, + { + "epoch": 0.41, + "learning_rate": 1.5905445490167356e-07, + "loss": 0.6916, + "step": 11231 + }, + { + "epoch": 0.41, + "learning_rate": 1.5905079283700148e-07, + "loss": 0.6916, + "step": 11232 + }, + { + "epoch": 0.41, + "learning_rate": 1.5904713077232944e-07, + "loss": 0.7007, + "step": 11233 + }, + { + "epoch": 0.41, + "learning_rate": 1.5904346870765736e-07, + "loss": 0.6865, + "step": 11234 + }, + { + "epoch": 0.41, + "learning_rate": 1.5903980664298532e-07, + "loss": 0.689, + "step": 11235 + }, + { + "epoch": 0.41, + "learning_rate": 1.5903614457831324e-07, + "loss": 0.6892, + "step": 11236 + }, + { + "epoch": 0.41, + "learning_rate": 1.590324825136412e-07, + "loss": 0.6971, + "step": 11237 + }, + { + "epoch": 0.41, + "learning_rate": 1.5902882044896912e-07, + "loss": 0.6973, + "step": 11238 + }, + { + "epoch": 0.41, + "learning_rate": 1.5902515838429708e-07, + "loss": 0.6973, + "step": 11239 + }, + { + "epoch": 0.41, + "learning_rate": 1.59021496319625e-07, + "loss": 0.6956, + "step": 11240 + }, + { + "epoch": 0.41, + "learning_rate": 1.5901783425495295e-07, + "loss": 0.6965, + "step": 11241 + }, + { + "epoch": 0.41, + "learning_rate": 1.5901417219028088e-07, + "loss": 0.6899, + "step": 11242 + }, + { + "epoch": 0.41, + "learning_rate": 1.590105101256088e-07, + "loss": 0.6944, + "step": 11243 + }, + { + "epoch": 0.41, + "learning_rate": 1.5900684806093673e-07, + "loss": 0.6985, + "step": 11244 + }, + { + "epoch": 0.41, + "learning_rate": 1.590031859962647e-07, + "loss": 0.6888, + "step": 11245 + }, + { + "epoch": 0.41, + "learning_rate": 1.5899952393159261e-07, + "loss": 0.6982, + "step": 11246 + }, + { + "epoch": 0.41, + "learning_rate": 1.5899586186692057e-07, + "loss": 0.6926, + "step": 11247 + }, + { + "epoch": 0.41, + "learning_rate": 1.589921998022485e-07, + "loss": 0.6935, + "step": 11248 + }, + { + "epoch": 0.41, + "learning_rate": 1.5898853773757642e-07, + "loss": 0.6876, + "step": 11249 + }, + { + "epoch": 0.41, + "learning_rate": 1.5898487567290437e-07, + "loss": 0.6912, + "step": 11250 + }, + { + "epoch": 0.41, + "learning_rate": 1.589812136082323e-07, + "loss": 0.691, + "step": 11251 + }, + { + "epoch": 0.41, + "learning_rate": 1.5897755154356025e-07, + "loss": 0.687, + "step": 11252 + }, + { + "epoch": 0.41, + "learning_rate": 1.5897388947888818e-07, + "loss": 0.6888, + "step": 11253 + }, + { + "epoch": 0.41, + "learning_rate": 1.5897022741421613e-07, + "loss": 0.6802, + "step": 11254 + }, + { + "epoch": 0.41, + "learning_rate": 1.5896656534954406e-07, + "loss": 0.6992, + "step": 11255 + }, + { + "epoch": 0.41, + "learning_rate": 1.5896290328487201e-07, + "loss": 0.6865, + "step": 11256 + }, + { + "epoch": 0.41, + "learning_rate": 1.5895924122019994e-07, + "loss": 0.6897, + "step": 11257 + }, + { + "epoch": 0.41, + "learning_rate": 1.589555791555279e-07, + "loss": 0.6862, + "step": 11258 + }, + { + "epoch": 0.41, + "learning_rate": 1.5895191709085582e-07, + "loss": 0.698, + "step": 11259 + }, + { + "epoch": 0.41, + "learning_rate": 1.5894825502618377e-07, + "loss": 0.693, + "step": 11260 + }, + { + "epoch": 0.41, + "learning_rate": 1.589445929615117e-07, + "loss": 0.6928, + "step": 11261 + }, + { + "epoch": 0.41, + "learning_rate": 1.5894093089683963e-07, + "loss": 0.68, + "step": 11262 + }, + { + "epoch": 0.41, + "learning_rate": 1.5893726883216755e-07, + "loss": 0.689, + "step": 11263 + }, + { + "epoch": 0.41, + "learning_rate": 1.589336067674955e-07, + "loss": 0.6839, + "step": 11264 + }, + { + "epoch": 0.41, + "learning_rate": 1.5892994470282343e-07, + "loss": 0.6985, + "step": 11265 + }, + { + "epoch": 0.41, + "learning_rate": 1.589262826381514e-07, + "loss": 0.7048, + "step": 11266 + }, + { + "epoch": 0.41, + "learning_rate": 1.5892262057347931e-07, + "loss": 0.6753, + "step": 11267 + }, + { + "epoch": 0.41, + "learning_rate": 1.5891895850880727e-07, + "loss": 0.7018, + "step": 11268 + }, + { + "epoch": 0.41, + "learning_rate": 1.589152964441352e-07, + "loss": 0.7122, + "step": 11269 + }, + { + "epoch": 0.41, + "learning_rate": 1.5891163437946312e-07, + "loss": 0.6911, + "step": 11270 + }, + { + "epoch": 0.41, + "learning_rate": 1.5890797231479107e-07, + "loss": 0.6866, + "step": 11271 + }, + { + "epoch": 0.41, + "learning_rate": 1.58904310250119e-07, + "loss": 0.6924, + "step": 11272 + }, + { + "epoch": 0.41, + "learning_rate": 1.5890064818544695e-07, + "loss": 0.7022, + "step": 11273 + }, + { + "epoch": 0.41, + "learning_rate": 1.5889698612077488e-07, + "loss": 0.6933, + "step": 11274 + }, + { + "epoch": 0.41, + "learning_rate": 1.5889332405610283e-07, + "loss": 0.6813, + "step": 11275 + }, + { + "epoch": 0.41, + "learning_rate": 1.5888966199143076e-07, + "loss": 0.6721, + "step": 11276 + }, + { + "epoch": 0.41, + "learning_rate": 1.588859999267587e-07, + "loss": 0.6896, + "step": 11277 + }, + { + "epoch": 0.41, + "learning_rate": 1.5888233786208664e-07, + "loss": 0.6695, + "step": 11278 + }, + { + "epoch": 0.41, + "learning_rate": 1.588786757974146e-07, + "loss": 0.6844, + "step": 11279 + }, + { + "epoch": 0.41, + "learning_rate": 1.588750137327425e-07, + "loss": 0.6997, + "step": 11280 + }, + { + "epoch": 0.41, + "learning_rate": 1.5887135166807045e-07, + "loss": 0.6928, + "step": 11281 + }, + { + "epoch": 0.41, + "learning_rate": 1.5886768960339837e-07, + "loss": 0.691, + "step": 11282 + }, + { + "epoch": 0.41, + "learning_rate": 1.5886402753872633e-07, + "loss": 0.7065, + "step": 11283 + }, + { + "epoch": 0.41, + "learning_rate": 1.5886036547405425e-07, + "loss": 0.7123, + "step": 11284 + }, + { + "epoch": 0.41, + "learning_rate": 1.588567034093822e-07, + "loss": 0.687, + "step": 11285 + }, + { + "epoch": 0.41, + "learning_rate": 1.5885304134471013e-07, + "loss": 0.7008, + "step": 11286 + }, + { + "epoch": 0.41, + "learning_rate": 1.5884937928003809e-07, + "loss": 0.6865, + "step": 11287 + }, + { + "epoch": 0.41, + "learning_rate": 1.58845717215366e-07, + "loss": 0.6839, + "step": 11288 + }, + { + "epoch": 0.41, + "learning_rate": 1.5884205515069397e-07, + "loss": 0.6879, + "step": 11289 + }, + { + "epoch": 0.41, + "learning_rate": 1.588383930860219e-07, + "loss": 0.6994, + "step": 11290 + }, + { + "epoch": 0.41, + "learning_rate": 1.5883473102134985e-07, + "loss": 0.697, + "step": 11291 + }, + { + "epoch": 0.41, + "learning_rate": 1.5883106895667777e-07, + "loss": 0.7031, + "step": 11292 + }, + { + "epoch": 0.41, + "learning_rate": 1.588274068920057e-07, + "loss": 0.7069, + "step": 11293 + }, + { + "epoch": 0.41, + "learning_rate": 1.5882374482733365e-07, + "loss": 0.6935, + "step": 11294 + }, + { + "epoch": 0.41, + "learning_rate": 1.5882008276266158e-07, + "loss": 0.7028, + "step": 11295 + }, + { + "epoch": 0.41, + "learning_rate": 1.5881642069798953e-07, + "loss": 0.7014, + "step": 11296 + }, + { + "epoch": 0.41, + "learning_rate": 1.5881275863331746e-07, + "loss": 0.7032, + "step": 11297 + }, + { + "epoch": 0.41, + "learning_rate": 1.5880909656864539e-07, + "loss": 0.6987, + "step": 11298 + }, + { + "epoch": 0.41, + "learning_rate": 1.588054345039733e-07, + "loss": 0.6711, + "step": 11299 + }, + { + "epoch": 0.41, + "learning_rate": 1.5880177243930127e-07, + "loss": 0.6985, + "step": 11300 + }, + { + "epoch": 0.41, + "learning_rate": 1.587981103746292e-07, + "loss": 0.6908, + "step": 11301 + }, + { + "epoch": 0.41, + "learning_rate": 1.5879444830995714e-07, + "loss": 0.6864, + "step": 11302 + }, + { + "epoch": 0.41, + "learning_rate": 1.5879078624528507e-07, + "loss": 0.7, + "step": 11303 + }, + { + "epoch": 0.41, + "learning_rate": 1.5878712418061302e-07, + "loss": 0.6914, + "step": 11304 + }, + { + "epoch": 0.41, + "learning_rate": 1.5878346211594095e-07, + "loss": 0.69, + "step": 11305 + }, + { + "epoch": 0.41, + "learning_rate": 1.587798000512689e-07, + "loss": 0.6947, + "step": 11306 + }, + { + "epoch": 0.41, + "learning_rate": 1.5877613798659683e-07, + "loss": 0.6874, + "step": 11307 + }, + { + "epoch": 0.41, + "learning_rate": 1.5877247592192478e-07, + "loss": 0.6948, + "step": 11308 + }, + { + "epoch": 0.41, + "learning_rate": 1.587688138572527e-07, + "loss": 0.6862, + "step": 11309 + }, + { + "epoch": 0.41, + "learning_rate": 1.5876515179258066e-07, + "loss": 0.6872, + "step": 11310 + }, + { + "epoch": 0.41, + "learning_rate": 1.587614897279086e-07, + "loss": 0.6961, + "step": 11311 + }, + { + "epoch": 0.41, + "learning_rate": 1.5875782766323654e-07, + "loss": 0.6913, + "step": 11312 + }, + { + "epoch": 0.41, + "learning_rate": 1.5875416559856447e-07, + "loss": 0.6927, + "step": 11313 + }, + { + "epoch": 0.41, + "learning_rate": 1.587505035338924e-07, + "loss": 0.6956, + "step": 11314 + }, + { + "epoch": 0.41, + "learning_rate": 1.5874684146922035e-07, + "loss": 0.7073, + "step": 11315 + }, + { + "epoch": 0.41, + "learning_rate": 1.5874317940454828e-07, + "loss": 0.6929, + "step": 11316 + }, + { + "epoch": 0.41, + "learning_rate": 1.587395173398762e-07, + "loss": 0.6828, + "step": 11317 + }, + { + "epoch": 0.41, + "learning_rate": 1.5873585527520416e-07, + "loss": 0.69, + "step": 11318 + }, + { + "epoch": 0.41, + "learning_rate": 1.5873219321053208e-07, + "loss": 0.7085, + "step": 11319 + }, + { + "epoch": 0.41, + "learning_rate": 1.5872853114586e-07, + "loss": 0.7045, + "step": 11320 + }, + { + "epoch": 0.41, + "learning_rate": 1.5872486908118796e-07, + "loss": 0.6737, + "step": 11321 + }, + { + "epoch": 0.41, + "learning_rate": 1.587212070165159e-07, + "loss": 0.6881, + "step": 11322 + }, + { + "epoch": 0.41, + "learning_rate": 1.5871754495184384e-07, + "loss": 0.6928, + "step": 11323 + }, + { + "epoch": 0.41, + "learning_rate": 1.5871388288717177e-07, + "loss": 0.7086, + "step": 11324 + }, + { + "epoch": 0.41, + "learning_rate": 1.5871022082249972e-07, + "loss": 0.6781, + "step": 11325 + }, + { + "epoch": 0.41, + "learning_rate": 1.5870655875782765e-07, + "loss": 0.6924, + "step": 11326 + }, + { + "epoch": 0.41, + "learning_rate": 1.587028966931556e-07, + "loss": 0.6899, + "step": 11327 + }, + { + "epoch": 0.41, + "learning_rate": 1.5869923462848353e-07, + "loss": 0.6976, + "step": 11328 + }, + { + "epoch": 0.41, + "learning_rate": 1.5869557256381148e-07, + "loss": 0.6905, + "step": 11329 + }, + { + "epoch": 0.41, + "learning_rate": 1.586919104991394e-07, + "loss": 0.6949, + "step": 11330 + }, + { + "epoch": 0.41, + "learning_rate": 1.5868824843446736e-07, + "loss": 0.6935, + "step": 11331 + }, + { + "epoch": 0.41, + "learning_rate": 1.586845863697953e-07, + "loss": 0.6974, + "step": 11332 + }, + { + "epoch": 0.41, + "learning_rate": 1.5868092430512324e-07, + "loss": 0.6997, + "step": 11333 + }, + { + "epoch": 0.41, + "learning_rate": 1.5867726224045114e-07, + "loss": 0.6811, + "step": 11334 + }, + { + "epoch": 0.41, + "learning_rate": 1.586736001757791e-07, + "loss": 0.6895, + "step": 11335 + }, + { + "epoch": 0.41, + "learning_rate": 1.5866993811110702e-07, + "loss": 0.7063, + "step": 11336 + }, + { + "epoch": 0.41, + "learning_rate": 1.5866627604643498e-07, + "loss": 0.7053, + "step": 11337 + }, + { + "epoch": 0.41, + "learning_rate": 1.586626139817629e-07, + "loss": 0.6901, + "step": 11338 + }, + { + "epoch": 0.41, + "learning_rate": 1.5865895191709086e-07, + "loss": 0.6962, + "step": 11339 + }, + { + "epoch": 0.41, + "learning_rate": 1.5865528985241878e-07, + "loss": 0.7044, + "step": 11340 + }, + { + "epoch": 0.41, + "learning_rate": 1.586516277877467e-07, + "loss": 0.6795, + "step": 11341 + }, + { + "epoch": 0.41, + "learning_rate": 1.5864796572307466e-07, + "loss": 0.6953, + "step": 11342 + }, + { + "epoch": 0.42, + "learning_rate": 1.586443036584026e-07, + "loss": 0.6817, + "step": 11343 + }, + { + "epoch": 0.42, + "learning_rate": 1.5864064159373054e-07, + "loss": 0.6894, + "step": 11344 + }, + { + "epoch": 0.42, + "learning_rate": 1.5863697952905847e-07, + "loss": 0.6989, + "step": 11345 + }, + { + "epoch": 0.42, + "learning_rate": 1.5863331746438642e-07, + "loss": 0.6937, + "step": 11346 + }, + { + "epoch": 0.42, + "learning_rate": 1.5862965539971435e-07, + "loss": 0.6996, + "step": 11347 + }, + { + "epoch": 0.42, + "learning_rate": 1.586259933350423e-07, + "loss": 0.7027, + "step": 11348 + }, + { + "epoch": 0.42, + "learning_rate": 1.5862233127037023e-07, + "loss": 0.6919, + "step": 11349 + }, + { + "epoch": 0.42, + "learning_rate": 1.5861866920569818e-07, + "loss": 0.6865, + "step": 11350 + }, + { + "epoch": 0.42, + "learning_rate": 1.586150071410261e-07, + "loss": 0.6889, + "step": 11351 + }, + { + "epoch": 0.42, + "learning_rate": 1.5861134507635404e-07, + "loss": 0.6877, + "step": 11352 + }, + { + "epoch": 0.42, + "learning_rate": 1.5860768301168196e-07, + "loss": 0.7016, + "step": 11353 + }, + { + "epoch": 0.42, + "learning_rate": 1.5860402094700992e-07, + "loss": 0.7003, + "step": 11354 + }, + { + "epoch": 0.42, + "learning_rate": 1.5860035888233784e-07, + "loss": 0.6934, + "step": 11355 + }, + { + "epoch": 0.42, + "learning_rate": 1.585966968176658e-07, + "loss": 0.697, + "step": 11356 + }, + { + "epoch": 0.42, + "learning_rate": 1.5859303475299372e-07, + "loss": 0.7043, + "step": 11357 + }, + { + "epoch": 0.42, + "learning_rate": 1.5858937268832168e-07, + "loss": 0.6965, + "step": 11358 + }, + { + "epoch": 0.42, + "learning_rate": 1.585857106236496e-07, + "loss": 0.6807, + "step": 11359 + }, + { + "epoch": 0.42, + "learning_rate": 1.5858204855897755e-07, + "loss": 0.6857, + "step": 11360 + }, + { + "epoch": 0.42, + "learning_rate": 1.5857838649430548e-07, + "loss": 0.6931, + "step": 11361 + }, + { + "epoch": 0.42, + "learning_rate": 1.5857472442963343e-07, + "loss": 0.6958, + "step": 11362 + }, + { + "epoch": 0.42, + "learning_rate": 1.5857106236496136e-07, + "loss": 0.6833, + "step": 11363 + }, + { + "epoch": 0.42, + "learning_rate": 1.585674003002893e-07, + "loss": 0.6884, + "step": 11364 + }, + { + "epoch": 0.42, + "learning_rate": 1.5856373823561724e-07, + "loss": 0.6842, + "step": 11365 + }, + { + "epoch": 0.42, + "learning_rate": 1.5856007617094517e-07, + "loss": 0.703, + "step": 11366 + }, + { + "epoch": 0.42, + "learning_rate": 1.5855641410627312e-07, + "loss": 0.701, + "step": 11367 + }, + { + "epoch": 0.42, + "learning_rate": 1.5855275204160105e-07, + "loss": 0.709, + "step": 11368 + }, + { + "epoch": 0.42, + "learning_rate": 1.58549089976929e-07, + "loss": 0.6764, + "step": 11369 + }, + { + "epoch": 0.42, + "learning_rate": 1.585454279122569e-07, + "loss": 0.693, + "step": 11370 + }, + { + "epoch": 0.42, + "learning_rate": 1.5854176584758485e-07, + "loss": 0.677, + "step": 11371 + }, + { + "epoch": 0.42, + "learning_rate": 1.5853810378291278e-07, + "loss": 0.6941, + "step": 11372 + }, + { + "epoch": 0.42, + "learning_rate": 1.5853444171824073e-07, + "loss": 0.6879, + "step": 11373 + }, + { + "epoch": 0.42, + "learning_rate": 1.5853077965356866e-07, + "loss": 0.6908, + "step": 11374 + }, + { + "epoch": 0.42, + "learning_rate": 1.5852711758889661e-07, + "loss": 0.6947, + "step": 11375 + }, + { + "epoch": 0.42, + "learning_rate": 1.5852345552422454e-07, + "loss": 0.6887, + "step": 11376 + }, + { + "epoch": 0.42, + "learning_rate": 1.585197934595525e-07, + "loss": 0.6895, + "step": 11377 + }, + { + "epoch": 0.42, + "learning_rate": 1.5851613139488042e-07, + "loss": 0.6836, + "step": 11378 + }, + { + "epoch": 0.42, + "learning_rate": 1.5851246933020837e-07, + "loss": 0.6975, + "step": 11379 + }, + { + "epoch": 0.42, + "learning_rate": 1.585088072655363e-07, + "loss": 0.7209, + "step": 11380 + }, + { + "epoch": 0.42, + "learning_rate": 1.5850514520086425e-07, + "loss": 0.6923, + "step": 11381 + }, + { + "epoch": 0.42, + "learning_rate": 1.5850148313619218e-07, + "loss": 0.6726, + "step": 11382 + }, + { + "epoch": 0.42, + "learning_rate": 1.5849782107152013e-07, + "loss": 0.6982, + "step": 11383 + }, + { + "epoch": 0.42, + "learning_rate": 1.5849415900684806e-07, + "loss": 0.676, + "step": 11384 + }, + { + "epoch": 0.42, + "learning_rate": 1.58490496942176e-07, + "loss": 0.682, + "step": 11385 + }, + { + "epoch": 0.42, + "learning_rate": 1.5848683487750394e-07, + "loss": 0.6981, + "step": 11386 + }, + { + "epoch": 0.42, + "learning_rate": 1.5848317281283187e-07, + "loss": 0.695, + "step": 11387 + }, + { + "epoch": 0.42, + "learning_rate": 1.584795107481598e-07, + "loss": 0.6889, + "step": 11388 + }, + { + "epoch": 0.42, + "learning_rate": 1.5847584868348775e-07, + "loss": 0.7091, + "step": 11389 + }, + { + "epoch": 0.42, + "learning_rate": 1.5847218661881567e-07, + "loss": 0.7014, + "step": 11390 + }, + { + "epoch": 0.42, + "learning_rate": 1.584685245541436e-07, + "loss": 0.6997, + "step": 11391 + }, + { + "epoch": 0.42, + "learning_rate": 1.5846486248947155e-07, + "loss": 0.7013, + "step": 11392 + }, + { + "epoch": 0.42, + "learning_rate": 1.5846120042479948e-07, + "loss": 0.6826, + "step": 11393 + }, + { + "epoch": 0.42, + "learning_rate": 1.5845753836012743e-07, + "loss": 0.6993, + "step": 11394 + }, + { + "epoch": 0.42, + "learning_rate": 1.5845387629545536e-07, + "loss": 0.6902, + "step": 11395 + }, + { + "epoch": 0.42, + "learning_rate": 1.584502142307833e-07, + "loss": 0.6767, + "step": 11396 + }, + { + "epoch": 0.42, + "learning_rate": 1.5844655216611124e-07, + "loss": 0.6928, + "step": 11397 + }, + { + "epoch": 0.42, + "learning_rate": 1.584428901014392e-07, + "loss": 0.6988, + "step": 11398 + }, + { + "epoch": 0.42, + "learning_rate": 1.5843922803676712e-07, + "loss": 0.6906, + "step": 11399 + }, + { + "epoch": 0.42, + "learning_rate": 1.5843556597209507e-07, + "loss": 0.6911, + "step": 11400 + }, + { + "epoch": 0.42, + "learning_rate": 1.58431903907423e-07, + "loss": 0.68, + "step": 11401 + }, + { + "epoch": 0.42, + "learning_rate": 1.5842824184275095e-07, + "loss": 0.6903, + "step": 11402 + }, + { + "epoch": 0.42, + "learning_rate": 1.5842457977807888e-07, + "loss": 0.7012, + "step": 11403 + }, + { + "epoch": 0.42, + "learning_rate": 1.5842091771340683e-07, + "loss": 0.6951, + "step": 11404 + }, + { + "epoch": 0.42, + "learning_rate": 1.5841725564873476e-07, + "loss": 0.7028, + "step": 11405 + }, + { + "epoch": 0.42, + "learning_rate": 1.5841359358406269e-07, + "loss": 0.7144, + "step": 11406 + }, + { + "epoch": 0.42, + "learning_rate": 1.584099315193906e-07, + "loss": 0.6919, + "step": 11407 + }, + { + "epoch": 0.42, + "learning_rate": 1.5840626945471857e-07, + "loss": 0.681, + "step": 11408 + }, + { + "epoch": 0.42, + "learning_rate": 1.584026073900465e-07, + "loss": 0.6892, + "step": 11409 + }, + { + "epoch": 0.42, + "learning_rate": 1.5839894532537445e-07, + "loss": 0.696, + "step": 11410 + }, + { + "epoch": 0.42, + "learning_rate": 1.5839528326070237e-07, + "loss": 0.6992, + "step": 11411 + }, + { + "epoch": 0.42, + "learning_rate": 1.583916211960303e-07, + "loss": 0.6932, + "step": 11412 + }, + { + "epoch": 0.42, + "learning_rate": 1.5838795913135825e-07, + "loss": 0.6962, + "step": 11413 + }, + { + "epoch": 0.42, + "learning_rate": 1.5838429706668618e-07, + "loss": 0.6983, + "step": 11414 + }, + { + "epoch": 0.42, + "learning_rate": 1.5838063500201413e-07, + "loss": 0.6811, + "step": 11415 + }, + { + "epoch": 0.42, + "learning_rate": 1.5837697293734206e-07, + "loss": 0.6865, + "step": 11416 + }, + { + "epoch": 0.42, + "learning_rate": 1.5837331087267e-07, + "loss": 0.6954, + "step": 11417 + }, + { + "epoch": 0.42, + "learning_rate": 1.5836964880799794e-07, + "loss": 0.6857, + "step": 11418 + }, + { + "epoch": 0.42, + "learning_rate": 1.583659867433259e-07, + "loss": 0.6953, + "step": 11419 + }, + { + "epoch": 0.42, + "learning_rate": 1.5836232467865382e-07, + "loss": 0.688, + "step": 11420 + }, + { + "epoch": 0.42, + "learning_rate": 1.5835866261398177e-07, + "loss": 0.6831, + "step": 11421 + }, + { + "epoch": 0.42, + "learning_rate": 1.583550005493097e-07, + "loss": 0.6933, + "step": 11422 + }, + { + "epoch": 0.42, + "learning_rate": 1.5835133848463765e-07, + "loss": 0.6897, + "step": 11423 + }, + { + "epoch": 0.42, + "learning_rate": 1.5834767641996555e-07, + "loss": 0.692, + "step": 11424 + }, + { + "epoch": 0.42, + "learning_rate": 1.583440143552935e-07, + "loss": 0.6848, + "step": 11425 + }, + { + "epoch": 0.42, + "learning_rate": 1.5834035229062143e-07, + "loss": 0.6883, + "step": 11426 + }, + { + "epoch": 0.42, + "learning_rate": 1.5833669022594938e-07, + "loss": 0.6878, + "step": 11427 + }, + { + "epoch": 0.42, + "learning_rate": 1.583330281612773e-07, + "loss": 0.686, + "step": 11428 + }, + { + "epoch": 0.42, + "learning_rate": 1.5832936609660526e-07, + "loss": 0.7035, + "step": 11429 + }, + { + "epoch": 0.42, + "learning_rate": 1.583257040319332e-07, + "loss": 0.6923, + "step": 11430 + }, + { + "epoch": 0.42, + "learning_rate": 1.5832204196726114e-07, + "loss": 0.6936, + "step": 11431 + }, + { + "epoch": 0.42, + "learning_rate": 1.5831837990258907e-07, + "loss": 0.6844, + "step": 11432 + }, + { + "epoch": 0.42, + "learning_rate": 1.5831471783791702e-07, + "loss": 0.6874, + "step": 11433 + }, + { + "epoch": 0.42, + "learning_rate": 1.5831105577324495e-07, + "loss": 0.6942, + "step": 11434 + }, + { + "epoch": 0.42, + "learning_rate": 1.5830739370857288e-07, + "loss": 0.6869, + "step": 11435 + }, + { + "epoch": 0.42, + "learning_rate": 1.5830373164390083e-07, + "loss": 0.7104, + "step": 11436 + }, + { + "epoch": 0.42, + "learning_rate": 1.5830006957922876e-07, + "loss": 0.6942, + "step": 11437 + }, + { + "epoch": 0.42, + "learning_rate": 1.582964075145567e-07, + "loss": 0.6903, + "step": 11438 + }, + { + "epoch": 0.42, + "learning_rate": 1.5829274544988464e-07, + "loss": 0.6828, + "step": 11439 + }, + { + "epoch": 0.42, + "learning_rate": 1.582890833852126e-07, + "loss": 0.6788, + "step": 11440 + }, + { + "epoch": 0.42, + "learning_rate": 1.5828542132054052e-07, + "loss": 0.6908, + "step": 11441 + }, + { + "epoch": 0.42, + "learning_rate": 1.5828175925586844e-07, + "loss": 0.6976, + "step": 11442 + }, + { + "epoch": 0.42, + "learning_rate": 1.5827809719119637e-07, + "loss": 0.6987, + "step": 11443 + }, + { + "epoch": 0.42, + "learning_rate": 1.5827443512652432e-07, + "loss": 0.69, + "step": 11444 + }, + { + "epoch": 0.42, + "learning_rate": 1.5827077306185225e-07, + "loss": 0.6919, + "step": 11445 + }, + { + "epoch": 0.42, + "learning_rate": 1.582671109971802e-07, + "loss": 0.6825, + "step": 11446 + }, + { + "epoch": 0.42, + "learning_rate": 1.5826344893250813e-07, + "loss": 0.6876, + "step": 11447 + }, + { + "epoch": 0.42, + "learning_rate": 1.5825978686783608e-07, + "loss": 0.6864, + "step": 11448 + }, + { + "epoch": 0.42, + "learning_rate": 1.58256124803164e-07, + "loss": 0.6999, + "step": 11449 + }, + { + "epoch": 0.42, + "learning_rate": 1.5825246273849196e-07, + "loss": 0.6815, + "step": 11450 + }, + { + "epoch": 0.42, + "learning_rate": 1.582488006738199e-07, + "loss": 0.686, + "step": 11451 + }, + { + "epoch": 0.42, + "learning_rate": 1.5824513860914784e-07, + "loss": 0.6921, + "step": 11452 + }, + { + "epoch": 0.42, + "learning_rate": 1.5824147654447577e-07, + "loss": 0.6851, + "step": 11453 + }, + { + "epoch": 0.42, + "learning_rate": 1.5823781447980372e-07, + "loss": 0.6884, + "step": 11454 + }, + { + "epoch": 0.42, + "learning_rate": 1.5823415241513165e-07, + "loss": 0.6803, + "step": 11455 + }, + { + "epoch": 0.42, + "learning_rate": 1.5823049035045958e-07, + "loss": 0.6921, + "step": 11456 + }, + { + "epoch": 0.42, + "learning_rate": 1.5822682828578753e-07, + "loss": 0.7008, + "step": 11457 + }, + { + "epoch": 0.42, + "learning_rate": 1.5822316622111546e-07, + "loss": 0.6964, + "step": 11458 + }, + { + "epoch": 0.42, + "learning_rate": 1.582195041564434e-07, + "loss": 0.6985, + "step": 11459 + }, + { + "epoch": 0.42, + "learning_rate": 1.5821584209177134e-07, + "loss": 0.7047, + "step": 11460 + }, + { + "epoch": 0.42, + "learning_rate": 1.5821218002709926e-07, + "loss": 0.688, + "step": 11461 + }, + { + "epoch": 0.42, + "learning_rate": 1.582085179624272e-07, + "loss": 0.6924, + "step": 11462 + }, + { + "epoch": 0.42, + "learning_rate": 1.5820485589775514e-07, + "loss": 0.6787, + "step": 11463 + }, + { + "epoch": 0.42, + "learning_rate": 1.5820119383308307e-07, + "loss": 0.6971, + "step": 11464 + }, + { + "epoch": 0.42, + "learning_rate": 1.5819753176841102e-07, + "loss": 0.6898, + "step": 11465 + }, + { + "epoch": 0.42, + "learning_rate": 1.5819386970373895e-07, + "loss": 0.7025, + "step": 11466 + }, + { + "epoch": 0.42, + "learning_rate": 1.581902076390669e-07, + "loss": 0.6865, + "step": 11467 + }, + { + "epoch": 0.42, + "learning_rate": 1.5818654557439483e-07, + "loss": 0.7061, + "step": 11468 + }, + { + "epoch": 0.42, + "learning_rate": 1.5818288350972278e-07, + "loss": 0.6955, + "step": 11469 + }, + { + "epoch": 0.42, + "learning_rate": 1.581792214450507e-07, + "loss": 0.6999, + "step": 11470 + }, + { + "epoch": 0.42, + "learning_rate": 1.5817555938037866e-07, + "loss": 0.6925, + "step": 11471 + }, + { + "epoch": 0.42, + "learning_rate": 1.581718973157066e-07, + "loss": 0.6892, + "step": 11472 + }, + { + "epoch": 0.42, + "learning_rate": 1.5816823525103454e-07, + "loss": 0.7027, + "step": 11473 + }, + { + "epoch": 0.42, + "learning_rate": 1.5816457318636247e-07, + "loss": 0.6869, + "step": 11474 + }, + { + "epoch": 0.42, + "learning_rate": 1.5816091112169042e-07, + "loss": 0.6952, + "step": 11475 + }, + { + "epoch": 0.42, + "learning_rate": 1.5815724905701835e-07, + "loss": 0.6937, + "step": 11476 + }, + { + "epoch": 0.42, + "learning_rate": 1.581535869923463e-07, + "loss": 0.7021, + "step": 11477 + }, + { + "epoch": 0.42, + "learning_rate": 1.581499249276742e-07, + "loss": 0.6861, + "step": 11478 + }, + { + "epoch": 0.42, + "learning_rate": 1.5814626286300215e-07, + "loss": 0.7052, + "step": 11479 + }, + { + "epoch": 0.42, + "learning_rate": 1.5814260079833008e-07, + "loss": 0.6984, + "step": 11480 + }, + { + "epoch": 0.42, + "learning_rate": 1.5813893873365803e-07, + "loss": 0.6856, + "step": 11481 + }, + { + "epoch": 0.42, + "learning_rate": 1.5813527666898596e-07, + "loss": 0.6852, + "step": 11482 + }, + { + "epoch": 0.42, + "learning_rate": 1.581316146043139e-07, + "loss": 0.7117, + "step": 11483 + }, + { + "epoch": 0.42, + "learning_rate": 1.5812795253964184e-07, + "loss": 0.6931, + "step": 11484 + }, + { + "epoch": 0.42, + "learning_rate": 1.5812429047496977e-07, + "loss": 0.6911, + "step": 11485 + }, + { + "epoch": 0.42, + "learning_rate": 1.5812062841029772e-07, + "loss": 0.6952, + "step": 11486 + }, + { + "epoch": 0.42, + "learning_rate": 1.5811696634562565e-07, + "loss": 0.7056, + "step": 11487 + }, + { + "epoch": 0.42, + "learning_rate": 1.581133042809536e-07, + "loss": 0.7021, + "step": 11488 + }, + { + "epoch": 0.42, + "learning_rate": 1.5810964221628153e-07, + "loss": 0.7114, + "step": 11489 + }, + { + "epoch": 0.42, + "learning_rate": 1.5810598015160948e-07, + "loss": 0.6995, + "step": 11490 + }, + { + "epoch": 0.42, + "learning_rate": 1.581023180869374e-07, + "loss": 0.684, + "step": 11491 + }, + { + "epoch": 0.42, + "learning_rate": 1.5809865602226536e-07, + "loss": 0.6872, + "step": 11492 + }, + { + "epoch": 0.42, + "learning_rate": 1.580949939575933e-07, + "loss": 0.6907, + "step": 11493 + }, + { + "epoch": 0.42, + "learning_rate": 1.5809133189292124e-07, + "loss": 0.7041, + "step": 11494 + }, + { + "epoch": 0.42, + "learning_rate": 1.5808766982824917e-07, + "loss": 0.6691, + "step": 11495 + }, + { + "epoch": 0.42, + "learning_rate": 1.580840077635771e-07, + "loss": 0.6832, + "step": 11496 + }, + { + "epoch": 0.42, + "learning_rate": 1.5808034569890502e-07, + "loss": 0.6989, + "step": 11497 + }, + { + "epoch": 0.42, + "learning_rate": 1.5807668363423297e-07, + "loss": 0.691, + "step": 11498 + }, + { + "epoch": 0.42, + "learning_rate": 1.580730215695609e-07, + "loss": 0.6924, + "step": 11499 + }, + { + "epoch": 0.42, + "learning_rate": 1.5806935950488885e-07, + "loss": 0.7214, + "step": 11500 + }, + { + "epoch": 0.42, + "learning_rate": 1.5806569744021678e-07, + "loss": 0.6938, + "step": 11501 + }, + { + "epoch": 0.42, + "learning_rate": 1.5806203537554473e-07, + "loss": 0.681, + "step": 11502 + }, + { + "epoch": 0.42, + "learning_rate": 1.5805837331087266e-07, + "loss": 0.6821, + "step": 11503 + }, + { + "epoch": 0.42, + "learning_rate": 1.5805471124620061e-07, + "loss": 0.6789, + "step": 11504 + }, + { + "epoch": 0.42, + "learning_rate": 1.5805104918152854e-07, + "loss": 0.6904, + "step": 11505 + }, + { + "epoch": 0.42, + "learning_rate": 1.5804738711685647e-07, + "loss": 0.7082, + "step": 11506 + }, + { + "epoch": 0.42, + "learning_rate": 1.5804372505218442e-07, + "loss": 0.6953, + "step": 11507 + }, + { + "epoch": 0.42, + "learning_rate": 1.5804006298751235e-07, + "loss": 0.6863, + "step": 11508 + }, + { + "epoch": 0.42, + "learning_rate": 1.580364009228403e-07, + "loss": 0.701, + "step": 11509 + }, + { + "epoch": 0.42, + "learning_rate": 1.5803273885816823e-07, + "loss": 0.7039, + "step": 11510 + }, + { + "epoch": 0.42, + "learning_rate": 1.5802907679349618e-07, + "loss": 0.7096, + "step": 11511 + }, + { + "epoch": 0.42, + "learning_rate": 1.580254147288241e-07, + "loss": 0.704, + "step": 11512 + }, + { + "epoch": 0.42, + "learning_rate": 1.5802175266415206e-07, + "loss": 0.6981, + "step": 11513 + }, + { + "epoch": 0.42, + "learning_rate": 1.5801809059947996e-07, + "loss": 0.6923, + "step": 11514 + }, + { + "epoch": 0.42, + "learning_rate": 1.580144285348079e-07, + "loss": 0.6857, + "step": 11515 + }, + { + "epoch": 0.42, + "learning_rate": 1.5801076647013584e-07, + "loss": 0.6871, + "step": 11516 + }, + { + "epoch": 0.42, + "learning_rate": 1.580071044054638e-07, + "loss": 0.6954, + "step": 11517 + }, + { + "epoch": 0.42, + "learning_rate": 1.5800344234079172e-07, + "loss": 0.6879, + "step": 11518 + }, + { + "epoch": 0.42, + "learning_rate": 1.5799978027611967e-07, + "loss": 0.697, + "step": 11519 + }, + { + "epoch": 0.42, + "learning_rate": 1.579961182114476e-07, + "loss": 0.698, + "step": 11520 + }, + { + "epoch": 0.42, + "learning_rate": 1.5799245614677555e-07, + "loss": 0.7157, + "step": 11521 + }, + { + "epoch": 0.42, + "learning_rate": 1.5798879408210348e-07, + "loss": 0.6855, + "step": 11522 + }, + { + "epoch": 0.42, + "learning_rate": 1.5798513201743143e-07, + "loss": 0.6796, + "step": 11523 + }, + { + "epoch": 0.42, + "learning_rate": 1.5798146995275936e-07, + "loss": 0.6929, + "step": 11524 + }, + { + "epoch": 0.42, + "learning_rate": 1.579778078880873e-07, + "loss": 0.6874, + "step": 11525 + }, + { + "epoch": 0.42, + "learning_rate": 1.5797414582341524e-07, + "loss": 0.6956, + "step": 11526 + }, + { + "epoch": 0.42, + "learning_rate": 1.5797048375874317e-07, + "loss": 0.7019, + "step": 11527 + }, + { + "epoch": 0.42, + "learning_rate": 1.5796682169407112e-07, + "loss": 0.6935, + "step": 11528 + }, + { + "epoch": 0.42, + "learning_rate": 1.5796315962939905e-07, + "loss": 0.6982, + "step": 11529 + }, + { + "epoch": 0.42, + "learning_rate": 1.57959497564727e-07, + "loss": 0.6849, + "step": 11530 + }, + { + "epoch": 0.42, + "learning_rate": 1.5795583550005493e-07, + "loss": 0.6812, + "step": 11531 + }, + { + "epoch": 0.42, + "learning_rate": 1.5795217343538285e-07, + "loss": 0.6897, + "step": 11532 + }, + { + "epoch": 0.42, + "learning_rate": 1.5794851137071078e-07, + "loss": 0.689, + "step": 11533 + }, + { + "epoch": 0.42, + "learning_rate": 1.5794484930603873e-07, + "loss": 0.6924, + "step": 11534 + }, + { + "epoch": 0.42, + "learning_rate": 1.5794118724136666e-07, + "loss": 0.6974, + "step": 11535 + }, + { + "epoch": 0.42, + "learning_rate": 1.579375251766946e-07, + "loss": 0.7077, + "step": 11536 + }, + { + "epoch": 0.42, + "learning_rate": 1.5793386311202254e-07, + "loss": 0.6986, + "step": 11537 + }, + { + "epoch": 0.42, + "learning_rate": 1.579302010473505e-07, + "loss": 0.6833, + "step": 11538 + }, + { + "epoch": 0.42, + "learning_rate": 1.5792653898267842e-07, + "loss": 0.6914, + "step": 11539 + }, + { + "epoch": 0.42, + "learning_rate": 1.5792287691800637e-07, + "loss": 0.6876, + "step": 11540 + }, + { + "epoch": 0.42, + "learning_rate": 1.579192148533343e-07, + "loss": 0.6935, + "step": 11541 + }, + { + "epoch": 0.42, + "learning_rate": 1.5791555278866225e-07, + "loss": 0.6945, + "step": 11542 + }, + { + "epoch": 0.42, + "learning_rate": 1.5791189072399018e-07, + "loss": 0.7188, + "step": 11543 + }, + { + "epoch": 0.42, + "learning_rate": 1.5790822865931813e-07, + "loss": 0.6962, + "step": 11544 + }, + { + "epoch": 0.42, + "learning_rate": 1.5790456659464606e-07, + "loss": 0.683, + "step": 11545 + }, + { + "epoch": 0.42, + "learning_rate": 1.57900904529974e-07, + "loss": 0.7077, + "step": 11546 + }, + { + "epoch": 0.42, + "learning_rate": 1.5789724246530194e-07, + "loss": 0.6809, + "step": 11547 + }, + { + "epoch": 0.42, + "learning_rate": 1.578935804006299e-07, + "loss": 0.6804, + "step": 11548 + }, + { + "epoch": 0.42, + "learning_rate": 1.5788991833595782e-07, + "loss": 0.6951, + "step": 11549 + }, + { + "epoch": 0.42, + "learning_rate": 1.5788625627128574e-07, + "loss": 0.694, + "step": 11550 + }, + { + "epoch": 0.42, + "learning_rate": 1.5788259420661367e-07, + "loss": 0.6828, + "step": 11551 + }, + { + "epoch": 0.42, + "learning_rate": 1.5787893214194162e-07, + "loss": 0.6748, + "step": 11552 + }, + { + "epoch": 0.42, + "learning_rate": 1.5787527007726955e-07, + "loss": 0.694, + "step": 11553 + }, + { + "epoch": 0.42, + "learning_rate": 1.5787160801259748e-07, + "loss": 0.691, + "step": 11554 + }, + { + "epoch": 0.42, + "learning_rate": 1.5786794594792543e-07, + "loss": 0.691, + "step": 11555 + }, + { + "epoch": 0.42, + "learning_rate": 1.5786428388325336e-07, + "loss": 0.7117, + "step": 11556 + }, + { + "epoch": 0.42, + "learning_rate": 1.578606218185813e-07, + "loss": 0.6821, + "step": 11557 + }, + { + "epoch": 0.42, + "learning_rate": 1.5785695975390924e-07, + "loss": 0.6835, + "step": 11558 + }, + { + "epoch": 0.42, + "learning_rate": 1.578532976892372e-07, + "loss": 0.7021, + "step": 11559 + }, + { + "epoch": 0.42, + "learning_rate": 1.5784963562456512e-07, + "loss": 0.6855, + "step": 11560 + }, + { + "epoch": 0.42, + "learning_rate": 1.5784597355989307e-07, + "loss": 0.6918, + "step": 11561 + }, + { + "epoch": 0.42, + "learning_rate": 1.57842311495221e-07, + "loss": 0.6869, + "step": 11562 + }, + { + "epoch": 0.42, + "learning_rate": 1.5783864943054895e-07, + "loss": 0.6829, + "step": 11563 + }, + { + "epoch": 0.42, + "learning_rate": 1.5783498736587688e-07, + "loss": 0.6909, + "step": 11564 + }, + { + "epoch": 0.42, + "learning_rate": 1.5783132530120483e-07, + "loss": 0.6835, + "step": 11565 + }, + { + "epoch": 0.42, + "learning_rate": 1.5782766323653276e-07, + "loss": 0.6848, + "step": 11566 + }, + { + "epoch": 0.42, + "learning_rate": 1.578240011718607e-07, + "loss": 0.6913, + "step": 11567 + }, + { + "epoch": 0.42, + "learning_rate": 1.578203391071886e-07, + "loss": 0.6879, + "step": 11568 + }, + { + "epoch": 0.42, + "learning_rate": 1.5781667704251656e-07, + "loss": 0.6991, + "step": 11569 + }, + { + "epoch": 0.42, + "learning_rate": 1.578130149778445e-07, + "loss": 0.6928, + "step": 11570 + }, + { + "epoch": 0.42, + "learning_rate": 1.5780935291317244e-07, + "loss": 0.703, + "step": 11571 + }, + { + "epoch": 0.42, + "learning_rate": 1.5780569084850037e-07, + "loss": 0.6876, + "step": 11572 + }, + { + "epoch": 0.42, + "learning_rate": 1.5780202878382832e-07, + "loss": 0.6934, + "step": 11573 + }, + { + "epoch": 0.42, + "learning_rate": 1.5779836671915625e-07, + "loss": 0.7003, + "step": 11574 + }, + { + "epoch": 0.42, + "learning_rate": 1.577947046544842e-07, + "loss": 0.7101, + "step": 11575 + }, + { + "epoch": 0.42, + "learning_rate": 1.5779104258981213e-07, + "loss": 0.7057, + "step": 11576 + }, + { + "epoch": 0.42, + "learning_rate": 1.5778738052514006e-07, + "loss": 0.6868, + "step": 11577 + }, + { + "epoch": 0.42, + "learning_rate": 1.57783718460468e-07, + "loss": 0.6803, + "step": 11578 + }, + { + "epoch": 0.42, + "learning_rate": 1.5778005639579594e-07, + "loss": 0.6733, + "step": 11579 + }, + { + "epoch": 0.42, + "learning_rate": 1.577763943311239e-07, + "loss": 0.6952, + "step": 11580 + }, + { + "epoch": 0.42, + "learning_rate": 1.5777273226645182e-07, + "loss": 0.7097, + "step": 11581 + }, + { + "epoch": 0.42, + "learning_rate": 1.5776907020177977e-07, + "loss": 0.6965, + "step": 11582 + }, + { + "epoch": 0.42, + "learning_rate": 1.577654081371077e-07, + "loss": 0.6863, + "step": 11583 + }, + { + "epoch": 0.42, + "learning_rate": 1.5776174607243565e-07, + "loss": 0.6833, + "step": 11584 + }, + { + "epoch": 0.42, + "learning_rate": 1.5775808400776358e-07, + "loss": 0.6889, + "step": 11585 + }, + { + "epoch": 0.42, + "learning_rate": 1.577544219430915e-07, + "loss": 0.6948, + "step": 11586 + }, + { + "epoch": 0.42, + "learning_rate": 1.5775075987841943e-07, + "loss": 0.7014, + "step": 11587 + }, + { + "epoch": 0.42, + "learning_rate": 1.5774709781374738e-07, + "loss": 0.6998, + "step": 11588 + }, + { + "epoch": 0.42, + "learning_rate": 1.577434357490753e-07, + "loss": 0.6861, + "step": 11589 + }, + { + "epoch": 0.42, + "learning_rate": 1.5773977368440326e-07, + "loss": 0.6898, + "step": 11590 + }, + { + "epoch": 0.42, + "learning_rate": 1.577361116197312e-07, + "loss": 0.6965, + "step": 11591 + }, + { + "epoch": 0.42, + "learning_rate": 1.5773244955505914e-07, + "loss": 0.6866, + "step": 11592 + }, + { + "epoch": 0.42, + "learning_rate": 1.5772878749038707e-07, + "loss": 0.7211, + "step": 11593 + }, + { + "epoch": 0.42, + "learning_rate": 1.5772512542571502e-07, + "loss": 0.6734, + "step": 11594 + }, + { + "epoch": 0.42, + "learning_rate": 1.5772146336104295e-07, + "loss": 0.6912, + "step": 11595 + }, + { + "epoch": 0.42, + "learning_rate": 1.577178012963709e-07, + "loss": 0.6857, + "step": 11596 + }, + { + "epoch": 0.42, + "learning_rate": 1.5771413923169883e-07, + "loss": 0.6869, + "step": 11597 + }, + { + "epoch": 0.42, + "learning_rate": 1.5771047716702675e-07, + "loss": 0.7092, + "step": 11598 + }, + { + "epoch": 0.42, + "learning_rate": 1.577068151023547e-07, + "loss": 0.6763, + "step": 11599 + }, + { + "epoch": 0.42, + "learning_rate": 1.5770315303768263e-07, + "loss": 0.6906, + "step": 11600 + }, + { + "epoch": 0.42, + "learning_rate": 1.576994909730106e-07, + "loss": 0.6898, + "step": 11601 + }, + { + "epoch": 0.42, + "learning_rate": 1.5769582890833851e-07, + "loss": 0.715, + "step": 11602 + }, + { + "epoch": 0.42, + "learning_rate": 1.5769216684366644e-07, + "loss": 0.6935, + "step": 11603 + }, + { + "epoch": 0.42, + "learning_rate": 1.5768850477899437e-07, + "loss": 0.6845, + "step": 11604 + }, + { + "epoch": 0.42, + "learning_rate": 1.5768484271432232e-07, + "loss": 0.6858, + "step": 11605 + }, + { + "epoch": 0.42, + "learning_rate": 1.5768118064965025e-07, + "loss": 0.6653, + "step": 11606 + }, + { + "epoch": 0.42, + "learning_rate": 1.576775185849782e-07, + "loss": 0.6773, + "step": 11607 + }, + { + "epoch": 0.42, + "learning_rate": 1.5767385652030613e-07, + "loss": 0.691, + "step": 11608 + }, + { + "epoch": 0.42, + "learning_rate": 1.5767019445563408e-07, + "loss": 0.6954, + "step": 11609 + }, + { + "epoch": 0.42, + "learning_rate": 1.57666532390962e-07, + "loss": 0.6867, + "step": 11610 + }, + { + "epoch": 0.42, + "learning_rate": 1.5766287032628996e-07, + "loss": 0.6821, + "step": 11611 + }, + { + "epoch": 0.42, + "learning_rate": 1.576592082616179e-07, + "loss": 0.7078, + "step": 11612 + }, + { + "epoch": 0.42, + "learning_rate": 1.5765554619694584e-07, + "loss": 0.6771, + "step": 11613 + }, + { + "epoch": 0.42, + "learning_rate": 1.5765188413227377e-07, + "loss": 0.6985, + "step": 11614 + }, + { + "epoch": 0.42, + "learning_rate": 1.5764822206760172e-07, + "loss": 0.7032, + "step": 11615 + }, + { + "epoch": 0.42, + "learning_rate": 1.5764456000292965e-07, + "loss": 0.6871, + "step": 11616 + }, + { + "epoch": 0.43, + "learning_rate": 1.576408979382576e-07, + "loss": 0.7122, + "step": 11617 + }, + { + "epoch": 0.43, + "learning_rate": 1.5763723587358553e-07, + "loss": 0.6935, + "step": 11618 + }, + { + "epoch": 0.43, + "learning_rate": 1.5763357380891345e-07, + "loss": 0.698, + "step": 11619 + }, + { + "epoch": 0.43, + "learning_rate": 1.576299117442414e-07, + "loss": 0.6785, + "step": 11620 + }, + { + "epoch": 0.43, + "learning_rate": 1.5762624967956933e-07, + "loss": 0.7028, + "step": 11621 + }, + { + "epoch": 0.43, + "learning_rate": 1.5762258761489726e-07, + "loss": 0.6865, + "step": 11622 + }, + { + "epoch": 0.43, + "learning_rate": 1.5761892555022521e-07, + "loss": 0.7066, + "step": 11623 + }, + { + "epoch": 0.43, + "learning_rate": 1.5761526348555314e-07, + "loss": 0.6888, + "step": 11624 + }, + { + "epoch": 0.43, + "learning_rate": 1.5761160142088107e-07, + "loss": 0.6796, + "step": 11625 + }, + { + "epoch": 0.43, + "learning_rate": 1.5760793935620902e-07, + "loss": 0.6857, + "step": 11626 + }, + { + "epoch": 0.43, + "learning_rate": 1.5760427729153695e-07, + "loss": 0.6974, + "step": 11627 + }, + { + "epoch": 0.43, + "learning_rate": 1.576006152268649e-07, + "loss": 0.6911, + "step": 11628 + }, + { + "epoch": 0.43, + "learning_rate": 1.5759695316219283e-07, + "loss": 0.7012, + "step": 11629 + }, + { + "epoch": 0.43, + "learning_rate": 1.5759329109752078e-07, + "loss": 0.6926, + "step": 11630 + }, + { + "epoch": 0.43, + "learning_rate": 1.575896290328487e-07, + "loss": 0.6763, + "step": 11631 + }, + { + "epoch": 0.43, + "learning_rate": 1.5758596696817666e-07, + "loss": 0.6922, + "step": 11632 + }, + { + "epoch": 0.43, + "learning_rate": 1.5758230490350459e-07, + "loss": 0.7004, + "step": 11633 + }, + { + "epoch": 0.43, + "learning_rate": 1.5757864283883254e-07, + "loss": 0.6893, + "step": 11634 + }, + { + "epoch": 0.43, + "learning_rate": 1.5757498077416047e-07, + "loss": 0.6927, + "step": 11635 + }, + { + "epoch": 0.43, + "learning_rate": 1.5757131870948842e-07, + "loss": 0.6756, + "step": 11636 + }, + { + "epoch": 0.43, + "learning_rate": 1.5756765664481635e-07, + "loss": 0.7051, + "step": 11637 + }, + { + "epoch": 0.43, + "learning_rate": 1.575639945801443e-07, + "loss": 0.679, + "step": 11638 + }, + { + "epoch": 0.43, + "learning_rate": 1.575603325154722e-07, + "loss": 0.6992, + "step": 11639 + }, + { + "epoch": 0.43, + "learning_rate": 1.5755667045080015e-07, + "loss": 0.6851, + "step": 11640 + }, + { + "epoch": 0.43, + "learning_rate": 1.5755300838612808e-07, + "loss": 0.6806, + "step": 11641 + }, + { + "epoch": 0.43, + "learning_rate": 1.5754934632145603e-07, + "loss": 0.689, + "step": 11642 + }, + { + "epoch": 0.43, + "learning_rate": 1.5754568425678396e-07, + "loss": 0.695, + "step": 11643 + }, + { + "epoch": 0.43, + "learning_rate": 1.575420221921119e-07, + "loss": 0.7028, + "step": 11644 + }, + { + "epoch": 0.43, + "learning_rate": 1.5753836012743984e-07, + "loss": 0.6877, + "step": 11645 + }, + { + "epoch": 0.43, + "learning_rate": 1.575346980627678e-07, + "loss": 0.6915, + "step": 11646 + }, + { + "epoch": 0.43, + "learning_rate": 1.5753103599809572e-07, + "loss": 0.7056, + "step": 11647 + }, + { + "epoch": 0.43, + "learning_rate": 1.5752737393342365e-07, + "loss": 0.7076, + "step": 11648 + }, + { + "epoch": 0.43, + "learning_rate": 1.575237118687516e-07, + "loss": 0.6868, + "step": 11649 + }, + { + "epoch": 0.43, + "learning_rate": 1.5752004980407953e-07, + "loss": 0.6902, + "step": 11650 + }, + { + "epoch": 0.43, + "learning_rate": 1.5751638773940748e-07, + "loss": 0.7006, + "step": 11651 + }, + { + "epoch": 0.43, + "learning_rate": 1.575127256747354e-07, + "loss": 0.7067, + "step": 11652 + }, + { + "epoch": 0.43, + "learning_rate": 1.5750906361006336e-07, + "loss": 0.6985, + "step": 11653 + }, + { + "epoch": 0.43, + "learning_rate": 1.5750540154539129e-07, + "loss": 0.7019, + "step": 11654 + }, + { + "epoch": 0.43, + "learning_rate": 1.5750173948071924e-07, + "loss": 0.6891, + "step": 11655 + }, + { + "epoch": 0.43, + "learning_rate": 1.5749807741604716e-07, + "loss": 0.6967, + "step": 11656 + }, + { + "epoch": 0.43, + "learning_rate": 1.574944153513751e-07, + "loss": 0.704, + "step": 11657 + }, + { + "epoch": 0.43, + "learning_rate": 1.5749075328670302e-07, + "loss": 0.6885, + "step": 11658 + }, + { + "epoch": 0.43, + "learning_rate": 1.5748709122203097e-07, + "loss": 0.705, + "step": 11659 + }, + { + "epoch": 0.43, + "learning_rate": 1.574834291573589e-07, + "loss": 0.6991, + "step": 11660 + }, + { + "epoch": 0.43, + "learning_rate": 1.5747976709268685e-07, + "loss": 0.6875, + "step": 11661 + }, + { + "epoch": 0.43, + "learning_rate": 1.5747610502801478e-07, + "loss": 0.6893, + "step": 11662 + }, + { + "epoch": 0.43, + "learning_rate": 1.5747244296334273e-07, + "loss": 0.6949, + "step": 11663 + }, + { + "epoch": 0.43, + "learning_rate": 1.5746878089867066e-07, + "loss": 0.7106, + "step": 11664 + }, + { + "epoch": 0.43, + "learning_rate": 1.574651188339986e-07, + "loss": 0.7198, + "step": 11665 + }, + { + "epoch": 0.43, + "learning_rate": 1.5746145676932654e-07, + "loss": 0.6773, + "step": 11666 + }, + { + "epoch": 0.43, + "learning_rate": 1.574577947046545e-07, + "loss": 0.6922, + "step": 11667 + }, + { + "epoch": 0.43, + "learning_rate": 1.5745413263998242e-07, + "loss": 0.6956, + "step": 11668 + }, + { + "epoch": 0.43, + "learning_rate": 1.5745047057531034e-07, + "loss": 0.6919, + "step": 11669 + }, + { + "epoch": 0.43, + "learning_rate": 1.574468085106383e-07, + "loss": 0.6963, + "step": 11670 + }, + { + "epoch": 0.43, + "learning_rate": 1.5744314644596622e-07, + "loss": 0.697, + "step": 11671 + }, + { + "epoch": 0.43, + "learning_rate": 1.5743948438129418e-07, + "loss": 0.6956, + "step": 11672 + }, + { + "epoch": 0.43, + "learning_rate": 1.574358223166221e-07, + "loss": 0.685, + "step": 11673 + }, + { + "epoch": 0.43, + "learning_rate": 1.5743216025195006e-07, + "loss": 0.6943, + "step": 11674 + }, + { + "epoch": 0.43, + "learning_rate": 1.5742849818727796e-07, + "loss": 0.6858, + "step": 11675 + }, + { + "epoch": 0.43, + "learning_rate": 1.574248361226059e-07, + "loss": 0.6987, + "step": 11676 + }, + { + "epoch": 0.43, + "learning_rate": 1.5742117405793384e-07, + "loss": 0.6742, + "step": 11677 + }, + { + "epoch": 0.43, + "learning_rate": 1.574175119932618e-07, + "loss": 0.6944, + "step": 11678 + }, + { + "epoch": 0.43, + "learning_rate": 1.5741384992858972e-07, + "loss": 0.6919, + "step": 11679 + }, + { + "epoch": 0.43, + "learning_rate": 1.5741018786391767e-07, + "loss": 0.6948, + "step": 11680 + }, + { + "epoch": 0.43, + "learning_rate": 1.574065257992456e-07, + "loss": 0.717, + "step": 11681 + }, + { + "epoch": 0.43, + "learning_rate": 1.5740286373457355e-07, + "loss": 0.6881, + "step": 11682 + }, + { + "epoch": 0.43, + "learning_rate": 1.5739920166990148e-07, + "loss": 0.6793, + "step": 11683 + }, + { + "epoch": 0.43, + "learning_rate": 1.5739553960522943e-07, + "loss": 0.7024, + "step": 11684 + }, + { + "epoch": 0.43, + "learning_rate": 1.5739187754055736e-07, + "loss": 0.7, + "step": 11685 + }, + { + "epoch": 0.43, + "learning_rate": 1.573882154758853e-07, + "loss": 0.7258, + "step": 11686 + }, + { + "epoch": 0.43, + "learning_rate": 1.5738455341121324e-07, + "loss": 0.7238, + "step": 11687 + }, + { + "epoch": 0.43, + "learning_rate": 1.573808913465412e-07, + "loss": 0.7004, + "step": 11688 + }, + { + "epoch": 0.43, + "learning_rate": 1.5737722928186912e-07, + "loss": 0.7, + "step": 11689 + }, + { + "epoch": 0.43, + "learning_rate": 1.5737356721719704e-07, + "loss": 0.6979, + "step": 11690 + }, + { + "epoch": 0.43, + "learning_rate": 1.57369905152525e-07, + "loss": 0.6989, + "step": 11691 + }, + { + "epoch": 0.43, + "learning_rate": 1.5736624308785292e-07, + "loss": 0.6675, + "step": 11692 + }, + { + "epoch": 0.43, + "learning_rate": 1.5736258102318085e-07, + "loss": 0.6848, + "step": 11693 + }, + { + "epoch": 0.43, + "learning_rate": 1.573589189585088e-07, + "loss": 0.6862, + "step": 11694 + }, + { + "epoch": 0.43, + "learning_rate": 1.5735525689383673e-07, + "loss": 0.7001, + "step": 11695 + }, + { + "epoch": 0.43, + "learning_rate": 1.5735159482916466e-07, + "loss": 0.6869, + "step": 11696 + }, + { + "epoch": 0.43, + "learning_rate": 1.573479327644926e-07, + "loss": 0.705, + "step": 11697 + }, + { + "epoch": 0.43, + "learning_rate": 1.5734427069982054e-07, + "loss": 0.6954, + "step": 11698 + }, + { + "epoch": 0.43, + "learning_rate": 1.573406086351485e-07, + "loss": 0.6959, + "step": 11699 + }, + { + "epoch": 0.43, + "learning_rate": 1.5733694657047642e-07, + "loss": 0.6947, + "step": 11700 + }, + { + "epoch": 0.43, + "learning_rate": 1.5733328450580437e-07, + "loss": 0.6916, + "step": 11701 + }, + { + "epoch": 0.43, + "learning_rate": 1.573296224411323e-07, + "loss": 0.706, + "step": 11702 + }, + { + "epoch": 0.43, + "learning_rate": 1.5732596037646025e-07, + "loss": 0.7116, + "step": 11703 + }, + { + "epoch": 0.43, + "learning_rate": 1.5732229831178818e-07, + "loss": 0.6916, + "step": 11704 + }, + { + "epoch": 0.43, + "learning_rate": 1.5731863624711613e-07, + "loss": 0.6957, + "step": 11705 + }, + { + "epoch": 0.43, + "learning_rate": 1.5731497418244406e-07, + "loss": 0.6774, + "step": 11706 + }, + { + "epoch": 0.43, + "learning_rate": 1.57311312117772e-07, + "loss": 0.6758, + "step": 11707 + }, + { + "epoch": 0.43, + "learning_rate": 1.5730765005309994e-07, + "loss": 0.7019, + "step": 11708 + }, + { + "epoch": 0.43, + "learning_rate": 1.573039879884279e-07, + "loss": 0.6723, + "step": 11709 + }, + { + "epoch": 0.43, + "learning_rate": 1.5730032592375582e-07, + "loss": 0.709, + "step": 11710 + }, + { + "epoch": 0.43, + "learning_rate": 1.5729666385908374e-07, + "loss": 0.7021, + "step": 11711 + }, + { + "epoch": 0.43, + "learning_rate": 1.5729300179441167e-07, + "loss": 0.6953, + "step": 11712 + }, + { + "epoch": 0.43, + "learning_rate": 1.5728933972973962e-07, + "loss": 0.6846, + "step": 11713 + }, + { + "epoch": 0.43, + "learning_rate": 1.5728567766506755e-07, + "loss": 0.7039, + "step": 11714 + }, + { + "epoch": 0.43, + "learning_rate": 1.572820156003955e-07, + "loss": 0.6878, + "step": 11715 + }, + { + "epoch": 0.43, + "learning_rate": 1.5727835353572343e-07, + "loss": 0.7228, + "step": 11716 + }, + { + "epoch": 0.43, + "learning_rate": 1.5727469147105138e-07, + "loss": 0.6915, + "step": 11717 + }, + { + "epoch": 0.43, + "learning_rate": 1.572710294063793e-07, + "loss": 0.6881, + "step": 11718 + }, + { + "epoch": 0.43, + "learning_rate": 1.5726736734170723e-07, + "loss": 0.7045, + "step": 11719 + }, + { + "epoch": 0.43, + "learning_rate": 1.572637052770352e-07, + "loss": 0.6966, + "step": 11720 + }, + { + "epoch": 0.43, + "learning_rate": 1.5726004321236311e-07, + "loss": 0.7109, + "step": 11721 + }, + { + "epoch": 0.43, + "learning_rate": 1.5725638114769107e-07, + "loss": 0.7073, + "step": 11722 + }, + { + "epoch": 0.43, + "learning_rate": 1.57252719083019e-07, + "loss": 0.6976, + "step": 11723 + }, + { + "epoch": 0.43, + "learning_rate": 1.5724905701834695e-07, + "loss": 0.6988, + "step": 11724 + }, + { + "epoch": 0.43, + "learning_rate": 1.5724539495367487e-07, + "loss": 0.6983, + "step": 11725 + }, + { + "epoch": 0.43, + "learning_rate": 1.5724173288900283e-07, + "loss": 0.6967, + "step": 11726 + }, + { + "epoch": 0.43, + "learning_rate": 1.5723807082433075e-07, + "loss": 0.6743, + "step": 11727 + }, + { + "epoch": 0.43, + "learning_rate": 1.572344087596587e-07, + "loss": 0.6826, + "step": 11728 + }, + { + "epoch": 0.43, + "learning_rate": 1.572307466949866e-07, + "loss": 0.6793, + "step": 11729 + }, + { + "epoch": 0.43, + "learning_rate": 1.5722708463031456e-07, + "loss": 0.687, + "step": 11730 + }, + { + "epoch": 0.43, + "learning_rate": 1.572234225656425e-07, + "loss": 0.6982, + "step": 11731 + }, + { + "epoch": 0.43, + "learning_rate": 1.5721976050097044e-07, + "loss": 0.7033, + "step": 11732 + }, + { + "epoch": 0.43, + "learning_rate": 1.5721609843629837e-07, + "loss": 0.6814, + "step": 11733 + }, + { + "epoch": 0.43, + "learning_rate": 1.5721243637162632e-07, + "loss": 0.6974, + "step": 11734 + }, + { + "epoch": 0.43, + "learning_rate": 1.5720877430695425e-07, + "loss": 0.6968, + "step": 11735 + }, + { + "epoch": 0.43, + "learning_rate": 1.572051122422822e-07, + "loss": 0.7035, + "step": 11736 + }, + { + "epoch": 0.43, + "learning_rate": 1.5720145017761013e-07, + "loss": 0.6933, + "step": 11737 + }, + { + "epoch": 0.43, + "learning_rate": 1.5719778811293808e-07, + "loss": 0.7051, + "step": 11738 + }, + { + "epoch": 0.43, + "learning_rate": 1.57194126048266e-07, + "loss": 0.7016, + "step": 11739 + }, + { + "epoch": 0.43, + "learning_rate": 1.5719046398359393e-07, + "loss": 0.6842, + "step": 11740 + }, + { + "epoch": 0.43, + "learning_rate": 1.571868019189219e-07, + "loss": 0.6902, + "step": 11741 + }, + { + "epoch": 0.43, + "learning_rate": 1.5718313985424981e-07, + "loss": 0.6913, + "step": 11742 + }, + { + "epoch": 0.43, + "learning_rate": 1.5717947778957777e-07, + "loss": 0.6738, + "step": 11743 + }, + { + "epoch": 0.43, + "learning_rate": 1.571758157249057e-07, + "loss": 0.692, + "step": 11744 + }, + { + "epoch": 0.43, + "learning_rate": 1.5717215366023365e-07, + "loss": 0.6883, + "step": 11745 + }, + { + "epoch": 0.43, + "learning_rate": 1.5716849159556157e-07, + "loss": 0.6919, + "step": 11746 + }, + { + "epoch": 0.43, + "learning_rate": 1.571648295308895e-07, + "loss": 0.6916, + "step": 11747 + }, + { + "epoch": 0.43, + "learning_rate": 1.5716116746621743e-07, + "loss": 0.6921, + "step": 11748 + }, + { + "epoch": 0.43, + "learning_rate": 1.5715750540154538e-07, + "loss": 0.6747, + "step": 11749 + }, + { + "epoch": 0.43, + "learning_rate": 1.571538433368733e-07, + "loss": 0.6912, + "step": 11750 + }, + { + "epoch": 0.43, + "learning_rate": 1.5715018127220126e-07, + "loss": 0.7057, + "step": 11751 + }, + { + "epoch": 0.43, + "learning_rate": 1.5714651920752919e-07, + "loss": 0.6962, + "step": 11752 + }, + { + "epoch": 0.43, + "learning_rate": 1.5714285714285714e-07, + "loss": 0.6924, + "step": 11753 + }, + { + "epoch": 0.43, + "learning_rate": 1.5713919507818507e-07, + "loss": 0.6998, + "step": 11754 + }, + { + "epoch": 0.43, + "learning_rate": 1.5713553301351302e-07, + "loss": 0.6995, + "step": 11755 + }, + { + "epoch": 0.43, + "learning_rate": 1.5713187094884095e-07, + "loss": 0.6901, + "step": 11756 + }, + { + "epoch": 0.43, + "learning_rate": 1.571282088841689e-07, + "loss": 0.6986, + "step": 11757 + }, + { + "epoch": 0.43, + "learning_rate": 1.5712454681949683e-07, + "loss": 0.7036, + "step": 11758 + }, + { + "epoch": 0.43, + "learning_rate": 1.5712088475482478e-07, + "loss": 0.6895, + "step": 11759 + }, + { + "epoch": 0.43, + "learning_rate": 1.571172226901527e-07, + "loss": 0.7077, + "step": 11760 + }, + { + "epoch": 0.43, + "learning_rate": 1.5711356062548063e-07, + "loss": 0.6889, + "step": 11761 + }, + { + "epoch": 0.43, + "learning_rate": 1.5710989856080859e-07, + "loss": 0.704, + "step": 11762 + }, + { + "epoch": 0.43, + "learning_rate": 1.571062364961365e-07, + "loss": 0.6983, + "step": 11763 + }, + { + "epoch": 0.43, + "learning_rate": 1.5710257443146447e-07, + "loss": 0.6847, + "step": 11764 + }, + { + "epoch": 0.43, + "learning_rate": 1.570989123667924e-07, + "loss": 0.6921, + "step": 11765 + }, + { + "epoch": 0.43, + "learning_rate": 1.5709525030212032e-07, + "loss": 0.7019, + "step": 11766 + }, + { + "epoch": 0.43, + "learning_rate": 1.5709158823744825e-07, + "loss": 0.6946, + "step": 11767 + }, + { + "epoch": 0.43, + "learning_rate": 1.570879261727762e-07, + "loss": 0.6805, + "step": 11768 + }, + { + "epoch": 0.43, + "learning_rate": 1.5708426410810413e-07, + "loss": 0.6717, + "step": 11769 + }, + { + "epoch": 0.43, + "learning_rate": 1.5708060204343208e-07, + "loss": 0.6748, + "step": 11770 + }, + { + "epoch": 0.43, + "learning_rate": 1.5707693997876e-07, + "loss": 0.681, + "step": 11771 + }, + { + "epoch": 0.43, + "learning_rate": 1.5707327791408796e-07, + "loss": 0.6939, + "step": 11772 + }, + { + "epoch": 0.43, + "learning_rate": 1.5706961584941589e-07, + "loss": 0.6873, + "step": 11773 + }, + { + "epoch": 0.43, + "learning_rate": 1.5706595378474384e-07, + "loss": 0.6998, + "step": 11774 + }, + { + "epoch": 0.43, + "learning_rate": 1.5706229172007177e-07, + "loss": 0.6727, + "step": 11775 + }, + { + "epoch": 0.43, + "learning_rate": 1.5705862965539972e-07, + "loss": 0.6999, + "step": 11776 + }, + { + "epoch": 0.43, + "learning_rate": 1.5705496759072764e-07, + "loss": 0.7006, + "step": 11777 + }, + { + "epoch": 0.43, + "learning_rate": 1.570513055260556e-07, + "loss": 0.6977, + "step": 11778 + }, + { + "epoch": 0.43, + "learning_rate": 1.5704764346138352e-07, + "loss": 0.6979, + "step": 11779 + }, + { + "epoch": 0.43, + "learning_rate": 1.5704398139671148e-07, + "loss": 0.6693, + "step": 11780 + }, + { + "epoch": 0.43, + "learning_rate": 1.570403193320394e-07, + "loss": 0.693, + "step": 11781 + }, + { + "epoch": 0.43, + "learning_rate": 1.5703665726736736e-07, + "loss": 0.6935, + "step": 11782 + }, + { + "epoch": 0.43, + "learning_rate": 1.5703299520269526e-07, + "loss": 0.7065, + "step": 11783 + }, + { + "epoch": 0.43, + "learning_rate": 1.570293331380232e-07, + "loss": 0.7016, + "step": 11784 + }, + { + "epoch": 0.43, + "learning_rate": 1.5702567107335114e-07, + "loss": 0.7057, + "step": 11785 + }, + { + "epoch": 0.43, + "learning_rate": 1.570220090086791e-07, + "loss": 0.6975, + "step": 11786 + }, + { + "epoch": 0.43, + "learning_rate": 1.5701834694400702e-07, + "loss": 0.6794, + "step": 11787 + }, + { + "epoch": 0.43, + "learning_rate": 1.5701468487933497e-07, + "loss": 0.6913, + "step": 11788 + }, + { + "epoch": 0.43, + "learning_rate": 1.570110228146629e-07, + "loss": 0.6983, + "step": 11789 + }, + { + "epoch": 0.43, + "learning_rate": 1.5700736074999082e-07, + "loss": 0.6805, + "step": 11790 + }, + { + "epoch": 0.43, + "learning_rate": 1.5700369868531878e-07, + "loss": 0.6691, + "step": 11791 + }, + { + "epoch": 0.43, + "learning_rate": 1.570000366206467e-07, + "loss": 0.6829, + "step": 11792 + }, + { + "epoch": 0.43, + "learning_rate": 1.5699637455597466e-07, + "loss": 0.6966, + "step": 11793 + }, + { + "epoch": 0.43, + "learning_rate": 1.5699271249130258e-07, + "loss": 0.6984, + "step": 11794 + }, + { + "epoch": 0.43, + "learning_rate": 1.5698905042663054e-07, + "loss": 0.6964, + "step": 11795 + }, + { + "epoch": 0.43, + "learning_rate": 1.5698538836195846e-07, + "loss": 0.6739, + "step": 11796 + }, + { + "epoch": 0.43, + "learning_rate": 1.5698172629728642e-07, + "loss": 0.6827, + "step": 11797 + }, + { + "epoch": 0.43, + "learning_rate": 1.5697806423261434e-07, + "loss": 0.6916, + "step": 11798 + }, + { + "epoch": 0.43, + "learning_rate": 1.569744021679423e-07, + "loss": 0.7131, + "step": 11799 + }, + { + "epoch": 0.43, + "learning_rate": 1.5697074010327022e-07, + "loss": 0.69, + "step": 11800 + }, + { + "epoch": 0.43, + "learning_rate": 1.5696707803859815e-07, + "loss": 0.6946, + "step": 11801 + }, + { + "epoch": 0.43, + "learning_rate": 1.5696341597392608e-07, + "loss": 0.696, + "step": 11802 + }, + { + "epoch": 0.43, + "learning_rate": 1.5695975390925403e-07, + "loss": 0.6861, + "step": 11803 + }, + { + "epoch": 0.43, + "learning_rate": 1.5695609184458196e-07, + "loss": 0.6875, + "step": 11804 + }, + { + "epoch": 0.43, + "learning_rate": 1.569524297799099e-07, + "loss": 0.7149, + "step": 11805 + }, + { + "epoch": 0.43, + "learning_rate": 1.5694876771523784e-07, + "loss": 0.6949, + "step": 11806 + }, + { + "epoch": 0.43, + "learning_rate": 1.569451056505658e-07, + "loss": 0.7055, + "step": 11807 + }, + { + "epoch": 0.43, + "learning_rate": 1.5694144358589372e-07, + "loss": 0.6878, + "step": 11808 + }, + { + "epoch": 0.43, + "learning_rate": 1.5693778152122167e-07, + "loss": 0.6994, + "step": 11809 + }, + { + "epoch": 0.43, + "learning_rate": 1.569341194565496e-07, + "loss": 0.7024, + "step": 11810 + }, + { + "epoch": 0.43, + "learning_rate": 1.5693045739187752e-07, + "loss": 0.6951, + "step": 11811 + }, + { + "epoch": 0.43, + "learning_rate": 1.5692679532720548e-07, + "loss": 0.6992, + "step": 11812 + }, + { + "epoch": 0.43, + "learning_rate": 1.569231332625334e-07, + "loss": 0.7015, + "step": 11813 + }, + { + "epoch": 0.43, + "learning_rate": 1.5691947119786136e-07, + "loss": 0.701, + "step": 11814 + }, + { + "epoch": 0.43, + "learning_rate": 1.5691580913318928e-07, + "loss": 0.6895, + "step": 11815 + }, + { + "epoch": 0.43, + "learning_rate": 1.5691214706851724e-07, + "loss": 0.7033, + "step": 11816 + }, + { + "epoch": 0.43, + "learning_rate": 1.5690848500384516e-07, + "loss": 0.6917, + "step": 11817 + }, + { + "epoch": 0.43, + "learning_rate": 1.5690482293917312e-07, + "loss": 0.6979, + "step": 11818 + }, + { + "epoch": 0.43, + "learning_rate": 1.5690116087450102e-07, + "loss": 0.6943, + "step": 11819 + }, + { + "epoch": 0.43, + "learning_rate": 1.5689749880982897e-07, + "loss": 0.6918, + "step": 11820 + }, + { + "epoch": 0.43, + "learning_rate": 1.568938367451569e-07, + "loss": 0.6795, + "step": 11821 + }, + { + "epoch": 0.43, + "learning_rate": 1.5689017468048485e-07, + "loss": 0.6994, + "step": 11822 + }, + { + "epoch": 0.43, + "learning_rate": 1.5688651261581278e-07, + "loss": 0.693, + "step": 11823 + }, + { + "epoch": 0.43, + "learning_rate": 1.5688285055114073e-07, + "loss": 0.7002, + "step": 11824 + }, + { + "epoch": 0.43, + "learning_rate": 1.5687918848646866e-07, + "loss": 0.6883, + "step": 11825 + }, + { + "epoch": 0.43, + "learning_rate": 1.568755264217966e-07, + "loss": 0.7027, + "step": 11826 + }, + { + "epoch": 0.43, + "learning_rate": 1.5687186435712454e-07, + "loss": 0.6827, + "step": 11827 + }, + { + "epoch": 0.43, + "learning_rate": 1.568682022924525e-07, + "loss": 0.7093, + "step": 11828 + }, + { + "epoch": 0.43, + "learning_rate": 1.5686454022778042e-07, + "loss": 0.7014, + "step": 11829 + }, + { + "epoch": 0.43, + "learning_rate": 1.5686087816310837e-07, + "loss": 0.7194, + "step": 11830 + }, + { + "epoch": 0.43, + "learning_rate": 1.568572160984363e-07, + "loss": 0.6941, + "step": 11831 + }, + { + "epoch": 0.43, + "learning_rate": 1.5685355403376422e-07, + "loss": 0.6903, + "step": 11832 + }, + { + "epoch": 0.43, + "learning_rate": 1.5684989196909217e-07, + "loss": 0.6818, + "step": 11833 + }, + { + "epoch": 0.43, + "learning_rate": 1.568462299044201e-07, + "loss": 0.7026, + "step": 11834 + }, + { + "epoch": 0.43, + "learning_rate": 1.5684256783974805e-07, + "loss": 0.7118, + "step": 11835 + }, + { + "epoch": 0.43, + "learning_rate": 1.5683890577507598e-07, + "loss": 0.6806, + "step": 11836 + }, + { + "epoch": 0.43, + "learning_rate": 1.568352437104039e-07, + "loss": 0.6882, + "step": 11837 + }, + { + "epoch": 0.43, + "learning_rate": 1.5683158164573183e-07, + "loss": 0.707, + "step": 11838 + }, + { + "epoch": 0.43, + "learning_rate": 1.568279195810598e-07, + "loss": 0.6937, + "step": 11839 + }, + { + "epoch": 0.43, + "learning_rate": 1.5682425751638771e-07, + "loss": 0.6917, + "step": 11840 + }, + { + "epoch": 0.43, + "learning_rate": 1.5682059545171567e-07, + "loss": 0.7047, + "step": 11841 + }, + { + "epoch": 0.43, + "learning_rate": 1.568169333870436e-07, + "loss": 0.7035, + "step": 11842 + }, + { + "epoch": 0.43, + "learning_rate": 1.5681327132237155e-07, + "loss": 0.6851, + "step": 11843 + }, + { + "epoch": 0.43, + "learning_rate": 1.5680960925769947e-07, + "loss": 0.711, + "step": 11844 + }, + { + "epoch": 0.43, + "learning_rate": 1.5680594719302743e-07, + "loss": 0.7003, + "step": 11845 + }, + { + "epoch": 0.43, + "learning_rate": 1.5680228512835535e-07, + "loss": 0.7252, + "step": 11846 + }, + { + "epoch": 0.43, + "learning_rate": 1.567986230636833e-07, + "loss": 0.6888, + "step": 11847 + }, + { + "epoch": 0.43, + "learning_rate": 1.5679496099901123e-07, + "loss": 0.7087, + "step": 11848 + }, + { + "epoch": 0.43, + "learning_rate": 1.567912989343392e-07, + "loss": 0.6962, + "step": 11849 + }, + { + "epoch": 0.43, + "learning_rate": 1.5678763686966711e-07, + "loss": 0.6922, + "step": 11850 + }, + { + "epoch": 0.43, + "learning_rate": 1.5678397480499507e-07, + "loss": 0.7035, + "step": 11851 + }, + { + "epoch": 0.43, + "learning_rate": 1.56780312740323e-07, + "loss": 0.6907, + "step": 11852 + }, + { + "epoch": 0.43, + "learning_rate": 1.5677665067565095e-07, + "loss": 0.6942, + "step": 11853 + }, + { + "epoch": 0.43, + "learning_rate": 1.5677298861097887e-07, + "loss": 0.6761, + "step": 11854 + }, + { + "epoch": 0.43, + "learning_rate": 1.567693265463068e-07, + "loss": 0.6916, + "step": 11855 + }, + { + "epoch": 0.43, + "learning_rate": 1.5676566448163473e-07, + "loss": 0.6874, + "step": 11856 + }, + { + "epoch": 0.43, + "learning_rate": 1.5676200241696268e-07, + "loss": 0.6803, + "step": 11857 + }, + { + "epoch": 0.43, + "learning_rate": 1.567583403522906e-07, + "loss": 0.6897, + "step": 11858 + }, + { + "epoch": 0.43, + "learning_rate": 1.5675467828761853e-07, + "loss": 0.7008, + "step": 11859 + }, + { + "epoch": 0.43, + "learning_rate": 1.567510162229465e-07, + "loss": 0.6868, + "step": 11860 + }, + { + "epoch": 0.43, + "learning_rate": 1.5674735415827441e-07, + "loss": 0.7087, + "step": 11861 + }, + { + "epoch": 0.43, + "learning_rate": 1.5674369209360237e-07, + "loss": 0.6972, + "step": 11862 + }, + { + "epoch": 0.43, + "learning_rate": 1.567400300289303e-07, + "loss": 0.701, + "step": 11863 + }, + { + "epoch": 0.43, + "learning_rate": 1.5673636796425825e-07, + "loss": 0.6795, + "step": 11864 + }, + { + "epoch": 0.43, + "learning_rate": 1.5673270589958617e-07, + "loss": 0.6929, + "step": 11865 + }, + { + "epoch": 0.43, + "learning_rate": 1.5672904383491413e-07, + "loss": 0.6835, + "step": 11866 + }, + { + "epoch": 0.43, + "learning_rate": 1.5672538177024205e-07, + "loss": 0.6934, + "step": 11867 + }, + { + "epoch": 0.43, + "learning_rate": 1.5672171970557e-07, + "loss": 0.7109, + "step": 11868 + }, + { + "epoch": 0.43, + "learning_rate": 1.5671805764089793e-07, + "loss": 0.6937, + "step": 11869 + }, + { + "epoch": 0.43, + "learning_rate": 1.5671439557622589e-07, + "loss": 0.7107, + "step": 11870 + }, + { + "epoch": 0.43, + "learning_rate": 1.567107335115538e-07, + "loss": 0.6816, + "step": 11871 + }, + { + "epoch": 0.43, + "learning_rate": 1.5670707144688177e-07, + "loss": 0.7208, + "step": 11872 + }, + { + "epoch": 0.43, + "learning_rate": 1.5670340938220967e-07, + "loss": 0.7115, + "step": 11873 + }, + { + "epoch": 0.43, + "learning_rate": 1.5669974731753762e-07, + "loss": 0.6798, + "step": 11874 + }, + { + "epoch": 0.43, + "learning_rate": 1.5669608525286555e-07, + "loss": 0.6874, + "step": 11875 + }, + { + "epoch": 0.43, + "learning_rate": 1.566924231881935e-07, + "loss": 0.6993, + "step": 11876 + }, + { + "epoch": 0.43, + "learning_rate": 1.5668876112352143e-07, + "loss": 0.6953, + "step": 11877 + }, + { + "epoch": 0.43, + "learning_rate": 1.5668509905884938e-07, + "loss": 0.6827, + "step": 11878 + }, + { + "epoch": 0.43, + "learning_rate": 1.566814369941773e-07, + "loss": 0.7015, + "step": 11879 + }, + { + "epoch": 0.43, + "learning_rate": 1.5667777492950526e-07, + "loss": 0.6964, + "step": 11880 + }, + { + "epoch": 0.43, + "learning_rate": 1.5667411286483319e-07, + "loss": 0.6811, + "step": 11881 + }, + { + "epoch": 0.43, + "learning_rate": 1.566704508001611e-07, + "loss": 0.6778, + "step": 11882 + }, + { + "epoch": 0.43, + "learning_rate": 1.5666678873548907e-07, + "loss": 0.6869, + "step": 11883 + }, + { + "epoch": 0.43, + "learning_rate": 1.56663126670817e-07, + "loss": 0.6797, + "step": 11884 + }, + { + "epoch": 0.43, + "learning_rate": 1.5665946460614495e-07, + "loss": 0.6941, + "step": 11885 + }, + { + "epoch": 0.43, + "learning_rate": 1.5665580254147287e-07, + "loss": 0.6784, + "step": 11886 + }, + { + "epoch": 0.43, + "learning_rate": 1.5665214047680083e-07, + "loss": 0.6956, + "step": 11887 + }, + { + "epoch": 0.43, + "learning_rate": 1.5664847841212875e-07, + "loss": 0.6942, + "step": 11888 + }, + { + "epoch": 0.43, + "learning_rate": 1.566448163474567e-07, + "loss": 0.7047, + "step": 11889 + }, + { + "epoch": 0.44, + "learning_rate": 1.5664115428278463e-07, + "loss": 0.6752, + "step": 11890 + }, + { + "epoch": 0.44, + "learning_rate": 1.5663749221811256e-07, + "loss": 0.7121, + "step": 11891 + }, + { + "epoch": 0.44, + "learning_rate": 1.5663383015344049e-07, + "loss": 0.6907, + "step": 11892 + }, + { + "epoch": 0.44, + "learning_rate": 1.5663016808876844e-07, + "loss": 0.6911, + "step": 11893 + }, + { + "epoch": 0.44, + "learning_rate": 1.5662650602409637e-07, + "loss": 0.6884, + "step": 11894 + }, + { + "epoch": 0.44, + "learning_rate": 1.5662284395942432e-07, + "loss": 0.706, + "step": 11895 + }, + { + "epoch": 0.44, + "learning_rate": 1.5661918189475224e-07, + "loss": 0.7023, + "step": 11896 + }, + { + "epoch": 0.44, + "learning_rate": 1.566155198300802e-07, + "loss": 0.6957, + "step": 11897 + }, + { + "epoch": 0.44, + "learning_rate": 1.5661185776540812e-07, + "loss": 0.7013, + "step": 11898 + }, + { + "epoch": 0.44, + "learning_rate": 1.5660819570073608e-07, + "loss": 0.7047, + "step": 11899 + }, + { + "epoch": 0.44, + "learning_rate": 1.56604533636064e-07, + "loss": 0.7027, + "step": 11900 + }, + { + "epoch": 0.44, + "learning_rate": 1.5660087157139196e-07, + "loss": 0.6894, + "step": 11901 + }, + { + "epoch": 0.44, + "learning_rate": 1.5659720950671988e-07, + "loss": 0.6949, + "step": 11902 + }, + { + "epoch": 0.44, + "learning_rate": 1.565935474420478e-07, + "loss": 0.7018, + "step": 11903 + }, + { + "epoch": 0.44, + "learning_rate": 1.5658988537737576e-07, + "loss": 0.6877, + "step": 11904 + }, + { + "epoch": 0.44, + "learning_rate": 1.565862233127037e-07, + "loss": 0.7017, + "step": 11905 + }, + { + "epoch": 0.44, + "learning_rate": 1.5658256124803164e-07, + "loss": 0.708, + "step": 11906 + }, + { + "epoch": 0.44, + "learning_rate": 1.5657889918335957e-07, + "loss": 0.6924, + "step": 11907 + }, + { + "epoch": 0.44, + "learning_rate": 1.5657523711868752e-07, + "loss": 0.6939, + "step": 11908 + }, + { + "epoch": 0.44, + "learning_rate": 1.5657157505401542e-07, + "loss": 0.681, + "step": 11909 + }, + { + "epoch": 0.44, + "learning_rate": 1.5656791298934338e-07, + "loss": 0.6823, + "step": 11910 + }, + { + "epoch": 0.44, + "learning_rate": 1.565642509246713e-07, + "loss": 0.7051, + "step": 11911 + }, + { + "epoch": 0.44, + "learning_rate": 1.5656058885999926e-07, + "loss": 0.6996, + "step": 11912 + }, + { + "epoch": 0.44, + "learning_rate": 1.5655692679532718e-07, + "loss": 0.7047, + "step": 11913 + }, + { + "epoch": 0.44, + "learning_rate": 1.5655326473065514e-07, + "loss": 0.7014, + "step": 11914 + }, + { + "epoch": 0.44, + "learning_rate": 1.5654960266598306e-07, + "loss": 0.705, + "step": 11915 + }, + { + "epoch": 0.44, + "learning_rate": 1.5654594060131102e-07, + "loss": 0.6986, + "step": 11916 + }, + { + "epoch": 0.44, + "learning_rate": 1.5654227853663894e-07, + "loss": 0.6584, + "step": 11917 + }, + { + "epoch": 0.44, + "learning_rate": 1.565386164719669e-07, + "loss": 0.6973, + "step": 11918 + }, + { + "epoch": 0.44, + "learning_rate": 1.5653495440729482e-07, + "loss": 0.6907, + "step": 11919 + }, + { + "epoch": 0.44, + "learning_rate": 1.5653129234262278e-07, + "loss": 0.6871, + "step": 11920 + }, + { + "epoch": 0.44, + "learning_rate": 1.565276302779507e-07, + "loss": 0.6939, + "step": 11921 + }, + { + "epoch": 0.44, + "learning_rate": 1.5652396821327866e-07, + "loss": 0.6677, + "step": 11922 + }, + { + "epoch": 0.44, + "learning_rate": 1.5652030614860658e-07, + "loss": 0.6882, + "step": 11923 + }, + { + "epoch": 0.44, + "learning_rate": 1.5651664408393454e-07, + "loss": 0.7004, + "step": 11924 + }, + { + "epoch": 0.44, + "learning_rate": 1.5651298201926246e-07, + "loss": 0.6946, + "step": 11925 + }, + { + "epoch": 0.44, + "learning_rate": 1.565093199545904e-07, + "loss": 0.688, + "step": 11926 + }, + { + "epoch": 0.44, + "learning_rate": 1.5650565788991832e-07, + "loss": 0.7024, + "step": 11927 + }, + { + "epoch": 0.44, + "learning_rate": 1.5650199582524627e-07, + "loss": 0.7006, + "step": 11928 + }, + { + "epoch": 0.44, + "learning_rate": 1.564983337605742e-07, + "loss": 0.6986, + "step": 11929 + }, + { + "epoch": 0.44, + "learning_rate": 1.5649467169590212e-07, + "loss": 0.6957, + "step": 11930 + }, + { + "epoch": 0.44, + "learning_rate": 1.5649100963123008e-07, + "loss": 0.6787, + "step": 11931 + }, + { + "epoch": 0.44, + "learning_rate": 1.56487347566558e-07, + "loss": 0.7135, + "step": 11932 + }, + { + "epoch": 0.44, + "learning_rate": 1.5648368550188596e-07, + "loss": 0.6978, + "step": 11933 + }, + { + "epoch": 0.44, + "learning_rate": 1.5648002343721388e-07, + "loss": 0.7076, + "step": 11934 + }, + { + "epoch": 0.44, + "learning_rate": 1.5647636137254184e-07, + "loss": 0.7001, + "step": 11935 + }, + { + "epoch": 0.44, + "learning_rate": 1.5647269930786976e-07, + "loss": 0.6974, + "step": 11936 + }, + { + "epoch": 0.44, + "learning_rate": 1.5646903724319772e-07, + "loss": 0.6832, + "step": 11937 + }, + { + "epoch": 0.44, + "learning_rate": 1.5646537517852564e-07, + "loss": 0.7194, + "step": 11938 + }, + { + "epoch": 0.44, + "learning_rate": 1.564617131138536e-07, + "loss": 0.6986, + "step": 11939 + }, + { + "epoch": 0.44, + "learning_rate": 1.5645805104918152e-07, + "loss": 0.6877, + "step": 11940 + }, + { + "epoch": 0.44, + "learning_rate": 1.5645438898450948e-07, + "loss": 0.6822, + "step": 11941 + }, + { + "epoch": 0.44, + "learning_rate": 1.564507269198374e-07, + "loss": 0.6913, + "step": 11942 + }, + { + "epoch": 0.44, + "learning_rate": 1.5644706485516536e-07, + "loss": 0.6891, + "step": 11943 + }, + { + "epoch": 0.44, + "learning_rate": 1.5644340279049328e-07, + "loss": 0.7211, + "step": 11944 + }, + { + "epoch": 0.44, + "learning_rate": 1.564397407258212e-07, + "loss": 0.7036, + "step": 11945 + }, + { + "epoch": 0.44, + "learning_rate": 1.5643607866114914e-07, + "loss": 0.6957, + "step": 11946 + }, + { + "epoch": 0.44, + "learning_rate": 1.564324165964771e-07, + "loss": 0.69, + "step": 11947 + }, + { + "epoch": 0.44, + "learning_rate": 1.5642875453180502e-07, + "loss": 0.6809, + "step": 11948 + }, + { + "epoch": 0.44, + "learning_rate": 1.5642509246713297e-07, + "loss": 0.709, + "step": 11949 + }, + { + "epoch": 0.44, + "learning_rate": 1.564214304024609e-07, + "loss": 0.6766, + "step": 11950 + }, + { + "epoch": 0.44, + "learning_rate": 1.5641776833778885e-07, + "loss": 0.7095, + "step": 11951 + }, + { + "epoch": 0.44, + "learning_rate": 1.5641410627311678e-07, + "loss": 0.6973, + "step": 11952 + }, + { + "epoch": 0.44, + "learning_rate": 1.564104442084447e-07, + "loss": 0.6968, + "step": 11953 + }, + { + "epoch": 0.44, + "learning_rate": 1.5640678214377265e-07, + "loss": 0.6924, + "step": 11954 + }, + { + "epoch": 0.44, + "learning_rate": 1.5640312007910058e-07, + "loss": 0.6773, + "step": 11955 + }, + { + "epoch": 0.44, + "learning_rate": 1.5639945801442853e-07, + "loss": 0.6994, + "step": 11956 + }, + { + "epoch": 0.44, + "learning_rate": 1.5639579594975646e-07, + "loss": 0.6874, + "step": 11957 + }, + { + "epoch": 0.44, + "learning_rate": 1.5639213388508441e-07, + "loss": 0.6867, + "step": 11958 + }, + { + "epoch": 0.44, + "learning_rate": 1.5638847182041234e-07, + "loss": 0.6888, + "step": 11959 + }, + { + "epoch": 0.44, + "learning_rate": 1.563848097557403e-07, + "loss": 0.6692, + "step": 11960 + }, + { + "epoch": 0.44, + "learning_rate": 1.5638114769106822e-07, + "loss": 0.6947, + "step": 11961 + }, + { + "epoch": 0.44, + "learning_rate": 1.5637748562639617e-07, + "loss": 0.6961, + "step": 11962 + }, + { + "epoch": 0.44, + "learning_rate": 1.5637382356172407e-07, + "loss": 0.6877, + "step": 11963 + }, + { + "epoch": 0.44, + "learning_rate": 1.5637016149705203e-07, + "loss": 0.7047, + "step": 11964 + }, + { + "epoch": 0.44, + "learning_rate": 1.5636649943237995e-07, + "loss": 0.6821, + "step": 11965 + }, + { + "epoch": 0.44, + "learning_rate": 1.563628373677079e-07, + "loss": 0.6865, + "step": 11966 + }, + { + "epoch": 0.44, + "learning_rate": 1.5635917530303583e-07, + "loss": 0.6878, + "step": 11967 + }, + { + "epoch": 0.44, + "learning_rate": 1.563555132383638e-07, + "loss": 0.6919, + "step": 11968 + }, + { + "epoch": 0.44, + "learning_rate": 1.5635185117369171e-07, + "loss": 0.6979, + "step": 11969 + }, + { + "epoch": 0.44, + "learning_rate": 1.5634818910901967e-07, + "loss": 0.7059, + "step": 11970 + }, + { + "epoch": 0.44, + "learning_rate": 1.563445270443476e-07, + "loss": 0.6913, + "step": 11971 + }, + { + "epoch": 0.44, + "learning_rate": 1.5634086497967555e-07, + "loss": 0.691, + "step": 11972 + }, + { + "epoch": 0.44, + "learning_rate": 1.5633720291500347e-07, + "loss": 0.6973, + "step": 11973 + }, + { + "epoch": 0.44, + "learning_rate": 1.563335408503314e-07, + "loss": 0.7045, + "step": 11974 + }, + { + "epoch": 0.44, + "learning_rate": 1.5632987878565935e-07, + "loss": 0.6891, + "step": 11975 + }, + { + "epoch": 0.44, + "learning_rate": 1.5632621672098728e-07, + "loss": 0.685, + "step": 11976 + }, + { + "epoch": 0.44, + "learning_rate": 1.5632255465631523e-07, + "loss": 0.6845, + "step": 11977 + }, + { + "epoch": 0.44, + "learning_rate": 1.5631889259164316e-07, + "loss": 0.6896, + "step": 11978 + }, + { + "epoch": 0.44, + "learning_rate": 1.5631523052697111e-07, + "loss": 0.687, + "step": 11979 + }, + { + "epoch": 0.44, + "learning_rate": 1.5631156846229901e-07, + "loss": 0.6998, + "step": 11980 + }, + { + "epoch": 0.44, + "learning_rate": 1.5630790639762697e-07, + "loss": 0.6711, + "step": 11981 + }, + { + "epoch": 0.44, + "learning_rate": 1.563042443329549e-07, + "loss": 0.7047, + "step": 11982 + }, + { + "epoch": 0.44, + "learning_rate": 1.5630058226828285e-07, + "loss": 0.682, + "step": 11983 + }, + { + "epoch": 0.44, + "learning_rate": 1.5629692020361077e-07, + "loss": 0.71, + "step": 11984 + }, + { + "epoch": 0.44, + "learning_rate": 1.5629325813893873e-07, + "loss": 0.689, + "step": 11985 + }, + { + "epoch": 0.44, + "learning_rate": 1.5628959607426665e-07, + "loss": 0.6679, + "step": 11986 + }, + { + "epoch": 0.44, + "learning_rate": 1.562859340095946e-07, + "loss": 0.702, + "step": 11987 + }, + { + "epoch": 0.44, + "learning_rate": 1.5628227194492253e-07, + "loss": 0.6874, + "step": 11988 + }, + { + "epoch": 0.44, + "learning_rate": 1.5627860988025049e-07, + "loss": 0.6991, + "step": 11989 + }, + { + "epoch": 0.44, + "learning_rate": 1.562749478155784e-07, + "loss": 0.713, + "step": 11990 + }, + { + "epoch": 0.44, + "learning_rate": 1.5627128575090637e-07, + "loss": 0.6901, + "step": 11991 + }, + { + "epoch": 0.44, + "learning_rate": 1.562676236862343e-07, + "loss": 0.6856, + "step": 11992 + }, + { + "epoch": 0.44, + "learning_rate": 1.5626396162156225e-07, + "loss": 0.7074, + "step": 11993 + }, + { + "epoch": 0.44, + "learning_rate": 1.5626029955689017e-07, + "loss": 0.6976, + "step": 11994 + }, + { + "epoch": 0.44, + "learning_rate": 1.5625663749221813e-07, + "loss": 0.7001, + "step": 11995 + }, + { + "epoch": 0.44, + "learning_rate": 1.5625297542754605e-07, + "loss": 0.6865, + "step": 11996 + }, + { + "epoch": 0.44, + "learning_rate": 1.5624931336287398e-07, + "loss": 0.7097, + "step": 11997 + }, + { + "epoch": 0.44, + "learning_rate": 1.562456512982019e-07, + "loss": 0.6891, + "step": 11998 + }, + { + "epoch": 0.44, + "learning_rate": 1.5624198923352986e-07, + "loss": 0.683, + "step": 11999 + }, + { + "epoch": 0.44, + "learning_rate": 1.5623832716885779e-07, + "loss": 0.6925, + "step": 12000 + }, + { + "epoch": 0.44, + "learning_rate": 1.562346651041857e-07, + "loss": 0.6886, + "step": 12001 + }, + { + "epoch": 0.44, + "learning_rate": 1.5623100303951367e-07, + "loss": 0.6875, + "step": 12002 + }, + { + "epoch": 0.44, + "learning_rate": 1.562273409748416e-07, + "loss": 0.7026, + "step": 12003 + }, + { + "epoch": 0.44, + "learning_rate": 1.5622367891016955e-07, + "loss": 0.7052, + "step": 12004 + }, + { + "epoch": 0.44, + "learning_rate": 1.5622001684549747e-07, + "loss": 0.7055, + "step": 12005 + }, + { + "epoch": 0.44, + "learning_rate": 1.5621635478082543e-07, + "loss": 0.69, + "step": 12006 + }, + { + "epoch": 0.44, + "learning_rate": 1.5621269271615335e-07, + "loss": 0.6897, + "step": 12007 + }, + { + "epoch": 0.44, + "learning_rate": 1.562090306514813e-07, + "loss": 0.6752, + "step": 12008 + }, + { + "epoch": 0.44, + "learning_rate": 1.5620536858680923e-07, + "loss": 0.7001, + "step": 12009 + }, + { + "epoch": 0.44, + "learning_rate": 1.5620170652213719e-07, + "loss": 0.694, + "step": 12010 + }, + { + "epoch": 0.44, + "learning_rate": 1.561980444574651e-07, + "loss": 0.6803, + "step": 12011 + }, + { + "epoch": 0.44, + "learning_rate": 1.5619438239279306e-07, + "loss": 0.6973, + "step": 12012 + }, + { + "epoch": 0.44, + "learning_rate": 1.56190720328121e-07, + "loss": 0.6854, + "step": 12013 + }, + { + "epoch": 0.44, + "learning_rate": 1.5618705826344894e-07, + "loss": 0.6881, + "step": 12014 + }, + { + "epoch": 0.44, + "learning_rate": 1.5618339619877687e-07, + "loss": 0.7085, + "step": 12015 + }, + { + "epoch": 0.44, + "learning_rate": 1.561797341341048e-07, + "loss": 0.6978, + "step": 12016 + }, + { + "epoch": 0.44, + "learning_rate": 1.5617607206943272e-07, + "loss": 0.6915, + "step": 12017 + }, + { + "epoch": 0.44, + "learning_rate": 1.5617241000476068e-07, + "loss": 0.6999, + "step": 12018 + }, + { + "epoch": 0.44, + "learning_rate": 1.561687479400886e-07, + "loss": 0.6974, + "step": 12019 + }, + { + "epoch": 0.44, + "learning_rate": 1.5616508587541656e-07, + "loss": 0.708, + "step": 12020 + }, + { + "epoch": 0.44, + "learning_rate": 1.5616142381074448e-07, + "loss": 0.6858, + "step": 12021 + }, + { + "epoch": 0.44, + "learning_rate": 1.5615776174607244e-07, + "loss": 0.6862, + "step": 12022 + }, + { + "epoch": 0.44, + "learning_rate": 1.5615409968140036e-07, + "loss": 0.688, + "step": 12023 + }, + { + "epoch": 0.44, + "learning_rate": 1.561504376167283e-07, + "loss": 0.692, + "step": 12024 + }, + { + "epoch": 0.44, + "learning_rate": 1.5614677555205624e-07, + "loss": 0.6953, + "step": 12025 + }, + { + "epoch": 0.44, + "learning_rate": 1.5614311348738417e-07, + "loss": 0.6994, + "step": 12026 + }, + { + "epoch": 0.44, + "learning_rate": 1.5613945142271212e-07, + "loss": 0.7073, + "step": 12027 + }, + { + "epoch": 0.44, + "learning_rate": 1.5613578935804005e-07, + "loss": 0.6857, + "step": 12028 + }, + { + "epoch": 0.44, + "learning_rate": 1.56132127293368e-07, + "loss": 0.7031, + "step": 12029 + }, + { + "epoch": 0.44, + "learning_rate": 1.5612846522869593e-07, + "loss": 0.6871, + "step": 12030 + }, + { + "epoch": 0.44, + "learning_rate": 1.5612480316402388e-07, + "loss": 0.6893, + "step": 12031 + }, + { + "epoch": 0.44, + "learning_rate": 1.561211410993518e-07, + "loss": 0.6869, + "step": 12032 + }, + { + "epoch": 0.44, + "learning_rate": 1.5611747903467976e-07, + "loss": 0.6988, + "step": 12033 + }, + { + "epoch": 0.44, + "learning_rate": 1.5611381697000766e-07, + "loss": 0.7017, + "step": 12034 + }, + { + "epoch": 0.44, + "learning_rate": 1.5611015490533562e-07, + "loss": 0.7003, + "step": 12035 + }, + { + "epoch": 0.44, + "learning_rate": 1.5610649284066354e-07, + "loss": 0.6895, + "step": 12036 + }, + { + "epoch": 0.44, + "learning_rate": 1.561028307759915e-07, + "loss": 0.7012, + "step": 12037 + }, + { + "epoch": 0.44, + "learning_rate": 1.5609916871131942e-07, + "loss": 0.6919, + "step": 12038 + }, + { + "epoch": 0.44, + "learning_rate": 1.5609550664664738e-07, + "loss": 0.6787, + "step": 12039 + }, + { + "epoch": 0.44, + "learning_rate": 1.560918445819753e-07, + "loss": 0.6884, + "step": 12040 + }, + { + "epoch": 0.44, + "learning_rate": 1.5608818251730326e-07, + "loss": 0.6895, + "step": 12041 + }, + { + "epoch": 0.44, + "learning_rate": 1.5608452045263118e-07, + "loss": 0.7038, + "step": 12042 + }, + { + "epoch": 0.44, + "learning_rate": 1.5608085838795914e-07, + "loss": 0.6744, + "step": 12043 + }, + { + "epoch": 0.44, + "learning_rate": 1.5607719632328706e-07, + "loss": 0.6988, + "step": 12044 + }, + { + "epoch": 0.44, + "learning_rate": 1.56073534258615e-07, + "loss": 0.6889, + "step": 12045 + }, + { + "epoch": 0.44, + "learning_rate": 1.5606987219394294e-07, + "loss": 0.6972, + "step": 12046 + }, + { + "epoch": 0.44, + "learning_rate": 1.5606621012927087e-07, + "loss": 0.6825, + "step": 12047 + }, + { + "epoch": 0.44, + "learning_rate": 1.5606254806459882e-07, + "loss": 0.6757, + "step": 12048 + }, + { + "epoch": 0.44, + "learning_rate": 1.5605888599992675e-07, + "loss": 0.6917, + "step": 12049 + }, + { + "epoch": 0.44, + "learning_rate": 1.560552239352547e-07, + "loss": 0.7004, + "step": 12050 + }, + { + "epoch": 0.44, + "learning_rate": 1.5605156187058263e-07, + "loss": 0.6923, + "step": 12051 + }, + { + "epoch": 0.44, + "learning_rate": 1.5604789980591056e-07, + "loss": 0.6695, + "step": 12052 + }, + { + "epoch": 0.44, + "learning_rate": 1.5604423774123848e-07, + "loss": 0.6801, + "step": 12053 + }, + { + "epoch": 0.44, + "learning_rate": 1.5604057567656644e-07, + "loss": 0.7183, + "step": 12054 + }, + { + "epoch": 0.44, + "learning_rate": 1.5603691361189436e-07, + "loss": 0.6971, + "step": 12055 + }, + { + "epoch": 0.44, + "learning_rate": 1.5603325154722232e-07, + "loss": 0.701, + "step": 12056 + }, + { + "epoch": 0.44, + "learning_rate": 1.5602958948255024e-07, + "loss": 0.6846, + "step": 12057 + }, + { + "epoch": 0.44, + "learning_rate": 1.560259274178782e-07, + "loss": 0.6867, + "step": 12058 + }, + { + "epoch": 0.44, + "learning_rate": 1.5602226535320612e-07, + "loss": 0.6926, + "step": 12059 + }, + { + "epoch": 0.44, + "learning_rate": 1.5601860328853408e-07, + "loss": 0.692, + "step": 12060 + }, + { + "epoch": 0.44, + "learning_rate": 1.56014941223862e-07, + "loss": 0.7031, + "step": 12061 + }, + { + "epoch": 0.44, + "learning_rate": 1.5601127915918996e-07, + "loss": 0.715, + "step": 12062 + }, + { + "epoch": 0.44, + "learning_rate": 1.5600761709451788e-07, + "loss": 0.6913, + "step": 12063 + }, + { + "epoch": 0.44, + "learning_rate": 1.5600395502984584e-07, + "loss": 0.6869, + "step": 12064 + }, + { + "epoch": 0.44, + "learning_rate": 1.5600029296517376e-07, + "loss": 0.7011, + "step": 12065 + }, + { + "epoch": 0.44, + "learning_rate": 1.5599663090050172e-07, + "loss": 0.685, + "step": 12066 + }, + { + "epoch": 0.44, + "learning_rate": 1.5599296883582964e-07, + "loss": 0.6996, + "step": 12067 + }, + { + "epoch": 0.44, + "learning_rate": 1.5598930677115757e-07, + "loss": 0.6812, + "step": 12068 + }, + { + "epoch": 0.44, + "learning_rate": 1.5598564470648552e-07, + "loss": 0.6838, + "step": 12069 + }, + { + "epoch": 0.44, + "learning_rate": 1.5598198264181345e-07, + "loss": 0.7024, + "step": 12070 + }, + { + "epoch": 0.44, + "learning_rate": 1.5597832057714138e-07, + "loss": 0.6889, + "step": 12071 + }, + { + "epoch": 0.44, + "learning_rate": 1.559746585124693e-07, + "loss": 0.6862, + "step": 12072 + }, + { + "epoch": 0.44, + "learning_rate": 1.5597099644779725e-07, + "loss": 0.7119, + "step": 12073 + }, + { + "epoch": 0.44, + "learning_rate": 1.5596733438312518e-07, + "loss": 0.6928, + "step": 12074 + }, + { + "epoch": 0.44, + "learning_rate": 1.5596367231845313e-07, + "loss": 0.6841, + "step": 12075 + }, + { + "epoch": 0.44, + "learning_rate": 1.5596001025378106e-07, + "loss": 0.6945, + "step": 12076 + }, + { + "epoch": 0.44, + "learning_rate": 1.5595634818910901e-07, + "loss": 0.6918, + "step": 12077 + }, + { + "epoch": 0.44, + "learning_rate": 1.5595268612443694e-07, + "loss": 0.7053, + "step": 12078 + }, + { + "epoch": 0.44, + "learning_rate": 1.559490240597649e-07, + "loss": 0.7063, + "step": 12079 + }, + { + "epoch": 0.44, + "learning_rate": 1.5594536199509282e-07, + "loss": 0.7036, + "step": 12080 + }, + { + "epoch": 0.44, + "learning_rate": 1.5594169993042077e-07, + "loss": 0.6957, + "step": 12081 + }, + { + "epoch": 0.44, + "learning_rate": 1.559380378657487e-07, + "loss": 0.7096, + "step": 12082 + }, + { + "epoch": 0.44, + "learning_rate": 1.5593437580107665e-07, + "loss": 0.6881, + "step": 12083 + }, + { + "epoch": 0.44, + "learning_rate": 1.5593071373640458e-07, + "loss": 0.6991, + "step": 12084 + }, + { + "epoch": 0.44, + "learning_rate": 1.5592705167173253e-07, + "loss": 0.6984, + "step": 12085 + }, + { + "epoch": 0.44, + "learning_rate": 1.5592338960706046e-07, + "loss": 0.7113, + "step": 12086 + }, + { + "epoch": 0.44, + "learning_rate": 1.5591972754238841e-07, + "loss": 0.6923, + "step": 12087 + }, + { + "epoch": 0.44, + "learning_rate": 1.5591606547771631e-07, + "loss": 0.6981, + "step": 12088 + }, + { + "epoch": 0.44, + "learning_rate": 1.5591240341304427e-07, + "loss": 0.6996, + "step": 12089 + }, + { + "epoch": 0.44, + "learning_rate": 1.559087413483722e-07, + "loss": 0.6957, + "step": 12090 + }, + { + "epoch": 0.44, + "learning_rate": 1.5590507928370015e-07, + "loss": 0.7007, + "step": 12091 + }, + { + "epoch": 0.44, + "learning_rate": 1.5590141721902807e-07, + "loss": 0.6713, + "step": 12092 + }, + { + "epoch": 0.44, + "learning_rate": 1.5589775515435603e-07, + "loss": 0.708, + "step": 12093 + }, + { + "epoch": 0.44, + "learning_rate": 1.5589409308968395e-07, + "loss": 0.6899, + "step": 12094 + }, + { + "epoch": 0.44, + "learning_rate": 1.5589043102501188e-07, + "loss": 0.7194, + "step": 12095 + }, + { + "epoch": 0.44, + "learning_rate": 1.5588676896033983e-07, + "loss": 0.6876, + "step": 12096 + }, + { + "epoch": 0.44, + "learning_rate": 1.5588310689566776e-07, + "loss": 0.705, + "step": 12097 + }, + { + "epoch": 0.44, + "learning_rate": 1.5587944483099571e-07, + "loss": 0.675, + "step": 12098 + }, + { + "epoch": 0.44, + "learning_rate": 1.5587578276632364e-07, + "loss": 0.6867, + "step": 12099 + }, + { + "epoch": 0.44, + "learning_rate": 1.558721207016516e-07, + "loss": 0.6959, + "step": 12100 + }, + { + "epoch": 0.44, + "learning_rate": 1.5586845863697952e-07, + "loss": 0.7113, + "step": 12101 + }, + { + "epoch": 0.44, + "learning_rate": 1.5586479657230747e-07, + "loss": 0.7113, + "step": 12102 + }, + { + "epoch": 0.44, + "learning_rate": 1.558611345076354e-07, + "loss": 0.6818, + "step": 12103 + }, + { + "epoch": 0.44, + "learning_rate": 1.5585747244296335e-07, + "loss": 0.7094, + "step": 12104 + }, + { + "epoch": 0.44, + "learning_rate": 1.5585381037829128e-07, + "loss": 0.6796, + "step": 12105 + }, + { + "epoch": 0.44, + "learning_rate": 1.558501483136192e-07, + "loss": 0.6909, + "step": 12106 + }, + { + "epoch": 0.44, + "learning_rate": 1.5584648624894713e-07, + "loss": 0.7036, + "step": 12107 + }, + { + "epoch": 0.44, + "learning_rate": 1.5584282418427509e-07, + "loss": 0.7082, + "step": 12108 + }, + { + "epoch": 0.44, + "learning_rate": 1.55839162119603e-07, + "loss": 0.6855, + "step": 12109 + }, + { + "epoch": 0.44, + "learning_rate": 1.5583550005493097e-07, + "loss": 0.6797, + "step": 12110 + }, + { + "epoch": 0.44, + "learning_rate": 1.558318379902589e-07, + "loss": 0.7008, + "step": 12111 + }, + { + "epoch": 0.44, + "learning_rate": 1.5582817592558685e-07, + "loss": 0.6831, + "step": 12112 + }, + { + "epoch": 0.44, + "learning_rate": 1.5582451386091477e-07, + "loss": 0.6887, + "step": 12113 + }, + { + "epoch": 0.44, + "learning_rate": 1.5582085179624273e-07, + "loss": 0.6813, + "step": 12114 + }, + { + "epoch": 0.44, + "learning_rate": 1.5581718973157065e-07, + "loss": 0.6903, + "step": 12115 + }, + { + "epoch": 0.44, + "learning_rate": 1.5581352766689858e-07, + "loss": 0.7172, + "step": 12116 + }, + { + "epoch": 0.44, + "learning_rate": 1.5580986560222653e-07, + "loss": 0.694, + "step": 12117 + }, + { + "epoch": 0.44, + "learning_rate": 1.5580620353755446e-07, + "loss": 0.6804, + "step": 12118 + }, + { + "epoch": 0.44, + "learning_rate": 1.558025414728824e-07, + "loss": 0.6991, + "step": 12119 + }, + { + "epoch": 0.44, + "learning_rate": 1.5579887940821034e-07, + "loss": 0.6814, + "step": 12120 + }, + { + "epoch": 0.44, + "learning_rate": 1.557952173435383e-07, + "loss": 0.6892, + "step": 12121 + }, + { + "epoch": 0.44, + "learning_rate": 1.5579155527886622e-07, + "loss": 0.6818, + "step": 12122 + }, + { + "epoch": 0.44, + "learning_rate": 1.5578789321419417e-07, + "loss": 0.6999, + "step": 12123 + }, + { + "epoch": 0.44, + "learning_rate": 1.5578423114952207e-07, + "loss": 0.6846, + "step": 12124 + }, + { + "epoch": 0.44, + "learning_rate": 1.5578056908485003e-07, + "loss": 0.6859, + "step": 12125 + }, + { + "epoch": 0.44, + "learning_rate": 1.5577690702017795e-07, + "loss": 0.7048, + "step": 12126 + }, + { + "epoch": 0.44, + "learning_rate": 1.557732449555059e-07, + "loss": 0.6949, + "step": 12127 + }, + { + "epoch": 0.44, + "learning_rate": 1.5576958289083383e-07, + "loss": 0.6887, + "step": 12128 + }, + { + "epoch": 0.44, + "learning_rate": 1.5576592082616179e-07, + "loss": 0.7015, + "step": 12129 + }, + { + "epoch": 0.44, + "learning_rate": 1.557622587614897e-07, + "loss": 0.6861, + "step": 12130 + }, + { + "epoch": 0.44, + "learning_rate": 1.5575859669681766e-07, + "loss": 0.7058, + "step": 12131 + }, + { + "epoch": 0.44, + "learning_rate": 1.557549346321456e-07, + "loss": 0.6973, + "step": 12132 + }, + { + "epoch": 0.44, + "learning_rate": 1.5575127256747354e-07, + "loss": 0.7067, + "step": 12133 + }, + { + "epoch": 0.44, + "learning_rate": 1.5574761050280147e-07, + "loss": 0.6855, + "step": 12134 + }, + { + "epoch": 0.44, + "learning_rate": 1.5574394843812942e-07, + "loss": 0.6979, + "step": 12135 + }, + { + "epoch": 0.44, + "learning_rate": 1.5574028637345735e-07, + "loss": 0.691, + "step": 12136 + }, + { + "epoch": 0.44, + "learning_rate": 1.557366243087853e-07, + "loss": 0.6675, + "step": 12137 + }, + { + "epoch": 0.44, + "learning_rate": 1.5573296224411323e-07, + "loss": 0.6713, + "step": 12138 + }, + { + "epoch": 0.44, + "learning_rate": 1.5572930017944116e-07, + "loss": 0.6637, + "step": 12139 + }, + { + "epoch": 0.44, + "learning_rate": 1.557256381147691e-07, + "loss": 0.7195, + "step": 12140 + }, + { + "epoch": 0.44, + "learning_rate": 1.5572197605009704e-07, + "loss": 0.6886, + "step": 12141 + }, + { + "epoch": 0.44, + "learning_rate": 1.5571831398542496e-07, + "loss": 0.7053, + "step": 12142 + }, + { + "epoch": 0.44, + "learning_rate": 1.557146519207529e-07, + "loss": 0.6819, + "step": 12143 + }, + { + "epoch": 0.44, + "learning_rate": 1.5571098985608084e-07, + "loss": 0.6908, + "step": 12144 + }, + { + "epoch": 0.44, + "learning_rate": 1.5570732779140877e-07, + "loss": 0.6852, + "step": 12145 + }, + { + "epoch": 0.44, + "learning_rate": 1.5570366572673672e-07, + "loss": 0.7046, + "step": 12146 + }, + { + "epoch": 0.44, + "learning_rate": 1.5570000366206465e-07, + "loss": 0.6963, + "step": 12147 + }, + { + "epoch": 0.44, + "learning_rate": 1.556963415973926e-07, + "loss": 0.6885, + "step": 12148 + }, + { + "epoch": 0.44, + "learning_rate": 1.5569267953272053e-07, + "loss": 0.7073, + "step": 12149 + }, + { + "epoch": 0.44, + "learning_rate": 1.5568901746804848e-07, + "loss": 0.6802, + "step": 12150 + }, + { + "epoch": 0.44, + "learning_rate": 1.556853554033764e-07, + "loss": 0.7117, + "step": 12151 + }, + { + "epoch": 0.44, + "learning_rate": 1.5568169333870436e-07, + "loss": 0.7041, + "step": 12152 + }, + { + "epoch": 0.44, + "learning_rate": 1.556780312740323e-07, + "loss": 0.6984, + "step": 12153 + }, + { + "epoch": 0.44, + "learning_rate": 1.5567436920936024e-07, + "loss": 0.6874, + "step": 12154 + }, + { + "epoch": 0.44, + "learning_rate": 1.5567070714468817e-07, + "loss": 0.6807, + "step": 12155 + }, + { + "epoch": 0.44, + "learning_rate": 1.5566704508001612e-07, + "loss": 0.7037, + "step": 12156 + }, + { + "epoch": 0.44, + "learning_rate": 1.5566338301534405e-07, + "loss": 0.7133, + "step": 12157 + }, + { + "epoch": 0.44, + "learning_rate": 1.55659720950672e-07, + "loss": 0.6916, + "step": 12158 + }, + { + "epoch": 0.44, + "learning_rate": 1.5565605888599993e-07, + "loss": 0.6812, + "step": 12159 + }, + { + "epoch": 0.44, + "learning_rate": 1.5565239682132786e-07, + "loss": 0.6803, + "step": 12160 + }, + { + "epoch": 0.44, + "learning_rate": 1.5564873475665578e-07, + "loss": 0.6907, + "step": 12161 + }, + { + "epoch": 0.44, + "learning_rate": 1.5564507269198374e-07, + "loss": 0.706, + "step": 12162 + }, + { + "epoch": 0.45, + "learning_rate": 1.5564141062731166e-07, + "loss": 0.6804, + "step": 12163 + }, + { + "epoch": 0.45, + "learning_rate": 1.5563774856263962e-07, + "loss": 0.6946, + "step": 12164 + }, + { + "epoch": 0.45, + "learning_rate": 1.5563408649796754e-07, + "loss": 0.6835, + "step": 12165 + }, + { + "epoch": 0.45, + "learning_rate": 1.5563042443329547e-07, + "loss": 0.7012, + "step": 12166 + }, + { + "epoch": 0.45, + "learning_rate": 1.5562676236862342e-07, + "loss": 0.6931, + "step": 12167 + }, + { + "epoch": 0.45, + "learning_rate": 1.5562310030395135e-07, + "loss": 0.6945, + "step": 12168 + }, + { + "epoch": 0.45, + "learning_rate": 1.556194382392793e-07, + "loss": 0.6914, + "step": 12169 + }, + { + "epoch": 0.45, + "learning_rate": 1.5561577617460723e-07, + "loss": 0.6993, + "step": 12170 + }, + { + "epoch": 0.45, + "learning_rate": 1.5561211410993518e-07, + "loss": 0.6975, + "step": 12171 + }, + { + "epoch": 0.45, + "learning_rate": 1.556084520452631e-07, + "loss": 0.706, + "step": 12172 + }, + { + "epoch": 0.45, + "learning_rate": 1.5560478998059106e-07, + "loss": 0.6927, + "step": 12173 + }, + { + "epoch": 0.45, + "learning_rate": 1.55601127915919e-07, + "loss": 0.6917, + "step": 12174 + }, + { + "epoch": 0.45, + "learning_rate": 1.5559746585124694e-07, + "loss": 0.681, + "step": 12175 + }, + { + "epoch": 0.45, + "learning_rate": 1.5559380378657487e-07, + "loss": 0.6889, + "step": 12176 + }, + { + "epoch": 0.45, + "learning_rate": 1.5559014172190282e-07, + "loss": 0.7115, + "step": 12177 + }, + { + "epoch": 0.45, + "learning_rate": 1.5558647965723072e-07, + "loss": 0.6906, + "step": 12178 + }, + { + "epoch": 0.45, + "learning_rate": 1.5558281759255868e-07, + "loss": 0.6836, + "step": 12179 + }, + { + "epoch": 0.45, + "learning_rate": 1.555791555278866e-07, + "loss": 0.6982, + "step": 12180 + }, + { + "epoch": 0.45, + "learning_rate": 1.5557549346321456e-07, + "loss": 0.6942, + "step": 12181 + }, + { + "epoch": 0.45, + "learning_rate": 1.5557183139854248e-07, + "loss": 0.6955, + "step": 12182 + }, + { + "epoch": 0.45, + "learning_rate": 1.5556816933387044e-07, + "loss": 0.7057, + "step": 12183 + }, + { + "epoch": 0.45, + "learning_rate": 1.5556450726919836e-07, + "loss": 0.685, + "step": 12184 + }, + { + "epoch": 0.45, + "learning_rate": 1.5556084520452632e-07, + "loss": 0.6997, + "step": 12185 + }, + { + "epoch": 0.45, + "learning_rate": 1.5555718313985424e-07, + "loss": 0.698, + "step": 12186 + }, + { + "epoch": 0.45, + "learning_rate": 1.5555352107518217e-07, + "loss": 0.6789, + "step": 12187 + }, + { + "epoch": 0.45, + "learning_rate": 1.5554985901051012e-07, + "loss": 0.674, + "step": 12188 + }, + { + "epoch": 0.45, + "learning_rate": 1.5554619694583805e-07, + "loss": 0.6998, + "step": 12189 + }, + { + "epoch": 0.45, + "learning_rate": 1.55542534881166e-07, + "loss": 0.6701, + "step": 12190 + }, + { + "epoch": 0.45, + "learning_rate": 1.5553887281649393e-07, + "loss": 0.6856, + "step": 12191 + }, + { + "epoch": 0.45, + "learning_rate": 1.5553521075182188e-07, + "loss": 0.7127, + "step": 12192 + }, + { + "epoch": 0.45, + "learning_rate": 1.555315486871498e-07, + "loss": 0.6938, + "step": 12193 + }, + { + "epoch": 0.45, + "learning_rate": 1.5552788662247776e-07, + "loss": 0.6929, + "step": 12194 + }, + { + "epoch": 0.45, + "learning_rate": 1.555242245578057e-07, + "loss": 0.6882, + "step": 12195 + }, + { + "epoch": 0.45, + "learning_rate": 1.5552056249313361e-07, + "loss": 0.6955, + "step": 12196 + }, + { + "epoch": 0.45, + "learning_rate": 1.5551690042846154e-07, + "loss": 0.6895, + "step": 12197 + }, + { + "epoch": 0.45, + "learning_rate": 1.555132383637895e-07, + "loss": 0.6823, + "step": 12198 + }, + { + "epoch": 0.45, + "learning_rate": 1.5550957629911742e-07, + "loss": 0.7091, + "step": 12199 + }, + { + "epoch": 0.45, + "learning_rate": 1.5550591423444537e-07, + "loss": 0.6782, + "step": 12200 + }, + { + "epoch": 0.45, + "learning_rate": 1.555022521697733e-07, + "loss": 0.6994, + "step": 12201 + }, + { + "epoch": 0.45, + "learning_rate": 1.5549859010510125e-07, + "loss": 0.6787, + "step": 12202 + }, + { + "epoch": 0.45, + "learning_rate": 1.5549492804042918e-07, + "loss": 0.6917, + "step": 12203 + }, + { + "epoch": 0.45, + "learning_rate": 1.5549126597575713e-07, + "loss": 0.7029, + "step": 12204 + }, + { + "epoch": 0.45, + "learning_rate": 1.5548760391108506e-07, + "loss": 0.6921, + "step": 12205 + }, + { + "epoch": 0.45, + "learning_rate": 1.5548394184641301e-07, + "loss": 0.6937, + "step": 12206 + }, + { + "epoch": 0.45, + "learning_rate": 1.5548027978174094e-07, + "loss": 0.6872, + "step": 12207 + }, + { + "epoch": 0.45, + "learning_rate": 1.554766177170689e-07, + "loss": 0.7217, + "step": 12208 + }, + { + "epoch": 0.45, + "learning_rate": 1.5547295565239682e-07, + "loss": 0.6852, + "step": 12209 + }, + { + "epoch": 0.45, + "learning_rate": 1.5546929358772475e-07, + "loss": 0.6989, + "step": 12210 + }, + { + "epoch": 0.45, + "learning_rate": 1.554656315230527e-07, + "loss": 0.7204, + "step": 12211 + }, + { + "epoch": 0.45, + "learning_rate": 1.5546196945838063e-07, + "loss": 0.6874, + "step": 12212 + }, + { + "epoch": 0.45, + "learning_rate": 1.5545830739370858e-07, + "loss": 0.6901, + "step": 12213 + }, + { + "epoch": 0.45, + "learning_rate": 1.5545464532903648e-07, + "loss": 0.6877, + "step": 12214 + }, + { + "epoch": 0.45, + "learning_rate": 1.5545098326436443e-07, + "loss": 0.6863, + "step": 12215 + }, + { + "epoch": 0.45, + "learning_rate": 1.5544732119969236e-07, + "loss": 0.6887, + "step": 12216 + }, + { + "epoch": 0.45, + "learning_rate": 1.5544365913502031e-07, + "loss": 0.7073, + "step": 12217 + }, + { + "epoch": 0.45, + "learning_rate": 1.5543999707034824e-07, + "loss": 0.6797, + "step": 12218 + }, + { + "epoch": 0.45, + "learning_rate": 1.554363350056762e-07, + "loss": 0.7006, + "step": 12219 + }, + { + "epoch": 0.45, + "learning_rate": 1.5543267294100412e-07, + "loss": 0.6826, + "step": 12220 + }, + { + "epoch": 0.45, + "learning_rate": 1.5542901087633207e-07, + "loss": 0.7089, + "step": 12221 + }, + { + "epoch": 0.45, + "learning_rate": 1.5542534881166e-07, + "loss": 0.6778, + "step": 12222 + }, + { + "epoch": 0.45, + "learning_rate": 1.5542168674698795e-07, + "loss": 0.684, + "step": 12223 + }, + { + "epoch": 0.45, + "learning_rate": 1.5541802468231588e-07, + "loss": 0.6758, + "step": 12224 + }, + { + "epoch": 0.45, + "learning_rate": 1.5541436261764383e-07, + "loss": 0.6823, + "step": 12225 + }, + { + "epoch": 0.45, + "learning_rate": 1.5541070055297176e-07, + "loss": 0.6889, + "step": 12226 + }, + { + "epoch": 0.45, + "learning_rate": 1.554070384882997e-07, + "loss": 0.6877, + "step": 12227 + }, + { + "epoch": 0.45, + "learning_rate": 1.5540337642362764e-07, + "loss": 0.6914, + "step": 12228 + }, + { + "epoch": 0.45, + "learning_rate": 1.553997143589556e-07, + "loss": 0.6809, + "step": 12229 + }, + { + "epoch": 0.45, + "learning_rate": 1.5539605229428352e-07, + "loss": 0.6993, + "step": 12230 + }, + { + "epoch": 0.45, + "learning_rate": 1.5539239022961145e-07, + "loss": 0.6952, + "step": 12231 + }, + { + "epoch": 0.45, + "learning_rate": 1.5538872816493937e-07, + "loss": 0.7003, + "step": 12232 + }, + { + "epoch": 0.45, + "learning_rate": 1.5538506610026733e-07, + "loss": 0.7106, + "step": 12233 + }, + { + "epoch": 0.45, + "learning_rate": 1.5538140403559525e-07, + "loss": 0.6813, + "step": 12234 + }, + { + "epoch": 0.45, + "learning_rate": 1.553777419709232e-07, + "loss": 0.6981, + "step": 12235 + }, + { + "epoch": 0.45, + "learning_rate": 1.5537407990625113e-07, + "loss": 0.6956, + "step": 12236 + }, + { + "epoch": 0.45, + "learning_rate": 1.5537041784157906e-07, + "loss": 0.6813, + "step": 12237 + }, + { + "epoch": 0.45, + "learning_rate": 1.55366755776907e-07, + "loss": 0.6766, + "step": 12238 + }, + { + "epoch": 0.45, + "learning_rate": 1.5536309371223494e-07, + "loss": 0.6706, + "step": 12239 + }, + { + "epoch": 0.45, + "learning_rate": 1.553594316475629e-07, + "loss": 0.7014, + "step": 12240 + }, + { + "epoch": 0.45, + "learning_rate": 1.5535576958289082e-07, + "loss": 0.6949, + "step": 12241 + }, + { + "epoch": 0.45, + "learning_rate": 1.5535210751821877e-07, + "loss": 0.7004, + "step": 12242 + }, + { + "epoch": 0.45, + "learning_rate": 1.553484454535467e-07, + "loss": 0.6959, + "step": 12243 + }, + { + "epoch": 0.45, + "learning_rate": 1.5534478338887465e-07, + "loss": 0.6721, + "step": 12244 + }, + { + "epoch": 0.45, + "learning_rate": 1.5534112132420258e-07, + "loss": 0.6885, + "step": 12245 + }, + { + "epoch": 0.45, + "learning_rate": 1.5533745925953053e-07, + "loss": 0.7089, + "step": 12246 + }, + { + "epoch": 0.45, + "learning_rate": 1.5533379719485846e-07, + "loss": 0.7086, + "step": 12247 + }, + { + "epoch": 0.45, + "learning_rate": 1.553301351301864e-07, + "loss": 0.7009, + "step": 12248 + }, + { + "epoch": 0.45, + "learning_rate": 1.5532647306551434e-07, + "loss": 0.6928, + "step": 12249 + }, + { + "epoch": 0.45, + "learning_rate": 1.5532281100084226e-07, + "loss": 0.6897, + "step": 12250 + }, + { + "epoch": 0.45, + "learning_rate": 1.553191489361702e-07, + "loss": 0.7071, + "step": 12251 + }, + { + "epoch": 0.45, + "learning_rate": 1.5531548687149814e-07, + "loss": 0.6949, + "step": 12252 + }, + { + "epoch": 0.45, + "learning_rate": 1.5531182480682607e-07, + "loss": 0.7092, + "step": 12253 + }, + { + "epoch": 0.45, + "learning_rate": 1.5530816274215402e-07, + "loss": 0.6984, + "step": 12254 + }, + { + "epoch": 0.45, + "learning_rate": 1.5530450067748195e-07, + "loss": 0.6888, + "step": 12255 + }, + { + "epoch": 0.45, + "learning_rate": 1.553008386128099e-07, + "loss": 0.6866, + "step": 12256 + }, + { + "epoch": 0.45, + "learning_rate": 1.5529717654813783e-07, + "loss": 0.6886, + "step": 12257 + }, + { + "epoch": 0.45, + "learning_rate": 1.5529351448346576e-07, + "loss": 0.7235, + "step": 12258 + }, + { + "epoch": 0.45, + "learning_rate": 1.552898524187937e-07, + "loss": 0.7085, + "step": 12259 + }, + { + "epoch": 0.45, + "learning_rate": 1.5528619035412164e-07, + "loss": 0.6919, + "step": 12260 + }, + { + "epoch": 0.45, + "learning_rate": 1.552825282894496e-07, + "loss": 0.6936, + "step": 12261 + }, + { + "epoch": 0.45, + "learning_rate": 1.5527886622477752e-07, + "loss": 0.7042, + "step": 12262 + }, + { + "epoch": 0.45, + "learning_rate": 1.5527520416010547e-07, + "loss": 0.6986, + "step": 12263 + }, + { + "epoch": 0.45, + "learning_rate": 1.552715420954334e-07, + "loss": 0.6797, + "step": 12264 + }, + { + "epoch": 0.45, + "learning_rate": 1.5526788003076135e-07, + "loss": 0.6843, + "step": 12265 + }, + { + "epoch": 0.45, + "learning_rate": 1.5526421796608928e-07, + "loss": 0.6882, + "step": 12266 + }, + { + "epoch": 0.45, + "learning_rate": 1.5526055590141723e-07, + "loss": 0.6981, + "step": 12267 + }, + { + "epoch": 0.45, + "learning_rate": 1.5525689383674513e-07, + "loss": 0.7041, + "step": 12268 + }, + { + "epoch": 0.45, + "learning_rate": 1.5525323177207308e-07, + "loss": 0.6796, + "step": 12269 + }, + { + "epoch": 0.45, + "learning_rate": 1.55249569707401e-07, + "loss": 0.6877, + "step": 12270 + }, + { + "epoch": 0.45, + "learning_rate": 1.5524590764272896e-07, + "loss": 0.6981, + "step": 12271 + }, + { + "epoch": 0.45, + "learning_rate": 1.552422455780569e-07, + "loss": 0.6951, + "step": 12272 + }, + { + "epoch": 0.45, + "learning_rate": 1.5523858351338484e-07, + "loss": 0.7076, + "step": 12273 + }, + { + "epoch": 0.45, + "learning_rate": 1.5523492144871277e-07, + "loss": 0.6964, + "step": 12274 + }, + { + "epoch": 0.45, + "learning_rate": 1.5523125938404072e-07, + "loss": 0.6966, + "step": 12275 + }, + { + "epoch": 0.45, + "learning_rate": 1.5522759731936865e-07, + "loss": 0.6869, + "step": 12276 + }, + { + "epoch": 0.45, + "learning_rate": 1.552239352546966e-07, + "loss": 0.6708, + "step": 12277 + }, + { + "epoch": 0.45, + "learning_rate": 1.5522027319002453e-07, + "loss": 0.686, + "step": 12278 + }, + { + "epoch": 0.45, + "learning_rate": 1.5521661112535248e-07, + "loss": 0.6921, + "step": 12279 + }, + { + "epoch": 0.45, + "learning_rate": 1.552129490606804e-07, + "loss": 0.6917, + "step": 12280 + }, + { + "epoch": 0.45, + "learning_rate": 1.5520928699600834e-07, + "loss": 0.6839, + "step": 12281 + }, + { + "epoch": 0.45, + "learning_rate": 1.552056249313363e-07, + "loss": 0.6889, + "step": 12282 + }, + { + "epoch": 0.45, + "learning_rate": 1.5520196286666422e-07, + "loss": 0.6921, + "step": 12283 + }, + { + "epoch": 0.45, + "learning_rate": 1.5519830080199217e-07, + "loss": 0.6902, + "step": 12284 + }, + { + "epoch": 0.45, + "learning_rate": 1.551946387373201e-07, + "loss": 0.6746, + "step": 12285 + }, + { + "epoch": 0.45, + "learning_rate": 1.5519097667264802e-07, + "loss": 0.6966, + "step": 12286 + }, + { + "epoch": 0.45, + "learning_rate": 1.5518731460797595e-07, + "loss": 0.7081, + "step": 12287 + }, + { + "epoch": 0.45, + "learning_rate": 1.551836525433039e-07, + "loss": 0.6942, + "step": 12288 + }, + { + "epoch": 0.45, + "learning_rate": 1.5517999047863183e-07, + "loss": 0.6951, + "step": 12289 + }, + { + "epoch": 0.45, + "learning_rate": 1.5517632841395978e-07, + "loss": 0.6997, + "step": 12290 + }, + { + "epoch": 0.45, + "learning_rate": 1.551726663492877e-07, + "loss": 0.6886, + "step": 12291 + }, + { + "epoch": 0.45, + "learning_rate": 1.5516900428461566e-07, + "loss": 0.6842, + "step": 12292 + }, + { + "epoch": 0.45, + "learning_rate": 1.551653422199436e-07, + "loss": 0.7047, + "step": 12293 + }, + { + "epoch": 0.45, + "learning_rate": 1.5516168015527154e-07, + "loss": 0.7003, + "step": 12294 + }, + { + "epoch": 0.45, + "learning_rate": 1.5515801809059947e-07, + "loss": 0.6919, + "step": 12295 + }, + { + "epoch": 0.45, + "learning_rate": 1.5515435602592742e-07, + "loss": 0.6944, + "step": 12296 + }, + { + "epoch": 0.45, + "learning_rate": 1.5515069396125535e-07, + "loss": 0.7, + "step": 12297 + }, + { + "epoch": 0.45, + "learning_rate": 1.551470318965833e-07, + "loss": 0.6867, + "step": 12298 + }, + { + "epoch": 0.45, + "learning_rate": 1.5514336983191123e-07, + "loss": 0.708, + "step": 12299 + }, + { + "epoch": 0.45, + "learning_rate": 1.5513970776723918e-07, + "loss": 0.6932, + "step": 12300 + }, + { + "epoch": 0.45, + "learning_rate": 1.551360457025671e-07, + "loss": 0.7024, + "step": 12301 + }, + { + "epoch": 0.45, + "learning_rate": 1.5513238363789504e-07, + "loss": 0.7055, + "step": 12302 + }, + { + "epoch": 0.45, + "learning_rate": 1.55128721573223e-07, + "loss": 0.6818, + "step": 12303 + }, + { + "epoch": 0.45, + "learning_rate": 1.5512505950855092e-07, + "loss": 0.6972, + "step": 12304 + }, + { + "epoch": 0.45, + "learning_rate": 1.5512139744387884e-07, + "loss": 0.7011, + "step": 12305 + }, + { + "epoch": 0.45, + "learning_rate": 1.551177353792068e-07, + "loss": 0.6949, + "step": 12306 + }, + { + "epoch": 0.45, + "learning_rate": 1.5511407331453472e-07, + "loss": 0.6972, + "step": 12307 + }, + { + "epoch": 0.45, + "learning_rate": 1.5511041124986265e-07, + "loss": 0.6913, + "step": 12308 + }, + { + "epoch": 0.45, + "learning_rate": 1.551067491851906e-07, + "loss": 0.6987, + "step": 12309 + }, + { + "epoch": 0.45, + "learning_rate": 1.5510308712051853e-07, + "loss": 0.698, + "step": 12310 + }, + { + "epoch": 0.45, + "learning_rate": 1.5509942505584648e-07, + "loss": 0.6751, + "step": 12311 + }, + { + "epoch": 0.45, + "learning_rate": 1.550957629911744e-07, + "loss": 0.6971, + "step": 12312 + }, + { + "epoch": 0.45, + "learning_rate": 1.5509210092650236e-07, + "loss": 0.6936, + "step": 12313 + }, + { + "epoch": 0.45, + "learning_rate": 1.550884388618303e-07, + "loss": 0.6904, + "step": 12314 + }, + { + "epoch": 0.45, + "learning_rate": 1.5508477679715824e-07, + "loss": 0.693, + "step": 12315 + }, + { + "epoch": 0.45, + "learning_rate": 1.5508111473248617e-07, + "loss": 0.695, + "step": 12316 + }, + { + "epoch": 0.45, + "learning_rate": 1.5507745266781412e-07, + "loss": 0.6874, + "step": 12317 + }, + { + "epoch": 0.45, + "learning_rate": 1.5507379060314205e-07, + "loss": 0.6897, + "step": 12318 + }, + { + "epoch": 0.45, + "learning_rate": 1.5507012853847e-07, + "loss": 0.717, + "step": 12319 + }, + { + "epoch": 0.45, + "learning_rate": 1.5506646647379793e-07, + "loss": 0.6938, + "step": 12320 + }, + { + "epoch": 0.45, + "learning_rate": 1.5506280440912588e-07, + "loss": 0.7016, + "step": 12321 + }, + { + "epoch": 0.45, + "learning_rate": 1.5505914234445378e-07, + "loss": 0.7056, + "step": 12322 + }, + { + "epoch": 0.45, + "learning_rate": 1.5505548027978173e-07, + "loss": 0.6897, + "step": 12323 + }, + { + "epoch": 0.45, + "learning_rate": 1.5505181821510966e-07, + "loss": 0.6862, + "step": 12324 + }, + { + "epoch": 0.45, + "learning_rate": 1.5504815615043761e-07, + "loss": 0.6784, + "step": 12325 + }, + { + "epoch": 0.45, + "learning_rate": 1.5504449408576554e-07, + "loss": 0.6969, + "step": 12326 + }, + { + "epoch": 0.45, + "learning_rate": 1.550408320210935e-07, + "loss": 0.6802, + "step": 12327 + }, + { + "epoch": 0.45, + "learning_rate": 1.5503716995642142e-07, + "loss": 0.6995, + "step": 12328 + }, + { + "epoch": 0.45, + "learning_rate": 1.5503350789174935e-07, + "loss": 0.6944, + "step": 12329 + }, + { + "epoch": 0.45, + "learning_rate": 1.550298458270773e-07, + "loss": 0.68, + "step": 12330 + }, + { + "epoch": 0.45, + "learning_rate": 1.5502618376240523e-07, + "loss": 0.6885, + "step": 12331 + }, + { + "epoch": 0.45, + "learning_rate": 1.5502252169773318e-07, + "loss": 0.7097, + "step": 12332 + }, + { + "epoch": 0.45, + "learning_rate": 1.550188596330611e-07, + "loss": 0.69, + "step": 12333 + }, + { + "epoch": 0.45, + "learning_rate": 1.5501519756838906e-07, + "loss": 0.6983, + "step": 12334 + }, + { + "epoch": 0.45, + "learning_rate": 1.55011535503717e-07, + "loss": 0.7159, + "step": 12335 + }, + { + "epoch": 0.45, + "learning_rate": 1.5500787343904494e-07, + "loss": 0.7038, + "step": 12336 + }, + { + "epoch": 0.45, + "learning_rate": 1.5500421137437287e-07, + "loss": 0.6829, + "step": 12337 + }, + { + "epoch": 0.45, + "learning_rate": 1.5500054930970082e-07, + "loss": 0.6958, + "step": 12338 + }, + { + "epoch": 0.45, + "learning_rate": 1.5499688724502875e-07, + "loss": 0.6944, + "step": 12339 + }, + { + "epoch": 0.45, + "learning_rate": 1.5499322518035667e-07, + "loss": 0.7019, + "step": 12340 + }, + { + "epoch": 0.45, + "learning_rate": 1.549895631156846e-07, + "loss": 0.6679, + "step": 12341 + }, + { + "epoch": 0.45, + "learning_rate": 1.5498590105101255e-07, + "loss": 0.7094, + "step": 12342 + }, + { + "epoch": 0.45, + "learning_rate": 1.5498223898634048e-07, + "loss": 0.6967, + "step": 12343 + }, + { + "epoch": 0.45, + "learning_rate": 1.5497857692166843e-07, + "loss": 0.6783, + "step": 12344 + }, + { + "epoch": 0.45, + "learning_rate": 1.5497491485699636e-07, + "loss": 0.6921, + "step": 12345 + }, + { + "epoch": 0.45, + "learning_rate": 1.549712527923243e-07, + "loss": 0.6814, + "step": 12346 + }, + { + "epoch": 0.45, + "learning_rate": 1.5496759072765224e-07, + "loss": 0.6847, + "step": 12347 + }, + { + "epoch": 0.45, + "learning_rate": 1.549639286629802e-07, + "loss": 0.6797, + "step": 12348 + }, + { + "epoch": 0.45, + "learning_rate": 1.5496026659830812e-07, + "loss": 0.7122, + "step": 12349 + }, + { + "epoch": 0.45, + "learning_rate": 1.5495660453363607e-07, + "loss": 0.6823, + "step": 12350 + }, + { + "epoch": 0.45, + "learning_rate": 1.54952942468964e-07, + "loss": 0.688, + "step": 12351 + }, + { + "epoch": 0.45, + "learning_rate": 1.5494928040429193e-07, + "loss": 0.6927, + "step": 12352 + }, + { + "epoch": 0.45, + "learning_rate": 1.5494561833961988e-07, + "loss": 0.7007, + "step": 12353 + }, + { + "epoch": 0.45, + "learning_rate": 1.549419562749478e-07, + "loss": 0.6799, + "step": 12354 + }, + { + "epoch": 0.45, + "learning_rate": 1.5493829421027576e-07, + "loss": 0.6968, + "step": 12355 + }, + { + "epoch": 0.45, + "learning_rate": 1.5493463214560369e-07, + "loss": 0.6798, + "step": 12356 + }, + { + "epoch": 0.45, + "learning_rate": 1.5493097008093164e-07, + "loss": 0.7005, + "step": 12357 + }, + { + "epoch": 0.45, + "learning_rate": 1.5492730801625954e-07, + "loss": 0.7067, + "step": 12358 + }, + { + "epoch": 0.45, + "learning_rate": 1.549236459515875e-07, + "loss": 0.6964, + "step": 12359 + }, + { + "epoch": 0.45, + "learning_rate": 1.5491998388691542e-07, + "loss": 0.6996, + "step": 12360 + }, + { + "epoch": 0.45, + "learning_rate": 1.5491632182224337e-07, + "loss": 0.7045, + "step": 12361 + }, + { + "epoch": 0.45, + "learning_rate": 1.549126597575713e-07, + "loss": 0.6953, + "step": 12362 + }, + { + "epoch": 0.45, + "learning_rate": 1.5490899769289925e-07, + "loss": 0.6974, + "step": 12363 + }, + { + "epoch": 0.45, + "learning_rate": 1.5490533562822718e-07, + "loss": 0.6964, + "step": 12364 + }, + { + "epoch": 0.45, + "learning_rate": 1.5490167356355513e-07, + "loss": 0.678, + "step": 12365 + }, + { + "epoch": 0.45, + "learning_rate": 1.5489801149888306e-07, + "loss": 0.7024, + "step": 12366 + }, + { + "epoch": 0.45, + "learning_rate": 1.54894349434211e-07, + "loss": 0.6891, + "step": 12367 + }, + { + "epoch": 0.45, + "learning_rate": 1.5489068736953894e-07, + "loss": 0.6801, + "step": 12368 + }, + { + "epoch": 0.45, + "learning_rate": 1.548870253048669e-07, + "loss": 0.6829, + "step": 12369 + }, + { + "epoch": 0.45, + "learning_rate": 1.5488336324019482e-07, + "loss": 0.6913, + "step": 12370 + }, + { + "epoch": 0.45, + "learning_rate": 1.5487970117552277e-07, + "loss": 0.6853, + "step": 12371 + }, + { + "epoch": 0.45, + "learning_rate": 1.548760391108507e-07, + "loss": 0.6884, + "step": 12372 + }, + { + "epoch": 0.45, + "learning_rate": 1.5487237704617862e-07, + "loss": 0.7, + "step": 12373 + }, + { + "epoch": 0.45, + "learning_rate": 1.5486871498150658e-07, + "loss": 0.7054, + "step": 12374 + }, + { + "epoch": 0.45, + "learning_rate": 1.548650529168345e-07, + "loss": 0.6863, + "step": 12375 + }, + { + "epoch": 0.45, + "learning_rate": 1.5486139085216243e-07, + "loss": 0.6906, + "step": 12376 + }, + { + "epoch": 0.45, + "learning_rate": 1.5485772878749038e-07, + "loss": 0.6948, + "step": 12377 + }, + { + "epoch": 0.45, + "learning_rate": 1.548540667228183e-07, + "loss": 0.7048, + "step": 12378 + }, + { + "epoch": 0.45, + "learning_rate": 1.5485040465814624e-07, + "loss": 0.6836, + "step": 12379 + }, + { + "epoch": 0.45, + "learning_rate": 1.548467425934742e-07, + "loss": 0.7081, + "step": 12380 + }, + { + "epoch": 0.45, + "learning_rate": 1.5484308052880212e-07, + "loss": 0.6789, + "step": 12381 + }, + { + "epoch": 0.45, + "learning_rate": 1.5483941846413007e-07, + "loss": 0.7187, + "step": 12382 + }, + { + "epoch": 0.45, + "learning_rate": 1.54835756399458e-07, + "loss": 0.687, + "step": 12383 + }, + { + "epoch": 0.45, + "learning_rate": 1.5483209433478595e-07, + "loss": 0.7026, + "step": 12384 + }, + { + "epoch": 0.45, + "learning_rate": 1.5482843227011388e-07, + "loss": 0.6967, + "step": 12385 + }, + { + "epoch": 0.45, + "learning_rate": 1.5482477020544183e-07, + "loss": 0.6792, + "step": 12386 + }, + { + "epoch": 0.45, + "learning_rate": 1.5482110814076976e-07, + "loss": 0.6961, + "step": 12387 + }, + { + "epoch": 0.45, + "learning_rate": 1.548174460760977e-07, + "loss": 0.6848, + "step": 12388 + }, + { + "epoch": 0.45, + "learning_rate": 1.5481378401142564e-07, + "loss": 0.6829, + "step": 12389 + }, + { + "epoch": 0.45, + "learning_rate": 1.548101219467536e-07, + "loss": 0.7015, + "step": 12390 + }, + { + "epoch": 0.45, + "learning_rate": 1.5480645988208152e-07, + "loss": 0.6871, + "step": 12391 + }, + { + "epoch": 0.45, + "learning_rate": 1.5480279781740947e-07, + "loss": 0.6974, + "step": 12392 + }, + { + "epoch": 0.45, + "learning_rate": 1.5479913575273737e-07, + "loss": 0.7059, + "step": 12393 + }, + { + "epoch": 0.45, + "learning_rate": 1.5479547368806532e-07, + "loss": 0.6899, + "step": 12394 + }, + { + "epoch": 0.45, + "learning_rate": 1.5479181162339325e-07, + "loss": 0.6993, + "step": 12395 + }, + { + "epoch": 0.45, + "learning_rate": 1.547881495587212e-07, + "loss": 0.735, + "step": 12396 + }, + { + "epoch": 0.45, + "learning_rate": 1.5478448749404913e-07, + "loss": 0.7038, + "step": 12397 + }, + { + "epoch": 0.45, + "learning_rate": 1.5478082542937708e-07, + "loss": 0.6946, + "step": 12398 + }, + { + "epoch": 0.45, + "learning_rate": 1.54777163364705e-07, + "loss": 0.6975, + "step": 12399 + }, + { + "epoch": 0.45, + "learning_rate": 1.5477350130003294e-07, + "loss": 0.7061, + "step": 12400 + }, + { + "epoch": 0.45, + "learning_rate": 1.547698392353609e-07, + "loss": 0.6794, + "step": 12401 + }, + { + "epoch": 0.45, + "learning_rate": 1.5476617717068882e-07, + "loss": 0.705, + "step": 12402 + }, + { + "epoch": 0.45, + "learning_rate": 1.5476251510601677e-07, + "loss": 0.6816, + "step": 12403 + }, + { + "epoch": 0.45, + "learning_rate": 1.547588530413447e-07, + "loss": 0.7097, + "step": 12404 + }, + { + "epoch": 0.45, + "learning_rate": 1.5475519097667265e-07, + "loss": 0.6982, + "step": 12405 + }, + { + "epoch": 0.45, + "learning_rate": 1.5475152891200058e-07, + "loss": 0.672, + "step": 12406 + }, + { + "epoch": 0.45, + "learning_rate": 1.5474786684732853e-07, + "loss": 0.692, + "step": 12407 + }, + { + "epoch": 0.45, + "learning_rate": 1.5474420478265646e-07, + "loss": 0.6996, + "step": 12408 + }, + { + "epoch": 0.45, + "learning_rate": 1.547405427179844e-07, + "loss": 0.7039, + "step": 12409 + }, + { + "epoch": 0.45, + "learning_rate": 1.5473688065331234e-07, + "loss": 0.6954, + "step": 12410 + }, + { + "epoch": 0.45, + "learning_rate": 1.5473321858864026e-07, + "loss": 0.6719, + "step": 12411 + }, + { + "epoch": 0.45, + "learning_rate": 1.547295565239682e-07, + "loss": 0.678, + "step": 12412 + }, + { + "epoch": 0.45, + "learning_rate": 1.5472589445929614e-07, + "loss": 0.6854, + "step": 12413 + }, + { + "epoch": 0.45, + "learning_rate": 1.5472223239462407e-07, + "loss": 0.6882, + "step": 12414 + }, + { + "epoch": 0.45, + "learning_rate": 1.5471857032995202e-07, + "loss": 0.6898, + "step": 12415 + }, + { + "epoch": 0.45, + "learning_rate": 1.5471490826527995e-07, + "loss": 0.6939, + "step": 12416 + }, + { + "epoch": 0.45, + "learning_rate": 1.547112462006079e-07, + "loss": 0.6779, + "step": 12417 + }, + { + "epoch": 0.45, + "learning_rate": 1.5470758413593583e-07, + "loss": 0.6814, + "step": 12418 + }, + { + "epoch": 0.45, + "learning_rate": 1.5470392207126378e-07, + "loss": 0.7072, + "step": 12419 + }, + { + "epoch": 0.45, + "learning_rate": 1.547002600065917e-07, + "loss": 0.6908, + "step": 12420 + }, + { + "epoch": 0.45, + "learning_rate": 1.5469659794191966e-07, + "loss": 0.6889, + "step": 12421 + }, + { + "epoch": 0.45, + "learning_rate": 1.546929358772476e-07, + "loss": 0.6797, + "step": 12422 + }, + { + "epoch": 0.45, + "learning_rate": 1.5468927381257552e-07, + "loss": 0.6865, + "step": 12423 + }, + { + "epoch": 0.45, + "learning_rate": 1.5468561174790347e-07, + "loss": 0.6984, + "step": 12424 + }, + { + "epoch": 0.45, + "learning_rate": 1.546819496832314e-07, + "loss": 0.6955, + "step": 12425 + }, + { + "epoch": 0.45, + "learning_rate": 1.5467828761855935e-07, + "loss": 0.7168, + "step": 12426 + }, + { + "epoch": 0.45, + "learning_rate": 1.5467462555388727e-07, + "loss": 0.6722, + "step": 12427 + }, + { + "epoch": 0.45, + "learning_rate": 1.5467096348921523e-07, + "loss": 0.6908, + "step": 12428 + }, + { + "epoch": 0.45, + "learning_rate": 1.5466730142454313e-07, + "loss": 0.6919, + "step": 12429 + }, + { + "epoch": 0.45, + "learning_rate": 1.5466363935987108e-07, + "loss": 0.6955, + "step": 12430 + }, + { + "epoch": 0.45, + "learning_rate": 1.54659977295199e-07, + "loss": 0.6834, + "step": 12431 + }, + { + "epoch": 0.45, + "learning_rate": 1.5465631523052696e-07, + "loss": 0.6969, + "step": 12432 + }, + { + "epoch": 0.45, + "learning_rate": 1.546526531658549e-07, + "loss": 0.6861, + "step": 12433 + }, + { + "epoch": 0.45, + "learning_rate": 1.5464899110118284e-07, + "loss": 0.7004, + "step": 12434 + }, + { + "epoch": 0.45, + "learning_rate": 1.5464532903651077e-07, + "loss": 0.6956, + "step": 12435 + }, + { + "epoch": 0.45, + "learning_rate": 1.5464166697183872e-07, + "loss": 0.6855, + "step": 12436 + }, + { + "epoch": 0.46, + "learning_rate": 1.5463800490716665e-07, + "loss": 0.6976, + "step": 12437 + }, + { + "epoch": 0.46, + "learning_rate": 1.546343428424946e-07, + "loss": 0.7024, + "step": 12438 + }, + { + "epoch": 0.46, + "learning_rate": 1.5463068077782253e-07, + "loss": 0.6897, + "step": 12439 + }, + { + "epoch": 0.46, + "learning_rate": 1.5462701871315048e-07, + "loss": 0.6984, + "step": 12440 + }, + { + "epoch": 0.46, + "learning_rate": 1.546233566484784e-07, + "loss": 0.7022, + "step": 12441 + }, + { + "epoch": 0.46, + "learning_rate": 1.5461969458380636e-07, + "loss": 0.7071, + "step": 12442 + }, + { + "epoch": 0.46, + "learning_rate": 1.546160325191343e-07, + "loss": 0.6858, + "step": 12443 + }, + { + "epoch": 0.46, + "learning_rate": 1.5461237045446221e-07, + "loss": 0.7017, + "step": 12444 + }, + { + "epoch": 0.46, + "learning_rate": 1.5460870838979017e-07, + "loss": 0.6795, + "step": 12445 + }, + { + "epoch": 0.46, + "learning_rate": 1.546050463251181e-07, + "loss": 0.6979, + "step": 12446 + }, + { + "epoch": 0.46, + "learning_rate": 1.5460138426044602e-07, + "loss": 0.6855, + "step": 12447 + }, + { + "epoch": 0.46, + "learning_rate": 1.5459772219577397e-07, + "loss": 0.7075, + "step": 12448 + }, + { + "epoch": 0.46, + "learning_rate": 1.545940601311019e-07, + "loss": 0.6936, + "step": 12449 + }, + { + "epoch": 0.46, + "learning_rate": 1.5459039806642983e-07, + "loss": 0.6913, + "step": 12450 + }, + { + "epoch": 0.46, + "learning_rate": 1.5458673600175778e-07, + "loss": 0.6878, + "step": 12451 + }, + { + "epoch": 0.46, + "learning_rate": 1.545830739370857e-07, + "loss": 0.6947, + "step": 12452 + }, + { + "epoch": 0.46, + "learning_rate": 1.5457941187241366e-07, + "loss": 0.6776, + "step": 12453 + }, + { + "epoch": 0.46, + "learning_rate": 1.545757498077416e-07, + "loss": 0.7066, + "step": 12454 + }, + { + "epoch": 0.46, + "learning_rate": 1.5457208774306954e-07, + "loss": 0.6997, + "step": 12455 + }, + { + "epoch": 0.46, + "learning_rate": 1.5456842567839747e-07, + "loss": 0.7055, + "step": 12456 + }, + { + "epoch": 0.46, + "learning_rate": 1.5456476361372542e-07, + "loss": 0.706, + "step": 12457 + }, + { + "epoch": 0.46, + "learning_rate": 1.5456110154905335e-07, + "loss": 0.7046, + "step": 12458 + }, + { + "epoch": 0.46, + "learning_rate": 1.545574394843813e-07, + "loss": 0.6843, + "step": 12459 + }, + { + "epoch": 0.46, + "learning_rate": 1.5455377741970923e-07, + "loss": 0.6911, + "step": 12460 + }, + { + "epoch": 0.46, + "learning_rate": 1.5455011535503718e-07, + "loss": 0.691, + "step": 12461 + }, + { + "epoch": 0.46, + "learning_rate": 1.545464532903651e-07, + "loss": 0.7136, + "step": 12462 + }, + { + "epoch": 0.46, + "learning_rate": 1.5454279122569306e-07, + "loss": 0.6921, + "step": 12463 + }, + { + "epoch": 0.46, + "learning_rate": 1.5453912916102099e-07, + "loss": 0.6934, + "step": 12464 + }, + { + "epoch": 0.46, + "learning_rate": 1.545354670963489e-07, + "loss": 0.7001, + "step": 12465 + }, + { + "epoch": 0.46, + "learning_rate": 1.5453180503167684e-07, + "loss": 0.7077, + "step": 12466 + }, + { + "epoch": 0.46, + "learning_rate": 1.545281429670048e-07, + "loss": 0.6959, + "step": 12467 + }, + { + "epoch": 0.46, + "learning_rate": 1.5452448090233272e-07, + "loss": 0.6896, + "step": 12468 + }, + { + "epoch": 0.46, + "learning_rate": 1.5452081883766067e-07, + "loss": 0.6995, + "step": 12469 + }, + { + "epoch": 0.46, + "learning_rate": 1.545171567729886e-07, + "loss": 0.6944, + "step": 12470 + }, + { + "epoch": 0.46, + "learning_rate": 1.5451349470831653e-07, + "loss": 0.6991, + "step": 12471 + }, + { + "epoch": 0.46, + "learning_rate": 1.5450983264364448e-07, + "loss": 0.6921, + "step": 12472 + }, + { + "epoch": 0.46, + "learning_rate": 1.545061705789724e-07, + "loss": 0.6871, + "step": 12473 + }, + { + "epoch": 0.46, + "learning_rate": 1.5450250851430036e-07, + "loss": 0.6925, + "step": 12474 + }, + { + "epoch": 0.46, + "learning_rate": 1.5449884644962829e-07, + "loss": 0.6917, + "step": 12475 + }, + { + "epoch": 0.46, + "learning_rate": 1.5449518438495624e-07, + "loss": 0.7031, + "step": 12476 + }, + { + "epoch": 0.46, + "learning_rate": 1.5449152232028417e-07, + "loss": 0.6814, + "step": 12477 + }, + { + "epoch": 0.46, + "learning_rate": 1.5448786025561212e-07, + "loss": 0.7011, + "step": 12478 + }, + { + "epoch": 0.46, + "learning_rate": 1.5448419819094005e-07, + "loss": 0.6874, + "step": 12479 + }, + { + "epoch": 0.46, + "learning_rate": 1.54480536126268e-07, + "loss": 0.7196, + "step": 12480 + }, + { + "epoch": 0.46, + "learning_rate": 1.5447687406159593e-07, + "loss": 0.6901, + "step": 12481 + }, + { + "epoch": 0.46, + "learning_rate": 1.5447321199692388e-07, + "loss": 0.7159, + "step": 12482 + }, + { + "epoch": 0.46, + "learning_rate": 1.5446954993225178e-07, + "loss": 0.6825, + "step": 12483 + }, + { + "epoch": 0.46, + "learning_rate": 1.5446588786757973e-07, + "loss": 0.6859, + "step": 12484 + }, + { + "epoch": 0.46, + "learning_rate": 1.5446222580290766e-07, + "loss": 0.6813, + "step": 12485 + }, + { + "epoch": 0.46, + "learning_rate": 1.544585637382356e-07, + "loss": 0.6854, + "step": 12486 + }, + { + "epoch": 0.46, + "learning_rate": 1.5445490167356354e-07, + "loss": 0.6956, + "step": 12487 + }, + { + "epoch": 0.46, + "learning_rate": 1.544512396088915e-07, + "loss": 0.6838, + "step": 12488 + }, + { + "epoch": 0.46, + "learning_rate": 1.5444757754421942e-07, + "loss": 0.6809, + "step": 12489 + }, + { + "epoch": 0.46, + "learning_rate": 1.5444391547954737e-07, + "loss": 0.6891, + "step": 12490 + }, + { + "epoch": 0.46, + "learning_rate": 1.544402534148753e-07, + "loss": 0.6693, + "step": 12491 + }, + { + "epoch": 0.46, + "learning_rate": 1.5443659135020325e-07, + "loss": 0.6888, + "step": 12492 + }, + { + "epoch": 0.46, + "learning_rate": 1.5443292928553118e-07, + "loss": 0.6842, + "step": 12493 + }, + { + "epoch": 0.46, + "learning_rate": 1.544292672208591e-07, + "loss": 0.6762, + "step": 12494 + }, + { + "epoch": 0.46, + "learning_rate": 1.5442560515618706e-07, + "loss": 0.6933, + "step": 12495 + }, + { + "epoch": 0.46, + "learning_rate": 1.5442194309151498e-07, + "loss": 0.6992, + "step": 12496 + }, + { + "epoch": 0.46, + "learning_rate": 1.5441828102684294e-07, + "loss": 0.6965, + "step": 12497 + }, + { + "epoch": 0.46, + "learning_rate": 1.5441461896217086e-07, + "loss": 0.6777, + "step": 12498 + }, + { + "epoch": 0.46, + "learning_rate": 1.5441095689749882e-07, + "loss": 0.6867, + "step": 12499 + }, + { + "epoch": 0.46, + "learning_rate": 1.5440729483282674e-07, + "loss": 0.6966, + "step": 12500 + }, + { + "epoch": 0.46, + "learning_rate": 1.5440363276815467e-07, + "loss": 0.726, + "step": 12501 + }, + { + "epoch": 0.46, + "learning_rate": 1.543999707034826e-07, + "loss": 0.6736, + "step": 12502 + }, + { + "epoch": 0.46, + "learning_rate": 1.5439630863881055e-07, + "loss": 0.7149, + "step": 12503 + }, + { + "epoch": 0.46, + "learning_rate": 1.5439264657413848e-07, + "loss": 0.6872, + "step": 12504 + }, + { + "epoch": 0.46, + "learning_rate": 1.5438898450946643e-07, + "loss": 0.6835, + "step": 12505 + }, + { + "epoch": 0.46, + "learning_rate": 1.5438532244479436e-07, + "loss": 0.6994, + "step": 12506 + }, + { + "epoch": 0.46, + "learning_rate": 1.543816603801223e-07, + "loss": 0.7011, + "step": 12507 + }, + { + "epoch": 0.46, + "learning_rate": 1.5437799831545024e-07, + "loss": 0.6905, + "step": 12508 + }, + { + "epoch": 0.46, + "learning_rate": 1.543743362507782e-07, + "loss": 0.6976, + "step": 12509 + }, + { + "epoch": 0.46, + "learning_rate": 1.5437067418610612e-07, + "loss": 0.7021, + "step": 12510 + }, + { + "epoch": 0.46, + "learning_rate": 1.5436701212143407e-07, + "loss": 0.6946, + "step": 12511 + }, + { + "epoch": 0.46, + "learning_rate": 1.54363350056762e-07, + "loss": 0.6879, + "step": 12512 + }, + { + "epoch": 0.46, + "learning_rate": 1.5435968799208995e-07, + "loss": 0.6859, + "step": 12513 + }, + { + "epoch": 0.46, + "learning_rate": 1.5435602592741788e-07, + "loss": 0.6952, + "step": 12514 + }, + { + "epoch": 0.46, + "learning_rate": 1.543523638627458e-07, + "loss": 0.6796, + "step": 12515 + }, + { + "epoch": 0.46, + "learning_rate": 1.5434870179807376e-07, + "loss": 0.6803, + "step": 12516 + }, + { + "epoch": 0.46, + "learning_rate": 1.5434503973340168e-07, + "loss": 0.6816, + "step": 12517 + }, + { + "epoch": 0.46, + "learning_rate": 1.5434137766872964e-07, + "loss": 0.687, + "step": 12518 + }, + { + "epoch": 0.46, + "learning_rate": 1.5433771560405756e-07, + "loss": 0.6907, + "step": 12519 + }, + { + "epoch": 0.46, + "learning_rate": 1.543340535393855e-07, + "loss": 0.7149, + "step": 12520 + }, + { + "epoch": 0.46, + "learning_rate": 1.5433039147471342e-07, + "loss": 0.6782, + "step": 12521 + }, + { + "epoch": 0.46, + "learning_rate": 1.5432672941004137e-07, + "loss": 0.6941, + "step": 12522 + }, + { + "epoch": 0.46, + "learning_rate": 1.543230673453693e-07, + "loss": 0.7002, + "step": 12523 + }, + { + "epoch": 0.46, + "learning_rate": 1.5431940528069725e-07, + "loss": 0.7115, + "step": 12524 + }, + { + "epoch": 0.46, + "learning_rate": 1.5431574321602518e-07, + "loss": 0.6841, + "step": 12525 + }, + { + "epoch": 0.46, + "learning_rate": 1.5431208115135313e-07, + "loss": 0.71, + "step": 12526 + }, + { + "epoch": 0.46, + "learning_rate": 1.5430841908668106e-07, + "loss": 0.6947, + "step": 12527 + }, + { + "epoch": 0.46, + "learning_rate": 1.54304757022009e-07, + "loss": 0.6888, + "step": 12528 + }, + { + "epoch": 0.46, + "learning_rate": 1.5430109495733694e-07, + "loss": 0.7162, + "step": 12529 + }, + { + "epoch": 0.46, + "learning_rate": 1.542974328926649e-07, + "loss": 0.7065, + "step": 12530 + }, + { + "epoch": 0.46, + "learning_rate": 1.5429377082799282e-07, + "loss": 0.7, + "step": 12531 + }, + { + "epoch": 0.46, + "learning_rate": 1.5429010876332077e-07, + "loss": 0.6858, + "step": 12532 + }, + { + "epoch": 0.46, + "learning_rate": 1.542864466986487e-07, + "loss": 0.698, + "step": 12533 + }, + { + "epoch": 0.46, + "learning_rate": 1.5428278463397665e-07, + "loss": 0.7045, + "step": 12534 + }, + { + "epoch": 0.46, + "learning_rate": 1.5427912256930458e-07, + "loss": 0.6868, + "step": 12535 + }, + { + "epoch": 0.46, + "learning_rate": 1.542754605046325e-07, + "loss": 0.7016, + "step": 12536 + }, + { + "epoch": 0.46, + "learning_rate": 1.5427179843996043e-07, + "loss": 0.6917, + "step": 12537 + }, + { + "epoch": 0.46, + "learning_rate": 1.5426813637528838e-07, + "loss": 0.7029, + "step": 12538 + }, + { + "epoch": 0.46, + "learning_rate": 1.542644743106163e-07, + "loss": 0.6964, + "step": 12539 + }, + { + "epoch": 0.46, + "learning_rate": 1.5426081224594426e-07, + "loss": 0.6769, + "step": 12540 + }, + { + "epoch": 0.46, + "learning_rate": 1.542571501812722e-07, + "loss": 0.6934, + "step": 12541 + }, + { + "epoch": 0.46, + "learning_rate": 1.5425348811660012e-07, + "loss": 0.6857, + "step": 12542 + }, + { + "epoch": 0.46, + "learning_rate": 1.5424982605192807e-07, + "loss": 0.6988, + "step": 12543 + }, + { + "epoch": 0.46, + "learning_rate": 1.54246163987256e-07, + "loss": 0.716, + "step": 12544 + }, + { + "epoch": 0.46, + "learning_rate": 1.5424250192258395e-07, + "loss": 0.6875, + "step": 12545 + }, + { + "epoch": 0.46, + "learning_rate": 1.5423883985791188e-07, + "loss": 0.6972, + "step": 12546 + }, + { + "epoch": 0.46, + "learning_rate": 1.5423517779323983e-07, + "loss": 0.7116, + "step": 12547 + }, + { + "epoch": 0.46, + "learning_rate": 1.5423151572856775e-07, + "loss": 0.6811, + "step": 12548 + }, + { + "epoch": 0.46, + "learning_rate": 1.542278536638957e-07, + "loss": 0.6884, + "step": 12549 + }, + { + "epoch": 0.46, + "learning_rate": 1.5422419159922363e-07, + "loss": 0.6963, + "step": 12550 + }, + { + "epoch": 0.46, + "learning_rate": 1.542205295345516e-07, + "loss": 0.6875, + "step": 12551 + }, + { + "epoch": 0.46, + "learning_rate": 1.5421686746987951e-07, + "loss": 0.6899, + "step": 12552 + }, + { + "epoch": 0.46, + "learning_rate": 1.5421320540520747e-07, + "loss": 0.6951, + "step": 12553 + }, + { + "epoch": 0.46, + "learning_rate": 1.542095433405354e-07, + "loss": 0.7063, + "step": 12554 + }, + { + "epoch": 0.46, + "learning_rate": 1.5420588127586332e-07, + "loss": 0.6825, + "step": 12555 + }, + { + "epoch": 0.46, + "learning_rate": 1.5420221921119125e-07, + "loss": 0.7001, + "step": 12556 + }, + { + "epoch": 0.46, + "learning_rate": 1.541985571465192e-07, + "loss": 0.7078, + "step": 12557 + }, + { + "epoch": 0.46, + "learning_rate": 1.5419489508184713e-07, + "loss": 0.7185, + "step": 12558 + }, + { + "epoch": 0.46, + "learning_rate": 1.5419123301717508e-07, + "loss": 0.6785, + "step": 12559 + }, + { + "epoch": 0.46, + "learning_rate": 1.54187570952503e-07, + "loss": 0.7035, + "step": 12560 + }, + { + "epoch": 0.46, + "learning_rate": 1.5418390888783096e-07, + "loss": 0.6884, + "step": 12561 + }, + { + "epoch": 0.46, + "learning_rate": 1.541802468231589e-07, + "loss": 0.6893, + "step": 12562 + }, + { + "epoch": 0.46, + "learning_rate": 1.5417658475848681e-07, + "loss": 0.693, + "step": 12563 + }, + { + "epoch": 0.46, + "learning_rate": 1.5417292269381477e-07, + "loss": 0.6905, + "step": 12564 + }, + { + "epoch": 0.46, + "learning_rate": 1.541692606291427e-07, + "loss": 0.6937, + "step": 12565 + }, + { + "epoch": 0.46, + "learning_rate": 1.5416559856447065e-07, + "loss": 0.7102, + "step": 12566 + }, + { + "epoch": 0.46, + "learning_rate": 1.5416193649979857e-07, + "loss": 0.6868, + "step": 12567 + }, + { + "epoch": 0.46, + "learning_rate": 1.5415827443512653e-07, + "loss": 0.6921, + "step": 12568 + }, + { + "epoch": 0.46, + "learning_rate": 1.5415461237045445e-07, + "loss": 0.6703, + "step": 12569 + }, + { + "epoch": 0.46, + "learning_rate": 1.541509503057824e-07, + "loss": 0.704, + "step": 12570 + }, + { + "epoch": 0.46, + "learning_rate": 1.5414728824111033e-07, + "loss": 0.6771, + "step": 12571 + }, + { + "epoch": 0.46, + "learning_rate": 1.5414362617643829e-07, + "loss": 0.7083, + "step": 12572 + }, + { + "epoch": 0.46, + "learning_rate": 1.541399641117662e-07, + "loss": 0.6963, + "step": 12573 + }, + { + "epoch": 0.46, + "learning_rate": 1.5413630204709414e-07, + "loss": 0.6871, + "step": 12574 + }, + { + "epoch": 0.46, + "learning_rate": 1.5413263998242207e-07, + "loss": 0.6889, + "step": 12575 + }, + { + "epoch": 0.46, + "learning_rate": 1.5412897791775002e-07, + "loss": 0.6875, + "step": 12576 + }, + { + "epoch": 0.46, + "learning_rate": 1.5412531585307795e-07, + "loss": 0.6775, + "step": 12577 + }, + { + "epoch": 0.46, + "learning_rate": 1.541216537884059e-07, + "loss": 0.6843, + "step": 12578 + }, + { + "epoch": 0.46, + "learning_rate": 1.5411799172373383e-07, + "loss": 0.699, + "step": 12579 + }, + { + "epoch": 0.46, + "learning_rate": 1.5411432965906178e-07, + "loss": 0.6919, + "step": 12580 + }, + { + "epoch": 0.46, + "learning_rate": 1.541106675943897e-07, + "loss": 0.692, + "step": 12581 + }, + { + "epoch": 0.46, + "learning_rate": 1.5410700552971766e-07, + "loss": 0.6922, + "step": 12582 + }, + { + "epoch": 0.46, + "learning_rate": 1.5410334346504559e-07, + "loss": 0.7223, + "step": 12583 + }, + { + "epoch": 0.46, + "learning_rate": 1.5409968140037354e-07, + "loss": 0.6833, + "step": 12584 + }, + { + "epoch": 0.46, + "learning_rate": 1.5409601933570147e-07, + "loss": 0.6942, + "step": 12585 + }, + { + "epoch": 0.46, + "learning_rate": 1.540923572710294e-07, + "loss": 0.698, + "step": 12586 + }, + { + "epoch": 0.46, + "learning_rate": 1.5408869520635735e-07, + "loss": 0.6898, + "step": 12587 + }, + { + "epoch": 0.46, + "learning_rate": 1.5408503314168527e-07, + "loss": 0.7012, + "step": 12588 + }, + { + "epoch": 0.46, + "learning_rate": 1.5408137107701323e-07, + "loss": 0.7065, + "step": 12589 + }, + { + "epoch": 0.46, + "learning_rate": 1.5407770901234115e-07, + "loss": 0.6937, + "step": 12590 + }, + { + "epoch": 0.46, + "learning_rate": 1.5407404694766908e-07, + "loss": 0.7071, + "step": 12591 + }, + { + "epoch": 0.46, + "learning_rate": 1.54070384882997e-07, + "loss": 0.6928, + "step": 12592 + }, + { + "epoch": 0.46, + "learning_rate": 1.5406672281832496e-07, + "loss": 0.7057, + "step": 12593 + }, + { + "epoch": 0.46, + "learning_rate": 1.5406306075365289e-07, + "loss": 0.7025, + "step": 12594 + }, + { + "epoch": 0.46, + "learning_rate": 1.5405939868898084e-07, + "loss": 0.6686, + "step": 12595 + }, + { + "epoch": 0.46, + "learning_rate": 1.5405573662430877e-07, + "loss": 0.6925, + "step": 12596 + }, + { + "epoch": 0.46, + "learning_rate": 1.5405207455963672e-07, + "loss": 0.6749, + "step": 12597 + }, + { + "epoch": 0.46, + "learning_rate": 1.5404841249496465e-07, + "loss": 0.7042, + "step": 12598 + }, + { + "epoch": 0.46, + "learning_rate": 1.540447504302926e-07, + "loss": 0.697, + "step": 12599 + }, + { + "epoch": 0.46, + "learning_rate": 1.5404108836562053e-07, + "loss": 0.6943, + "step": 12600 + }, + { + "epoch": 0.46, + "learning_rate": 1.5403742630094848e-07, + "loss": 0.7051, + "step": 12601 + }, + { + "epoch": 0.46, + "learning_rate": 1.540337642362764e-07, + "loss": 0.6963, + "step": 12602 + }, + { + "epoch": 0.46, + "learning_rate": 1.5403010217160436e-07, + "loss": 0.6842, + "step": 12603 + }, + { + "epoch": 0.46, + "learning_rate": 1.5402644010693229e-07, + "loss": 0.6838, + "step": 12604 + }, + { + "epoch": 0.46, + "learning_rate": 1.5402277804226024e-07, + "loss": 0.6887, + "step": 12605 + }, + { + "epoch": 0.46, + "learning_rate": 1.5401911597758816e-07, + "loss": 0.6871, + "step": 12606 + }, + { + "epoch": 0.46, + "learning_rate": 1.540154539129161e-07, + "loss": 0.6857, + "step": 12607 + }, + { + "epoch": 0.46, + "learning_rate": 1.5401179184824404e-07, + "loss": 0.6859, + "step": 12608 + }, + { + "epoch": 0.46, + "learning_rate": 1.5400812978357197e-07, + "loss": 0.6822, + "step": 12609 + }, + { + "epoch": 0.46, + "learning_rate": 1.540044677188999e-07, + "loss": 0.6995, + "step": 12610 + }, + { + "epoch": 0.46, + "learning_rate": 1.5400080565422785e-07, + "loss": 0.7017, + "step": 12611 + }, + { + "epoch": 0.46, + "learning_rate": 1.5399714358955578e-07, + "loss": 0.6897, + "step": 12612 + }, + { + "epoch": 0.46, + "learning_rate": 1.539934815248837e-07, + "loss": 0.6778, + "step": 12613 + }, + { + "epoch": 0.46, + "learning_rate": 1.5398981946021166e-07, + "loss": 0.6895, + "step": 12614 + }, + { + "epoch": 0.46, + "learning_rate": 1.5398615739553958e-07, + "loss": 0.6976, + "step": 12615 + }, + { + "epoch": 0.46, + "learning_rate": 1.5398249533086754e-07, + "loss": 0.6862, + "step": 12616 + }, + { + "epoch": 0.46, + "learning_rate": 1.5397883326619546e-07, + "loss": 0.6929, + "step": 12617 + }, + { + "epoch": 0.46, + "learning_rate": 1.5397517120152342e-07, + "loss": 0.6666, + "step": 12618 + }, + { + "epoch": 0.46, + "learning_rate": 1.5397150913685134e-07, + "loss": 0.6761, + "step": 12619 + }, + { + "epoch": 0.46, + "learning_rate": 1.539678470721793e-07, + "loss": 0.6975, + "step": 12620 + }, + { + "epoch": 0.46, + "learning_rate": 1.5396418500750722e-07, + "loss": 0.6936, + "step": 12621 + }, + { + "epoch": 0.46, + "learning_rate": 1.5396052294283518e-07, + "loss": 0.6965, + "step": 12622 + }, + { + "epoch": 0.46, + "learning_rate": 1.539568608781631e-07, + "loss": 0.7, + "step": 12623 + }, + { + "epoch": 0.46, + "learning_rate": 1.5395319881349106e-07, + "loss": 0.6868, + "step": 12624 + }, + { + "epoch": 0.46, + "learning_rate": 1.5394953674881898e-07, + "loss": 0.7065, + "step": 12625 + }, + { + "epoch": 0.46, + "learning_rate": 1.5394587468414694e-07, + "loss": 0.6696, + "step": 12626 + }, + { + "epoch": 0.46, + "learning_rate": 1.5394221261947484e-07, + "loss": 0.6955, + "step": 12627 + }, + { + "epoch": 0.46, + "learning_rate": 1.539385505548028e-07, + "loss": 0.6909, + "step": 12628 + }, + { + "epoch": 0.46, + "learning_rate": 1.5393488849013072e-07, + "loss": 0.6873, + "step": 12629 + }, + { + "epoch": 0.46, + "learning_rate": 1.5393122642545867e-07, + "loss": 0.7016, + "step": 12630 + }, + { + "epoch": 0.46, + "learning_rate": 1.539275643607866e-07, + "loss": 0.6942, + "step": 12631 + }, + { + "epoch": 0.46, + "learning_rate": 1.5392390229611455e-07, + "loss": 0.6815, + "step": 12632 + }, + { + "epoch": 0.46, + "learning_rate": 1.5392024023144248e-07, + "loss": 0.7016, + "step": 12633 + }, + { + "epoch": 0.46, + "learning_rate": 1.539165781667704e-07, + "loss": 0.7106, + "step": 12634 + }, + { + "epoch": 0.46, + "learning_rate": 1.5391291610209836e-07, + "loss": 0.6903, + "step": 12635 + }, + { + "epoch": 0.46, + "learning_rate": 1.5390925403742628e-07, + "loss": 0.6876, + "step": 12636 + }, + { + "epoch": 0.46, + "learning_rate": 1.5390559197275424e-07, + "loss": 0.7039, + "step": 12637 + }, + { + "epoch": 0.46, + "learning_rate": 1.5390192990808216e-07, + "loss": 0.6987, + "step": 12638 + }, + { + "epoch": 0.46, + "learning_rate": 1.5389826784341012e-07, + "loss": 0.6889, + "step": 12639 + }, + { + "epoch": 0.46, + "learning_rate": 1.5389460577873804e-07, + "loss": 0.6846, + "step": 12640 + }, + { + "epoch": 0.46, + "learning_rate": 1.53890943714066e-07, + "loss": 0.6874, + "step": 12641 + }, + { + "epoch": 0.46, + "learning_rate": 1.5388728164939392e-07, + "loss": 0.6895, + "step": 12642 + }, + { + "epoch": 0.46, + "learning_rate": 1.5388361958472188e-07, + "loss": 0.6924, + "step": 12643 + }, + { + "epoch": 0.46, + "learning_rate": 1.538799575200498e-07, + "loss": 0.7108, + "step": 12644 + }, + { + "epoch": 0.46, + "learning_rate": 1.5387629545537773e-07, + "loss": 0.6886, + "step": 12645 + }, + { + "epoch": 0.46, + "learning_rate": 1.5387263339070566e-07, + "loss": 0.69, + "step": 12646 + }, + { + "epoch": 0.46, + "learning_rate": 1.538689713260336e-07, + "loss": 0.6912, + "step": 12647 + }, + { + "epoch": 0.46, + "learning_rate": 1.5386530926136154e-07, + "loss": 0.7076, + "step": 12648 + }, + { + "epoch": 0.46, + "learning_rate": 1.538616471966895e-07, + "loss": 0.6962, + "step": 12649 + }, + { + "epoch": 0.46, + "learning_rate": 1.5385798513201742e-07, + "loss": 0.6766, + "step": 12650 + }, + { + "epoch": 0.46, + "learning_rate": 1.5385432306734537e-07, + "loss": 0.6946, + "step": 12651 + }, + { + "epoch": 0.46, + "learning_rate": 1.538506610026733e-07, + "loss": 0.69, + "step": 12652 + }, + { + "epoch": 0.46, + "learning_rate": 1.5384699893800125e-07, + "loss": 0.6941, + "step": 12653 + }, + { + "epoch": 0.46, + "learning_rate": 1.5384333687332918e-07, + "loss": 0.6978, + "step": 12654 + }, + { + "epoch": 0.46, + "learning_rate": 1.5383967480865713e-07, + "loss": 0.6885, + "step": 12655 + }, + { + "epoch": 0.46, + "learning_rate": 1.5383601274398506e-07, + "loss": 0.7023, + "step": 12656 + }, + { + "epoch": 0.46, + "learning_rate": 1.5383235067931298e-07, + "loss": 0.6925, + "step": 12657 + }, + { + "epoch": 0.46, + "learning_rate": 1.5382868861464094e-07, + "loss": 0.6958, + "step": 12658 + }, + { + "epoch": 0.46, + "learning_rate": 1.5382502654996886e-07, + "loss": 0.6801, + "step": 12659 + }, + { + "epoch": 0.46, + "learning_rate": 1.5382136448529682e-07, + "loss": 0.6831, + "step": 12660 + }, + { + "epoch": 0.46, + "learning_rate": 1.5381770242062474e-07, + "loss": 0.6943, + "step": 12661 + }, + { + "epoch": 0.46, + "learning_rate": 1.538140403559527e-07, + "loss": 0.6956, + "step": 12662 + }, + { + "epoch": 0.46, + "learning_rate": 1.538103782912806e-07, + "loss": 0.6965, + "step": 12663 + }, + { + "epoch": 0.46, + "learning_rate": 1.5380671622660855e-07, + "loss": 0.6788, + "step": 12664 + }, + { + "epoch": 0.46, + "learning_rate": 1.5380305416193648e-07, + "loss": 0.6845, + "step": 12665 + }, + { + "epoch": 0.46, + "learning_rate": 1.5379939209726443e-07, + "loss": 0.693, + "step": 12666 + }, + { + "epoch": 0.46, + "learning_rate": 1.5379573003259235e-07, + "loss": 0.6939, + "step": 12667 + }, + { + "epoch": 0.46, + "learning_rate": 1.537920679679203e-07, + "loss": 0.6997, + "step": 12668 + }, + { + "epoch": 0.46, + "learning_rate": 1.5378840590324823e-07, + "loss": 0.703, + "step": 12669 + }, + { + "epoch": 0.46, + "learning_rate": 1.537847438385762e-07, + "loss": 0.7005, + "step": 12670 + }, + { + "epoch": 0.46, + "learning_rate": 1.5378108177390411e-07, + "loss": 0.6874, + "step": 12671 + }, + { + "epoch": 0.46, + "learning_rate": 1.5377741970923207e-07, + "loss": 0.6879, + "step": 12672 + }, + { + "epoch": 0.46, + "learning_rate": 1.5377375764456e-07, + "loss": 0.6995, + "step": 12673 + }, + { + "epoch": 0.46, + "learning_rate": 1.5377009557988795e-07, + "loss": 0.6965, + "step": 12674 + }, + { + "epoch": 0.46, + "learning_rate": 1.5376643351521587e-07, + "loss": 0.6769, + "step": 12675 + }, + { + "epoch": 0.46, + "learning_rate": 1.5376277145054383e-07, + "loss": 0.696, + "step": 12676 + }, + { + "epoch": 0.46, + "learning_rate": 1.5375910938587175e-07, + "loss": 0.7014, + "step": 12677 + }, + { + "epoch": 0.46, + "learning_rate": 1.5375544732119968e-07, + "loss": 0.6916, + "step": 12678 + }, + { + "epoch": 0.46, + "learning_rate": 1.5375178525652763e-07, + "loss": 0.6904, + "step": 12679 + }, + { + "epoch": 0.46, + "learning_rate": 1.5374812319185556e-07, + "loss": 0.6964, + "step": 12680 + }, + { + "epoch": 0.46, + "learning_rate": 1.537444611271835e-07, + "loss": 0.6943, + "step": 12681 + }, + { + "epoch": 0.46, + "learning_rate": 1.5374079906251144e-07, + "loss": 0.6857, + "step": 12682 + }, + { + "epoch": 0.46, + "learning_rate": 1.5373713699783937e-07, + "loss": 0.6895, + "step": 12683 + }, + { + "epoch": 0.46, + "learning_rate": 1.537334749331673e-07, + "loss": 0.682, + "step": 12684 + }, + { + "epoch": 0.46, + "learning_rate": 1.5372981286849525e-07, + "loss": 0.6747, + "step": 12685 + }, + { + "epoch": 0.46, + "learning_rate": 1.5372615080382317e-07, + "loss": 0.7076, + "step": 12686 + }, + { + "epoch": 0.46, + "learning_rate": 1.5372248873915113e-07, + "loss": 0.7071, + "step": 12687 + }, + { + "epoch": 0.46, + "learning_rate": 1.5371882667447905e-07, + "loss": 0.6947, + "step": 12688 + }, + { + "epoch": 0.46, + "learning_rate": 1.53715164609807e-07, + "loss": 0.6917, + "step": 12689 + }, + { + "epoch": 0.46, + "learning_rate": 1.5371150254513493e-07, + "loss": 0.6965, + "step": 12690 + }, + { + "epoch": 0.46, + "learning_rate": 1.5370784048046289e-07, + "loss": 0.6944, + "step": 12691 + }, + { + "epoch": 0.46, + "learning_rate": 1.5370417841579081e-07, + "loss": 0.6889, + "step": 12692 + }, + { + "epoch": 0.46, + "learning_rate": 1.5370051635111877e-07, + "loss": 0.6798, + "step": 12693 + }, + { + "epoch": 0.46, + "learning_rate": 1.536968542864467e-07, + "loss": 0.699, + "step": 12694 + }, + { + "epoch": 0.46, + "learning_rate": 1.5369319222177465e-07, + "loss": 0.6848, + "step": 12695 + }, + { + "epoch": 0.46, + "learning_rate": 1.5368953015710257e-07, + "loss": 0.6806, + "step": 12696 + }, + { + "epoch": 0.46, + "learning_rate": 1.5368586809243053e-07, + "loss": 0.6966, + "step": 12697 + }, + { + "epoch": 0.46, + "learning_rate": 1.5368220602775845e-07, + "loss": 0.6902, + "step": 12698 + }, + { + "epoch": 0.46, + "learning_rate": 1.5367854396308638e-07, + "loss": 0.6851, + "step": 12699 + }, + { + "epoch": 0.46, + "learning_rate": 1.536748818984143e-07, + "loss": 0.708, + "step": 12700 + }, + { + "epoch": 0.46, + "learning_rate": 1.5367121983374226e-07, + "loss": 0.6749, + "step": 12701 + }, + { + "epoch": 0.46, + "learning_rate": 1.5366755776907019e-07, + "loss": 0.7089, + "step": 12702 + }, + { + "epoch": 0.46, + "learning_rate": 1.5366389570439814e-07, + "loss": 0.6809, + "step": 12703 + }, + { + "epoch": 0.46, + "learning_rate": 1.5366023363972607e-07, + "loss": 0.6936, + "step": 12704 + }, + { + "epoch": 0.46, + "learning_rate": 1.53656571575054e-07, + "loss": 0.7055, + "step": 12705 + }, + { + "epoch": 0.46, + "learning_rate": 1.5365290951038195e-07, + "loss": 0.6875, + "step": 12706 + }, + { + "epoch": 0.46, + "learning_rate": 1.5364924744570987e-07, + "loss": 0.7091, + "step": 12707 + }, + { + "epoch": 0.46, + "learning_rate": 1.5364558538103783e-07, + "loss": 0.6947, + "step": 12708 + }, + { + "epoch": 0.46, + "learning_rate": 1.5364192331636575e-07, + "loss": 0.6873, + "step": 12709 + }, + { + "epoch": 0.47, + "learning_rate": 1.536382612516937e-07, + "loss": 0.697, + "step": 12710 + }, + { + "epoch": 0.47, + "learning_rate": 1.5363459918702163e-07, + "loss": 0.7074, + "step": 12711 + }, + { + "epoch": 0.47, + "learning_rate": 1.5363093712234959e-07, + "loss": 0.6942, + "step": 12712 + }, + { + "epoch": 0.47, + "learning_rate": 1.536272750576775e-07, + "loss": 0.6868, + "step": 12713 + }, + { + "epoch": 0.47, + "learning_rate": 1.5362361299300547e-07, + "loss": 0.6898, + "step": 12714 + }, + { + "epoch": 0.47, + "learning_rate": 1.536199509283334e-07, + "loss": 0.7099, + "step": 12715 + }, + { + "epoch": 0.47, + "learning_rate": 1.5361628886366135e-07, + "loss": 0.6811, + "step": 12716 + }, + { + "epoch": 0.47, + "learning_rate": 1.5361262679898925e-07, + "loss": 0.6949, + "step": 12717 + }, + { + "epoch": 0.47, + "learning_rate": 1.536089647343172e-07, + "loss": 0.6842, + "step": 12718 + }, + { + "epoch": 0.47, + "learning_rate": 1.5360530266964513e-07, + "loss": 0.6844, + "step": 12719 + }, + { + "epoch": 0.47, + "learning_rate": 1.5360164060497308e-07, + "loss": 0.6761, + "step": 12720 + }, + { + "epoch": 0.47, + "learning_rate": 1.53597978540301e-07, + "loss": 0.6995, + "step": 12721 + }, + { + "epoch": 0.47, + "learning_rate": 1.5359431647562896e-07, + "loss": 0.7036, + "step": 12722 + }, + { + "epoch": 0.47, + "learning_rate": 1.5359065441095689e-07, + "loss": 0.6814, + "step": 12723 + }, + { + "epoch": 0.47, + "learning_rate": 1.5358699234628484e-07, + "loss": 0.7035, + "step": 12724 + }, + { + "epoch": 0.47, + "learning_rate": 1.5358333028161276e-07, + "loss": 0.6981, + "step": 12725 + }, + { + "epoch": 0.47, + "learning_rate": 1.5357966821694072e-07, + "loss": 0.6951, + "step": 12726 + }, + { + "epoch": 0.47, + "learning_rate": 1.5357600615226864e-07, + "loss": 0.6878, + "step": 12727 + }, + { + "epoch": 0.47, + "learning_rate": 1.5357234408759657e-07, + "loss": 0.7003, + "step": 12728 + }, + { + "epoch": 0.47, + "learning_rate": 1.5356868202292452e-07, + "loss": 0.6962, + "step": 12729 + }, + { + "epoch": 0.47, + "learning_rate": 1.5356501995825245e-07, + "loss": 0.6925, + "step": 12730 + }, + { + "epoch": 0.47, + "learning_rate": 1.535613578935804e-07, + "loss": 0.6963, + "step": 12731 + }, + { + "epoch": 0.47, + "learning_rate": 1.5355769582890833e-07, + "loss": 0.6951, + "step": 12732 + }, + { + "epoch": 0.47, + "learning_rate": 1.5355403376423628e-07, + "loss": 0.694, + "step": 12733 + }, + { + "epoch": 0.47, + "learning_rate": 1.535503716995642e-07, + "loss": 0.6907, + "step": 12734 + }, + { + "epoch": 0.47, + "learning_rate": 1.5354670963489214e-07, + "loss": 0.6821, + "step": 12735 + }, + { + "epoch": 0.47, + "learning_rate": 1.5354304757022006e-07, + "loss": 0.7015, + "step": 12736 + }, + { + "epoch": 0.47, + "learning_rate": 1.5353938550554802e-07, + "loss": 0.6892, + "step": 12737 + }, + { + "epoch": 0.47, + "learning_rate": 1.5353572344087594e-07, + "loss": 0.6797, + "step": 12738 + }, + { + "epoch": 0.47, + "learning_rate": 1.535320613762039e-07, + "loss": 0.7019, + "step": 12739 + }, + { + "epoch": 0.47, + "learning_rate": 1.5352839931153182e-07, + "loss": 0.6952, + "step": 12740 + }, + { + "epoch": 0.47, + "learning_rate": 1.5352473724685978e-07, + "loss": 0.701, + "step": 12741 + }, + { + "epoch": 0.47, + "learning_rate": 1.535210751821877e-07, + "loss": 0.7043, + "step": 12742 + }, + { + "epoch": 0.47, + "learning_rate": 1.5351741311751566e-07, + "loss": 0.6868, + "step": 12743 + }, + { + "epoch": 0.47, + "learning_rate": 1.5351375105284358e-07, + "loss": 0.6895, + "step": 12744 + }, + { + "epoch": 0.47, + "learning_rate": 1.5351008898817154e-07, + "loss": 0.7041, + "step": 12745 + }, + { + "epoch": 0.47, + "learning_rate": 1.5350642692349946e-07, + "loss": 0.7058, + "step": 12746 + }, + { + "epoch": 0.47, + "learning_rate": 1.5350276485882742e-07, + "loss": 0.6945, + "step": 12747 + }, + { + "epoch": 0.47, + "learning_rate": 1.5349910279415534e-07, + "loss": 0.692, + "step": 12748 + }, + { + "epoch": 0.47, + "learning_rate": 1.5349544072948327e-07, + "loss": 0.6853, + "step": 12749 + }, + { + "epoch": 0.47, + "learning_rate": 1.5349177866481122e-07, + "loss": 0.6983, + "step": 12750 + }, + { + "epoch": 0.47, + "learning_rate": 1.5348811660013915e-07, + "loss": 0.6949, + "step": 12751 + }, + { + "epoch": 0.47, + "learning_rate": 1.5348445453546708e-07, + "loss": 0.698, + "step": 12752 + }, + { + "epoch": 0.47, + "learning_rate": 1.5348079247079503e-07, + "loss": 0.6935, + "step": 12753 + }, + { + "epoch": 0.47, + "learning_rate": 1.5347713040612296e-07, + "loss": 0.6958, + "step": 12754 + }, + { + "epoch": 0.47, + "learning_rate": 1.5347346834145088e-07, + "loss": 0.6848, + "step": 12755 + }, + { + "epoch": 0.47, + "learning_rate": 1.5346980627677884e-07, + "loss": 0.6926, + "step": 12756 + }, + { + "epoch": 0.47, + "learning_rate": 1.5346614421210676e-07, + "loss": 0.6936, + "step": 12757 + }, + { + "epoch": 0.47, + "learning_rate": 1.5346248214743472e-07, + "loss": 0.689, + "step": 12758 + }, + { + "epoch": 0.47, + "learning_rate": 1.5345882008276264e-07, + "loss": 0.6955, + "step": 12759 + }, + { + "epoch": 0.47, + "learning_rate": 1.534551580180906e-07, + "loss": 0.6837, + "step": 12760 + }, + { + "epoch": 0.47, + "learning_rate": 1.5345149595341852e-07, + "loss": 0.6718, + "step": 12761 + }, + { + "epoch": 0.47, + "learning_rate": 1.5344783388874648e-07, + "loss": 0.6936, + "step": 12762 + }, + { + "epoch": 0.47, + "learning_rate": 1.534441718240744e-07, + "loss": 0.6956, + "step": 12763 + }, + { + "epoch": 0.47, + "learning_rate": 1.5344050975940236e-07, + "loss": 0.6869, + "step": 12764 + }, + { + "epoch": 0.47, + "learning_rate": 1.5343684769473028e-07, + "loss": 0.6834, + "step": 12765 + }, + { + "epoch": 0.47, + "learning_rate": 1.5343318563005824e-07, + "loss": 0.6858, + "step": 12766 + }, + { + "epoch": 0.47, + "learning_rate": 1.5342952356538616e-07, + "loss": 0.6907, + "step": 12767 + }, + { + "epoch": 0.47, + "learning_rate": 1.5342586150071412e-07, + "loss": 0.6942, + "step": 12768 + }, + { + "epoch": 0.47, + "learning_rate": 1.5342219943604204e-07, + "loss": 0.6949, + "step": 12769 + }, + { + "epoch": 0.47, + "learning_rate": 1.5341853737136997e-07, + "loss": 0.6921, + "step": 12770 + }, + { + "epoch": 0.47, + "learning_rate": 1.534148753066979e-07, + "loss": 0.6881, + "step": 12771 + }, + { + "epoch": 0.47, + "learning_rate": 1.5341121324202585e-07, + "loss": 0.684, + "step": 12772 + }, + { + "epoch": 0.47, + "learning_rate": 1.5340755117735378e-07, + "loss": 0.7016, + "step": 12773 + }, + { + "epoch": 0.47, + "learning_rate": 1.5340388911268173e-07, + "loss": 0.7009, + "step": 12774 + }, + { + "epoch": 0.47, + "learning_rate": 1.5340022704800966e-07, + "loss": 0.6977, + "step": 12775 + }, + { + "epoch": 0.47, + "learning_rate": 1.5339656498333758e-07, + "loss": 0.6881, + "step": 12776 + }, + { + "epoch": 0.47, + "learning_rate": 1.5339290291866554e-07, + "loss": 0.7, + "step": 12777 + }, + { + "epoch": 0.47, + "learning_rate": 1.5338924085399346e-07, + "loss": 0.6998, + "step": 12778 + }, + { + "epoch": 0.47, + "learning_rate": 1.5338557878932142e-07, + "loss": 0.7064, + "step": 12779 + }, + { + "epoch": 0.47, + "learning_rate": 1.5338191672464934e-07, + "loss": 0.6898, + "step": 12780 + }, + { + "epoch": 0.47, + "learning_rate": 1.533782546599773e-07, + "loss": 0.7137, + "step": 12781 + }, + { + "epoch": 0.47, + "learning_rate": 1.5337459259530522e-07, + "loss": 0.694, + "step": 12782 + }, + { + "epoch": 0.47, + "learning_rate": 1.5337093053063317e-07, + "loss": 0.6772, + "step": 12783 + }, + { + "epoch": 0.47, + "learning_rate": 1.533672684659611e-07, + "loss": 0.6962, + "step": 12784 + }, + { + "epoch": 0.47, + "learning_rate": 1.5336360640128905e-07, + "loss": 0.6926, + "step": 12785 + }, + { + "epoch": 0.47, + "learning_rate": 1.5335994433661698e-07, + "loss": 0.6945, + "step": 12786 + }, + { + "epoch": 0.47, + "learning_rate": 1.5335628227194493e-07, + "loss": 0.6868, + "step": 12787 + }, + { + "epoch": 0.47, + "learning_rate": 1.5335262020727283e-07, + "loss": 0.6878, + "step": 12788 + }, + { + "epoch": 0.47, + "learning_rate": 1.533489581426008e-07, + "loss": 0.6808, + "step": 12789 + }, + { + "epoch": 0.47, + "learning_rate": 1.5334529607792871e-07, + "loss": 0.7058, + "step": 12790 + }, + { + "epoch": 0.47, + "learning_rate": 1.5334163401325667e-07, + "loss": 0.717, + "step": 12791 + }, + { + "epoch": 0.47, + "learning_rate": 1.533379719485846e-07, + "loss": 0.6833, + "step": 12792 + }, + { + "epoch": 0.47, + "learning_rate": 1.5333430988391255e-07, + "loss": 0.6933, + "step": 12793 + }, + { + "epoch": 0.47, + "learning_rate": 1.5333064781924047e-07, + "loss": 0.6863, + "step": 12794 + }, + { + "epoch": 0.47, + "learning_rate": 1.5332698575456843e-07, + "loss": 0.6857, + "step": 12795 + }, + { + "epoch": 0.47, + "learning_rate": 1.5332332368989635e-07, + "loss": 0.6867, + "step": 12796 + }, + { + "epoch": 0.47, + "learning_rate": 1.533196616252243e-07, + "loss": 0.6971, + "step": 12797 + }, + { + "epoch": 0.47, + "learning_rate": 1.5331599956055223e-07, + "loss": 0.6894, + "step": 12798 + }, + { + "epoch": 0.47, + "learning_rate": 1.5331233749588016e-07, + "loss": 0.7024, + "step": 12799 + }, + { + "epoch": 0.47, + "learning_rate": 1.5330867543120811e-07, + "loss": 0.6929, + "step": 12800 + }, + { + "epoch": 0.47, + "learning_rate": 1.5330501336653604e-07, + "loss": 0.7003, + "step": 12801 + }, + { + "epoch": 0.47, + "learning_rate": 1.53301351301864e-07, + "loss": 0.68, + "step": 12802 + }, + { + "epoch": 0.47, + "learning_rate": 1.5329768923719192e-07, + "loss": 0.6918, + "step": 12803 + }, + { + "epoch": 0.47, + "learning_rate": 1.5329402717251987e-07, + "loss": 0.6757, + "step": 12804 + }, + { + "epoch": 0.47, + "learning_rate": 1.532903651078478e-07, + "loss": 0.6953, + "step": 12805 + }, + { + "epoch": 0.47, + "learning_rate": 1.5328670304317573e-07, + "loss": 0.6874, + "step": 12806 + }, + { + "epoch": 0.47, + "learning_rate": 1.5328304097850365e-07, + "loss": 0.6775, + "step": 12807 + }, + { + "epoch": 0.47, + "learning_rate": 1.532793789138316e-07, + "loss": 0.6977, + "step": 12808 + }, + { + "epoch": 0.47, + "learning_rate": 1.5327571684915953e-07, + "loss": 0.6886, + "step": 12809 + }, + { + "epoch": 0.47, + "learning_rate": 1.532720547844875e-07, + "loss": 0.7131, + "step": 12810 + }, + { + "epoch": 0.47, + "learning_rate": 1.5326839271981541e-07, + "loss": 0.7022, + "step": 12811 + }, + { + "epoch": 0.47, + "learning_rate": 1.5326473065514337e-07, + "loss": 0.706, + "step": 12812 + }, + { + "epoch": 0.47, + "learning_rate": 1.532610685904713e-07, + "loss": 0.7084, + "step": 12813 + }, + { + "epoch": 0.47, + "learning_rate": 1.5325740652579925e-07, + "loss": 0.6987, + "step": 12814 + }, + { + "epoch": 0.47, + "learning_rate": 1.5325374446112717e-07, + "loss": 0.6891, + "step": 12815 + }, + { + "epoch": 0.47, + "learning_rate": 1.5325008239645513e-07, + "loss": 0.6935, + "step": 12816 + }, + { + "epoch": 0.47, + "learning_rate": 1.5324642033178305e-07, + "loss": 0.6962, + "step": 12817 + }, + { + "epoch": 0.47, + "learning_rate": 1.53242758267111e-07, + "loss": 0.6955, + "step": 12818 + }, + { + "epoch": 0.47, + "learning_rate": 1.5323909620243893e-07, + "loss": 0.7001, + "step": 12819 + }, + { + "epoch": 0.47, + "learning_rate": 1.5323543413776686e-07, + "loss": 0.6878, + "step": 12820 + }, + { + "epoch": 0.47, + "learning_rate": 1.532317720730948e-07, + "loss": 0.6902, + "step": 12821 + }, + { + "epoch": 0.47, + "learning_rate": 1.5322811000842274e-07, + "loss": 0.7038, + "step": 12822 + }, + { + "epoch": 0.47, + "learning_rate": 1.532244479437507e-07, + "loss": 0.7046, + "step": 12823 + }, + { + "epoch": 0.47, + "learning_rate": 1.5322078587907862e-07, + "loss": 0.7049, + "step": 12824 + }, + { + "epoch": 0.47, + "learning_rate": 1.5321712381440655e-07, + "loss": 0.6959, + "step": 12825 + }, + { + "epoch": 0.47, + "learning_rate": 1.5321346174973447e-07, + "loss": 0.6867, + "step": 12826 + }, + { + "epoch": 0.47, + "learning_rate": 1.5320979968506243e-07, + "loss": 0.6906, + "step": 12827 + }, + { + "epoch": 0.47, + "learning_rate": 1.5320613762039035e-07, + "loss": 0.702, + "step": 12828 + }, + { + "epoch": 0.47, + "learning_rate": 1.532024755557183e-07, + "loss": 0.6852, + "step": 12829 + }, + { + "epoch": 0.47, + "learning_rate": 1.5319881349104623e-07, + "loss": 0.7119, + "step": 12830 + }, + { + "epoch": 0.47, + "learning_rate": 1.5319515142637419e-07, + "loss": 0.7005, + "step": 12831 + }, + { + "epoch": 0.47, + "learning_rate": 1.531914893617021e-07, + "loss": 0.6867, + "step": 12832 + }, + { + "epoch": 0.47, + "learning_rate": 1.5318782729703007e-07, + "loss": 0.716, + "step": 12833 + }, + { + "epoch": 0.47, + "learning_rate": 1.53184165232358e-07, + "loss": 0.6824, + "step": 12834 + }, + { + "epoch": 0.47, + "learning_rate": 1.5318050316768595e-07, + "loss": 0.7143, + "step": 12835 + }, + { + "epoch": 0.47, + "learning_rate": 1.5317684110301387e-07, + "loss": 0.6949, + "step": 12836 + }, + { + "epoch": 0.47, + "learning_rate": 1.5317317903834183e-07, + "loss": 0.6973, + "step": 12837 + }, + { + "epoch": 0.47, + "learning_rate": 1.5316951697366975e-07, + "loss": 0.6898, + "step": 12838 + }, + { + "epoch": 0.47, + "learning_rate": 1.531658549089977e-07, + "loss": 0.7052, + "step": 12839 + }, + { + "epoch": 0.47, + "learning_rate": 1.5316219284432563e-07, + "loss": 0.6854, + "step": 12840 + }, + { + "epoch": 0.47, + "learning_rate": 1.5315853077965358e-07, + "loss": 0.7059, + "step": 12841 + }, + { + "epoch": 0.47, + "learning_rate": 1.5315486871498149e-07, + "loss": 0.69, + "step": 12842 + }, + { + "epoch": 0.47, + "learning_rate": 1.5315120665030944e-07, + "loss": 0.6718, + "step": 12843 + }, + { + "epoch": 0.47, + "learning_rate": 1.5314754458563736e-07, + "loss": 0.7006, + "step": 12844 + }, + { + "epoch": 0.47, + "learning_rate": 1.5314388252096532e-07, + "loss": 0.6883, + "step": 12845 + }, + { + "epoch": 0.47, + "learning_rate": 1.5314022045629324e-07, + "loss": 0.6882, + "step": 12846 + }, + { + "epoch": 0.47, + "learning_rate": 1.5313655839162117e-07, + "loss": 0.7075, + "step": 12847 + }, + { + "epoch": 0.47, + "learning_rate": 1.5313289632694912e-07, + "loss": 0.7105, + "step": 12848 + }, + { + "epoch": 0.47, + "learning_rate": 1.5312923426227705e-07, + "loss": 0.6928, + "step": 12849 + }, + { + "epoch": 0.47, + "learning_rate": 1.53125572197605e-07, + "loss": 0.6757, + "step": 12850 + }, + { + "epoch": 0.47, + "learning_rate": 1.5312191013293293e-07, + "loss": 0.7083, + "step": 12851 + }, + { + "epoch": 0.47, + "learning_rate": 1.5311824806826088e-07, + "loss": 0.6873, + "step": 12852 + }, + { + "epoch": 0.47, + "learning_rate": 1.531145860035888e-07, + "loss": 0.6907, + "step": 12853 + }, + { + "epoch": 0.47, + "learning_rate": 1.5311092393891676e-07, + "loss": 0.6976, + "step": 12854 + }, + { + "epoch": 0.47, + "learning_rate": 1.531072618742447e-07, + "loss": 0.6903, + "step": 12855 + }, + { + "epoch": 0.47, + "learning_rate": 1.5310359980957264e-07, + "loss": 0.7069, + "step": 12856 + }, + { + "epoch": 0.47, + "learning_rate": 1.5309993774490057e-07, + "loss": 0.6812, + "step": 12857 + }, + { + "epoch": 0.47, + "learning_rate": 1.5309627568022852e-07, + "loss": 0.67, + "step": 12858 + }, + { + "epoch": 0.47, + "learning_rate": 1.5309261361555645e-07, + "loss": 0.6942, + "step": 12859 + }, + { + "epoch": 0.47, + "learning_rate": 1.5308895155088438e-07, + "loss": 0.7029, + "step": 12860 + }, + { + "epoch": 0.47, + "learning_rate": 1.530852894862123e-07, + "loss": 0.7002, + "step": 12861 + }, + { + "epoch": 0.47, + "learning_rate": 1.5308162742154026e-07, + "loss": 0.6912, + "step": 12862 + }, + { + "epoch": 0.47, + "learning_rate": 1.5307796535686818e-07, + "loss": 0.6901, + "step": 12863 + }, + { + "epoch": 0.47, + "learning_rate": 1.5307430329219614e-07, + "loss": 0.6945, + "step": 12864 + }, + { + "epoch": 0.47, + "learning_rate": 1.5307064122752406e-07, + "loss": 0.7057, + "step": 12865 + }, + { + "epoch": 0.47, + "learning_rate": 1.5306697916285202e-07, + "loss": 0.7004, + "step": 12866 + }, + { + "epoch": 0.47, + "learning_rate": 1.5306331709817994e-07, + "loss": 0.6955, + "step": 12867 + }, + { + "epoch": 0.47, + "learning_rate": 1.530596550335079e-07, + "loss": 0.6907, + "step": 12868 + }, + { + "epoch": 0.47, + "learning_rate": 1.5305599296883582e-07, + "loss": 0.7029, + "step": 12869 + }, + { + "epoch": 0.47, + "learning_rate": 1.5305233090416375e-07, + "loss": 0.672, + "step": 12870 + }, + { + "epoch": 0.47, + "learning_rate": 1.530486688394917e-07, + "loss": 0.6845, + "step": 12871 + }, + { + "epoch": 0.47, + "learning_rate": 1.5304500677481963e-07, + "loss": 0.6902, + "step": 12872 + }, + { + "epoch": 0.47, + "learning_rate": 1.5304134471014758e-07, + "loss": 0.6873, + "step": 12873 + }, + { + "epoch": 0.47, + "learning_rate": 1.530376826454755e-07, + "loss": 0.676, + "step": 12874 + }, + { + "epoch": 0.47, + "learning_rate": 1.5303402058080346e-07, + "loss": 0.6791, + "step": 12875 + }, + { + "epoch": 0.47, + "learning_rate": 1.530303585161314e-07, + "loss": 0.6821, + "step": 12876 + }, + { + "epoch": 0.47, + "learning_rate": 1.5302669645145934e-07, + "loss": 0.699, + "step": 12877 + }, + { + "epoch": 0.47, + "learning_rate": 1.5302303438678724e-07, + "loss": 0.7067, + "step": 12878 + }, + { + "epoch": 0.47, + "learning_rate": 1.530193723221152e-07, + "loss": 0.6821, + "step": 12879 + }, + { + "epoch": 0.47, + "learning_rate": 1.5301571025744312e-07, + "loss": 0.7027, + "step": 12880 + }, + { + "epoch": 0.47, + "learning_rate": 1.5301204819277108e-07, + "loss": 0.6985, + "step": 12881 + }, + { + "epoch": 0.47, + "learning_rate": 1.53008386128099e-07, + "loss": 0.6897, + "step": 12882 + }, + { + "epoch": 0.47, + "learning_rate": 1.5300472406342696e-07, + "loss": 0.691, + "step": 12883 + }, + { + "epoch": 0.47, + "learning_rate": 1.5300106199875488e-07, + "loss": 0.7098, + "step": 12884 + }, + { + "epoch": 0.47, + "learning_rate": 1.5299739993408284e-07, + "loss": 0.697, + "step": 12885 + }, + { + "epoch": 0.47, + "learning_rate": 1.5299373786941076e-07, + "loss": 0.6891, + "step": 12886 + }, + { + "epoch": 0.47, + "learning_rate": 1.5299007580473872e-07, + "loss": 0.6829, + "step": 12887 + }, + { + "epoch": 0.47, + "learning_rate": 1.5298641374006664e-07, + "loss": 0.7302, + "step": 12888 + }, + { + "epoch": 0.47, + "learning_rate": 1.529827516753946e-07, + "loss": 0.6856, + "step": 12889 + }, + { + "epoch": 0.47, + "learning_rate": 1.5297908961072252e-07, + "loss": 0.6766, + "step": 12890 + }, + { + "epoch": 0.47, + "learning_rate": 1.5297542754605045e-07, + "loss": 0.6885, + "step": 12891 + }, + { + "epoch": 0.47, + "learning_rate": 1.529717654813784e-07, + "loss": 0.6882, + "step": 12892 + }, + { + "epoch": 0.47, + "learning_rate": 1.5296810341670633e-07, + "loss": 0.683, + "step": 12893 + }, + { + "epoch": 0.47, + "learning_rate": 1.5296444135203428e-07, + "loss": 0.691, + "step": 12894 + }, + { + "epoch": 0.47, + "learning_rate": 1.529607792873622e-07, + "loss": 0.68, + "step": 12895 + }, + { + "epoch": 0.47, + "learning_rate": 1.5295711722269014e-07, + "loss": 0.6812, + "step": 12896 + }, + { + "epoch": 0.47, + "learning_rate": 1.5295345515801806e-07, + "loss": 0.6851, + "step": 12897 + }, + { + "epoch": 0.47, + "learning_rate": 1.5294979309334602e-07, + "loss": 0.7057, + "step": 12898 + }, + { + "epoch": 0.47, + "learning_rate": 1.5294613102867394e-07, + "loss": 0.7155, + "step": 12899 + }, + { + "epoch": 0.47, + "learning_rate": 1.529424689640019e-07, + "loss": 0.6801, + "step": 12900 + }, + { + "epoch": 0.47, + "learning_rate": 1.5293880689932982e-07, + "loss": 0.7058, + "step": 12901 + }, + { + "epoch": 0.47, + "learning_rate": 1.5293514483465777e-07, + "loss": 0.6951, + "step": 12902 + }, + { + "epoch": 0.47, + "learning_rate": 1.529314827699857e-07, + "loss": 0.6891, + "step": 12903 + }, + { + "epoch": 0.47, + "learning_rate": 1.5292782070531365e-07, + "loss": 0.6836, + "step": 12904 + }, + { + "epoch": 0.47, + "learning_rate": 1.5292415864064158e-07, + "loss": 0.7054, + "step": 12905 + }, + { + "epoch": 0.47, + "learning_rate": 1.5292049657596953e-07, + "loss": 0.6915, + "step": 12906 + }, + { + "epoch": 0.47, + "learning_rate": 1.5291683451129746e-07, + "loss": 0.6895, + "step": 12907 + }, + { + "epoch": 0.47, + "learning_rate": 1.5291317244662541e-07, + "loss": 0.6815, + "step": 12908 + }, + { + "epoch": 0.47, + "learning_rate": 1.5290951038195334e-07, + "loss": 0.7061, + "step": 12909 + }, + { + "epoch": 0.47, + "learning_rate": 1.529058483172813e-07, + "loss": 0.6949, + "step": 12910 + }, + { + "epoch": 0.47, + "learning_rate": 1.5290218625260922e-07, + "loss": 0.6932, + "step": 12911 + }, + { + "epoch": 0.47, + "learning_rate": 1.5289852418793717e-07, + "loss": 0.6885, + "step": 12912 + }, + { + "epoch": 0.47, + "learning_rate": 1.528948621232651e-07, + "loss": 0.6732, + "step": 12913 + }, + { + "epoch": 0.47, + "learning_rate": 1.5289120005859303e-07, + "loss": 0.699, + "step": 12914 + }, + { + "epoch": 0.47, + "learning_rate": 1.5288753799392095e-07, + "loss": 0.6883, + "step": 12915 + }, + { + "epoch": 0.47, + "learning_rate": 1.528838759292489e-07, + "loss": 0.7026, + "step": 12916 + }, + { + "epoch": 0.47, + "learning_rate": 1.5288021386457683e-07, + "loss": 0.6954, + "step": 12917 + }, + { + "epoch": 0.47, + "learning_rate": 1.5287655179990476e-07, + "loss": 0.6955, + "step": 12918 + }, + { + "epoch": 0.47, + "learning_rate": 1.5287288973523271e-07, + "loss": 0.6935, + "step": 12919 + }, + { + "epoch": 0.47, + "learning_rate": 1.5286922767056064e-07, + "loss": 0.7007, + "step": 12920 + }, + { + "epoch": 0.47, + "learning_rate": 1.528655656058886e-07, + "loss": 0.6848, + "step": 12921 + }, + { + "epoch": 0.47, + "learning_rate": 1.5286190354121652e-07, + "loss": 0.6968, + "step": 12922 + }, + { + "epoch": 0.47, + "learning_rate": 1.5285824147654447e-07, + "loss": 0.6867, + "step": 12923 + }, + { + "epoch": 0.47, + "learning_rate": 1.528545794118724e-07, + "loss": 0.6794, + "step": 12924 + }, + { + "epoch": 0.47, + "learning_rate": 1.5285091734720035e-07, + "loss": 0.7006, + "step": 12925 + }, + { + "epoch": 0.47, + "learning_rate": 1.5284725528252828e-07, + "loss": 0.723, + "step": 12926 + }, + { + "epoch": 0.47, + "learning_rate": 1.5284359321785623e-07, + "loss": 0.6809, + "step": 12927 + }, + { + "epoch": 0.47, + "learning_rate": 1.5283993115318416e-07, + "loss": 0.6971, + "step": 12928 + }, + { + "epoch": 0.47, + "learning_rate": 1.5283626908851211e-07, + "loss": 0.6956, + "step": 12929 + }, + { + "epoch": 0.47, + "learning_rate": 1.5283260702384004e-07, + "loss": 0.7159, + "step": 12930 + }, + { + "epoch": 0.47, + "learning_rate": 1.52828944959168e-07, + "loss": 0.7134, + "step": 12931 + }, + { + "epoch": 0.47, + "learning_rate": 1.528252828944959e-07, + "loss": 0.7189, + "step": 12932 + }, + { + "epoch": 0.47, + "learning_rate": 1.5282162082982385e-07, + "loss": 0.6882, + "step": 12933 + }, + { + "epoch": 0.47, + "learning_rate": 1.5281795876515177e-07, + "loss": 0.7, + "step": 12934 + }, + { + "epoch": 0.47, + "learning_rate": 1.5281429670047973e-07, + "loss": 0.7042, + "step": 12935 + }, + { + "epoch": 0.47, + "learning_rate": 1.5281063463580765e-07, + "loss": 0.6855, + "step": 12936 + }, + { + "epoch": 0.47, + "learning_rate": 1.528069725711356e-07, + "loss": 0.6944, + "step": 12937 + }, + { + "epoch": 0.47, + "learning_rate": 1.5280331050646353e-07, + "loss": 0.6895, + "step": 12938 + }, + { + "epoch": 0.47, + "learning_rate": 1.5279964844179149e-07, + "loss": 0.6852, + "step": 12939 + }, + { + "epoch": 0.47, + "learning_rate": 1.527959863771194e-07, + "loss": 0.6876, + "step": 12940 + }, + { + "epoch": 0.47, + "learning_rate": 1.5279232431244734e-07, + "loss": 0.7057, + "step": 12941 + }, + { + "epoch": 0.47, + "learning_rate": 1.527886622477753e-07, + "loss": 0.6954, + "step": 12942 + }, + { + "epoch": 0.47, + "learning_rate": 1.5278500018310322e-07, + "loss": 0.6928, + "step": 12943 + }, + { + "epoch": 0.47, + "learning_rate": 1.5278133811843117e-07, + "loss": 0.6886, + "step": 12944 + }, + { + "epoch": 0.47, + "learning_rate": 1.527776760537591e-07, + "loss": 0.7245, + "step": 12945 + }, + { + "epoch": 0.47, + "learning_rate": 1.5277401398908705e-07, + "loss": 0.7054, + "step": 12946 + }, + { + "epoch": 0.47, + "learning_rate": 1.5277035192441498e-07, + "loss": 0.6784, + "step": 12947 + }, + { + "epoch": 0.47, + "learning_rate": 1.5276668985974293e-07, + "loss": 0.6867, + "step": 12948 + }, + { + "epoch": 0.47, + "learning_rate": 1.5276302779507086e-07, + "loss": 0.7034, + "step": 12949 + }, + { + "epoch": 0.47, + "learning_rate": 1.5275936573039879e-07, + "loss": 0.6999, + "step": 12950 + }, + { + "epoch": 0.47, + "learning_rate": 1.527557036657267e-07, + "loss": 0.7022, + "step": 12951 + }, + { + "epoch": 0.47, + "learning_rate": 1.5275204160105467e-07, + "loss": 0.7163, + "step": 12952 + }, + { + "epoch": 0.47, + "learning_rate": 1.527483795363826e-07, + "loss": 0.6787, + "step": 12953 + }, + { + "epoch": 0.47, + "learning_rate": 1.5274471747171055e-07, + "loss": 0.684, + "step": 12954 + }, + { + "epoch": 0.47, + "learning_rate": 1.5274105540703847e-07, + "loss": 0.6743, + "step": 12955 + }, + { + "epoch": 0.47, + "learning_rate": 1.5273739334236643e-07, + "loss": 0.6935, + "step": 12956 + }, + { + "epoch": 0.47, + "learning_rate": 1.5273373127769435e-07, + "loss": 0.6988, + "step": 12957 + }, + { + "epoch": 0.47, + "learning_rate": 1.527300692130223e-07, + "loss": 0.6863, + "step": 12958 + }, + { + "epoch": 0.47, + "learning_rate": 1.5272640714835023e-07, + "loss": 0.6835, + "step": 12959 + }, + { + "epoch": 0.47, + "learning_rate": 1.5272274508367818e-07, + "loss": 0.6917, + "step": 12960 + }, + { + "epoch": 0.47, + "learning_rate": 1.527190830190061e-07, + "loss": 0.6812, + "step": 12961 + }, + { + "epoch": 0.47, + "learning_rate": 1.5271542095433404e-07, + "loss": 0.6935, + "step": 12962 + }, + { + "epoch": 0.47, + "learning_rate": 1.52711758889662e-07, + "loss": 0.6828, + "step": 12963 + }, + { + "epoch": 0.47, + "learning_rate": 1.5270809682498992e-07, + "loss": 0.7015, + "step": 12964 + }, + { + "epoch": 0.47, + "learning_rate": 1.5270443476031787e-07, + "loss": 0.7073, + "step": 12965 + }, + { + "epoch": 0.47, + "learning_rate": 1.527007726956458e-07, + "loss": 0.6872, + "step": 12966 + }, + { + "epoch": 0.47, + "learning_rate": 1.5269711063097375e-07, + "loss": 0.6813, + "step": 12967 + }, + { + "epoch": 0.47, + "learning_rate": 1.5269344856630165e-07, + "loss": 0.6911, + "step": 12968 + }, + { + "epoch": 0.47, + "learning_rate": 1.526897865016296e-07, + "loss": 0.6918, + "step": 12969 + }, + { + "epoch": 0.47, + "learning_rate": 1.5268612443695753e-07, + "loss": 0.686, + "step": 12970 + }, + { + "epoch": 0.47, + "learning_rate": 1.5268246237228548e-07, + "loss": 0.6995, + "step": 12971 + }, + { + "epoch": 0.47, + "learning_rate": 1.526788003076134e-07, + "loss": 0.692, + "step": 12972 + }, + { + "epoch": 0.47, + "learning_rate": 1.5267513824294136e-07, + "loss": 0.7048, + "step": 12973 + }, + { + "epoch": 0.47, + "learning_rate": 1.526714761782693e-07, + "loss": 0.7019, + "step": 12974 + }, + { + "epoch": 0.47, + "learning_rate": 1.5266781411359724e-07, + "loss": 0.6824, + "step": 12975 + }, + { + "epoch": 0.47, + "learning_rate": 1.5266415204892517e-07, + "loss": 0.692, + "step": 12976 + }, + { + "epoch": 0.47, + "learning_rate": 1.5266048998425312e-07, + "loss": 0.7004, + "step": 12977 + }, + { + "epoch": 0.47, + "learning_rate": 1.5265682791958105e-07, + "loss": 0.6878, + "step": 12978 + }, + { + "epoch": 0.47, + "learning_rate": 1.52653165854909e-07, + "loss": 0.6984, + "step": 12979 + }, + { + "epoch": 0.47, + "learning_rate": 1.5264950379023693e-07, + "loss": 0.6952, + "step": 12980 + }, + { + "epoch": 0.47, + "learning_rate": 1.5264584172556488e-07, + "loss": 0.7012, + "step": 12981 + }, + { + "epoch": 0.47, + "learning_rate": 1.526421796608928e-07, + "loss": 0.6861, + "step": 12982 + }, + { + "epoch": 0.48, + "learning_rate": 1.5263851759622076e-07, + "loss": 0.6886, + "step": 12983 + }, + { + "epoch": 0.48, + "learning_rate": 1.526348555315487e-07, + "loss": 0.6864, + "step": 12984 + }, + { + "epoch": 0.48, + "learning_rate": 1.5263119346687662e-07, + "loss": 0.6998, + "step": 12985 + }, + { + "epoch": 0.48, + "learning_rate": 1.5262753140220454e-07, + "loss": 0.6832, + "step": 12986 + }, + { + "epoch": 0.48, + "learning_rate": 1.526238693375325e-07, + "loss": 0.6803, + "step": 12987 + }, + { + "epoch": 0.48, + "learning_rate": 1.5262020727286042e-07, + "loss": 0.6939, + "step": 12988 + }, + { + "epoch": 0.48, + "learning_rate": 1.5261654520818835e-07, + "loss": 0.6828, + "step": 12989 + }, + { + "epoch": 0.48, + "learning_rate": 1.526128831435163e-07, + "loss": 0.6852, + "step": 12990 + }, + { + "epoch": 0.48, + "learning_rate": 1.5260922107884423e-07, + "loss": 0.696, + "step": 12991 + }, + { + "epoch": 0.48, + "learning_rate": 1.5260555901417218e-07, + "loss": 0.6898, + "step": 12992 + }, + { + "epoch": 0.48, + "learning_rate": 1.526018969495001e-07, + "loss": 0.6813, + "step": 12993 + }, + { + "epoch": 0.48, + "learning_rate": 1.5259823488482806e-07, + "loss": 0.7051, + "step": 12994 + }, + { + "epoch": 0.48, + "learning_rate": 1.52594572820156e-07, + "loss": 0.6852, + "step": 12995 + }, + { + "epoch": 0.48, + "learning_rate": 1.5259091075548394e-07, + "loss": 0.6978, + "step": 12996 + }, + { + "epoch": 0.48, + "learning_rate": 1.5258724869081187e-07, + "loss": 0.6902, + "step": 12997 + }, + { + "epoch": 0.48, + "learning_rate": 1.5258358662613982e-07, + "loss": 0.7029, + "step": 12998 + }, + { + "epoch": 0.48, + "learning_rate": 1.5257992456146775e-07, + "loss": 0.6889, + "step": 12999 + }, + { + "epoch": 0.48, + "learning_rate": 1.525762624967957e-07, + "loss": 0.6842, + "step": 13000 + }, + { + "epoch": 0.48, + "learning_rate": 1.5257260043212363e-07, + "loss": 0.6955, + "step": 13001 + }, + { + "epoch": 0.48, + "learning_rate": 1.5256893836745158e-07, + "loss": 0.6914, + "step": 13002 + }, + { + "epoch": 0.48, + "learning_rate": 1.525652763027795e-07, + "loss": 0.7026, + "step": 13003 + }, + { + "epoch": 0.48, + "learning_rate": 1.5256161423810744e-07, + "loss": 0.7059, + "step": 13004 + }, + { + "epoch": 0.48, + "learning_rate": 1.5255795217343536e-07, + "loss": 0.7024, + "step": 13005 + }, + { + "epoch": 0.48, + "learning_rate": 1.5255429010876332e-07, + "loss": 0.6851, + "step": 13006 + }, + { + "epoch": 0.48, + "learning_rate": 1.5255062804409124e-07, + "loss": 0.7109, + "step": 13007 + }, + { + "epoch": 0.48, + "learning_rate": 1.525469659794192e-07, + "loss": 0.6852, + "step": 13008 + }, + { + "epoch": 0.48, + "learning_rate": 1.5254330391474712e-07, + "loss": 0.703, + "step": 13009 + }, + { + "epoch": 0.48, + "learning_rate": 1.5253964185007508e-07, + "loss": 0.6954, + "step": 13010 + }, + { + "epoch": 0.48, + "learning_rate": 1.52535979785403e-07, + "loss": 0.6874, + "step": 13011 + }, + { + "epoch": 0.48, + "learning_rate": 1.5253231772073093e-07, + "loss": 0.6947, + "step": 13012 + }, + { + "epoch": 0.48, + "learning_rate": 1.5252865565605888e-07, + "loss": 0.6914, + "step": 13013 + }, + { + "epoch": 0.48, + "learning_rate": 1.525249935913868e-07, + "loss": 0.6924, + "step": 13014 + }, + { + "epoch": 0.48, + "learning_rate": 1.5252133152671476e-07, + "loss": 0.6875, + "step": 13015 + }, + { + "epoch": 0.48, + "learning_rate": 1.525176694620427e-07, + "loss": 0.6912, + "step": 13016 + }, + { + "epoch": 0.48, + "learning_rate": 1.5251400739737064e-07, + "loss": 0.6939, + "step": 13017 + }, + { + "epoch": 0.48, + "learning_rate": 1.5251034533269857e-07, + "loss": 0.6945, + "step": 13018 + }, + { + "epoch": 0.48, + "learning_rate": 1.5250668326802652e-07, + "loss": 0.7062, + "step": 13019 + }, + { + "epoch": 0.48, + "learning_rate": 1.5250302120335445e-07, + "loss": 0.6936, + "step": 13020 + }, + { + "epoch": 0.48, + "learning_rate": 1.524993591386824e-07, + "loss": 0.6967, + "step": 13021 + }, + { + "epoch": 0.48, + "learning_rate": 1.524956970740103e-07, + "loss": 0.6887, + "step": 13022 + }, + { + "epoch": 0.48, + "learning_rate": 1.5249203500933825e-07, + "loss": 0.6874, + "step": 13023 + }, + { + "epoch": 0.48, + "learning_rate": 1.5248837294466618e-07, + "loss": 0.6854, + "step": 13024 + }, + { + "epoch": 0.48, + "learning_rate": 1.5248471087999413e-07, + "loss": 0.687, + "step": 13025 + }, + { + "epoch": 0.48, + "learning_rate": 1.5248104881532206e-07, + "loss": 0.684, + "step": 13026 + }, + { + "epoch": 0.48, + "learning_rate": 1.5247738675065001e-07, + "loss": 0.6939, + "step": 13027 + }, + { + "epoch": 0.48, + "learning_rate": 1.5247372468597794e-07, + "loss": 0.7004, + "step": 13028 + }, + { + "epoch": 0.48, + "learning_rate": 1.524700626213059e-07, + "loss": 0.6912, + "step": 13029 + }, + { + "epoch": 0.48, + "learning_rate": 1.5246640055663382e-07, + "loss": 0.695, + "step": 13030 + }, + { + "epoch": 0.48, + "learning_rate": 1.5246273849196177e-07, + "loss": 0.6956, + "step": 13031 + }, + { + "epoch": 0.48, + "learning_rate": 1.524590764272897e-07, + "loss": 0.6999, + "step": 13032 + }, + { + "epoch": 0.48, + "learning_rate": 1.5245541436261763e-07, + "loss": 0.6931, + "step": 13033 + }, + { + "epoch": 0.48, + "learning_rate": 1.5245175229794558e-07, + "loss": 0.7052, + "step": 13034 + }, + { + "epoch": 0.48, + "learning_rate": 1.524480902332735e-07, + "loss": 0.6973, + "step": 13035 + }, + { + "epoch": 0.48, + "learning_rate": 1.5244442816860146e-07, + "loss": 0.6835, + "step": 13036 + }, + { + "epoch": 0.48, + "learning_rate": 1.524407661039294e-07, + "loss": 0.6947, + "step": 13037 + }, + { + "epoch": 0.48, + "learning_rate": 1.5243710403925734e-07, + "loss": 0.7067, + "step": 13038 + }, + { + "epoch": 0.48, + "learning_rate": 1.5243344197458527e-07, + "loss": 0.6887, + "step": 13039 + }, + { + "epoch": 0.48, + "learning_rate": 1.524297799099132e-07, + "loss": 0.7019, + "step": 13040 + }, + { + "epoch": 0.48, + "learning_rate": 1.5242611784524112e-07, + "loss": 0.6944, + "step": 13041 + }, + { + "epoch": 0.48, + "learning_rate": 1.5242245578056907e-07, + "loss": 0.7085, + "step": 13042 + }, + { + "epoch": 0.48, + "learning_rate": 1.52418793715897e-07, + "loss": 0.7045, + "step": 13043 + }, + { + "epoch": 0.48, + "learning_rate": 1.5241513165122495e-07, + "loss": 0.6826, + "step": 13044 + }, + { + "epoch": 0.48, + "learning_rate": 1.5241146958655288e-07, + "loss": 0.6847, + "step": 13045 + }, + { + "epoch": 0.48, + "learning_rate": 1.5240780752188083e-07, + "loss": 0.6963, + "step": 13046 + }, + { + "epoch": 0.48, + "learning_rate": 1.5240414545720876e-07, + "loss": 0.6922, + "step": 13047 + }, + { + "epoch": 0.48, + "learning_rate": 1.5240048339253671e-07, + "loss": 0.6984, + "step": 13048 + }, + { + "epoch": 0.48, + "learning_rate": 1.5239682132786464e-07, + "loss": 0.6839, + "step": 13049 + }, + { + "epoch": 0.48, + "learning_rate": 1.523931592631926e-07, + "loss": 0.6876, + "step": 13050 + }, + { + "epoch": 0.48, + "learning_rate": 1.5238949719852052e-07, + "loss": 0.706, + "step": 13051 + }, + { + "epoch": 0.48, + "learning_rate": 1.5238583513384847e-07, + "loss": 0.6991, + "step": 13052 + }, + { + "epoch": 0.48, + "learning_rate": 1.523821730691764e-07, + "loss": 0.6978, + "step": 13053 + }, + { + "epoch": 0.48, + "learning_rate": 1.5237851100450435e-07, + "loss": 0.6777, + "step": 13054 + }, + { + "epoch": 0.48, + "learning_rate": 1.5237484893983228e-07, + "loss": 0.6897, + "step": 13055 + }, + { + "epoch": 0.48, + "learning_rate": 1.523711868751602e-07, + "loss": 0.6864, + "step": 13056 + }, + { + "epoch": 0.48, + "learning_rate": 1.5236752481048816e-07, + "loss": 0.7007, + "step": 13057 + }, + { + "epoch": 0.48, + "learning_rate": 1.5236386274581609e-07, + "loss": 0.6829, + "step": 13058 + }, + { + "epoch": 0.48, + "learning_rate": 1.52360200681144e-07, + "loss": 0.6876, + "step": 13059 + }, + { + "epoch": 0.48, + "learning_rate": 1.5235653861647194e-07, + "loss": 0.6989, + "step": 13060 + }, + { + "epoch": 0.48, + "learning_rate": 1.523528765517999e-07, + "loss": 0.7043, + "step": 13061 + }, + { + "epoch": 0.48, + "learning_rate": 1.5234921448712782e-07, + "loss": 0.7064, + "step": 13062 + }, + { + "epoch": 0.48, + "learning_rate": 1.5234555242245577e-07, + "loss": 0.682, + "step": 13063 + }, + { + "epoch": 0.48, + "learning_rate": 1.523418903577837e-07, + "loss": 0.7029, + "step": 13064 + }, + { + "epoch": 0.48, + "learning_rate": 1.5233822829311165e-07, + "loss": 0.6772, + "step": 13065 + }, + { + "epoch": 0.48, + "learning_rate": 1.5233456622843958e-07, + "loss": 0.6833, + "step": 13066 + }, + { + "epoch": 0.48, + "learning_rate": 1.5233090416376753e-07, + "loss": 0.6868, + "step": 13067 + }, + { + "epoch": 0.48, + "learning_rate": 1.5232724209909546e-07, + "loss": 0.689, + "step": 13068 + }, + { + "epoch": 0.48, + "learning_rate": 1.523235800344234e-07, + "loss": 0.707, + "step": 13069 + }, + { + "epoch": 0.48, + "learning_rate": 1.5231991796975134e-07, + "loss": 0.681, + "step": 13070 + }, + { + "epoch": 0.48, + "learning_rate": 1.523162559050793e-07, + "loss": 0.6893, + "step": 13071 + }, + { + "epoch": 0.48, + "learning_rate": 1.5231259384040722e-07, + "loss": 0.701, + "step": 13072 + }, + { + "epoch": 0.48, + "learning_rate": 1.5230893177573517e-07, + "loss": 0.679, + "step": 13073 + }, + { + "epoch": 0.48, + "learning_rate": 1.523052697110631e-07, + "loss": 0.6929, + "step": 13074 + }, + { + "epoch": 0.48, + "learning_rate": 1.5230160764639105e-07, + "loss": 0.6961, + "step": 13075 + }, + { + "epoch": 0.48, + "learning_rate": 1.5229794558171895e-07, + "loss": 0.6957, + "step": 13076 + }, + { + "epoch": 0.48, + "learning_rate": 1.522942835170469e-07, + "loss": 0.6839, + "step": 13077 + }, + { + "epoch": 0.48, + "learning_rate": 1.5229062145237483e-07, + "loss": 0.6907, + "step": 13078 + }, + { + "epoch": 0.48, + "learning_rate": 1.5228695938770278e-07, + "loss": 0.6921, + "step": 13079 + }, + { + "epoch": 0.48, + "learning_rate": 1.522832973230307e-07, + "loss": 0.6923, + "step": 13080 + }, + { + "epoch": 0.48, + "learning_rate": 1.5227963525835866e-07, + "loss": 0.6868, + "step": 13081 + }, + { + "epoch": 0.48, + "learning_rate": 1.522759731936866e-07, + "loss": 0.6996, + "step": 13082 + }, + { + "epoch": 0.48, + "learning_rate": 1.5227231112901452e-07, + "loss": 0.6833, + "step": 13083 + }, + { + "epoch": 0.48, + "learning_rate": 1.5226864906434247e-07, + "loss": 0.6867, + "step": 13084 + }, + { + "epoch": 0.48, + "learning_rate": 1.522649869996704e-07, + "loss": 0.6871, + "step": 13085 + }, + { + "epoch": 0.48, + "learning_rate": 1.5226132493499835e-07, + "loss": 0.6799, + "step": 13086 + }, + { + "epoch": 0.48, + "learning_rate": 1.5225766287032628e-07, + "loss": 0.6882, + "step": 13087 + }, + { + "epoch": 0.48, + "learning_rate": 1.5225400080565423e-07, + "loss": 0.6946, + "step": 13088 + }, + { + "epoch": 0.48, + "learning_rate": 1.5225033874098216e-07, + "loss": 0.6985, + "step": 13089 + }, + { + "epoch": 0.48, + "learning_rate": 1.522466766763101e-07, + "loss": 0.6968, + "step": 13090 + }, + { + "epoch": 0.48, + "learning_rate": 1.5224301461163804e-07, + "loss": 0.6861, + "step": 13091 + }, + { + "epoch": 0.48, + "learning_rate": 1.52239352546966e-07, + "loss": 0.6837, + "step": 13092 + }, + { + "epoch": 0.48, + "learning_rate": 1.5223569048229392e-07, + "loss": 0.6932, + "step": 13093 + }, + { + "epoch": 0.48, + "learning_rate": 1.5223202841762184e-07, + "loss": 0.6775, + "step": 13094 + }, + { + "epoch": 0.48, + "learning_rate": 1.5222836635294977e-07, + "loss": 0.6808, + "step": 13095 + }, + { + "epoch": 0.48, + "learning_rate": 1.5222470428827772e-07, + "loss": 0.6967, + "step": 13096 + }, + { + "epoch": 0.48, + "learning_rate": 1.5222104222360565e-07, + "loss": 0.6842, + "step": 13097 + }, + { + "epoch": 0.48, + "learning_rate": 1.522173801589336e-07, + "loss": 0.7146, + "step": 13098 + }, + { + "epoch": 0.48, + "learning_rate": 1.5221371809426153e-07, + "loss": 0.6906, + "step": 13099 + }, + { + "epoch": 0.48, + "learning_rate": 1.5221005602958948e-07, + "loss": 0.6943, + "step": 13100 + }, + { + "epoch": 0.48, + "learning_rate": 1.522063939649174e-07, + "loss": 0.6894, + "step": 13101 + }, + { + "epoch": 0.48, + "learning_rate": 1.5220273190024536e-07, + "loss": 0.6852, + "step": 13102 + }, + { + "epoch": 0.48, + "learning_rate": 1.521990698355733e-07, + "loss": 0.698, + "step": 13103 + }, + { + "epoch": 0.48, + "learning_rate": 1.5219540777090122e-07, + "loss": 0.7025, + "step": 13104 + }, + { + "epoch": 0.48, + "learning_rate": 1.5219174570622917e-07, + "loss": 0.6952, + "step": 13105 + }, + { + "epoch": 0.48, + "learning_rate": 1.521880836415571e-07, + "loss": 0.6979, + "step": 13106 + }, + { + "epoch": 0.48, + "learning_rate": 1.5218442157688505e-07, + "loss": 0.7005, + "step": 13107 + }, + { + "epoch": 0.48, + "learning_rate": 1.5218075951221298e-07, + "loss": 0.7101, + "step": 13108 + }, + { + "epoch": 0.48, + "learning_rate": 1.5217709744754093e-07, + "loss": 0.6989, + "step": 13109 + }, + { + "epoch": 0.48, + "learning_rate": 1.5217343538286886e-07, + "loss": 0.7096, + "step": 13110 + }, + { + "epoch": 0.48, + "learning_rate": 1.521697733181968e-07, + "loss": 0.6916, + "step": 13111 + }, + { + "epoch": 0.48, + "learning_rate": 1.521661112535247e-07, + "loss": 0.6866, + "step": 13112 + }, + { + "epoch": 0.48, + "learning_rate": 1.5216244918885266e-07, + "loss": 0.7004, + "step": 13113 + }, + { + "epoch": 0.48, + "learning_rate": 1.521587871241806e-07, + "loss": 0.7018, + "step": 13114 + }, + { + "epoch": 0.48, + "learning_rate": 1.5215512505950854e-07, + "loss": 0.6829, + "step": 13115 + }, + { + "epoch": 0.48, + "learning_rate": 1.5215146299483647e-07, + "loss": 0.6982, + "step": 13116 + }, + { + "epoch": 0.48, + "learning_rate": 1.5214780093016442e-07, + "loss": 0.7039, + "step": 13117 + }, + { + "epoch": 0.48, + "learning_rate": 1.5214413886549235e-07, + "loss": 0.6894, + "step": 13118 + }, + { + "epoch": 0.48, + "learning_rate": 1.521404768008203e-07, + "loss": 0.6931, + "step": 13119 + }, + { + "epoch": 0.48, + "learning_rate": 1.5213681473614823e-07, + "loss": 0.6966, + "step": 13120 + }, + { + "epoch": 0.48, + "learning_rate": 1.5213315267147618e-07, + "loss": 0.7052, + "step": 13121 + }, + { + "epoch": 0.48, + "learning_rate": 1.521294906068041e-07, + "loss": 0.7027, + "step": 13122 + }, + { + "epoch": 0.48, + "learning_rate": 1.5212582854213206e-07, + "loss": 0.6861, + "step": 13123 + }, + { + "epoch": 0.48, + "learning_rate": 1.5212216647746e-07, + "loss": 0.6973, + "step": 13124 + }, + { + "epoch": 0.48, + "learning_rate": 1.5211850441278794e-07, + "loss": 0.68, + "step": 13125 + }, + { + "epoch": 0.48, + "learning_rate": 1.5211484234811587e-07, + "loss": 0.6814, + "step": 13126 + }, + { + "epoch": 0.48, + "learning_rate": 1.521111802834438e-07, + "loss": 0.6994, + "step": 13127 + }, + { + "epoch": 0.48, + "learning_rate": 1.5210751821877175e-07, + "loss": 0.6919, + "step": 13128 + }, + { + "epoch": 0.48, + "learning_rate": 1.5210385615409968e-07, + "loss": 0.6887, + "step": 13129 + }, + { + "epoch": 0.48, + "learning_rate": 1.521001940894276e-07, + "loss": 0.6777, + "step": 13130 + }, + { + "epoch": 0.48, + "learning_rate": 1.5209653202475553e-07, + "loss": 0.6883, + "step": 13131 + }, + { + "epoch": 0.48, + "learning_rate": 1.5209286996008348e-07, + "loss": 0.6894, + "step": 13132 + }, + { + "epoch": 0.48, + "learning_rate": 1.520892078954114e-07, + "loss": 0.7136, + "step": 13133 + }, + { + "epoch": 0.48, + "learning_rate": 1.5208554583073936e-07, + "loss": 0.6898, + "step": 13134 + }, + { + "epoch": 0.48, + "learning_rate": 1.520818837660673e-07, + "loss": 0.707, + "step": 13135 + }, + { + "epoch": 0.48, + "learning_rate": 1.5207822170139524e-07, + "loss": 0.6874, + "step": 13136 + }, + { + "epoch": 0.48, + "learning_rate": 1.5207455963672317e-07, + "loss": 0.6837, + "step": 13137 + }, + { + "epoch": 0.48, + "learning_rate": 1.5207089757205112e-07, + "loss": 0.7025, + "step": 13138 + }, + { + "epoch": 0.48, + "learning_rate": 1.5206723550737905e-07, + "loss": 0.7029, + "step": 13139 + }, + { + "epoch": 0.48, + "learning_rate": 1.52063573442707e-07, + "loss": 0.6877, + "step": 13140 + }, + { + "epoch": 0.48, + "learning_rate": 1.5205991137803493e-07, + "loss": 0.7072, + "step": 13141 + }, + { + "epoch": 0.48, + "learning_rate": 1.5205624931336288e-07, + "loss": 0.6908, + "step": 13142 + }, + { + "epoch": 0.48, + "learning_rate": 1.520525872486908e-07, + "loss": 0.6938, + "step": 13143 + }, + { + "epoch": 0.48, + "learning_rate": 1.5204892518401876e-07, + "loss": 0.6906, + "step": 13144 + }, + { + "epoch": 0.48, + "learning_rate": 1.520452631193467e-07, + "loss": 0.6941, + "step": 13145 + }, + { + "epoch": 0.48, + "learning_rate": 1.5204160105467464e-07, + "loss": 0.7008, + "step": 13146 + }, + { + "epoch": 0.48, + "learning_rate": 1.5203793899000254e-07, + "loss": 0.6798, + "step": 13147 + }, + { + "epoch": 0.48, + "learning_rate": 1.520342769253305e-07, + "loss": 0.7099, + "step": 13148 + }, + { + "epoch": 0.48, + "learning_rate": 1.5203061486065842e-07, + "loss": 0.6858, + "step": 13149 + }, + { + "epoch": 0.48, + "learning_rate": 1.5202695279598637e-07, + "loss": 0.7029, + "step": 13150 + }, + { + "epoch": 0.48, + "learning_rate": 1.520232907313143e-07, + "loss": 0.6842, + "step": 13151 + }, + { + "epoch": 0.48, + "learning_rate": 1.5201962866664225e-07, + "loss": 0.7075, + "step": 13152 + }, + { + "epoch": 0.48, + "learning_rate": 1.5201596660197018e-07, + "loss": 0.6951, + "step": 13153 + }, + { + "epoch": 0.48, + "learning_rate": 1.520123045372981e-07, + "loss": 0.6851, + "step": 13154 + }, + { + "epoch": 0.48, + "learning_rate": 1.5200864247262606e-07, + "loss": 0.6764, + "step": 13155 + }, + { + "epoch": 0.48, + "learning_rate": 1.52004980407954e-07, + "loss": 0.6985, + "step": 13156 + }, + { + "epoch": 0.48, + "learning_rate": 1.5200131834328194e-07, + "loss": 0.6806, + "step": 13157 + }, + { + "epoch": 0.48, + "learning_rate": 1.5199765627860987e-07, + "loss": 0.6907, + "step": 13158 + }, + { + "epoch": 0.48, + "learning_rate": 1.5199399421393782e-07, + "loss": 0.6875, + "step": 13159 + }, + { + "epoch": 0.48, + "learning_rate": 1.5199033214926575e-07, + "loss": 0.6882, + "step": 13160 + }, + { + "epoch": 0.48, + "learning_rate": 1.519866700845937e-07, + "loss": 0.6828, + "step": 13161 + }, + { + "epoch": 0.48, + "learning_rate": 1.5198300801992163e-07, + "loss": 0.6968, + "step": 13162 + }, + { + "epoch": 0.48, + "learning_rate": 1.5197934595524958e-07, + "loss": 0.6821, + "step": 13163 + }, + { + "epoch": 0.48, + "learning_rate": 1.519756838905775e-07, + "loss": 0.7137, + "step": 13164 + }, + { + "epoch": 0.48, + "learning_rate": 1.5197202182590543e-07, + "loss": 0.6825, + "step": 13165 + }, + { + "epoch": 0.48, + "learning_rate": 1.5196835976123336e-07, + "loss": 0.6977, + "step": 13166 + }, + { + "epoch": 0.48, + "learning_rate": 1.5196469769656131e-07, + "loss": 0.725, + "step": 13167 + }, + { + "epoch": 0.48, + "learning_rate": 1.5196103563188924e-07, + "loss": 0.7063, + "step": 13168 + }, + { + "epoch": 0.48, + "learning_rate": 1.519573735672172e-07, + "loss": 0.6847, + "step": 13169 + }, + { + "epoch": 0.48, + "learning_rate": 1.5195371150254512e-07, + "loss": 0.7008, + "step": 13170 + }, + { + "epoch": 0.48, + "learning_rate": 1.5195004943787307e-07, + "loss": 0.6928, + "step": 13171 + }, + { + "epoch": 0.48, + "learning_rate": 1.51946387373201e-07, + "loss": 0.6991, + "step": 13172 + }, + { + "epoch": 0.48, + "learning_rate": 1.5194272530852895e-07, + "loss": 0.6958, + "step": 13173 + }, + { + "epoch": 0.48, + "learning_rate": 1.5193906324385688e-07, + "loss": 0.6856, + "step": 13174 + }, + { + "epoch": 0.48, + "learning_rate": 1.519354011791848e-07, + "loss": 0.7001, + "step": 13175 + }, + { + "epoch": 0.48, + "learning_rate": 1.5193173911451276e-07, + "loss": 0.6939, + "step": 13176 + }, + { + "epoch": 0.48, + "learning_rate": 1.5192807704984069e-07, + "loss": 0.6905, + "step": 13177 + }, + { + "epoch": 0.48, + "learning_rate": 1.5192441498516864e-07, + "loss": 0.6876, + "step": 13178 + }, + { + "epoch": 0.48, + "learning_rate": 1.5192075292049657e-07, + "loss": 0.6804, + "step": 13179 + }, + { + "epoch": 0.48, + "learning_rate": 1.5191709085582452e-07, + "loss": 0.6946, + "step": 13180 + }, + { + "epoch": 0.48, + "learning_rate": 1.5191342879115245e-07, + "loss": 0.6906, + "step": 13181 + }, + { + "epoch": 0.48, + "learning_rate": 1.519097667264804e-07, + "loss": 0.6886, + "step": 13182 + }, + { + "epoch": 0.48, + "learning_rate": 1.519061046618083e-07, + "loss": 0.704, + "step": 13183 + }, + { + "epoch": 0.48, + "learning_rate": 1.5190244259713625e-07, + "loss": 0.7037, + "step": 13184 + }, + { + "epoch": 0.48, + "learning_rate": 1.5189878053246418e-07, + "loss": 0.7021, + "step": 13185 + }, + { + "epoch": 0.48, + "learning_rate": 1.5189511846779213e-07, + "loss": 0.6989, + "step": 13186 + }, + { + "epoch": 0.48, + "learning_rate": 1.5189145640312006e-07, + "loss": 0.6818, + "step": 13187 + }, + { + "epoch": 0.48, + "learning_rate": 1.51887794338448e-07, + "loss": 0.676, + "step": 13188 + }, + { + "epoch": 0.48, + "learning_rate": 1.5188413227377594e-07, + "loss": 0.6988, + "step": 13189 + }, + { + "epoch": 0.48, + "learning_rate": 1.518804702091039e-07, + "loss": 0.6941, + "step": 13190 + }, + { + "epoch": 0.48, + "learning_rate": 1.5187680814443182e-07, + "loss": 0.7002, + "step": 13191 + }, + { + "epoch": 0.48, + "learning_rate": 1.5187314607975977e-07, + "loss": 0.7014, + "step": 13192 + }, + { + "epoch": 0.48, + "learning_rate": 1.518694840150877e-07, + "loss": 0.6859, + "step": 13193 + }, + { + "epoch": 0.48, + "learning_rate": 1.5186582195041565e-07, + "loss": 0.6949, + "step": 13194 + }, + { + "epoch": 0.48, + "learning_rate": 1.5186215988574358e-07, + "loss": 0.6859, + "step": 13195 + }, + { + "epoch": 0.48, + "learning_rate": 1.5185849782107153e-07, + "loss": 0.697, + "step": 13196 + }, + { + "epoch": 0.48, + "learning_rate": 1.5185483575639946e-07, + "loss": 0.6876, + "step": 13197 + }, + { + "epoch": 0.48, + "learning_rate": 1.5185117369172738e-07, + "loss": 0.7056, + "step": 13198 + }, + { + "epoch": 0.48, + "learning_rate": 1.5184751162705534e-07, + "loss": 0.6829, + "step": 13199 + }, + { + "epoch": 0.48, + "learning_rate": 1.5184384956238326e-07, + "loss": 0.6922, + "step": 13200 + }, + { + "epoch": 0.48, + "learning_rate": 1.518401874977112e-07, + "loss": 0.6915, + "step": 13201 + }, + { + "epoch": 0.48, + "learning_rate": 1.5183652543303912e-07, + "loss": 0.6964, + "step": 13202 + }, + { + "epoch": 0.48, + "learning_rate": 1.5183286336836707e-07, + "loss": 0.7035, + "step": 13203 + }, + { + "epoch": 0.48, + "learning_rate": 1.51829201303695e-07, + "loss": 0.6726, + "step": 13204 + }, + { + "epoch": 0.48, + "learning_rate": 1.5182553923902295e-07, + "loss": 0.7022, + "step": 13205 + }, + { + "epoch": 0.48, + "learning_rate": 1.5182187717435088e-07, + "loss": 0.7171, + "step": 13206 + }, + { + "epoch": 0.48, + "learning_rate": 1.5181821510967883e-07, + "loss": 0.7059, + "step": 13207 + }, + { + "epoch": 0.48, + "learning_rate": 1.5181455304500676e-07, + "loss": 0.6857, + "step": 13208 + }, + { + "epoch": 0.48, + "learning_rate": 1.518108909803347e-07, + "loss": 0.6981, + "step": 13209 + }, + { + "epoch": 0.48, + "learning_rate": 1.5180722891566264e-07, + "loss": 0.7038, + "step": 13210 + }, + { + "epoch": 0.48, + "learning_rate": 1.518035668509906e-07, + "loss": 0.7055, + "step": 13211 + }, + { + "epoch": 0.48, + "learning_rate": 1.5179990478631852e-07, + "loss": 0.6882, + "step": 13212 + }, + { + "epoch": 0.48, + "learning_rate": 1.5179624272164647e-07, + "loss": 0.6863, + "step": 13213 + }, + { + "epoch": 0.48, + "learning_rate": 1.517925806569744e-07, + "loss": 0.6828, + "step": 13214 + }, + { + "epoch": 0.48, + "learning_rate": 1.5178891859230235e-07, + "loss": 0.6991, + "step": 13215 + }, + { + "epoch": 0.48, + "learning_rate": 1.5178525652763028e-07, + "loss": 0.6942, + "step": 13216 + }, + { + "epoch": 0.48, + "learning_rate": 1.5178159446295823e-07, + "loss": 0.6842, + "step": 13217 + }, + { + "epoch": 0.48, + "learning_rate": 1.5177793239828616e-07, + "loss": 0.7016, + "step": 13218 + }, + { + "epoch": 0.48, + "learning_rate": 1.5177427033361408e-07, + "loss": 0.6855, + "step": 13219 + }, + { + "epoch": 0.48, + "learning_rate": 1.51770608268942e-07, + "loss": 0.6793, + "step": 13220 + }, + { + "epoch": 0.48, + "learning_rate": 1.5176694620426996e-07, + "loss": 0.6908, + "step": 13221 + }, + { + "epoch": 0.48, + "learning_rate": 1.517632841395979e-07, + "loss": 0.6967, + "step": 13222 + }, + { + "epoch": 0.48, + "learning_rate": 1.5175962207492584e-07, + "loss": 0.7022, + "step": 13223 + }, + { + "epoch": 0.48, + "learning_rate": 1.5175596001025377e-07, + "loss": 0.6943, + "step": 13224 + }, + { + "epoch": 0.48, + "learning_rate": 1.517522979455817e-07, + "loss": 0.7003, + "step": 13225 + }, + { + "epoch": 0.48, + "learning_rate": 1.5174863588090965e-07, + "loss": 0.6967, + "step": 13226 + }, + { + "epoch": 0.48, + "learning_rate": 1.5174497381623758e-07, + "loss": 0.6963, + "step": 13227 + }, + { + "epoch": 0.48, + "learning_rate": 1.5174131175156553e-07, + "loss": 0.6923, + "step": 13228 + }, + { + "epoch": 0.48, + "learning_rate": 1.5173764968689346e-07, + "loss": 0.6845, + "step": 13229 + }, + { + "epoch": 0.48, + "learning_rate": 1.517339876222214e-07, + "loss": 0.6985, + "step": 13230 + }, + { + "epoch": 0.48, + "learning_rate": 1.5173032555754934e-07, + "loss": 0.7081, + "step": 13231 + }, + { + "epoch": 0.48, + "learning_rate": 1.517266634928773e-07, + "loss": 0.6985, + "step": 13232 + }, + { + "epoch": 0.48, + "learning_rate": 1.5172300142820522e-07, + "loss": 0.6825, + "step": 13233 + }, + { + "epoch": 0.48, + "learning_rate": 1.5171933936353317e-07, + "loss": 0.6972, + "step": 13234 + }, + { + "epoch": 0.48, + "learning_rate": 1.517156772988611e-07, + "loss": 0.6798, + "step": 13235 + }, + { + "epoch": 0.48, + "learning_rate": 1.5171201523418905e-07, + "loss": 0.6932, + "step": 13236 + }, + { + "epoch": 0.48, + "learning_rate": 1.5170835316951695e-07, + "loss": 0.6794, + "step": 13237 + }, + { + "epoch": 0.48, + "learning_rate": 1.517046911048449e-07, + "loss": 0.6995, + "step": 13238 + }, + { + "epoch": 0.48, + "learning_rate": 1.5170102904017283e-07, + "loss": 0.7072, + "step": 13239 + }, + { + "epoch": 0.48, + "learning_rate": 1.5169736697550078e-07, + "loss": 0.6983, + "step": 13240 + }, + { + "epoch": 0.48, + "learning_rate": 1.516937049108287e-07, + "loss": 0.6921, + "step": 13241 + }, + { + "epoch": 0.48, + "learning_rate": 1.5169004284615666e-07, + "loss": 0.6883, + "step": 13242 + }, + { + "epoch": 0.48, + "learning_rate": 1.516863807814846e-07, + "loss": 0.697, + "step": 13243 + }, + { + "epoch": 0.48, + "learning_rate": 1.5168271871681254e-07, + "loss": 0.7057, + "step": 13244 + }, + { + "epoch": 0.48, + "learning_rate": 1.5167905665214047e-07, + "loss": 0.6992, + "step": 13245 + }, + { + "epoch": 0.48, + "learning_rate": 1.516753945874684e-07, + "loss": 0.7037, + "step": 13246 + }, + { + "epoch": 0.48, + "learning_rate": 1.5167173252279635e-07, + "loss": 0.6704, + "step": 13247 + }, + { + "epoch": 0.48, + "learning_rate": 1.5166807045812428e-07, + "loss": 0.699, + "step": 13248 + }, + { + "epoch": 0.48, + "learning_rate": 1.5166440839345223e-07, + "loss": 0.7065, + "step": 13249 + }, + { + "epoch": 0.48, + "learning_rate": 1.5166074632878016e-07, + "loss": 0.6982, + "step": 13250 + }, + { + "epoch": 0.48, + "learning_rate": 1.516570842641081e-07, + "loss": 0.6893, + "step": 13251 + }, + { + "epoch": 0.48, + "learning_rate": 1.5165342219943604e-07, + "loss": 0.709, + "step": 13252 + }, + { + "epoch": 0.48, + "learning_rate": 1.51649760134764e-07, + "loss": 0.6945, + "step": 13253 + }, + { + "epoch": 0.48, + "learning_rate": 1.5164609807009192e-07, + "loss": 0.7099, + "step": 13254 + }, + { + "epoch": 0.48, + "learning_rate": 1.5164243600541984e-07, + "loss": 0.692, + "step": 13255 + }, + { + "epoch": 0.48, + "learning_rate": 1.5163877394074777e-07, + "loss": 0.7003, + "step": 13256 + }, + { + "epoch": 0.49, + "learning_rate": 1.5163511187607572e-07, + "loss": 0.6973, + "step": 13257 + }, + { + "epoch": 0.49, + "learning_rate": 1.5163144981140365e-07, + "loss": 0.6828, + "step": 13258 + }, + { + "epoch": 0.49, + "learning_rate": 1.516277877467316e-07, + "loss": 0.6834, + "step": 13259 + }, + { + "epoch": 0.49, + "learning_rate": 1.5162412568205953e-07, + "loss": 0.7034, + "step": 13260 + }, + { + "epoch": 0.49, + "learning_rate": 1.5162046361738748e-07, + "loss": 0.6945, + "step": 13261 + }, + { + "epoch": 0.49, + "learning_rate": 1.516168015527154e-07, + "loss": 0.7083, + "step": 13262 + }, + { + "epoch": 0.49, + "learning_rate": 1.5161313948804336e-07, + "loss": 0.7116, + "step": 13263 + }, + { + "epoch": 0.49, + "learning_rate": 1.516094774233713e-07, + "loss": 0.6993, + "step": 13264 + }, + { + "epoch": 0.49, + "learning_rate": 1.5160581535869924e-07, + "loss": 0.6947, + "step": 13265 + }, + { + "epoch": 0.49, + "learning_rate": 1.5160215329402717e-07, + "loss": 0.6972, + "step": 13266 + }, + { + "epoch": 0.49, + "learning_rate": 1.515984912293551e-07, + "loss": 0.6914, + "step": 13267 + }, + { + "epoch": 0.49, + "learning_rate": 1.5159482916468305e-07, + "loss": 0.6994, + "step": 13268 + }, + { + "epoch": 0.49, + "learning_rate": 1.5159116710001097e-07, + "loss": 0.6932, + "step": 13269 + }, + { + "epoch": 0.49, + "learning_rate": 1.5158750503533893e-07, + "loss": 0.6951, + "step": 13270 + }, + { + "epoch": 0.49, + "learning_rate": 1.5158384297066685e-07, + "loss": 0.6999, + "step": 13271 + }, + { + "epoch": 0.49, + "learning_rate": 1.515801809059948e-07, + "loss": 0.6983, + "step": 13272 + }, + { + "epoch": 0.49, + "learning_rate": 1.515765188413227e-07, + "loss": 0.6823, + "step": 13273 + }, + { + "epoch": 0.49, + "learning_rate": 1.5157285677665066e-07, + "loss": 0.7047, + "step": 13274 + }, + { + "epoch": 0.49, + "learning_rate": 1.515691947119786e-07, + "loss": 0.701, + "step": 13275 + }, + { + "epoch": 0.49, + "learning_rate": 1.5156553264730654e-07, + "loss": 0.6902, + "step": 13276 + }, + { + "epoch": 0.49, + "learning_rate": 1.5156187058263447e-07, + "loss": 0.6912, + "step": 13277 + }, + { + "epoch": 0.49, + "learning_rate": 1.5155820851796242e-07, + "loss": 0.687, + "step": 13278 + }, + { + "epoch": 0.49, + "learning_rate": 1.5155454645329035e-07, + "loss": 0.7022, + "step": 13279 + }, + { + "epoch": 0.49, + "learning_rate": 1.515508843886183e-07, + "loss": 0.6938, + "step": 13280 + }, + { + "epoch": 0.49, + "learning_rate": 1.5154722232394623e-07, + "loss": 0.704, + "step": 13281 + }, + { + "epoch": 0.49, + "learning_rate": 1.5154356025927418e-07, + "loss": 0.6942, + "step": 13282 + }, + { + "epoch": 0.49, + "learning_rate": 1.515398981946021e-07, + "loss": 0.6909, + "step": 13283 + }, + { + "epoch": 0.49, + "learning_rate": 1.5153623612993006e-07, + "loss": 0.6828, + "step": 13284 + }, + { + "epoch": 0.49, + "learning_rate": 1.5153257406525799e-07, + "loss": 0.6909, + "step": 13285 + }, + { + "epoch": 0.49, + "learning_rate": 1.5152891200058594e-07, + "loss": 0.6972, + "step": 13286 + }, + { + "epoch": 0.49, + "learning_rate": 1.5152524993591387e-07, + "loss": 0.7057, + "step": 13287 + }, + { + "epoch": 0.49, + "learning_rate": 1.5152158787124182e-07, + "loss": 0.6864, + "step": 13288 + }, + { + "epoch": 0.49, + "learning_rate": 1.5151792580656975e-07, + "loss": 0.6853, + "step": 13289 + }, + { + "epoch": 0.49, + "learning_rate": 1.5151426374189767e-07, + "loss": 0.6851, + "step": 13290 + }, + { + "epoch": 0.49, + "learning_rate": 1.515106016772256e-07, + "loss": 0.7041, + "step": 13291 + }, + { + "epoch": 0.49, + "learning_rate": 1.5150693961255355e-07, + "loss": 0.699, + "step": 13292 + }, + { + "epoch": 0.49, + "learning_rate": 1.5150327754788148e-07, + "loss": 0.7149, + "step": 13293 + }, + { + "epoch": 0.49, + "learning_rate": 1.5149961548320943e-07, + "loss": 0.6933, + "step": 13294 + }, + { + "epoch": 0.49, + "learning_rate": 1.5149595341853736e-07, + "loss": 0.7142, + "step": 13295 + }, + { + "epoch": 0.49, + "learning_rate": 1.5149229135386529e-07, + "loss": 0.6851, + "step": 13296 + }, + { + "epoch": 0.49, + "learning_rate": 1.5148862928919324e-07, + "loss": 0.7224, + "step": 13297 + }, + { + "epoch": 0.49, + "learning_rate": 1.5148496722452117e-07, + "loss": 0.6853, + "step": 13298 + }, + { + "epoch": 0.49, + "learning_rate": 1.5148130515984912e-07, + "loss": 0.6944, + "step": 13299 + }, + { + "epoch": 0.49, + "learning_rate": 1.5147764309517705e-07, + "loss": 0.6988, + "step": 13300 + }, + { + "epoch": 0.49, + "learning_rate": 1.51473981030505e-07, + "loss": 0.6962, + "step": 13301 + }, + { + "epoch": 0.49, + "learning_rate": 1.5147031896583293e-07, + "loss": 0.6961, + "step": 13302 + }, + { + "epoch": 0.49, + "learning_rate": 1.5146665690116088e-07, + "loss": 0.6863, + "step": 13303 + }, + { + "epoch": 0.49, + "learning_rate": 1.514629948364888e-07, + "loss": 0.6934, + "step": 13304 + }, + { + "epoch": 0.49, + "learning_rate": 1.5145933277181676e-07, + "loss": 0.6888, + "step": 13305 + }, + { + "epoch": 0.49, + "learning_rate": 1.5145567070714469e-07, + "loss": 0.6803, + "step": 13306 + }, + { + "epoch": 0.49, + "learning_rate": 1.5145200864247264e-07, + "loss": 0.6949, + "step": 13307 + }, + { + "epoch": 0.49, + "learning_rate": 1.5144834657780057e-07, + "loss": 0.6846, + "step": 13308 + }, + { + "epoch": 0.49, + "learning_rate": 1.514446845131285e-07, + "loss": 0.6833, + "step": 13309 + }, + { + "epoch": 0.49, + "learning_rate": 1.5144102244845642e-07, + "loss": 0.7062, + "step": 13310 + }, + { + "epoch": 0.49, + "learning_rate": 1.5143736038378437e-07, + "loss": 0.7127, + "step": 13311 + }, + { + "epoch": 0.49, + "learning_rate": 1.514336983191123e-07, + "loss": 0.6781, + "step": 13312 + }, + { + "epoch": 0.49, + "learning_rate": 1.5143003625444025e-07, + "loss": 0.6949, + "step": 13313 + }, + { + "epoch": 0.49, + "learning_rate": 1.5142637418976818e-07, + "loss": 0.6882, + "step": 13314 + }, + { + "epoch": 0.49, + "learning_rate": 1.5142271212509613e-07, + "loss": 0.6927, + "step": 13315 + }, + { + "epoch": 0.49, + "learning_rate": 1.5141905006042406e-07, + "loss": 0.7092, + "step": 13316 + }, + { + "epoch": 0.49, + "learning_rate": 1.5141538799575199e-07, + "loss": 0.6989, + "step": 13317 + }, + { + "epoch": 0.49, + "learning_rate": 1.5141172593107994e-07, + "loss": 0.7083, + "step": 13318 + }, + { + "epoch": 0.49, + "learning_rate": 1.5140806386640786e-07, + "loss": 0.7008, + "step": 13319 + }, + { + "epoch": 0.49, + "learning_rate": 1.5140440180173582e-07, + "loss": 0.6932, + "step": 13320 + }, + { + "epoch": 0.49, + "learning_rate": 1.5140073973706374e-07, + "loss": 0.7081, + "step": 13321 + }, + { + "epoch": 0.49, + "learning_rate": 1.513970776723917e-07, + "loss": 0.7111, + "step": 13322 + }, + { + "epoch": 0.49, + "learning_rate": 1.5139341560771962e-07, + "loss": 0.6893, + "step": 13323 + }, + { + "epoch": 0.49, + "learning_rate": 1.5138975354304758e-07, + "loss": 0.6851, + "step": 13324 + }, + { + "epoch": 0.49, + "learning_rate": 1.513860914783755e-07, + "loss": 0.7032, + "step": 13325 + }, + { + "epoch": 0.49, + "learning_rate": 1.5138242941370346e-07, + "loss": 0.6923, + "step": 13326 + }, + { + "epoch": 0.49, + "learning_rate": 1.5137876734903136e-07, + "loss": 0.6914, + "step": 13327 + }, + { + "epoch": 0.49, + "learning_rate": 1.513751052843593e-07, + "loss": 0.6879, + "step": 13328 + }, + { + "epoch": 0.49, + "learning_rate": 1.5137144321968724e-07, + "loss": 0.697, + "step": 13329 + }, + { + "epoch": 0.49, + "learning_rate": 1.513677811550152e-07, + "loss": 0.6879, + "step": 13330 + }, + { + "epoch": 0.49, + "learning_rate": 1.5136411909034312e-07, + "loss": 0.6892, + "step": 13331 + }, + { + "epoch": 0.49, + "learning_rate": 1.5136045702567107e-07, + "loss": 0.6938, + "step": 13332 + }, + { + "epoch": 0.49, + "learning_rate": 1.51356794960999e-07, + "loss": 0.6851, + "step": 13333 + }, + { + "epoch": 0.49, + "learning_rate": 1.5135313289632695e-07, + "loss": 0.6871, + "step": 13334 + }, + { + "epoch": 0.49, + "learning_rate": 1.5134947083165488e-07, + "loss": 0.6939, + "step": 13335 + }, + { + "epoch": 0.49, + "learning_rate": 1.5134580876698283e-07, + "loss": 0.7031, + "step": 13336 + }, + { + "epoch": 0.49, + "learning_rate": 1.5134214670231076e-07, + "loss": 0.6905, + "step": 13337 + }, + { + "epoch": 0.49, + "learning_rate": 1.5133848463763868e-07, + "loss": 0.6748, + "step": 13338 + }, + { + "epoch": 0.49, + "learning_rate": 1.5133482257296664e-07, + "loss": 0.6981, + "step": 13339 + }, + { + "epoch": 0.49, + "learning_rate": 1.5133116050829456e-07, + "loss": 0.6938, + "step": 13340 + }, + { + "epoch": 0.49, + "learning_rate": 1.5132749844362252e-07, + "loss": 0.6952, + "step": 13341 + }, + { + "epoch": 0.49, + "learning_rate": 1.5132383637895044e-07, + "loss": 0.6798, + "step": 13342 + }, + { + "epoch": 0.49, + "learning_rate": 1.513201743142784e-07, + "loss": 0.6945, + "step": 13343 + }, + { + "epoch": 0.49, + "learning_rate": 1.5131651224960632e-07, + "loss": 0.7043, + "step": 13344 + }, + { + "epoch": 0.49, + "learning_rate": 1.5131285018493425e-07, + "loss": 0.7015, + "step": 13345 + }, + { + "epoch": 0.49, + "learning_rate": 1.5130918812026218e-07, + "loss": 0.7077, + "step": 13346 + }, + { + "epoch": 0.49, + "learning_rate": 1.5130552605559013e-07, + "loss": 0.6875, + "step": 13347 + }, + { + "epoch": 0.49, + "learning_rate": 1.5130186399091806e-07, + "loss": 0.7042, + "step": 13348 + }, + { + "epoch": 0.49, + "learning_rate": 1.51298201926246e-07, + "loss": 0.6932, + "step": 13349 + }, + { + "epoch": 0.49, + "learning_rate": 1.5129453986157394e-07, + "loss": 0.6814, + "step": 13350 + }, + { + "epoch": 0.49, + "learning_rate": 1.512908777969019e-07, + "loss": 0.682, + "step": 13351 + }, + { + "epoch": 0.49, + "learning_rate": 1.5128721573222982e-07, + "loss": 0.6955, + "step": 13352 + }, + { + "epoch": 0.49, + "learning_rate": 1.5128355366755777e-07, + "loss": 0.6901, + "step": 13353 + }, + { + "epoch": 0.49, + "learning_rate": 1.512798916028857e-07, + "loss": 0.6955, + "step": 13354 + }, + { + "epoch": 0.49, + "learning_rate": 1.5127622953821365e-07, + "loss": 0.6925, + "step": 13355 + }, + { + "epoch": 0.49, + "learning_rate": 1.5127256747354158e-07, + "loss": 0.6949, + "step": 13356 + }, + { + "epoch": 0.49, + "learning_rate": 1.5126890540886953e-07, + "loss": 0.6853, + "step": 13357 + }, + { + "epoch": 0.49, + "learning_rate": 1.5126524334419746e-07, + "loss": 0.696, + "step": 13358 + }, + { + "epoch": 0.49, + "learning_rate": 1.512615812795254e-07, + "loss": 0.6844, + "step": 13359 + }, + { + "epoch": 0.49, + "learning_rate": 1.5125791921485334e-07, + "loss": 0.6838, + "step": 13360 + }, + { + "epoch": 0.49, + "learning_rate": 1.5125425715018126e-07, + "loss": 0.699, + "step": 13361 + }, + { + "epoch": 0.49, + "learning_rate": 1.5125059508550922e-07, + "loss": 0.6911, + "step": 13362 + }, + { + "epoch": 0.49, + "learning_rate": 1.5124693302083714e-07, + "loss": 0.6931, + "step": 13363 + }, + { + "epoch": 0.49, + "learning_rate": 1.5124327095616507e-07, + "loss": 0.6945, + "step": 13364 + }, + { + "epoch": 0.49, + "learning_rate": 1.5123960889149302e-07, + "loss": 0.6985, + "step": 13365 + }, + { + "epoch": 0.49, + "learning_rate": 1.5123594682682095e-07, + "loss": 0.687, + "step": 13366 + }, + { + "epoch": 0.49, + "learning_rate": 1.5123228476214888e-07, + "loss": 0.6858, + "step": 13367 + }, + { + "epoch": 0.49, + "learning_rate": 1.5122862269747683e-07, + "loss": 0.6836, + "step": 13368 + }, + { + "epoch": 0.49, + "learning_rate": 1.5122496063280476e-07, + "loss": 0.6867, + "step": 13369 + }, + { + "epoch": 0.49, + "learning_rate": 1.512212985681327e-07, + "loss": 0.6875, + "step": 13370 + }, + { + "epoch": 0.49, + "learning_rate": 1.5121763650346064e-07, + "loss": 0.676, + "step": 13371 + }, + { + "epoch": 0.49, + "learning_rate": 1.512139744387886e-07, + "loss": 0.6878, + "step": 13372 + }, + { + "epoch": 0.49, + "learning_rate": 1.5121031237411652e-07, + "loss": 0.6991, + "step": 13373 + }, + { + "epoch": 0.49, + "learning_rate": 1.5120665030944447e-07, + "loss": 0.7052, + "step": 13374 + }, + { + "epoch": 0.49, + "learning_rate": 1.512029882447724e-07, + "loss": 0.6894, + "step": 13375 + }, + { + "epoch": 0.49, + "learning_rate": 1.5119932618010035e-07, + "loss": 0.6886, + "step": 13376 + }, + { + "epoch": 0.49, + "learning_rate": 1.5119566411542827e-07, + "loss": 0.6877, + "step": 13377 + }, + { + "epoch": 0.49, + "learning_rate": 1.5119200205075623e-07, + "loss": 0.6949, + "step": 13378 + }, + { + "epoch": 0.49, + "learning_rate": 1.5118833998608415e-07, + "loss": 0.711, + "step": 13379 + }, + { + "epoch": 0.49, + "learning_rate": 1.511846779214121e-07, + "loss": 0.7077, + "step": 13380 + }, + { + "epoch": 0.49, + "learning_rate": 1.5118101585674e-07, + "loss": 0.6957, + "step": 13381 + }, + { + "epoch": 0.49, + "learning_rate": 1.5117735379206796e-07, + "loss": 0.6938, + "step": 13382 + }, + { + "epoch": 0.49, + "learning_rate": 1.511736917273959e-07, + "loss": 0.6825, + "step": 13383 + }, + { + "epoch": 0.49, + "learning_rate": 1.5117002966272384e-07, + "loss": 0.685, + "step": 13384 + }, + { + "epoch": 0.49, + "learning_rate": 1.5116636759805177e-07, + "loss": 0.6943, + "step": 13385 + }, + { + "epoch": 0.49, + "learning_rate": 1.5116270553337972e-07, + "loss": 0.7035, + "step": 13386 + }, + { + "epoch": 0.49, + "learning_rate": 1.5115904346870765e-07, + "loss": 0.6852, + "step": 13387 + }, + { + "epoch": 0.49, + "learning_rate": 1.5115538140403557e-07, + "loss": 0.6921, + "step": 13388 + }, + { + "epoch": 0.49, + "learning_rate": 1.5115171933936353e-07, + "loss": 0.7057, + "step": 13389 + }, + { + "epoch": 0.49, + "learning_rate": 1.5114805727469145e-07, + "loss": 0.7085, + "step": 13390 + }, + { + "epoch": 0.49, + "learning_rate": 1.511443952100194e-07, + "loss": 0.6874, + "step": 13391 + }, + { + "epoch": 0.49, + "learning_rate": 1.5114073314534733e-07, + "loss": 0.6969, + "step": 13392 + }, + { + "epoch": 0.49, + "learning_rate": 1.511370710806753e-07, + "loss": 0.6893, + "step": 13393 + }, + { + "epoch": 0.49, + "learning_rate": 1.5113340901600321e-07, + "loss": 0.694, + "step": 13394 + }, + { + "epoch": 0.49, + "learning_rate": 1.5112974695133117e-07, + "loss": 0.6854, + "step": 13395 + }, + { + "epoch": 0.49, + "learning_rate": 1.511260848866591e-07, + "loss": 0.6866, + "step": 13396 + }, + { + "epoch": 0.49, + "learning_rate": 1.5112242282198705e-07, + "loss": 0.7011, + "step": 13397 + }, + { + "epoch": 0.49, + "learning_rate": 1.5111876075731497e-07, + "loss": 0.6912, + "step": 13398 + }, + { + "epoch": 0.49, + "learning_rate": 1.511150986926429e-07, + "loss": 0.6936, + "step": 13399 + }, + { + "epoch": 0.49, + "learning_rate": 1.5111143662797083e-07, + "loss": 0.6957, + "step": 13400 + }, + { + "epoch": 0.49, + "learning_rate": 1.5110777456329878e-07, + "loss": 0.6889, + "step": 13401 + }, + { + "epoch": 0.49, + "learning_rate": 1.511041124986267e-07, + "loss": 0.6911, + "step": 13402 + }, + { + "epoch": 0.49, + "learning_rate": 1.5110045043395466e-07, + "loss": 0.6788, + "step": 13403 + }, + { + "epoch": 0.49, + "learning_rate": 1.510967883692826e-07, + "loss": 0.7082, + "step": 13404 + }, + { + "epoch": 0.49, + "learning_rate": 1.5109312630461054e-07, + "loss": 0.702, + "step": 13405 + }, + { + "epoch": 0.49, + "learning_rate": 1.5108946423993847e-07, + "loss": 0.6862, + "step": 13406 + }, + { + "epoch": 0.49, + "learning_rate": 1.5108580217526642e-07, + "loss": 0.6969, + "step": 13407 + }, + { + "epoch": 0.49, + "learning_rate": 1.5108214011059435e-07, + "loss": 0.698, + "step": 13408 + }, + { + "epoch": 0.49, + "learning_rate": 1.5107847804592227e-07, + "loss": 0.6963, + "step": 13409 + }, + { + "epoch": 0.49, + "learning_rate": 1.5107481598125023e-07, + "loss": 0.6893, + "step": 13410 + }, + { + "epoch": 0.49, + "learning_rate": 1.5107115391657815e-07, + "loss": 0.6975, + "step": 13411 + }, + { + "epoch": 0.49, + "learning_rate": 1.510674918519061e-07, + "loss": 0.6791, + "step": 13412 + }, + { + "epoch": 0.49, + "learning_rate": 1.5106382978723403e-07, + "loss": 0.691, + "step": 13413 + }, + { + "epoch": 0.49, + "learning_rate": 1.5106016772256199e-07, + "loss": 0.7002, + "step": 13414 + }, + { + "epoch": 0.49, + "learning_rate": 1.510565056578899e-07, + "loss": 0.7139, + "step": 13415 + }, + { + "epoch": 0.49, + "learning_rate": 1.5105284359321787e-07, + "loss": 0.6944, + "step": 13416 + }, + { + "epoch": 0.49, + "learning_rate": 1.5104918152854577e-07, + "loss": 0.7162, + "step": 13417 + }, + { + "epoch": 0.49, + "learning_rate": 1.5104551946387372e-07, + "loss": 0.6907, + "step": 13418 + }, + { + "epoch": 0.49, + "learning_rate": 1.5104185739920165e-07, + "loss": 0.7062, + "step": 13419 + }, + { + "epoch": 0.49, + "learning_rate": 1.510381953345296e-07, + "loss": 0.6897, + "step": 13420 + }, + { + "epoch": 0.49, + "learning_rate": 1.5103453326985753e-07, + "loss": 0.6902, + "step": 13421 + }, + { + "epoch": 0.49, + "learning_rate": 1.5103087120518548e-07, + "loss": 0.6972, + "step": 13422 + }, + { + "epoch": 0.49, + "learning_rate": 1.510272091405134e-07, + "loss": 0.7062, + "step": 13423 + }, + { + "epoch": 0.49, + "learning_rate": 1.5102354707584136e-07, + "loss": 0.7119, + "step": 13424 + }, + { + "epoch": 0.49, + "learning_rate": 1.5101988501116929e-07, + "loss": 0.6828, + "step": 13425 + }, + { + "epoch": 0.49, + "learning_rate": 1.5101622294649724e-07, + "loss": 0.6944, + "step": 13426 + }, + { + "epoch": 0.49, + "learning_rate": 1.5101256088182517e-07, + "loss": 0.6861, + "step": 13427 + }, + { + "epoch": 0.49, + "learning_rate": 1.5100889881715312e-07, + "loss": 0.6899, + "step": 13428 + }, + { + "epoch": 0.49, + "learning_rate": 1.5100523675248105e-07, + "loss": 0.6946, + "step": 13429 + }, + { + "epoch": 0.49, + "learning_rate": 1.51001574687809e-07, + "loss": 0.6981, + "step": 13430 + }, + { + "epoch": 0.49, + "learning_rate": 1.5099791262313693e-07, + "loss": 0.6858, + "step": 13431 + }, + { + "epoch": 0.49, + "learning_rate": 1.5099425055846485e-07, + "loss": 0.6912, + "step": 13432 + }, + { + "epoch": 0.49, + "learning_rate": 1.509905884937928e-07, + "loss": 0.7036, + "step": 13433 + }, + { + "epoch": 0.49, + "learning_rate": 1.5098692642912073e-07, + "loss": 0.6988, + "step": 13434 + }, + { + "epoch": 0.49, + "learning_rate": 1.5098326436444866e-07, + "loss": 0.6999, + "step": 13435 + }, + { + "epoch": 0.49, + "learning_rate": 1.509796022997766e-07, + "loss": 0.6903, + "step": 13436 + }, + { + "epoch": 0.49, + "learning_rate": 1.5097594023510454e-07, + "loss": 0.6909, + "step": 13437 + }, + { + "epoch": 0.49, + "learning_rate": 1.5097227817043246e-07, + "loss": 0.6902, + "step": 13438 + }, + { + "epoch": 0.49, + "learning_rate": 1.5096861610576042e-07, + "loss": 0.6982, + "step": 13439 + }, + { + "epoch": 0.49, + "learning_rate": 1.5096495404108834e-07, + "loss": 0.6817, + "step": 13440 + }, + { + "epoch": 0.49, + "learning_rate": 1.509612919764163e-07, + "loss": 0.7029, + "step": 13441 + }, + { + "epoch": 0.49, + "learning_rate": 1.5095762991174422e-07, + "loss": 0.6863, + "step": 13442 + }, + { + "epoch": 0.49, + "learning_rate": 1.5095396784707218e-07, + "loss": 0.6818, + "step": 13443 + }, + { + "epoch": 0.49, + "learning_rate": 1.509503057824001e-07, + "loss": 0.695, + "step": 13444 + }, + { + "epoch": 0.49, + "learning_rate": 1.5094664371772806e-07, + "loss": 0.6895, + "step": 13445 + }, + { + "epoch": 0.49, + "learning_rate": 1.5094298165305598e-07, + "loss": 0.687, + "step": 13446 + }, + { + "epoch": 0.49, + "learning_rate": 1.5093931958838394e-07, + "loss": 0.6826, + "step": 13447 + }, + { + "epoch": 0.49, + "learning_rate": 1.5093565752371186e-07, + "loss": 0.6935, + "step": 13448 + }, + { + "epoch": 0.49, + "learning_rate": 1.5093199545903982e-07, + "loss": 0.6883, + "step": 13449 + }, + { + "epoch": 0.49, + "learning_rate": 1.5092833339436774e-07, + "loss": 0.6872, + "step": 13450 + }, + { + "epoch": 0.49, + "learning_rate": 1.509246713296957e-07, + "loss": 0.7069, + "step": 13451 + }, + { + "epoch": 0.49, + "learning_rate": 1.5092100926502362e-07, + "loss": 0.6891, + "step": 13452 + }, + { + "epoch": 0.49, + "learning_rate": 1.5091734720035155e-07, + "loss": 0.7008, + "step": 13453 + }, + { + "epoch": 0.49, + "learning_rate": 1.5091368513567948e-07, + "loss": 0.6968, + "step": 13454 + }, + { + "epoch": 0.49, + "learning_rate": 1.5091002307100743e-07, + "loss": 0.6871, + "step": 13455 + }, + { + "epoch": 0.49, + "learning_rate": 1.5090636100633536e-07, + "loss": 0.6882, + "step": 13456 + }, + { + "epoch": 0.49, + "learning_rate": 1.509026989416633e-07, + "loss": 0.7068, + "step": 13457 + }, + { + "epoch": 0.49, + "learning_rate": 1.5089903687699124e-07, + "loss": 0.7038, + "step": 13458 + }, + { + "epoch": 0.49, + "learning_rate": 1.5089537481231916e-07, + "loss": 0.6844, + "step": 13459 + }, + { + "epoch": 0.49, + "learning_rate": 1.5089171274764712e-07, + "loss": 0.6876, + "step": 13460 + }, + { + "epoch": 0.49, + "learning_rate": 1.5088805068297504e-07, + "loss": 0.689, + "step": 13461 + }, + { + "epoch": 0.49, + "learning_rate": 1.50884388618303e-07, + "loss": 0.7267, + "step": 13462 + }, + { + "epoch": 0.49, + "learning_rate": 1.5088072655363092e-07, + "loss": 0.7045, + "step": 13463 + }, + { + "epoch": 0.49, + "learning_rate": 1.5087706448895888e-07, + "loss": 0.6864, + "step": 13464 + }, + { + "epoch": 0.49, + "learning_rate": 1.508734024242868e-07, + "loss": 0.6964, + "step": 13465 + }, + { + "epoch": 0.49, + "learning_rate": 1.5086974035961476e-07, + "loss": 0.6813, + "step": 13466 + }, + { + "epoch": 0.49, + "learning_rate": 1.5086607829494268e-07, + "loss": 0.7091, + "step": 13467 + }, + { + "epoch": 0.49, + "learning_rate": 1.5086241623027064e-07, + "loss": 0.6931, + "step": 13468 + }, + { + "epoch": 0.49, + "learning_rate": 1.5085875416559856e-07, + "loss": 0.6872, + "step": 13469 + }, + { + "epoch": 0.49, + "learning_rate": 1.5085509210092652e-07, + "loss": 0.6951, + "step": 13470 + }, + { + "epoch": 0.49, + "learning_rate": 1.5085143003625442e-07, + "loss": 0.6871, + "step": 13471 + }, + { + "epoch": 0.49, + "learning_rate": 1.5084776797158237e-07, + "loss": 0.6758, + "step": 13472 + }, + { + "epoch": 0.49, + "learning_rate": 1.508441059069103e-07, + "loss": 0.697, + "step": 13473 + }, + { + "epoch": 0.49, + "learning_rate": 1.5084044384223825e-07, + "loss": 0.6971, + "step": 13474 + }, + { + "epoch": 0.49, + "learning_rate": 1.5083678177756618e-07, + "loss": 0.6872, + "step": 13475 + }, + { + "epoch": 0.49, + "learning_rate": 1.5083311971289413e-07, + "loss": 0.6905, + "step": 13476 + }, + { + "epoch": 0.49, + "learning_rate": 1.5082945764822206e-07, + "loss": 0.6881, + "step": 13477 + }, + { + "epoch": 0.49, + "learning_rate": 1.5082579558355e-07, + "loss": 0.6804, + "step": 13478 + }, + { + "epoch": 0.49, + "learning_rate": 1.5082213351887794e-07, + "loss": 0.6967, + "step": 13479 + }, + { + "epoch": 0.49, + "learning_rate": 1.5081847145420586e-07, + "loss": 0.696, + "step": 13480 + }, + { + "epoch": 0.49, + "learning_rate": 1.5081480938953382e-07, + "loss": 0.6896, + "step": 13481 + }, + { + "epoch": 0.49, + "learning_rate": 1.5081114732486174e-07, + "loss": 0.6928, + "step": 13482 + }, + { + "epoch": 0.49, + "learning_rate": 1.508074852601897e-07, + "loss": 0.6816, + "step": 13483 + }, + { + "epoch": 0.49, + "learning_rate": 1.5080382319551762e-07, + "loss": 0.7, + "step": 13484 + }, + { + "epoch": 0.49, + "learning_rate": 1.5080016113084558e-07, + "loss": 0.6996, + "step": 13485 + }, + { + "epoch": 0.49, + "learning_rate": 1.507964990661735e-07, + "loss": 0.6972, + "step": 13486 + }, + { + "epoch": 0.49, + "learning_rate": 1.5079283700150146e-07, + "loss": 0.6857, + "step": 13487 + }, + { + "epoch": 0.49, + "learning_rate": 1.5078917493682938e-07, + "loss": 0.6955, + "step": 13488 + }, + { + "epoch": 0.49, + "learning_rate": 1.507855128721573e-07, + "loss": 0.685, + "step": 13489 + }, + { + "epoch": 0.49, + "learning_rate": 1.5078185080748524e-07, + "loss": 0.6863, + "step": 13490 + }, + { + "epoch": 0.49, + "learning_rate": 1.507781887428132e-07, + "loss": 0.6899, + "step": 13491 + }, + { + "epoch": 0.49, + "learning_rate": 1.5077452667814112e-07, + "loss": 0.6816, + "step": 13492 + }, + { + "epoch": 0.49, + "learning_rate": 1.5077086461346907e-07, + "loss": 0.6914, + "step": 13493 + }, + { + "epoch": 0.49, + "learning_rate": 1.50767202548797e-07, + "loss": 0.6872, + "step": 13494 + }, + { + "epoch": 0.49, + "learning_rate": 1.5076354048412495e-07, + "loss": 0.688, + "step": 13495 + }, + { + "epoch": 0.49, + "learning_rate": 1.5075987841945287e-07, + "loss": 0.6914, + "step": 13496 + }, + { + "epoch": 0.49, + "learning_rate": 1.5075621635478083e-07, + "loss": 0.7121, + "step": 13497 + }, + { + "epoch": 0.49, + "learning_rate": 1.5075255429010875e-07, + "loss": 0.692, + "step": 13498 + }, + { + "epoch": 0.49, + "learning_rate": 1.507488922254367e-07, + "loss": 0.6895, + "step": 13499 + }, + { + "epoch": 0.49, + "learning_rate": 1.5074523016076463e-07, + "loss": 0.6875, + "step": 13500 + }, + { + "epoch": 0.49, + "learning_rate": 1.507415680960926e-07, + "loss": 0.7052, + "step": 13501 + }, + { + "epoch": 0.49, + "learning_rate": 1.5073790603142051e-07, + "loss": 0.6916, + "step": 13502 + }, + { + "epoch": 0.49, + "learning_rate": 1.5073424396674844e-07, + "loss": 0.6986, + "step": 13503 + }, + { + "epoch": 0.49, + "learning_rate": 1.507305819020764e-07, + "loss": 0.6812, + "step": 13504 + }, + { + "epoch": 0.49, + "learning_rate": 1.5072691983740432e-07, + "loss": 0.6907, + "step": 13505 + }, + { + "epoch": 0.49, + "learning_rate": 1.5072325777273227e-07, + "loss": 0.6888, + "step": 13506 + }, + { + "epoch": 0.49, + "learning_rate": 1.5071959570806017e-07, + "loss": 0.7048, + "step": 13507 + }, + { + "epoch": 0.49, + "learning_rate": 1.5071593364338813e-07, + "loss": 0.6927, + "step": 13508 + }, + { + "epoch": 0.49, + "learning_rate": 1.5071227157871605e-07, + "loss": 0.7003, + "step": 13509 + }, + { + "epoch": 0.49, + "learning_rate": 1.50708609514044e-07, + "loss": 0.6846, + "step": 13510 + }, + { + "epoch": 0.49, + "learning_rate": 1.5070494744937193e-07, + "loss": 0.692, + "step": 13511 + }, + { + "epoch": 0.49, + "learning_rate": 1.507012853846999e-07, + "loss": 0.6969, + "step": 13512 + }, + { + "epoch": 0.49, + "learning_rate": 1.5069762332002781e-07, + "loss": 0.697, + "step": 13513 + }, + { + "epoch": 0.49, + "learning_rate": 1.5069396125535577e-07, + "loss": 0.6921, + "step": 13514 + }, + { + "epoch": 0.49, + "learning_rate": 1.506902991906837e-07, + "loss": 0.6868, + "step": 13515 + }, + { + "epoch": 0.49, + "learning_rate": 1.5068663712601165e-07, + "loss": 0.6955, + "step": 13516 + }, + { + "epoch": 0.49, + "learning_rate": 1.5068297506133957e-07, + "loss": 0.6909, + "step": 13517 + }, + { + "epoch": 0.49, + "learning_rate": 1.5067931299666753e-07, + "loss": 0.6951, + "step": 13518 + }, + { + "epoch": 0.49, + "learning_rate": 1.5067565093199545e-07, + "loss": 0.6892, + "step": 13519 + }, + { + "epoch": 0.49, + "learning_rate": 1.506719888673234e-07, + "loss": 0.6956, + "step": 13520 + }, + { + "epoch": 0.49, + "learning_rate": 1.5066832680265133e-07, + "loss": 0.6816, + "step": 13521 + }, + { + "epoch": 0.49, + "learning_rate": 1.5066466473797929e-07, + "loss": 0.6969, + "step": 13522 + }, + { + "epoch": 0.49, + "learning_rate": 1.5066100267330721e-07, + "loss": 0.7031, + "step": 13523 + }, + { + "epoch": 0.49, + "learning_rate": 1.5065734060863514e-07, + "loss": 0.6932, + "step": 13524 + }, + { + "epoch": 0.49, + "learning_rate": 1.5065367854396307e-07, + "loss": 0.7017, + "step": 13525 + }, + { + "epoch": 0.49, + "learning_rate": 1.5065001647929102e-07, + "loss": 0.6999, + "step": 13526 + }, + { + "epoch": 0.49, + "learning_rate": 1.5064635441461895e-07, + "loss": 0.696, + "step": 13527 + }, + { + "epoch": 0.49, + "learning_rate": 1.506426923499469e-07, + "loss": 0.6959, + "step": 13528 + }, + { + "epoch": 0.49, + "learning_rate": 1.5063903028527483e-07, + "loss": 0.7026, + "step": 13529 + }, + { + "epoch": 0.5, + "learning_rate": 1.5063536822060275e-07, + "loss": 0.6935, + "step": 13530 + }, + { + "epoch": 0.5, + "learning_rate": 1.506317061559307e-07, + "loss": 0.7022, + "step": 13531 + }, + { + "epoch": 0.5, + "learning_rate": 1.5062804409125863e-07, + "loss": 0.6989, + "step": 13532 + }, + { + "epoch": 0.5, + "learning_rate": 1.5062438202658659e-07, + "loss": 0.6968, + "step": 13533 + }, + { + "epoch": 0.5, + "learning_rate": 1.506207199619145e-07, + "loss": 0.6818, + "step": 13534 + }, + { + "epoch": 0.5, + "learning_rate": 1.5061705789724247e-07, + "loss": 0.6875, + "step": 13535 + }, + { + "epoch": 0.5, + "learning_rate": 1.506133958325704e-07, + "loss": 0.6921, + "step": 13536 + }, + { + "epoch": 0.5, + "learning_rate": 1.5060973376789835e-07, + "loss": 0.6933, + "step": 13537 + }, + { + "epoch": 0.5, + "learning_rate": 1.5060607170322627e-07, + "loss": 0.6937, + "step": 13538 + }, + { + "epoch": 0.5, + "learning_rate": 1.5060240963855423e-07, + "loss": 0.7015, + "step": 13539 + }, + { + "epoch": 0.5, + "learning_rate": 1.5059874757388215e-07, + "loss": 0.6895, + "step": 13540 + }, + { + "epoch": 0.5, + "learning_rate": 1.505950855092101e-07, + "loss": 0.6891, + "step": 13541 + }, + { + "epoch": 0.5, + "learning_rate": 1.50591423444538e-07, + "loss": 0.6971, + "step": 13542 + }, + { + "epoch": 0.5, + "learning_rate": 1.5058776137986596e-07, + "loss": 0.6813, + "step": 13543 + }, + { + "epoch": 0.5, + "learning_rate": 1.5058409931519389e-07, + "loss": 0.6923, + "step": 13544 + }, + { + "epoch": 0.5, + "learning_rate": 1.5058043725052184e-07, + "loss": 0.6931, + "step": 13545 + }, + { + "epoch": 0.5, + "learning_rate": 1.5057677518584977e-07, + "loss": 0.6836, + "step": 13546 + }, + { + "epoch": 0.5, + "learning_rate": 1.5057311312117772e-07, + "loss": 0.6935, + "step": 13547 + }, + { + "epoch": 0.5, + "learning_rate": 1.5056945105650565e-07, + "loss": 0.6951, + "step": 13548 + }, + { + "epoch": 0.5, + "learning_rate": 1.505657889918336e-07, + "loss": 0.7086, + "step": 13549 + }, + { + "epoch": 0.5, + "learning_rate": 1.5056212692716153e-07, + "loss": 0.7158, + "step": 13550 + }, + { + "epoch": 0.5, + "learning_rate": 1.5055846486248945e-07, + "loss": 0.6982, + "step": 13551 + }, + { + "epoch": 0.5, + "learning_rate": 1.505548027978174e-07, + "loss": 0.6923, + "step": 13552 + }, + { + "epoch": 0.5, + "learning_rate": 1.5055114073314533e-07, + "loss": 0.6828, + "step": 13553 + }, + { + "epoch": 0.5, + "learning_rate": 1.5054747866847328e-07, + "loss": 0.6963, + "step": 13554 + }, + { + "epoch": 0.5, + "learning_rate": 1.505438166038012e-07, + "loss": 0.7, + "step": 13555 + }, + { + "epoch": 0.5, + "learning_rate": 1.5054015453912916e-07, + "loss": 0.6982, + "step": 13556 + }, + { + "epoch": 0.5, + "learning_rate": 1.505364924744571e-07, + "loss": 0.6874, + "step": 13557 + }, + { + "epoch": 0.5, + "learning_rate": 1.5053283040978504e-07, + "loss": 0.686, + "step": 13558 + }, + { + "epoch": 0.5, + "learning_rate": 1.5052916834511297e-07, + "loss": 0.6953, + "step": 13559 + }, + { + "epoch": 0.5, + "learning_rate": 1.505255062804409e-07, + "loss": 0.6829, + "step": 13560 + }, + { + "epoch": 0.5, + "learning_rate": 1.5052184421576882e-07, + "loss": 0.6976, + "step": 13561 + }, + { + "epoch": 0.5, + "learning_rate": 1.5051818215109678e-07, + "loss": 0.6975, + "step": 13562 + }, + { + "epoch": 0.5, + "learning_rate": 1.505145200864247e-07, + "loss": 0.7043, + "step": 13563 + }, + { + "epoch": 0.5, + "learning_rate": 1.5051085802175266e-07, + "loss": 0.7106, + "step": 13564 + }, + { + "epoch": 0.5, + "learning_rate": 1.5050719595708058e-07, + "loss": 0.6922, + "step": 13565 + }, + { + "epoch": 0.5, + "learning_rate": 1.5050353389240854e-07, + "loss": 0.7014, + "step": 13566 + }, + { + "epoch": 0.5, + "learning_rate": 1.5049987182773646e-07, + "loss": 0.7036, + "step": 13567 + }, + { + "epoch": 0.5, + "learning_rate": 1.5049620976306442e-07, + "loss": 0.6943, + "step": 13568 + }, + { + "epoch": 0.5, + "learning_rate": 1.5049254769839234e-07, + "loss": 0.6928, + "step": 13569 + }, + { + "epoch": 0.5, + "learning_rate": 1.504888856337203e-07, + "loss": 0.6789, + "step": 13570 + }, + { + "epoch": 0.5, + "learning_rate": 1.5048522356904822e-07, + "loss": 0.6876, + "step": 13571 + }, + { + "epoch": 0.5, + "learning_rate": 1.5048156150437618e-07, + "loss": 0.6902, + "step": 13572 + }, + { + "epoch": 0.5, + "learning_rate": 1.504778994397041e-07, + "loss": 0.6965, + "step": 13573 + }, + { + "epoch": 0.5, + "learning_rate": 1.5047423737503203e-07, + "loss": 0.6889, + "step": 13574 + }, + { + "epoch": 0.5, + "learning_rate": 1.5047057531035998e-07, + "loss": 0.6898, + "step": 13575 + }, + { + "epoch": 0.5, + "learning_rate": 1.504669132456879e-07, + "loss": 0.697, + "step": 13576 + }, + { + "epoch": 0.5, + "learning_rate": 1.5046325118101586e-07, + "loss": 0.6895, + "step": 13577 + }, + { + "epoch": 0.5, + "learning_rate": 1.5045958911634376e-07, + "loss": 0.6844, + "step": 13578 + }, + { + "epoch": 0.5, + "learning_rate": 1.5045592705167172e-07, + "loss": 0.6947, + "step": 13579 + }, + { + "epoch": 0.5, + "learning_rate": 1.5045226498699964e-07, + "loss": 0.6662, + "step": 13580 + }, + { + "epoch": 0.5, + "learning_rate": 1.504486029223276e-07, + "loss": 0.6908, + "step": 13581 + }, + { + "epoch": 0.5, + "learning_rate": 1.5044494085765552e-07, + "loss": 0.7029, + "step": 13582 + }, + { + "epoch": 0.5, + "learning_rate": 1.5044127879298348e-07, + "loss": 0.6962, + "step": 13583 + }, + { + "epoch": 0.5, + "learning_rate": 1.504376167283114e-07, + "loss": 0.714, + "step": 13584 + }, + { + "epoch": 0.5, + "learning_rate": 1.5043395466363936e-07, + "loss": 0.6925, + "step": 13585 + }, + { + "epoch": 0.5, + "learning_rate": 1.5043029259896728e-07, + "loss": 0.7011, + "step": 13586 + }, + { + "epoch": 0.5, + "learning_rate": 1.5042663053429524e-07, + "loss": 0.6844, + "step": 13587 + }, + { + "epoch": 0.5, + "learning_rate": 1.5042296846962316e-07, + "loss": 0.6905, + "step": 13588 + }, + { + "epoch": 0.5, + "learning_rate": 1.5041930640495112e-07, + "loss": 0.6892, + "step": 13589 + }, + { + "epoch": 0.5, + "learning_rate": 1.5041564434027904e-07, + "loss": 0.6949, + "step": 13590 + }, + { + "epoch": 0.5, + "learning_rate": 1.50411982275607e-07, + "loss": 0.6916, + "step": 13591 + }, + { + "epoch": 0.5, + "learning_rate": 1.5040832021093492e-07, + "loss": 0.6872, + "step": 13592 + }, + { + "epoch": 0.5, + "learning_rate": 1.5040465814626288e-07, + "loss": 0.7029, + "step": 13593 + }, + { + "epoch": 0.5, + "learning_rate": 1.504009960815908e-07, + "loss": 0.6912, + "step": 13594 + }, + { + "epoch": 0.5, + "learning_rate": 1.5039733401691873e-07, + "loss": 0.6901, + "step": 13595 + }, + { + "epoch": 0.5, + "learning_rate": 1.5039367195224666e-07, + "loss": 0.683, + "step": 13596 + }, + { + "epoch": 0.5, + "learning_rate": 1.503900098875746e-07, + "loss": 0.6951, + "step": 13597 + }, + { + "epoch": 0.5, + "learning_rate": 1.5038634782290254e-07, + "loss": 0.6893, + "step": 13598 + }, + { + "epoch": 0.5, + "learning_rate": 1.503826857582305e-07, + "loss": 0.6935, + "step": 13599 + }, + { + "epoch": 0.5, + "learning_rate": 1.5037902369355842e-07, + "loss": 0.6856, + "step": 13600 + }, + { + "epoch": 0.5, + "learning_rate": 1.5037536162888634e-07, + "loss": 0.6916, + "step": 13601 + }, + { + "epoch": 0.5, + "learning_rate": 1.503716995642143e-07, + "loss": 0.6945, + "step": 13602 + }, + { + "epoch": 0.5, + "learning_rate": 1.5036803749954222e-07, + "loss": 0.6952, + "step": 13603 + }, + { + "epoch": 0.5, + "learning_rate": 1.5036437543487018e-07, + "loss": 0.6876, + "step": 13604 + }, + { + "epoch": 0.5, + "learning_rate": 1.503607133701981e-07, + "loss": 0.6943, + "step": 13605 + }, + { + "epoch": 0.5, + "learning_rate": 1.5035705130552606e-07, + "loss": 0.6963, + "step": 13606 + }, + { + "epoch": 0.5, + "learning_rate": 1.5035338924085398e-07, + "loss": 0.6887, + "step": 13607 + }, + { + "epoch": 0.5, + "learning_rate": 1.5034972717618194e-07, + "loss": 0.6904, + "step": 13608 + }, + { + "epoch": 0.5, + "learning_rate": 1.5034606511150986e-07, + "loss": 0.6767, + "step": 13609 + }, + { + "epoch": 0.5, + "learning_rate": 1.5034240304683782e-07, + "loss": 0.6936, + "step": 13610 + }, + { + "epoch": 0.5, + "learning_rate": 1.5033874098216574e-07, + "loss": 0.7051, + "step": 13611 + }, + { + "epoch": 0.5, + "learning_rate": 1.503350789174937e-07, + "loss": 0.7017, + "step": 13612 + }, + { + "epoch": 0.5, + "learning_rate": 1.5033141685282162e-07, + "loss": 0.6974, + "step": 13613 + }, + { + "epoch": 0.5, + "learning_rate": 1.5032775478814955e-07, + "loss": 0.6907, + "step": 13614 + }, + { + "epoch": 0.5, + "learning_rate": 1.5032409272347747e-07, + "loss": 0.7016, + "step": 13615 + }, + { + "epoch": 0.5, + "learning_rate": 1.5032043065880543e-07, + "loss": 0.7068, + "step": 13616 + }, + { + "epoch": 0.5, + "learning_rate": 1.5031676859413335e-07, + "loss": 0.6901, + "step": 13617 + }, + { + "epoch": 0.5, + "learning_rate": 1.503131065294613e-07, + "loss": 0.7043, + "step": 13618 + }, + { + "epoch": 0.5, + "learning_rate": 1.5030944446478923e-07, + "loss": 0.6892, + "step": 13619 + }, + { + "epoch": 0.5, + "learning_rate": 1.503057824001172e-07, + "loss": 0.695, + "step": 13620 + }, + { + "epoch": 0.5, + "learning_rate": 1.5030212033544511e-07, + "loss": 0.6839, + "step": 13621 + }, + { + "epoch": 0.5, + "learning_rate": 1.5029845827077304e-07, + "loss": 0.693, + "step": 13622 + }, + { + "epoch": 0.5, + "learning_rate": 1.50294796206101e-07, + "loss": 0.6925, + "step": 13623 + }, + { + "epoch": 0.5, + "learning_rate": 1.5029113414142892e-07, + "loss": 0.6891, + "step": 13624 + }, + { + "epoch": 0.5, + "learning_rate": 1.5028747207675687e-07, + "loss": 0.6913, + "step": 13625 + }, + { + "epoch": 0.5, + "learning_rate": 1.502838100120848e-07, + "loss": 0.6899, + "step": 13626 + }, + { + "epoch": 0.5, + "learning_rate": 1.5028014794741275e-07, + "loss": 0.6981, + "step": 13627 + }, + { + "epoch": 0.5, + "learning_rate": 1.5027648588274068e-07, + "loss": 0.6931, + "step": 13628 + }, + { + "epoch": 0.5, + "learning_rate": 1.5027282381806863e-07, + "loss": 0.7001, + "step": 13629 + }, + { + "epoch": 0.5, + "learning_rate": 1.5026916175339656e-07, + "loss": 0.7009, + "step": 13630 + }, + { + "epoch": 0.5, + "learning_rate": 1.5026549968872451e-07, + "loss": 0.6883, + "step": 13631 + }, + { + "epoch": 0.5, + "learning_rate": 1.5026183762405241e-07, + "loss": 0.6836, + "step": 13632 + }, + { + "epoch": 0.5, + "learning_rate": 1.5025817555938037e-07, + "loss": 0.6914, + "step": 13633 + }, + { + "epoch": 0.5, + "learning_rate": 1.502545134947083e-07, + "loss": 0.7, + "step": 13634 + }, + { + "epoch": 0.5, + "learning_rate": 1.5025085143003625e-07, + "loss": 0.6852, + "step": 13635 + }, + { + "epoch": 0.5, + "learning_rate": 1.5024718936536417e-07, + "loss": 0.6902, + "step": 13636 + }, + { + "epoch": 0.5, + "learning_rate": 1.5024352730069213e-07, + "loss": 0.7056, + "step": 13637 + }, + { + "epoch": 0.5, + "learning_rate": 1.5023986523602005e-07, + "loss": 0.6961, + "step": 13638 + }, + { + "epoch": 0.5, + "learning_rate": 1.50236203171348e-07, + "loss": 0.69, + "step": 13639 + }, + { + "epoch": 0.5, + "learning_rate": 1.5023254110667593e-07, + "loss": 0.6789, + "step": 13640 + }, + { + "epoch": 0.5, + "learning_rate": 1.5022887904200389e-07, + "loss": 0.6858, + "step": 13641 + }, + { + "epoch": 0.5, + "learning_rate": 1.5022521697733181e-07, + "loss": 0.7017, + "step": 13642 + }, + { + "epoch": 0.5, + "learning_rate": 1.5022155491265977e-07, + "loss": 0.6951, + "step": 13643 + }, + { + "epoch": 0.5, + "learning_rate": 1.502178928479877e-07, + "loss": 0.6794, + "step": 13644 + }, + { + "epoch": 0.5, + "learning_rate": 1.5021423078331562e-07, + "loss": 0.6906, + "step": 13645 + }, + { + "epoch": 0.5, + "learning_rate": 1.5021056871864357e-07, + "loss": 0.6995, + "step": 13646 + }, + { + "epoch": 0.5, + "learning_rate": 1.502069066539715e-07, + "loss": 0.7072, + "step": 13647 + }, + { + "epoch": 0.5, + "learning_rate": 1.5020324458929945e-07, + "loss": 0.6862, + "step": 13648 + }, + { + "epoch": 0.5, + "learning_rate": 1.5019958252462738e-07, + "loss": 0.687, + "step": 13649 + }, + { + "epoch": 0.5, + "learning_rate": 1.501959204599553e-07, + "loss": 0.6879, + "step": 13650 + }, + { + "epoch": 0.5, + "learning_rate": 1.5019225839528323e-07, + "loss": 0.7019, + "step": 13651 + }, + { + "epoch": 0.5, + "learning_rate": 1.5018859633061119e-07, + "loss": 0.6927, + "step": 13652 + }, + { + "epoch": 0.5, + "learning_rate": 1.501849342659391e-07, + "loss": 0.6855, + "step": 13653 + }, + { + "epoch": 0.5, + "learning_rate": 1.5018127220126707e-07, + "loss": 0.6996, + "step": 13654 + }, + { + "epoch": 0.5, + "learning_rate": 1.50177610136595e-07, + "loss": 0.7081, + "step": 13655 + }, + { + "epoch": 0.5, + "learning_rate": 1.5017394807192295e-07, + "loss": 0.6991, + "step": 13656 + }, + { + "epoch": 0.5, + "learning_rate": 1.5017028600725087e-07, + "loss": 0.6848, + "step": 13657 + }, + { + "epoch": 0.5, + "learning_rate": 1.5016662394257883e-07, + "loss": 0.6977, + "step": 13658 + }, + { + "epoch": 0.5, + "learning_rate": 1.5016296187790675e-07, + "loss": 0.6905, + "step": 13659 + }, + { + "epoch": 0.5, + "learning_rate": 1.501592998132347e-07, + "loss": 0.6967, + "step": 13660 + }, + { + "epoch": 0.5, + "learning_rate": 1.5015563774856263e-07, + "loss": 0.6893, + "step": 13661 + }, + { + "epoch": 0.5, + "learning_rate": 1.5015197568389059e-07, + "loss": 0.6997, + "step": 13662 + }, + { + "epoch": 0.5, + "learning_rate": 1.501483136192185e-07, + "loss": 0.6891, + "step": 13663 + }, + { + "epoch": 0.5, + "learning_rate": 1.5014465155454647e-07, + "loss": 0.6931, + "step": 13664 + }, + { + "epoch": 0.5, + "learning_rate": 1.501409894898744e-07, + "loss": 0.6881, + "step": 13665 + }, + { + "epoch": 0.5, + "learning_rate": 1.5013732742520232e-07, + "loss": 0.7017, + "step": 13666 + }, + { + "epoch": 0.5, + "learning_rate": 1.5013366536053027e-07, + "loss": 0.6812, + "step": 13667 + }, + { + "epoch": 0.5, + "learning_rate": 1.501300032958582e-07, + "loss": 0.6949, + "step": 13668 + }, + { + "epoch": 0.5, + "learning_rate": 1.5012634123118613e-07, + "loss": 0.6963, + "step": 13669 + }, + { + "epoch": 0.5, + "learning_rate": 1.5012267916651408e-07, + "loss": 0.6913, + "step": 13670 + }, + { + "epoch": 0.5, + "learning_rate": 1.50119017101842e-07, + "loss": 0.6821, + "step": 13671 + }, + { + "epoch": 0.5, + "learning_rate": 1.5011535503716993e-07, + "loss": 0.6827, + "step": 13672 + }, + { + "epoch": 0.5, + "learning_rate": 1.5011169297249788e-07, + "loss": 0.6951, + "step": 13673 + }, + { + "epoch": 0.5, + "learning_rate": 1.501080309078258e-07, + "loss": 0.6896, + "step": 13674 + }, + { + "epoch": 0.5, + "learning_rate": 1.5010436884315376e-07, + "loss": 0.6915, + "step": 13675 + }, + { + "epoch": 0.5, + "learning_rate": 1.501007067784817e-07, + "loss": 0.6994, + "step": 13676 + }, + { + "epoch": 0.5, + "learning_rate": 1.5009704471380964e-07, + "loss": 0.7018, + "step": 13677 + }, + { + "epoch": 0.5, + "learning_rate": 1.5009338264913757e-07, + "loss": 0.6905, + "step": 13678 + }, + { + "epoch": 0.5, + "learning_rate": 1.5008972058446552e-07, + "loss": 0.6868, + "step": 13679 + }, + { + "epoch": 0.5, + "learning_rate": 1.5008605851979345e-07, + "loss": 0.6937, + "step": 13680 + }, + { + "epoch": 0.5, + "learning_rate": 1.500823964551214e-07, + "loss": 0.6921, + "step": 13681 + }, + { + "epoch": 0.5, + "learning_rate": 1.5007873439044933e-07, + "loss": 0.6907, + "step": 13682 + }, + { + "epoch": 0.5, + "learning_rate": 1.5007507232577728e-07, + "loss": 0.6941, + "step": 13683 + }, + { + "epoch": 0.5, + "learning_rate": 1.500714102611052e-07, + "loss": 0.6957, + "step": 13684 + }, + { + "epoch": 0.5, + "learning_rate": 1.5006774819643316e-07, + "loss": 0.6954, + "step": 13685 + }, + { + "epoch": 0.5, + "learning_rate": 1.5006408613176106e-07, + "loss": 0.6834, + "step": 13686 + }, + { + "epoch": 0.5, + "learning_rate": 1.5006042406708902e-07, + "loss": 0.709, + "step": 13687 + }, + { + "epoch": 0.5, + "learning_rate": 1.5005676200241694e-07, + "loss": 0.6827, + "step": 13688 + }, + { + "epoch": 0.5, + "learning_rate": 1.500530999377449e-07, + "loss": 0.6955, + "step": 13689 + }, + { + "epoch": 0.5, + "learning_rate": 1.5004943787307282e-07, + "loss": 0.6874, + "step": 13690 + }, + { + "epoch": 0.5, + "learning_rate": 1.5004577580840078e-07, + "loss": 0.6926, + "step": 13691 + }, + { + "epoch": 0.5, + "learning_rate": 1.500421137437287e-07, + "loss": 0.6874, + "step": 13692 + }, + { + "epoch": 0.5, + "learning_rate": 1.5003845167905663e-07, + "loss": 0.6819, + "step": 13693 + }, + { + "epoch": 0.5, + "learning_rate": 1.5003478961438458e-07, + "loss": 0.7074, + "step": 13694 + }, + { + "epoch": 0.5, + "learning_rate": 1.500311275497125e-07, + "loss": 0.6921, + "step": 13695 + }, + { + "epoch": 0.5, + "learning_rate": 1.5002746548504046e-07, + "loss": 0.7034, + "step": 13696 + }, + { + "epoch": 0.5, + "learning_rate": 1.500238034203684e-07, + "loss": 0.6853, + "step": 13697 + }, + { + "epoch": 0.5, + "learning_rate": 1.5002014135569634e-07, + "loss": 0.6967, + "step": 13698 + }, + { + "epoch": 0.5, + "learning_rate": 1.5001647929102427e-07, + "loss": 0.6933, + "step": 13699 + }, + { + "epoch": 0.5, + "learning_rate": 1.5001281722635222e-07, + "loss": 0.6972, + "step": 13700 + }, + { + "epoch": 0.5, + "learning_rate": 1.5000915516168015e-07, + "loss": 0.6887, + "step": 13701 + }, + { + "epoch": 0.5, + "learning_rate": 1.500054930970081e-07, + "loss": 0.6815, + "step": 13702 + }, + { + "epoch": 0.5, + "learning_rate": 1.5000183103233603e-07, + "loss": 0.7006, + "step": 13703 + }, + { + "epoch": 0.5, + "learning_rate": 1.4999816896766396e-07, + "loss": 0.684, + "step": 13704 + }, + { + "epoch": 0.5, + "learning_rate": 1.4999450690299188e-07, + "loss": 0.6779, + "step": 13705 + }, + { + "epoch": 0.5, + "learning_rate": 1.4999084483831984e-07, + "loss": 0.6875, + "step": 13706 + }, + { + "epoch": 0.5, + "learning_rate": 1.4998718277364776e-07, + "loss": 0.6942, + "step": 13707 + }, + { + "epoch": 0.5, + "learning_rate": 1.4998352070897572e-07, + "loss": 0.691, + "step": 13708 + }, + { + "epoch": 0.5, + "learning_rate": 1.4997985864430364e-07, + "loss": 0.6943, + "step": 13709 + }, + { + "epoch": 0.5, + "learning_rate": 1.499761965796316e-07, + "loss": 0.7032, + "step": 13710 + }, + { + "epoch": 0.5, + "learning_rate": 1.4997253451495952e-07, + "loss": 0.7004, + "step": 13711 + }, + { + "epoch": 0.5, + "learning_rate": 1.4996887245028748e-07, + "loss": 0.6936, + "step": 13712 + }, + { + "epoch": 0.5, + "learning_rate": 1.499652103856154e-07, + "loss": 0.6962, + "step": 13713 + }, + { + "epoch": 0.5, + "learning_rate": 1.4996154832094336e-07, + "loss": 0.6923, + "step": 13714 + }, + { + "epoch": 0.5, + "learning_rate": 1.4995788625627128e-07, + "loss": 0.7084, + "step": 13715 + }, + { + "epoch": 0.5, + "learning_rate": 1.499542241915992e-07, + "loss": 0.6985, + "step": 13716 + }, + { + "epoch": 0.5, + "learning_rate": 1.4995056212692716e-07, + "loss": 0.6896, + "step": 13717 + }, + { + "epoch": 0.5, + "learning_rate": 1.499469000622551e-07, + "loss": 0.6789, + "step": 13718 + }, + { + "epoch": 0.5, + "learning_rate": 1.4994323799758304e-07, + "loss": 0.7028, + "step": 13719 + }, + { + "epoch": 0.5, + "learning_rate": 1.4993957593291097e-07, + "loss": 0.7004, + "step": 13720 + }, + { + "epoch": 0.5, + "learning_rate": 1.4993591386823892e-07, + "loss": 0.6919, + "step": 13721 + }, + { + "epoch": 0.5, + "learning_rate": 1.4993225180356682e-07, + "loss": 0.6995, + "step": 13722 + }, + { + "epoch": 0.5, + "learning_rate": 1.4992858973889478e-07, + "loss": 0.6946, + "step": 13723 + }, + { + "epoch": 0.5, + "learning_rate": 1.499249276742227e-07, + "loss": 0.6981, + "step": 13724 + }, + { + "epoch": 0.5, + "learning_rate": 1.4992126560955066e-07, + "loss": 0.6879, + "step": 13725 + }, + { + "epoch": 0.5, + "learning_rate": 1.4991760354487858e-07, + "loss": 0.69, + "step": 13726 + }, + { + "epoch": 0.5, + "learning_rate": 1.4991394148020654e-07, + "loss": 0.6875, + "step": 13727 + }, + { + "epoch": 0.5, + "learning_rate": 1.4991027941553446e-07, + "loss": 0.6825, + "step": 13728 + }, + { + "epoch": 0.5, + "learning_rate": 1.4990661735086242e-07, + "loss": 0.6947, + "step": 13729 + }, + { + "epoch": 0.5, + "learning_rate": 1.4990295528619034e-07, + "loss": 0.6874, + "step": 13730 + }, + { + "epoch": 0.5, + "learning_rate": 1.498992932215183e-07, + "loss": 0.7013, + "step": 13731 + }, + { + "epoch": 0.5, + "learning_rate": 1.4989563115684622e-07, + "loss": 0.7011, + "step": 13732 + }, + { + "epoch": 0.5, + "learning_rate": 1.4989196909217417e-07, + "loss": 0.6775, + "step": 13733 + }, + { + "epoch": 0.5, + "learning_rate": 1.498883070275021e-07, + "loss": 0.6958, + "step": 13734 + }, + { + "epoch": 0.5, + "learning_rate": 1.4988464496283005e-07, + "loss": 0.6917, + "step": 13735 + }, + { + "epoch": 0.5, + "learning_rate": 1.4988098289815798e-07, + "loss": 0.701, + "step": 13736 + }, + { + "epoch": 0.5, + "learning_rate": 1.498773208334859e-07, + "loss": 0.6958, + "step": 13737 + }, + { + "epoch": 0.5, + "learning_rate": 1.4987365876881386e-07, + "loss": 0.6975, + "step": 13738 + }, + { + "epoch": 0.5, + "learning_rate": 1.498699967041418e-07, + "loss": 0.6937, + "step": 13739 + }, + { + "epoch": 0.5, + "learning_rate": 1.4986633463946971e-07, + "loss": 0.6914, + "step": 13740 + }, + { + "epoch": 0.5, + "learning_rate": 1.4986267257479767e-07, + "loss": 0.6976, + "step": 13741 + }, + { + "epoch": 0.5, + "learning_rate": 1.498590105101256e-07, + "loss": 0.6936, + "step": 13742 + }, + { + "epoch": 0.5, + "learning_rate": 1.4985534844545352e-07, + "loss": 0.6925, + "step": 13743 + }, + { + "epoch": 0.5, + "learning_rate": 1.4985168638078147e-07, + "loss": 0.6957, + "step": 13744 + }, + { + "epoch": 0.5, + "learning_rate": 1.498480243161094e-07, + "loss": 0.6924, + "step": 13745 + }, + { + "epoch": 0.5, + "learning_rate": 1.4984436225143735e-07, + "loss": 0.6853, + "step": 13746 + }, + { + "epoch": 0.5, + "learning_rate": 1.4984070018676528e-07, + "loss": 0.6751, + "step": 13747 + }, + { + "epoch": 0.5, + "learning_rate": 1.4983703812209323e-07, + "loss": 0.7026, + "step": 13748 + }, + { + "epoch": 0.5, + "learning_rate": 1.4983337605742116e-07, + "loss": 0.6921, + "step": 13749 + }, + { + "epoch": 0.5, + "learning_rate": 1.4982971399274911e-07, + "loss": 0.6999, + "step": 13750 + }, + { + "epoch": 0.5, + "learning_rate": 1.4982605192807704e-07, + "loss": 0.6874, + "step": 13751 + }, + { + "epoch": 0.5, + "learning_rate": 1.49822389863405e-07, + "loss": 0.6947, + "step": 13752 + }, + { + "epoch": 0.5, + "learning_rate": 1.4981872779873292e-07, + "loss": 0.6986, + "step": 13753 + }, + { + "epoch": 0.5, + "learning_rate": 1.4981506573406087e-07, + "loss": 0.701, + "step": 13754 + }, + { + "epoch": 0.5, + "learning_rate": 1.498114036693888e-07, + "loss": 0.6846, + "step": 13755 + }, + { + "epoch": 0.5, + "learning_rate": 1.4980774160471675e-07, + "loss": 0.7077, + "step": 13756 + }, + { + "epoch": 0.5, + "learning_rate": 1.4980407954004468e-07, + "loss": 0.6944, + "step": 13757 + }, + { + "epoch": 0.5, + "learning_rate": 1.498004174753726e-07, + "loss": 0.6804, + "step": 13758 + }, + { + "epoch": 0.5, + "learning_rate": 1.4979675541070053e-07, + "loss": 0.6876, + "step": 13759 + }, + { + "epoch": 0.5, + "learning_rate": 1.4979309334602849e-07, + "loss": 0.6957, + "step": 13760 + }, + { + "epoch": 0.5, + "learning_rate": 1.4978943128135641e-07, + "loss": 0.7187, + "step": 13761 + }, + { + "epoch": 0.5, + "learning_rate": 1.4978576921668437e-07, + "loss": 0.6993, + "step": 13762 + }, + { + "epoch": 0.5, + "learning_rate": 1.497821071520123e-07, + "loss": 0.6972, + "step": 13763 + }, + { + "epoch": 0.5, + "learning_rate": 1.4977844508734022e-07, + "loss": 0.6836, + "step": 13764 + }, + { + "epoch": 0.5, + "learning_rate": 1.4977478302266817e-07, + "loss": 0.6888, + "step": 13765 + }, + { + "epoch": 0.5, + "learning_rate": 1.497711209579961e-07, + "loss": 0.696, + "step": 13766 + }, + { + "epoch": 0.5, + "learning_rate": 1.4976745889332405e-07, + "loss": 0.7016, + "step": 13767 + }, + { + "epoch": 0.5, + "learning_rate": 1.4976379682865198e-07, + "loss": 0.692, + "step": 13768 + }, + { + "epoch": 0.5, + "learning_rate": 1.4976013476397993e-07, + "loss": 0.689, + "step": 13769 + }, + { + "epoch": 0.5, + "learning_rate": 1.4975647269930786e-07, + "loss": 0.6937, + "step": 13770 + }, + { + "epoch": 0.5, + "learning_rate": 1.497528106346358e-07, + "loss": 0.6992, + "step": 13771 + }, + { + "epoch": 0.5, + "learning_rate": 1.4974914856996374e-07, + "loss": 0.6902, + "step": 13772 + }, + { + "epoch": 0.5, + "learning_rate": 1.497454865052917e-07, + "loss": 0.6927, + "step": 13773 + }, + { + "epoch": 0.5, + "learning_rate": 1.4974182444061962e-07, + "loss": 0.694, + "step": 13774 + }, + { + "epoch": 0.5, + "learning_rate": 1.4973816237594757e-07, + "loss": 0.6965, + "step": 13775 + }, + { + "epoch": 0.5, + "learning_rate": 1.4973450031127547e-07, + "loss": 0.6984, + "step": 13776 + }, + { + "epoch": 0.5, + "learning_rate": 1.4973083824660343e-07, + "loss": 0.7029, + "step": 13777 + }, + { + "epoch": 0.5, + "learning_rate": 1.4972717618193135e-07, + "loss": 0.6893, + "step": 13778 + }, + { + "epoch": 0.5, + "learning_rate": 1.497235141172593e-07, + "loss": 0.6923, + "step": 13779 + }, + { + "epoch": 0.5, + "learning_rate": 1.4971985205258723e-07, + "loss": 0.6742, + "step": 13780 + }, + { + "epoch": 0.5, + "learning_rate": 1.4971618998791519e-07, + "loss": 0.6923, + "step": 13781 + }, + { + "epoch": 0.5, + "learning_rate": 1.497125279232431e-07, + "loss": 0.682, + "step": 13782 + }, + { + "epoch": 0.5, + "learning_rate": 1.4970886585857107e-07, + "loss": 0.6853, + "step": 13783 + }, + { + "epoch": 0.5, + "learning_rate": 1.49705203793899e-07, + "loss": 0.6973, + "step": 13784 + }, + { + "epoch": 0.5, + "learning_rate": 1.4970154172922695e-07, + "loss": 0.7038, + "step": 13785 + }, + { + "epoch": 0.5, + "learning_rate": 1.4969787966455487e-07, + "loss": 0.6954, + "step": 13786 + }, + { + "epoch": 0.5, + "learning_rate": 1.496942175998828e-07, + "loss": 0.6914, + "step": 13787 + }, + { + "epoch": 0.5, + "learning_rate": 1.4969055553521075e-07, + "loss": 0.6892, + "step": 13788 + }, + { + "epoch": 0.5, + "learning_rate": 1.4968689347053868e-07, + "loss": 0.7026, + "step": 13789 + }, + { + "epoch": 0.5, + "learning_rate": 1.4968323140586663e-07, + "loss": 0.6985, + "step": 13790 + }, + { + "epoch": 0.5, + "learning_rate": 1.4967956934119456e-07, + "loss": 0.6946, + "step": 13791 + }, + { + "epoch": 0.5, + "learning_rate": 1.496759072765225e-07, + "loss": 0.6949, + "step": 13792 + }, + { + "epoch": 0.5, + "learning_rate": 1.4967224521185044e-07, + "loss": 0.6946, + "step": 13793 + }, + { + "epoch": 0.5, + "learning_rate": 1.4966858314717836e-07, + "loss": 0.7025, + "step": 13794 + }, + { + "epoch": 0.5, + "learning_rate": 1.496649210825063e-07, + "loss": 0.7079, + "step": 13795 + }, + { + "epoch": 0.5, + "learning_rate": 1.4966125901783424e-07, + "loss": 0.683, + "step": 13796 + }, + { + "epoch": 0.5, + "learning_rate": 1.4965759695316217e-07, + "loss": 0.6951, + "step": 13797 + }, + { + "epoch": 0.5, + "learning_rate": 1.4965393488849012e-07, + "loss": 0.6877, + "step": 13798 + }, + { + "epoch": 0.5, + "learning_rate": 1.4965027282381805e-07, + "loss": 0.685, + "step": 13799 + }, + { + "epoch": 0.5, + "learning_rate": 1.49646610759146e-07, + "loss": 0.692, + "step": 13800 + }, + { + "epoch": 0.5, + "learning_rate": 1.4964294869447393e-07, + "loss": 0.6878, + "step": 13801 + }, + { + "epoch": 0.5, + "learning_rate": 1.4963928662980188e-07, + "loss": 0.7006, + "step": 13802 + }, + { + "epoch": 0.51, + "learning_rate": 1.496356245651298e-07, + "loss": 0.6813, + "step": 13803 + }, + { + "epoch": 0.51, + "learning_rate": 1.4963196250045776e-07, + "loss": 0.6964, + "step": 13804 + }, + { + "epoch": 0.51, + "learning_rate": 1.496283004357857e-07, + "loss": 0.7028, + "step": 13805 + }, + { + "epoch": 0.51, + "learning_rate": 1.4962463837111364e-07, + "loss": 0.7003, + "step": 13806 + }, + { + "epoch": 0.51, + "learning_rate": 1.4962097630644157e-07, + "loss": 0.6943, + "step": 13807 + }, + { + "epoch": 0.51, + "learning_rate": 1.496173142417695e-07, + "loss": 0.6961, + "step": 13808 + }, + { + "epoch": 0.51, + "learning_rate": 1.4961365217709745e-07, + "loss": 0.6891, + "step": 13809 + }, + { + "epoch": 0.51, + "learning_rate": 1.4960999011242538e-07, + "loss": 0.6905, + "step": 13810 + }, + { + "epoch": 0.51, + "learning_rate": 1.4960632804775333e-07, + "loss": 0.6943, + "step": 13811 + }, + { + "epoch": 0.51, + "learning_rate": 1.4960266598308126e-07, + "loss": 0.7043, + "step": 13812 + }, + { + "epoch": 0.51, + "learning_rate": 1.4959900391840918e-07, + "loss": 0.6935, + "step": 13813 + }, + { + "epoch": 0.51, + "learning_rate": 1.495953418537371e-07, + "loss": 0.6948, + "step": 13814 + }, + { + "epoch": 0.51, + "learning_rate": 1.4959167978906506e-07, + "loss": 0.7067, + "step": 13815 + }, + { + "epoch": 0.51, + "learning_rate": 1.49588017724393e-07, + "loss": 0.7092, + "step": 13816 + }, + { + "epoch": 0.51, + "learning_rate": 1.4958435565972094e-07, + "loss": 0.6812, + "step": 13817 + }, + { + "epoch": 0.51, + "learning_rate": 1.4958069359504887e-07, + "loss": 0.6749, + "step": 13818 + }, + { + "epoch": 0.51, + "learning_rate": 1.4957703153037682e-07, + "loss": 0.7075, + "step": 13819 + }, + { + "epoch": 0.51, + "learning_rate": 1.4957336946570475e-07, + "loss": 0.7014, + "step": 13820 + }, + { + "epoch": 0.51, + "learning_rate": 1.495697074010327e-07, + "loss": 0.6972, + "step": 13821 + }, + { + "epoch": 0.51, + "learning_rate": 1.4956604533636063e-07, + "loss": 0.6906, + "step": 13822 + }, + { + "epoch": 0.51, + "learning_rate": 1.4956238327168858e-07, + "loss": 0.7021, + "step": 13823 + }, + { + "epoch": 0.51, + "learning_rate": 1.495587212070165e-07, + "loss": 0.6875, + "step": 13824 + }, + { + "epoch": 0.51, + "learning_rate": 1.4955505914234446e-07, + "loss": 0.6862, + "step": 13825 + }, + { + "epoch": 0.51, + "learning_rate": 1.495513970776724e-07, + "loss": 0.687, + "step": 13826 + }, + { + "epoch": 0.51, + "learning_rate": 1.4954773501300034e-07, + "loss": 0.6829, + "step": 13827 + }, + { + "epoch": 0.51, + "learning_rate": 1.4954407294832827e-07, + "loss": 0.6931, + "step": 13828 + }, + { + "epoch": 0.51, + "learning_rate": 1.4954041088365622e-07, + "loss": 0.6869, + "step": 13829 + }, + { + "epoch": 0.51, + "learning_rate": 1.4953674881898412e-07, + "loss": 0.6875, + "step": 13830 + }, + { + "epoch": 0.51, + "learning_rate": 1.4953308675431208e-07, + "loss": 0.6949, + "step": 13831 + }, + { + "epoch": 0.51, + "learning_rate": 1.4952942468964e-07, + "loss": 0.6976, + "step": 13832 + }, + { + "epoch": 0.51, + "learning_rate": 1.4952576262496796e-07, + "loss": 0.6866, + "step": 13833 + }, + { + "epoch": 0.51, + "learning_rate": 1.4952210056029588e-07, + "loss": 0.7123, + "step": 13834 + }, + { + "epoch": 0.51, + "learning_rate": 1.495184384956238e-07, + "loss": 0.6903, + "step": 13835 + }, + { + "epoch": 0.51, + "learning_rate": 1.4951477643095176e-07, + "loss": 0.7002, + "step": 13836 + }, + { + "epoch": 0.51, + "learning_rate": 1.495111143662797e-07, + "loss": 0.6934, + "step": 13837 + }, + { + "epoch": 0.51, + "learning_rate": 1.4950745230160764e-07, + "loss": 0.689, + "step": 13838 + }, + { + "epoch": 0.51, + "learning_rate": 1.4950379023693557e-07, + "loss": 0.6953, + "step": 13839 + }, + { + "epoch": 0.51, + "learning_rate": 1.4950012817226352e-07, + "loss": 0.6976, + "step": 13840 + }, + { + "epoch": 0.51, + "learning_rate": 1.4949646610759145e-07, + "loss": 0.7034, + "step": 13841 + }, + { + "epoch": 0.51, + "learning_rate": 1.494928040429194e-07, + "loss": 0.6835, + "step": 13842 + }, + { + "epoch": 0.51, + "learning_rate": 1.4948914197824733e-07, + "loss": 0.6925, + "step": 13843 + }, + { + "epoch": 0.51, + "learning_rate": 1.4948547991357528e-07, + "loss": 0.6871, + "step": 13844 + }, + { + "epoch": 0.51, + "learning_rate": 1.494818178489032e-07, + "loss": 0.6876, + "step": 13845 + }, + { + "epoch": 0.51, + "learning_rate": 1.4947815578423116e-07, + "loss": 0.6811, + "step": 13846 + }, + { + "epoch": 0.51, + "learning_rate": 1.494744937195591e-07, + "loss": 0.6969, + "step": 13847 + }, + { + "epoch": 0.51, + "learning_rate": 1.4947083165488702e-07, + "loss": 0.6815, + "step": 13848 + }, + { + "epoch": 0.51, + "learning_rate": 1.4946716959021494e-07, + "loss": 0.6985, + "step": 13849 + }, + { + "epoch": 0.51, + "learning_rate": 1.494635075255429e-07, + "loss": 0.6885, + "step": 13850 + }, + { + "epoch": 0.51, + "learning_rate": 1.4945984546087082e-07, + "loss": 0.6883, + "step": 13851 + }, + { + "epoch": 0.51, + "learning_rate": 1.4945618339619877e-07, + "loss": 0.6949, + "step": 13852 + }, + { + "epoch": 0.51, + "learning_rate": 1.494525213315267e-07, + "loss": 0.6887, + "step": 13853 + }, + { + "epoch": 0.51, + "learning_rate": 1.4944885926685465e-07, + "loss": 0.6945, + "step": 13854 + }, + { + "epoch": 0.51, + "learning_rate": 1.4944519720218258e-07, + "loss": 0.6977, + "step": 13855 + }, + { + "epoch": 0.51, + "learning_rate": 1.4944153513751053e-07, + "loss": 0.6964, + "step": 13856 + }, + { + "epoch": 0.51, + "learning_rate": 1.4943787307283846e-07, + "loss": 0.6913, + "step": 13857 + }, + { + "epoch": 0.51, + "learning_rate": 1.494342110081664e-07, + "loss": 0.7094, + "step": 13858 + }, + { + "epoch": 0.51, + "learning_rate": 1.4943054894349434e-07, + "loss": 0.6932, + "step": 13859 + }, + { + "epoch": 0.51, + "learning_rate": 1.4942688687882227e-07, + "loss": 0.6967, + "step": 13860 + }, + { + "epoch": 0.51, + "learning_rate": 1.4942322481415022e-07, + "loss": 0.7064, + "step": 13861 + }, + { + "epoch": 0.51, + "learning_rate": 1.4941956274947815e-07, + "loss": 0.6922, + "step": 13862 + }, + { + "epoch": 0.51, + "learning_rate": 1.494159006848061e-07, + "loss": 0.6852, + "step": 13863 + }, + { + "epoch": 0.51, + "learning_rate": 1.4941223862013403e-07, + "loss": 0.6936, + "step": 13864 + }, + { + "epoch": 0.51, + "learning_rate": 1.4940857655546198e-07, + "loss": 0.7031, + "step": 13865 + }, + { + "epoch": 0.51, + "learning_rate": 1.4940491449078988e-07, + "loss": 0.6888, + "step": 13866 + }, + { + "epoch": 0.51, + "learning_rate": 1.4940125242611783e-07, + "loss": 0.6993, + "step": 13867 + }, + { + "epoch": 0.51, + "learning_rate": 1.4939759036144576e-07, + "loss": 0.6885, + "step": 13868 + }, + { + "epoch": 0.51, + "learning_rate": 1.4939392829677371e-07, + "loss": 0.6889, + "step": 13869 + }, + { + "epoch": 0.51, + "learning_rate": 1.4939026623210164e-07, + "loss": 0.6972, + "step": 13870 + }, + { + "epoch": 0.51, + "learning_rate": 1.493866041674296e-07, + "loss": 0.7197, + "step": 13871 + }, + { + "epoch": 0.51, + "learning_rate": 1.4938294210275752e-07, + "loss": 0.7054, + "step": 13872 + }, + { + "epoch": 0.51, + "learning_rate": 1.4937928003808547e-07, + "loss": 0.6869, + "step": 13873 + }, + { + "epoch": 0.51, + "learning_rate": 1.493756179734134e-07, + "loss": 0.6969, + "step": 13874 + }, + { + "epoch": 0.51, + "learning_rate": 1.4937195590874135e-07, + "loss": 0.6797, + "step": 13875 + }, + { + "epoch": 0.51, + "learning_rate": 1.4936829384406928e-07, + "loss": 0.7019, + "step": 13876 + }, + { + "epoch": 0.51, + "learning_rate": 1.4936463177939723e-07, + "loss": 0.6964, + "step": 13877 + }, + { + "epoch": 0.51, + "learning_rate": 1.4936096971472516e-07, + "loss": 0.696, + "step": 13878 + }, + { + "epoch": 0.51, + "learning_rate": 1.4935730765005309e-07, + "loss": 0.6875, + "step": 13879 + }, + { + "epoch": 0.51, + "learning_rate": 1.4935364558538104e-07, + "loss": 0.6858, + "step": 13880 + }, + { + "epoch": 0.51, + "learning_rate": 1.4934998352070897e-07, + "loss": 0.6877, + "step": 13881 + }, + { + "epoch": 0.51, + "learning_rate": 1.4934632145603692e-07, + "loss": 0.6928, + "step": 13882 + }, + { + "epoch": 0.51, + "learning_rate": 1.4934265939136485e-07, + "loss": 0.6952, + "step": 13883 + }, + { + "epoch": 0.51, + "learning_rate": 1.4933899732669277e-07, + "loss": 0.6843, + "step": 13884 + }, + { + "epoch": 0.51, + "learning_rate": 1.493353352620207e-07, + "loss": 0.6971, + "step": 13885 + }, + { + "epoch": 0.51, + "learning_rate": 1.4933167319734865e-07, + "loss": 0.705, + "step": 13886 + }, + { + "epoch": 0.51, + "learning_rate": 1.4932801113267658e-07, + "loss": 0.6948, + "step": 13887 + }, + { + "epoch": 0.51, + "learning_rate": 1.4932434906800453e-07, + "loss": 0.6725, + "step": 13888 + }, + { + "epoch": 0.51, + "learning_rate": 1.4932068700333246e-07, + "loss": 0.7015, + "step": 13889 + }, + { + "epoch": 0.51, + "learning_rate": 1.493170249386604e-07, + "loss": 0.6897, + "step": 13890 + }, + { + "epoch": 0.51, + "learning_rate": 1.4931336287398834e-07, + "loss": 0.6721, + "step": 13891 + }, + { + "epoch": 0.51, + "learning_rate": 1.493097008093163e-07, + "loss": 0.6805, + "step": 13892 + }, + { + "epoch": 0.51, + "learning_rate": 1.4930603874464422e-07, + "loss": 0.6853, + "step": 13893 + }, + { + "epoch": 0.51, + "learning_rate": 1.4930237667997217e-07, + "loss": 0.7068, + "step": 13894 + }, + { + "epoch": 0.51, + "learning_rate": 1.492987146153001e-07, + "loss": 0.6961, + "step": 13895 + }, + { + "epoch": 0.51, + "learning_rate": 1.4929505255062805e-07, + "loss": 0.6927, + "step": 13896 + }, + { + "epoch": 0.51, + "learning_rate": 1.4929139048595598e-07, + "loss": 0.7081, + "step": 13897 + }, + { + "epoch": 0.51, + "learning_rate": 1.4928772842128393e-07, + "loss": 0.6935, + "step": 13898 + }, + { + "epoch": 0.51, + "learning_rate": 1.4928406635661186e-07, + "loss": 0.7001, + "step": 13899 + }, + { + "epoch": 0.51, + "learning_rate": 1.492804042919398e-07, + "loss": 0.6941, + "step": 13900 + }, + { + "epoch": 0.51, + "learning_rate": 1.492767422272677e-07, + "loss": 0.6957, + "step": 13901 + }, + { + "epoch": 0.51, + "learning_rate": 1.4927308016259567e-07, + "loss": 0.6819, + "step": 13902 + }, + { + "epoch": 0.51, + "learning_rate": 1.492694180979236e-07, + "loss": 0.695, + "step": 13903 + }, + { + "epoch": 0.51, + "learning_rate": 1.4926575603325155e-07, + "loss": 0.677, + "step": 13904 + }, + { + "epoch": 0.51, + "learning_rate": 1.4926209396857947e-07, + "loss": 0.7001, + "step": 13905 + }, + { + "epoch": 0.51, + "learning_rate": 1.492584319039074e-07, + "loss": 0.7061, + "step": 13906 + }, + { + "epoch": 0.51, + "learning_rate": 1.4925476983923535e-07, + "loss": 0.6951, + "step": 13907 + }, + { + "epoch": 0.51, + "learning_rate": 1.4925110777456328e-07, + "loss": 0.6919, + "step": 13908 + }, + { + "epoch": 0.51, + "learning_rate": 1.4924744570989123e-07, + "loss": 0.6995, + "step": 13909 + }, + { + "epoch": 0.51, + "learning_rate": 1.4924378364521916e-07, + "loss": 0.674, + "step": 13910 + }, + { + "epoch": 0.51, + "learning_rate": 1.492401215805471e-07, + "loss": 0.6852, + "step": 13911 + }, + { + "epoch": 0.51, + "learning_rate": 1.4923645951587504e-07, + "loss": 0.6867, + "step": 13912 + }, + { + "epoch": 0.51, + "learning_rate": 1.49232797451203e-07, + "loss": 0.6826, + "step": 13913 + }, + { + "epoch": 0.51, + "learning_rate": 1.4922913538653092e-07, + "loss": 0.7005, + "step": 13914 + }, + { + "epoch": 0.51, + "learning_rate": 1.4922547332185887e-07, + "loss": 0.702, + "step": 13915 + }, + { + "epoch": 0.51, + "learning_rate": 1.492218112571868e-07, + "loss": 0.6803, + "step": 13916 + }, + { + "epoch": 0.51, + "learning_rate": 1.4921814919251475e-07, + "loss": 0.6869, + "step": 13917 + }, + { + "epoch": 0.51, + "learning_rate": 1.4921448712784268e-07, + "loss": 0.6935, + "step": 13918 + }, + { + "epoch": 0.51, + "learning_rate": 1.492108250631706e-07, + "loss": 0.6887, + "step": 13919 + }, + { + "epoch": 0.51, + "learning_rate": 1.4920716299849853e-07, + "loss": 0.6872, + "step": 13920 + }, + { + "epoch": 0.51, + "learning_rate": 1.4920350093382648e-07, + "loss": 0.6979, + "step": 13921 + }, + { + "epoch": 0.51, + "learning_rate": 1.491998388691544e-07, + "loss": 0.7168, + "step": 13922 + }, + { + "epoch": 0.51, + "learning_rate": 1.4919617680448236e-07, + "loss": 0.684, + "step": 13923 + }, + { + "epoch": 0.51, + "learning_rate": 1.491925147398103e-07, + "loss": 0.6834, + "step": 13924 + }, + { + "epoch": 0.51, + "learning_rate": 1.4918885267513824e-07, + "loss": 0.7025, + "step": 13925 + }, + { + "epoch": 0.51, + "learning_rate": 1.4918519061046617e-07, + "loss": 0.692, + "step": 13926 + }, + { + "epoch": 0.51, + "learning_rate": 1.4918152854579412e-07, + "loss": 0.6957, + "step": 13927 + }, + { + "epoch": 0.51, + "learning_rate": 1.4917786648112205e-07, + "loss": 0.6809, + "step": 13928 + }, + { + "epoch": 0.51, + "learning_rate": 1.4917420441644998e-07, + "loss": 0.685, + "step": 13929 + }, + { + "epoch": 0.51, + "learning_rate": 1.4917054235177793e-07, + "loss": 0.6894, + "step": 13930 + }, + { + "epoch": 0.51, + "learning_rate": 1.4916688028710586e-07, + "loss": 0.6985, + "step": 13931 + }, + { + "epoch": 0.51, + "learning_rate": 1.491632182224338e-07, + "loss": 0.6913, + "step": 13932 + }, + { + "epoch": 0.51, + "learning_rate": 1.4915955615776174e-07, + "loss": 0.7214, + "step": 13933 + }, + { + "epoch": 0.51, + "learning_rate": 1.491558940930897e-07, + "loss": 0.6959, + "step": 13934 + }, + { + "epoch": 0.51, + "learning_rate": 1.4915223202841762e-07, + "loss": 0.6952, + "step": 13935 + }, + { + "epoch": 0.51, + "learning_rate": 1.4914856996374557e-07, + "loss": 0.7054, + "step": 13936 + }, + { + "epoch": 0.51, + "learning_rate": 1.4914490789907347e-07, + "loss": 0.6978, + "step": 13937 + }, + { + "epoch": 0.51, + "learning_rate": 1.4914124583440142e-07, + "loss": 0.6923, + "step": 13938 + }, + { + "epoch": 0.51, + "learning_rate": 1.4913758376972935e-07, + "loss": 0.6914, + "step": 13939 + }, + { + "epoch": 0.51, + "learning_rate": 1.491339217050573e-07, + "loss": 0.6806, + "step": 13940 + }, + { + "epoch": 0.51, + "learning_rate": 1.4913025964038523e-07, + "loss": 0.6854, + "step": 13941 + }, + { + "epoch": 0.51, + "learning_rate": 1.4912659757571318e-07, + "loss": 0.685, + "step": 13942 + }, + { + "epoch": 0.51, + "learning_rate": 1.491229355110411e-07, + "loss": 0.6823, + "step": 13943 + }, + { + "epoch": 0.51, + "learning_rate": 1.4911927344636906e-07, + "loss": 0.6912, + "step": 13944 + }, + { + "epoch": 0.51, + "learning_rate": 1.49115611381697e-07, + "loss": 0.704, + "step": 13945 + }, + { + "epoch": 0.51, + "learning_rate": 1.4911194931702494e-07, + "loss": 0.7037, + "step": 13946 + }, + { + "epoch": 0.51, + "learning_rate": 1.4910828725235287e-07, + "loss": 0.6874, + "step": 13947 + }, + { + "epoch": 0.51, + "learning_rate": 1.4910462518768082e-07, + "loss": 0.6969, + "step": 13948 + }, + { + "epoch": 0.51, + "learning_rate": 1.4910096312300875e-07, + "loss": 0.6879, + "step": 13949 + }, + { + "epoch": 0.51, + "learning_rate": 1.4909730105833668e-07, + "loss": 0.6938, + "step": 13950 + }, + { + "epoch": 0.51, + "learning_rate": 1.4909363899366463e-07, + "loss": 0.6903, + "step": 13951 + }, + { + "epoch": 0.51, + "learning_rate": 1.4908997692899256e-07, + "loss": 0.6816, + "step": 13952 + }, + { + "epoch": 0.51, + "learning_rate": 1.490863148643205e-07, + "loss": 0.6992, + "step": 13953 + }, + { + "epoch": 0.51, + "learning_rate": 1.4908265279964844e-07, + "loss": 0.6967, + "step": 13954 + }, + { + "epoch": 0.51, + "learning_rate": 1.4907899073497636e-07, + "loss": 0.698, + "step": 13955 + }, + { + "epoch": 0.51, + "learning_rate": 1.490753286703043e-07, + "loss": 0.6936, + "step": 13956 + }, + { + "epoch": 0.51, + "learning_rate": 1.4907166660563224e-07, + "loss": 0.688, + "step": 13957 + }, + { + "epoch": 0.51, + "learning_rate": 1.4906800454096017e-07, + "loss": 0.6925, + "step": 13958 + }, + { + "epoch": 0.51, + "learning_rate": 1.4906434247628812e-07, + "loss": 0.6931, + "step": 13959 + }, + { + "epoch": 0.51, + "learning_rate": 1.4906068041161605e-07, + "loss": 0.6949, + "step": 13960 + }, + { + "epoch": 0.51, + "learning_rate": 1.49057018346944e-07, + "loss": 0.6971, + "step": 13961 + }, + { + "epoch": 0.51, + "learning_rate": 1.4905335628227193e-07, + "loss": 0.6835, + "step": 13962 + }, + { + "epoch": 0.51, + "learning_rate": 1.4904969421759988e-07, + "loss": 0.681, + "step": 13963 + }, + { + "epoch": 0.51, + "learning_rate": 1.490460321529278e-07, + "loss": 0.6954, + "step": 13964 + }, + { + "epoch": 0.51, + "learning_rate": 1.4904237008825576e-07, + "loss": 0.7032, + "step": 13965 + }, + { + "epoch": 0.51, + "learning_rate": 1.490387080235837e-07, + "loss": 0.6945, + "step": 13966 + }, + { + "epoch": 0.51, + "learning_rate": 1.4903504595891164e-07, + "loss": 0.7024, + "step": 13967 + }, + { + "epoch": 0.51, + "learning_rate": 1.4903138389423957e-07, + "loss": 0.6853, + "step": 13968 + }, + { + "epoch": 0.51, + "learning_rate": 1.4902772182956752e-07, + "loss": 0.7084, + "step": 13969 + }, + { + "epoch": 0.51, + "learning_rate": 1.4902405976489545e-07, + "loss": 0.7026, + "step": 13970 + }, + { + "epoch": 0.51, + "learning_rate": 1.4902039770022337e-07, + "loss": 0.6984, + "step": 13971 + }, + { + "epoch": 0.51, + "learning_rate": 1.4901673563555133e-07, + "loss": 0.7067, + "step": 13972 + }, + { + "epoch": 0.51, + "learning_rate": 1.4901307357087925e-07, + "loss": 0.7031, + "step": 13973 + }, + { + "epoch": 0.51, + "learning_rate": 1.4900941150620718e-07, + "loss": 0.6924, + "step": 13974 + }, + { + "epoch": 0.51, + "learning_rate": 1.4900574944153513e-07, + "loss": 0.6882, + "step": 13975 + }, + { + "epoch": 0.51, + "learning_rate": 1.4900208737686306e-07, + "loss": 0.6977, + "step": 13976 + }, + { + "epoch": 0.51, + "learning_rate": 1.48998425312191e-07, + "loss": 0.6855, + "step": 13977 + }, + { + "epoch": 0.51, + "learning_rate": 1.4899476324751894e-07, + "loss": 0.6985, + "step": 13978 + }, + { + "epoch": 0.51, + "learning_rate": 1.4899110118284687e-07, + "loss": 0.698, + "step": 13979 + }, + { + "epoch": 0.51, + "learning_rate": 1.4898743911817482e-07, + "loss": 0.6956, + "step": 13980 + }, + { + "epoch": 0.51, + "learning_rate": 1.4898377705350275e-07, + "loss": 0.6912, + "step": 13981 + }, + { + "epoch": 0.51, + "learning_rate": 1.489801149888307e-07, + "loss": 0.701, + "step": 13982 + }, + { + "epoch": 0.51, + "learning_rate": 1.4897645292415863e-07, + "loss": 0.6946, + "step": 13983 + }, + { + "epoch": 0.51, + "learning_rate": 1.4897279085948658e-07, + "loss": 0.6984, + "step": 13984 + }, + { + "epoch": 0.51, + "learning_rate": 1.489691287948145e-07, + "loss": 0.6903, + "step": 13985 + }, + { + "epoch": 0.51, + "learning_rate": 1.4896546673014246e-07, + "loss": 0.6946, + "step": 13986 + }, + { + "epoch": 0.51, + "learning_rate": 1.489618046654704e-07, + "loss": 0.6922, + "step": 13987 + }, + { + "epoch": 0.51, + "learning_rate": 1.4895814260079834e-07, + "loss": 0.6905, + "step": 13988 + }, + { + "epoch": 0.51, + "learning_rate": 1.4895448053612627e-07, + "loss": 0.6948, + "step": 13989 + }, + { + "epoch": 0.51, + "learning_rate": 1.4895081847145422e-07, + "loss": 0.6988, + "step": 13990 + }, + { + "epoch": 0.51, + "learning_rate": 1.4894715640678212e-07, + "loss": 0.7019, + "step": 13991 + }, + { + "epoch": 0.51, + "learning_rate": 1.4894349434211007e-07, + "loss": 0.6973, + "step": 13992 + }, + { + "epoch": 0.51, + "learning_rate": 1.48939832277438e-07, + "loss": 0.6915, + "step": 13993 + }, + { + "epoch": 0.51, + "learning_rate": 1.4893617021276595e-07, + "loss": 0.6856, + "step": 13994 + }, + { + "epoch": 0.51, + "learning_rate": 1.4893250814809388e-07, + "loss": 0.689, + "step": 13995 + }, + { + "epoch": 0.51, + "learning_rate": 1.4892884608342183e-07, + "loss": 0.6944, + "step": 13996 + }, + { + "epoch": 0.51, + "learning_rate": 1.4892518401874976e-07, + "loss": 0.696, + "step": 13997 + }, + { + "epoch": 0.51, + "learning_rate": 1.4892152195407771e-07, + "loss": 0.7098, + "step": 13998 + }, + { + "epoch": 0.51, + "learning_rate": 1.4891785988940564e-07, + "loss": 0.676, + "step": 13999 + }, + { + "epoch": 0.51, + "learning_rate": 1.4891419782473357e-07, + "loss": 0.6938, + "step": 14000 + }, + { + "epoch": 0.51, + "learning_rate": 1.4891053576006152e-07, + "loss": 0.6862, + "step": 14001 + }, + { + "epoch": 0.51, + "learning_rate": 1.4890687369538945e-07, + "loss": 0.6886, + "step": 14002 + }, + { + "epoch": 0.51, + "learning_rate": 1.489032116307174e-07, + "loss": 0.7077, + "step": 14003 + }, + { + "epoch": 0.51, + "learning_rate": 1.4889954956604533e-07, + "loss": 0.6876, + "step": 14004 + }, + { + "epoch": 0.51, + "learning_rate": 1.4889588750137328e-07, + "loss": 0.6964, + "step": 14005 + }, + { + "epoch": 0.51, + "learning_rate": 1.488922254367012e-07, + "loss": 0.6911, + "step": 14006 + }, + { + "epoch": 0.51, + "learning_rate": 1.4888856337202916e-07, + "loss": 0.6944, + "step": 14007 + }, + { + "epoch": 0.51, + "learning_rate": 1.4888490130735709e-07, + "loss": 0.6984, + "step": 14008 + }, + { + "epoch": 0.51, + "learning_rate": 1.48881239242685e-07, + "loss": 0.6849, + "step": 14009 + }, + { + "epoch": 0.51, + "learning_rate": 1.4887757717801294e-07, + "loss": 0.6848, + "step": 14010 + }, + { + "epoch": 0.51, + "learning_rate": 1.488739151133409e-07, + "loss": 0.7057, + "step": 14011 + }, + { + "epoch": 0.51, + "learning_rate": 1.4887025304866882e-07, + "loss": 0.7155, + "step": 14012 + }, + { + "epoch": 0.51, + "learning_rate": 1.4886659098399677e-07, + "loss": 0.6995, + "step": 14013 + }, + { + "epoch": 0.51, + "learning_rate": 1.488629289193247e-07, + "loss": 0.6957, + "step": 14014 + }, + { + "epoch": 0.51, + "learning_rate": 1.4885926685465265e-07, + "loss": 0.6979, + "step": 14015 + }, + { + "epoch": 0.51, + "learning_rate": 1.4885560478998058e-07, + "loss": 0.693, + "step": 14016 + }, + { + "epoch": 0.51, + "learning_rate": 1.4885194272530853e-07, + "loss": 0.7019, + "step": 14017 + }, + { + "epoch": 0.51, + "learning_rate": 1.4884828066063646e-07, + "loss": 0.6882, + "step": 14018 + }, + { + "epoch": 0.51, + "learning_rate": 1.488446185959644e-07, + "loss": 0.6915, + "step": 14019 + }, + { + "epoch": 0.51, + "learning_rate": 1.4884095653129234e-07, + "loss": 0.6984, + "step": 14020 + }, + { + "epoch": 0.51, + "learning_rate": 1.4883729446662027e-07, + "loss": 0.6998, + "step": 14021 + }, + { + "epoch": 0.51, + "learning_rate": 1.4883363240194822e-07, + "loss": 0.6997, + "step": 14022 + }, + { + "epoch": 0.51, + "learning_rate": 1.4882997033727615e-07, + "loss": 0.6981, + "step": 14023 + }, + { + "epoch": 0.51, + "learning_rate": 1.488263082726041e-07, + "loss": 0.689, + "step": 14024 + }, + { + "epoch": 0.51, + "learning_rate": 1.4882264620793203e-07, + "loss": 0.6985, + "step": 14025 + }, + { + "epoch": 0.51, + "learning_rate": 1.4881898414325998e-07, + "loss": 0.6998, + "step": 14026 + }, + { + "epoch": 0.51, + "learning_rate": 1.4881532207858788e-07, + "loss": 0.684, + "step": 14027 + }, + { + "epoch": 0.51, + "learning_rate": 1.4881166001391583e-07, + "loss": 0.6831, + "step": 14028 + }, + { + "epoch": 0.51, + "learning_rate": 1.4880799794924376e-07, + "loss": 0.703, + "step": 14029 + }, + { + "epoch": 0.51, + "learning_rate": 1.488043358845717e-07, + "loss": 0.6971, + "step": 14030 + }, + { + "epoch": 0.51, + "learning_rate": 1.4880067381989964e-07, + "loss": 0.6896, + "step": 14031 + }, + { + "epoch": 0.51, + "learning_rate": 1.487970117552276e-07, + "loss": 0.6954, + "step": 14032 + }, + { + "epoch": 0.51, + "learning_rate": 1.4879334969055552e-07, + "loss": 0.6917, + "step": 14033 + }, + { + "epoch": 0.51, + "learning_rate": 1.4878968762588347e-07, + "loss": 0.6987, + "step": 14034 + }, + { + "epoch": 0.51, + "learning_rate": 1.487860255612114e-07, + "loss": 0.6971, + "step": 14035 + }, + { + "epoch": 0.51, + "learning_rate": 1.4878236349653935e-07, + "loss": 0.6909, + "step": 14036 + }, + { + "epoch": 0.51, + "learning_rate": 1.4877870143186728e-07, + "loss": 0.6851, + "step": 14037 + }, + { + "epoch": 0.51, + "learning_rate": 1.4877503936719523e-07, + "loss": 0.688, + "step": 14038 + }, + { + "epoch": 0.51, + "learning_rate": 1.4877137730252316e-07, + "loss": 0.6928, + "step": 14039 + }, + { + "epoch": 0.51, + "learning_rate": 1.487677152378511e-07, + "loss": 0.7028, + "step": 14040 + }, + { + "epoch": 0.51, + "learning_rate": 1.4876405317317904e-07, + "loss": 0.7082, + "step": 14041 + }, + { + "epoch": 0.51, + "learning_rate": 1.4876039110850696e-07, + "loss": 0.6993, + "step": 14042 + }, + { + "epoch": 0.51, + "learning_rate": 1.4875672904383492e-07, + "loss": 0.684, + "step": 14043 + }, + { + "epoch": 0.51, + "learning_rate": 1.4875306697916284e-07, + "loss": 0.6886, + "step": 14044 + }, + { + "epoch": 0.51, + "learning_rate": 1.4874940491449077e-07, + "loss": 0.6971, + "step": 14045 + }, + { + "epoch": 0.51, + "learning_rate": 1.4874574284981872e-07, + "loss": 0.6881, + "step": 14046 + }, + { + "epoch": 0.51, + "learning_rate": 1.4874208078514665e-07, + "loss": 0.6896, + "step": 14047 + }, + { + "epoch": 0.51, + "learning_rate": 1.4873841872047458e-07, + "loss": 0.7005, + "step": 14048 + }, + { + "epoch": 0.51, + "learning_rate": 1.4873475665580253e-07, + "loss": 0.6936, + "step": 14049 + }, + { + "epoch": 0.51, + "learning_rate": 1.4873109459113046e-07, + "loss": 0.6773, + "step": 14050 + }, + { + "epoch": 0.51, + "learning_rate": 1.487274325264584e-07, + "loss": 0.7032, + "step": 14051 + }, + { + "epoch": 0.51, + "learning_rate": 1.4872377046178634e-07, + "loss": 0.6837, + "step": 14052 + }, + { + "epoch": 0.51, + "learning_rate": 1.487201083971143e-07, + "loss": 0.698, + "step": 14053 + }, + { + "epoch": 0.51, + "learning_rate": 1.4871644633244222e-07, + "loss": 0.6905, + "step": 14054 + }, + { + "epoch": 0.51, + "learning_rate": 1.4871278426777017e-07, + "loss": 0.6873, + "step": 14055 + }, + { + "epoch": 0.51, + "learning_rate": 1.487091222030981e-07, + "loss": 0.7003, + "step": 14056 + }, + { + "epoch": 0.51, + "learning_rate": 1.4870546013842605e-07, + "loss": 0.6932, + "step": 14057 + }, + { + "epoch": 0.51, + "learning_rate": 1.4870179807375398e-07, + "loss": 0.6959, + "step": 14058 + }, + { + "epoch": 0.51, + "learning_rate": 1.4869813600908193e-07, + "loss": 0.6855, + "step": 14059 + }, + { + "epoch": 0.51, + "learning_rate": 1.4869447394440986e-07, + "loss": 0.6908, + "step": 14060 + }, + { + "epoch": 0.51, + "learning_rate": 1.486908118797378e-07, + "loss": 0.7012, + "step": 14061 + }, + { + "epoch": 0.51, + "learning_rate": 1.4868714981506574e-07, + "loss": 0.6969, + "step": 14062 + }, + { + "epoch": 0.51, + "learning_rate": 1.4868348775039366e-07, + "loss": 0.7037, + "step": 14063 + }, + { + "epoch": 0.51, + "learning_rate": 1.486798256857216e-07, + "loss": 0.7111, + "step": 14064 + }, + { + "epoch": 0.51, + "learning_rate": 1.4867616362104954e-07, + "loss": 0.6844, + "step": 14065 + }, + { + "epoch": 0.51, + "learning_rate": 1.4867250155637747e-07, + "loss": 0.7069, + "step": 14066 + }, + { + "epoch": 0.51, + "learning_rate": 1.4866883949170542e-07, + "loss": 0.7062, + "step": 14067 + }, + { + "epoch": 0.51, + "learning_rate": 1.4866517742703335e-07, + "loss": 0.6939, + "step": 14068 + }, + { + "epoch": 0.51, + "learning_rate": 1.486615153623613e-07, + "loss": 0.6902, + "step": 14069 + }, + { + "epoch": 0.51, + "learning_rate": 1.4865785329768923e-07, + "loss": 0.7058, + "step": 14070 + }, + { + "epoch": 0.51, + "learning_rate": 1.4865419123301716e-07, + "loss": 0.6865, + "step": 14071 + }, + { + "epoch": 0.51, + "learning_rate": 1.486505291683451e-07, + "loss": 0.686, + "step": 14072 + }, + { + "epoch": 0.51, + "learning_rate": 1.4864686710367304e-07, + "loss": 0.6963, + "step": 14073 + }, + { + "epoch": 0.51, + "learning_rate": 1.48643205039001e-07, + "loss": 0.6908, + "step": 14074 + }, + { + "epoch": 0.51, + "learning_rate": 1.4863954297432892e-07, + "loss": 0.6996, + "step": 14075 + }, + { + "epoch": 0.52, + "learning_rate": 1.4863588090965687e-07, + "loss": 0.7035, + "step": 14076 + }, + { + "epoch": 0.52, + "learning_rate": 1.486322188449848e-07, + "loss": 0.6894, + "step": 14077 + }, + { + "epoch": 0.52, + "learning_rate": 1.4862855678031275e-07, + "loss": 0.69, + "step": 14078 + }, + { + "epoch": 0.52, + "learning_rate": 1.4862489471564068e-07, + "loss": 0.6998, + "step": 14079 + }, + { + "epoch": 0.52, + "learning_rate": 1.4862123265096863e-07, + "loss": 0.6916, + "step": 14080 + }, + { + "epoch": 0.52, + "learning_rate": 1.4861757058629653e-07, + "loss": 0.7208, + "step": 14081 + }, + { + "epoch": 0.52, + "learning_rate": 1.4861390852162448e-07, + "loss": 0.7084, + "step": 14082 + }, + { + "epoch": 0.52, + "learning_rate": 1.486102464569524e-07, + "loss": 0.6857, + "step": 14083 + }, + { + "epoch": 0.52, + "learning_rate": 1.4860658439228036e-07, + "loss": 0.6992, + "step": 14084 + }, + { + "epoch": 0.52, + "learning_rate": 1.486029223276083e-07, + "loss": 0.6906, + "step": 14085 + }, + { + "epoch": 0.52, + "learning_rate": 1.4859926026293624e-07, + "loss": 0.6981, + "step": 14086 + }, + { + "epoch": 0.52, + "learning_rate": 1.4859559819826417e-07, + "loss": 0.7073, + "step": 14087 + }, + { + "epoch": 0.52, + "learning_rate": 1.4859193613359212e-07, + "loss": 0.701, + "step": 14088 + }, + { + "epoch": 0.52, + "learning_rate": 1.4858827406892005e-07, + "loss": 0.6902, + "step": 14089 + }, + { + "epoch": 0.52, + "learning_rate": 1.48584612004248e-07, + "loss": 0.6895, + "step": 14090 + }, + { + "epoch": 0.52, + "learning_rate": 1.4858094993957593e-07, + "loss": 0.7088, + "step": 14091 + }, + { + "epoch": 0.52, + "learning_rate": 1.4857728787490385e-07, + "loss": 0.6956, + "step": 14092 + }, + { + "epoch": 0.52, + "learning_rate": 1.485736258102318e-07, + "loss": 0.702, + "step": 14093 + }, + { + "epoch": 0.52, + "learning_rate": 1.4856996374555973e-07, + "loss": 0.6934, + "step": 14094 + }, + { + "epoch": 0.52, + "learning_rate": 1.485663016808877e-07, + "loss": 0.6929, + "step": 14095 + }, + { + "epoch": 0.52, + "learning_rate": 1.4856263961621561e-07, + "loss": 0.6919, + "step": 14096 + }, + { + "epoch": 0.52, + "learning_rate": 1.4855897755154357e-07, + "loss": 0.6968, + "step": 14097 + }, + { + "epoch": 0.52, + "learning_rate": 1.485553154868715e-07, + "loss": 0.7006, + "step": 14098 + }, + { + "epoch": 0.52, + "learning_rate": 1.4855165342219942e-07, + "loss": 0.6929, + "step": 14099 + }, + { + "epoch": 0.52, + "learning_rate": 1.4854799135752735e-07, + "loss": 0.7061, + "step": 14100 + }, + { + "epoch": 0.52, + "learning_rate": 1.485443292928553e-07, + "loss": 0.6774, + "step": 14101 + }, + { + "epoch": 0.52, + "learning_rate": 1.4854066722818323e-07, + "loss": 0.6962, + "step": 14102 + }, + { + "epoch": 0.52, + "learning_rate": 1.4853700516351118e-07, + "loss": 0.6859, + "step": 14103 + }, + { + "epoch": 0.52, + "learning_rate": 1.485333430988391e-07, + "loss": 0.6838, + "step": 14104 + }, + { + "epoch": 0.52, + "learning_rate": 1.4852968103416706e-07, + "loss": 0.6857, + "step": 14105 + }, + { + "epoch": 0.52, + "learning_rate": 1.48526018969495e-07, + "loss": 0.6844, + "step": 14106 + }, + { + "epoch": 0.52, + "learning_rate": 1.4852235690482294e-07, + "loss": 0.6995, + "step": 14107 + }, + { + "epoch": 0.52, + "learning_rate": 1.4851869484015087e-07, + "loss": 0.6941, + "step": 14108 + }, + { + "epoch": 0.52, + "learning_rate": 1.4851503277547882e-07, + "loss": 0.6885, + "step": 14109 + }, + { + "epoch": 0.52, + "learning_rate": 1.4851137071080675e-07, + "loss": 0.6977, + "step": 14110 + }, + { + "epoch": 0.52, + "learning_rate": 1.485077086461347e-07, + "loss": 0.6886, + "step": 14111 + }, + { + "epoch": 0.52, + "learning_rate": 1.4850404658146263e-07, + "loss": 0.6872, + "step": 14112 + }, + { + "epoch": 0.52, + "learning_rate": 1.4850038451679055e-07, + "loss": 0.6975, + "step": 14113 + }, + { + "epoch": 0.52, + "learning_rate": 1.484967224521185e-07, + "loss": 0.6999, + "step": 14114 + }, + { + "epoch": 0.52, + "learning_rate": 1.4849306038744643e-07, + "loss": 0.6927, + "step": 14115 + }, + { + "epoch": 0.52, + "learning_rate": 1.4848939832277439e-07, + "loss": 0.6931, + "step": 14116 + }, + { + "epoch": 0.52, + "learning_rate": 1.4848573625810231e-07, + "loss": 0.6912, + "step": 14117 + }, + { + "epoch": 0.52, + "learning_rate": 1.4848207419343024e-07, + "loss": 0.6949, + "step": 14118 + }, + { + "epoch": 0.52, + "learning_rate": 1.4847841212875817e-07, + "loss": 0.6906, + "step": 14119 + }, + { + "epoch": 0.52, + "learning_rate": 1.4847475006408612e-07, + "loss": 0.6929, + "step": 14120 + }, + { + "epoch": 0.52, + "learning_rate": 1.4847108799941405e-07, + "loss": 0.7023, + "step": 14121 + }, + { + "epoch": 0.52, + "learning_rate": 1.48467425934742e-07, + "loss": 0.6821, + "step": 14122 + }, + { + "epoch": 0.52, + "learning_rate": 1.4846376387006993e-07, + "loss": 0.699, + "step": 14123 + }, + { + "epoch": 0.52, + "learning_rate": 1.4846010180539788e-07, + "loss": 0.698, + "step": 14124 + }, + { + "epoch": 0.52, + "learning_rate": 1.484564397407258e-07, + "loss": 0.7031, + "step": 14125 + }, + { + "epoch": 0.52, + "learning_rate": 1.4845277767605376e-07, + "loss": 0.6852, + "step": 14126 + }, + { + "epoch": 0.52, + "learning_rate": 1.4844911561138169e-07, + "loss": 0.6944, + "step": 14127 + }, + { + "epoch": 0.52, + "learning_rate": 1.4844545354670964e-07, + "loss": 0.6961, + "step": 14128 + }, + { + "epoch": 0.52, + "learning_rate": 1.4844179148203757e-07, + "loss": 0.6995, + "step": 14129 + }, + { + "epoch": 0.52, + "learning_rate": 1.4843812941736552e-07, + "loss": 0.7018, + "step": 14130 + }, + { + "epoch": 0.52, + "learning_rate": 1.4843446735269345e-07, + "loss": 0.703, + "step": 14131 + }, + { + "epoch": 0.52, + "learning_rate": 1.484308052880214e-07, + "loss": 0.6911, + "step": 14132 + }, + { + "epoch": 0.52, + "learning_rate": 1.4842714322334933e-07, + "loss": 0.6913, + "step": 14133 + }, + { + "epoch": 0.52, + "learning_rate": 1.4842348115867728e-07, + "loss": 0.6808, + "step": 14134 + }, + { + "epoch": 0.52, + "learning_rate": 1.4841981909400518e-07, + "loss": 0.6997, + "step": 14135 + }, + { + "epoch": 0.52, + "learning_rate": 1.4841615702933313e-07, + "loss": 0.6917, + "step": 14136 + }, + { + "epoch": 0.52, + "learning_rate": 1.4841249496466106e-07, + "loss": 0.7125, + "step": 14137 + }, + { + "epoch": 0.52, + "learning_rate": 1.48408832899989e-07, + "loss": 0.6955, + "step": 14138 + }, + { + "epoch": 0.52, + "learning_rate": 1.4840517083531694e-07, + "loss": 0.6828, + "step": 14139 + }, + { + "epoch": 0.52, + "learning_rate": 1.484015087706449e-07, + "loss": 0.6971, + "step": 14140 + }, + { + "epoch": 0.52, + "learning_rate": 1.4839784670597282e-07, + "loss": 0.6953, + "step": 14141 + }, + { + "epoch": 0.52, + "learning_rate": 1.4839418464130075e-07, + "loss": 0.6916, + "step": 14142 + }, + { + "epoch": 0.52, + "learning_rate": 1.483905225766287e-07, + "loss": 0.6972, + "step": 14143 + }, + { + "epoch": 0.52, + "learning_rate": 1.4838686051195663e-07, + "loss": 0.6865, + "step": 14144 + }, + { + "epoch": 0.52, + "learning_rate": 1.4838319844728458e-07, + "loss": 0.693, + "step": 14145 + }, + { + "epoch": 0.52, + "learning_rate": 1.483795363826125e-07, + "loss": 0.6996, + "step": 14146 + }, + { + "epoch": 0.52, + "learning_rate": 1.4837587431794046e-07, + "loss": 0.6975, + "step": 14147 + }, + { + "epoch": 0.52, + "learning_rate": 1.4837221225326838e-07, + "loss": 0.694, + "step": 14148 + }, + { + "epoch": 0.52, + "learning_rate": 1.4836855018859634e-07, + "loss": 0.6879, + "step": 14149 + }, + { + "epoch": 0.52, + "learning_rate": 1.4836488812392426e-07, + "loss": 0.6829, + "step": 14150 + }, + { + "epoch": 0.52, + "learning_rate": 1.4836122605925222e-07, + "loss": 0.6989, + "step": 14151 + }, + { + "epoch": 0.52, + "learning_rate": 1.4835756399458014e-07, + "loss": 0.6883, + "step": 14152 + }, + { + "epoch": 0.52, + "learning_rate": 1.4835390192990807e-07, + "loss": 0.7059, + "step": 14153 + }, + { + "epoch": 0.52, + "learning_rate": 1.48350239865236e-07, + "loss": 0.6879, + "step": 14154 + }, + { + "epoch": 0.52, + "learning_rate": 1.4834657780056395e-07, + "loss": 0.6922, + "step": 14155 + }, + { + "epoch": 0.52, + "learning_rate": 1.4834291573589188e-07, + "loss": 0.698, + "step": 14156 + }, + { + "epoch": 0.52, + "learning_rate": 1.4833925367121983e-07, + "loss": 0.6966, + "step": 14157 + }, + { + "epoch": 0.52, + "learning_rate": 1.4833559160654776e-07, + "loss": 0.703, + "step": 14158 + }, + { + "epoch": 0.52, + "learning_rate": 1.483319295418757e-07, + "loss": 0.6971, + "step": 14159 + }, + { + "epoch": 0.52, + "learning_rate": 1.4832826747720364e-07, + "loss": 0.6979, + "step": 14160 + }, + { + "epoch": 0.52, + "learning_rate": 1.483246054125316e-07, + "loss": 0.6896, + "step": 14161 + }, + { + "epoch": 0.52, + "learning_rate": 1.4832094334785952e-07, + "loss": 0.6972, + "step": 14162 + }, + { + "epoch": 0.52, + "learning_rate": 1.4831728128318744e-07, + "loss": 0.6861, + "step": 14163 + }, + { + "epoch": 0.52, + "learning_rate": 1.483136192185154e-07, + "loss": 0.7084, + "step": 14164 + }, + { + "epoch": 0.52, + "learning_rate": 1.4830995715384332e-07, + "loss": 0.6882, + "step": 14165 + }, + { + "epoch": 0.52, + "learning_rate": 1.4830629508917128e-07, + "loss": 0.6986, + "step": 14166 + }, + { + "epoch": 0.52, + "learning_rate": 1.483026330244992e-07, + "loss": 0.6783, + "step": 14167 + }, + { + "epoch": 0.52, + "learning_rate": 1.4829897095982716e-07, + "loss": 0.7029, + "step": 14168 + }, + { + "epoch": 0.52, + "learning_rate": 1.4829530889515508e-07, + "loss": 0.6929, + "step": 14169 + }, + { + "epoch": 0.52, + "learning_rate": 1.4829164683048304e-07, + "loss": 0.6973, + "step": 14170 + }, + { + "epoch": 0.52, + "learning_rate": 1.4828798476581094e-07, + "loss": 0.6918, + "step": 14171 + }, + { + "epoch": 0.52, + "learning_rate": 1.482843227011389e-07, + "loss": 0.7011, + "step": 14172 + }, + { + "epoch": 0.52, + "learning_rate": 1.4828066063646682e-07, + "loss": 0.6908, + "step": 14173 + }, + { + "epoch": 0.52, + "learning_rate": 1.4827699857179477e-07, + "loss": 0.6867, + "step": 14174 + }, + { + "epoch": 0.52, + "learning_rate": 1.482733365071227e-07, + "loss": 0.6902, + "step": 14175 + }, + { + "epoch": 0.52, + "learning_rate": 1.4826967444245065e-07, + "loss": 0.6927, + "step": 14176 + }, + { + "epoch": 0.52, + "learning_rate": 1.4826601237777858e-07, + "loss": 0.7051, + "step": 14177 + }, + { + "epoch": 0.52, + "learning_rate": 1.4826235031310653e-07, + "loss": 0.6913, + "step": 14178 + }, + { + "epoch": 0.52, + "learning_rate": 1.4825868824843446e-07, + "loss": 0.6884, + "step": 14179 + }, + { + "epoch": 0.52, + "learning_rate": 1.482550261837624e-07, + "loss": 0.6936, + "step": 14180 + }, + { + "epoch": 0.52, + "learning_rate": 1.4825136411909034e-07, + "loss": 0.6838, + "step": 14181 + }, + { + "epoch": 0.52, + "learning_rate": 1.482477020544183e-07, + "loss": 0.6959, + "step": 14182 + }, + { + "epoch": 0.52, + "learning_rate": 1.4824403998974622e-07, + "loss": 0.682, + "step": 14183 + }, + { + "epoch": 0.52, + "learning_rate": 1.4824037792507414e-07, + "loss": 0.6933, + "step": 14184 + }, + { + "epoch": 0.52, + "learning_rate": 1.482367158604021e-07, + "loss": 0.6945, + "step": 14185 + }, + { + "epoch": 0.52, + "learning_rate": 1.4823305379573002e-07, + "loss": 0.6865, + "step": 14186 + }, + { + "epoch": 0.52, + "learning_rate": 1.4822939173105798e-07, + "loss": 0.6927, + "step": 14187 + }, + { + "epoch": 0.52, + "learning_rate": 1.482257296663859e-07, + "loss": 0.6973, + "step": 14188 + }, + { + "epoch": 0.52, + "learning_rate": 1.4822206760171383e-07, + "loss": 0.6913, + "step": 14189 + }, + { + "epoch": 0.52, + "learning_rate": 1.4821840553704176e-07, + "loss": 0.6816, + "step": 14190 + }, + { + "epoch": 0.52, + "learning_rate": 1.482147434723697e-07, + "loss": 0.6939, + "step": 14191 + }, + { + "epoch": 0.52, + "learning_rate": 1.4821108140769764e-07, + "loss": 0.6962, + "step": 14192 + }, + { + "epoch": 0.52, + "learning_rate": 1.482074193430256e-07, + "loss": 0.7006, + "step": 14193 + }, + { + "epoch": 0.52, + "learning_rate": 1.4820375727835352e-07, + "loss": 0.6985, + "step": 14194 + }, + { + "epoch": 0.52, + "learning_rate": 1.4820009521368147e-07, + "loss": 0.6825, + "step": 14195 + }, + { + "epoch": 0.52, + "learning_rate": 1.481964331490094e-07, + "loss": 0.6931, + "step": 14196 + }, + { + "epoch": 0.52, + "learning_rate": 1.4819277108433735e-07, + "loss": 0.7093, + "step": 14197 + }, + { + "epoch": 0.52, + "learning_rate": 1.4818910901966528e-07, + "loss": 0.6989, + "step": 14198 + }, + { + "epoch": 0.52, + "learning_rate": 1.4818544695499323e-07, + "loss": 0.6977, + "step": 14199 + }, + { + "epoch": 0.52, + "learning_rate": 1.4818178489032116e-07, + "loss": 0.686, + "step": 14200 + }, + { + "epoch": 0.52, + "learning_rate": 1.481781228256491e-07, + "loss": 0.6942, + "step": 14201 + }, + { + "epoch": 0.52, + "learning_rate": 1.4817446076097704e-07, + "loss": 0.7002, + "step": 14202 + }, + { + "epoch": 0.52, + "learning_rate": 1.48170798696305e-07, + "loss": 0.6898, + "step": 14203 + }, + { + "epoch": 0.52, + "learning_rate": 1.4816713663163292e-07, + "loss": 0.6997, + "step": 14204 + }, + { + "epoch": 0.52, + "learning_rate": 1.4816347456696087e-07, + "loss": 0.678, + "step": 14205 + }, + { + "epoch": 0.52, + "learning_rate": 1.481598125022888e-07, + "loss": 0.691, + "step": 14206 + }, + { + "epoch": 0.52, + "learning_rate": 1.4815615043761672e-07, + "loss": 0.7059, + "step": 14207 + }, + { + "epoch": 0.52, + "learning_rate": 1.4815248837294465e-07, + "loss": 0.7139, + "step": 14208 + }, + { + "epoch": 0.52, + "learning_rate": 1.481488263082726e-07, + "loss": 0.6846, + "step": 14209 + }, + { + "epoch": 0.52, + "learning_rate": 1.4814516424360053e-07, + "loss": 0.6987, + "step": 14210 + }, + { + "epoch": 0.52, + "learning_rate": 1.4814150217892845e-07, + "loss": 0.6801, + "step": 14211 + }, + { + "epoch": 0.52, + "learning_rate": 1.481378401142564e-07, + "loss": 0.6798, + "step": 14212 + }, + { + "epoch": 0.52, + "learning_rate": 1.4813417804958433e-07, + "loss": 0.6851, + "step": 14213 + }, + { + "epoch": 0.52, + "learning_rate": 1.481305159849123e-07, + "loss": 0.7052, + "step": 14214 + }, + { + "epoch": 0.52, + "learning_rate": 1.4812685392024021e-07, + "loss": 0.6878, + "step": 14215 + }, + { + "epoch": 0.52, + "learning_rate": 1.4812319185556817e-07, + "loss": 0.6905, + "step": 14216 + }, + { + "epoch": 0.52, + "learning_rate": 1.481195297908961e-07, + "loss": 0.693, + "step": 14217 + }, + { + "epoch": 0.52, + "learning_rate": 1.4811586772622405e-07, + "loss": 0.6968, + "step": 14218 + }, + { + "epoch": 0.52, + "learning_rate": 1.4811220566155197e-07, + "loss": 0.7037, + "step": 14219 + }, + { + "epoch": 0.52, + "learning_rate": 1.4810854359687993e-07, + "loss": 0.6925, + "step": 14220 + }, + { + "epoch": 0.52, + "learning_rate": 1.4810488153220785e-07, + "loss": 0.6859, + "step": 14221 + }, + { + "epoch": 0.52, + "learning_rate": 1.481012194675358e-07, + "loss": 0.6982, + "step": 14222 + }, + { + "epoch": 0.52, + "learning_rate": 1.4809755740286373e-07, + "loss": 0.6842, + "step": 14223 + }, + { + "epoch": 0.52, + "learning_rate": 1.480938953381917e-07, + "loss": 0.6921, + "step": 14224 + }, + { + "epoch": 0.52, + "learning_rate": 1.480902332735196e-07, + "loss": 0.6987, + "step": 14225 + }, + { + "epoch": 0.52, + "learning_rate": 1.4808657120884754e-07, + "loss": 0.6906, + "step": 14226 + }, + { + "epoch": 0.52, + "learning_rate": 1.4808290914417547e-07, + "loss": 0.6921, + "step": 14227 + }, + { + "epoch": 0.52, + "learning_rate": 1.4807924707950342e-07, + "loss": 0.6813, + "step": 14228 + }, + { + "epoch": 0.52, + "learning_rate": 1.4807558501483135e-07, + "loss": 0.6903, + "step": 14229 + }, + { + "epoch": 0.52, + "learning_rate": 1.480719229501593e-07, + "loss": 0.697, + "step": 14230 + }, + { + "epoch": 0.52, + "learning_rate": 1.4806826088548723e-07, + "loss": 0.691, + "step": 14231 + }, + { + "epoch": 0.52, + "learning_rate": 1.4806459882081518e-07, + "loss": 0.6847, + "step": 14232 + }, + { + "epoch": 0.52, + "learning_rate": 1.480609367561431e-07, + "loss": 0.7167, + "step": 14233 + }, + { + "epoch": 0.52, + "learning_rate": 1.4805727469147103e-07, + "loss": 0.6812, + "step": 14234 + }, + { + "epoch": 0.52, + "learning_rate": 1.4805361262679899e-07, + "loss": 0.7037, + "step": 14235 + }, + { + "epoch": 0.52, + "learning_rate": 1.4804995056212691e-07, + "loss": 0.6938, + "step": 14236 + }, + { + "epoch": 0.52, + "learning_rate": 1.4804628849745487e-07, + "loss": 0.6894, + "step": 14237 + }, + { + "epoch": 0.52, + "learning_rate": 1.480426264327828e-07, + "loss": 0.6937, + "step": 14238 + }, + { + "epoch": 0.52, + "learning_rate": 1.4803896436811075e-07, + "loss": 0.7125, + "step": 14239 + }, + { + "epoch": 0.52, + "learning_rate": 1.4803530230343867e-07, + "loss": 0.7012, + "step": 14240 + }, + { + "epoch": 0.52, + "learning_rate": 1.4803164023876663e-07, + "loss": 0.7056, + "step": 14241 + }, + { + "epoch": 0.52, + "learning_rate": 1.4802797817409455e-07, + "loss": 0.698, + "step": 14242 + }, + { + "epoch": 0.52, + "learning_rate": 1.4802431610942248e-07, + "loss": 0.6857, + "step": 14243 + }, + { + "epoch": 0.52, + "learning_rate": 1.480206540447504e-07, + "loss": 0.6771, + "step": 14244 + }, + { + "epoch": 0.52, + "learning_rate": 1.4801699198007836e-07, + "loss": 0.6938, + "step": 14245 + }, + { + "epoch": 0.52, + "learning_rate": 1.4801332991540629e-07, + "loss": 0.7038, + "step": 14246 + }, + { + "epoch": 0.52, + "learning_rate": 1.4800966785073424e-07, + "loss": 0.6885, + "step": 14247 + }, + { + "epoch": 0.52, + "learning_rate": 1.4800600578606217e-07, + "loss": 0.6887, + "step": 14248 + }, + { + "epoch": 0.52, + "learning_rate": 1.4800234372139012e-07, + "loss": 0.6877, + "step": 14249 + }, + { + "epoch": 0.52, + "learning_rate": 1.4799868165671805e-07, + "loss": 0.6976, + "step": 14250 + }, + { + "epoch": 0.52, + "learning_rate": 1.47995019592046e-07, + "loss": 0.6972, + "step": 14251 + }, + { + "epoch": 0.52, + "learning_rate": 1.4799135752737393e-07, + "loss": 0.6946, + "step": 14252 + }, + { + "epoch": 0.52, + "learning_rate": 1.4798769546270188e-07, + "loss": 0.6979, + "step": 14253 + }, + { + "epoch": 0.52, + "learning_rate": 1.479840333980298e-07, + "loss": 0.6802, + "step": 14254 + }, + { + "epoch": 0.52, + "learning_rate": 1.4798037133335773e-07, + "loss": 0.6956, + "step": 14255 + }, + { + "epoch": 0.52, + "learning_rate": 1.4797670926868569e-07, + "loss": 0.7088, + "step": 14256 + }, + { + "epoch": 0.52, + "learning_rate": 1.479730472040136e-07, + "loss": 0.7122, + "step": 14257 + }, + { + "epoch": 0.52, + "learning_rate": 1.4796938513934157e-07, + "loss": 0.689, + "step": 14258 + }, + { + "epoch": 0.52, + "learning_rate": 1.479657230746695e-07, + "loss": 0.6871, + "step": 14259 + }, + { + "epoch": 0.52, + "learning_rate": 1.4796206100999745e-07, + "loss": 0.6933, + "step": 14260 + }, + { + "epoch": 0.52, + "learning_rate": 1.4795839894532535e-07, + "loss": 0.6918, + "step": 14261 + }, + { + "epoch": 0.52, + "learning_rate": 1.479547368806533e-07, + "loss": 0.6963, + "step": 14262 + }, + { + "epoch": 0.52, + "learning_rate": 1.4795107481598123e-07, + "loss": 0.6908, + "step": 14263 + }, + { + "epoch": 0.52, + "learning_rate": 1.4794741275130918e-07, + "loss": 0.6997, + "step": 14264 + }, + { + "epoch": 0.52, + "learning_rate": 1.479437506866371e-07, + "loss": 0.6882, + "step": 14265 + }, + { + "epoch": 0.52, + "learning_rate": 1.4794008862196506e-07, + "loss": 0.684, + "step": 14266 + }, + { + "epoch": 0.52, + "learning_rate": 1.4793642655729298e-07, + "loss": 0.6921, + "step": 14267 + }, + { + "epoch": 0.52, + "learning_rate": 1.4793276449262094e-07, + "loss": 0.6853, + "step": 14268 + }, + { + "epoch": 0.52, + "learning_rate": 1.4792910242794886e-07, + "loss": 0.6933, + "step": 14269 + }, + { + "epoch": 0.52, + "learning_rate": 1.4792544036327682e-07, + "loss": 0.6978, + "step": 14270 + }, + { + "epoch": 0.52, + "learning_rate": 1.4792177829860474e-07, + "loss": 0.6773, + "step": 14271 + }, + { + "epoch": 0.52, + "learning_rate": 1.479181162339327e-07, + "loss": 0.709, + "step": 14272 + }, + { + "epoch": 0.52, + "learning_rate": 1.4791445416926062e-07, + "loss": 0.6921, + "step": 14273 + }, + { + "epoch": 0.52, + "learning_rate": 1.4791079210458858e-07, + "loss": 0.7027, + "step": 14274 + }, + { + "epoch": 0.52, + "learning_rate": 1.479071300399165e-07, + "loss": 0.6825, + "step": 14275 + }, + { + "epoch": 0.52, + "learning_rate": 1.4790346797524446e-07, + "loss": 0.6923, + "step": 14276 + }, + { + "epoch": 0.52, + "learning_rate": 1.4789980591057238e-07, + "loss": 0.6942, + "step": 14277 + }, + { + "epoch": 0.52, + "learning_rate": 1.478961438459003e-07, + "loss": 0.6855, + "step": 14278 + }, + { + "epoch": 0.52, + "learning_rate": 1.4789248178122824e-07, + "loss": 0.6844, + "step": 14279 + }, + { + "epoch": 0.52, + "learning_rate": 1.478888197165562e-07, + "loss": 0.6848, + "step": 14280 + }, + { + "epoch": 0.52, + "learning_rate": 1.4788515765188412e-07, + "loss": 0.6811, + "step": 14281 + }, + { + "epoch": 0.52, + "learning_rate": 1.4788149558721204e-07, + "loss": 0.6967, + "step": 14282 + }, + { + "epoch": 0.52, + "learning_rate": 1.4787783352254e-07, + "loss": 0.6891, + "step": 14283 + }, + { + "epoch": 0.52, + "learning_rate": 1.4787417145786792e-07, + "loss": 0.6882, + "step": 14284 + }, + { + "epoch": 0.52, + "learning_rate": 1.4787050939319588e-07, + "loss": 0.6946, + "step": 14285 + }, + { + "epoch": 0.52, + "learning_rate": 1.478668473285238e-07, + "loss": 0.6857, + "step": 14286 + }, + { + "epoch": 0.52, + "learning_rate": 1.4786318526385176e-07, + "loss": 0.7004, + "step": 14287 + }, + { + "epoch": 0.52, + "learning_rate": 1.4785952319917968e-07, + "loss": 0.6946, + "step": 14288 + }, + { + "epoch": 0.52, + "learning_rate": 1.4785586113450764e-07, + "loss": 0.6742, + "step": 14289 + }, + { + "epoch": 0.52, + "learning_rate": 1.4785219906983556e-07, + "loss": 0.7027, + "step": 14290 + }, + { + "epoch": 0.52, + "learning_rate": 1.4784853700516352e-07, + "loss": 0.6965, + "step": 14291 + }, + { + "epoch": 0.52, + "learning_rate": 1.4784487494049144e-07, + "loss": 0.6935, + "step": 14292 + }, + { + "epoch": 0.52, + "learning_rate": 1.478412128758194e-07, + "loss": 0.6969, + "step": 14293 + }, + { + "epoch": 0.52, + "learning_rate": 1.4783755081114732e-07, + "loss": 0.6845, + "step": 14294 + }, + { + "epoch": 0.52, + "learning_rate": 1.4783388874647528e-07, + "loss": 0.6999, + "step": 14295 + }, + { + "epoch": 0.52, + "learning_rate": 1.4783022668180318e-07, + "loss": 0.6947, + "step": 14296 + }, + { + "epoch": 0.52, + "learning_rate": 1.4782656461713113e-07, + "loss": 0.6869, + "step": 14297 + }, + { + "epoch": 0.52, + "learning_rate": 1.4782290255245906e-07, + "loss": 0.704, + "step": 14298 + }, + { + "epoch": 0.52, + "learning_rate": 1.47819240487787e-07, + "loss": 0.6986, + "step": 14299 + }, + { + "epoch": 0.52, + "learning_rate": 1.4781557842311494e-07, + "loss": 0.6898, + "step": 14300 + }, + { + "epoch": 0.52, + "learning_rate": 1.478119163584429e-07, + "loss": 0.7073, + "step": 14301 + }, + { + "epoch": 0.52, + "learning_rate": 1.4780825429377082e-07, + "loss": 0.6842, + "step": 14302 + }, + { + "epoch": 0.52, + "learning_rate": 1.4780459222909877e-07, + "loss": 0.692, + "step": 14303 + }, + { + "epoch": 0.52, + "learning_rate": 1.478009301644267e-07, + "loss": 0.6908, + "step": 14304 + }, + { + "epoch": 0.52, + "learning_rate": 1.4779726809975462e-07, + "loss": 0.6977, + "step": 14305 + }, + { + "epoch": 0.52, + "learning_rate": 1.4779360603508258e-07, + "loss": 0.7142, + "step": 14306 + }, + { + "epoch": 0.52, + "learning_rate": 1.477899439704105e-07, + "loss": 0.6959, + "step": 14307 + }, + { + "epoch": 0.52, + "learning_rate": 1.4778628190573846e-07, + "loss": 0.6906, + "step": 14308 + }, + { + "epoch": 0.52, + "learning_rate": 1.4778261984106638e-07, + "loss": 0.6942, + "step": 14309 + }, + { + "epoch": 0.52, + "learning_rate": 1.4777895777639434e-07, + "loss": 0.6887, + "step": 14310 + }, + { + "epoch": 0.52, + "learning_rate": 1.4777529571172226e-07, + "loss": 0.7002, + "step": 14311 + }, + { + "epoch": 0.52, + "learning_rate": 1.4777163364705022e-07, + "loss": 0.6831, + "step": 14312 + }, + { + "epoch": 0.52, + "learning_rate": 1.4776797158237814e-07, + "loss": 0.6837, + "step": 14313 + }, + { + "epoch": 0.52, + "learning_rate": 1.4776430951770607e-07, + "loss": 0.6921, + "step": 14314 + }, + { + "epoch": 0.52, + "learning_rate": 1.47760647453034e-07, + "loss": 0.6978, + "step": 14315 + }, + { + "epoch": 0.52, + "learning_rate": 1.4775698538836195e-07, + "loss": 0.6946, + "step": 14316 + }, + { + "epoch": 0.52, + "learning_rate": 1.4775332332368988e-07, + "loss": 0.6912, + "step": 14317 + }, + { + "epoch": 0.52, + "learning_rate": 1.4774966125901783e-07, + "loss": 0.6906, + "step": 14318 + }, + { + "epoch": 0.52, + "learning_rate": 1.4774599919434576e-07, + "loss": 0.6953, + "step": 14319 + }, + { + "epoch": 0.52, + "learning_rate": 1.477423371296737e-07, + "loss": 0.7023, + "step": 14320 + }, + { + "epoch": 0.52, + "learning_rate": 1.4773867506500164e-07, + "loss": 0.6994, + "step": 14321 + }, + { + "epoch": 0.52, + "learning_rate": 1.477350130003296e-07, + "loss": 0.6904, + "step": 14322 + }, + { + "epoch": 0.52, + "learning_rate": 1.4773135093565752e-07, + "loss": 0.6897, + "step": 14323 + }, + { + "epoch": 0.52, + "learning_rate": 1.4772768887098547e-07, + "loss": 0.7045, + "step": 14324 + }, + { + "epoch": 0.52, + "learning_rate": 1.477240268063134e-07, + "loss": 0.6958, + "step": 14325 + }, + { + "epoch": 0.52, + "learning_rate": 1.4772036474164132e-07, + "loss": 0.6939, + "step": 14326 + }, + { + "epoch": 0.52, + "learning_rate": 1.4771670267696927e-07, + "loss": 0.7068, + "step": 14327 + }, + { + "epoch": 0.52, + "learning_rate": 1.477130406122972e-07, + "loss": 0.6905, + "step": 14328 + }, + { + "epoch": 0.52, + "learning_rate": 1.4770937854762515e-07, + "loss": 0.683, + "step": 14329 + }, + { + "epoch": 0.52, + "learning_rate": 1.4770571648295308e-07, + "loss": 0.711, + "step": 14330 + }, + { + "epoch": 0.52, + "learning_rate": 1.4770205441828103e-07, + "loss": 0.6964, + "step": 14331 + }, + { + "epoch": 0.52, + "learning_rate": 1.4769839235360893e-07, + "loss": 0.701, + "step": 14332 + }, + { + "epoch": 0.52, + "learning_rate": 1.476947302889369e-07, + "loss": 0.6688, + "step": 14333 + }, + { + "epoch": 0.52, + "learning_rate": 1.4769106822426481e-07, + "loss": 0.6994, + "step": 14334 + }, + { + "epoch": 0.52, + "learning_rate": 1.4768740615959277e-07, + "loss": 0.6966, + "step": 14335 + }, + { + "epoch": 0.52, + "learning_rate": 1.476837440949207e-07, + "loss": 0.6954, + "step": 14336 + }, + { + "epoch": 0.52, + "learning_rate": 1.4768008203024865e-07, + "loss": 0.6912, + "step": 14337 + }, + { + "epoch": 0.52, + "learning_rate": 1.4767641996557657e-07, + "loss": 0.7009, + "step": 14338 + }, + { + "epoch": 0.52, + "learning_rate": 1.4767275790090453e-07, + "loss": 0.6891, + "step": 14339 + }, + { + "epoch": 0.52, + "learning_rate": 1.4766909583623245e-07, + "loss": 0.6963, + "step": 14340 + }, + { + "epoch": 0.52, + "learning_rate": 1.476654337715604e-07, + "loss": 0.688, + "step": 14341 + }, + { + "epoch": 0.52, + "learning_rate": 1.4766177170688833e-07, + "loss": 0.6783, + "step": 14342 + }, + { + "epoch": 0.52, + "learning_rate": 1.476581096422163e-07, + "loss": 0.6991, + "step": 14343 + }, + { + "epoch": 0.52, + "learning_rate": 1.4765444757754421e-07, + "loss": 0.686, + "step": 14344 + }, + { + "epoch": 0.52, + "learning_rate": 1.4765078551287217e-07, + "loss": 0.6944, + "step": 14345 + }, + { + "epoch": 0.52, + "learning_rate": 1.476471234482001e-07, + "loss": 0.6992, + "step": 14346 + }, + { + "epoch": 0.52, + "learning_rate": 1.4764346138352805e-07, + "loss": 0.6952, + "step": 14347 + }, + { + "epoch": 0.52, + "learning_rate": 1.4763979931885597e-07, + "loss": 0.6981, + "step": 14348 + }, + { + "epoch": 0.52, + "learning_rate": 1.476361372541839e-07, + "loss": 0.6844, + "step": 14349 + }, + { + "epoch": 0.53, + "learning_rate": 1.4763247518951183e-07, + "loss": 0.6886, + "step": 14350 + }, + { + "epoch": 0.53, + "learning_rate": 1.4762881312483978e-07, + "loss": 0.6959, + "step": 14351 + }, + { + "epoch": 0.53, + "learning_rate": 1.476251510601677e-07, + "loss": 0.6842, + "step": 14352 + }, + { + "epoch": 0.53, + "learning_rate": 1.4762148899549563e-07, + "loss": 0.6977, + "step": 14353 + }, + { + "epoch": 0.53, + "learning_rate": 1.4761782693082359e-07, + "loss": 0.6923, + "step": 14354 + }, + { + "epoch": 0.53, + "learning_rate": 1.4761416486615151e-07, + "loss": 0.6831, + "step": 14355 + }, + { + "epoch": 0.53, + "learning_rate": 1.4761050280147947e-07, + "loss": 0.6986, + "step": 14356 + }, + { + "epoch": 0.53, + "learning_rate": 1.476068407368074e-07, + "loss": 0.688, + "step": 14357 + }, + { + "epoch": 0.53, + "learning_rate": 1.4760317867213535e-07, + "loss": 0.6994, + "step": 14358 + }, + { + "epoch": 0.53, + "learning_rate": 1.4759951660746327e-07, + "loss": 0.6907, + "step": 14359 + }, + { + "epoch": 0.53, + "learning_rate": 1.4759585454279123e-07, + "loss": 0.7073, + "step": 14360 + }, + { + "epoch": 0.53, + "learning_rate": 1.4759219247811915e-07, + "loss": 0.6893, + "step": 14361 + }, + { + "epoch": 0.53, + "learning_rate": 1.475885304134471e-07, + "loss": 0.6947, + "step": 14362 + }, + { + "epoch": 0.53, + "learning_rate": 1.4758486834877503e-07, + "loss": 0.6875, + "step": 14363 + }, + { + "epoch": 0.53, + "learning_rate": 1.4758120628410299e-07, + "loss": 0.6972, + "step": 14364 + }, + { + "epoch": 0.53, + "learning_rate": 1.475775442194309e-07, + "loss": 0.6853, + "step": 14365 + }, + { + "epoch": 0.53, + "learning_rate": 1.4757388215475887e-07, + "loss": 0.7009, + "step": 14366 + }, + { + "epoch": 0.53, + "learning_rate": 1.475702200900868e-07, + "loss": 0.703, + "step": 14367 + }, + { + "epoch": 0.53, + "learning_rate": 1.4756655802541472e-07, + "loss": 0.6891, + "step": 14368 + }, + { + "epoch": 0.53, + "learning_rate": 1.4756289596074265e-07, + "loss": 0.6994, + "step": 14369 + }, + { + "epoch": 0.53, + "learning_rate": 1.475592338960706e-07, + "loss": 0.6935, + "step": 14370 + }, + { + "epoch": 0.53, + "learning_rate": 1.4755557183139853e-07, + "loss": 0.7025, + "step": 14371 + }, + { + "epoch": 0.53, + "learning_rate": 1.4755190976672648e-07, + "loss": 0.6892, + "step": 14372 + }, + { + "epoch": 0.53, + "learning_rate": 1.475482477020544e-07, + "loss": 0.6811, + "step": 14373 + }, + { + "epoch": 0.53, + "learning_rate": 1.4754458563738236e-07, + "loss": 0.683, + "step": 14374 + }, + { + "epoch": 0.53, + "learning_rate": 1.4754092357271029e-07, + "loss": 0.6919, + "step": 14375 + }, + { + "epoch": 0.53, + "learning_rate": 1.475372615080382e-07, + "loss": 0.6946, + "step": 14376 + }, + { + "epoch": 0.53, + "learning_rate": 1.4753359944336617e-07, + "loss": 0.6949, + "step": 14377 + }, + { + "epoch": 0.53, + "learning_rate": 1.475299373786941e-07, + "loss": 0.7022, + "step": 14378 + }, + { + "epoch": 0.53, + "learning_rate": 1.4752627531402205e-07, + "loss": 0.6868, + "step": 14379 + }, + { + "epoch": 0.53, + "learning_rate": 1.4752261324934997e-07, + "loss": 0.6901, + "step": 14380 + }, + { + "epoch": 0.53, + "learning_rate": 1.4751895118467793e-07, + "loss": 0.6998, + "step": 14381 + }, + { + "epoch": 0.53, + "learning_rate": 1.4751528912000585e-07, + "loss": 0.6966, + "step": 14382 + }, + { + "epoch": 0.53, + "learning_rate": 1.475116270553338e-07, + "loss": 0.6983, + "step": 14383 + }, + { + "epoch": 0.53, + "learning_rate": 1.4750796499066173e-07, + "loss": 0.6902, + "step": 14384 + }, + { + "epoch": 0.53, + "learning_rate": 1.4750430292598968e-07, + "loss": 0.6931, + "step": 14385 + }, + { + "epoch": 0.53, + "learning_rate": 1.4750064086131758e-07, + "loss": 0.7128, + "step": 14386 + }, + { + "epoch": 0.53, + "learning_rate": 1.4749697879664554e-07, + "loss": 0.6799, + "step": 14387 + }, + { + "epoch": 0.53, + "learning_rate": 1.4749331673197346e-07, + "loss": 0.7051, + "step": 14388 + }, + { + "epoch": 0.53, + "learning_rate": 1.4748965466730142e-07, + "loss": 0.6838, + "step": 14389 + }, + { + "epoch": 0.53, + "learning_rate": 1.4748599260262934e-07, + "loss": 0.7048, + "step": 14390 + }, + { + "epoch": 0.53, + "learning_rate": 1.474823305379573e-07, + "loss": 0.6969, + "step": 14391 + }, + { + "epoch": 0.53, + "learning_rate": 1.4747866847328522e-07, + "loss": 0.6896, + "step": 14392 + }, + { + "epoch": 0.53, + "learning_rate": 1.4747500640861318e-07, + "loss": 0.7035, + "step": 14393 + }, + { + "epoch": 0.53, + "learning_rate": 1.474713443439411e-07, + "loss": 0.6927, + "step": 14394 + }, + { + "epoch": 0.53, + "learning_rate": 1.4746768227926906e-07, + "loss": 0.6844, + "step": 14395 + }, + { + "epoch": 0.53, + "learning_rate": 1.4746402021459698e-07, + "loss": 0.6932, + "step": 14396 + }, + { + "epoch": 0.53, + "learning_rate": 1.474603581499249e-07, + "loss": 0.6852, + "step": 14397 + }, + { + "epoch": 0.53, + "learning_rate": 1.4745669608525286e-07, + "loss": 0.6875, + "step": 14398 + }, + { + "epoch": 0.53, + "learning_rate": 1.474530340205808e-07, + "loss": 0.6941, + "step": 14399 + }, + { + "epoch": 0.53, + "learning_rate": 1.4744937195590874e-07, + "loss": 0.6835, + "step": 14400 + }, + { + "epoch": 0.53, + "learning_rate": 1.4744570989123667e-07, + "loss": 0.6874, + "step": 14401 + }, + { + "epoch": 0.53, + "learning_rate": 1.4744204782656462e-07, + "loss": 0.6956, + "step": 14402 + }, + { + "epoch": 0.53, + "learning_rate": 1.4743838576189255e-07, + "loss": 0.6926, + "step": 14403 + }, + { + "epoch": 0.53, + "learning_rate": 1.4743472369722048e-07, + "loss": 0.693, + "step": 14404 + }, + { + "epoch": 0.53, + "learning_rate": 1.474310616325484e-07, + "loss": 0.6865, + "step": 14405 + }, + { + "epoch": 0.53, + "learning_rate": 1.4742739956787636e-07, + "loss": 0.6866, + "step": 14406 + }, + { + "epoch": 0.53, + "learning_rate": 1.4742373750320428e-07, + "loss": 0.6758, + "step": 14407 + }, + { + "epoch": 0.53, + "learning_rate": 1.4742007543853224e-07, + "loss": 0.6852, + "step": 14408 + }, + { + "epoch": 0.53, + "learning_rate": 1.4741641337386016e-07, + "loss": 0.6965, + "step": 14409 + }, + { + "epoch": 0.53, + "learning_rate": 1.4741275130918812e-07, + "loss": 0.705, + "step": 14410 + }, + { + "epoch": 0.53, + "learning_rate": 1.4740908924451604e-07, + "loss": 0.6958, + "step": 14411 + }, + { + "epoch": 0.53, + "learning_rate": 1.47405427179844e-07, + "loss": 0.7009, + "step": 14412 + }, + { + "epoch": 0.53, + "learning_rate": 1.4740176511517192e-07, + "loss": 0.68, + "step": 14413 + }, + { + "epoch": 0.53, + "learning_rate": 1.4739810305049988e-07, + "loss": 0.7031, + "step": 14414 + }, + { + "epoch": 0.53, + "learning_rate": 1.473944409858278e-07, + "loss": 0.7003, + "step": 14415 + }, + { + "epoch": 0.53, + "learning_rate": 1.4739077892115576e-07, + "loss": 0.6835, + "step": 14416 + }, + { + "epoch": 0.53, + "learning_rate": 1.4738711685648368e-07, + "loss": 0.7025, + "step": 14417 + }, + { + "epoch": 0.53, + "learning_rate": 1.4738345479181164e-07, + "loss": 0.6901, + "step": 14418 + }, + { + "epoch": 0.53, + "learning_rate": 1.4737979272713956e-07, + "loss": 0.7075, + "step": 14419 + }, + { + "epoch": 0.53, + "learning_rate": 1.473761306624675e-07, + "loss": 0.6898, + "step": 14420 + }, + { + "epoch": 0.53, + "learning_rate": 1.4737246859779544e-07, + "loss": 0.6869, + "step": 14421 + }, + { + "epoch": 0.53, + "learning_rate": 1.4736880653312337e-07, + "loss": 0.6882, + "step": 14422 + }, + { + "epoch": 0.53, + "learning_rate": 1.473651444684513e-07, + "loss": 0.6816, + "step": 14423 + }, + { + "epoch": 0.53, + "learning_rate": 1.4736148240377922e-07, + "loss": 0.698, + "step": 14424 + }, + { + "epoch": 0.53, + "learning_rate": 1.4735782033910718e-07, + "loss": 0.6852, + "step": 14425 + }, + { + "epoch": 0.53, + "learning_rate": 1.473541582744351e-07, + "loss": 0.6772, + "step": 14426 + }, + { + "epoch": 0.53, + "learning_rate": 1.4735049620976306e-07, + "loss": 0.7105, + "step": 14427 + }, + { + "epoch": 0.53, + "learning_rate": 1.4734683414509098e-07, + "loss": 0.7039, + "step": 14428 + }, + { + "epoch": 0.53, + "learning_rate": 1.4734317208041894e-07, + "loss": 0.6886, + "step": 14429 + }, + { + "epoch": 0.53, + "learning_rate": 1.4733951001574686e-07, + "loss": 0.6914, + "step": 14430 + }, + { + "epoch": 0.53, + "learning_rate": 1.4733584795107482e-07, + "loss": 0.6882, + "step": 14431 + }, + { + "epoch": 0.53, + "learning_rate": 1.4733218588640274e-07, + "loss": 0.7054, + "step": 14432 + }, + { + "epoch": 0.53, + "learning_rate": 1.473285238217307e-07, + "loss": 0.6916, + "step": 14433 + }, + { + "epoch": 0.53, + "learning_rate": 1.4732486175705862e-07, + "loss": 0.6815, + "step": 14434 + }, + { + "epoch": 0.53, + "learning_rate": 1.4732119969238658e-07, + "loss": 0.6948, + "step": 14435 + }, + { + "epoch": 0.53, + "learning_rate": 1.473175376277145e-07, + "loss": 0.6842, + "step": 14436 + }, + { + "epoch": 0.53, + "learning_rate": 1.4731387556304246e-07, + "loss": 0.6964, + "step": 14437 + }, + { + "epoch": 0.53, + "learning_rate": 1.4731021349837038e-07, + "loss": 0.6878, + "step": 14438 + }, + { + "epoch": 0.53, + "learning_rate": 1.4730655143369834e-07, + "loss": 0.7007, + "step": 14439 + }, + { + "epoch": 0.53, + "learning_rate": 1.4730288936902624e-07, + "loss": 0.702, + "step": 14440 + }, + { + "epoch": 0.53, + "learning_rate": 1.472992273043542e-07, + "loss": 0.7027, + "step": 14441 + }, + { + "epoch": 0.53, + "learning_rate": 1.4729556523968212e-07, + "loss": 0.6931, + "step": 14442 + }, + { + "epoch": 0.53, + "learning_rate": 1.4729190317501007e-07, + "loss": 0.6941, + "step": 14443 + }, + { + "epoch": 0.53, + "learning_rate": 1.47288241110338e-07, + "loss": 0.6858, + "step": 14444 + }, + { + "epoch": 0.53, + "learning_rate": 1.4728457904566595e-07, + "loss": 0.6943, + "step": 14445 + }, + { + "epoch": 0.53, + "learning_rate": 1.4728091698099387e-07, + "loss": 0.6964, + "step": 14446 + }, + { + "epoch": 0.53, + "learning_rate": 1.472772549163218e-07, + "loss": 0.6944, + "step": 14447 + }, + { + "epoch": 0.53, + "learning_rate": 1.4727359285164975e-07, + "loss": 0.6808, + "step": 14448 + }, + { + "epoch": 0.53, + "learning_rate": 1.4726993078697768e-07, + "loss": 0.6744, + "step": 14449 + }, + { + "epoch": 0.53, + "learning_rate": 1.4726626872230563e-07, + "loss": 0.6781, + "step": 14450 + }, + { + "epoch": 0.53, + "learning_rate": 1.4726260665763356e-07, + "loss": 0.7074, + "step": 14451 + }, + { + "epoch": 0.53, + "learning_rate": 1.4725894459296151e-07, + "loss": 0.6853, + "step": 14452 + }, + { + "epoch": 0.53, + "learning_rate": 1.4725528252828944e-07, + "loss": 0.6903, + "step": 14453 + }, + { + "epoch": 0.53, + "learning_rate": 1.472516204636174e-07, + "loss": 0.6812, + "step": 14454 + }, + { + "epoch": 0.53, + "learning_rate": 1.4724795839894532e-07, + "loss": 0.7027, + "step": 14455 + }, + { + "epoch": 0.53, + "learning_rate": 1.4724429633427327e-07, + "loss": 0.6974, + "step": 14456 + }, + { + "epoch": 0.53, + "learning_rate": 1.472406342696012e-07, + "loss": 0.7063, + "step": 14457 + }, + { + "epoch": 0.53, + "learning_rate": 1.4723697220492913e-07, + "loss": 0.6874, + "step": 14458 + }, + { + "epoch": 0.53, + "learning_rate": 1.4723331014025705e-07, + "loss": 0.6775, + "step": 14459 + }, + { + "epoch": 0.53, + "learning_rate": 1.47229648075585e-07, + "loss": 0.6916, + "step": 14460 + }, + { + "epoch": 0.53, + "learning_rate": 1.4722598601091293e-07, + "loss": 0.6851, + "step": 14461 + }, + { + "epoch": 0.53, + "learning_rate": 1.472223239462409e-07, + "loss": 0.6942, + "step": 14462 + }, + { + "epoch": 0.53, + "learning_rate": 1.4721866188156881e-07, + "loss": 0.6947, + "step": 14463 + }, + { + "epoch": 0.53, + "learning_rate": 1.4721499981689677e-07, + "loss": 0.698, + "step": 14464 + }, + { + "epoch": 0.53, + "learning_rate": 1.472113377522247e-07, + "loss": 0.6834, + "step": 14465 + }, + { + "epoch": 0.53, + "learning_rate": 1.4720767568755265e-07, + "loss": 0.6975, + "step": 14466 + }, + { + "epoch": 0.53, + "learning_rate": 1.4720401362288057e-07, + "loss": 0.7013, + "step": 14467 + }, + { + "epoch": 0.53, + "learning_rate": 1.472003515582085e-07, + "loss": 0.6903, + "step": 14468 + }, + { + "epoch": 0.53, + "learning_rate": 1.4719668949353645e-07, + "loss": 0.6985, + "step": 14469 + }, + { + "epoch": 0.53, + "learning_rate": 1.4719302742886438e-07, + "loss": 0.6908, + "step": 14470 + }, + { + "epoch": 0.53, + "learning_rate": 1.4718936536419233e-07, + "loss": 0.6885, + "step": 14471 + }, + { + "epoch": 0.53, + "learning_rate": 1.4718570329952026e-07, + "loss": 0.6857, + "step": 14472 + }, + { + "epoch": 0.53, + "learning_rate": 1.471820412348482e-07, + "loss": 0.6856, + "step": 14473 + }, + { + "epoch": 0.53, + "learning_rate": 1.4717837917017614e-07, + "loss": 0.6977, + "step": 14474 + }, + { + "epoch": 0.53, + "learning_rate": 1.471747171055041e-07, + "loss": 0.6778, + "step": 14475 + }, + { + "epoch": 0.53, + "learning_rate": 1.47171055040832e-07, + "loss": 0.6921, + "step": 14476 + }, + { + "epoch": 0.53, + "learning_rate": 1.4716739297615995e-07, + "loss": 0.7018, + "step": 14477 + }, + { + "epoch": 0.53, + "learning_rate": 1.4716373091148787e-07, + "loss": 0.6902, + "step": 14478 + }, + { + "epoch": 0.53, + "learning_rate": 1.4716006884681583e-07, + "loss": 0.6851, + "step": 14479 + }, + { + "epoch": 0.53, + "learning_rate": 1.4715640678214375e-07, + "loss": 0.7005, + "step": 14480 + }, + { + "epoch": 0.53, + "learning_rate": 1.471527447174717e-07, + "loss": 0.6886, + "step": 14481 + }, + { + "epoch": 0.53, + "learning_rate": 1.4714908265279963e-07, + "loss": 0.6921, + "step": 14482 + }, + { + "epoch": 0.53, + "learning_rate": 1.4714542058812759e-07, + "loss": 0.701, + "step": 14483 + }, + { + "epoch": 0.53, + "learning_rate": 1.471417585234555e-07, + "loss": 0.6914, + "step": 14484 + }, + { + "epoch": 0.53, + "learning_rate": 1.4713809645878347e-07, + "loss": 0.7086, + "step": 14485 + }, + { + "epoch": 0.53, + "learning_rate": 1.471344343941114e-07, + "loss": 0.6936, + "step": 14486 + }, + { + "epoch": 0.53, + "learning_rate": 1.4713077232943935e-07, + "loss": 0.6893, + "step": 14487 + }, + { + "epoch": 0.53, + "learning_rate": 1.4712711026476727e-07, + "loss": 0.7013, + "step": 14488 + }, + { + "epoch": 0.53, + "learning_rate": 1.4712344820009523e-07, + "loss": 0.6989, + "step": 14489 + }, + { + "epoch": 0.53, + "learning_rate": 1.4711978613542315e-07, + "loss": 0.6928, + "step": 14490 + }, + { + "epoch": 0.53, + "learning_rate": 1.4711612407075108e-07, + "loss": 0.6965, + "step": 14491 + }, + { + "epoch": 0.53, + "learning_rate": 1.4711246200607903e-07, + "loss": 0.6934, + "step": 14492 + }, + { + "epoch": 0.53, + "learning_rate": 1.4710879994140696e-07, + "loss": 0.6941, + "step": 14493 + }, + { + "epoch": 0.53, + "learning_rate": 1.4710513787673489e-07, + "loss": 0.6949, + "step": 14494 + }, + { + "epoch": 0.53, + "learning_rate": 1.471014758120628e-07, + "loss": 0.6957, + "step": 14495 + }, + { + "epoch": 0.53, + "learning_rate": 1.4709781374739077e-07, + "loss": 0.692, + "step": 14496 + }, + { + "epoch": 0.53, + "learning_rate": 1.470941516827187e-07, + "loss": 0.6953, + "step": 14497 + }, + { + "epoch": 0.53, + "learning_rate": 1.4709048961804665e-07, + "loss": 0.7022, + "step": 14498 + }, + { + "epoch": 0.53, + "learning_rate": 1.4708682755337457e-07, + "loss": 0.712, + "step": 14499 + }, + { + "epoch": 0.53, + "learning_rate": 1.4708316548870253e-07, + "loss": 0.6904, + "step": 14500 + }, + { + "epoch": 0.53, + "learning_rate": 1.4707950342403045e-07, + "loss": 0.6894, + "step": 14501 + }, + { + "epoch": 0.53, + "learning_rate": 1.470758413593584e-07, + "loss": 0.6885, + "step": 14502 + }, + { + "epoch": 0.53, + "learning_rate": 1.4707217929468633e-07, + "loss": 0.6913, + "step": 14503 + }, + { + "epoch": 0.53, + "learning_rate": 1.4706851723001428e-07, + "loss": 0.7035, + "step": 14504 + }, + { + "epoch": 0.53, + "learning_rate": 1.470648551653422e-07, + "loss": 0.6813, + "step": 14505 + }, + { + "epoch": 0.53, + "learning_rate": 1.4706119310067016e-07, + "loss": 0.6918, + "step": 14506 + }, + { + "epoch": 0.53, + "learning_rate": 1.470575310359981e-07, + "loss": 0.6865, + "step": 14507 + }, + { + "epoch": 0.53, + "learning_rate": 1.4705386897132604e-07, + "loss": 0.7019, + "step": 14508 + }, + { + "epoch": 0.53, + "learning_rate": 1.4705020690665397e-07, + "loss": 0.7019, + "step": 14509 + }, + { + "epoch": 0.53, + "learning_rate": 1.4704654484198192e-07, + "loss": 0.6758, + "step": 14510 + }, + { + "epoch": 0.53, + "learning_rate": 1.4704288277730985e-07, + "loss": 0.692, + "step": 14511 + }, + { + "epoch": 0.53, + "learning_rate": 1.4703922071263778e-07, + "loss": 0.6966, + "step": 14512 + }, + { + "epoch": 0.53, + "learning_rate": 1.470355586479657e-07, + "loss": 0.6842, + "step": 14513 + }, + { + "epoch": 0.53, + "learning_rate": 1.4703189658329366e-07, + "loss": 0.6976, + "step": 14514 + }, + { + "epoch": 0.53, + "learning_rate": 1.4702823451862158e-07, + "loss": 0.6866, + "step": 14515 + }, + { + "epoch": 0.53, + "learning_rate": 1.4702457245394954e-07, + "loss": 0.6993, + "step": 14516 + }, + { + "epoch": 0.53, + "learning_rate": 1.4702091038927746e-07, + "loss": 0.7028, + "step": 14517 + }, + { + "epoch": 0.53, + "learning_rate": 1.470172483246054e-07, + "loss": 0.7043, + "step": 14518 + }, + { + "epoch": 0.53, + "learning_rate": 1.4701358625993334e-07, + "loss": 0.6941, + "step": 14519 + }, + { + "epoch": 0.53, + "learning_rate": 1.4700992419526127e-07, + "loss": 0.6816, + "step": 14520 + }, + { + "epoch": 0.53, + "learning_rate": 1.4700626213058922e-07, + "loss": 0.689, + "step": 14521 + }, + { + "epoch": 0.53, + "learning_rate": 1.4700260006591715e-07, + "loss": 0.6882, + "step": 14522 + }, + { + "epoch": 0.53, + "learning_rate": 1.469989380012451e-07, + "loss": 0.706, + "step": 14523 + }, + { + "epoch": 0.53, + "learning_rate": 1.4699527593657303e-07, + "loss": 0.6927, + "step": 14524 + }, + { + "epoch": 0.53, + "learning_rate": 1.4699161387190098e-07, + "loss": 0.6842, + "step": 14525 + }, + { + "epoch": 0.53, + "learning_rate": 1.469879518072289e-07, + "loss": 0.7022, + "step": 14526 + }, + { + "epoch": 0.53, + "learning_rate": 1.4698428974255686e-07, + "loss": 0.6933, + "step": 14527 + }, + { + "epoch": 0.53, + "learning_rate": 1.469806276778848e-07, + "loss": 0.6888, + "step": 14528 + }, + { + "epoch": 0.53, + "learning_rate": 1.4697696561321274e-07, + "loss": 0.6955, + "step": 14529 + }, + { + "epoch": 0.53, + "learning_rate": 1.4697330354854064e-07, + "loss": 0.68, + "step": 14530 + }, + { + "epoch": 0.53, + "learning_rate": 1.469696414838686e-07, + "loss": 0.7007, + "step": 14531 + }, + { + "epoch": 0.53, + "learning_rate": 1.4696597941919652e-07, + "loss": 0.6944, + "step": 14532 + }, + { + "epoch": 0.53, + "learning_rate": 1.4696231735452448e-07, + "loss": 0.6949, + "step": 14533 + }, + { + "epoch": 0.53, + "learning_rate": 1.469586552898524e-07, + "loss": 0.706, + "step": 14534 + }, + { + "epoch": 0.53, + "learning_rate": 1.4695499322518036e-07, + "loss": 0.7027, + "step": 14535 + }, + { + "epoch": 0.53, + "learning_rate": 1.4695133116050828e-07, + "loss": 0.698, + "step": 14536 + }, + { + "epoch": 0.53, + "learning_rate": 1.4694766909583624e-07, + "loss": 0.6959, + "step": 14537 + }, + { + "epoch": 0.53, + "learning_rate": 1.4694400703116416e-07, + "loss": 0.6774, + "step": 14538 + }, + { + "epoch": 0.53, + "learning_rate": 1.469403449664921e-07, + "loss": 0.6856, + "step": 14539 + }, + { + "epoch": 0.53, + "learning_rate": 1.4693668290182004e-07, + "loss": 0.6962, + "step": 14540 + }, + { + "epoch": 0.53, + "learning_rate": 1.4693302083714797e-07, + "loss": 0.6982, + "step": 14541 + }, + { + "epoch": 0.53, + "learning_rate": 1.4692935877247592e-07, + "loss": 0.7082, + "step": 14542 + }, + { + "epoch": 0.53, + "learning_rate": 1.4692569670780385e-07, + "loss": 0.6965, + "step": 14543 + }, + { + "epoch": 0.53, + "learning_rate": 1.469220346431318e-07, + "loss": 0.6915, + "step": 14544 + }, + { + "epoch": 0.53, + "learning_rate": 1.4691837257845973e-07, + "loss": 0.704, + "step": 14545 + }, + { + "epoch": 0.53, + "learning_rate": 1.4691471051378768e-07, + "loss": 0.7008, + "step": 14546 + }, + { + "epoch": 0.53, + "learning_rate": 1.469110484491156e-07, + "loss": 0.6879, + "step": 14547 + }, + { + "epoch": 0.53, + "learning_rate": 1.4690738638444354e-07, + "loss": 0.6883, + "step": 14548 + }, + { + "epoch": 0.53, + "learning_rate": 1.4690372431977146e-07, + "loss": 0.6865, + "step": 14549 + }, + { + "epoch": 0.53, + "learning_rate": 1.4690006225509942e-07, + "loss": 0.6947, + "step": 14550 + }, + { + "epoch": 0.53, + "learning_rate": 1.4689640019042734e-07, + "loss": 0.6937, + "step": 14551 + }, + { + "epoch": 0.53, + "learning_rate": 1.468927381257553e-07, + "loss": 0.6707, + "step": 14552 + }, + { + "epoch": 0.53, + "learning_rate": 1.4688907606108322e-07, + "loss": 0.6987, + "step": 14553 + }, + { + "epoch": 0.53, + "learning_rate": 1.4688541399641118e-07, + "loss": 0.6957, + "step": 14554 + }, + { + "epoch": 0.53, + "learning_rate": 1.468817519317391e-07, + "loss": 0.6999, + "step": 14555 + }, + { + "epoch": 0.53, + "learning_rate": 1.4687808986706706e-07, + "loss": 0.7039, + "step": 14556 + }, + { + "epoch": 0.53, + "learning_rate": 1.4687442780239498e-07, + "loss": 0.7044, + "step": 14557 + }, + { + "epoch": 0.53, + "learning_rate": 1.4687076573772294e-07, + "loss": 0.6883, + "step": 14558 + }, + { + "epoch": 0.53, + "learning_rate": 1.4686710367305086e-07, + "loss": 0.684, + "step": 14559 + }, + { + "epoch": 0.53, + "learning_rate": 1.4686344160837881e-07, + "loss": 0.6887, + "step": 14560 + }, + { + "epoch": 0.53, + "learning_rate": 1.4685977954370674e-07, + "loss": 0.691, + "step": 14561 + }, + { + "epoch": 0.53, + "learning_rate": 1.4685611747903467e-07, + "loss": 0.6995, + "step": 14562 + }, + { + "epoch": 0.53, + "learning_rate": 1.4685245541436262e-07, + "loss": 0.6899, + "step": 14563 + }, + { + "epoch": 0.53, + "learning_rate": 1.4684879334969055e-07, + "loss": 0.6985, + "step": 14564 + }, + { + "epoch": 0.53, + "learning_rate": 1.468451312850185e-07, + "loss": 0.7105, + "step": 14565 + }, + { + "epoch": 0.53, + "learning_rate": 1.468414692203464e-07, + "loss": 0.6868, + "step": 14566 + }, + { + "epoch": 0.53, + "learning_rate": 1.4683780715567435e-07, + "loss": 0.711, + "step": 14567 + }, + { + "epoch": 0.53, + "learning_rate": 1.4683414509100228e-07, + "loss": 0.7075, + "step": 14568 + }, + { + "epoch": 0.53, + "learning_rate": 1.4683048302633023e-07, + "loss": 0.7035, + "step": 14569 + }, + { + "epoch": 0.53, + "learning_rate": 1.4682682096165816e-07, + "loss": 0.6866, + "step": 14570 + }, + { + "epoch": 0.53, + "learning_rate": 1.4682315889698611e-07, + "loss": 0.696, + "step": 14571 + }, + { + "epoch": 0.53, + "learning_rate": 1.4681949683231404e-07, + "loss": 0.7149, + "step": 14572 + }, + { + "epoch": 0.53, + "learning_rate": 1.46815834767642e-07, + "loss": 0.7048, + "step": 14573 + }, + { + "epoch": 0.53, + "learning_rate": 1.4681217270296992e-07, + "loss": 0.6917, + "step": 14574 + }, + { + "epoch": 0.53, + "learning_rate": 1.4680851063829787e-07, + "loss": 0.6937, + "step": 14575 + }, + { + "epoch": 0.53, + "learning_rate": 1.468048485736258e-07, + "loss": 0.7038, + "step": 14576 + }, + { + "epoch": 0.53, + "learning_rate": 1.4680118650895375e-07, + "loss": 0.682, + "step": 14577 + }, + { + "epoch": 0.53, + "learning_rate": 1.4679752444428168e-07, + "loss": 0.677, + "step": 14578 + }, + { + "epoch": 0.53, + "learning_rate": 1.4679386237960963e-07, + "loss": 0.6911, + "step": 14579 + }, + { + "epoch": 0.53, + "learning_rate": 1.4679020031493756e-07, + "loss": 0.6674, + "step": 14580 + }, + { + "epoch": 0.53, + "learning_rate": 1.4678653825026551e-07, + "loss": 0.6876, + "step": 14581 + }, + { + "epoch": 0.53, + "learning_rate": 1.4678287618559344e-07, + "loss": 0.6858, + "step": 14582 + }, + { + "epoch": 0.53, + "learning_rate": 1.4677921412092137e-07, + "loss": 0.7019, + "step": 14583 + }, + { + "epoch": 0.53, + "learning_rate": 1.467755520562493e-07, + "loss": 0.7112, + "step": 14584 + }, + { + "epoch": 0.53, + "learning_rate": 1.4677188999157725e-07, + "loss": 0.7022, + "step": 14585 + }, + { + "epoch": 0.53, + "learning_rate": 1.4676822792690517e-07, + "loss": 0.6941, + "step": 14586 + }, + { + "epoch": 0.53, + "learning_rate": 1.4676456586223313e-07, + "loss": 0.6746, + "step": 14587 + }, + { + "epoch": 0.53, + "learning_rate": 1.4676090379756105e-07, + "loss": 0.6961, + "step": 14588 + }, + { + "epoch": 0.53, + "learning_rate": 1.4675724173288898e-07, + "loss": 0.6979, + "step": 14589 + }, + { + "epoch": 0.53, + "learning_rate": 1.4675357966821693e-07, + "loss": 0.703, + "step": 14590 + }, + { + "epoch": 0.53, + "learning_rate": 1.4674991760354486e-07, + "loss": 0.6982, + "step": 14591 + }, + { + "epoch": 0.53, + "learning_rate": 1.467462555388728e-07, + "loss": 0.6922, + "step": 14592 + }, + { + "epoch": 0.53, + "learning_rate": 1.4674259347420074e-07, + "loss": 0.6891, + "step": 14593 + }, + { + "epoch": 0.53, + "learning_rate": 1.467389314095287e-07, + "loss": 0.6943, + "step": 14594 + }, + { + "epoch": 0.53, + "learning_rate": 1.4673526934485662e-07, + "loss": 0.7012, + "step": 14595 + }, + { + "epoch": 0.53, + "learning_rate": 1.4673160728018457e-07, + "loss": 0.7072, + "step": 14596 + }, + { + "epoch": 0.53, + "learning_rate": 1.467279452155125e-07, + "loss": 0.6892, + "step": 14597 + }, + { + "epoch": 0.53, + "learning_rate": 1.4672428315084045e-07, + "loss": 0.6792, + "step": 14598 + }, + { + "epoch": 0.53, + "learning_rate": 1.4672062108616838e-07, + "loss": 0.6928, + "step": 14599 + }, + { + "epoch": 0.53, + "learning_rate": 1.4671695902149633e-07, + "loss": 0.6991, + "step": 14600 + }, + { + "epoch": 0.53, + "learning_rate": 1.4671329695682426e-07, + "loss": 0.6999, + "step": 14601 + }, + { + "epoch": 0.53, + "learning_rate": 1.4670963489215219e-07, + "loss": 0.7082, + "step": 14602 + }, + { + "epoch": 0.53, + "learning_rate": 1.467059728274801e-07, + "loss": 0.6789, + "step": 14603 + }, + { + "epoch": 0.53, + "learning_rate": 1.4670231076280807e-07, + "loss": 0.6942, + "step": 14604 + }, + { + "epoch": 0.53, + "learning_rate": 1.46698648698136e-07, + "loss": 0.6925, + "step": 14605 + }, + { + "epoch": 0.53, + "learning_rate": 1.4669498663346395e-07, + "loss": 0.6863, + "step": 14606 + }, + { + "epoch": 0.53, + "learning_rate": 1.4669132456879187e-07, + "loss": 0.6835, + "step": 14607 + }, + { + "epoch": 0.53, + "learning_rate": 1.4668766250411983e-07, + "loss": 0.6893, + "step": 14608 + }, + { + "epoch": 0.53, + "learning_rate": 1.4668400043944775e-07, + "loss": 0.6933, + "step": 14609 + }, + { + "epoch": 0.53, + "learning_rate": 1.4668033837477568e-07, + "loss": 0.7073, + "step": 14610 + }, + { + "epoch": 0.53, + "learning_rate": 1.4667667631010363e-07, + "loss": 0.6821, + "step": 14611 + }, + { + "epoch": 0.53, + "learning_rate": 1.4667301424543156e-07, + "loss": 0.7066, + "step": 14612 + }, + { + "epoch": 0.53, + "learning_rate": 1.466693521807595e-07, + "loss": 0.6934, + "step": 14613 + }, + { + "epoch": 0.53, + "learning_rate": 1.4666569011608744e-07, + "loss": 0.6986, + "step": 14614 + }, + { + "epoch": 0.53, + "learning_rate": 1.466620280514154e-07, + "loss": 0.6902, + "step": 14615 + }, + { + "epoch": 0.53, + "learning_rate": 1.4665836598674332e-07, + "loss": 0.7002, + "step": 14616 + }, + { + "epoch": 0.53, + "learning_rate": 1.4665470392207127e-07, + "loss": 0.6967, + "step": 14617 + }, + { + "epoch": 0.53, + "learning_rate": 1.466510418573992e-07, + "loss": 0.6892, + "step": 14618 + }, + { + "epoch": 0.53, + "learning_rate": 1.4664737979272715e-07, + "loss": 0.6938, + "step": 14619 + }, + { + "epoch": 0.53, + "learning_rate": 1.4664371772805505e-07, + "loss": 0.6966, + "step": 14620 + }, + { + "epoch": 0.53, + "learning_rate": 1.46640055663383e-07, + "loss": 0.6933, + "step": 14621 + }, + { + "epoch": 0.53, + "learning_rate": 1.4663639359871093e-07, + "loss": 0.6889, + "step": 14622 + }, + { + "epoch": 0.54, + "learning_rate": 1.4663273153403888e-07, + "loss": 0.6919, + "step": 14623 + }, + { + "epoch": 0.54, + "learning_rate": 1.466290694693668e-07, + "loss": 0.7018, + "step": 14624 + }, + { + "epoch": 0.54, + "learning_rate": 1.4662540740469476e-07, + "loss": 0.699, + "step": 14625 + }, + { + "epoch": 0.54, + "learning_rate": 1.466217453400227e-07, + "loss": 0.6973, + "step": 14626 + }, + { + "epoch": 0.54, + "learning_rate": 1.4661808327535064e-07, + "loss": 0.6851, + "step": 14627 + }, + { + "epoch": 0.54, + "learning_rate": 1.4661442121067857e-07, + "loss": 0.6738, + "step": 14628 + }, + { + "epoch": 0.54, + "learning_rate": 1.4661075914600652e-07, + "loss": 0.6887, + "step": 14629 + }, + { + "epoch": 0.54, + "learning_rate": 1.4660709708133445e-07, + "loss": 0.6836, + "step": 14630 + }, + { + "epoch": 0.54, + "learning_rate": 1.466034350166624e-07, + "loss": 0.6896, + "step": 14631 + }, + { + "epoch": 0.54, + "learning_rate": 1.4659977295199033e-07, + "loss": 0.6879, + "step": 14632 + }, + { + "epoch": 0.54, + "learning_rate": 1.4659611088731826e-07, + "loss": 0.6868, + "step": 14633 + }, + { + "epoch": 0.54, + "learning_rate": 1.465924488226462e-07, + "loss": 0.6885, + "step": 14634 + }, + { + "epoch": 0.54, + "learning_rate": 1.4658878675797414e-07, + "loss": 0.7075, + "step": 14635 + }, + { + "epoch": 0.54, + "learning_rate": 1.465851246933021e-07, + "loss": 0.7089, + "step": 14636 + }, + { + "epoch": 0.54, + "learning_rate": 1.4658146262863002e-07, + "loss": 0.686, + "step": 14637 + }, + { + "epoch": 0.54, + "learning_rate": 1.4657780056395794e-07, + "loss": 0.6952, + "step": 14638 + }, + { + "epoch": 0.54, + "learning_rate": 1.4657413849928587e-07, + "loss": 0.696, + "step": 14639 + }, + { + "epoch": 0.54, + "learning_rate": 1.4657047643461382e-07, + "loss": 0.6856, + "step": 14640 + }, + { + "epoch": 0.54, + "learning_rate": 1.4656681436994175e-07, + "loss": 0.6903, + "step": 14641 + }, + { + "epoch": 0.54, + "learning_rate": 1.465631523052697e-07, + "loss": 0.7042, + "step": 14642 + }, + { + "epoch": 0.54, + "learning_rate": 1.4655949024059763e-07, + "loss": 0.6814, + "step": 14643 + }, + { + "epoch": 0.54, + "learning_rate": 1.4655582817592558e-07, + "loss": 0.6949, + "step": 14644 + }, + { + "epoch": 0.54, + "learning_rate": 1.465521661112535e-07, + "loss": 0.6816, + "step": 14645 + }, + { + "epoch": 0.54, + "learning_rate": 1.4654850404658146e-07, + "loss": 0.6944, + "step": 14646 + }, + { + "epoch": 0.54, + "learning_rate": 1.465448419819094e-07, + "loss": 0.6799, + "step": 14647 + }, + { + "epoch": 0.54, + "learning_rate": 1.4654117991723734e-07, + "loss": 0.6791, + "step": 14648 + }, + { + "epoch": 0.54, + "learning_rate": 1.4653751785256527e-07, + "loss": 0.6909, + "step": 14649 + }, + { + "epoch": 0.54, + "learning_rate": 1.4653385578789322e-07, + "loss": 0.679, + "step": 14650 + }, + { + "epoch": 0.54, + "learning_rate": 1.4653019372322115e-07, + "loss": 0.6812, + "step": 14651 + }, + { + "epoch": 0.54, + "learning_rate": 1.465265316585491e-07, + "loss": 0.6985, + "step": 14652 + }, + { + "epoch": 0.54, + "learning_rate": 1.4652286959387703e-07, + "loss": 0.6999, + "step": 14653 + }, + { + "epoch": 0.54, + "learning_rate": 1.4651920752920496e-07, + "loss": 0.7148, + "step": 14654 + }, + { + "epoch": 0.54, + "learning_rate": 1.465155454645329e-07, + "loss": 0.6857, + "step": 14655 + }, + { + "epoch": 0.54, + "learning_rate": 1.4651188339986084e-07, + "loss": 0.6881, + "step": 14656 + }, + { + "epoch": 0.54, + "learning_rate": 1.4650822133518876e-07, + "loss": 0.7068, + "step": 14657 + }, + { + "epoch": 0.54, + "learning_rate": 1.4650455927051672e-07, + "loss": 0.6972, + "step": 14658 + }, + { + "epoch": 0.54, + "learning_rate": 1.4650089720584464e-07, + "loss": 0.7015, + "step": 14659 + }, + { + "epoch": 0.54, + "learning_rate": 1.4649723514117257e-07, + "loss": 0.6918, + "step": 14660 + }, + { + "epoch": 0.54, + "learning_rate": 1.4649357307650052e-07, + "loss": 0.6976, + "step": 14661 + }, + { + "epoch": 0.54, + "learning_rate": 1.4648991101182845e-07, + "loss": 0.7036, + "step": 14662 + }, + { + "epoch": 0.54, + "learning_rate": 1.464862489471564e-07, + "loss": 0.6814, + "step": 14663 + }, + { + "epoch": 0.54, + "learning_rate": 1.4648258688248433e-07, + "loss": 0.6973, + "step": 14664 + }, + { + "epoch": 0.54, + "learning_rate": 1.4647892481781228e-07, + "loss": 0.6869, + "step": 14665 + }, + { + "epoch": 0.54, + "learning_rate": 1.464752627531402e-07, + "loss": 0.6929, + "step": 14666 + }, + { + "epoch": 0.54, + "learning_rate": 1.4647160068846816e-07, + "loss": 0.6939, + "step": 14667 + }, + { + "epoch": 0.54, + "learning_rate": 1.464679386237961e-07, + "loss": 0.6812, + "step": 14668 + }, + { + "epoch": 0.54, + "learning_rate": 1.4646427655912404e-07, + "loss": 0.684, + "step": 14669 + }, + { + "epoch": 0.54, + "learning_rate": 1.4646061449445197e-07, + "loss": 0.6846, + "step": 14670 + }, + { + "epoch": 0.54, + "learning_rate": 1.4645695242977992e-07, + "loss": 0.6974, + "step": 14671 + }, + { + "epoch": 0.54, + "learning_rate": 1.4645329036510785e-07, + "loss": 0.6707, + "step": 14672 + }, + { + "epoch": 0.54, + "learning_rate": 1.4644962830043578e-07, + "loss": 0.6904, + "step": 14673 + }, + { + "epoch": 0.54, + "learning_rate": 1.464459662357637e-07, + "loss": 0.6918, + "step": 14674 + }, + { + "epoch": 0.54, + "learning_rate": 1.4644230417109166e-07, + "loss": 0.6757, + "step": 14675 + }, + { + "epoch": 0.54, + "learning_rate": 1.4643864210641958e-07, + "loss": 0.6956, + "step": 14676 + }, + { + "epoch": 0.54, + "learning_rate": 1.4643498004174754e-07, + "loss": 0.687, + "step": 14677 + }, + { + "epoch": 0.54, + "learning_rate": 1.4643131797707546e-07, + "loss": 0.7069, + "step": 14678 + }, + { + "epoch": 0.54, + "learning_rate": 1.4642765591240341e-07, + "loss": 0.7134, + "step": 14679 + }, + { + "epoch": 0.54, + "learning_rate": 1.4642399384773134e-07, + "loss": 0.7094, + "step": 14680 + }, + { + "epoch": 0.54, + "learning_rate": 1.4642033178305927e-07, + "loss": 0.7068, + "step": 14681 + }, + { + "epoch": 0.54, + "learning_rate": 1.4641666971838722e-07, + "loss": 0.7034, + "step": 14682 + }, + { + "epoch": 0.54, + "learning_rate": 1.4641300765371515e-07, + "loss": 0.6823, + "step": 14683 + }, + { + "epoch": 0.54, + "learning_rate": 1.464093455890431e-07, + "loss": 0.6789, + "step": 14684 + }, + { + "epoch": 0.54, + "learning_rate": 1.4640568352437103e-07, + "loss": 0.6798, + "step": 14685 + }, + { + "epoch": 0.54, + "learning_rate": 1.4640202145969898e-07, + "loss": 0.6938, + "step": 14686 + }, + { + "epoch": 0.54, + "learning_rate": 1.463983593950269e-07, + "loss": 0.7025, + "step": 14687 + }, + { + "epoch": 0.54, + "learning_rate": 1.4639469733035486e-07, + "loss": 0.7044, + "step": 14688 + }, + { + "epoch": 0.54, + "learning_rate": 1.463910352656828e-07, + "loss": 0.6894, + "step": 14689 + }, + { + "epoch": 0.54, + "learning_rate": 1.4638737320101074e-07, + "loss": 0.6841, + "step": 14690 + }, + { + "epoch": 0.54, + "learning_rate": 1.4638371113633864e-07, + "loss": 0.7009, + "step": 14691 + }, + { + "epoch": 0.54, + "learning_rate": 1.463800490716666e-07, + "loss": 0.6938, + "step": 14692 + }, + { + "epoch": 0.54, + "learning_rate": 1.4637638700699452e-07, + "loss": 0.7028, + "step": 14693 + }, + { + "epoch": 0.54, + "learning_rate": 1.4637272494232247e-07, + "loss": 0.7013, + "step": 14694 + }, + { + "epoch": 0.54, + "learning_rate": 1.463690628776504e-07, + "loss": 0.6876, + "step": 14695 + }, + { + "epoch": 0.54, + "learning_rate": 1.4636540081297835e-07, + "loss": 0.6834, + "step": 14696 + }, + { + "epoch": 0.54, + "learning_rate": 1.4636173874830628e-07, + "loss": 0.6956, + "step": 14697 + }, + { + "epoch": 0.54, + "learning_rate": 1.4635807668363423e-07, + "loss": 0.6963, + "step": 14698 + }, + { + "epoch": 0.54, + "learning_rate": 1.4635441461896216e-07, + "loss": 0.6917, + "step": 14699 + }, + { + "epoch": 0.54, + "learning_rate": 1.4635075255429011e-07, + "loss": 0.6944, + "step": 14700 + }, + { + "epoch": 0.54, + "learning_rate": 1.4634709048961804e-07, + "loss": 0.6949, + "step": 14701 + }, + { + "epoch": 0.54, + "learning_rate": 1.46343428424946e-07, + "loss": 0.6921, + "step": 14702 + }, + { + "epoch": 0.54, + "learning_rate": 1.4633976636027392e-07, + "loss": 0.707, + "step": 14703 + }, + { + "epoch": 0.54, + "learning_rate": 1.4633610429560185e-07, + "loss": 0.685, + "step": 14704 + }, + { + "epoch": 0.54, + "learning_rate": 1.463324422309298e-07, + "loss": 0.6905, + "step": 14705 + }, + { + "epoch": 0.54, + "learning_rate": 1.4632878016625773e-07, + "loss": 0.6906, + "step": 14706 + }, + { + "epoch": 0.54, + "learning_rate": 1.4632511810158568e-07, + "loss": 0.6946, + "step": 14707 + }, + { + "epoch": 0.54, + "learning_rate": 1.463214560369136e-07, + "loss": 0.6898, + "step": 14708 + }, + { + "epoch": 0.54, + "learning_rate": 1.4631779397224153e-07, + "loss": 0.6896, + "step": 14709 + }, + { + "epoch": 0.54, + "learning_rate": 1.4631413190756946e-07, + "loss": 0.6912, + "step": 14710 + }, + { + "epoch": 0.54, + "learning_rate": 1.4631046984289741e-07, + "loss": 0.7008, + "step": 14711 + }, + { + "epoch": 0.54, + "learning_rate": 1.4630680777822534e-07, + "loss": 0.6957, + "step": 14712 + }, + { + "epoch": 0.54, + "learning_rate": 1.463031457135533e-07, + "loss": 0.6949, + "step": 14713 + }, + { + "epoch": 0.54, + "learning_rate": 1.4629948364888122e-07, + "loss": 0.703, + "step": 14714 + }, + { + "epoch": 0.54, + "learning_rate": 1.4629582158420917e-07, + "loss": 0.6999, + "step": 14715 + }, + { + "epoch": 0.54, + "learning_rate": 1.462921595195371e-07, + "loss": 0.6911, + "step": 14716 + }, + { + "epoch": 0.54, + "learning_rate": 1.4628849745486505e-07, + "loss": 0.6889, + "step": 14717 + }, + { + "epoch": 0.54, + "learning_rate": 1.4628483539019298e-07, + "loss": 0.6918, + "step": 14718 + }, + { + "epoch": 0.54, + "learning_rate": 1.4628117332552093e-07, + "loss": 0.6982, + "step": 14719 + }, + { + "epoch": 0.54, + "learning_rate": 1.4627751126084886e-07, + "loss": 0.6952, + "step": 14720 + }, + { + "epoch": 0.54, + "learning_rate": 1.462738491961768e-07, + "loss": 0.7015, + "step": 14721 + }, + { + "epoch": 0.54, + "learning_rate": 1.4627018713150474e-07, + "loss": 0.6897, + "step": 14722 + }, + { + "epoch": 0.54, + "learning_rate": 1.462665250668327e-07, + "loss": 0.6862, + "step": 14723 + }, + { + "epoch": 0.54, + "learning_rate": 1.4626286300216062e-07, + "loss": 0.6959, + "step": 14724 + }, + { + "epoch": 0.54, + "learning_rate": 1.4625920093748855e-07, + "loss": 0.7007, + "step": 14725 + }, + { + "epoch": 0.54, + "learning_rate": 1.462555388728165e-07, + "loss": 0.6968, + "step": 14726 + }, + { + "epoch": 0.54, + "learning_rate": 1.4625187680814443e-07, + "loss": 0.6977, + "step": 14727 + }, + { + "epoch": 0.54, + "learning_rate": 1.4624821474347235e-07, + "loss": 0.7068, + "step": 14728 + }, + { + "epoch": 0.54, + "learning_rate": 1.462445526788003e-07, + "loss": 0.6854, + "step": 14729 + }, + { + "epoch": 0.54, + "learning_rate": 1.4624089061412823e-07, + "loss": 0.6944, + "step": 14730 + }, + { + "epoch": 0.54, + "learning_rate": 1.4623722854945616e-07, + "loss": 0.6935, + "step": 14731 + }, + { + "epoch": 0.54, + "learning_rate": 1.462335664847841e-07, + "loss": 0.704, + "step": 14732 + }, + { + "epoch": 0.54, + "learning_rate": 1.4622990442011204e-07, + "loss": 0.6929, + "step": 14733 + }, + { + "epoch": 0.54, + "learning_rate": 1.4622624235544e-07, + "loss": 0.6947, + "step": 14734 + }, + { + "epoch": 0.54, + "learning_rate": 1.4622258029076792e-07, + "loss": 0.7038, + "step": 14735 + }, + { + "epoch": 0.54, + "learning_rate": 1.4621891822609587e-07, + "loss": 0.6907, + "step": 14736 + }, + { + "epoch": 0.54, + "learning_rate": 1.462152561614238e-07, + "loss": 0.7156, + "step": 14737 + }, + { + "epoch": 0.54, + "learning_rate": 1.4621159409675175e-07, + "loss": 0.7071, + "step": 14738 + }, + { + "epoch": 0.54, + "learning_rate": 1.4620793203207968e-07, + "loss": 0.6929, + "step": 14739 + }, + { + "epoch": 0.54, + "learning_rate": 1.4620426996740763e-07, + "loss": 0.6963, + "step": 14740 + }, + { + "epoch": 0.54, + "learning_rate": 1.4620060790273556e-07, + "loss": 0.6901, + "step": 14741 + }, + { + "epoch": 0.54, + "learning_rate": 1.461969458380635e-07, + "loss": 0.6842, + "step": 14742 + }, + { + "epoch": 0.54, + "learning_rate": 1.4619328377339144e-07, + "loss": 0.6848, + "step": 14743 + }, + { + "epoch": 0.54, + "learning_rate": 1.461896217087194e-07, + "loss": 0.6962, + "step": 14744 + }, + { + "epoch": 0.54, + "learning_rate": 1.461859596440473e-07, + "loss": 0.6921, + "step": 14745 + }, + { + "epoch": 0.54, + "learning_rate": 1.4618229757937524e-07, + "loss": 0.6851, + "step": 14746 + }, + { + "epoch": 0.54, + "learning_rate": 1.4617863551470317e-07, + "loss": 0.6927, + "step": 14747 + }, + { + "epoch": 0.54, + "learning_rate": 1.4617497345003112e-07, + "loss": 0.6852, + "step": 14748 + }, + { + "epoch": 0.54, + "learning_rate": 1.4617131138535905e-07, + "loss": 0.6982, + "step": 14749 + }, + { + "epoch": 0.54, + "learning_rate": 1.46167649320687e-07, + "loss": 0.6818, + "step": 14750 + }, + { + "epoch": 0.54, + "learning_rate": 1.4616398725601493e-07, + "loss": 0.6874, + "step": 14751 + }, + { + "epoch": 0.54, + "learning_rate": 1.4616032519134286e-07, + "loss": 0.6946, + "step": 14752 + }, + { + "epoch": 0.54, + "learning_rate": 1.461566631266708e-07, + "loss": 0.6996, + "step": 14753 + }, + { + "epoch": 0.54, + "learning_rate": 1.4615300106199874e-07, + "loss": 0.7006, + "step": 14754 + }, + { + "epoch": 0.54, + "learning_rate": 1.461493389973267e-07, + "loss": 0.6821, + "step": 14755 + }, + { + "epoch": 0.54, + "learning_rate": 1.4614567693265462e-07, + "loss": 0.7013, + "step": 14756 + }, + { + "epoch": 0.54, + "learning_rate": 1.4614201486798257e-07, + "loss": 0.7017, + "step": 14757 + }, + { + "epoch": 0.54, + "learning_rate": 1.461383528033105e-07, + "loss": 0.6732, + "step": 14758 + }, + { + "epoch": 0.54, + "learning_rate": 1.4613469073863845e-07, + "loss": 0.7015, + "step": 14759 + }, + { + "epoch": 0.54, + "learning_rate": 1.4613102867396638e-07, + "loss": 0.6976, + "step": 14760 + }, + { + "epoch": 0.54, + "learning_rate": 1.4612736660929433e-07, + "loss": 0.6953, + "step": 14761 + }, + { + "epoch": 0.54, + "learning_rate": 1.4612370454462226e-07, + "loss": 0.6902, + "step": 14762 + }, + { + "epoch": 0.54, + "learning_rate": 1.4612004247995018e-07, + "loss": 0.6866, + "step": 14763 + }, + { + "epoch": 0.54, + "learning_rate": 1.461163804152781e-07, + "loss": 0.6932, + "step": 14764 + }, + { + "epoch": 0.54, + "learning_rate": 1.4611271835060606e-07, + "loss": 0.6991, + "step": 14765 + }, + { + "epoch": 0.54, + "learning_rate": 1.46109056285934e-07, + "loss": 0.6877, + "step": 14766 + }, + { + "epoch": 0.54, + "learning_rate": 1.4610539422126194e-07, + "loss": 0.6854, + "step": 14767 + }, + { + "epoch": 0.54, + "learning_rate": 1.4610173215658987e-07, + "loss": 0.7011, + "step": 14768 + }, + { + "epoch": 0.54, + "learning_rate": 1.4609807009191782e-07, + "loss": 0.693, + "step": 14769 + }, + { + "epoch": 0.54, + "learning_rate": 1.4609440802724575e-07, + "loss": 0.6826, + "step": 14770 + }, + { + "epoch": 0.54, + "learning_rate": 1.460907459625737e-07, + "loss": 0.6922, + "step": 14771 + }, + { + "epoch": 0.54, + "learning_rate": 1.4608708389790163e-07, + "loss": 0.6898, + "step": 14772 + }, + { + "epoch": 0.54, + "learning_rate": 1.4608342183322958e-07, + "loss": 0.6887, + "step": 14773 + }, + { + "epoch": 0.54, + "learning_rate": 1.460797597685575e-07, + "loss": 0.6913, + "step": 14774 + }, + { + "epoch": 0.54, + "learning_rate": 1.4607609770388544e-07, + "loss": 0.6997, + "step": 14775 + }, + { + "epoch": 0.54, + "learning_rate": 1.460724356392134e-07, + "loss": 0.7112, + "step": 14776 + }, + { + "epoch": 0.54, + "learning_rate": 1.4606877357454132e-07, + "loss": 0.6822, + "step": 14777 + }, + { + "epoch": 0.54, + "learning_rate": 1.4606511150986927e-07, + "loss": 0.7011, + "step": 14778 + }, + { + "epoch": 0.54, + "learning_rate": 1.460614494451972e-07, + "loss": 0.6762, + "step": 14779 + }, + { + "epoch": 0.54, + "learning_rate": 1.4605778738052515e-07, + "loss": 0.7026, + "step": 14780 + }, + { + "epoch": 0.54, + "learning_rate": 1.4605412531585305e-07, + "loss": 0.695, + "step": 14781 + }, + { + "epoch": 0.54, + "learning_rate": 1.46050463251181e-07, + "loss": 0.6957, + "step": 14782 + }, + { + "epoch": 0.54, + "learning_rate": 1.4604680118650893e-07, + "loss": 0.675, + "step": 14783 + }, + { + "epoch": 0.54, + "learning_rate": 1.4604313912183688e-07, + "loss": 0.6925, + "step": 14784 + }, + { + "epoch": 0.54, + "learning_rate": 1.460394770571648e-07, + "loss": 0.6972, + "step": 14785 + }, + { + "epoch": 0.54, + "learning_rate": 1.4603581499249276e-07, + "loss": 0.6842, + "step": 14786 + }, + { + "epoch": 0.54, + "learning_rate": 1.460321529278207e-07, + "loss": 0.6981, + "step": 14787 + }, + { + "epoch": 0.54, + "learning_rate": 1.4602849086314864e-07, + "loss": 0.6888, + "step": 14788 + }, + { + "epoch": 0.54, + "learning_rate": 1.4602482879847657e-07, + "loss": 0.6939, + "step": 14789 + }, + { + "epoch": 0.54, + "learning_rate": 1.4602116673380452e-07, + "loss": 0.6958, + "step": 14790 + }, + { + "epoch": 0.54, + "learning_rate": 1.4601750466913245e-07, + "loss": 0.6963, + "step": 14791 + }, + { + "epoch": 0.54, + "learning_rate": 1.460138426044604e-07, + "loss": 0.6755, + "step": 14792 + }, + { + "epoch": 0.54, + "learning_rate": 1.4601018053978833e-07, + "loss": 0.6986, + "step": 14793 + }, + { + "epoch": 0.54, + "learning_rate": 1.4600651847511628e-07, + "loss": 0.6859, + "step": 14794 + }, + { + "epoch": 0.54, + "learning_rate": 1.460028564104442e-07, + "loss": 0.6853, + "step": 14795 + }, + { + "epoch": 0.54, + "learning_rate": 1.4599919434577214e-07, + "loss": 0.7064, + "step": 14796 + }, + { + "epoch": 0.54, + "learning_rate": 1.459955322811001e-07, + "loss": 0.6936, + "step": 14797 + }, + { + "epoch": 0.54, + "learning_rate": 1.4599187021642802e-07, + "loss": 0.6898, + "step": 14798 + }, + { + "epoch": 0.54, + "learning_rate": 1.4598820815175594e-07, + "loss": 0.6985, + "step": 14799 + }, + { + "epoch": 0.54, + "learning_rate": 1.459845460870839e-07, + "loss": 0.6924, + "step": 14800 + }, + { + "epoch": 0.54, + "learning_rate": 1.4598088402241182e-07, + "loss": 0.6846, + "step": 14801 + }, + { + "epoch": 0.54, + "learning_rate": 1.4597722195773975e-07, + "loss": 0.7035, + "step": 14802 + }, + { + "epoch": 0.54, + "learning_rate": 1.459735598930677e-07, + "loss": 0.705, + "step": 14803 + }, + { + "epoch": 0.54, + "learning_rate": 1.4596989782839563e-07, + "loss": 0.6952, + "step": 14804 + }, + { + "epoch": 0.54, + "learning_rate": 1.4596623576372358e-07, + "loss": 0.7005, + "step": 14805 + }, + { + "epoch": 0.54, + "learning_rate": 1.459625736990515e-07, + "loss": 0.6887, + "step": 14806 + }, + { + "epoch": 0.54, + "learning_rate": 1.4595891163437946e-07, + "loss": 0.6945, + "step": 14807 + }, + { + "epoch": 0.54, + "learning_rate": 1.459552495697074e-07, + "loss": 0.6919, + "step": 14808 + }, + { + "epoch": 0.54, + "learning_rate": 1.4595158750503534e-07, + "loss": 0.7091, + "step": 14809 + }, + { + "epoch": 0.54, + "learning_rate": 1.4594792544036327e-07, + "loss": 0.6814, + "step": 14810 + }, + { + "epoch": 0.54, + "learning_rate": 1.4594426337569122e-07, + "loss": 0.704, + "step": 14811 + }, + { + "epoch": 0.54, + "learning_rate": 1.4594060131101915e-07, + "loss": 0.6916, + "step": 14812 + }, + { + "epoch": 0.54, + "learning_rate": 1.459369392463471e-07, + "loss": 0.6968, + "step": 14813 + }, + { + "epoch": 0.54, + "learning_rate": 1.4593327718167503e-07, + "loss": 0.6888, + "step": 14814 + }, + { + "epoch": 0.54, + "learning_rate": 1.4592961511700298e-07, + "loss": 0.6962, + "step": 14815 + }, + { + "epoch": 0.54, + "learning_rate": 1.459259530523309e-07, + "loss": 0.7021, + "step": 14816 + }, + { + "epoch": 0.54, + "learning_rate": 1.4592229098765883e-07, + "loss": 0.7, + "step": 14817 + }, + { + "epoch": 0.54, + "learning_rate": 1.4591862892298676e-07, + "loss": 0.6849, + "step": 14818 + }, + { + "epoch": 0.54, + "learning_rate": 1.4591496685831471e-07, + "loss": 0.6857, + "step": 14819 + }, + { + "epoch": 0.54, + "learning_rate": 1.4591130479364264e-07, + "loss": 0.698, + "step": 14820 + }, + { + "epoch": 0.54, + "learning_rate": 1.459076427289706e-07, + "loss": 0.7014, + "step": 14821 + }, + { + "epoch": 0.54, + "learning_rate": 1.4590398066429852e-07, + "loss": 0.7116, + "step": 14822 + }, + { + "epoch": 0.54, + "learning_rate": 1.4590031859962645e-07, + "loss": 0.6925, + "step": 14823 + }, + { + "epoch": 0.54, + "learning_rate": 1.458966565349544e-07, + "loss": 0.7046, + "step": 14824 + }, + { + "epoch": 0.54, + "learning_rate": 1.4589299447028233e-07, + "loss": 0.7054, + "step": 14825 + }, + { + "epoch": 0.54, + "learning_rate": 1.4588933240561028e-07, + "loss": 0.6867, + "step": 14826 + }, + { + "epoch": 0.54, + "learning_rate": 1.458856703409382e-07, + "loss": 0.6891, + "step": 14827 + }, + { + "epoch": 0.54, + "learning_rate": 1.4588200827626616e-07, + "loss": 0.6901, + "step": 14828 + }, + { + "epoch": 0.54, + "learning_rate": 1.4587834621159409e-07, + "loss": 0.6919, + "step": 14829 + }, + { + "epoch": 0.54, + "learning_rate": 1.4587468414692204e-07, + "loss": 0.6845, + "step": 14830 + }, + { + "epoch": 0.54, + "learning_rate": 1.4587102208224997e-07, + "loss": 0.693, + "step": 14831 + }, + { + "epoch": 0.54, + "learning_rate": 1.4586736001757792e-07, + "loss": 0.6919, + "step": 14832 + }, + { + "epoch": 0.54, + "learning_rate": 1.4586369795290585e-07, + "loss": 0.6926, + "step": 14833 + }, + { + "epoch": 0.54, + "learning_rate": 1.458600358882338e-07, + "loss": 0.7048, + "step": 14834 + }, + { + "epoch": 0.54, + "learning_rate": 1.458563738235617e-07, + "loss": 0.6997, + "step": 14835 + }, + { + "epoch": 0.54, + "learning_rate": 1.4585271175888965e-07, + "loss": 0.6948, + "step": 14836 + }, + { + "epoch": 0.54, + "learning_rate": 1.4584904969421758e-07, + "loss": 0.7012, + "step": 14837 + }, + { + "epoch": 0.54, + "learning_rate": 1.4584538762954553e-07, + "loss": 0.6912, + "step": 14838 + }, + { + "epoch": 0.54, + "learning_rate": 1.4584172556487346e-07, + "loss": 0.6888, + "step": 14839 + }, + { + "epoch": 0.54, + "learning_rate": 1.458380635002014e-07, + "loss": 0.6995, + "step": 14840 + }, + { + "epoch": 0.54, + "learning_rate": 1.4583440143552934e-07, + "loss": 0.7046, + "step": 14841 + }, + { + "epoch": 0.54, + "learning_rate": 1.458307393708573e-07, + "loss": 0.6918, + "step": 14842 + }, + { + "epoch": 0.54, + "learning_rate": 1.4582707730618522e-07, + "loss": 0.7062, + "step": 14843 + }, + { + "epoch": 0.54, + "learning_rate": 1.4582341524151317e-07, + "loss": 0.6939, + "step": 14844 + }, + { + "epoch": 0.54, + "learning_rate": 1.458197531768411e-07, + "loss": 0.7039, + "step": 14845 + }, + { + "epoch": 0.54, + "learning_rate": 1.4581609111216903e-07, + "loss": 0.7008, + "step": 14846 + }, + { + "epoch": 0.54, + "learning_rate": 1.4581242904749698e-07, + "loss": 0.6945, + "step": 14847 + }, + { + "epoch": 0.54, + "learning_rate": 1.458087669828249e-07, + "loss": 0.6963, + "step": 14848 + }, + { + "epoch": 0.54, + "learning_rate": 1.4580510491815286e-07, + "loss": 0.6948, + "step": 14849 + }, + { + "epoch": 0.54, + "learning_rate": 1.4580144285348079e-07, + "loss": 0.6954, + "step": 14850 + }, + { + "epoch": 0.54, + "learning_rate": 1.4579778078880874e-07, + "loss": 0.6995, + "step": 14851 + }, + { + "epoch": 0.54, + "learning_rate": 1.4579411872413667e-07, + "loss": 0.6826, + "step": 14852 + }, + { + "epoch": 0.54, + "learning_rate": 1.457904566594646e-07, + "loss": 0.6881, + "step": 14853 + }, + { + "epoch": 0.54, + "learning_rate": 1.4578679459479252e-07, + "loss": 0.6957, + "step": 14854 + }, + { + "epoch": 0.54, + "learning_rate": 1.4578313253012047e-07, + "loss": 0.7005, + "step": 14855 + }, + { + "epoch": 0.54, + "learning_rate": 1.457794704654484e-07, + "loss": 0.6954, + "step": 14856 + }, + { + "epoch": 0.54, + "learning_rate": 1.4577580840077635e-07, + "loss": 0.6918, + "step": 14857 + }, + { + "epoch": 0.54, + "learning_rate": 1.4577214633610428e-07, + "loss": 0.6961, + "step": 14858 + }, + { + "epoch": 0.54, + "learning_rate": 1.4576848427143223e-07, + "loss": 0.7018, + "step": 14859 + }, + { + "epoch": 0.54, + "learning_rate": 1.4576482220676016e-07, + "loss": 0.7012, + "step": 14860 + }, + { + "epoch": 0.54, + "learning_rate": 1.457611601420881e-07, + "loss": 0.6974, + "step": 14861 + }, + { + "epoch": 0.54, + "learning_rate": 1.4575749807741604e-07, + "loss": 0.6944, + "step": 14862 + }, + { + "epoch": 0.54, + "learning_rate": 1.45753836012744e-07, + "loss": 0.7023, + "step": 14863 + }, + { + "epoch": 0.54, + "learning_rate": 1.4575017394807192e-07, + "loss": 0.6955, + "step": 14864 + }, + { + "epoch": 0.54, + "learning_rate": 1.4574651188339987e-07, + "loss": 0.6969, + "step": 14865 + }, + { + "epoch": 0.54, + "learning_rate": 1.457428498187278e-07, + "loss": 0.6952, + "step": 14866 + }, + { + "epoch": 0.54, + "learning_rate": 1.4573918775405572e-07, + "loss": 0.6925, + "step": 14867 + }, + { + "epoch": 0.54, + "learning_rate": 1.4573552568938368e-07, + "loss": 0.6986, + "step": 14868 + }, + { + "epoch": 0.54, + "learning_rate": 1.457318636247116e-07, + "loss": 0.6886, + "step": 14869 + }, + { + "epoch": 0.54, + "learning_rate": 1.4572820156003956e-07, + "loss": 0.6929, + "step": 14870 + }, + { + "epoch": 0.54, + "learning_rate": 1.4572453949536748e-07, + "loss": 0.7055, + "step": 14871 + }, + { + "epoch": 0.54, + "learning_rate": 1.457208774306954e-07, + "loss": 0.6872, + "step": 14872 + }, + { + "epoch": 0.54, + "learning_rate": 1.4571721536602334e-07, + "loss": 0.6877, + "step": 14873 + }, + { + "epoch": 0.54, + "learning_rate": 1.457135533013513e-07, + "loss": 0.6993, + "step": 14874 + }, + { + "epoch": 0.54, + "learning_rate": 1.4570989123667922e-07, + "loss": 0.6928, + "step": 14875 + }, + { + "epoch": 0.54, + "learning_rate": 1.4570622917200717e-07, + "loss": 0.6888, + "step": 14876 + }, + { + "epoch": 0.54, + "learning_rate": 1.457025671073351e-07, + "loss": 0.6919, + "step": 14877 + }, + { + "epoch": 0.54, + "learning_rate": 1.4569890504266305e-07, + "loss": 0.6874, + "step": 14878 + }, + { + "epoch": 0.54, + "learning_rate": 1.4569524297799098e-07, + "loss": 0.6742, + "step": 14879 + }, + { + "epoch": 0.54, + "learning_rate": 1.4569158091331893e-07, + "loss": 0.6905, + "step": 14880 + }, + { + "epoch": 0.54, + "learning_rate": 1.4568791884864686e-07, + "loss": 0.6741, + "step": 14881 + }, + { + "epoch": 0.54, + "learning_rate": 1.456842567839748e-07, + "loss": 0.6959, + "step": 14882 + }, + { + "epoch": 0.54, + "learning_rate": 1.4568059471930274e-07, + "loss": 0.681, + "step": 14883 + }, + { + "epoch": 0.54, + "learning_rate": 1.456769326546307e-07, + "loss": 0.6788, + "step": 14884 + }, + { + "epoch": 0.54, + "learning_rate": 1.4567327058995862e-07, + "loss": 0.6655, + "step": 14885 + }, + { + "epoch": 0.54, + "learning_rate": 1.4566960852528657e-07, + "loss": 0.6885, + "step": 14886 + }, + { + "epoch": 0.54, + "learning_rate": 1.456659464606145e-07, + "loss": 0.6858, + "step": 14887 + }, + { + "epoch": 0.54, + "learning_rate": 1.4566228439594242e-07, + "loss": 0.683, + "step": 14888 + }, + { + "epoch": 0.54, + "learning_rate": 1.4565862233127035e-07, + "loss": 0.6934, + "step": 14889 + }, + { + "epoch": 0.54, + "learning_rate": 1.456549602665983e-07, + "loss": 0.6957, + "step": 14890 + }, + { + "epoch": 0.54, + "learning_rate": 1.4565129820192623e-07, + "loss": 0.6905, + "step": 14891 + }, + { + "epoch": 0.54, + "learning_rate": 1.4564763613725418e-07, + "loss": 0.6918, + "step": 14892 + }, + { + "epoch": 0.54, + "learning_rate": 1.456439740725821e-07, + "loss": 0.6923, + "step": 14893 + }, + { + "epoch": 0.54, + "learning_rate": 1.4564031200791004e-07, + "loss": 0.6846, + "step": 14894 + }, + { + "epoch": 0.54, + "learning_rate": 1.45636649943238e-07, + "loss": 0.7015, + "step": 14895 + }, + { + "epoch": 0.55, + "learning_rate": 1.4563298787856592e-07, + "loss": 0.6976, + "step": 14896 + }, + { + "epoch": 0.55, + "learning_rate": 1.4562932581389387e-07, + "loss": 0.6862, + "step": 14897 + }, + { + "epoch": 0.55, + "learning_rate": 1.456256637492218e-07, + "loss": 0.6953, + "step": 14898 + }, + { + "epoch": 0.55, + "learning_rate": 1.4562200168454975e-07, + "loss": 0.6865, + "step": 14899 + }, + { + "epoch": 0.55, + "learning_rate": 1.4561833961987768e-07, + "loss": 0.6939, + "step": 14900 + }, + { + "epoch": 0.55, + "learning_rate": 1.4561467755520563e-07, + "loss": 0.6831, + "step": 14901 + }, + { + "epoch": 0.55, + "learning_rate": 1.4561101549053356e-07, + "loss": 0.6928, + "step": 14902 + }, + { + "epoch": 0.55, + "learning_rate": 1.456073534258615e-07, + "loss": 0.685, + "step": 14903 + }, + { + "epoch": 0.55, + "learning_rate": 1.4560369136118944e-07, + "loss": 0.6967, + "step": 14904 + }, + { + "epoch": 0.55, + "learning_rate": 1.456000292965174e-07, + "loss": 0.6944, + "step": 14905 + }, + { + "epoch": 0.55, + "learning_rate": 1.4559636723184532e-07, + "loss": 0.6823, + "step": 14906 + }, + { + "epoch": 0.55, + "learning_rate": 1.4559270516717324e-07, + "loss": 0.6828, + "step": 14907 + }, + { + "epoch": 0.55, + "learning_rate": 1.4558904310250117e-07, + "loss": 0.7008, + "step": 14908 + }, + { + "epoch": 0.55, + "learning_rate": 1.4558538103782912e-07, + "loss": 0.6851, + "step": 14909 + }, + { + "epoch": 0.55, + "learning_rate": 1.4558171897315705e-07, + "loss": 0.6888, + "step": 14910 + }, + { + "epoch": 0.55, + "learning_rate": 1.45578056908485e-07, + "loss": 0.6916, + "step": 14911 + }, + { + "epoch": 0.55, + "learning_rate": 1.4557439484381293e-07, + "loss": 0.6964, + "step": 14912 + }, + { + "epoch": 0.55, + "learning_rate": 1.4557073277914088e-07, + "loss": 0.6923, + "step": 14913 + }, + { + "epoch": 0.55, + "learning_rate": 1.455670707144688e-07, + "loss": 0.7122, + "step": 14914 + }, + { + "epoch": 0.55, + "learning_rate": 1.4556340864979674e-07, + "loss": 0.6912, + "step": 14915 + }, + { + "epoch": 0.55, + "learning_rate": 1.455597465851247e-07, + "loss": 0.6961, + "step": 14916 + }, + { + "epoch": 0.55, + "learning_rate": 1.4555608452045262e-07, + "loss": 0.6943, + "step": 14917 + }, + { + "epoch": 0.55, + "learning_rate": 1.4555242245578057e-07, + "loss": 0.6977, + "step": 14918 + }, + { + "epoch": 0.55, + "learning_rate": 1.455487603911085e-07, + "loss": 0.6793, + "step": 14919 + }, + { + "epoch": 0.55, + "learning_rate": 1.4554509832643645e-07, + "loss": 0.6903, + "step": 14920 + }, + { + "epoch": 0.55, + "learning_rate": 1.4554143626176437e-07, + "loss": 0.6882, + "step": 14921 + }, + { + "epoch": 0.55, + "learning_rate": 1.4553777419709233e-07, + "loss": 0.6752, + "step": 14922 + }, + { + "epoch": 0.55, + "learning_rate": 1.4553411213242025e-07, + "loss": 0.6748, + "step": 14923 + }, + { + "epoch": 0.55, + "learning_rate": 1.455304500677482e-07, + "loss": 0.697, + "step": 14924 + }, + { + "epoch": 0.55, + "learning_rate": 1.455267880030761e-07, + "loss": 0.6901, + "step": 14925 + }, + { + "epoch": 0.55, + "learning_rate": 1.4552312593840406e-07, + "loss": 0.7239, + "step": 14926 + }, + { + "epoch": 0.55, + "learning_rate": 1.45519463873732e-07, + "loss": 0.7038, + "step": 14927 + }, + { + "epoch": 0.55, + "learning_rate": 1.4551580180905994e-07, + "loss": 0.6962, + "step": 14928 + }, + { + "epoch": 0.55, + "learning_rate": 1.4551213974438787e-07, + "loss": 0.6863, + "step": 14929 + }, + { + "epoch": 0.55, + "learning_rate": 1.4550847767971582e-07, + "loss": 0.6976, + "step": 14930 + }, + { + "epoch": 0.55, + "learning_rate": 1.4550481561504375e-07, + "loss": 0.6934, + "step": 14931 + }, + { + "epoch": 0.55, + "learning_rate": 1.455011535503717e-07, + "loss": 0.6846, + "step": 14932 + }, + { + "epoch": 0.55, + "learning_rate": 1.4549749148569963e-07, + "loss": 0.688, + "step": 14933 + }, + { + "epoch": 0.55, + "learning_rate": 1.4549382942102758e-07, + "loss": 0.7122, + "step": 14934 + }, + { + "epoch": 0.55, + "learning_rate": 1.454901673563555e-07, + "loss": 0.6897, + "step": 14935 + }, + { + "epoch": 0.55, + "learning_rate": 1.4548650529168346e-07, + "loss": 0.7192, + "step": 14936 + }, + { + "epoch": 0.55, + "learning_rate": 1.454828432270114e-07, + "loss": 0.7096, + "step": 14937 + }, + { + "epoch": 0.55, + "learning_rate": 1.4547918116233931e-07, + "loss": 0.6951, + "step": 14938 + }, + { + "epoch": 0.55, + "learning_rate": 1.4547551909766727e-07, + "loss": 0.6938, + "step": 14939 + }, + { + "epoch": 0.55, + "learning_rate": 1.454718570329952e-07, + "loss": 0.6782, + "step": 14940 + }, + { + "epoch": 0.55, + "learning_rate": 1.4546819496832315e-07, + "loss": 0.6899, + "step": 14941 + }, + { + "epoch": 0.55, + "learning_rate": 1.4546453290365107e-07, + "loss": 0.704, + "step": 14942 + }, + { + "epoch": 0.55, + "learning_rate": 1.45460870838979e-07, + "loss": 0.693, + "step": 14943 + }, + { + "epoch": 0.55, + "learning_rate": 1.4545720877430693e-07, + "loss": 0.7072, + "step": 14944 + }, + { + "epoch": 0.55, + "learning_rate": 1.4545354670963488e-07, + "loss": 0.687, + "step": 14945 + }, + { + "epoch": 0.55, + "learning_rate": 1.454498846449628e-07, + "loss": 0.6796, + "step": 14946 + }, + { + "epoch": 0.55, + "learning_rate": 1.4544622258029076e-07, + "loss": 0.6936, + "step": 14947 + }, + { + "epoch": 0.55, + "learning_rate": 1.4544256051561869e-07, + "loss": 0.6908, + "step": 14948 + }, + { + "epoch": 0.55, + "learning_rate": 1.4543889845094664e-07, + "loss": 0.6834, + "step": 14949 + }, + { + "epoch": 0.55, + "learning_rate": 1.4543523638627457e-07, + "loss": 0.6741, + "step": 14950 + }, + { + "epoch": 0.55, + "learning_rate": 1.4543157432160252e-07, + "loss": 0.7062, + "step": 14951 + }, + { + "epoch": 0.55, + "learning_rate": 1.4542791225693045e-07, + "loss": 0.6852, + "step": 14952 + }, + { + "epoch": 0.55, + "learning_rate": 1.454242501922584e-07, + "loss": 0.7057, + "step": 14953 + }, + { + "epoch": 0.55, + "learning_rate": 1.4542058812758633e-07, + "loss": 0.6727, + "step": 14954 + }, + { + "epoch": 0.55, + "learning_rate": 1.4541692606291428e-07, + "loss": 0.6949, + "step": 14955 + }, + { + "epoch": 0.55, + "learning_rate": 1.454132639982422e-07, + "loss": 0.6953, + "step": 14956 + }, + { + "epoch": 0.55, + "learning_rate": 1.4540960193357016e-07, + "loss": 0.6948, + "step": 14957 + }, + { + "epoch": 0.55, + "learning_rate": 1.4540593986889809e-07, + "loss": 0.6977, + "step": 14958 + }, + { + "epoch": 0.55, + "learning_rate": 1.45402277804226e-07, + "loss": 0.6946, + "step": 14959 + }, + { + "epoch": 0.55, + "learning_rate": 1.4539861573955397e-07, + "loss": 0.688, + "step": 14960 + }, + { + "epoch": 0.55, + "learning_rate": 1.453949536748819e-07, + "loss": 0.6921, + "step": 14961 + }, + { + "epoch": 0.55, + "learning_rate": 1.4539129161020982e-07, + "loss": 0.7095, + "step": 14962 + }, + { + "epoch": 0.55, + "learning_rate": 1.4538762954553777e-07, + "loss": 0.6916, + "step": 14963 + }, + { + "epoch": 0.55, + "learning_rate": 1.453839674808657e-07, + "loss": 0.6789, + "step": 14964 + }, + { + "epoch": 0.55, + "learning_rate": 1.4538030541619363e-07, + "loss": 0.7007, + "step": 14965 + }, + { + "epoch": 0.55, + "learning_rate": 1.4537664335152158e-07, + "loss": 0.704, + "step": 14966 + }, + { + "epoch": 0.55, + "learning_rate": 1.453729812868495e-07, + "loss": 0.699, + "step": 14967 + }, + { + "epoch": 0.55, + "learning_rate": 1.4536931922217746e-07, + "loss": 0.6912, + "step": 14968 + }, + { + "epoch": 0.55, + "learning_rate": 1.4536565715750539e-07, + "loss": 0.7037, + "step": 14969 + }, + { + "epoch": 0.55, + "learning_rate": 1.4536199509283334e-07, + "loss": 0.6981, + "step": 14970 + }, + { + "epoch": 0.55, + "learning_rate": 1.4535833302816127e-07, + "loss": 0.7008, + "step": 14971 + }, + { + "epoch": 0.55, + "learning_rate": 1.4535467096348922e-07, + "loss": 0.7042, + "step": 14972 + }, + { + "epoch": 0.55, + "learning_rate": 1.4535100889881715e-07, + "loss": 0.699, + "step": 14973 + }, + { + "epoch": 0.55, + "learning_rate": 1.453473468341451e-07, + "loss": 0.7122, + "step": 14974 + }, + { + "epoch": 0.55, + "learning_rate": 1.4534368476947303e-07, + "loss": 0.7034, + "step": 14975 + }, + { + "epoch": 0.55, + "learning_rate": 1.4534002270480098e-07, + "loss": 0.6932, + "step": 14976 + }, + { + "epoch": 0.55, + "learning_rate": 1.453363606401289e-07, + "loss": 0.6886, + "step": 14977 + }, + { + "epoch": 0.55, + "learning_rate": 1.4533269857545686e-07, + "loss": 0.6907, + "step": 14978 + }, + { + "epoch": 0.55, + "learning_rate": 1.4532903651078476e-07, + "loss": 0.6937, + "step": 14979 + }, + { + "epoch": 0.55, + "learning_rate": 1.453253744461127e-07, + "loss": 0.692, + "step": 14980 + }, + { + "epoch": 0.55, + "learning_rate": 1.4532171238144064e-07, + "loss": 0.6895, + "step": 14981 + }, + { + "epoch": 0.55, + "learning_rate": 1.453180503167686e-07, + "loss": 0.7011, + "step": 14982 + }, + { + "epoch": 0.55, + "learning_rate": 1.4531438825209652e-07, + "loss": 0.7006, + "step": 14983 + }, + { + "epoch": 0.55, + "learning_rate": 1.4531072618742447e-07, + "loss": 0.7049, + "step": 14984 + }, + { + "epoch": 0.55, + "learning_rate": 1.453070641227524e-07, + "loss": 0.6919, + "step": 14985 + }, + { + "epoch": 0.55, + "learning_rate": 1.4530340205808032e-07, + "loss": 0.703, + "step": 14986 + }, + { + "epoch": 0.55, + "learning_rate": 1.4529973999340828e-07, + "loss": 0.6879, + "step": 14987 + }, + { + "epoch": 0.55, + "learning_rate": 1.452960779287362e-07, + "loss": 0.6912, + "step": 14988 + }, + { + "epoch": 0.55, + "learning_rate": 1.4529241586406416e-07, + "loss": 0.7027, + "step": 14989 + }, + { + "epoch": 0.55, + "learning_rate": 1.4528875379939208e-07, + "loss": 0.6975, + "step": 14990 + }, + { + "epoch": 0.55, + "learning_rate": 1.4528509173472004e-07, + "loss": 0.684, + "step": 14991 + }, + { + "epoch": 0.55, + "learning_rate": 1.4528142967004796e-07, + "loss": 0.7085, + "step": 14992 + }, + { + "epoch": 0.55, + "learning_rate": 1.4527776760537592e-07, + "loss": 0.6896, + "step": 14993 + }, + { + "epoch": 0.55, + "learning_rate": 1.4527410554070384e-07, + "loss": 0.6893, + "step": 14994 + }, + { + "epoch": 0.55, + "learning_rate": 1.452704434760318e-07, + "loss": 0.6915, + "step": 14995 + }, + { + "epoch": 0.55, + "learning_rate": 1.4526678141135972e-07, + "loss": 0.7191, + "step": 14996 + }, + { + "epoch": 0.55, + "learning_rate": 1.4526311934668765e-07, + "loss": 0.7007, + "step": 14997 + }, + { + "epoch": 0.55, + "learning_rate": 1.4525945728201558e-07, + "loss": 0.6937, + "step": 14998 + }, + { + "epoch": 0.55, + "learning_rate": 1.4525579521734353e-07, + "loss": 0.6776, + "step": 14999 + }, + { + "epoch": 0.55, + "learning_rate": 1.4525213315267146e-07, + "loss": 0.7077, + "step": 15000 + }, + { + "epoch": 0.55, + "learning_rate": 1.452484710879994e-07, + "loss": 0.6789, + "step": 15001 + }, + { + "epoch": 0.55, + "learning_rate": 1.4524480902332734e-07, + "loss": 0.6877, + "step": 15002 + }, + { + "epoch": 0.55, + "learning_rate": 1.452411469586553e-07, + "loss": 0.6992, + "step": 15003 + }, + { + "epoch": 0.55, + "learning_rate": 1.4523748489398322e-07, + "loss": 0.6981, + "step": 15004 + }, + { + "epoch": 0.55, + "learning_rate": 1.4523382282931117e-07, + "loss": 0.6806, + "step": 15005 + }, + { + "epoch": 0.55, + "learning_rate": 1.452301607646391e-07, + "loss": 0.6657, + "step": 15006 + }, + { + "epoch": 0.55, + "learning_rate": 1.4522649869996705e-07, + "loss": 0.7043, + "step": 15007 + }, + { + "epoch": 0.55, + "learning_rate": 1.4522283663529498e-07, + "loss": 0.6819, + "step": 15008 + }, + { + "epoch": 0.55, + "learning_rate": 1.452191745706229e-07, + "loss": 0.69, + "step": 15009 + }, + { + "epoch": 0.55, + "learning_rate": 1.4521551250595086e-07, + "loss": 0.6943, + "step": 15010 + }, + { + "epoch": 0.55, + "learning_rate": 1.4521185044127878e-07, + "loss": 0.6938, + "step": 15011 + }, + { + "epoch": 0.55, + "learning_rate": 1.4520818837660674e-07, + "loss": 0.6892, + "step": 15012 + }, + { + "epoch": 0.55, + "learning_rate": 1.4520452631193466e-07, + "loss": 0.695, + "step": 15013 + }, + { + "epoch": 0.55, + "learning_rate": 1.4520086424726262e-07, + "loss": 0.6978, + "step": 15014 + }, + { + "epoch": 0.55, + "learning_rate": 1.4519720218259052e-07, + "loss": 0.7098, + "step": 15015 + }, + { + "epoch": 0.55, + "learning_rate": 1.4519354011791847e-07, + "loss": 0.6953, + "step": 15016 + }, + { + "epoch": 0.55, + "learning_rate": 1.451898780532464e-07, + "loss": 0.6933, + "step": 15017 + }, + { + "epoch": 0.55, + "learning_rate": 1.4518621598857435e-07, + "loss": 0.6998, + "step": 15018 + }, + { + "epoch": 0.55, + "learning_rate": 1.4518255392390228e-07, + "loss": 0.7156, + "step": 15019 + }, + { + "epoch": 0.55, + "learning_rate": 1.4517889185923023e-07, + "loss": 0.6931, + "step": 15020 + }, + { + "epoch": 0.55, + "learning_rate": 1.4517522979455816e-07, + "loss": 0.6909, + "step": 15021 + }, + { + "epoch": 0.55, + "learning_rate": 1.451715677298861e-07, + "loss": 0.6868, + "step": 15022 + }, + { + "epoch": 0.55, + "learning_rate": 1.4516790566521404e-07, + "loss": 0.6979, + "step": 15023 + }, + { + "epoch": 0.55, + "learning_rate": 1.45164243600542e-07, + "loss": 0.6859, + "step": 15024 + }, + { + "epoch": 0.55, + "learning_rate": 1.4516058153586992e-07, + "loss": 0.7005, + "step": 15025 + }, + { + "epoch": 0.55, + "learning_rate": 1.4515691947119787e-07, + "loss": 0.7026, + "step": 15026 + }, + { + "epoch": 0.55, + "learning_rate": 1.451532574065258e-07, + "loss": 0.6894, + "step": 15027 + }, + { + "epoch": 0.55, + "learning_rate": 1.4514959534185375e-07, + "loss": 0.6963, + "step": 15028 + }, + { + "epoch": 0.55, + "learning_rate": 1.4514593327718168e-07, + "loss": 0.6892, + "step": 15029 + }, + { + "epoch": 0.55, + "learning_rate": 1.451422712125096e-07, + "loss": 0.6687, + "step": 15030 + }, + { + "epoch": 0.55, + "learning_rate": 1.4513860914783756e-07, + "loss": 0.6908, + "step": 15031 + }, + { + "epoch": 0.55, + "learning_rate": 1.4513494708316548e-07, + "loss": 0.6861, + "step": 15032 + }, + { + "epoch": 0.55, + "learning_rate": 1.451312850184934e-07, + "loss": 0.6893, + "step": 15033 + }, + { + "epoch": 0.55, + "learning_rate": 1.4512762295382136e-07, + "loss": 0.6958, + "step": 15034 + }, + { + "epoch": 0.55, + "learning_rate": 1.451239608891493e-07, + "loss": 0.7069, + "step": 15035 + }, + { + "epoch": 0.55, + "learning_rate": 1.4512029882447722e-07, + "loss": 0.688, + "step": 15036 + }, + { + "epoch": 0.55, + "learning_rate": 1.4511663675980517e-07, + "loss": 0.6936, + "step": 15037 + }, + { + "epoch": 0.55, + "learning_rate": 1.451129746951331e-07, + "loss": 0.6796, + "step": 15038 + }, + { + "epoch": 0.55, + "learning_rate": 1.4510931263046105e-07, + "loss": 0.6706, + "step": 15039 + }, + { + "epoch": 0.55, + "learning_rate": 1.4510565056578897e-07, + "loss": 0.6942, + "step": 15040 + }, + { + "epoch": 0.55, + "learning_rate": 1.4510198850111693e-07, + "loss": 0.696, + "step": 15041 + }, + { + "epoch": 0.55, + "learning_rate": 1.4509832643644485e-07, + "loss": 0.7028, + "step": 15042 + }, + { + "epoch": 0.55, + "learning_rate": 1.450946643717728e-07, + "loss": 0.6906, + "step": 15043 + }, + { + "epoch": 0.55, + "learning_rate": 1.4509100230710073e-07, + "loss": 0.7014, + "step": 15044 + }, + { + "epoch": 0.55, + "learning_rate": 1.450873402424287e-07, + "loss": 0.6997, + "step": 15045 + }, + { + "epoch": 0.55, + "learning_rate": 1.4508367817775661e-07, + "loss": 0.6887, + "step": 15046 + }, + { + "epoch": 0.55, + "learning_rate": 1.4508001611308457e-07, + "loss": 0.6961, + "step": 15047 + }, + { + "epoch": 0.55, + "learning_rate": 1.450763540484125e-07, + "loss": 0.6996, + "step": 15048 + }, + { + "epoch": 0.55, + "learning_rate": 1.4507269198374045e-07, + "loss": 0.7076, + "step": 15049 + }, + { + "epoch": 0.55, + "learning_rate": 1.4506902991906835e-07, + "loss": 0.7108, + "step": 15050 + }, + { + "epoch": 0.55, + "learning_rate": 1.450653678543963e-07, + "loss": 0.6883, + "step": 15051 + }, + { + "epoch": 0.55, + "learning_rate": 1.4506170578972423e-07, + "loss": 0.7012, + "step": 15052 + }, + { + "epoch": 0.55, + "learning_rate": 1.4505804372505218e-07, + "loss": 0.6913, + "step": 15053 + }, + { + "epoch": 0.55, + "learning_rate": 1.450543816603801e-07, + "loss": 0.6922, + "step": 15054 + }, + { + "epoch": 0.55, + "learning_rate": 1.4505071959570806e-07, + "loss": 0.6888, + "step": 15055 + }, + { + "epoch": 0.55, + "learning_rate": 1.45047057531036e-07, + "loss": 0.6831, + "step": 15056 + }, + { + "epoch": 0.55, + "learning_rate": 1.4504339546636391e-07, + "loss": 0.6912, + "step": 15057 + }, + { + "epoch": 0.55, + "learning_rate": 1.4503973340169187e-07, + "loss": 0.6952, + "step": 15058 + }, + { + "epoch": 0.55, + "learning_rate": 1.450360713370198e-07, + "loss": 0.6997, + "step": 15059 + }, + { + "epoch": 0.55, + "learning_rate": 1.4503240927234775e-07, + "loss": 0.6971, + "step": 15060 + }, + { + "epoch": 0.55, + "learning_rate": 1.4502874720767567e-07, + "loss": 0.6799, + "step": 15061 + }, + { + "epoch": 0.55, + "learning_rate": 1.4502508514300363e-07, + "loss": 0.696, + "step": 15062 + }, + { + "epoch": 0.55, + "learning_rate": 1.4502142307833155e-07, + "loss": 0.6931, + "step": 15063 + }, + { + "epoch": 0.55, + "learning_rate": 1.450177610136595e-07, + "loss": 0.6905, + "step": 15064 + }, + { + "epoch": 0.55, + "learning_rate": 1.4501409894898743e-07, + "loss": 0.6914, + "step": 15065 + }, + { + "epoch": 0.55, + "learning_rate": 1.4501043688431539e-07, + "loss": 0.6929, + "step": 15066 + }, + { + "epoch": 0.55, + "learning_rate": 1.450067748196433e-07, + "loss": 0.689, + "step": 15067 + }, + { + "epoch": 0.55, + "learning_rate": 1.4500311275497124e-07, + "loss": 0.7053, + "step": 15068 + }, + { + "epoch": 0.55, + "learning_rate": 1.4499945069029917e-07, + "loss": 0.7036, + "step": 15069 + }, + { + "epoch": 0.55, + "learning_rate": 1.4499578862562712e-07, + "loss": 0.686, + "step": 15070 + }, + { + "epoch": 0.55, + "learning_rate": 1.4499212656095505e-07, + "loss": 0.6954, + "step": 15071 + }, + { + "epoch": 0.55, + "learning_rate": 1.44988464496283e-07, + "loss": 0.6879, + "step": 15072 + }, + { + "epoch": 0.55, + "learning_rate": 1.4498480243161093e-07, + "loss": 0.6908, + "step": 15073 + }, + { + "epoch": 0.55, + "learning_rate": 1.4498114036693888e-07, + "loss": 0.6963, + "step": 15074 + }, + { + "epoch": 0.55, + "learning_rate": 1.449774783022668e-07, + "loss": 0.6915, + "step": 15075 + }, + { + "epoch": 0.55, + "learning_rate": 1.4497381623759476e-07, + "loss": 0.6916, + "step": 15076 + }, + { + "epoch": 0.55, + "learning_rate": 1.4497015417292269e-07, + "loss": 0.6914, + "step": 15077 + }, + { + "epoch": 0.55, + "learning_rate": 1.4496649210825064e-07, + "loss": 0.7003, + "step": 15078 + }, + { + "epoch": 0.55, + "learning_rate": 1.4496283004357857e-07, + "loss": 0.7047, + "step": 15079 + }, + { + "epoch": 0.55, + "learning_rate": 1.449591679789065e-07, + "loss": 0.6957, + "step": 15080 + }, + { + "epoch": 0.55, + "learning_rate": 1.4495550591423445e-07, + "loss": 0.705, + "step": 15081 + }, + { + "epoch": 0.55, + "learning_rate": 1.4495184384956237e-07, + "loss": 0.6837, + "step": 15082 + }, + { + "epoch": 0.55, + "learning_rate": 1.4494818178489033e-07, + "loss": 0.695, + "step": 15083 + }, + { + "epoch": 0.55, + "learning_rate": 1.4494451972021825e-07, + "loss": 0.7072, + "step": 15084 + }, + { + "epoch": 0.55, + "learning_rate": 1.449408576555462e-07, + "loss": 0.6899, + "step": 15085 + }, + { + "epoch": 0.55, + "learning_rate": 1.449371955908741e-07, + "loss": 0.6962, + "step": 15086 + }, + { + "epoch": 0.55, + "learning_rate": 1.4493353352620206e-07, + "loss": 0.6862, + "step": 15087 + }, + { + "epoch": 0.55, + "learning_rate": 1.4492987146152999e-07, + "loss": 0.6937, + "step": 15088 + }, + { + "epoch": 0.55, + "learning_rate": 1.4492620939685794e-07, + "loss": 0.6943, + "step": 15089 + }, + { + "epoch": 0.55, + "learning_rate": 1.4492254733218587e-07, + "loss": 0.71, + "step": 15090 + }, + { + "epoch": 0.55, + "learning_rate": 1.4491888526751382e-07, + "loss": 0.6896, + "step": 15091 + }, + { + "epoch": 0.55, + "learning_rate": 1.4491522320284175e-07, + "loss": 0.6904, + "step": 15092 + }, + { + "epoch": 0.55, + "learning_rate": 1.449115611381697e-07, + "loss": 0.7027, + "step": 15093 + }, + { + "epoch": 0.55, + "learning_rate": 1.4490789907349763e-07, + "loss": 0.6909, + "step": 15094 + }, + { + "epoch": 0.55, + "learning_rate": 1.4490423700882558e-07, + "loss": 0.7117, + "step": 15095 + }, + { + "epoch": 0.55, + "learning_rate": 1.449005749441535e-07, + "loss": 0.6751, + "step": 15096 + }, + { + "epoch": 0.55, + "learning_rate": 1.4489691287948146e-07, + "loss": 0.6837, + "step": 15097 + }, + { + "epoch": 0.55, + "learning_rate": 1.4489325081480938e-07, + "loss": 0.6939, + "step": 15098 + }, + { + "epoch": 0.55, + "learning_rate": 1.4488958875013734e-07, + "loss": 0.7037, + "step": 15099 + }, + { + "epoch": 0.55, + "learning_rate": 1.4488592668546526e-07, + "loss": 0.6941, + "step": 15100 + }, + { + "epoch": 0.55, + "learning_rate": 1.448822646207932e-07, + "loss": 0.6983, + "step": 15101 + }, + { + "epoch": 0.55, + "learning_rate": 1.4487860255612114e-07, + "loss": 0.7075, + "step": 15102 + }, + { + "epoch": 0.55, + "learning_rate": 1.4487494049144907e-07, + "loss": 0.6897, + "step": 15103 + }, + { + "epoch": 0.55, + "learning_rate": 1.44871278426777e-07, + "loss": 0.6875, + "step": 15104 + }, + { + "epoch": 0.55, + "learning_rate": 1.4486761636210495e-07, + "loss": 0.71, + "step": 15105 + }, + { + "epoch": 0.55, + "learning_rate": 1.4486395429743288e-07, + "loss": 0.6828, + "step": 15106 + }, + { + "epoch": 0.55, + "learning_rate": 1.448602922327608e-07, + "loss": 0.6853, + "step": 15107 + }, + { + "epoch": 0.55, + "learning_rate": 1.4485663016808876e-07, + "loss": 0.6867, + "step": 15108 + }, + { + "epoch": 0.55, + "learning_rate": 1.4485296810341668e-07, + "loss": 0.6939, + "step": 15109 + }, + { + "epoch": 0.55, + "learning_rate": 1.4484930603874464e-07, + "loss": 0.6905, + "step": 15110 + }, + { + "epoch": 0.55, + "learning_rate": 1.4484564397407256e-07, + "loss": 0.712, + "step": 15111 + }, + { + "epoch": 0.55, + "learning_rate": 1.4484198190940052e-07, + "loss": 0.69, + "step": 15112 + }, + { + "epoch": 0.55, + "learning_rate": 1.4483831984472844e-07, + "loss": 0.7037, + "step": 15113 + }, + { + "epoch": 0.55, + "learning_rate": 1.448346577800564e-07, + "loss": 0.6954, + "step": 15114 + }, + { + "epoch": 0.55, + "learning_rate": 1.4483099571538432e-07, + "loss": 0.6981, + "step": 15115 + }, + { + "epoch": 0.55, + "learning_rate": 1.4482733365071228e-07, + "loss": 0.6882, + "step": 15116 + }, + { + "epoch": 0.55, + "learning_rate": 1.448236715860402e-07, + "loss": 0.688, + "step": 15117 + }, + { + "epoch": 0.55, + "learning_rate": 1.4482000952136816e-07, + "loss": 0.7026, + "step": 15118 + }, + { + "epoch": 0.55, + "learning_rate": 1.4481634745669608e-07, + "loss": 0.6877, + "step": 15119 + }, + { + "epoch": 0.55, + "learning_rate": 1.4481268539202404e-07, + "loss": 0.6966, + "step": 15120 + }, + { + "epoch": 0.55, + "learning_rate": 1.4480902332735196e-07, + "loss": 0.6931, + "step": 15121 + }, + { + "epoch": 0.55, + "learning_rate": 1.448053612626799e-07, + "loss": 0.7008, + "step": 15122 + }, + { + "epoch": 0.55, + "learning_rate": 1.4480169919800782e-07, + "loss": 0.6741, + "step": 15123 + }, + { + "epoch": 0.55, + "learning_rate": 1.4479803713333577e-07, + "loss": 0.6702, + "step": 15124 + }, + { + "epoch": 0.55, + "learning_rate": 1.447943750686637e-07, + "loss": 0.6986, + "step": 15125 + }, + { + "epoch": 0.55, + "learning_rate": 1.4479071300399165e-07, + "loss": 0.6854, + "step": 15126 + }, + { + "epoch": 0.55, + "learning_rate": 1.4478705093931958e-07, + "loss": 0.7063, + "step": 15127 + }, + { + "epoch": 0.55, + "learning_rate": 1.447833888746475e-07, + "loss": 0.7012, + "step": 15128 + }, + { + "epoch": 0.55, + "learning_rate": 1.4477972680997546e-07, + "loss": 0.6983, + "step": 15129 + }, + { + "epoch": 0.55, + "learning_rate": 1.4477606474530338e-07, + "loss": 0.6814, + "step": 15130 + }, + { + "epoch": 0.55, + "learning_rate": 1.4477240268063134e-07, + "loss": 0.701, + "step": 15131 + }, + { + "epoch": 0.55, + "learning_rate": 1.4476874061595926e-07, + "loss": 0.6955, + "step": 15132 + }, + { + "epoch": 0.55, + "learning_rate": 1.4476507855128722e-07, + "loss": 0.6853, + "step": 15133 + }, + { + "epoch": 0.55, + "learning_rate": 1.4476141648661514e-07, + "loss": 0.6953, + "step": 15134 + }, + { + "epoch": 0.55, + "learning_rate": 1.447577544219431e-07, + "loss": 0.6895, + "step": 15135 + }, + { + "epoch": 0.55, + "learning_rate": 1.4475409235727102e-07, + "loss": 0.6998, + "step": 15136 + }, + { + "epoch": 0.55, + "learning_rate": 1.4475043029259898e-07, + "loss": 0.6842, + "step": 15137 + }, + { + "epoch": 0.55, + "learning_rate": 1.447467682279269e-07, + "loss": 0.6766, + "step": 15138 + }, + { + "epoch": 0.55, + "learning_rate": 1.4474310616325486e-07, + "loss": 0.7019, + "step": 15139 + }, + { + "epoch": 0.55, + "learning_rate": 1.4473944409858276e-07, + "loss": 0.6919, + "step": 15140 + }, + { + "epoch": 0.55, + "learning_rate": 1.447357820339107e-07, + "loss": 0.6904, + "step": 15141 + }, + { + "epoch": 0.55, + "learning_rate": 1.4473211996923864e-07, + "loss": 0.6937, + "step": 15142 + }, + { + "epoch": 0.55, + "learning_rate": 1.447284579045666e-07, + "loss": 0.7041, + "step": 15143 + }, + { + "epoch": 0.55, + "learning_rate": 1.4472479583989452e-07, + "loss": 0.7032, + "step": 15144 + }, + { + "epoch": 0.55, + "learning_rate": 1.4472113377522247e-07, + "loss": 0.7062, + "step": 15145 + }, + { + "epoch": 0.55, + "learning_rate": 1.447174717105504e-07, + "loss": 0.6947, + "step": 15146 + }, + { + "epoch": 0.55, + "learning_rate": 1.4471380964587835e-07, + "loss": 0.6969, + "step": 15147 + }, + { + "epoch": 0.55, + "learning_rate": 1.4471014758120628e-07, + "loss": 0.6982, + "step": 15148 + }, + { + "epoch": 0.55, + "learning_rate": 1.4470648551653423e-07, + "loss": 0.6943, + "step": 15149 + }, + { + "epoch": 0.55, + "learning_rate": 1.4470282345186216e-07, + "loss": 0.6942, + "step": 15150 + }, + { + "epoch": 0.55, + "learning_rate": 1.4469916138719008e-07, + "loss": 0.6943, + "step": 15151 + }, + { + "epoch": 0.55, + "learning_rate": 1.4469549932251804e-07, + "loss": 0.6902, + "step": 15152 + }, + { + "epoch": 0.55, + "learning_rate": 1.4469183725784596e-07, + "loss": 0.6928, + "step": 15153 + }, + { + "epoch": 0.55, + "learning_rate": 1.4468817519317391e-07, + "loss": 0.696, + "step": 15154 + }, + { + "epoch": 0.55, + "learning_rate": 1.4468451312850184e-07, + "loss": 0.7031, + "step": 15155 + }, + { + "epoch": 0.55, + "learning_rate": 1.446808510638298e-07, + "loss": 0.6968, + "step": 15156 + }, + { + "epoch": 0.55, + "learning_rate": 1.4467718899915772e-07, + "loss": 0.6819, + "step": 15157 + }, + { + "epoch": 0.55, + "learning_rate": 1.4467352693448565e-07, + "loss": 0.6871, + "step": 15158 + }, + { + "epoch": 0.55, + "learning_rate": 1.4466986486981357e-07, + "loss": 0.6845, + "step": 15159 + }, + { + "epoch": 0.55, + "learning_rate": 1.4466620280514153e-07, + "loss": 0.7024, + "step": 15160 + }, + { + "epoch": 0.55, + "learning_rate": 1.4466254074046945e-07, + "loss": 0.6891, + "step": 15161 + }, + { + "epoch": 0.55, + "learning_rate": 1.446588786757974e-07, + "loss": 0.6851, + "step": 15162 + }, + { + "epoch": 0.55, + "learning_rate": 1.4465521661112533e-07, + "loss": 0.6807, + "step": 15163 + }, + { + "epoch": 0.55, + "learning_rate": 1.446515545464533e-07, + "loss": 0.6989, + "step": 15164 + }, + { + "epoch": 0.55, + "learning_rate": 1.4464789248178121e-07, + "loss": 0.7042, + "step": 15165 + }, + { + "epoch": 0.55, + "learning_rate": 1.4464423041710917e-07, + "loss": 0.6809, + "step": 15166 + }, + { + "epoch": 0.55, + "learning_rate": 1.446405683524371e-07, + "loss": 0.6903, + "step": 15167 + }, + { + "epoch": 0.55, + "learning_rate": 1.4463690628776505e-07, + "loss": 0.6876, + "step": 15168 + }, + { + "epoch": 0.55, + "learning_rate": 1.4463324422309297e-07, + "loss": 0.7009, + "step": 15169 + }, + { + "epoch": 0.56, + "learning_rate": 1.4462958215842093e-07, + "loss": 0.6874, + "step": 15170 + }, + { + "epoch": 0.56, + "learning_rate": 1.4462592009374885e-07, + "loss": 0.6912, + "step": 15171 + }, + { + "epoch": 0.56, + "learning_rate": 1.4462225802907678e-07, + "loss": 0.6904, + "step": 15172 + }, + { + "epoch": 0.56, + "learning_rate": 1.4461859596440473e-07, + "loss": 0.6984, + "step": 15173 + }, + { + "epoch": 0.56, + "learning_rate": 1.4461493389973266e-07, + "loss": 0.6858, + "step": 15174 + }, + { + "epoch": 0.56, + "learning_rate": 1.4461127183506061e-07, + "loss": 0.6847, + "step": 15175 + }, + { + "epoch": 0.56, + "learning_rate": 1.4460760977038854e-07, + "loss": 0.6962, + "step": 15176 + }, + { + "epoch": 0.56, + "learning_rate": 1.4460394770571647e-07, + "loss": 0.7006, + "step": 15177 + }, + { + "epoch": 0.56, + "learning_rate": 1.446002856410444e-07, + "loss": 0.6917, + "step": 15178 + }, + { + "epoch": 0.56, + "learning_rate": 1.4459662357637235e-07, + "loss": 0.6836, + "step": 15179 + }, + { + "epoch": 0.56, + "learning_rate": 1.4459296151170027e-07, + "loss": 0.6836, + "step": 15180 + }, + { + "epoch": 0.56, + "learning_rate": 1.4458929944702823e-07, + "loss": 0.6826, + "step": 15181 + }, + { + "epoch": 0.56, + "learning_rate": 1.4458563738235615e-07, + "loss": 0.701, + "step": 15182 + }, + { + "epoch": 0.56, + "learning_rate": 1.445819753176841e-07, + "loss": 0.7044, + "step": 15183 + }, + { + "epoch": 0.56, + "learning_rate": 1.4457831325301203e-07, + "loss": 0.6985, + "step": 15184 + }, + { + "epoch": 0.56, + "learning_rate": 1.4457465118833999e-07, + "loss": 0.6932, + "step": 15185 + }, + { + "epoch": 0.56, + "learning_rate": 1.445709891236679e-07, + "loss": 0.6857, + "step": 15186 + }, + { + "epoch": 0.56, + "learning_rate": 1.4456732705899587e-07, + "loss": 0.6801, + "step": 15187 + }, + { + "epoch": 0.56, + "learning_rate": 1.445636649943238e-07, + "loss": 0.7043, + "step": 15188 + }, + { + "epoch": 0.56, + "learning_rate": 1.4456000292965175e-07, + "loss": 0.6938, + "step": 15189 + }, + { + "epoch": 0.56, + "learning_rate": 1.4455634086497967e-07, + "loss": 0.6845, + "step": 15190 + }, + { + "epoch": 0.56, + "learning_rate": 1.4455267880030763e-07, + "loss": 0.6888, + "step": 15191 + }, + { + "epoch": 0.56, + "learning_rate": 1.4454901673563555e-07, + "loss": 0.6874, + "step": 15192 + }, + { + "epoch": 0.56, + "learning_rate": 1.445453546709635e-07, + "loss": 0.6906, + "step": 15193 + }, + { + "epoch": 0.56, + "learning_rate": 1.445416926062914e-07, + "loss": 0.7003, + "step": 15194 + }, + { + "epoch": 0.56, + "learning_rate": 1.4453803054161936e-07, + "loss": 0.6868, + "step": 15195 + }, + { + "epoch": 0.56, + "learning_rate": 1.4453436847694729e-07, + "loss": 0.6963, + "step": 15196 + }, + { + "epoch": 0.56, + "learning_rate": 1.4453070641227524e-07, + "loss": 0.7082, + "step": 15197 + }, + { + "epoch": 0.56, + "learning_rate": 1.4452704434760317e-07, + "loss": 0.6914, + "step": 15198 + }, + { + "epoch": 0.56, + "learning_rate": 1.445233822829311e-07, + "loss": 0.7002, + "step": 15199 + }, + { + "epoch": 0.56, + "learning_rate": 1.4451972021825905e-07, + "loss": 0.6789, + "step": 15200 + }, + { + "epoch": 0.56, + "learning_rate": 1.4451605815358697e-07, + "loss": 0.6955, + "step": 15201 + }, + { + "epoch": 0.56, + "learning_rate": 1.4451239608891493e-07, + "loss": 0.6965, + "step": 15202 + }, + { + "epoch": 0.56, + "learning_rate": 1.4450873402424285e-07, + "loss": 0.6973, + "step": 15203 + }, + { + "epoch": 0.56, + "learning_rate": 1.445050719595708e-07, + "loss": 0.7052, + "step": 15204 + }, + { + "epoch": 0.56, + "learning_rate": 1.4450140989489873e-07, + "loss": 0.7068, + "step": 15205 + }, + { + "epoch": 0.56, + "learning_rate": 1.4449774783022669e-07, + "loss": 0.6989, + "step": 15206 + }, + { + "epoch": 0.56, + "learning_rate": 1.444940857655546e-07, + "loss": 0.6965, + "step": 15207 + }, + { + "epoch": 0.56, + "learning_rate": 1.4449042370088257e-07, + "loss": 0.6884, + "step": 15208 + }, + { + "epoch": 0.56, + "learning_rate": 1.444867616362105e-07, + "loss": 0.6995, + "step": 15209 + }, + { + "epoch": 0.56, + "learning_rate": 1.4448309957153845e-07, + "loss": 0.6863, + "step": 15210 + }, + { + "epoch": 0.56, + "learning_rate": 1.4447943750686637e-07, + "loss": 0.6932, + "step": 15211 + }, + { + "epoch": 0.56, + "learning_rate": 1.444757754421943e-07, + "loss": 0.6892, + "step": 15212 + }, + { + "epoch": 0.56, + "learning_rate": 1.4447211337752223e-07, + "loss": 0.7054, + "step": 15213 + }, + { + "epoch": 0.56, + "learning_rate": 1.4446845131285018e-07, + "loss": 0.671, + "step": 15214 + }, + { + "epoch": 0.56, + "learning_rate": 1.444647892481781e-07, + "loss": 0.7106, + "step": 15215 + }, + { + "epoch": 0.56, + "learning_rate": 1.4446112718350606e-07, + "loss": 0.6879, + "step": 15216 + }, + { + "epoch": 0.56, + "learning_rate": 1.4445746511883398e-07, + "loss": 0.7057, + "step": 15217 + }, + { + "epoch": 0.56, + "learning_rate": 1.4445380305416194e-07, + "loss": 0.6783, + "step": 15218 + }, + { + "epoch": 0.56, + "learning_rate": 1.4445014098948986e-07, + "loss": 0.6853, + "step": 15219 + }, + { + "epoch": 0.56, + "learning_rate": 1.4444647892481782e-07, + "loss": 0.7009, + "step": 15220 + }, + { + "epoch": 0.56, + "learning_rate": 1.4444281686014574e-07, + "loss": 0.69, + "step": 15221 + }, + { + "epoch": 0.56, + "learning_rate": 1.4443915479547367e-07, + "loss": 0.6821, + "step": 15222 + }, + { + "epoch": 0.56, + "learning_rate": 1.4443549273080162e-07, + "loss": 0.6865, + "step": 15223 + }, + { + "epoch": 0.56, + "learning_rate": 1.4443183066612955e-07, + "loss": 0.6962, + "step": 15224 + }, + { + "epoch": 0.56, + "learning_rate": 1.444281686014575e-07, + "loss": 0.7001, + "step": 15225 + }, + { + "epoch": 0.56, + "learning_rate": 1.4442450653678543e-07, + "loss": 0.692, + "step": 15226 + }, + { + "epoch": 0.56, + "learning_rate": 1.4442084447211338e-07, + "loss": 0.6958, + "step": 15227 + }, + { + "epoch": 0.56, + "learning_rate": 1.444171824074413e-07, + "loss": 0.6944, + "step": 15228 + }, + { + "epoch": 0.56, + "learning_rate": 1.4441352034276926e-07, + "loss": 0.6884, + "step": 15229 + }, + { + "epoch": 0.56, + "learning_rate": 1.4440985827809716e-07, + "loss": 0.6853, + "step": 15230 + }, + { + "epoch": 0.56, + "learning_rate": 1.4440619621342512e-07, + "loss": 0.7014, + "step": 15231 + }, + { + "epoch": 0.56, + "learning_rate": 1.4440253414875304e-07, + "loss": 0.6684, + "step": 15232 + }, + { + "epoch": 0.56, + "learning_rate": 1.44398872084081e-07, + "loss": 0.6962, + "step": 15233 + }, + { + "epoch": 0.56, + "learning_rate": 1.4439521001940892e-07, + "loss": 0.695, + "step": 15234 + }, + { + "epoch": 0.56, + "learning_rate": 1.4439154795473688e-07, + "loss": 0.6979, + "step": 15235 + }, + { + "epoch": 0.56, + "learning_rate": 1.443878858900648e-07, + "loss": 0.7106, + "step": 15236 + }, + { + "epoch": 0.56, + "learning_rate": 1.4438422382539276e-07, + "loss": 0.6888, + "step": 15237 + }, + { + "epoch": 0.56, + "learning_rate": 1.4438056176072068e-07, + "loss": 0.6836, + "step": 15238 + }, + { + "epoch": 0.56, + "learning_rate": 1.4437689969604864e-07, + "loss": 0.7035, + "step": 15239 + }, + { + "epoch": 0.56, + "learning_rate": 1.4437323763137656e-07, + "loss": 0.6996, + "step": 15240 + }, + { + "epoch": 0.56, + "learning_rate": 1.4436957556670452e-07, + "loss": 0.7028, + "step": 15241 + }, + { + "epoch": 0.56, + "learning_rate": 1.4436591350203244e-07, + "loss": 0.686, + "step": 15242 + }, + { + "epoch": 0.56, + "learning_rate": 1.4436225143736037e-07, + "loss": 0.6887, + "step": 15243 + }, + { + "epoch": 0.56, + "learning_rate": 1.4435858937268832e-07, + "loss": 0.6713, + "step": 15244 + }, + { + "epoch": 0.56, + "learning_rate": 1.4435492730801625e-07, + "loss": 0.6857, + "step": 15245 + }, + { + "epoch": 0.56, + "learning_rate": 1.443512652433442e-07, + "loss": 0.6933, + "step": 15246 + }, + { + "epoch": 0.56, + "learning_rate": 1.4434760317867213e-07, + "loss": 0.7106, + "step": 15247 + }, + { + "epoch": 0.56, + "learning_rate": 1.4434394111400006e-07, + "loss": 0.6855, + "step": 15248 + }, + { + "epoch": 0.56, + "learning_rate": 1.4434027904932798e-07, + "loss": 0.6971, + "step": 15249 + }, + { + "epoch": 0.56, + "learning_rate": 1.4433661698465594e-07, + "loss": 0.6834, + "step": 15250 + }, + { + "epoch": 0.56, + "learning_rate": 1.4433295491998386e-07, + "loss": 0.6987, + "step": 15251 + }, + { + "epoch": 0.56, + "learning_rate": 1.4432929285531182e-07, + "loss": 0.6952, + "step": 15252 + }, + { + "epoch": 0.56, + "learning_rate": 1.4432563079063974e-07, + "loss": 0.694, + "step": 15253 + }, + { + "epoch": 0.56, + "learning_rate": 1.443219687259677e-07, + "loss": 0.7132, + "step": 15254 + }, + { + "epoch": 0.56, + "learning_rate": 1.4431830666129562e-07, + "loss": 0.691, + "step": 15255 + }, + { + "epoch": 0.56, + "learning_rate": 1.4431464459662358e-07, + "loss": 0.6807, + "step": 15256 + }, + { + "epoch": 0.56, + "learning_rate": 1.443109825319515e-07, + "loss": 0.7032, + "step": 15257 + }, + { + "epoch": 0.56, + "learning_rate": 1.4430732046727946e-07, + "loss": 0.6842, + "step": 15258 + }, + { + "epoch": 0.56, + "learning_rate": 1.4430365840260738e-07, + "loss": 0.7162, + "step": 15259 + }, + { + "epoch": 0.56, + "learning_rate": 1.4429999633793534e-07, + "loss": 0.6914, + "step": 15260 + }, + { + "epoch": 0.56, + "learning_rate": 1.4429633427326326e-07, + "loss": 0.6972, + "step": 15261 + }, + { + "epoch": 0.56, + "learning_rate": 1.4429267220859122e-07, + "loss": 0.6978, + "step": 15262 + }, + { + "epoch": 0.56, + "learning_rate": 1.4428901014391914e-07, + "loss": 0.7038, + "step": 15263 + }, + { + "epoch": 0.56, + "learning_rate": 1.442853480792471e-07, + "loss": 0.6917, + "step": 15264 + }, + { + "epoch": 0.56, + "learning_rate": 1.4428168601457502e-07, + "loss": 0.6974, + "step": 15265 + }, + { + "epoch": 0.56, + "learning_rate": 1.4427802394990295e-07, + "loss": 0.6989, + "step": 15266 + }, + { + "epoch": 0.56, + "learning_rate": 1.4427436188523088e-07, + "loss": 0.6905, + "step": 15267 + }, + { + "epoch": 0.56, + "learning_rate": 1.4427069982055883e-07, + "loss": 0.6779, + "step": 15268 + }, + { + "epoch": 0.56, + "learning_rate": 1.4426703775588676e-07, + "loss": 0.6782, + "step": 15269 + }, + { + "epoch": 0.56, + "learning_rate": 1.4426337569121468e-07, + "loss": 0.6985, + "step": 15270 + }, + { + "epoch": 0.56, + "learning_rate": 1.4425971362654264e-07, + "loss": 0.6836, + "step": 15271 + }, + { + "epoch": 0.56, + "learning_rate": 1.4425605156187056e-07, + "loss": 0.6787, + "step": 15272 + }, + { + "epoch": 0.56, + "learning_rate": 1.4425238949719851e-07, + "loss": 0.6926, + "step": 15273 + }, + { + "epoch": 0.56, + "learning_rate": 1.4424872743252644e-07, + "loss": 0.6939, + "step": 15274 + }, + { + "epoch": 0.56, + "learning_rate": 1.442450653678544e-07, + "loss": 0.7075, + "step": 15275 + }, + { + "epoch": 0.56, + "learning_rate": 1.4424140330318232e-07, + "loss": 0.6849, + "step": 15276 + }, + { + "epoch": 0.56, + "learning_rate": 1.4423774123851027e-07, + "loss": 0.6973, + "step": 15277 + }, + { + "epoch": 0.56, + "learning_rate": 1.442340791738382e-07, + "loss": 0.6833, + "step": 15278 + }, + { + "epoch": 0.56, + "learning_rate": 1.4423041710916615e-07, + "loss": 0.7106, + "step": 15279 + }, + { + "epoch": 0.56, + "learning_rate": 1.4422675504449408e-07, + "loss": 0.6933, + "step": 15280 + }, + { + "epoch": 0.56, + "learning_rate": 1.4422309297982203e-07, + "loss": 0.686, + "step": 15281 + }, + { + "epoch": 0.56, + "learning_rate": 1.4421943091514996e-07, + "loss": 0.6887, + "step": 15282 + }, + { + "epoch": 0.56, + "learning_rate": 1.4421576885047791e-07, + "loss": 0.6784, + "step": 15283 + }, + { + "epoch": 0.56, + "learning_rate": 1.4421210678580581e-07, + "loss": 0.7015, + "step": 15284 + }, + { + "epoch": 0.56, + "learning_rate": 1.4420844472113377e-07, + "loss": 0.6861, + "step": 15285 + }, + { + "epoch": 0.56, + "learning_rate": 1.442047826564617e-07, + "loss": 0.6975, + "step": 15286 + }, + { + "epoch": 0.56, + "learning_rate": 1.4420112059178965e-07, + "loss": 0.6825, + "step": 15287 + }, + { + "epoch": 0.56, + "learning_rate": 1.4419745852711757e-07, + "loss": 0.7066, + "step": 15288 + }, + { + "epoch": 0.56, + "learning_rate": 1.4419379646244553e-07, + "loss": 0.7009, + "step": 15289 + }, + { + "epoch": 0.56, + "learning_rate": 1.4419013439777345e-07, + "loss": 0.6899, + "step": 15290 + }, + { + "epoch": 0.56, + "learning_rate": 1.441864723331014e-07, + "loss": 0.6886, + "step": 15291 + }, + { + "epoch": 0.56, + "learning_rate": 1.4418281026842933e-07, + "loss": 0.6998, + "step": 15292 + }, + { + "epoch": 0.56, + "learning_rate": 1.4417914820375726e-07, + "loss": 0.6866, + "step": 15293 + }, + { + "epoch": 0.56, + "learning_rate": 1.4417548613908521e-07, + "loss": 0.695, + "step": 15294 + }, + { + "epoch": 0.56, + "learning_rate": 1.4417182407441314e-07, + "loss": 0.6978, + "step": 15295 + }, + { + "epoch": 0.56, + "learning_rate": 1.441681620097411e-07, + "loss": 0.6989, + "step": 15296 + }, + { + "epoch": 0.56, + "learning_rate": 1.4416449994506902e-07, + "loss": 0.6927, + "step": 15297 + }, + { + "epoch": 0.56, + "learning_rate": 1.4416083788039697e-07, + "loss": 0.6897, + "step": 15298 + }, + { + "epoch": 0.56, + "learning_rate": 1.441571758157249e-07, + "loss": 0.6964, + "step": 15299 + }, + { + "epoch": 0.56, + "learning_rate": 1.4415351375105285e-07, + "loss": 0.6972, + "step": 15300 + }, + { + "epoch": 0.56, + "learning_rate": 1.4414985168638078e-07, + "loss": 0.6955, + "step": 15301 + }, + { + "epoch": 0.56, + "learning_rate": 1.441461896217087e-07, + "loss": 0.7229, + "step": 15302 + }, + { + "epoch": 0.56, + "learning_rate": 1.4414252755703663e-07, + "loss": 0.7056, + "step": 15303 + }, + { + "epoch": 0.56, + "learning_rate": 1.4413886549236459e-07, + "loss": 0.6964, + "step": 15304 + }, + { + "epoch": 0.56, + "learning_rate": 1.4413520342769251e-07, + "loss": 0.6897, + "step": 15305 + }, + { + "epoch": 0.56, + "learning_rate": 1.4413154136302047e-07, + "loss": 0.7026, + "step": 15306 + }, + { + "epoch": 0.56, + "learning_rate": 1.441278792983484e-07, + "loss": 0.7037, + "step": 15307 + }, + { + "epoch": 0.56, + "learning_rate": 1.4412421723367635e-07, + "loss": 0.6977, + "step": 15308 + }, + { + "epoch": 0.56, + "learning_rate": 1.4412055516900427e-07, + "loss": 0.71, + "step": 15309 + }, + { + "epoch": 0.56, + "learning_rate": 1.4411689310433223e-07, + "loss": 0.6926, + "step": 15310 + }, + { + "epoch": 0.56, + "learning_rate": 1.4411323103966015e-07, + "loss": 0.6995, + "step": 15311 + }, + { + "epoch": 0.56, + "learning_rate": 1.441095689749881e-07, + "loss": 0.6945, + "step": 15312 + }, + { + "epoch": 0.56, + "learning_rate": 1.4410590691031603e-07, + "loss": 0.7031, + "step": 15313 + }, + { + "epoch": 0.56, + "learning_rate": 1.4410224484564396e-07, + "loss": 0.7017, + "step": 15314 + }, + { + "epoch": 0.56, + "learning_rate": 1.440985827809719e-07, + "loss": 0.6925, + "step": 15315 + }, + { + "epoch": 0.56, + "learning_rate": 1.4409492071629984e-07, + "loss": 0.6824, + "step": 15316 + }, + { + "epoch": 0.56, + "learning_rate": 1.440912586516278e-07, + "loss": 0.6983, + "step": 15317 + }, + { + "epoch": 0.56, + "learning_rate": 1.4408759658695572e-07, + "loss": 0.6887, + "step": 15318 + }, + { + "epoch": 0.56, + "learning_rate": 1.4408393452228367e-07, + "loss": 0.6945, + "step": 15319 + }, + { + "epoch": 0.56, + "learning_rate": 1.4408027245761157e-07, + "loss": 0.7026, + "step": 15320 + }, + { + "epoch": 0.56, + "learning_rate": 1.4407661039293953e-07, + "loss": 0.7006, + "step": 15321 + }, + { + "epoch": 0.56, + "learning_rate": 1.4407294832826745e-07, + "loss": 0.6943, + "step": 15322 + }, + { + "epoch": 0.56, + "learning_rate": 1.440692862635954e-07, + "loss": 0.71, + "step": 15323 + }, + { + "epoch": 0.56, + "learning_rate": 1.4406562419892333e-07, + "loss": 0.6913, + "step": 15324 + }, + { + "epoch": 0.56, + "learning_rate": 1.4406196213425129e-07, + "loss": 0.6959, + "step": 15325 + }, + { + "epoch": 0.56, + "learning_rate": 1.440583000695792e-07, + "loss": 0.7002, + "step": 15326 + }, + { + "epoch": 0.56, + "learning_rate": 1.4405463800490717e-07, + "loss": 0.7073, + "step": 15327 + }, + { + "epoch": 0.56, + "learning_rate": 1.440509759402351e-07, + "loss": 0.688, + "step": 15328 + }, + { + "epoch": 0.56, + "learning_rate": 1.4404731387556305e-07, + "loss": 0.7029, + "step": 15329 + }, + { + "epoch": 0.56, + "learning_rate": 1.4404365181089097e-07, + "loss": 0.7005, + "step": 15330 + }, + { + "epoch": 0.56, + "learning_rate": 1.4403998974621892e-07, + "loss": 0.6965, + "step": 15331 + }, + { + "epoch": 0.56, + "learning_rate": 1.4403632768154685e-07, + "loss": 0.6963, + "step": 15332 + }, + { + "epoch": 0.56, + "learning_rate": 1.440326656168748e-07, + "loss": 0.6947, + "step": 15333 + }, + { + "epoch": 0.56, + "learning_rate": 1.4402900355220273e-07, + "loss": 0.6817, + "step": 15334 + }, + { + "epoch": 0.56, + "learning_rate": 1.4402534148753068e-07, + "loss": 0.6948, + "step": 15335 + }, + { + "epoch": 0.56, + "learning_rate": 1.440216794228586e-07, + "loss": 0.6968, + "step": 15336 + }, + { + "epoch": 0.56, + "learning_rate": 1.4401801735818654e-07, + "loss": 0.6999, + "step": 15337 + }, + { + "epoch": 0.56, + "learning_rate": 1.4401435529351446e-07, + "loss": 0.6923, + "step": 15338 + }, + { + "epoch": 0.56, + "learning_rate": 1.4401069322884242e-07, + "loss": 0.688, + "step": 15339 + }, + { + "epoch": 0.56, + "learning_rate": 1.4400703116417034e-07, + "loss": 0.7028, + "step": 15340 + }, + { + "epoch": 0.56, + "learning_rate": 1.4400336909949827e-07, + "loss": 0.6881, + "step": 15341 + }, + { + "epoch": 0.56, + "learning_rate": 1.4399970703482622e-07, + "loss": 0.7064, + "step": 15342 + }, + { + "epoch": 0.56, + "learning_rate": 1.4399604497015415e-07, + "loss": 0.6811, + "step": 15343 + }, + { + "epoch": 0.56, + "learning_rate": 1.439923829054821e-07, + "loss": 0.6906, + "step": 15344 + }, + { + "epoch": 0.56, + "learning_rate": 1.4398872084081003e-07, + "loss": 0.7042, + "step": 15345 + }, + { + "epoch": 0.56, + "learning_rate": 1.4398505877613798e-07, + "loss": 0.6969, + "step": 15346 + }, + { + "epoch": 0.56, + "learning_rate": 1.439813967114659e-07, + "loss": 0.7009, + "step": 15347 + }, + { + "epoch": 0.56, + "learning_rate": 1.4397773464679386e-07, + "loss": 0.691, + "step": 15348 + }, + { + "epoch": 0.56, + "learning_rate": 1.439740725821218e-07, + "loss": 0.6841, + "step": 15349 + }, + { + "epoch": 0.56, + "learning_rate": 1.4397041051744974e-07, + "loss": 0.6934, + "step": 15350 + }, + { + "epoch": 0.56, + "learning_rate": 1.4396674845277767e-07, + "loss": 0.686, + "step": 15351 + }, + { + "epoch": 0.56, + "learning_rate": 1.4396308638810562e-07, + "loss": 0.7069, + "step": 15352 + }, + { + "epoch": 0.56, + "learning_rate": 1.4395942432343355e-07, + "loss": 0.69, + "step": 15353 + }, + { + "epoch": 0.56, + "learning_rate": 1.439557622587615e-07, + "loss": 0.6852, + "step": 15354 + }, + { + "epoch": 0.56, + "learning_rate": 1.4395210019408943e-07, + "loss": 0.6957, + "step": 15355 + }, + { + "epoch": 0.56, + "learning_rate": 1.4394843812941736e-07, + "loss": 0.6977, + "step": 15356 + }, + { + "epoch": 0.56, + "learning_rate": 1.4394477606474528e-07, + "loss": 0.6888, + "step": 15357 + }, + { + "epoch": 0.56, + "learning_rate": 1.4394111400007324e-07, + "loss": 0.6887, + "step": 15358 + }, + { + "epoch": 0.56, + "learning_rate": 1.4393745193540116e-07, + "loss": 0.6923, + "step": 15359 + }, + { + "epoch": 0.56, + "learning_rate": 1.4393378987072912e-07, + "loss": 0.6943, + "step": 15360 + }, + { + "epoch": 0.56, + "learning_rate": 1.4393012780605704e-07, + "loss": 0.7056, + "step": 15361 + }, + { + "epoch": 0.56, + "learning_rate": 1.43926465741385e-07, + "loss": 0.6871, + "step": 15362 + }, + { + "epoch": 0.56, + "learning_rate": 1.4392280367671292e-07, + "loss": 0.6921, + "step": 15363 + }, + { + "epoch": 0.56, + "learning_rate": 1.4391914161204085e-07, + "loss": 0.6788, + "step": 15364 + }, + { + "epoch": 0.56, + "learning_rate": 1.439154795473688e-07, + "loss": 0.7048, + "step": 15365 + }, + { + "epoch": 0.56, + "learning_rate": 1.4391181748269673e-07, + "loss": 0.6676, + "step": 15366 + }, + { + "epoch": 0.56, + "learning_rate": 1.4390815541802468e-07, + "loss": 0.6863, + "step": 15367 + }, + { + "epoch": 0.56, + "learning_rate": 1.439044933533526e-07, + "loss": 0.6911, + "step": 15368 + }, + { + "epoch": 0.56, + "learning_rate": 1.4390083128868056e-07, + "loss": 0.6839, + "step": 15369 + }, + { + "epoch": 0.56, + "learning_rate": 1.438971692240085e-07, + "loss": 0.6908, + "step": 15370 + }, + { + "epoch": 0.56, + "learning_rate": 1.4389350715933644e-07, + "loss": 0.6842, + "step": 15371 + }, + { + "epoch": 0.56, + "learning_rate": 1.4388984509466437e-07, + "loss": 0.6974, + "step": 15372 + }, + { + "epoch": 0.56, + "learning_rate": 1.4388618302999232e-07, + "loss": 0.698, + "step": 15373 + }, + { + "epoch": 0.56, + "learning_rate": 1.4388252096532022e-07, + "loss": 0.6872, + "step": 15374 + }, + { + "epoch": 0.56, + "learning_rate": 1.4387885890064818e-07, + "loss": 0.711, + "step": 15375 + }, + { + "epoch": 0.56, + "learning_rate": 1.438751968359761e-07, + "loss": 0.6977, + "step": 15376 + }, + { + "epoch": 0.56, + "learning_rate": 1.4387153477130406e-07, + "loss": 0.6862, + "step": 15377 + }, + { + "epoch": 0.56, + "learning_rate": 1.4386787270663198e-07, + "loss": 0.6866, + "step": 15378 + }, + { + "epoch": 0.56, + "learning_rate": 1.4386421064195994e-07, + "loss": 0.7005, + "step": 15379 + }, + { + "epoch": 0.56, + "learning_rate": 1.4386054857728786e-07, + "loss": 0.6869, + "step": 15380 + }, + { + "epoch": 0.56, + "learning_rate": 1.4385688651261582e-07, + "loss": 0.7012, + "step": 15381 + }, + { + "epoch": 0.56, + "learning_rate": 1.4385322444794374e-07, + "loss": 0.6869, + "step": 15382 + }, + { + "epoch": 0.56, + "learning_rate": 1.438495623832717e-07, + "loss": 0.7047, + "step": 15383 + }, + { + "epoch": 0.56, + "learning_rate": 1.4384590031859962e-07, + "loss": 0.6959, + "step": 15384 + }, + { + "epoch": 0.56, + "learning_rate": 1.4384223825392755e-07, + "loss": 0.6929, + "step": 15385 + }, + { + "epoch": 0.56, + "learning_rate": 1.438385761892555e-07, + "loss": 0.7016, + "step": 15386 + }, + { + "epoch": 0.56, + "learning_rate": 1.4383491412458343e-07, + "loss": 0.6939, + "step": 15387 + }, + { + "epoch": 0.56, + "learning_rate": 1.4383125205991138e-07, + "loss": 0.7004, + "step": 15388 + }, + { + "epoch": 0.56, + "learning_rate": 1.438275899952393e-07, + "loss": 0.6986, + "step": 15389 + }, + { + "epoch": 0.56, + "learning_rate": 1.4382392793056726e-07, + "loss": 0.6739, + "step": 15390 + }, + { + "epoch": 0.56, + "learning_rate": 1.438202658658952e-07, + "loss": 0.6883, + "step": 15391 + }, + { + "epoch": 0.56, + "learning_rate": 1.4381660380122311e-07, + "loss": 0.6948, + "step": 15392 + }, + { + "epoch": 0.56, + "learning_rate": 1.4381294173655104e-07, + "loss": 0.7048, + "step": 15393 + }, + { + "epoch": 0.56, + "learning_rate": 1.43809279671879e-07, + "loss": 0.6964, + "step": 15394 + }, + { + "epoch": 0.56, + "learning_rate": 1.4380561760720692e-07, + "loss": 0.688, + "step": 15395 + }, + { + "epoch": 0.56, + "learning_rate": 1.4380195554253487e-07, + "loss": 0.694, + "step": 15396 + }, + { + "epoch": 0.56, + "learning_rate": 1.437982934778628e-07, + "loss": 0.6886, + "step": 15397 + }, + { + "epoch": 0.56, + "learning_rate": 1.4379463141319075e-07, + "loss": 0.6863, + "step": 15398 + }, + { + "epoch": 0.56, + "learning_rate": 1.4379096934851868e-07, + "loss": 0.6954, + "step": 15399 + }, + { + "epoch": 0.56, + "learning_rate": 1.4378730728384663e-07, + "loss": 0.6833, + "step": 15400 + }, + { + "epoch": 0.56, + "learning_rate": 1.4378364521917456e-07, + "loss": 0.6998, + "step": 15401 + }, + { + "epoch": 0.56, + "learning_rate": 1.4377998315450251e-07, + "loss": 0.6975, + "step": 15402 + }, + { + "epoch": 0.56, + "learning_rate": 1.4377632108983044e-07, + "loss": 0.6844, + "step": 15403 + }, + { + "epoch": 0.56, + "learning_rate": 1.437726590251584e-07, + "loss": 0.6875, + "step": 15404 + }, + { + "epoch": 0.56, + "learning_rate": 1.4376899696048632e-07, + "loss": 0.6859, + "step": 15405 + }, + { + "epoch": 0.56, + "learning_rate": 1.4376533489581427e-07, + "loss": 0.6909, + "step": 15406 + }, + { + "epoch": 0.56, + "learning_rate": 1.437616728311422e-07, + "loss": 0.6968, + "step": 15407 + }, + { + "epoch": 0.56, + "learning_rate": 1.4375801076647013e-07, + "loss": 0.6943, + "step": 15408 + }, + { + "epoch": 0.56, + "learning_rate": 1.4375434870179808e-07, + "loss": 0.7035, + "step": 15409 + }, + { + "epoch": 0.56, + "learning_rate": 1.43750686637126e-07, + "loss": 0.6781, + "step": 15410 + }, + { + "epoch": 0.56, + "learning_rate": 1.4374702457245393e-07, + "loss": 0.6895, + "step": 15411 + }, + { + "epoch": 0.56, + "learning_rate": 1.4374336250778186e-07, + "loss": 0.6802, + "step": 15412 + }, + { + "epoch": 0.56, + "learning_rate": 1.4373970044310981e-07, + "loss": 0.6867, + "step": 15413 + }, + { + "epoch": 0.56, + "learning_rate": 1.4373603837843774e-07, + "loss": 0.6932, + "step": 15414 + }, + { + "epoch": 0.56, + "learning_rate": 1.437323763137657e-07, + "loss": 0.6873, + "step": 15415 + }, + { + "epoch": 0.56, + "learning_rate": 1.4372871424909362e-07, + "loss": 0.691, + "step": 15416 + }, + { + "epoch": 0.56, + "learning_rate": 1.4372505218442157e-07, + "loss": 0.6872, + "step": 15417 + }, + { + "epoch": 0.56, + "learning_rate": 1.437213901197495e-07, + "loss": 0.6892, + "step": 15418 + }, + { + "epoch": 0.56, + "learning_rate": 1.4371772805507745e-07, + "loss": 0.6992, + "step": 15419 + }, + { + "epoch": 0.56, + "learning_rate": 1.4371406599040538e-07, + "loss": 0.6933, + "step": 15420 + }, + { + "epoch": 0.56, + "learning_rate": 1.4371040392573333e-07, + "loss": 0.7066, + "step": 15421 + }, + { + "epoch": 0.56, + "learning_rate": 1.4370674186106126e-07, + "loss": 0.6971, + "step": 15422 + }, + { + "epoch": 0.56, + "learning_rate": 1.437030797963892e-07, + "loss": 0.6919, + "step": 15423 + }, + { + "epoch": 0.56, + "learning_rate": 1.4369941773171714e-07, + "loss": 0.6958, + "step": 15424 + }, + { + "epoch": 0.56, + "learning_rate": 1.436957556670451e-07, + "loss": 0.6897, + "step": 15425 + }, + { + "epoch": 0.56, + "learning_rate": 1.4369209360237302e-07, + "loss": 0.6827, + "step": 15426 + }, + { + "epoch": 0.56, + "learning_rate": 1.4368843153770097e-07, + "loss": 0.6862, + "step": 15427 + }, + { + "epoch": 0.56, + "learning_rate": 1.4368476947302887e-07, + "loss": 0.6773, + "step": 15428 + }, + { + "epoch": 0.56, + "learning_rate": 1.4368110740835683e-07, + "loss": 0.6879, + "step": 15429 + }, + { + "epoch": 0.56, + "learning_rate": 1.4367744534368475e-07, + "loss": 0.7079, + "step": 15430 + }, + { + "epoch": 0.56, + "learning_rate": 1.436737832790127e-07, + "loss": 0.6936, + "step": 15431 + }, + { + "epoch": 0.56, + "learning_rate": 1.4367012121434063e-07, + "loss": 0.6919, + "step": 15432 + }, + { + "epoch": 0.56, + "learning_rate": 1.4366645914966859e-07, + "loss": 0.6868, + "step": 15433 + }, + { + "epoch": 0.56, + "learning_rate": 1.436627970849965e-07, + "loss": 0.6966, + "step": 15434 + }, + { + "epoch": 0.56, + "learning_rate": 1.4365913502032444e-07, + "loss": 0.6927, + "step": 15435 + }, + { + "epoch": 0.56, + "learning_rate": 1.436554729556524e-07, + "loss": 0.7085, + "step": 15436 + }, + { + "epoch": 0.56, + "learning_rate": 1.4365181089098032e-07, + "loss": 0.6985, + "step": 15437 + }, + { + "epoch": 0.56, + "learning_rate": 1.4364814882630827e-07, + "loss": 0.692, + "step": 15438 + }, + { + "epoch": 0.56, + "learning_rate": 1.436444867616362e-07, + "loss": 0.6912, + "step": 15439 + }, + { + "epoch": 0.56, + "learning_rate": 1.4364082469696415e-07, + "loss": 0.6953, + "step": 15440 + }, + { + "epoch": 0.56, + "learning_rate": 1.4363716263229208e-07, + "loss": 0.709, + "step": 15441 + }, + { + "epoch": 0.56, + "learning_rate": 1.4363350056762003e-07, + "loss": 0.6839, + "step": 15442 + }, + { + "epoch": 0.57, + "learning_rate": 1.4362983850294796e-07, + "loss": 0.6852, + "step": 15443 + }, + { + "epoch": 0.57, + "learning_rate": 1.436261764382759e-07, + "loss": 0.7075, + "step": 15444 + }, + { + "epoch": 0.57, + "learning_rate": 1.436225143736038e-07, + "loss": 0.684, + "step": 15445 + }, + { + "epoch": 0.57, + "learning_rate": 1.4361885230893177e-07, + "loss": 0.7035, + "step": 15446 + }, + { + "epoch": 0.57, + "learning_rate": 1.436151902442597e-07, + "loss": 0.6914, + "step": 15447 + }, + { + "epoch": 0.57, + "learning_rate": 1.4361152817958765e-07, + "loss": 0.6945, + "step": 15448 + }, + { + "epoch": 0.57, + "learning_rate": 1.4360786611491557e-07, + "loss": 0.6912, + "step": 15449 + }, + { + "epoch": 0.57, + "learning_rate": 1.4360420405024352e-07, + "loss": 0.6873, + "step": 15450 + }, + { + "epoch": 0.57, + "learning_rate": 1.4360054198557145e-07, + "loss": 0.7078, + "step": 15451 + }, + { + "epoch": 0.57, + "learning_rate": 1.435968799208994e-07, + "loss": 0.6953, + "step": 15452 + }, + { + "epoch": 0.57, + "learning_rate": 1.4359321785622733e-07, + "loss": 0.6919, + "step": 15453 + }, + { + "epoch": 0.57, + "learning_rate": 1.4358955579155528e-07, + "loss": 0.6905, + "step": 15454 + }, + { + "epoch": 0.57, + "learning_rate": 1.435858937268832e-07, + "loss": 0.6892, + "step": 15455 + }, + { + "epoch": 0.57, + "learning_rate": 1.4358223166221114e-07, + "loss": 0.6954, + "step": 15456 + }, + { + "epoch": 0.57, + "learning_rate": 1.435785695975391e-07, + "loss": 0.692, + "step": 15457 + }, + { + "epoch": 0.57, + "learning_rate": 1.4357490753286702e-07, + "loss": 0.6981, + "step": 15458 + }, + { + "epoch": 0.57, + "learning_rate": 1.4357124546819497e-07, + "loss": 0.6991, + "step": 15459 + }, + { + "epoch": 0.57, + "learning_rate": 1.435675834035229e-07, + "loss": 0.6942, + "step": 15460 + }, + { + "epoch": 0.57, + "learning_rate": 1.4356392133885085e-07, + "loss": 0.7074, + "step": 15461 + }, + { + "epoch": 0.57, + "learning_rate": 1.4356025927417878e-07, + "loss": 0.6933, + "step": 15462 + }, + { + "epoch": 0.57, + "learning_rate": 1.435565972095067e-07, + "loss": 0.7109, + "step": 15463 + }, + { + "epoch": 0.57, + "learning_rate": 1.4355293514483463e-07, + "loss": 0.6853, + "step": 15464 + }, + { + "epoch": 0.57, + "learning_rate": 1.4354927308016258e-07, + "loss": 0.6944, + "step": 15465 + }, + { + "epoch": 0.57, + "learning_rate": 1.435456110154905e-07, + "loss": 0.6847, + "step": 15466 + }, + { + "epoch": 0.57, + "learning_rate": 1.4354194895081846e-07, + "loss": 0.6703, + "step": 15467 + }, + { + "epoch": 0.57, + "learning_rate": 1.435382868861464e-07, + "loss": 0.6967, + "step": 15468 + }, + { + "epoch": 0.57, + "learning_rate": 1.4353462482147434e-07, + "loss": 0.6965, + "step": 15469 + }, + { + "epoch": 0.57, + "learning_rate": 1.4353096275680227e-07, + "loss": 0.6817, + "step": 15470 + }, + { + "epoch": 0.57, + "learning_rate": 1.4352730069213022e-07, + "loss": 0.6881, + "step": 15471 + }, + { + "epoch": 0.57, + "learning_rate": 1.4352363862745815e-07, + "loss": 0.693, + "step": 15472 + }, + { + "epoch": 0.57, + "learning_rate": 1.435199765627861e-07, + "loss": 0.7043, + "step": 15473 + }, + { + "epoch": 0.57, + "learning_rate": 1.4351631449811403e-07, + "loss": 0.7021, + "step": 15474 + }, + { + "epoch": 0.57, + "learning_rate": 1.4351265243344198e-07, + "loss": 0.7045, + "step": 15475 + }, + { + "epoch": 0.57, + "learning_rate": 1.435089903687699e-07, + "loss": 0.6738, + "step": 15476 + }, + { + "epoch": 0.57, + "learning_rate": 1.4350532830409786e-07, + "loss": 0.6992, + "step": 15477 + }, + { + "epoch": 0.57, + "learning_rate": 1.435016662394258e-07, + "loss": 0.6881, + "step": 15478 + }, + { + "epoch": 0.57, + "learning_rate": 1.4349800417475372e-07, + "loss": 0.6909, + "step": 15479 + }, + { + "epoch": 0.57, + "learning_rate": 1.4349434211008167e-07, + "loss": 0.6759, + "step": 15480 + }, + { + "epoch": 0.57, + "learning_rate": 1.434906800454096e-07, + "loss": 0.6852, + "step": 15481 + }, + { + "epoch": 0.57, + "learning_rate": 1.4348701798073752e-07, + "loss": 0.7032, + "step": 15482 + }, + { + "epoch": 0.57, + "learning_rate": 1.4348335591606545e-07, + "loss": 0.6978, + "step": 15483 + }, + { + "epoch": 0.57, + "learning_rate": 1.434796938513934e-07, + "loss": 0.6997, + "step": 15484 + }, + { + "epoch": 0.57, + "learning_rate": 1.4347603178672133e-07, + "loss": 0.6848, + "step": 15485 + }, + { + "epoch": 0.57, + "learning_rate": 1.4347236972204928e-07, + "loss": 0.6926, + "step": 15486 + }, + { + "epoch": 0.57, + "learning_rate": 1.434687076573772e-07, + "loss": 0.6876, + "step": 15487 + }, + { + "epoch": 0.57, + "learning_rate": 1.4346504559270516e-07, + "loss": 0.6927, + "step": 15488 + }, + { + "epoch": 0.57, + "learning_rate": 1.434613835280331e-07, + "loss": 0.6897, + "step": 15489 + }, + { + "epoch": 0.57, + "learning_rate": 1.4345772146336104e-07, + "loss": 0.687, + "step": 15490 + }, + { + "epoch": 0.57, + "learning_rate": 1.4345405939868897e-07, + "loss": 0.7, + "step": 15491 + }, + { + "epoch": 0.57, + "learning_rate": 1.4345039733401692e-07, + "loss": 0.6931, + "step": 15492 + }, + { + "epoch": 0.57, + "learning_rate": 1.4344673526934485e-07, + "loss": 0.7079, + "step": 15493 + }, + { + "epoch": 0.57, + "learning_rate": 1.434430732046728e-07, + "loss": 0.6954, + "step": 15494 + }, + { + "epoch": 0.57, + "learning_rate": 1.4343941114000073e-07, + "loss": 0.6992, + "step": 15495 + }, + { + "epoch": 0.57, + "learning_rate": 1.4343574907532868e-07, + "loss": 0.684, + "step": 15496 + }, + { + "epoch": 0.57, + "learning_rate": 1.434320870106566e-07, + "loss": 0.6929, + "step": 15497 + }, + { + "epoch": 0.57, + "learning_rate": 1.4342842494598456e-07, + "loss": 0.6968, + "step": 15498 + }, + { + "epoch": 0.57, + "learning_rate": 1.4342476288131246e-07, + "loss": 0.684, + "step": 15499 + }, + { + "epoch": 0.57, + "learning_rate": 1.4342110081664042e-07, + "loss": 0.6723, + "step": 15500 + }, + { + "epoch": 0.57, + "learning_rate": 1.4341743875196834e-07, + "loss": 0.6897, + "step": 15501 + }, + { + "epoch": 0.57, + "learning_rate": 1.434137766872963e-07, + "loss": 0.6873, + "step": 15502 + }, + { + "epoch": 0.57, + "learning_rate": 1.4341011462262422e-07, + "loss": 0.6956, + "step": 15503 + }, + { + "epoch": 0.57, + "learning_rate": 1.4340645255795218e-07, + "loss": 0.6958, + "step": 15504 + }, + { + "epoch": 0.57, + "learning_rate": 1.434027904932801e-07, + "loss": 0.6927, + "step": 15505 + }, + { + "epoch": 0.57, + "learning_rate": 1.4339912842860803e-07, + "loss": 0.7078, + "step": 15506 + }, + { + "epoch": 0.57, + "learning_rate": 1.4339546636393598e-07, + "loss": 0.684, + "step": 15507 + }, + { + "epoch": 0.57, + "learning_rate": 1.433918042992639e-07, + "loss": 0.6824, + "step": 15508 + }, + { + "epoch": 0.57, + "learning_rate": 1.4338814223459186e-07, + "loss": 0.6951, + "step": 15509 + }, + { + "epoch": 0.57, + "learning_rate": 1.433844801699198e-07, + "loss": 0.694, + "step": 15510 + }, + { + "epoch": 0.57, + "learning_rate": 1.4338081810524774e-07, + "loss": 0.6865, + "step": 15511 + }, + { + "epoch": 0.57, + "learning_rate": 1.4337715604057567e-07, + "loss": 0.6918, + "step": 15512 + }, + { + "epoch": 0.57, + "learning_rate": 1.4337349397590362e-07, + "loss": 0.6757, + "step": 15513 + }, + { + "epoch": 0.57, + "learning_rate": 1.4336983191123155e-07, + "loss": 0.7005, + "step": 15514 + }, + { + "epoch": 0.57, + "learning_rate": 1.433661698465595e-07, + "loss": 0.6955, + "step": 15515 + }, + { + "epoch": 0.57, + "learning_rate": 1.4336250778188743e-07, + "loss": 0.6988, + "step": 15516 + }, + { + "epoch": 0.57, + "learning_rate": 1.4335884571721535e-07, + "loss": 0.6972, + "step": 15517 + }, + { + "epoch": 0.57, + "learning_rate": 1.4335518365254328e-07, + "loss": 0.6777, + "step": 15518 + }, + { + "epoch": 0.57, + "learning_rate": 1.4335152158787123e-07, + "loss": 0.6833, + "step": 15519 + }, + { + "epoch": 0.57, + "learning_rate": 1.4334785952319916e-07, + "loss": 0.6918, + "step": 15520 + }, + { + "epoch": 0.57, + "learning_rate": 1.4334419745852711e-07, + "loss": 0.6944, + "step": 15521 + }, + { + "epoch": 0.57, + "learning_rate": 1.4334053539385504e-07, + "loss": 0.6978, + "step": 15522 + }, + { + "epoch": 0.57, + "learning_rate": 1.43336873329183e-07, + "loss": 0.6934, + "step": 15523 + }, + { + "epoch": 0.57, + "learning_rate": 1.4333321126451092e-07, + "loss": 0.6967, + "step": 15524 + }, + { + "epoch": 0.57, + "learning_rate": 1.4332954919983887e-07, + "loss": 0.6856, + "step": 15525 + }, + { + "epoch": 0.57, + "learning_rate": 1.433258871351668e-07, + "loss": 0.6896, + "step": 15526 + }, + { + "epoch": 0.57, + "learning_rate": 1.4332222507049473e-07, + "loss": 0.6923, + "step": 15527 + }, + { + "epoch": 0.57, + "learning_rate": 1.4331856300582268e-07, + "loss": 0.6872, + "step": 15528 + }, + { + "epoch": 0.57, + "learning_rate": 1.433149009411506e-07, + "loss": 0.6942, + "step": 15529 + }, + { + "epoch": 0.57, + "learning_rate": 1.4331123887647856e-07, + "loss": 0.6948, + "step": 15530 + }, + { + "epoch": 0.57, + "learning_rate": 1.433075768118065e-07, + "loss": 0.6964, + "step": 15531 + }, + { + "epoch": 0.57, + "learning_rate": 1.4330391474713444e-07, + "loss": 0.6689, + "step": 15532 + }, + { + "epoch": 0.57, + "learning_rate": 1.4330025268246237e-07, + "loss": 0.7027, + "step": 15533 + }, + { + "epoch": 0.57, + "learning_rate": 1.4329659061779032e-07, + "loss": 0.7022, + "step": 15534 + }, + { + "epoch": 0.57, + "learning_rate": 1.4329292855311822e-07, + "loss": 0.6981, + "step": 15535 + }, + { + "epoch": 0.57, + "learning_rate": 1.4328926648844617e-07, + "loss": 0.696, + "step": 15536 + }, + { + "epoch": 0.57, + "learning_rate": 1.432856044237741e-07, + "loss": 0.6933, + "step": 15537 + }, + { + "epoch": 0.57, + "learning_rate": 1.4328194235910205e-07, + "loss": 0.6815, + "step": 15538 + }, + { + "epoch": 0.57, + "learning_rate": 1.4327828029442998e-07, + "loss": 0.7101, + "step": 15539 + }, + { + "epoch": 0.57, + "learning_rate": 1.4327461822975793e-07, + "loss": 0.6946, + "step": 15540 + }, + { + "epoch": 0.57, + "learning_rate": 1.4327095616508586e-07, + "loss": 0.6959, + "step": 15541 + }, + { + "epoch": 0.57, + "learning_rate": 1.432672941004138e-07, + "loss": 0.6986, + "step": 15542 + }, + { + "epoch": 0.57, + "learning_rate": 1.4326363203574174e-07, + "loss": 0.6998, + "step": 15543 + }, + { + "epoch": 0.57, + "learning_rate": 1.432599699710697e-07, + "loss": 0.6946, + "step": 15544 + }, + { + "epoch": 0.57, + "learning_rate": 1.4325630790639762e-07, + "loss": 0.6901, + "step": 15545 + }, + { + "epoch": 0.57, + "learning_rate": 1.4325264584172557e-07, + "loss": 0.7044, + "step": 15546 + }, + { + "epoch": 0.57, + "learning_rate": 1.432489837770535e-07, + "loss": 0.6906, + "step": 15547 + }, + { + "epoch": 0.57, + "learning_rate": 1.4324532171238145e-07, + "loss": 0.7043, + "step": 15548 + }, + { + "epoch": 0.57, + "learning_rate": 1.4324165964770938e-07, + "loss": 0.6981, + "step": 15549 + }, + { + "epoch": 0.57, + "learning_rate": 1.432379975830373e-07, + "loss": 0.6981, + "step": 15550 + }, + { + "epoch": 0.57, + "learning_rate": 1.4323433551836526e-07, + "loss": 0.7008, + "step": 15551 + }, + { + "epoch": 0.57, + "learning_rate": 1.4323067345369319e-07, + "loss": 0.7045, + "step": 15552 + }, + { + "epoch": 0.57, + "learning_rate": 1.432270113890211e-07, + "loss": 0.6917, + "step": 15553 + }, + { + "epoch": 0.57, + "learning_rate": 1.4322334932434904e-07, + "loss": 0.6822, + "step": 15554 + }, + { + "epoch": 0.57, + "learning_rate": 1.43219687259677e-07, + "loss": 0.6949, + "step": 15555 + }, + { + "epoch": 0.57, + "learning_rate": 1.4321602519500492e-07, + "loss": 0.6946, + "step": 15556 + }, + { + "epoch": 0.57, + "learning_rate": 1.4321236313033287e-07, + "loss": 0.6944, + "step": 15557 + }, + { + "epoch": 0.57, + "learning_rate": 1.432087010656608e-07, + "loss": 0.6851, + "step": 15558 + }, + { + "epoch": 0.57, + "learning_rate": 1.4320503900098875e-07, + "loss": 0.6948, + "step": 15559 + }, + { + "epoch": 0.57, + "learning_rate": 1.4320137693631668e-07, + "loss": 0.6996, + "step": 15560 + }, + { + "epoch": 0.57, + "learning_rate": 1.4319771487164463e-07, + "loss": 0.6925, + "step": 15561 + }, + { + "epoch": 0.57, + "learning_rate": 1.4319405280697256e-07, + "loss": 0.7015, + "step": 15562 + }, + { + "epoch": 0.57, + "learning_rate": 1.431903907423005e-07, + "loss": 0.6801, + "step": 15563 + }, + { + "epoch": 0.57, + "learning_rate": 1.4318672867762844e-07, + "loss": 0.6973, + "step": 15564 + }, + { + "epoch": 0.57, + "learning_rate": 1.431830666129564e-07, + "loss": 0.6824, + "step": 15565 + }, + { + "epoch": 0.57, + "learning_rate": 1.4317940454828432e-07, + "loss": 0.6927, + "step": 15566 + }, + { + "epoch": 0.57, + "learning_rate": 1.4317574248361227e-07, + "loss": 0.7036, + "step": 15567 + }, + { + "epoch": 0.57, + "learning_rate": 1.431720804189402e-07, + "loss": 0.6901, + "step": 15568 + }, + { + "epoch": 0.57, + "learning_rate": 1.4316841835426815e-07, + "loss": 0.6936, + "step": 15569 + }, + { + "epoch": 0.57, + "learning_rate": 1.4316475628959608e-07, + "loss": 0.6977, + "step": 15570 + }, + { + "epoch": 0.57, + "learning_rate": 1.43161094224924e-07, + "loss": 0.712, + "step": 15571 + }, + { + "epoch": 0.57, + "learning_rate": 1.4315743216025193e-07, + "loss": 0.7028, + "step": 15572 + }, + { + "epoch": 0.57, + "learning_rate": 1.4315377009557988e-07, + "loss": 0.6947, + "step": 15573 + }, + { + "epoch": 0.57, + "learning_rate": 1.431501080309078e-07, + "loss": 0.6745, + "step": 15574 + }, + { + "epoch": 0.57, + "learning_rate": 1.4314644596623576e-07, + "loss": 0.6877, + "step": 15575 + }, + { + "epoch": 0.57, + "learning_rate": 1.431427839015637e-07, + "loss": 0.7009, + "step": 15576 + }, + { + "epoch": 0.57, + "learning_rate": 1.4313912183689162e-07, + "loss": 0.7013, + "step": 15577 + }, + { + "epoch": 0.57, + "learning_rate": 1.4313545977221957e-07, + "loss": 0.6972, + "step": 15578 + }, + { + "epoch": 0.57, + "learning_rate": 1.431317977075475e-07, + "loss": 0.7016, + "step": 15579 + }, + { + "epoch": 0.57, + "learning_rate": 1.4312813564287545e-07, + "loss": 0.6955, + "step": 15580 + }, + { + "epoch": 0.57, + "learning_rate": 1.4312447357820338e-07, + "loss": 0.6942, + "step": 15581 + }, + { + "epoch": 0.57, + "learning_rate": 1.4312081151353133e-07, + "loss": 0.6968, + "step": 15582 + }, + { + "epoch": 0.57, + "learning_rate": 1.4311714944885926e-07, + "loss": 0.6913, + "step": 15583 + }, + { + "epoch": 0.57, + "learning_rate": 1.431134873841872e-07, + "loss": 0.702, + "step": 15584 + }, + { + "epoch": 0.57, + "learning_rate": 1.4310982531951514e-07, + "loss": 0.6951, + "step": 15585 + }, + { + "epoch": 0.57, + "learning_rate": 1.431061632548431e-07, + "loss": 0.6985, + "step": 15586 + }, + { + "epoch": 0.57, + "learning_rate": 1.4310250119017102e-07, + "loss": 0.7, + "step": 15587 + }, + { + "epoch": 0.57, + "learning_rate": 1.4309883912549897e-07, + "loss": 0.6935, + "step": 15588 + }, + { + "epoch": 0.57, + "learning_rate": 1.4309517706082687e-07, + "loss": 0.6957, + "step": 15589 + }, + { + "epoch": 0.57, + "learning_rate": 1.4309151499615482e-07, + "loss": 0.6984, + "step": 15590 + }, + { + "epoch": 0.57, + "learning_rate": 1.4308785293148275e-07, + "loss": 0.6735, + "step": 15591 + }, + { + "epoch": 0.57, + "learning_rate": 1.430841908668107e-07, + "loss": 0.6965, + "step": 15592 + }, + { + "epoch": 0.57, + "learning_rate": 1.4308052880213863e-07, + "loss": 0.6914, + "step": 15593 + }, + { + "epoch": 0.57, + "learning_rate": 1.4307686673746658e-07, + "loss": 0.6927, + "step": 15594 + }, + { + "epoch": 0.57, + "learning_rate": 1.430732046727945e-07, + "loss": 0.6936, + "step": 15595 + }, + { + "epoch": 0.57, + "learning_rate": 1.4306954260812246e-07, + "loss": 0.6945, + "step": 15596 + }, + { + "epoch": 0.57, + "learning_rate": 1.430658805434504e-07, + "loss": 0.6935, + "step": 15597 + }, + { + "epoch": 0.57, + "learning_rate": 1.4306221847877832e-07, + "loss": 0.7021, + "step": 15598 + }, + { + "epoch": 0.57, + "learning_rate": 1.4305855641410627e-07, + "loss": 0.7052, + "step": 15599 + }, + { + "epoch": 0.57, + "learning_rate": 1.430548943494342e-07, + "loss": 0.68, + "step": 15600 + }, + { + "epoch": 0.57, + "learning_rate": 1.4305123228476215e-07, + "loss": 0.6951, + "step": 15601 + }, + { + "epoch": 0.57, + "learning_rate": 1.4304757022009008e-07, + "loss": 0.6954, + "step": 15602 + }, + { + "epoch": 0.57, + "learning_rate": 1.4304390815541803e-07, + "loss": 0.7102, + "step": 15603 + }, + { + "epoch": 0.57, + "learning_rate": 1.4304024609074596e-07, + "loss": 0.6869, + "step": 15604 + }, + { + "epoch": 0.57, + "learning_rate": 1.430365840260739e-07, + "loss": 0.6897, + "step": 15605 + }, + { + "epoch": 0.57, + "learning_rate": 1.4303292196140184e-07, + "loss": 0.6826, + "step": 15606 + }, + { + "epoch": 0.57, + "learning_rate": 1.4302925989672976e-07, + "loss": 0.6997, + "step": 15607 + }, + { + "epoch": 0.57, + "learning_rate": 1.430255978320577e-07, + "loss": 0.6946, + "step": 15608 + }, + { + "epoch": 0.57, + "learning_rate": 1.4302193576738564e-07, + "loss": 0.6931, + "step": 15609 + }, + { + "epoch": 0.57, + "learning_rate": 1.4301827370271357e-07, + "loss": 0.6938, + "step": 15610 + }, + { + "epoch": 0.57, + "learning_rate": 1.4301461163804152e-07, + "loss": 0.6942, + "step": 15611 + }, + { + "epoch": 0.57, + "learning_rate": 1.4301094957336945e-07, + "loss": 0.7057, + "step": 15612 + }, + { + "epoch": 0.57, + "learning_rate": 1.430072875086974e-07, + "loss": 0.7064, + "step": 15613 + }, + { + "epoch": 0.57, + "learning_rate": 1.4300362544402533e-07, + "loss": 0.689, + "step": 15614 + }, + { + "epoch": 0.57, + "learning_rate": 1.4299996337935328e-07, + "loss": 0.707, + "step": 15615 + }, + { + "epoch": 0.57, + "learning_rate": 1.429963013146812e-07, + "loss": 0.6807, + "step": 15616 + }, + { + "epoch": 0.57, + "learning_rate": 1.4299263925000916e-07, + "loss": 0.689, + "step": 15617 + }, + { + "epoch": 0.57, + "learning_rate": 1.429889771853371e-07, + "loss": 0.7033, + "step": 15618 + }, + { + "epoch": 0.57, + "learning_rate": 1.4298531512066502e-07, + "loss": 0.6923, + "step": 15619 + }, + { + "epoch": 0.57, + "learning_rate": 1.4298165305599297e-07, + "loss": 0.6976, + "step": 15620 + }, + { + "epoch": 0.57, + "learning_rate": 1.429779909913209e-07, + "loss": 0.6909, + "step": 15621 + }, + { + "epoch": 0.57, + "learning_rate": 1.4297432892664885e-07, + "loss": 0.6998, + "step": 15622 + }, + { + "epoch": 0.57, + "learning_rate": 1.4297066686197678e-07, + "loss": 0.6893, + "step": 15623 + }, + { + "epoch": 0.57, + "learning_rate": 1.4296700479730473e-07, + "loss": 0.6899, + "step": 15624 + }, + { + "epoch": 0.57, + "learning_rate": 1.4296334273263263e-07, + "loss": 0.7065, + "step": 15625 + }, + { + "epoch": 0.57, + "learning_rate": 1.4295968066796058e-07, + "loss": 0.716, + "step": 15626 + }, + { + "epoch": 0.57, + "learning_rate": 1.429560186032885e-07, + "loss": 0.6899, + "step": 15627 + }, + { + "epoch": 0.57, + "learning_rate": 1.4295235653861646e-07, + "loss": 0.6884, + "step": 15628 + }, + { + "epoch": 0.57, + "learning_rate": 1.429486944739444e-07, + "loss": 0.6968, + "step": 15629 + }, + { + "epoch": 0.57, + "learning_rate": 1.4294503240927234e-07, + "loss": 0.7004, + "step": 15630 + }, + { + "epoch": 0.57, + "learning_rate": 1.4294137034460027e-07, + "loss": 0.6972, + "step": 15631 + }, + { + "epoch": 0.57, + "learning_rate": 1.4293770827992822e-07, + "loss": 0.6809, + "step": 15632 + }, + { + "epoch": 0.57, + "learning_rate": 1.4293404621525615e-07, + "loss": 0.6662, + "step": 15633 + }, + { + "epoch": 0.57, + "learning_rate": 1.429303841505841e-07, + "loss": 0.6735, + "step": 15634 + }, + { + "epoch": 0.57, + "learning_rate": 1.4292672208591203e-07, + "loss": 0.6817, + "step": 15635 + }, + { + "epoch": 0.57, + "learning_rate": 1.4292306002123998e-07, + "loss": 0.692, + "step": 15636 + }, + { + "epoch": 0.57, + "learning_rate": 1.429193979565679e-07, + "loss": 0.708, + "step": 15637 + }, + { + "epoch": 0.57, + "learning_rate": 1.4291573589189586e-07, + "loss": 0.7186, + "step": 15638 + }, + { + "epoch": 0.57, + "learning_rate": 1.429120738272238e-07, + "loss": 0.7069, + "step": 15639 + }, + { + "epoch": 0.57, + "learning_rate": 1.4290841176255174e-07, + "loss": 0.697, + "step": 15640 + }, + { + "epoch": 0.57, + "learning_rate": 1.4290474969787967e-07, + "loss": 0.6965, + "step": 15641 + }, + { + "epoch": 0.57, + "learning_rate": 1.429010876332076e-07, + "loss": 0.6991, + "step": 15642 + }, + { + "epoch": 0.57, + "learning_rate": 1.4289742556853552e-07, + "loss": 0.7032, + "step": 15643 + }, + { + "epoch": 0.57, + "learning_rate": 1.4289376350386347e-07, + "loss": 0.6944, + "step": 15644 + }, + { + "epoch": 0.57, + "learning_rate": 1.428901014391914e-07, + "loss": 0.6869, + "step": 15645 + }, + { + "epoch": 0.57, + "learning_rate": 1.4288643937451935e-07, + "loss": 0.6984, + "step": 15646 + }, + { + "epoch": 0.57, + "learning_rate": 1.4288277730984728e-07, + "loss": 0.6826, + "step": 15647 + }, + { + "epoch": 0.57, + "learning_rate": 1.428791152451752e-07, + "loss": 0.693, + "step": 15648 + }, + { + "epoch": 0.57, + "learning_rate": 1.4287545318050316e-07, + "loss": 0.6651, + "step": 15649 + }, + { + "epoch": 0.57, + "learning_rate": 1.428717911158311e-07, + "loss": 0.7023, + "step": 15650 + }, + { + "epoch": 0.57, + "learning_rate": 1.4286812905115904e-07, + "loss": 0.6987, + "step": 15651 + }, + { + "epoch": 0.57, + "learning_rate": 1.4286446698648697e-07, + "loss": 0.7062, + "step": 15652 + }, + { + "epoch": 0.57, + "learning_rate": 1.4286080492181492e-07, + "loss": 0.7072, + "step": 15653 + }, + { + "epoch": 0.57, + "learning_rate": 1.4285714285714285e-07, + "loss": 0.6858, + "step": 15654 + }, + { + "epoch": 0.57, + "learning_rate": 1.428534807924708e-07, + "loss": 0.6928, + "step": 15655 + }, + { + "epoch": 0.57, + "learning_rate": 1.4284981872779873e-07, + "loss": 0.6869, + "step": 15656 + }, + { + "epoch": 0.57, + "learning_rate": 1.4284615666312668e-07, + "loss": 0.6816, + "step": 15657 + }, + { + "epoch": 0.57, + "learning_rate": 1.428424945984546e-07, + "loss": 0.6865, + "step": 15658 + }, + { + "epoch": 0.57, + "learning_rate": 1.4283883253378256e-07, + "loss": 0.6948, + "step": 15659 + }, + { + "epoch": 0.57, + "learning_rate": 1.4283517046911049e-07, + "loss": 0.6826, + "step": 15660 + }, + { + "epoch": 0.57, + "learning_rate": 1.428315084044384e-07, + "loss": 0.6957, + "step": 15661 + }, + { + "epoch": 0.57, + "learning_rate": 1.4282784633976634e-07, + "loss": 0.6845, + "step": 15662 + }, + { + "epoch": 0.57, + "learning_rate": 1.428241842750943e-07, + "loss": 0.6995, + "step": 15663 + }, + { + "epoch": 0.57, + "learning_rate": 1.4282052221042222e-07, + "loss": 0.7041, + "step": 15664 + }, + { + "epoch": 0.57, + "learning_rate": 1.4281686014575017e-07, + "loss": 0.6874, + "step": 15665 + }, + { + "epoch": 0.57, + "learning_rate": 1.428131980810781e-07, + "loss": 0.7197, + "step": 15666 + }, + { + "epoch": 0.57, + "learning_rate": 1.4280953601640605e-07, + "loss": 0.6923, + "step": 15667 + }, + { + "epoch": 0.57, + "learning_rate": 1.4280587395173398e-07, + "loss": 0.6956, + "step": 15668 + }, + { + "epoch": 0.57, + "learning_rate": 1.428022118870619e-07, + "loss": 0.7025, + "step": 15669 + }, + { + "epoch": 0.57, + "learning_rate": 1.4279854982238986e-07, + "loss": 0.6957, + "step": 15670 + }, + { + "epoch": 0.57, + "learning_rate": 1.4279488775771779e-07, + "loss": 0.6874, + "step": 15671 + }, + { + "epoch": 0.57, + "learning_rate": 1.4279122569304574e-07, + "loss": 0.6949, + "step": 15672 + }, + { + "epoch": 0.57, + "learning_rate": 1.4278756362837367e-07, + "loss": 0.6908, + "step": 15673 + }, + { + "epoch": 0.57, + "learning_rate": 1.4278390156370162e-07, + "loss": 0.6999, + "step": 15674 + }, + { + "epoch": 0.57, + "learning_rate": 1.4278023949902955e-07, + "loss": 0.7077, + "step": 15675 + }, + { + "epoch": 0.57, + "learning_rate": 1.427765774343575e-07, + "loss": 0.6969, + "step": 15676 + }, + { + "epoch": 0.57, + "learning_rate": 1.4277291536968543e-07, + "loss": 0.6863, + "step": 15677 + }, + { + "epoch": 0.57, + "learning_rate": 1.4276925330501338e-07, + "loss": 0.6948, + "step": 15678 + }, + { + "epoch": 0.57, + "learning_rate": 1.4276559124034128e-07, + "loss": 0.7025, + "step": 15679 + }, + { + "epoch": 0.57, + "learning_rate": 1.4276192917566923e-07, + "loss": 0.6877, + "step": 15680 + }, + { + "epoch": 0.57, + "learning_rate": 1.4275826711099716e-07, + "loss": 0.7003, + "step": 15681 + }, + { + "epoch": 0.57, + "learning_rate": 1.427546050463251e-07, + "loss": 0.6867, + "step": 15682 + }, + { + "epoch": 0.57, + "learning_rate": 1.4275094298165304e-07, + "loss": 0.6918, + "step": 15683 + }, + { + "epoch": 0.57, + "learning_rate": 1.42747280916981e-07, + "loss": 0.7037, + "step": 15684 + }, + { + "epoch": 0.57, + "learning_rate": 1.4274361885230892e-07, + "loss": 0.6979, + "step": 15685 + }, + { + "epoch": 0.57, + "learning_rate": 1.4273995678763687e-07, + "loss": 0.6888, + "step": 15686 + }, + { + "epoch": 0.57, + "learning_rate": 1.427362947229648e-07, + "loss": 0.6963, + "step": 15687 + }, + { + "epoch": 0.57, + "learning_rate": 1.4273263265829275e-07, + "loss": 0.6873, + "step": 15688 + }, + { + "epoch": 0.57, + "learning_rate": 1.4272897059362068e-07, + "loss": 0.7111, + "step": 15689 + }, + { + "epoch": 0.57, + "learning_rate": 1.427253085289486e-07, + "loss": 0.6966, + "step": 15690 + }, + { + "epoch": 0.57, + "learning_rate": 1.4272164646427656e-07, + "loss": 0.6949, + "step": 15691 + }, + { + "epoch": 0.57, + "learning_rate": 1.4271798439960448e-07, + "loss": 0.6886, + "step": 15692 + }, + { + "epoch": 0.57, + "learning_rate": 1.4271432233493244e-07, + "loss": 0.6919, + "step": 15693 + }, + { + "epoch": 0.57, + "learning_rate": 1.4271066027026036e-07, + "loss": 0.683, + "step": 15694 + }, + { + "epoch": 0.57, + "learning_rate": 1.4270699820558832e-07, + "loss": 0.6985, + "step": 15695 + }, + { + "epoch": 0.57, + "learning_rate": 1.4270333614091624e-07, + "loss": 0.7027, + "step": 15696 + }, + { + "epoch": 0.57, + "learning_rate": 1.4269967407624417e-07, + "loss": 0.6932, + "step": 15697 + }, + { + "epoch": 0.57, + "learning_rate": 1.426960120115721e-07, + "loss": 0.7056, + "step": 15698 + }, + { + "epoch": 0.57, + "learning_rate": 1.4269234994690005e-07, + "loss": 0.6864, + "step": 15699 + }, + { + "epoch": 0.57, + "learning_rate": 1.4268868788222798e-07, + "loss": 0.6904, + "step": 15700 + }, + { + "epoch": 0.57, + "learning_rate": 1.4268502581755593e-07, + "loss": 0.6963, + "step": 15701 + }, + { + "epoch": 0.57, + "learning_rate": 1.4268136375288386e-07, + "loss": 0.6873, + "step": 15702 + }, + { + "epoch": 0.57, + "learning_rate": 1.426777016882118e-07, + "loss": 0.6831, + "step": 15703 + }, + { + "epoch": 0.57, + "learning_rate": 1.4267403962353974e-07, + "loss": 0.7017, + "step": 15704 + }, + { + "epoch": 0.57, + "learning_rate": 1.426703775588677e-07, + "loss": 0.6929, + "step": 15705 + }, + { + "epoch": 0.57, + "learning_rate": 1.4266671549419562e-07, + "loss": 0.6987, + "step": 15706 + }, + { + "epoch": 0.57, + "learning_rate": 1.4266305342952357e-07, + "loss": 0.695, + "step": 15707 + }, + { + "epoch": 0.57, + "learning_rate": 1.426593913648515e-07, + "loss": 0.6944, + "step": 15708 + }, + { + "epoch": 0.57, + "learning_rate": 1.4265572930017945e-07, + "loss": 0.6891, + "step": 15709 + }, + { + "epoch": 0.57, + "learning_rate": 1.4265206723550738e-07, + "loss": 0.7116, + "step": 15710 + }, + { + "epoch": 0.57, + "learning_rate": 1.4264840517083533e-07, + "loss": 0.6959, + "step": 15711 + }, + { + "epoch": 0.57, + "learning_rate": 1.4264474310616326e-07, + "loss": 0.6937, + "step": 15712 + }, + { + "epoch": 0.57, + "learning_rate": 1.4264108104149118e-07, + "loss": 0.6946, + "step": 15713 + }, + { + "epoch": 0.57, + "learning_rate": 1.4263741897681914e-07, + "loss": 0.6912, + "step": 15714 + }, + { + "epoch": 0.57, + "learning_rate": 1.4263375691214706e-07, + "loss": 0.696, + "step": 15715 + }, + { + "epoch": 0.58, + "learning_rate": 1.42630094847475e-07, + "loss": 0.7012, + "step": 15716 + }, + { + "epoch": 0.58, + "learning_rate": 1.4262643278280294e-07, + "loss": 0.69, + "step": 15717 + }, + { + "epoch": 0.58, + "learning_rate": 1.4262277071813087e-07, + "loss": 0.6916, + "step": 15718 + }, + { + "epoch": 0.58, + "learning_rate": 1.426191086534588e-07, + "loss": 0.6936, + "step": 15719 + }, + { + "epoch": 0.58, + "learning_rate": 1.4261544658878675e-07, + "loss": 0.6883, + "step": 15720 + }, + { + "epoch": 0.58, + "learning_rate": 1.4261178452411468e-07, + "loss": 0.6882, + "step": 15721 + }, + { + "epoch": 0.58, + "learning_rate": 1.4260812245944263e-07, + "loss": 0.7009, + "step": 15722 + }, + { + "epoch": 0.58, + "learning_rate": 1.4260446039477056e-07, + "loss": 0.6887, + "step": 15723 + }, + { + "epoch": 0.58, + "learning_rate": 1.426007983300985e-07, + "loss": 0.6842, + "step": 15724 + }, + { + "epoch": 0.58, + "learning_rate": 1.4259713626542644e-07, + "loss": 0.702, + "step": 15725 + }, + { + "epoch": 0.58, + "learning_rate": 1.425934742007544e-07, + "loss": 0.6977, + "step": 15726 + }, + { + "epoch": 0.58, + "learning_rate": 1.4258981213608232e-07, + "loss": 0.6915, + "step": 15727 + }, + { + "epoch": 0.58, + "learning_rate": 1.4258615007141027e-07, + "loss": 0.6975, + "step": 15728 + }, + { + "epoch": 0.58, + "learning_rate": 1.425824880067382e-07, + "loss": 0.7066, + "step": 15729 + }, + { + "epoch": 0.58, + "learning_rate": 1.4257882594206615e-07, + "loss": 0.6909, + "step": 15730 + }, + { + "epoch": 0.58, + "learning_rate": 1.4257516387739408e-07, + "loss": 0.7314, + "step": 15731 + }, + { + "epoch": 0.58, + "learning_rate": 1.4257150181272203e-07, + "loss": 0.6988, + "step": 15732 + }, + { + "epoch": 0.58, + "learning_rate": 1.4256783974804993e-07, + "loss": 0.6925, + "step": 15733 + }, + { + "epoch": 0.58, + "learning_rate": 1.4256417768337788e-07, + "loss": 0.6885, + "step": 15734 + }, + { + "epoch": 0.58, + "learning_rate": 1.425605156187058e-07, + "loss": 0.6919, + "step": 15735 + }, + { + "epoch": 0.58, + "learning_rate": 1.4255685355403376e-07, + "loss": 0.7011, + "step": 15736 + }, + { + "epoch": 0.58, + "learning_rate": 1.425531914893617e-07, + "loss": 0.6941, + "step": 15737 + }, + { + "epoch": 0.58, + "learning_rate": 1.4254952942468964e-07, + "loss": 0.6969, + "step": 15738 + }, + { + "epoch": 0.58, + "learning_rate": 1.4254586736001757e-07, + "loss": 0.6939, + "step": 15739 + }, + { + "epoch": 0.58, + "learning_rate": 1.425422052953455e-07, + "loss": 0.693, + "step": 15740 + }, + { + "epoch": 0.58, + "learning_rate": 1.4253854323067345e-07, + "loss": 0.6866, + "step": 15741 + }, + { + "epoch": 0.58, + "learning_rate": 1.4253488116600138e-07, + "loss": 0.6967, + "step": 15742 + }, + { + "epoch": 0.58, + "learning_rate": 1.4253121910132933e-07, + "loss": 0.68, + "step": 15743 + }, + { + "epoch": 0.58, + "learning_rate": 1.4252755703665726e-07, + "loss": 0.7097, + "step": 15744 + }, + { + "epoch": 0.58, + "learning_rate": 1.425238949719852e-07, + "loss": 0.6948, + "step": 15745 + }, + { + "epoch": 0.58, + "learning_rate": 1.4252023290731314e-07, + "loss": 0.6758, + "step": 15746 + }, + { + "epoch": 0.58, + "learning_rate": 1.425165708426411e-07, + "loss": 0.693, + "step": 15747 + }, + { + "epoch": 0.58, + "learning_rate": 1.4251290877796901e-07, + "loss": 0.6902, + "step": 15748 + }, + { + "epoch": 0.58, + "learning_rate": 1.4250924671329697e-07, + "loss": 0.6936, + "step": 15749 + }, + { + "epoch": 0.58, + "learning_rate": 1.425055846486249e-07, + "loss": 0.6986, + "step": 15750 + }, + { + "epoch": 0.58, + "learning_rate": 1.4250192258395282e-07, + "loss": 0.6903, + "step": 15751 + }, + { + "epoch": 0.58, + "learning_rate": 1.4249826051928075e-07, + "loss": 0.6933, + "step": 15752 + }, + { + "epoch": 0.58, + "learning_rate": 1.424945984546087e-07, + "loss": 0.6932, + "step": 15753 + }, + { + "epoch": 0.58, + "learning_rate": 1.4249093638993663e-07, + "loss": 0.6975, + "step": 15754 + }, + { + "epoch": 0.58, + "learning_rate": 1.4248727432526458e-07, + "loss": 0.6822, + "step": 15755 + }, + { + "epoch": 0.58, + "learning_rate": 1.424836122605925e-07, + "loss": 0.6862, + "step": 15756 + }, + { + "epoch": 0.58, + "learning_rate": 1.4247995019592046e-07, + "loss": 0.6925, + "step": 15757 + }, + { + "epoch": 0.58, + "learning_rate": 1.424762881312484e-07, + "loss": 0.6832, + "step": 15758 + }, + { + "epoch": 0.58, + "learning_rate": 1.4247262606657634e-07, + "loss": 0.6965, + "step": 15759 + }, + { + "epoch": 0.58, + "learning_rate": 1.4246896400190427e-07, + "loss": 0.6768, + "step": 15760 + }, + { + "epoch": 0.58, + "learning_rate": 1.424653019372322e-07, + "loss": 0.6929, + "step": 15761 + }, + { + "epoch": 0.58, + "learning_rate": 1.4246163987256015e-07, + "loss": 0.6942, + "step": 15762 + }, + { + "epoch": 0.58, + "learning_rate": 1.4245797780788807e-07, + "loss": 0.6853, + "step": 15763 + }, + { + "epoch": 0.58, + "learning_rate": 1.4245431574321603e-07, + "loss": 0.6986, + "step": 15764 + }, + { + "epoch": 0.58, + "learning_rate": 1.4245065367854395e-07, + "loss": 0.7086, + "step": 15765 + }, + { + "epoch": 0.58, + "learning_rate": 1.424469916138719e-07, + "loss": 0.6898, + "step": 15766 + }, + { + "epoch": 0.58, + "learning_rate": 1.4244332954919983e-07, + "loss": 0.6926, + "step": 15767 + }, + { + "epoch": 0.58, + "learning_rate": 1.424396674845278e-07, + "loss": 0.6995, + "step": 15768 + }, + { + "epoch": 0.58, + "learning_rate": 1.424360054198557e-07, + "loss": 0.6835, + "step": 15769 + }, + { + "epoch": 0.58, + "learning_rate": 1.4243234335518364e-07, + "loss": 0.6939, + "step": 15770 + }, + { + "epoch": 0.58, + "learning_rate": 1.4242868129051157e-07, + "loss": 0.7029, + "step": 15771 + }, + { + "epoch": 0.58, + "learning_rate": 1.4242501922583952e-07, + "loss": 0.6934, + "step": 15772 + }, + { + "epoch": 0.58, + "learning_rate": 1.4242135716116745e-07, + "loss": 0.6839, + "step": 15773 + }, + { + "epoch": 0.58, + "learning_rate": 1.424176950964954e-07, + "loss": 0.6909, + "step": 15774 + }, + { + "epoch": 0.58, + "learning_rate": 1.4241403303182333e-07, + "loss": 0.6914, + "step": 15775 + }, + { + "epoch": 0.58, + "learning_rate": 1.4241037096715128e-07, + "loss": 0.7161, + "step": 15776 + }, + { + "epoch": 0.58, + "learning_rate": 1.424067089024792e-07, + "loss": 0.6988, + "step": 15777 + }, + { + "epoch": 0.58, + "learning_rate": 1.4240304683780716e-07, + "loss": 0.6873, + "step": 15778 + }, + { + "epoch": 0.58, + "learning_rate": 1.4239938477313509e-07, + "loss": 0.6894, + "step": 15779 + }, + { + "epoch": 0.58, + "learning_rate": 1.4239572270846304e-07, + "loss": 0.6958, + "step": 15780 + }, + { + "epoch": 0.58, + "learning_rate": 1.4239206064379097e-07, + "loss": 0.7022, + "step": 15781 + }, + { + "epoch": 0.58, + "learning_rate": 1.4238839857911892e-07, + "loss": 0.6797, + "step": 15782 + }, + { + "epoch": 0.58, + "learning_rate": 1.4238473651444685e-07, + "loss": 0.702, + "step": 15783 + }, + { + "epoch": 0.58, + "learning_rate": 1.4238107444977477e-07, + "loss": 0.6821, + "step": 15784 + }, + { + "epoch": 0.58, + "learning_rate": 1.4237741238510273e-07, + "loss": 0.6796, + "step": 15785 + }, + { + "epoch": 0.58, + "learning_rate": 1.4237375032043065e-07, + "loss": 0.704, + "step": 15786 + }, + { + "epoch": 0.58, + "learning_rate": 1.4237008825575858e-07, + "loss": 0.6799, + "step": 15787 + }, + { + "epoch": 0.58, + "learning_rate": 1.4236642619108653e-07, + "loss": 0.6874, + "step": 15788 + }, + { + "epoch": 0.58, + "learning_rate": 1.4236276412641446e-07, + "loss": 0.6928, + "step": 15789 + }, + { + "epoch": 0.58, + "learning_rate": 1.4235910206174239e-07, + "loss": 0.694, + "step": 15790 + }, + { + "epoch": 0.58, + "learning_rate": 1.4235543999707034e-07, + "loss": 0.7016, + "step": 15791 + }, + { + "epoch": 0.58, + "learning_rate": 1.4235177793239827e-07, + "loss": 0.6927, + "step": 15792 + }, + { + "epoch": 0.58, + "learning_rate": 1.4234811586772622e-07, + "loss": 0.6902, + "step": 15793 + }, + { + "epoch": 0.58, + "learning_rate": 1.4234445380305415e-07, + "loss": 0.6944, + "step": 15794 + }, + { + "epoch": 0.58, + "learning_rate": 1.423407917383821e-07, + "loss": 0.6986, + "step": 15795 + }, + { + "epoch": 0.58, + "learning_rate": 1.4233712967371003e-07, + "loss": 0.6952, + "step": 15796 + }, + { + "epoch": 0.58, + "learning_rate": 1.4233346760903798e-07, + "loss": 0.6985, + "step": 15797 + }, + { + "epoch": 0.58, + "learning_rate": 1.423298055443659e-07, + "loss": 0.6916, + "step": 15798 + }, + { + "epoch": 0.58, + "learning_rate": 1.4232614347969386e-07, + "loss": 0.6906, + "step": 15799 + }, + { + "epoch": 0.58, + "learning_rate": 1.4232248141502179e-07, + "loss": 0.7016, + "step": 15800 + }, + { + "epoch": 0.58, + "learning_rate": 1.4231881935034974e-07, + "loss": 0.693, + "step": 15801 + }, + { + "epoch": 0.58, + "learning_rate": 1.4231515728567767e-07, + "loss": 0.6995, + "step": 15802 + }, + { + "epoch": 0.58, + "learning_rate": 1.4231149522100562e-07, + "loss": 0.6883, + "step": 15803 + }, + { + "epoch": 0.58, + "learning_rate": 1.4230783315633355e-07, + "loss": 0.7057, + "step": 15804 + }, + { + "epoch": 0.58, + "learning_rate": 1.4230417109166147e-07, + "loss": 0.6924, + "step": 15805 + }, + { + "epoch": 0.58, + "learning_rate": 1.423005090269894e-07, + "loss": 0.6771, + "step": 15806 + }, + { + "epoch": 0.58, + "learning_rate": 1.4229684696231735e-07, + "loss": 0.6884, + "step": 15807 + }, + { + "epoch": 0.58, + "learning_rate": 1.4229318489764528e-07, + "loss": 0.7006, + "step": 15808 + }, + { + "epoch": 0.58, + "learning_rate": 1.4228952283297323e-07, + "loss": 0.7039, + "step": 15809 + }, + { + "epoch": 0.58, + "learning_rate": 1.4228586076830116e-07, + "loss": 0.6994, + "step": 15810 + }, + { + "epoch": 0.58, + "learning_rate": 1.4228219870362908e-07, + "loss": 0.694, + "step": 15811 + }, + { + "epoch": 0.58, + "learning_rate": 1.4227853663895704e-07, + "loss": 0.6904, + "step": 15812 + }, + { + "epoch": 0.58, + "learning_rate": 1.4227487457428496e-07, + "loss": 0.7059, + "step": 15813 + }, + { + "epoch": 0.58, + "learning_rate": 1.4227121250961292e-07, + "loss": 0.6935, + "step": 15814 + }, + { + "epoch": 0.58, + "learning_rate": 1.4226755044494084e-07, + "loss": 0.6919, + "step": 15815 + }, + { + "epoch": 0.58, + "learning_rate": 1.422638883802688e-07, + "loss": 0.6907, + "step": 15816 + }, + { + "epoch": 0.58, + "learning_rate": 1.4226022631559672e-07, + "loss": 0.7015, + "step": 15817 + }, + { + "epoch": 0.58, + "learning_rate": 1.4225656425092468e-07, + "loss": 0.6927, + "step": 15818 + }, + { + "epoch": 0.58, + "learning_rate": 1.422529021862526e-07, + "loss": 0.6923, + "step": 15819 + }, + { + "epoch": 0.58, + "learning_rate": 1.4224924012158056e-07, + "loss": 0.6868, + "step": 15820 + }, + { + "epoch": 0.58, + "learning_rate": 1.4224557805690848e-07, + "loss": 0.6974, + "step": 15821 + }, + { + "epoch": 0.58, + "learning_rate": 1.422419159922364e-07, + "loss": 0.6858, + "step": 15822 + }, + { + "epoch": 0.58, + "learning_rate": 1.4223825392756434e-07, + "loss": 0.6922, + "step": 15823 + }, + { + "epoch": 0.58, + "learning_rate": 1.422345918628923e-07, + "loss": 0.6896, + "step": 15824 + }, + { + "epoch": 0.58, + "learning_rate": 1.4223092979822022e-07, + "loss": 0.6988, + "step": 15825 + }, + { + "epoch": 0.58, + "learning_rate": 1.4222726773354817e-07, + "loss": 0.6924, + "step": 15826 + }, + { + "epoch": 0.58, + "learning_rate": 1.422236056688761e-07, + "loss": 0.6842, + "step": 15827 + }, + { + "epoch": 0.58, + "learning_rate": 1.4221994360420405e-07, + "loss": 0.7031, + "step": 15828 + }, + { + "epoch": 0.58, + "learning_rate": 1.4221628153953198e-07, + "loss": 0.6978, + "step": 15829 + }, + { + "epoch": 0.58, + "learning_rate": 1.4221261947485993e-07, + "loss": 0.7239, + "step": 15830 + }, + { + "epoch": 0.58, + "learning_rate": 1.4220895741018786e-07, + "loss": 0.7107, + "step": 15831 + }, + { + "epoch": 0.58, + "learning_rate": 1.4220529534551578e-07, + "loss": 0.6824, + "step": 15832 + }, + { + "epoch": 0.58, + "learning_rate": 1.4220163328084374e-07, + "loss": 0.6929, + "step": 15833 + }, + { + "epoch": 0.58, + "learning_rate": 1.4219797121617166e-07, + "loss": 0.6969, + "step": 15834 + }, + { + "epoch": 0.58, + "learning_rate": 1.4219430915149962e-07, + "loss": 0.6906, + "step": 15835 + }, + { + "epoch": 0.58, + "learning_rate": 1.4219064708682754e-07, + "loss": 0.6975, + "step": 15836 + }, + { + "epoch": 0.58, + "learning_rate": 1.421869850221555e-07, + "loss": 0.6938, + "step": 15837 + }, + { + "epoch": 0.58, + "learning_rate": 1.4218332295748342e-07, + "loss": 0.6866, + "step": 15838 + }, + { + "epoch": 0.58, + "learning_rate": 1.4217966089281138e-07, + "loss": 0.6959, + "step": 15839 + }, + { + "epoch": 0.58, + "learning_rate": 1.4217599882813928e-07, + "loss": 0.6854, + "step": 15840 + }, + { + "epoch": 0.58, + "learning_rate": 1.4217233676346723e-07, + "loss": 0.6995, + "step": 15841 + }, + { + "epoch": 0.58, + "learning_rate": 1.4216867469879516e-07, + "loss": 0.6896, + "step": 15842 + }, + { + "epoch": 0.58, + "learning_rate": 1.421650126341231e-07, + "loss": 0.6878, + "step": 15843 + }, + { + "epoch": 0.58, + "learning_rate": 1.4216135056945104e-07, + "loss": 0.6974, + "step": 15844 + }, + { + "epoch": 0.58, + "learning_rate": 1.42157688504779e-07, + "loss": 0.6874, + "step": 15845 + }, + { + "epoch": 0.58, + "learning_rate": 1.4215402644010692e-07, + "loss": 0.7015, + "step": 15846 + }, + { + "epoch": 0.58, + "learning_rate": 1.4215036437543487e-07, + "loss": 0.6956, + "step": 15847 + }, + { + "epoch": 0.58, + "learning_rate": 1.421467023107628e-07, + "loss": 0.6952, + "step": 15848 + }, + { + "epoch": 0.58, + "learning_rate": 1.4214304024609075e-07, + "loss": 0.69, + "step": 15849 + }, + { + "epoch": 0.58, + "learning_rate": 1.4213937818141868e-07, + "loss": 0.6865, + "step": 15850 + }, + { + "epoch": 0.58, + "learning_rate": 1.4213571611674663e-07, + "loss": 0.7006, + "step": 15851 + }, + { + "epoch": 0.58, + "learning_rate": 1.4213205405207456e-07, + "loss": 0.6953, + "step": 15852 + }, + { + "epoch": 0.58, + "learning_rate": 1.421283919874025e-07, + "loss": 0.6813, + "step": 15853 + }, + { + "epoch": 0.58, + "learning_rate": 1.4212472992273044e-07, + "loss": 0.6869, + "step": 15854 + }, + { + "epoch": 0.58, + "learning_rate": 1.4212106785805836e-07, + "loss": 0.6773, + "step": 15855 + }, + { + "epoch": 0.58, + "learning_rate": 1.4211740579338632e-07, + "loss": 0.6958, + "step": 15856 + }, + { + "epoch": 0.58, + "learning_rate": 1.4211374372871424e-07, + "loss": 0.7001, + "step": 15857 + }, + { + "epoch": 0.58, + "learning_rate": 1.4211008166404217e-07, + "loss": 0.7019, + "step": 15858 + }, + { + "epoch": 0.58, + "learning_rate": 1.421064195993701e-07, + "loss": 0.6789, + "step": 15859 + }, + { + "epoch": 0.58, + "learning_rate": 1.4210275753469805e-07, + "loss": 0.6839, + "step": 15860 + }, + { + "epoch": 0.58, + "learning_rate": 1.4209909547002598e-07, + "loss": 0.6966, + "step": 15861 + }, + { + "epoch": 0.58, + "learning_rate": 1.4209543340535393e-07, + "loss": 0.6903, + "step": 15862 + }, + { + "epoch": 0.58, + "learning_rate": 1.4209177134068186e-07, + "loss": 0.7023, + "step": 15863 + }, + { + "epoch": 0.58, + "learning_rate": 1.420881092760098e-07, + "loss": 0.6906, + "step": 15864 + }, + { + "epoch": 0.58, + "learning_rate": 1.4208444721133774e-07, + "loss": 0.6897, + "step": 15865 + }, + { + "epoch": 0.58, + "learning_rate": 1.420807851466657e-07, + "loss": 0.7037, + "step": 15866 + }, + { + "epoch": 0.58, + "learning_rate": 1.4207712308199361e-07, + "loss": 0.6819, + "step": 15867 + }, + { + "epoch": 0.58, + "learning_rate": 1.4207346101732157e-07, + "loss": 0.6915, + "step": 15868 + }, + { + "epoch": 0.58, + "learning_rate": 1.420697989526495e-07, + "loss": 0.6816, + "step": 15869 + }, + { + "epoch": 0.58, + "learning_rate": 1.4206613688797745e-07, + "loss": 0.7093, + "step": 15870 + }, + { + "epoch": 0.58, + "learning_rate": 1.4206247482330537e-07, + "loss": 0.6837, + "step": 15871 + }, + { + "epoch": 0.58, + "learning_rate": 1.4205881275863333e-07, + "loss": 0.6928, + "step": 15872 + }, + { + "epoch": 0.58, + "learning_rate": 1.4205515069396125e-07, + "loss": 0.6839, + "step": 15873 + }, + { + "epoch": 0.58, + "learning_rate": 1.420514886292892e-07, + "loss": 0.6912, + "step": 15874 + }, + { + "epoch": 0.58, + "learning_rate": 1.4204782656461713e-07, + "loss": 0.6935, + "step": 15875 + }, + { + "epoch": 0.58, + "learning_rate": 1.4204416449994506e-07, + "loss": 0.6978, + "step": 15876 + }, + { + "epoch": 0.58, + "learning_rate": 1.42040502435273e-07, + "loss": 0.6895, + "step": 15877 + }, + { + "epoch": 0.58, + "learning_rate": 1.4203684037060094e-07, + "loss": 0.6938, + "step": 15878 + }, + { + "epoch": 0.58, + "learning_rate": 1.4203317830592887e-07, + "loss": 0.6953, + "step": 15879 + }, + { + "epoch": 0.58, + "learning_rate": 1.4202951624125682e-07, + "loss": 0.6889, + "step": 15880 + }, + { + "epoch": 0.58, + "learning_rate": 1.4202585417658475e-07, + "loss": 0.6959, + "step": 15881 + }, + { + "epoch": 0.58, + "learning_rate": 1.4202219211191267e-07, + "loss": 0.6845, + "step": 15882 + }, + { + "epoch": 0.58, + "learning_rate": 1.4201853004724063e-07, + "loss": 0.6934, + "step": 15883 + }, + { + "epoch": 0.58, + "learning_rate": 1.4201486798256855e-07, + "loss": 0.6964, + "step": 15884 + }, + { + "epoch": 0.58, + "learning_rate": 1.420112059178965e-07, + "loss": 0.696, + "step": 15885 + }, + { + "epoch": 0.58, + "learning_rate": 1.4200754385322443e-07, + "loss": 0.6911, + "step": 15886 + }, + { + "epoch": 0.58, + "learning_rate": 1.420038817885524e-07, + "loss": 0.6859, + "step": 15887 + }, + { + "epoch": 0.58, + "learning_rate": 1.4200021972388031e-07, + "loss": 0.6978, + "step": 15888 + }, + { + "epoch": 0.58, + "learning_rate": 1.4199655765920827e-07, + "loss": 0.6954, + "step": 15889 + }, + { + "epoch": 0.58, + "learning_rate": 1.419928955945362e-07, + "loss": 0.6913, + "step": 15890 + }, + { + "epoch": 0.58, + "learning_rate": 1.4198923352986415e-07, + "loss": 0.6942, + "step": 15891 + }, + { + "epoch": 0.58, + "learning_rate": 1.4198557146519207e-07, + "loss": 0.6838, + "step": 15892 + }, + { + "epoch": 0.58, + "learning_rate": 1.4198190940052003e-07, + "loss": 0.6975, + "step": 15893 + }, + { + "epoch": 0.58, + "learning_rate": 1.4197824733584793e-07, + "loss": 0.704, + "step": 15894 + }, + { + "epoch": 0.58, + "learning_rate": 1.4197458527117588e-07, + "loss": 0.6809, + "step": 15895 + }, + { + "epoch": 0.58, + "learning_rate": 1.419709232065038e-07, + "loss": 0.6811, + "step": 15896 + }, + { + "epoch": 0.58, + "learning_rate": 1.4196726114183176e-07, + "loss": 0.684, + "step": 15897 + }, + { + "epoch": 0.58, + "learning_rate": 1.4196359907715969e-07, + "loss": 0.6847, + "step": 15898 + }, + { + "epoch": 0.58, + "learning_rate": 1.4195993701248764e-07, + "loss": 0.6903, + "step": 15899 + }, + { + "epoch": 0.58, + "learning_rate": 1.4195627494781557e-07, + "loss": 0.7029, + "step": 15900 + }, + { + "epoch": 0.58, + "learning_rate": 1.4195261288314352e-07, + "loss": 0.6906, + "step": 15901 + }, + { + "epoch": 0.58, + "learning_rate": 1.4194895081847145e-07, + "loss": 0.6936, + "step": 15902 + }, + { + "epoch": 0.58, + "learning_rate": 1.4194528875379937e-07, + "loss": 0.6981, + "step": 15903 + }, + { + "epoch": 0.58, + "learning_rate": 1.4194162668912733e-07, + "loss": 0.696, + "step": 15904 + }, + { + "epoch": 0.58, + "learning_rate": 1.4193796462445525e-07, + "loss": 0.6981, + "step": 15905 + }, + { + "epoch": 0.58, + "learning_rate": 1.419343025597832e-07, + "loss": 0.6947, + "step": 15906 + }, + { + "epoch": 0.58, + "learning_rate": 1.4193064049511113e-07, + "loss": 0.6967, + "step": 15907 + }, + { + "epoch": 0.58, + "learning_rate": 1.4192697843043909e-07, + "loss": 0.6939, + "step": 15908 + }, + { + "epoch": 0.58, + "learning_rate": 1.41923316365767e-07, + "loss": 0.6854, + "step": 15909 + }, + { + "epoch": 0.58, + "learning_rate": 1.4191965430109497e-07, + "loss": 0.694, + "step": 15910 + }, + { + "epoch": 0.58, + "learning_rate": 1.419159922364229e-07, + "loss": 0.6856, + "step": 15911 + }, + { + "epoch": 0.58, + "learning_rate": 1.4191233017175082e-07, + "loss": 0.6874, + "step": 15912 + }, + { + "epoch": 0.58, + "learning_rate": 1.4190866810707875e-07, + "loss": 0.6772, + "step": 15913 + }, + { + "epoch": 0.58, + "learning_rate": 1.419050060424067e-07, + "loss": 0.6867, + "step": 15914 + }, + { + "epoch": 0.58, + "learning_rate": 1.4190134397773463e-07, + "loss": 0.6851, + "step": 15915 + }, + { + "epoch": 0.58, + "learning_rate": 1.4189768191306258e-07, + "loss": 0.6959, + "step": 15916 + }, + { + "epoch": 0.58, + "learning_rate": 1.418940198483905e-07, + "loss": 0.6829, + "step": 15917 + }, + { + "epoch": 0.58, + "learning_rate": 1.4189035778371846e-07, + "loss": 0.6868, + "step": 15918 + }, + { + "epoch": 0.58, + "learning_rate": 1.4188669571904639e-07, + "loss": 0.6802, + "step": 15919 + }, + { + "epoch": 0.58, + "learning_rate": 1.4188303365437434e-07, + "loss": 0.701, + "step": 15920 + }, + { + "epoch": 0.58, + "learning_rate": 1.4187937158970227e-07, + "loss": 0.7053, + "step": 15921 + }, + { + "epoch": 0.58, + "learning_rate": 1.4187570952503022e-07, + "loss": 0.6915, + "step": 15922 + }, + { + "epoch": 0.58, + "learning_rate": 1.4187204746035815e-07, + "loss": 0.6871, + "step": 15923 + }, + { + "epoch": 0.58, + "learning_rate": 1.418683853956861e-07, + "loss": 0.6917, + "step": 15924 + }, + { + "epoch": 0.58, + "learning_rate": 1.4186472333101402e-07, + "loss": 0.676, + "step": 15925 + }, + { + "epoch": 0.58, + "learning_rate": 1.4186106126634195e-07, + "loss": 0.689, + "step": 15926 + }, + { + "epoch": 0.58, + "learning_rate": 1.418573992016699e-07, + "loss": 0.6867, + "step": 15927 + }, + { + "epoch": 0.58, + "learning_rate": 1.4185373713699783e-07, + "loss": 0.6833, + "step": 15928 + }, + { + "epoch": 0.58, + "learning_rate": 1.4185007507232578e-07, + "loss": 0.6815, + "step": 15929 + }, + { + "epoch": 0.58, + "learning_rate": 1.4184641300765368e-07, + "loss": 0.6924, + "step": 15930 + }, + { + "epoch": 0.58, + "learning_rate": 1.4184275094298164e-07, + "loss": 0.6851, + "step": 15931 + }, + { + "epoch": 0.58, + "learning_rate": 1.4183908887830956e-07, + "loss": 0.6899, + "step": 15932 + }, + { + "epoch": 0.58, + "learning_rate": 1.4183542681363752e-07, + "loss": 0.6912, + "step": 15933 + }, + { + "epoch": 0.58, + "learning_rate": 1.4183176474896544e-07, + "loss": 0.7043, + "step": 15934 + }, + { + "epoch": 0.58, + "learning_rate": 1.418281026842934e-07, + "loss": 0.6977, + "step": 15935 + }, + { + "epoch": 0.58, + "learning_rate": 1.4182444061962132e-07, + "loss": 0.6893, + "step": 15936 + }, + { + "epoch": 0.58, + "learning_rate": 1.4182077855494928e-07, + "loss": 0.7006, + "step": 15937 + }, + { + "epoch": 0.58, + "learning_rate": 1.418171164902772e-07, + "loss": 0.6942, + "step": 15938 + }, + { + "epoch": 0.58, + "learning_rate": 1.4181345442560516e-07, + "loss": 0.7035, + "step": 15939 + }, + { + "epoch": 0.58, + "learning_rate": 1.4180979236093308e-07, + "loss": 0.6961, + "step": 15940 + }, + { + "epoch": 0.58, + "learning_rate": 1.4180613029626104e-07, + "loss": 0.6931, + "step": 15941 + }, + { + "epoch": 0.58, + "learning_rate": 1.4180246823158896e-07, + "loss": 0.6875, + "step": 15942 + }, + { + "epoch": 0.58, + "learning_rate": 1.4179880616691692e-07, + "loss": 0.6995, + "step": 15943 + }, + { + "epoch": 0.58, + "learning_rate": 1.4179514410224484e-07, + "loss": 0.7016, + "step": 15944 + }, + { + "epoch": 0.58, + "learning_rate": 1.417914820375728e-07, + "loss": 0.6804, + "step": 15945 + }, + { + "epoch": 0.58, + "learning_rate": 1.4178781997290072e-07, + "loss": 0.692, + "step": 15946 + }, + { + "epoch": 0.58, + "learning_rate": 1.4178415790822865e-07, + "loss": 0.701, + "step": 15947 + }, + { + "epoch": 0.58, + "learning_rate": 1.4178049584355658e-07, + "loss": 0.6986, + "step": 15948 + }, + { + "epoch": 0.58, + "learning_rate": 1.4177683377888453e-07, + "loss": 0.6887, + "step": 15949 + }, + { + "epoch": 0.58, + "learning_rate": 1.4177317171421246e-07, + "loss": 0.6902, + "step": 15950 + }, + { + "epoch": 0.58, + "learning_rate": 1.417695096495404e-07, + "loss": 0.6818, + "step": 15951 + }, + { + "epoch": 0.58, + "learning_rate": 1.4176584758486834e-07, + "loss": 0.6978, + "step": 15952 + }, + { + "epoch": 0.58, + "learning_rate": 1.4176218552019626e-07, + "loss": 0.6875, + "step": 15953 + }, + { + "epoch": 0.58, + "learning_rate": 1.4175852345552422e-07, + "loss": 0.6934, + "step": 15954 + }, + { + "epoch": 0.58, + "learning_rate": 1.4175486139085214e-07, + "loss": 0.6999, + "step": 15955 + }, + { + "epoch": 0.58, + "learning_rate": 1.417511993261801e-07, + "loss": 0.6889, + "step": 15956 + }, + { + "epoch": 0.58, + "learning_rate": 1.4174753726150802e-07, + "loss": 0.7013, + "step": 15957 + }, + { + "epoch": 0.58, + "learning_rate": 1.4174387519683598e-07, + "loss": 0.693, + "step": 15958 + }, + { + "epoch": 0.58, + "learning_rate": 1.417402131321639e-07, + "loss": 0.6955, + "step": 15959 + }, + { + "epoch": 0.58, + "learning_rate": 1.4173655106749186e-07, + "loss": 0.6901, + "step": 15960 + }, + { + "epoch": 0.58, + "learning_rate": 1.4173288900281978e-07, + "loss": 0.6931, + "step": 15961 + }, + { + "epoch": 0.58, + "learning_rate": 1.4172922693814774e-07, + "loss": 0.6869, + "step": 15962 + }, + { + "epoch": 0.58, + "learning_rate": 1.4172556487347566e-07, + "loss": 0.6788, + "step": 15963 + }, + { + "epoch": 0.58, + "learning_rate": 1.4172190280880362e-07, + "loss": 0.6971, + "step": 15964 + }, + { + "epoch": 0.58, + "learning_rate": 1.4171824074413154e-07, + "loss": 0.7014, + "step": 15965 + }, + { + "epoch": 0.58, + "learning_rate": 1.4171457867945947e-07, + "loss": 0.6818, + "step": 15966 + }, + { + "epoch": 0.58, + "learning_rate": 1.417109166147874e-07, + "loss": 0.7054, + "step": 15967 + }, + { + "epoch": 0.58, + "learning_rate": 1.4170725455011535e-07, + "loss": 0.7025, + "step": 15968 + }, + { + "epoch": 0.58, + "learning_rate": 1.4170359248544328e-07, + "loss": 0.6937, + "step": 15969 + }, + { + "epoch": 0.58, + "learning_rate": 1.4169993042077123e-07, + "loss": 0.6971, + "step": 15970 + }, + { + "epoch": 0.58, + "learning_rate": 1.4169626835609916e-07, + "loss": 0.6938, + "step": 15971 + }, + { + "epoch": 0.58, + "learning_rate": 1.416926062914271e-07, + "loss": 0.6989, + "step": 15972 + }, + { + "epoch": 0.58, + "learning_rate": 1.4168894422675504e-07, + "loss": 0.6912, + "step": 15973 + }, + { + "epoch": 0.58, + "learning_rate": 1.4168528216208296e-07, + "loss": 0.6909, + "step": 15974 + }, + { + "epoch": 0.58, + "learning_rate": 1.4168162009741092e-07, + "loss": 0.6963, + "step": 15975 + }, + { + "epoch": 0.58, + "learning_rate": 1.4167795803273884e-07, + "loss": 0.6873, + "step": 15976 + }, + { + "epoch": 0.58, + "learning_rate": 1.416742959680668e-07, + "loss": 0.6836, + "step": 15977 + }, + { + "epoch": 0.58, + "learning_rate": 1.4167063390339472e-07, + "loss": 0.6924, + "step": 15978 + }, + { + "epoch": 0.58, + "learning_rate": 1.4166697183872268e-07, + "loss": 0.6928, + "step": 15979 + }, + { + "epoch": 0.58, + "learning_rate": 1.416633097740506e-07, + "loss": 0.7088, + "step": 15980 + }, + { + "epoch": 0.58, + "learning_rate": 1.4165964770937856e-07, + "loss": 0.6936, + "step": 15981 + }, + { + "epoch": 0.58, + "learning_rate": 1.4165598564470648e-07, + "loss": 0.6896, + "step": 15982 + }, + { + "epoch": 0.58, + "learning_rate": 1.4165232358003443e-07, + "loss": 0.6926, + "step": 15983 + }, + { + "epoch": 0.58, + "learning_rate": 1.4164866151536234e-07, + "loss": 0.6925, + "step": 15984 + }, + { + "epoch": 0.58, + "learning_rate": 1.416449994506903e-07, + "loss": 0.687, + "step": 15985 + }, + { + "epoch": 0.58, + "learning_rate": 1.4164133738601821e-07, + "loss": 0.697, + "step": 15986 + }, + { + "epoch": 0.58, + "learning_rate": 1.4163767532134617e-07, + "loss": 0.6875, + "step": 15987 + }, + { + "epoch": 0.58, + "learning_rate": 1.416340132566741e-07, + "loss": 0.6985, + "step": 15988 + }, + { + "epoch": 0.58, + "learning_rate": 1.4163035119200205e-07, + "loss": 0.6931, + "step": 15989 + }, + { + "epoch": 0.59, + "learning_rate": 1.4162668912732997e-07, + "loss": 0.698, + "step": 15990 + }, + { + "epoch": 0.59, + "learning_rate": 1.4162302706265793e-07, + "loss": 0.6841, + "step": 15991 + }, + { + "epoch": 0.59, + "learning_rate": 1.4161936499798585e-07, + "loss": 0.6963, + "step": 15992 + }, + { + "epoch": 0.59, + "learning_rate": 1.416157029333138e-07, + "loss": 0.6937, + "step": 15993 + }, + { + "epoch": 0.59, + "learning_rate": 1.4161204086864173e-07, + "loss": 0.7047, + "step": 15994 + }, + { + "epoch": 0.59, + "learning_rate": 1.416083788039697e-07, + "loss": 0.6779, + "step": 15995 + }, + { + "epoch": 0.59, + "learning_rate": 1.4160471673929761e-07, + "loss": 0.7038, + "step": 15996 + }, + { + "epoch": 0.59, + "learning_rate": 1.4160105467462554e-07, + "loss": 0.6947, + "step": 15997 + }, + { + "epoch": 0.59, + "learning_rate": 1.415973926099535e-07, + "loss": 0.699, + "step": 15998 + }, + { + "epoch": 0.59, + "learning_rate": 1.4159373054528142e-07, + "loss": 0.6924, + "step": 15999 + }, + { + "epoch": 0.59, + "learning_rate": 1.4159006848060937e-07, + "loss": 0.6759, + "step": 16000 + }, + { + "epoch": 0.59, + "learning_rate": 1.415864064159373e-07, + "loss": 0.6798, + "step": 16001 + }, + { + "epoch": 0.59, + "learning_rate": 1.4158274435126523e-07, + "loss": 0.6882, + "step": 16002 + }, + { + "epoch": 0.59, + "learning_rate": 1.4157908228659315e-07, + "loss": 0.6962, + "step": 16003 + }, + { + "epoch": 0.59, + "learning_rate": 1.415754202219211e-07, + "loss": 0.679, + "step": 16004 + }, + { + "epoch": 0.59, + "learning_rate": 1.4157175815724903e-07, + "loss": 0.6955, + "step": 16005 + }, + { + "epoch": 0.59, + "learning_rate": 1.41568096092577e-07, + "loss": 0.7016, + "step": 16006 + }, + { + "epoch": 0.59, + "learning_rate": 1.4156443402790491e-07, + "loss": 0.6901, + "step": 16007 + }, + { + "epoch": 0.59, + "learning_rate": 1.4156077196323287e-07, + "loss": 0.7012, + "step": 16008 + }, + { + "epoch": 0.59, + "learning_rate": 1.415571098985608e-07, + "loss": 0.6927, + "step": 16009 + }, + { + "epoch": 0.59, + "learning_rate": 1.4155344783388875e-07, + "loss": 0.6962, + "step": 16010 + }, + { + "epoch": 0.59, + "learning_rate": 1.4154978576921667e-07, + "loss": 0.6948, + "step": 16011 + }, + { + "epoch": 0.59, + "learning_rate": 1.4154612370454463e-07, + "loss": 0.6934, + "step": 16012 + }, + { + "epoch": 0.59, + "learning_rate": 1.4154246163987255e-07, + "loss": 0.7033, + "step": 16013 + }, + { + "epoch": 0.59, + "learning_rate": 1.415387995752005e-07, + "loss": 0.6881, + "step": 16014 + }, + { + "epoch": 0.59, + "learning_rate": 1.4153513751052843e-07, + "loss": 0.695, + "step": 16015 + }, + { + "epoch": 0.59, + "learning_rate": 1.4153147544585639e-07, + "loss": 0.6873, + "step": 16016 + }, + { + "epoch": 0.59, + "learning_rate": 1.415278133811843e-07, + "loss": 0.6987, + "step": 16017 + }, + { + "epoch": 0.59, + "learning_rate": 1.4152415131651224e-07, + "loss": 0.6787, + "step": 16018 + }, + { + "epoch": 0.59, + "learning_rate": 1.415204892518402e-07, + "loss": 0.7047, + "step": 16019 + }, + { + "epoch": 0.59, + "learning_rate": 1.4151682718716812e-07, + "loss": 0.6861, + "step": 16020 + }, + { + "epoch": 0.59, + "learning_rate": 1.4151316512249605e-07, + "loss": 0.696, + "step": 16021 + }, + { + "epoch": 0.59, + "learning_rate": 1.41509503057824e-07, + "loss": 0.6913, + "step": 16022 + }, + { + "epoch": 0.59, + "learning_rate": 1.4150584099315193e-07, + "loss": 0.6874, + "step": 16023 + }, + { + "epoch": 0.59, + "learning_rate": 1.4150217892847985e-07, + "loss": 0.7005, + "step": 16024 + }, + { + "epoch": 0.59, + "learning_rate": 1.414985168638078e-07, + "loss": 0.6887, + "step": 16025 + }, + { + "epoch": 0.59, + "learning_rate": 1.4149485479913573e-07, + "loss": 0.6893, + "step": 16026 + }, + { + "epoch": 0.59, + "learning_rate": 1.4149119273446369e-07, + "loss": 0.6918, + "step": 16027 + }, + { + "epoch": 0.59, + "learning_rate": 1.414875306697916e-07, + "loss": 0.7048, + "step": 16028 + }, + { + "epoch": 0.59, + "learning_rate": 1.4148386860511957e-07, + "loss": 0.6919, + "step": 16029 + }, + { + "epoch": 0.59, + "learning_rate": 1.414802065404475e-07, + "loss": 0.6727, + "step": 16030 + }, + { + "epoch": 0.59, + "learning_rate": 1.4147654447577545e-07, + "loss": 0.6941, + "step": 16031 + }, + { + "epoch": 0.59, + "learning_rate": 1.4147288241110337e-07, + "loss": 0.6953, + "step": 16032 + }, + { + "epoch": 0.59, + "learning_rate": 1.4146922034643133e-07, + "loss": 0.7043, + "step": 16033 + }, + { + "epoch": 0.59, + "learning_rate": 1.4146555828175925e-07, + "loss": 0.7069, + "step": 16034 + }, + { + "epoch": 0.59, + "learning_rate": 1.414618962170872e-07, + "loss": 0.6897, + "step": 16035 + }, + { + "epoch": 0.59, + "learning_rate": 1.4145823415241513e-07, + "loss": 0.7019, + "step": 16036 + }, + { + "epoch": 0.59, + "learning_rate": 1.4145457208774309e-07, + "loss": 0.7002, + "step": 16037 + }, + { + "epoch": 0.59, + "learning_rate": 1.4145091002307099e-07, + "loss": 0.6831, + "step": 16038 + }, + { + "epoch": 0.59, + "learning_rate": 1.4144724795839894e-07, + "loss": 0.6881, + "step": 16039 + }, + { + "epoch": 0.59, + "learning_rate": 1.4144358589372687e-07, + "loss": 0.6899, + "step": 16040 + }, + { + "epoch": 0.59, + "learning_rate": 1.4143992382905482e-07, + "loss": 0.6832, + "step": 16041 + }, + { + "epoch": 0.59, + "learning_rate": 1.4143626176438275e-07, + "loss": 0.6879, + "step": 16042 + }, + { + "epoch": 0.59, + "learning_rate": 1.414325996997107e-07, + "loss": 0.7017, + "step": 16043 + }, + { + "epoch": 0.59, + "learning_rate": 1.4142893763503862e-07, + "loss": 0.6937, + "step": 16044 + }, + { + "epoch": 0.59, + "learning_rate": 1.4142527557036655e-07, + "loss": 0.6942, + "step": 16045 + }, + { + "epoch": 0.59, + "learning_rate": 1.414216135056945e-07, + "loss": 0.6948, + "step": 16046 + }, + { + "epoch": 0.59, + "learning_rate": 1.4141795144102243e-07, + "loss": 0.6911, + "step": 16047 + }, + { + "epoch": 0.59, + "learning_rate": 1.4141428937635038e-07, + "loss": 0.6767, + "step": 16048 + }, + { + "epoch": 0.59, + "learning_rate": 1.414106273116783e-07, + "loss": 0.6908, + "step": 16049 + }, + { + "epoch": 0.59, + "learning_rate": 1.4140696524700626e-07, + "loss": 0.7066, + "step": 16050 + }, + { + "epoch": 0.59, + "learning_rate": 1.414033031823342e-07, + "loss": 0.6933, + "step": 16051 + }, + { + "epoch": 0.59, + "learning_rate": 1.4139964111766214e-07, + "loss": 0.6953, + "step": 16052 + }, + { + "epoch": 0.59, + "learning_rate": 1.4139597905299007e-07, + "loss": 0.7034, + "step": 16053 + }, + { + "epoch": 0.59, + "learning_rate": 1.4139231698831802e-07, + "loss": 0.7067, + "step": 16054 + }, + { + "epoch": 0.59, + "learning_rate": 1.4138865492364595e-07, + "loss": 0.696, + "step": 16055 + }, + { + "epoch": 0.59, + "learning_rate": 1.4138499285897388e-07, + "loss": 0.7072, + "step": 16056 + }, + { + "epoch": 0.59, + "learning_rate": 1.413813307943018e-07, + "loss": 0.6977, + "step": 16057 + }, + { + "epoch": 0.59, + "learning_rate": 1.4137766872962976e-07, + "loss": 0.693, + "step": 16058 + }, + { + "epoch": 0.59, + "learning_rate": 1.4137400666495768e-07, + "loss": 0.693, + "step": 16059 + }, + { + "epoch": 0.59, + "learning_rate": 1.4137034460028564e-07, + "loss": 0.7008, + "step": 16060 + }, + { + "epoch": 0.59, + "learning_rate": 1.4136668253561356e-07, + "loss": 0.7078, + "step": 16061 + }, + { + "epoch": 0.59, + "learning_rate": 1.4136302047094152e-07, + "loss": 0.693, + "step": 16062 + }, + { + "epoch": 0.59, + "learning_rate": 1.4135935840626944e-07, + "loss": 0.6955, + "step": 16063 + }, + { + "epoch": 0.59, + "learning_rate": 1.413556963415974e-07, + "loss": 0.6995, + "step": 16064 + }, + { + "epoch": 0.59, + "learning_rate": 1.4135203427692532e-07, + "loss": 0.6922, + "step": 16065 + }, + { + "epoch": 0.59, + "learning_rate": 1.4134837221225328e-07, + "loss": 0.687, + "step": 16066 + }, + { + "epoch": 0.59, + "learning_rate": 1.413447101475812e-07, + "loss": 0.6977, + "step": 16067 + }, + { + "epoch": 0.59, + "learning_rate": 1.4134104808290913e-07, + "loss": 0.6983, + "step": 16068 + }, + { + "epoch": 0.59, + "learning_rate": 1.4133738601823708e-07, + "loss": 0.6998, + "step": 16069 + }, + { + "epoch": 0.59, + "learning_rate": 1.41333723953565e-07, + "loss": 0.6951, + "step": 16070 + }, + { + "epoch": 0.59, + "learning_rate": 1.4133006188889296e-07, + "loss": 0.6707, + "step": 16071 + }, + { + "epoch": 0.59, + "learning_rate": 1.413263998242209e-07, + "loss": 0.703, + "step": 16072 + }, + { + "epoch": 0.59, + "learning_rate": 1.4132273775954884e-07, + "loss": 0.6907, + "step": 16073 + }, + { + "epoch": 0.59, + "learning_rate": 1.4131907569487674e-07, + "loss": 0.6929, + "step": 16074 + }, + { + "epoch": 0.59, + "learning_rate": 1.413154136302047e-07, + "loss": 0.6772, + "step": 16075 + }, + { + "epoch": 0.59, + "learning_rate": 1.4131175156553262e-07, + "loss": 0.6893, + "step": 16076 + }, + { + "epoch": 0.59, + "learning_rate": 1.4130808950086058e-07, + "loss": 0.6746, + "step": 16077 + }, + { + "epoch": 0.59, + "learning_rate": 1.413044274361885e-07, + "loss": 0.6779, + "step": 16078 + }, + { + "epoch": 0.59, + "learning_rate": 1.4130076537151646e-07, + "loss": 0.6978, + "step": 16079 + }, + { + "epoch": 0.59, + "learning_rate": 1.4129710330684438e-07, + "loss": 0.688, + "step": 16080 + }, + { + "epoch": 0.59, + "learning_rate": 1.4129344124217234e-07, + "loss": 0.6936, + "step": 16081 + }, + { + "epoch": 0.59, + "learning_rate": 1.4128977917750026e-07, + "loss": 0.6892, + "step": 16082 + }, + { + "epoch": 0.59, + "learning_rate": 1.4128611711282822e-07, + "loss": 0.7001, + "step": 16083 + }, + { + "epoch": 0.59, + "learning_rate": 1.4128245504815614e-07, + "loss": 0.6872, + "step": 16084 + }, + { + "epoch": 0.59, + "learning_rate": 1.412787929834841e-07, + "loss": 0.6953, + "step": 16085 + }, + { + "epoch": 0.59, + "learning_rate": 1.4127513091881202e-07, + "loss": 0.7009, + "step": 16086 + }, + { + "epoch": 0.59, + "learning_rate": 1.4127146885413998e-07, + "loss": 0.6913, + "step": 16087 + }, + { + "epoch": 0.59, + "learning_rate": 1.412678067894679e-07, + "loss": 0.6951, + "step": 16088 + }, + { + "epoch": 0.59, + "learning_rate": 1.4126414472479583e-07, + "loss": 0.6997, + "step": 16089 + }, + { + "epoch": 0.59, + "learning_rate": 1.4126048266012378e-07, + "loss": 0.6951, + "step": 16090 + }, + { + "epoch": 0.59, + "learning_rate": 1.412568205954517e-07, + "loss": 0.6922, + "step": 16091 + }, + { + "epoch": 0.59, + "learning_rate": 1.4125315853077964e-07, + "loss": 0.7005, + "step": 16092 + }, + { + "epoch": 0.59, + "learning_rate": 1.412494964661076e-07, + "loss": 0.7076, + "step": 16093 + }, + { + "epoch": 0.59, + "learning_rate": 1.4124583440143552e-07, + "loss": 0.6976, + "step": 16094 + }, + { + "epoch": 0.59, + "learning_rate": 1.4124217233676344e-07, + "loss": 0.6804, + "step": 16095 + }, + { + "epoch": 0.59, + "learning_rate": 1.412385102720914e-07, + "loss": 0.6928, + "step": 16096 + }, + { + "epoch": 0.59, + "learning_rate": 1.4123484820741932e-07, + "loss": 0.7086, + "step": 16097 + }, + { + "epoch": 0.59, + "learning_rate": 1.4123118614274728e-07, + "loss": 0.695, + "step": 16098 + }, + { + "epoch": 0.59, + "learning_rate": 1.412275240780752e-07, + "loss": 0.7011, + "step": 16099 + }, + { + "epoch": 0.59, + "learning_rate": 1.4122386201340316e-07, + "loss": 0.6962, + "step": 16100 + }, + { + "epoch": 0.59, + "learning_rate": 1.4122019994873108e-07, + "loss": 0.6936, + "step": 16101 + }, + { + "epoch": 0.59, + "learning_rate": 1.4121653788405903e-07, + "loss": 0.6995, + "step": 16102 + }, + { + "epoch": 0.59, + "learning_rate": 1.4121287581938696e-07, + "loss": 0.6825, + "step": 16103 + }, + { + "epoch": 0.59, + "learning_rate": 1.4120921375471491e-07, + "loss": 0.6861, + "step": 16104 + }, + { + "epoch": 0.59, + "learning_rate": 1.4120555169004284e-07, + "loss": 0.7028, + "step": 16105 + }, + { + "epoch": 0.59, + "learning_rate": 1.412018896253708e-07, + "loss": 0.6961, + "step": 16106 + }, + { + "epoch": 0.59, + "learning_rate": 1.4119822756069872e-07, + "loss": 0.6962, + "step": 16107 + }, + { + "epoch": 0.59, + "learning_rate": 1.4119456549602667e-07, + "loss": 0.6989, + "step": 16108 + }, + { + "epoch": 0.59, + "learning_rate": 1.411909034313546e-07, + "loss": 0.6953, + "step": 16109 + }, + { + "epoch": 0.59, + "learning_rate": 1.4118724136668253e-07, + "loss": 0.6877, + "step": 16110 + }, + { + "epoch": 0.59, + "learning_rate": 1.4118357930201045e-07, + "loss": 0.6948, + "step": 16111 + }, + { + "epoch": 0.59, + "learning_rate": 1.411799172373384e-07, + "loss": 0.6901, + "step": 16112 + }, + { + "epoch": 0.59, + "learning_rate": 1.4117625517266633e-07, + "loss": 0.6901, + "step": 16113 + }, + { + "epoch": 0.59, + "learning_rate": 1.411725931079943e-07, + "loss": 0.6963, + "step": 16114 + }, + { + "epoch": 0.59, + "learning_rate": 1.4116893104332221e-07, + "loss": 0.697, + "step": 16115 + }, + { + "epoch": 0.59, + "learning_rate": 1.4116526897865014e-07, + "loss": 0.7035, + "step": 16116 + }, + { + "epoch": 0.59, + "learning_rate": 1.411616069139781e-07, + "loss": 0.6948, + "step": 16117 + }, + { + "epoch": 0.59, + "learning_rate": 1.4115794484930602e-07, + "loss": 0.6864, + "step": 16118 + }, + { + "epoch": 0.59, + "learning_rate": 1.4115428278463397e-07, + "loss": 0.686, + "step": 16119 + }, + { + "epoch": 0.59, + "learning_rate": 1.411506207199619e-07, + "loss": 0.6944, + "step": 16120 + }, + { + "epoch": 0.59, + "learning_rate": 1.4114695865528985e-07, + "loss": 0.6804, + "step": 16121 + }, + { + "epoch": 0.59, + "learning_rate": 1.4114329659061778e-07, + "loss": 0.7, + "step": 16122 + }, + { + "epoch": 0.59, + "learning_rate": 1.4113963452594573e-07, + "loss": 0.6849, + "step": 16123 + }, + { + "epoch": 0.59, + "learning_rate": 1.4113597246127366e-07, + "loss": 0.6862, + "step": 16124 + }, + { + "epoch": 0.59, + "learning_rate": 1.4113231039660161e-07, + "loss": 0.6806, + "step": 16125 + }, + { + "epoch": 0.59, + "learning_rate": 1.4112864833192954e-07, + "loss": 0.6862, + "step": 16126 + }, + { + "epoch": 0.59, + "learning_rate": 1.411249862672575e-07, + "loss": 0.6799, + "step": 16127 + }, + { + "epoch": 0.59, + "learning_rate": 1.411213242025854e-07, + "loss": 0.6858, + "step": 16128 + }, + { + "epoch": 0.59, + "learning_rate": 1.4111766213791335e-07, + "loss": 0.6921, + "step": 16129 + }, + { + "epoch": 0.59, + "learning_rate": 1.4111400007324127e-07, + "loss": 0.6989, + "step": 16130 + }, + { + "epoch": 0.59, + "learning_rate": 1.4111033800856923e-07, + "loss": 0.6892, + "step": 16131 + }, + { + "epoch": 0.59, + "learning_rate": 1.4110667594389715e-07, + "loss": 0.699, + "step": 16132 + }, + { + "epoch": 0.59, + "learning_rate": 1.411030138792251e-07, + "loss": 0.6871, + "step": 16133 + }, + { + "epoch": 0.59, + "learning_rate": 1.4109935181455303e-07, + "loss": 0.7023, + "step": 16134 + }, + { + "epoch": 0.59, + "learning_rate": 1.4109568974988099e-07, + "loss": 0.6927, + "step": 16135 + }, + { + "epoch": 0.59, + "learning_rate": 1.410920276852089e-07, + "loss": 0.6872, + "step": 16136 + }, + { + "epoch": 0.59, + "learning_rate": 1.4108836562053687e-07, + "loss": 0.6984, + "step": 16137 + }, + { + "epoch": 0.59, + "learning_rate": 1.410847035558648e-07, + "loss": 0.6921, + "step": 16138 + }, + { + "epoch": 0.59, + "learning_rate": 1.4108104149119272e-07, + "loss": 0.7006, + "step": 16139 + }, + { + "epoch": 0.59, + "learning_rate": 1.4107737942652067e-07, + "loss": 0.683, + "step": 16140 + }, + { + "epoch": 0.59, + "learning_rate": 1.410737173618486e-07, + "loss": 0.7076, + "step": 16141 + }, + { + "epoch": 0.59, + "learning_rate": 1.4107005529717655e-07, + "loss": 0.6875, + "step": 16142 + }, + { + "epoch": 0.59, + "learning_rate": 1.4106639323250448e-07, + "loss": 0.6885, + "step": 16143 + }, + { + "epoch": 0.59, + "learning_rate": 1.4106273116783243e-07, + "loss": 0.6882, + "step": 16144 + }, + { + "epoch": 0.59, + "learning_rate": 1.4105906910316036e-07, + "loss": 0.6945, + "step": 16145 + }, + { + "epoch": 0.59, + "learning_rate": 1.4105540703848829e-07, + "loss": 0.6874, + "step": 16146 + }, + { + "epoch": 0.59, + "learning_rate": 1.410517449738162e-07, + "loss": 0.6933, + "step": 16147 + }, + { + "epoch": 0.59, + "learning_rate": 1.4104808290914417e-07, + "loss": 0.6845, + "step": 16148 + }, + { + "epoch": 0.59, + "learning_rate": 1.410444208444721e-07, + "loss": 0.7045, + "step": 16149 + }, + { + "epoch": 0.59, + "learning_rate": 1.4104075877980005e-07, + "loss": 0.7037, + "step": 16150 + }, + { + "epoch": 0.59, + "learning_rate": 1.4103709671512797e-07, + "loss": 0.6965, + "step": 16151 + }, + { + "epoch": 0.59, + "learning_rate": 1.4103343465045593e-07, + "loss": 0.6844, + "step": 16152 + }, + { + "epoch": 0.59, + "learning_rate": 1.4102977258578385e-07, + "loss": 0.684, + "step": 16153 + }, + { + "epoch": 0.59, + "learning_rate": 1.410261105211118e-07, + "loss": 0.6976, + "step": 16154 + }, + { + "epoch": 0.59, + "learning_rate": 1.4102244845643973e-07, + "loss": 0.6946, + "step": 16155 + }, + { + "epoch": 0.59, + "learning_rate": 1.4101878639176769e-07, + "loss": 0.6928, + "step": 16156 + }, + { + "epoch": 0.59, + "learning_rate": 1.410151243270956e-07, + "loss": 0.6887, + "step": 16157 + }, + { + "epoch": 0.59, + "learning_rate": 1.4101146226242357e-07, + "loss": 0.7053, + "step": 16158 + }, + { + "epoch": 0.59, + "learning_rate": 1.410078001977515e-07, + "loss": 0.6988, + "step": 16159 + }, + { + "epoch": 0.59, + "learning_rate": 1.4100413813307942e-07, + "loss": 0.6898, + "step": 16160 + }, + { + "epoch": 0.59, + "learning_rate": 1.4100047606840737e-07, + "loss": 0.6797, + "step": 16161 + }, + { + "epoch": 0.59, + "learning_rate": 1.409968140037353e-07, + "loss": 0.6845, + "step": 16162 + }, + { + "epoch": 0.59, + "learning_rate": 1.4099315193906325e-07, + "loss": 0.7081, + "step": 16163 + }, + { + "epoch": 0.59, + "learning_rate": 1.4098948987439118e-07, + "loss": 0.6888, + "step": 16164 + }, + { + "epoch": 0.59, + "learning_rate": 1.409858278097191e-07, + "loss": 0.6989, + "step": 16165 + }, + { + "epoch": 0.59, + "learning_rate": 1.4098216574504703e-07, + "loss": 0.6984, + "step": 16166 + }, + { + "epoch": 0.59, + "learning_rate": 1.4097850368037498e-07, + "loss": 0.6933, + "step": 16167 + }, + { + "epoch": 0.59, + "learning_rate": 1.409748416157029e-07, + "loss": 0.6918, + "step": 16168 + }, + { + "epoch": 0.59, + "learning_rate": 1.4097117955103086e-07, + "loss": 0.697, + "step": 16169 + }, + { + "epoch": 0.59, + "learning_rate": 1.409675174863588e-07, + "loss": 0.7018, + "step": 16170 + }, + { + "epoch": 0.59, + "learning_rate": 1.4096385542168674e-07, + "loss": 0.7046, + "step": 16171 + }, + { + "epoch": 0.59, + "learning_rate": 1.4096019335701467e-07, + "loss": 0.687, + "step": 16172 + }, + { + "epoch": 0.59, + "learning_rate": 1.4095653129234262e-07, + "loss": 0.7099, + "step": 16173 + }, + { + "epoch": 0.59, + "learning_rate": 1.4095286922767055e-07, + "loss": 0.6887, + "step": 16174 + }, + { + "epoch": 0.59, + "learning_rate": 1.409492071629985e-07, + "loss": 0.6801, + "step": 16175 + }, + { + "epoch": 0.59, + "learning_rate": 1.4094554509832643e-07, + "loss": 0.6957, + "step": 16176 + }, + { + "epoch": 0.59, + "learning_rate": 1.4094188303365438e-07, + "loss": 0.6947, + "step": 16177 + }, + { + "epoch": 0.59, + "learning_rate": 1.409382209689823e-07, + "loss": 0.6927, + "step": 16178 + }, + { + "epoch": 0.59, + "learning_rate": 1.4093455890431026e-07, + "loss": 0.6957, + "step": 16179 + }, + { + "epoch": 0.59, + "learning_rate": 1.409308968396382e-07, + "loss": 0.7007, + "step": 16180 + }, + { + "epoch": 0.59, + "learning_rate": 1.4092723477496614e-07, + "loss": 0.701, + "step": 16181 + }, + { + "epoch": 0.59, + "learning_rate": 1.4092357271029404e-07, + "loss": 0.7034, + "step": 16182 + }, + { + "epoch": 0.59, + "learning_rate": 1.40919910645622e-07, + "loss": 0.6836, + "step": 16183 + }, + { + "epoch": 0.59, + "learning_rate": 1.4091624858094992e-07, + "loss": 0.6863, + "step": 16184 + }, + { + "epoch": 0.59, + "learning_rate": 1.4091258651627788e-07, + "loss": 0.696, + "step": 16185 + }, + { + "epoch": 0.59, + "learning_rate": 1.409089244516058e-07, + "loss": 0.693, + "step": 16186 + }, + { + "epoch": 0.59, + "learning_rate": 1.4090526238693373e-07, + "loss": 0.6918, + "step": 16187 + }, + { + "epoch": 0.59, + "learning_rate": 1.4090160032226168e-07, + "loss": 0.6884, + "step": 16188 + }, + { + "epoch": 0.59, + "learning_rate": 1.408979382575896e-07, + "loss": 0.6878, + "step": 16189 + }, + { + "epoch": 0.59, + "learning_rate": 1.4089427619291756e-07, + "loss": 0.691, + "step": 16190 + }, + { + "epoch": 0.59, + "learning_rate": 1.408906141282455e-07, + "loss": 0.6853, + "step": 16191 + }, + { + "epoch": 0.59, + "learning_rate": 1.4088695206357344e-07, + "loss": 0.6859, + "step": 16192 + }, + { + "epoch": 0.59, + "learning_rate": 1.4088328999890137e-07, + "loss": 0.6987, + "step": 16193 + }, + { + "epoch": 0.59, + "learning_rate": 1.4087962793422932e-07, + "loss": 0.6982, + "step": 16194 + }, + { + "epoch": 0.59, + "learning_rate": 1.4087596586955725e-07, + "loss": 0.6911, + "step": 16195 + }, + { + "epoch": 0.59, + "learning_rate": 1.408723038048852e-07, + "loss": 0.7003, + "step": 16196 + }, + { + "epoch": 0.59, + "learning_rate": 1.4086864174021313e-07, + "loss": 0.6996, + "step": 16197 + }, + { + "epoch": 0.59, + "learning_rate": 1.4086497967554108e-07, + "loss": 0.688, + "step": 16198 + }, + { + "epoch": 0.59, + "learning_rate": 1.40861317610869e-07, + "loss": 0.6939, + "step": 16199 + }, + { + "epoch": 0.59, + "learning_rate": 1.4085765554619694e-07, + "loss": 0.678, + "step": 16200 + }, + { + "epoch": 0.59, + "learning_rate": 1.4085399348152486e-07, + "loss": 0.6774, + "step": 16201 + }, + { + "epoch": 0.59, + "learning_rate": 1.4085033141685282e-07, + "loss": 0.687, + "step": 16202 + }, + { + "epoch": 0.59, + "learning_rate": 1.4084666935218074e-07, + "loss": 0.7077, + "step": 16203 + }, + { + "epoch": 0.59, + "learning_rate": 1.408430072875087e-07, + "loss": 0.6943, + "step": 16204 + }, + { + "epoch": 0.59, + "learning_rate": 1.4083934522283662e-07, + "loss": 0.6886, + "step": 16205 + }, + { + "epoch": 0.59, + "learning_rate": 1.4083568315816458e-07, + "loss": 0.6937, + "step": 16206 + }, + { + "epoch": 0.59, + "learning_rate": 1.408320210934925e-07, + "loss": 0.704, + "step": 16207 + }, + { + "epoch": 0.59, + "learning_rate": 1.4082835902882046e-07, + "loss": 0.703, + "step": 16208 + }, + { + "epoch": 0.59, + "learning_rate": 1.4082469696414838e-07, + "loss": 0.688, + "step": 16209 + }, + { + "epoch": 0.59, + "learning_rate": 1.408210348994763e-07, + "loss": 0.6883, + "step": 16210 + }, + { + "epoch": 0.59, + "learning_rate": 1.4081737283480426e-07, + "loss": 0.6921, + "step": 16211 + }, + { + "epoch": 0.59, + "learning_rate": 1.408137107701322e-07, + "loss": 0.6989, + "step": 16212 + }, + { + "epoch": 0.59, + "learning_rate": 1.4081004870546014e-07, + "loss": 0.6869, + "step": 16213 + }, + { + "epoch": 0.59, + "learning_rate": 1.4080638664078807e-07, + "loss": 0.6973, + "step": 16214 + }, + { + "epoch": 0.59, + "learning_rate": 1.4080272457611602e-07, + "loss": 0.6927, + "step": 16215 + }, + { + "epoch": 0.59, + "learning_rate": 1.4079906251144395e-07, + "loss": 0.689, + "step": 16216 + }, + { + "epoch": 0.59, + "learning_rate": 1.4079540044677188e-07, + "loss": 0.6974, + "step": 16217 + }, + { + "epoch": 0.59, + "learning_rate": 1.407917383820998e-07, + "loss": 0.6966, + "step": 16218 + }, + { + "epoch": 0.59, + "learning_rate": 1.4078807631742776e-07, + "loss": 0.6911, + "step": 16219 + }, + { + "epoch": 0.59, + "learning_rate": 1.4078441425275568e-07, + "loss": 0.6884, + "step": 16220 + }, + { + "epoch": 0.59, + "learning_rate": 1.4078075218808363e-07, + "loss": 0.6892, + "step": 16221 + }, + { + "epoch": 0.59, + "learning_rate": 1.4077709012341156e-07, + "loss": 0.6858, + "step": 16222 + }, + { + "epoch": 0.59, + "learning_rate": 1.4077342805873951e-07, + "loss": 0.6984, + "step": 16223 + }, + { + "epoch": 0.59, + "learning_rate": 1.4076976599406744e-07, + "loss": 0.6849, + "step": 16224 + }, + { + "epoch": 0.59, + "learning_rate": 1.407661039293954e-07, + "loss": 0.6886, + "step": 16225 + }, + { + "epoch": 0.59, + "learning_rate": 1.4076244186472332e-07, + "loss": 0.6653, + "step": 16226 + }, + { + "epoch": 0.59, + "learning_rate": 1.4075877980005127e-07, + "loss": 0.6971, + "step": 16227 + }, + { + "epoch": 0.59, + "learning_rate": 1.407551177353792e-07, + "loss": 0.696, + "step": 16228 + }, + { + "epoch": 0.59, + "learning_rate": 1.4075145567070715e-07, + "loss": 0.6873, + "step": 16229 + }, + { + "epoch": 0.59, + "learning_rate": 1.4074779360603508e-07, + "loss": 0.6965, + "step": 16230 + }, + { + "epoch": 0.59, + "learning_rate": 1.40744131541363e-07, + "loss": 0.7004, + "step": 16231 + }, + { + "epoch": 0.59, + "learning_rate": 1.4074046947669096e-07, + "loss": 0.6931, + "step": 16232 + }, + { + "epoch": 0.59, + "learning_rate": 1.407368074120189e-07, + "loss": 0.6867, + "step": 16233 + }, + { + "epoch": 0.59, + "learning_rate": 1.4073314534734684e-07, + "loss": 0.686, + "step": 16234 + }, + { + "epoch": 0.59, + "learning_rate": 1.4072948328267477e-07, + "loss": 0.6948, + "step": 16235 + }, + { + "epoch": 0.59, + "learning_rate": 1.407258212180027e-07, + "loss": 0.695, + "step": 16236 + }, + { + "epoch": 0.59, + "learning_rate": 1.4072215915333062e-07, + "loss": 0.6962, + "step": 16237 + }, + { + "epoch": 0.59, + "learning_rate": 1.4071849708865857e-07, + "loss": 0.6888, + "step": 16238 + }, + { + "epoch": 0.59, + "learning_rate": 1.407148350239865e-07, + "loss": 0.6917, + "step": 16239 + }, + { + "epoch": 0.59, + "learning_rate": 1.4071117295931445e-07, + "loss": 0.6814, + "step": 16240 + }, + { + "epoch": 0.59, + "learning_rate": 1.4070751089464238e-07, + "loss": 0.6955, + "step": 16241 + }, + { + "epoch": 0.59, + "learning_rate": 1.4070384882997033e-07, + "loss": 0.6936, + "step": 16242 + }, + { + "epoch": 0.59, + "learning_rate": 1.4070018676529826e-07, + "loss": 0.6959, + "step": 16243 + }, + { + "epoch": 0.59, + "learning_rate": 1.4069652470062621e-07, + "loss": 0.6983, + "step": 16244 + }, + { + "epoch": 0.59, + "learning_rate": 1.4069286263595414e-07, + "loss": 0.6913, + "step": 16245 + }, + { + "epoch": 0.59, + "learning_rate": 1.406892005712821e-07, + "loss": 0.6993, + "step": 16246 + }, + { + "epoch": 0.59, + "learning_rate": 1.4068553850661002e-07, + "loss": 0.6991, + "step": 16247 + }, + { + "epoch": 0.59, + "learning_rate": 1.4068187644193797e-07, + "loss": 0.6881, + "step": 16248 + }, + { + "epoch": 0.59, + "learning_rate": 1.406782143772659e-07, + "loss": 0.6802, + "step": 16249 + }, + { + "epoch": 0.59, + "learning_rate": 1.4067455231259385e-07, + "loss": 0.6917, + "step": 16250 + }, + { + "epoch": 0.59, + "learning_rate": 1.4067089024792178e-07, + "loss": 0.6974, + "step": 16251 + }, + { + "epoch": 0.59, + "learning_rate": 1.4066722818324973e-07, + "loss": 0.6901, + "step": 16252 + }, + { + "epoch": 0.59, + "learning_rate": 1.4066356611857763e-07, + "loss": 0.6885, + "step": 16253 + }, + { + "epoch": 0.59, + "learning_rate": 1.4065990405390559e-07, + "loss": 0.701, + "step": 16254 + }, + { + "epoch": 0.59, + "learning_rate": 1.406562419892335e-07, + "loss": 0.6852, + "step": 16255 + }, + { + "epoch": 0.59, + "learning_rate": 1.4065257992456147e-07, + "loss": 0.691, + "step": 16256 + }, + { + "epoch": 0.59, + "learning_rate": 1.406489178598894e-07, + "loss": 0.6889, + "step": 16257 + }, + { + "epoch": 0.59, + "learning_rate": 1.4064525579521732e-07, + "loss": 0.6939, + "step": 16258 + }, + { + "epoch": 0.59, + "learning_rate": 1.4064159373054527e-07, + "loss": 0.6941, + "step": 16259 + }, + { + "epoch": 0.59, + "learning_rate": 1.406379316658732e-07, + "loss": 0.6933, + "step": 16260 + }, + { + "epoch": 0.59, + "learning_rate": 1.4063426960120115e-07, + "loss": 0.6962, + "step": 16261 + }, + { + "epoch": 0.59, + "learning_rate": 1.4063060753652908e-07, + "loss": 0.699, + "step": 16262 + }, + { + "epoch": 0.6, + "learning_rate": 1.4062694547185703e-07, + "loss": 0.6874, + "step": 16263 + }, + { + "epoch": 0.6, + "learning_rate": 1.4062328340718496e-07, + "loss": 0.7027, + "step": 16264 + }, + { + "epoch": 0.6, + "learning_rate": 1.406196213425129e-07, + "loss": 0.6761, + "step": 16265 + }, + { + "epoch": 0.6, + "learning_rate": 1.4061595927784084e-07, + "loss": 0.6908, + "step": 16266 + }, + { + "epoch": 0.6, + "learning_rate": 1.406122972131688e-07, + "loss": 0.7067, + "step": 16267 + }, + { + "epoch": 0.6, + "learning_rate": 1.4060863514849672e-07, + "loss": 0.6984, + "step": 16268 + }, + { + "epoch": 0.6, + "learning_rate": 1.4060497308382467e-07, + "loss": 0.6979, + "step": 16269 + }, + { + "epoch": 0.6, + "learning_rate": 1.406013110191526e-07, + "loss": 0.7031, + "step": 16270 + }, + { + "epoch": 0.6, + "learning_rate": 1.4059764895448053e-07, + "loss": 0.6884, + "step": 16271 + }, + { + "epoch": 0.6, + "learning_rate": 1.4059398688980845e-07, + "loss": 0.6919, + "step": 16272 + }, + { + "epoch": 0.6, + "learning_rate": 1.405903248251364e-07, + "loss": 0.7045, + "step": 16273 + }, + { + "epoch": 0.6, + "learning_rate": 1.4058666276046433e-07, + "loss": 0.6852, + "step": 16274 + }, + { + "epoch": 0.6, + "learning_rate": 1.4058300069579229e-07, + "loss": 0.6989, + "step": 16275 + }, + { + "epoch": 0.6, + "learning_rate": 1.405793386311202e-07, + "loss": 0.7077, + "step": 16276 + }, + { + "epoch": 0.6, + "learning_rate": 1.4057567656644817e-07, + "loss": 0.6907, + "step": 16277 + }, + { + "epoch": 0.6, + "learning_rate": 1.405720145017761e-07, + "loss": 0.6903, + "step": 16278 + }, + { + "epoch": 0.6, + "learning_rate": 1.4056835243710404e-07, + "loss": 0.6991, + "step": 16279 + }, + { + "epoch": 0.6, + "learning_rate": 1.4056469037243197e-07, + "loss": 0.6941, + "step": 16280 + }, + { + "epoch": 0.6, + "learning_rate": 1.405610283077599e-07, + "loss": 0.6802, + "step": 16281 + }, + { + "epoch": 0.6, + "learning_rate": 1.4055736624308785e-07, + "loss": 0.698, + "step": 16282 + }, + { + "epoch": 0.6, + "learning_rate": 1.4055370417841578e-07, + "loss": 0.6812, + "step": 16283 + }, + { + "epoch": 0.6, + "learning_rate": 1.4055004211374373e-07, + "loss": 0.7017, + "step": 16284 + }, + { + "epoch": 0.6, + "learning_rate": 1.4054638004907166e-07, + "loss": 0.6837, + "step": 16285 + }, + { + "epoch": 0.6, + "learning_rate": 1.405427179843996e-07, + "loss": 0.7028, + "step": 16286 + }, + { + "epoch": 0.6, + "learning_rate": 1.4053905591972754e-07, + "loss": 0.7019, + "step": 16287 + }, + { + "epoch": 0.6, + "learning_rate": 1.405353938550555e-07, + "loss": 0.6926, + "step": 16288 + }, + { + "epoch": 0.6, + "learning_rate": 1.405317317903834e-07, + "loss": 0.6996, + "step": 16289 + }, + { + "epoch": 0.6, + "learning_rate": 1.4052806972571134e-07, + "loss": 0.6777, + "step": 16290 + }, + { + "epoch": 0.6, + "learning_rate": 1.4052440766103927e-07, + "loss": 0.672, + "step": 16291 + }, + { + "epoch": 0.6, + "learning_rate": 1.4052074559636722e-07, + "loss": 0.6897, + "step": 16292 + }, + { + "epoch": 0.6, + "learning_rate": 1.4051708353169515e-07, + "loss": 0.6748, + "step": 16293 + }, + { + "epoch": 0.6, + "learning_rate": 1.405134214670231e-07, + "loss": 0.7004, + "step": 16294 + }, + { + "epoch": 0.6, + "learning_rate": 1.4050975940235103e-07, + "loss": 0.6844, + "step": 16295 + }, + { + "epoch": 0.6, + "learning_rate": 1.4050609733767898e-07, + "loss": 0.6885, + "step": 16296 + }, + { + "epoch": 0.6, + "learning_rate": 1.405024352730069e-07, + "loss": 0.6945, + "step": 16297 + }, + { + "epoch": 0.6, + "learning_rate": 1.4049877320833486e-07, + "loss": 0.6888, + "step": 16298 + }, + { + "epoch": 0.6, + "learning_rate": 1.404951111436628e-07, + "loss": 0.7, + "step": 16299 + }, + { + "epoch": 0.6, + "learning_rate": 1.4049144907899074e-07, + "loss": 0.6898, + "step": 16300 + }, + { + "epoch": 0.6, + "learning_rate": 1.4048778701431867e-07, + "loss": 0.6912, + "step": 16301 + }, + { + "epoch": 0.6, + "learning_rate": 1.404841249496466e-07, + "loss": 0.689, + "step": 16302 + }, + { + "epoch": 0.6, + "learning_rate": 1.4048046288497455e-07, + "loss": 0.6941, + "step": 16303 + }, + { + "epoch": 0.6, + "learning_rate": 1.4047680082030248e-07, + "loss": 0.6857, + "step": 16304 + }, + { + "epoch": 0.6, + "learning_rate": 1.4047313875563043e-07, + "loss": 0.6899, + "step": 16305 + }, + { + "epoch": 0.6, + "learning_rate": 1.4046947669095836e-07, + "loss": 0.689, + "step": 16306 + }, + { + "epoch": 0.6, + "learning_rate": 1.4046581462628628e-07, + "loss": 0.695, + "step": 16307 + }, + { + "epoch": 0.6, + "learning_rate": 1.404621525616142e-07, + "loss": 0.6933, + "step": 16308 + }, + { + "epoch": 0.6, + "learning_rate": 1.4045849049694216e-07, + "loss": 0.6857, + "step": 16309 + }, + { + "epoch": 0.6, + "learning_rate": 1.404548284322701e-07, + "loss": 0.6945, + "step": 16310 + }, + { + "epoch": 0.6, + "learning_rate": 1.4045116636759804e-07, + "loss": 0.7123, + "step": 16311 + }, + { + "epoch": 0.6, + "learning_rate": 1.4044750430292597e-07, + "loss": 0.6961, + "step": 16312 + }, + { + "epoch": 0.6, + "learning_rate": 1.4044384223825392e-07, + "loss": 0.7067, + "step": 16313 + }, + { + "epoch": 0.6, + "learning_rate": 1.4044018017358185e-07, + "loss": 0.6995, + "step": 16314 + }, + { + "epoch": 0.6, + "learning_rate": 1.404365181089098e-07, + "loss": 0.7024, + "step": 16315 + }, + { + "epoch": 0.6, + "learning_rate": 1.4043285604423773e-07, + "loss": 0.692, + "step": 16316 + }, + { + "epoch": 0.6, + "learning_rate": 1.4042919397956568e-07, + "loss": 0.6899, + "step": 16317 + }, + { + "epoch": 0.6, + "learning_rate": 1.404255319148936e-07, + "loss": 0.688, + "step": 16318 + }, + { + "epoch": 0.6, + "learning_rate": 1.4042186985022156e-07, + "loss": 0.6993, + "step": 16319 + }, + { + "epoch": 0.6, + "learning_rate": 1.404182077855495e-07, + "loss": 0.6938, + "step": 16320 + }, + { + "epoch": 0.6, + "learning_rate": 1.4041454572087744e-07, + "loss": 0.6891, + "step": 16321 + }, + { + "epoch": 0.6, + "learning_rate": 1.4041088365620537e-07, + "loss": 0.6953, + "step": 16322 + }, + { + "epoch": 0.6, + "learning_rate": 1.404072215915333e-07, + "loss": 0.6791, + "step": 16323 + }, + { + "epoch": 0.6, + "learning_rate": 1.4040355952686125e-07, + "loss": 0.7006, + "step": 16324 + }, + { + "epoch": 0.6, + "learning_rate": 1.4039989746218918e-07, + "loss": 0.6941, + "step": 16325 + }, + { + "epoch": 0.6, + "learning_rate": 1.403962353975171e-07, + "loss": 0.6971, + "step": 16326 + }, + { + "epoch": 0.6, + "learning_rate": 1.4039257333284506e-07, + "loss": 0.681, + "step": 16327 + }, + { + "epoch": 0.6, + "learning_rate": 1.4038891126817298e-07, + "loss": 0.6923, + "step": 16328 + }, + { + "epoch": 0.6, + "learning_rate": 1.403852492035009e-07, + "loss": 0.6899, + "step": 16329 + }, + { + "epoch": 0.6, + "learning_rate": 1.4038158713882886e-07, + "loss": 0.7101, + "step": 16330 + }, + { + "epoch": 0.6, + "learning_rate": 1.403779250741568e-07, + "loss": 0.697, + "step": 16331 + }, + { + "epoch": 0.6, + "learning_rate": 1.4037426300948474e-07, + "loss": 0.6856, + "step": 16332 + }, + { + "epoch": 0.6, + "learning_rate": 1.4037060094481267e-07, + "loss": 0.6907, + "step": 16333 + }, + { + "epoch": 0.6, + "learning_rate": 1.4036693888014062e-07, + "loss": 0.7104, + "step": 16334 + }, + { + "epoch": 0.6, + "learning_rate": 1.4036327681546855e-07, + "loss": 0.6846, + "step": 16335 + }, + { + "epoch": 0.6, + "learning_rate": 1.403596147507965e-07, + "loss": 0.6853, + "step": 16336 + }, + { + "epoch": 0.6, + "learning_rate": 1.4035595268612443e-07, + "loss": 0.6907, + "step": 16337 + }, + { + "epoch": 0.6, + "learning_rate": 1.4035229062145238e-07, + "loss": 0.6911, + "step": 16338 + }, + { + "epoch": 0.6, + "learning_rate": 1.403486285567803e-07, + "loss": 0.7085, + "step": 16339 + }, + { + "epoch": 0.6, + "learning_rate": 1.4034496649210826e-07, + "loss": 0.694, + "step": 16340 + }, + { + "epoch": 0.6, + "learning_rate": 1.403413044274362e-07, + "loss": 0.6952, + "step": 16341 + }, + { + "epoch": 0.6, + "learning_rate": 1.4033764236276414e-07, + "loss": 0.6913, + "step": 16342 + }, + { + "epoch": 0.6, + "learning_rate": 1.4033398029809204e-07, + "loss": 0.7006, + "step": 16343 + }, + { + "epoch": 0.6, + "learning_rate": 1.4033031823342e-07, + "loss": 0.6861, + "step": 16344 + }, + { + "epoch": 0.6, + "learning_rate": 1.4032665616874792e-07, + "loss": 0.6931, + "step": 16345 + }, + { + "epoch": 0.6, + "learning_rate": 1.4032299410407587e-07, + "loss": 0.6922, + "step": 16346 + }, + { + "epoch": 0.6, + "learning_rate": 1.403193320394038e-07, + "loss": 0.686, + "step": 16347 + }, + { + "epoch": 0.6, + "learning_rate": 1.4031566997473175e-07, + "loss": 0.6852, + "step": 16348 + }, + { + "epoch": 0.6, + "learning_rate": 1.4031200791005968e-07, + "loss": 0.6936, + "step": 16349 + }, + { + "epoch": 0.6, + "learning_rate": 1.4030834584538763e-07, + "loss": 0.6872, + "step": 16350 + }, + { + "epoch": 0.6, + "learning_rate": 1.4030468378071556e-07, + "loss": 0.6936, + "step": 16351 + }, + { + "epoch": 0.6, + "learning_rate": 1.403010217160435e-07, + "loss": 0.6949, + "step": 16352 + }, + { + "epoch": 0.6, + "learning_rate": 1.4029735965137144e-07, + "loss": 0.6979, + "step": 16353 + }, + { + "epoch": 0.6, + "learning_rate": 1.4029369758669937e-07, + "loss": 0.671, + "step": 16354 + }, + { + "epoch": 0.6, + "learning_rate": 1.4029003552202732e-07, + "loss": 0.7002, + "step": 16355 + }, + { + "epoch": 0.6, + "learning_rate": 1.4028637345735525e-07, + "loss": 0.6936, + "step": 16356 + }, + { + "epoch": 0.6, + "learning_rate": 1.402827113926832e-07, + "loss": 0.6906, + "step": 16357 + }, + { + "epoch": 0.6, + "learning_rate": 1.4027904932801113e-07, + "loss": 0.6908, + "step": 16358 + }, + { + "epoch": 0.6, + "learning_rate": 1.4027538726333908e-07, + "loss": 0.7001, + "step": 16359 + }, + { + "epoch": 0.6, + "learning_rate": 1.40271725198667e-07, + "loss": 0.6976, + "step": 16360 + }, + { + "epoch": 0.6, + "learning_rate": 1.4026806313399493e-07, + "loss": 0.6957, + "step": 16361 + }, + { + "epoch": 0.6, + "learning_rate": 1.4026440106932286e-07, + "loss": 0.6939, + "step": 16362 + }, + { + "epoch": 0.6, + "learning_rate": 1.4026073900465081e-07, + "loss": 0.7001, + "step": 16363 + }, + { + "epoch": 0.6, + "learning_rate": 1.4025707693997874e-07, + "loss": 0.6937, + "step": 16364 + }, + { + "epoch": 0.6, + "learning_rate": 1.402534148753067e-07, + "loss": 0.6944, + "step": 16365 + }, + { + "epoch": 0.6, + "learning_rate": 1.4024975281063462e-07, + "loss": 0.6878, + "step": 16366 + }, + { + "epoch": 0.6, + "learning_rate": 1.4024609074596257e-07, + "loss": 0.6806, + "step": 16367 + }, + { + "epoch": 0.6, + "learning_rate": 1.402424286812905e-07, + "loss": 0.6962, + "step": 16368 + }, + { + "epoch": 0.6, + "learning_rate": 1.4023876661661845e-07, + "loss": 0.698, + "step": 16369 + }, + { + "epoch": 0.6, + "learning_rate": 1.4023510455194638e-07, + "loss": 0.7006, + "step": 16370 + }, + { + "epoch": 0.6, + "learning_rate": 1.4023144248727433e-07, + "loss": 0.6668, + "step": 16371 + }, + { + "epoch": 0.6, + "learning_rate": 1.4022778042260226e-07, + "loss": 0.6855, + "step": 16372 + }, + { + "epoch": 0.6, + "learning_rate": 1.4022411835793019e-07, + "loss": 0.6886, + "step": 16373 + }, + { + "epoch": 0.6, + "learning_rate": 1.4022045629325814e-07, + "loss": 0.6882, + "step": 16374 + }, + { + "epoch": 0.6, + "learning_rate": 1.4021679422858607e-07, + "loss": 0.6909, + "step": 16375 + }, + { + "epoch": 0.6, + "learning_rate": 1.4021313216391402e-07, + "loss": 0.6913, + "step": 16376 + }, + { + "epoch": 0.6, + "learning_rate": 1.4020947009924195e-07, + "loss": 0.6955, + "step": 16377 + }, + { + "epoch": 0.6, + "learning_rate": 1.402058080345699e-07, + "loss": 0.694, + "step": 16378 + }, + { + "epoch": 0.6, + "learning_rate": 1.402021459698978e-07, + "loss": 0.695, + "step": 16379 + }, + { + "epoch": 0.6, + "learning_rate": 1.4019848390522575e-07, + "loss": 0.694, + "step": 16380 + }, + { + "epoch": 0.6, + "learning_rate": 1.4019482184055368e-07, + "loss": 0.697, + "step": 16381 + }, + { + "epoch": 0.6, + "learning_rate": 1.4019115977588163e-07, + "loss": 0.6888, + "step": 16382 + }, + { + "epoch": 0.6, + "learning_rate": 1.4018749771120956e-07, + "loss": 0.6947, + "step": 16383 + }, + { + "epoch": 0.6, + "learning_rate": 1.401838356465375e-07, + "loss": 0.6965, + "step": 16384 + }, + { + "epoch": 0.6, + "learning_rate": 1.4018017358186544e-07, + "loss": 0.6884, + "step": 16385 + }, + { + "epoch": 0.6, + "learning_rate": 1.401765115171934e-07, + "loss": 0.703, + "step": 16386 + }, + { + "epoch": 0.6, + "learning_rate": 1.4017284945252132e-07, + "loss": 0.702, + "step": 16387 + }, + { + "epoch": 0.6, + "learning_rate": 1.4016918738784927e-07, + "loss": 0.6858, + "step": 16388 + }, + { + "epoch": 0.6, + "learning_rate": 1.401655253231772e-07, + "loss": 0.6895, + "step": 16389 + }, + { + "epoch": 0.6, + "learning_rate": 1.4016186325850515e-07, + "loss": 0.6879, + "step": 16390 + }, + { + "epoch": 0.6, + "learning_rate": 1.4015820119383308e-07, + "loss": 0.6974, + "step": 16391 + }, + { + "epoch": 0.6, + "learning_rate": 1.4015453912916103e-07, + "loss": 0.6919, + "step": 16392 + }, + { + "epoch": 0.6, + "learning_rate": 1.4015087706448896e-07, + "loss": 0.6901, + "step": 16393 + }, + { + "epoch": 0.6, + "learning_rate": 1.4014721499981689e-07, + "loss": 0.696, + "step": 16394 + }, + { + "epoch": 0.6, + "learning_rate": 1.4014355293514484e-07, + "loss": 0.6994, + "step": 16395 + }, + { + "epoch": 0.6, + "learning_rate": 1.4013989087047277e-07, + "loss": 0.6897, + "step": 16396 + }, + { + "epoch": 0.6, + "learning_rate": 1.401362288058007e-07, + "loss": 0.6867, + "step": 16397 + }, + { + "epoch": 0.6, + "learning_rate": 1.4013256674112865e-07, + "loss": 0.6908, + "step": 16398 + }, + { + "epoch": 0.6, + "learning_rate": 1.4012890467645657e-07, + "loss": 0.7019, + "step": 16399 + }, + { + "epoch": 0.6, + "learning_rate": 1.401252426117845e-07, + "loss": 0.6955, + "step": 16400 + }, + { + "epoch": 0.6, + "learning_rate": 1.4012158054711245e-07, + "loss": 0.7002, + "step": 16401 + }, + { + "epoch": 0.6, + "learning_rate": 1.4011791848244038e-07, + "loss": 0.6897, + "step": 16402 + }, + { + "epoch": 0.6, + "learning_rate": 1.4011425641776833e-07, + "loss": 0.6913, + "step": 16403 + }, + { + "epoch": 0.6, + "learning_rate": 1.4011059435309626e-07, + "loss": 0.69, + "step": 16404 + }, + { + "epoch": 0.6, + "learning_rate": 1.401069322884242e-07, + "loss": 0.676, + "step": 16405 + }, + { + "epoch": 0.6, + "learning_rate": 1.4010327022375214e-07, + "loss": 0.6833, + "step": 16406 + }, + { + "epoch": 0.6, + "learning_rate": 1.400996081590801e-07, + "loss": 0.6884, + "step": 16407 + }, + { + "epoch": 0.6, + "learning_rate": 1.4009594609440802e-07, + "loss": 0.6849, + "step": 16408 + }, + { + "epoch": 0.6, + "learning_rate": 1.4009228402973597e-07, + "loss": 0.7003, + "step": 16409 + }, + { + "epoch": 0.6, + "learning_rate": 1.400886219650639e-07, + "loss": 0.7059, + "step": 16410 + }, + { + "epoch": 0.6, + "learning_rate": 1.4008495990039185e-07, + "loss": 0.6946, + "step": 16411 + }, + { + "epoch": 0.6, + "learning_rate": 1.4008129783571978e-07, + "loss": 0.6795, + "step": 16412 + }, + { + "epoch": 0.6, + "learning_rate": 1.4007763577104773e-07, + "loss": 0.6898, + "step": 16413 + }, + { + "epoch": 0.6, + "learning_rate": 1.4007397370637566e-07, + "loss": 0.689, + "step": 16414 + }, + { + "epoch": 0.6, + "learning_rate": 1.4007031164170358e-07, + "loss": 0.6838, + "step": 16415 + }, + { + "epoch": 0.6, + "learning_rate": 1.400666495770315e-07, + "loss": 0.6877, + "step": 16416 + }, + { + "epoch": 0.6, + "learning_rate": 1.4006298751235946e-07, + "loss": 0.695, + "step": 16417 + }, + { + "epoch": 0.6, + "learning_rate": 1.400593254476874e-07, + "loss": 0.6976, + "step": 16418 + }, + { + "epoch": 0.6, + "learning_rate": 1.4005566338301534e-07, + "loss": 0.7076, + "step": 16419 + }, + { + "epoch": 0.6, + "learning_rate": 1.4005200131834327e-07, + "loss": 0.701, + "step": 16420 + }, + { + "epoch": 0.6, + "learning_rate": 1.4004833925367122e-07, + "loss": 0.6843, + "step": 16421 + }, + { + "epoch": 0.6, + "learning_rate": 1.4004467718899915e-07, + "loss": 0.6856, + "step": 16422 + }, + { + "epoch": 0.6, + "learning_rate": 1.4004101512432708e-07, + "loss": 0.7008, + "step": 16423 + }, + { + "epoch": 0.6, + "learning_rate": 1.4003735305965503e-07, + "loss": 0.6858, + "step": 16424 + }, + { + "epoch": 0.6, + "learning_rate": 1.4003369099498296e-07, + "loss": 0.6944, + "step": 16425 + }, + { + "epoch": 0.6, + "learning_rate": 1.400300289303109e-07, + "loss": 0.6951, + "step": 16426 + }, + { + "epoch": 0.6, + "learning_rate": 1.4002636686563884e-07, + "loss": 0.6965, + "step": 16427 + }, + { + "epoch": 0.6, + "learning_rate": 1.400227048009668e-07, + "loss": 0.6937, + "step": 16428 + }, + { + "epoch": 0.6, + "learning_rate": 1.4001904273629472e-07, + "loss": 0.6974, + "step": 16429 + }, + { + "epoch": 0.6, + "learning_rate": 1.4001538067162267e-07, + "loss": 0.6896, + "step": 16430 + }, + { + "epoch": 0.6, + "learning_rate": 1.400117186069506e-07, + "loss": 0.6883, + "step": 16431 + }, + { + "epoch": 0.6, + "learning_rate": 1.4000805654227855e-07, + "loss": 0.7089, + "step": 16432 + }, + { + "epoch": 0.6, + "learning_rate": 1.4000439447760645e-07, + "loss": 0.7024, + "step": 16433 + }, + { + "epoch": 0.6, + "learning_rate": 1.400007324129344e-07, + "loss": 0.6954, + "step": 16434 + }, + { + "epoch": 0.6, + "learning_rate": 1.3999707034826233e-07, + "loss": 0.6956, + "step": 16435 + }, + { + "epoch": 0.6, + "learning_rate": 1.3999340828359028e-07, + "loss": 0.6906, + "step": 16436 + }, + { + "epoch": 0.6, + "learning_rate": 1.399897462189182e-07, + "loss": 0.6898, + "step": 16437 + }, + { + "epoch": 0.6, + "learning_rate": 1.3998608415424616e-07, + "loss": 0.6916, + "step": 16438 + }, + { + "epoch": 0.6, + "learning_rate": 1.399824220895741e-07, + "loss": 0.7103, + "step": 16439 + }, + { + "epoch": 0.6, + "learning_rate": 1.3997876002490204e-07, + "loss": 0.696, + "step": 16440 + }, + { + "epoch": 0.6, + "learning_rate": 1.3997509796022997e-07, + "loss": 0.6836, + "step": 16441 + }, + { + "epoch": 0.6, + "learning_rate": 1.3997143589555792e-07, + "loss": 0.6876, + "step": 16442 + }, + { + "epoch": 0.6, + "learning_rate": 1.3996777383088585e-07, + "loss": 0.6994, + "step": 16443 + }, + { + "epoch": 0.6, + "learning_rate": 1.3996411176621378e-07, + "loss": 0.692, + "step": 16444 + }, + { + "epoch": 0.6, + "learning_rate": 1.3996044970154173e-07, + "loss": 0.677, + "step": 16445 + }, + { + "epoch": 0.6, + "learning_rate": 1.3995678763686966e-07, + "loss": 0.6957, + "step": 16446 + }, + { + "epoch": 0.6, + "learning_rate": 1.399531255721976e-07, + "loss": 0.6914, + "step": 16447 + }, + { + "epoch": 0.6, + "learning_rate": 1.3994946350752554e-07, + "loss": 0.6864, + "step": 16448 + }, + { + "epoch": 0.6, + "learning_rate": 1.399458014428535e-07, + "loss": 0.707, + "step": 16449 + }, + { + "epoch": 0.6, + "learning_rate": 1.3994213937818142e-07, + "loss": 0.7056, + "step": 16450 + }, + { + "epoch": 0.6, + "learning_rate": 1.3993847731350934e-07, + "loss": 0.6969, + "step": 16451 + }, + { + "epoch": 0.6, + "learning_rate": 1.3993481524883727e-07, + "loss": 0.6936, + "step": 16452 + }, + { + "epoch": 0.6, + "learning_rate": 1.3993115318416522e-07, + "loss": 0.7079, + "step": 16453 + }, + { + "epoch": 0.6, + "learning_rate": 1.3992749111949315e-07, + "loss": 0.699, + "step": 16454 + }, + { + "epoch": 0.6, + "learning_rate": 1.399238290548211e-07, + "loss": 0.7008, + "step": 16455 + }, + { + "epoch": 0.6, + "learning_rate": 1.3992016699014903e-07, + "loss": 0.684, + "step": 16456 + }, + { + "epoch": 0.6, + "learning_rate": 1.3991650492547698e-07, + "loss": 0.6861, + "step": 16457 + }, + { + "epoch": 0.6, + "learning_rate": 1.399128428608049e-07, + "loss": 0.695, + "step": 16458 + }, + { + "epoch": 0.6, + "learning_rate": 1.3990918079613286e-07, + "loss": 0.6887, + "step": 16459 + }, + { + "epoch": 0.6, + "learning_rate": 1.399055187314608e-07, + "loss": 0.6956, + "step": 16460 + }, + { + "epoch": 0.6, + "learning_rate": 1.3990185666678874e-07, + "loss": 0.6981, + "step": 16461 + }, + { + "epoch": 0.6, + "learning_rate": 1.3989819460211667e-07, + "loss": 0.7088, + "step": 16462 + }, + { + "epoch": 0.6, + "learning_rate": 1.3989453253744462e-07, + "loss": 0.6855, + "step": 16463 + }, + { + "epoch": 0.6, + "learning_rate": 1.3989087047277255e-07, + "loss": 0.6942, + "step": 16464 + }, + { + "epoch": 0.6, + "learning_rate": 1.3988720840810047e-07, + "loss": 0.7194, + "step": 16465 + }, + { + "epoch": 0.6, + "learning_rate": 1.3988354634342843e-07, + "loss": 0.6992, + "step": 16466 + }, + { + "epoch": 0.6, + "learning_rate": 1.3987988427875635e-07, + "loss": 0.6881, + "step": 16467 + }, + { + "epoch": 0.6, + "learning_rate": 1.398762222140843e-07, + "loss": 0.6981, + "step": 16468 + }, + { + "epoch": 0.6, + "learning_rate": 1.3987256014941223e-07, + "loss": 0.6854, + "step": 16469 + }, + { + "epoch": 0.6, + "learning_rate": 1.3986889808474016e-07, + "loss": 0.6876, + "step": 16470 + }, + { + "epoch": 0.6, + "learning_rate": 1.398652360200681e-07, + "loss": 0.68, + "step": 16471 + }, + { + "epoch": 0.6, + "learning_rate": 1.3986157395539604e-07, + "loss": 0.698, + "step": 16472 + }, + { + "epoch": 0.6, + "learning_rate": 1.3985791189072397e-07, + "loss": 0.689, + "step": 16473 + }, + { + "epoch": 0.6, + "learning_rate": 1.3985424982605192e-07, + "loss": 0.705, + "step": 16474 + }, + { + "epoch": 0.6, + "learning_rate": 1.3985058776137985e-07, + "loss": 0.6978, + "step": 16475 + }, + { + "epoch": 0.6, + "learning_rate": 1.398469256967078e-07, + "loss": 0.6965, + "step": 16476 + }, + { + "epoch": 0.6, + "learning_rate": 1.3984326363203573e-07, + "loss": 0.6884, + "step": 16477 + }, + { + "epoch": 0.6, + "learning_rate": 1.3983960156736368e-07, + "loss": 0.6793, + "step": 16478 + }, + { + "epoch": 0.6, + "learning_rate": 1.398359395026916e-07, + "loss": 0.6927, + "step": 16479 + }, + { + "epoch": 0.6, + "learning_rate": 1.3983227743801956e-07, + "loss": 0.7054, + "step": 16480 + }, + { + "epoch": 0.6, + "learning_rate": 1.398286153733475e-07, + "loss": 0.7105, + "step": 16481 + }, + { + "epoch": 0.6, + "learning_rate": 1.3982495330867544e-07, + "loss": 0.6867, + "step": 16482 + }, + { + "epoch": 0.6, + "learning_rate": 1.3982129124400337e-07, + "loss": 0.6987, + "step": 16483 + }, + { + "epoch": 0.6, + "learning_rate": 1.3981762917933132e-07, + "loss": 0.7114, + "step": 16484 + }, + { + "epoch": 0.6, + "learning_rate": 1.3981396711465925e-07, + "loss": 0.7012, + "step": 16485 + }, + { + "epoch": 0.6, + "learning_rate": 1.398103050499872e-07, + "loss": 0.7085, + "step": 16486 + }, + { + "epoch": 0.6, + "learning_rate": 1.398066429853151e-07, + "loss": 0.6939, + "step": 16487 + }, + { + "epoch": 0.6, + "learning_rate": 1.3980298092064305e-07, + "loss": 0.6994, + "step": 16488 + }, + { + "epoch": 0.6, + "learning_rate": 1.3979931885597098e-07, + "loss": 0.6847, + "step": 16489 + }, + { + "epoch": 0.6, + "learning_rate": 1.3979565679129893e-07, + "loss": 0.6883, + "step": 16490 + }, + { + "epoch": 0.6, + "learning_rate": 1.3979199472662686e-07, + "loss": 0.6963, + "step": 16491 + }, + { + "epoch": 0.6, + "learning_rate": 1.397883326619548e-07, + "loss": 0.6974, + "step": 16492 + }, + { + "epoch": 0.6, + "learning_rate": 1.3978467059728274e-07, + "loss": 0.6995, + "step": 16493 + }, + { + "epoch": 0.6, + "learning_rate": 1.3978100853261067e-07, + "loss": 0.6976, + "step": 16494 + }, + { + "epoch": 0.6, + "learning_rate": 1.3977734646793862e-07, + "loss": 0.7073, + "step": 16495 + }, + { + "epoch": 0.6, + "learning_rate": 1.3977368440326655e-07, + "loss": 0.7043, + "step": 16496 + }, + { + "epoch": 0.6, + "learning_rate": 1.397700223385945e-07, + "loss": 0.6954, + "step": 16497 + }, + { + "epoch": 0.6, + "learning_rate": 1.3976636027392243e-07, + "loss": 0.6874, + "step": 16498 + }, + { + "epoch": 0.6, + "learning_rate": 1.3976269820925038e-07, + "loss": 0.6956, + "step": 16499 + }, + { + "epoch": 0.6, + "learning_rate": 1.397590361445783e-07, + "loss": 0.688, + "step": 16500 + }, + { + "epoch": 0.6, + "learning_rate": 1.3975537407990626e-07, + "loss": 0.696, + "step": 16501 + }, + { + "epoch": 0.6, + "learning_rate": 1.3975171201523419e-07, + "loss": 0.7089, + "step": 16502 + }, + { + "epoch": 0.6, + "learning_rate": 1.3974804995056214e-07, + "loss": 0.7014, + "step": 16503 + }, + { + "epoch": 0.6, + "learning_rate": 1.3974438788589007e-07, + "loss": 0.6766, + "step": 16504 + }, + { + "epoch": 0.6, + "learning_rate": 1.39740725821218e-07, + "loss": 0.6872, + "step": 16505 + }, + { + "epoch": 0.6, + "learning_rate": 1.3973706375654592e-07, + "loss": 0.7052, + "step": 16506 + }, + { + "epoch": 0.6, + "learning_rate": 1.3973340169187387e-07, + "loss": 0.7034, + "step": 16507 + }, + { + "epoch": 0.6, + "learning_rate": 1.397297396272018e-07, + "loss": 0.7108, + "step": 16508 + }, + { + "epoch": 0.6, + "learning_rate": 1.3972607756252975e-07, + "loss": 0.6869, + "step": 16509 + }, + { + "epoch": 0.6, + "learning_rate": 1.3972241549785768e-07, + "loss": 0.6911, + "step": 16510 + }, + { + "epoch": 0.6, + "learning_rate": 1.3971875343318563e-07, + "loss": 0.7045, + "step": 16511 + }, + { + "epoch": 0.6, + "learning_rate": 1.3971509136851356e-07, + "loss": 0.6926, + "step": 16512 + }, + { + "epoch": 0.6, + "learning_rate": 1.397114293038415e-07, + "loss": 0.6899, + "step": 16513 + }, + { + "epoch": 0.6, + "learning_rate": 1.3970776723916944e-07, + "loss": 0.6999, + "step": 16514 + }, + { + "epoch": 0.6, + "learning_rate": 1.3970410517449737e-07, + "loss": 0.6935, + "step": 16515 + }, + { + "epoch": 0.6, + "learning_rate": 1.3970044310982532e-07, + "loss": 0.689, + "step": 16516 + }, + { + "epoch": 0.6, + "learning_rate": 1.3969678104515325e-07, + "loss": 0.6949, + "step": 16517 + }, + { + "epoch": 0.6, + "learning_rate": 1.396931189804812e-07, + "loss": 0.6809, + "step": 16518 + }, + { + "epoch": 0.6, + "learning_rate": 1.3968945691580912e-07, + "loss": 0.6935, + "step": 16519 + }, + { + "epoch": 0.6, + "learning_rate": 1.3968579485113708e-07, + "loss": 0.6871, + "step": 16520 + }, + { + "epoch": 0.6, + "learning_rate": 1.39682132786465e-07, + "loss": 0.6976, + "step": 16521 + }, + { + "epoch": 0.6, + "learning_rate": 1.3967847072179296e-07, + "loss": 0.6985, + "step": 16522 + }, + { + "epoch": 0.6, + "learning_rate": 1.3967480865712086e-07, + "loss": 0.7033, + "step": 16523 + }, + { + "epoch": 0.6, + "learning_rate": 1.396711465924488e-07, + "loss": 0.6888, + "step": 16524 + }, + { + "epoch": 0.6, + "learning_rate": 1.3966748452777674e-07, + "loss": 0.688, + "step": 16525 + }, + { + "epoch": 0.6, + "learning_rate": 1.396638224631047e-07, + "loss": 0.69, + "step": 16526 + }, + { + "epoch": 0.6, + "learning_rate": 1.3966016039843262e-07, + "loss": 0.6998, + "step": 16527 + }, + { + "epoch": 0.6, + "learning_rate": 1.3965649833376057e-07, + "loss": 0.6961, + "step": 16528 + }, + { + "epoch": 0.6, + "learning_rate": 1.396528362690885e-07, + "loss": 0.6874, + "step": 16529 + }, + { + "epoch": 0.6, + "learning_rate": 1.3964917420441645e-07, + "loss": 0.697, + "step": 16530 + }, + { + "epoch": 0.6, + "learning_rate": 1.3964551213974438e-07, + "loss": 0.7131, + "step": 16531 + }, + { + "epoch": 0.6, + "learning_rate": 1.3964185007507233e-07, + "loss": 0.7012, + "step": 16532 + }, + { + "epoch": 0.6, + "learning_rate": 1.3963818801040026e-07, + "loss": 0.6944, + "step": 16533 + }, + { + "epoch": 0.6, + "learning_rate": 1.396345259457282e-07, + "loss": 0.7022, + "step": 16534 + }, + { + "epoch": 0.6, + "learning_rate": 1.3963086388105614e-07, + "loss": 0.6917, + "step": 16535 + }, + { + "epoch": 0.61, + "learning_rate": 1.3962720181638406e-07, + "loss": 0.6918, + "step": 16536 + }, + { + "epoch": 0.61, + "learning_rate": 1.3962353975171202e-07, + "loss": 0.6884, + "step": 16537 + }, + { + "epoch": 0.61, + "learning_rate": 1.3961987768703994e-07, + "loss": 0.697, + "step": 16538 + }, + { + "epoch": 0.61, + "learning_rate": 1.396162156223679e-07, + "loss": 0.6945, + "step": 16539 + }, + { + "epoch": 0.61, + "learning_rate": 1.3961255355769582e-07, + "loss": 0.6889, + "step": 16540 + }, + { + "epoch": 0.61, + "learning_rate": 1.3960889149302375e-07, + "loss": 0.6869, + "step": 16541 + }, + { + "epoch": 0.61, + "learning_rate": 1.3960522942835168e-07, + "loss": 0.7022, + "step": 16542 + }, + { + "epoch": 0.61, + "learning_rate": 1.3960156736367963e-07, + "loss": 0.6809, + "step": 16543 + }, + { + "epoch": 0.61, + "learning_rate": 1.3959790529900756e-07, + "loss": 0.7087, + "step": 16544 + }, + { + "epoch": 0.61, + "learning_rate": 1.395942432343355e-07, + "loss": 0.7034, + "step": 16545 + }, + { + "epoch": 0.61, + "learning_rate": 1.3959058116966344e-07, + "loss": 0.6889, + "step": 16546 + }, + { + "epoch": 0.61, + "learning_rate": 1.395869191049914e-07, + "loss": 0.683, + "step": 16547 + }, + { + "epoch": 0.61, + "learning_rate": 1.3958325704031932e-07, + "loss": 0.6896, + "step": 16548 + }, + { + "epoch": 0.61, + "learning_rate": 1.3957959497564727e-07, + "loss": 0.6864, + "step": 16549 + }, + { + "epoch": 0.61, + "learning_rate": 1.395759329109752e-07, + "loss": 0.7039, + "step": 16550 + }, + { + "epoch": 0.61, + "learning_rate": 1.3957227084630315e-07, + "loss": 0.6983, + "step": 16551 + }, + { + "epoch": 0.61, + "learning_rate": 1.3956860878163108e-07, + "loss": 0.7005, + "step": 16552 + }, + { + "epoch": 0.61, + "learning_rate": 1.3956494671695903e-07, + "loss": 0.6987, + "step": 16553 + }, + { + "epoch": 0.61, + "learning_rate": 1.3956128465228696e-07, + "loss": 0.6815, + "step": 16554 + }, + { + "epoch": 0.61, + "learning_rate": 1.395576225876149e-07, + "loss": 0.6952, + "step": 16555 + }, + { + "epoch": 0.61, + "learning_rate": 1.3955396052294284e-07, + "loss": 0.6978, + "step": 16556 + }, + { + "epoch": 0.61, + "learning_rate": 1.395502984582708e-07, + "loss": 0.7022, + "step": 16557 + }, + { + "epoch": 0.61, + "learning_rate": 1.3954663639359872e-07, + "loss": 0.7005, + "step": 16558 + }, + { + "epoch": 0.61, + "learning_rate": 1.3954297432892664e-07, + "loss": 0.691, + "step": 16559 + }, + { + "epoch": 0.61, + "learning_rate": 1.3953931226425457e-07, + "loss": 0.682, + "step": 16560 + }, + { + "epoch": 0.61, + "learning_rate": 1.3953565019958252e-07, + "loss": 0.7065, + "step": 16561 + }, + { + "epoch": 0.61, + "learning_rate": 1.3953198813491045e-07, + "loss": 0.7022, + "step": 16562 + }, + { + "epoch": 0.61, + "learning_rate": 1.3952832607023838e-07, + "loss": 0.6897, + "step": 16563 + }, + { + "epoch": 0.61, + "learning_rate": 1.3952466400556633e-07, + "loss": 0.6832, + "step": 16564 + }, + { + "epoch": 0.61, + "learning_rate": 1.3952100194089426e-07, + "loss": 0.6915, + "step": 16565 + }, + { + "epoch": 0.61, + "learning_rate": 1.395173398762222e-07, + "loss": 0.7006, + "step": 16566 + }, + { + "epoch": 0.61, + "learning_rate": 1.3951367781155014e-07, + "loss": 0.7133, + "step": 16567 + }, + { + "epoch": 0.61, + "learning_rate": 1.395100157468781e-07, + "loss": 0.6964, + "step": 16568 + }, + { + "epoch": 0.61, + "learning_rate": 1.3950635368220602e-07, + "loss": 0.691, + "step": 16569 + }, + { + "epoch": 0.61, + "learning_rate": 1.3950269161753397e-07, + "loss": 0.6788, + "step": 16570 + }, + { + "epoch": 0.61, + "learning_rate": 1.394990295528619e-07, + "loss": 0.6965, + "step": 16571 + }, + { + "epoch": 0.61, + "learning_rate": 1.3949536748818985e-07, + "loss": 0.692, + "step": 16572 + }, + { + "epoch": 0.61, + "learning_rate": 1.3949170542351778e-07, + "loss": 0.6899, + "step": 16573 + }, + { + "epoch": 0.61, + "learning_rate": 1.3948804335884573e-07, + "loss": 0.694, + "step": 16574 + }, + { + "epoch": 0.61, + "learning_rate": 1.3948438129417366e-07, + "loss": 0.6952, + "step": 16575 + }, + { + "epoch": 0.61, + "learning_rate": 1.394807192295016e-07, + "loss": 0.6899, + "step": 16576 + }, + { + "epoch": 0.61, + "learning_rate": 1.394770571648295e-07, + "loss": 0.6811, + "step": 16577 + }, + { + "epoch": 0.61, + "learning_rate": 1.3947339510015746e-07, + "loss": 0.691, + "step": 16578 + }, + { + "epoch": 0.61, + "learning_rate": 1.394697330354854e-07, + "loss": 0.6907, + "step": 16579 + }, + { + "epoch": 0.61, + "learning_rate": 1.3946607097081334e-07, + "loss": 0.6837, + "step": 16580 + }, + { + "epoch": 0.61, + "learning_rate": 1.3946240890614127e-07, + "loss": 0.7007, + "step": 16581 + }, + { + "epoch": 0.61, + "learning_rate": 1.3945874684146922e-07, + "loss": 0.6899, + "step": 16582 + }, + { + "epoch": 0.61, + "learning_rate": 1.3945508477679715e-07, + "loss": 0.6884, + "step": 16583 + }, + { + "epoch": 0.61, + "learning_rate": 1.394514227121251e-07, + "loss": 0.7045, + "step": 16584 + }, + { + "epoch": 0.61, + "learning_rate": 1.3944776064745303e-07, + "loss": 0.6947, + "step": 16585 + }, + { + "epoch": 0.61, + "learning_rate": 1.3944409858278095e-07, + "loss": 0.6987, + "step": 16586 + }, + { + "epoch": 0.61, + "learning_rate": 1.394404365181089e-07, + "loss": 0.69, + "step": 16587 + }, + { + "epoch": 0.61, + "learning_rate": 1.3943677445343683e-07, + "loss": 0.6955, + "step": 16588 + }, + { + "epoch": 0.61, + "learning_rate": 1.394331123887648e-07, + "loss": 0.7041, + "step": 16589 + }, + { + "epoch": 0.61, + "learning_rate": 1.3942945032409271e-07, + "loss": 0.6861, + "step": 16590 + }, + { + "epoch": 0.61, + "learning_rate": 1.3942578825942067e-07, + "loss": 0.7001, + "step": 16591 + }, + { + "epoch": 0.61, + "learning_rate": 1.394221261947486e-07, + "loss": 0.7018, + "step": 16592 + }, + { + "epoch": 0.61, + "learning_rate": 1.3941846413007655e-07, + "loss": 0.6957, + "step": 16593 + }, + { + "epoch": 0.61, + "learning_rate": 1.3941480206540445e-07, + "loss": 0.7015, + "step": 16594 + }, + { + "epoch": 0.61, + "learning_rate": 1.394111400007324e-07, + "loss": 0.6978, + "step": 16595 + }, + { + "epoch": 0.61, + "learning_rate": 1.3940747793606033e-07, + "loss": 0.6915, + "step": 16596 + }, + { + "epoch": 0.61, + "learning_rate": 1.3940381587138828e-07, + "loss": 0.6871, + "step": 16597 + }, + { + "epoch": 0.61, + "learning_rate": 1.394001538067162e-07, + "loss": 0.6944, + "step": 16598 + }, + { + "epoch": 0.61, + "learning_rate": 1.3939649174204416e-07, + "loss": 0.7019, + "step": 16599 + }, + { + "epoch": 0.61, + "learning_rate": 1.393928296773721e-07, + "loss": 0.6914, + "step": 16600 + }, + { + "epoch": 0.61, + "learning_rate": 1.3938916761270004e-07, + "loss": 0.6993, + "step": 16601 + }, + { + "epoch": 0.61, + "learning_rate": 1.3938550554802797e-07, + "loss": 0.6904, + "step": 16602 + }, + { + "epoch": 0.61, + "learning_rate": 1.3938184348335592e-07, + "loss": 0.6867, + "step": 16603 + }, + { + "epoch": 0.61, + "learning_rate": 1.3937818141868385e-07, + "loss": 0.6981, + "step": 16604 + }, + { + "epoch": 0.61, + "learning_rate": 1.393745193540118e-07, + "loss": 0.6902, + "step": 16605 + }, + { + "epoch": 0.61, + "learning_rate": 1.3937085728933973e-07, + "loss": 0.6991, + "step": 16606 + }, + { + "epoch": 0.61, + "learning_rate": 1.3936719522466765e-07, + "loss": 0.6925, + "step": 16607 + }, + { + "epoch": 0.61, + "learning_rate": 1.393635331599956e-07, + "loss": 0.7005, + "step": 16608 + }, + { + "epoch": 0.61, + "learning_rate": 1.3935987109532353e-07, + "loss": 0.695, + "step": 16609 + }, + { + "epoch": 0.61, + "learning_rate": 1.3935620903065149e-07, + "loss": 0.6931, + "step": 16610 + }, + { + "epoch": 0.61, + "learning_rate": 1.393525469659794e-07, + "loss": 0.6956, + "step": 16611 + }, + { + "epoch": 0.61, + "learning_rate": 1.3934888490130734e-07, + "loss": 0.6872, + "step": 16612 + }, + { + "epoch": 0.61, + "learning_rate": 1.3934522283663527e-07, + "loss": 0.6964, + "step": 16613 + }, + { + "epoch": 0.61, + "learning_rate": 1.3934156077196322e-07, + "loss": 0.7094, + "step": 16614 + }, + { + "epoch": 0.61, + "learning_rate": 1.3933789870729115e-07, + "loss": 0.6957, + "step": 16615 + }, + { + "epoch": 0.61, + "learning_rate": 1.393342366426191e-07, + "loss": 0.7041, + "step": 16616 + }, + { + "epoch": 0.61, + "learning_rate": 1.3933057457794703e-07, + "loss": 0.698, + "step": 16617 + }, + { + "epoch": 0.61, + "learning_rate": 1.3932691251327498e-07, + "loss": 0.7029, + "step": 16618 + }, + { + "epoch": 0.61, + "learning_rate": 1.393232504486029e-07, + "loss": 0.6786, + "step": 16619 + }, + { + "epoch": 0.61, + "learning_rate": 1.3931958838393086e-07, + "loss": 0.6853, + "step": 16620 + }, + { + "epoch": 0.61, + "learning_rate": 1.3931592631925879e-07, + "loss": 0.6915, + "step": 16621 + }, + { + "epoch": 0.61, + "learning_rate": 1.3931226425458674e-07, + "loss": 0.6966, + "step": 16622 + }, + { + "epoch": 0.61, + "learning_rate": 1.3930860218991467e-07, + "loss": 0.7024, + "step": 16623 + }, + { + "epoch": 0.61, + "learning_rate": 1.3930494012524262e-07, + "loss": 0.685, + "step": 16624 + }, + { + "epoch": 0.61, + "learning_rate": 1.3930127806057055e-07, + "loss": 0.6895, + "step": 16625 + }, + { + "epoch": 0.61, + "learning_rate": 1.392976159958985e-07, + "loss": 0.6963, + "step": 16626 + }, + { + "epoch": 0.61, + "learning_rate": 1.3929395393122643e-07, + "loss": 0.7007, + "step": 16627 + }, + { + "epoch": 0.61, + "learning_rate": 1.3929029186655438e-07, + "loss": 0.6996, + "step": 16628 + }, + { + "epoch": 0.61, + "learning_rate": 1.392866298018823e-07, + "loss": 0.7105, + "step": 16629 + }, + { + "epoch": 0.61, + "learning_rate": 1.3928296773721023e-07, + "loss": 0.6823, + "step": 16630 + }, + { + "epoch": 0.61, + "learning_rate": 1.3927930567253816e-07, + "loss": 0.6919, + "step": 16631 + }, + { + "epoch": 0.61, + "learning_rate": 1.392756436078661e-07, + "loss": 0.6974, + "step": 16632 + }, + { + "epoch": 0.61, + "learning_rate": 1.3927198154319404e-07, + "loss": 0.7013, + "step": 16633 + }, + { + "epoch": 0.61, + "learning_rate": 1.3926831947852197e-07, + "loss": 0.7004, + "step": 16634 + }, + { + "epoch": 0.61, + "learning_rate": 1.3926465741384992e-07, + "loss": 0.7034, + "step": 16635 + }, + { + "epoch": 0.61, + "learning_rate": 1.3926099534917785e-07, + "loss": 0.6936, + "step": 16636 + }, + { + "epoch": 0.61, + "learning_rate": 1.392573332845058e-07, + "loss": 0.693, + "step": 16637 + }, + { + "epoch": 0.61, + "learning_rate": 1.3925367121983372e-07, + "loss": 0.6969, + "step": 16638 + }, + { + "epoch": 0.61, + "learning_rate": 1.3925000915516168e-07, + "loss": 0.6923, + "step": 16639 + }, + { + "epoch": 0.61, + "learning_rate": 1.392463470904896e-07, + "loss": 0.6684, + "step": 16640 + }, + { + "epoch": 0.61, + "learning_rate": 1.3924268502581756e-07, + "loss": 0.6912, + "step": 16641 + }, + { + "epoch": 0.61, + "learning_rate": 1.3923902296114548e-07, + "loss": 0.7049, + "step": 16642 + }, + { + "epoch": 0.61, + "learning_rate": 1.3923536089647344e-07, + "loss": 0.7022, + "step": 16643 + }, + { + "epoch": 0.61, + "learning_rate": 1.3923169883180136e-07, + "loss": 0.6901, + "step": 16644 + }, + { + "epoch": 0.61, + "learning_rate": 1.3922803676712932e-07, + "loss": 0.6982, + "step": 16645 + }, + { + "epoch": 0.61, + "learning_rate": 1.3922437470245724e-07, + "loss": 0.6916, + "step": 16646 + }, + { + "epoch": 0.61, + "learning_rate": 1.392207126377852e-07, + "loss": 0.6991, + "step": 16647 + }, + { + "epoch": 0.61, + "learning_rate": 1.392170505731131e-07, + "loss": 0.6862, + "step": 16648 + }, + { + "epoch": 0.61, + "learning_rate": 1.3921338850844105e-07, + "loss": 0.6953, + "step": 16649 + }, + { + "epoch": 0.61, + "learning_rate": 1.3920972644376898e-07, + "loss": 0.6927, + "step": 16650 + }, + { + "epoch": 0.61, + "learning_rate": 1.3920606437909693e-07, + "loss": 0.6844, + "step": 16651 + }, + { + "epoch": 0.61, + "learning_rate": 1.3920240231442486e-07, + "loss": 0.6813, + "step": 16652 + }, + { + "epoch": 0.61, + "learning_rate": 1.391987402497528e-07, + "loss": 0.6855, + "step": 16653 + }, + { + "epoch": 0.61, + "learning_rate": 1.3919507818508074e-07, + "loss": 0.6974, + "step": 16654 + }, + { + "epoch": 0.61, + "learning_rate": 1.391914161204087e-07, + "loss": 0.6993, + "step": 16655 + }, + { + "epoch": 0.61, + "learning_rate": 1.3918775405573662e-07, + "loss": 0.7074, + "step": 16656 + }, + { + "epoch": 0.61, + "learning_rate": 1.3918409199106454e-07, + "loss": 0.694, + "step": 16657 + }, + { + "epoch": 0.61, + "learning_rate": 1.391804299263925e-07, + "loss": 0.6908, + "step": 16658 + }, + { + "epoch": 0.61, + "learning_rate": 1.3917676786172042e-07, + "loss": 0.7029, + "step": 16659 + }, + { + "epoch": 0.61, + "learning_rate": 1.3917310579704838e-07, + "loss": 0.6908, + "step": 16660 + }, + { + "epoch": 0.61, + "learning_rate": 1.391694437323763e-07, + "loss": 0.696, + "step": 16661 + }, + { + "epoch": 0.61, + "learning_rate": 1.3916578166770426e-07, + "loss": 0.68, + "step": 16662 + }, + { + "epoch": 0.61, + "learning_rate": 1.3916211960303218e-07, + "loss": 0.6897, + "step": 16663 + }, + { + "epoch": 0.61, + "learning_rate": 1.3915845753836014e-07, + "loss": 0.7018, + "step": 16664 + }, + { + "epoch": 0.61, + "learning_rate": 1.3915479547368806e-07, + "loss": 0.6901, + "step": 16665 + }, + { + "epoch": 0.61, + "learning_rate": 1.39151133409016e-07, + "loss": 0.6878, + "step": 16666 + }, + { + "epoch": 0.61, + "learning_rate": 1.3914747134434392e-07, + "loss": 0.7099, + "step": 16667 + }, + { + "epoch": 0.61, + "learning_rate": 1.3914380927967187e-07, + "loss": 0.6969, + "step": 16668 + }, + { + "epoch": 0.61, + "learning_rate": 1.391401472149998e-07, + "loss": 0.6893, + "step": 16669 + }, + { + "epoch": 0.61, + "learning_rate": 1.3913648515032775e-07, + "loss": 0.6911, + "step": 16670 + }, + { + "epoch": 0.61, + "learning_rate": 1.3913282308565568e-07, + "loss": 0.7003, + "step": 16671 + }, + { + "epoch": 0.61, + "learning_rate": 1.3912916102098363e-07, + "loss": 0.7059, + "step": 16672 + }, + { + "epoch": 0.61, + "learning_rate": 1.3912549895631156e-07, + "loss": 0.6882, + "step": 16673 + }, + { + "epoch": 0.61, + "learning_rate": 1.391218368916395e-07, + "loss": 0.703, + "step": 16674 + }, + { + "epoch": 0.61, + "learning_rate": 1.3911817482696744e-07, + "loss": 0.6942, + "step": 16675 + }, + { + "epoch": 0.61, + "learning_rate": 1.391145127622954e-07, + "loss": 0.6925, + "step": 16676 + }, + { + "epoch": 0.61, + "learning_rate": 1.3911085069762332e-07, + "loss": 0.706, + "step": 16677 + }, + { + "epoch": 0.61, + "learning_rate": 1.3910718863295124e-07, + "loss": 0.681, + "step": 16678 + }, + { + "epoch": 0.61, + "learning_rate": 1.391035265682792e-07, + "loss": 0.6764, + "step": 16679 + }, + { + "epoch": 0.61, + "learning_rate": 1.3909986450360712e-07, + "loss": 0.702, + "step": 16680 + }, + { + "epoch": 0.61, + "learning_rate": 1.3909620243893508e-07, + "loss": 0.6882, + "step": 16681 + }, + { + "epoch": 0.61, + "learning_rate": 1.39092540374263e-07, + "loss": 0.6879, + "step": 16682 + }, + { + "epoch": 0.61, + "learning_rate": 1.3908887830959096e-07, + "loss": 0.696, + "step": 16683 + }, + { + "epoch": 0.61, + "learning_rate": 1.3908521624491886e-07, + "loss": 0.6989, + "step": 16684 + }, + { + "epoch": 0.61, + "learning_rate": 1.390815541802468e-07, + "loss": 0.6855, + "step": 16685 + }, + { + "epoch": 0.61, + "learning_rate": 1.3907789211557474e-07, + "loss": 0.7138, + "step": 16686 + }, + { + "epoch": 0.61, + "learning_rate": 1.390742300509027e-07, + "loss": 0.6949, + "step": 16687 + }, + { + "epoch": 0.61, + "learning_rate": 1.3907056798623062e-07, + "loss": 0.6887, + "step": 16688 + }, + { + "epoch": 0.61, + "learning_rate": 1.3906690592155857e-07, + "loss": 0.6932, + "step": 16689 + }, + { + "epoch": 0.61, + "learning_rate": 1.390632438568865e-07, + "loss": 0.7149, + "step": 16690 + }, + { + "epoch": 0.61, + "learning_rate": 1.3905958179221445e-07, + "loss": 0.7139, + "step": 16691 + }, + { + "epoch": 0.61, + "learning_rate": 1.3905591972754238e-07, + "loss": 0.6773, + "step": 16692 + }, + { + "epoch": 0.61, + "learning_rate": 1.3905225766287033e-07, + "loss": 0.7051, + "step": 16693 + }, + { + "epoch": 0.61, + "learning_rate": 1.3904859559819826e-07, + "loss": 0.6917, + "step": 16694 + }, + { + "epoch": 0.61, + "learning_rate": 1.390449335335262e-07, + "loss": 0.6843, + "step": 16695 + }, + { + "epoch": 0.61, + "learning_rate": 1.3904127146885413e-07, + "loss": 0.7061, + "step": 16696 + }, + { + "epoch": 0.61, + "learning_rate": 1.390376094041821e-07, + "loss": 0.6957, + "step": 16697 + }, + { + "epoch": 0.61, + "learning_rate": 1.3903394733951001e-07, + "loss": 0.6968, + "step": 16698 + }, + { + "epoch": 0.61, + "learning_rate": 1.3903028527483797e-07, + "loss": 0.6866, + "step": 16699 + }, + { + "epoch": 0.61, + "learning_rate": 1.390266232101659e-07, + "loss": 0.6925, + "step": 16700 + }, + { + "epoch": 0.61, + "learning_rate": 1.3902296114549382e-07, + "loss": 0.6779, + "step": 16701 + }, + { + "epoch": 0.61, + "learning_rate": 1.3901929908082175e-07, + "loss": 0.6849, + "step": 16702 + }, + { + "epoch": 0.61, + "learning_rate": 1.390156370161497e-07, + "loss": 0.6945, + "step": 16703 + }, + { + "epoch": 0.61, + "learning_rate": 1.3901197495147763e-07, + "loss": 0.6925, + "step": 16704 + }, + { + "epoch": 0.61, + "learning_rate": 1.3900831288680555e-07, + "loss": 0.6883, + "step": 16705 + }, + { + "epoch": 0.61, + "learning_rate": 1.390046508221335e-07, + "loss": 0.6882, + "step": 16706 + }, + { + "epoch": 0.61, + "learning_rate": 1.3900098875746143e-07, + "loss": 0.688, + "step": 16707 + }, + { + "epoch": 0.61, + "learning_rate": 1.389973266927894e-07, + "loss": 0.7034, + "step": 16708 + }, + { + "epoch": 0.61, + "learning_rate": 1.3899366462811731e-07, + "loss": 0.6884, + "step": 16709 + }, + { + "epoch": 0.61, + "learning_rate": 1.3899000256344527e-07, + "loss": 0.6952, + "step": 16710 + }, + { + "epoch": 0.61, + "learning_rate": 1.389863404987732e-07, + "loss": 0.6839, + "step": 16711 + }, + { + "epoch": 0.61, + "learning_rate": 1.3898267843410115e-07, + "loss": 0.6904, + "step": 16712 + }, + { + "epoch": 0.61, + "learning_rate": 1.3897901636942907e-07, + "loss": 0.6886, + "step": 16713 + }, + { + "epoch": 0.61, + "learning_rate": 1.3897535430475703e-07, + "loss": 0.7007, + "step": 16714 + }, + { + "epoch": 0.61, + "learning_rate": 1.3897169224008495e-07, + "loss": 0.6831, + "step": 16715 + }, + { + "epoch": 0.61, + "learning_rate": 1.389680301754129e-07, + "loss": 0.6664, + "step": 16716 + }, + { + "epoch": 0.61, + "learning_rate": 1.3896436811074083e-07, + "loss": 0.7038, + "step": 16717 + }, + { + "epoch": 0.61, + "learning_rate": 1.389607060460688e-07, + "loss": 0.6898, + "step": 16718 + }, + { + "epoch": 0.61, + "learning_rate": 1.3895704398139671e-07, + "loss": 0.6956, + "step": 16719 + }, + { + "epoch": 0.61, + "learning_rate": 1.3895338191672464e-07, + "loss": 0.6811, + "step": 16720 + }, + { + "epoch": 0.61, + "learning_rate": 1.3894971985205257e-07, + "loss": 0.7062, + "step": 16721 + }, + { + "epoch": 0.61, + "learning_rate": 1.3894605778738052e-07, + "loss": 0.7023, + "step": 16722 + }, + { + "epoch": 0.61, + "learning_rate": 1.3894239572270845e-07, + "loss": 0.6947, + "step": 16723 + }, + { + "epoch": 0.61, + "learning_rate": 1.389387336580364e-07, + "loss": 0.6827, + "step": 16724 + }, + { + "epoch": 0.61, + "learning_rate": 1.3893507159336433e-07, + "loss": 0.6981, + "step": 16725 + }, + { + "epoch": 0.61, + "learning_rate": 1.3893140952869228e-07, + "loss": 0.7078, + "step": 16726 + }, + { + "epoch": 0.61, + "learning_rate": 1.389277474640202e-07, + "loss": 0.6892, + "step": 16727 + }, + { + "epoch": 0.61, + "learning_rate": 1.3892408539934813e-07, + "loss": 0.7001, + "step": 16728 + }, + { + "epoch": 0.61, + "learning_rate": 1.3892042333467609e-07, + "loss": 0.6981, + "step": 16729 + }, + { + "epoch": 0.61, + "learning_rate": 1.38916761270004e-07, + "loss": 0.7043, + "step": 16730 + }, + { + "epoch": 0.61, + "learning_rate": 1.3891309920533197e-07, + "loss": 0.7031, + "step": 16731 + }, + { + "epoch": 0.61, + "learning_rate": 1.389094371406599e-07, + "loss": 0.6914, + "step": 16732 + }, + { + "epoch": 0.61, + "learning_rate": 1.3890577507598785e-07, + "loss": 0.6927, + "step": 16733 + }, + { + "epoch": 0.61, + "learning_rate": 1.3890211301131577e-07, + "loss": 0.6868, + "step": 16734 + }, + { + "epoch": 0.61, + "learning_rate": 1.3889845094664373e-07, + "loss": 0.7065, + "step": 16735 + }, + { + "epoch": 0.61, + "learning_rate": 1.3889478888197165e-07, + "loss": 0.706, + "step": 16736 + }, + { + "epoch": 0.61, + "learning_rate": 1.388911268172996e-07, + "loss": 0.695, + "step": 16737 + }, + { + "epoch": 0.61, + "learning_rate": 1.388874647526275e-07, + "loss": 0.6883, + "step": 16738 + }, + { + "epoch": 0.61, + "learning_rate": 1.3888380268795546e-07, + "loss": 0.7046, + "step": 16739 + }, + { + "epoch": 0.61, + "learning_rate": 1.3888014062328339e-07, + "loss": 0.6877, + "step": 16740 + }, + { + "epoch": 0.61, + "learning_rate": 1.3887647855861134e-07, + "loss": 0.692, + "step": 16741 + }, + { + "epoch": 0.61, + "learning_rate": 1.3887281649393927e-07, + "loss": 0.6974, + "step": 16742 + }, + { + "epoch": 0.61, + "learning_rate": 1.3886915442926722e-07, + "loss": 0.6957, + "step": 16743 + }, + { + "epoch": 0.61, + "learning_rate": 1.3886549236459515e-07, + "loss": 0.6943, + "step": 16744 + }, + { + "epoch": 0.61, + "learning_rate": 1.388618302999231e-07, + "loss": 0.69, + "step": 16745 + }, + { + "epoch": 0.61, + "learning_rate": 1.3885816823525103e-07, + "loss": 0.6806, + "step": 16746 + }, + { + "epoch": 0.61, + "learning_rate": 1.3885450617057898e-07, + "loss": 0.704, + "step": 16747 + }, + { + "epoch": 0.61, + "learning_rate": 1.388508441059069e-07, + "loss": 0.7006, + "step": 16748 + }, + { + "epoch": 0.61, + "learning_rate": 1.3884718204123483e-07, + "loss": 0.6967, + "step": 16749 + }, + { + "epoch": 0.61, + "learning_rate": 1.3884351997656279e-07, + "loss": 0.6995, + "step": 16750 + }, + { + "epoch": 0.61, + "learning_rate": 1.388398579118907e-07, + "loss": 0.6906, + "step": 16751 + }, + { + "epoch": 0.61, + "learning_rate": 1.3883619584721867e-07, + "loss": 0.6882, + "step": 16752 + }, + { + "epoch": 0.61, + "learning_rate": 1.388325337825466e-07, + "loss": 0.7299, + "step": 16753 + }, + { + "epoch": 0.61, + "learning_rate": 1.3882887171787454e-07, + "loss": 0.6982, + "step": 16754 + }, + { + "epoch": 0.61, + "learning_rate": 1.3882520965320247e-07, + "loss": 0.7004, + "step": 16755 + }, + { + "epoch": 0.61, + "learning_rate": 1.388215475885304e-07, + "loss": 0.7126, + "step": 16756 + }, + { + "epoch": 0.61, + "learning_rate": 1.3881788552385832e-07, + "loss": 0.6879, + "step": 16757 + }, + { + "epoch": 0.61, + "learning_rate": 1.3881422345918628e-07, + "loss": 0.6954, + "step": 16758 + }, + { + "epoch": 0.61, + "learning_rate": 1.388105613945142e-07, + "loss": 0.6898, + "step": 16759 + }, + { + "epoch": 0.61, + "learning_rate": 1.3880689932984216e-07, + "loss": 0.6972, + "step": 16760 + }, + { + "epoch": 0.61, + "learning_rate": 1.3880323726517008e-07, + "loss": 0.6878, + "step": 16761 + }, + { + "epoch": 0.61, + "learning_rate": 1.3879957520049804e-07, + "loss": 0.7118, + "step": 16762 + }, + { + "epoch": 0.61, + "learning_rate": 1.3879591313582596e-07, + "loss": 0.6954, + "step": 16763 + }, + { + "epoch": 0.61, + "learning_rate": 1.3879225107115392e-07, + "loss": 0.6959, + "step": 16764 + }, + { + "epoch": 0.61, + "learning_rate": 1.3878858900648184e-07, + "loss": 0.6833, + "step": 16765 + }, + { + "epoch": 0.61, + "learning_rate": 1.387849269418098e-07, + "loss": 0.6946, + "step": 16766 + }, + { + "epoch": 0.61, + "learning_rate": 1.3878126487713772e-07, + "loss": 0.6963, + "step": 16767 + }, + { + "epoch": 0.61, + "learning_rate": 1.3877760281246568e-07, + "loss": 0.6793, + "step": 16768 + }, + { + "epoch": 0.61, + "learning_rate": 1.387739407477936e-07, + "loss": 0.6982, + "step": 16769 + }, + { + "epoch": 0.61, + "learning_rate": 1.3877027868312156e-07, + "loss": 0.7045, + "step": 16770 + }, + { + "epoch": 0.61, + "learning_rate": 1.3876661661844948e-07, + "loss": 0.6928, + "step": 16771 + }, + { + "epoch": 0.61, + "learning_rate": 1.387629545537774e-07, + "loss": 0.682, + "step": 16772 + }, + { + "epoch": 0.61, + "learning_rate": 1.3875929248910536e-07, + "loss": 0.6854, + "step": 16773 + }, + { + "epoch": 0.61, + "learning_rate": 1.387556304244333e-07, + "loss": 0.6844, + "step": 16774 + }, + { + "epoch": 0.61, + "learning_rate": 1.3875196835976122e-07, + "loss": 0.6894, + "step": 16775 + }, + { + "epoch": 0.61, + "learning_rate": 1.3874830629508914e-07, + "loss": 0.692, + "step": 16776 + }, + { + "epoch": 0.61, + "learning_rate": 1.387446442304171e-07, + "loss": 0.6876, + "step": 16777 + }, + { + "epoch": 0.61, + "learning_rate": 1.3874098216574502e-07, + "loss": 0.698, + "step": 16778 + }, + { + "epoch": 0.61, + "learning_rate": 1.3873732010107298e-07, + "loss": 0.6897, + "step": 16779 + }, + { + "epoch": 0.61, + "learning_rate": 1.387336580364009e-07, + "loss": 0.686, + "step": 16780 + }, + { + "epoch": 0.61, + "learning_rate": 1.3872999597172886e-07, + "loss": 0.6982, + "step": 16781 + }, + { + "epoch": 0.61, + "learning_rate": 1.3872633390705678e-07, + "loss": 0.6915, + "step": 16782 + }, + { + "epoch": 0.61, + "learning_rate": 1.3872267184238474e-07, + "loss": 0.6986, + "step": 16783 + }, + { + "epoch": 0.61, + "learning_rate": 1.3871900977771266e-07, + "loss": 0.6966, + "step": 16784 + }, + { + "epoch": 0.61, + "learning_rate": 1.3871534771304062e-07, + "loss": 0.6909, + "step": 16785 + }, + { + "epoch": 0.61, + "learning_rate": 1.3871168564836854e-07, + "loss": 0.6987, + "step": 16786 + }, + { + "epoch": 0.61, + "learning_rate": 1.387080235836965e-07, + "loss": 0.693, + "step": 16787 + }, + { + "epoch": 0.61, + "learning_rate": 1.3870436151902442e-07, + "loss": 0.7065, + "step": 16788 + }, + { + "epoch": 0.61, + "learning_rate": 1.3870069945435238e-07, + "loss": 0.6924, + "step": 16789 + }, + { + "epoch": 0.61, + "learning_rate": 1.386970373896803e-07, + "loss": 0.6865, + "step": 16790 + }, + { + "epoch": 0.61, + "learning_rate": 1.3869337532500826e-07, + "loss": 0.7029, + "step": 16791 + }, + { + "epoch": 0.61, + "learning_rate": 1.3868971326033616e-07, + "loss": 0.6914, + "step": 16792 + }, + { + "epoch": 0.61, + "learning_rate": 1.386860511956641e-07, + "loss": 0.6981, + "step": 16793 + }, + { + "epoch": 0.61, + "learning_rate": 1.3868238913099204e-07, + "loss": 0.6862, + "step": 16794 + }, + { + "epoch": 0.61, + "learning_rate": 1.3867872706632e-07, + "loss": 0.7051, + "step": 16795 + }, + { + "epoch": 0.61, + "learning_rate": 1.3867506500164792e-07, + "loss": 0.6951, + "step": 16796 + }, + { + "epoch": 0.61, + "learning_rate": 1.3867140293697587e-07, + "loss": 0.6917, + "step": 16797 + }, + { + "epoch": 0.61, + "learning_rate": 1.386677408723038e-07, + "loss": 0.6905, + "step": 16798 + }, + { + "epoch": 0.61, + "learning_rate": 1.3866407880763172e-07, + "loss": 0.6809, + "step": 16799 + }, + { + "epoch": 0.61, + "learning_rate": 1.3866041674295968e-07, + "loss": 0.7069, + "step": 16800 + }, + { + "epoch": 0.61, + "learning_rate": 1.386567546782876e-07, + "loss": 0.6912, + "step": 16801 + }, + { + "epoch": 0.61, + "learning_rate": 1.3865309261361556e-07, + "loss": 0.6945, + "step": 16802 + }, + { + "epoch": 0.61, + "learning_rate": 1.3864943054894348e-07, + "loss": 0.681, + "step": 16803 + }, + { + "epoch": 0.61, + "learning_rate": 1.3864576848427144e-07, + "loss": 0.687, + "step": 16804 + }, + { + "epoch": 0.61, + "learning_rate": 1.3864210641959936e-07, + "loss": 0.6888, + "step": 16805 + }, + { + "epoch": 0.61, + "learning_rate": 1.3863844435492732e-07, + "loss": 0.684, + "step": 16806 + }, + { + "epoch": 0.61, + "learning_rate": 1.3863478229025524e-07, + "loss": 0.6799, + "step": 16807 + }, + { + "epoch": 0.61, + "learning_rate": 1.386311202255832e-07, + "loss": 0.6941, + "step": 16808 + }, + { + "epoch": 0.61, + "learning_rate": 1.3862745816091112e-07, + "loss": 0.6908, + "step": 16809 + }, + { + "epoch": 0.62, + "learning_rate": 1.3862379609623905e-07, + "loss": 0.6677, + "step": 16810 + }, + { + "epoch": 0.62, + "learning_rate": 1.3862013403156698e-07, + "loss": 0.6955, + "step": 16811 + }, + { + "epoch": 0.62, + "learning_rate": 1.3861647196689493e-07, + "loss": 0.6705, + "step": 16812 + }, + { + "epoch": 0.62, + "learning_rate": 1.3861280990222286e-07, + "loss": 0.6972, + "step": 16813 + }, + { + "epoch": 0.62, + "learning_rate": 1.386091478375508e-07, + "loss": 0.7036, + "step": 16814 + }, + { + "epoch": 0.62, + "learning_rate": 1.3860548577287873e-07, + "loss": 0.7044, + "step": 16815 + }, + { + "epoch": 0.62, + "learning_rate": 1.386018237082067e-07, + "loss": 0.6899, + "step": 16816 + }, + { + "epoch": 0.62, + "learning_rate": 1.3859816164353461e-07, + "loss": 0.6871, + "step": 16817 + }, + { + "epoch": 0.62, + "learning_rate": 1.3859449957886257e-07, + "loss": 0.7026, + "step": 16818 + }, + { + "epoch": 0.62, + "learning_rate": 1.385908375141905e-07, + "loss": 0.6823, + "step": 16819 + }, + { + "epoch": 0.62, + "learning_rate": 1.3858717544951842e-07, + "loss": 0.6884, + "step": 16820 + }, + { + "epoch": 0.62, + "learning_rate": 1.3858351338484637e-07, + "loss": 0.6981, + "step": 16821 + }, + { + "epoch": 0.62, + "learning_rate": 1.385798513201743e-07, + "loss": 0.6828, + "step": 16822 + }, + { + "epoch": 0.62, + "learning_rate": 1.3857618925550225e-07, + "loss": 0.6941, + "step": 16823 + }, + { + "epoch": 0.62, + "learning_rate": 1.3857252719083018e-07, + "loss": 0.6995, + "step": 16824 + }, + { + "epoch": 0.62, + "learning_rate": 1.3856886512615813e-07, + "loss": 0.6877, + "step": 16825 + }, + { + "epoch": 0.62, + "learning_rate": 1.3856520306148606e-07, + "loss": 0.6835, + "step": 16826 + }, + { + "epoch": 0.62, + "learning_rate": 1.3856154099681401e-07, + "loss": 0.7001, + "step": 16827 + }, + { + "epoch": 0.62, + "learning_rate": 1.3855787893214191e-07, + "loss": 0.7113, + "step": 16828 + }, + { + "epoch": 0.62, + "learning_rate": 1.3855421686746987e-07, + "loss": 0.6793, + "step": 16829 + }, + { + "epoch": 0.62, + "learning_rate": 1.385505548027978e-07, + "loss": 0.7054, + "step": 16830 + }, + { + "epoch": 0.62, + "learning_rate": 1.3854689273812575e-07, + "loss": 0.7005, + "step": 16831 + }, + { + "epoch": 0.62, + "learning_rate": 1.3854323067345367e-07, + "loss": 0.696, + "step": 16832 + }, + { + "epoch": 0.62, + "learning_rate": 1.3853956860878163e-07, + "loss": 0.6958, + "step": 16833 + }, + { + "epoch": 0.62, + "learning_rate": 1.3853590654410955e-07, + "loss": 0.6782, + "step": 16834 + }, + { + "epoch": 0.62, + "learning_rate": 1.385322444794375e-07, + "loss": 0.703, + "step": 16835 + }, + { + "epoch": 0.62, + "learning_rate": 1.3852858241476543e-07, + "loss": 0.6956, + "step": 16836 + }, + { + "epoch": 0.62, + "learning_rate": 1.385249203500934e-07, + "loss": 0.689, + "step": 16837 + }, + { + "epoch": 0.62, + "learning_rate": 1.3852125828542131e-07, + "loss": 0.6838, + "step": 16838 + }, + { + "epoch": 0.62, + "learning_rate": 1.3851759622074927e-07, + "loss": 0.6978, + "step": 16839 + }, + { + "epoch": 0.62, + "learning_rate": 1.385139341560772e-07, + "loss": 0.6932, + "step": 16840 + }, + { + "epoch": 0.62, + "learning_rate": 1.3851027209140515e-07, + "loss": 0.6981, + "step": 16841 + }, + { + "epoch": 0.62, + "learning_rate": 1.3850661002673307e-07, + "loss": 0.7007, + "step": 16842 + }, + { + "epoch": 0.62, + "learning_rate": 1.38502947962061e-07, + "loss": 0.6913, + "step": 16843 + }, + { + "epoch": 0.62, + "learning_rate": 1.3849928589738895e-07, + "loss": 0.6803, + "step": 16844 + }, + { + "epoch": 0.62, + "learning_rate": 1.3849562383271688e-07, + "loss": 0.6872, + "step": 16845 + }, + { + "epoch": 0.62, + "learning_rate": 1.384919617680448e-07, + "loss": 0.6892, + "step": 16846 + }, + { + "epoch": 0.62, + "learning_rate": 1.3848829970337273e-07, + "loss": 0.7048, + "step": 16847 + }, + { + "epoch": 0.62, + "learning_rate": 1.3848463763870069e-07, + "loss": 0.6732, + "step": 16848 + }, + { + "epoch": 0.62, + "learning_rate": 1.384809755740286e-07, + "loss": 0.6916, + "step": 16849 + }, + { + "epoch": 0.62, + "learning_rate": 1.3847731350935657e-07, + "loss": 0.69, + "step": 16850 + }, + { + "epoch": 0.62, + "learning_rate": 1.384736514446845e-07, + "loss": 0.7062, + "step": 16851 + }, + { + "epoch": 0.62, + "learning_rate": 1.3846998938001245e-07, + "loss": 0.6953, + "step": 16852 + }, + { + "epoch": 0.62, + "learning_rate": 1.3846632731534037e-07, + "loss": 0.7001, + "step": 16853 + }, + { + "epoch": 0.62, + "learning_rate": 1.3846266525066833e-07, + "loss": 0.6928, + "step": 16854 + }, + { + "epoch": 0.62, + "learning_rate": 1.3845900318599625e-07, + "loss": 0.6933, + "step": 16855 + }, + { + "epoch": 0.62, + "learning_rate": 1.384553411213242e-07, + "loss": 0.6854, + "step": 16856 + }, + { + "epoch": 0.62, + "learning_rate": 1.3845167905665213e-07, + "loss": 0.7034, + "step": 16857 + }, + { + "epoch": 0.62, + "learning_rate": 1.3844801699198009e-07, + "loss": 0.6953, + "step": 16858 + }, + { + "epoch": 0.62, + "learning_rate": 1.38444354927308e-07, + "loss": 0.6874, + "step": 16859 + }, + { + "epoch": 0.62, + "learning_rate": 1.3844069286263597e-07, + "loss": 0.6925, + "step": 16860 + }, + { + "epoch": 0.62, + "learning_rate": 1.384370307979639e-07, + "loss": 0.6761, + "step": 16861 + }, + { + "epoch": 0.62, + "learning_rate": 1.3843336873329185e-07, + "loss": 0.6955, + "step": 16862 + }, + { + "epoch": 0.62, + "learning_rate": 1.3842970666861977e-07, + "loss": 0.6799, + "step": 16863 + }, + { + "epoch": 0.62, + "learning_rate": 1.384260446039477e-07, + "loss": 0.6805, + "step": 16864 + }, + { + "epoch": 0.62, + "learning_rate": 1.3842238253927563e-07, + "loss": 0.6833, + "step": 16865 + }, + { + "epoch": 0.62, + "learning_rate": 1.3841872047460358e-07, + "loss": 0.6908, + "step": 16866 + }, + { + "epoch": 0.62, + "learning_rate": 1.384150584099315e-07, + "loss": 0.677, + "step": 16867 + }, + { + "epoch": 0.62, + "learning_rate": 1.3841139634525946e-07, + "loss": 0.6895, + "step": 16868 + }, + { + "epoch": 0.62, + "learning_rate": 1.3840773428058739e-07, + "loss": 0.6974, + "step": 16869 + }, + { + "epoch": 0.62, + "learning_rate": 1.384040722159153e-07, + "loss": 0.6861, + "step": 16870 + }, + { + "epoch": 0.62, + "learning_rate": 1.3840041015124327e-07, + "loss": 0.6912, + "step": 16871 + }, + { + "epoch": 0.62, + "learning_rate": 1.383967480865712e-07, + "loss": 0.6945, + "step": 16872 + }, + { + "epoch": 0.62, + "learning_rate": 1.3839308602189914e-07, + "loss": 0.7097, + "step": 16873 + }, + { + "epoch": 0.62, + "learning_rate": 1.3838942395722707e-07, + "loss": 0.6975, + "step": 16874 + }, + { + "epoch": 0.62, + "learning_rate": 1.3838576189255502e-07, + "loss": 0.6789, + "step": 16875 + }, + { + "epoch": 0.62, + "learning_rate": 1.3838209982788295e-07, + "loss": 0.6774, + "step": 16876 + }, + { + "epoch": 0.62, + "learning_rate": 1.383784377632109e-07, + "loss": 0.6859, + "step": 16877 + }, + { + "epoch": 0.62, + "learning_rate": 1.3837477569853883e-07, + "loss": 0.6946, + "step": 16878 + }, + { + "epoch": 0.62, + "learning_rate": 1.3837111363386678e-07, + "loss": 0.6938, + "step": 16879 + }, + { + "epoch": 0.62, + "learning_rate": 1.383674515691947e-07, + "loss": 0.6915, + "step": 16880 + }, + { + "epoch": 0.62, + "learning_rate": 1.3836378950452266e-07, + "loss": 0.6975, + "step": 16881 + }, + { + "epoch": 0.62, + "learning_rate": 1.3836012743985056e-07, + "loss": 0.6834, + "step": 16882 + }, + { + "epoch": 0.62, + "learning_rate": 1.3835646537517852e-07, + "loss": 0.7003, + "step": 16883 + }, + { + "epoch": 0.62, + "learning_rate": 1.3835280331050644e-07, + "loss": 0.694, + "step": 16884 + }, + { + "epoch": 0.62, + "learning_rate": 1.383491412458344e-07, + "loss": 0.6995, + "step": 16885 + }, + { + "epoch": 0.62, + "learning_rate": 1.3834547918116232e-07, + "loss": 0.6825, + "step": 16886 + }, + { + "epoch": 0.62, + "learning_rate": 1.3834181711649028e-07, + "loss": 0.6919, + "step": 16887 + }, + { + "epoch": 0.62, + "learning_rate": 1.383381550518182e-07, + "loss": 0.712, + "step": 16888 + }, + { + "epoch": 0.62, + "learning_rate": 1.3833449298714616e-07, + "loss": 0.6944, + "step": 16889 + }, + { + "epoch": 0.62, + "learning_rate": 1.3833083092247408e-07, + "loss": 0.6817, + "step": 16890 + }, + { + "epoch": 0.62, + "learning_rate": 1.38327168857802e-07, + "loss": 0.6797, + "step": 16891 + }, + { + "epoch": 0.62, + "learning_rate": 1.3832350679312996e-07, + "loss": 0.6983, + "step": 16892 + }, + { + "epoch": 0.62, + "learning_rate": 1.383198447284579e-07, + "loss": 0.6849, + "step": 16893 + }, + { + "epoch": 0.62, + "learning_rate": 1.3831618266378584e-07, + "loss": 0.6984, + "step": 16894 + }, + { + "epoch": 0.62, + "learning_rate": 1.3831252059911377e-07, + "loss": 0.7032, + "step": 16895 + }, + { + "epoch": 0.62, + "learning_rate": 1.3830885853444172e-07, + "loss": 0.6902, + "step": 16896 + }, + { + "epoch": 0.62, + "learning_rate": 1.3830519646976965e-07, + "loss": 0.7026, + "step": 16897 + }, + { + "epoch": 0.62, + "learning_rate": 1.383015344050976e-07, + "loss": 0.6834, + "step": 16898 + }, + { + "epoch": 0.62, + "learning_rate": 1.3829787234042553e-07, + "loss": 0.6923, + "step": 16899 + }, + { + "epoch": 0.62, + "learning_rate": 1.3829421027575346e-07, + "loss": 0.7003, + "step": 16900 + }, + { + "epoch": 0.62, + "learning_rate": 1.3829054821108138e-07, + "loss": 0.6931, + "step": 16901 + }, + { + "epoch": 0.62, + "learning_rate": 1.3828688614640934e-07, + "loss": 0.6881, + "step": 16902 + }, + { + "epoch": 0.62, + "learning_rate": 1.3828322408173726e-07, + "loss": 0.7009, + "step": 16903 + }, + { + "epoch": 0.62, + "learning_rate": 1.3827956201706522e-07, + "loss": 0.6931, + "step": 16904 + }, + { + "epoch": 0.62, + "learning_rate": 1.3827589995239314e-07, + "loss": 0.6922, + "step": 16905 + }, + { + "epoch": 0.62, + "learning_rate": 1.382722378877211e-07, + "loss": 0.6974, + "step": 16906 + }, + { + "epoch": 0.62, + "learning_rate": 1.3826857582304902e-07, + "loss": 0.7005, + "step": 16907 + }, + { + "epoch": 0.62, + "learning_rate": 1.3826491375837698e-07, + "loss": 0.6953, + "step": 16908 + }, + { + "epoch": 0.62, + "learning_rate": 1.382612516937049e-07, + "loss": 0.6914, + "step": 16909 + }, + { + "epoch": 0.62, + "learning_rate": 1.3825758962903286e-07, + "loss": 0.7065, + "step": 16910 + }, + { + "epoch": 0.62, + "learning_rate": 1.3825392756436078e-07, + "loss": 0.6912, + "step": 16911 + }, + { + "epoch": 0.62, + "learning_rate": 1.3825026549968874e-07, + "loss": 0.6899, + "step": 16912 + }, + { + "epoch": 0.62, + "learning_rate": 1.3824660343501666e-07, + "loss": 0.6883, + "step": 16913 + }, + { + "epoch": 0.62, + "learning_rate": 1.382429413703446e-07, + "loss": 0.6929, + "step": 16914 + }, + { + "epoch": 0.62, + "learning_rate": 1.3823927930567254e-07, + "loss": 0.683, + "step": 16915 + }, + { + "epoch": 0.62, + "learning_rate": 1.3823561724100047e-07, + "loss": 0.6904, + "step": 16916 + }, + { + "epoch": 0.62, + "learning_rate": 1.3823195517632842e-07, + "loss": 0.6957, + "step": 16917 + }, + { + "epoch": 0.62, + "learning_rate": 1.3822829311165632e-07, + "loss": 0.7006, + "step": 16918 + }, + { + "epoch": 0.62, + "learning_rate": 1.3822463104698428e-07, + "loss": 0.6919, + "step": 16919 + }, + { + "epoch": 0.62, + "learning_rate": 1.382209689823122e-07, + "loss": 0.6968, + "step": 16920 + }, + { + "epoch": 0.62, + "learning_rate": 1.3821730691764016e-07, + "loss": 0.6926, + "step": 16921 + }, + { + "epoch": 0.62, + "learning_rate": 1.3821364485296808e-07, + "loss": 0.6987, + "step": 16922 + }, + { + "epoch": 0.62, + "learning_rate": 1.3820998278829604e-07, + "loss": 0.6873, + "step": 16923 + }, + { + "epoch": 0.62, + "learning_rate": 1.3820632072362396e-07, + "loss": 0.7004, + "step": 16924 + }, + { + "epoch": 0.62, + "learning_rate": 1.3820265865895192e-07, + "loss": 0.696, + "step": 16925 + }, + { + "epoch": 0.62, + "learning_rate": 1.3819899659427984e-07, + "loss": 0.7016, + "step": 16926 + }, + { + "epoch": 0.62, + "learning_rate": 1.381953345296078e-07, + "loss": 0.6963, + "step": 16927 + }, + { + "epoch": 0.62, + "learning_rate": 1.3819167246493572e-07, + "loss": 0.681, + "step": 16928 + }, + { + "epoch": 0.62, + "learning_rate": 1.3818801040026368e-07, + "loss": 0.6909, + "step": 16929 + }, + { + "epoch": 0.62, + "learning_rate": 1.381843483355916e-07, + "loss": 0.6847, + "step": 16930 + }, + { + "epoch": 0.62, + "learning_rate": 1.3818068627091955e-07, + "loss": 0.7087, + "step": 16931 + }, + { + "epoch": 0.62, + "learning_rate": 1.3817702420624748e-07, + "loss": 0.6961, + "step": 16932 + }, + { + "epoch": 0.62, + "learning_rate": 1.3817336214157543e-07, + "loss": 0.6902, + "step": 16933 + }, + { + "epoch": 0.62, + "learning_rate": 1.3816970007690336e-07, + "loss": 0.7065, + "step": 16934 + }, + { + "epoch": 0.62, + "learning_rate": 1.381660380122313e-07, + "loss": 0.6967, + "step": 16935 + }, + { + "epoch": 0.62, + "learning_rate": 1.3816237594755921e-07, + "loss": 0.7069, + "step": 16936 + }, + { + "epoch": 0.62, + "learning_rate": 1.3815871388288717e-07, + "loss": 0.6932, + "step": 16937 + }, + { + "epoch": 0.62, + "learning_rate": 1.381550518182151e-07, + "loss": 0.6956, + "step": 16938 + }, + { + "epoch": 0.62, + "learning_rate": 1.3815138975354305e-07, + "loss": 0.6857, + "step": 16939 + }, + { + "epoch": 0.62, + "learning_rate": 1.3814772768887097e-07, + "loss": 0.6955, + "step": 16940 + }, + { + "epoch": 0.62, + "learning_rate": 1.381440656241989e-07, + "loss": 0.6778, + "step": 16941 + }, + { + "epoch": 0.62, + "learning_rate": 1.3814040355952685e-07, + "loss": 0.6885, + "step": 16942 + }, + { + "epoch": 0.62, + "learning_rate": 1.3813674149485478e-07, + "loss": 0.6954, + "step": 16943 + }, + { + "epoch": 0.62, + "learning_rate": 1.3813307943018273e-07, + "loss": 0.6898, + "step": 16944 + }, + { + "epoch": 0.62, + "learning_rate": 1.3812941736551066e-07, + "loss": 0.6925, + "step": 16945 + }, + { + "epoch": 0.62, + "learning_rate": 1.3812575530083861e-07, + "loss": 0.7088, + "step": 16946 + }, + { + "epoch": 0.62, + "learning_rate": 1.3812209323616654e-07, + "loss": 0.6927, + "step": 16947 + }, + { + "epoch": 0.62, + "learning_rate": 1.381184311714945e-07, + "loss": 0.6951, + "step": 16948 + }, + { + "epoch": 0.62, + "learning_rate": 1.3811476910682242e-07, + "loss": 0.6964, + "step": 16949 + }, + { + "epoch": 0.62, + "learning_rate": 1.3811110704215037e-07, + "loss": 0.6978, + "step": 16950 + }, + { + "epoch": 0.62, + "learning_rate": 1.381074449774783e-07, + "loss": 0.6878, + "step": 16951 + }, + { + "epoch": 0.62, + "learning_rate": 1.3810378291280625e-07, + "loss": 0.6926, + "step": 16952 + }, + { + "epoch": 0.62, + "learning_rate": 1.3810012084813418e-07, + "loss": 0.6916, + "step": 16953 + }, + { + "epoch": 0.62, + "learning_rate": 1.380964587834621e-07, + "loss": 0.6955, + "step": 16954 + }, + { + "epoch": 0.62, + "learning_rate": 1.3809279671879003e-07, + "loss": 0.7039, + "step": 16955 + }, + { + "epoch": 0.62, + "learning_rate": 1.38089134654118e-07, + "loss": 0.6806, + "step": 16956 + }, + { + "epoch": 0.62, + "learning_rate": 1.3808547258944591e-07, + "loss": 0.6814, + "step": 16957 + }, + { + "epoch": 0.62, + "learning_rate": 1.3808181052477387e-07, + "loss": 0.6858, + "step": 16958 + }, + { + "epoch": 0.62, + "learning_rate": 1.380781484601018e-07, + "loss": 0.6806, + "step": 16959 + }, + { + "epoch": 0.62, + "learning_rate": 1.3807448639542975e-07, + "loss": 0.6813, + "step": 16960 + }, + { + "epoch": 0.62, + "learning_rate": 1.3807082433075767e-07, + "loss": 0.6858, + "step": 16961 + }, + { + "epoch": 0.62, + "learning_rate": 1.380671622660856e-07, + "loss": 0.6934, + "step": 16962 + }, + { + "epoch": 0.62, + "learning_rate": 1.3806350020141355e-07, + "loss": 0.6887, + "step": 16963 + }, + { + "epoch": 0.62, + "learning_rate": 1.3805983813674148e-07, + "loss": 0.6917, + "step": 16964 + }, + { + "epoch": 0.62, + "learning_rate": 1.3805617607206943e-07, + "loss": 0.6978, + "step": 16965 + }, + { + "epoch": 0.62, + "learning_rate": 1.3805251400739736e-07, + "loss": 0.7075, + "step": 16966 + }, + { + "epoch": 0.62, + "learning_rate": 1.380488519427253e-07, + "loss": 0.6852, + "step": 16967 + }, + { + "epoch": 0.62, + "learning_rate": 1.3804518987805324e-07, + "loss": 0.6913, + "step": 16968 + }, + { + "epoch": 0.62, + "learning_rate": 1.380415278133812e-07, + "loss": 0.6823, + "step": 16969 + }, + { + "epoch": 0.62, + "learning_rate": 1.3803786574870912e-07, + "loss": 0.7075, + "step": 16970 + }, + { + "epoch": 0.62, + "learning_rate": 1.3803420368403707e-07, + "loss": 0.6869, + "step": 16971 + }, + { + "epoch": 0.62, + "learning_rate": 1.3803054161936497e-07, + "loss": 0.6966, + "step": 16972 + }, + { + "epoch": 0.62, + "learning_rate": 1.3802687955469293e-07, + "loss": 0.6924, + "step": 16973 + }, + { + "epoch": 0.62, + "learning_rate": 1.3802321749002085e-07, + "loss": 0.6932, + "step": 16974 + }, + { + "epoch": 0.62, + "learning_rate": 1.380195554253488e-07, + "loss": 0.7, + "step": 16975 + }, + { + "epoch": 0.62, + "learning_rate": 1.3801589336067673e-07, + "loss": 0.6874, + "step": 16976 + }, + { + "epoch": 0.62, + "learning_rate": 1.3801223129600469e-07, + "loss": 0.6943, + "step": 16977 + }, + { + "epoch": 0.62, + "learning_rate": 1.380085692313326e-07, + "loss": 0.6958, + "step": 16978 + }, + { + "epoch": 0.62, + "learning_rate": 1.3800490716666057e-07, + "loss": 0.6882, + "step": 16979 + }, + { + "epoch": 0.62, + "learning_rate": 1.380012451019885e-07, + "loss": 0.7061, + "step": 16980 + }, + { + "epoch": 0.62, + "learning_rate": 1.3799758303731645e-07, + "loss": 0.6958, + "step": 16981 + }, + { + "epoch": 0.62, + "learning_rate": 1.3799392097264437e-07, + "loss": 0.6939, + "step": 16982 + }, + { + "epoch": 0.62, + "learning_rate": 1.3799025890797233e-07, + "loss": 0.6882, + "step": 16983 + }, + { + "epoch": 0.62, + "learning_rate": 1.3798659684330025e-07, + "loss": 0.6965, + "step": 16984 + }, + { + "epoch": 0.62, + "learning_rate": 1.3798293477862818e-07, + "loss": 0.7003, + "step": 16985 + }, + { + "epoch": 0.62, + "learning_rate": 1.3797927271395613e-07, + "loss": 0.6853, + "step": 16986 + }, + { + "epoch": 0.62, + "learning_rate": 1.3797561064928406e-07, + "loss": 0.6908, + "step": 16987 + }, + { + "epoch": 0.62, + "learning_rate": 1.37971948584612e-07, + "loss": 0.6965, + "step": 16988 + }, + { + "epoch": 0.62, + "learning_rate": 1.379682865199399e-07, + "loss": 0.7021, + "step": 16989 + }, + { + "epoch": 0.62, + "learning_rate": 1.3796462445526787e-07, + "loss": 0.6975, + "step": 16990 + }, + { + "epoch": 0.62, + "learning_rate": 1.379609623905958e-07, + "loss": 0.6929, + "step": 16991 + }, + { + "epoch": 0.62, + "learning_rate": 1.3795730032592375e-07, + "loss": 0.6918, + "step": 16992 + }, + { + "epoch": 0.62, + "learning_rate": 1.3795363826125167e-07, + "loss": 0.6956, + "step": 16993 + }, + { + "epoch": 0.62, + "learning_rate": 1.3794997619657962e-07, + "loss": 0.6913, + "step": 16994 + }, + { + "epoch": 0.62, + "learning_rate": 1.3794631413190755e-07, + "loss": 0.6875, + "step": 16995 + }, + { + "epoch": 0.62, + "learning_rate": 1.379426520672355e-07, + "loss": 0.6968, + "step": 16996 + }, + { + "epoch": 0.62, + "learning_rate": 1.3793899000256343e-07, + "loss": 0.7006, + "step": 16997 + }, + { + "epoch": 0.62, + "learning_rate": 1.3793532793789138e-07, + "loss": 0.7055, + "step": 16998 + }, + { + "epoch": 0.62, + "learning_rate": 1.379316658732193e-07, + "loss": 0.6974, + "step": 16999 + }, + { + "epoch": 0.62, + "learning_rate": 1.3792800380854726e-07, + "loss": 0.696, + "step": 17000 + }, + { + "epoch": 0.62, + "learning_rate": 1.379243417438752e-07, + "loss": 0.6903, + "step": 17001 + }, + { + "epoch": 0.62, + "learning_rate": 1.3792067967920314e-07, + "loss": 0.6877, + "step": 17002 + }, + { + "epoch": 0.62, + "learning_rate": 1.3791701761453107e-07, + "loss": 0.7031, + "step": 17003 + }, + { + "epoch": 0.62, + "learning_rate": 1.3791335554985902e-07, + "loss": 0.6951, + "step": 17004 + }, + { + "epoch": 0.62, + "learning_rate": 1.3790969348518695e-07, + "loss": 0.7072, + "step": 17005 + }, + { + "epoch": 0.62, + "learning_rate": 1.3790603142051488e-07, + "loss": 0.6971, + "step": 17006 + }, + { + "epoch": 0.62, + "learning_rate": 1.379023693558428e-07, + "loss": 0.6999, + "step": 17007 + }, + { + "epoch": 0.62, + "learning_rate": 1.3789870729117076e-07, + "loss": 0.7018, + "step": 17008 + }, + { + "epoch": 0.62, + "learning_rate": 1.3789504522649868e-07, + "loss": 0.6987, + "step": 17009 + }, + { + "epoch": 0.62, + "learning_rate": 1.3789138316182664e-07, + "loss": 0.6899, + "step": 17010 + }, + { + "epoch": 0.62, + "learning_rate": 1.3788772109715456e-07, + "loss": 0.677, + "step": 17011 + }, + { + "epoch": 0.62, + "learning_rate": 1.378840590324825e-07, + "loss": 0.6991, + "step": 17012 + }, + { + "epoch": 0.62, + "learning_rate": 1.3788039696781044e-07, + "loss": 0.6734, + "step": 17013 + }, + { + "epoch": 0.62, + "learning_rate": 1.3787673490313837e-07, + "loss": 0.6933, + "step": 17014 + }, + { + "epoch": 0.62, + "learning_rate": 1.3787307283846632e-07, + "loss": 0.6879, + "step": 17015 + }, + { + "epoch": 0.62, + "learning_rate": 1.3786941077379425e-07, + "loss": 0.6806, + "step": 17016 + }, + { + "epoch": 0.62, + "learning_rate": 1.378657487091222e-07, + "loss": 0.7026, + "step": 17017 + }, + { + "epoch": 0.62, + "learning_rate": 1.3786208664445013e-07, + "loss": 0.6877, + "step": 17018 + }, + { + "epoch": 0.62, + "learning_rate": 1.3785842457977808e-07, + "loss": 0.7011, + "step": 17019 + }, + { + "epoch": 0.62, + "learning_rate": 1.37854762515106e-07, + "loss": 0.6885, + "step": 17020 + }, + { + "epoch": 0.62, + "learning_rate": 1.3785110045043396e-07, + "loss": 0.6986, + "step": 17021 + }, + { + "epoch": 0.62, + "learning_rate": 1.378474383857619e-07, + "loss": 0.6975, + "step": 17022 + }, + { + "epoch": 0.62, + "learning_rate": 1.3784377632108984e-07, + "loss": 0.6847, + "step": 17023 + }, + { + "epoch": 0.62, + "learning_rate": 1.3784011425641777e-07, + "loss": 0.6989, + "step": 17024 + }, + { + "epoch": 0.62, + "learning_rate": 1.378364521917457e-07, + "loss": 0.6877, + "step": 17025 + }, + { + "epoch": 0.62, + "learning_rate": 1.3783279012707362e-07, + "loss": 0.6941, + "step": 17026 + }, + { + "epoch": 0.62, + "learning_rate": 1.3782912806240158e-07, + "loss": 0.6971, + "step": 17027 + }, + { + "epoch": 0.62, + "learning_rate": 1.378254659977295e-07, + "loss": 0.696, + "step": 17028 + }, + { + "epoch": 0.62, + "learning_rate": 1.3782180393305746e-07, + "loss": 0.6894, + "step": 17029 + }, + { + "epoch": 0.62, + "learning_rate": 1.3781814186838538e-07, + "loss": 0.6881, + "step": 17030 + }, + { + "epoch": 0.62, + "learning_rate": 1.3781447980371334e-07, + "loss": 0.7001, + "step": 17031 + }, + { + "epoch": 0.62, + "learning_rate": 1.3781081773904126e-07, + "loss": 0.7079, + "step": 17032 + }, + { + "epoch": 0.62, + "learning_rate": 1.378071556743692e-07, + "loss": 0.6934, + "step": 17033 + }, + { + "epoch": 0.62, + "learning_rate": 1.3780349360969714e-07, + "loss": 0.6917, + "step": 17034 + }, + { + "epoch": 0.62, + "learning_rate": 1.3779983154502507e-07, + "loss": 0.6959, + "step": 17035 + }, + { + "epoch": 0.62, + "learning_rate": 1.3779616948035302e-07, + "loss": 0.6981, + "step": 17036 + }, + { + "epoch": 0.62, + "learning_rate": 1.3779250741568095e-07, + "loss": 0.7015, + "step": 17037 + }, + { + "epoch": 0.62, + "learning_rate": 1.377888453510089e-07, + "loss": 0.6869, + "step": 17038 + }, + { + "epoch": 0.62, + "learning_rate": 1.3778518328633683e-07, + "loss": 0.6967, + "step": 17039 + }, + { + "epoch": 0.62, + "learning_rate": 1.3778152122166478e-07, + "loss": 0.6889, + "step": 17040 + }, + { + "epoch": 0.62, + "learning_rate": 1.377778591569927e-07, + "loss": 0.7081, + "step": 17041 + }, + { + "epoch": 0.62, + "learning_rate": 1.3777419709232066e-07, + "loss": 0.6956, + "step": 17042 + }, + { + "epoch": 0.62, + "learning_rate": 1.3777053502764856e-07, + "loss": 0.6846, + "step": 17043 + }, + { + "epoch": 0.62, + "learning_rate": 1.3776687296297652e-07, + "loss": 0.6907, + "step": 17044 + }, + { + "epoch": 0.62, + "learning_rate": 1.3776321089830444e-07, + "loss": 0.6934, + "step": 17045 + }, + { + "epoch": 0.62, + "learning_rate": 1.377595488336324e-07, + "loss": 0.6888, + "step": 17046 + }, + { + "epoch": 0.62, + "learning_rate": 1.3775588676896032e-07, + "loss": 0.6978, + "step": 17047 + }, + { + "epoch": 0.62, + "learning_rate": 1.3775222470428828e-07, + "loss": 0.6845, + "step": 17048 + }, + { + "epoch": 0.62, + "learning_rate": 1.377485626396162e-07, + "loss": 0.674, + "step": 17049 + }, + { + "epoch": 0.62, + "learning_rate": 1.3774490057494415e-07, + "loss": 0.6993, + "step": 17050 + }, + { + "epoch": 0.62, + "learning_rate": 1.3774123851027208e-07, + "loss": 0.6931, + "step": 17051 + }, + { + "epoch": 0.62, + "learning_rate": 1.3773757644560003e-07, + "loss": 0.6914, + "step": 17052 + }, + { + "epoch": 0.62, + "learning_rate": 1.3773391438092796e-07, + "loss": 0.6895, + "step": 17053 + }, + { + "epoch": 0.62, + "learning_rate": 1.3773025231625591e-07, + "loss": 0.6845, + "step": 17054 + }, + { + "epoch": 0.62, + "learning_rate": 1.3772659025158384e-07, + "loss": 0.6963, + "step": 17055 + }, + { + "epoch": 0.62, + "learning_rate": 1.3772292818691177e-07, + "loss": 0.6867, + "step": 17056 + }, + { + "epoch": 0.62, + "learning_rate": 1.3771926612223972e-07, + "loss": 0.7051, + "step": 17057 + }, + { + "epoch": 0.62, + "learning_rate": 1.3771560405756765e-07, + "loss": 0.6944, + "step": 17058 + }, + { + "epoch": 0.62, + "learning_rate": 1.377119419928956e-07, + "loss": 0.6944, + "step": 17059 + }, + { + "epoch": 0.62, + "learning_rate": 1.3770827992822353e-07, + "loss": 0.6849, + "step": 17060 + }, + { + "epoch": 0.62, + "learning_rate": 1.3770461786355145e-07, + "loss": 0.7112, + "step": 17061 + }, + { + "epoch": 0.62, + "learning_rate": 1.3770095579887938e-07, + "loss": 0.7, + "step": 17062 + }, + { + "epoch": 0.62, + "learning_rate": 1.3769729373420733e-07, + "loss": 0.6889, + "step": 17063 + }, + { + "epoch": 0.62, + "learning_rate": 1.3769363166953526e-07, + "loss": 0.6939, + "step": 17064 + }, + { + "epoch": 0.62, + "learning_rate": 1.3768996960486321e-07, + "loss": 0.6861, + "step": 17065 + }, + { + "epoch": 0.62, + "learning_rate": 1.3768630754019114e-07, + "loss": 0.6992, + "step": 17066 + }, + { + "epoch": 0.62, + "learning_rate": 1.376826454755191e-07, + "loss": 0.6853, + "step": 17067 + }, + { + "epoch": 0.62, + "learning_rate": 1.3767898341084702e-07, + "loss": 0.703, + "step": 17068 + }, + { + "epoch": 0.62, + "learning_rate": 1.3767532134617497e-07, + "loss": 0.6964, + "step": 17069 + }, + { + "epoch": 0.62, + "learning_rate": 1.376716592815029e-07, + "loss": 0.6962, + "step": 17070 + }, + { + "epoch": 0.62, + "learning_rate": 1.3766799721683085e-07, + "loss": 0.6895, + "step": 17071 + }, + { + "epoch": 0.62, + "learning_rate": 1.3766433515215878e-07, + "loss": 0.711, + "step": 17072 + }, + { + "epoch": 0.62, + "learning_rate": 1.3766067308748673e-07, + "loss": 0.6914, + "step": 17073 + }, + { + "epoch": 0.62, + "learning_rate": 1.3765701102281466e-07, + "loss": 0.6632, + "step": 17074 + }, + { + "epoch": 0.62, + "learning_rate": 1.3765334895814261e-07, + "loss": 0.6936, + "step": 17075 + }, + { + "epoch": 0.62, + "learning_rate": 1.3764968689347054e-07, + "loss": 0.6952, + "step": 17076 + }, + { + "epoch": 0.62, + "learning_rate": 1.3764602482879847e-07, + "loss": 0.7023, + "step": 17077 + }, + { + "epoch": 0.62, + "learning_rate": 1.3764236276412642e-07, + "loss": 0.7014, + "step": 17078 + }, + { + "epoch": 0.62, + "learning_rate": 1.3763870069945435e-07, + "loss": 0.6933, + "step": 17079 + }, + { + "epoch": 0.62, + "learning_rate": 1.3763503863478227e-07, + "loss": 0.7033, + "step": 17080 + }, + { + "epoch": 0.62, + "learning_rate": 1.3763137657011023e-07, + "loss": 0.7046, + "step": 17081 + }, + { + "epoch": 0.62, + "learning_rate": 1.3762771450543815e-07, + "loss": 0.6833, + "step": 17082 + }, + { + "epoch": 0.63, + "learning_rate": 1.3762405244076608e-07, + "loss": 0.6968, + "step": 17083 + }, + { + "epoch": 0.63, + "learning_rate": 1.3762039037609403e-07, + "loss": 0.6946, + "step": 17084 + }, + { + "epoch": 0.63, + "learning_rate": 1.3761672831142196e-07, + "loss": 0.7012, + "step": 17085 + }, + { + "epoch": 0.63, + "learning_rate": 1.376130662467499e-07, + "loss": 0.6814, + "step": 17086 + }, + { + "epoch": 0.63, + "learning_rate": 1.3760940418207784e-07, + "loss": 0.6874, + "step": 17087 + }, + { + "epoch": 0.63, + "learning_rate": 1.376057421174058e-07, + "loss": 0.6872, + "step": 17088 + }, + { + "epoch": 0.63, + "learning_rate": 1.3760208005273372e-07, + "loss": 0.6907, + "step": 17089 + }, + { + "epoch": 0.63, + "learning_rate": 1.3759841798806167e-07, + "loss": 0.6903, + "step": 17090 + }, + { + "epoch": 0.63, + "learning_rate": 1.375947559233896e-07, + "loss": 0.6995, + "step": 17091 + }, + { + "epoch": 0.63, + "learning_rate": 1.3759109385871755e-07, + "loss": 0.6883, + "step": 17092 + }, + { + "epoch": 0.63, + "learning_rate": 1.3758743179404548e-07, + "loss": 0.6834, + "step": 17093 + }, + { + "epoch": 0.63, + "learning_rate": 1.3758376972937343e-07, + "loss": 0.6988, + "step": 17094 + }, + { + "epoch": 0.63, + "learning_rate": 1.3758010766470136e-07, + "loss": 0.6995, + "step": 17095 + }, + { + "epoch": 0.63, + "learning_rate": 1.375764456000293e-07, + "loss": 0.6869, + "step": 17096 + }, + { + "epoch": 0.63, + "learning_rate": 1.375727835353572e-07, + "loss": 0.6864, + "step": 17097 + }, + { + "epoch": 0.63, + "learning_rate": 1.3756912147068517e-07, + "loss": 0.7018, + "step": 17098 + }, + { + "epoch": 0.63, + "learning_rate": 1.375654594060131e-07, + "loss": 0.6888, + "step": 17099 + }, + { + "epoch": 0.63, + "learning_rate": 1.3756179734134105e-07, + "loss": 0.6955, + "step": 17100 + }, + { + "epoch": 0.63, + "learning_rate": 1.3755813527666897e-07, + "loss": 0.6984, + "step": 17101 + }, + { + "epoch": 0.63, + "learning_rate": 1.3755447321199693e-07, + "loss": 0.6853, + "step": 17102 + }, + { + "epoch": 0.63, + "learning_rate": 1.3755081114732485e-07, + "loss": 0.6988, + "step": 17103 + }, + { + "epoch": 0.63, + "learning_rate": 1.3754714908265278e-07, + "loss": 0.6974, + "step": 17104 + }, + { + "epoch": 0.63, + "learning_rate": 1.3754348701798073e-07, + "loss": 0.6883, + "step": 17105 + }, + { + "epoch": 0.63, + "learning_rate": 1.3753982495330866e-07, + "loss": 0.6942, + "step": 17106 + }, + { + "epoch": 0.63, + "learning_rate": 1.375361628886366e-07, + "loss": 0.7117, + "step": 17107 + }, + { + "epoch": 0.63, + "learning_rate": 1.3753250082396454e-07, + "loss": 0.7004, + "step": 17108 + }, + { + "epoch": 0.63, + "learning_rate": 1.375288387592925e-07, + "loss": 0.7003, + "step": 17109 + }, + { + "epoch": 0.63, + "learning_rate": 1.3752517669462042e-07, + "loss": 0.6898, + "step": 17110 + }, + { + "epoch": 0.63, + "learning_rate": 1.3752151462994837e-07, + "loss": 0.6804, + "step": 17111 + }, + { + "epoch": 0.63, + "learning_rate": 1.375178525652763e-07, + "loss": 0.6888, + "step": 17112 + }, + { + "epoch": 0.63, + "learning_rate": 1.3751419050060425e-07, + "loss": 0.6843, + "step": 17113 + }, + { + "epoch": 0.63, + "learning_rate": 1.3751052843593218e-07, + "loss": 0.6818, + "step": 17114 + }, + { + "epoch": 0.63, + "learning_rate": 1.375068663712601e-07, + "loss": 0.6776, + "step": 17115 + }, + { + "epoch": 0.63, + "learning_rate": 1.3750320430658803e-07, + "loss": 0.6836, + "step": 17116 + }, + { + "epoch": 0.63, + "learning_rate": 1.3749954224191598e-07, + "loss": 0.6771, + "step": 17117 + }, + { + "epoch": 0.63, + "learning_rate": 1.374958801772439e-07, + "loss": 0.6865, + "step": 17118 + }, + { + "epoch": 0.63, + "learning_rate": 1.3749221811257186e-07, + "loss": 0.6906, + "step": 17119 + }, + { + "epoch": 0.63, + "learning_rate": 1.374885560478998e-07, + "loss": 0.6948, + "step": 17120 + }, + { + "epoch": 0.63, + "learning_rate": 1.3748489398322774e-07, + "loss": 0.6725, + "step": 17121 + }, + { + "epoch": 0.63, + "learning_rate": 1.3748123191855567e-07, + "loss": 0.6974, + "step": 17122 + }, + { + "epoch": 0.63, + "learning_rate": 1.3747756985388362e-07, + "loss": 0.6989, + "step": 17123 + }, + { + "epoch": 0.63, + "learning_rate": 1.3747390778921155e-07, + "loss": 0.694, + "step": 17124 + }, + { + "epoch": 0.63, + "learning_rate": 1.374702457245395e-07, + "loss": 0.6804, + "step": 17125 + }, + { + "epoch": 0.63, + "learning_rate": 1.3746658365986743e-07, + "loss": 0.6861, + "step": 17126 + }, + { + "epoch": 0.63, + "learning_rate": 1.3746292159519536e-07, + "loss": 0.6997, + "step": 17127 + }, + { + "epoch": 0.63, + "learning_rate": 1.374592595305233e-07, + "loss": 0.7112, + "step": 17128 + }, + { + "epoch": 0.63, + "learning_rate": 1.3745559746585124e-07, + "loss": 0.6863, + "step": 17129 + }, + { + "epoch": 0.63, + "learning_rate": 1.374519354011792e-07, + "loss": 0.6932, + "step": 17130 + }, + { + "epoch": 0.63, + "learning_rate": 1.3744827333650712e-07, + "loss": 0.7033, + "step": 17131 + }, + { + "epoch": 0.63, + "learning_rate": 1.3744461127183507e-07, + "loss": 0.7138, + "step": 17132 + }, + { + "epoch": 0.63, + "learning_rate": 1.3744094920716297e-07, + "loss": 0.695, + "step": 17133 + }, + { + "epoch": 0.63, + "learning_rate": 1.3743728714249092e-07, + "loss": 0.6953, + "step": 17134 + }, + { + "epoch": 0.63, + "learning_rate": 1.3743362507781885e-07, + "loss": 0.6995, + "step": 17135 + }, + { + "epoch": 0.63, + "learning_rate": 1.374299630131468e-07, + "loss": 0.7025, + "step": 17136 + }, + { + "epoch": 0.63, + "learning_rate": 1.3742630094847473e-07, + "loss": 0.6846, + "step": 17137 + }, + { + "epoch": 0.63, + "learning_rate": 1.3742263888380268e-07, + "loss": 0.6951, + "step": 17138 + }, + { + "epoch": 0.63, + "learning_rate": 1.374189768191306e-07, + "loss": 0.6968, + "step": 17139 + }, + { + "epoch": 0.63, + "learning_rate": 1.3741531475445856e-07, + "loss": 0.6962, + "step": 17140 + }, + { + "epoch": 0.63, + "learning_rate": 1.374116526897865e-07, + "loss": 0.6937, + "step": 17141 + }, + { + "epoch": 0.63, + "learning_rate": 1.3740799062511444e-07, + "loss": 0.6905, + "step": 17142 + }, + { + "epoch": 0.63, + "learning_rate": 1.3740432856044237e-07, + "loss": 0.6921, + "step": 17143 + }, + { + "epoch": 0.63, + "learning_rate": 1.3740066649577032e-07, + "loss": 0.6917, + "step": 17144 + }, + { + "epoch": 0.63, + "learning_rate": 1.3739700443109825e-07, + "loss": 0.7106, + "step": 17145 + }, + { + "epoch": 0.63, + "learning_rate": 1.373933423664262e-07, + "loss": 0.6916, + "step": 17146 + }, + { + "epoch": 0.63, + "learning_rate": 1.3738968030175413e-07, + "loss": 0.6909, + "step": 17147 + }, + { + "epoch": 0.63, + "learning_rate": 1.3738601823708206e-07, + "loss": 0.6997, + "step": 17148 + }, + { + "epoch": 0.63, + "learning_rate": 1.3738235617241e-07, + "loss": 0.6836, + "step": 17149 + }, + { + "epoch": 0.63, + "learning_rate": 1.3737869410773794e-07, + "loss": 0.695, + "step": 17150 + }, + { + "epoch": 0.63, + "learning_rate": 1.3737503204306586e-07, + "loss": 0.6855, + "step": 17151 + }, + { + "epoch": 0.63, + "learning_rate": 1.3737136997839382e-07, + "loss": 0.6927, + "step": 17152 + }, + { + "epoch": 0.63, + "learning_rate": 1.3736770791372174e-07, + "loss": 0.6999, + "step": 17153 + }, + { + "epoch": 0.63, + "learning_rate": 1.3736404584904967e-07, + "loss": 0.7042, + "step": 17154 + }, + { + "epoch": 0.63, + "learning_rate": 1.3736038378437762e-07, + "loss": 0.6975, + "step": 17155 + }, + { + "epoch": 0.63, + "learning_rate": 1.3735672171970555e-07, + "loss": 0.7005, + "step": 17156 + }, + { + "epoch": 0.63, + "learning_rate": 1.373530596550335e-07, + "loss": 0.6849, + "step": 17157 + }, + { + "epoch": 0.63, + "learning_rate": 1.3734939759036143e-07, + "loss": 0.6793, + "step": 17158 + }, + { + "epoch": 0.63, + "learning_rate": 1.3734573552568938e-07, + "loss": 0.6919, + "step": 17159 + }, + { + "epoch": 0.63, + "learning_rate": 1.373420734610173e-07, + "loss": 0.6901, + "step": 17160 + }, + { + "epoch": 0.63, + "learning_rate": 1.3733841139634526e-07, + "loss": 0.7051, + "step": 17161 + }, + { + "epoch": 0.63, + "learning_rate": 1.373347493316732e-07, + "loss": 0.7123, + "step": 17162 + }, + { + "epoch": 0.63, + "learning_rate": 1.3733108726700114e-07, + "loss": 0.693, + "step": 17163 + }, + { + "epoch": 0.63, + "learning_rate": 1.3732742520232907e-07, + "loss": 0.684, + "step": 17164 + }, + { + "epoch": 0.63, + "learning_rate": 1.3732376313765702e-07, + "loss": 0.7052, + "step": 17165 + }, + { + "epoch": 0.63, + "learning_rate": 1.3732010107298495e-07, + "loss": 0.7026, + "step": 17166 + }, + { + "epoch": 0.63, + "learning_rate": 1.373164390083129e-07, + "loss": 0.6938, + "step": 17167 + }, + { + "epoch": 0.63, + "learning_rate": 1.3731277694364083e-07, + "loss": 0.7014, + "step": 17168 + }, + { + "epoch": 0.63, + "learning_rate": 1.3730911487896876e-07, + "loss": 0.6895, + "step": 17169 + }, + { + "epoch": 0.63, + "learning_rate": 1.3730545281429668e-07, + "loss": 0.6954, + "step": 17170 + }, + { + "epoch": 0.63, + "learning_rate": 1.3730179074962463e-07, + "loss": 0.691, + "step": 17171 + }, + { + "epoch": 0.63, + "learning_rate": 1.3729812868495256e-07, + "loss": 0.6779, + "step": 17172 + }, + { + "epoch": 0.63, + "learning_rate": 1.3729446662028051e-07, + "loss": 0.6956, + "step": 17173 + }, + { + "epoch": 0.63, + "learning_rate": 1.3729080455560844e-07, + "loss": 0.6854, + "step": 17174 + }, + { + "epoch": 0.63, + "learning_rate": 1.3728714249093637e-07, + "loss": 0.7009, + "step": 17175 + }, + { + "epoch": 0.63, + "learning_rate": 1.3728348042626432e-07, + "loss": 0.6965, + "step": 17176 + }, + { + "epoch": 0.63, + "learning_rate": 1.3727981836159225e-07, + "loss": 0.6876, + "step": 17177 + }, + { + "epoch": 0.63, + "learning_rate": 1.372761562969202e-07, + "loss": 0.6842, + "step": 17178 + }, + { + "epoch": 0.63, + "learning_rate": 1.3727249423224813e-07, + "loss": 0.6907, + "step": 17179 + }, + { + "epoch": 0.63, + "learning_rate": 1.3726883216757608e-07, + "loss": 0.6893, + "step": 17180 + }, + { + "epoch": 0.63, + "learning_rate": 1.37265170102904e-07, + "loss": 0.7026, + "step": 17181 + }, + { + "epoch": 0.63, + "learning_rate": 1.3726150803823196e-07, + "loss": 0.6892, + "step": 17182 + }, + { + "epoch": 0.63, + "learning_rate": 1.372578459735599e-07, + "loss": 0.6878, + "step": 17183 + }, + { + "epoch": 0.63, + "learning_rate": 1.3725418390888784e-07, + "loss": 0.6875, + "step": 17184 + }, + { + "epoch": 0.63, + "learning_rate": 1.3725052184421577e-07, + "loss": 0.6979, + "step": 17185 + }, + { + "epoch": 0.63, + "learning_rate": 1.3724685977954372e-07, + "loss": 0.6972, + "step": 17186 + }, + { + "epoch": 0.63, + "learning_rate": 1.3724319771487162e-07, + "loss": 0.6946, + "step": 17187 + }, + { + "epoch": 0.63, + "learning_rate": 1.3723953565019957e-07, + "loss": 0.6996, + "step": 17188 + }, + { + "epoch": 0.63, + "learning_rate": 1.372358735855275e-07, + "loss": 0.6914, + "step": 17189 + }, + { + "epoch": 0.63, + "learning_rate": 1.3723221152085545e-07, + "loss": 0.6943, + "step": 17190 + }, + { + "epoch": 0.63, + "learning_rate": 1.3722854945618338e-07, + "loss": 0.6826, + "step": 17191 + }, + { + "epoch": 0.63, + "learning_rate": 1.3722488739151133e-07, + "loss": 0.6895, + "step": 17192 + }, + { + "epoch": 0.63, + "learning_rate": 1.3722122532683926e-07, + "loss": 0.6884, + "step": 17193 + }, + { + "epoch": 0.63, + "learning_rate": 1.3721756326216721e-07, + "loss": 0.6935, + "step": 17194 + }, + { + "epoch": 0.63, + "learning_rate": 1.3721390119749514e-07, + "loss": 0.7047, + "step": 17195 + }, + { + "epoch": 0.63, + "learning_rate": 1.372102391328231e-07, + "loss": 0.7032, + "step": 17196 + }, + { + "epoch": 0.63, + "learning_rate": 1.3720657706815102e-07, + "loss": 0.6802, + "step": 17197 + }, + { + "epoch": 0.63, + "learning_rate": 1.3720291500347895e-07, + "loss": 0.6951, + "step": 17198 + }, + { + "epoch": 0.63, + "learning_rate": 1.371992529388069e-07, + "loss": 0.6923, + "step": 17199 + }, + { + "epoch": 0.63, + "learning_rate": 1.3719559087413483e-07, + "loss": 0.6882, + "step": 17200 + }, + { + "epoch": 0.63, + "learning_rate": 1.3719192880946278e-07, + "loss": 0.6963, + "step": 17201 + }, + { + "epoch": 0.63, + "learning_rate": 1.371882667447907e-07, + "loss": 0.706, + "step": 17202 + }, + { + "epoch": 0.63, + "learning_rate": 1.3718460468011866e-07, + "loss": 0.6877, + "step": 17203 + }, + { + "epoch": 0.63, + "learning_rate": 1.3718094261544659e-07, + "loss": 0.6992, + "step": 17204 + }, + { + "epoch": 0.63, + "learning_rate": 1.371772805507745e-07, + "loss": 0.6931, + "step": 17205 + }, + { + "epoch": 0.63, + "learning_rate": 1.3717361848610244e-07, + "loss": 0.6865, + "step": 17206 + }, + { + "epoch": 0.63, + "learning_rate": 1.371699564214304e-07, + "loss": 0.7108, + "step": 17207 + }, + { + "epoch": 0.63, + "learning_rate": 1.3716629435675832e-07, + "loss": 0.6838, + "step": 17208 + }, + { + "epoch": 0.63, + "learning_rate": 1.3716263229208627e-07, + "loss": 0.69, + "step": 17209 + }, + { + "epoch": 0.63, + "learning_rate": 1.371589702274142e-07, + "loss": 0.7088, + "step": 17210 + }, + { + "epoch": 0.63, + "learning_rate": 1.3715530816274215e-07, + "loss": 0.706, + "step": 17211 + }, + { + "epoch": 0.63, + "learning_rate": 1.3715164609807008e-07, + "loss": 0.6833, + "step": 17212 + }, + { + "epoch": 0.63, + "learning_rate": 1.3714798403339803e-07, + "loss": 0.6954, + "step": 17213 + }, + { + "epoch": 0.63, + "learning_rate": 1.3714432196872596e-07, + "loss": 0.6923, + "step": 17214 + }, + { + "epoch": 0.63, + "learning_rate": 1.371406599040539e-07, + "loss": 0.6945, + "step": 17215 + }, + { + "epoch": 0.63, + "learning_rate": 1.3713699783938184e-07, + "loss": 0.7027, + "step": 17216 + }, + { + "epoch": 0.63, + "learning_rate": 1.371333357747098e-07, + "loss": 0.7029, + "step": 17217 + }, + { + "epoch": 0.63, + "learning_rate": 1.3712967371003772e-07, + "loss": 0.6886, + "step": 17218 + }, + { + "epoch": 0.63, + "learning_rate": 1.3712601164536565e-07, + "loss": 0.6939, + "step": 17219 + }, + { + "epoch": 0.63, + "learning_rate": 1.371223495806936e-07, + "loss": 0.6979, + "step": 17220 + }, + { + "epoch": 0.63, + "learning_rate": 1.3711868751602153e-07, + "loss": 0.693, + "step": 17221 + }, + { + "epoch": 0.63, + "learning_rate": 1.3711502545134948e-07, + "loss": 0.7083, + "step": 17222 + }, + { + "epoch": 0.63, + "learning_rate": 1.371113633866774e-07, + "loss": 0.7009, + "step": 17223 + }, + { + "epoch": 0.63, + "learning_rate": 1.3710770132200533e-07, + "loss": 0.6964, + "step": 17224 + }, + { + "epoch": 0.63, + "learning_rate": 1.3710403925733326e-07, + "loss": 0.6811, + "step": 17225 + }, + { + "epoch": 0.63, + "learning_rate": 1.371003771926612e-07, + "loss": 0.6978, + "step": 17226 + }, + { + "epoch": 0.63, + "learning_rate": 1.3709671512798914e-07, + "loss": 0.6928, + "step": 17227 + }, + { + "epoch": 0.63, + "learning_rate": 1.370930530633171e-07, + "loss": 0.6888, + "step": 17228 + }, + { + "epoch": 0.63, + "learning_rate": 1.3708939099864502e-07, + "loss": 0.692, + "step": 17229 + }, + { + "epoch": 0.63, + "learning_rate": 1.3708572893397297e-07, + "loss": 0.6953, + "step": 17230 + }, + { + "epoch": 0.63, + "learning_rate": 1.370820668693009e-07, + "loss": 0.6932, + "step": 17231 + }, + { + "epoch": 0.63, + "learning_rate": 1.3707840480462885e-07, + "loss": 0.7019, + "step": 17232 + }, + { + "epoch": 0.63, + "learning_rate": 1.3707474273995678e-07, + "loss": 0.7026, + "step": 17233 + }, + { + "epoch": 0.63, + "learning_rate": 1.3707108067528473e-07, + "loss": 0.682, + "step": 17234 + }, + { + "epoch": 0.63, + "learning_rate": 1.3706741861061266e-07, + "loss": 0.6948, + "step": 17235 + }, + { + "epoch": 0.63, + "learning_rate": 1.370637565459406e-07, + "loss": 0.6893, + "step": 17236 + }, + { + "epoch": 0.63, + "learning_rate": 1.3706009448126854e-07, + "loss": 0.6926, + "step": 17237 + }, + { + "epoch": 0.63, + "learning_rate": 1.370564324165965e-07, + "loss": 0.698, + "step": 17238 + }, + { + "epoch": 0.63, + "learning_rate": 1.3705277035192442e-07, + "loss": 0.6884, + "step": 17239 + }, + { + "epoch": 0.63, + "learning_rate": 1.3704910828725234e-07, + "loss": 0.6834, + "step": 17240 + }, + { + "epoch": 0.63, + "learning_rate": 1.3704544622258027e-07, + "loss": 0.6865, + "step": 17241 + }, + { + "epoch": 0.63, + "learning_rate": 1.3704178415790822e-07, + "loss": 0.7036, + "step": 17242 + }, + { + "epoch": 0.63, + "learning_rate": 1.3703812209323615e-07, + "loss": 0.6912, + "step": 17243 + }, + { + "epoch": 0.63, + "learning_rate": 1.370344600285641e-07, + "loss": 0.6848, + "step": 17244 + }, + { + "epoch": 0.63, + "learning_rate": 1.3703079796389203e-07, + "loss": 0.7071, + "step": 17245 + }, + { + "epoch": 0.63, + "learning_rate": 1.3702713589921996e-07, + "loss": 0.6967, + "step": 17246 + }, + { + "epoch": 0.63, + "learning_rate": 1.370234738345479e-07, + "loss": 0.6955, + "step": 17247 + }, + { + "epoch": 0.63, + "learning_rate": 1.3701981176987584e-07, + "loss": 0.7061, + "step": 17248 + }, + { + "epoch": 0.63, + "learning_rate": 1.370161497052038e-07, + "loss": 0.7171, + "step": 17249 + }, + { + "epoch": 0.63, + "learning_rate": 1.3701248764053172e-07, + "loss": 0.7145, + "step": 17250 + }, + { + "epoch": 0.63, + "learning_rate": 1.3700882557585967e-07, + "loss": 0.706, + "step": 17251 + }, + { + "epoch": 0.63, + "learning_rate": 1.370051635111876e-07, + "loss": 0.6876, + "step": 17252 + }, + { + "epoch": 0.63, + "learning_rate": 1.3700150144651555e-07, + "loss": 0.702, + "step": 17253 + }, + { + "epoch": 0.63, + "learning_rate": 1.3699783938184348e-07, + "loss": 0.6933, + "step": 17254 + }, + { + "epoch": 0.63, + "learning_rate": 1.3699417731717143e-07, + "loss": 0.6961, + "step": 17255 + }, + { + "epoch": 0.63, + "learning_rate": 1.3699051525249936e-07, + "loss": 0.7002, + "step": 17256 + }, + { + "epoch": 0.63, + "learning_rate": 1.369868531878273e-07, + "loss": 0.6829, + "step": 17257 + }, + { + "epoch": 0.63, + "learning_rate": 1.3698319112315524e-07, + "loss": 0.6835, + "step": 17258 + }, + { + "epoch": 0.63, + "learning_rate": 1.3697952905848316e-07, + "loss": 0.7006, + "step": 17259 + }, + { + "epoch": 0.63, + "learning_rate": 1.369758669938111e-07, + "loss": 0.6999, + "step": 17260 + }, + { + "epoch": 0.63, + "learning_rate": 1.3697220492913904e-07, + "loss": 0.6909, + "step": 17261 + }, + { + "epoch": 0.63, + "learning_rate": 1.3696854286446697e-07, + "loss": 0.6815, + "step": 17262 + }, + { + "epoch": 0.63, + "learning_rate": 1.3696488079979492e-07, + "loss": 0.7005, + "step": 17263 + }, + { + "epoch": 0.63, + "learning_rate": 1.3696121873512285e-07, + "loss": 0.7062, + "step": 17264 + }, + { + "epoch": 0.63, + "learning_rate": 1.369575566704508e-07, + "loss": 0.6929, + "step": 17265 + }, + { + "epoch": 0.63, + "learning_rate": 1.3695389460577873e-07, + "loss": 0.6907, + "step": 17266 + }, + { + "epoch": 0.63, + "learning_rate": 1.3695023254110666e-07, + "loss": 0.7089, + "step": 17267 + }, + { + "epoch": 0.63, + "learning_rate": 1.369465704764346e-07, + "loss": 0.6903, + "step": 17268 + }, + { + "epoch": 0.63, + "learning_rate": 1.3694290841176254e-07, + "loss": 0.6854, + "step": 17269 + }, + { + "epoch": 0.63, + "learning_rate": 1.369392463470905e-07, + "loss": 0.7059, + "step": 17270 + }, + { + "epoch": 0.63, + "learning_rate": 1.3693558428241842e-07, + "loss": 0.6931, + "step": 17271 + }, + { + "epoch": 0.63, + "learning_rate": 1.3693192221774637e-07, + "loss": 0.7044, + "step": 17272 + }, + { + "epoch": 0.63, + "learning_rate": 1.369282601530743e-07, + "loss": 0.6951, + "step": 17273 + }, + { + "epoch": 0.63, + "learning_rate": 1.3692459808840225e-07, + "loss": 0.6747, + "step": 17274 + }, + { + "epoch": 0.63, + "learning_rate": 1.3692093602373018e-07, + "loss": 0.6957, + "step": 17275 + }, + { + "epoch": 0.63, + "learning_rate": 1.3691727395905813e-07, + "loss": 0.6885, + "step": 17276 + }, + { + "epoch": 0.63, + "learning_rate": 1.3691361189438603e-07, + "loss": 0.693, + "step": 17277 + }, + { + "epoch": 0.63, + "learning_rate": 1.3690994982971398e-07, + "loss": 0.6948, + "step": 17278 + }, + { + "epoch": 0.63, + "learning_rate": 1.369062877650419e-07, + "loss": 0.6975, + "step": 17279 + }, + { + "epoch": 0.63, + "learning_rate": 1.3690262570036986e-07, + "loss": 0.6935, + "step": 17280 + }, + { + "epoch": 0.63, + "learning_rate": 1.368989636356978e-07, + "loss": 0.7029, + "step": 17281 + }, + { + "epoch": 0.63, + "learning_rate": 1.3689530157102574e-07, + "loss": 0.6903, + "step": 17282 + }, + { + "epoch": 0.63, + "learning_rate": 1.3689163950635367e-07, + "loss": 0.7008, + "step": 17283 + }, + { + "epoch": 0.63, + "learning_rate": 1.3688797744168162e-07, + "loss": 0.6948, + "step": 17284 + }, + { + "epoch": 0.63, + "learning_rate": 1.3688431537700955e-07, + "loss": 0.6958, + "step": 17285 + }, + { + "epoch": 0.63, + "learning_rate": 1.368806533123375e-07, + "loss": 0.7001, + "step": 17286 + }, + { + "epoch": 0.63, + "learning_rate": 1.3687699124766543e-07, + "loss": 0.6713, + "step": 17287 + }, + { + "epoch": 0.63, + "learning_rate": 1.3687332918299338e-07, + "loss": 0.6866, + "step": 17288 + }, + { + "epoch": 0.63, + "learning_rate": 1.368696671183213e-07, + "loss": 0.683, + "step": 17289 + }, + { + "epoch": 0.63, + "learning_rate": 1.3686600505364923e-07, + "loss": 0.685, + "step": 17290 + }, + { + "epoch": 0.63, + "learning_rate": 1.368623429889772e-07, + "loss": 0.6803, + "step": 17291 + }, + { + "epoch": 0.63, + "learning_rate": 1.3685868092430511e-07, + "loss": 0.692, + "step": 17292 + }, + { + "epoch": 0.63, + "learning_rate": 1.3685501885963307e-07, + "loss": 0.6997, + "step": 17293 + }, + { + "epoch": 0.63, + "learning_rate": 1.36851356794961e-07, + "loss": 0.7011, + "step": 17294 + }, + { + "epoch": 0.63, + "learning_rate": 1.3684769473028892e-07, + "loss": 0.6949, + "step": 17295 + }, + { + "epoch": 0.63, + "learning_rate": 1.3684403266561685e-07, + "loss": 0.6895, + "step": 17296 + }, + { + "epoch": 0.63, + "learning_rate": 1.368403706009448e-07, + "loss": 0.6948, + "step": 17297 + }, + { + "epoch": 0.63, + "learning_rate": 1.3683670853627273e-07, + "loss": 0.6878, + "step": 17298 + }, + { + "epoch": 0.63, + "learning_rate": 1.3683304647160068e-07, + "loss": 0.6916, + "step": 17299 + }, + { + "epoch": 0.63, + "learning_rate": 1.368293844069286e-07, + "loss": 0.6844, + "step": 17300 + }, + { + "epoch": 0.63, + "learning_rate": 1.3682572234225656e-07, + "loss": 0.7023, + "step": 17301 + }, + { + "epoch": 0.63, + "learning_rate": 1.368220602775845e-07, + "loss": 0.7054, + "step": 17302 + }, + { + "epoch": 0.63, + "learning_rate": 1.3681839821291244e-07, + "loss": 0.6894, + "step": 17303 + }, + { + "epoch": 0.63, + "learning_rate": 1.3681473614824037e-07, + "loss": 0.6928, + "step": 17304 + }, + { + "epoch": 0.63, + "learning_rate": 1.3681107408356832e-07, + "loss": 0.6948, + "step": 17305 + }, + { + "epoch": 0.63, + "learning_rate": 1.3680741201889625e-07, + "loss": 0.6913, + "step": 17306 + }, + { + "epoch": 0.63, + "learning_rate": 1.368037499542242e-07, + "loss": 0.688, + "step": 17307 + }, + { + "epoch": 0.63, + "learning_rate": 1.3680008788955213e-07, + "loss": 0.7022, + "step": 17308 + }, + { + "epoch": 0.63, + "learning_rate": 1.3679642582488008e-07, + "loss": 0.6949, + "step": 17309 + }, + { + "epoch": 0.63, + "learning_rate": 1.36792763760208e-07, + "loss": 0.6862, + "step": 17310 + }, + { + "epoch": 0.63, + "learning_rate": 1.3678910169553593e-07, + "loss": 0.6997, + "step": 17311 + }, + { + "epoch": 0.63, + "learning_rate": 1.367854396308639e-07, + "loss": 0.6992, + "step": 17312 + }, + { + "epoch": 0.63, + "learning_rate": 1.3678177756619181e-07, + "loss": 0.7007, + "step": 17313 + }, + { + "epoch": 0.63, + "learning_rate": 1.3677811550151974e-07, + "loss": 0.6819, + "step": 17314 + }, + { + "epoch": 0.63, + "learning_rate": 1.367744534368477e-07, + "loss": 0.7023, + "step": 17315 + }, + { + "epoch": 0.63, + "learning_rate": 1.3677079137217562e-07, + "loss": 0.6962, + "step": 17316 + }, + { + "epoch": 0.63, + "learning_rate": 1.3676712930750355e-07, + "loss": 0.7076, + "step": 17317 + }, + { + "epoch": 0.63, + "learning_rate": 1.367634672428315e-07, + "loss": 0.6887, + "step": 17318 + }, + { + "epoch": 0.63, + "learning_rate": 1.3675980517815943e-07, + "loss": 0.6991, + "step": 17319 + }, + { + "epoch": 0.63, + "learning_rate": 1.3675614311348738e-07, + "loss": 0.7004, + "step": 17320 + }, + { + "epoch": 0.63, + "learning_rate": 1.367524810488153e-07, + "loss": 0.7048, + "step": 17321 + }, + { + "epoch": 0.63, + "learning_rate": 1.3674881898414326e-07, + "loss": 0.6897, + "step": 17322 + }, + { + "epoch": 0.63, + "learning_rate": 1.3674515691947119e-07, + "loss": 0.6972, + "step": 17323 + }, + { + "epoch": 0.63, + "learning_rate": 1.3674149485479914e-07, + "loss": 0.6975, + "step": 17324 + }, + { + "epoch": 0.63, + "learning_rate": 1.3673783279012707e-07, + "loss": 0.6996, + "step": 17325 + }, + { + "epoch": 0.63, + "learning_rate": 1.3673417072545502e-07, + "loss": 0.7032, + "step": 17326 + }, + { + "epoch": 0.63, + "learning_rate": 1.3673050866078295e-07, + "loss": 0.6787, + "step": 17327 + }, + { + "epoch": 0.63, + "learning_rate": 1.367268465961109e-07, + "loss": 0.6923, + "step": 17328 + }, + { + "epoch": 0.63, + "learning_rate": 1.3672318453143883e-07, + "loss": 0.6803, + "step": 17329 + }, + { + "epoch": 0.63, + "learning_rate": 1.3671952246676678e-07, + "loss": 0.6962, + "step": 17330 + }, + { + "epoch": 0.63, + "learning_rate": 1.3671586040209468e-07, + "loss": 0.6923, + "step": 17331 + }, + { + "epoch": 0.63, + "learning_rate": 1.3671219833742263e-07, + "loss": 0.704, + "step": 17332 + }, + { + "epoch": 0.63, + "learning_rate": 1.3670853627275056e-07, + "loss": 0.6893, + "step": 17333 + }, + { + "epoch": 0.63, + "learning_rate": 1.367048742080785e-07, + "loss": 0.6981, + "step": 17334 + }, + { + "epoch": 0.63, + "learning_rate": 1.3670121214340644e-07, + "loss": 0.709, + "step": 17335 + }, + { + "epoch": 0.63, + "learning_rate": 1.366975500787344e-07, + "loss": 0.6839, + "step": 17336 + }, + { + "epoch": 0.63, + "learning_rate": 1.3669388801406232e-07, + "loss": 0.6808, + "step": 17337 + }, + { + "epoch": 0.63, + "learning_rate": 1.3669022594939025e-07, + "loss": 0.7098, + "step": 17338 + }, + { + "epoch": 0.63, + "learning_rate": 1.366865638847182e-07, + "loss": 0.6853, + "step": 17339 + }, + { + "epoch": 0.63, + "learning_rate": 1.3668290182004613e-07, + "loss": 0.686, + "step": 17340 + }, + { + "epoch": 0.63, + "learning_rate": 1.3667923975537408e-07, + "loss": 0.7108, + "step": 17341 + }, + { + "epoch": 0.63, + "learning_rate": 1.36675577690702e-07, + "loss": 0.6763, + "step": 17342 + }, + { + "epoch": 0.63, + "learning_rate": 1.3667191562602996e-07, + "loss": 0.7064, + "step": 17343 + }, + { + "epoch": 0.63, + "learning_rate": 1.3666825356135789e-07, + "loss": 0.6933, + "step": 17344 + }, + { + "epoch": 0.63, + "learning_rate": 1.3666459149668584e-07, + "loss": 0.6789, + "step": 17345 + }, + { + "epoch": 0.63, + "learning_rate": 1.3666092943201377e-07, + "loss": 0.6909, + "step": 17346 + }, + { + "epoch": 0.63, + "learning_rate": 1.3665726736734172e-07, + "loss": 0.696, + "step": 17347 + }, + { + "epoch": 0.63, + "learning_rate": 1.3665360530266964e-07, + "loss": 0.6969, + "step": 17348 + }, + { + "epoch": 0.63, + "learning_rate": 1.3664994323799757e-07, + "loss": 0.678, + "step": 17349 + }, + { + "epoch": 0.63, + "learning_rate": 1.366462811733255e-07, + "loss": 0.6865, + "step": 17350 + }, + { + "epoch": 0.63, + "learning_rate": 1.3664261910865345e-07, + "loss": 0.6863, + "step": 17351 + }, + { + "epoch": 0.63, + "learning_rate": 1.3663895704398138e-07, + "loss": 0.6992, + "step": 17352 + }, + { + "epoch": 0.63, + "learning_rate": 1.3663529497930933e-07, + "loss": 0.6851, + "step": 17353 + }, + { + "epoch": 0.63, + "learning_rate": 1.3663163291463726e-07, + "loss": 0.7008, + "step": 17354 + }, + { + "epoch": 0.63, + "learning_rate": 1.366279708499652e-07, + "loss": 0.7011, + "step": 17355 + }, + { + "epoch": 0.64, + "learning_rate": 1.3662430878529314e-07, + "loss": 0.6985, + "step": 17356 + }, + { + "epoch": 0.64, + "learning_rate": 1.366206467206211e-07, + "loss": 0.6951, + "step": 17357 + }, + { + "epoch": 0.64, + "learning_rate": 1.3661698465594902e-07, + "loss": 0.6918, + "step": 17358 + }, + { + "epoch": 0.64, + "learning_rate": 1.3661332259127697e-07, + "loss": 0.7019, + "step": 17359 + }, + { + "epoch": 0.64, + "learning_rate": 1.366096605266049e-07, + "loss": 0.7005, + "step": 17360 + }, + { + "epoch": 0.64, + "learning_rate": 1.3660599846193282e-07, + "loss": 0.6862, + "step": 17361 + }, + { + "epoch": 0.64, + "learning_rate": 1.3660233639726078e-07, + "loss": 0.6983, + "step": 17362 + }, + { + "epoch": 0.64, + "learning_rate": 1.365986743325887e-07, + "loss": 0.7069, + "step": 17363 + }, + { + "epoch": 0.64, + "learning_rate": 1.3659501226791666e-07, + "loss": 0.6981, + "step": 17364 + }, + { + "epoch": 0.64, + "learning_rate": 1.3659135020324458e-07, + "loss": 0.6949, + "step": 17365 + }, + { + "epoch": 0.64, + "learning_rate": 1.365876881385725e-07, + "loss": 0.6987, + "step": 17366 + }, + { + "epoch": 0.64, + "learning_rate": 1.3658402607390044e-07, + "loss": 0.6763, + "step": 17367 + }, + { + "epoch": 0.64, + "learning_rate": 1.365803640092284e-07, + "loss": 0.6915, + "step": 17368 + }, + { + "epoch": 0.64, + "learning_rate": 1.3657670194455632e-07, + "loss": 0.691, + "step": 17369 + }, + { + "epoch": 0.64, + "learning_rate": 1.3657303987988427e-07, + "loss": 0.6844, + "step": 17370 + }, + { + "epoch": 0.64, + "learning_rate": 1.365693778152122e-07, + "loss": 0.7016, + "step": 17371 + }, + { + "epoch": 0.64, + "learning_rate": 1.3656571575054015e-07, + "loss": 0.6954, + "step": 17372 + }, + { + "epoch": 0.64, + "learning_rate": 1.3656205368586808e-07, + "loss": 0.6873, + "step": 17373 + }, + { + "epoch": 0.64, + "learning_rate": 1.3655839162119603e-07, + "loss": 0.6774, + "step": 17374 + }, + { + "epoch": 0.64, + "learning_rate": 1.3655472955652396e-07, + "loss": 0.6899, + "step": 17375 + }, + { + "epoch": 0.64, + "learning_rate": 1.365510674918519e-07, + "loss": 0.6839, + "step": 17376 + }, + { + "epoch": 0.64, + "learning_rate": 1.3654740542717984e-07, + "loss": 0.6811, + "step": 17377 + }, + { + "epoch": 0.64, + "learning_rate": 1.365437433625078e-07, + "loss": 0.6808, + "step": 17378 + }, + { + "epoch": 0.64, + "learning_rate": 1.3654008129783572e-07, + "loss": 0.6898, + "step": 17379 + }, + { + "epoch": 0.64, + "learning_rate": 1.3653641923316367e-07, + "loss": 0.6881, + "step": 17380 + }, + { + "epoch": 0.64, + "learning_rate": 1.365327571684916e-07, + "loss": 0.6899, + "step": 17381 + }, + { + "epoch": 0.64, + "learning_rate": 1.3652909510381952e-07, + "loss": 0.6839, + "step": 17382 + }, + { + "epoch": 0.64, + "learning_rate": 1.3652543303914748e-07, + "loss": 0.682, + "step": 17383 + }, + { + "epoch": 0.64, + "learning_rate": 1.365217709744754e-07, + "loss": 0.6952, + "step": 17384 + }, + { + "epoch": 0.64, + "learning_rate": 1.3651810890980333e-07, + "loss": 0.688, + "step": 17385 + }, + { + "epoch": 0.64, + "learning_rate": 1.3651444684513128e-07, + "loss": 0.6764, + "step": 17386 + }, + { + "epoch": 0.64, + "learning_rate": 1.365107847804592e-07, + "loss": 0.686, + "step": 17387 + }, + { + "epoch": 0.64, + "learning_rate": 1.3650712271578714e-07, + "loss": 0.6902, + "step": 17388 + }, + { + "epoch": 0.64, + "learning_rate": 1.365034606511151e-07, + "loss": 0.6804, + "step": 17389 + }, + { + "epoch": 0.64, + "learning_rate": 1.3649979858644302e-07, + "loss": 0.7067, + "step": 17390 + }, + { + "epoch": 0.64, + "learning_rate": 1.3649613652177097e-07, + "loss": 0.6799, + "step": 17391 + }, + { + "epoch": 0.64, + "learning_rate": 1.364924744570989e-07, + "loss": 0.7019, + "step": 17392 + }, + { + "epoch": 0.64, + "learning_rate": 1.3648881239242685e-07, + "loss": 0.6889, + "step": 17393 + }, + { + "epoch": 0.64, + "learning_rate": 1.3648515032775478e-07, + "loss": 0.6964, + "step": 17394 + }, + { + "epoch": 0.64, + "learning_rate": 1.3648148826308273e-07, + "loss": 0.6875, + "step": 17395 + }, + { + "epoch": 0.64, + "learning_rate": 1.3647782619841066e-07, + "loss": 0.6944, + "step": 17396 + }, + { + "epoch": 0.64, + "learning_rate": 1.364741641337386e-07, + "loss": 0.6828, + "step": 17397 + }, + { + "epoch": 0.64, + "learning_rate": 1.3647050206906654e-07, + "loss": 0.721, + "step": 17398 + }, + { + "epoch": 0.64, + "learning_rate": 1.364668400043945e-07, + "loss": 0.6909, + "step": 17399 + }, + { + "epoch": 0.64, + "learning_rate": 1.3646317793972242e-07, + "loss": 0.6873, + "step": 17400 + }, + { + "epoch": 0.64, + "learning_rate": 1.3645951587505037e-07, + "loss": 0.6921, + "step": 17401 + }, + { + "epoch": 0.64, + "learning_rate": 1.3645585381037827e-07, + "loss": 0.6965, + "step": 17402 + }, + { + "epoch": 0.64, + "learning_rate": 1.3645219174570622e-07, + "loss": 0.6818, + "step": 17403 + }, + { + "epoch": 0.64, + "learning_rate": 1.3644852968103415e-07, + "loss": 0.6884, + "step": 17404 + }, + { + "epoch": 0.64, + "learning_rate": 1.364448676163621e-07, + "loss": 0.6843, + "step": 17405 + }, + { + "epoch": 0.64, + "learning_rate": 1.3644120555169003e-07, + "loss": 0.6877, + "step": 17406 + }, + { + "epoch": 0.64, + "learning_rate": 1.3643754348701798e-07, + "loss": 0.6885, + "step": 17407 + }, + { + "epoch": 0.64, + "learning_rate": 1.364338814223459e-07, + "loss": 0.6984, + "step": 17408 + }, + { + "epoch": 0.64, + "learning_rate": 1.3643021935767383e-07, + "loss": 0.693, + "step": 17409 + }, + { + "epoch": 0.64, + "learning_rate": 1.364265572930018e-07, + "loss": 0.695, + "step": 17410 + }, + { + "epoch": 0.64, + "learning_rate": 1.3642289522832971e-07, + "loss": 0.6919, + "step": 17411 + }, + { + "epoch": 0.64, + "learning_rate": 1.3641923316365767e-07, + "loss": 0.6955, + "step": 17412 + }, + { + "epoch": 0.64, + "learning_rate": 1.364155710989856e-07, + "loss": 0.6846, + "step": 17413 + }, + { + "epoch": 0.64, + "learning_rate": 1.3641190903431355e-07, + "loss": 0.7114, + "step": 17414 + }, + { + "epoch": 0.64, + "learning_rate": 1.3640824696964147e-07, + "loss": 0.6928, + "step": 17415 + }, + { + "epoch": 0.64, + "learning_rate": 1.3640458490496943e-07, + "loss": 0.691, + "step": 17416 + }, + { + "epoch": 0.64, + "learning_rate": 1.3640092284029735e-07, + "loss": 0.6957, + "step": 17417 + }, + { + "epoch": 0.64, + "learning_rate": 1.363972607756253e-07, + "loss": 0.681, + "step": 17418 + }, + { + "epoch": 0.64, + "learning_rate": 1.3639359871095323e-07, + "loss": 0.7004, + "step": 17419 + }, + { + "epoch": 0.64, + "learning_rate": 1.3638993664628116e-07, + "loss": 0.7002, + "step": 17420 + }, + { + "epoch": 0.64, + "learning_rate": 1.363862745816091e-07, + "loss": 0.69, + "step": 17421 + }, + { + "epoch": 0.64, + "learning_rate": 1.3638261251693704e-07, + "loss": 0.687, + "step": 17422 + }, + { + "epoch": 0.64, + "learning_rate": 1.3637895045226497e-07, + "loss": 0.6917, + "step": 17423 + }, + { + "epoch": 0.64, + "learning_rate": 1.3637528838759292e-07, + "loss": 0.6951, + "step": 17424 + }, + { + "epoch": 0.64, + "learning_rate": 1.3637162632292085e-07, + "loss": 0.6804, + "step": 17425 + }, + { + "epoch": 0.64, + "learning_rate": 1.363679642582488e-07, + "loss": 0.6852, + "step": 17426 + }, + { + "epoch": 0.64, + "learning_rate": 1.3636430219357673e-07, + "loss": 0.6908, + "step": 17427 + }, + { + "epoch": 0.64, + "learning_rate": 1.3636064012890468e-07, + "loss": 0.7063, + "step": 17428 + }, + { + "epoch": 0.64, + "learning_rate": 1.363569780642326e-07, + "loss": 0.696, + "step": 17429 + }, + { + "epoch": 0.64, + "learning_rate": 1.3635331599956056e-07, + "loss": 0.7178, + "step": 17430 + }, + { + "epoch": 0.64, + "learning_rate": 1.363496539348885e-07, + "loss": 0.6967, + "step": 17431 + }, + { + "epoch": 0.64, + "learning_rate": 1.3634599187021641e-07, + "loss": 0.6837, + "step": 17432 + }, + { + "epoch": 0.64, + "learning_rate": 1.3634232980554437e-07, + "loss": 0.6938, + "step": 17433 + }, + { + "epoch": 0.64, + "learning_rate": 1.363386677408723e-07, + "loss": 0.69, + "step": 17434 + }, + { + "epoch": 0.64, + "learning_rate": 1.3633500567620025e-07, + "loss": 0.6951, + "step": 17435 + }, + { + "epoch": 0.64, + "learning_rate": 1.3633134361152817e-07, + "loss": 0.6827, + "step": 17436 + }, + { + "epoch": 0.64, + "learning_rate": 1.3632768154685613e-07, + "loss": 0.6829, + "step": 17437 + }, + { + "epoch": 0.64, + "learning_rate": 1.3632401948218403e-07, + "loss": 0.687, + "step": 17438 + }, + { + "epoch": 0.64, + "learning_rate": 1.3632035741751198e-07, + "loss": 0.6998, + "step": 17439 + }, + { + "epoch": 0.64, + "learning_rate": 1.363166953528399e-07, + "loss": 0.6955, + "step": 17440 + }, + { + "epoch": 0.64, + "learning_rate": 1.3631303328816786e-07, + "loss": 0.6859, + "step": 17441 + }, + { + "epoch": 0.64, + "learning_rate": 1.3630937122349579e-07, + "loss": 0.7005, + "step": 17442 + }, + { + "epoch": 0.64, + "learning_rate": 1.3630570915882374e-07, + "loss": 0.695, + "step": 17443 + }, + { + "epoch": 0.64, + "learning_rate": 1.3630204709415167e-07, + "loss": 0.7014, + "step": 17444 + }, + { + "epoch": 0.64, + "learning_rate": 1.3629838502947962e-07, + "loss": 0.7007, + "step": 17445 + }, + { + "epoch": 0.64, + "learning_rate": 1.3629472296480755e-07, + "loss": 0.6958, + "step": 17446 + }, + { + "epoch": 0.64, + "learning_rate": 1.362910609001355e-07, + "loss": 0.7029, + "step": 17447 + }, + { + "epoch": 0.64, + "learning_rate": 1.3628739883546343e-07, + "loss": 0.696, + "step": 17448 + }, + { + "epoch": 0.64, + "learning_rate": 1.3628373677079138e-07, + "loss": 0.7097, + "step": 17449 + }, + { + "epoch": 0.64, + "learning_rate": 1.362800747061193e-07, + "loss": 0.6921, + "step": 17450 + }, + { + "epoch": 0.64, + "learning_rate": 1.3627641264144726e-07, + "loss": 0.6942, + "step": 17451 + }, + { + "epoch": 0.64, + "learning_rate": 1.3627275057677519e-07, + "loss": 0.7041, + "step": 17452 + }, + { + "epoch": 0.64, + "learning_rate": 1.362690885121031e-07, + "loss": 0.6872, + "step": 17453 + }, + { + "epoch": 0.64, + "learning_rate": 1.3626542644743107e-07, + "loss": 0.6993, + "step": 17454 + }, + { + "epoch": 0.64, + "learning_rate": 1.36261764382759e-07, + "loss": 0.6846, + "step": 17455 + }, + { + "epoch": 0.64, + "learning_rate": 1.3625810231808692e-07, + "loss": 0.6911, + "step": 17456 + }, + { + "epoch": 0.64, + "learning_rate": 1.3625444025341487e-07, + "loss": 0.6864, + "step": 17457 + }, + { + "epoch": 0.64, + "learning_rate": 1.362507781887428e-07, + "loss": 0.6813, + "step": 17458 + }, + { + "epoch": 0.64, + "learning_rate": 1.3624711612407073e-07, + "loss": 0.6943, + "step": 17459 + }, + { + "epoch": 0.64, + "learning_rate": 1.3624345405939868e-07, + "loss": 0.6849, + "step": 17460 + }, + { + "epoch": 0.64, + "learning_rate": 1.362397919947266e-07, + "loss": 0.6829, + "step": 17461 + }, + { + "epoch": 0.64, + "learning_rate": 1.3623612993005456e-07, + "loss": 0.6919, + "step": 17462 + }, + { + "epoch": 0.64, + "learning_rate": 1.3623246786538249e-07, + "loss": 0.6977, + "step": 17463 + }, + { + "epoch": 0.64, + "learning_rate": 1.3622880580071044e-07, + "loss": 0.6764, + "step": 17464 + }, + { + "epoch": 0.64, + "learning_rate": 1.3622514373603837e-07, + "loss": 0.6855, + "step": 17465 + }, + { + "epoch": 0.64, + "learning_rate": 1.3622148167136632e-07, + "loss": 0.6901, + "step": 17466 + }, + { + "epoch": 0.64, + "learning_rate": 1.3621781960669424e-07, + "loss": 0.6961, + "step": 17467 + }, + { + "epoch": 0.64, + "learning_rate": 1.362141575420222e-07, + "loss": 0.6765, + "step": 17468 + }, + { + "epoch": 0.64, + "learning_rate": 1.3621049547735012e-07, + "loss": 0.6835, + "step": 17469 + }, + { + "epoch": 0.64, + "learning_rate": 1.3620683341267808e-07, + "loss": 0.7013, + "step": 17470 + }, + { + "epoch": 0.64, + "learning_rate": 1.36203171348006e-07, + "loss": 0.693, + "step": 17471 + }, + { + "epoch": 0.64, + "learning_rate": 1.3619950928333396e-07, + "loss": 0.7008, + "step": 17472 + }, + { + "epoch": 0.64, + "learning_rate": 1.3619584721866188e-07, + "loss": 0.7022, + "step": 17473 + }, + { + "epoch": 0.64, + "learning_rate": 1.361921851539898e-07, + "loss": 0.6934, + "step": 17474 + }, + { + "epoch": 0.64, + "learning_rate": 1.3618852308931774e-07, + "loss": 0.6951, + "step": 17475 + }, + { + "epoch": 0.64, + "learning_rate": 1.361848610246457e-07, + "loss": 0.704, + "step": 17476 + }, + { + "epoch": 0.64, + "learning_rate": 1.3618119895997362e-07, + "loss": 0.6914, + "step": 17477 + }, + { + "epoch": 0.64, + "learning_rate": 1.3617753689530157e-07, + "loss": 0.6875, + "step": 17478 + }, + { + "epoch": 0.64, + "learning_rate": 1.361738748306295e-07, + "loss": 0.681, + "step": 17479 + }, + { + "epoch": 0.64, + "learning_rate": 1.3617021276595742e-07, + "loss": 0.6904, + "step": 17480 + }, + { + "epoch": 0.64, + "learning_rate": 1.3616655070128538e-07, + "loss": 0.694, + "step": 17481 + }, + { + "epoch": 0.64, + "learning_rate": 1.361628886366133e-07, + "loss": 0.6979, + "step": 17482 + }, + { + "epoch": 0.64, + "learning_rate": 1.3615922657194126e-07, + "loss": 0.6953, + "step": 17483 + }, + { + "epoch": 0.64, + "learning_rate": 1.3615556450726918e-07, + "loss": 0.6988, + "step": 17484 + }, + { + "epoch": 0.64, + "learning_rate": 1.3615190244259714e-07, + "loss": 0.7003, + "step": 17485 + }, + { + "epoch": 0.64, + "learning_rate": 1.3614824037792506e-07, + "loss": 0.7014, + "step": 17486 + }, + { + "epoch": 0.64, + "learning_rate": 1.3614457831325302e-07, + "loss": 0.6954, + "step": 17487 + }, + { + "epoch": 0.64, + "learning_rate": 1.3614091624858094e-07, + "loss": 0.6795, + "step": 17488 + }, + { + "epoch": 0.64, + "learning_rate": 1.361372541839089e-07, + "loss": 0.6904, + "step": 17489 + }, + { + "epoch": 0.64, + "learning_rate": 1.3613359211923682e-07, + "loss": 0.6946, + "step": 17490 + }, + { + "epoch": 0.64, + "learning_rate": 1.3612993005456478e-07, + "loss": 0.7054, + "step": 17491 + }, + { + "epoch": 0.64, + "learning_rate": 1.3612626798989268e-07, + "loss": 0.693, + "step": 17492 + }, + { + "epoch": 0.64, + "learning_rate": 1.3612260592522063e-07, + "loss": 0.6976, + "step": 17493 + }, + { + "epoch": 0.64, + "learning_rate": 1.3611894386054856e-07, + "loss": 0.6976, + "step": 17494 + }, + { + "epoch": 0.64, + "learning_rate": 1.361152817958765e-07, + "loss": 0.6927, + "step": 17495 + }, + { + "epoch": 0.64, + "learning_rate": 1.3611161973120444e-07, + "loss": 0.7141, + "step": 17496 + }, + { + "epoch": 0.64, + "learning_rate": 1.361079576665324e-07, + "loss": 0.698, + "step": 17497 + }, + { + "epoch": 0.64, + "learning_rate": 1.3610429560186032e-07, + "loss": 0.7056, + "step": 17498 + }, + { + "epoch": 0.64, + "learning_rate": 1.3610063353718827e-07, + "loss": 0.6943, + "step": 17499 + }, + { + "epoch": 0.64, + "learning_rate": 1.360969714725162e-07, + "loss": 0.6975, + "step": 17500 + }, + { + "epoch": 0.64, + "learning_rate": 1.3609330940784415e-07, + "loss": 0.7035, + "step": 17501 + }, + { + "epoch": 0.64, + "learning_rate": 1.3608964734317208e-07, + "loss": 0.6977, + "step": 17502 + }, + { + "epoch": 0.64, + "learning_rate": 1.360859852785e-07, + "loss": 0.6867, + "step": 17503 + }, + { + "epoch": 0.64, + "learning_rate": 1.3608232321382796e-07, + "loss": 0.7034, + "step": 17504 + }, + { + "epoch": 0.64, + "learning_rate": 1.3607866114915588e-07, + "loss": 0.7007, + "step": 17505 + }, + { + "epoch": 0.64, + "learning_rate": 1.3607499908448384e-07, + "loss": 0.6893, + "step": 17506 + }, + { + "epoch": 0.64, + "learning_rate": 1.3607133701981176e-07, + "loss": 0.6979, + "step": 17507 + }, + { + "epoch": 0.64, + "learning_rate": 1.3606767495513972e-07, + "loss": 0.6936, + "step": 17508 + }, + { + "epoch": 0.64, + "learning_rate": 1.3606401289046764e-07, + "loss": 0.6957, + "step": 17509 + }, + { + "epoch": 0.64, + "learning_rate": 1.3606035082579557e-07, + "loss": 0.6907, + "step": 17510 + }, + { + "epoch": 0.64, + "learning_rate": 1.360566887611235e-07, + "loss": 0.694, + "step": 17511 + }, + { + "epoch": 0.64, + "learning_rate": 1.3605302669645145e-07, + "loss": 0.6933, + "step": 17512 + }, + { + "epoch": 0.64, + "learning_rate": 1.3604936463177938e-07, + "loss": 0.6821, + "step": 17513 + }, + { + "epoch": 0.64, + "learning_rate": 1.3604570256710733e-07, + "loss": 0.6938, + "step": 17514 + }, + { + "epoch": 0.64, + "learning_rate": 1.3604204050243526e-07, + "loss": 0.7027, + "step": 17515 + }, + { + "epoch": 0.64, + "learning_rate": 1.360383784377632e-07, + "loss": 0.683, + "step": 17516 + }, + { + "epoch": 0.64, + "learning_rate": 1.3603471637309114e-07, + "loss": 0.6955, + "step": 17517 + }, + { + "epoch": 0.64, + "learning_rate": 1.360310543084191e-07, + "loss": 0.6999, + "step": 17518 + }, + { + "epoch": 0.64, + "learning_rate": 1.3602739224374702e-07, + "loss": 0.7041, + "step": 17519 + }, + { + "epoch": 0.64, + "learning_rate": 1.3602373017907497e-07, + "loss": 0.6997, + "step": 17520 + }, + { + "epoch": 0.64, + "learning_rate": 1.360200681144029e-07, + "loss": 0.692, + "step": 17521 + }, + { + "epoch": 0.64, + "learning_rate": 1.3601640604973085e-07, + "loss": 0.6912, + "step": 17522 + }, + { + "epoch": 0.64, + "learning_rate": 1.3601274398505878e-07, + "loss": 0.6969, + "step": 17523 + }, + { + "epoch": 0.64, + "learning_rate": 1.360090819203867e-07, + "loss": 0.6962, + "step": 17524 + }, + { + "epoch": 0.64, + "learning_rate": 1.3600541985571465e-07, + "loss": 0.6994, + "step": 17525 + }, + { + "epoch": 0.64, + "learning_rate": 1.3600175779104258e-07, + "loss": 0.6949, + "step": 17526 + }, + { + "epoch": 0.64, + "learning_rate": 1.3599809572637053e-07, + "loss": 0.6993, + "step": 17527 + }, + { + "epoch": 0.64, + "learning_rate": 1.3599443366169846e-07, + "loss": 0.688, + "step": 17528 + }, + { + "epoch": 0.64, + "learning_rate": 1.359907715970264e-07, + "loss": 0.7011, + "step": 17529 + }, + { + "epoch": 0.64, + "learning_rate": 1.3598710953235431e-07, + "loss": 0.6781, + "step": 17530 + }, + { + "epoch": 0.64, + "learning_rate": 1.3598344746768227e-07, + "loss": 0.6992, + "step": 17531 + }, + { + "epoch": 0.64, + "learning_rate": 1.359797854030102e-07, + "loss": 0.6896, + "step": 17532 + }, + { + "epoch": 0.64, + "learning_rate": 1.3597612333833815e-07, + "loss": 0.6947, + "step": 17533 + }, + { + "epoch": 0.64, + "learning_rate": 1.3597246127366607e-07, + "loss": 0.6903, + "step": 17534 + }, + { + "epoch": 0.64, + "learning_rate": 1.3596879920899403e-07, + "loss": 0.6939, + "step": 17535 + }, + { + "epoch": 0.64, + "learning_rate": 1.3596513714432195e-07, + "loss": 0.6874, + "step": 17536 + }, + { + "epoch": 0.64, + "learning_rate": 1.359614750796499e-07, + "loss": 0.6912, + "step": 17537 + }, + { + "epoch": 0.64, + "learning_rate": 1.3595781301497783e-07, + "loss": 0.6843, + "step": 17538 + }, + { + "epoch": 0.64, + "learning_rate": 1.359541509503058e-07, + "loss": 0.7041, + "step": 17539 + }, + { + "epoch": 0.64, + "learning_rate": 1.3595048888563371e-07, + "loss": 0.6943, + "step": 17540 + }, + { + "epoch": 0.64, + "learning_rate": 1.3594682682096167e-07, + "loss": 0.701, + "step": 17541 + }, + { + "epoch": 0.64, + "learning_rate": 1.359431647562896e-07, + "loss": 0.6866, + "step": 17542 + }, + { + "epoch": 0.64, + "learning_rate": 1.3593950269161755e-07, + "loss": 0.6968, + "step": 17543 + }, + { + "epoch": 0.64, + "learning_rate": 1.3593584062694547e-07, + "loss": 0.6902, + "step": 17544 + }, + { + "epoch": 0.64, + "learning_rate": 1.3593217856227343e-07, + "loss": 0.6909, + "step": 17545 + }, + { + "epoch": 0.64, + "learning_rate": 1.3592851649760133e-07, + "loss": 0.703, + "step": 17546 + }, + { + "epoch": 0.64, + "learning_rate": 1.3592485443292928e-07, + "loss": 0.692, + "step": 17547 + }, + { + "epoch": 0.64, + "learning_rate": 1.359211923682572e-07, + "loss": 0.6974, + "step": 17548 + }, + { + "epoch": 0.64, + "learning_rate": 1.3591753030358516e-07, + "loss": 0.697, + "step": 17549 + }, + { + "epoch": 0.64, + "learning_rate": 1.359138682389131e-07, + "loss": 0.6904, + "step": 17550 + }, + { + "epoch": 0.64, + "learning_rate": 1.3591020617424101e-07, + "loss": 0.6969, + "step": 17551 + }, + { + "epoch": 0.64, + "learning_rate": 1.3590654410956897e-07, + "loss": 0.6891, + "step": 17552 + }, + { + "epoch": 0.64, + "learning_rate": 1.359028820448969e-07, + "loss": 0.6907, + "step": 17553 + }, + { + "epoch": 0.64, + "learning_rate": 1.3589921998022485e-07, + "loss": 0.705, + "step": 17554 + }, + { + "epoch": 0.64, + "learning_rate": 1.3589555791555277e-07, + "loss": 0.6871, + "step": 17555 + }, + { + "epoch": 0.64, + "learning_rate": 1.3589189585088073e-07, + "loss": 0.6849, + "step": 17556 + }, + { + "epoch": 0.64, + "learning_rate": 1.3588823378620865e-07, + "loss": 0.6851, + "step": 17557 + }, + { + "epoch": 0.64, + "learning_rate": 1.358845717215366e-07, + "loss": 0.6743, + "step": 17558 + }, + { + "epoch": 0.64, + "learning_rate": 1.3588090965686453e-07, + "loss": 0.6871, + "step": 17559 + }, + { + "epoch": 0.64, + "learning_rate": 1.3587724759219249e-07, + "loss": 0.7005, + "step": 17560 + }, + { + "epoch": 0.64, + "learning_rate": 1.358735855275204e-07, + "loss": 0.6974, + "step": 17561 + }, + { + "epoch": 0.64, + "learning_rate": 1.3586992346284837e-07, + "loss": 0.6993, + "step": 17562 + }, + { + "epoch": 0.64, + "learning_rate": 1.358662613981763e-07, + "loss": 0.6937, + "step": 17563 + }, + { + "epoch": 0.64, + "learning_rate": 1.3586259933350422e-07, + "loss": 0.687, + "step": 17564 + }, + { + "epoch": 0.64, + "learning_rate": 1.3585893726883215e-07, + "loss": 0.6922, + "step": 17565 + }, + { + "epoch": 0.64, + "learning_rate": 1.358552752041601e-07, + "loss": 0.6974, + "step": 17566 + }, + { + "epoch": 0.64, + "learning_rate": 1.3585161313948803e-07, + "loss": 0.6819, + "step": 17567 + }, + { + "epoch": 0.64, + "learning_rate": 1.3584795107481598e-07, + "loss": 0.6888, + "step": 17568 + }, + { + "epoch": 0.64, + "learning_rate": 1.358442890101439e-07, + "loss": 0.6975, + "step": 17569 + }, + { + "epoch": 0.64, + "learning_rate": 1.3584062694547186e-07, + "loss": 0.677, + "step": 17570 + }, + { + "epoch": 0.64, + "learning_rate": 1.3583696488079979e-07, + "loss": 0.7031, + "step": 17571 + }, + { + "epoch": 0.64, + "learning_rate": 1.3583330281612774e-07, + "loss": 0.7031, + "step": 17572 + }, + { + "epoch": 0.64, + "learning_rate": 1.3582964075145567e-07, + "loss": 0.6911, + "step": 17573 + }, + { + "epoch": 0.64, + "learning_rate": 1.358259786867836e-07, + "loss": 0.7106, + "step": 17574 + }, + { + "epoch": 0.64, + "learning_rate": 1.3582231662211155e-07, + "loss": 0.6935, + "step": 17575 + }, + { + "epoch": 0.64, + "learning_rate": 1.3581865455743947e-07, + "loss": 0.6939, + "step": 17576 + }, + { + "epoch": 0.64, + "learning_rate": 1.3581499249276743e-07, + "loss": 0.6989, + "step": 17577 + }, + { + "epoch": 0.64, + "learning_rate": 1.3581133042809535e-07, + "loss": 0.7017, + "step": 17578 + }, + { + "epoch": 0.64, + "learning_rate": 1.358076683634233e-07, + "loss": 0.7051, + "step": 17579 + }, + { + "epoch": 0.64, + "learning_rate": 1.3580400629875123e-07, + "loss": 0.695, + "step": 17580 + }, + { + "epoch": 0.64, + "learning_rate": 1.3580034423407919e-07, + "loss": 0.6848, + "step": 17581 + }, + { + "epoch": 0.64, + "learning_rate": 1.3579668216940709e-07, + "loss": 0.6943, + "step": 17582 + }, + { + "epoch": 0.64, + "learning_rate": 1.3579302010473504e-07, + "loss": 0.6859, + "step": 17583 + }, + { + "epoch": 0.64, + "learning_rate": 1.3578935804006297e-07, + "loss": 0.6842, + "step": 17584 + }, + { + "epoch": 0.64, + "learning_rate": 1.3578569597539092e-07, + "loss": 0.699, + "step": 17585 + }, + { + "epoch": 0.64, + "learning_rate": 1.3578203391071884e-07, + "loss": 0.7085, + "step": 17586 + }, + { + "epoch": 0.64, + "learning_rate": 1.357783718460468e-07, + "loss": 0.6824, + "step": 17587 + }, + { + "epoch": 0.64, + "learning_rate": 1.3577470978137472e-07, + "loss": 0.6887, + "step": 17588 + }, + { + "epoch": 0.64, + "learning_rate": 1.3577104771670268e-07, + "loss": 0.6987, + "step": 17589 + }, + { + "epoch": 0.64, + "learning_rate": 1.357673856520306e-07, + "loss": 0.6901, + "step": 17590 + }, + { + "epoch": 0.64, + "learning_rate": 1.3576372358735856e-07, + "loss": 0.693, + "step": 17591 + }, + { + "epoch": 0.64, + "learning_rate": 1.3576006152268648e-07, + "loss": 0.6857, + "step": 17592 + }, + { + "epoch": 0.64, + "learning_rate": 1.3575639945801444e-07, + "loss": 0.6977, + "step": 17593 + }, + { + "epoch": 0.64, + "learning_rate": 1.3575273739334236e-07, + "loss": 0.6868, + "step": 17594 + }, + { + "epoch": 0.64, + "learning_rate": 1.357490753286703e-07, + "loss": 0.6909, + "step": 17595 + }, + { + "epoch": 0.64, + "learning_rate": 1.3574541326399824e-07, + "loss": 0.6937, + "step": 17596 + }, + { + "epoch": 0.64, + "learning_rate": 1.3574175119932617e-07, + "loss": 0.7008, + "step": 17597 + }, + { + "epoch": 0.64, + "learning_rate": 1.3573808913465412e-07, + "loss": 0.6926, + "step": 17598 + }, + { + "epoch": 0.64, + "learning_rate": 1.3573442706998205e-07, + "loss": 0.7082, + "step": 17599 + }, + { + "epoch": 0.64, + "learning_rate": 1.3573076500530998e-07, + "loss": 0.6996, + "step": 17600 + }, + { + "epoch": 0.64, + "learning_rate": 1.357271029406379e-07, + "loss": 0.6893, + "step": 17601 + }, + { + "epoch": 0.64, + "learning_rate": 1.3572344087596586e-07, + "loss": 0.6947, + "step": 17602 + }, + { + "epoch": 0.64, + "learning_rate": 1.3571977881129378e-07, + "loss": 0.6874, + "step": 17603 + }, + { + "epoch": 0.64, + "learning_rate": 1.3571611674662174e-07, + "loss": 0.6849, + "step": 17604 + }, + { + "epoch": 0.64, + "learning_rate": 1.3571245468194966e-07, + "loss": 0.6911, + "step": 17605 + }, + { + "epoch": 0.64, + "learning_rate": 1.3570879261727762e-07, + "loss": 0.6918, + "step": 17606 + }, + { + "epoch": 0.64, + "learning_rate": 1.3570513055260554e-07, + "loss": 0.7044, + "step": 17607 + }, + { + "epoch": 0.64, + "learning_rate": 1.357014684879335e-07, + "loss": 0.7037, + "step": 17608 + }, + { + "epoch": 0.64, + "learning_rate": 1.3569780642326142e-07, + "loss": 0.7019, + "step": 17609 + }, + { + "epoch": 0.64, + "learning_rate": 1.3569414435858938e-07, + "loss": 0.6728, + "step": 17610 + }, + { + "epoch": 0.64, + "learning_rate": 1.356904822939173e-07, + "loss": 0.6905, + "step": 17611 + }, + { + "epoch": 0.64, + "learning_rate": 1.3568682022924526e-07, + "loss": 0.6993, + "step": 17612 + }, + { + "epoch": 0.64, + "learning_rate": 1.3568315816457318e-07, + "loss": 0.7011, + "step": 17613 + }, + { + "epoch": 0.64, + "learning_rate": 1.3567949609990114e-07, + "loss": 0.6983, + "step": 17614 + }, + { + "epoch": 0.64, + "learning_rate": 1.3567583403522906e-07, + "loss": 0.7006, + "step": 17615 + }, + { + "epoch": 0.64, + "learning_rate": 1.3567217197055702e-07, + "loss": 0.6852, + "step": 17616 + }, + { + "epoch": 0.64, + "learning_rate": 1.3566850990588494e-07, + "loss": 0.6997, + "step": 17617 + }, + { + "epoch": 0.64, + "learning_rate": 1.3566484784121287e-07, + "loss": 0.6903, + "step": 17618 + }, + { + "epoch": 0.64, + "learning_rate": 1.356611857765408e-07, + "loss": 0.6869, + "step": 17619 + }, + { + "epoch": 0.64, + "learning_rate": 1.3565752371186875e-07, + "loss": 0.689, + "step": 17620 + }, + { + "epoch": 0.64, + "learning_rate": 1.3565386164719668e-07, + "loss": 0.6931, + "step": 17621 + }, + { + "epoch": 0.64, + "learning_rate": 1.356501995825246e-07, + "loss": 0.696, + "step": 17622 + }, + { + "epoch": 0.64, + "learning_rate": 1.3564653751785256e-07, + "loss": 0.6909, + "step": 17623 + }, + { + "epoch": 0.64, + "learning_rate": 1.3564287545318048e-07, + "loss": 0.6832, + "step": 17624 + }, + { + "epoch": 0.64, + "learning_rate": 1.3563921338850844e-07, + "loss": 0.6881, + "step": 17625 + }, + { + "epoch": 0.64, + "learning_rate": 1.3563555132383636e-07, + "loss": 0.6991, + "step": 17626 + }, + { + "epoch": 0.64, + "learning_rate": 1.3563188925916432e-07, + "loss": 0.7077, + "step": 17627 + }, + { + "epoch": 0.64, + "learning_rate": 1.3562822719449224e-07, + "loss": 0.6873, + "step": 17628 + }, + { + "epoch": 0.64, + "learning_rate": 1.356245651298202e-07, + "loss": 0.6958, + "step": 17629 + }, + { + "epoch": 0.65, + "learning_rate": 1.3562090306514812e-07, + "loss": 0.6898, + "step": 17630 + }, + { + "epoch": 0.65, + "learning_rate": 1.3561724100047608e-07, + "loss": 0.6849, + "step": 17631 + }, + { + "epoch": 0.65, + "learning_rate": 1.35613578935804e-07, + "loss": 0.7029, + "step": 17632 + }, + { + "epoch": 0.65, + "learning_rate": 1.3560991687113196e-07, + "loss": 0.6834, + "step": 17633 + }, + { + "epoch": 0.65, + "learning_rate": 1.3560625480645988e-07, + "loss": 0.7013, + "step": 17634 + }, + { + "epoch": 0.65, + "learning_rate": 1.3560259274178784e-07, + "loss": 0.6945, + "step": 17635 + }, + { + "epoch": 0.65, + "learning_rate": 1.3559893067711574e-07, + "loss": 0.6965, + "step": 17636 + }, + { + "epoch": 0.65, + "learning_rate": 1.355952686124437e-07, + "loss": 0.6869, + "step": 17637 + }, + { + "epoch": 0.65, + "learning_rate": 1.3559160654777162e-07, + "loss": 0.706, + "step": 17638 + }, + { + "epoch": 0.65, + "learning_rate": 1.3558794448309957e-07, + "loss": 0.6951, + "step": 17639 + }, + { + "epoch": 0.65, + "learning_rate": 1.355842824184275e-07, + "loss": 0.6993, + "step": 17640 + }, + { + "epoch": 0.65, + "learning_rate": 1.3558062035375545e-07, + "loss": 0.7037, + "step": 17641 + }, + { + "epoch": 0.65, + "learning_rate": 1.3557695828908338e-07, + "loss": 0.6932, + "step": 17642 + }, + { + "epoch": 0.65, + "learning_rate": 1.3557329622441133e-07, + "loss": 0.6974, + "step": 17643 + }, + { + "epoch": 0.65, + "learning_rate": 1.3556963415973925e-07, + "loss": 0.6776, + "step": 17644 + }, + { + "epoch": 0.65, + "learning_rate": 1.3556597209506718e-07, + "loss": 0.7066, + "step": 17645 + }, + { + "epoch": 0.65, + "learning_rate": 1.3556231003039513e-07, + "loss": 0.6936, + "step": 17646 + }, + { + "epoch": 0.65, + "learning_rate": 1.3555864796572306e-07, + "loss": 0.6985, + "step": 17647 + }, + { + "epoch": 0.65, + "learning_rate": 1.3555498590105101e-07, + "loss": 0.6719, + "step": 17648 + }, + { + "epoch": 0.65, + "learning_rate": 1.3555132383637894e-07, + "loss": 0.6946, + "step": 17649 + }, + { + "epoch": 0.65, + "learning_rate": 1.355476617717069e-07, + "loss": 0.6831, + "step": 17650 + }, + { + "epoch": 0.65, + "learning_rate": 1.3554399970703482e-07, + "loss": 0.6974, + "step": 17651 + }, + { + "epoch": 0.65, + "learning_rate": 1.3554033764236277e-07, + "loss": 0.7037, + "step": 17652 + }, + { + "epoch": 0.65, + "learning_rate": 1.355366755776907e-07, + "loss": 0.6965, + "step": 17653 + }, + { + "epoch": 0.65, + "learning_rate": 1.3553301351301863e-07, + "loss": 0.6933, + "step": 17654 + }, + { + "epoch": 0.65, + "learning_rate": 1.3552935144834655e-07, + "loss": 0.6821, + "step": 17655 + }, + { + "epoch": 0.65, + "learning_rate": 1.355256893836745e-07, + "loss": 0.6965, + "step": 17656 + }, + { + "epoch": 0.65, + "learning_rate": 1.3552202731900243e-07, + "loss": 0.6889, + "step": 17657 + }, + { + "epoch": 0.65, + "learning_rate": 1.355183652543304e-07, + "loss": 0.7094, + "step": 17658 + }, + { + "epoch": 0.65, + "learning_rate": 1.3551470318965831e-07, + "loss": 0.6991, + "step": 17659 + }, + { + "epoch": 0.65, + "learning_rate": 1.3551104112498627e-07, + "loss": 0.6944, + "step": 17660 + }, + { + "epoch": 0.65, + "learning_rate": 1.355073790603142e-07, + "loss": 0.6939, + "step": 17661 + }, + { + "epoch": 0.65, + "learning_rate": 1.3550371699564215e-07, + "loss": 0.6925, + "step": 17662 + }, + { + "epoch": 0.65, + "learning_rate": 1.3550005493097007e-07, + "loss": 0.6919, + "step": 17663 + }, + { + "epoch": 0.65, + "learning_rate": 1.3549639286629803e-07, + "loss": 0.6936, + "step": 17664 + }, + { + "epoch": 0.65, + "learning_rate": 1.3549273080162595e-07, + "loss": 0.7126, + "step": 17665 + }, + { + "epoch": 0.65, + "learning_rate": 1.3548906873695388e-07, + "loss": 0.7035, + "step": 17666 + }, + { + "epoch": 0.65, + "learning_rate": 1.3548540667228183e-07, + "loss": 0.6888, + "step": 17667 + }, + { + "epoch": 0.65, + "learning_rate": 1.3548174460760976e-07, + "loss": 0.6963, + "step": 17668 + }, + { + "epoch": 0.65, + "learning_rate": 1.3547808254293771e-07, + "loss": 0.7041, + "step": 17669 + }, + { + "epoch": 0.65, + "learning_rate": 1.3547442047826564e-07, + "loss": 0.6758, + "step": 17670 + }, + { + "epoch": 0.65, + "learning_rate": 1.354707584135936e-07, + "loss": 0.6858, + "step": 17671 + }, + { + "epoch": 0.65, + "learning_rate": 1.354670963489215e-07, + "loss": 0.7056, + "step": 17672 + }, + { + "epoch": 0.65, + "learning_rate": 1.3546343428424945e-07, + "loss": 0.7015, + "step": 17673 + }, + { + "epoch": 0.65, + "learning_rate": 1.3545977221957737e-07, + "loss": 0.705, + "step": 17674 + }, + { + "epoch": 0.65, + "learning_rate": 1.3545611015490533e-07, + "loss": 0.6973, + "step": 17675 + }, + { + "epoch": 0.65, + "learning_rate": 1.3545244809023325e-07, + "loss": 0.695, + "step": 17676 + }, + { + "epoch": 0.65, + "learning_rate": 1.354487860255612e-07, + "loss": 0.694, + "step": 17677 + }, + { + "epoch": 0.65, + "learning_rate": 1.3544512396088913e-07, + "loss": 0.7002, + "step": 17678 + }, + { + "epoch": 0.65, + "learning_rate": 1.3544146189621709e-07, + "loss": 0.69, + "step": 17679 + }, + { + "epoch": 0.65, + "learning_rate": 1.35437799831545e-07, + "loss": 0.7044, + "step": 17680 + }, + { + "epoch": 0.65, + "learning_rate": 1.3543413776687297e-07, + "loss": 0.6917, + "step": 17681 + }, + { + "epoch": 0.65, + "learning_rate": 1.354304757022009e-07, + "loss": 0.6969, + "step": 17682 + }, + { + "epoch": 0.65, + "learning_rate": 1.3542681363752885e-07, + "loss": 0.7049, + "step": 17683 + }, + { + "epoch": 0.65, + "learning_rate": 1.3542315157285677e-07, + "loss": 0.7002, + "step": 17684 + }, + { + "epoch": 0.65, + "learning_rate": 1.3541948950818473e-07, + "loss": 0.6928, + "step": 17685 + }, + { + "epoch": 0.65, + "learning_rate": 1.3541582744351265e-07, + "loss": 0.691, + "step": 17686 + }, + { + "epoch": 0.65, + "learning_rate": 1.354121653788406e-07, + "loss": 0.6891, + "step": 17687 + }, + { + "epoch": 0.65, + "learning_rate": 1.3540850331416853e-07, + "loss": 0.678, + "step": 17688 + }, + { + "epoch": 0.65, + "learning_rate": 1.3540484124949646e-07, + "loss": 0.6908, + "step": 17689 + }, + { + "epoch": 0.65, + "learning_rate": 1.3540117918482439e-07, + "loss": 0.6963, + "step": 17690 + }, + { + "epoch": 0.65, + "learning_rate": 1.3539751712015234e-07, + "loss": 0.7002, + "step": 17691 + }, + { + "epoch": 0.65, + "learning_rate": 1.3539385505548027e-07, + "loss": 0.6887, + "step": 17692 + }, + { + "epoch": 0.65, + "learning_rate": 1.353901929908082e-07, + "loss": 0.7013, + "step": 17693 + }, + { + "epoch": 0.65, + "learning_rate": 1.3538653092613615e-07, + "loss": 0.7034, + "step": 17694 + }, + { + "epoch": 0.65, + "learning_rate": 1.3538286886146407e-07, + "loss": 0.6952, + "step": 17695 + }, + { + "epoch": 0.65, + "learning_rate": 1.3537920679679203e-07, + "loss": 0.6951, + "step": 17696 + }, + { + "epoch": 0.65, + "learning_rate": 1.3537554473211995e-07, + "loss": 0.6844, + "step": 17697 + }, + { + "epoch": 0.65, + "learning_rate": 1.353718826674479e-07, + "loss": 0.7099, + "step": 17698 + }, + { + "epoch": 0.65, + "learning_rate": 1.3536822060277583e-07, + "loss": 0.6838, + "step": 17699 + }, + { + "epoch": 0.65, + "learning_rate": 1.3536455853810379e-07, + "loss": 0.6907, + "step": 17700 + }, + { + "epoch": 0.65, + "learning_rate": 1.353608964734317e-07, + "loss": 0.6851, + "step": 17701 + }, + { + "epoch": 0.65, + "learning_rate": 1.3535723440875966e-07, + "loss": 0.703, + "step": 17702 + }, + { + "epoch": 0.65, + "learning_rate": 1.353535723440876e-07, + "loss": 0.6973, + "step": 17703 + }, + { + "epoch": 0.65, + "learning_rate": 1.3534991027941554e-07, + "loss": 0.7025, + "step": 17704 + }, + { + "epoch": 0.65, + "learning_rate": 1.3534624821474347e-07, + "loss": 0.6881, + "step": 17705 + }, + { + "epoch": 0.65, + "learning_rate": 1.3534258615007142e-07, + "loss": 0.6938, + "step": 17706 + }, + { + "epoch": 0.65, + "learning_rate": 1.3533892408539935e-07, + "loss": 0.6912, + "step": 17707 + }, + { + "epoch": 0.65, + "learning_rate": 1.3533526202072728e-07, + "loss": 0.7038, + "step": 17708 + }, + { + "epoch": 0.65, + "learning_rate": 1.353315999560552e-07, + "loss": 0.6942, + "step": 17709 + }, + { + "epoch": 0.65, + "learning_rate": 1.3532793789138316e-07, + "loss": 0.6857, + "step": 17710 + }, + { + "epoch": 0.65, + "learning_rate": 1.3532427582671108e-07, + "loss": 0.6959, + "step": 17711 + }, + { + "epoch": 0.65, + "learning_rate": 1.3532061376203904e-07, + "loss": 0.7011, + "step": 17712 + }, + { + "epoch": 0.65, + "learning_rate": 1.3531695169736696e-07, + "loss": 0.6957, + "step": 17713 + }, + { + "epoch": 0.65, + "learning_rate": 1.3531328963269492e-07, + "loss": 0.6881, + "step": 17714 + }, + { + "epoch": 0.65, + "learning_rate": 1.3530962756802284e-07, + "loss": 0.6875, + "step": 17715 + }, + { + "epoch": 0.65, + "learning_rate": 1.3530596550335077e-07, + "loss": 0.6967, + "step": 17716 + }, + { + "epoch": 0.65, + "learning_rate": 1.3530230343867872e-07, + "loss": 0.7032, + "step": 17717 + }, + { + "epoch": 0.65, + "learning_rate": 1.3529864137400665e-07, + "loss": 0.6956, + "step": 17718 + }, + { + "epoch": 0.65, + "learning_rate": 1.352949793093346e-07, + "loss": 0.685, + "step": 17719 + }, + { + "epoch": 0.65, + "learning_rate": 1.3529131724466253e-07, + "loss": 0.6896, + "step": 17720 + }, + { + "epoch": 0.65, + "learning_rate": 1.3528765517999048e-07, + "loss": 0.6962, + "step": 17721 + }, + { + "epoch": 0.65, + "learning_rate": 1.352839931153184e-07, + "loss": 0.6956, + "step": 17722 + }, + { + "epoch": 0.65, + "learning_rate": 1.3528033105064636e-07, + "loss": 0.6857, + "step": 17723 + }, + { + "epoch": 0.65, + "learning_rate": 1.352766689859743e-07, + "loss": 0.697, + "step": 17724 + }, + { + "epoch": 0.65, + "learning_rate": 1.3527300692130224e-07, + "loss": 0.6893, + "step": 17725 + }, + { + "epoch": 0.65, + "learning_rate": 1.3526934485663014e-07, + "loss": 0.6843, + "step": 17726 + }, + { + "epoch": 0.65, + "learning_rate": 1.352656827919581e-07, + "loss": 0.6856, + "step": 17727 + }, + { + "epoch": 0.65, + "learning_rate": 1.3526202072728602e-07, + "loss": 0.7053, + "step": 17728 + }, + { + "epoch": 0.65, + "learning_rate": 1.3525835866261398e-07, + "loss": 0.7007, + "step": 17729 + }, + { + "epoch": 0.65, + "learning_rate": 1.352546965979419e-07, + "loss": 0.6956, + "step": 17730 + }, + { + "epoch": 0.65, + "learning_rate": 1.3525103453326986e-07, + "loss": 0.6915, + "step": 17731 + }, + { + "epoch": 0.65, + "learning_rate": 1.3524737246859778e-07, + "loss": 0.6861, + "step": 17732 + }, + { + "epoch": 0.65, + "learning_rate": 1.3524371040392574e-07, + "loss": 0.6921, + "step": 17733 + }, + { + "epoch": 0.65, + "learning_rate": 1.3524004833925366e-07, + "loss": 0.6853, + "step": 17734 + }, + { + "epoch": 0.65, + "learning_rate": 1.3523638627458162e-07, + "loss": 0.6981, + "step": 17735 + }, + { + "epoch": 0.65, + "learning_rate": 1.3523272420990954e-07, + "loss": 0.6918, + "step": 17736 + }, + { + "epoch": 0.65, + "learning_rate": 1.3522906214523747e-07, + "loss": 0.689, + "step": 17737 + }, + { + "epoch": 0.65, + "learning_rate": 1.3522540008056542e-07, + "loss": 0.705, + "step": 17738 + }, + { + "epoch": 0.65, + "learning_rate": 1.3522173801589335e-07, + "loss": 0.7019, + "step": 17739 + }, + { + "epoch": 0.65, + "learning_rate": 1.352180759512213e-07, + "loss": 0.6946, + "step": 17740 + }, + { + "epoch": 0.65, + "learning_rate": 1.3521441388654923e-07, + "loss": 0.691, + "step": 17741 + }, + { + "epoch": 0.65, + "learning_rate": 1.3521075182187718e-07, + "loss": 0.6918, + "step": 17742 + }, + { + "epoch": 0.65, + "learning_rate": 1.3520708975720508e-07, + "loss": 0.6988, + "step": 17743 + }, + { + "epoch": 0.65, + "learning_rate": 1.3520342769253304e-07, + "loss": 0.6982, + "step": 17744 + }, + { + "epoch": 0.65, + "learning_rate": 1.3519976562786096e-07, + "loss": 0.7006, + "step": 17745 + }, + { + "epoch": 0.65, + "learning_rate": 1.3519610356318892e-07, + "loss": 0.7023, + "step": 17746 + }, + { + "epoch": 0.65, + "learning_rate": 1.3519244149851684e-07, + "loss": 0.6896, + "step": 17747 + }, + { + "epoch": 0.65, + "learning_rate": 1.351887794338448e-07, + "loss": 0.687, + "step": 17748 + }, + { + "epoch": 0.65, + "learning_rate": 1.3518511736917272e-07, + "loss": 0.6881, + "step": 17749 + }, + { + "epoch": 0.65, + "learning_rate": 1.3518145530450068e-07, + "loss": 0.6944, + "step": 17750 + }, + { + "epoch": 0.65, + "learning_rate": 1.351777932398286e-07, + "loss": 0.6997, + "step": 17751 + }, + { + "epoch": 0.65, + "learning_rate": 1.3517413117515656e-07, + "loss": 0.6946, + "step": 17752 + }, + { + "epoch": 0.65, + "learning_rate": 1.3517046911048448e-07, + "loss": 0.7092, + "step": 17753 + }, + { + "epoch": 0.65, + "learning_rate": 1.3516680704581244e-07, + "loss": 0.7007, + "step": 17754 + }, + { + "epoch": 0.65, + "learning_rate": 1.3516314498114036e-07, + "loss": 0.6818, + "step": 17755 + }, + { + "epoch": 0.65, + "learning_rate": 1.3515948291646832e-07, + "loss": 0.6957, + "step": 17756 + }, + { + "epoch": 0.65, + "learning_rate": 1.3515582085179624e-07, + "loss": 0.6867, + "step": 17757 + }, + { + "epoch": 0.65, + "learning_rate": 1.351521587871242e-07, + "loss": 0.7062, + "step": 17758 + }, + { + "epoch": 0.65, + "learning_rate": 1.3514849672245212e-07, + "loss": 0.6882, + "step": 17759 + }, + { + "epoch": 0.65, + "learning_rate": 1.3514483465778005e-07, + "loss": 0.6859, + "step": 17760 + }, + { + "epoch": 0.65, + "learning_rate": 1.3514117259310798e-07, + "loss": 0.6929, + "step": 17761 + }, + { + "epoch": 0.65, + "learning_rate": 1.3513751052843593e-07, + "loss": 0.6986, + "step": 17762 + }, + { + "epoch": 0.65, + "learning_rate": 1.3513384846376386e-07, + "loss": 0.6799, + "step": 17763 + }, + { + "epoch": 0.65, + "learning_rate": 1.3513018639909178e-07, + "loss": 0.6853, + "step": 17764 + }, + { + "epoch": 0.65, + "learning_rate": 1.3512652433441973e-07, + "loss": 0.6838, + "step": 17765 + }, + { + "epoch": 0.65, + "learning_rate": 1.3512286226974766e-07, + "loss": 0.6939, + "step": 17766 + }, + { + "epoch": 0.65, + "learning_rate": 1.3511920020507561e-07, + "loss": 0.6967, + "step": 17767 + }, + { + "epoch": 0.65, + "learning_rate": 1.3511553814040354e-07, + "loss": 0.6866, + "step": 17768 + }, + { + "epoch": 0.65, + "learning_rate": 1.351118760757315e-07, + "loss": 0.6886, + "step": 17769 + }, + { + "epoch": 0.65, + "learning_rate": 1.3510821401105942e-07, + "loss": 0.6997, + "step": 17770 + }, + { + "epoch": 0.65, + "learning_rate": 1.3510455194638737e-07, + "loss": 0.6969, + "step": 17771 + }, + { + "epoch": 0.65, + "learning_rate": 1.351008898817153e-07, + "loss": 0.6942, + "step": 17772 + }, + { + "epoch": 0.65, + "learning_rate": 1.3509722781704325e-07, + "loss": 0.6891, + "step": 17773 + }, + { + "epoch": 0.65, + "learning_rate": 1.3509356575237118e-07, + "loss": 0.6961, + "step": 17774 + }, + { + "epoch": 0.65, + "learning_rate": 1.3508990368769913e-07, + "loss": 0.6893, + "step": 17775 + }, + { + "epoch": 0.65, + "learning_rate": 1.3508624162302706e-07, + "loss": 0.6981, + "step": 17776 + }, + { + "epoch": 0.65, + "learning_rate": 1.3508257955835501e-07, + "loss": 0.6965, + "step": 17777 + }, + { + "epoch": 0.65, + "learning_rate": 1.3507891749368294e-07, + "loss": 0.6991, + "step": 17778 + }, + { + "epoch": 0.65, + "learning_rate": 1.3507525542901087e-07, + "loss": 0.6895, + "step": 17779 + }, + { + "epoch": 0.65, + "learning_rate": 1.350715933643388e-07, + "loss": 0.69, + "step": 17780 + }, + { + "epoch": 0.65, + "learning_rate": 1.3506793129966675e-07, + "loss": 0.6963, + "step": 17781 + }, + { + "epoch": 0.65, + "learning_rate": 1.3506426923499467e-07, + "loss": 0.6892, + "step": 17782 + }, + { + "epoch": 0.65, + "learning_rate": 1.3506060717032263e-07, + "loss": 0.6863, + "step": 17783 + }, + { + "epoch": 0.65, + "learning_rate": 1.3505694510565055e-07, + "loss": 0.6989, + "step": 17784 + }, + { + "epoch": 0.65, + "learning_rate": 1.350532830409785e-07, + "loss": 0.6911, + "step": 17785 + }, + { + "epoch": 0.65, + "learning_rate": 1.3504962097630643e-07, + "loss": 0.6874, + "step": 17786 + }, + { + "epoch": 0.65, + "learning_rate": 1.3504595891163436e-07, + "loss": 0.6814, + "step": 17787 + }, + { + "epoch": 0.65, + "learning_rate": 1.3504229684696231e-07, + "loss": 0.6866, + "step": 17788 + }, + { + "epoch": 0.65, + "learning_rate": 1.3503863478229024e-07, + "loss": 0.6864, + "step": 17789 + }, + { + "epoch": 0.65, + "learning_rate": 1.350349727176182e-07, + "loss": 0.6818, + "step": 17790 + }, + { + "epoch": 0.65, + "learning_rate": 1.3503131065294612e-07, + "loss": 0.6956, + "step": 17791 + }, + { + "epoch": 0.65, + "learning_rate": 1.3502764858827407e-07, + "loss": 0.6953, + "step": 17792 + }, + { + "epoch": 0.65, + "learning_rate": 1.35023986523602e-07, + "loss": 0.7032, + "step": 17793 + }, + { + "epoch": 0.65, + "learning_rate": 1.3502032445892995e-07, + "loss": 0.6949, + "step": 17794 + }, + { + "epoch": 0.65, + "learning_rate": 1.3501666239425788e-07, + "loss": 0.6935, + "step": 17795 + }, + { + "epoch": 0.65, + "learning_rate": 1.3501300032958583e-07, + "loss": 0.6956, + "step": 17796 + }, + { + "epoch": 0.65, + "learning_rate": 1.3500933826491373e-07, + "loss": 0.7047, + "step": 17797 + }, + { + "epoch": 0.65, + "learning_rate": 1.3500567620024169e-07, + "loss": 0.6822, + "step": 17798 + }, + { + "epoch": 0.65, + "learning_rate": 1.350020141355696e-07, + "loss": 0.6833, + "step": 17799 + }, + { + "epoch": 0.65, + "learning_rate": 1.3499835207089757e-07, + "loss": 0.7001, + "step": 17800 + }, + { + "epoch": 0.65, + "learning_rate": 1.349946900062255e-07, + "loss": 0.6859, + "step": 17801 + }, + { + "epoch": 0.65, + "learning_rate": 1.3499102794155345e-07, + "loss": 0.6748, + "step": 17802 + }, + { + "epoch": 0.65, + "learning_rate": 1.3498736587688137e-07, + "loss": 0.7006, + "step": 17803 + }, + { + "epoch": 0.65, + "learning_rate": 1.3498370381220933e-07, + "loss": 0.6912, + "step": 17804 + }, + { + "epoch": 0.65, + "learning_rate": 1.3498004174753725e-07, + "loss": 0.6982, + "step": 17805 + }, + { + "epoch": 0.65, + "learning_rate": 1.349763796828652e-07, + "loss": 0.699, + "step": 17806 + }, + { + "epoch": 0.65, + "learning_rate": 1.3497271761819313e-07, + "loss": 0.6909, + "step": 17807 + }, + { + "epoch": 0.65, + "learning_rate": 1.3496905555352106e-07, + "loss": 0.697, + "step": 17808 + }, + { + "epoch": 0.65, + "learning_rate": 1.34965393488849e-07, + "loss": 0.7055, + "step": 17809 + }, + { + "epoch": 0.65, + "learning_rate": 1.3496173142417694e-07, + "loss": 0.6886, + "step": 17810 + }, + { + "epoch": 0.65, + "learning_rate": 1.349580693595049e-07, + "loss": 0.6916, + "step": 17811 + }, + { + "epoch": 0.65, + "learning_rate": 1.3495440729483282e-07, + "loss": 0.6997, + "step": 17812 + }, + { + "epoch": 0.65, + "learning_rate": 1.3495074523016077e-07, + "loss": 0.7048, + "step": 17813 + }, + { + "epoch": 0.65, + "learning_rate": 1.349470831654887e-07, + "loss": 0.6951, + "step": 17814 + }, + { + "epoch": 0.65, + "learning_rate": 1.3494342110081663e-07, + "loss": 0.6934, + "step": 17815 + }, + { + "epoch": 0.65, + "learning_rate": 1.3493975903614455e-07, + "loss": 0.692, + "step": 17816 + }, + { + "epoch": 0.65, + "learning_rate": 1.349360969714725e-07, + "loss": 0.7084, + "step": 17817 + }, + { + "epoch": 0.65, + "learning_rate": 1.3493243490680043e-07, + "loss": 0.6861, + "step": 17818 + }, + { + "epoch": 0.65, + "learning_rate": 1.3492877284212839e-07, + "loss": 0.6945, + "step": 17819 + }, + { + "epoch": 0.65, + "learning_rate": 1.349251107774563e-07, + "loss": 0.6951, + "step": 17820 + }, + { + "epoch": 0.65, + "learning_rate": 1.3492144871278426e-07, + "loss": 0.6897, + "step": 17821 + }, + { + "epoch": 0.65, + "learning_rate": 1.349177866481122e-07, + "loss": 0.7006, + "step": 17822 + }, + { + "epoch": 0.65, + "learning_rate": 1.3491412458344014e-07, + "loss": 0.6876, + "step": 17823 + }, + { + "epoch": 0.65, + "learning_rate": 1.3491046251876807e-07, + "loss": 0.698, + "step": 17824 + }, + { + "epoch": 0.65, + "learning_rate": 1.3490680045409602e-07, + "loss": 0.6827, + "step": 17825 + }, + { + "epoch": 0.65, + "learning_rate": 1.3490313838942395e-07, + "loss": 0.6922, + "step": 17826 + }, + { + "epoch": 0.65, + "learning_rate": 1.348994763247519e-07, + "loss": 0.6922, + "step": 17827 + }, + { + "epoch": 0.65, + "learning_rate": 1.3489581426007983e-07, + "loss": 0.701, + "step": 17828 + }, + { + "epoch": 0.65, + "learning_rate": 1.3489215219540778e-07, + "loss": 0.6942, + "step": 17829 + }, + { + "epoch": 0.65, + "learning_rate": 1.348884901307357e-07, + "loss": 0.6979, + "step": 17830 + }, + { + "epoch": 0.65, + "learning_rate": 1.3488482806606364e-07, + "loss": 0.6914, + "step": 17831 + }, + { + "epoch": 0.65, + "learning_rate": 1.348811660013916e-07, + "loss": 0.6999, + "step": 17832 + }, + { + "epoch": 0.65, + "learning_rate": 1.3487750393671952e-07, + "loss": 0.6864, + "step": 17833 + }, + { + "epoch": 0.65, + "learning_rate": 1.3487384187204744e-07, + "loss": 0.6921, + "step": 17834 + }, + { + "epoch": 0.65, + "learning_rate": 1.3487017980737537e-07, + "loss": 0.6927, + "step": 17835 + }, + { + "epoch": 0.65, + "learning_rate": 1.3486651774270332e-07, + "loss": 0.6972, + "step": 17836 + }, + { + "epoch": 0.65, + "learning_rate": 1.3486285567803125e-07, + "loss": 0.6985, + "step": 17837 + }, + { + "epoch": 0.65, + "learning_rate": 1.348591936133592e-07, + "loss": 0.7069, + "step": 17838 + }, + { + "epoch": 0.65, + "learning_rate": 1.3485553154868713e-07, + "loss": 0.6961, + "step": 17839 + }, + { + "epoch": 0.65, + "learning_rate": 1.3485186948401508e-07, + "loss": 0.709, + "step": 17840 + }, + { + "epoch": 0.65, + "learning_rate": 1.34848207419343e-07, + "loss": 0.6816, + "step": 17841 + }, + { + "epoch": 0.65, + "learning_rate": 1.3484454535467096e-07, + "loss": 0.6882, + "step": 17842 + }, + { + "epoch": 0.65, + "learning_rate": 1.348408832899989e-07, + "loss": 0.686, + "step": 17843 + }, + { + "epoch": 0.65, + "learning_rate": 1.3483722122532684e-07, + "loss": 0.6931, + "step": 17844 + }, + { + "epoch": 0.65, + "learning_rate": 1.3483355916065477e-07, + "loss": 0.7038, + "step": 17845 + }, + { + "epoch": 0.65, + "learning_rate": 1.3482989709598272e-07, + "loss": 0.7081, + "step": 17846 + }, + { + "epoch": 0.65, + "learning_rate": 1.3482623503131065e-07, + "loss": 0.6962, + "step": 17847 + }, + { + "epoch": 0.65, + "learning_rate": 1.348225729666386e-07, + "loss": 0.695, + "step": 17848 + }, + { + "epoch": 0.65, + "learning_rate": 1.3481891090196653e-07, + "loss": 0.691, + "step": 17849 + }, + { + "epoch": 0.65, + "learning_rate": 1.3481524883729448e-07, + "loss": 0.6916, + "step": 17850 + }, + { + "epoch": 0.65, + "learning_rate": 1.3481158677262238e-07, + "loss": 0.6879, + "step": 17851 + }, + { + "epoch": 0.65, + "learning_rate": 1.3480792470795034e-07, + "loss": 0.6926, + "step": 17852 + }, + { + "epoch": 0.65, + "learning_rate": 1.3480426264327826e-07, + "loss": 0.703, + "step": 17853 + }, + { + "epoch": 0.65, + "learning_rate": 1.3480060057860622e-07, + "loss": 0.693, + "step": 17854 + }, + { + "epoch": 0.65, + "learning_rate": 1.3479693851393414e-07, + "loss": 0.6813, + "step": 17855 + }, + { + "epoch": 0.65, + "learning_rate": 1.347932764492621e-07, + "loss": 0.6902, + "step": 17856 + }, + { + "epoch": 0.65, + "learning_rate": 1.3478961438459002e-07, + "loss": 0.6814, + "step": 17857 + }, + { + "epoch": 0.65, + "learning_rate": 1.3478595231991795e-07, + "loss": 0.6991, + "step": 17858 + }, + { + "epoch": 0.65, + "learning_rate": 1.347822902552459e-07, + "loss": 0.6959, + "step": 17859 + }, + { + "epoch": 0.65, + "learning_rate": 1.3477862819057383e-07, + "loss": 0.6951, + "step": 17860 + }, + { + "epoch": 0.65, + "learning_rate": 1.3477496612590178e-07, + "loss": 0.6988, + "step": 17861 + }, + { + "epoch": 0.65, + "learning_rate": 1.347713040612297e-07, + "loss": 0.6948, + "step": 17862 + }, + { + "epoch": 0.65, + "learning_rate": 1.3476764199655766e-07, + "loss": 0.6888, + "step": 17863 + }, + { + "epoch": 0.65, + "learning_rate": 1.347639799318856e-07, + "loss": 0.6826, + "step": 17864 + }, + { + "epoch": 0.65, + "learning_rate": 1.3476031786721354e-07, + "loss": 0.7029, + "step": 17865 + }, + { + "epoch": 0.65, + "learning_rate": 1.3475665580254147e-07, + "loss": 0.7005, + "step": 17866 + }, + { + "epoch": 0.65, + "learning_rate": 1.3475299373786942e-07, + "loss": 0.6926, + "step": 17867 + }, + { + "epoch": 0.65, + "learning_rate": 1.3474933167319735e-07, + "loss": 0.7091, + "step": 17868 + }, + { + "epoch": 0.65, + "learning_rate": 1.3474566960852528e-07, + "loss": 0.6923, + "step": 17869 + }, + { + "epoch": 0.65, + "learning_rate": 1.347420075438532e-07, + "loss": 0.6934, + "step": 17870 + }, + { + "epoch": 0.65, + "learning_rate": 1.3473834547918116e-07, + "loss": 0.6845, + "step": 17871 + }, + { + "epoch": 0.65, + "learning_rate": 1.3473468341450908e-07, + "loss": 0.684, + "step": 17872 + }, + { + "epoch": 0.65, + "learning_rate": 1.3473102134983704e-07, + "loss": 0.692, + "step": 17873 + }, + { + "epoch": 0.65, + "learning_rate": 1.3472735928516496e-07, + "loss": 0.7008, + "step": 17874 + }, + { + "epoch": 0.65, + "learning_rate": 1.3472369722049292e-07, + "loss": 0.6845, + "step": 17875 + }, + { + "epoch": 0.65, + "learning_rate": 1.3472003515582084e-07, + "loss": 0.6936, + "step": 17876 + }, + { + "epoch": 0.65, + "learning_rate": 1.347163730911488e-07, + "loss": 0.6994, + "step": 17877 + }, + { + "epoch": 0.65, + "learning_rate": 1.3471271102647672e-07, + "loss": 0.6902, + "step": 17878 + }, + { + "epoch": 0.65, + "learning_rate": 1.3470904896180465e-07, + "loss": 0.6866, + "step": 17879 + }, + { + "epoch": 0.65, + "learning_rate": 1.347053868971326e-07, + "loss": 0.6904, + "step": 17880 + }, + { + "epoch": 0.65, + "learning_rate": 1.3470172483246053e-07, + "loss": 0.6968, + "step": 17881 + }, + { + "epoch": 0.65, + "learning_rate": 1.3469806276778848e-07, + "loss": 0.687, + "step": 17882 + }, + { + "epoch": 0.65, + "learning_rate": 1.346944007031164e-07, + "loss": 0.683, + "step": 17883 + }, + { + "epoch": 0.65, + "learning_rate": 1.3469073863844436e-07, + "loss": 0.6896, + "step": 17884 + }, + { + "epoch": 0.65, + "learning_rate": 1.346870765737723e-07, + "loss": 0.6896, + "step": 17885 + }, + { + "epoch": 0.65, + "learning_rate": 1.3468341450910024e-07, + "loss": 0.6934, + "step": 17886 + }, + { + "epoch": 0.65, + "learning_rate": 1.3467975244442814e-07, + "loss": 0.6898, + "step": 17887 + }, + { + "epoch": 0.65, + "learning_rate": 1.346760903797561e-07, + "loss": 0.6986, + "step": 17888 + }, + { + "epoch": 0.65, + "learning_rate": 1.3467242831508402e-07, + "loss": 0.6956, + "step": 17889 + }, + { + "epoch": 0.65, + "learning_rate": 1.3466876625041197e-07, + "loss": 0.6973, + "step": 17890 + }, + { + "epoch": 0.65, + "learning_rate": 1.346651041857399e-07, + "loss": 0.6854, + "step": 17891 + }, + { + "epoch": 0.65, + "learning_rate": 1.3466144212106785e-07, + "loss": 0.6985, + "step": 17892 + }, + { + "epoch": 0.65, + "learning_rate": 1.3465778005639578e-07, + "loss": 0.6922, + "step": 17893 + }, + { + "epoch": 0.65, + "learning_rate": 1.3465411799172373e-07, + "loss": 0.6913, + "step": 17894 + }, + { + "epoch": 0.65, + "learning_rate": 1.3465045592705166e-07, + "loss": 0.6824, + "step": 17895 + }, + { + "epoch": 0.65, + "learning_rate": 1.3464679386237961e-07, + "loss": 0.6975, + "step": 17896 + }, + { + "epoch": 0.65, + "learning_rate": 1.3464313179770754e-07, + "loss": 0.6881, + "step": 17897 + }, + { + "epoch": 0.65, + "learning_rate": 1.346394697330355e-07, + "loss": 0.6826, + "step": 17898 + }, + { + "epoch": 0.65, + "learning_rate": 1.3463580766836342e-07, + "loss": 0.6981, + "step": 17899 + }, + { + "epoch": 0.65, + "learning_rate": 1.3463214560369137e-07, + "loss": 0.6942, + "step": 17900 + }, + { + "epoch": 0.65, + "learning_rate": 1.346284835390193e-07, + "loss": 0.7042, + "step": 17901 + }, + { + "epoch": 0.65, + "learning_rate": 1.3462482147434723e-07, + "loss": 0.698, + "step": 17902 + }, + { + "epoch": 0.66, + "learning_rate": 1.3462115940967518e-07, + "loss": 0.688, + "step": 17903 + }, + { + "epoch": 0.66, + "learning_rate": 1.346174973450031e-07, + "loss": 0.6959, + "step": 17904 + }, + { + "epoch": 0.66, + "learning_rate": 1.3461383528033103e-07, + "loss": 0.6885, + "step": 17905 + }, + { + "epoch": 0.66, + "learning_rate": 1.3461017321565896e-07, + "loss": 0.6897, + "step": 17906 + }, + { + "epoch": 0.66, + "learning_rate": 1.3460651115098691e-07, + "loss": 0.699, + "step": 17907 + }, + { + "epoch": 0.66, + "learning_rate": 1.3460284908631484e-07, + "loss": 0.6866, + "step": 17908 + }, + { + "epoch": 0.66, + "learning_rate": 1.345991870216428e-07, + "loss": 0.704, + "step": 17909 + }, + { + "epoch": 0.66, + "learning_rate": 1.3459552495697072e-07, + "loss": 0.691, + "step": 17910 + }, + { + "epoch": 0.66, + "learning_rate": 1.3459186289229867e-07, + "loss": 0.6797, + "step": 17911 + }, + { + "epoch": 0.66, + "learning_rate": 1.345882008276266e-07, + "loss": 0.687, + "step": 17912 + }, + { + "epoch": 0.66, + "learning_rate": 1.3458453876295455e-07, + "loss": 0.6802, + "step": 17913 + }, + { + "epoch": 0.66, + "learning_rate": 1.3458087669828248e-07, + "loss": 0.699, + "step": 17914 + }, + { + "epoch": 0.66, + "learning_rate": 1.3457721463361043e-07, + "loss": 0.6907, + "step": 17915 + }, + { + "epoch": 0.66, + "learning_rate": 1.3457355256893836e-07, + "loss": 0.6971, + "step": 17916 + }, + { + "epoch": 0.66, + "learning_rate": 1.345698905042663e-07, + "loss": 0.6935, + "step": 17917 + }, + { + "epoch": 0.66, + "learning_rate": 1.3456622843959424e-07, + "loss": 0.7101, + "step": 17918 + }, + { + "epoch": 0.66, + "learning_rate": 1.345625663749222e-07, + "loss": 0.6829, + "step": 17919 + }, + { + "epoch": 0.66, + "learning_rate": 1.3455890431025012e-07, + "loss": 0.7043, + "step": 17920 + }, + { + "epoch": 0.66, + "learning_rate": 1.3455524224557807e-07, + "loss": 0.6951, + "step": 17921 + }, + { + "epoch": 0.66, + "learning_rate": 1.34551580180906e-07, + "loss": 0.7064, + "step": 17922 + }, + { + "epoch": 0.66, + "learning_rate": 1.3454791811623393e-07, + "loss": 0.7031, + "step": 17923 + }, + { + "epoch": 0.66, + "learning_rate": 1.3454425605156185e-07, + "loss": 0.6908, + "step": 17924 + }, + { + "epoch": 0.66, + "learning_rate": 1.345405939868898e-07, + "loss": 0.6965, + "step": 17925 + }, + { + "epoch": 0.66, + "learning_rate": 1.3453693192221773e-07, + "loss": 0.6782, + "step": 17926 + }, + { + "epoch": 0.66, + "learning_rate": 1.3453326985754569e-07, + "loss": 0.7017, + "step": 17927 + }, + { + "epoch": 0.66, + "learning_rate": 1.345296077928736e-07, + "loss": 0.6949, + "step": 17928 + }, + { + "epoch": 0.66, + "learning_rate": 1.3452594572820154e-07, + "loss": 0.6889, + "step": 17929 + }, + { + "epoch": 0.66, + "learning_rate": 1.345222836635295e-07, + "loss": 0.693, + "step": 17930 + }, + { + "epoch": 0.66, + "learning_rate": 1.3451862159885742e-07, + "loss": 0.6789, + "step": 17931 + }, + { + "epoch": 0.66, + "learning_rate": 1.3451495953418537e-07, + "loss": 0.6933, + "step": 17932 + }, + { + "epoch": 0.66, + "learning_rate": 1.345112974695133e-07, + "loss": 0.7075, + "step": 17933 + }, + { + "epoch": 0.66, + "learning_rate": 1.3450763540484125e-07, + "loss": 0.6918, + "step": 17934 + }, + { + "epoch": 0.66, + "learning_rate": 1.3450397334016918e-07, + "loss": 0.6922, + "step": 17935 + }, + { + "epoch": 0.66, + "learning_rate": 1.3450031127549713e-07, + "loss": 0.688, + "step": 17936 + }, + { + "epoch": 0.66, + "learning_rate": 1.3449664921082506e-07, + "loss": 0.6714, + "step": 17937 + }, + { + "epoch": 0.66, + "learning_rate": 1.34492987146153e-07, + "loss": 0.7078, + "step": 17938 + }, + { + "epoch": 0.66, + "learning_rate": 1.3448932508148094e-07, + "loss": 0.686, + "step": 17939 + }, + { + "epoch": 0.66, + "learning_rate": 1.344856630168089e-07, + "loss": 0.6721, + "step": 17940 + }, + { + "epoch": 0.66, + "learning_rate": 1.344820009521368e-07, + "loss": 0.687, + "step": 17941 + }, + { + "epoch": 0.66, + "learning_rate": 1.3447833888746474e-07, + "loss": 0.6827, + "step": 17942 + }, + { + "epoch": 0.66, + "learning_rate": 1.3447467682279267e-07, + "loss": 0.7012, + "step": 17943 + }, + { + "epoch": 0.66, + "learning_rate": 1.3447101475812062e-07, + "loss": 0.7017, + "step": 17944 + }, + { + "epoch": 0.66, + "learning_rate": 1.3446735269344855e-07, + "loss": 0.6845, + "step": 17945 + }, + { + "epoch": 0.66, + "learning_rate": 1.344636906287765e-07, + "loss": 0.694, + "step": 17946 + }, + { + "epoch": 0.66, + "learning_rate": 1.3446002856410443e-07, + "loss": 0.6866, + "step": 17947 + }, + { + "epoch": 0.66, + "learning_rate": 1.3445636649943238e-07, + "loss": 0.7032, + "step": 17948 + }, + { + "epoch": 0.66, + "learning_rate": 1.344527044347603e-07, + "loss": 0.688, + "step": 17949 + }, + { + "epoch": 0.66, + "learning_rate": 1.3444904237008824e-07, + "loss": 0.6923, + "step": 17950 + }, + { + "epoch": 0.66, + "learning_rate": 1.344453803054162e-07, + "loss": 0.697, + "step": 17951 + }, + { + "epoch": 0.66, + "learning_rate": 1.3444171824074412e-07, + "loss": 0.6898, + "step": 17952 + }, + { + "epoch": 0.66, + "learning_rate": 1.3443805617607207e-07, + "loss": 0.6665, + "step": 17953 + }, + { + "epoch": 0.66, + "learning_rate": 1.344343941114e-07, + "loss": 0.6854, + "step": 17954 + }, + { + "epoch": 0.66, + "learning_rate": 1.3443073204672795e-07, + "loss": 0.706, + "step": 17955 + }, + { + "epoch": 0.66, + "learning_rate": 1.3442706998205588e-07, + "loss": 0.696, + "step": 17956 + }, + { + "epoch": 0.66, + "learning_rate": 1.3442340791738383e-07, + "loss": 0.683, + "step": 17957 + }, + { + "epoch": 0.66, + "learning_rate": 1.3441974585271176e-07, + "loss": 0.7042, + "step": 17958 + }, + { + "epoch": 0.66, + "learning_rate": 1.3441608378803968e-07, + "loss": 0.6906, + "step": 17959 + }, + { + "epoch": 0.66, + "learning_rate": 1.344124217233676e-07, + "loss": 0.6933, + "step": 17960 + }, + { + "epoch": 0.66, + "learning_rate": 1.3440875965869556e-07, + "loss": 0.6923, + "step": 17961 + }, + { + "epoch": 0.66, + "learning_rate": 1.344050975940235e-07, + "loss": 0.6973, + "step": 17962 + }, + { + "epoch": 0.66, + "learning_rate": 1.3440143552935144e-07, + "loss": 0.6878, + "step": 17963 + }, + { + "epoch": 0.66, + "learning_rate": 1.3439777346467937e-07, + "loss": 0.6948, + "step": 17964 + }, + { + "epoch": 0.66, + "learning_rate": 1.3439411140000732e-07, + "loss": 0.686, + "step": 17965 + }, + { + "epoch": 0.66, + "learning_rate": 1.3439044933533525e-07, + "loss": 0.695, + "step": 17966 + }, + { + "epoch": 0.66, + "learning_rate": 1.343867872706632e-07, + "loss": 0.7025, + "step": 17967 + }, + { + "epoch": 0.66, + "learning_rate": 1.3438312520599113e-07, + "loss": 0.6778, + "step": 17968 + }, + { + "epoch": 0.66, + "learning_rate": 1.3437946314131908e-07, + "loss": 0.6937, + "step": 17969 + }, + { + "epoch": 0.66, + "learning_rate": 1.34375801076647e-07, + "loss": 0.6901, + "step": 17970 + }, + { + "epoch": 0.66, + "learning_rate": 1.3437213901197494e-07, + "loss": 0.7033, + "step": 17971 + }, + { + "epoch": 0.66, + "learning_rate": 1.343684769473029e-07, + "loss": 0.6911, + "step": 17972 + }, + { + "epoch": 0.66, + "learning_rate": 1.3436481488263082e-07, + "loss": 0.6968, + "step": 17973 + }, + { + "epoch": 0.66, + "learning_rate": 1.3436115281795877e-07, + "loss": 0.6944, + "step": 17974 + }, + { + "epoch": 0.66, + "learning_rate": 1.343574907532867e-07, + "loss": 0.7072, + "step": 17975 + }, + { + "epoch": 0.66, + "learning_rate": 1.3435382868861465e-07, + "loss": 0.7003, + "step": 17976 + }, + { + "epoch": 0.66, + "learning_rate": 1.3435016662394255e-07, + "loss": 0.696, + "step": 17977 + }, + { + "epoch": 0.66, + "learning_rate": 1.343465045592705e-07, + "loss": 0.6971, + "step": 17978 + }, + { + "epoch": 0.66, + "learning_rate": 1.3434284249459843e-07, + "loss": 0.6856, + "step": 17979 + }, + { + "epoch": 0.66, + "learning_rate": 1.3433918042992638e-07, + "loss": 0.6868, + "step": 17980 + }, + { + "epoch": 0.66, + "learning_rate": 1.343355183652543e-07, + "loss": 0.7035, + "step": 17981 + }, + { + "epoch": 0.66, + "learning_rate": 1.3433185630058226e-07, + "loss": 0.6959, + "step": 17982 + }, + { + "epoch": 0.66, + "learning_rate": 1.343281942359102e-07, + "loss": 0.6911, + "step": 17983 + }, + { + "epoch": 0.66, + "learning_rate": 1.3432453217123814e-07, + "loss": 0.6801, + "step": 17984 + }, + { + "epoch": 0.66, + "learning_rate": 1.3432087010656607e-07, + "loss": 0.6901, + "step": 17985 + }, + { + "epoch": 0.66, + "learning_rate": 1.3431720804189402e-07, + "loss": 0.6827, + "step": 17986 + }, + { + "epoch": 0.66, + "learning_rate": 1.3431354597722195e-07, + "loss": 0.6978, + "step": 17987 + }, + { + "epoch": 0.66, + "learning_rate": 1.343098839125499e-07, + "loss": 0.7004, + "step": 17988 + }, + { + "epoch": 0.66, + "learning_rate": 1.3430622184787783e-07, + "loss": 0.7044, + "step": 17989 + }, + { + "epoch": 0.66, + "learning_rate": 1.3430255978320578e-07, + "loss": 0.695, + "step": 17990 + }, + { + "epoch": 0.66, + "learning_rate": 1.342988977185337e-07, + "loss": 0.6958, + "step": 17991 + }, + { + "epoch": 0.66, + "learning_rate": 1.3429523565386166e-07, + "loss": 0.6885, + "step": 17992 + }, + { + "epoch": 0.66, + "learning_rate": 1.342915735891896e-07, + "loss": 0.7054, + "step": 17993 + }, + { + "epoch": 0.66, + "learning_rate": 1.3428791152451752e-07, + "loss": 0.701, + "step": 17994 + }, + { + "epoch": 0.66, + "learning_rate": 1.3428424945984544e-07, + "loss": 0.7017, + "step": 17995 + }, + { + "epoch": 0.66, + "learning_rate": 1.342805873951734e-07, + "loss": 0.7233, + "step": 17996 + }, + { + "epoch": 0.66, + "learning_rate": 1.3427692533050132e-07, + "loss": 0.6987, + "step": 17997 + }, + { + "epoch": 0.66, + "learning_rate": 1.3427326326582928e-07, + "loss": 0.6922, + "step": 17998 + }, + { + "epoch": 0.66, + "learning_rate": 1.342696012011572e-07, + "loss": 0.6875, + "step": 17999 + }, + { + "epoch": 0.66, + "learning_rate": 1.3426593913648513e-07, + "loss": 0.6927, + "step": 18000 + }, + { + "epoch": 0.66, + "learning_rate": 1.3426227707181308e-07, + "loss": 0.6891, + "step": 18001 + }, + { + "epoch": 0.66, + "learning_rate": 1.34258615007141e-07, + "loss": 0.6894, + "step": 18002 + }, + { + "epoch": 0.66, + "learning_rate": 1.3425495294246896e-07, + "loss": 0.7, + "step": 18003 + }, + { + "epoch": 0.66, + "learning_rate": 1.342512908777969e-07, + "loss": 0.6741, + "step": 18004 + }, + { + "epoch": 0.66, + "learning_rate": 1.3424762881312484e-07, + "loss": 0.6899, + "step": 18005 + }, + { + "epoch": 0.66, + "learning_rate": 1.3424396674845277e-07, + "loss": 0.6701, + "step": 18006 + }, + { + "epoch": 0.66, + "learning_rate": 1.3424030468378072e-07, + "loss": 0.69, + "step": 18007 + }, + { + "epoch": 0.66, + "learning_rate": 1.3423664261910865e-07, + "loss": 0.6836, + "step": 18008 + }, + { + "epoch": 0.66, + "learning_rate": 1.342329805544366e-07, + "loss": 0.6962, + "step": 18009 + }, + { + "epoch": 0.66, + "learning_rate": 1.3422931848976453e-07, + "loss": 0.6831, + "step": 18010 + }, + { + "epoch": 0.66, + "learning_rate": 1.3422565642509248e-07, + "loss": 0.6966, + "step": 18011 + }, + { + "epoch": 0.66, + "learning_rate": 1.342219943604204e-07, + "loss": 0.6896, + "step": 18012 + }, + { + "epoch": 0.66, + "learning_rate": 1.3421833229574833e-07, + "loss": 0.6952, + "step": 18013 + }, + { + "epoch": 0.66, + "learning_rate": 1.3421467023107626e-07, + "loss": 0.6831, + "step": 18014 + }, + { + "epoch": 0.66, + "learning_rate": 1.3421100816640421e-07, + "loss": 0.6928, + "step": 18015 + }, + { + "epoch": 0.66, + "learning_rate": 1.3420734610173214e-07, + "loss": 0.6893, + "step": 18016 + }, + { + "epoch": 0.66, + "learning_rate": 1.342036840370601e-07, + "loss": 0.6991, + "step": 18017 + }, + { + "epoch": 0.66, + "learning_rate": 1.3420002197238802e-07, + "loss": 0.6855, + "step": 18018 + }, + { + "epoch": 0.66, + "learning_rate": 1.3419635990771597e-07, + "loss": 0.6852, + "step": 18019 + }, + { + "epoch": 0.66, + "learning_rate": 1.341926978430439e-07, + "loss": 0.6931, + "step": 18020 + }, + { + "epoch": 0.66, + "learning_rate": 1.3418903577837183e-07, + "loss": 0.6963, + "step": 18021 + }, + { + "epoch": 0.66, + "learning_rate": 1.3418537371369978e-07, + "loss": 0.6917, + "step": 18022 + }, + { + "epoch": 0.66, + "learning_rate": 1.341817116490277e-07, + "loss": 0.6766, + "step": 18023 + }, + { + "epoch": 0.66, + "learning_rate": 1.3417804958435566e-07, + "loss": 0.6818, + "step": 18024 + }, + { + "epoch": 0.66, + "learning_rate": 1.341743875196836e-07, + "loss": 0.6889, + "step": 18025 + }, + { + "epoch": 0.66, + "learning_rate": 1.3417072545501154e-07, + "loss": 0.6974, + "step": 18026 + }, + { + "epoch": 0.66, + "learning_rate": 1.3416706339033947e-07, + "loss": 0.6931, + "step": 18027 + }, + { + "epoch": 0.66, + "learning_rate": 1.3416340132566742e-07, + "loss": 0.6996, + "step": 18028 + }, + { + "epoch": 0.66, + "learning_rate": 1.3415973926099535e-07, + "loss": 0.6914, + "step": 18029 + }, + { + "epoch": 0.66, + "learning_rate": 1.341560771963233e-07, + "loss": 0.6875, + "step": 18030 + }, + { + "epoch": 0.66, + "learning_rate": 1.341524151316512e-07, + "loss": 0.694, + "step": 18031 + }, + { + "epoch": 0.66, + "learning_rate": 1.3414875306697915e-07, + "loss": 0.6892, + "step": 18032 + }, + { + "epoch": 0.66, + "learning_rate": 1.3414509100230708e-07, + "loss": 0.703, + "step": 18033 + }, + { + "epoch": 0.66, + "learning_rate": 1.3414142893763503e-07, + "loss": 0.697, + "step": 18034 + }, + { + "epoch": 0.66, + "learning_rate": 1.3413776687296296e-07, + "loss": 0.6933, + "step": 18035 + }, + { + "epoch": 0.66, + "learning_rate": 1.341341048082909e-07, + "loss": 0.6958, + "step": 18036 + }, + { + "epoch": 0.66, + "learning_rate": 1.3413044274361884e-07, + "loss": 0.683, + "step": 18037 + }, + { + "epoch": 0.66, + "learning_rate": 1.341267806789468e-07, + "loss": 0.6854, + "step": 18038 + }, + { + "epoch": 0.66, + "learning_rate": 1.3412311861427472e-07, + "loss": 0.6751, + "step": 18039 + }, + { + "epoch": 0.66, + "learning_rate": 1.3411945654960267e-07, + "loss": 0.6937, + "step": 18040 + }, + { + "epoch": 0.66, + "learning_rate": 1.341157944849306e-07, + "loss": 0.6958, + "step": 18041 + }, + { + "epoch": 0.66, + "learning_rate": 1.3411213242025853e-07, + "loss": 0.6919, + "step": 18042 + }, + { + "epoch": 0.66, + "learning_rate": 1.3410847035558648e-07, + "loss": 0.7012, + "step": 18043 + }, + { + "epoch": 0.66, + "learning_rate": 1.341048082909144e-07, + "loss": 0.7081, + "step": 18044 + }, + { + "epoch": 0.66, + "learning_rate": 1.3410114622624236e-07, + "loss": 0.7021, + "step": 18045 + }, + { + "epoch": 0.66, + "learning_rate": 1.3409748416157029e-07, + "loss": 0.7139, + "step": 18046 + }, + { + "epoch": 0.66, + "learning_rate": 1.3409382209689824e-07, + "loss": 0.6837, + "step": 18047 + }, + { + "epoch": 0.66, + "learning_rate": 1.3409016003222617e-07, + "loss": 0.6822, + "step": 18048 + }, + { + "epoch": 0.66, + "learning_rate": 1.340864979675541e-07, + "loss": 0.7095, + "step": 18049 + }, + { + "epoch": 0.66, + "learning_rate": 1.3408283590288202e-07, + "loss": 0.6916, + "step": 18050 + }, + { + "epoch": 0.66, + "learning_rate": 1.3407917383820997e-07, + "loss": 0.7102, + "step": 18051 + }, + { + "epoch": 0.66, + "learning_rate": 1.340755117735379e-07, + "loss": 0.6884, + "step": 18052 + }, + { + "epoch": 0.66, + "learning_rate": 1.3407184970886585e-07, + "loss": 0.6824, + "step": 18053 + }, + { + "epoch": 0.66, + "learning_rate": 1.3406818764419378e-07, + "loss": 0.6937, + "step": 18054 + }, + { + "epoch": 0.66, + "learning_rate": 1.3406452557952173e-07, + "loss": 0.6984, + "step": 18055 + }, + { + "epoch": 0.66, + "learning_rate": 1.3406086351484966e-07, + "loss": 0.6856, + "step": 18056 + }, + { + "epoch": 0.66, + "learning_rate": 1.340572014501776e-07, + "loss": 0.7037, + "step": 18057 + }, + { + "epoch": 0.66, + "learning_rate": 1.3405353938550554e-07, + "loss": 0.6904, + "step": 18058 + }, + { + "epoch": 0.66, + "learning_rate": 1.340498773208335e-07, + "loss": 0.6775, + "step": 18059 + }, + { + "epoch": 0.66, + "learning_rate": 1.3404621525616142e-07, + "loss": 0.6798, + "step": 18060 + }, + { + "epoch": 0.66, + "learning_rate": 1.3404255319148937e-07, + "loss": 0.6914, + "step": 18061 + }, + { + "epoch": 0.66, + "learning_rate": 1.340388911268173e-07, + "loss": 0.6928, + "step": 18062 + }, + { + "epoch": 0.66, + "learning_rate": 1.3403522906214525e-07, + "loss": 0.6866, + "step": 18063 + }, + { + "epoch": 0.66, + "learning_rate": 1.3403156699747318e-07, + "loss": 0.6956, + "step": 18064 + }, + { + "epoch": 0.66, + "learning_rate": 1.340279049328011e-07, + "loss": 0.6987, + "step": 18065 + }, + { + "epoch": 0.66, + "learning_rate": 1.3402424286812906e-07, + "loss": 0.6862, + "step": 18066 + }, + { + "epoch": 0.66, + "learning_rate": 1.3402058080345698e-07, + "loss": 0.6986, + "step": 18067 + }, + { + "epoch": 0.66, + "learning_rate": 1.340169187387849e-07, + "loss": 0.7077, + "step": 18068 + }, + { + "epoch": 0.66, + "learning_rate": 1.3401325667411286e-07, + "loss": 0.7112, + "step": 18069 + }, + { + "epoch": 0.66, + "learning_rate": 1.340095946094408e-07, + "loss": 0.6858, + "step": 18070 + }, + { + "epoch": 0.66, + "learning_rate": 1.3400593254476872e-07, + "loss": 0.6922, + "step": 18071 + }, + { + "epoch": 0.66, + "learning_rate": 1.3400227048009667e-07, + "loss": 0.6946, + "step": 18072 + }, + { + "epoch": 0.66, + "learning_rate": 1.339986084154246e-07, + "loss": 0.6996, + "step": 18073 + }, + { + "epoch": 0.66, + "learning_rate": 1.3399494635075255e-07, + "loss": 0.7028, + "step": 18074 + }, + { + "epoch": 0.66, + "learning_rate": 1.3399128428608048e-07, + "loss": 0.6942, + "step": 18075 + }, + { + "epoch": 0.66, + "learning_rate": 1.3398762222140843e-07, + "loss": 0.6895, + "step": 18076 + }, + { + "epoch": 0.66, + "learning_rate": 1.3398396015673636e-07, + "loss": 0.6994, + "step": 18077 + }, + { + "epoch": 0.66, + "learning_rate": 1.339802980920643e-07, + "loss": 0.6866, + "step": 18078 + }, + { + "epoch": 0.66, + "learning_rate": 1.3397663602739224e-07, + "loss": 0.6923, + "step": 18079 + }, + { + "epoch": 0.66, + "learning_rate": 1.339729739627202e-07, + "loss": 0.6995, + "step": 18080 + }, + { + "epoch": 0.66, + "learning_rate": 1.3396931189804812e-07, + "loss": 0.7027, + "step": 18081 + }, + { + "epoch": 0.66, + "learning_rate": 1.3396564983337607e-07, + "loss": 0.7014, + "step": 18082 + }, + { + "epoch": 0.66, + "learning_rate": 1.33961987768704e-07, + "loss": 0.6921, + "step": 18083 + }, + { + "epoch": 0.66, + "learning_rate": 1.3395832570403195e-07, + "loss": 0.6818, + "step": 18084 + }, + { + "epoch": 0.66, + "learning_rate": 1.3395466363935985e-07, + "loss": 0.6963, + "step": 18085 + }, + { + "epoch": 0.66, + "learning_rate": 1.339510015746878e-07, + "loss": 0.6906, + "step": 18086 + }, + { + "epoch": 0.66, + "learning_rate": 1.3394733951001573e-07, + "loss": 0.7014, + "step": 18087 + }, + { + "epoch": 0.66, + "learning_rate": 1.3394367744534368e-07, + "loss": 0.7051, + "step": 18088 + }, + { + "epoch": 0.66, + "learning_rate": 1.339400153806716e-07, + "loss": 0.6983, + "step": 18089 + }, + { + "epoch": 0.66, + "learning_rate": 1.3393635331599956e-07, + "loss": 0.6779, + "step": 18090 + }, + { + "epoch": 0.66, + "learning_rate": 1.339326912513275e-07, + "loss": 0.6916, + "step": 18091 + }, + { + "epoch": 0.66, + "learning_rate": 1.3392902918665542e-07, + "loss": 0.7133, + "step": 18092 + }, + { + "epoch": 0.66, + "learning_rate": 1.3392536712198337e-07, + "loss": 0.6755, + "step": 18093 + }, + { + "epoch": 0.66, + "learning_rate": 1.339217050573113e-07, + "loss": 0.698, + "step": 18094 + }, + { + "epoch": 0.66, + "learning_rate": 1.3391804299263925e-07, + "loss": 0.6756, + "step": 18095 + }, + { + "epoch": 0.66, + "learning_rate": 1.3391438092796718e-07, + "loss": 0.6881, + "step": 18096 + }, + { + "epoch": 0.66, + "learning_rate": 1.3391071886329513e-07, + "loss": 0.6838, + "step": 18097 + }, + { + "epoch": 0.66, + "learning_rate": 1.3390705679862306e-07, + "loss": 0.6873, + "step": 18098 + }, + { + "epoch": 0.66, + "learning_rate": 1.33903394733951e-07, + "loss": 0.6866, + "step": 18099 + }, + { + "epoch": 0.66, + "learning_rate": 1.3389973266927894e-07, + "loss": 0.6918, + "step": 18100 + }, + { + "epoch": 0.66, + "learning_rate": 1.338960706046069e-07, + "loss": 0.6904, + "step": 18101 + }, + { + "epoch": 0.66, + "learning_rate": 1.3389240853993482e-07, + "loss": 0.6858, + "step": 18102 + }, + { + "epoch": 0.66, + "learning_rate": 1.3388874647526274e-07, + "loss": 0.7088, + "step": 18103 + }, + { + "epoch": 0.66, + "learning_rate": 1.3388508441059067e-07, + "loss": 0.7069, + "step": 18104 + }, + { + "epoch": 0.66, + "learning_rate": 1.3388142234591862e-07, + "loss": 0.6886, + "step": 18105 + }, + { + "epoch": 0.66, + "learning_rate": 1.3387776028124655e-07, + "loss": 0.6791, + "step": 18106 + }, + { + "epoch": 0.66, + "learning_rate": 1.338740982165745e-07, + "loss": 0.687, + "step": 18107 + }, + { + "epoch": 0.66, + "learning_rate": 1.3387043615190243e-07, + "loss": 0.7079, + "step": 18108 + }, + { + "epoch": 0.66, + "learning_rate": 1.3386677408723038e-07, + "loss": 0.6869, + "step": 18109 + }, + { + "epoch": 0.66, + "learning_rate": 1.338631120225583e-07, + "loss": 0.7077, + "step": 18110 + }, + { + "epoch": 0.66, + "learning_rate": 1.3385944995788626e-07, + "loss": 0.6997, + "step": 18111 + }, + { + "epoch": 0.66, + "learning_rate": 1.338557878932142e-07, + "loss": 0.688, + "step": 18112 + }, + { + "epoch": 0.66, + "learning_rate": 1.3385212582854212e-07, + "loss": 0.6911, + "step": 18113 + }, + { + "epoch": 0.66, + "learning_rate": 1.3384846376387007e-07, + "loss": 0.6832, + "step": 18114 + }, + { + "epoch": 0.66, + "learning_rate": 1.33844801699198e-07, + "loss": 0.7029, + "step": 18115 + }, + { + "epoch": 0.66, + "learning_rate": 1.3384113963452595e-07, + "loss": 0.6996, + "step": 18116 + }, + { + "epoch": 0.66, + "learning_rate": 1.3383747756985388e-07, + "loss": 0.6922, + "step": 18117 + }, + { + "epoch": 0.66, + "learning_rate": 1.3383381550518183e-07, + "loss": 0.705, + "step": 18118 + }, + { + "epoch": 0.66, + "learning_rate": 1.3383015344050975e-07, + "loss": 0.6909, + "step": 18119 + }, + { + "epoch": 0.66, + "learning_rate": 1.338264913758377e-07, + "loss": 0.6883, + "step": 18120 + }, + { + "epoch": 0.66, + "learning_rate": 1.338228293111656e-07, + "loss": 0.7006, + "step": 18121 + }, + { + "epoch": 0.66, + "learning_rate": 1.3381916724649356e-07, + "loss": 0.7043, + "step": 18122 + }, + { + "epoch": 0.66, + "learning_rate": 1.338155051818215e-07, + "loss": 0.7141, + "step": 18123 + }, + { + "epoch": 0.66, + "learning_rate": 1.3381184311714944e-07, + "loss": 0.6932, + "step": 18124 + }, + { + "epoch": 0.66, + "learning_rate": 1.3380818105247737e-07, + "loss": 0.6908, + "step": 18125 + }, + { + "epoch": 0.66, + "learning_rate": 1.3380451898780532e-07, + "loss": 0.7002, + "step": 18126 + }, + { + "epoch": 0.66, + "learning_rate": 1.3380085692313325e-07, + "loss": 0.688, + "step": 18127 + }, + { + "epoch": 0.66, + "learning_rate": 1.337971948584612e-07, + "loss": 0.6949, + "step": 18128 + }, + { + "epoch": 0.66, + "learning_rate": 1.3379353279378913e-07, + "loss": 0.6789, + "step": 18129 + }, + { + "epoch": 0.66, + "learning_rate": 1.3378987072911708e-07, + "loss": 0.6991, + "step": 18130 + }, + { + "epoch": 0.66, + "learning_rate": 1.33786208664445e-07, + "loss": 0.7155, + "step": 18131 + }, + { + "epoch": 0.66, + "learning_rate": 1.3378254659977296e-07, + "loss": 0.7004, + "step": 18132 + }, + { + "epoch": 0.66, + "learning_rate": 1.337788845351009e-07, + "loss": 0.6922, + "step": 18133 + }, + { + "epoch": 0.66, + "learning_rate": 1.3377522247042884e-07, + "loss": 0.6855, + "step": 18134 + }, + { + "epoch": 0.66, + "learning_rate": 1.3377156040575677e-07, + "loss": 0.6936, + "step": 18135 + }, + { + "epoch": 0.66, + "learning_rate": 1.337678983410847e-07, + "loss": 0.6969, + "step": 18136 + }, + { + "epoch": 0.66, + "learning_rate": 1.3376423627641265e-07, + "loss": 0.7019, + "step": 18137 + }, + { + "epoch": 0.66, + "learning_rate": 1.3376057421174057e-07, + "loss": 0.6942, + "step": 18138 + }, + { + "epoch": 0.66, + "learning_rate": 1.337569121470685e-07, + "loss": 0.697, + "step": 18139 + }, + { + "epoch": 0.66, + "learning_rate": 1.3375325008239643e-07, + "loss": 0.7036, + "step": 18140 + }, + { + "epoch": 0.66, + "learning_rate": 1.3374958801772438e-07, + "loss": 0.7014, + "step": 18141 + }, + { + "epoch": 0.66, + "learning_rate": 1.337459259530523e-07, + "loss": 0.7025, + "step": 18142 + }, + { + "epoch": 0.66, + "learning_rate": 1.3374226388838026e-07, + "loss": 0.7033, + "step": 18143 + }, + { + "epoch": 0.66, + "learning_rate": 1.337386018237082e-07, + "loss": 0.7149, + "step": 18144 + }, + { + "epoch": 0.66, + "learning_rate": 1.3373493975903614e-07, + "loss": 0.7048, + "step": 18145 + }, + { + "epoch": 0.66, + "learning_rate": 1.3373127769436407e-07, + "loss": 0.6889, + "step": 18146 + }, + { + "epoch": 0.66, + "learning_rate": 1.3372761562969202e-07, + "loss": 0.6884, + "step": 18147 + }, + { + "epoch": 0.66, + "learning_rate": 1.3372395356501995e-07, + "loss": 0.6926, + "step": 18148 + }, + { + "epoch": 0.66, + "learning_rate": 1.337202915003479e-07, + "loss": 0.6887, + "step": 18149 + }, + { + "epoch": 0.66, + "learning_rate": 1.3371662943567583e-07, + "loss": 0.6986, + "step": 18150 + }, + { + "epoch": 0.66, + "learning_rate": 1.3371296737100378e-07, + "loss": 0.6955, + "step": 18151 + }, + { + "epoch": 0.66, + "learning_rate": 1.337093053063317e-07, + "loss": 0.6872, + "step": 18152 + }, + { + "epoch": 0.66, + "learning_rate": 1.3370564324165966e-07, + "loss": 0.689, + "step": 18153 + }, + { + "epoch": 0.66, + "learning_rate": 1.3370198117698759e-07, + "loss": 0.689, + "step": 18154 + }, + { + "epoch": 0.66, + "learning_rate": 1.3369831911231554e-07, + "loss": 0.7051, + "step": 18155 + }, + { + "epoch": 0.66, + "learning_rate": 1.3369465704764344e-07, + "loss": 0.6937, + "step": 18156 + }, + { + "epoch": 0.66, + "learning_rate": 1.336909949829714e-07, + "loss": 0.6996, + "step": 18157 + }, + { + "epoch": 0.66, + "learning_rate": 1.3368733291829932e-07, + "loss": 0.6959, + "step": 18158 + }, + { + "epoch": 0.66, + "learning_rate": 1.3368367085362727e-07, + "loss": 0.6919, + "step": 18159 + }, + { + "epoch": 0.66, + "learning_rate": 1.336800087889552e-07, + "loss": 0.6812, + "step": 18160 + }, + { + "epoch": 0.66, + "learning_rate": 1.3367634672428315e-07, + "loss": 0.6957, + "step": 18161 + }, + { + "epoch": 0.66, + "learning_rate": 1.3367268465961108e-07, + "loss": 0.7005, + "step": 18162 + }, + { + "epoch": 0.66, + "learning_rate": 1.33669022594939e-07, + "loss": 0.6803, + "step": 18163 + }, + { + "epoch": 0.66, + "learning_rate": 1.3366536053026696e-07, + "loss": 0.6794, + "step": 18164 + }, + { + "epoch": 0.66, + "learning_rate": 1.3366169846559489e-07, + "loss": 0.6821, + "step": 18165 + }, + { + "epoch": 0.66, + "learning_rate": 1.3365803640092284e-07, + "loss": 0.6874, + "step": 18166 + }, + { + "epoch": 0.66, + "learning_rate": 1.3365437433625077e-07, + "loss": 0.6931, + "step": 18167 + }, + { + "epoch": 0.66, + "learning_rate": 1.3365071227157872e-07, + "loss": 0.6875, + "step": 18168 + }, + { + "epoch": 0.66, + "learning_rate": 1.3364705020690665e-07, + "loss": 0.6852, + "step": 18169 + }, + { + "epoch": 0.66, + "learning_rate": 1.336433881422346e-07, + "loss": 0.6869, + "step": 18170 + }, + { + "epoch": 0.66, + "learning_rate": 1.3363972607756253e-07, + "loss": 0.7052, + "step": 18171 + }, + { + "epoch": 0.66, + "learning_rate": 1.3363606401289048e-07, + "loss": 0.6877, + "step": 18172 + }, + { + "epoch": 0.66, + "learning_rate": 1.336324019482184e-07, + "loss": 0.7112, + "step": 18173 + }, + { + "epoch": 0.66, + "learning_rate": 1.3362873988354633e-07, + "loss": 0.6959, + "step": 18174 + }, + { + "epoch": 0.66, + "learning_rate": 1.3362507781887426e-07, + "loss": 0.6888, + "step": 18175 + }, + { + "epoch": 0.67, + "learning_rate": 1.336214157542022e-07, + "loss": 0.7029, + "step": 18176 + }, + { + "epoch": 0.67, + "learning_rate": 1.3361775368953014e-07, + "loss": 0.6899, + "step": 18177 + }, + { + "epoch": 0.67, + "learning_rate": 1.336140916248581e-07, + "loss": 0.6895, + "step": 18178 + }, + { + "epoch": 0.67, + "learning_rate": 1.3361042956018602e-07, + "loss": 0.688, + "step": 18179 + }, + { + "epoch": 0.67, + "learning_rate": 1.3360676749551397e-07, + "loss": 0.7077, + "step": 18180 + }, + { + "epoch": 0.67, + "learning_rate": 1.336031054308419e-07, + "loss": 0.6764, + "step": 18181 + }, + { + "epoch": 0.67, + "learning_rate": 1.3359944336616985e-07, + "loss": 0.6727, + "step": 18182 + }, + { + "epoch": 0.67, + "learning_rate": 1.3359578130149778e-07, + "loss": 0.681, + "step": 18183 + }, + { + "epoch": 0.67, + "learning_rate": 1.335921192368257e-07, + "loss": 0.7033, + "step": 18184 + }, + { + "epoch": 0.67, + "learning_rate": 1.3358845717215366e-07, + "loss": 0.6752, + "step": 18185 + }, + { + "epoch": 0.67, + "learning_rate": 1.3358479510748158e-07, + "loss": 0.6939, + "step": 18186 + }, + { + "epoch": 0.67, + "learning_rate": 1.3358113304280954e-07, + "loss": 0.7049, + "step": 18187 + }, + { + "epoch": 0.67, + "learning_rate": 1.3357747097813746e-07, + "loss": 0.6762, + "step": 18188 + }, + { + "epoch": 0.67, + "learning_rate": 1.3357380891346542e-07, + "loss": 0.6894, + "step": 18189 + }, + { + "epoch": 0.67, + "learning_rate": 1.3357014684879334e-07, + "loss": 0.7007, + "step": 18190 + }, + { + "epoch": 0.67, + "learning_rate": 1.335664847841213e-07, + "loss": 0.7026, + "step": 18191 + }, + { + "epoch": 0.67, + "learning_rate": 1.335628227194492e-07, + "loss": 0.6979, + "step": 18192 + }, + { + "epoch": 0.67, + "learning_rate": 1.3355916065477715e-07, + "loss": 0.6866, + "step": 18193 + }, + { + "epoch": 0.67, + "learning_rate": 1.3355549859010508e-07, + "loss": 0.6916, + "step": 18194 + }, + { + "epoch": 0.67, + "learning_rate": 1.3355183652543303e-07, + "loss": 0.6947, + "step": 18195 + }, + { + "epoch": 0.67, + "learning_rate": 1.3354817446076096e-07, + "loss": 0.6936, + "step": 18196 + }, + { + "epoch": 0.67, + "learning_rate": 1.335445123960889e-07, + "loss": 0.6899, + "step": 18197 + }, + { + "epoch": 0.67, + "learning_rate": 1.3354085033141684e-07, + "loss": 0.6959, + "step": 18198 + }, + { + "epoch": 0.67, + "learning_rate": 1.335371882667448e-07, + "loss": 0.7003, + "step": 18199 + }, + { + "epoch": 0.67, + "learning_rate": 1.3353352620207272e-07, + "loss": 0.6892, + "step": 18200 + }, + { + "epoch": 0.67, + "learning_rate": 1.3352986413740067e-07, + "loss": 0.6945, + "step": 18201 + }, + { + "epoch": 0.67, + "learning_rate": 1.335262020727286e-07, + "loss": 0.6717, + "step": 18202 + }, + { + "epoch": 0.67, + "learning_rate": 1.3352254000805655e-07, + "loss": 0.7007, + "step": 18203 + }, + { + "epoch": 0.67, + "learning_rate": 1.3351887794338448e-07, + "loss": 0.676, + "step": 18204 + }, + { + "epoch": 0.67, + "learning_rate": 1.3351521587871243e-07, + "loss": 0.698, + "step": 18205 + }, + { + "epoch": 0.67, + "learning_rate": 1.3351155381404036e-07, + "loss": 0.7079, + "step": 18206 + }, + { + "epoch": 0.67, + "learning_rate": 1.3350789174936828e-07, + "loss": 0.6933, + "step": 18207 + }, + { + "epoch": 0.67, + "learning_rate": 1.3350422968469624e-07, + "loss": 0.6934, + "step": 18208 + }, + { + "epoch": 0.67, + "learning_rate": 1.3350056762002416e-07, + "loss": 0.6938, + "step": 18209 + }, + { + "epoch": 0.67, + "learning_rate": 1.334969055553521e-07, + "loss": 0.6866, + "step": 18210 + }, + { + "epoch": 0.67, + "learning_rate": 1.3349324349068002e-07, + "loss": 0.6984, + "step": 18211 + }, + { + "epoch": 0.67, + "learning_rate": 1.3348958142600797e-07, + "loss": 0.7051, + "step": 18212 + }, + { + "epoch": 0.67, + "learning_rate": 1.334859193613359e-07, + "loss": 0.6888, + "step": 18213 + }, + { + "epoch": 0.67, + "learning_rate": 1.3348225729666385e-07, + "loss": 0.6909, + "step": 18214 + }, + { + "epoch": 0.67, + "learning_rate": 1.3347859523199178e-07, + "loss": 0.6882, + "step": 18215 + }, + { + "epoch": 0.67, + "learning_rate": 1.3347493316731973e-07, + "loss": 0.6841, + "step": 18216 + }, + { + "epoch": 0.67, + "learning_rate": 1.3347127110264766e-07, + "loss": 0.6974, + "step": 18217 + }, + { + "epoch": 0.67, + "learning_rate": 1.334676090379756e-07, + "loss": 0.6838, + "step": 18218 + }, + { + "epoch": 0.67, + "learning_rate": 1.3346394697330354e-07, + "loss": 0.6995, + "step": 18219 + }, + { + "epoch": 0.67, + "learning_rate": 1.334602849086315e-07, + "loss": 0.6945, + "step": 18220 + }, + { + "epoch": 0.67, + "learning_rate": 1.3345662284395942e-07, + "loss": 0.6889, + "step": 18221 + }, + { + "epoch": 0.67, + "learning_rate": 1.3345296077928737e-07, + "loss": 0.6863, + "step": 18222 + }, + { + "epoch": 0.67, + "learning_rate": 1.334492987146153e-07, + "loss": 0.7004, + "step": 18223 + }, + { + "epoch": 0.67, + "learning_rate": 1.3344563664994325e-07, + "loss": 0.6718, + "step": 18224 + }, + { + "epoch": 0.67, + "learning_rate": 1.3344197458527118e-07, + "loss": 0.6918, + "step": 18225 + }, + { + "epoch": 0.67, + "learning_rate": 1.3343831252059913e-07, + "loss": 0.7092, + "step": 18226 + }, + { + "epoch": 0.67, + "learning_rate": 1.3343465045592706e-07, + "loss": 0.6894, + "step": 18227 + }, + { + "epoch": 0.67, + "learning_rate": 1.3343098839125498e-07, + "loss": 0.6987, + "step": 18228 + }, + { + "epoch": 0.67, + "learning_rate": 1.334273263265829e-07, + "loss": 0.701, + "step": 18229 + }, + { + "epoch": 0.67, + "learning_rate": 1.3342366426191086e-07, + "loss": 0.6928, + "step": 18230 + }, + { + "epoch": 0.67, + "learning_rate": 1.334200021972388e-07, + "loss": 0.7061, + "step": 18231 + }, + { + "epoch": 0.67, + "learning_rate": 1.3341634013256674e-07, + "loss": 0.6936, + "step": 18232 + }, + { + "epoch": 0.67, + "learning_rate": 1.3341267806789467e-07, + "loss": 0.697, + "step": 18233 + }, + { + "epoch": 0.67, + "learning_rate": 1.334090160032226e-07, + "loss": 0.7218, + "step": 18234 + }, + { + "epoch": 0.67, + "learning_rate": 1.3340535393855055e-07, + "loss": 0.6892, + "step": 18235 + }, + { + "epoch": 0.67, + "learning_rate": 1.3340169187387848e-07, + "loss": 0.6976, + "step": 18236 + }, + { + "epoch": 0.67, + "learning_rate": 1.3339802980920643e-07, + "loss": 0.6939, + "step": 18237 + }, + { + "epoch": 0.67, + "learning_rate": 1.3339436774453435e-07, + "loss": 0.6887, + "step": 18238 + }, + { + "epoch": 0.67, + "learning_rate": 1.333907056798623e-07, + "loss": 0.7022, + "step": 18239 + }, + { + "epoch": 0.67, + "learning_rate": 1.3338704361519023e-07, + "loss": 0.6921, + "step": 18240 + }, + { + "epoch": 0.67, + "learning_rate": 1.333833815505182e-07, + "loss": 0.695, + "step": 18241 + }, + { + "epoch": 0.67, + "learning_rate": 1.3337971948584611e-07, + "loss": 0.719, + "step": 18242 + }, + { + "epoch": 0.67, + "learning_rate": 1.3337605742117407e-07, + "loss": 0.6949, + "step": 18243 + }, + { + "epoch": 0.67, + "learning_rate": 1.33372395356502e-07, + "loss": 0.6988, + "step": 18244 + }, + { + "epoch": 0.67, + "learning_rate": 1.3336873329182995e-07, + "loss": 0.697, + "step": 18245 + }, + { + "epoch": 0.67, + "learning_rate": 1.3336507122715785e-07, + "loss": 0.6871, + "step": 18246 + }, + { + "epoch": 0.67, + "learning_rate": 1.333614091624858e-07, + "loss": 0.6955, + "step": 18247 + }, + { + "epoch": 0.67, + "learning_rate": 1.3335774709781373e-07, + "loss": 0.6941, + "step": 18248 + }, + { + "epoch": 0.67, + "learning_rate": 1.3335408503314168e-07, + "loss": 0.6967, + "step": 18249 + }, + { + "epoch": 0.67, + "learning_rate": 1.333504229684696e-07, + "loss": 0.6812, + "step": 18250 + }, + { + "epoch": 0.67, + "learning_rate": 1.3334676090379756e-07, + "loss": 0.6997, + "step": 18251 + }, + { + "epoch": 0.67, + "learning_rate": 1.333430988391255e-07, + "loss": 0.6948, + "step": 18252 + }, + { + "epoch": 0.67, + "learning_rate": 1.3333943677445344e-07, + "loss": 0.6904, + "step": 18253 + }, + { + "epoch": 0.67, + "learning_rate": 1.3333577470978137e-07, + "loss": 0.6881, + "step": 18254 + }, + { + "epoch": 0.67, + "learning_rate": 1.333321126451093e-07, + "loss": 0.6865, + "step": 18255 + }, + { + "epoch": 0.67, + "learning_rate": 1.3332845058043725e-07, + "loss": 0.6912, + "step": 18256 + }, + { + "epoch": 0.67, + "learning_rate": 1.3332478851576517e-07, + "loss": 0.6893, + "step": 18257 + }, + { + "epoch": 0.67, + "learning_rate": 1.3332112645109313e-07, + "loss": 0.6853, + "step": 18258 + }, + { + "epoch": 0.67, + "learning_rate": 1.3331746438642105e-07, + "loss": 0.682, + "step": 18259 + }, + { + "epoch": 0.67, + "learning_rate": 1.33313802321749e-07, + "loss": 0.6828, + "step": 18260 + }, + { + "epoch": 0.67, + "learning_rate": 1.3331014025707693e-07, + "loss": 0.6991, + "step": 18261 + }, + { + "epoch": 0.67, + "learning_rate": 1.333064781924049e-07, + "loss": 0.6895, + "step": 18262 + }, + { + "epoch": 0.67, + "learning_rate": 1.3330281612773281e-07, + "loss": 0.71, + "step": 18263 + }, + { + "epoch": 0.67, + "learning_rate": 1.3329915406306074e-07, + "loss": 0.6959, + "step": 18264 + }, + { + "epoch": 0.67, + "learning_rate": 1.3329549199838867e-07, + "loss": 0.7066, + "step": 18265 + }, + { + "epoch": 0.67, + "learning_rate": 1.3329182993371662e-07, + "loss": 0.697, + "step": 18266 + }, + { + "epoch": 0.67, + "learning_rate": 1.3328816786904455e-07, + "loss": 0.7026, + "step": 18267 + }, + { + "epoch": 0.67, + "learning_rate": 1.332845058043725e-07, + "loss": 0.696, + "step": 18268 + }, + { + "epoch": 0.67, + "learning_rate": 1.3328084373970043e-07, + "loss": 0.6909, + "step": 18269 + }, + { + "epoch": 0.67, + "learning_rate": 1.3327718167502838e-07, + "loss": 0.6803, + "step": 18270 + }, + { + "epoch": 0.67, + "learning_rate": 1.332735196103563e-07, + "loss": 0.6929, + "step": 18271 + }, + { + "epoch": 0.67, + "learning_rate": 1.3326985754568426e-07, + "loss": 0.6833, + "step": 18272 + }, + { + "epoch": 0.67, + "learning_rate": 1.3326619548101219e-07, + "loss": 0.6876, + "step": 18273 + }, + { + "epoch": 0.67, + "learning_rate": 1.3326253341634014e-07, + "loss": 0.6897, + "step": 18274 + }, + { + "epoch": 0.67, + "learning_rate": 1.3325887135166807e-07, + "loss": 0.6827, + "step": 18275 + }, + { + "epoch": 0.67, + "learning_rate": 1.3325520928699602e-07, + "loss": 0.6885, + "step": 18276 + }, + { + "epoch": 0.67, + "learning_rate": 1.3325154722232395e-07, + "loss": 0.7012, + "step": 18277 + }, + { + "epoch": 0.67, + "learning_rate": 1.3324788515765187e-07, + "loss": 0.7105, + "step": 18278 + }, + { + "epoch": 0.67, + "learning_rate": 1.3324422309297983e-07, + "loss": 0.6928, + "step": 18279 + }, + { + "epoch": 0.67, + "learning_rate": 1.3324056102830775e-07, + "loss": 0.6849, + "step": 18280 + }, + { + "epoch": 0.67, + "learning_rate": 1.332368989636357e-07, + "loss": 0.6932, + "step": 18281 + }, + { + "epoch": 0.67, + "learning_rate": 1.332332368989636e-07, + "loss": 0.6938, + "step": 18282 + }, + { + "epoch": 0.67, + "learning_rate": 1.3322957483429156e-07, + "loss": 0.6914, + "step": 18283 + }, + { + "epoch": 0.67, + "learning_rate": 1.3322591276961949e-07, + "loss": 0.6939, + "step": 18284 + }, + { + "epoch": 0.67, + "learning_rate": 1.3322225070494744e-07, + "loss": 0.6991, + "step": 18285 + }, + { + "epoch": 0.67, + "learning_rate": 1.3321858864027537e-07, + "loss": 0.7, + "step": 18286 + }, + { + "epoch": 0.67, + "learning_rate": 1.3321492657560332e-07, + "loss": 0.6936, + "step": 18287 + }, + { + "epoch": 0.67, + "learning_rate": 1.3321126451093125e-07, + "loss": 0.6989, + "step": 18288 + }, + { + "epoch": 0.67, + "learning_rate": 1.332076024462592e-07, + "loss": 0.688, + "step": 18289 + }, + { + "epoch": 0.67, + "learning_rate": 1.3320394038158713e-07, + "loss": 0.7028, + "step": 18290 + }, + { + "epoch": 0.67, + "learning_rate": 1.3320027831691508e-07, + "loss": 0.6906, + "step": 18291 + }, + { + "epoch": 0.67, + "learning_rate": 1.33196616252243e-07, + "loss": 0.7053, + "step": 18292 + }, + { + "epoch": 0.67, + "learning_rate": 1.3319295418757096e-07, + "loss": 0.6974, + "step": 18293 + }, + { + "epoch": 0.67, + "learning_rate": 1.3318929212289889e-07, + "loss": 0.6806, + "step": 18294 + }, + { + "epoch": 0.67, + "learning_rate": 1.3318563005822684e-07, + "loss": 0.6766, + "step": 18295 + }, + { + "epoch": 0.67, + "learning_rate": 1.3318196799355476e-07, + "loss": 0.6877, + "step": 18296 + }, + { + "epoch": 0.67, + "learning_rate": 1.3317830592888272e-07, + "loss": 0.6849, + "step": 18297 + }, + { + "epoch": 0.67, + "learning_rate": 1.3317464386421064e-07, + "loss": 0.6917, + "step": 18298 + }, + { + "epoch": 0.67, + "learning_rate": 1.3317098179953857e-07, + "loss": 0.6899, + "step": 18299 + }, + { + "epoch": 0.67, + "learning_rate": 1.331673197348665e-07, + "loss": 0.6712, + "step": 18300 + }, + { + "epoch": 0.67, + "learning_rate": 1.3316365767019445e-07, + "loss": 0.6935, + "step": 18301 + }, + { + "epoch": 0.67, + "learning_rate": 1.3315999560552238e-07, + "loss": 0.688, + "step": 18302 + }, + { + "epoch": 0.67, + "learning_rate": 1.3315633354085033e-07, + "loss": 0.6812, + "step": 18303 + }, + { + "epoch": 0.67, + "learning_rate": 1.3315267147617826e-07, + "loss": 0.6861, + "step": 18304 + }, + { + "epoch": 0.67, + "learning_rate": 1.3314900941150618e-07, + "loss": 0.697, + "step": 18305 + }, + { + "epoch": 0.67, + "learning_rate": 1.3314534734683414e-07, + "loss": 0.6875, + "step": 18306 + }, + { + "epoch": 0.67, + "learning_rate": 1.3314168528216206e-07, + "loss": 0.6897, + "step": 18307 + }, + { + "epoch": 0.67, + "learning_rate": 1.3313802321749002e-07, + "loss": 0.6982, + "step": 18308 + }, + { + "epoch": 0.67, + "learning_rate": 1.3313436115281794e-07, + "loss": 0.6912, + "step": 18309 + }, + { + "epoch": 0.67, + "learning_rate": 1.331306990881459e-07, + "loss": 0.6853, + "step": 18310 + }, + { + "epoch": 0.67, + "learning_rate": 1.3312703702347382e-07, + "loss": 0.7034, + "step": 18311 + }, + { + "epoch": 0.67, + "learning_rate": 1.3312337495880178e-07, + "loss": 0.6975, + "step": 18312 + }, + { + "epoch": 0.67, + "learning_rate": 1.331197128941297e-07, + "loss": 0.6962, + "step": 18313 + }, + { + "epoch": 0.67, + "learning_rate": 1.3311605082945766e-07, + "loss": 0.692, + "step": 18314 + }, + { + "epoch": 0.67, + "learning_rate": 1.3311238876478558e-07, + "loss": 0.6941, + "step": 18315 + }, + { + "epoch": 0.67, + "learning_rate": 1.3310872670011354e-07, + "loss": 0.6875, + "step": 18316 + }, + { + "epoch": 0.67, + "learning_rate": 1.3310506463544146e-07, + "loss": 0.6942, + "step": 18317 + }, + { + "epoch": 0.67, + "learning_rate": 1.331014025707694e-07, + "loss": 0.7008, + "step": 18318 + }, + { + "epoch": 0.67, + "learning_rate": 1.3309774050609732e-07, + "loss": 0.6948, + "step": 18319 + }, + { + "epoch": 0.67, + "learning_rate": 1.3309407844142527e-07, + "loss": 0.7103, + "step": 18320 + }, + { + "epoch": 0.67, + "learning_rate": 1.330904163767532e-07, + "loss": 0.6881, + "step": 18321 + }, + { + "epoch": 0.67, + "learning_rate": 1.3308675431208115e-07, + "loss": 0.6965, + "step": 18322 + }, + { + "epoch": 0.67, + "learning_rate": 1.3308309224740908e-07, + "loss": 0.6956, + "step": 18323 + }, + { + "epoch": 0.67, + "learning_rate": 1.3307943018273703e-07, + "loss": 0.6961, + "step": 18324 + }, + { + "epoch": 0.67, + "learning_rate": 1.3307576811806496e-07, + "loss": 0.6892, + "step": 18325 + }, + { + "epoch": 0.67, + "learning_rate": 1.3307210605339288e-07, + "loss": 0.6922, + "step": 18326 + }, + { + "epoch": 0.67, + "learning_rate": 1.3306844398872084e-07, + "loss": 0.6848, + "step": 18327 + }, + { + "epoch": 0.67, + "learning_rate": 1.3306478192404876e-07, + "loss": 0.6971, + "step": 18328 + }, + { + "epoch": 0.67, + "learning_rate": 1.3306111985937672e-07, + "loss": 0.693, + "step": 18329 + }, + { + "epoch": 0.67, + "learning_rate": 1.3305745779470464e-07, + "loss": 0.7037, + "step": 18330 + }, + { + "epoch": 0.67, + "learning_rate": 1.330537957300326e-07, + "loss": 0.7165, + "step": 18331 + }, + { + "epoch": 0.67, + "learning_rate": 1.3305013366536052e-07, + "loss": 0.6866, + "step": 18332 + }, + { + "epoch": 0.67, + "learning_rate": 1.3304647160068848e-07, + "loss": 0.6943, + "step": 18333 + }, + { + "epoch": 0.67, + "learning_rate": 1.330428095360164e-07, + "loss": 0.6843, + "step": 18334 + }, + { + "epoch": 0.67, + "learning_rate": 1.3303914747134436e-07, + "loss": 0.6851, + "step": 18335 + }, + { + "epoch": 0.67, + "learning_rate": 1.3303548540667226e-07, + "loss": 0.6715, + "step": 18336 + }, + { + "epoch": 0.67, + "learning_rate": 1.330318233420002e-07, + "loss": 0.6897, + "step": 18337 + }, + { + "epoch": 0.67, + "learning_rate": 1.3302816127732814e-07, + "loss": 0.698, + "step": 18338 + }, + { + "epoch": 0.67, + "learning_rate": 1.330244992126561e-07, + "loss": 0.687, + "step": 18339 + }, + { + "epoch": 0.67, + "learning_rate": 1.3302083714798402e-07, + "loss": 0.6988, + "step": 18340 + }, + { + "epoch": 0.67, + "learning_rate": 1.3301717508331197e-07, + "loss": 0.6927, + "step": 18341 + }, + { + "epoch": 0.67, + "learning_rate": 1.330135130186399e-07, + "loss": 0.6856, + "step": 18342 + }, + { + "epoch": 0.67, + "learning_rate": 1.3300985095396785e-07, + "loss": 0.6917, + "step": 18343 + }, + { + "epoch": 0.67, + "learning_rate": 1.3300618888929578e-07, + "loss": 0.6921, + "step": 18344 + }, + { + "epoch": 0.67, + "learning_rate": 1.3300252682462373e-07, + "loss": 0.6957, + "step": 18345 + }, + { + "epoch": 0.67, + "learning_rate": 1.3299886475995166e-07, + "loss": 0.6834, + "step": 18346 + }, + { + "epoch": 0.67, + "learning_rate": 1.329952026952796e-07, + "loss": 0.6913, + "step": 18347 + }, + { + "epoch": 0.67, + "learning_rate": 1.3299154063060754e-07, + "loss": 0.6877, + "step": 18348 + }, + { + "epoch": 0.67, + "learning_rate": 1.3298787856593546e-07, + "loss": 0.6804, + "step": 18349 + }, + { + "epoch": 0.67, + "learning_rate": 1.3298421650126342e-07, + "loss": 0.7046, + "step": 18350 + }, + { + "epoch": 0.67, + "learning_rate": 1.3298055443659134e-07, + "loss": 0.6813, + "step": 18351 + }, + { + "epoch": 0.67, + "learning_rate": 1.329768923719193e-07, + "loss": 0.6911, + "step": 18352 + }, + { + "epoch": 0.67, + "learning_rate": 1.3297323030724722e-07, + "loss": 0.688, + "step": 18353 + }, + { + "epoch": 0.67, + "learning_rate": 1.3296956824257515e-07, + "loss": 0.6847, + "step": 18354 + }, + { + "epoch": 0.67, + "learning_rate": 1.3296590617790308e-07, + "loss": 0.7103, + "step": 18355 + }, + { + "epoch": 0.67, + "learning_rate": 1.3296224411323103e-07, + "loss": 0.6913, + "step": 18356 + }, + { + "epoch": 0.67, + "learning_rate": 1.3295858204855896e-07, + "loss": 0.697, + "step": 18357 + }, + { + "epoch": 0.67, + "learning_rate": 1.329549199838869e-07, + "loss": 0.6988, + "step": 18358 + }, + { + "epoch": 0.67, + "learning_rate": 1.3295125791921483e-07, + "loss": 0.6901, + "step": 18359 + }, + { + "epoch": 0.67, + "learning_rate": 1.329475958545428e-07, + "loss": 0.696, + "step": 18360 + }, + { + "epoch": 0.67, + "learning_rate": 1.3294393378987071e-07, + "loss": 0.689, + "step": 18361 + }, + { + "epoch": 0.67, + "learning_rate": 1.3294027172519867e-07, + "loss": 0.6918, + "step": 18362 + }, + { + "epoch": 0.67, + "learning_rate": 1.329366096605266e-07, + "loss": 0.7072, + "step": 18363 + }, + { + "epoch": 0.67, + "learning_rate": 1.3293294759585455e-07, + "loss": 0.6937, + "step": 18364 + }, + { + "epoch": 0.67, + "learning_rate": 1.3292928553118247e-07, + "loss": 0.6977, + "step": 18365 + }, + { + "epoch": 0.67, + "learning_rate": 1.3292562346651043e-07, + "loss": 0.6817, + "step": 18366 + }, + { + "epoch": 0.67, + "learning_rate": 1.3292196140183835e-07, + "loss": 0.6991, + "step": 18367 + }, + { + "epoch": 0.67, + "learning_rate": 1.329182993371663e-07, + "loss": 0.7013, + "step": 18368 + }, + { + "epoch": 0.67, + "learning_rate": 1.3291463727249423e-07, + "loss": 0.6924, + "step": 18369 + }, + { + "epoch": 0.67, + "learning_rate": 1.3291097520782216e-07, + "loss": 0.6912, + "step": 18370 + }, + { + "epoch": 0.67, + "learning_rate": 1.3290731314315011e-07, + "loss": 0.6837, + "step": 18371 + }, + { + "epoch": 0.67, + "learning_rate": 1.3290365107847804e-07, + "loss": 0.7011, + "step": 18372 + }, + { + "epoch": 0.67, + "learning_rate": 1.3289998901380597e-07, + "loss": 0.6965, + "step": 18373 + }, + { + "epoch": 0.67, + "learning_rate": 1.3289632694913392e-07, + "loss": 0.6967, + "step": 18374 + }, + { + "epoch": 0.67, + "learning_rate": 1.3289266488446185e-07, + "loss": 0.6906, + "step": 18375 + }, + { + "epoch": 0.67, + "learning_rate": 1.3288900281978977e-07, + "loss": 0.7001, + "step": 18376 + }, + { + "epoch": 0.67, + "learning_rate": 1.3288534075511773e-07, + "loss": 0.7037, + "step": 18377 + }, + { + "epoch": 0.67, + "learning_rate": 1.3288167869044565e-07, + "loss": 0.7055, + "step": 18378 + }, + { + "epoch": 0.67, + "learning_rate": 1.328780166257736e-07, + "loss": 0.713, + "step": 18379 + }, + { + "epoch": 0.67, + "learning_rate": 1.3287435456110153e-07, + "loss": 0.7086, + "step": 18380 + }, + { + "epoch": 0.67, + "learning_rate": 1.328706924964295e-07, + "loss": 0.69, + "step": 18381 + }, + { + "epoch": 0.67, + "learning_rate": 1.3286703043175741e-07, + "loss": 0.6976, + "step": 18382 + }, + { + "epoch": 0.67, + "learning_rate": 1.3286336836708537e-07, + "loss": 0.7032, + "step": 18383 + }, + { + "epoch": 0.67, + "learning_rate": 1.328597063024133e-07, + "loss": 0.708, + "step": 18384 + }, + { + "epoch": 0.67, + "learning_rate": 1.3285604423774125e-07, + "loss": 0.6923, + "step": 18385 + }, + { + "epoch": 0.67, + "learning_rate": 1.3285238217306917e-07, + "loss": 0.6946, + "step": 18386 + }, + { + "epoch": 0.67, + "learning_rate": 1.3284872010839713e-07, + "loss": 0.6941, + "step": 18387 + }, + { + "epoch": 0.67, + "learning_rate": 1.3284505804372505e-07, + "loss": 0.6914, + "step": 18388 + }, + { + "epoch": 0.67, + "learning_rate": 1.32841395979053e-07, + "loss": 0.7194, + "step": 18389 + }, + { + "epoch": 0.67, + "learning_rate": 1.328377339143809e-07, + "loss": 0.6907, + "step": 18390 + }, + { + "epoch": 0.67, + "learning_rate": 1.3283407184970886e-07, + "loss": 0.6925, + "step": 18391 + }, + { + "epoch": 0.67, + "learning_rate": 1.3283040978503679e-07, + "loss": 0.6913, + "step": 18392 + }, + { + "epoch": 0.67, + "learning_rate": 1.3282674772036474e-07, + "loss": 0.6911, + "step": 18393 + }, + { + "epoch": 0.67, + "learning_rate": 1.3282308565569267e-07, + "loss": 0.6976, + "step": 18394 + }, + { + "epoch": 0.67, + "learning_rate": 1.3281942359102062e-07, + "loss": 0.7058, + "step": 18395 + }, + { + "epoch": 0.67, + "learning_rate": 1.3281576152634855e-07, + "loss": 0.6926, + "step": 18396 + }, + { + "epoch": 0.67, + "learning_rate": 1.3281209946167647e-07, + "loss": 0.6881, + "step": 18397 + }, + { + "epoch": 0.67, + "learning_rate": 1.3280843739700443e-07, + "loss": 0.7027, + "step": 18398 + }, + { + "epoch": 0.67, + "learning_rate": 1.3280477533233235e-07, + "loss": 0.7108, + "step": 18399 + }, + { + "epoch": 0.67, + "learning_rate": 1.328011132676603e-07, + "loss": 0.6855, + "step": 18400 + }, + { + "epoch": 0.67, + "learning_rate": 1.3279745120298823e-07, + "loss": 0.6876, + "step": 18401 + }, + { + "epoch": 0.67, + "learning_rate": 1.3279378913831619e-07, + "loss": 0.6999, + "step": 18402 + }, + { + "epoch": 0.67, + "learning_rate": 1.327901270736441e-07, + "loss": 0.6923, + "step": 18403 + }, + { + "epoch": 0.67, + "learning_rate": 1.3278646500897207e-07, + "loss": 0.6821, + "step": 18404 + }, + { + "epoch": 0.67, + "learning_rate": 1.327828029443e-07, + "loss": 0.6927, + "step": 18405 + }, + { + "epoch": 0.67, + "learning_rate": 1.3277914087962795e-07, + "loss": 0.6902, + "step": 18406 + }, + { + "epoch": 0.67, + "learning_rate": 1.3277547881495587e-07, + "loss": 0.6954, + "step": 18407 + }, + { + "epoch": 0.67, + "learning_rate": 1.327718167502838e-07, + "loss": 0.6866, + "step": 18408 + }, + { + "epoch": 0.67, + "learning_rate": 1.3276815468561173e-07, + "loss": 0.6878, + "step": 18409 + }, + { + "epoch": 0.67, + "learning_rate": 1.3276449262093968e-07, + "loss": 0.6832, + "step": 18410 + }, + { + "epoch": 0.67, + "learning_rate": 1.327608305562676e-07, + "loss": 0.6806, + "step": 18411 + }, + { + "epoch": 0.67, + "learning_rate": 1.3275716849159556e-07, + "loss": 0.6909, + "step": 18412 + }, + { + "epoch": 0.67, + "learning_rate": 1.3275350642692349e-07, + "loss": 0.6888, + "step": 18413 + }, + { + "epoch": 0.67, + "learning_rate": 1.3274984436225144e-07, + "loss": 0.6941, + "step": 18414 + }, + { + "epoch": 0.67, + "learning_rate": 1.3274618229757936e-07, + "loss": 0.6932, + "step": 18415 + }, + { + "epoch": 0.67, + "learning_rate": 1.3274252023290732e-07, + "loss": 0.699, + "step": 18416 + }, + { + "epoch": 0.67, + "learning_rate": 1.3273885816823524e-07, + "loss": 0.6865, + "step": 18417 + }, + { + "epoch": 0.67, + "learning_rate": 1.327351961035632e-07, + "loss": 0.6922, + "step": 18418 + }, + { + "epoch": 0.67, + "learning_rate": 1.3273153403889112e-07, + "loss": 0.6897, + "step": 18419 + }, + { + "epoch": 0.67, + "learning_rate": 1.3272787197421905e-07, + "loss": 0.6879, + "step": 18420 + }, + { + "epoch": 0.67, + "learning_rate": 1.32724209909547e-07, + "loss": 0.6972, + "step": 18421 + }, + { + "epoch": 0.67, + "learning_rate": 1.3272054784487493e-07, + "loss": 0.6945, + "step": 18422 + }, + { + "epoch": 0.67, + "learning_rate": 1.3271688578020288e-07, + "loss": 0.7041, + "step": 18423 + }, + { + "epoch": 0.67, + "learning_rate": 1.327132237155308e-07, + "loss": 0.7043, + "step": 18424 + }, + { + "epoch": 0.67, + "learning_rate": 1.3270956165085876e-07, + "loss": 0.6976, + "step": 18425 + }, + { + "epoch": 0.67, + "learning_rate": 1.3270589958618666e-07, + "loss": 0.6782, + "step": 18426 + }, + { + "epoch": 0.67, + "learning_rate": 1.3270223752151462e-07, + "loss": 0.7016, + "step": 18427 + }, + { + "epoch": 0.67, + "learning_rate": 1.3269857545684254e-07, + "loss": 0.6702, + "step": 18428 + }, + { + "epoch": 0.67, + "learning_rate": 1.326949133921705e-07, + "loss": 0.7038, + "step": 18429 + }, + { + "epoch": 0.67, + "learning_rate": 1.3269125132749842e-07, + "loss": 0.6921, + "step": 18430 + }, + { + "epoch": 0.67, + "learning_rate": 1.3268758926282638e-07, + "loss": 0.6849, + "step": 18431 + }, + { + "epoch": 0.67, + "learning_rate": 1.326839271981543e-07, + "loss": 0.6903, + "step": 18432 + }, + { + "epoch": 0.67, + "learning_rate": 1.3268026513348226e-07, + "loss": 0.7207, + "step": 18433 + }, + { + "epoch": 0.67, + "learning_rate": 1.3267660306881018e-07, + "loss": 0.6987, + "step": 18434 + }, + { + "epoch": 0.67, + "learning_rate": 1.3267294100413814e-07, + "loss": 0.692, + "step": 18435 + }, + { + "epoch": 0.67, + "learning_rate": 1.3266927893946606e-07, + "loss": 0.6762, + "step": 18436 + }, + { + "epoch": 0.67, + "learning_rate": 1.3266561687479402e-07, + "loss": 0.7113, + "step": 18437 + }, + { + "epoch": 0.67, + "learning_rate": 1.3266195481012194e-07, + "loss": 0.6932, + "step": 18438 + }, + { + "epoch": 0.67, + "learning_rate": 1.326582927454499e-07, + "loss": 0.703, + "step": 18439 + }, + { + "epoch": 0.67, + "learning_rate": 1.3265463068077782e-07, + "loss": 0.7008, + "step": 18440 + }, + { + "epoch": 0.67, + "learning_rate": 1.3265096861610575e-07, + "loss": 0.7054, + "step": 18441 + }, + { + "epoch": 0.67, + "learning_rate": 1.326473065514337e-07, + "loss": 0.6924, + "step": 18442 + }, + { + "epoch": 0.67, + "learning_rate": 1.3264364448676163e-07, + "loss": 0.6915, + "step": 18443 + }, + { + "epoch": 0.67, + "learning_rate": 1.3263998242208956e-07, + "loss": 0.71, + "step": 18444 + }, + { + "epoch": 0.67, + "learning_rate": 1.326363203574175e-07, + "loss": 0.6799, + "step": 18445 + }, + { + "epoch": 0.67, + "learning_rate": 1.3263265829274544e-07, + "loss": 0.6915, + "step": 18446 + }, + { + "epoch": 0.67, + "learning_rate": 1.3262899622807336e-07, + "loss": 0.6912, + "step": 18447 + }, + { + "epoch": 0.67, + "learning_rate": 1.3262533416340132e-07, + "loss": 0.6966, + "step": 18448 + }, + { + "epoch": 0.67, + "learning_rate": 1.3262167209872924e-07, + "loss": 0.686, + "step": 18449 + }, + { + "epoch": 0.68, + "learning_rate": 1.326180100340572e-07, + "loss": 0.6876, + "step": 18450 + }, + { + "epoch": 0.68, + "learning_rate": 1.3261434796938512e-07, + "loss": 0.7121, + "step": 18451 + }, + { + "epoch": 0.68, + "learning_rate": 1.3261068590471308e-07, + "loss": 0.6869, + "step": 18452 + }, + { + "epoch": 0.68, + "learning_rate": 1.32607023840041e-07, + "loss": 0.6862, + "step": 18453 + }, + { + "epoch": 0.68, + "learning_rate": 1.3260336177536896e-07, + "loss": 0.7047, + "step": 18454 + }, + { + "epoch": 0.68, + "learning_rate": 1.3259969971069688e-07, + "loss": 0.6739, + "step": 18455 + }, + { + "epoch": 0.68, + "learning_rate": 1.3259603764602484e-07, + "loss": 0.6816, + "step": 18456 + }, + { + "epoch": 0.68, + "learning_rate": 1.3259237558135276e-07, + "loss": 0.6913, + "step": 18457 + }, + { + "epoch": 0.68, + "learning_rate": 1.3258871351668072e-07, + "loss": 0.6933, + "step": 18458 + }, + { + "epoch": 0.68, + "learning_rate": 1.3258505145200864e-07, + "loss": 0.6881, + "step": 18459 + }, + { + "epoch": 0.68, + "learning_rate": 1.325813893873366e-07, + "loss": 0.6943, + "step": 18460 + }, + { + "epoch": 0.68, + "learning_rate": 1.3257772732266452e-07, + "loss": 0.7016, + "step": 18461 + }, + { + "epoch": 0.68, + "learning_rate": 1.3257406525799245e-07, + "loss": 0.6894, + "step": 18462 + }, + { + "epoch": 0.68, + "learning_rate": 1.3257040319332038e-07, + "loss": 0.6933, + "step": 18463 + }, + { + "epoch": 0.68, + "learning_rate": 1.3256674112864833e-07, + "loss": 0.6981, + "step": 18464 + }, + { + "epoch": 0.68, + "learning_rate": 1.3256307906397626e-07, + "loss": 0.6845, + "step": 18465 + }, + { + "epoch": 0.68, + "learning_rate": 1.325594169993042e-07, + "loss": 0.6939, + "step": 18466 + }, + { + "epoch": 0.68, + "learning_rate": 1.3255575493463214e-07, + "loss": 0.685, + "step": 18467 + }, + { + "epoch": 0.68, + "learning_rate": 1.3255209286996006e-07, + "loss": 0.6929, + "step": 18468 + }, + { + "epoch": 0.68, + "learning_rate": 1.3254843080528802e-07, + "loss": 0.7116, + "step": 18469 + }, + { + "epoch": 0.68, + "learning_rate": 1.3254476874061594e-07, + "loss": 0.6741, + "step": 18470 + }, + { + "epoch": 0.68, + "learning_rate": 1.325411066759439e-07, + "loss": 0.6821, + "step": 18471 + }, + { + "epoch": 0.68, + "learning_rate": 1.3253744461127182e-07, + "loss": 0.6975, + "step": 18472 + }, + { + "epoch": 0.68, + "learning_rate": 1.3253378254659977e-07, + "loss": 0.6923, + "step": 18473 + }, + { + "epoch": 0.68, + "learning_rate": 1.325301204819277e-07, + "loss": 0.698, + "step": 18474 + }, + { + "epoch": 0.68, + "learning_rate": 1.3252645841725565e-07, + "loss": 0.7059, + "step": 18475 + }, + { + "epoch": 0.68, + "learning_rate": 1.3252279635258358e-07, + "loss": 0.6973, + "step": 18476 + }, + { + "epoch": 0.68, + "learning_rate": 1.3251913428791153e-07, + "loss": 0.6945, + "step": 18477 + }, + { + "epoch": 0.68, + "learning_rate": 1.3251547222323946e-07, + "loss": 0.6973, + "step": 18478 + }, + { + "epoch": 0.68, + "learning_rate": 1.3251181015856741e-07, + "loss": 0.7029, + "step": 18479 + }, + { + "epoch": 0.68, + "learning_rate": 1.3250814809389531e-07, + "loss": 0.6929, + "step": 18480 + }, + { + "epoch": 0.68, + "learning_rate": 1.3250448602922327e-07, + "loss": 0.7143, + "step": 18481 + }, + { + "epoch": 0.68, + "learning_rate": 1.325008239645512e-07, + "loss": 0.6991, + "step": 18482 + }, + { + "epoch": 0.68, + "learning_rate": 1.3249716189987915e-07, + "loss": 0.7055, + "step": 18483 + }, + { + "epoch": 0.68, + "learning_rate": 1.3249349983520707e-07, + "loss": 0.684, + "step": 18484 + }, + { + "epoch": 0.68, + "learning_rate": 1.3248983777053503e-07, + "loss": 0.6955, + "step": 18485 + }, + { + "epoch": 0.68, + "learning_rate": 1.3248617570586295e-07, + "loss": 0.6941, + "step": 18486 + }, + { + "epoch": 0.68, + "learning_rate": 1.324825136411909e-07, + "loss": 0.6886, + "step": 18487 + }, + { + "epoch": 0.68, + "learning_rate": 1.3247885157651883e-07, + "loss": 0.7022, + "step": 18488 + }, + { + "epoch": 0.68, + "learning_rate": 1.324751895118468e-07, + "loss": 0.6918, + "step": 18489 + }, + { + "epoch": 0.68, + "learning_rate": 1.3247152744717471e-07, + "loss": 0.6851, + "step": 18490 + }, + { + "epoch": 0.68, + "learning_rate": 1.3246786538250264e-07, + "loss": 0.6974, + "step": 18491 + }, + { + "epoch": 0.68, + "learning_rate": 1.324642033178306e-07, + "loss": 0.6903, + "step": 18492 + }, + { + "epoch": 0.68, + "learning_rate": 1.3246054125315852e-07, + "loss": 0.6888, + "step": 18493 + }, + { + "epoch": 0.68, + "learning_rate": 1.3245687918848647e-07, + "loss": 0.6885, + "step": 18494 + }, + { + "epoch": 0.68, + "learning_rate": 1.324532171238144e-07, + "loss": 0.6843, + "step": 18495 + }, + { + "epoch": 0.68, + "learning_rate": 1.3244955505914235e-07, + "loss": 0.6851, + "step": 18496 + }, + { + "epoch": 0.68, + "learning_rate": 1.3244589299447028e-07, + "loss": 0.6983, + "step": 18497 + }, + { + "epoch": 0.68, + "learning_rate": 1.324422309297982e-07, + "loss": 0.707, + "step": 18498 + }, + { + "epoch": 0.68, + "learning_rate": 1.3243856886512613e-07, + "loss": 0.685, + "step": 18499 + }, + { + "epoch": 0.68, + "learning_rate": 1.324349068004541e-07, + "loss": 0.7179, + "step": 18500 + }, + { + "epoch": 0.68, + "learning_rate": 1.3243124473578201e-07, + "loss": 0.6913, + "step": 18501 + }, + { + "epoch": 0.68, + "learning_rate": 1.3242758267110997e-07, + "loss": 0.6793, + "step": 18502 + }, + { + "epoch": 0.68, + "learning_rate": 1.324239206064379e-07, + "loss": 0.6844, + "step": 18503 + }, + { + "epoch": 0.68, + "learning_rate": 1.3242025854176585e-07, + "loss": 0.687, + "step": 18504 + }, + { + "epoch": 0.68, + "learning_rate": 1.3241659647709377e-07, + "loss": 0.6964, + "step": 18505 + }, + { + "epoch": 0.68, + "learning_rate": 1.3241293441242173e-07, + "loss": 0.6854, + "step": 18506 + }, + { + "epoch": 0.68, + "learning_rate": 1.3240927234774965e-07, + "loss": 0.6992, + "step": 18507 + }, + { + "epoch": 0.68, + "learning_rate": 1.324056102830776e-07, + "loss": 0.6791, + "step": 18508 + }, + { + "epoch": 0.68, + "learning_rate": 1.3240194821840553e-07, + "loss": 0.6941, + "step": 18509 + }, + { + "epoch": 0.68, + "learning_rate": 1.3239828615373349e-07, + "loss": 0.6876, + "step": 18510 + }, + { + "epoch": 0.68, + "learning_rate": 1.323946240890614e-07, + "loss": 0.6961, + "step": 18511 + }, + { + "epoch": 0.68, + "learning_rate": 1.3239096202438934e-07, + "loss": 0.6949, + "step": 18512 + }, + { + "epoch": 0.68, + "learning_rate": 1.323872999597173e-07, + "loss": 0.6925, + "step": 18513 + }, + { + "epoch": 0.68, + "learning_rate": 1.3238363789504522e-07, + "loss": 0.6941, + "step": 18514 + }, + { + "epoch": 0.68, + "learning_rate": 1.3237997583037315e-07, + "loss": 0.712, + "step": 18515 + }, + { + "epoch": 0.68, + "learning_rate": 1.323763137657011e-07, + "loss": 0.7009, + "step": 18516 + }, + { + "epoch": 0.68, + "learning_rate": 1.3237265170102903e-07, + "loss": 0.702, + "step": 18517 + }, + { + "epoch": 0.68, + "learning_rate": 1.3236898963635695e-07, + "loss": 0.6902, + "step": 18518 + }, + { + "epoch": 0.68, + "learning_rate": 1.323653275716849e-07, + "loss": 0.7004, + "step": 18519 + }, + { + "epoch": 0.68, + "learning_rate": 1.3236166550701283e-07, + "loss": 0.6959, + "step": 18520 + }, + { + "epoch": 0.68, + "learning_rate": 1.3235800344234079e-07, + "loss": 0.6736, + "step": 18521 + }, + { + "epoch": 0.68, + "learning_rate": 1.323543413776687e-07, + "loss": 0.7081, + "step": 18522 + }, + { + "epoch": 0.68, + "learning_rate": 1.3235067931299667e-07, + "loss": 0.6955, + "step": 18523 + }, + { + "epoch": 0.68, + "learning_rate": 1.323470172483246e-07, + "loss": 0.6878, + "step": 18524 + }, + { + "epoch": 0.68, + "learning_rate": 1.3234335518365255e-07, + "loss": 0.7036, + "step": 18525 + }, + { + "epoch": 0.68, + "learning_rate": 1.3233969311898047e-07, + "loss": 0.6806, + "step": 18526 + }, + { + "epoch": 0.68, + "learning_rate": 1.3233603105430843e-07, + "loss": 0.6972, + "step": 18527 + }, + { + "epoch": 0.68, + "learning_rate": 1.3233236898963635e-07, + "loss": 0.6919, + "step": 18528 + }, + { + "epoch": 0.68, + "learning_rate": 1.323287069249643e-07, + "loss": 0.6931, + "step": 18529 + }, + { + "epoch": 0.68, + "learning_rate": 1.3232504486029223e-07, + "loss": 0.6953, + "step": 18530 + }, + { + "epoch": 0.68, + "learning_rate": 1.3232138279562018e-07, + "loss": 0.6796, + "step": 18531 + }, + { + "epoch": 0.68, + "learning_rate": 1.323177207309481e-07, + "loss": 0.7078, + "step": 18532 + }, + { + "epoch": 0.68, + "learning_rate": 1.3231405866627604e-07, + "loss": 0.6869, + "step": 18533 + }, + { + "epoch": 0.68, + "learning_rate": 1.3231039660160397e-07, + "loss": 0.6921, + "step": 18534 + }, + { + "epoch": 0.68, + "learning_rate": 1.3230673453693192e-07, + "loss": 0.6804, + "step": 18535 + }, + { + "epoch": 0.68, + "learning_rate": 1.3230307247225984e-07, + "loss": 0.6849, + "step": 18536 + }, + { + "epoch": 0.68, + "learning_rate": 1.322994104075878e-07, + "loss": 0.682, + "step": 18537 + }, + { + "epoch": 0.68, + "learning_rate": 1.3229574834291572e-07, + "loss": 0.7008, + "step": 18538 + }, + { + "epoch": 0.68, + "learning_rate": 1.3229208627824365e-07, + "loss": 0.6942, + "step": 18539 + }, + { + "epoch": 0.68, + "learning_rate": 1.322884242135716e-07, + "loss": 0.6973, + "step": 18540 + }, + { + "epoch": 0.68, + "learning_rate": 1.3228476214889953e-07, + "loss": 0.6779, + "step": 18541 + }, + { + "epoch": 0.68, + "learning_rate": 1.3228110008422748e-07, + "loss": 0.6792, + "step": 18542 + }, + { + "epoch": 0.68, + "learning_rate": 1.322774380195554e-07, + "loss": 0.6845, + "step": 18543 + }, + { + "epoch": 0.68, + "learning_rate": 1.3227377595488336e-07, + "loss": 0.6926, + "step": 18544 + }, + { + "epoch": 0.68, + "learning_rate": 1.322701138902113e-07, + "loss": 0.6759, + "step": 18545 + }, + { + "epoch": 0.68, + "learning_rate": 1.3226645182553924e-07, + "loss": 0.7064, + "step": 18546 + }, + { + "epoch": 0.68, + "learning_rate": 1.3226278976086717e-07, + "loss": 0.6858, + "step": 18547 + }, + { + "epoch": 0.68, + "learning_rate": 1.3225912769619512e-07, + "loss": 0.698, + "step": 18548 + }, + { + "epoch": 0.68, + "learning_rate": 1.3225546563152305e-07, + "loss": 0.6923, + "step": 18549 + }, + { + "epoch": 0.68, + "learning_rate": 1.32251803566851e-07, + "loss": 0.6924, + "step": 18550 + }, + { + "epoch": 0.68, + "learning_rate": 1.322481415021789e-07, + "loss": 0.6899, + "step": 18551 + }, + { + "epoch": 0.68, + "learning_rate": 1.3224447943750686e-07, + "loss": 0.7115, + "step": 18552 + }, + { + "epoch": 0.68, + "learning_rate": 1.3224081737283478e-07, + "loss": 0.6942, + "step": 18553 + }, + { + "epoch": 0.68, + "learning_rate": 1.3223715530816274e-07, + "loss": 0.6871, + "step": 18554 + }, + { + "epoch": 0.68, + "learning_rate": 1.3223349324349066e-07, + "loss": 0.6862, + "step": 18555 + }, + { + "epoch": 0.68, + "learning_rate": 1.3222983117881862e-07, + "loss": 0.6997, + "step": 18556 + }, + { + "epoch": 0.68, + "learning_rate": 1.3222616911414654e-07, + "loss": 0.7, + "step": 18557 + }, + { + "epoch": 0.68, + "learning_rate": 1.322225070494745e-07, + "loss": 0.6871, + "step": 18558 + }, + { + "epoch": 0.68, + "learning_rate": 1.3221884498480242e-07, + "loss": 0.6929, + "step": 18559 + }, + { + "epoch": 0.68, + "learning_rate": 1.3221518292013038e-07, + "loss": 0.7067, + "step": 18560 + }, + { + "epoch": 0.68, + "learning_rate": 1.322115208554583e-07, + "loss": 0.6844, + "step": 18561 + }, + { + "epoch": 0.68, + "learning_rate": 1.3220785879078623e-07, + "loss": 0.6912, + "step": 18562 + }, + { + "epoch": 0.68, + "learning_rate": 1.3220419672611418e-07, + "loss": 0.6972, + "step": 18563 + }, + { + "epoch": 0.68, + "learning_rate": 1.322005346614421e-07, + "loss": 0.6958, + "step": 18564 + }, + { + "epoch": 0.68, + "learning_rate": 1.3219687259677006e-07, + "loss": 0.7007, + "step": 18565 + }, + { + "epoch": 0.68, + "learning_rate": 1.32193210532098e-07, + "loss": 0.6786, + "step": 18566 + }, + { + "epoch": 0.68, + "learning_rate": 1.3218954846742594e-07, + "loss": 0.6858, + "step": 18567 + }, + { + "epoch": 0.68, + "learning_rate": 1.3218588640275387e-07, + "loss": 0.6852, + "step": 18568 + }, + { + "epoch": 0.68, + "learning_rate": 1.321822243380818e-07, + "loss": 0.6907, + "step": 18569 + }, + { + "epoch": 0.68, + "learning_rate": 1.3217856227340972e-07, + "loss": 0.6773, + "step": 18570 + }, + { + "epoch": 0.68, + "learning_rate": 1.3217490020873768e-07, + "loss": 0.6953, + "step": 18571 + }, + { + "epoch": 0.68, + "learning_rate": 1.321712381440656e-07, + "loss": 0.6883, + "step": 18572 + }, + { + "epoch": 0.68, + "learning_rate": 1.3216757607939356e-07, + "loss": 0.6933, + "step": 18573 + }, + { + "epoch": 0.68, + "learning_rate": 1.3216391401472148e-07, + "loss": 0.6832, + "step": 18574 + }, + { + "epoch": 0.68, + "learning_rate": 1.3216025195004944e-07, + "loss": 0.6789, + "step": 18575 + }, + { + "epoch": 0.68, + "learning_rate": 1.3215658988537736e-07, + "loss": 0.6976, + "step": 18576 + }, + { + "epoch": 0.68, + "learning_rate": 1.3215292782070532e-07, + "loss": 0.6907, + "step": 18577 + }, + { + "epoch": 0.68, + "learning_rate": 1.3214926575603324e-07, + "loss": 0.7013, + "step": 18578 + }, + { + "epoch": 0.68, + "learning_rate": 1.321456036913612e-07, + "loss": 0.694, + "step": 18579 + }, + { + "epoch": 0.68, + "learning_rate": 1.3214194162668912e-07, + "loss": 0.7019, + "step": 18580 + }, + { + "epoch": 0.68, + "learning_rate": 1.3213827956201708e-07, + "loss": 0.6981, + "step": 18581 + }, + { + "epoch": 0.68, + "learning_rate": 1.32134617497345e-07, + "loss": 0.6982, + "step": 18582 + }, + { + "epoch": 0.68, + "learning_rate": 1.3213095543267293e-07, + "loss": 0.7006, + "step": 18583 + }, + { + "epoch": 0.68, + "learning_rate": 1.3212729336800088e-07, + "loss": 0.6939, + "step": 18584 + }, + { + "epoch": 0.68, + "learning_rate": 1.321236313033288e-07, + "loss": 0.6866, + "step": 18585 + }, + { + "epoch": 0.68, + "learning_rate": 1.3211996923865676e-07, + "loss": 0.6886, + "step": 18586 + }, + { + "epoch": 0.68, + "learning_rate": 1.321163071739847e-07, + "loss": 0.6885, + "step": 18587 + }, + { + "epoch": 0.68, + "learning_rate": 1.3211264510931262e-07, + "loss": 0.6817, + "step": 18588 + }, + { + "epoch": 0.68, + "learning_rate": 1.3210898304464054e-07, + "loss": 0.6911, + "step": 18589 + }, + { + "epoch": 0.68, + "learning_rate": 1.321053209799685e-07, + "loss": 0.6916, + "step": 18590 + }, + { + "epoch": 0.68, + "learning_rate": 1.3210165891529642e-07, + "loss": 0.6928, + "step": 18591 + }, + { + "epoch": 0.68, + "learning_rate": 1.3209799685062438e-07, + "loss": 0.6805, + "step": 18592 + }, + { + "epoch": 0.68, + "learning_rate": 1.320943347859523e-07, + "loss": 0.7045, + "step": 18593 + }, + { + "epoch": 0.68, + "learning_rate": 1.3209067272128025e-07, + "loss": 0.6815, + "step": 18594 + }, + { + "epoch": 0.68, + "learning_rate": 1.3208701065660818e-07, + "loss": 0.6898, + "step": 18595 + }, + { + "epoch": 0.68, + "learning_rate": 1.3208334859193613e-07, + "loss": 0.6956, + "step": 18596 + }, + { + "epoch": 0.68, + "learning_rate": 1.3207968652726406e-07, + "loss": 0.6932, + "step": 18597 + }, + { + "epoch": 0.68, + "learning_rate": 1.3207602446259201e-07, + "loss": 0.7079, + "step": 18598 + }, + { + "epoch": 0.68, + "learning_rate": 1.3207236239791994e-07, + "loss": 0.7001, + "step": 18599 + }, + { + "epoch": 0.68, + "learning_rate": 1.320687003332479e-07, + "loss": 0.6973, + "step": 18600 + }, + { + "epoch": 0.68, + "learning_rate": 1.3206503826857582e-07, + "loss": 0.6936, + "step": 18601 + }, + { + "epoch": 0.68, + "learning_rate": 1.3206137620390377e-07, + "loss": 0.7066, + "step": 18602 + }, + { + "epoch": 0.68, + "learning_rate": 1.320577141392317e-07, + "loss": 0.6966, + "step": 18603 + }, + { + "epoch": 0.68, + "learning_rate": 1.3205405207455963e-07, + "loss": 0.6772, + "step": 18604 + }, + { + "epoch": 0.68, + "learning_rate": 1.3205039000988755e-07, + "loss": 0.6819, + "step": 18605 + }, + { + "epoch": 0.68, + "learning_rate": 1.320467279452155e-07, + "loss": 0.6783, + "step": 18606 + }, + { + "epoch": 0.68, + "learning_rate": 1.3204306588054343e-07, + "loss": 0.6907, + "step": 18607 + }, + { + "epoch": 0.68, + "learning_rate": 1.320394038158714e-07, + "loss": 0.689, + "step": 18608 + }, + { + "epoch": 0.68, + "learning_rate": 1.3203574175119931e-07, + "loss": 0.6764, + "step": 18609 + }, + { + "epoch": 0.68, + "learning_rate": 1.3203207968652724e-07, + "loss": 0.7033, + "step": 18610 + }, + { + "epoch": 0.68, + "learning_rate": 1.320284176218552e-07, + "loss": 0.6955, + "step": 18611 + }, + { + "epoch": 0.68, + "learning_rate": 1.3202475555718312e-07, + "loss": 0.714, + "step": 18612 + }, + { + "epoch": 0.68, + "learning_rate": 1.3202109349251107e-07, + "loss": 0.6859, + "step": 18613 + }, + { + "epoch": 0.68, + "learning_rate": 1.32017431427839e-07, + "loss": 0.7055, + "step": 18614 + }, + { + "epoch": 0.68, + "learning_rate": 1.3201376936316695e-07, + "loss": 0.6973, + "step": 18615 + }, + { + "epoch": 0.68, + "learning_rate": 1.3201010729849488e-07, + "loss": 0.7053, + "step": 18616 + }, + { + "epoch": 0.68, + "learning_rate": 1.3200644523382283e-07, + "loss": 0.7, + "step": 18617 + }, + { + "epoch": 0.68, + "learning_rate": 1.3200278316915076e-07, + "loss": 0.6979, + "step": 18618 + }, + { + "epoch": 0.68, + "learning_rate": 1.3199912110447871e-07, + "loss": 0.6855, + "step": 18619 + }, + { + "epoch": 0.68, + "learning_rate": 1.3199545903980664e-07, + "loss": 0.6723, + "step": 18620 + }, + { + "epoch": 0.68, + "learning_rate": 1.319917969751346e-07, + "loss": 0.6831, + "step": 18621 + }, + { + "epoch": 0.68, + "learning_rate": 1.3198813491046252e-07, + "loss": 0.6804, + "step": 18622 + }, + { + "epoch": 0.68, + "learning_rate": 1.3198447284579045e-07, + "loss": 0.6852, + "step": 18623 + }, + { + "epoch": 0.68, + "learning_rate": 1.3198081078111837e-07, + "loss": 0.6836, + "step": 18624 + }, + { + "epoch": 0.68, + "learning_rate": 1.3197714871644633e-07, + "loss": 0.6951, + "step": 18625 + }, + { + "epoch": 0.68, + "learning_rate": 1.3197348665177425e-07, + "loss": 0.6998, + "step": 18626 + }, + { + "epoch": 0.68, + "learning_rate": 1.319698245871022e-07, + "loss": 0.7113, + "step": 18627 + }, + { + "epoch": 0.68, + "learning_rate": 1.3196616252243013e-07, + "loss": 0.7025, + "step": 18628 + }, + { + "epoch": 0.68, + "learning_rate": 1.3196250045775809e-07, + "loss": 0.6883, + "step": 18629 + }, + { + "epoch": 0.68, + "learning_rate": 1.31958838393086e-07, + "loss": 0.6898, + "step": 18630 + }, + { + "epoch": 0.68, + "learning_rate": 1.3195517632841397e-07, + "loss": 0.698, + "step": 18631 + }, + { + "epoch": 0.68, + "learning_rate": 1.319515142637419e-07, + "loss": 0.6804, + "step": 18632 + }, + { + "epoch": 0.68, + "learning_rate": 1.3194785219906982e-07, + "loss": 0.6941, + "step": 18633 + }, + { + "epoch": 0.68, + "learning_rate": 1.3194419013439777e-07, + "loss": 0.6778, + "step": 18634 + }, + { + "epoch": 0.68, + "learning_rate": 1.319405280697257e-07, + "loss": 0.6844, + "step": 18635 + }, + { + "epoch": 0.68, + "learning_rate": 1.3193686600505365e-07, + "loss": 0.6654, + "step": 18636 + }, + { + "epoch": 0.68, + "learning_rate": 1.3193320394038158e-07, + "loss": 0.6897, + "step": 18637 + }, + { + "epoch": 0.68, + "learning_rate": 1.3192954187570953e-07, + "loss": 0.6969, + "step": 18638 + }, + { + "epoch": 0.68, + "learning_rate": 1.3192587981103746e-07, + "loss": 0.6885, + "step": 18639 + }, + { + "epoch": 0.68, + "learning_rate": 1.319222177463654e-07, + "loss": 0.6958, + "step": 18640 + }, + { + "epoch": 0.68, + "learning_rate": 1.319185556816933e-07, + "loss": 0.6813, + "step": 18641 + }, + { + "epoch": 0.68, + "learning_rate": 1.3191489361702127e-07, + "loss": 0.6941, + "step": 18642 + }, + { + "epoch": 0.68, + "learning_rate": 1.319112315523492e-07, + "loss": 0.6806, + "step": 18643 + }, + { + "epoch": 0.68, + "learning_rate": 1.3190756948767715e-07, + "loss": 0.7062, + "step": 18644 + }, + { + "epoch": 0.68, + "learning_rate": 1.3190390742300507e-07, + "loss": 0.6877, + "step": 18645 + }, + { + "epoch": 0.68, + "learning_rate": 1.3190024535833303e-07, + "loss": 0.6844, + "step": 18646 + }, + { + "epoch": 0.68, + "learning_rate": 1.3189658329366095e-07, + "loss": 0.6893, + "step": 18647 + }, + { + "epoch": 0.68, + "learning_rate": 1.318929212289889e-07, + "loss": 0.6943, + "step": 18648 + }, + { + "epoch": 0.68, + "learning_rate": 1.3188925916431683e-07, + "loss": 0.6981, + "step": 18649 + }, + { + "epoch": 0.68, + "learning_rate": 1.3188559709964478e-07, + "loss": 0.6926, + "step": 18650 + }, + { + "epoch": 0.68, + "learning_rate": 1.318819350349727e-07, + "loss": 0.708, + "step": 18651 + }, + { + "epoch": 0.68, + "learning_rate": 1.3187827297030066e-07, + "loss": 0.694, + "step": 18652 + }, + { + "epoch": 0.68, + "learning_rate": 1.318746109056286e-07, + "loss": 0.6819, + "step": 18653 + }, + { + "epoch": 0.68, + "learning_rate": 1.3187094884095652e-07, + "loss": 0.7105, + "step": 18654 + }, + { + "epoch": 0.68, + "learning_rate": 1.3186728677628447e-07, + "loss": 0.6791, + "step": 18655 + }, + { + "epoch": 0.68, + "learning_rate": 1.318636247116124e-07, + "loss": 0.6849, + "step": 18656 + }, + { + "epoch": 0.68, + "learning_rate": 1.3185996264694035e-07, + "loss": 0.6999, + "step": 18657 + }, + { + "epoch": 0.68, + "learning_rate": 1.3185630058226828e-07, + "loss": 0.6762, + "step": 18658 + }, + { + "epoch": 0.68, + "learning_rate": 1.318526385175962e-07, + "loss": 0.667, + "step": 18659 + }, + { + "epoch": 0.68, + "learning_rate": 1.3184897645292413e-07, + "loss": 0.6889, + "step": 18660 + }, + { + "epoch": 0.68, + "learning_rate": 1.3184531438825208e-07, + "loss": 0.6891, + "step": 18661 + }, + { + "epoch": 0.68, + "learning_rate": 1.3184165232358e-07, + "loss": 0.6997, + "step": 18662 + }, + { + "epoch": 0.68, + "learning_rate": 1.3183799025890796e-07, + "loss": 0.705, + "step": 18663 + }, + { + "epoch": 0.68, + "learning_rate": 1.318343281942359e-07, + "loss": 0.6646, + "step": 18664 + }, + { + "epoch": 0.68, + "learning_rate": 1.3183066612956384e-07, + "loss": 0.6864, + "step": 18665 + }, + { + "epoch": 0.68, + "learning_rate": 1.3182700406489177e-07, + "loss": 0.6897, + "step": 18666 + }, + { + "epoch": 0.68, + "learning_rate": 1.3182334200021972e-07, + "loss": 0.6986, + "step": 18667 + }, + { + "epoch": 0.68, + "learning_rate": 1.3181967993554765e-07, + "loss": 0.6795, + "step": 18668 + }, + { + "epoch": 0.68, + "learning_rate": 1.318160178708756e-07, + "loss": 0.6831, + "step": 18669 + }, + { + "epoch": 0.68, + "learning_rate": 1.3181235580620353e-07, + "loss": 0.6778, + "step": 18670 + }, + { + "epoch": 0.68, + "learning_rate": 1.3180869374153148e-07, + "loss": 0.7028, + "step": 18671 + }, + { + "epoch": 0.68, + "learning_rate": 1.318050316768594e-07, + "loss": 0.6865, + "step": 18672 + }, + { + "epoch": 0.68, + "learning_rate": 1.3180136961218736e-07, + "loss": 0.6959, + "step": 18673 + }, + { + "epoch": 0.68, + "learning_rate": 1.317977075475153e-07, + "loss": 0.7015, + "step": 18674 + }, + { + "epoch": 0.68, + "learning_rate": 1.3179404548284322e-07, + "loss": 0.6876, + "step": 18675 + }, + { + "epoch": 0.68, + "learning_rate": 1.3179038341817117e-07, + "loss": 0.6777, + "step": 18676 + }, + { + "epoch": 0.68, + "learning_rate": 1.317867213534991e-07, + "loss": 0.6789, + "step": 18677 + }, + { + "epoch": 0.68, + "learning_rate": 1.3178305928882702e-07, + "loss": 0.6965, + "step": 18678 + }, + { + "epoch": 0.68, + "learning_rate": 1.3177939722415498e-07, + "loss": 0.7004, + "step": 18679 + }, + { + "epoch": 0.68, + "learning_rate": 1.317757351594829e-07, + "loss": 0.6994, + "step": 18680 + }, + { + "epoch": 0.68, + "learning_rate": 1.3177207309481083e-07, + "loss": 0.7012, + "step": 18681 + }, + { + "epoch": 0.68, + "learning_rate": 1.3176841103013878e-07, + "loss": 0.7014, + "step": 18682 + }, + { + "epoch": 0.68, + "learning_rate": 1.317647489654667e-07, + "loss": 0.7094, + "step": 18683 + }, + { + "epoch": 0.68, + "learning_rate": 1.3176108690079466e-07, + "loss": 0.6972, + "step": 18684 + }, + { + "epoch": 0.68, + "learning_rate": 1.317574248361226e-07, + "loss": 0.6964, + "step": 18685 + }, + { + "epoch": 0.68, + "learning_rate": 1.3175376277145054e-07, + "loss": 0.6873, + "step": 18686 + }, + { + "epoch": 0.68, + "learning_rate": 1.3175010070677847e-07, + "loss": 0.7044, + "step": 18687 + }, + { + "epoch": 0.68, + "learning_rate": 1.3174643864210642e-07, + "loss": 0.6993, + "step": 18688 + }, + { + "epoch": 0.68, + "learning_rate": 1.3174277657743435e-07, + "loss": 0.6867, + "step": 18689 + }, + { + "epoch": 0.68, + "learning_rate": 1.317391145127623e-07, + "loss": 0.7066, + "step": 18690 + }, + { + "epoch": 0.68, + "learning_rate": 1.3173545244809023e-07, + "loss": 0.7036, + "step": 18691 + }, + { + "epoch": 0.68, + "learning_rate": 1.3173179038341818e-07, + "loss": 0.7152, + "step": 18692 + }, + { + "epoch": 0.68, + "learning_rate": 1.317281283187461e-07, + "loss": 0.6874, + "step": 18693 + }, + { + "epoch": 0.68, + "learning_rate": 1.3172446625407406e-07, + "loss": 0.701, + "step": 18694 + }, + { + "epoch": 0.68, + "learning_rate": 1.3172080418940196e-07, + "loss": 0.683, + "step": 18695 + }, + { + "epoch": 0.68, + "learning_rate": 1.3171714212472992e-07, + "loss": 0.7042, + "step": 18696 + }, + { + "epoch": 0.68, + "learning_rate": 1.3171348006005784e-07, + "loss": 0.6866, + "step": 18697 + }, + { + "epoch": 0.68, + "learning_rate": 1.317098179953858e-07, + "loss": 0.6949, + "step": 18698 + }, + { + "epoch": 0.68, + "learning_rate": 1.3170615593071372e-07, + "loss": 0.6931, + "step": 18699 + }, + { + "epoch": 0.68, + "learning_rate": 1.3170249386604168e-07, + "loss": 0.7057, + "step": 18700 + }, + { + "epoch": 0.68, + "learning_rate": 1.316988318013696e-07, + "loss": 0.6909, + "step": 18701 + }, + { + "epoch": 0.68, + "learning_rate": 1.3169516973669756e-07, + "loss": 0.6965, + "step": 18702 + }, + { + "epoch": 0.68, + "learning_rate": 1.3169150767202548e-07, + "loss": 0.6864, + "step": 18703 + }, + { + "epoch": 0.68, + "learning_rate": 1.316878456073534e-07, + "loss": 0.6913, + "step": 18704 + }, + { + "epoch": 0.68, + "learning_rate": 1.3168418354268136e-07, + "loss": 0.689, + "step": 18705 + }, + { + "epoch": 0.68, + "learning_rate": 1.316805214780093e-07, + "loss": 0.7089, + "step": 18706 + }, + { + "epoch": 0.68, + "learning_rate": 1.3167685941333724e-07, + "loss": 0.6983, + "step": 18707 + }, + { + "epoch": 0.68, + "learning_rate": 1.3167319734866517e-07, + "loss": 0.6821, + "step": 18708 + }, + { + "epoch": 0.68, + "learning_rate": 1.3166953528399312e-07, + "loss": 0.6995, + "step": 18709 + }, + { + "epoch": 0.68, + "learning_rate": 1.3166587321932105e-07, + "loss": 0.6885, + "step": 18710 + }, + { + "epoch": 0.68, + "learning_rate": 1.31662211154649e-07, + "loss": 0.7033, + "step": 18711 + }, + { + "epoch": 0.68, + "learning_rate": 1.3165854908997693e-07, + "loss": 0.6949, + "step": 18712 + }, + { + "epoch": 0.68, + "learning_rate": 1.3165488702530485e-07, + "loss": 0.6883, + "step": 18713 + }, + { + "epoch": 0.68, + "learning_rate": 1.3165122496063278e-07, + "loss": 0.6739, + "step": 18714 + }, + { + "epoch": 0.68, + "learning_rate": 1.3164756289596073e-07, + "loss": 0.6847, + "step": 18715 + }, + { + "epoch": 0.68, + "learning_rate": 1.3164390083128866e-07, + "loss": 0.7018, + "step": 18716 + }, + { + "epoch": 0.68, + "learning_rate": 1.3164023876661661e-07, + "loss": 0.6806, + "step": 18717 + }, + { + "epoch": 0.68, + "learning_rate": 1.3163657670194454e-07, + "loss": 0.7016, + "step": 18718 + }, + { + "epoch": 0.68, + "learning_rate": 1.316329146372725e-07, + "loss": 0.6905, + "step": 18719 + }, + { + "epoch": 0.68, + "learning_rate": 1.3162925257260042e-07, + "loss": 0.6985, + "step": 18720 + }, + { + "epoch": 0.68, + "learning_rate": 1.3162559050792837e-07, + "loss": 0.6867, + "step": 18721 + }, + { + "epoch": 0.68, + "learning_rate": 1.316219284432563e-07, + "loss": 0.6935, + "step": 18722 + }, + { + "epoch": 0.69, + "learning_rate": 1.3161826637858425e-07, + "loss": 0.7109, + "step": 18723 + }, + { + "epoch": 0.69, + "learning_rate": 1.3161460431391218e-07, + "loss": 0.6799, + "step": 18724 + }, + { + "epoch": 0.69, + "learning_rate": 1.316109422492401e-07, + "loss": 0.705, + "step": 18725 + }, + { + "epoch": 0.69, + "learning_rate": 1.3160728018456806e-07, + "loss": 0.6922, + "step": 18726 + }, + { + "epoch": 0.69, + "learning_rate": 1.31603618119896e-07, + "loss": 0.6989, + "step": 18727 + }, + { + "epoch": 0.69, + "learning_rate": 1.3159995605522394e-07, + "loss": 0.6996, + "step": 18728 + }, + { + "epoch": 0.69, + "learning_rate": 1.3159629399055187e-07, + "loss": 0.6832, + "step": 18729 + }, + { + "epoch": 0.69, + "learning_rate": 1.3159263192587982e-07, + "loss": 0.6895, + "step": 18730 + }, + { + "epoch": 0.69, + "learning_rate": 1.3158896986120772e-07, + "loss": 0.6898, + "step": 18731 + }, + { + "epoch": 0.69, + "learning_rate": 1.3158530779653567e-07, + "loss": 0.6948, + "step": 18732 + }, + { + "epoch": 0.69, + "learning_rate": 1.315816457318636e-07, + "loss": 0.6971, + "step": 18733 + }, + { + "epoch": 0.69, + "learning_rate": 1.3157798366719155e-07, + "loss": 0.6915, + "step": 18734 + }, + { + "epoch": 0.69, + "learning_rate": 1.3157432160251948e-07, + "loss": 0.6868, + "step": 18735 + }, + { + "epoch": 0.69, + "learning_rate": 1.3157065953784743e-07, + "loss": 0.6889, + "step": 18736 + }, + { + "epoch": 0.69, + "learning_rate": 1.3156699747317536e-07, + "loss": 0.6868, + "step": 18737 + }, + { + "epoch": 0.69, + "learning_rate": 1.3156333540850331e-07, + "loss": 0.6892, + "step": 18738 + }, + { + "epoch": 0.69, + "learning_rate": 1.3155967334383124e-07, + "loss": 0.7054, + "step": 18739 + }, + { + "epoch": 0.69, + "learning_rate": 1.315560112791592e-07, + "loss": 0.6961, + "step": 18740 + }, + { + "epoch": 0.69, + "learning_rate": 1.3155234921448712e-07, + "loss": 0.6986, + "step": 18741 + }, + { + "epoch": 0.69, + "learning_rate": 1.3154868714981507e-07, + "loss": 0.7222, + "step": 18742 + }, + { + "epoch": 0.69, + "learning_rate": 1.31545025085143e-07, + "loss": 0.7036, + "step": 18743 + }, + { + "epoch": 0.69, + "learning_rate": 1.3154136302047095e-07, + "loss": 0.6877, + "step": 18744 + }, + { + "epoch": 0.69, + "learning_rate": 1.3153770095579888e-07, + "loss": 0.6876, + "step": 18745 + }, + { + "epoch": 0.69, + "learning_rate": 1.315340388911268e-07, + "loss": 0.6947, + "step": 18746 + }, + { + "epoch": 0.69, + "learning_rate": 1.3153037682645476e-07, + "loss": 0.6779, + "step": 18747 + }, + { + "epoch": 0.69, + "learning_rate": 1.3152671476178269e-07, + "loss": 0.6944, + "step": 18748 + }, + { + "epoch": 0.69, + "learning_rate": 1.315230526971106e-07, + "loss": 0.692, + "step": 18749 + }, + { + "epoch": 0.69, + "learning_rate": 1.3151939063243857e-07, + "loss": 0.6882, + "step": 18750 + }, + { + "epoch": 0.69, + "learning_rate": 1.315157285677665e-07, + "loss": 0.6956, + "step": 18751 + }, + { + "epoch": 0.69, + "learning_rate": 1.3151206650309442e-07, + "loss": 0.7009, + "step": 18752 + }, + { + "epoch": 0.69, + "learning_rate": 1.3150840443842237e-07, + "loss": 0.6861, + "step": 18753 + }, + { + "epoch": 0.69, + "learning_rate": 1.315047423737503e-07, + "loss": 0.6738, + "step": 18754 + }, + { + "epoch": 0.69, + "learning_rate": 1.3150108030907825e-07, + "loss": 0.6871, + "step": 18755 + }, + { + "epoch": 0.69, + "learning_rate": 1.3149741824440618e-07, + "loss": 0.6911, + "step": 18756 + }, + { + "epoch": 0.69, + "learning_rate": 1.3149375617973413e-07, + "loss": 0.7006, + "step": 18757 + }, + { + "epoch": 0.69, + "learning_rate": 1.3149009411506206e-07, + "loss": 0.6917, + "step": 18758 + }, + { + "epoch": 0.69, + "learning_rate": 1.3148643205039e-07, + "loss": 0.6987, + "step": 18759 + }, + { + "epoch": 0.69, + "learning_rate": 1.3148276998571794e-07, + "loss": 0.6961, + "step": 18760 + }, + { + "epoch": 0.69, + "learning_rate": 1.314791079210459e-07, + "loss": 0.6861, + "step": 18761 + }, + { + "epoch": 0.69, + "learning_rate": 1.3147544585637382e-07, + "loss": 0.6883, + "step": 18762 + }, + { + "epoch": 0.69, + "learning_rate": 1.3147178379170177e-07, + "loss": 0.6942, + "step": 18763 + }, + { + "epoch": 0.69, + "learning_rate": 1.314681217270297e-07, + "loss": 0.6945, + "step": 18764 + }, + { + "epoch": 0.69, + "learning_rate": 1.3146445966235765e-07, + "loss": 0.6958, + "step": 18765 + }, + { + "epoch": 0.69, + "learning_rate": 1.3146079759768558e-07, + "loss": 0.6946, + "step": 18766 + }, + { + "epoch": 0.69, + "learning_rate": 1.314571355330135e-07, + "loss": 0.6969, + "step": 18767 + }, + { + "epoch": 0.69, + "learning_rate": 1.3145347346834143e-07, + "loss": 0.6938, + "step": 18768 + }, + { + "epoch": 0.69, + "learning_rate": 1.3144981140366939e-07, + "loss": 0.7039, + "step": 18769 + }, + { + "epoch": 0.69, + "learning_rate": 1.314461493389973e-07, + "loss": 0.6961, + "step": 18770 + }, + { + "epoch": 0.69, + "learning_rate": 1.3144248727432526e-07, + "loss": 0.7002, + "step": 18771 + }, + { + "epoch": 0.69, + "learning_rate": 1.314388252096532e-07, + "loss": 0.6861, + "step": 18772 + }, + { + "epoch": 0.69, + "learning_rate": 1.3143516314498114e-07, + "loss": 0.69, + "step": 18773 + }, + { + "epoch": 0.69, + "learning_rate": 1.3143150108030907e-07, + "loss": 0.6879, + "step": 18774 + }, + { + "epoch": 0.69, + "learning_rate": 1.31427839015637e-07, + "loss": 0.7016, + "step": 18775 + }, + { + "epoch": 0.69, + "learning_rate": 1.3142417695096495e-07, + "loss": 0.6894, + "step": 18776 + }, + { + "epoch": 0.69, + "learning_rate": 1.3142051488629288e-07, + "loss": 0.6954, + "step": 18777 + }, + { + "epoch": 0.69, + "learning_rate": 1.3141685282162083e-07, + "loss": 0.6846, + "step": 18778 + }, + { + "epoch": 0.69, + "learning_rate": 1.3141319075694876e-07, + "loss": 0.6825, + "step": 18779 + }, + { + "epoch": 0.69, + "learning_rate": 1.314095286922767e-07, + "loss": 0.7036, + "step": 18780 + }, + { + "epoch": 0.69, + "learning_rate": 1.3140586662760464e-07, + "loss": 0.6957, + "step": 18781 + }, + { + "epoch": 0.69, + "learning_rate": 1.314022045629326e-07, + "loss": 0.6901, + "step": 18782 + }, + { + "epoch": 0.69, + "learning_rate": 1.3139854249826052e-07, + "loss": 0.6918, + "step": 18783 + }, + { + "epoch": 0.69, + "learning_rate": 1.3139488043358847e-07, + "loss": 0.6948, + "step": 18784 + }, + { + "epoch": 0.69, + "learning_rate": 1.3139121836891637e-07, + "loss": 0.7015, + "step": 18785 + }, + { + "epoch": 0.69, + "learning_rate": 1.3138755630424432e-07, + "loss": 0.6896, + "step": 18786 + }, + { + "epoch": 0.69, + "learning_rate": 1.3138389423957225e-07, + "loss": 0.6831, + "step": 18787 + }, + { + "epoch": 0.69, + "learning_rate": 1.313802321749002e-07, + "loss": 0.6842, + "step": 18788 + }, + { + "epoch": 0.69, + "learning_rate": 1.3137657011022813e-07, + "loss": 0.691, + "step": 18789 + }, + { + "epoch": 0.69, + "learning_rate": 1.3137290804555608e-07, + "loss": 0.6985, + "step": 18790 + }, + { + "epoch": 0.69, + "learning_rate": 1.31369245980884e-07, + "loss": 0.6944, + "step": 18791 + }, + { + "epoch": 0.69, + "learning_rate": 1.3136558391621196e-07, + "loss": 0.6983, + "step": 18792 + }, + { + "epoch": 0.69, + "learning_rate": 1.313619218515399e-07, + "loss": 0.6706, + "step": 18793 + }, + { + "epoch": 0.69, + "learning_rate": 1.3135825978686784e-07, + "loss": 0.6929, + "step": 18794 + }, + { + "epoch": 0.69, + "learning_rate": 1.3135459772219577e-07, + "loss": 0.6917, + "step": 18795 + }, + { + "epoch": 0.69, + "learning_rate": 1.313509356575237e-07, + "loss": 0.6943, + "step": 18796 + }, + { + "epoch": 0.69, + "learning_rate": 1.3134727359285165e-07, + "loss": 0.6818, + "step": 18797 + }, + { + "epoch": 0.69, + "learning_rate": 1.3134361152817958e-07, + "loss": 0.6924, + "step": 18798 + }, + { + "epoch": 0.69, + "learning_rate": 1.3133994946350753e-07, + "loss": 0.691, + "step": 18799 + }, + { + "epoch": 0.69, + "learning_rate": 1.3133628739883546e-07, + "loss": 0.6906, + "step": 18800 + }, + { + "epoch": 0.69, + "learning_rate": 1.313326253341634e-07, + "loss": 0.7022, + "step": 18801 + }, + { + "epoch": 0.69, + "learning_rate": 1.3132896326949134e-07, + "loss": 0.7048, + "step": 18802 + }, + { + "epoch": 0.69, + "learning_rate": 1.3132530120481926e-07, + "loss": 0.6954, + "step": 18803 + }, + { + "epoch": 0.69, + "learning_rate": 1.313216391401472e-07, + "loss": 0.7028, + "step": 18804 + }, + { + "epoch": 0.69, + "learning_rate": 1.3131797707547514e-07, + "loss": 0.688, + "step": 18805 + }, + { + "epoch": 0.69, + "learning_rate": 1.3131431501080307e-07, + "loss": 0.6959, + "step": 18806 + }, + { + "epoch": 0.69, + "learning_rate": 1.3131065294613102e-07, + "loss": 0.6932, + "step": 18807 + }, + { + "epoch": 0.69, + "learning_rate": 1.3130699088145895e-07, + "loss": 0.6811, + "step": 18808 + }, + { + "epoch": 0.69, + "learning_rate": 1.313033288167869e-07, + "loss": 0.6953, + "step": 18809 + }, + { + "epoch": 0.69, + "learning_rate": 1.3129966675211483e-07, + "loss": 0.6929, + "step": 18810 + }, + { + "epoch": 0.69, + "learning_rate": 1.3129600468744278e-07, + "loss": 0.7017, + "step": 18811 + }, + { + "epoch": 0.69, + "learning_rate": 1.312923426227707e-07, + "loss": 0.6934, + "step": 18812 + }, + { + "epoch": 0.69, + "learning_rate": 1.3128868055809866e-07, + "loss": 0.6876, + "step": 18813 + }, + { + "epoch": 0.69, + "learning_rate": 1.312850184934266e-07, + "loss": 0.6856, + "step": 18814 + }, + { + "epoch": 0.69, + "learning_rate": 1.3128135642875454e-07, + "loss": 0.695, + "step": 18815 + }, + { + "epoch": 0.69, + "learning_rate": 1.3127769436408247e-07, + "loss": 0.6786, + "step": 18816 + }, + { + "epoch": 0.69, + "learning_rate": 1.312740322994104e-07, + "loss": 0.6874, + "step": 18817 + }, + { + "epoch": 0.69, + "learning_rate": 1.3127037023473835e-07, + "loss": 0.6774, + "step": 18818 + }, + { + "epoch": 0.69, + "learning_rate": 1.3126670817006628e-07, + "loss": 0.6947, + "step": 18819 + }, + { + "epoch": 0.69, + "learning_rate": 1.3126304610539423e-07, + "loss": 0.7003, + "step": 18820 + }, + { + "epoch": 0.69, + "learning_rate": 1.3125938404072216e-07, + "loss": 0.6919, + "step": 18821 + }, + { + "epoch": 0.69, + "learning_rate": 1.3125572197605008e-07, + "loss": 0.6967, + "step": 18822 + }, + { + "epoch": 0.69, + "learning_rate": 1.31252059911378e-07, + "loss": 0.6906, + "step": 18823 + }, + { + "epoch": 0.69, + "learning_rate": 1.3124839784670596e-07, + "loss": 0.6793, + "step": 18824 + }, + { + "epoch": 0.69, + "learning_rate": 1.312447357820339e-07, + "loss": 0.6933, + "step": 18825 + }, + { + "epoch": 0.69, + "learning_rate": 1.3124107371736184e-07, + "loss": 0.6936, + "step": 18826 + }, + { + "epoch": 0.69, + "learning_rate": 1.3123741165268977e-07, + "loss": 0.6861, + "step": 18827 + }, + { + "epoch": 0.69, + "learning_rate": 1.3123374958801772e-07, + "loss": 0.6986, + "step": 18828 + }, + { + "epoch": 0.69, + "learning_rate": 1.3123008752334565e-07, + "loss": 0.697, + "step": 18829 + }, + { + "epoch": 0.69, + "learning_rate": 1.312264254586736e-07, + "loss": 0.6818, + "step": 18830 + }, + { + "epoch": 0.69, + "learning_rate": 1.3122276339400153e-07, + "loss": 0.713, + "step": 18831 + }, + { + "epoch": 0.69, + "learning_rate": 1.3121910132932948e-07, + "loss": 0.6891, + "step": 18832 + }, + { + "epoch": 0.69, + "learning_rate": 1.312154392646574e-07, + "loss": 0.7037, + "step": 18833 + }, + { + "epoch": 0.69, + "learning_rate": 1.3121177719998536e-07, + "loss": 0.6919, + "step": 18834 + }, + { + "epoch": 0.69, + "learning_rate": 1.312081151353133e-07, + "loss": 0.6856, + "step": 18835 + }, + { + "epoch": 0.69, + "learning_rate": 1.3120445307064124e-07, + "loss": 0.688, + "step": 18836 + }, + { + "epoch": 0.69, + "learning_rate": 1.3120079100596917e-07, + "loss": 0.6994, + "step": 18837 + }, + { + "epoch": 0.69, + "learning_rate": 1.3119712894129712e-07, + "loss": 0.6876, + "step": 18838 + }, + { + "epoch": 0.69, + "learning_rate": 1.3119346687662502e-07, + "loss": 0.6996, + "step": 18839 + }, + { + "epoch": 0.69, + "learning_rate": 1.3118980481195297e-07, + "loss": 0.6955, + "step": 18840 + }, + { + "epoch": 0.69, + "learning_rate": 1.311861427472809e-07, + "loss": 0.6996, + "step": 18841 + }, + { + "epoch": 0.69, + "learning_rate": 1.3118248068260885e-07, + "loss": 0.6974, + "step": 18842 + }, + { + "epoch": 0.69, + "learning_rate": 1.3117881861793678e-07, + "loss": 0.6906, + "step": 18843 + }, + { + "epoch": 0.69, + "learning_rate": 1.311751565532647e-07, + "loss": 0.6959, + "step": 18844 + }, + { + "epoch": 0.69, + "learning_rate": 1.3117149448859266e-07, + "loss": 0.6866, + "step": 18845 + }, + { + "epoch": 0.69, + "learning_rate": 1.311678324239206e-07, + "loss": 0.6834, + "step": 18846 + }, + { + "epoch": 0.69, + "learning_rate": 1.3116417035924854e-07, + "loss": 0.6957, + "step": 18847 + }, + { + "epoch": 0.69, + "learning_rate": 1.3116050829457647e-07, + "loss": 0.6869, + "step": 18848 + }, + { + "epoch": 0.69, + "learning_rate": 1.3115684622990442e-07, + "loss": 0.6944, + "step": 18849 + }, + { + "epoch": 0.69, + "learning_rate": 1.3115318416523235e-07, + "loss": 0.6838, + "step": 18850 + }, + { + "epoch": 0.69, + "learning_rate": 1.311495221005603e-07, + "loss": 0.7, + "step": 18851 + }, + { + "epoch": 0.69, + "learning_rate": 1.3114586003588823e-07, + "loss": 0.6687, + "step": 18852 + }, + { + "epoch": 0.69, + "learning_rate": 1.3114219797121618e-07, + "loss": 0.6957, + "step": 18853 + }, + { + "epoch": 0.69, + "learning_rate": 1.311385359065441e-07, + "loss": 0.6998, + "step": 18854 + }, + { + "epoch": 0.69, + "learning_rate": 1.3113487384187206e-07, + "loss": 0.693, + "step": 18855 + }, + { + "epoch": 0.69, + "learning_rate": 1.311312117772e-07, + "loss": 0.6958, + "step": 18856 + }, + { + "epoch": 0.69, + "learning_rate": 1.3112754971252791e-07, + "loss": 0.6873, + "step": 18857 + }, + { + "epoch": 0.69, + "learning_rate": 1.3112388764785584e-07, + "loss": 0.7015, + "step": 18858 + }, + { + "epoch": 0.69, + "learning_rate": 1.311202255831838e-07, + "loss": 0.6995, + "step": 18859 + }, + { + "epoch": 0.69, + "learning_rate": 1.3111656351851172e-07, + "loss": 0.7027, + "step": 18860 + }, + { + "epoch": 0.69, + "learning_rate": 1.3111290145383967e-07, + "loss": 0.6981, + "step": 18861 + }, + { + "epoch": 0.69, + "learning_rate": 1.311092393891676e-07, + "loss": 0.6648, + "step": 18862 + }, + { + "epoch": 0.69, + "learning_rate": 1.3110557732449555e-07, + "loss": 0.6958, + "step": 18863 + }, + { + "epoch": 0.69, + "learning_rate": 1.3110191525982348e-07, + "loss": 0.6778, + "step": 18864 + }, + { + "epoch": 0.69, + "learning_rate": 1.3109825319515143e-07, + "loss": 0.6846, + "step": 18865 + }, + { + "epoch": 0.69, + "learning_rate": 1.3109459113047936e-07, + "loss": 0.6882, + "step": 18866 + }, + { + "epoch": 0.69, + "learning_rate": 1.3109092906580729e-07, + "loss": 0.6995, + "step": 18867 + }, + { + "epoch": 0.69, + "learning_rate": 1.3108726700113524e-07, + "loss": 0.6985, + "step": 18868 + }, + { + "epoch": 0.69, + "learning_rate": 1.3108360493646317e-07, + "loss": 0.683, + "step": 18869 + }, + { + "epoch": 0.69, + "learning_rate": 1.3107994287179112e-07, + "loss": 0.6907, + "step": 18870 + }, + { + "epoch": 0.69, + "learning_rate": 1.3107628080711905e-07, + "loss": 0.6942, + "step": 18871 + }, + { + "epoch": 0.69, + "learning_rate": 1.31072618742447e-07, + "loss": 0.69, + "step": 18872 + }, + { + "epoch": 0.69, + "learning_rate": 1.3106895667777493e-07, + "loss": 0.7018, + "step": 18873 + }, + { + "epoch": 0.69, + "learning_rate": 1.3106529461310288e-07, + "loss": 0.6769, + "step": 18874 + }, + { + "epoch": 0.69, + "learning_rate": 1.3106163254843078e-07, + "loss": 0.683, + "step": 18875 + }, + { + "epoch": 0.69, + "learning_rate": 1.3105797048375873e-07, + "loss": 0.6931, + "step": 18876 + }, + { + "epoch": 0.69, + "learning_rate": 1.3105430841908666e-07, + "loss": 0.6911, + "step": 18877 + }, + { + "epoch": 0.69, + "learning_rate": 1.310506463544146e-07, + "loss": 0.6909, + "step": 18878 + }, + { + "epoch": 0.69, + "learning_rate": 1.3104698428974254e-07, + "loss": 0.6791, + "step": 18879 + }, + { + "epoch": 0.69, + "learning_rate": 1.310433222250705e-07, + "loss": 0.6849, + "step": 18880 + }, + { + "epoch": 0.69, + "learning_rate": 1.3103966016039842e-07, + "loss": 0.6966, + "step": 18881 + }, + { + "epoch": 0.69, + "learning_rate": 1.3103599809572637e-07, + "loss": 0.6946, + "step": 18882 + }, + { + "epoch": 0.69, + "learning_rate": 1.310323360310543e-07, + "loss": 0.6784, + "step": 18883 + }, + { + "epoch": 0.69, + "learning_rate": 1.3102867396638225e-07, + "loss": 0.6898, + "step": 18884 + }, + { + "epoch": 0.69, + "learning_rate": 1.3102501190171018e-07, + "loss": 0.6802, + "step": 18885 + }, + { + "epoch": 0.69, + "learning_rate": 1.3102134983703813e-07, + "loss": 0.689, + "step": 18886 + }, + { + "epoch": 0.69, + "learning_rate": 1.3101768777236606e-07, + "loss": 0.6928, + "step": 18887 + }, + { + "epoch": 0.69, + "learning_rate": 1.3101402570769399e-07, + "loss": 0.6908, + "step": 18888 + }, + { + "epoch": 0.69, + "learning_rate": 1.3101036364302194e-07, + "loss": 0.6902, + "step": 18889 + }, + { + "epoch": 0.69, + "learning_rate": 1.3100670157834986e-07, + "loss": 0.6998, + "step": 18890 + }, + { + "epoch": 0.69, + "learning_rate": 1.3100303951367782e-07, + "loss": 0.7054, + "step": 18891 + }, + { + "epoch": 0.69, + "learning_rate": 1.3099937744900574e-07, + "loss": 0.6742, + "step": 18892 + }, + { + "epoch": 0.69, + "learning_rate": 1.3099571538433367e-07, + "loss": 0.6893, + "step": 18893 + }, + { + "epoch": 0.69, + "learning_rate": 1.309920533196616e-07, + "loss": 0.6955, + "step": 18894 + }, + { + "epoch": 0.69, + "learning_rate": 1.3098839125498955e-07, + "loss": 0.7022, + "step": 18895 + }, + { + "epoch": 0.69, + "learning_rate": 1.3098472919031748e-07, + "loss": 0.6954, + "step": 18896 + }, + { + "epoch": 0.69, + "learning_rate": 1.3098106712564543e-07, + "loss": 0.679, + "step": 18897 + }, + { + "epoch": 0.69, + "learning_rate": 1.3097740506097336e-07, + "loss": 0.7034, + "step": 18898 + }, + { + "epoch": 0.69, + "learning_rate": 1.309737429963013e-07, + "loss": 0.7016, + "step": 18899 + }, + { + "epoch": 0.69, + "learning_rate": 1.3097008093162924e-07, + "loss": 0.6959, + "step": 18900 + }, + { + "epoch": 0.69, + "learning_rate": 1.309664188669572e-07, + "loss": 0.692, + "step": 18901 + }, + { + "epoch": 0.69, + "learning_rate": 1.3096275680228512e-07, + "loss": 0.6755, + "step": 18902 + }, + { + "epoch": 0.69, + "learning_rate": 1.3095909473761307e-07, + "loss": 0.6885, + "step": 18903 + }, + { + "epoch": 0.69, + "learning_rate": 1.30955432672941e-07, + "loss": 0.6872, + "step": 18904 + }, + { + "epoch": 0.69, + "learning_rate": 1.3095177060826895e-07, + "loss": 0.6951, + "step": 18905 + }, + { + "epoch": 0.69, + "learning_rate": 1.3094810854359688e-07, + "loss": 0.6877, + "step": 18906 + }, + { + "epoch": 0.69, + "learning_rate": 1.3094444647892483e-07, + "loss": 0.69, + "step": 18907 + }, + { + "epoch": 0.69, + "learning_rate": 1.3094078441425276e-07, + "loss": 0.6939, + "step": 18908 + }, + { + "epoch": 0.69, + "learning_rate": 1.309371223495807e-07, + "loss": 0.6957, + "step": 18909 + }, + { + "epoch": 0.69, + "learning_rate": 1.309334602849086e-07, + "loss": 0.7125, + "step": 18910 + }, + { + "epoch": 0.69, + "learning_rate": 1.3092979822023656e-07, + "loss": 0.6853, + "step": 18911 + }, + { + "epoch": 0.69, + "learning_rate": 1.309261361555645e-07, + "loss": 0.6925, + "step": 18912 + }, + { + "epoch": 0.69, + "learning_rate": 1.3092247409089244e-07, + "loss": 0.6844, + "step": 18913 + }, + { + "epoch": 0.69, + "learning_rate": 1.3091881202622037e-07, + "loss": 0.7096, + "step": 18914 + }, + { + "epoch": 0.69, + "learning_rate": 1.309151499615483e-07, + "loss": 0.6836, + "step": 18915 + }, + { + "epoch": 0.69, + "learning_rate": 1.3091148789687625e-07, + "loss": 0.6905, + "step": 18916 + }, + { + "epoch": 0.69, + "learning_rate": 1.3090782583220418e-07, + "loss": 0.6775, + "step": 18917 + }, + { + "epoch": 0.69, + "learning_rate": 1.3090416376753213e-07, + "loss": 0.695, + "step": 18918 + }, + { + "epoch": 0.69, + "learning_rate": 1.3090050170286006e-07, + "loss": 0.7088, + "step": 18919 + }, + { + "epoch": 0.69, + "learning_rate": 1.30896839638188e-07, + "loss": 0.693, + "step": 18920 + }, + { + "epoch": 0.69, + "learning_rate": 1.3089317757351594e-07, + "loss": 0.6955, + "step": 18921 + }, + { + "epoch": 0.69, + "learning_rate": 1.308895155088439e-07, + "loss": 0.702, + "step": 18922 + }, + { + "epoch": 0.69, + "learning_rate": 1.3088585344417182e-07, + "loss": 0.6952, + "step": 18923 + }, + { + "epoch": 0.69, + "learning_rate": 1.3088219137949977e-07, + "loss": 0.6988, + "step": 18924 + }, + { + "epoch": 0.69, + "learning_rate": 1.308785293148277e-07, + "loss": 0.688, + "step": 18925 + }, + { + "epoch": 0.69, + "learning_rate": 1.3087486725015565e-07, + "loss": 0.6943, + "step": 18926 + }, + { + "epoch": 0.69, + "learning_rate": 1.3087120518548358e-07, + "loss": 0.682, + "step": 18927 + }, + { + "epoch": 0.69, + "learning_rate": 1.308675431208115e-07, + "loss": 0.6869, + "step": 18928 + }, + { + "epoch": 0.69, + "learning_rate": 1.3086388105613943e-07, + "loss": 0.6824, + "step": 18929 + }, + { + "epoch": 0.69, + "learning_rate": 1.3086021899146738e-07, + "loss": 0.6922, + "step": 18930 + }, + { + "epoch": 0.69, + "learning_rate": 1.308565569267953e-07, + "loss": 0.6985, + "step": 18931 + }, + { + "epoch": 0.69, + "learning_rate": 1.3085289486212326e-07, + "loss": 0.6899, + "step": 18932 + }, + { + "epoch": 0.69, + "learning_rate": 1.308492327974512e-07, + "loss": 0.6881, + "step": 18933 + }, + { + "epoch": 0.69, + "learning_rate": 1.3084557073277914e-07, + "loss": 0.6877, + "step": 18934 + }, + { + "epoch": 0.69, + "learning_rate": 1.3084190866810707e-07, + "loss": 0.7034, + "step": 18935 + }, + { + "epoch": 0.69, + "learning_rate": 1.3083824660343502e-07, + "loss": 0.6989, + "step": 18936 + }, + { + "epoch": 0.69, + "learning_rate": 1.3083458453876295e-07, + "loss": 0.7015, + "step": 18937 + }, + { + "epoch": 0.69, + "learning_rate": 1.3083092247409088e-07, + "loss": 0.7075, + "step": 18938 + }, + { + "epoch": 0.69, + "learning_rate": 1.3082726040941883e-07, + "loss": 0.6955, + "step": 18939 + }, + { + "epoch": 0.69, + "learning_rate": 1.3082359834474676e-07, + "loss": 0.7152, + "step": 18940 + }, + { + "epoch": 0.69, + "learning_rate": 1.308199362800747e-07, + "loss": 0.6944, + "step": 18941 + }, + { + "epoch": 0.69, + "learning_rate": 1.3081627421540264e-07, + "loss": 0.6943, + "step": 18942 + }, + { + "epoch": 0.69, + "learning_rate": 1.308126121507306e-07, + "loss": 0.6935, + "step": 18943 + }, + { + "epoch": 0.69, + "learning_rate": 1.3080895008605852e-07, + "loss": 0.704, + "step": 18944 + }, + { + "epoch": 0.69, + "learning_rate": 1.3080528802138647e-07, + "loss": 0.6943, + "step": 18945 + }, + { + "epoch": 0.69, + "learning_rate": 1.3080162595671437e-07, + "loss": 0.6941, + "step": 18946 + }, + { + "epoch": 0.69, + "learning_rate": 1.3079796389204232e-07, + "loss": 0.6912, + "step": 18947 + }, + { + "epoch": 0.69, + "learning_rate": 1.3079430182737025e-07, + "loss": 0.6867, + "step": 18948 + }, + { + "epoch": 0.69, + "learning_rate": 1.307906397626982e-07, + "loss": 0.695, + "step": 18949 + }, + { + "epoch": 0.69, + "learning_rate": 1.3078697769802613e-07, + "loss": 0.7017, + "step": 18950 + }, + { + "epoch": 0.69, + "learning_rate": 1.3078331563335408e-07, + "loss": 0.7064, + "step": 18951 + }, + { + "epoch": 0.69, + "learning_rate": 1.30779653568682e-07, + "loss": 0.6843, + "step": 18952 + }, + { + "epoch": 0.69, + "learning_rate": 1.3077599150400996e-07, + "loss": 0.7012, + "step": 18953 + }, + { + "epoch": 0.69, + "learning_rate": 1.307723294393379e-07, + "loss": 0.6994, + "step": 18954 + }, + { + "epoch": 0.69, + "learning_rate": 1.3076866737466584e-07, + "loss": 0.6902, + "step": 18955 + }, + { + "epoch": 0.69, + "learning_rate": 1.3076500530999377e-07, + "loss": 0.6782, + "step": 18956 + }, + { + "epoch": 0.69, + "learning_rate": 1.3076134324532172e-07, + "loss": 0.6889, + "step": 18957 + }, + { + "epoch": 0.69, + "learning_rate": 1.3075768118064965e-07, + "loss": 0.6896, + "step": 18958 + }, + { + "epoch": 0.69, + "learning_rate": 1.3075401911597757e-07, + "loss": 0.6785, + "step": 18959 + }, + { + "epoch": 0.69, + "learning_rate": 1.3075035705130553e-07, + "loss": 0.6958, + "step": 18960 + }, + { + "epoch": 0.69, + "learning_rate": 1.3074669498663345e-07, + "loss": 0.7092, + "step": 18961 + }, + { + "epoch": 0.69, + "learning_rate": 1.307430329219614e-07, + "loss": 0.6994, + "step": 18962 + }, + { + "epoch": 0.69, + "learning_rate": 1.3073937085728933e-07, + "loss": 0.6823, + "step": 18963 + }, + { + "epoch": 0.69, + "learning_rate": 1.3073570879261726e-07, + "loss": 0.6866, + "step": 18964 + }, + { + "epoch": 0.69, + "learning_rate": 1.307320467279452e-07, + "loss": 0.7028, + "step": 18965 + }, + { + "epoch": 0.69, + "learning_rate": 1.3072838466327314e-07, + "loss": 0.6976, + "step": 18966 + }, + { + "epoch": 0.69, + "learning_rate": 1.3072472259860107e-07, + "loss": 0.6894, + "step": 18967 + }, + { + "epoch": 0.69, + "learning_rate": 1.3072106053392902e-07, + "loss": 0.6891, + "step": 18968 + }, + { + "epoch": 0.69, + "learning_rate": 1.3071739846925695e-07, + "loss": 0.6984, + "step": 18969 + }, + { + "epoch": 0.69, + "learning_rate": 1.307137364045849e-07, + "loss": 0.6885, + "step": 18970 + }, + { + "epoch": 0.69, + "learning_rate": 1.3071007433991283e-07, + "loss": 0.6831, + "step": 18971 + }, + { + "epoch": 0.69, + "learning_rate": 1.3070641227524078e-07, + "loss": 0.6991, + "step": 18972 + }, + { + "epoch": 0.69, + "learning_rate": 1.307027502105687e-07, + "loss": 0.7044, + "step": 18973 + }, + { + "epoch": 0.69, + "learning_rate": 1.3069908814589666e-07, + "loss": 0.7134, + "step": 18974 + }, + { + "epoch": 0.69, + "learning_rate": 1.306954260812246e-07, + "loss": 0.6966, + "step": 18975 + }, + { + "epoch": 0.69, + "learning_rate": 1.3069176401655254e-07, + "loss": 0.6867, + "step": 18976 + }, + { + "epoch": 0.69, + "learning_rate": 1.3068810195188047e-07, + "loss": 0.6948, + "step": 18977 + }, + { + "epoch": 0.69, + "learning_rate": 1.3068443988720842e-07, + "loss": 0.695, + "step": 18978 + }, + { + "epoch": 0.69, + "learning_rate": 1.3068077782253635e-07, + "loss": 0.6908, + "step": 18979 + }, + { + "epoch": 0.69, + "learning_rate": 1.306771157578643e-07, + "loss": 0.6865, + "step": 18980 + }, + { + "epoch": 0.69, + "learning_rate": 1.3067345369319223e-07, + "loss": 0.6921, + "step": 18981 + }, + { + "epoch": 0.69, + "learning_rate": 1.3066979162852015e-07, + "loss": 0.6854, + "step": 18982 + }, + { + "epoch": 0.69, + "learning_rate": 1.3066612956384808e-07, + "loss": 0.6906, + "step": 18983 + }, + { + "epoch": 0.69, + "learning_rate": 1.3066246749917603e-07, + "loss": 0.6778, + "step": 18984 + }, + { + "epoch": 0.69, + "learning_rate": 1.3065880543450396e-07, + "loss": 0.6981, + "step": 18985 + }, + { + "epoch": 0.69, + "learning_rate": 1.3065514336983189e-07, + "loss": 0.6749, + "step": 18986 + }, + { + "epoch": 0.69, + "learning_rate": 1.3065148130515984e-07, + "loss": 0.6989, + "step": 18987 + }, + { + "epoch": 0.69, + "learning_rate": 1.3064781924048777e-07, + "loss": 0.6919, + "step": 18988 + }, + { + "epoch": 0.69, + "learning_rate": 1.3064415717581572e-07, + "loss": 0.6831, + "step": 18989 + }, + { + "epoch": 0.69, + "learning_rate": 1.3064049511114365e-07, + "loss": 0.6999, + "step": 18990 + }, + { + "epoch": 0.69, + "learning_rate": 1.306368330464716e-07, + "loss": 0.7055, + "step": 18991 + }, + { + "epoch": 0.69, + "learning_rate": 1.3063317098179953e-07, + "loss": 0.7073, + "step": 18992 + }, + { + "epoch": 0.69, + "learning_rate": 1.3062950891712748e-07, + "loss": 0.6995, + "step": 18993 + }, + { + "epoch": 0.69, + "learning_rate": 1.306258468524554e-07, + "loss": 0.6995, + "step": 18994 + }, + { + "epoch": 0.69, + "learning_rate": 1.3062218478778336e-07, + "loss": 0.6833, + "step": 18995 + }, + { + "epoch": 0.7, + "learning_rate": 1.3061852272311129e-07, + "loss": 0.6946, + "step": 18996 + }, + { + "epoch": 0.7, + "learning_rate": 1.3061486065843924e-07, + "loss": 0.6934, + "step": 18997 + }, + { + "epoch": 0.7, + "learning_rate": 1.3061119859376717e-07, + "loss": 0.6902, + "step": 18998 + }, + { + "epoch": 0.7, + "learning_rate": 1.3060753652909512e-07, + "loss": 0.704, + "step": 18999 + }, + { + "epoch": 0.7, + "learning_rate": 1.3060387446442302e-07, + "loss": 0.6924, + "step": 19000 + }, + { + "epoch": 0.7, + "learning_rate": 1.3060021239975097e-07, + "loss": 0.7023, + "step": 19001 + }, + { + "epoch": 0.7, + "learning_rate": 1.305965503350789e-07, + "loss": 0.71, + "step": 19002 + }, + { + "epoch": 0.7, + "learning_rate": 1.3059288827040685e-07, + "loss": 0.6937, + "step": 19003 + }, + { + "epoch": 0.7, + "learning_rate": 1.3058922620573478e-07, + "loss": 0.7156, + "step": 19004 + }, + { + "epoch": 0.7, + "learning_rate": 1.3058556414106273e-07, + "loss": 0.6948, + "step": 19005 + }, + { + "epoch": 0.7, + "learning_rate": 1.3058190207639066e-07, + "loss": 0.7049, + "step": 19006 + }, + { + "epoch": 0.7, + "learning_rate": 1.305782400117186e-07, + "loss": 0.686, + "step": 19007 + }, + { + "epoch": 0.7, + "learning_rate": 1.3057457794704654e-07, + "loss": 0.6969, + "step": 19008 + }, + { + "epoch": 0.7, + "learning_rate": 1.3057091588237446e-07, + "loss": 0.6906, + "step": 19009 + }, + { + "epoch": 0.7, + "learning_rate": 1.3056725381770242e-07, + "loss": 0.6908, + "step": 19010 + }, + { + "epoch": 0.7, + "learning_rate": 1.3056359175303034e-07, + "loss": 0.6978, + "step": 19011 + }, + { + "epoch": 0.7, + "learning_rate": 1.305599296883583e-07, + "loss": 0.6914, + "step": 19012 + }, + { + "epoch": 0.7, + "learning_rate": 1.3055626762368622e-07, + "loss": 0.6911, + "step": 19013 + }, + { + "epoch": 0.7, + "learning_rate": 1.3055260555901418e-07, + "loss": 0.7013, + "step": 19014 + }, + { + "epoch": 0.7, + "learning_rate": 1.305489434943421e-07, + "loss": 0.6945, + "step": 19015 + }, + { + "epoch": 0.7, + "learning_rate": 1.3054528142967006e-07, + "loss": 0.6891, + "step": 19016 + }, + { + "epoch": 0.7, + "learning_rate": 1.3054161936499798e-07, + "loss": 0.7096, + "step": 19017 + }, + { + "epoch": 0.7, + "learning_rate": 1.305379573003259e-07, + "loss": 0.695, + "step": 19018 + }, + { + "epoch": 0.7, + "learning_rate": 1.3053429523565384e-07, + "loss": 0.6902, + "step": 19019 + }, + { + "epoch": 0.7, + "learning_rate": 1.305306331709818e-07, + "loss": 0.6822, + "step": 19020 + }, + { + "epoch": 0.7, + "learning_rate": 1.3052697110630972e-07, + "loss": 0.6884, + "step": 19021 + }, + { + "epoch": 0.7, + "learning_rate": 1.3052330904163767e-07, + "loss": 0.6896, + "step": 19022 + }, + { + "epoch": 0.7, + "learning_rate": 1.305196469769656e-07, + "loss": 0.6862, + "step": 19023 + }, + { + "epoch": 0.7, + "learning_rate": 1.3051598491229355e-07, + "loss": 0.6954, + "step": 19024 + }, + { + "epoch": 0.7, + "learning_rate": 1.3051232284762148e-07, + "loss": 0.6796, + "step": 19025 + }, + { + "epoch": 0.7, + "learning_rate": 1.3050866078294943e-07, + "loss": 0.6879, + "step": 19026 + }, + { + "epoch": 0.7, + "learning_rate": 1.3050499871827736e-07, + "loss": 0.6636, + "step": 19027 + }, + { + "epoch": 0.7, + "learning_rate": 1.305013366536053e-07, + "loss": 0.7087, + "step": 19028 + }, + { + "epoch": 0.7, + "learning_rate": 1.3049767458893324e-07, + "loss": 0.6992, + "step": 19029 + }, + { + "epoch": 0.7, + "learning_rate": 1.3049401252426116e-07, + "loss": 0.6911, + "step": 19030 + }, + { + "epoch": 0.7, + "learning_rate": 1.3049035045958912e-07, + "loss": 0.7081, + "step": 19031 + }, + { + "epoch": 0.7, + "learning_rate": 1.3048668839491704e-07, + "loss": 0.6958, + "step": 19032 + }, + { + "epoch": 0.7, + "learning_rate": 1.30483026330245e-07, + "loss": 0.6983, + "step": 19033 + }, + { + "epoch": 0.7, + "learning_rate": 1.3047936426557292e-07, + "loss": 0.6853, + "step": 19034 + }, + { + "epoch": 0.7, + "learning_rate": 1.3047570220090088e-07, + "loss": 0.6826, + "step": 19035 + }, + { + "epoch": 0.7, + "learning_rate": 1.3047204013622878e-07, + "loss": 0.701, + "step": 19036 + }, + { + "epoch": 0.7, + "learning_rate": 1.3046837807155673e-07, + "loss": 0.6938, + "step": 19037 + }, + { + "epoch": 0.7, + "learning_rate": 1.3046471600688466e-07, + "loss": 0.688, + "step": 19038 + }, + { + "epoch": 0.7, + "learning_rate": 1.304610539422126e-07, + "loss": 0.6935, + "step": 19039 + }, + { + "epoch": 0.7, + "learning_rate": 1.3045739187754054e-07, + "loss": 0.7058, + "step": 19040 + }, + { + "epoch": 0.7, + "learning_rate": 1.304537298128685e-07, + "loss": 0.6984, + "step": 19041 + }, + { + "epoch": 0.7, + "learning_rate": 1.3045006774819642e-07, + "loss": 0.682, + "step": 19042 + }, + { + "epoch": 0.7, + "learning_rate": 1.3044640568352437e-07, + "loss": 0.6843, + "step": 19043 + }, + { + "epoch": 0.7, + "learning_rate": 1.304427436188523e-07, + "loss": 0.6923, + "step": 19044 + }, + { + "epoch": 0.7, + "learning_rate": 1.3043908155418025e-07, + "loss": 0.6933, + "step": 19045 + }, + { + "epoch": 0.7, + "learning_rate": 1.3043541948950818e-07, + "loss": 0.6942, + "step": 19046 + }, + { + "epoch": 0.7, + "learning_rate": 1.3043175742483613e-07, + "loss": 0.6845, + "step": 19047 + }, + { + "epoch": 0.7, + "learning_rate": 1.3042809536016406e-07, + "loss": 0.7036, + "step": 19048 + }, + { + "epoch": 0.7, + "learning_rate": 1.30424433295492e-07, + "loss": 0.6746, + "step": 19049 + }, + { + "epoch": 0.7, + "learning_rate": 1.3042077123081994e-07, + "loss": 0.6831, + "step": 19050 + }, + { + "epoch": 0.7, + "learning_rate": 1.304171091661479e-07, + "loss": 0.6919, + "step": 19051 + }, + { + "epoch": 0.7, + "learning_rate": 1.3041344710147582e-07, + "loss": 0.7025, + "step": 19052 + }, + { + "epoch": 0.7, + "learning_rate": 1.3040978503680374e-07, + "loss": 0.6935, + "step": 19053 + }, + { + "epoch": 0.7, + "learning_rate": 1.3040612297213167e-07, + "loss": 0.685, + "step": 19054 + }, + { + "epoch": 0.7, + "learning_rate": 1.3040246090745962e-07, + "loss": 0.6894, + "step": 19055 + }, + { + "epoch": 0.7, + "learning_rate": 1.3039879884278755e-07, + "loss": 0.674, + "step": 19056 + }, + { + "epoch": 0.7, + "learning_rate": 1.3039513677811548e-07, + "loss": 0.69, + "step": 19057 + }, + { + "epoch": 0.7, + "learning_rate": 1.3039147471344343e-07, + "loss": 0.6939, + "step": 19058 + }, + { + "epoch": 0.7, + "learning_rate": 1.3038781264877136e-07, + "loss": 0.6955, + "step": 19059 + }, + { + "epoch": 0.7, + "learning_rate": 1.303841505840993e-07, + "loss": 0.6967, + "step": 19060 + }, + { + "epoch": 0.7, + "learning_rate": 1.3038048851942724e-07, + "loss": 0.6912, + "step": 19061 + }, + { + "epoch": 0.7, + "learning_rate": 1.303768264547552e-07, + "loss": 0.6939, + "step": 19062 + }, + { + "epoch": 0.7, + "learning_rate": 1.3037316439008312e-07, + "loss": 0.6834, + "step": 19063 + }, + { + "epoch": 0.7, + "learning_rate": 1.3036950232541107e-07, + "loss": 0.6821, + "step": 19064 + }, + { + "epoch": 0.7, + "learning_rate": 1.30365840260739e-07, + "loss": 0.688, + "step": 19065 + }, + { + "epoch": 0.7, + "learning_rate": 1.3036217819606695e-07, + "loss": 0.7107, + "step": 19066 + }, + { + "epoch": 0.7, + "learning_rate": 1.3035851613139487e-07, + "loss": 0.6979, + "step": 19067 + }, + { + "epoch": 0.7, + "learning_rate": 1.3035485406672283e-07, + "loss": 0.696, + "step": 19068 + }, + { + "epoch": 0.7, + "learning_rate": 1.3035119200205075e-07, + "loss": 0.6898, + "step": 19069 + }, + { + "epoch": 0.7, + "learning_rate": 1.303475299373787e-07, + "loss": 0.715, + "step": 19070 + }, + { + "epoch": 0.7, + "learning_rate": 1.3034386787270663e-07, + "loss": 0.7048, + "step": 19071 + }, + { + "epoch": 0.7, + "learning_rate": 1.3034020580803456e-07, + "loss": 0.6959, + "step": 19072 + }, + { + "epoch": 0.7, + "learning_rate": 1.303365437433625e-07, + "loss": 0.6773, + "step": 19073 + }, + { + "epoch": 0.7, + "learning_rate": 1.3033288167869044e-07, + "loss": 0.7016, + "step": 19074 + }, + { + "epoch": 0.7, + "learning_rate": 1.3032921961401837e-07, + "loss": 0.6893, + "step": 19075 + }, + { + "epoch": 0.7, + "learning_rate": 1.3032555754934632e-07, + "loss": 0.7011, + "step": 19076 + }, + { + "epoch": 0.7, + "learning_rate": 1.3032189548467425e-07, + "loss": 0.6895, + "step": 19077 + }, + { + "epoch": 0.7, + "learning_rate": 1.303182334200022e-07, + "loss": 0.7013, + "step": 19078 + }, + { + "epoch": 0.7, + "learning_rate": 1.3031457135533013e-07, + "loss": 0.6871, + "step": 19079 + }, + { + "epoch": 0.7, + "learning_rate": 1.3031090929065805e-07, + "loss": 0.6826, + "step": 19080 + }, + { + "epoch": 0.7, + "learning_rate": 1.30307247225986e-07, + "loss": 0.6819, + "step": 19081 + }, + { + "epoch": 0.7, + "learning_rate": 1.3030358516131393e-07, + "loss": 0.6781, + "step": 19082 + }, + { + "epoch": 0.7, + "learning_rate": 1.302999230966419e-07, + "loss": 0.696, + "step": 19083 + }, + { + "epoch": 0.7, + "learning_rate": 1.3029626103196981e-07, + "loss": 0.6886, + "step": 19084 + }, + { + "epoch": 0.7, + "learning_rate": 1.3029259896729777e-07, + "loss": 0.7014, + "step": 19085 + }, + { + "epoch": 0.7, + "learning_rate": 1.302889369026257e-07, + "loss": 0.6946, + "step": 19086 + }, + { + "epoch": 0.7, + "learning_rate": 1.3028527483795365e-07, + "loss": 0.6863, + "step": 19087 + }, + { + "epoch": 0.7, + "learning_rate": 1.3028161277328157e-07, + "loss": 0.6956, + "step": 19088 + }, + { + "epoch": 0.7, + "learning_rate": 1.3027795070860953e-07, + "loss": 0.6678, + "step": 19089 + }, + { + "epoch": 0.7, + "learning_rate": 1.3027428864393743e-07, + "loss": 0.6849, + "step": 19090 + }, + { + "epoch": 0.7, + "learning_rate": 1.3027062657926538e-07, + "loss": 0.7027, + "step": 19091 + }, + { + "epoch": 0.7, + "learning_rate": 1.302669645145933e-07, + "loss": 0.699, + "step": 19092 + }, + { + "epoch": 0.7, + "learning_rate": 1.3026330244992126e-07, + "loss": 0.7121, + "step": 19093 + }, + { + "epoch": 0.7, + "learning_rate": 1.302596403852492e-07, + "loss": 0.6857, + "step": 19094 + }, + { + "epoch": 0.7, + "learning_rate": 1.3025597832057714e-07, + "loss": 0.6831, + "step": 19095 + }, + { + "epoch": 0.7, + "learning_rate": 1.3025231625590507e-07, + "loss": 0.6909, + "step": 19096 + }, + { + "epoch": 0.7, + "learning_rate": 1.3024865419123302e-07, + "loss": 0.6942, + "step": 19097 + }, + { + "epoch": 0.7, + "learning_rate": 1.3024499212656095e-07, + "loss": 0.6975, + "step": 19098 + }, + { + "epoch": 0.7, + "learning_rate": 1.302413300618889e-07, + "loss": 0.6889, + "step": 19099 + }, + { + "epoch": 0.7, + "learning_rate": 1.3023766799721683e-07, + "loss": 0.6891, + "step": 19100 + }, + { + "epoch": 0.7, + "learning_rate": 1.3023400593254475e-07, + "loss": 0.714, + "step": 19101 + }, + { + "epoch": 0.7, + "learning_rate": 1.302303438678727e-07, + "loss": 0.6846, + "step": 19102 + }, + { + "epoch": 0.7, + "learning_rate": 1.3022668180320063e-07, + "loss": 0.6903, + "step": 19103 + }, + { + "epoch": 0.7, + "learning_rate": 1.3022301973852859e-07, + "loss": 0.6873, + "step": 19104 + }, + { + "epoch": 0.7, + "learning_rate": 1.302193576738565e-07, + "loss": 0.6882, + "step": 19105 + }, + { + "epoch": 0.7, + "learning_rate": 1.3021569560918447e-07, + "loss": 0.6857, + "step": 19106 + }, + { + "epoch": 0.7, + "learning_rate": 1.302120335445124e-07, + "loss": 0.6918, + "step": 19107 + }, + { + "epoch": 0.7, + "learning_rate": 1.3020837147984032e-07, + "loss": 0.677, + "step": 19108 + }, + { + "epoch": 0.7, + "learning_rate": 1.3020470941516825e-07, + "loss": 0.696, + "step": 19109 + }, + { + "epoch": 0.7, + "learning_rate": 1.302010473504962e-07, + "loss": 0.6896, + "step": 19110 + }, + { + "epoch": 0.7, + "learning_rate": 1.3019738528582413e-07, + "loss": 0.6885, + "step": 19111 + }, + { + "epoch": 0.7, + "learning_rate": 1.3019372322115208e-07, + "loss": 0.704, + "step": 19112 + }, + { + "epoch": 0.7, + "learning_rate": 1.3019006115648e-07, + "loss": 0.7054, + "step": 19113 + }, + { + "epoch": 0.7, + "learning_rate": 1.3018639909180796e-07, + "loss": 0.7178, + "step": 19114 + }, + { + "epoch": 0.7, + "learning_rate": 1.3018273702713589e-07, + "loss": 0.6908, + "step": 19115 + }, + { + "epoch": 0.7, + "learning_rate": 1.3017907496246384e-07, + "loss": 0.7027, + "step": 19116 + }, + { + "epoch": 0.7, + "learning_rate": 1.3017541289779177e-07, + "loss": 0.6911, + "step": 19117 + }, + { + "epoch": 0.7, + "learning_rate": 1.3017175083311972e-07, + "loss": 0.6946, + "step": 19118 + }, + { + "epoch": 0.7, + "learning_rate": 1.3016808876844765e-07, + "loss": 0.6696, + "step": 19119 + }, + { + "epoch": 0.7, + "learning_rate": 1.301644267037756e-07, + "loss": 0.6824, + "step": 19120 + }, + { + "epoch": 0.7, + "learning_rate": 1.3016076463910353e-07, + "loss": 0.6952, + "step": 19121 + }, + { + "epoch": 0.7, + "learning_rate": 1.3015710257443148e-07, + "loss": 0.6993, + "step": 19122 + }, + { + "epoch": 0.7, + "learning_rate": 1.301534405097594e-07, + "loss": 0.7069, + "step": 19123 + }, + { + "epoch": 0.7, + "learning_rate": 1.3014977844508733e-07, + "loss": 0.6894, + "step": 19124 + }, + { + "epoch": 0.7, + "learning_rate": 1.3014611638041528e-07, + "loss": 0.6957, + "step": 19125 + }, + { + "epoch": 0.7, + "learning_rate": 1.301424543157432e-07, + "loss": 0.6948, + "step": 19126 + }, + { + "epoch": 0.7, + "learning_rate": 1.3013879225107114e-07, + "loss": 0.6924, + "step": 19127 + }, + { + "epoch": 0.7, + "learning_rate": 1.3013513018639907e-07, + "loss": 0.7034, + "step": 19128 + }, + { + "epoch": 0.7, + "learning_rate": 1.3013146812172702e-07, + "loss": 0.6933, + "step": 19129 + }, + { + "epoch": 0.7, + "learning_rate": 1.3012780605705494e-07, + "loss": 0.6928, + "step": 19130 + }, + { + "epoch": 0.7, + "learning_rate": 1.301241439923829e-07, + "loss": 0.6907, + "step": 19131 + }, + { + "epoch": 0.7, + "learning_rate": 1.3012048192771082e-07, + "loss": 0.6808, + "step": 19132 + }, + { + "epoch": 0.7, + "learning_rate": 1.3011681986303878e-07, + "loss": 0.6962, + "step": 19133 + }, + { + "epoch": 0.7, + "learning_rate": 1.301131577983667e-07, + "loss": 0.6983, + "step": 19134 + }, + { + "epoch": 0.7, + "learning_rate": 1.3010949573369466e-07, + "loss": 0.6951, + "step": 19135 + }, + { + "epoch": 0.7, + "learning_rate": 1.3010583366902258e-07, + "loss": 0.6848, + "step": 19136 + }, + { + "epoch": 0.7, + "learning_rate": 1.3010217160435054e-07, + "loss": 0.7011, + "step": 19137 + }, + { + "epoch": 0.7, + "learning_rate": 1.3009850953967846e-07, + "loss": 0.6897, + "step": 19138 + }, + { + "epoch": 0.7, + "learning_rate": 1.3009484747500642e-07, + "loss": 0.7037, + "step": 19139 + }, + { + "epoch": 0.7, + "learning_rate": 1.3009118541033434e-07, + "loss": 0.699, + "step": 19140 + }, + { + "epoch": 0.7, + "learning_rate": 1.300875233456623e-07, + "loss": 0.693, + "step": 19141 + }, + { + "epoch": 0.7, + "learning_rate": 1.3008386128099022e-07, + "loss": 0.6871, + "step": 19142 + }, + { + "epoch": 0.7, + "learning_rate": 1.3008019921631818e-07, + "loss": 0.6931, + "step": 19143 + }, + { + "epoch": 0.7, + "learning_rate": 1.3007653715164608e-07, + "loss": 0.6737, + "step": 19144 + }, + { + "epoch": 0.7, + "learning_rate": 1.3007287508697403e-07, + "loss": 0.7, + "step": 19145 + }, + { + "epoch": 0.7, + "learning_rate": 1.3006921302230196e-07, + "loss": 0.6796, + "step": 19146 + }, + { + "epoch": 0.7, + "learning_rate": 1.300655509576299e-07, + "loss": 0.7035, + "step": 19147 + }, + { + "epoch": 0.7, + "learning_rate": 1.3006188889295784e-07, + "loss": 0.6929, + "step": 19148 + }, + { + "epoch": 0.7, + "learning_rate": 1.300582268282858e-07, + "loss": 0.6856, + "step": 19149 + }, + { + "epoch": 0.7, + "learning_rate": 1.3005456476361372e-07, + "loss": 0.6954, + "step": 19150 + }, + { + "epoch": 0.7, + "learning_rate": 1.3005090269894164e-07, + "loss": 0.6845, + "step": 19151 + }, + { + "epoch": 0.7, + "learning_rate": 1.300472406342696e-07, + "loss": 0.6944, + "step": 19152 + }, + { + "epoch": 0.7, + "learning_rate": 1.3004357856959752e-07, + "loss": 0.6867, + "step": 19153 + }, + { + "epoch": 0.7, + "learning_rate": 1.3003991650492548e-07, + "loss": 0.7064, + "step": 19154 + }, + { + "epoch": 0.7, + "learning_rate": 1.300362544402534e-07, + "loss": 0.6851, + "step": 19155 + }, + { + "epoch": 0.7, + "learning_rate": 1.3003259237558136e-07, + "loss": 0.6972, + "step": 19156 + }, + { + "epoch": 0.7, + "learning_rate": 1.3002893031090928e-07, + "loss": 0.694, + "step": 19157 + }, + { + "epoch": 0.7, + "learning_rate": 1.3002526824623724e-07, + "loss": 0.6914, + "step": 19158 + }, + { + "epoch": 0.7, + "learning_rate": 1.3002160618156516e-07, + "loss": 0.6886, + "step": 19159 + }, + { + "epoch": 0.7, + "learning_rate": 1.3001794411689312e-07, + "loss": 0.7049, + "step": 19160 + }, + { + "epoch": 0.7, + "learning_rate": 1.3001428205222104e-07, + "loss": 0.6804, + "step": 19161 + }, + { + "epoch": 0.7, + "learning_rate": 1.3001061998754897e-07, + "loss": 0.7102, + "step": 19162 + }, + { + "epoch": 0.7, + "learning_rate": 1.300069579228769e-07, + "loss": 0.6941, + "step": 19163 + }, + { + "epoch": 0.7, + "learning_rate": 1.3000329585820485e-07, + "loss": 0.6858, + "step": 19164 + }, + { + "epoch": 0.7, + "learning_rate": 1.2999963379353278e-07, + "loss": 0.6923, + "step": 19165 + }, + { + "epoch": 0.7, + "learning_rate": 1.2999597172886073e-07, + "loss": 0.6944, + "step": 19166 + }, + { + "epoch": 0.7, + "learning_rate": 1.2999230966418866e-07, + "loss": 0.6934, + "step": 19167 + }, + { + "epoch": 0.7, + "learning_rate": 1.299886475995166e-07, + "loss": 0.6974, + "step": 19168 + }, + { + "epoch": 0.7, + "learning_rate": 1.2998498553484454e-07, + "loss": 0.6832, + "step": 19169 + }, + { + "epoch": 0.7, + "learning_rate": 1.299813234701725e-07, + "loss": 0.693, + "step": 19170 + }, + { + "epoch": 0.7, + "learning_rate": 1.2997766140550042e-07, + "loss": 0.6937, + "step": 19171 + }, + { + "epoch": 0.7, + "learning_rate": 1.2997399934082834e-07, + "loss": 0.7161, + "step": 19172 + }, + { + "epoch": 0.7, + "learning_rate": 1.299703372761563e-07, + "loss": 0.6991, + "step": 19173 + }, + { + "epoch": 0.7, + "learning_rate": 1.2996667521148422e-07, + "loss": 0.7005, + "step": 19174 + }, + { + "epoch": 0.7, + "learning_rate": 1.2996301314681218e-07, + "loss": 0.6921, + "step": 19175 + }, + { + "epoch": 0.7, + "learning_rate": 1.299593510821401e-07, + "loss": 0.6875, + "step": 19176 + }, + { + "epoch": 0.7, + "learning_rate": 1.2995568901746806e-07, + "loss": 0.6993, + "step": 19177 + }, + { + "epoch": 0.7, + "learning_rate": 1.2995202695279598e-07, + "loss": 0.6939, + "step": 19178 + }, + { + "epoch": 0.7, + "learning_rate": 1.2994836488812394e-07, + "loss": 0.6858, + "step": 19179 + }, + { + "epoch": 0.7, + "learning_rate": 1.2994470282345184e-07, + "loss": 0.7066, + "step": 19180 + }, + { + "epoch": 0.7, + "learning_rate": 1.299410407587798e-07, + "loss": 0.6878, + "step": 19181 + }, + { + "epoch": 0.7, + "learning_rate": 1.2993737869410772e-07, + "loss": 0.6961, + "step": 19182 + }, + { + "epoch": 0.7, + "learning_rate": 1.2993371662943567e-07, + "loss": 0.6878, + "step": 19183 + }, + { + "epoch": 0.7, + "learning_rate": 1.299300545647636e-07, + "loss": 0.6976, + "step": 19184 + }, + { + "epoch": 0.7, + "learning_rate": 1.2992639250009155e-07, + "loss": 0.687, + "step": 19185 + }, + { + "epoch": 0.7, + "learning_rate": 1.2992273043541947e-07, + "loss": 0.6993, + "step": 19186 + }, + { + "epoch": 0.7, + "learning_rate": 1.2991906837074743e-07, + "loss": 0.6943, + "step": 19187 + }, + { + "epoch": 0.7, + "learning_rate": 1.2991540630607535e-07, + "loss": 0.6909, + "step": 19188 + }, + { + "epoch": 0.7, + "learning_rate": 1.299117442414033e-07, + "loss": 0.7014, + "step": 19189 + }, + { + "epoch": 0.7, + "learning_rate": 1.2990808217673123e-07, + "loss": 0.6869, + "step": 19190 + }, + { + "epoch": 0.7, + "learning_rate": 1.299044201120592e-07, + "loss": 0.6667, + "step": 19191 + }, + { + "epoch": 0.7, + "learning_rate": 1.2990075804738711e-07, + "loss": 0.6893, + "step": 19192 + }, + { + "epoch": 0.7, + "learning_rate": 1.2989709598271507e-07, + "loss": 0.7027, + "step": 19193 + }, + { + "epoch": 0.7, + "learning_rate": 1.29893433918043e-07, + "loss": 0.7008, + "step": 19194 + }, + { + "epoch": 0.7, + "learning_rate": 1.2988977185337092e-07, + "loss": 0.6938, + "step": 19195 + }, + { + "epoch": 0.7, + "learning_rate": 1.2988610978869887e-07, + "loss": 0.6986, + "step": 19196 + }, + { + "epoch": 0.7, + "learning_rate": 1.298824477240268e-07, + "loss": 0.7071, + "step": 19197 + }, + { + "epoch": 0.7, + "learning_rate": 1.2987878565935473e-07, + "loss": 0.6868, + "step": 19198 + }, + { + "epoch": 0.7, + "learning_rate": 1.2987512359468265e-07, + "loss": 0.6885, + "step": 19199 + }, + { + "epoch": 0.7, + "learning_rate": 1.298714615300106e-07, + "loss": 0.695, + "step": 19200 + }, + { + "epoch": 0.7, + "learning_rate": 1.2986779946533853e-07, + "loss": 0.694, + "step": 19201 + }, + { + "epoch": 0.7, + "learning_rate": 1.298641374006665e-07, + "loss": 0.6856, + "step": 19202 + }, + { + "epoch": 0.7, + "learning_rate": 1.2986047533599441e-07, + "loss": 0.6982, + "step": 19203 + }, + { + "epoch": 0.7, + "learning_rate": 1.2985681327132237e-07, + "loss": 0.684, + "step": 19204 + }, + { + "epoch": 0.7, + "learning_rate": 1.298531512066503e-07, + "loss": 0.695, + "step": 19205 + }, + { + "epoch": 0.7, + "learning_rate": 1.2984948914197825e-07, + "loss": 0.7052, + "step": 19206 + }, + { + "epoch": 0.7, + "learning_rate": 1.2984582707730617e-07, + "loss": 0.6964, + "step": 19207 + }, + { + "epoch": 0.7, + "learning_rate": 1.2984216501263413e-07, + "loss": 0.697, + "step": 19208 + }, + { + "epoch": 0.7, + "learning_rate": 1.2983850294796205e-07, + "loss": 0.7004, + "step": 19209 + }, + { + "epoch": 0.7, + "learning_rate": 1.2983484088329e-07, + "loss": 0.6848, + "step": 19210 + }, + { + "epoch": 0.7, + "learning_rate": 1.2983117881861793e-07, + "loss": 0.6861, + "step": 19211 + }, + { + "epoch": 0.7, + "learning_rate": 1.2982751675394589e-07, + "loss": 0.6955, + "step": 19212 + }, + { + "epoch": 0.7, + "learning_rate": 1.2982385468927381e-07, + "loss": 0.6901, + "step": 19213 + }, + { + "epoch": 0.7, + "learning_rate": 1.2982019262460177e-07, + "loss": 0.6977, + "step": 19214 + }, + { + "epoch": 0.7, + "learning_rate": 1.298165305599297e-07, + "loss": 0.6953, + "step": 19215 + }, + { + "epoch": 0.7, + "learning_rate": 1.2981286849525762e-07, + "loss": 0.6835, + "step": 19216 + }, + { + "epoch": 0.7, + "learning_rate": 1.2980920643058555e-07, + "loss": 0.7044, + "step": 19217 + }, + { + "epoch": 0.7, + "learning_rate": 1.298055443659135e-07, + "loss": 0.6909, + "step": 19218 + }, + { + "epoch": 0.7, + "learning_rate": 1.2980188230124143e-07, + "loss": 0.694, + "step": 19219 + }, + { + "epoch": 0.7, + "learning_rate": 1.2979822023656938e-07, + "loss": 0.6796, + "step": 19220 + }, + { + "epoch": 0.7, + "learning_rate": 1.297945581718973e-07, + "loss": 0.696, + "step": 19221 + }, + { + "epoch": 0.7, + "learning_rate": 1.2979089610722523e-07, + "loss": 0.6951, + "step": 19222 + }, + { + "epoch": 0.7, + "learning_rate": 1.2978723404255319e-07, + "loss": 0.6944, + "step": 19223 + }, + { + "epoch": 0.7, + "learning_rate": 1.297835719778811e-07, + "loss": 0.6861, + "step": 19224 + }, + { + "epoch": 0.7, + "learning_rate": 1.2977990991320907e-07, + "loss": 0.6933, + "step": 19225 + }, + { + "epoch": 0.7, + "learning_rate": 1.29776247848537e-07, + "loss": 0.6799, + "step": 19226 + }, + { + "epoch": 0.7, + "learning_rate": 1.2977258578386495e-07, + "loss": 0.6832, + "step": 19227 + }, + { + "epoch": 0.7, + "learning_rate": 1.2976892371919287e-07, + "loss": 0.6873, + "step": 19228 + }, + { + "epoch": 0.7, + "learning_rate": 1.2976526165452083e-07, + "loss": 0.6976, + "step": 19229 + }, + { + "epoch": 0.7, + "learning_rate": 1.2976159958984875e-07, + "loss": 0.6967, + "step": 19230 + }, + { + "epoch": 0.7, + "learning_rate": 1.297579375251767e-07, + "loss": 0.7013, + "step": 19231 + }, + { + "epoch": 0.7, + "learning_rate": 1.2975427546050463e-07, + "loss": 0.6979, + "step": 19232 + }, + { + "epoch": 0.7, + "learning_rate": 1.2975061339583259e-07, + "loss": 0.6939, + "step": 19233 + }, + { + "epoch": 0.7, + "learning_rate": 1.2974695133116049e-07, + "loss": 0.6923, + "step": 19234 + }, + { + "epoch": 0.7, + "learning_rate": 1.2974328926648844e-07, + "loss": 0.6929, + "step": 19235 + }, + { + "epoch": 0.7, + "learning_rate": 1.2973962720181637e-07, + "loss": 0.6954, + "step": 19236 + }, + { + "epoch": 0.7, + "learning_rate": 1.2973596513714432e-07, + "loss": 0.6794, + "step": 19237 + }, + { + "epoch": 0.7, + "learning_rate": 1.2973230307247225e-07, + "loss": 0.6922, + "step": 19238 + }, + { + "epoch": 0.7, + "learning_rate": 1.297286410078002e-07, + "loss": 0.6794, + "step": 19239 + }, + { + "epoch": 0.7, + "learning_rate": 1.2972497894312813e-07, + "loss": 0.6867, + "step": 19240 + }, + { + "epoch": 0.7, + "learning_rate": 1.2972131687845608e-07, + "loss": 0.7101, + "step": 19241 + }, + { + "epoch": 0.7, + "learning_rate": 1.29717654813784e-07, + "loss": 0.6933, + "step": 19242 + }, + { + "epoch": 0.7, + "learning_rate": 1.2971399274911193e-07, + "loss": 0.6778, + "step": 19243 + }, + { + "epoch": 0.7, + "learning_rate": 1.2971033068443988e-07, + "loss": 0.6996, + "step": 19244 + }, + { + "epoch": 0.7, + "learning_rate": 1.297066686197678e-07, + "loss": 0.6891, + "step": 19245 + }, + { + "epoch": 0.7, + "learning_rate": 1.2970300655509576e-07, + "loss": 0.7105, + "step": 19246 + }, + { + "epoch": 0.7, + "learning_rate": 1.296993444904237e-07, + "loss": 0.6987, + "step": 19247 + }, + { + "epoch": 0.7, + "learning_rate": 1.2969568242575164e-07, + "loss": 0.7019, + "step": 19248 + }, + { + "epoch": 0.7, + "learning_rate": 1.2969202036107957e-07, + "loss": 0.6897, + "step": 19249 + }, + { + "epoch": 0.7, + "learning_rate": 1.2968835829640752e-07, + "loss": 0.6921, + "step": 19250 + }, + { + "epoch": 0.7, + "learning_rate": 1.2968469623173545e-07, + "loss": 0.7102, + "step": 19251 + }, + { + "epoch": 0.7, + "learning_rate": 1.2968103416706338e-07, + "loss": 0.6881, + "step": 19252 + }, + { + "epoch": 0.7, + "learning_rate": 1.296773721023913e-07, + "loss": 0.7012, + "step": 19253 + }, + { + "epoch": 0.7, + "learning_rate": 1.2967371003771926e-07, + "loss": 0.6872, + "step": 19254 + }, + { + "epoch": 0.7, + "learning_rate": 1.2967004797304718e-07, + "loss": 0.71, + "step": 19255 + }, + { + "epoch": 0.7, + "learning_rate": 1.2966638590837514e-07, + "loss": 0.7044, + "step": 19256 + }, + { + "epoch": 0.7, + "learning_rate": 1.2966272384370306e-07, + "loss": 0.7008, + "step": 19257 + }, + { + "epoch": 0.7, + "learning_rate": 1.2965906177903102e-07, + "loss": 0.7008, + "step": 19258 + }, + { + "epoch": 0.7, + "learning_rate": 1.2965539971435894e-07, + "loss": 0.6931, + "step": 19259 + }, + { + "epoch": 0.7, + "learning_rate": 1.296517376496869e-07, + "loss": 0.6965, + "step": 19260 + }, + { + "epoch": 0.7, + "learning_rate": 1.2964807558501482e-07, + "loss": 0.6924, + "step": 19261 + }, + { + "epoch": 0.7, + "learning_rate": 1.2964441352034278e-07, + "loss": 0.6877, + "step": 19262 + }, + { + "epoch": 0.7, + "learning_rate": 1.296407514556707e-07, + "loss": 0.6939, + "step": 19263 + }, + { + "epoch": 0.7, + "learning_rate": 1.2963708939099866e-07, + "loss": 0.6909, + "step": 19264 + }, + { + "epoch": 0.7, + "learning_rate": 1.2963342732632658e-07, + "loss": 0.6997, + "step": 19265 + }, + { + "epoch": 0.7, + "learning_rate": 1.296297652616545e-07, + "loss": 0.6874, + "step": 19266 + }, + { + "epoch": 0.7, + "learning_rate": 1.2962610319698246e-07, + "loss": 0.6845, + "step": 19267 + }, + { + "epoch": 0.7, + "learning_rate": 1.296224411323104e-07, + "loss": 0.6999, + "step": 19268 + }, + { + "epoch": 0.7, + "learning_rate": 1.2961877906763834e-07, + "loss": 0.6827, + "step": 19269 + }, + { + "epoch": 0.71, + "learning_rate": 1.2961511700296624e-07, + "loss": 0.6845, + "step": 19270 + }, + { + "epoch": 0.71, + "learning_rate": 1.296114549382942e-07, + "loss": 0.6967, + "step": 19271 + }, + { + "epoch": 0.71, + "learning_rate": 1.2960779287362212e-07, + "loss": 0.6919, + "step": 19272 + }, + { + "epoch": 0.71, + "learning_rate": 1.2960413080895008e-07, + "loss": 0.6839, + "step": 19273 + }, + { + "epoch": 0.71, + "learning_rate": 1.29600468744278e-07, + "loss": 0.6956, + "step": 19274 + }, + { + "epoch": 0.71, + "learning_rate": 1.2959680667960596e-07, + "loss": 0.6959, + "step": 19275 + }, + { + "epoch": 0.71, + "learning_rate": 1.2959314461493388e-07, + "loss": 0.699, + "step": 19276 + }, + { + "epoch": 0.71, + "learning_rate": 1.2958948255026184e-07, + "loss": 0.6879, + "step": 19277 + }, + { + "epoch": 0.71, + "learning_rate": 1.2958582048558976e-07, + "loss": 0.699, + "step": 19278 + }, + { + "epoch": 0.71, + "learning_rate": 1.2958215842091772e-07, + "loss": 0.7047, + "step": 19279 + }, + { + "epoch": 0.71, + "learning_rate": 1.2957849635624564e-07, + "loss": 0.7033, + "step": 19280 + }, + { + "epoch": 0.71, + "learning_rate": 1.295748342915736e-07, + "loss": 0.696, + "step": 19281 + }, + { + "epoch": 0.71, + "learning_rate": 1.2957117222690152e-07, + "loss": 0.6848, + "step": 19282 + }, + { + "epoch": 0.71, + "learning_rate": 1.2956751016222948e-07, + "loss": 0.7007, + "step": 19283 + }, + { + "epoch": 0.71, + "learning_rate": 1.295638480975574e-07, + "loss": 0.6928, + "step": 19284 + }, + { + "epoch": 0.71, + "learning_rate": 1.2956018603288536e-07, + "loss": 0.6863, + "step": 19285 + }, + { + "epoch": 0.71, + "learning_rate": 1.2955652396821328e-07, + "loss": 0.6929, + "step": 19286 + }, + { + "epoch": 0.71, + "learning_rate": 1.295528619035412e-07, + "loss": 0.6783, + "step": 19287 + }, + { + "epoch": 0.71, + "learning_rate": 1.2954919983886914e-07, + "loss": 0.6929, + "step": 19288 + }, + { + "epoch": 0.71, + "learning_rate": 1.295455377741971e-07, + "loss": 0.7049, + "step": 19289 + }, + { + "epoch": 0.71, + "learning_rate": 1.2954187570952502e-07, + "loss": 0.6846, + "step": 19290 + }, + { + "epoch": 0.71, + "learning_rate": 1.2953821364485297e-07, + "loss": 0.6874, + "step": 19291 + }, + { + "epoch": 0.71, + "learning_rate": 1.295345515801809e-07, + "loss": 0.6813, + "step": 19292 + }, + { + "epoch": 0.71, + "learning_rate": 1.2953088951550882e-07, + "loss": 0.6803, + "step": 19293 + }, + { + "epoch": 0.71, + "learning_rate": 1.2952722745083678e-07, + "loss": 0.7026, + "step": 19294 + }, + { + "epoch": 0.71, + "learning_rate": 1.295235653861647e-07, + "loss": 0.705, + "step": 19295 + }, + { + "epoch": 0.71, + "learning_rate": 1.2951990332149266e-07, + "loss": 0.6945, + "step": 19296 + }, + { + "epoch": 0.71, + "learning_rate": 1.2951624125682058e-07, + "loss": 0.6871, + "step": 19297 + }, + { + "epoch": 0.71, + "learning_rate": 1.2951257919214854e-07, + "loss": 0.6969, + "step": 19298 + }, + { + "epoch": 0.71, + "learning_rate": 1.2950891712747646e-07, + "loss": 0.6895, + "step": 19299 + }, + { + "epoch": 0.71, + "learning_rate": 1.2950525506280442e-07, + "loss": 0.6889, + "step": 19300 + }, + { + "epoch": 0.71, + "learning_rate": 1.2950159299813234e-07, + "loss": 0.6836, + "step": 19301 + }, + { + "epoch": 0.71, + "learning_rate": 1.294979309334603e-07, + "loss": 0.684, + "step": 19302 + }, + { + "epoch": 0.71, + "learning_rate": 1.2949426886878822e-07, + "loss": 0.6958, + "step": 19303 + }, + { + "epoch": 0.71, + "learning_rate": 1.2949060680411617e-07, + "loss": 0.6807, + "step": 19304 + }, + { + "epoch": 0.71, + "learning_rate": 1.2948694473944408e-07, + "loss": 0.6983, + "step": 19305 + }, + { + "epoch": 0.71, + "learning_rate": 1.2948328267477203e-07, + "loss": 0.6982, + "step": 19306 + }, + { + "epoch": 0.71, + "learning_rate": 1.2947962061009995e-07, + "loss": 0.6949, + "step": 19307 + }, + { + "epoch": 0.71, + "learning_rate": 1.294759585454279e-07, + "loss": 0.69, + "step": 19308 + }, + { + "epoch": 0.71, + "learning_rate": 1.2947229648075583e-07, + "loss": 0.6782, + "step": 19309 + }, + { + "epoch": 0.71, + "learning_rate": 1.294686344160838e-07, + "loss": 0.6929, + "step": 19310 + }, + { + "epoch": 0.71, + "learning_rate": 1.2946497235141171e-07, + "loss": 0.6922, + "step": 19311 + }, + { + "epoch": 0.71, + "learning_rate": 1.2946131028673967e-07, + "loss": 0.6729, + "step": 19312 + }, + { + "epoch": 0.71, + "learning_rate": 1.294576482220676e-07, + "loss": 0.6914, + "step": 19313 + }, + { + "epoch": 0.71, + "learning_rate": 1.2945398615739552e-07, + "loss": 0.6876, + "step": 19314 + }, + { + "epoch": 0.71, + "learning_rate": 1.2945032409272347e-07, + "loss": 0.6872, + "step": 19315 + }, + { + "epoch": 0.71, + "learning_rate": 1.294466620280514e-07, + "loss": 0.6928, + "step": 19316 + }, + { + "epoch": 0.71, + "learning_rate": 1.2944299996337935e-07, + "loss": 0.6873, + "step": 19317 + }, + { + "epoch": 0.71, + "learning_rate": 1.2943933789870728e-07, + "loss": 0.6932, + "step": 19318 + }, + { + "epoch": 0.71, + "learning_rate": 1.2943567583403523e-07, + "loss": 0.6839, + "step": 19319 + }, + { + "epoch": 0.71, + "learning_rate": 1.2943201376936316e-07, + "loss": 0.6899, + "step": 19320 + }, + { + "epoch": 0.71, + "learning_rate": 1.2942835170469111e-07, + "loss": 0.6981, + "step": 19321 + }, + { + "epoch": 0.71, + "learning_rate": 1.2942468964001904e-07, + "loss": 0.6908, + "step": 19322 + }, + { + "epoch": 0.71, + "learning_rate": 1.2942102757534697e-07, + "loss": 0.6838, + "step": 19323 + }, + { + "epoch": 0.71, + "learning_rate": 1.294173655106749e-07, + "loss": 0.7023, + "step": 19324 + }, + { + "epoch": 0.71, + "learning_rate": 1.2941370344600285e-07, + "loss": 0.6916, + "step": 19325 + }, + { + "epoch": 0.71, + "learning_rate": 1.2941004138133077e-07, + "loss": 0.6861, + "step": 19326 + }, + { + "epoch": 0.71, + "learning_rate": 1.2940637931665873e-07, + "loss": 0.698, + "step": 19327 + }, + { + "epoch": 0.71, + "learning_rate": 1.2940271725198665e-07, + "loss": 0.6791, + "step": 19328 + }, + { + "epoch": 0.71, + "learning_rate": 1.293990551873146e-07, + "loss": 0.7032, + "step": 19329 + }, + { + "epoch": 0.71, + "learning_rate": 1.2939539312264253e-07, + "loss": 0.6986, + "step": 19330 + }, + { + "epoch": 0.71, + "learning_rate": 1.2939173105797049e-07, + "loss": 0.6985, + "step": 19331 + }, + { + "epoch": 0.71, + "learning_rate": 1.2938806899329841e-07, + "loss": 0.6921, + "step": 19332 + }, + { + "epoch": 0.71, + "learning_rate": 1.2938440692862637e-07, + "loss": 0.6978, + "step": 19333 + }, + { + "epoch": 0.71, + "learning_rate": 1.293807448639543e-07, + "loss": 0.699, + "step": 19334 + }, + { + "epoch": 0.71, + "learning_rate": 1.2937708279928225e-07, + "loss": 0.7049, + "step": 19335 + }, + { + "epoch": 0.71, + "learning_rate": 1.2937342073461017e-07, + "loss": 0.6904, + "step": 19336 + }, + { + "epoch": 0.71, + "learning_rate": 1.293697586699381e-07, + "loss": 0.6919, + "step": 19337 + }, + { + "epoch": 0.71, + "learning_rate": 1.2936609660526605e-07, + "loss": 0.7142, + "step": 19338 + }, + { + "epoch": 0.71, + "learning_rate": 1.2936243454059398e-07, + "loss": 0.689, + "step": 19339 + }, + { + "epoch": 0.71, + "learning_rate": 1.2935877247592193e-07, + "loss": 0.7008, + "step": 19340 + }, + { + "epoch": 0.71, + "learning_rate": 1.2935511041124983e-07, + "loss": 0.6896, + "step": 19341 + }, + { + "epoch": 0.71, + "learning_rate": 1.2935144834657779e-07, + "loss": 0.6918, + "step": 19342 + }, + { + "epoch": 0.71, + "learning_rate": 1.293477862819057e-07, + "loss": 0.7007, + "step": 19343 + }, + { + "epoch": 0.71, + "learning_rate": 1.2934412421723367e-07, + "loss": 0.6956, + "step": 19344 + }, + { + "epoch": 0.71, + "learning_rate": 1.293404621525616e-07, + "loss": 0.6978, + "step": 19345 + }, + { + "epoch": 0.71, + "learning_rate": 1.2933680008788955e-07, + "loss": 0.6731, + "step": 19346 + }, + { + "epoch": 0.71, + "learning_rate": 1.2933313802321747e-07, + "loss": 0.691, + "step": 19347 + }, + { + "epoch": 0.71, + "learning_rate": 1.2932947595854543e-07, + "loss": 0.7067, + "step": 19348 + }, + { + "epoch": 0.71, + "learning_rate": 1.2932581389387335e-07, + "loss": 0.6999, + "step": 19349 + }, + { + "epoch": 0.71, + "learning_rate": 1.293221518292013e-07, + "loss": 0.6878, + "step": 19350 + }, + { + "epoch": 0.71, + "learning_rate": 1.2931848976452923e-07, + "loss": 0.6865, + "step": 19351 + }, + { + "epoch": 0.71, + "learning_rate": 1.2931482769985719e-07, + "loss": 0.6974, + "step": 19352 + }, + { + "epoch": 0.71, + "learning_rate": 1.293111656351851e-07, + "loss": 0.6927, + "step": 19353 + }, + { + "epoch": 0.71, + "learning_rate": 1.2930750357051307e-07, + "loss": 0.6996, + "step": 19354 + }, + { + "epoch": 0.71, + "learning_rate": 1.29303841505841e-07, + "loss": 0.6975, + "step": 19355 + }, + { + "epoch": 0.71, + "learning_rate": 1.2930017944116895e-07, + "loss": 0.6949, + "step": 19356 + }, + { + "epoch": 0.71, + "learning_rate": 1.2929651737649687e-07, + "loss": 0.7017, + "step": 19357 + }, + { + "epoch": 0.71, + "learning_rate": 1.292928553118248e-07, + "loss": 0.7069, + "step": 19358 + }, + { + "epoch": 0.71, + "learning_rate": 1.2928919324715273e-07, + "loss": 0.6874, + "step": 19359 + }, + { + "epoch": 0.71, + "learning_rate": 1.2928553118248068e-07, + "loss": 0.6894, + "step": 19360 + }, + { + "epoch": 0.71, + "learning_rate": 1.292818691178086e-07, + "loss": 0.6863, + "step": 19361 + }, + { + "epoch": 0.71, + "learning_rate": 1.2927820705313656e-07, + "loss": 0.6832, + "step": 19362 + }, + { + "epoch": 0.71, + "learning_rate": 1.2927454498846449e-07, + "loss": 0.6895, + "step": 19363 + }, + { + "epoch": 0.71, + "learning_rate": 1.292708829237924e-07, + "loss": 0.7001, + "step": 19364 + }, + { + "epoch": 0.71, + "learning_rate": 1.2926722085912036e-07, + "loss": 0.7015, + "step": 19365 + }, + { + "epoch": 0.71, + "learning_rate": 1.292635587944483e-07, + "loss": 0.6769, + "step": 19366 + }, + { + "epoch": 0.71, + "learning_rate": 1.2925989672977624e-07, + "loss": 0.7048, + "step": 19367 + }, + { + "epoch": 0.71, + "learning_rate": 1.2925623466510417e-07, + "loss": 0.6952, + "step": 19368 + }, + { + "epoch": 0.71, + "learning_rate": 1.2925257260043212e-07, + "loss": 0.7123, + "step": 19369 + }, + { + "epoch": 0.71, + "learning_rate": 1.2924891053576005e-07, + "loss": 0.692, + "step": 19370 + }, + { + "epoch": 0.71, + "learning_rate": 1.29245248471088e-07, + "loss": 0.6944, + "step": 19371 + }, + { + "epoch": 0.71, + "learning_rate": 1.2924158640641593e-07, + "loss": 0.6916, + "step": 19372 + }, + { + "epoch": 0.71, + "learning_rate": 1.2923792434174388e-07, + "loss": 0.6894, + "step": 19373 + }, + { + "epoch": 0.71, + "learning_rate": 1.292342622770718e-07, + "loss": 0.6953, + "step": 19374 + }, + { + "epoch": 0.71, + "learning_rate": 1.2923060021239976e-07, + "loss": 0.6865, + "step": 19375 + }, + { + "epoch": 0.71, + "learning_rate": 1.292269381477277e-07, + "loss": 0.6969, + "step": 19376 + }, + { + "epoch": 0.71, + "learning_rate": 1.2922327608305562e-07, + "loss": 0.705, + "step": 19377 + }, + { + "epoch": 0.71, + "learning_rate": 1.2921961401838354e-07, + "loss": 0.6915, + "step": 19378 + }, + { + "epoch": 0.71, + "learning_rate": 1.292159519537115e-07, + "loss": 0.6897, + "step": 19379 + }, + { + "epoch": 0.71, + "learning_rate": 1.2921228988903942e-07, + "loss": 0.6943, + "step": 19380 + }, + { + "epoch": 0.71, + "learning_rate": 1.2920862782436738e-07, + "loss": 0.6913, + "step": 19381 + }, + { + "epoch": 0.71, + "learning_rate": 1.292049657596953e-07, + "loss": 0.6994, + "step": 19382 + }, + { + "epoch": 0.71, + "learning_rate": 1.2920130369502326e-07, + "loss": 0.7073, + "step": 19383 + }, + { + "epoch": 0.71, + "learning_rate": 1.2919764163035118e-07, + "loss": 0.6943, + "step": 19384 + }, + { + "epoch": 0.71, + "learning_rate": 1.291939795656791e-07, + "loss": 0.6986, + "step": 19385 + }, + { + "epoch": 0.71, + "learning_rate": 1.2919031750100706e-07, + "loss": 0.6818, + "step": 19386 + }, + { + "epoch": 0.71, + "learning_rate": 1.29186655436335e-07, + "loss": 0.7168, + "step": 19387 + }, + { + "epoch": 0.71, + "learning_rate": 1.2918299337166294e-07, + "loss": 0.7033, + "step": 19388 + }, + { + "epoch": 0.71, + "learning_rate": 1.2917933130699087e-07, + "loss": 0.6971, + "step": 19389 + }, + { + "epoch": 0.71, + "learning_rate": 1.2917566924231882e-07, + "loss": 0.7056, + "step": 19390 + }, + { + "epoch": 0.71, + "learning_rate": 1.2917200717764675e-07, + "loss": 0.692, + "step": 19391 + }, + { + "epoch": 0.71, + "learning_rate": 1.291683451129747e-07, + "loss": 0.6821, + "step": 19392 + }, + { + "epoch": 0.71, + "learning_rate": 1.2916468304830263e-07, + "loss": 0.7028, + "step": 19393 + }, + { + "epoch": 0.71, + "learning_rate": 1.2916102098363058e-07, + "loss": 0.6925, + "step": 19394 + }, + { + "epoch": 0.71, + "learning_rate": 1.2915735891895848e-07, + "loss": 0.6948, + "step": 19395 + }, + { + "epoch": 0.71, + "learning_rate": 1.2915369685428644e-07, + "loss": 0.7007, + "step": 19396 + }, + { + "epoch": 0.71, + "learning_rate": 1.2915003478961436e-07, + "loss": 0.6873, + "step": 19397 + }, + { + "epoch": 0.71, + "learning_rate": 1.2914637272494232e-07, + "loss": 0.7096, + "step": 19398 + }, + { + "epoch": 0.71, + "learning_rate": 1.2914271066027024e-07, + "loss": 0.6955, + "step": 19399 + }, + { + "epoch": 0.71, + "learning_rate": 1.291390485955982e-07, + "loss": 0.7023, + "step": 19400 + }, + { + "epoch": 0.71, + "learning_rate": 1.2913538653092612e-07, + "loss": 0.6907, + "step": 19401 + }, + { + "epoch": 0.71, + "learning_rate": 1.2913172446625408e-07, + "loss": 0.689, + "step": 19402 + }, + { + "epoch": 0.71, + "learning_rate": 1.29128062401582e-07, + "loss": 0.7004, + "step": 19403 + }, + { + "epoch": 0.71, + "learning_rate": 1.2912440033690996e-07, + "loss": 0.6855, + "step": 19404 + }, + { + "epoch": 0.71, + "learning_rate": 1.2912073827223788e-07, + "loss": 0.702, + "step": 19405 + }, + { + "epoch": 0.71, + "learning_rate": 1.2911707620756584e-07, + "loss": 0.689, + "step": 19406 + }, + { + "epoch": 0.71, + "learning_rate": 1.2911341414289376e-07, + "loss": 0.6952, + "step": 19407 + }, + { + "epoch": 0.71, + "learning_rate": 1.291097520782217e-07, + "loss": 0.6971, + "step": 19408 + }, + { + "epoch": 0.71, + "learning_rate": 1.2910609001354964e-07, + "loss": 0.6896, + "step": 19409 + }, + { + "epoch": 0.71, + "learning_rate": 1.2910242794887757e-07, + "loss": 0.7071, + "step": 19410 + }, + { + "epoch": 0.71, + "learning_rate": 1.2909876588420552e-07, + "loss": 0.6894, + "step": 19411 + }, + { + "epoch": 0.71, + "learning_rate": 1.2909510381953345e-07, + "loss": 0.6941, + "step": 19412 + }, + { + "epoch": 0.71, + "learning_rate": 1.2909144175486138e-07, + "loss": 0.7004, + "step": 19413 + }, + { + "epoch": 0.71, + "learning_rate": 1.290877796901893e-07, + "loss": 0.6851, + "step": 19414 + }, + { + "epoch": 0.71, + "learning_rate": 1.2908411762551726e-07, + "loss": 0.6865, + "step": 19415 + }, + { + "epoch": 0.71, + "learning_rate": 1.2908045556084518e-07, + "loss": 0.6971, + "step": 19416 + }, + { + "epoch": 0.71, + "learning_rate": 1.2907679349617314e-07, + "loss": 0.6825, + "step": 19417 + }, + { + "epoch": 0.71, + "learning_rate": 1.2907313143150106e-07, + "loss": 0.69, + "step": 19418 + }, + { + "epoch": 0.71, + "learning_rate": 1.2906946936682902e-07, + "loss": 0.6976, + "step": 19419 + }, + { + "epoch": 0.71, + "learning_rate": 1.2906580730215694e-07, + "loss": 0.6945, + "step": 19420 + }, + { + "epoch": 0.71, + "learning_rate": 1.290621452374849e-07, + "loss": 0.7, + "step": 19421 + }, + { + "epoch": 0.71, + "learning_rate": 1.2905848317281282e-07, + "loss": 0.6956, + "step": 19422 + }, + { + "epoch": 0.71, + "learning_rate": 1.2905482110814077e-07, + "loss": 0.6917, + "step": 19423 + }, + { + "epoch": 0.71, + "learning_rate": 1.290511590434687e-07, + "loss": 0.6882, + "step": 19424 + }, + { + "epoch": 0.71, + "learning_rate": 1.2904749697879665e-07, + "loss": 0.6917, + "step": 19425 + }, + { + "epoch": 0.71, + "learning_rate": 1.2904383491412458e-07, + "loss": 0.6922, + "step": 19426 + }, + { + "epoch": 0.71, + "learning_rate": 1.2904017284945253e-07, + "loss": 0.6964, + "step": 19427 + }, + { + "epoch": 0.71, + "learning_rate": 1.2903651078478046e-07, + "loss": 0.6977, + "step": 19428 + }, + { + "epoch": 0.71, + "learning_rate": 1.290328487201084e-07, + "loss": 0.6786, + "step": 19429 + }, + { + "epoch": 0.71, + "learning_rate": 1.2902918665543634e-07, + "loss": 0.6902, + "step": 19430 + }, + { + "epoch": 0.71, + "learning_rate": 1.2902552459076427e-07, + "loss": 0.6782, + "step": 19431 + }, + { + "epoch": 0.71, + "learning_rate": 1.290218625260922e-07, + "loss": 0.6947, + "step": 19432 + }, + { + "epoch": 0.71, + "learning_rate": 1.2901820046142015e-07, + "loss": 0.711, + "step": 19433 + }, + { + "epoch": 0.71, + "learning_rate": 1.2901453839674807e-07, + "loss": 0.6882, + "step": 19434 + }, + { + "epoch": 0.71, + "learning_rate": 1.29010876332076e-07, + "loss": 0.6787, + "step": 19435 + }, + { + "epoch": 0.71, + "learning_rate": 1.2900721426740395e-07, + "loss": 0.6816, + "step": 19436 + }, + { + "epoch": 0.71, + "learning_rate": 1.2900355220273188e-07, + "loss": 0.6896, + "step": 19437 + }, + { + "epoch": 0.71, + "learning_rate": 1.2899989013805983e-07, + "loss": 0.6931, + "step": 19438 + }, + { + "epoch": 0.71, + "learning_rate": 1.2899622807338776e-07, + "loss": 0.6794, + "step": 19439 + }, + { + "epoch": 0.71, + "learning_rate": 1.2899256600871571e-07, + "loss": 0.689, + "step": 19440 + }, + { + "epoch": 0.71, + "learning_rate": 1.2898890394404364e-07, + "loss": 0.7046, + "step": 19441 + }, + { + "epoch": 0.71, + "learning_rate": 1.289852418793716e-07, + "loss": 0.6955, + "step": 19442 + }, + { + "epoch": 0.71, + "learning_rate": 1.2898157981469952e-07, + "loss": 0.6942, + "step": 19443 + }, + { + "epoch": 0.71, + "learning_rate": 1.2897791775002747e-07, + "loss": 0.6928, + "step": 19444 + }, + { + "epoch": 0.71, + "learning_rate": 1.289742556853554e-07, + "loss": 0.6813, + "step": 19445 + }, + { + "epoch": 0.71, + "learning_rate": 1.2897059362068335e-07, + "loss": 0.6951, + "step": 19446 + }, + { + "epoch": 0.71, + "learning_rate": 1.2896693155601128e-07, + "loss": 0.7029, + "step": 19447 + }, + { + "epoch": 0.71, + "learning_rate": 1.2896326949133923e-07, + "loss": 0.6907, + "step": 19448 + }, + { + "epoch": 0.71, + "learning_rate": 1.2895960742666713e-07, + "loss": 0.6905, + "step": 19449 + }, + { + "epoch": 0.71, + "learning_rate": 1.2895594536199509e-07, + "loss": 0.7072, + "step": 19450 + }, + { + "epoch": 0.71, + "learning_rate": 1.2895228329732301e-07, + "loss": 0.6983, + "step": 19451 + }, + { + "epoch": 0.71, + "learning_rate": 1.2894862123265097e-07, + "loss": 0.6962, + "step": 19452 + }, + { + "epoch": 0.71, + "learning_rate": 1.289449591679789e-07, + "loss": 0.6864, + "step": 19453 + }, + { + "epoch": 0.71, + "learning_rate": 1.2894129710330685e-07, + "loss": 0.7062, + "step": 19454 + }, + { + "epoch": 0.71, + "learning_rate": 1.2893763503863477e-07, + "loss": 0.7042, + "step": 19455 + }, + { + "epoch": 0.71, + "learning_rate": 1.289339729739627e-07, + "loss": 0.6829, + "step": 19456 + }, + { + "epoch": 0.71, + "learning_rate": 1.2893031090929065e-07, + "loss": 0.6973, + "step": 19457 + }, + { + "epoch": 0.71, + "learning_rate": 1.2892664884461858e-07, + "loss": 0.6848, + "step": 19458 + }, + { + "epoch": 0.71, + "learning_rate": 1.2892298677994653e-07, + "loss": 0.6875, + "step": 19459 + }, + { + "epoch": 0.71, + "learning_rate": 1.2891932471527446e-07, + "loss": 0.6859, + "step": 19460 + }, + { + "epoch": 0.71, + "learning_rate": 1.289156626506024e-07, + "loss": 0.6911, + "step": 19461 + }, + { + "epoch": 0.71, + "learning_rate": 1.2891200058593034e-07, + "loss": 0.7052, + "step": 19462 + }, + { + "epoch": 0.71, + "learning_rate": 1.289083385212583e-07, + "loss": 0.6927, + "step": 19463 + }, + { + "epoch": 0.71, + "learning_rate": 1.2890467645658622e-07, + "loss": 0.6879, + "step": 19464 + }, + { + "epoch": 0.71, + "learning_rate": 1.2890101439191417e-07, + "loss": 0.6939, + "step": 19465 + }, + { + "epoch": 0.71, + "learning_rate": 1.288973523272421e-07, + "loss": 0.6944, + "step": 19466 + }, + { + "epoch": 0.71, + "learning_rate": 1.2889369026257003e-07, + "loss": 0.6892, + "step": 19467 + }, + { + "epoch": 0.71, + "learning_rate": 1.2889002819789795e-07, + "loss": 0.689, + "step": 19468 + }, + { + "epoch": 0.71, + "learning_rate": 1.288863661332259e-07, + "loss": 0.6988, + "step": 19469 + }, + { + "epoch": 0.71, + "learning_rate": 1.2888270406855383e-07, + "loss": 0.6922, + "step": 19470 + }, + { + "epoch": 0.71, + "learning_rate": 1.2887904200388179e-07, + "loss": 0.6897, + "step": 19471 + }, + { + "epoch": 0.71, + "learning_rate": 1.288753799392097e-07, + "loss": 0.6939, + "step": 19472 + }, + { + "epoch": 0.71, + "learning_rate": 1.2887171787453767e-07, + "loss": 0.6796, + "step": 19473 + }, + { + "epoch": 0.71, + "learning_rate": 1.288680558098656e-07, + "loss": 0.691, + "step": 19474 + }, + { + "epoch": 0.71, + "learning_rate": 1.2886439374519355e-07, + "loss": 0.6986, + "step": 19475 + }, + { + "epoch": 0.71, + "learning_rate": 1.2886073168052147e-07, + "loss": 0.6873, + "step": 19476 + }, + { + "epoch": 0.71, + "learning_rate": 1.2885706961584943e-07, + "loss": 0.6991, + "step": 19477 + }, + { + "epoch": 0.71, + "learning_rate": 1.2885340755117735e-07, + "loss": 0.6928, + "step": 19478 + }, + { + "epoch": 0.71, + "learning_rate": 1.2884974548650528e-07, + "loss": 0.6872, + "step": 19479 + }, + { + "epoch": 0.71, + "learning_rate": 1.2884608342183323e-07, + "loss": 0.6723, + "step": 19480 + }, + { + "epoch": 0.71, + "learning_rate": 1.2884242135716116e-07, + "loss": 0.7038, + "step": 19481 + }, + { + "epoch": 0.71, + "learning_rate": 1.288387592924891e-07, + "loss": 0.7042, + "step": 19482 + }, + { + "epoch": 0.71, + "learning_rate": 1.2883509722781704e-07, + "loss": 0.6977, + "step": 19483 + }, + { + "epoch": 0.71, + "learning_rate": 1.28831435163145e-07, + "loss": 0.6952, + "step": 19484 + }, + { + "epoch": 0.71, + "learning_rate": 1.288277730984729e-07, + "loss": 0.6959, + "step": 19485 + }, + { + "epoch": 0.71, + "learning_rate": 1.2882411103380084e-07, + "loss": 0.6907, + "step": 19486 + }, + { + "epoch": 0.71, + "learning_rate": 1.2882044896912877e-07, + "loss": 0.6968, + "step": 19487 + }, + { + "epoch": 0.71, + "learning_rate": 1.2881678690445672e-07, + "loss": 0.676, + "step": 19488 + }, + { + "epoch": 0.71, + "learning_rate": 1.2881312483978465e-07, + "loss": 0.6887, + "step": 19489 + }, + { + "epoch": 0.71, + "learning_rate": 1.288094627751126e-07, + "loss": 0.7008, + "step": 19490 + }, + { + "epoch": 0.71, + "learning_rate": 1.2880580071044053e-07, + "loss": 0.6979, + "step": 19491 + }, + { + "epoch": 0.71, + "learning_rate": 1.2880213864576848e-07, + "loss": 0.6873, + "step": 19492 + }, + { + "epoch": 0.71, + "learning_rate": 1.287984765810964e-07, + "loss": 0.6914, + "step": 19493 + }, + { + "epoch": 0.71, + "learning_rate": 1.2879481451642436e-07, + "loss": 0.6882, + "step": 19494 + }, + { + "epoch": 0.71, + "learning_rate": 1.287911524517523e-07, + "loss": 0.7027, + "step": 19495 + }, + { + "epoch": 0.71, + "learning_rate": 1.2878749038708024e-07, + "loss": 0.6886, + "step": 19496 + }, + { + "epoch": 0.71, + "learning_rate": 1.2878382832240817e-07, + "loss": 0.7128, + "step": 19497 + }, + { + "epoch": 0.71, + "learning_rate": 1.2878016625773612e-07, + "loss": 0.6934, + "step": 19498 + }, + { + "epoch": 0.71, + "learning_rate": 1.2877650419306405e-07, + "loss": 0.6863, + "step": 19499 + }, + { + "epoch": 0.71, + "learning_rate": 1.2877284212839198e-07, + "loss": 0.6812, + "step": 19500 + }, + { + "epoch": 0.71, + "learning_rate": 1.2876918006371993e-07, + "loss": 0.6808, + "step": 19501 + }, + { + "epoch": 0.71, + "learning_rate": 1.2876551799904786e-07, + "loss": 0.6957, + "step": 19502 + }, + { + "epoch": 0.71, + "learning_rate": 1.2876185593437578e-07, + "loss": 0.679, + "step": 19503 + }, + { + "epoch": 0.71, + "learning_rate": 1.2875819386970374e-07, + "loss": 0.6937, + "step": 19504 + }, + { + "epoch": 0.71, + "learning_rate": 1.2875453180503166e-07, + "loss": 0.6961, + "step": 19505 + }, + { + "epoch": 0.71, + "learning_rate": 1.287508697403596e-07, + "loss": 0.6948, + "step": 19506 + }, + { + "epoch": 0.71, + "learning_rate": 1.2874720767568754e-07, + "loss": 0.7081, + "step": 19507 + }, + { + "epoch": 0.71, + "learning_rate": 1.2874354561101547e-07, + "loss": 0.6859, + "step": 19508 + }, + { + "epoch": 0.71, + "learning_rate": 1.2873988354634342e-07, + "loss": 0.69, + "step": 19509 + }, + { + "epoch": 0.71, + "learning_rate": 1.2873622148167135e-07, + "loss": 0.697, + "step": 19510 + }, + { + "epoch": 0.71, + "learning_rate": 1.287325594169993e-07, + "loss": 0.6893, + "step": 19511 + }, + { + "epoch": 0.71, + "learning_rate": 1.2872889735232723e-07, + "loss": 0.6971, + "step": 19512 + }, + { + "epoch": 0.71, + "learning_rate": 1.2872523528765518e-07, + "loss": 0.6996, + "step": 19513 + }, + { + "epoch": 0.71, + "learning_rate": 1.287215732229831e-07, + "loss": 0.6752, + "step": 19514 + }, + { + "epoch": 0.71, + "learning_rate": 1.2871791115831106e-07, + "loss": 0.705, + "step": 19515 + }, + { + "epoch": 0.71, + "learning_rate": 1.28714249093639e-07, + "loss": 0.6924, + "step": 19516 + }, + { + "epoch": 0.71, + "learning_rate": 1.2871058702896694e-07, + "loss": 0.6969, + "step": 19517 + }, + { + "epoch": 0.71, + "learning_rate": 1.2870692496429487e-07, + "loss": 0.6821, + "step": 19518 + }, + { + "epoch": 0.71, + "learning_rate": 1.2870326289962282e-07, + "loss": 0.6914, + "step": 19519 + }, + { + "epoch": 0.71, + "learning_rate": 1.2869960083495075e-07, + "loss": 0.692, + "step": 19520 + }, + { + "epoch": 0.71, + "learning_rate": 1.2869593877027868e-07, + "loss": 0.698, + "step": 19521 + }, + { + "epoch": 0.71, + "learning_rate": 1.286922767056066e-07, + "loss": 0.686, + "step": 19522 + }, + { + "epoch": 0.71, + "learning_rate": 1.2868861464093456e-07, + "loss": 0.6889, + "step": 19523 + }, + { + "epoch": 0.71, + "learning_rate": 1.2868495257626248e-07, + "loss": 0.6889, + "step": 19524 + }, + { + "epoch": 0.71, + "learning_rate": 1.2868129051159044e-07, + "loss": 0.6986, + "step": 19525 + }, + { + "epoch": 0.71, + "learning_rate": 1.2867762844691836e-07, + "loss": 0.6893, + "step": 19526 + }, + { + "epoch": 0.71, + "learning_rate": 1.286739663822463e-07, + "loss": 0.7025, + "step": 19527 + }, + { + "epoch": 0.71, + "learning_rate": 1.2867030431757424e-07, + "loss": 0.6895, + "step": 19528 + }, + { + "epoch": 0.71, + "learning_rate": 1.2866664225290217e-07, + "loss": 0.6867, + "step": 19529 + }, + { + "epoch": 0.71, + "learning_rate": 1.2866298018823012e-07, + "loss": 0.7019, + "step": 19530 + }, + { + "epoch": 0.71, + "learning_rate": 1.2865931812355805e-07, + "loss": 0.698, + "step": 19531 + }, + { + "epoch": 0.71, + "learning_rate": 1.28655656058886e-07, + "loss": 0.6842, + "step": 19532 + }, + { + "epoch": 0.71, + "learning_rate": 1.2865199399421393e-07, + "loss": 0.6955, + "step": 19533 + }, + { + "epoch": 0.71, + "learning_rate": 1.2864833192954188e-07, + "loss": 0.6744, + "step": 19534 + }, + { + "epoch": 0.71, + "learning_rate": 1.286446698648698e-07, + "loss": 0.6936, + "step": 19535 + }, + { + "epoch": 0.71, + "learning_rate": 1.2864100780019776e-07, + "loss": 0.691, + "step": 19536 + }, + { + "epoch": 0.71, + "learning_rate": 1.286373457355257e-07, + "loss": 0.6992, + "step": 19537 + }, + { + "epoch": 0.71, + "learning_rate": 1.2863368367085364e-07, + "loss": 0.7117, + "step": 19538 + }, + { + "epoch": 0.71, + "learning_rate": 1.2863002160618154e-07, + "loss": 0.6977, + "step": 19539 + }, + { + "epoch": 0.71, + "learning_rate": 1.286263595415095e-07, + "loss": 0.6919, + "step": 19540 + }, + { + "epoch": 0.71, + "learning_rate": 1.2862269747683742e-07, + "loss": 0.6801, + "step": 19541 + }, + { + "epoch": 0.71, + "learning_rate": 1.2861903541216537e-07, + "loss": 0.6873, + "step": 19542 + }, + { + "epoch": 0.72, + "learning_rate": 1.286153733474933e-07, + "loss": 0.7008, + "step": 19543 + }, + { + "epoch": 0.72, + "learning_rate": 1.2861171128282125e-07, + "loss": 0.6964, + "step": 19544 + }, + { + "epoch": 0.72, + "learning_rate": 1.2860804921814918e-07, + "loss": 0.6909, + "step": 19545 + }, + { + "epoch": 0.72, + "learning_rate": 1.2860438715347713e-07, + "loss": 0.6926, + "step": 19546 + }, + { + "epoch": 0.72, + "learning_rate": 1.2860072508880506e-07, + "loss": 0.6862, + "step": 19547 + }, + { + "epoch": 0.72, + "learning_rate": 1.28597063024133e-07, + "loss": 0.7218, + "step": 19548 + }, + { + "epoch": 0.72, + "learning_rate": 1.2859340095946094e-07, + "loss": 0.6923, + "step": 19549 + }, + { + "epoch": 0.72, + "learning_rate": 1.2858973889478887e-07, + "loss": 0.6813, + "step": 19550 + }, + { + "epoch": 0.72, + "learning_rate": 1.2858607683011682e-07, + "loss": 0.7062, + "step": 19551 + }, + { + "epoch": 0.72, + "learning_rate": 1.2858241476544475e-07, + "loss": 0.7152, + "step": 19552 + }, + { + "epoch": 0.72, + "learning_rate": 1.285787527007727e-07, + "loss": 0.6961, + "step": 19553 + }, + { + "epoch": 0.72, + "learning_rate": 1.2857509063610063e-07, + "loss": 0.7044, + "step": 19554 + }, + { + "epoch": 0.72, + "learning_rate": 1.2857142857142858e-07, + "loss": 0.7022, + "step": 19555 + }, + { + "epoch": 0.72, + "learning_rate": 1.285677665067565e-07, + "loss": 0.6923, + "step": 19556 + }, + { + "epoch": 0.72, + "learning_rate": 1.2856410444208443e-07, + "loss": 0.6907, + "step": 19557 + }, + { + "epoch": 0.72, + "learning_rate": 1.2856044237741236e-07, + "loss": 0.6723, + "step": 19558 + }, + { + "epoch": 0.72, + "learning_rate": 1.2855678031274031e-07, + "loss": 0.7029, + "step": 19559 + }, + { + "epoch": 0.72, + "learning_rate": 1.2855311824806824e-07, + "loss": 0.6916, + "step": 19560 + }, + { + "epoch": 0.72, + "learning_rate": 1.285494561833962e-07, + "loss": 0.6989, + "step": 19561 + }, + { + "epoch": 0.72, + "learning_rate": 1.2854579411872412e-07, + "loss": 0.7137, + "step": 19562 + }, + { + "epoch": 0.72, + "learning_rate": 1.2854213205405207e-07, + "loss": 0.6852, + "step": 19563 + }, + { + "epoch": 0.72, + "learning_rate": 1.2853846998938e-07, + "loss": 0.6836, + "step": 19564 + }, + { + "epoch": 0.72, + "learning_rate": 1.2853480792470795e-07, + "loss": 0.6934, + "step": 19565 + }, + { + "epoch": 0.72, + "learning_rate": 1.2853114586003588e-07, + "loss": 0.6904, + "step": 19566 + }, + { + "epoch": 0.72, + "learning_rate": 1.2852748379536383e-07, + "loss": 0.6922, + "step": 19567 + }, + { + "epoch": 0.72, + "learning_rate": 1.2852382173069176e-07, + "loss": 0.7014, + "step": 19568 + }, + { + "epoch": 0.72, + "learning_rate": 1.2852015966601971e-07, + "loss": 0.6774, + "step": 19569 + }, + { + "epoch": 0.72, + "learning_rate": 1.2851649760134764e-07, + "loss": 0.681, + "step": 19570 + }, + { + "epoch": 0.72, + "learning_rate": 1.2851283553667557e-07, + "loss": 0.699, + "step": 19571 + }, + { + "epoch": 0.72, + "learning_rate": 1.2850917347200352e-07, + "loss": 0.6974, + "step": 19572 + }, + { + "epoch": 0.72, + "learning_rate": 1.2850551140733145e-07, + "loss": 0.7053, + "step": 19573 + }, + { + "epoch": 0.72, + "learning_rate": 1.285018493426594e-07, + "loss": 0.6938, + "step": 19574 + }, + { + "epoch": 0.72, + "learning_rate": 1.2849818727798733e-07, + "loss": 0.6875, + "step": 19575 + }, + { + "epoch": 0.72, + "learning_rate": 1.2849452521331525e-07, + "loss": 0.7011, + "step": 19576 + }, + { + "epoch": 0.72, + "learning_rate": 1.2849086314864318e-07, + "loss": 0.6966, + "step": 19577 + }, + { + "epoch": 0.72, + "learning_rate": 1.2848720108397113e-07, + "loss": 0.6873, + "step": 19578 + }, + { + "epoch": 0.72, + "learning_rate": 1.2848353901929906e-07, + "loss": 0.6816, + "step": 19579 + }, + { + "epoch": 0.72, + "learning_rate": 1.28479876954627e-07, + "loss": 0.6986, + "step": 19580 + }, + { + "epoch": 0.72, + "learning_rate": 1.2847621488995494e-07, + "loss": 0.713, + "step": 19581 + }, + { + "epoch": 0.72, + "learning_rate": 1.284725528252829e-07, + "loss": 0.6905, + "step": 19582 + }, + { + "epoch": 0.72, + "learning_rate": 1.2846889076061082e-07, + "loss": 0.6758, + "step": 19583 + }, + { + "epoch": 0.72, + "learning_rate": 1.2846522869593877e-07, + "loss": 0.687, + "step": 19584 + }, + { + "epoch": 0.72, + "learning_rate": 1.284615666312667e-07, + "loss": 0.6956, + "step": 19585 + }, + { + "epoch": 0.72, + "learning_rate": 1.2845790456659465e-07, + "loss": 0.7013, + "step": 19586 + }, + { + "epoch": 0.72, + "learning_rate": 1.2845424250192258e-07, + "loss": 0.6953, + "step": 19587 + }, + { + "epoch": 0.72, + "learning_rate": 1.2845058043725053e-07, + "loss": 0.7067, + "step": 19588 + }, + { + "epoch": 0.72, + "learning_rate": 1.2844691837257846e-07, + "loss": 0.6903, + "step": 19589 + }, + { + "epoch": 0.72, + "learning_rate": 1.284432563079064e-07, + "loss": 0.6934, + "step": 19590 + }, + { + "epoch": 0.72, + "learning_rate": 1.2843959424323434e-07, + "loss": 0.6967, + "step": 19591 + }, + { + "epoch": 0.72, + "learning_rate": 1.2843593217856227e-07, + "loss": 0.7039, + "step": 19592 + }, + { + "epoch": 0.72, + "learning_rate": 1.284322701138902e-07, + "loss": 0.7161, + "step": 19593 + }, + { + "epoch": 0.72, + "learning_rate": 1.2842860804921815e-07, + "loss": 0.6961, + "step": 19594 + }, + { + "epoch": 0.72, + "learning_rate": 1.2842494598454607e-07, + "loss": 0.6845, + "step": 19595 + }, + { + "epoch": 0.72, + "learning_rate": 1.2842128391987403e-07, + "loss": 0.7086, + "step": 19596 + }, + { + "epoch": 0.72, + "learning_rate": 1.2841762185520195e-07, + "loss": 0.696, + "step": 19597 + }, + { + "epoch": 0.72, + "learning_rate": 1.2841395979052988e-07, + "loss": 0.7004, + "step": 19598 + }, + { + "epoch": 0.72, + "learning_rate": 1.2841029772585783e-07, + "loss": 0.6898, + "step": 19599 + }, + { + "epoch": 0.72, + "learning_rate": 1.2840663566118576e-07, + "loss": 0.6892, + "step": 19600 + }, + { + "epoch": 0.72, + "learning_rate": 1.284029735965137e-07, + "loss": 0.691, + "step": 19601 + }, + { + "epoch": 0.72, + "learning_rate": 1.2839931153184164e-07, + "loss": 0.6809, + "step": 19602 + }, + { + "epoch": 0.72, + "learning_rate": 1.283956494671696e-07, + "loss": 0.6891, + "step": 19603 + }, + { + "epoch": 0.72, + "learning_rate": 1.2839198740249752e-07, + "loss": 0.6866, + "step": 19604 + }, + { + "epoch": 0.72, + "learning_rate": 1.2838832533782547e-07, + "loss": 0.7071, + "step": 19605 + }, + { + "epoch": 0.72, + "learning_rate": 1.283846632731534e-07, + "loss": 0.6809, + "step": 19606 + }, + { + "epoch": 0.72, + "learning_rate": 1.2838100120848135e-07, + "loss": 0.6949, + "step": 19607 + }, + { + "epoch": 0.72, + "learning_rate": 1.2837733914380928e-07, + "loss": 0.6869, + "step": 19608 + }, + { + "epoch": 0.72, + "learning_rate": 1.2837367707913723e-07, + "loss": 0.6909, + "step": 19609 + }, + { + "epoch": 0.72, + "learning_rate": 1.2837001501446516e-07, + "loss": 0.6931, + "step": 19610 + }, + { + "epoch": 0.72, + "learning_rate": 1.2836635294979308e-07, + "loss": 0.6816, + "step": 19611 + }, + { + "epoch": 0.72, + "learning_rate": 1.28362690885121e-07, + "loss": 0.7005, + "step": 19612 + }, + { + "epoch": 0.72, + "learning_rate": 1.2835902882044896e-07, + "loss": 0.69, + "step": 19613 + }, + { + "epoch": 0.72, + "learning_rate": 1.283553667557769e-07, + "loss": 0.6719, + "step": 19614 + }, + { + "epoch": 0.72, + "learning_rate": 1.2835170469110484e-07, + "loss": 0.7003, + "step": 19615 + }, + { + "epoch": 0.72, + "learning_rate": 1.2834804262643277e-07, + "loss": 0.6843, + "step": 19616 + }, + { + "epoch": 0.72, + "learning_rate": 1.2834438056176072e-07, + "loss": 0.7033, + "step": 19617 + }, + { + "epoch": 0.72, + "learning_rate": 1.2834071849708865e-07, + "loss": 0.6923, + "step": 19618 + }, + { + "epoch": 0.72, + "learning_rate": 1.2833705643241658e-07, + "loss": 0.6904, + "step": 19619 + }, + { + "epoch": 0.72, + "learning_rate": 1.2833339436774453e-07, + "loss": 0.6986, + "step": 19620 + }, + { + "epoch": 0.72, + "learning_rate": 1.2832973230307246e-07, + "loss": 0.6879, + "step": 19621 + }, + { + "epoch": 0.72, + "learning_rate": 1.283260702384004e-07, + "loss": 0.6804, + "step": 19622 + }, + { + "epoch": 0.72, + "learning_rate": 1.2832240817372834e-07, + "loss": 0.7035, + "step": 19623 + }, + { + "epoch": 0.72, + "learning_rate": 1.283187461090563e-07, + "loss": 0.6882, + "step": 19624 + }, + { + "epoch": 0.72, + "learning_rate": 1.2831508404438422e-07, + "loss": 0.695, + "step": 19625 + }, + { + "epoch": 0.72, + "learning_rate": 1.2831142197971217e-07, + "loss": 0.7069, + "step": 19626 + }, + { + "epoch": 0.72, + "learning_rate": 1.283077599150401e-07, + "loss": 0.6956, + "step": 19627 + }, + { + "epoch": 0.72, + "learning_rate": 1.2830409785036805e-07, + "loss": 0.6813, + "step": 19628 + }, + { + "epoch": 0.72, + "learning_rate": 1.2830043578569595e-07, + "loss": 0.7007, + "step": 19629 + }, + { + "epoch": 0.72, + "learning_rate": 1.282967737210239e-07, + "loss": 0.692, + "step": 19630 + }, + { + "epoch": 0.72, + "learning_rate": 1.2829311165635183e-07, + "loss": 0.6851, + "step": 19631 + }, + { + "epoch": 0.72, + "learning_rate": 1.2828944959167978e-07, + "loss": 0.6963, + "step": 19632 + }, + { + "epoch": 0.72, + "learning_rate": 1.282857875270077e-07, + "loss": 0.6813, + "step": 19633 + }, + { + "epoch": 0.72, + "learning_rate": 1.2828212546233566e-07, + "loss": 0.6841, + "step": 19634 + }, + { + "epoch": 0.72, + "learning_rate": 1.282784633976636e-07, + "loss": 0.6959, + "step": 19635 + }, + { + "epoch": 0.72, + "learning_rate": 1.2827480133299154e-07, + "loss": 0.6978, + "step": 19636 + }, + { + "epoch": 0.72, + "learning_rate": 1.2827113926831947e-07, + "loss": 0.6953, + "step": 19637 + }, + { + "epoch": 0.72, + "learning_rate": 1.2826747720364742e-07, + "loss": 0.696, + "step": 19638 + }, + { + "epoch": 0.72, + "learning_rate": 1.2826381513897535e-07, + "loss": 0.7003, + "step": 19639 + }, + { + "epoch": 0.72, + "learning_rate": 1.282601530743033e-07, + "loss": 0.6814, + "step": 19640 + }, + { + "epoch": 0.72, + "learning_rate": 1.2825649100963123e-07, + "loss": 0.6773, + "step": 19641 + }, + { + "epoch": 0.72, + "learning_rate": 1.2825282894495916e-07, + "loss": 0.6936, + "step": 19642 + }, + { + "epoch": 0.72, + "learning_rate": 1.282491668802871e-07, + "loss": 0.6923, + "step": 19643 + }, + { + "epoch": 0.72, + "learning_rate": 1.2824550481561504e-07, + "loss": 0.6909, + "step": 19644 + }, + { + "epoch": 0.72, + "learning_rate": 1.28241842750943e-07, + "loss": 0.7014, + "step": 19645 + }, + { + "epoch": 0.72, + "learning_rate": 1.2823818068627092e-07, + "loss": 0.6959, + "step": 19646 + }, + { + "epoch": 0.72, + "learning_rate": 1.2823451862159884e-07, + "loss": 0.7068, + "step": 19647 + }, + { + "epoch": 0.72, + "learning_rate": 1.2823085655692677e-07, + "loss": 0.676, + "step": 19648 + }, + { + "epoch": 0.72, + "learning_rate": 1.2822719449225472e-07, + "loss": 0.685, + "step": 19649 + }, + { + "epoch": 0.72, + "learning_rate": 1.2822353242758265e-07, + "loss": 0.6955, + "step": 19650 + }, + { + "epoch": 0.72, + "learning_rate": 1.282198703629106e-07, + "loss": 0.6964, + "step": 19651 + }, + { + "epoch": 0.72, + "learning_rate": 1.2821620829823853e-07, + "loss": 0.694, + "step": 19652 + }, + { + "epoch": 0.72, + "learning_rate": 1.2821254623356648e-07, + "loss": 0.6918, + "step": 19653 + }, + { + "epoch": 0.72, + "learning_rate": 1.282088841688944e-07, + "loss": 0.6815, + "step": 19654 + }, + { + "epoch": 0.72, + "learning_rate": 1.2820522210422236e-07, + "loss": 0.6932, + "step": 19655 + }, + { + "epoch": 0.72, + "learning_rate": 1.282015600395503e-07, + "loss": 0.6948, + "step": 19656 + }, + { + "epoch": 0.72, + "learning_rate": 1.2819789797487824e-07, + "loss": 0.6978, + "step": 19657 + }, + { + "epoch": 0.72, + "learning_rate": 1.2819423591020617e-07, + "loss": 0.6963, + "step": 19658 + }, + { + "epoch": 0.72, + "learning_rate": 1.2819057384553412e-07, + "loss": 0.6977, + "step": 19659 + }, + { + "epoch": 0.72, + "learning_rate": 1.2818691178086205e-07, + "loss": 0.7073, + "step": 19660 + }, + { + "epoch": 0.72, + "learning_rate": 1.2818324971619e-07, + "loss": 0.6922, + "step": 19661 + }, + { + "epoch": 0.72, + "learning_rate": 1.2817958765151793e-07, + "loss": 0.717, + "step": 19662 + }, + { + "epoch": 0.72, + "learning_rate": 1.2817592558684585e-07, + "loss": 0.6868, + "step": 19663 + }, + { + "epoch": 0.72, + "learning_rate": 1.281722635221738e-07, + "loss": 0.6957, + "step": 19664 + }, + { + "epoch": 0.72, + "learning_rate": 1.2816860145750173e-07, + "loss": 0.6812, + "step": 19665 + }, + { + "epoch": 0.72, + "learning_rate": 1.2816493939282966e-07, + "loss": 0.6863, + "step": 19666 + }, + { + "epoch": 0.72, + "learning_rate": 1.2816127732815761e-07, + "loss": 0.7041, + "step": 19667 + }, + { + "epoch": 0.72, + "learning_rate": 1.2815761526348554e-07, + "loss": 0.6968, + "step": 19668 + }, + { + "epoch": 0.72, + "learning_rate": 1.2815395319881347e-07, + "loss": 0.6991, + "step": 19669 + }, + { + "epoch": 0.72, + "learning_rate": 1.2815029113414142e-07, + "loss": 0.6964, + "step": 19670 + }, + { + "epoch": 0.72, + "learning_rate": 1.2814662906946935e-07, + "loss": 0.6989, + "step": 19671 + }, + { + "epoch": 0.72, + "learning_rate": 1.281429670047973e-07, + "loss": 0.6948, + "step": 19672 + }, + { + "epoch": 0.72, + "learning_rate": 1.2813930494012523e-07, + "loss": 0.6898, + "step": 19673 + }, + { + "epoch": 0.72, + "learning_rate": 1.2813564287545318e-07, + "loss": 0.7024, + "step": 19674 + }, + { + "epoch": 0.72, + "learning_rate": 1.281319808107811e-07, + "loss": 0.6923, + "step": 19675 + }, + { + "epoch": 0.72, + "learning_rate": 1.2812831874610906e-07, + "loss": 0.6904, + "step": 19676 + }, + { + "epoch": 0.72, + "learning_rate": 1.28124656681437e-07, + "loss": 0.6964, + "step": 19677 + }, + { + "epoch": 0.72, + "learning_rate": 1.2812099461676494e-07, + "loss": 0.7038, + "step": 19678 + }, + { + "epoch": 0.72, + "learning_rate": 1.2811733255209287e-07, + "loss": 0.7053, + "step": 19679 + }, + { + "epoch": 0.72, + "learning_rate": 1.2811367048742082e-07, + "loss": 0.699, + "step": 19680 + }, + { + "epoch": 0.72, + "learning_rate": 1.2811000842274875e-07, + "loss": 0.6973, + "step": 19681 + }, + { + "epoch": 0.72, + "learning_rate": 1.2810634635807667e-07, + "loss": 0.6972, + "step": 19682 + }, + { + "epoch": 0.72, + "learning_rate": 1.281026842934046e-07, + "loss": 0.6877, + "step": 19683 + }, + { + "epoch": 0.72, + "learning_rate": 1.2809902222873255e-07, + "loss": 0.6972, + "step": 19684 + }, + { + "epoch": 0.72, + "learning_rate": 1.2809536016406048e-07, + "loss": 0.7008, + "step": 19685 + }, + { + "epoch": 0.72, + "learning_rate": 1.2809169809938843e-07, + "loss": 0.7, + "step": 19686 + }, + { + "epoch": 0.72, + "learning_rate": 1.2808803603471636e-07, + "loss": 0.6904, + "step": 19687 + }, + { + "epoch": 0.72, + "learning_rate": 1.2808437397004431e-07, + "loss": 0.6928, + "step": 19688 + }, + { + "epoch": 0.72, + "learning_rate": 1.2808071190537224e-07, + "loss": 0.6844, + "step": 19689 + }, + { + "epoch": 0.72, + "learning_rate": 1.2807704984070017e-07, + "loss": 0.6882, + "step": 19690 + }, + { + "epoch": 0.72, + "learning_rate": 1.2807338777602812e-07, + "loss": 0.6988, + "step": 19691 + }, + { + "epoch": 0.72, + "learning_rate": 1.2806972571135605e-07, + "loss": 0.691, + "step": 19692 + }, + { + "epoch": 0.72, + "learning_rate": 1.28066063646684e-07, + "loss": 0.6905, + "step": 19693 + }, + { + "epoch": 0.72, + "learning_rate": 1.2806240158201193e-07, + "loss": 0.6875, + "step": 19694 + }, + { + "epoch": 0.72, + "learning_rate": 1.2805873951733988e-07, + "loss": 0.6927, + "step": 19695 + }, + { + "epoch": 0.72, + "learning_rate": 1.280550774526678e-07, + "loss": 0.6924, + "step": 19696 + }, + { + "epoch": 0.72, + "learning_rate": 1.2805141538799576e-07, + "loss": 0.6955, + "step": 19697 + }, + { + "epoch": 0.72, + "learning_rate": 1.2804775332332369e-07, + "loss": 0.6915, + "step": 19698 + }, + { + "epoch": 0.72, + "learning_rate": 1.2804409125865164e-07, + "loss": 0.672, + "step": 19699 + }, + { + "epoch": 0.72, + "learning_rate": 1.2804042919397954e-07, + "loss": 0.6908, + "step": 19700 + }, + { + "epoch": 0.72, + "learning_rate": 1.280367671293075e-07, + "loss": 0.6726, + "step": 19701 + }, + { + "epoch": 0.72, + "learning_rate": 1.2803310506463542e-07, + "loss": 0.6846, + "step": 19702 + }, + { + "epoch": 0.72, + "learning_rate": 1.2802944299996337e-07, + "loss": 0.6966, + "step": 19703 + }, + { + "epoch": 0.72, + "learning_rate": 1.280257809352913e-07, + "loss": 0.6939, + "step": 19704 + }, + { + "epoch": 0.72, + "learning_rate": 1.2802211887061925e-07, + "loss": 0.6824, + "step": 19705 + }, + { + "epoch": 0.72, + "learning_rate": 1.2801845680594718e-07, + "loss": 0.6894, + "step": 19706 + }, + { + "epoch": 0.72, + "learning_rate": 1.2801479474127513e-07, + "loss": 0.6884, + "step": 19707 + }, + { + "epoch": 0.72, + "learning_rate": 1.2801113267660306e-07, + "loss": 0.7088, + "step": 19708 + }, + { + "epoch": 0.72, + "learning_rate": 1.28007470611931e-07, + "loss": 0.6962, + "step": 19709 + }, + { + "epoch": 0.72, + "learning_rate": 1.2800380854725894e-07, + "loss": 0.6942, + "step": 19710 + }, + { + "epoch": 0.72, + "learning_rate": 1.280001464825869e-07, + "loss": 0.687, + "step": 19711 + }, + { + "epoch": 0.72, + "learning_rate": 1.2799648441791482e-07, + "loss": 0.6822, + "step": 19712 + }, + { + "epoch": 0.72, + "learning_rate": 1.2799282235324275e-07, + "loss": 0.6914, + "step": 19713 + }, + { + "epoch": 0.72, + "learning_rate": 1.279891602885707e-07, + "loss": 0.6911, + "step": 19714 + }, + { + "epoch": 0.72, + "learning_rate": 1.2798549822389863e-07, + "loss": 0.6901, + "step": 19715 + }, + { + "epoch": 0.72, + "learning_rate": 1.2798183615922658e-07, + "loss": 0.7023, + "step": 19716 + }, + { + "epoch": 0.72, + "learning_rate": 1.279781740945545e-07, + "loss": 0.6879, + "step": 19717 + }, + { + "epoch": 0.72, + "learning_rate": 1.2797451202988243e-07, + "loss": 0.6851, + "step": 19718 + }, + { + "epoch": 0.72, + "learning_rate": 1.2797084996521036e-07, + "loss": 0.6823, + "step": 19719 + }, + { + "epoch": 0.72, + "learning_rate": 1.279671879005383e-07, + "loss": 0.6898, + "step": 19720 + }, + { + "epoch": 0.72, + "learning_rate": 1.2796352583586624e-07, + "loss": 0.7093, + "step": 19721 + }, + { + "epoch": 0.72, + "learning_rate": 1.279598637711942e-07, + "loss": 0.6932, + "step": 19722 + }, + { + "epoch": 0.72, + "learning_rate": 1.2795620170652212e-07, + "loss": 0.7044, + "step": 19723 + }, + { + "epoch": 0.72, + "learning_rate": 1.2795253964185007e-07, + "loss": 0.6912, + "step": 19724 + }, + { + "epoch": 0.72, + "learning_rate": 1.27948877577178e-07, + "loss": 0.7077, + "step": 19725 + }, + { + "epoch": 0.72, + "learning_rate": 1.2794521551250595e-07, + "loss": 0.6901, + "step": 19726 + }, + { + "epoch": 0.72, + "learning_rate": 1.2794155344783388e-07, + "loss": 0.7005, + "step": 19727 + }, + { + "epoch": 0.72, + "learning_rate": 1.2793789138316183e-07, + "loss": 0.692, + "step": 19728 + }, + { + "epoch": 0.72, + "learning_rate": 1.2793422931848976e-07, + "loss": 0.6918, + "step": 19729 + }, + { + "epoch": 0.72, + "learning_rate": 1.279305672538177e-07, + "loss": 0.6917, + "step": 19730 + }, + { + "epoch": 0.72, + "learning_rate": 1.2792690518914564e-07, + "loss": 0.695, + "step": 19731 + }, + { + "epoch": 0.72, + "learning_rate": 1.279232431244736e-07, + "loss": 0.6813, + "step": 19732 + }, + { + "epoch": 0.72, + "learning_rate": 1.2791958105980152e-07, + "loss": 0.6992, + "step": 19733 + }, + { + "epoch": 0.72, + "learning_rate": 1.2791591899512944e-07, + "loss": 0.6875, + "step": 19734 + }, + { + "epoch": 0.72, + "learning_rate": 1.279122569304574e-07, + "loss": 0.6918, + "step": 19735 + }, + { + "epoch": 0.72, + "learning_rate": 1.2790859486578532e-07, + "loss": 0.6904, + "step": 19736 + }, + { + "epoch": 0.72, + "learning_rate": 1.2790493280111325e-07, + "loss": 0.7023, + "step": 19737 + }, + { + "epoch": 0.72, + "learning_rate": 1.279012707364412e-07, + "loss": 0.6787, + "step": 19738 + }, + { + "epoch": 0.72, + "learning_rate": 1.2789760867176913e-07, + "loss": 0.6927, + "step": 19739 + }, + { + "epoch": 0.72, + "learning_rate": 1.2789394660709706e-07, + "loss": 0.7044, + "step": 19740 + }, + { + "epoch": 0.72, + "learning_rate": 1.27890284542425e-07, + "loss": 0.6921, + "step": 19741 + }, + { + "epoch": 0.72, + "learning_rate": 1.2788662247775294e-07, + "loss": 0.6973, + "step": 19742 + }, + { + "epoch": 0.72, + "learning_rate": 1.278829604130809e-07, + "loss": 0.6892, + "step": 19743 + }, + { + "epoch": 0.72, + "learning_rate": 1.2787929834840882e-07, + "loss": 0.6879, + "step": 19744 + }, + { + "epoch": 0.72, + "learning_rate": 1.2787563628373677e-07, + "loss": 0.67, + "step": 19745 + }, + { + "epoch": 0.72, + "learning_rate": 1.278719742190647e-07, + "loss": 0.6912, + "step": 19746 + }, + { + "epoch": 0.72, + "learning_rate": 1.2786831215439265e-07, + "loss": 0.6978, + "step": 19747 + }, + { + "epoch": 0.72, + "learning_rate": 1.2786465008972058e-07, + "loss": 0.6977, + "step": 19748 + }, + { + "epoch": 0.72, + "learning_rate": 1.2786098802504853e-07, + "loss": 0.6856, + "step": 19749 + }, + { + "epoch": 0.72, + "learning_rate": 1.2785732596037646e-07, + "loss": 0.6912, + "step": 19750 + }, + { + "epoch": 0.72, + "learning_rate": 1.278536638957044e-07, + "loss": 0.6946, + "step": 19751 + }, + { + "epoch": 0.72, + "learning_rate": 1.2785000183103234e-07, + "loss": 0.6877, + "step": 19752 + }, + { + "epoch": 0.72, + "learning_rate": 1.278463397663603e-07, + "loss": 0.6971, + "step": 19753 + }, + { + "epoch": 0.72, + "learning_rate": 1.278426777016882e-07, + "loss": 0.7039, + "step": 19754 + }, + { + "epoch": 0.72, + "learning_rate": 1.2783901563701614e-07, + "loss": 0.6915, + "step": 19755 + }, + { + "epoch": 0.72, + "learning_rate": 1.2783535357234407e-07, + "loss": 0.7282, + "step": 19756 + }, + { + "epoch": 0.72, + "learning_rate": 1.2783169150767202e-07, + "loss": 0.6854, + "step": 19757 + }, + { + "epoch": 0.72, + "learning_rate": 1.2782802944299995e-07, + "loss": 0.6637, + "step": 19758 + }, + { + "epoch": 0.72, + "learning_rate": 1.278243673783279e-07, + "loss": 0.6946, + "step": 19759 + }, + { + "epoch": 0.72, + "learning_rate": 1.2782070531365583e-07, + "loss": 0.6883, + "step": 19760 + }, + { + "epoch": 0.72, + "learning_rate": 1.2781704324898376e-07, + "loss": 0.6844, + "step": 19761 + }, + { + "epoch": 0.72, + "learning_rate": 1.278133811843117e-07, + "loss": 0.6956, + "step": 19762 + }, + { + "epoch": 0.72, + "learning_rate": 1.2780971911963964e-07, + "loss": 0.6834, + "step": 19763 + }, + { + "epoch": 0.72, + "learning_rate": 1.278060570549676e-07, + "loss": 0.6905, + "step": 19764 + }, + { + "epoch": 0.72, + "learning_rate": 1.2780239499029552e-07, + "loss": 0.7034, + "step": 19765 + }, + { + "epoch": 0.72, + "learning_rate": 1.2779873292562347e-07, + "loss": 0.6989, + "step": 19766 + }, + { + "epoch": 0.72, + "learning_rate": 1.277950708609514e-07, + "loss": 0.6967, + "step": 19767 + }, + { + "epoch": 0.72, + "learning_rate": 1.2779140879627935e-07, + "loss": 0.6775, + "step": 19768 + }, + { + "epoch": 0.72, + "learning_rate": 1.2778774673160728e-07, + "loss": 0.6936, + "step": 19769 + }, + { + "epoch": 0.72, + "learning_rate": 1.2778408466693523e-07, + "loss": 0.6898, + "step": 19770 + }, + { + "epoch": 0.72, + "learning_rate": 1.2778042260226316e-07, + "loss": 0.7018, + "step": 19771 + }, + { + "epoch": 0.72, + "learning_rate": 1.2777676053759108e-07, + "loss": 0.7078, + "step": 19772 + }, + { + "epoch": 0.72, + "learning_rate": 1.27773098472919e-07, + "loss": 0.6961, + "step": 19773 + }, + { + "epoch": 0.72, + "learning_rate": 1.2776943640824696e-07, + "loss": 0.6861, + "step": 19774 + }, + { + "epoch": 0.72, + "learning_rate": 1.277657743435749e-07, + "loss": 0.6855, + "step": 19775 + }, + { + "epoch": 0.72, + "learning_rate": 1.2776211227890284e-07, + "loss": 0.6895, + "step": 19776 + }, + { + "epoch": 0.72, + "learning_rate": 1.2775845021423077e-07, + "loss": 0.697, + "step": 19777 + }, + { + "epoch": 0.72, + "learning_rate": 1.2775478814955872e-07, + "loss": 0.6919, + "step": 19778 + }, + { + "epoch": 0.72, + "learning_rate": 1.2775112608488665e-07, + "loss": 0.6963, + "step": 19779 + }, + { + "epoch": 0.72, + "learning_rate": 1.277474640202146e-07, + "loss": 0.7036, + "step": 19780 + }, + { + "epoch": 0.72, + "learning_rate": 1.2774380195554253e-07, + "loss": 0.6907, + "step": 19781 + }, + { + "epoch": 0.72, + "learning_rate": 1.2774013989087048e-07, + "loss": 0.6983, + "step": 19782 + }, + { + "epoch": 0.72, + "learning_rate": 1.277364778261984e-07, + "loss": 0.7013, + "step": 19783 + }, + { + "epoch": 0.72, + "learning_rate": 1.2773281576152633e-07, + "loss": 0.6805, + "step": 19784 + }, + { + "epoch": 0.72, + "learning_rate": 1.277291536968543e-07, + "loss": 0.6894, + "step": 19785 + }, + { + "epoch": 0.72, + "learning_rate": 1.2772549163218221e-07, + "loss": 0.6752, + "step": 19786 + }, + { + "epoch": 0.72, + "learning_rate": 1.2772182956751017e-07, + "loss": 0.6813, + "step": 19787 + }, + { + "epoch": 0.72, + "learning_rate": 1.277181675028381e-07, + "loss": 0.694, + "step": 19788 + }, + { + "epoch": 0.72, + "learning_rate": 1.2771450543816605e-07, + "loss": 0.7068, + "step": 19789 + }, + { + "epoch": 0.72, + "learning_rate": 1.2771084337349395e-07, + "loss": 0.7035, + "step": 19790 + }, + { + "epoch": 0.72, + "learning_rate": 1.277071813088219e-07, + "loss": 0.6936, + "step": 19791 + }, + { + "epoch": 0.72, + "learning_rate": 1.2770351924414983e-07, + "loss": 0.6936, + "step": 19792 + }, + { + "epoch": 0.72, + "learning_rate": 1.2769985717947778e-07, + "loss": 0.6932, + "step": 19793 + }, + { + "epoch": 0.72, + "learning_rate": 1.276961951148057e-07, + "loss": 0.7023, + "step": 19794 + }, + { + "epoch": 0.72, + "learning_rate": 1.2769253305013366e-07, + "loss": 0.6921, + "step": 19795 + }, + { + "epoch": 0.72, + "learning_rate": 1.276888709854616e-07, + "loss": 0.702, + "step": 19796 + }, + { + "epoch": 0.72, + "learning_rate": 1.2768520892078954e-07, + "loss": 0.6977, + "step": 19797 + }, + { + "epoch": 0.72, + "learning_rate": 1.2768154685611747e-07, + "loss": 0.698, + "step": 19798 + }, + { + "epoch": 0.72, + "learning_rate": 1.2767788479144542e-07, + "loss": 0.6961, + "step": 19799 + }, + { + "epoch": 0.72, + "learning_rate": 1.2767422272677335e-07, + "loss": 0.6959, + "step": 19800 + }, + { + "epoch": 0.72, + "learning_rate": 1.276705606621013e-07, + "loss": 0.6902, + "step": 19801 + }, + { + "epoch": 0.72, + "learning_rate": 1.2766689859742923e-07, + "loss": 0.6944, + "step": 19802 + }, + { + "epoch": 0.72, + "learning_rate": 1.2766323653275718e-07, + "loss": 0.699, + "step": 19803 + }, + { + "epoch": 0.72, + "learning_rate": 1.276595744680851e-07, + "loss": 0.6937, + "step": 19804 + }, + { + "epoch": 0.72, + "learning_rate": 1.2765591240341303e-07, + "loss": 0.695, + "step": 19805 + }, + { + "epoch": 0.72, + "learning_rate": 1.2765225033874099e-07, + "loss": 0.6954, + "step": 19806 + }, + { + "epoch": 0.72, + "learning_rate": 1.2764858827406891e-07, + "loss": 0.6875, + "step": 19807 + }, + { + "epoch": 0.72, + "learning_rate": 1.2764492620939684e-07, + "loss": 0.6966, + "step": 19808 + }, + { + "epoch": 0.72, + "learning_rate": 1.276412641447248e-07, + "loss": 0.7004, + "step": 19809 + }, + { + "epoch": 0.72, + "learning_rate": 1.2763760208005272e-07, + "loss": 0.6958, + "step": 19810 + }, + { + "epoch": 0.72, + "learning_rate": 1.2763394001538065e-07, + "loss": 0.6869, + "step": 19811 + }, + { + "epoch": 0.72, + "learning_rate": 1.276302779507086e-07, + "loss": 0.6992, + "step": 19812 + }, + { + "epoch": 0.72, + "learning_rate": 1.2762661588603653e-07, + "loss": 0.6959, + "step": 19813 + }, + { + "epoch": 0.72, + "learning_rate": 1.2762295382136448e-07, + "loss": 0.685, + "step": 19814 + }, + { + "epoch": 0.72, + "learning_rate": 1.276192917566924e-07, + "loss": 0.6892, + "step": 19815 + }, + { + "epoch": 0.73, + "learning_rate": 1.2761562969202036e-07, + "loss": 0.6918, + "step": 19816 + }, + { + "epoch": 0.73, + "learning_rate": 1.2761196762734829e-07, + "loss": 0.7157, + "step": 19817 + }, + { + "epoch": 0.73, + "learning_rate": 1.2760830556267624e-07, + "loss": 0.6765, + "step": 19818 + }, + { + "epoch": 0.73, + "learning_rate": 1.2760464349800417e-07, + "loss": 0.6997, + "step": 19819 + }, + { + "epoch": 0.73, + "learning_rate": 1.2760098143333212e-07, + "loss": 0.6913, + "step": 19820 + }, + { + "epoch": 0.73, + "learning_rate": 1.2759731936866005e-07, + "loss": 0.6949, + "step": 19821 + }, + { + "epoch": 0.73, + "learning_rate": 1.27593657303988e-07, + "loss": 0.6974, + "step": 19822 + }, + { + "epoch": 0.73, + "learning_rate": 1.2758999523931593e-07, + "loss": 0.7006, + "step": 19823 + }, + { + "epoch": 0.73, + "learning_rate": 1.2758633317464388e-07, + "loss": 0.7076, + "step": 19824 + }, + { + "epoch": 0.73, + "learning_rate": 1.275826711099718e-07, + "loss": 0.7043, + "step": 19825 + }, + { + "epoch": 0.73, + "learning_rate": 1.2757900904529973e-07, + "loss": 0.6882, + "step": 19826 + }, + { + "epoch": 0.73, + "learning_rate": 1.2757534698062766e-07, + "loss": 0.7095, + "step": 19827 + }, + { + "epoch": 0.73, + "learning_rate": 1.275716849159556e-07, + "loss": 0.6894, + "step": 19828 + }, + { + "epoch": 0.73, + "learning_rate": 1.2756802285128354e-07, + "loss": 0.7053, + "step": 19829 + }, + { + "epoch": 0.73, + "learning_rate": 1.275643607866115e-07, + "loss": 0.6767, + "step": 19830 + }, + { + "epoch": 0.73, + "learning_rate": 1.2756069872193942e-07, + "loss": 0.6898, + "step": 19831 + }, + { + "epoch": 0.73, + "learning_rate": 1.2755703665726735e-07, + "loss": 0.6995, + "step": 19832 + }, + { + "epoch": 0.73, + "learning_rate": 1.275533745925953e-07, + "loss": 0.7024, + "step": 19833 + }, + { + "epoch": 0.73, + "learning_rate": 1.2754971252792323e-07, + "loss": 0.6954, + "step": 19834 + }, + { + "epoch": 0.73, + "learning_rate": 1.2754605046325118e-07, + "loss": 0.6976, + "step": 19835 + }, + { + "epoch": 0.73, + "learning_rate": 1.275423883985791e-07, + "loss": 0.7004, + "step": 19836 + }, + { + "epoch": 0.73, + "learning_rate": 1.2753872633390706e-07, + "loss": 0.7093, + "step": 19837 + }, + { + "epoch": 0.73, + "learning_rate": 1.2753506426923498e-07, + "loss": 0.6844, + "step": 19838 + }, + { + "epoch": 0.73, + "learning_rate": 1.2753140220456294e-07, + "loss": 0.7009, + "step": 19839 + }, + { + "epoch": 0.73, + "learning_rate": 1.2752774013989086e-07, + "loss": 0.6892, + "step": 19840 + }, + { + "epoch": 0.73, + "learning_rate": 1.2752407807521882e-07, + "loss": 0.6906, + "step": 19841 + }, + { + "epoch": 0.73, + "learning_rate": 1.2752041601054674e-07, + "loss": 0.6911, + "step": 19842 + }, + { + "epoch": 0.73, + "learning_rate": 1.275167539458747e-07, + "loss": 0.6887, + "step": 19843 + }, + { + "epoch": 0.73, + "learning_rate": 1.275130918812026e-07, + "loss": 0.6949, + "step": 19844 + }, + { + "epoch": 0.73, + "learning_rate": 1.2750942981653055e-07, + "loss": 0.6812, + "step": 19845 + }, + { + "epoch": 0.73, + "learning_rate": 1.2750576775185848e-07, + "loss": 0.6881, + "step": 19846 + }, + { + "epoch": 0.73, + "learning_rate": 1.2750210568718643e-07, + "loss": 0.6992, + "step": 19847 + }, + { + "epoch": 0.73, + "learning_rate": 1.2749844362251436e-07, + "loss": 0.7071, + "step": 19848 + }, + { + "epoch": 0.73, + "learning_rate": 1.274947815578423e-07, + "loss": 0.6804, + "step": 19849 + }, + { + "epoch": 0.73, + "learning_rate": 1.2749111949317024e-07, + "loss": 0.685, + "step": 19850 + }, + { + "epoch": 0.73, + "learning_rate": 1.274874574284982e-07, + "loss": 0.6879, + "step": 19851 + }, + { + "epoch": 0.73, + "learning_rate": 1.2748379536382612e-07, + "loss": 0.6836, + "step": 19852 + }, + { + "epoch": 0.73, + "learning_rate": 1.2748013329915407e-07, + "loss": 0.6952, + "step": 19853 + }, + { + "epoch": 0.73, + "learning_rate": 1.27476471234482e-07, + "loss": 0.7091, + "step": 19854 + }, + { + "epoch": 0.73, + "learning_rate": 1.2747280916980992e-07, + "loss": 0.6956, + "step": 19855 + }, + { + "epoch": 0.73, + "learning_rate": 1.2746914710513788e-07, + "loss": 0.7067, + "step": 19856 + }, + { + "epoch": 0.73, + "learning_rate": 1.274654850404658e-07, + "loss": 0.7005, + "step": 19857 + }, + { + "epoch": 0.73, + "learning_rate": 1.2746182297579376e-07, + "loss": 0.695, + "step": 19858 + }, + { + "epoch": 0.73, + "learning_rate": 1.2745816091112168e-07, + "loss": 0.6792, + "step": 19859 + }, + { + "epoch": 0.73, + "learning_rate": 1.2745449884644964e-07, + "loss": 0.6917, + "step": 19860 + }, + { + "epoch": 0.73, + "learning_rate": 1.2745083678177756e-07, + "loss": 0.6839, + "step": 19861 + }, + { + "epoch": 0.73, + "learning_rate": 1.274471747171055e-07, + "loss": 0.6826, + "step": 19862 + }, + { + "epoch": 0.73, + "learning_rate": 1.2744351265243342e-07, + "loss": 0.693, + "step": 19863 + }, + { + "epoch": 0.73, + "learning_rate": 1.2743985058776137e-07, + "loss": 0.7009, + "step": 19864 + }, + { + "epoch": 0.73, + "learning_rate": 1.274361885230893e-07, + "loss": 0.7015, + "step": 19865 + }, + { + "epoch": 0.73, + "learning_rate": 1.2743252645841725e-07, + "loss": 0.7005, + "step": 19866 + }, + { + "epoch": 0.73, + "learning_rate": 1.2742886439374518e-07, + "loss": 0.6967, + "step": 19867 + }, + { + "epoch": 0.73, + "learning_rate": 1.2742520232907313e-07, + "loss": 0.6864, + "step": 19868 + }, + { + "epoch": 0.73, + "learning_rate": 1.2742154026440106e-07, + "loss": 0.7012, + "step": 19869 + }, + { + "epoch": 0.73, + "learning_rate": 1.27417878199729e-07, + "loss": 0.7087, + "step": 19870 + }, + { + "epoch": 0.73, + "learning_rate": 1.2741421613505694e-07, + "loss": 0.6829, + "step": 19871 + }, + { + "epoch": 0.73, + "learning_rate": 1.274105540703849e-07, + "loss": 0.6976, + "step": 19872 + }, + { + "epoch": 0.73, + "learning_rate": 1.2740689200571282e-07, + "loss": 0.6848, + "step": 19873 + }, + { + "epoch": 0.73, + "learning_rate": 1.2740322994104077e-07, + "loss": 0.6859, + "step": 19874 + }, + { + "epoch": 0.73, + "learning_rate": 1.273995678763687e-07, + "loss": 0.7038, + "step": 19875 + }, + { + "epoch": 0.73, + "learning_rate": 1.2739590581169662e-07, + "loss": 0.6942, + "step": 19876 + }, + { + "epoch": 0.73, + "learning_rate": 1.2739224374702458e-07, + "loss": 0.695, + "step": 19877 + }, + { + "epoch": 0.73, + "learning_rate": 1.273885816823525e-07, + "loss": 0.6972, + "step": 19878 + }, + { + "epoch": 0.73, + "learning_rate": 1.2738491961768046e-07, + "loss": 0.6935, + "step": 19879 + }, + { + "epoch": 0.73, + "learning_rate": 1.2738125755300838e-07, + "loss": 0.6894, + "step": 19880 + }, + { + "epoch": 0.73, + "learning_rate": 1.273775954883363e-07, + "loss": 0.6896, + "step": 19881 + }, + { + "epoch": 0.73, + "learning_rate": 1.2737393342366424e-07, + "loss": 0.6828, + "step": 19882 + }, + { + "epoch": 0.73, + "learning_rate": 1.273702713589922e-07, + "loss": 0.695, + "step": 19883 + }, + { + "epoch": 0.73, + "learning_rate": 1.2736660929432012e-07, + "loss": 0.7019, + "step": 19884 + }, + { + "epoch": 0.73, + "learning_rate": 1.2736294722964807e-07, + "loss": 0.6968, + "step": 19885 + }, + { + "epoch": 0.73, + "learning_rate": 1.27359285164976e-07, + "loss": 0.6875, + "step": 19886 + }, + { + "epoch": 0.73, + "learning_rate": 1.2735562310030395e-07, + "loss": 0.6921, + "step": 19887 + }, + { + "epoch": 0.73, + "learning_rate": 1.2735196103563188e-07, + "loss": 0.7068, + "step": 19888 + }, + { + "epoch": 0.73, + "learning_rate": 1.2734829897095983e-07, + "loss": 0.6906, + "step": 19889 + }, + { + "epoch": 0.73, + "learning_rate": 1.2734463690628776e-07, + "loss": 0.6942, + "step": 19890 + }, + { + "epoch": 0.73, + "learning_rate": 1.273409748416157e-07, + "loss": 0.7035, + "step": 19891 + }, + { + "epoch": 0.73, + "learning_rate": 1.2733731277694364e-07, + "loss": 0.6959, + "step": 19892 + }, + { + "epoch": 0.73, + "learning_rate": 1.273336507122716e-07, + "loss": 0.6832, + "step": 19893 + }, + { + "epoch": 0.73, + "learning_rate": 1.2732998864759952e-07, + "loss": 0.68, + "step": 19894 + }, + { + "epoch": 0.73, + "learning_rate": 1.2732632658292747e-07, + "loss": 0.692, + "step": 19895 + }, + { + "epoch": 0.73, + "learning_rate": 1.273226645182554e-07, + "loss": 0.7021, + "step": 19896 + }, + { + "epoch": 0.73, + "learning_rate": 1.2731900245358335e-07, + "loss": 0.707, + "step": 19897 + }, + { + "epoch": 0.73, + "learning_rate": 1.2731534038891125e-07, + "loss": 0.7087, + "step": 19898 + }, + { + "epoch": 0.73, + "learning_rate": 1.273116783242392e-07, + "loss": 0.6898, + "step": 19899 + }, + { + "epoch": 0.73, + "learning_rate": 1.2730801625956713e-07, + "loss": 0.6903, + "step": 19900 + }, + { + "epoch": 0.73, + "learning_rate": 1.2730435419489508e-07, + "loss": 0.691, + "step": 19901 + }, + { + "epoch": 0.73, + "learning_rate": 1.27300692130223e-07, + "loss": 0.686, + "step": 19902 + }, + { + "epoch": 0.73, + "learning_rate": 1.2729703006555093e-07, + "loss": 0.6988, + "step": 19903 + }, + { + "epoch": 0.73, + "learning_rate": 1.272933680008789e-07, + "loss": 0.6996, + "step": 19904 + }, + { + "epoch": 0.73, + "learning_rate": 1.2728970593620681e-07, + "loss": 0.6913, + "step": 19905 + }, + { + "epoch": 0.73, + "learning_rate": 1.2728604387153477e-07, + "loss": 0.6799, + "step": 19906 + }, + { + "epoch": 0.73, + "learning_rate": 1.272823818068627e-07, + "loss": 0.6856, + "step": 19907 + }, + { + "epoch": 0.73, + "learning_rate": 1.2727871974219065e-07, + "loss": 0.6907, + "step": 19908 + }, + { + "epoch": 0.73, + "learning_rate": 1.2727505767751857e-07, + "loss": 0.6982, + "step": 19909 + }, + { + "epoch": 0.73, + "learning_rate": 1.2727139561284653e-07, + "loss": 0.7087, + "step": 19910 + }, + { + "epoch": 0.73, + "learning_rate": 1.2726773354817445e-07, + "loss": 0.6922, + "step": 19911 + }, + { + "epoch": 0.73, + "learning_rate": 1.272640714835024e-07, + "loss": 0.6952, + "step": 19912 + }, + { + "epoch": 0.73, + "learning_rate": 1.2726040941883033e-07, + "loss": 0.6795, + "step": 19913 + }, + { + "epoch": 0.73, + "learning_rate": 1.272567473541583e-07, + "loss": 0.7027, + "step": 19914 + }, + { + "epoch": 0.73, + "learning_rate": 1.2725308528948621e-07, + "loss": 0.7008, + "step": 19915 + }, + { + "epoch": 0.73, + "learning_rate": 1.2724942322481414e-07, + "loss": 0.7014, + "step": 19916 + }, + { + "epoch": 0.73, + "learning_rate": 1.2724576116014207e-07, + "loss": 0.6995, + "step": 19917 + }, + { + "epoch": 0.73, + "learning_rate": 1.2724209909547002e-07, + "loss": 0.6906, + "step": 19918 + }, + { + "epoch": 0.73, + "learning_rate": 1.2723843703079795e-07, + "loss": 0.7054, + "step": 19919 + }, + { + "epoch": 0.73, + "learning_rate": 1.272347749661259e-07, + "loss": 0.7042, + "step": 19920 + }, + { + "epoch": 0.73, + "learning_rate": 1.2723111290145383e-07, + "loss": 0.6801, + "step": 19921 + }, + { + "epoch": 0.73, + "learning_rate": 1.2722745083678178e-07, + "loss": 0.6796, + "step": 19922 + }, + { + "epoch": 0.73, + "learning_rate": 1.272237887721097e-07, + "loss": 0.7024, + "step": 19923 + }, + { + "epoch": 0.73, + "learning_rate": 1.2722012670743766e-07, + "loss": 0.6889, + "step": 19924 + }, + { + "epoch": 0.73, + "learning_rate": 1.2721646464276559e-07, + "loss": 0.6927, + "step": 19925 + }, + { + "epoch": 0.73, + "learning_rate": 1.2721280257809351e-07, + "loss": 0.6953, + "step": 19926 + }, + { + "epoch": 0.73, + "learning_rate": 1.2720914051342147e-07, + "loss": 0.6688, + "step": 19927 + }, + { + "epoch": 0.73, + "learning_rate": 1.272054784487494e-07, + "loss": 0.6926, + "step": 19928 + }, + { + "epoch": 0.73, + "learning_rate": 1.2720181638407735e-07, + "loss": 0.7008, + "step": 19929 + }, + { + "epoch": 0.73, + "learning_rate": 1.2719815431940527e-07, + "loss": 0.6944, + "step": 19930 + }, + { + "epoch": 0.73, + "learning_rate": 1.2719449225473323e-07, + "loss": 0.7006, + "step": 19931 + }, + { + "epoch": 0.73, + "learning_rate": 1.2719083019006115e-07, + "loss": 0.7044, + "step": 19932 + }, + { + "epoch": 0.73, + "learning_rate": 1.271871681253891e-07, + "loss": 0.691, + "step": 19933 + }, + { + "epoch": 0.73, + "learning_rate": 1.27183506060717e-07, + "loss": 0.6886, + "step": 19934 + }, + { + "epoch": 0.73, + "learning_rate": 1.2717984399604496e-07, + "loss": 0.6862, + "step": 19935 + }, + { + "epoch": 0.73, + "learning_rate": 1.2717618193137289e-07, + "loss": 0.6932, + "step": 19936 + }, + { + "epoch": 0.73, + "learning_rate": 1.2717251986670084e-07, + "loss": 0.7023, + "step": 19937 + }, + { + "epoch": 0.73, + "learning_rate": 1.2716885780202877e-07, + "loss": 0.7199, + "step": 19938 + }, + { + "epoch": 0.73, + "learning_rate": 1.2716519573735672e-07, + "loss": 0.6856, + "step": 19939 + }, + { + "epoch": 0.73, + "learning_rate": 1.2716153367268465e-07, + "loss": 0.7001, + "step": 19940 + }, + { + "epoch": 0.73, + "learning_rate": 1.271578716080126e-07, + "loss": 0.6888, + "step": 19941 + }, + { + "epoch": 0.73, + "learning_rate": 1.2715420954334053e-07, + "loss": 0.6937, + "step": 19942 + }, + { + "epoch": 0.73, + "learning_rate": 1.2715054747866848e-07, + "loss": 0.6858, + "step": 19943 + }, + { + "epoch": 0.73, + "learning_rate": 1.271468854139964e-07, + "loss": 0.6855, + "step": 19944 + }, + { + "epoch": 0.73, + "learning_rate": 1.2714322334932436e-07, + "loss": 0.6836, + "step": 19945 + }, + { + "epoch": 0.73, + "learning_rate": 1.2713956128465229e-07, + "loss": 0.6818, + "step": 19946 + }, + { + "epoch": 0.73, + "learning_rate": 1.271358992199802e-07, + "loss": 0.69, + "step": 19947 + }, + { + "epoch": 0.73, + "learning_rate": 1.2713223715530817e-07, + "loss": 0.6952, + "step": 19948 + }, + { + "epoch": 0.73, + "learning_rate": 1.271285750906361e-07, + "loss": 0.6901, + "step": 19949 + }, + { + "epoch": 0.73, + "learning_rate": 1.2712491302596405e-07, + "loss": 0.6932, + "step": 19950 + }, + { + "epoch": 0.73, + "learning_rate": 1.2712125096129197e-07, + "loss": 0.6777, + "step": 19951 + }, + { + "epoch": 0.73, + "learning_rate": 1.271175888966199e-07, + "loss": 0.6908, + "step": 19952 + }, + { + "epoch": 0.73, + "learning_rate": 1.2711392683194783e-07, + "loss": 0.6797, + "step": 19953 + }, + { + "epoch": 0.73, + "learning_rate": 1.2711026476727578e-07, + "loss": 0.6839, + "step": 19954 + }, + { + "epoch": 0.73, + "learning_rate": 1.271066027026037e-07, + "loss": 0.6877, + "step": 19955 + }, + { + "epoch": 0.73, + "learning_rate": 1.2710294063793166e-07, + "loss": 0.7072, + "step": 19956 + }, + { + "epoch": 0.73, + "learning_rate": 1.2709927857325959e-07, + "loss": 0.6951, + "step": 19957 + }, + { + "epoch": 0.73, + "learning_rate": 1.2709561650858754e-07, + "loss": 0.6928, + "step": 19958 + }, + { + "epoch": 0.73, + "learning_rate": 1.2709195444391546e-07, + "loss": 0.6889, + "step": 19959 + }, + { + "epoch": 0.73, + "learning_rate": 1.2708829237924342e-07, + "loss": 0.6862, + "step": 19960 + }, + { + "epoch": 0.73, + "learning_rate": 1.2708463031457134e-07, + "loss": 0.6987, + "step": 19961 + }, + { + "epoch": 0.73, + "learning_rate": 1.270809682498993e-07, + "loss": 0.6878, + "step": 19962 + }, + { + "epoch": 0.73, + "learning_rate": 1.2707730618522722e-07, + "loss": 0.7078, + "step": 19963 + }, + { + "epoch": 0.73, + "learning_rate": 1.2707364412055518e-07, + "loss": 0.6976, + "step": 19964 + }, + { + "epoch": 0.73, + "learning_rate": 1.270699820558831e-07, + "loss": 0.6916, + "step": 19965 + }, + { + "epoch": 0.73, + "learning_rate": 1.2706631999121106e-07, + "loss": 0.6993, + "step": 19966 + }, + { + "epoch": 0.73, + "learning_rate": 1.2706265792653898e-07, + "loss": 0.691, + "step": 19967 + }, + { + "epoch": 0.73, + "learning_rate": 1.2705899586186694e-07, + "loss": 0.6907, + "step": 19968 + }, + { + "epoch": 0.73, + "learning_rate": 1.2705533379719486e-07, + "loss": 0.6936, + "step": 19969 + }, + { + "epoch": 0.73, + "learning_rate": 1.270516717325228e-07, + "loss": 0.7001, + "step": 19970 + }, + { + "epoch": 0.73, + "learning_rate": 1.2704800966785072e-07, + "loss": 0.6979, + "step": 19971 + }, + { + "epoch": 0.73, + "learning_rate": 1.2704434760317867e-07, + "loss": 0.702, + "step": 19972 + }, + { + "epoch": 0.73, + "learning_rate": 1.270406855385066e-07, + "loss": 0.694, + "step": 19973 + }, + { + "epoch": 0.73, + "learning_rate": 1.2703702347383452e-07, + "loss": 0.7007, + "step": 19974 + }, + { + "epoch": 0.73, + "learning_rate": 1.2703336140916248e-07, + "loss": 0.6971, + "step": 19975 + }, + { + "epoch": 0.73, + "learning_rate": 1.270296993444904e-07, + "loss": 0.6994, + "step": 19976 + }, + { + "epoch": 0.73, + "learning_rate": 1.2702603727981836e-07, + "loss": 0.697, + "step": 19977 + }, + { + "epoch": 0.73, + "learning_rate": 1.2702237521514628e-07, + "loss": 0.6785, + "step": 19978 + }, + { + "epoch": 0.73, + "learning_rate": 1.2701871315047424e-07, + "loss": 0.6898, + "step": 19979 + }, + { + "epoch": 0.73, + "learning_rate": 1.2701505108580216e-07, + "loss": 0.6888, + "step": 19980 + }, + { + "epoch": 0.73, + "learning_rate": 1.2701138902113012e-07, + "loss": 0.6842, + "step": 19981 + }, + { + "epoch": 0.73, + "learning_rate": 1.2700772695645804e-07, + "loss": 0.6975, + "step": 19982 + }, + { + "epoch": 0.73, + "learning_rate": 1.27004064891786e-07, + "loss": 0.6914, + "step": 19983 + }, + { + "epoch": 0.73, + "learning_rate": 1.2700040282711392e-07, + "loss": 0.6894, + "step": 19984 + }, + { + "epoch": 0.73, + "learning_rate": 1.2699674076244188e-07, + "loss": 0.6935, + "step": 19985 + }, + { + "epoch": 0.73, + "learning_rate": 1.269930786977698e-07, + "loss": 0.6858, + "step": 19986 + }, + { + "epoch": 0.73, + "learning_rate": 1.2698941663309776e-07, + "loss": 0.6944, + "step": 19987 + }, + { + "epoch": 0.73, + "learning_rate": 1.2698575456842566e-07, + "loss": 0.7053, + "step": 19988 + }, + { + "epoch": 0.73, + "learning_rate": 1.269820925037536e-07, + "loss": 0.7022, + "step": 19989 + }, + { + "epoch": 0.73, + "learning_rate": 1.2697843043908154e-07, + "loss": 0.6876, + "step": 19990 + }, + { + "epoch": 0.73, + "learning_rate": 1.269747683744095e-07, + "loss": 0.6758, + "step": 19991 + }, + { + "epoch": 0.73, + "learning_rate": 1.2697110630973742e-07, + "loss": 0.6926, + "step": 19992 + }, + { + "epoch": 0.73, + "learning_rate": 1.2696744424506537e-07, + "loss": 0.7059, + "step": 19993 + }, + { + "epoch": 0.73, + "learning_rate": 1.269637821803933e-07, + "loss": 0.6954, + "step": 19994 + }, + { + "epoch": 0.73, + "learning_rate": 1.2696012011572125e-07, + "loss": 0.6879, + "step": 19995 + }, + { + "epoch": 0.73, + "learning_rate": 1.2695645805104918e-07, + "loss": 0.6967, + "step": 19996 + }, + { + "epoch": 0.73, + "learning_rate": 1.269527959863771e-07, + "loss": 0.6966, + "step": 19997 + }, + { + "epoch": 0.73, + "learning_rate": 1.2694913392170506e-07, + "loss": 0.706, + "step": 19998 + }, + { + "epoch": 0.73, + "learning_rate": 1.2694547185703298e-07, + "loss": 0.7023, + "step": 19999 + }, + { + "epoch": 0.73, + "learning_rate": 1.2694180979236094e-07, + "loss": 0.6849, + "step": 20000 + }, + { + "epoch": 0.73, + "learning_rate": 1.2693814772768886e-07, + "loss": 0.6821, + "step": 20001 + }, + { + "epoch": 0.73, + "learning_rate": 1.2693448566301682e-07, + "loss": 0.6827, + "step": 20002 + }, + { + "epoch": 0.73, + "learning_rate": 1.2693082359834474e-07, + "loss": 0.6981, + "step": 20003 + }, + { + "epoch": 0.73, + "learning_rate": 1.269271615336727e-07, + "loss": 0.6932, + "step": 20004 + }, + { + "epoch": 0.73, + "learning_rate": 1.2692349946900062e-07, + "loss": 0.6883, + "step": 20005 + }, + { + "epoch": 0.73, + "learning_rate": 1.2691983740432855e-07, + "loss": 0.6838, + "step": 20006 + }, + { + "epoch": 0.73, + "learning_rate": 1.2691617533965648e-07, + "loss": 0.6905, + "step": 20007 + }, + { + "epoch": 0.73, + "learning_rate": 1.2691251327498443e-07, + "loss": 0.6949, + "step": 20008 + }, + { + "epoch": 0.73, + "learning_rate": 1.2690885121031236e-07, + "loss": 0.6901, + "step": 20009 + }, + { + "epoch": 0.73, + "learning_rate": 1.269051891456403e-07, + "loss": 0.6939, + "step": 20010 + }, + { + "epoch": 0.73, + "learning_rate": 1.2690152708096824e-07, + "loss": 0.7068, + "step": 20011 + }, + { + "epoch": 0.73, + "learning_rate": 1.268978650162962e-07, + "loss": 0.6826, + "step": 20012 + }, + { + "epoch": 0.73, + "learning_rate": 1.2689420295162412e-07, + "loss": 0.6901, + "step": 20013 + }, + { + "epoch": 0.73, + "learning_rate": 1.2689054088695207e-07, + "loss": 0.712, + "step": 20014 + }, + { + "epoch": 0.73, + "learning_rate": 1.2688687882228e-07, + "loss": 0.6979, + "step": 20015 + }, + { + "epoch": 0.73, + "learning_rate": 1.2688321675760795e-07, + "loss": 0.6737, + "step": 20016 + }, + { + "epoch": 0.73, + "learning_rate": 1.2687955469293587e-07, + "loss": 0.7053, + "step": 20017 + }, + { + "epoch": 0.73, + "learning_rate": 1.268758926282638e-07, + "loss": 0.6886, + "step": 20018 + }, + { + "epoch": 0.73, + "learning_rate": 1.2687223056359175e-07, + "loss": 0.6836, + "step": 20019 + }, + { + "epoch": 0.73, + "learning_rate": 1.2686856849891968e-07, + "loss": 0.6924, + "step": 20020 + }, + { + "epoch": 0.73, + "learning_rate": 1.2686490643424763e-07, + "loss": 0.6914, + "step": 20021 + }, + { + "epoch": 0.73, + "learning_rate": 1.2686124436957556e-07, + "loss": 0.7131, + "step": 20022 + }, + { + "epoch": 0.73, + "learning_rate": 1.2685758230490351e-07, + "loss": 0.6934, + "step": 20023 + }, + { + "epoch": 0.73, + "learning_rate": 1.2685392024023141e-07, + "loss": 0.694, + "step": 20024 + }, + { + "epoch": 0.73, + "learning_rate": 1.2685025817555937e-07, + "loss": 0.6991, + "step": 20025 + }, + { + "epoch": 0.73, + "learning_rate": 1.268465961108873e-07, + "loss": 0.692, + "step": 20026 + }, + { + "epoch": 0.73, + "learning_rate": 1.2684293404621525e-07, + "loss": 0.7036, + "step": 20027 + }, + { + "epoch": 0.73, + "learning_rate": 1.2683927198154317e-07, + "loss": 0.7056, + "step": 20028 + }, + { + "epoch": 0.73, + "learning_rate": 1.2683560991687113e-07, + "loss": 0.6957, + "step": 20029 + }, + { + "epoch": 0.73, + "learning_rate": 1.2683194785219905e-07, + "loss": 0.6848, + "step": 20030 + }, + { + "epoch": 0.73, + "learning_rate": 1.26828285787527e-07, + "loss": 0.6954, + "step": 20031 + }, + { + "epoch": 0.73, + "learning_rate": 1.2682462372285493e-07, + "loss": 0.687, + "step": 20032 + }, + { + "epoch": 0.73, + "learning_rate": 1.268209616581829e-07, + "loss": 0.6942, + "step": 20033 + }, + { + "epoch": 0.73, + "learning_rate": 1.2681729959351081e-07, + "loss": 0.6822, + "step": 20034 + }, + { + "epoch": 0.73, + "learning_rate": 1.2681363752883877e-07, + "loss": 0.7181, + "step": 20035 + }, + { + "epoch": 0.73, + "learning_rate": 1.268099754641667e-07, + "loss": 0.6972, + "step": 20036 + }, + { + "epoch": 0.73, + "learning_rate": 1.2680631339949465e-07, + "loss": 0.6763, + "step": 20037 + }, + { + "epoch": 0.73, + "learning_rate": 1.2680265133482257e-07, + "loss": 0.7032, + "step": 20038 + }, + { + "epoch": 0.73, + "learning_rate": 1.2679898927015053e-07, + "loss": 0.6885, + "step": 20039 + }, + { + "epoch": 0.73, + "learning_rate": 1.2679532720547845e-07, + "loss": 0.7014, + "step": 20040 + }, + { + "epoch": 0.73, + "learning_rate": 1.2679166514080638e-07, + "loss": 0.7001, + "step": 20041 + }, + { + "epoch": 0.73, + "learning_rate": 1.267880030761343e-07, + "loss": 0.6931, + "step": 20042 + }, + { + "epoch": 0.73, + "learning_rate": 1.2678434101146226e-07, + "loss": 0.7011, + "step": 20043 + }, + { + "epoch": 0.73, + "learning_rate": 1.2678067894679019e-07, + "loss": 0.6925, + "step": 20044 + }, + { + "epoch": 0.73, + "learning_rate": 1.2677701688211811e-07, + "loss": 0.6842, + "step": 20045 + }, + { + "epoch": 0.73, + "learning_rate": 1.2677335481744607e-07, + "loss": 0.7005, + "step": 20046 + }, + { + "epoch": 0.73, + "learning_rate": 1.26769692752774e-07, + "loss": 0.6919, + "step": 20047 + }, + { + "epoch": 0.73, + "learning_rate": 1.2676603068810195e-07, + "loss": 0.6929, + "step": 20048 + }, + { + "epoch": 0.73, + "learning_rate": 1.2676236862342987e-07, + "loss": 0.687, + "step": 20049 + }, + { + "epoch": 0.73, + "learning_rate": 1.2675870655875783e-07, + "loss": 0.6984, + "step": 20050 + }, + { + "epoch": 0.73, + "learning_rate": 1.2675504449408575e-07, + "loss": 0.6936, + "step": 20051 + }, + { + "epoch": 0.73, + "learning_rate": 1.267513824294137e-07, + "loss": 0.7012, + "step": 20052 + }, + { + "epoch": 0.73, + "learning_rate": 1.2674772036474163e-07, + "loss": 0.6789, + "step": 20053 + }, + { + "epoch": 0.73, + "learning_rate": 1.2674405830006959e-07, + "loss": 0.6896, + "step": 20054 + }, + { + "epoch": 0.73, + "learning_rate": 1.267403962353975e-07, + "loss": 0.692, + "step": 20055 + }, + { + "epoch": 0.73, + "learning_rate": 1.2673673417072547e-07, + "loss": 0.695, + "step": 20056 + }, + { + "epoch": 0.73, + "learning_rate": 1.267330721060534e-07, + "loss": 0.6926, + "step": 20057 + }, + { + "epoch": 0.73, + "learning_rate": 1.2672941004138135e-07, + "loss": 0.695, + "step": 20058 + }, + { + "epoch": 0.73, + "learning_rate": 1.2672574797670925e-07, + "loss": 0.6929, + "step": 20059 + }, + { + "epoch": 0.73, + "learning_rate": 1.267220859120372e-07, + "loss": 0.6924, + "step": 20060 + }, + { + "epoch": 0.73, + "learning_rate": 1.2671842384736513e-07, + "loss": 0.6937, + "step": 20061 + }, + { + "epoch": 0.73, + "learning_rate": 1.2671476178269308e-07, + "loss": 0.7027, + "step": 20062 + }, + { + "epoch": 0.73, + "learning_rate": 1.26711099718021e-07, + "loss": 0.6944, + "step": 20063 + }, + { + "epoch": 0.73, + "learning_rate": 1.2670743765334896e-07, + "loss": 0.6924, + "step": 20064 + }, + { + "epoch": 0.73, + "learning_rate": 1.2670377558867689e-07, + "loss": 0.6872, + "step": 20065 + }, + { + "epoch": 0.73, + "learning_rate": 1.2670011352400484e-07, + "loss": 0.6959, + "step": 20066 + }, + { + "epoch": 0.73, + "learning_rate": 1.2669645145933277e-07, + "loss": 0.6998, + "step": 20067 + }, + { + "epoch": 0.73, + "learning_rate": 1.266927893946607e-07, + "loss": 0.6901, + "step": 20068 + }, + { + "epoch": 0.73, + "learning_rate": 1.2668912732998865e-07, + "loss": 0.6914, + "step": 20069 + }, + { + "epoch": 0.73, + "learning_rate": 1.2668546526531657e-07, + "loss": 0.6768, + "step": 20070 + }, + { + "epoch": 0.73, + "learning_rate": 1.2668180320064453e-07, + "loss": 0.6992, + "step": 20071 + }, + { + "epoch": 0.73, + "learning_rate": 1.2667814113597245e-07, + "loss": 0.714, + "step": 20072 + }, + { + "epoch": 0.73, + "learning_rate": 1.266744790713004e-07, + "loss": 0.6991, + "step": 20073 + }, + { + "epoch": 0.73, + "learning_rate": 1.2667081700662833e-07, + "loss": 0.6919, + "step": 20074 + }, + { + "epoch": 0.73, + "learning_rate": 1.2666715494195628e-07, + "loss": 0.6939, + "step": 20075 + }, + { + "epoch": 0.73, + "learning_rate": 1.266634928772842e-07, + "loss": 0.6905, + "step": 20076 + }, + { + "epoch": 0.73, + "learning_rate": 1.2665983081261214e-07, + "loss": 0.6823, + "step": 20077 + }, + { + "epoch": 0.73, + "learning_rate": 1.2665616874794006e-07, + "loss": 0.7012, + "step": 20078 + }, + { + "epoch": 0.73, + "learning_rate": 1.2665250668326802e-07, + "loss": 0.6885, + "step": 20079 + }, + { + "epoch": 0.73, + "learning_rate": 1.2664884461859594e-07, + "loss": 0.6916, + "step": 20080 + }, + { + "epoch": 0.73, + "learning_rate": 1.266451825539239e-07, + "loss": 0.6988, + "step": 20081 + }, + { + "epoch": 0.73, + "learning_rate": 1.2664152048925182e-07, + "loss": 0.6837, + "step": 20082 + }, + { + "epoch": 0.73, + "learning_rate": 1.2663785842457978e-07, + "loss": 0.7116, + "step": 20083 + }, + { + "epoch": 0.73, + "learning_rate": 1.266341963599077e-07, + "loss": 0.7075, + "step": 20084 + }, + { + "epoch": 0.73, + "learning_rate": 1.2663053429523566e-07, + "loss": 0.6882, + "step": 20085 + }, + { + "epoch": 0.73, + "learning_rate": 1.2662687223056358e-07, + "loss": 0.7132, + "step": 20086 + }, + { + "epoch": 0.73, + "learning_rate": 1.2662321016589154e-07, + "loss": 0.6934, + "step": 20087 + }, + { + "epoch": 0.73, + "learning_rate": 1.2661954810121946e-07, + "loss": 0.7245, + "step": 20088 + }, + { + "epoch": 0.73, + "learning_rate": 1.266158860365474e-07, + "loss": 0.7055, + "step": 20089 + }, + { + "epoch": 0.74, + "learning_rate": 1.2661222397187534e-07, + "loss": 0.6865, + "step": 20090 + }, + { + "epoch": 0.74, + "learning_rate": 1.2660856190720327e-07, + "loss": 0.7038, + "step": 20091 + }, + { + "epoch": 0.74, + "learning_rate": 1.2660489984253122e-07, + "loss": 0.6981, + "step": 20092 + }, + { + "epoch": 0.74, + "learning_rate": 1.2660123777785915e-07, + "loss": 0.6856, + "step": 20093 + }, + { + "epoch": 0.74, + "learning_rate": 1.265975757131871e-07, + "loss": 0.6944, + "step": 20094 + }, + { + "epoch": 0.74, + "learning_rate": 1.26593913648515e-07, + "loss": 0.6901, + "step": 20095 + }, + { + "epoch": 0.74, + "learning_rate": 1.2659025158384296e-07, + "loss": 0.6906, + "step": 20096 + }, + { + "epoch": 0.74, + "learning_rate": 1.2658658951917088e-07, + "loss": 0.6971, + "step": 20097 + }, + { + "epoch": 0.74, + "learning_rate": 1.2658292745449884e-07, + "loss": 0.6945, + "step": 20098 + }, + { + "epoch": 0.74, + "learning_rate": 1.2657926538982676e-07, + "loss": 0.7011, + "step": 20099 + }, + { + "epoch": 0.74, + "learning_rate": 1.2657560332515472e-07, + "loss": 0.6733, + "step": 20100 + }, + { + "epoch": 0.74, + "learning_rate": 1.2657194126048264e-07, + "loss": 0.706, + "step": 20101 + }, + { + "epoch": 0.74, + "learning_rate": 1.265682791958106e-07, + "loss": 0.7021, + "step": 20102 + }, + { + "epoch": 0.74, + "learning_rate": 1.2656461713113852e-07, + "loss": 0.6895, + "step": 20103 + }, + { + "epoch": 0.74, + "learning_rate": 1.2656095506646648e-07, + "loss": 0.6996, + "step": 20104 + }, + { + "epoch": 0.74, + "learning_rate": 1.265572930017944e-07, + "loss": 0.7046, + "step": 20105 + }, + { + "epoch": 0.74, + "learning_rate": 1.2655363093712236e-07, + "loss": 0.6984, + "step": 20106 + }, + { + "epoch": 0.74, + "learning_rate": 1.2654996887245028e-07, + "loss": 0.7048, + "step": 20107 + }, + { + "epoch": 0.74, + "learning_rate": 1.2654630680777824e-07, + "loss": 0.6884, + "step": 20108 + }, + { + "epoch": 0.74, + "learning_rate": 1.2654264474310616e-07, + "loss": 0.679, + "step": 20109 + }, + { + "epoch": 0.74, + "learning_rate": 1.2653898267843412e-07, + "loss": 0.6828, + "step": 20110 + }, + { + "epoch": 0.74, + "learning_rate": 1.2653532061376204e-07, + "loss": 0.6873, + "step": 20111 + }, + { + "epoch": 0.74, + "learning_rate": 1.2653165854908997e-07, + "loss": 0.7012, + "step": 20112 + }, + { + "epoch": 0.74, + "learning_rate": 1.265279964844179e-07, + "loss": 0.7053, + "step": 20113 + }, + { + "epoch": 0.74, + "learning_rate": 1.2652433441974585e-07, + "loss": 0.6921, + "step": 20114 + }, + { + "epoch": 0.74, + "learning_rate": 1.2652067235507378e-07, + "loss": 0.689, + "step": 20115 + }, + { + "epoch": 0.74, + "learning_rate": 1.265170102904017e-07, + "loss": 0.6919, + "step": 20116 + }, + { + "epoch": 0.74, + "learning_rate": 1.2651334822572966e-07, + "loss": 0.6941, + "step": 20117 + }, + { + "epoch": 0.74, + "learning_rate": 1.2650968616105758e-07, + "loss": 0.7078, + "step": 20118 + }, + { + "epoch": 0.74, + "learning_rate": 1.2650602409638554e-07, + "loss": 0.7006, + "step": 20119 + }, + { + "epoch": 0.74, + "learning_rate": 1.2650236203171346e-07, + "loss": 0.6837, + "step": 20120 + }, + { + "epoch": 0.74, + "learning_rate": 1.2649869996704142e-07, + "loss": 0.6794, + "step": 20121 + }, + { + "epoch": 0.74, + "learning_rate": 1.2649503790236934e-07, + "loss": 0.6978, + "step": 20122 + }, + { + "epoch": 0.74, + "learning_rate": 1.264913758376973e-07, + "loss": 0.6893, + "step": 20123 + }, + { + "epoch": 0.74, + "learning_rate": 1.2648771377302522e-07, + "loss": 0.683, + "step": 20124 + }, + { + "epoch": 0.74, + "learning_rate": 1.2648405170835318e-07, + "loss": 0.7103, + "step": 20125 + }, + { + "epoch": 0.74, + "learning_rate": 1.264803896436811e-07, + "loss": 0.7001, + "step": 20126 + }, + { + "epoch": 0.74, + "learning_rate": 1.2647672757900906e-07, + "loss": 0.7022, + "step": 20127 + }, + { + "epoch": 0.74, + "learning_rate": 1.2647306551433698e-07, + "loss": 0.6956, + "step": 20128 + }, + { + "epoch": 0.74, + "learning_rate": 1.2646940344966494e-07, + "loss": 0.696, + "step": 20129 + }, + { + "epoch": 0.74, + "learning_rate": 1.2646574138499286e-07, + "loss": 0.687, + "step": 20130 + }, + { + "epoch": 0.74, + "learning_rate": 1.264620793203208e-07, + "loss": 0.6921, + "step": 20131 + }, + { + "epoch": 0.74, + "learning_rate": 1.2645841725564872e-07, + "loss": 0.6864, + "step": 20132 + }, + { + "epoch": 0.74, + "learning_rate": 1.2645475519097667e-07, + "loss": 0.6829, + "step": 20133 + }, + { + "epoch": 0.74, + "learning_rate": 1.264510931263046e-07, + "loss": 0.6899, + "step": 20134 + }, + { + "epoch": 0.74, + "learning_rate": 1.2644743106163255e-07, + "loss": 0.6839, + "step": 20135 + }, + { + "epoch": 0.74, + "learning_rate": 1.2644376899696047e-07, + "loss": 0.6891, + "step": 20136 + }, + { + "epoch": 0.74, + "learning_rate": 1.2644010693228843e-07, + "loss": 0.6922, + "step": 20137 + }, + { + "epoch": 0.74, + "learning_rate": 1.2643644486761635e-07, + "loss": 0.7069, + "step": 20138 + }, + { + "epoch": 0.74, + "learning_rate": 1.2643278280294428e-07, + "loss": 0.7005, + "step": 20139 + }, + { + "epoch": 0.74, + "learning_rate": 1.2642912073827223e-07, + "loss": 0.6898, + "step": 20140 + }, + { + "epoch": 0.74, + "learning_rate": 1.2642545867360016e-07, + "loss": 0.6885, + "step": 20141 + }, + { + "epoch": 0.74, + "learning_rate": 1.2642179660892811e-07, + "loss": 0.6833, + "step": 20142 + }, + { + "epoch": 0.74, + "learning_rate": 1.2641813454425604e-07, + "loss": 0.6794, + "step": 20143 + }, + { + "epoch": 0.74, + "learning_rate": 1.26414472479584e-07, + "loss": 0.6896, + "step": 20144 + }, + { + "epoch": 0.74, + "learning_rate": 1.2641081041491192e-07, + "loss": 0.6896, + "step": 20145 + }, + { + "epoch": 0.74, + "learning_rate": 1.2640714835023987e-07, + "loss": 0.6948, + "step": 20146 + }, + { + "epoch": 0.74, + "learning_rate": 1.264034862855678e-07, + "loss": 0.6985, + "step": 20147 + }, + { + "epoch": 0.74, + "learning_rate": 1.2639982422089575e-07, + "loss": 0.6949, + "step": 20148 + }, + { + "epoch": 0.74, + "learning_rate": 1.2639616215622365e-07, + "loss": 0.692, + "step": 20149 + }, + { + "epoch": 0.74, + "learning_rate": 1.263925000915516e-07, + "loss": 0.7084, + "step": 20150 + }, + { + "epoch": 0.74, + "learning_rate": 1.2638883802687953e-07, + "loss": 0.7021, + "step": 20151 + }, + { + "epoch": 0.74, + "learning_rate": 1.263851759622075e-07, + "loss": 0.6968, + "step": 20152 + }, + { + "epoch": 0.74, + "learning_rate": 1.2638151389753541e-07, + "loss": 0.6981, + "step": 20153 + }, + { + "epoch": 0.74, + "learning_rate": 1.2637785183286337e-07, + "loss": 0.6975, + "step": 20154 + }, + { + "epoch": 0.74, + "learning_rate": 1.263741897681913e-07, + "loss": 0.7015, + "step": 20155 + }, + { + "epoch": 0.74, + "learning_rate": 1.2637052770351925e-07, + "loss": 0.6955, + "step": 20156 + }, + { + "epoch": 0.74, + "learning_rate": 1.2636686563884717e-07, + "loss": 0.7036, + "step": 20157 + }, + { + "epoch": 0.74, + "learning_rate": 1.2636320357417513e-07, + "loss": 0.7008, + "step": 20158 + }, + { + "epoch": 0.74, + "learning_rate": 1.2635954150950305e-07, + "loss": 0.7094, + "step": 20159 + }, + { + "epoch": 0.74, + "learning_rate": 1.2635587944483098e-07, + "loss": 0.6911, + "step": 20160 + }, + { + "epoch": 0.74, + "learning_rate": 1.2635221738015893e-07, + "loss": 0.6909, + "step": 20161 + }, + { + "epoch": 0.74, + "learning_rate": 1.2634855531548686e-07, + "loss": 0.7041, + "step": 20162 + }, + { + "epoch": 0.74, + "learning_rate": 1.2634489325081481e-07, + "loss": 0.6823, + "step": 20163 + }, + { + "epoch": 0.74, + "learning_rate": 1.2634123118614274e-07, + "loss": 0.6981, + "step": 20164 + }, + { + "epoch": 0.74, + "learning_rate": 1.263375691214707e-07, + "loss": 0.6943, + "step": 20165 + }, + { + "epoch": 0.74, + "learning_rate": 1.2633390705679862e-07, + "loss": 0.6696, + "step": 20166 + }, + { + "epoch": 0.74, + "learning_rate": 1.2633024499212655e-07, + "loss": 0.7104, + "step": 20167 + }, + { + "epoch": 0.74, + "learning_rate": 1.2632658292745447e-07, + "loss": 0.6851, + "step": 20168 + }, + { + "epoch": 0.74, + "learning_rate": 1.2632292086278243e-07, + "loss": 0.6962, + "step": 20169 + }, + { + "epoch": 0.74, + "learning_rate": 1.2631925879811035e-07, + "loss": 0.6856, + "step": 20170 + }, + { + "epoch": 0.74, + "learning_rate": 1.263155967334383e-07, + "loss": 0.6923, + "step": 20171 + }, + { + "epoch": 0.74, + "learning_rate": 1.2631193466876623e-07, + "loss": 0.6798, + "step": 20172 + }, + { + "epoch": 0.74, + "learning_rate": 1.2630827260409419e-07, + "loss": 0.6803, + "step": 20173 + }, + { + "epoch": 0.74, + "learning_rate": 1.263046105394221e-07, + "loss": 0.6978, + "step": 20174 + }, + { + "epoch": 0.74, + "learning_rate": 1.2630094847475007e-07, + "loss": 0.6968, + "step": 20175 + }, + { + "epoch": 0.74, + "learning_rate": 1.26297286410078e-07, + "loss": 0.6916, + "step": 20176 + }, + { + "epoch": 0.74, + "learning_rate": 1.2629362434540595e-07, + "loss": 0.6973, + "step": 20177 + }, + { + "epoch": 0.74, + "learning_rate": 1.2628996228073387e-07, + "loss": 0.6956, + "step": 20178 + }, + { + "epoch": 0.74, + "learning_rate": 1.2628630021606183e-07, + "loss": 0.6882, + "step": 20179 + }, + { + "epoch": 0.74, + "learning_rate": 1.2628263815138975e-07, + "loss": 0.6892, + "step": 20180 + }, + { + "epoch": 0.74, + "learning_rate": 1.262789760867177e-07, + "loss": 0.6874, + "step": 20181 + }, + { + "epoch": 0.74, + "learning_rate": 1.2627531402204563e-07, + "loss": 0.7175, + "step": 20182 + }, + { + "epoch": 0.74, + "learning_rate": 1.2627165195737356e-07, + "loss": 0.6907, + "step": 20183 + }, + { + "epoch": 0.74, + "learning_rate": 1.262679898927015e-07, + "loss": 0.6921, + "step": 20184 + }, + { + "epoch": 0.74, + "learning_rate": 1.2626432782802944e-07, + "loss": 0.6946, + "step": 20185 + }, + { + "epoch": 0.74, + "learning_rate": 1.2626066576335737e-07, + "loss": 0.7047, + "step": 20186 + }, + { + "epoch": 0.74, + "learning_rate": 1.262570036986853e-07, + "loss": 0.694, + "step": 20187 + }, + { + "epoch": 0.74, + "learning_rate": 1.2625334163401325e-07, + "loss": 0.6973, + "step": 20188 + }, + { + "epoch": 0.74, + "learning_rate": 1.2624967956934117e-07, + "loss": 0.6934, + "step": 20189 + }, + { + "epoch": 0.74, + "learning_rate": 1.2624601750466913e-07, + "loss": 0.704, + "step": 20190 + }, + { + "epoch": 0.74, + "learning_rate": 1.2624235543999705e-07, + "loss": 0.7008, + "step": 20191 + }, + { + "epoch": 0.74, + "learning_rate": 1.26238693375325e-07, + "loss": 0.693, + "step": 20192 + }, + { + "epoch": 0.74, + "learning_rate": 1.2623503131065293e-07, + "loss": 0.6945, + "step": 20193 + }, + { + "epoch": 0.74, + "learning_rate": 1.2623136924598088e-07, + "loss": 0.6922, + "step": 20194 + }, + { + "epoch": 0.74, + "learning_rate": 1.262277071813088e-07, + "loss": 0.6767, + "step": 20195 + }, + { + "epoch": 0.74, + "learning_rate": 1.2622404511663676e-07, + "loss": 0.6811, + "step": 20196 + }, + { + "epoch": 0.74, + "learning_rate": 1.262203830519647e-07, + "loss": 0.6865, + "step": 20197 + }, + { + "epoch": 0.74, + "learning_rate": 1.2621672098729264e-07, + "loss": 0.6839, + "step": 20198 + }, + { + "epoch": 0.74, + "learning_rate": 1.2621305892262057e-07, + "loss": 0.6966, + "step": 20199 + }, + { + "epoch": 0.74, + "learning_rate": 1.2620939685794852e-07, + "loss": 0.6986, + "step": 20200 + }, + { + "epoch": 0.74, + "learning_rate": 1.2620573479327645e-07, + "loss": 0.6953, + "step": 20201 + }, + { + "epoch": 0.74, + "learning_rate": 1.262020727286044e-07, + "loss": 0.6903, + "step": 20202 + }, + { + "epoch": 0.74, + "learning_rate": 1.261984106639323e-07, + "loss": 0.6852, + "step": 20203 + }, + { + "epoch": 0.74, + "learning_rate": 1.2619474859926026e-07, + "loss": 0.681, + "step": 20204 + }, + { + "epoch": 0.74, + "learning_rate": 1.2619108653458818e-07, + "loss": 0.6785, + "step": 20205 + }, + { + "epoch": 0.74, + "learning_rate": 1.2618742446991614e-07, + "loss": 0.6751, + "step": 20206 + }, + { + "epoch": 0.74, + "learning_rate": 1.2618376240524406e-07, + "loss": 0.6764, + "step": 20207 + }, + { + "epoch": 0.74, + "learning_rate": 1.2618010034057202e-07, + "loss": 0.6818, + "step": 20208 + }, + { + "epoch": 0.74, + "learning_rate": 1.2617643827589994e-07, + "loss": 0.6888, + "step": 20209 + }, + { + "epoch": 0.74, + "learning_rate": 1.2617277621122787e-07, + "loss": 0.6903, + "step": 20210 + }, + { + "epoch": 0.74, + "learning_rate": 1.2616911414655582e-07, + "loss": 0.6902, + "step": 20211 + }, + { + "epoch": 0.74, + "learning_rate": 1.2616545208188375e-07, + "loss": 0.7053, + "step": 20212 + }, + { + "epoch": 0.74, + "learning_rate": 1.261617900172117e-07, + "loss": 0.6889, + "step": 20213 + }, + { + "epoch": 0.74, + "learning_rate": 1.2615812795253963e-07, + "loss": 0.6862, + "step": 20214 + }, + { + "epoch": 0.74, + "learning_rate": 1.2615446588786758e-07, + "loss": 0.7175, + "step": 20215 + }, + { + "epoch": 0.74, + "learning_rate": 1.261508038231955e-07, + "loss": 0.6898, + "step": 20216 + }, + { + "epoch": 0.74, + "learning_rate": 1.2614714175852346e-07, + "loss": 0.7018, + "step": 20217 + }, + { + "epoch": 0.74, + "learning_rate": 1.261434796938514e-07, + "loss": 0.6888, + "step": 20218 + }, + { + "epoch": 0.74, + "learning_rate": 1.2613981762917934e-07, + "loss": 0.6946, + "step": 20219 + }, + { + "epoch": 0.74, + "learning_rate": 1.2613615556450727e-07, + "loss": 0.7051, + "step": 20220 + }, + { + "epoch": 0.74, + "learning_rate": 1.261324934998352e-07, + "loss": 0.7058, + "step": 20221 + }, + { + "epoch": 0.74, + "learning_rate": 1.2612883143516312e-07, + "loss": 0.6961, + "step": 20222 + }, + { + "epoch": 0.74, + "learning_rate": 1.2612516937049108e-07, + "loss": 0.6918, + "step": 20223 + }, + { + "epoch": 0.74, + "learning_rate": 1.26121507305819e-07, + "loss": 0.6956, + "step": 20224 + }, + { + "epoch": 0.74, + "learning_rate": 1.2611784524114696e-07, + "loss": 0.6814, + "step": 20225 + }, + { + "epoch": 0.74, + "learning_rate": 1.2611418317647488e-07, + "loss": 0.7002, + "step": 20226 + }, + { + "epoch": 0.74, + "learning_rate": 1.2611052111180284e-07, + "loss": 0.6909, + "step": 20227 + }, + { + "epoch": 0.74, + "learning_rate": 1.2610685904713076e-07, + "loss": 0.6967, + "step": 20228 + }, + { + "epoch": 0.74, + "learning_rate": 1.2610319698245872e-07, + "loss": 0.6939, + "step": 20229 + }, + { + "epoch": 0.74, + "learning_rate": 1.2609953491778664e-07, + "loss": 0.6989, + "step": 20230 + }, + { + "epoch": 0.74, + "learning_rate": 1.2609587285311457e-07, + "loss": 0.6929, + "step": 20231 + }, + { + "epoch": 0.74, + "learning_rate": 1.2609221078844252e-07, + "loss": 0.6871, + "step": 20232 + }, + { + "epoch": 0.74, + "learning_rate": 1.2608854872377045e-07, + "loss": 0.6871, + "step": 20233 + }, + { + "epoch": 0.74, + "learning_rate": 1.260848866590984e-07, + "loss": 0.6831, + "step": 20234 + }, + { + "epoch": 0.74, + "learning_rate": 1.2608122459442633e-07, + "loss": 0.678, + "step": 20235 + }, + { + "epoch": 0.74, + "learning_rate": 1.2607756252975428e-07, + "loss": 0.6938, + "step": 20236 + }, + { + "epoch": 0.74, + "learning_rate": 1.260739004650822e-07, + "loss": 0.6956, + "step": 20237 + }, + { + "epoch": 0.74, + "learning_rate": 1.2607023840041016e-07, + "loss": 0.6902, + "step": 20238 + }, + { + "epoch": 0.74, + "learning_rate": 1.2606657633573806e-07, + "loss": 0.696, + "step": 20239 + }, + { + "epoch": 0.74, + "learning_rate": 1.2606291427106602e-07, + "loss": 0.7035, + "step": 20240 + }, + { + "epoch": 0.74, + "learning_rate": 1.2605925220639394e-07, + "loss": 0.6916, + "step": 20241 + }, + { + "epoch": 0.74, + "learning_rate": 1.260555901417219e-07, + "loss": 0.6959, + "step": 20242 + }, + { + "epoch": 0.74, + "learning_rate": 1.2605192807704982e-07, + "loss": 0.6895, + "step": 20243 + }, + { + "epoch": 0.74, + "learning_rate": 1.2604826601237778e-07, + "loss": 0.6813, + "step": 20244 + }, + { + "epoch": 0.74, + "learning_rate": 1.260446039477057e-07, + "loss": 0.6897, + "step": 20245 + }, + { + "epoch": 0.74, + "learning_rate": 1.2604094188303366e-07, + "loss": 0.6877, + "step": 20246 + }, + { + "epoch": 0.74, + "learning_rate": 1.2603727981836158e-07, + "loss": 0.7006, + "step": 20247 + }, + { + "epoch": 0.74, + "learning_rate": 1.2603361775368954e-07, + "loss": 0.6982, + "step": 20248 + }, + { + "epoch": 0.74, + "learning_rate": 1.2602995568901746e-07, + "loss": 0.6937, + "step": 20249 + }, + { + "epoch": 0.74, + "learning_rate": 1.2602629362434541e-07, + "loss": 0.6923, + "step": 20250 + }, + { + "epoch": 0.74, + "learning_rate": 1.2602263155967334e-07, + "loss": 0.6752, + "step": 20251 + }, + { + "epoch": 0.74, + "learning_rate": 1.2601896949500127e-07, + "loss": 0.6905, + "step": 20252 + }, + { + "epoch": 0.74, + "learning_rate": 1.2601530743032922e-07, + "loss": 0.7034, + "step": 20253 + }, + { + "epoch": 0.74, + "learning_rate": 1.2601164536565715e-07, + "loss": 0.6902, + "step": 20254 + }, + { + "epoch": 0.74, + "learning_rate": 1.260079833009851e-07, + "loss": 0.7008, + "step": 20255 + }, + { + "epoch": 0.74, + "learning_rate": 1.2600432123631303e-07, + "loss": 0.677, + "step": 20256 + }, + { + "epoch": 0.74, + "learning_rate": 1.2600065917164095e-07, + "loss": 0.6833, + "step": 20257 + }, + { + "epoch": 0.74, + "learning_rate": 1.2599699710696888e-07, + "loss": 0.6983, + "step": 20258 + }, + { + "epoch": 0.74, + "learning_rate": 1.2599333504229683e-07, + "loss": 0.6945, + "step": 20259 + }, + { + "epoch": 0.74, + "learning_rate": 1.2598967297762476e-07, + "loss": 0.6975, + "step": 20260 + }, + { + "epoch": 0.74, + "learning_rate": 1.2598601091295271e-07, + "loss": 0.6937, + "step": 20261 + }, + { + "epoch": 0.74, + "learning_rate": 1.2598234884828064e-07, + "loss": 0.6919, + "step": 20262 + }, + { + "epoch": 0.74, + "learning_rate": 1.259786867836086e-07, + "loss": 0.6926, + "step": 20263 + }, + { + "epoch": 0.74, + "learning_rate": 1.2597502471893652e-07, + "loss": 0.7019, + "step": 20264 + }, + { + "epoch": 0.74, + "learning_rate": 1.2597136265426447e-07, + "loss": 0.7016, + "step": 20265 + }, + { + "epoch": 0.74, + "learning_rate": 1.259677005895924e-07, + "loss": 0.6895, + "step": 20266 + }, + { + "epoch": 0.74, + "learning_rate": 1.2596403852492035e-07, + "loss": 0.6994, + "step": 20267 + }, + { + "epoch": 0.74, + "learning_rate": 1.2596037646024828e-07, + "loss": 0.703, + "step": 20268 + }, + { + "epoch": 0.74, + "learning_rate": 1.2595671439557623e-07, + "loss": 0.6876, + "step": 20269 + }, + { + "epoch": 0.74, + "learning_rate": 1.2595305233090416e-07, + "loss": 0.6799, + "step": 20270 + }, + { + "epoch": 0.74, + "learning_rate": 1.2594939026623211e-07, + "loss": 0.6913, + "step": 20271 + }, + { + "epoch": 0.74, + "learning_rate": 1.2594572820156004e-07, + "loss": 0.6885, + "step": 20272 + }, + { + "epoch": 0.74, + "learning_rate": 1.25942066136888e-07, + "loss": 0.6863, + "step": 20273 + }, + { + "epoch": 0.74, + "learning_rate": 1.2593840407221592e-07, + "loss": 0.6906, + "step": 20274 + }, + { + "epoch": 0.74, + "learning_rate": 1.2593474200754385e-07, + "loss": 0.6885, + "step": 20275 + }, + { + "epoch": 0.74, + "learning_rate": 1.2593107994287177e-07, + "loss": 0.688, + "step": 20276 + }, + { + "epoch": 0.74, + "learning_rate": 1.2592741787819973e-07, + "loss": 0.6884, + "step": 20277 + }, + { + "epoch": 0.74, + "learning_rate": 1.2592375581352765e-07, + "loss": 0.6805, + "step": 20278 + }, + { + "epoch": 0.74, + "learning_rate": 1.259200937488556e-07, + "loss": 0.7084, + "step": 20279 + }, + { + "epoch": 0.74, + "learning_rate": 1.2591643168418353e-07, + "loss": 0.6846, + "step": 20280 + }, + { + "epoch": 0.74, + "learning_rate": 1.2591276961951146e-07, + "loss": 0.6956, + "step": 20281 + }, + { + "epoch": 0.74, + "learning_rate": 1.2590910755483941e-07, + "loss": 0.6942, + "step": 20282 + }, + { + "epoch": 0.74, + "learning_rate": 1.2590544549016734e-07, + "loss": 0.6945, + "step": 20283 + }, + { + "epoch": 0.74, + "learning_rate": 1.259017834254953e-07, + "loss": 0.6964, + "step": 20284 + }, + { + "epoch": 0.74, + "learning_rate": 1.2589812136082322e-07, + "loss": 0.6987, + "step": 20285 + }, + { + "epoch": 0.74, + "learning_rate": 1.2589445929615117e-07, + "loss": 0.7066, + "step": 20286 + }, + { + "epoch": 0.74, + "learning_rate": 1.258907972314791e-07, + "loss": 0.6871, + "step": 20287 + }, + { + "epoch": 0.74, + "learning_rate": 1.2588713516680705e-07, + "loss": 0.6786, + "step": 20288 + }, + { + "epoch": 0.74, + "learning_rate": 1.2588347310213498e-07, + "loss": 0.6895, + "step": 20289 + }, + { + "epoch": 0.74, + "learning_rate": 1.2587981103746293e-07, + "loss": 0.6957, + "step": 20290 + }, + { + "epoch": 0.74, + "learning_rate": 1.2587614897279086e-07, + "loss": 0.7015, + "step": 20291 + }, + { + "epoch": 0.74, + "learning_rate": 1.258724869081188e-07, + "loss": 0.6992, + "step": 20292 + }, + { + "epoch": 0.74, + "learning_rate": 1.258688248434467e-07, + "loss": 0.6755, + "step": 20293 + }, + { + "epoch": 0.74, + "learning_rate": 1.2586516277877467e-07, + "loss": 0.6838, + "step": 20294 + }, + { + "epoch": 0.74, + "learning_rate": 1.258615007141026e-07, + "loss": 0.6882, + "step": 20295 + }, + { + "epoch": 0.74, + "learning_rate": 1.2585783864943055e-07, + "loss": 0.6837, + "step": 20296 + }, + { + "epoch": 0.74, + "learning_rate": 1.2585417658475847e-07, + "loss": 0.7004, + "step": 20297 + }, + { + "epoch": 0.74, + "learning_rate": 1.2585051452008643e-07, + "loss": 0.6916, + "step": 20298 + }, + { + "epoch": 0.74, + "learning_rate": 1.2584685245541435e-07, + "loss": 0.6775, + "step": 20299 + }, + { + "epoch": 0.74, + "learning_rate": 1.258431903907423e-07, + "loss": 0.6907, + "step": 20300 + }, + { + "epoch": 0.74, + "learning_rate": 1.2583952832607023e-07, + "loss": 0.6778, + "step": 20301 + }, + { + "epoch": 0.74, + "learning_rate": 1.2583586626139816e-07, + "loss": 0.6987, + "step": 20302 + }, + { + "epoch": 0.74, + "learning_rate": 1.258322041967261e-07, + "loss": 0.7101, + "step": 20303 + }, + { + "epoch": 0.74, + "learning_rate": 1.2582854213205404e-07, + "loss": 0.6864, + "step": 20304 + }, + { + "epoch": 0.74, + "learning_rate": 1.25824880067382e-07, + "loss": 0.7027, + "step": 20305 + }, + { + "epoch": 0.74, + "learning_rate": 1.2582121800270992e-07, + "loss": 0.6956, + "step": 20306 + }, + { + "epoch": 0.74, + "learning_rate": 1.2581755593803787e-07, + "loss": 0.6928, + "step": 20307 + }, + { + "epoch": 0.74, + "learning_rate": 1.258138938733658e-07, + "loss": 0.6904, + "step": 20308 + }, + { + "epoch": 0.74, + "learning_rate": 1.2581023180869375e-07, + "loss": 0.6857, + "step": 20309 + }, + { + "epoch": 0.74, + "learning_rate": 1.2580656974402168e-07, + "loss": 0.699, + "step": 20310 + }, + { + "epoch": 0.74, + "learning_rate": 1.258029076793496e-07, + "loss": 0.7024, + "step": 20311 + }, + { + "epoch": 0.74, + "learning_rate": 1.2579924561467753e-07, + "loss": 0.7063, + "step": 20312 + }, + { + "epoch": 0.74, + "learning_rate": 1.2579558355000548e-07, + "loss": 0.682, + "step": 20313 + }, + { + "epoch": 0.74, + "learning_rate": 1.257919214853334e-07, + "loss": 0.7053, + "step": 20314 + }, + { + "epoch": 0.74, + "learning_rate": 1.2578825942066136e-07, + "loss": 0.7005, + "step": 20315 + }, + { + "epoch": 0.74, + "learning_rate": 1.257845973559893e-07, + "loss": 0.6879, + "step": 20316 + }, + { + "epoch": 0.74, + "learning_rate": 1.2578093529131724e-07, + "loss": 0.7021, + "step": 20317 + }, + { + "epoch": 0.74, + "learning_rate": 1.2577727322664517e-07, + "loss": 0.7067, + "step": 20318 + }, + { + "epoch": 0.74, + "learning_rate": 1.2577361116197312e-07, + "loss": 0.7033, + "step": 20319 + }, + { + "epoch": 0.74, + "learning_rate": 1.2576994909730105e-07, + "loss": 0.6881, + "step": 20320 + }, + { + "epoch": 0.74, + "learning_rate": 1.25766287032629e-07, + "loss": 0.6884, + "step": 20321 + }, + { + "epoch": 0.74, + "learning_rate": 1.2576262496795693e-07, + "loss": 0.6899, + "step": 20322 + }, + { + "epoch": 0.74, + "learning_rate": 1.2575896290328486e-07, + "loss": 0.689, + "step": 20323 + }, + { + "epoch": 0.74, + "learning_rate": 1.257553008386128e-07, + "loss": 0.7129, + "step": 20324 + }, + { + "epoch": 0.74, + "learning_rate": 1.2575163877394074e-07, + "loss": 0.6984, + "step": 20325 + }, + { + "epoch": 0.74, + "learning_rate": 1.257479767092687e-07, + "loss": 0.6924, + "step": 20326 + }, + { + "epoch": 0.74, + "learning_rate": 1.2574431464459662e-07, + "loss": 0.7047, + "step": 20327 + }, + { + "epoch": 0.74, + "learning_rate": 1.2574065257992457e-07, + "loss": 0.6897, + "step": 20328 + }, + { + "epoch": 0.74, + "learning_rate": 1.2573699051525247e-07, + "loss": 0.6841, + "step": 20329 + }, + { + "epoch": 0.74, + "learning_rate": 1.2573332845058042e-07, + "loss": 0.7063, + "step": 20330 + }, + { + "epoch": 0.74, + "learning_rate": 1.2572966638590835e-07, + "loss": 0.702, + "step": 20331 + }, + { + "epoch": 0.74, + "learning_rate": 1.257260043212363e-07, + "loss": 0.6971, + "step": 20332 + }, + { + "epoch": 0.74, + "learning_rate": 1.2572234225656423e-07, + "loss": 0.689, + "step": 20333 + }, + { + "epoch": 0.74, + "learning_rate": 1.2571868019189218e-07, + "loss": 0.686, + "step": 20334 + }, + { + "epoch": 0.74, + "learning_rate": 1.257150181272201e-07, + "loss": 0.7094, + "step": 20335 + }, + { + "epoch": 0.74, + "learning_rate": 1.2571135606254806e-07, + "loss": 0.6858, + "step": 20336 + }, + { + "epoch": 0.74, + "learning_rate": 1.25707693997876e-07, + "loss": 0.6876, + "step": 20337 + }, + { + "epoch": 0.74, + "learning_rate": 1.2570403193320394e-07, + "loss": 0.6941, + "step": 20338 + }, + { + "epoch": 0.74, + "learning_rate": 1.2570036986853187e-07, + "loss": 0.6927, + "step": 20339 + }, + { + "epoch": 0.74, + "learning_rate": 1.2569670780385982e-07, + "loss": 0.6958, + "step": 20340 + }, + { + "epoch": 0.74, + "learning_rate": 1.2569304573918775e-07, + "loss": 0.6926, + "step": 20341 + }, + { + "epoch": 0.74, + "learning_rate": 1.256893836745157e-07, + "loss": 0.6851, + "step": 20342 + }, + { + "epoch": 0.74, + "learning_rate": 1.2568572160984363e-07, + "loss": 0.6996, + "step": 20343 + }, + { + "epoch": 0.74, + "learning_rate": 1.2568205954517158e-07, + "loss": 0.6963, + "step": 20344 + }, + { + "epoch": 0.74, + "learning_rate": 1.256783974804995e-07, + "loss": 0.6985, + "step": 20345 + }, + { + "epoch": 0.74, + "learning_rate": 1.2567473541582744e-07, + "loss": 0.6997, + "step": 20346 + }, + { + "epoch": 0.74, + "learning_rate": 1.2567107335115536e-07, + "loss": 0.6881, + "step": 20347 + }, + { + "epoch": 0.74, + "learning_rate": 1.2566741128648332e-07, + "loss": 0.6822, + "step": 20348 + }, + { + "epoch": 0.74, + "learning_rate": 1.2566374922181124e-07, + "loss": 0.6866, + "step": 20349 + }, + { + "epoch": 0.74, + "learning_rate": 1.256600871571392e-07, + "loss": 0.6952, + "step": 20350 + }, + { + "epoch": 0.74, + "learning_rate": 1.2565642509246712e-07, + "loss": 0.6899, + "step": 20351 + }, + { + "epoch": 0.74, + "learning_rate": 1.2565276302779505e-07, + "loss": 0.6968, + "step": 20352 + }, + { + "epoch": 0.74, + "learning_rate": 1.25649100963123e-07, + "loss": 0.6926, + "step": 20353 + }, + { + "epoch": 0.74, + "learning_rate": 1.2564543889845093e-07, + "loss": 0.6777, + "step": 20354 + }, + { + "epoch": 0.74, + "learning_rate": 1.2564177683377888e-07, + "loss": 0.692, + "step": 20355 + }, + { + "epoch": 0.74, + "learning_rate": 1.256381147691068e-07, + "loss": 0.6785, + "step": 20356 + }, + { + "epoch": 0.74, + "learning_rate": 1.2563445270443476e-07, + "loss": 0.6942, + "step": 20357 + }, + { + "epoch": 0.74, + "learning_rate": 1.256307906397627e-07, + "loss": 0.6984, + "step": 20358 + }, + { + "epoch": 0.74, + "learning_rate": 1.2562712857509064e-07, + "loss": 0.7093, + "step": 20359 + }, + { + "epoch": 0.74, + "learning_rate": 1.2562346651041857e-07, + "loss": 0.6876, + "step": 20360 + }, + { + "epoch": 0.74, + "learning_rate": 1.2561980444574652e-07, + "loss": 0.6956, + "step": 20361 + }, + { + "epoch": 0.74, + "learning_rate": 1.2561614238107445e-07, + "loss": 0.6934, + "step": 20362 + }, + { + "epoch": 0.75, + "learning_rate": 1.256124803164024e-07, + "loss": 0.696, + "step": 20363 + }, + { + "epoch": 0.75, + "learning_rate": 1.2560881825173033e-07, + "loss": 0.6967, + "step": 20364 + }, + { + "epoch": 0.75, + "learning_rate": 1.2560515618705826e-07, + "loss": 0.6921, + "step": 20365 + }, + { + "epoch": 0.75, + "learning_rate": 1.2560149412238618e-07, + "loss": 0.6993, + "step": 20366 + }, + { + "epoch": 0.75, + "learning_rate": 1.2559783205771414e-07, + "loss": 0.6963, + "step": 20367 + }, + { + "epoch": 0.75, + "learning_rate": 1.2559416999304206e-07, + "loss": 0.6885, + "step": 20368 + }, + { + "epoch": 0.75, + "learning_rate": 1.2559050792837002e-07, + "loss": 0.7009, + "step": 20369 + }, + { + "epoch": 0.75, + "learning_rate": 1.2558684586369794e-07, + "loss": 0.7014, + "step": 20370 + }, + { + "epoch": 0.75, + "learning_rate": 1.255831837990259e-07, + "loss": 0.6939, + "step": 20371 + }, + { + "epoch": 0.75, + "learning_rate": 1.2557952173435382e-07, + "loss": 0.6893, + "step": 20372 + }, + { + "epoch": 0.75, + "learning_rate": 1.2557585966968175e-07, + "loss": 0.7044, + "step": 20373 + }, + { + "epoch": 0.75, + "learning_rate": 1.255721976050097e-07, + "loss": 0.6951, + "step": 20374 + }, + { + "epoch": 0.75, + "learning_rate": 1.2556853554033763e-07, + "loss": 0.7043, + "step": 20375 + }, + { + "epoch": 0.75, + "learning_rate": 1.2556487347566558e-07, + "loss": 0.6917, + "step": 20376 + }, + { + "epoch": 0.75, + "learning_rate": 1.255612114109935e-07, + "loss": 0.6758, + "step": 20377 + }, + { + "epoch": 0.75, + "learning_rate": 1.2555754934632146e-07, + "loss": 0.6961, + "step": 20378 + }, + { + "epoch": 0.75, + "learning_rate": 1.255538872816494e-07, + "loss": 0.7006, + "step": 20379 + }, + { + "epoch": 0.75, + "learning_rate": 1.2555022521697734e-07, + "loss": 0.7014, + "step": 20380 + }, + { + "epoch": 0.75, + "learning_rate": 1.2554656315230527e-07, + "loss": 0.6983, + "step": 20381 + }, + { + "epoch": 0.75, + "learning_rate": 1.2554290108763322e-07, + "loss": 0.7039, + "step": 20382 + }, + { + "epoch": 0.75, + "learning_rate": 1.2553923902296112e-07, + "loss": 0.7012, + "step": 20383 + }, + { + "epoch": 0.75, + "learning_rate": 1.2553557695828907e-07, + "loss": 0.6977, + "step": 20384 + }, + { + "epoch": 0.75, + "learning_rate": 1.25531914893617e-07, + "loss": 0.6898, + "step": 20385 + }, + { + "epoch": 0.75, + "learning_rate": 1.2552825282894495e-07, + "loss": 0.6901, + "step": 20386 + }, + { + "epoch": 0.75, + "learning_rate": 1.2552459076427288e-07, + "loss": 0.6943, + "step": 20387 + }, + { + "epoch": 0.75, + "learning_rate": 1.2552092869960083e-07, + "loss": 0.6896, + "step": 20388 + }, + { + "epoch": 0.75, + "learning_rate": 1.2551726663492876e-07, + "loss": 0.6997, + "step": 20389 + }, + { + "epoch": 0.75, + "learning_rate": 1.2551360457025671e-07, + "loss": 0.6937, + "step": 20390 + }, + { + "epoch": 0.75, + "learning_rate": 1.2550994250558464e-07, + "loss": 0.6911, + "step": 20391 + }, + { + "epoch": 0.75, + "learning_rate": 1.255062804409126e-07, + "loss": 0.6939, + "step": 20392 + }, + { + "epoch": 0.75, + "learning_rate": 1.2550261837624052e-07, + "loss": 0.6876, + "step": 20393 + }, + { + "epoch": 0.75, + "learning_rate": 1.2549895631156845e-07, + "loss": 0.6847, + "step": 20394 + }, + { + "epoch": 0.75, + "learning_rate": 1.254952942468964e-07, + "loss": 0.69, + "step": 20395 + }, + { + "epoch": 0.75, + "learning_rate": 1.2549163218222433e-07, + "loss": 0.7059, + "step": 20396 + }, + { + "epoch": 0.75, + "learning_rate": 1.2548797011755228e-07, + "loss": 0.6992, + "step": 20397 + }, + { + "epoch": 0.75, + "learning_rate": 1.254843080528802e-07, + "loss": 0.6916, + "step": 20398 + }, + { + "epoch": 0.75, + "learning_rate": 1.2548064598820816e-07, + "loss": 0.6847, + "step": 20399 + }, + { + "epoch": 0.75, + "learning_rate": 1.2547698392353609e-07, + "loss": 0.6951, + "step": 20400 + }, + { + "epoch": 0.75, + "learning_rate": 1.2547332185886401e-07, + "loss": 0.6897, + "step": 20401 + }, + { + "epoch": 0.75, + "learning_rate": 1.2546965979419194e-07, + "loss": 0.688, + "step": 20402 + }, + { + "epoch": 0.75, + "learning_rate": 1.254659977295199e-07, + "loss": 0.6787, + "step": 20403 + }, + { + "epoch": 0.75, + "learning_rate": 1.2546233566484782e-07, + "loss": 0.6983, + "step": 20404 + }, + { + "epoch": 0.75, + "learning_rate": 1.2545867360017577e-07, + "loss": 0.7004, + "step": 20405 + }, + { + "epoch": 0.75, + "learning_rate": 1.254550115355037e-07, + "loss": 0.6971, + "step": 20406 + }, + { + "epoch": 0.75, + "learning_rate": 1.2545134947083165e-07, + "loss": 0.7094, + "step": 20407 + }, + { + "epoch": 0.75, + "learning_rate": 1.2544768740615958e-07, + "loss": 0.7052, + "step": 20408 + }, + { + "epoch": 0.75, + "learning_rate": 1.2544402534148753e-07, + "loss": 0.6837, + "step": 20409 + }, + { + "epoch": 0.75, + "learning_rate": 1.2544036327681546e-07, + "loss": 0.6885, + "step": 20410 + }, + { + "epoch": 0.75, + "learning_rate": 1.254367012121434e-07, + "loss": 0.7031, + "step": 20411 + }, + { + "epoch": 0.75, + "learning_rate": 1.2543303914747134e-07, + "loss": 0.693, + "step": 20412 + }, + { + "epoch": 0.75, + "learning_rate": 1.254293770827993e-07, + "loss": 0.694, + "step": 20413 + }, + { + "epoch": 0.75, + "learning_rate": 1.2542571501812722e-07, + "loss": 0.6931, + "step": 20414 + }, + { + "epoch": 0.75, + "learning_rate": 1.2542205295345517e-07, + "loss": 0.6969, + "step": 20415 + }, + { + "epoch": 0.75, + "learning_rate": 1.254183908887831e-07, + "loss": 0.6982, + "step": 20416 + }, + { + "epoch": 0.75, + "learning_rate": 1.2541472882411103e-07, + "loss": 0.6899, + "step": 20417 + }, + { + "epoch": 0.75, + "learning_rate": 1.2541106675943898e-07, + "loss": 0.6936, + "step": 20418 + }, + { + "epoch": 0.75, + "learning_rate": 1.254074046947669e-07, + "loss": 0.7005, + "step": 20419 + }, + { + "epoch": 0.75, + "learning_rate": 1.2540374263009483e-07, + "loss": 0.7071, + "step": 20420 + }, + { + "epoch": 0.75, + "learning_rate": 1.2540008056542279e-07, + "loss": 0.6827, + "step": 20421 + }, + { + "epoch": 0.75, + "learning_rate": 1.253964185007507e-07, + "loss": 0.6881, + "step": 20422 + }, + { + "epoch": 0.75, + "learning_rate": 1.2539275643607864e-07, + "loss": 0.6978, + "step": 20423 + }, + { + "epoch": 0.75, + "learning_rate": 1.253890943714066e-07, + "loss": 0.6825, + "step": 20424 + }, + { + "epoch": 0.75, + "learning_rate": 1.2538543230673452e-07, + "loss": 0.6921, + "step": 20425 + }, + { + "epoch": 0.75, + "learning_rate": 1.2538177024206247e-07, + "loss": 0.69, + "step": 20426 + }, + { + "epoch": 0.75, + "learning_rate": 1.253781081773904e-07, + "loss": 0.7, + "step": 20427 + }, + { + "epoch": 0.75, + "learning_rate": 1.2537444611271835e-07, + "loss": 0.6979, + "step": 20428 + }, + { + "epoch": 0.75, + "learning_rate": 1.2537078404804628e-07, + "loss": 0.69, + "step": 20429 + }, + { + "epoch": 0.75, + "learning_rate": 1.2536712198337423e-07, + "loss": 0.6967, + "step": 20430 + }, + { + "epoch": 0.75, + "learning_rate": 1.2536345991870216e-07, + "loss": 0.7023, + "step": 20431 + }, + { + "epoch": 0.75, + "learning_rate": 1.253597978540301e-07, + "loss": 0.6931, + "step": 20432 + }, + { + "epoch": 0.75, + "learning_rate": 1.2535613578935804e-07, + "loss": 0.7042, + "step": 20433 + }, + { + "epoch": 0.75, + "learning_rate": 1.25352473724686e-07, + "loss": 0.6861, + "step": 20434 + }, + { + "epoch": 0.75, + "learning_rate": 1.2534881166001392e-07, + "loss": 0.6991, + "step": 20435 + }, + { + "epoch": 0.75, + "learning_rate": 1.2534514959534184e-07, + "loss": 0.6993, + "step": 20436 + }, + { + "epoch": 0.75, + "learning_rate": 1.2534148753066977e-07, + "loss": 0.6775, + "step": 20437 + }, + { + "epoch": 0.75, + "learning_rate": 1.2533782546599772e-07, + "loss": 0.7004, + "step": 20438 + }, + { + "epoch": 0.75, + "learning_rate": 1.2533416340132565e-07, + "loss": 0.7104, + "step": 20439 + }, + { + "epoch": 0.75, + "learning_rate": 1.253305013366536e-07, + "loss": 0.6964, + "step": 20440 + }, + { + "epoch": 0.75, + "learning_rate": 1.2532683927198153e-07, + "loss": 0.6845, + "step": 20441 + }, + { + "epoch": 0.75, + "learning_rate": 1.2532317720730948e-07, + "loss": 0.6953, + "step": 20442 + }, + { + "epoch": 0.75, + "learning_rate": 1.253195151426374e-07, + "loss": 0.6733, + "step": 20443 + }, + { + "epoch": 0.75, + "learning_rate": 1.2531585307796534e-07, + "loss": 0.6905, + "step": 20444 + }, + { + "epoch": 0.75, + "learning_rate": 1.253121910132933e-07, + "loss": 0.7019, + "step": 20445 + }, + { + "epoch": 0.75, + "learning_rate": 1.2530852894862122e-07, + "loss": 0.694, + "step": 20446 + }, + { + "epoch": 0.75, + "learning_rate": 1.2530486688394917e-07, + "loss": 0.6874, + "step": 20447 + }, + { + "epoch": 0.75, + "learning_rate": 1.253012048192771e-07, + "loss": 0.6868, + "step": 20448 + }, + { + "epoch": 0.75, + "learning_rate": 1.2529754275460505e-07, + "loss": 0.7051, + "step": 20449 + }, + { + "epoch": 0.75, + "learning_rate": 1.2529388068993298e-07, + "loss": 0.6821, + "step": 20450 + }, + { + "epoch": 0.75, + "learning_rate": 1.2529021862526093e-07, + "loss": 0.7033, + "step": 20451 + }, + { + "epoch": 0.75, + "learning_rate": 1.2528655656058886e-07, + "loss": 0.6896, + "step": 20452 + }, + { + "epoch": 0.75, + "learning_rate": 1.252828944959168e-07, + "loss": 0.6888, + "step": 20453 + }, + { + "epoch": 0.75, + "learning_rate": 1.252792324312447e-07, + "loss": 0.6907, + "step": 20454 + }, + { + "epoch": 0.75, + "learning_rate": 1.2527557036657266e-07, + "loss": 0.7005, + "step": 20455 + }, + { + "epoch": 0.75, + "learning_rate": 1.252719083019006e-07, + "loss": 0.6829, + "step": 20456 + }, + { + "epoch": 0.75, + "learning_rate": 1.2526824623722854e-07, + "loss": 0.6796, + "step": 20457 + }, + { + "epoch": 0.75, + "learning_rate": 1.2526458417255647e-07, + "loss": 0.6889, + "step": 20458 + }, + { + "epoch": 0.75, + "learning_rate": 1.2526092210788442e-07, + "loss": 0.6634, + "step": 20459 + }, + { + "epoch": 0.75, + "learning_rate": 1.2525726004321235e-07, + "loss": 0.6937, + "step": 20460 + }, + { + "epoch": 0.75, + "learning_rate": 1.252535979785403e-07, + "loss": 0.6956, + "step": 20461 + }, + { + "epoch": 0.75, + "learning_rate": 1.2524993591386823e-07, + "loss": 0.697, + "step": 20462 + }, + { + "epoch": 0.75, + "learning_rate": 1.2524627384919618e-07, + "loss": 0.7014, + "step": 20463 + }, + { + "epoch": 0.75, + "learning_rate": 1.252426117845241e-07, + "loss": 0.7112, + "step": 20464 + }, + { + "epoch": 0.75, + "learning_rate": 1.2523894971985204e-07, + "loss": 0.707, + "step": 20465 + }, + { + "epoch": 0.75, + "learning_rate": 1.2523528765518e-07, + "loss": 0.6845, + "step": 20466 + }, + { + "epoch": 0.75, + "learning_rate": 1.2523162559050792e-07, + "loss": 0.6923, + "step": 20467 + }, + { + "epoch": 0.75, + "learning_rate": 1.2522796352583587e-07, + "loss": 0.6838, + "step": 20468 + }, + { + "epoch": 0.75, + "learning_rate": 1.252243014611638e-07, + "loss": 0.6952, + "step": 20469 + }, + { + "epoch": 0.75, + "learning_rate": 1.2522063939649175e-07, + "loss": 0.7088, + "step": 20470 + }, + { + "epoch": 0.75, + "learning_rate": 1.2521697733181968e-07, + "loss": 0.7021, + "step": 20471 + }, + { + "epoch": 0.75, + "learning_rate": 1.252133152671476e-07, + "loss": 0.6957, + "step": 20472 + }, + { + "epoch": 0.75, + "learning_rate": 1.2520965320247553e-07, + "loss": 0.6844, + "step": 20473 + }, + { + "epoch": 0.75, + "learning_rate": 1.2520599113780348e-07, + "loss": 0.7022, + "step": 20474 + }, + { + "epoch": 0.75, + "learning_rate": 1.252023290731314e-07, + "loss": 0.6989, + "step": 20475 + }, + { + "epoch": 0.75, + "learning_rate": 1.2519866700845936e-07, + "loss": 0.7197, + "step": 20476 + }, + { + "epoch": 0.75, + "learning_rate": 1.251950049437873e-07, + "loss": 0.6849, + "step": 20477 + }, + { + "epoch": 0.75, + "learning_rate": 1.2519134287911524e-07, + "loss": 0.685, + "step": 20478 + }, + { + "epoch": 0.75, + "learning_rate": 1.2518768081444317e-07, + "loss": 0.7006, + "step": 20479 + }, + { + "epoch": 0.75, + "learning_rate": 1.2518401874977112e-07, + "loss": 0.6796, + "step": 20480 + }, + { + "epoch": 0.75, + "learning_rate": 1.2518035668509905e-07, + "loss": 0.6893, + "step": 20481 + }, + { + "epoch": 0.75, + "learning_rate": 1.25176694620427e-07, + "loss": 0.6985, + "step": 20482 + }, + { + "epoch": 0.75, + "learning_rate": 1.2517303255575493e-07, + "loss": 0.6991, + "step": 20483 + }, + { + "epoch": 0.75, + "learning_rate": 1.2516937049108288e-07, + "loss": 0.6891, + "step": 20484 + }, + { + "epoch": 0.75, + "learning_rate": 1.251657084264108e-07, + "loss": 0.6974, + "step": 20485 + }, + { + "epoch": 0.75, + "learning_rate": 1.2516204636173876e-07, + "loss": 0.6982, + "step": 20486 + }, + { + "epoch": 0.75, + "learning_rate": 1.251583842970667e-07, + "loss": 0.6947, + "step": 20487 + }, + { + "epoch": 0.75, + "learning_rate": 1.2515472223239462e-07, + "loss": 0.6927, + "step": 20488 + }, + { + "epoch": 0.75, + "learning_rate": 1.2515106016772257e-07, + "loss": 0.6888, + "step": 20489 + }, + { + "epoch": 0.75, + "learning_rate": 1.251473981030505e-07, + "loss": 0.6953, + "step": 20490 + }, + { + "epoch": 0.75, + "learning_rate": 1.2514373603837842e-07, + "loss": 0.699, + "step": 20491 + }, + { + "epoch": 0.75, + "learning_rate": 1.2514007397370635e-07, + "loss": 0.6945, + "step": 20492 + }, + { + "epoch": 0.75, + "learning_rate": 1.251364119090343e-07, + "loss": 0.6943, + "step": 20493 + }, + { + "epoch": 0.75, + "learning_rate": 1.2513274984436223e-07, + "loss": 0.7046, + "step": 20494 + }, + { + "epoch": 0.75, + "learning_rate": 1.2512908777969018e-07, + "loss": 0.7016, + "step": 20495 + }, + { + "epoch": 0.75, + "learning_rate": 1.251254257150181e-07, + "loss": 0.7016, + "step": 20496 + }, + { + "epoch": 0.75, + "learning_rate": 1.2512176365034606e-07, + "loss": 0.7002, + "step": 20497 + }, + { + "epoch": 0.75, + "learning_rate": 1.25118101585674e-07, + "loss": 0.6951, + "step": 20498 + }, + { + "epoch": 0.75, + "learning_rate": 1.2511443952100194e-07, + "loss": 0.6887, + "step": 20499 + }, + { + "epoch": 0.75, + "learning_rate": 1.2511077745632987e-07, + "loss": 0.6989, + "step": 20500 + }, + { + "epoch": 0.75, + "learning_rate": 1.2510711539165782e-07, + "loss": 0.7041, + "step": 20501 + }, + { + "epoch": 0.75, + "learning_rate": 1.2510345332698575e-07, + "loss": 0.7013, + "step": 20502 + }, + { + "epoch": 0.75, + "learning_rate": 1.250997912623137e-07, + "loss": 0.6957, + "step": 20503 + }, + { + "epoch": 0.75, + "learning_rate": 1.2509612919764163e-07, + "loss": 0.684, + "step": 20504 + }, + { + "epoch": 0.75, + "learning_rate": 1.2509246713296958e-07, + "loss": 0.7015, + "step": 20505 + }, + { + "epoch": 0.75, + "learning_rate": 1.250888050682975e-07, + "loss": 0.6828, + "step": 20506 + }, + { + "epoch": 0.75, + "learning_rate": 1.2508514300362546e-07, + "loss": 0.6946, + "step": 20507 + }, + { + "epoch": 0.75, + "learning_rate": 1.2508148093895336e-07, + "loss": 0.6954, + "step": 20508 + }, + { + "epoch": 0.75, + "learning_rate": 1.2507781887428131e-07, + "loss": 0.6953, + "step": 20509 + }, + { + "epoch": 0.75, + "learning_rate": 1.2507415680960924e-07, + "loss": 0.691, + "step": 20510 + }, + { + "epoch": 0.75, + "learning_rate": 1.250704947449372e-07, + "loss": 0.6972, + "step": 20511 + }, + { + "epoch": 0.75, + "learning_rate": 1.2506683268026512e-07, + "loss": 0.6878, + "step": 20512 + }, + { + "epoch": 0.75, + "learning_rate": 1.2506317061559307e-07, + "loss": 0.6838, + "step": 20513 + }, + { + "epoch": 0.75, + "learning_rate": 1.25059508550921e-07, + "loss": 0.6986, + "step": 20514 + }, + { + "epoch": 0.75, + "learning_rate": 1.2505584648624893e-07, + "loss": 0.6786, + "step": 20515 + }, + { + "epoch": 0.75, + "learning_rate": 1.2505218442157688e-07, + "loss": 0.7026, + "step": 20516 + }, + { + "epoch": 0.75, + "learning_rate": 1.250485223569048e-07, + "loss": 0.6901, + "step": 20517 + }, + { + "epoch": 0.75, + "learning_rate": 1.2504486029223276e-07, + "loss": 0.6942, + "step": 20518 + }, + { + "epoch": 0.75, + "learning_rate": 1.2504119822756069e-07, + "loss": 0.6888, + "step": 20519 + }, + { + "epoch": 0.75, + "learning_rate": 1.2503753616288864e-07, + "loss": 0.6872, + "step": 20520 + }, + { + "epoch": 0.75, + "learning_rate": 1.2503387409821657e-07, + "loss": 0.692, + "step": 20521 + }, + { + "epoch": 0.75, + "learning_rate": 1.2503021203354452e-07, + "loss": 0.692, + "step": 20522 + }, + { + "epoch": 0.75, + "learning_rate": 1.2502654996887245e-07, + "loss": 0.6903, + "step": 20523 + }, + { + "epoch": 0.75, + "learning_rate": 1.250228879042004e-07, + "loss": 0.7026, + "step": 20524 + }, + { + "epoch": 0.75, + "learning_rate": 1.2501922583952833e-07, + "loss": 0.7013, + "step": 20525 + }, + { + "epoch": 0.75, + "learning_rate": 1.2501556377485625e-07, + "loss": 0.6807, + "step": 20526 + }, + { + "epoch": 0.75, + "learning_rate": 1.2501190171018418e-07, + "loss": 0.7005, + "step": 20527 + }, + { + "epoch": 0.75, + "learning_rate": 1.2500823964551213e-07, + "loss": 0.6755, + "step": 20528 + }, + { + "epoch": 0.75, + "learning_rate": 1.2500457758084006e-07, + "loss": 0.689, + "step": 20529 + }, + { + "epoch": 0.75, + "learning_rate": 1.25000915516168e-07, + "loss": 0.7041, + "step": 20530 + }, + { + "epoch": 0.75, + "learning_rate": 1.2499725345149594e-07, + "loss": 0.6916, + "step": 20531 + }, + { + "epoch": 0.75, + "learning_rate": 1.249935913868239e-07, + "loss": 0.6848, + "step": 20532 + }, + { + "epoch": 0.75, + "learning_rate": 1.2498992932215182e-07, + "loss": 0.6888, + "step": 20533 + }, + { + "epoch": 0.75, + "learning_rate": 1.2498626725747977e-07, + "loss": 0.6996, + "step": 20534 + }, + { + "epoch": 0.75, + "learning_rate": 1.249826051928077e-07, + "loss": 0.6873, + "step": 20535 + }, + { + "epoch": 0.75, + "learning_rate": 1.2497894312813563e-07, + "loss": 0.6903, + "step": 20536 + }, + { + "epoch": 0.75, + "learning_rate": 1.2497528106346358e-07, + "loss": 0.7094, + "step": 20537 + }, + { + "epoch": 0.75, + "learning_rate": 1.249716189987915e-07, + "loss": 0.6846, + "step": 20538 + }, + { + "epoch": 0.75, + "learning_rate": 1.2496795693411946e-07, + "loss": 0.7027, + "step": 20539 + }, + { + "epoch": 0.75, + "learning_rate": 1.2496429486944739e-07, + "loss": 0.6913, + "step": 20540 + }, + { + "epoch": 0.75, + "learning_rate": 1.2496063280477534e-07, + "loss": 0.6849, + "step": 20541 + }, + { + "epoch": 0.75, + "learning_rate": 1.2495697074010327e-07, + "loss": 0.7018, + "step": 20542 + }, + { + "epoch": 0.75, + "learning_rate": 1.2495330867543122e-07, + "loss": 0.6881, + "step": 20543 + }, + { + "epoch": 0.75, + "learning_rate": 1.2494964661075912e-07, + "loss": 0.6943, + "step": 20544 + }, + { + "epoch": 0.75, + "learning_rate": 1.2494598454608707e-07, + "loss": 0.6959, + "step": 20545 + }, + { + "epoch": 0.75, + "learning_rate": 1.24942322481415e-07, + "loss": 0.6884, + "step": 20546 + }, + { + "epoch": 0.75, + "learning_rate": 1.2493866041674295e-07, + "loss": 0.6999, + "step": 20547 + }, + { + "epoch": 0.75, + "learning_rate": 1.2493499835207088e-07, + "loss": 0.7001, + "step": 20548 + }, + { + "epoch": 0.75, + "learning_rate": 1.2493133628739883e-07, + "loss": 0.6885, + "step": 20549 + }, + { + "epoch": 0.75, + "learning_rate": 1.2492767422272676e-07, + "loss": 0.6899, + "step": 20550 + }, + { + "epoch": 0.75, + "learning_rate": 1.249240121580547e-07, + "loss": 0.6947, + "step": 20551 + }, + { + "epoch": 0.75, + "learning_rate": 1.2492035009338264e-07, + "loss": 0.687, + "step": 20552 + }, + { + "epoch": 0.75, + "learning_rate": 1.249166880287106e-07, + "loss": 0.6955, + "step": 20553 + }, + { + "epoch": 0.75, + "learning_rate": 1.2491302596403852e-07, + "loss": 0.6876, + "step": 20554 + }, + { + "epoch": 0.75, + "learning_rate": 1.2490936389936647e-07, + "loss": 0.6947, + "step": 20555 + }, + { + "epoch": 0.75, + "learning_rate": 1.249057018346944e-07, + "loss": 0.6879, + "step": 20556 + }, + { + "epoch": 0.75, + "learning_rate": 1.2490203977002235e-07, + "loss": 0.6844, + "step": 20557 + }, + { + "epoch": 0.75, + "learning_rate": 1.2489837770535028e-07, + "loss": 0.6903, + "step": 20558 + }, + { + "epoch": 0.75, + "learning_rate": 1.248947156406782e-07, + "loss": 0.7024, + "step": 20559 + }, + { + "epoch": 0.75, + "learning_rate": 1.2489105357600616e-07, + "loss": 0.6777, + "step": 20560 + }, + { + "epoch": 0.75, + "learning_rate": 1.2488739151133408e-07, + "loss": 0.693, + "step": 20561 + }, + { + "epoch": 0.75, + "learning_rate": 1.24883729446662e-07, + "loss": 0.6919, + "step": 20562 + }, + { + "epoch": 0.75, + "learning_rate": 1.2488006738198994e-07, + "loss": 0.6964, + "step": 20563 + }, + { + "epoch": 0.75, + "learning_rate": 1.248764053173179e-07, + "loss": 0.6843, + "step": 20564 + }, + { + "epoch": 0.75, + "learning_rate": 1.2487274325264582e-07, + "loss": 0.6902, + "step": 20565 + }, + { + "epoch": 0.75, + "learning_rate": 1.2486908118797377e-07, + "loss": 0.6867, + "step": 20566 + }, + { + "epoch": 0.75, + "learning_rate": 1.248654191233017e-07, + "loss": 0.6991, + "step": 20567 + }, + { + "epoch": 0.75, + "learning_rate": 1.2486175705862965e-07, + "loss": 0.6936, + "step": 20568 + }, + { + "epoch": 0.75, + "learning_rate": 1.2485809499395758e-07, + "loss": 0.6966, + "step": 20569 + }, + { + "epoch": 0.75, + "learning_rate": 1.2485443292928553e-07, + "loss": 0.691, + "step": 20570 + }, + { + "epoch": 0.75, + "learning_rate": 1.2485077086461346e-07, + "loss": 0.705, + "step": 20571 + }, + { + "epoch": 0.75, + "learning_rate": 1.248471087999414e-07, + "loss": 0.6867, + "step": 20572 + }, + { + "epoch": 0.75, + "learning_rate": 1.2484344673526934e-07, + "loss": 0.6961, + "step": 20573 + }, + { + "epoch": 0.75, + "learning_rate": 1.248397846705973e-07, + "loss": 0.7051, + "step": 20574 + }, + { + "epoch": 0.75, + "learning_rate": 1.2483612260592522e-07, + "loss": 0.6812, + "step": 20575 + }, + { + "epoch": 0.75, + "learning_rate": 1.2483246054125317e-07, + "loss": 0.7046, + "step": 20576 + }, + { + "epoch": 0.75, + "learning_rate": 1.248287984765811e-07, + "loss": 0.6935, + "step": 20577 + }, + { + "epoch": 0.75, + "learning_rate": 1.2482513641190905e-07, + "loss": 0.6948, + "step": 20578 + }, + { + "epoch": 0.75, + "learning_rate": 1.2482147434723698e-07, + "loss": 0.7064, + "step": 20579 + }, + { + "epoch": 0.75, + "learning_rate": 1.248178122825649e-07, + "loss": 0.7004, + "step": 20580 + }, + { + "epoch": 0.75, + "learning_rate": 1.2481415021789283e-07, + "loss": 0.6942, + "step": 20581 + }, + { + "epoch": 0.75, + "learning_rate": 1.2481048815322078e-07, + "loss": 0.7011, + "step": 20582 + }, + { + "epoch": 0.75, + "learning_rate": 1.248068260885487e-07, + "loss": 0.6986, + "step": 20583 + }, + { + "epoch": 0.75, + "learning_rate": 1.2480316402387666e-07, + "loss": 0.685, + "step": 20584 + }, + { + "epoch": 0.75, + "learning_rate": 1.247995019592046e-07, + "loss": 0.6905, + "step": 20585 + }, + { + "epoch": 0.75, + "learning_rate": 1.2479583989453252e-07, + "loss": 0.6959, + "step": 20586 + }, + { + "epoch": 0.75, + "learning_rate": 1.2479217782986047e-07, + "loss": 0.6873, + "step": 20587 + }, + { + "epoch": 0.75, + "learning_rate": 1.247885157651884e-07, + "loss": 0.6999, + "step": 20588 + }, + { + "epoch": 0.75, + "learning_rate": 1.2478485370051635e-07, + "loss": 0.6903, + "step": 20589 + }, + { + "epoch": 0.75, + "learning_rate": 1.2478119163584428e-07, + "loss": 0.7056, + "step": 20590 + }, + { + "epoch": 0.75, + "learning_rate": 1.2477752957117223e-07, + "loss": 0.7001, + "step": 20591 + }, + { + "epoch": 0.75, + "learning_rate": 1.2477386750650016e-07, + "loss": 0.6943, + "step": 20592 + }, + { + "epoch": 0.75, + "learning_rate": 1.247702054418281e-07, + "loss": 0.7023, + "step": 20593 + }, + { + "epoch": 0.75, + "learning_rate": 1.2476654337715604e-07, + "loss": 0.6989, + "step": 20594 + }, + { + "epoch": 0.75, + "learning_rate": 1.24762881312484e-07, + "loss": 0.6832, + "step": 20595 + }, + { + "epoch": 0.75, + "learning_rate": 1.2475921924781192e-07, + "loss": 0.6937, + "step": 20596 + }, + { + "epoch": 0.75, + "learning_rate": 1.2475555718313987e-07, + "loss": 0.6782, + "step": 20597 + }, + { + "epoch": 0.75, + "learning_rate": 1.2475189511846777e-07, + "loss": 0.6996, + "step": 20598 + }, + { + "epoch": 0.75, + "learning_rate": 1.2474823305379572e-07, + "loss": 0.7102, + "step": 20599 + }, + { + "epoch": 0.75, + "learning_rate": 1.2474457098912365e-07, + "loss": 0.6909, + "step": 20600 + }, + { + "epoch": 0.75, + "learning_rate": 1.247409089244516e-07, + "loss": 0.7034, + "step": 20601 + }, + { + "epoch": 0.75, + "learning_rate": 1.2473724685977953e-07, + "loss": 0.6981, + "step": 20602 + }, + { + "epoch": 0.75, + "learning_rate": 1.2473358479510748e-07, + "loss": 0.6951, + "step": 20603 + }, + { + "epoch": 0.75, + "learning_rate": 1.247299227304354e-07, + "loss": 0.7021, + "step": 20604 + }, + { + "epoch": 0.75, + "learning_rate": 1.2472626066576336e-07, + "loss": 0.6942, + "step": 20605 + }, + { + "epoch": 0.75, + "learning_rate": 1.247225986010913e-07, + "loss": 0.6946, + "step": 20606 + }, + { + "epoch": 0.75, + "learning_rate": 1.2471893653641922e-07, + "loss": 0.697, + "step": 20607 + }, + { + "epoch": 0.75, + "learning_rate": 1.2471527447174717e-07, + "loss": 0.6852, + "step": 20608 + }, + { + "epoch": 0.75, + "learning_rate": 1.247116124070751e-07, + "loss": 0.7028, + "step": 20609 + }, + { + "epoch": 0.75, + "learning_rate": 1.2470795034240305e-07, + "loss": 0.7098, + "step": 20610 + }, + { + "epoch": 0.75, + "learning_rate": 1.2470428827773097e-07, + "loss": 0.6996, + "step": 20611 + }, + { + "epoch": 0.75, + "learning_rate": 1.2470062621305893e-07, + "loss": 0.6843, + "step": 20612 + }, + { + "epoch": 0.75, + "learning_rate": 1.2469696414838685e-07, + "loss": 0.6948, + "step": 20613 + }, + { + "epoch": 0.75, + "learning_rate": 1.246933020837148e-07, + "loss": 0.6964, + "step": 20614 + }, + { + "epoch": 0.75, + "learning_rate": 1.2468964001904273e-07, + "loss": 0.705, + "step": 20615 + }, + { + "epoch": 0.75, + "learning_rate": 1.2468597795437066e-07, + "loss": 0.6951, + "step": 20616 + }, + { + "epoch": 0.75, + "learning_rate": 1.246823158896986e-07, + "loss": 0.692, + "step": 20617 + }, + { + "epoch": 0.75, + "learning_rate": 1.2467865382502654e-07, + "loss": 0.6938, + "step": 20618 + }, + { + "epoch": 0.75, + "learning_rate": 1.2467499176035447e-07, + "loss": 0.6915, + "step": 20619 + }, + { + "epoch": 0.75, + "learning_rate": 1.2467132969568242e-07, + "loss": 0.6912, + "step": 20620 + }, + { + "epoch": 0.75, + "learning_rate": 1.2466766763101035e-07, + "loss": 0.6927, + "step": 20621 + }, + { + "epoch": 0.75, + "learning_rate": 1.246640055663383e-07, + "loss": 0.6936, + "step": 20622 + }, + { + "epoch": 0.75, + "learning_rate": 1.2466034350166623e-07, + "loss": 0.694, + "step": 20623 + }, + { + "epoch": 0.75, + "learning_rate": 1.2465668143699418e-07, + "loss": 0.6961, + "step": 20624 + }, + { + "epoch": 0.75, + "learning_rate": 1.246530193723221e-07, + "loss": 0.6959, + "step": 20625 + }, + { + "epoch": 0.75, + "learning_rate": 1.2464935730765006e-07, + "loss": 0.7029, + "step": 20626 + }, + { + "epoch": 0.75, + "learning_rate": 1.24645695242978e-07, + "loss": 0.6843, + "step": 20627 + }, + { + "epoch": 0.75, + "learning_rate": 1.2464203317830594e-07, + "loss": 0.6889, + "step": 20628 + }, + { + "epoch": 0.75, + "learning_rate": 1.2463837111363387e-07, + "loss": 0.7012, + "step": 20629 + }, + { + "epoch": 0.75, + "learning_rate": 1.246347090489618e-07, + "loss": 0.691, + "step": 20630 + }, + { + "epoch": 0.75, + "learning_rate": 1.2463104698428975e-07, + "loss": 0.702, + "step": 20631 + }, + { + "epoch": 0.75, + "learning_rate": 1.2462738491961767e-07, + "loss": 0.6869, + "step": 20632 + }, + { + "epoch": 0.75, + "learning_rate": 1.2462372285494563e-07, + "loss": 0.6998, + "step": 20633 + }, + { + "epoch": 0.75, + "learning_rate": 1.2462006079027353e-07, + "loss": 0.6919, + "step": 20634 + }, + { + "epoch": 0.75, + "learning_rate": 1.2461639872560148e-07, + "loss": 0.6958, + "step": 20635 + }, + { + "epoch": 0.76, + "learning_rate": 1.246127366609294e-07, + "loss": 0.6935, + "step": 20636 + }, + { + "epoch": 0.76, + "learning_rate": 1.2460907459625736e-07, + "loss": 0.6951, + "step": 20637 + }, + { + "epoch": 0.76, + "learning_rate": 1.2460541253158529e-07, + "loss": 0.6979, + "step": 20638 + }, + { + "epoch": 0.76, + "learning_rate": 1.2460175046691324e-07, + "loss": 0.7058, + "step": 20639 + }, + { + "epoch": 0.76, + "learning_rate": 1.2459808840224117e-07, + "loss": 0.6997, + "step": 20640 + }, + { + "epoch": 0.76, + "learning_rate": 1.2459442633756912e-07, + "loss": 0.6858, + "step": 20641 + }, + { + "epoch": 0.76, + "learning_rate": 1.2459076427289705e-07, + "loss": 0.6858, + "step": 20642 + }, + { + "epoch": 0.76, + "learning_rate": 1.24587102208225e-07, + "loss": 0.6957, + "step": 20643 + }, + { + "epoch": 0.76, + "learning_rate": 1.2458344014355293e-07, + "loss": 0.6939, + "step": 20644 + }, + { + "epoch": 0.76, + "learning_rate": 1.2457977807888088e-07, + "loss": 0.6953, + "step": 20645 + }, + { + "epoch": 0.76, + "learning_rate": 1.245761160142088e-07, + "loss": 0.6675, + "step": 20646 + }, + { + "epoch": 0.76, + "learning_rate": 1.2457245394953676e-07, + "loss": 0.6977, + "step": 20647 + }, + { + "epoch": 0.76, + "learning_rate": 1.2456879188486469e-07, + "loss": 0.696, + "step": 20648 + }, + { + "epoch": 0.76, + "learning_rate": 1.2456512982019264e-07, + "loss": 0.688, + "step": 20649 + }, + { + "epoch": 0.76, + "learning_rate": 1.2456146775552057e-07, + "loss": 0.6851, + "step": 20650 + }, + { + "epoch": 0.76, + "learning_rate": 1.245578056908485e-07, + "loss": 0.699, + "step": 20651 + }, + { + "epoch": 0.76, + "learning_rate": 1.2455414362617642e-07, + "loss": 0.6951, + "step": 20652 + }, + { + "epoch": 0.76, + "learning_rate": 1.2455048156150437e-07, + "loss": 0.6925, + "step": 20653 + }, + { + "epoch": 0.76, + "learning_rate": 1.245468194968323e-07, + "loss": 0.6948, + "step": 20654 + }, + { + "epoch": 0.76, + "learning_rate": 1.2454315743216025e-07, + "loss": 0.6831, + "step": 20655 + }, + { + "epoch": 0.76, + "learning_rate": 1.2453949536748818e-07, + "loss": 0.6955, + "step": 20656 + }, + { + "epoch": 0.76, + "learning_rate": 1.245358333028161e-07, + "loss": 0.6885, + "step": 20657 + }, + { + "epoch": 0.76, + "learning_rate": 1.2453217123814406e-07, + "loss": 0.6894, + "step": 20658 + }, + { + "epoch": 0.76, + "learning_rate": 1.2452850917347199e-07, + "loss": 0.7065, + "step": 20659 + }, + { + "epoch": 0.76, + "learning_rate": 1.2452484710879994e-07, + "loss": 0.685, + "step": 20660 + }, + { + "epoch": 0.76, + "learning_rate": 1.2452118504412787e-07, + "loss": 0.697, + "step": 20661 + }, + { + "epoch": 0.76, + "learning_rate": 1.2451752297945582e-07, + "loss": 0.686, + "step": 20662 + }, + { + "epoch": 0.76, + "learning_rate": 1.2451386091478375e-07, + "loss": 0.684, + "step": 20663 + }, + { + "epoch": 0.76, + "learning_rate": 1.245101988501117e-07, + "loss": 0.6894, + "step": 20664 + }, + { + "epoch": 0.76, + "learning_rate": 1.2450653678543963e-07, + "loss": 0.7007, + "step": 20665 + }, + { + "epoch": 0.76, + "learning_rate": 1.2450287472076758e-07, + "loss": 0.7028, + "step": 20666 + }, + { + "epoch": 0.76, + "learning_rate": 1.244992126560955e-07, + "loss": 0.688, + "step": 20667 + }, + { + "epoch": 0.76, + "learning_rate": 1.2449555059142346e-07, + "loss": 0.6992, + "step": 20668 + }, + { + "epoch": 0.76, + "learning_rate": 1.2449188852675138e-07, + "loss": 0.7026, + "step": 20669 + }, + { + "epoch": 0.76, + "learning_rate": 1.244882264620793e-07, + "loss": 0.6825, + "step": 20670 + }, + { + "epoch": 0.76, + "learning_rate": 1.2448456439740724e-07, + "loss": 0.6894, + "step": 20671 + }, + { + "epoch": 0.76, + "learning_rate": 1.244809023327352e-07, + "loss": 0.6947, + "step": 20672 + }, + { + "epoch": 0.76, + "learning_rate": 1.2447724026806312e-07, + "loss": 0.7064, + "step": 20673 + }, + { + "epoch": 0.76, + "learning_rate": 1.2447357820339107e-07, + "loss": 0.6921, + "step": 20674 + }, + { + "epoch": 0.76, + "learning_rate": 1.24469916138719e-07, + "loss": 0.6883, + "step": 20675 + }, + { + "epoch": 0.76, + "learning_rate": 1.2446625407404695e-07, + "loss": 0.6792, + "step": 20676 + }, + { + "epoch": 0.76, + "learning_rate": 1.2446259200937488e-07, + "loss": 0.7026, + "step": 20677 + }, + { + "epoch": 0.76, + "learning_rate": 1.244589299447028e-07, + "loss": 0.7106, + "step": 20678 + }, + { + "epoch": 0.76, + "learning_rate": 1.2445526788003076e-07, + "loss": 0.6982, + "step": 20679 + }, + { + "epoch": 0.76, + "learning_rate": 1.2445160581535868e-07, + "loss": 0.6957, + "step": 20680 + }, + { + "epoch": 0.76, + "learning_rate": 1.2444794375068664e-07, + "loss": 0.6907, + "step": 20681 + }, + { + "epoch": 0.76, + "learning_rate": 1.2444428168601456e-07, + "loss": 0.7006, + "step": 20682 + }, + { + "epoch": 0.76, + "learning_rate": 1.2444061962134252e-07, + "loss": 0.7068, + "step": 20683 + }, + { + "epoch": 0.76, + "learning_rate": 1.2443695755667044e-07, + "loss": 0.6982, + "step": 20684 + }, + { + "epoch": 0.76, + "learning_rate": 1.244332954919984e-07, + "loss": 0.6858, + "step": 20685 + }, + { + "epoch": 0.76, + "learning_rate": 1.2442963342732632e-07, + "loss": 0.6822, + "step": 20686 + }, + { + "epoch": 0.76, + "learning_rate": 1.2442597136265428e-07, + "loss": 0.694, + "step": 20687 + }, + { + "epoch": 0.76, + "learning_rate": 1.2442230929798218e-07, + "loss": 0.6779, + "step": 20688 + }, + { + "epoch": 0.76, + "learning_rate": 1.2441864723331013e-07, + "loss": 0.6961, + "step": 20689 + }, + { + "epoch": 0.76, + "learning_rate": 1.2441498516863806e-07, + "loss": 0.7012, + "step": 20690 + }, + { + "epoch": 0.76, + "learning_rate": 1.24411323103966e-07, + "loss": 0.6908, + "step": 20691 + }, + { + "epoch": 0.76, + "learning_rate": 1.2440766103929394e-07, + "loss": 0.7003, + "step": 20692 + }, + { + "epoch": 0.76, + "learning_rate": 1.244039989746219e-07, + "loss": 0.688, + "step": 20693 + }, + { + "epoch": 0.76, + "learning_rate": 1.2440033690994982e-07, + "loss": 0.6946, + "step": 20694 + }, + { + "epoch": 0.76, + "learning_rate": 1.2439667484527777e-07, + "loss": 0.698, + "step": 20695 + }, + { + "epoch": 0.76, + "learning_rate": 1.243930127806057e-07, + "loss": 0.6898, + "step": 20696 + }, + { + "epoch": 0.76, + "learning_rate": 1.2438935071593365e-07, + "loss": 0.6921, + "step": 20697 + }, + { + "epoch": 0.76, + "learning_rate": 1.2438568865126158e-07, + "loss": 0.6824, + "step": 20698 + }, + { + "epoch": 0.76, + "learning_rate": 1.2438202658658953e-07, + "loss": 0.6958, + "step": 20699 + }, + { + "epoch": 0.76, + "learning_rate": 1.2437836452191746e-07, + "loss": 0.6993, + "step": 20700 + }, + { + "epoch": 0.76, + "learning_rate": 1.2437470245724538e-07, + "loss": 0.6918, + "step": 20701 + }, + { + "epoch": 0.76, + "learning_rate": 1.2437104039257334e-07, + "loss": 0.6969, + "step": 20702 + }, + { + "epoch": 0.76, + "learning_rate": 1.2436737832790126e-07, + "loss": 0.6941, + "step": 20703 + }, + { + "epoch": 0.76, + "learning_rate": 1.2436371626322922e-07, + "loss": 0.6848, + "step": 20704 + }, + { + "epoch": 0.76, + "learning_rate": 1.2436005419855714e-07, + "loss": 0.6889, + "step": 20705 + }, + { + "epoch": 0.76, + "learning_rate": 1.2435639213388507e-07, + "loss": 0.6885, + "step": 20706 + }, + { + "epoch": 0.76, + "learning_rate": 1.24352730069213e-07, + "loss": 0.7075, + "step": 20707 + }, + { + "epoch": 0.76, + "learning_rate": 1.2434906800454095e-07, + "loss": 0.6966, + "step": 20708 + }, + { + "epoch": 0.76, + "learning_rate": 1.2434540593986888e-07, + "loss": 0.6865, + "step": 20709 + }, + { + "epoch": 0.76, + "learning_rate": 1.2434174387519683e-07, + "loss": 0.6855, + "step": 20710 + }, + { + "epoch": 0.76, + "learning_rate": 1.2433808181052476e-07, + "loss": 0.6895, + "step": 20711 + }, + { + "epoch": 0.76, + "learning_rate": 1.243344197458527e-07, + "loss": 0.6873, + "step": 20712 + }, + { + "epoch": 0.76, + "learning_rate": 1.2433075768118064e-07, + "loss": 0.7015, + "step": 20713 + }, + { + "epoch": 0.76, + "learning_rate": 1.243270956165086e-07, + "loss": 0.7001, + "step": 20714 + }, + { + "epoch": 0.76, + "learning_rate": 1.2432343355183652e-07, + "loss": 0.708, + "step": 20715 + }, + { + "epoch": 0.76, + "learning_rate": 1.2431977148716447e-07, + "loss": 0.6881, + "step": 20716 + }, + { + "epoch": 0.76, + "learning_rate": 1.243161094224924e-07, + "loss": 0.6993, + "step": 20717 + }, + { + "epoch": 0.76, + "learning_rate": 1.2431244735782035e-07, + "loss": 0.685, + "step": 20718 + }, + { + "epoch": 0.76, + "learning_rate": 1.2430878529314828e-07, + "loss": 0.7003, + "step": 20719 + }, + { + "epoch": 0.76, + "learning_rate": 1.2430512322847623e-07, + "loss": 0.6897, + "step": 20720 + }, + { + "epoch": 0.76, + "learning_rate": 1.2430146116380416e-07, + "loss": 0.7034, + "step": 20721 + }, + { + "epoch": 0.76, + "learning_rate": 1.2429779909913208e-07, + "loss": 0.6957, + "step": 20722 + }, + { + "epoch": 0.76, + "learning_rate": 1.2429413703446004e-07, + "loss": 0.691, + "step": 20723 + }, + { + "epoch": 0.76, + "learning_rate": 1.2429047496978796e-07, + "loss": 0.7095, + "step": 20724 + }, + { + "epoch": 0.76, + "learning_rate": 1.242868129051159e-07, + "loss": 0.6908, + "step": 20725 + }, + { + "epoch": 0.76, + "learning_rate": 1.2428315084044384e-07, + "loss": 0.6819, + "step": 20726 + }, + { + "epoch": 0.76, + "learning_rate": 1.2427948877577177e-07, + "loss": 0.7012, + "step": 20727 + }, + { + "epoch": 0.76, + "learning_rate": 1.242758267110997e-07, + "loss": 0.683, + "step": 20728 + }, + { + "epoch": 0.76, + "learning_rate": 1.2427216464642765e-07, + "loss": 0.6961, + "step": 20729 + }, + { + "epoch": 0.76, + "learning_rate": 1.2426850258175557e-07, + "loss": 0.71, + "step": 20730 + }, + { + "epoch": 0.76, + "learning_rate": 1.2426484051708353e-07, + "loss": 0.6902, + "step": 20731 + }, + { + "epoch": 0.76, + "learning_rate": 1.2426117845241145e-07, + "loss": 0.6813, + "step": 20732 + }, + { + "epoch": 0.76, + "learning_rate": 1.242575163877394e-07, + "loss": 0.7054, + "step": 20733 + }, + { + "epoch": 0.76, + "learning_rate": 1.2425385432306733e-07, + "loss": 0.6853, + "step": 20734 + }, + { + "epoch": 0.76, + "learning_rate": 1.242501922583953e-07, + "loss": 0.7029, + "step": 20735 + }, + { + "epoch": 0.76, + "learning_rate": 1.2424653019372321e-07, + "loss": 0.7099, + "step": 20736 + }, + { + "epoch": 0.76, + "learning_rate": 1.2424286812905117e-07, + "loss": 0.6982, + "step": 20737 + }, + { + "epoch": 0.76, + "learning_rate": 1.242392060643791e-07, + "loss": 0.688, + "step": 20738 + }, + { + "epoch": 0.76, + "learning_rate": 1.2423554399970705e-07, + "loss": 0.6948, + "step": 20739 + }, + { + "epoch": 0.76, + "learning_rate": 1.2423188193503497e-07, + "loss": 0.6958, + "step": 20740 + }, + { + "epoch": 0.76, + "learning_rate": 1.2422821987036293e-07, + "loss": 0.6945, + "step": 20741 + }, + { + "epoch": 0.76, + "learning_rate": 1.2422455780569083e-07, + "loss": 0.6955, + "step": 20742 + }, + { + "epoch": 0.76, + "learning_rate": 1.2422089574101878e-07, + "loss": 0.6897, + "step": 20743 + }, + { + "epoch": 0.76, + "learning_rate": 1.242172336763467e-07, + "loss": 0.6881, + "step": 20744 + }, + { + "epoch": 0.76, + "learning_rate": 1.2421357161167466e-07, + "loss": 0.6965, + "step": 20745 + }, + { + "epoch": 0.76, + "learning_rate": 1.242099095470026e-07, + "loss": 0.6879, + "step": 20746 + }, + { + "epoch": 0.76, + "learning_rate": 1.2420624748233054e-07, + "loss": 0.6924, + "step": 20747 + }, + { + "epoch": 0.76, + "learning_rate": 1.2420258541765847e-07, + "loss": 0.6938, + "step": 20748 + }, + { + "epoch": 0.76, + "learning_rate": 1.241989233529864e-07, + "loss": 0.6866, + "step": 20749 + }, + { + "epoch": 0.76, + "learning_rate": 1.2419526128831435e-07, + "loss": 0.692, + "step": 20750 + }, + { + "epoch": 0.76, + "learning_rate": 1.2419159922364227e-07, + "loss": 0.6991, + "step": 20751 + }, + { + "epoch": 0.76, + "learning_rate": 1.2418793715897023e-07, + "loss": 0.7097, + "step": 20752 + }, + { + "epoch": 0.76, + "learning_rate": 1.2418427509429815e-07, + "loss": 0.6947, + "step": 20753 + }, + { + "epoch": 0.76, + "learning_rate": 1.241806130296261e-07, + "loss": 0.7095, + "step": 20754 + }, + { + "epoch": 0.76, + "learning_rate": 1.2417695096495403e-07, + "loss": 0.6845, + "step": 20755 + }, + { + "epoch": 0.76, + "learning_rate": 1.2417328890028199e-07, + "loss": 0.6854, + "step": 20756 + }, + { + "epoch": 0.76, + "learning_rate": 1.2416962683560991e-07, + "loss": 0.701, + "step": 20757 + }, + { + "epoch": 0.76, + "learning_rate": 1.2416596477093787e-07, + "loss": 0.6776, + "step": 20758 + }, + { + "epoch": 0.76, + "learning_rate": 1.241623027062658e-07, + "loss": 0.6928, + "step": 20759 + }, + { + "epoch": 0.76, + "learning_rate": 1.2415864064159372e-07, + "loss": 0.7016, + "step": 20760 + }, + { + "epoch": 0.76, + "learning_rate": 1.2415497857692165e-07, + "loss": 0.682, + "step": 20761 + }, + { + "epoch": 0.76, + "learning_rate": 1.241513165122496e-07, + "loss": 0.6811, + "step": 20762 + }, + { + "epoch": 0.76, + "learning_rate": 1.2414765444757753e-07, + "loss": 0.6953, + "step": 20763 + }, + { + "epoch": 0.76, + "learning_rate": 1.2414399238290548e-07, + "loss": 0.6896, + "step": 20764 + }, + { + "epoch": 0.76, + "learning_rate": 1.241403303182334e-07, + "loss": 0.6924, + "step": 20765 + }, + { + "epoch": 0.76, + "learning_rate": 1.2413666825356136e-07, + "loss": 0.6953, + "step": 20766 + }, + { + "epoch": 0.76, + "learning_rate": 1.2413300618888929e-07, + "loss": 0.6944, + "step": 20767 + }, + { + "epoch": 0.76, + "learning_rate": 1.2412934412421724e-07, + "loss": 0.6949, + "step": 20768 + }, + { + "epoch": 0.76, + "learning_rate": 1.2412568205954517e-07, + "loss": 0.6968, + "step": 20769 + }, + { + "epoch": 0.76, + "learning_rate": 1.2412201999487312e-07, + "loss": 0.6995, + "step": 20770 + }, + { + "epoch": 0.76, + "learning_rate": 1.2411835793020105e-07, + "loss": 0.7164, + "step": 20771 + }, + { + "epoch": 0.76, + "learning_rate": 1.2411469586552897e-07, + "loss": 0.6949, + "step": 20772 + }, + { + "epoch": 0.76, + "learning_rate": 1.2411103380085693e-07, + "loss": 0.6933, + "step": 20773 + }, + { + "epoch": 0.76, + "learning_rate": 1.2410737173618485e-07, + "loss": 0.6932, + "step": 20774 + }, + { + "epoch": 0.76, + "learning_rate": 1.241037096715128e-07, + "loss": 0.6936, + "step": 20775 + }, + { + "epoch": 0.76, + "learning_rate": 1.2410004760684073e-07, + "loss": 0.692, + "step": 20776 + }, + { + "epoch": 0.76, + "learning_rate": 1.2409638554216869e-07, + "loss": 0.6977, + "step": 20777 + }, + { + "epoch": 0.76, + "learning_rate": 1.2409272347749659e-07, + "loss": 0.693, + "step": 20778 + }, + { + "epoch": 0.76, + "learning_rate": 1.2408906141282454e-07, + "loss": 0.6909, + "step": 20779 + }, + { + "epoch": 0.76, + "learning_rate": 1.2408539934815247e-07, + "loss": 0.7027, + "step": 20780 + }, + { + "epoch": 0.76, + "learning_rate": 1.2408173728348042e-07, + "loss": 0.6968, + "step": 20781 + }, + { + "epoch": 0.76, + "learning_rate": 1.2407807521880835e-07, + "loss": 0.6945, + "step": 20782 + }, + { + "epoch": 0.76, + "learning_rate": 1.240744131541363e-07, + "loss": 0.6944, + "step": 20783 + }, + { + "epoch": 0.76, + "learning_rate": 1.2407075108946423e-07, + "loss": 0.6975, + "step": 20784 + }, + { + "epoch": 0.76, + "learning_rate": 1.2406708902479218e-07, + "loss": 0.7126, + "step": 20785 + }, + { + "epoch": 0.76, + "learning_rate": 1.240634269601201e-07, + "loss": 0.6999, + "step": 20786 + }, + { + "epoch": 0.76, + "learning_rate": 1.2405976489544806e-07, + "loss": 0.6937, + "step": 20787 + }, + { + "epoch": 0.76, + "learning_rate": 1.2405610283077598e-07, + "loss": 0.6924, + "step": 20788 + }, + { + "epoch": 0.76, + "learning_rate": 1.2405244076610394e-07, + "loss": 0.694, + "step": 20789 + }, + { + "epoch": 0.76, + "learning_rate": 1.2404877870143186e-07, + "loss": 0.6966, + "step": 20790 + }, + { + "epoch": 0.76, + "learning_rate": 1.2404511663675982e-07, + "loss": 0.7012, + "step": 20791 + }, + { + "epoch": 0.76, + "learning_rate": 1.2404145457208774e-07, + "loss": 0.7035, + "step": 20792 + }, + { + "epoch": 0.76, + "learning_rate": 1.2403779250741567e-07, + "loss": 0.7012, + "step": 20793 + }, + { + "epoch": 0.76, + "learning_rate": 1.2403413044274362e-07, + "loss": 0.7031, + "step": 20794 + }, + { + "epoch": 0.76, + "learning_rate": 1.2403046837807155e-07, + "loss": 0.6932, + "step": 20795 + }, + { + "epoch": 0.76, + "learning_rate": 1.2402680631339948e-07, + "loss": 0.6926, + "step": 20796 + }, + { + "epoch": 0.76, + "learning_rate": 1.2402314424872743e-07, + "loss": 0.6932, + "step": 20797 + }, + { + "epoch": 0.76, + "learning_rate": 1.2401948218405536e-07, + "loss": 0.6855, + "step": 20798 + }, + { + "epoch": 0.76, + "learning_rate": 1.2401582011938328e-07, + "loss": 0.7007, + "step": 20799 + }, + { + "epoch": 0.76, + "learning_rate": 1.2401215805471124e-07, + "loss": 0.7051, + "step": 20800 + }, + { + "epoch": 0.76, + "learning_rate": 1.2400849599003916e-07, + "loss": 0.6944, + "step": 20801 + }, + { + "epoch": 0.76, + "learning_rate": 1.2400483392536712e-07, + "loss": 0.6858, + "step": 20802 + }, + { + "epoch": 0.76, + "learning_rate": 1.2400117186069504e-07, + "loss": 0.6986, + "step": 20803 + }, + { + "epoch": 0.76, + "learning_rate": 1.23997509796023e-07, + "loss": 0.7098, + "step": 20804 + }, + { + "epoch": 0.76, + "learning_rate": 1.2399384773135092e-07, + "loss": 0.6975, + "step": 20805 + }, + { + "epoch": 0.76, + "learning_rate": 1.2399018566667888e-07, + "loss": 0.6853, + "step": 20806 + }, + { + "epoch": 0.76, + "learning_rate": 1.239865236020068e-07, + "loss": 0.6948, + "step": 20807 + }, + { + "epoch": 0.76, + "learning_rate": 1.2398286153733476e-07, + "loss": 0.6861, + "step": 20808 + }, + { + "epoch": 0.76, + "learning_rate": 1.2397919947266268e-07, + "loss": 0.7025, + "step": 20809 + }, + { + "epoch": 0.76, + "learning_rate": 1.2397553740799064e-07, + "loss": 0.6941, + "step": 20810 + }, + { + "epoch": 0.76, + "learning_rate": 1.2397187534331856e-07, + "loss": 0.7039, + "step": 20811 + }, + { + "epoch": 0.76, + "learning_rate": 1.2396821327864652e-07, + "loss": 0.6786, + "step": 20812 + }, + { + "epoch": 0.76, + "learning_rate": 1.2396455121397444e-07, + "loss": 0.6991, + "step": 20813 + }, + { + "epoch": 0.76, + "learning_rate": 1.2396088914930237e-07, + "loss": 0.7038, + "step": 20814 + }, + { + "epoch": 0.76, + "learning_rate": 1.239572270846303e-07, + "loss": 0.6841, + "step": 20815 + }, + { + "epoch": 0.76, + "learning_rate": 1.2395356501995825e-07, + "loss": 0.6963, + "step": 20816 + }, + { + "epoch": 0.76, + "learning_rate": 1.2394990295528618e-07, + "loss": 0.6822, + "step": 20817 + }, + { + "epoch": 0.76, + "learning_rate": 1.2394624089061413e-07, + "loss": 0.7023, + "step": 20818 + }, + { + "epoch": 0.76, + "learning_rate": 1.2394257882594206e-07, + "loss": 0.6858, + "step": 20819 + }, + { + "epoch": 0.76, + "learning_rate": 1.2393891676126998e-07, + "loss": 0.6853, + "step": 20820 + }, + { + "epoch": 0.76, + "learning_rate": 1.2393525469659794e-07, + "loss": 0.6964, + "step": 20821 + }, + { + "epoch": 0.76, + "learning_rate": 1.2393159263192586e-07, + "loss": 0.6902, + "step": 20822 + }, + { + "epoch": 0.76, + "learning_rate": 1.2392793056725382e-07, + "loss": 0.6974, + "step": 20823 + }, + { + "epoch": 0.76, + "learning_rate": 1.2392426850258174e-07, + "loss": 0.7067, + "step": 20824 + }, + { + "epoch": 0.76, + "learning_rate": 1.239206064379097e-07, + "loss": 0.6928, + "step": 20825 + }, + { + "epoch": 0.76, + "learning_rate": 1.2391694437323762e-07, + "loss": 0.6879, + "step": 20826 + }, + { + "epoch": 0.76, + "learning_rate": 1.2391328230856558e-07, + "loss": 0.6896, + "step": 20827 + }, + { + "epoch": 0.76, + "learning_rate": 1.239096202438935e-07, + "loss": 0.6971, + "step": 20828 + }, + { + "epoch": 0.76, + "learning_rate": 1.2390595817922146e-07, + "loss": 0.6892, + "step": 20829 + }, + { + "epoch": 0.76, + "learning_rate": 1.2390229611454938e-07, + "loss": 0.7082, + "step": 20830 + }, + { + "epoch": 0.76, + "learning_rate": 1.238986340498773e-07, + "loss": 0.6938, + "step": 20831 + }, + { + "epoch": 0.76, + "learning_rate": 1.2389497198520524e-07, + "loss": 0.6909, + "step": 20832 + }, + { + "epoch": 0.76, + "learning_rate": 1.238913099205332e-07, + "loss": 0.6983, + "step": 20833 + }, + { + "epoch": 0.76, + "learning_rate": 1.2388764785586112e-07, + "loss": 0.6936, + "step": 20834 + }, + { + "epoch": 0.76, + "learning_rate": 1.2388398579118907e-07, + "loss": 0.6805, + "step": 20835 + }, + { + "epoch": 0.76, + "learning_rate": 1.23880323726517e-07, + "loss": 0.6951, + "step": 20836 + }, + { + "epoch": 0.76, + "learning_rate": 1.2387666166184495e-07, + "loss": 0.68, + "step": 20837 + }, + { + "epoch": 0.76, + "learning_rate": 1.2387299959717288e-07, + "loss": 0.6934, + "step": 20838 + }, + { + "epoch": 0.76, + "learning_rate": 1.2386933753250083e-07, + "loss": 0.6994, + "step": 20839 + }, + { + "epoch": 0.76, + "learning_rate": 1.2386567546782876e-07, + "loss": 0.6967, + "step": 20840 + }, + { + "epoch": 0.76, + "learning_rate": 1.238620134031567e-07, + "loss": 0.7066, + "step": 20841 + }, + { + "epoch": 0.76, + "learning_rate": 1.2385835133848464e-07, + "loss": 0.6842, + "step": 20842 + }, + { + "epoch": 0.76, + "learning_rate": 1.2385468927381256e-07, + "loss": 0.6899, + "step": 20843 + }, + { + "epoch": 0.76, + "learning_rate": 1.2385102720914051e-07, + "loss": 0.7023, + "step": 20844 + }, + { + "epoch": 0.76, + "learning_rate": 1.2384736514446844e-07, + "loss": 0.7, + "step": 20845 + }, + { + "epoch": 0.76, + "learning_rate": 1.238437030797964e-07, + "loss": 0.7042, + "step": 20846 + }, + { + "epoch": 0.76, + "learning_rate": 1.2384004101512432e-07, + "loss": 0.6874, + "step": 20847 + }, + { + "epoch": 0.76, + "learning_rate": 1.2383637895045227e-07, + "loss": 0.6855, + "step": 20848 + }, + { + "epoch": 0.76, + "learning_rate": 1.2383271688578017e-07, + "loss": 0.6924, + "step": 20849 + }, + { + "epoch": 0.76, + "learning_rate": 1.2382905482110813e-07, + "loss": 0.6968, + "step": 20850 + }, + { + "epoch": 0.76, + "learning_rate": 1.2382539275643605e-07, + "loss": 0.697, + "step": 20851 + }, + { + "epoch": 0.76, + "learning_rate": 1.23821730691764e-07, + "loss": 0.6924, + "step": 20852 + }, + { + "epoch": 0.76, + "learning_rate": 1.2381806862709193e-07, + "loss": 0.6979, + "step": 20853 + }, + { + "epoch": 0.76, + "learning_rate": 1.238144065624199e-07, + "loss": 0.6863, + "step": 20854 + }, + { + "epoch": 0.76, + "learning_rate": 1.2381074449774781e-07, + "loss": 0.6936, + "step": 20855 + }, + { + "epoch": 0.76, + "learning_rate": 1.2380708243307577e-07, + "loss": 0.7069, + "step": 20856 + }, + { + "epoch": 0.76, + "learning_rate": 1.238034203684037e-07, + "loss": 0.6995, + "step": 20857 + }, + { + "epoch": 0.76, + "learning_rate": 1.2379975830373165e-07, + "loss": 0.6918, + "step": 20858 + }, + { + "epoch": 0.76, + "learning_rate": 1.2379609623905957e-07, + "loss": 0.6922, + "step": 20859 + }, + { + "epoch": 0.76, + "learning_rate": 1.2379243417438753e-07, + "loss": 0.704, + "step": 20860 + }, + { + "epoch": 0.76, + "learning_rate": 1.2378877210971545e-07, + "loss": 0.6963, + "step": 20861 + }, + { + "epoch": 0.76, + "learning_rate": 1.237851100450434e-07, + "loss": 0.705, + "step": 20862 + }, + { + "epoch": 0.76, + "learning_rate": 1.2378144798037133e-07, + "loss": 0.6957, + "step": 20863 + }, + { + "epoch": 0.76, + "learning_rate": 1.2377778591569926e-07, + "loss": 0.6954, + "step": 20864 + }, + { + "epoch": 0.76, + "learning_rate": 1.2377412385102721e-07, + "loss": 0.692, + "step": 20865 + }, + { + "epoch": 0.76, + "learning_rate": 1.2377046178635514e-07, + "loss": 0.6887, + "step": 20866 + }, + { + "epoch": 0.76, + "learning_rate": 1.2376679972168307e-07, + "loss": 0.6832, + "step": 20867 + }, + { + "epoch": 0.76, + "learning_rate": 1.2376313765701102e-07, + "loss": 0.6997, + "step": 20868 + }, + { + "epoch": 0.76, + "learning_rate": 1.2375947559233895e-07, + "loss": 0.6876, + "step": 20869 + }, + { + "epoch": 0.76, + "learning_rate": 1.2375581352766687e-07, + "loss": 0.7039, + "step": 20870 + }, + { + "epoch": 0.76, + "learning_rate": 1.2375215146299483e-07, + "loss": 0.6798, + "step": 20871 + }, + { + "epoch": 0.76, + "learning_rate": 1.2374848939832275e-07, + "loss": 0.6937, + "step": 20872 + }, + { + "epoch": 0.76, + "learning_rate": 1.237448273336507e-07, + "loss": 0.6976, + "step": 20873 + }, + { + "epoch": 0.76, + "learning_rate": 1.2374116526897863e-07, + "loss": 0.6929, + "step": 20874 + }, + { + "epoch": 0.76, + "learning_rate": 1.2373750320430659e-07, + "loss": 0.6753, + "step": 20875 + }, + { + "epoch": 0.76, + "learning_rate": 1.2373384113963451e-07, + "loss": 0.699, + "step": 20876 + }, + { + "epoch": 0.76, + "learning_rate": 1.2373017907496247e-07, + "loss": 0.6859, + "step": 20877 + }, + { + "epoch": 0.76, + "learning_rate": 1.237265170102904e-07, + "loss": 0.6884, + "step": 20878 + }, + { + "epoch": 0.76, + "learning_rate": 1.2372285494561835e-07, + "loss": 0.6899, + "step": 20879 + }, + { + "epoch": 0.76, + "learning_rate": 1.2371919288094627e-07, + "loss": 0.6882, + "step": 20880 + }, + { + "epoch": 0.76, + "learning_rate": 1.2371553081627423e-07, + "loss": 0.6915, + "step": 20881 + }, + { + "epoch": 0.76, + "learning_rate": 1.2371186875160215e-07, + "loss": 0.7042, + "step": 20882 + }, + { + "epoch": 0.76, + "learning_rate": 1.237082066869301e-07, + "loss": 0.7005, + "step": 20883 + }, + { + "epoch": 0.76, + "learning_rate": 1.2370454462225803e-07, + "loss": 0.7033, + "step": 20884 + }, + { + "epoch": 0.76, + "learning_rate": 1.2370088255758596e-07, + "loss": 0.6994, + "step": 20885 + }, + { + "epoch": 0.76, + "learning_rate": 1.2369722049291389e-07, + "loss": 0.6845, + "step": 20886 + }, + { + "epoch": 0.76, + "learning_rate": 1.2369355842824184e-07, + "loss": 0.6988, + "step": 20887 + }, + { + "epoch": 0.76, + "learning_rate": 1.2368989636356977e-07, + "loss": 0.7046, + "step": 20888 + }, + { + "epoch": 0.76, + "learning_rate": 1.2368623429889772e-07, + "loss": 0.6891, + "step": 20889 + }, + { + "epoch": 0.76, + "learning_rate": 1.2368257223422565e-07, + "loss": 0.6995, + "step": 20890 + }, + { + "epoch": 0.76, + "learning_rate": 1.2367891016955357e-07, + "loss": 0.6931, + "step": 20891 + }, + { + "epoch": 0.76, + "learning_rate": 1.2367524810488153e-07, + "loss": 0.6982, + "step": 20892 + }, + { + "epoch": 0.76, + "learning_rate": 1.2367158604020945e-07, + "loss": 0.6829, + "step": 20893 + }, + { + "epoch": 0.76, + "learning_rate": 1.236679239755374e-07, + "loss": 0.6713, + "step": 20894 + }, + { + "epoch": 0.76, + "learning_rate": 1.2366426191086533e-07, + "loss": 0.6979, + "step": 20895 + }, + { + "epoch": 0.76, + "learning_rate": 1.2366059984619329e-07, + "loss": 0.6987, + "step": 20896 + }, + { + "epoch": 0.76, + "learning_rate": 1.236569377815212e-07, + "loss": 0.6947, + "step": 20897 + }, + { + "epoch": 0.76, + "learning_rate": 1.2365327571684917e-07, + "loss": 0.6844, + "step": 20898 + }, + { + "epoch": 0.76, + "learning_rate": 1.236496136521771e-07, + "loss": 0.6943, + "step": 20899 + }, + { + "epoch": 0.76, + "learning_rate": 1.2364595158750505e-07, + "loss": 0.7037, + "step": 20900 + }, + { + "epoch": 0.76, + "learning_rate": 1.2364228952283297e-07, + "loss": 0.6971, + "step": 20901 + }, + { + "epoch": 0.76, + "learning_rate": 1.2363862745816092e-07, + "loss": 0.6756, + "step": 20902 + }, + { + "epoch": 0.76, + "learning_rate": 1.2363496539348883e-07, + "loss": 0.6925, + "step": 20903 + }, + { + "epoch": 0.76, + "learning_rate": 1.2363130332881678e-07, + "loss": 0.7089, + "step": 20904 + }, + { + "epoch": 0.76, + "learning_rate": 1.236276412641447e-07, + "loss": 0.7028, + "step": 20905 + }, + { + "epoch": 0.76, + "learning_rate": 1.2362397919947266e-07, + "loss": 0.6965, + "step": 20906 + }, + { + "epoch": 0.76, + "learning_rate": 1.2362031713480058e-07, + "loss": 0.6836, + "step": 20907 + }, + { + "epoch": 0.76, + "learning_rate": 1.2361665507012854e-07, + "loss": 0.6941, + "step": 20908 + }, + { + "epoch": 0.77, + "learning_rate": 1.2361299300545646e-07, + "loss": 0.6931, + "step": 20909 + }, + { + "epoch": 0.77, + "learning_rate": 1.2360933094078442e-07, + "loss": 0.6911, + "step": 20910 + }, + { + "epoch": 0.77, + "learning_rate": 1.2360566887611234e-07, + "loss": 0.7085, + "step": 20911 + }, + { + "epoch": 0.77, + "learning_rate": 1.236020068114403e-07, + "loss": 0.693, + "step": 20912 + }, + { + "epoch": 0.77, + "learning_rate": 1.2359834474676822e-07, + "loss": 0.6954, + "step": 20913 + }, + { + "epoch": 0.77, + "learning_rate": 1.2359468268209615e-07, + "loss": 0.691, + "step": 20914 + }, + { + "epoch": 0.77, + "learning_rate": 1.235910206174241e-07, + "loss": 0.6953, + "step": 20915 + }, + { + "epoch": 0.77, + "learning_rate": 1.2358735855275203e-07, + "loss": 0.6989, + "step": 20916 + }, + { + "epoch": 0.77, + "learning_rate": 1.2358369648807998e-07, + "loss": 0.7033, + "step": 20917 + }, + { + "epoch": 0.77, + "learning_rate": 1.235800344234079e-07, + "loss": 0.69, + "step": 20918 + }, + { + "epoch": 0.77, + "learning_rate": 1.2357637235873586e-07, + "loss": 0.6906, + "step": 20919 + }, + { + "epoch": 0.77, + "learning_rate": 1.235727102940638e-07, + "loss": 0.6904, + "step": 20920 + }, + { + "epoch": 0.77, + "learning_rate": 1.2356904822939172e-07, + "loss": 0.697, + "step": 20921 + }, + { + "epoch": 0.77, + "learning_rate": 1.2356538616471964e-07, + "loss": 0.6963, + "step": 20922 + }, + { + "epoch": 0.77, + "learning_rate": 1.235617241000476e-07, + "loss": 0.701, + "step": 20923 + }, + { + "epoch": 0.77, + "learning_rate": 1.2355806203537552e-07, + "loss": 0.6953, + "step": 20924 + }, + { + "epoch": 0.77, + "learning_rate": 1.2355439997070348e-07, + "loss": 0.6999, + "step": 20925 + }, + { + "epoch": 0.77, + "learning_rate": 1.235507379060314e-07, + "loss": 0.6962, + "step": 20926 + }, + { + "epoch": 0.77, + "learning_rate": 1.2354707584135936e-07, + "loss": 0.6866, + "step": 20927 + }, + { + "epoch": 0.77, + "learning_rate": 1.2354341377668728e-07, + "loss": 0.6896, + "step": 20928 + }, + { + "epoch": 0.77, + "learning_rate": 1.2353975171201524e-07, + "loss": 0.6939, + "step": 20929 + }, + { + "epoch": 0.77, + "learning_rate": 1.2353608964734316e-07, + "loss": 0.683, + "step": 20930 + }, + { + "epoch": 0.77, + "learning_rate": 1.2353242758267112e-07, + "loss": 0.6877, + "step": 20931 + }, + { + "epoch": 0.77, + "learning_rate": 1.2352876551799904e-07, + "loss": 0.7036, + "step": 20932 + }, + { + "epoch": 0.77, + "learning_rate": 1.23525103453327e-07, + "loss": 0.6763, + "step": 20933 + }, + { + "epoch": 0.77, + "learning_rate": 1.2352144138865492e-07, + "loss": 0.69, + "step": 20934 + }, + { + "epoch": 0.77, + "learning_rate": 1.2351777932398285e-07, + "loss": 0.7079, + "step": 20935 + }, + { + "epoch": 0.77, + "learning_rate": 1.235141172593108e-07, + "loss": 0.6941, + "step": 20936 + }, + { + "epoch": 0.77, + "learning_rate": 1.2351045519463873e-07, + "loss": 0.7092, + "step": 20937 + }, + { + "epoch": 0.77, + "learning_rate": 1.2350679312996668e-07, + "loss": 0.6868, + "step": 20938 + }, + { + "epoch": 0.77, + "learning_rate": 1.235031310652946e-07, + "loss": 0.7034, + "step": 20939 + }, + { + "epoch": 0.77, + "learning_rate": 1.2349946900062254e-07, + "loss": 0.6932, + "step": 20940 + }, + { + "epoch": 0.77, + "learning_rate": 1.2349580693595046e-07, + "loss": 0.6925, + "step": 20941 + }, + { + "epoch": 0.77, + "learning_rate": 1.2349214487127842e-07, + "loss": 0.7146, + "step": 20942 + }, + { + "epoch": 0.77, + "learning_rate": 1.2348848280660634e-07, + "loss": 0.6864, + "step": 20943 + }, + { + "epoch": 0.77, + "learning_rate": 1.234848207419343e-07, + "loss": 0.7012, + "step": 20944 + }, + { + "epoch": 0.77, + "learning_rate": 1.2348115867726222e-07, + "loss": 0.6873, + "step": 20945 + }, + { + "epoch": 0.77, + "learning_rate": 1.2347749661259018e-07, + "loss": 0.6825, + "step": 20946 + }, + { + "epoch": 0.77, + "learning_rate": 1.234738345479181e-07, + "loss": 0.6977, + "step": 20947 + }, + { + "epoch": 0.77, + "learning_rate": 1.2347017248324606e-07, + "loss": 0.693, + "step": 20948 + }, + { + "epoch": 0.77, + "learning_rate": 1.2346651041857398e-07, + "loss": 0.7038, + "step": 20949 + }, + { + "epoch": 0.77, + "learning_rate": 1.2346284835390194e-07, + "loss": 0.6977, + "step": 20950 + }, + { + "epoch": 0.77, + "learning_rate": 1.2345918628922986e-07, + "loss": 0.6936, + "step": 20951 + }, + { + "epoch": 0.77, + "learning_rate": 1.2345552422455782e-07, + "loss": 0.6835, + "step": 20952 + }, + { + "epoch": 0.77, + "learning_rate": 1.2345186215988574e-07, + "loss": 0.7067, + "step": 20953 + }, + { + "epoch": 0.77, + "learning_rate": 1.234482000952137e-07, + "loss": 0.7029, + "step": 20954 + }, + { + "epoch": 0.77, + "learning_rate": 1.2344453803054162e-07, + "loss": 0.6928, + "step": 20955 + }, + { + "epoch": 0.77, + "learning_rate": 1.2344087596586955e-07, + "loss": 0.6893, + "step": 20956 + }, + { + "epoch": 0.77, + "learning_rate": 1.2343721390119748e-07, + "loss": 0.7026, + "step": 20957 + }, + { + "epoch": 0.77, + "learning_rate": 1.2343355183652543e-07, + "loss": 0.6921, + "step": 20958 + }, + { + "epoch": 0.77, + "learning_rate": 1.2342988977185336e-07, + "loss": 0.6992, + "step": 20959 + }, + { + "epoch": 0.77, + "learning_rate": 1.234262277071813e-07, + "loss": 0.7043, + "step": 20960 + }, + { + "epoch": 0.77, + "learning_rate": 1.2342256564250924e-07, + "loss": 0.6962, + "step": 20961 + }, + { + "epoch": 0.77, + "learning_rate": 1.2341890357783716e-07, + "loss": 0.6959, + "step": 20962 + }, + { + "epoch": 0.77, + "learning_rate": 1.2341524151316512e-07, + "loss": 0.697, + "step": 20963 + }, + { + "epoch": 0.77, + "learning_rate": 1.2341157944849304e-07, + "loss": 0.6896, + "step": 20964 + }, + { + "epoch": 0.77, + "learning_rate": 1.23407917383821e-07, + "loss": 0.6839, + "step": 20965 + }, + { + "epoch": 0.77, + "learning_rate": 1.2340425531914892e-07, + "loss": 0.6972, + "step": 20966 + }, + { + "epoch": 0.77, + "learning_rate": 1.2340059325447687e-07, + "loss": 0.7008, + "step": 20967 + }, + { + "epoch": 0.77, + "learning_rate": 1.233969311898048e-07, + "loss": 0.6915, + "step": 20968 + }, + { + "epoch": 0.77, + "learning_rate": 1.2339326912513275e-07, + "loss": 0.7003, + "step": 20969 + }, + { + "epoch": 0.77, + "learning_rate": 1.2338960706046068e-07, + "loss": 0.6992, + "step": 20970 + }, + { + "epoch": 0.77, + "learning_rate": 1.2338594499578863e-07, + "loss": 0.6839, + "step": 20971 + }, + { + "epoch": 0.77, + "learning_rate": 1.2338228293111656e-07, + "loss": 0.6988, + "step": 20972 + }, + { + "epoch": 0.77, + "learning_rate": 1.2337862086644451e-07, + "loss": 0.699, + "step": 20973 + }, + { + "epoch": 0.77, + "learning_rate": 1.2337495880177244e-07, + "loss": 0.6855, + "step": 20974 + }, + { + "epoch": 0.77, + "learning_rate": 1.2337129673710037e-07, + "loss": 0.6986, + "step": 20975 + }, + { + "epoch": 0.77, + "learning_rate": 1.233676346724283e-07, + "loss": 0.6955, + "step": 20976 + }, + { + "epoch": 0.77, + "learning_rate": 1.2336397260775625e-07, + "loss": 0.6954, + "step": 20977 + }, + { + "epoch": 0.77, + "learning_rate": 1.2336031054308417e-07, + "loss": 0.6961, + "step": 20978 + }, + { + "epoch": 0.77, + "learning_rate": 1.2335664847841213e-07, + "loss": 0.6834, + "step": 20979 + }, + { + "epoch": 0.77, + "learning_rate": 1.2335298641374005e-07, + "loss": 0.6904, + "step": 20980 + }, + { + "epoch": 0.77, + "learning_rate": 1.23349324349068e-07, + "loss": 0.6941, + "step": 20981 + }, + { + "epoch": 0.77, + "learning_rate": 1.2334566228439593e-07, + "loss": 0.6941, + "step": 20982 + }, + { + "epoch": 0.77, + "learning_rate": 1.233420002197239e-07, + "loss": 0.6986, + "step": 20983 + }, + { + "epoch": 0.77, + "learning_rate": 1.2333833815505181e-07, + "loss": 0.7036, + "step": 20984 + }, + { + "epoch": 0.77, + "learning_rate": 1.2333467609037974e-07, + "loss": 0.6994, + "step": 20985 + }, + { + "epoch": 0.77, + "learning_rate": 1.233310140257077e-07, + "loss": 0.6702, + "step": 20986 + }, + { + "epoch": 0.77, + "learning_rate": 1.2332735196103562e-07, + "loss": 0.693, + "step": 20987 + }, + { + "epoch": 0.77, + "learning_rate": 1.2332368989636357e-07, + "loss": 0.6963, + "step": 20988 + }, + { + "epoch": 0.77, + "learning_rate": 1.233200278316915e-07, + "loss": 0.6939, + "step": 20989 + }, + { + "epoch": 0.77, + "learning_rate": 1.2331636576701945e-07, + "loss": 0.7015, + "step": 20990 + }, + { + "epoch": 0.77, + "learning_rate": 1.2331270370234738e-07, + "loss": 0.6981, + "step": 20991 + }, + { + "epoch": 0.77, + "learning_rate": 1.2330904163767533e-07, + "loss": 0.6907, + "step": 20992 + }, + { + "epoch": 0.77, + "learning_rate": 1.2330537957300323e-07, + "loss": 0.6788, + "step": 20993 + }, + { + "epoch": 0.77, + "learning_rate": 1.2330171750833119e-07, + "loss": 0.6919, + "step": 20994 + }, + { + "epoch": 0.77, + "learning_rate": 1.2329805544365911e-07, + "loss": 0.6785, + "step": 20995 + }, + { + "epoch": 0.77, + "learning_rate": 1.2329439337898707e-07, + "loss": 0.6983, + "step": 20996 + }, + { + "epoch": 0.77, + "learning_rate": 1.23290731314315e-07, + "loss": 0.6995, + "step": 20997 + }, + { + "epoch": 0.77, + "learning_rate": 1.2328706924964295e-07, + "loss": 0.703, + "step": 20998 + }, + { + "epoch": 0.77, + "learning_rate": 1.2328340718497087e-07, + "loss": 0.6972, + "step": 20999 + }, + { + "epoch": 0.77, + "learning_rate": 1.2327974512029883e-07, + "loss": 0.6883, + "step": 21000 + }, + { + "epoch": 0.77, + "learning_rate": 1.2327608305562675e-07, + "loss": 0.7039, + "step": 21001 + }, + { + "epoch": 0.77, + "learning_rate": 1.232724209909547e-07, + "loss": 0.694, + "step": 21002 + }, + { + "epoch": 0.77, + "learning_rate": 1.2326875892628263e-07, + "loss": 0.6883, + "step": 21003 + }, + { + "epoch": 0.77, + "learning_rate": 1.2326509686161059e-07, + "loss": 0.6873, + "step": 21004 + }, + { + "epoch": 0.77, + "learning_rate": 1.232614347969385e-07, + "loss": 0.6992, + "step": 21005 + }, + { + "epoch": 0.77, + "learning_rate": 1.2325777273226644e-07, + "loss": 0.6957, + "step": 21006 + }, + { + "epoch": 0.77, + "learning_rate": 1.232541106675944e-07, + "loss": 0.6927, + "step": 21007 + }, + { + "epoch": 0.77, + "learning_rate": 1.2325044860292232e-07, + "loss": 0.7027, + "step": 21008 + }, + { + "epoch": 0.77, + "learning_rate": 1.2324678653825027e-07, + "loss": 0.6634, + "step": 21009 + }, + { + "epoch": 0.77, + "learning_rate": 1.232431244735782e-07, + "loss": 0.6942, + "step": 21010 + }, + { + "epoch": 0.77, + "learning_rate": 1.2323946240890613e-07, + "loss": 0.68, + "step": 21011 + }, + { + "epoch": 0.77, + "learning_rate": 1.2323580034423405e-07, + "loss": 0.6923, + "step": 21012 + }, + { + "epoch": 0.77, + "learning_rate": 1.23232138279562e-07, + "loss": 0.7062, + "step": 21013 + }, + { + "epoch": 0.77, + "learning_rate": 1.2322847621488993e-07, + "loss": 0.6951, + "step": 21014 + }, + { + "epoch": 0.77, + "learning_rate": 1.2322481415021789e-07, + "loss": 0.6917, + "step": 21015 + }, + { + "epoch": 0.77, + "learning_rate": 1.232211520855458e-07, + "loss": 0.6892, + "step": 21016 + }, + { + "epoch": 0.77, + "learning_rate": 1.2321749002087377e-07, + "loss": 0.6813, + "step": 21017 + }, + { + "epoch": 0.77, + "learning_rate": 1.232138279562017e-07, + "loss": 0.7023, + "step": 21018 + }, + { + "epoch": 0.77, + "learning_rate": 1.2321016589152965e-07, + "loss": 0.6899, + "step": 21019 + }, + { + "epoch": 0.77, + "learning_rate": 1.2320650382685757e-07, + "loss": 0.6951, + "step": 21020 + }, + { + "epoch": 0.77, + "learning_rate": 1.2320284176218553e-07, + "loss": 0.6812, + "step": 21021 + }, + { + "epoch": 0.77, + "learning_rate": 1.2319917969751345e-07, + "loss": 0.6912, + "step": 21022 + }, + { + "epoch": 0.77, + "learning_rate": 1.231955176328414e-07, + "loss": 0.6911, + "step": 21023 + }, + { + "epoch": 0.77, + "learning_rate": 1.2319185556816933e-07, + "loss": 0.6874, + "step": 21024 + }, + { + "epoch": 0.77, + "learning_rate": 1.2318819350349728e-07, + "loss": 0.6807, + "step": 21025 + }, + { + "epoch": 0.77, + "learning_rate": 1.231845314388252e-07, + "loss": 0.7053, + "step": 21026 + }, + { + "epoch": 0.77, + "learning_rate": 1.2318086937415314e-07, + "loss": 0.6919, + "step": 21027 + }, + { + "epoch": 0.77, + "learning_rate": 1.231772073094811e-07, + "loss": 0.6973, + "step": 21028 + }, + { + "epoch": 0.77, + "learning_rate": 1.2317354524480902e-07, + "loss": 0.699, + "step": 21029 + }, + { + "epoch": 0.77, + "learning_rate": 1.2316988318013694e-07, + "loss": 0.6825, + "step": 21030 + }, + { + "epoch": 0.77, + "learning_rate": 1.231662211154649e-07, + "loss": 0.7056, + "step": 21031 + }, + { + "epoch": 0.77, + "learning_rate": 1.2316255905079282e-07, + "loss": 0.6881, + "step": 21032 + }, + { + "epoch": 0.77, + "learning_rate": 1.2315889698612075e-07, + "loss": 0.6948, + "step": 21033 + }, + { + "epoch": 0.77, + "learning_rate": 1.231552349214487e-07, + "loss": 0.6772, + "step": 21034 + }, + { + "epoch": 0.77, + "learning_rate": 1.2315157285677663e-07, + "loss": 0.6976, + "step": 21035 + }, + { + "epoch": 0.77, + "learning_rate": 1.2314791079210458e-07, + "loss": 0.7024, + "step": 21036 + }, + { + "epoch": 0.77, + "learning_rate": 1.231442487274325e-07, + "loss": 0.6935, + "step": 21037 + }, + { + "epoch": 0.77, + "learning_rate": 1.2314058666276046e-07, + "loss": 0.6913, + "step": 21038 + }, + { + "epoch": 0.77, + "learning_rate": 1.231369245980884e-07, + "loss": 0.6953, + "step": 21039 + }, + { + "epoch": 0.77, + "learning_rate": 1.2313326253341634e-07, + "loss": 0.6954, + "step": 21040 + }, + { + "epoch": 0.77, + "learning_rate": 1.2312960046874427e-07, + "loss": 0.6971, + "step": 21041 + }, + { + "epoch": 0.77, + "learning_rate": 1.2312593840407222e-07, + "loss": 0.6956, + "step": 21042 + }, + { + "epoch": 0.77, + "learning_rate": 1.2312227633940015e-07, + "loss": 0.6845, + "step": 21043 + }, + { + "epoch": 0.77, + "learning_rate": 1.231186142747281e-07, + "loss": 0.6869, + "step": 21044 + }, + { + "epoch": 0.77, + "learning_rate": 1.2311495221005603e-07, + "loss": 0.7015, + "step": 21045 + }, + { + "epoch": 0.77, + "learning_rate": 1.2311129014538398e-07, + "loss": 0.6719, + "step": 21046 + }, + { + "epoch": 0.77, + "learning_rate": 1.2310762808071188e-07, + "loss": 0.6947, + "step": 21047 + }, + { + "epoch": 0.77, + "learning_rate": 1.2310396601603984e-07, + "loss": 0.6929, + "step": 21048 + }, + { + "epoch": 0.77, + "learning_rate": 1.2310030395136776e-07, + "loss": 0.6913, + "step": 21049 + }, + { + "epoch": 0.77, + "learning_rate": 1.2309664188669572e-07, + "loss": 0.6942, + "step": 21050 + }, + { + "epoch": 0.77, + "learning_rate": 1.2309297982202364e-07, + "loss": 0.6909, + "step": 21051 + }, + { + "epoch": 0.77, + "learning_rate": 1.230893177573516e-07, + "loss": 0.7027, + "step": 21052 + }, + { + "epoch": 0.77, + "learning_rate": 1.2308565569267952e-07, + "loss": 0.6927, + "step": 21053 + }, + { + "epoch": 0.77, + "learning_rate": 1.2308199362800748e-07, + "loss": 0.6926, + "step": 21054 + }, + { + "epoch": 0.77, + "learning_rate": 1.230783315633354e-07, + "loss": 0.7076, + "step": 21055 + }, + { + "epoch": 0.77, + "learning_rate": 1.2307466949866333e-07, + "loss": 0.6931, + "step": 21056 + }, + { + "epoch": 0.77, + "learning_rate": 1.2307100743399128e-07, + "loss": 0.6873, + "step": 21057 + }, + { + "epoch": 0.77, + "learning_rate": 1.230673453693192e-07, + "loss": 0.6922, + "step": 21058 + }, + { + "epoch": 0.77, + "learning_rate": 1.2306368330464716e-07, + "loss": 0.6851, + "step": 21059 + }, + { + "epoch": 0.77, + "learning_rate": 1.230600212399751e-07, + "loss": 0.7001, + "step": 21060 + }, + { + "epoch": 0.77, + "learning_rate": 1.2305635917530304e-07, + "loss": 0.6948, + "step": 21061 + }, + { + "epoch": 0.77, + "learning_rate": 1.2305269711063097e-07, + "loss": 0.7108, + "step": 21062 + }, + { + "epoch": 0.77, + "learning_rate": 1.2304903504595892e-07, + "loss": 0.6962, + "step": 21063 + }, + { + "epoch": 0.77, + "learning_rate": 1.2304537298128685e-07, + "loss": 0.6932, + "step": 21064 + }, + { + "epoch": 0.77, + "learning_rate": 1.2304171091661478e-07, + "loss": 0.6843, + "step": 21065 + }, + { + "epoch": 0.77, + "learning_rate": 1.230380488519427e-07, + "loss": 0.6912, + "step": 21066 + }, + { + "epoch": 0.77, + "learning_rate": 1.2303438678727066e-07, + "loss": 0.6904, + "step": 21067 + }, + { + "epoch": 0.77, + "learning_rate": 1.2303072472259858e-07, + "loss": 0.694, + "step": 21068 + }, + { + "epoch": 0.77, + "learning_rate": 1.2302706265792654e-07, + "loss": 0.6966, + "step": 21069 + }, + { + "epoch": 0.77, + "learning_rate": 1.2302340059325446e-07, + "loss": 0.6905, + "step": 21070 + }, + { + "epoch": 0.77, + "learning_rate": 1.2301973852858242e-07, + "loss": 0.6912, + "step": 21071 + }, + { + "epoch": 0.77, + "learning_rate": 1.2301607646391034e-07, + "loss": 0.6916, + "step": 21072 + }, + { + "epoch": 0.77, + "learning_rate": 1.230124143992383e-07, + "loss": 0.6875, + "step": 21073 + }, + { + "epoch": 0.77, + "learning_rate": 1.2300875233456622e-07, + "loss": 0.6988, + "step": 21074 + }, + { + "epoch": 0.77, + "learning_rate": 1.2300509026989418e-07, + "loss": 0.6961, + "step": 21075 + }, + { + "epoch": 0.77, + "learning_rate": 1.230014282052221e-07, + "loss": 0.6932, + "step": 21076 + }, + { + "epoch": 0.77, + "learning_rate": 1.2299776614055003e-07, + "loss": 0.6838, + "step": 21077 + }, + { + "epoch": 0.77, + "learning_rate": 1.2299410407587798e-07, + "loss": 0.7043, + "step": 21078 + }, + { + "epoch": 0.77, + "learning_rate": 1.229904420112059e-07, + "loss": 0.6947, + "step": 21079 + }, + { + "epoch": 0.77, + "learning_rate": 1.2298677994653386e-07, + "loss": 0.7117, + "step": 21080 + }, + { + "epoch": 0.77, + "learning_rate": 1.229831178818618e-07, + "loss": 0.6879, + "step": 21081 + }, + { + "epoch": 0.77, + "learning_rate": 1.2297945581718974e-07, + "loss": 0.6829, + "step": 21082 + }, + { + "epoch": 0.77, + "learning_rate": 1.2297579375251764e-07, + "loss": 0.6994, + "step": 21083 + }, + { + "epoch": 0.77, + "learning_rate": 1.229721316878456e-07, + "loss": 0.6999, + "step": 21084 + }, + { + "epoch": 0.77, + "learning_rate": 1.2296846962317352e-07, + "loss": 0.6939, + "step": 21085 + }, + { + "epoch": 0.77, + "learning_rate": 1.2296480755850147e-07, + "loss": 0.6978, + "step": 21086 + }, + { + "epoch": 0.77, + "learning_rate": 1.229611454938294e-07, + "loss": 0.6786, + "step": 21087 + }, + { + "epoch": 0.77, + "learning_rate": 1.2295748342915735e-07, + "loss": 0.6882, + "step": 21088 + }, + { + "epoch": 0.77, + "learning_rate": 1.2295382136448528e-07, + "loss": 0.6991, + "step": 21089 + }, + { + "epoch": 0.77, + "learning_rate": 1.2295015929981323e-07, + "loss": 0.6753, + "step": 21090 + }, + { + "epoch": 0.77, + "learning_rate": 1.2294649723514116e-07, + "loss": 0.6915, + "step": 21091 + }, + { + "epoch": 0.77, + "learning_rate": 1.2294283517046911e-07, + "loss": 0.6924, + "step": 21092 + }, + { + "epoch": 0.77, + "learning_rate": 1.2293917310579704e-07, + "loss": 0.6956, + "step": 21093 + }, + { + "epoch": 0.77, + "learning_rate": 1.22935511041125e-07, + "loss": 0.6962, + "step": 21094 + }, + { + "epoch": 0.77, + "learning_rate": 1.2293184897645292e-07, + "loss": 0.6916, + "step": 21095 + }, + { + "epoch": 0.77, + "learning_rate": 1.2292818691178087e-07, + "loss": 0.6842, + "step": 21096 + }, + { + "epoch": 0.77, + "learning_rate": 1.229245248471088e-07, + "loss": 0.6773, + "step": 21097 + }, + { + "epoch": 0.77, + "learning_rate": 1.2292086278243673e-07, + "loss": 0.7007, + "step": 21098 + }, + { + "epoch": 0.77, + "learning_rate": 1.2291720071776468e-07, + "loss": 0.6899, + "step": 21099 + }, + { + "epoch": 0.77, + "learning_rate": 1.229135386530926e-07, + "loss": 0.6945, + "step": 21100 + }, + { + "epoch": 0.77, + "learning_rate": 1.2290987658842053e-07, + "loss": 0.6921, + "step": 21101 + }, + { + "epoch": 0.77, + "learning_rate": 1.229062145237485e-07, + "loss": 0.6958, + "step": 21102 + }, + { + "epoch": 0.77, + "learning_rate": 1.2290255245907641e-07, + "loss": 0.6927, + "step": 21103 + }, + { + "epoch": 0.77, + "learning_rate": 1.2289889039440434e-07, + "loss": 0.6931, + "step": 21104 + }, + { + "epoch": 0.77, + "learning_rate": 1.228952283297323e-07, + "loss": 0.6797, + "step": 21105 + }, + { + "epoch": 0.77, + "learning_rate": 1.2289156626506022e-07, + "loss": 0.6973, + "step": 21106 + }, + { + "epoch": 0.77, + "learning_rate": 1.2288790420038817e-07, + "loss": 0.6909, + "step": 21107 + }, + { + "epoch": 0.77, + "learning_rate": 1.228842421357161e-07, + "loss": 0.6933, + "step": 21108 + }, + { + "epoch": 0.77, + "learning_rate": 1.2288058007104405e-07, + "loss": 0.6811, + "step": 21109 + }, + { + "epoch": 0.77, + "learning_rate": 1.2287691800637198e-07, + "loss": 0.692, + "step": 21110 + }, + { + "epoch": 0.77, + "learning_rate": 1.2287325594169993e-07, + "loss": 0.6886, + "step": 21111 + }, + { + "epoch": 0.77, + "learning_rate": 1.2286959387702786e-07, + "loss": 0.695, + "step": 21112 + }, + { + "epoch": 0.77, + "learning_rate": 1.228659318123558e-07, + "loss": 0.7225, + "step": 21113 + }, + { + "epoch": 0.77, + "learning_rate": 1.2286226974768374e-07, + "loss": 0.7039, + "step": 21114 + }, + { + "epoch": 0.77, + "learning_rate": 1.228586076830117e-07, + "loss": 0.7007, + "step": 21115 + }, + { + "epoch": 0.77, + "learning_rate": 1.2285494561833962e-07, + "loss": 0.6939, + "step": 21116 + }, + { + "epoch": 0.77, + "learning_rate": 1.2285128355366757e-07, + "loss": 0.6986, + "step": 21117 + }, + { + "epoch": 0.77, + "learning_rate": 1.228476214889955e-07, + "loss": 0.7006, + "step": 21118 + }, + { + "epoch": 0.77, + "learning_rate": 1.2284395942432343e-07, + "loss": 0.7058, + "step": 21119 + }, + { + "epoch": 0.77, + "learning_rate": 1.2284029735965135e-07, + "loss": 0.6854, + "step": 21120 + }, + { + "epoch": 0.77, + "learning_rate": 1.228366352949793e-07, + "loss": 0.7082, + "step": 21121 + }, + { + "epoch": 0.77, + "learning_rate": 1.2283297323030723e-07, + "loss": 0.7042, + "step": 21122 + }, + { + "epoch": 0.77, + "learning_rate": 1.2282931116563519e-07, + "loss": 0.699, + "step": 21123 + }, + { + "epoch": 0.77, + "learning_rate": 1.228256491009631e-07, + "loss": 0.6994, + "step": 21124 + }, + { + "epoch": 0.77, + "learning_rate": 1.2282198703629107e-07, + "loss": 0.6924, + "step": 21125 + }, + { + "epoch": 0.77, + "learning_rate": 1.22818324971619e-07, + "loss": 0.7043, + "step": 21126 + }, + { + "epoch": 0.77, + "learning_rate": 1.2281466290694692e-07, + "loss": 0.7036, + "step": 21127 + }, + { + "epoch": 0.77, + "learning_rate": 1.2281100084227487e-07, + "loss": 0.6893, + "step": 21128 + }, + { + "epoch": 0.77, + "learning_rate": 1.228073387776028e-07, + "loss": 0.7081, + "step": 21129 + }, + { + "epoch": 0.77, + "learning_rate": 1.2280367671293075e-07, + "loss": 0.6884, + "step": 21130 + }, + { + "epoch": 0.77, + "learning_rate": 1.2280001464825868e-07, + "loss": 0.6972, + "step": 21131 + }, + { + "epoch": 0.77, + "learning_rate": 1.2279635258358663e-07, + "loss": 0.6957, + "step": 21132 + }, + { + "epoch": 0.77, + "learning_rate": 1.2279269051891456e-07, + "loss": 0.7078, + "step": 21133 + }, + { + "epoch": 0.77, + "learning_rate": 1.227890284542425e-07, + "loss": 0.693, + "step": 21134 + }, + { + "epoch": 0.77, + "learning_rate": 1.2278536638957044e-07, + "loss": 0.7017, + "step": 21135 + }, + { + "epoch": 0.77, + "learning_rate": 1.227817043248984e-07, + "loss": 0.6756, + "step": 21136 + }, + { + "epoch": 0.77, + "learning_rate": 1.227780422602263e-07, + "loss": 0.7001, + "step": 21137 + }, + { + "epoch": 0.77, + "learning_rate": 1.2277438019555425e-07, + "loss": 0.6908, + "step": 21138 + }, + { + "epoch": 0.77, + "learning_rate": 1.2277071813088217e-07, + "loss": 0.6719, + "step": 21139 + }, + { + "epoch": 0.77, + "learning_rate": 1.2276705606621013e-07, + "loss": 0.691, + "step": 21140 + }, + { + "epoch": 0.77, + "learning_rate": 1.2276339400153805e-07, + "loss": 0.6837, + "step": 21141 + }, + { + "epoch": 0.77, + "learning_rate": 1.22759731936866e-07, + "loss": 0.7208, + "step": 21142 + }, + { + "epoch": 0.77, + "learning_rate": 1.2275606987219393e-07, + "loss": 0.6945, + "step": 21143 + }, + { + "epoch": 0.77, + "learning_rate": 1.2275240780752188e-07, + "loss": 0.6878, + "step": 21144 + }, + { + "epoch": 0.77, + "learning_rate": 1.227487457428498e-07, + "loss": 0.6875, + "step": 21145 + }, + { + "epoch": 0.77, + "learning_rate": 1.2274508367817776e-07, + "loss": 0.6946, + "step": 21146 + }, + { + "epoch": 0.77, + "learning_rate": 1.227414216135057e-07, + "loss": 0.7026, + "step": 21147 + }, + { + "epoch": 0.77, + "learning_rate": 1.2273775954883362e-07, + "loss": 0.6966, + "step": 21148 + }, + { + "epoch": 0.77, + "learning_rate": 1.2273409748416157e-07, + "loss": 0.6973, + "step": 21149 + }, + { + "epoch": 0.77, + "learning_rate": 1.227304354194895e-07, + "loss": 0.6942, + "step": 21150 + }, + { + "epoch": 0.77, + "learning_rate": 1.2272677335481745e-07, + "loss": 0.6975, + "step": 21151 + }, + { + "epoch": 0.77, + "learning_rate": 1.2272311129014538e-07, + "loss": 0.6826, + "step": 21152 + }, + { + "epoch": 0.77, + "learning_rate": 1.2271944922547333e-07, + "loss": 0.6901, + "step": 21153 + }, + { + "epoch": 0.77, + "learning_rate": 1.2271578716080126e-07, + "loss": 0.6946, + "step": 21154 + }, + { + "epoch": 0.77, + "learning_rate": 1.2271212509612918e-07, + "loss": 0.6917, + "step": 21155 + }, + { + "epoch": 0.77, + "learning_rate": 1.227084630314571e-07, + "loss": 0.6947, + "step": 21156 + }, + { + "epoch": 0.77, + "learning_rate": 1.2270480096678506e-07, + "loss": 0.6806, + "step": 21157 + }, + { + "epoch": 0.77, + "learning_rate": 1.22701138902113e-07, + "loss": 0.7031, + "step": 21158 + }, + { + "epoch": 0.77, + "learning_rate": 1.2269747683744094e-07, + "loss": 0.6994, + "step": 21159 + }, + { + "epoch": 0.77, + "learning_rate": 1.2269381477276887e-07, + "loss": 0.6893, + "step": 21160 + }, + { + "epoch": 0.77, + "learning_rate": 1.2269015270809682e-07, + "loss": 0.6953, + "step": 21161 + }, + { + "epoch": 0.77, + "learning_rate": 1.2268649064342475e-07, + "loss": 0.6941, + "step": 21162 + }, + { + "epoch": 0.77, + "learning_rate": 1.226828285787527e-07, + "loss": 0.7014, + "step": 21163 + }, + { + "epoch": 0.77, + "learning_rate": 1.2267916651408063e-07, + "loss": 0.689, + "step": 21164 + }, + { + "epoch": 0.77, + "learning_rate": 1.2267550444940858e-07, + "loss": 0.6952, + "step": 21165 + }, + { + "epoch": 0.77, + "learning_rate": 1.226718423847365e-07, + "loss": 0.7066, + "step": 21166 + }, + { + "epoch": 0.77, + "learning_rate": 1.2266818032006446e-07, + "loss": 0.676, + "step": 21167 + }, + { + "epoch": 0.77, + "learning_rate": 1.226645182553924e-07, + "loss": 0.694, + "step": 21168 + }, + { + "epoch": 0.77, + "learning_rate": 1.2266085619072032e-07, + "loss": 0.695, + "step": 21169 + }, + { + "epoch": 0.77, + "learning_rate": 1.2265719412604827e-07, + "loss": 0.692, + "step": 21170 + }, + { + "epoch": 0.77, + "learning_rate": 1.226535320613762e-07, + "loss": 0.6843, + "step": 21171 + }, + { + "epoch": 0.77, + "learning_rate": 1.2264986999670415e-07, + "loss": 0.6956, + "step": 21172 + }, + { + "epoch": 0.77, + "learning_rate": 1.2264620793203208e-07, + "loss": 0.7106, + "step": 21173 + }, + { + "epoch": 0.77, + "learning_rate": 1.2264254586736e-07, + "loss": 0.7044, + "step": 21174 + }, + { + "epoch": 0.77, + "learning_rate": 1.2263888380268793e-07, + "loss": 0.7058, + "step": 21175 + }, + { + "epoch": 0.77, + "learning_rate": 1.2263522173801588e-07, + "loss": 0.6943, + "step": 21176 + }, + { + "epoch": 0.77, + "learning_rate": 1.226315596733438e-07, + "loss": 0.6806, + "step": 21177 + }, + { + "epoch": 0.77, + "learning_rate": 1.2262789760867176e-07, + "loss": 0.6814, + "step": 21178 + }, + { + "epoch": 0.77, + "learning_rate": 1.226242355439997e-07, + "loss": 0.6888, + "step": 21179 + }, + { + "epoch": 0.77, + "learning_rate": 1.2262057347932764e-07, + "loss": 0.7036, + "step": 21180 + }, + { + "epoch": 0.77, + "learning_rate": 1.2261691141465557e-07, + "loss": 0.6933, + "step": 21181 + }, + { + "epoch": 0.77, + "learning_rate": 1.2261324934998352e-07, + "loss": 0.6911, + "step": 21182 + }, + { + "epoch": 0.78, + "learning_rate": 1.2260958728531145e-07, + "loss": 0.6939, + "step": 21183 + }, + { + "epoch": 0.78, + "learning_rate": 1.226059252206394e-07, + "loss": 0.693, + "step": 21184 + }, + { + "epoch": 0.78, + "learning_rate": 1.2260226315596733e-07, + "loss": 0.6882, + "step": 21185 + }, + { + "epoch": 0.78, + "learning_rate": 1.2259860109129528e-07, + "loss": 0.7003, + "step": 21186 + }, + { + "epoch": 0.78, + "learning_rate": 1.225949390266232e-07, + "loss": 0.6943, + "step": 21187 + }, + { + "epoch": 0.78, + "learning_rate": 1.2259127696195116e-07, + "loss": 0.6854, + "step": 21188 + }, + { + "epoch": 0.78, + "learning_rate": 1.225876148972791e-07, + "loss": 0.6991, + "step": 21189 + }, + { + "epoch": 0.78, + "learning_rate": 1.2258395283260704e-07, + "loss": 0.6872, + "step": 21190 + }, + { + "epoch": 0.78, + "learning_rate": 1.2258029076793494e-07, + "loss": 0.6932, + "step": 21191 + }, + { + "epoch": 0.78, + "learning_rate": 1.225766287032629e-07, + "loss": 0.6911, + "step": 21192 + }, + { + "epoch": 0.78, + "learning_rate": 1.2257296663859082e-07, + "loss": 0.6874, + "step": 21193 + }, + { + "epoch": 0.78, + "learning_rate": 1.2256930457391878e-07, + "loss": 0.6917, + "step": 21194 + }, + { + "epoch": 0.78, + "learning_rate": 1.225656425092467e-07, + "loss": 0.6953, + "step": 21195 + }, + { + "epoch": 0.78, + "learning_rate": 1.2256198044457463e-07, + "loss": 0.6972, + "step": 21196 + }, + { + "epoch": 0.78, + "learning_rate": 1.2255831837990258e-07, + "loss": 0.6963, + "step": 21197 + }, + { + "epoch": 0.78, + "learning_rate": 1.225546563152305e-07, + "loss": 0.6917, + "step": 21198 + }, + { + "epoch": 0.78, + "learning_rate": 1.2255099425055846e-07, + "loss": 0.6909, + "step": 21199 + }, + { + "epoch": 0.78, + "learning_rate": 1.225473321858864e-07, + "loss": 0.6928, + "step": 21200 + }, + { + "epoch": 0.78, + "learning_rate": 1.2254367012121434e-07, + "loss": 0.6916, + "step": 21201 + }, + { + "epoch": 0.78, + "learning_rate": 1.2254000805654227e-07, + "loss": 0.6985, + "step": 21202 + }, + { + "epoch": 0.78, + "learning_rate": 1.2253634599187022e-07, + "loss": 0.706, + "step": 21203 + }, + { + "epoch": 0.78, + "learning_rate": 1.2253268392719815e-07, + "loss": 0.6897, + "step": 21204 + }, + { + "epoch": 0.78, + "learning_rate": 1.225290218625261e-07, + "loss": 0.7024, + "step": 21205 + }, + { + "epoch": 0.78, + "learning_rate": 1.2252535979785403e-07, + "loss": 0.6992, + "step": 21206 + }, + { + "epoch": 0.78, + "learning_rate": 1.2252169773318198e-07, + "loss": 0.6983, + "step": 21207 + }, + { + "epoch": 0.78, + "learning_rate": 1.2251803566850988e-07, + "loss": 0.6999, + "step": 21208 + }, + { + "epoch": 0.78, + "learning_rate": 1.2251437360383783e-07, + "loss": 0.6888, + "step": 21209 + }, + { + "epoch": 0.78, + "learning_rate": 1.2251071153916576e-07, + "loss": 0.6834, + "step": 21210 + }, + { + "epoch": 0.78, + "learning_rate": 1.2250704947449371e-07, + "loss": 0.6982, + "step": 21211 + }, + { + "epoch": 0.78, + "learning_rate": 1.2250338740982164e-07, + "loss": 0.6855, + "step": 21212 + }, + { + "epoch": 0.78, + "learning_rate": 1.224997253451496e-07, + "loss": 0.691, + "step": 21213 + }, + { + "epoch": 0.78, + "learning_rate": 1.2249606328047752e-07, + "loss": 0.7055, + "step": 21214 + }, + { + "epoch": 0.78, + "learning_rate": 1.2249240121580547e-07, + "loss": 0.6893, + "step": 21215 + }, + { + "epoch": 0.78, + "learning_rate": 1.224887391511334e-07, + "loss": 0.6922, + "step": 21216 + }, + { + "epoch": 0.78, + "learning_rate": 1.2248507708646135e-07, + "loss": 0.6847, + "step": 21217 + }, + { + "epoch": 0.78, + "learning_rate": 1.2248141502178928e-07, + "loss": 0.6794, + "step": 21218 + }, + { + "epoch": 0.78, + "learning_rate": 1.224777529571172e-07, + "loss": 0.6821, + "step": 21219 + }, + { + "epoch": 0.78, + "learning_rate": 1.2247409089244516e-07, + "loss": 0.7096, + "step": 21220 + }, + { + "epoch": 0.78, + "learning_rate": 1.224704288277731e-07, + "loss": 0.6968, + "step": 21221 + }, + { + "epoch": 0.78, + "learning_rate": 1.2246676676310104e-07, + "loss": 0.683, + "step": 21222 + }, + { + "epoch": 0.78, + "learning_rate": 1.2246310469842897e-07, + "loss": 0.7051, + "step": 21223 + }, + { + "epoch": 0.78, + "learning_rate": 1.2245944263375692e-07, + "loss": 0.7038, + "step": 21224 + }, + { + "epoch": 0.78, + "learning_rate": 1.2245578056908485e-07, + "loss": 0.7004, + "step": 21225 + }, + { + "epoch": 0.78, + "learning_rate": 1.2245211850441277e-07, + "loss": 0.6995, + "step": 21226 + }, + { + "epoch": 0.78, + "learning_rate": 1.224484564397407e-07, + "loss": 0.6964, + "step": 21227 + }, + { + "epoch": 0.78, + "learning_rate": 1.2244479437506865e-07, + "loss": 0.7029, + "step": 21228 + }, + { + "epoch": 0.78, + "learning_rate": 1.2244113231039658e-07, + "loss": 0.6988, + "step": 21229 + }, + { + "epoch": 0.78, + "learning_rate": 1.2243747024572453e-07, + "loss": 0.6933, + "step": 21230 + }, + { + "epoch": 0.78, + "learning_rate": 1.2243380818105246e-07, + "loss": 0.7047, + "step": 21231 + }, + { + "epoch": 0.78, + "learning_rate": 1.224301461163804e-07, + "loss": 0.6975, + "step": 21232 + }, + { + "epoch": 0.78, + "learning_rate": 1.2242648405170834e-07, + "loss": 0.6926, + "step": 21233 + }, + { + "epoch": 0.78, + "learning_rate": 1.224228219870363e-07, + "loss": 0.6883, + "step": 21234 + }, + { + "epoch": 0.78, + "learning_rate": 1.2241915992236422e-07, + "loss": 0.6835, + "step": 21235 + }, + { + "epoch": 0.78, + "learning_rate": 1.2241549785769217e-07, + "loss": 0.6813, + "step": 21236 + }, + { + "epoch": 0.78, + "learning_rate": 1.224118357930201e-07, + "loss": 0.6941, + "step": 21237 + }, + { + "epoch": 0.78, + "learning_rate": 1.2240817372834805e-07, + "loss": 0.6939, + "step": 21238 + }, + { + "epoch": 0.78, + "learning_rate": 1.2240451166367598e-07, + "loss": 0.7043, + "step": 21239 + }, + { + "epoch": 0.78, + "learning_rate": 1.224008495990039e-07, + "loss": 0.7004, + "step": 21240 + }, + { + "epoch": 0.78, + "learning_rate": 1.2239718753433186e-07, + "loss": 0.6936, + "step": 21241 + }, + { + "epoch": 0.78, + "learning_rate": 1.2239352546965979e-07, + "loss": 0.6985, + "step": 21242 + }, + { + "epoch": 0.78, + "learning_rate": 1.2238986340498774e-07, + "loss": 0.6791, + "step": 21243 + }, + { + "epoch": 0.78, + "learning_rate": 1.2238620134031567e-07, + "loss": 0.685, + "step": 21244 + }, + { + "epoch": 0.78, + "learning_rate": 1.223825392756436e-07, + "loss": 0.6903, + "step": 21245 + }, + { + "epoch": 0.78, + "learning_rate": 1.2237887721097152e-07, + "loss": 0.692, + "step": 21246 + }, + { + "epoch": 0.78, + "learning_rate": 1.2237521514629947e-07, + "loss": 0.7056, + "step": 21247 + }, + { + "epoch": 0.78, + "learning_rate": 1.223715530816274e-07, + "loss": 0.6964, + "step": 21248 + }, + { + "epoch": 0.78, + "learning_rate": 1.2236789101695535e-07, + "loss": 0.6906, + "step": 21249 + }, + { + "epoch": 0.78, + "learning_rate": 1.2236422895228328e-07, + "loss": 0.6983, + "step": 21250 + }, + { + "epoch": 0.78, + "learning_rate": 1.2236056688761123e-07, + "loss": 0.7037, + "step": 21251 + }, + { + "epoch": 0.78, + "learning_rate": 1.2235690482293916e-07, + "loss": 0.7139, + "step": 21252 + }, + { + "epoch": 0.78, + "learning_rate": 1.223532427582671e-07, + "loss": 0.7042, + "step": 21253 + }, + { + "epoch": 0.78, + "learning_rate": 1.2234958069359504e-07, + "loss": 0.6996, + "step": 21254 + }, + { + "epoch": 0.78, + "learning_rate": 1.22345918628923e-07, + "loss": 0.6998, + "step": 21255 + }, + { + "epoch": 0.78, + "learning_rate": 1.2234225656425092e-07, + "loss": 0.6856, + "step": 21256 + }, + { + "epoch": 0.78, + "learning_rate": 1.2233859449957887e-07, + "loss": 0.697, + "step": 21257 + }, + { + "epoch": 0.78, + "learning_rate": 1.223349324349068e-07, + "loss": 0.7026, + "step": 21258 + }, + { + "epoch": 0.78, + "learning_rate": 1.2233127037023475e-07, + "loss": 0.6952, + "step": 21259 + }, + { + "epoch": 0.78, + "learning_rate": 1.2232760830556268e-07, + "loss": 0.6873, + "step": 21260 + }, + { + "epoch": 0.78, + "learning_rate": 1.2232394624089063e-07, + "loss": 0.7007, + "step": 21261 + }, + { + "epoch": 0.78, + "learning_rate": 1.2232028417621853e-07, + "loss": 0.6952, + "step": 21262 + }, + { + "epoch": 0.78, + "learning_rate": 1.2231662211154648e-07, + "loss": 0.6985, + "step": 21263 + }, + { + "epoch": 0.78, + "learning_rate": 1.223129600468744e-07, + "loss": 0.7019, + "step": 21264 + }, + { + "epoch": 0.78, + "learning_rate": 1.2230929798220236e-07, + "loss": 0.6843, + "step": 21265 + }, + { + "epoch": 0.78, + "learning_rate": 1.223056359175303e-07, + "loss": 0.6929, + "step": 21266 + }, + { + "epoch": 0.78, + "learning_rate": 1.2230197385285822e-07, + "loss": 0.7046, + "step": 21267 + }, + { + "epoch": 0.78, + "learning_rate": 1.2229831178818617e-07, + "loss": 0.697, + "step": 21268 + }, + { + "epoch": 0.78, + "learning_rate": 1.222946497235141e-07, + "loss": 0.7036, + "step": 21269 + }, + { + "epoch": 0.78, + "learning_rate": 1.2229098765884205e-07, + "loss": 0.6955, + "step": 21270 + }, + { + "epoch": 0.78, + "learning_rate": 1.2228732559416998e-07, + "loss": 0.7, + "step": 21271 + }, + { + "epoch": 0.78, + "learning_rate": 1.2228366352949793e-07, + "loss": 0.7022, + "step": 21272 + }, + { + "epoch": 0.78, + "learning_rate": 1.2228000146482586e-07, + "loss": 0.6923, + "step": 21273 + }, + { + "epoch": 0.78, + "learning_rate": 1.222763394001538e-07, + "loss": 0.6925, + "step": 21274 + }, + { + "epoch": 0.78, + "learning_rate": 1.2227267733548174e-07, + "loss": 0.7002, + "step": 21275 + }, + { + "epoch": 0.78, + "learning_rate": 1.222690152708097e-07, + "loss": 0.7024, + "step": 21276 + }, + { + "epoch": 0.78, + "learning_rate": 1.2226535320613762e-07, + "loss": 0.6941, + "step": 21277 + }, + { + "epoch": 0.78, + "learning_rate": 1.2226169114146557e-07, + "loss": 0.7134, + "step": 21278 + }, + { + "epoch": 0.78, + "learning_rate": 1.222580290767935e-07, + "loss": 0.7045, + "step": 21279 + }, + { + "epoch": 0.78, + "learning_rate": 1.2225436701212142e-07, + "loss": 0.7056, + "step": 21280 + }, + { + "epoch": 0.78, + "learning_rate": 1.2225070494744935e-07, + "loss": 0.7033, + "step": 21281 + }, + { + "epoch": 0.78, + "learning_rate": 1.222470428827773e-07, + "loss": 0.6947, + "step": 21282 + }, + { + "epoch": 0.78, + "learning_rate": 1.2224338081810523e-07, + "loss": 0.7069, + "step": 21283 + }, + { + "epoch": 0.78, + "learning_rate": 1.2223971875343318e-07, + "loss": 0.693, + "step": 21284 + }, + { + "epoch": 0.78, + "learning_rate": 1.222360566887611e-07, + "loss": 0.6953, + "step": 21285 + }, + { + "epoch": 0.78, + "learning_rate": 1.2223239462408906e-07, + "loss": 0.6859, + "step": 21286 + }, + { + "epoch": 0.78, + "learning_rate": 1.22228732559417e-07, + "loss": 0.6853, + "step": 21287 + }, + { + "epoch": 0.78, + "learning_rate": 1.2222507049474494e-07, + "loss": 0.6964, + "step": 21288 + }, + { + "epoch": 0.78, + "learning_rate": 1.2222140843007287e-07, + "loss": 0.6874, + "step": 21289 + }, + { + "epoch": 0.78, + "learning_rate": 1.222177463654008e-07, + "loss": 0.6889, + "step": 21290 + }, + { + "epoch": 0.78, + "learning_rate": 1.2221408430072875e-07, + "loss": 0.7069, + "step": 21291 + }, + { + "epoch": 0.78, + "learning_rate": 1.2221042223605668e-07, + "loss": 0.6909, + "step": 21292 + }, + { + "epoch": 0.78, + "learning_rate": 1.2220676017138463e-07, + "loss": 0.6985, + "step": 21293 + }, + { + "epoch": 0.78, + "learning_rate": 1.2220309810671256e-07, + "loss": 0.6954, + "step": 21294 + }, + { + "epoch": 0.78, + "learning_rate": 1.221994360420405e-07, + "loss": 0.6955, + "step": 21295 + }, + { + "epoch": 0.78, + "learning_rate": 1.2219577397736844e-07, + "loss": 0.6984, + "step": 21296 + }, + { + "epoch": 0.78, + "learning_rate": 1.221921119126964e-07, + "loss": 0.6839, + "step": 21297 + }, + { + "epoch": 0.78, + "learning_rate": 1.221884498480243e-07, + "loss": 0.6997, + "step": 21298 + }, + { + "epoch": 0.78, + "learning_rate": 1.2218478778335224e-07, + "loss": 0.6908, + "step": 21299 + }, + { + "epoch": 0.78, + "learning_rate": 1.2218112571868017e-07, + "loss": 0.6859, + "step": 21300 + }, + { + "epoch": 0.78, + "learning_rate": 1.2217746365400812e-07, + "loss": 0.6826, + "step": 21301 + }, + { + "epoch": 0.78, + "learning_rate": 1.2217380158933605e-07, + "loss": 0.7083, + "step": 21302 + }, + { + "epoch": 0.78, + "learning_rate": 1.22170139524664e-07, + "loss": 0.6891, + "step": 21303 + }, + { + "epoch": 0.78, + "learning_rate": 1.2216647745999193e-07, + "loss": 0.687, + "step": 21304 + }, + { + "epoch": 0.78, + "learning_rate": 1.2216281539531988e-07, + "loss": 0.7072, + "step": 21305 + }, + { + "epoch": 0.78, + "learning_rate": 1.221591533306478e-07, + "loss": 0.7031, + "step": 21306 + }, + { + "epoch": 0.78, + "learning_rate": 1.2215549126597576e-07, + "loss": 0.7054, + "step": 21307 + }, + { + "epoch": 0.78, + "learning_rate": 1.221518292013037e-07, + "loss": 0.7007, + "step": 21308 + }, + { + "epoch": 0.78, + "learning_rate": 1.2214816713663164e-07, + "loss": 0.6917, + "step": 21309 + }, + { + "epoch": 0.78, + "learning_rate": 1.2214450507195957e-07, + "loss": 0.6959, + "step": 21310 + }, + { + "epoch": 0.78, + "learning_rate": 1.221408430072875e-07, + "loss": 0.6773, + "step": 21311 + }, + { + "epoch": 0.78, + "learning_rate": 1.2213718094261545e-07, + "loss": 0.689, + "step": 21312 + }, + { + "epoch": 0.78, + "learning_rate": 1.2213351887794338e-07, + "loss": 0.6932, + "step": 21313 + }, + { + "epoch": 0.78, + "learning_rate": 1.2212985681327133e-07, + "loss": 0.6853, + "step": 21314 + }, + { + "epoch": 0.78, + "learning_rate": 1.2212619474859926e-07, + "loss": 0.6881, + "step": 21315 + }, + { + "epoch": 0.78, + "learning_rate": 1.2212253268392718e-07, + "loss": 0.692, + "step": 21316 + }, + { + "epoch": 0.78, + "learning_rate": 1.221188706192551e-07, + "loss": 0.6942, + "step": 21317 + }, + { + "epoch": 0.78, + "learning_rate": 1.2211520855458306e-07, + "loss": 0.6847, + "step": 21318 + }, + { + "epoch": 0.78, + "learning_rate": 1.22111546489911e-07, + "loss": 0.6923, + "step": 21319 + }, + { + "epoch": 0.78, + "learning_rate": 1.2210788442523894e-07, + "loss": 0.691, + "step": 21320 + }, + { + "epoch": 0.78, + "learning_rate": 1.2210422236056687e-07, + "loss": 0.7007, + "step": 21321 + }, + { + "epoch": 0.78, + "learning_rate": 1.2210056029589482e-07, + "loss": 0.6979, + "step": 21322 + }, + { + "epoch": 0.78, + "learning_rate": 1.2209689823122275e-07, + "loss": 0.6998, + "step": 21323 + }, + { + "epoch": 0.78, + "learning_rate": 1.220932361665507e-07, + "loss": 0.6973, + "step": 21324 + }, + { + "epoch": 0.78, + "learning_rate": 1.2208957410187863e-07, + "loss": 0.7042, + "step": 21325 + }, + { + "epoch": 0.78, + "learning_rate": 1.2208591203720658e-07, + "loss": 0.6919, + "step": 21326 + }, + { + "epoch": 0.78, + "learning_rate": 1.220822499725345e-07, + "loss": 0.6923, + "step": 21327 + }, + { + "epoch": 0.78, + "learning_rate": 1.2207858790786246e-07, + "loss": 0.6799, + "step": 21328 + }, + { + "epoch": 0.78, + "learning_rate": 1.220749258431904e-07, + "loss": 0.6836, + "step": 21329 + }, + { + "epoch": 0.78, + "learning_rate": 1.2207126377851834e-07, + "loss": 0.6999, + "step": 21330 + }, + { + "epoch": 0.78, + "learning_rate": 1.2206760171384627e-07, + "loss": 0.6876, + "step": 21331 + }, + { + "epoch": 0.78, + "learning_rate": 1.2206393964917422e-07, + "loss": 0.7116, + "step": 21332 + }, + { + "epoch": 0.78, + "learning_rate": 1.2206027758450215e-07, + "loss": 0.7003, + "step": 21333 + }, + { + "epoch": 0.78, + "learning_rate": 1.2205661551983007e-07, + "loss": 0.6901, + "step": 21334 + }, + { + "epoch": 0.78, + "learning_rate": 1.22052953455158e-07, + "loss": 0.6902, + "step": 21335 + }, + { + "epoch": 0.78, + "learning_rate": 1.2204929139048595e-07, + "loss": 0.6867, + "step": 21336 + }, + { + "epoch": 0.78, + "learning_rate": 1.2204562932581388e-07, + "loss": 0.6965, + "step": 21337 + }, + { + "epoch": 0.78, + "learning_rate": 1.220419672611418e-07, + "loss": 0.706, + "step": 21338 + }, + { + "epoch": 0.78, + "learning_rate": 1.2203830519646976e-07, + "loss": 0.6967, + "step": 21339 + }, + { + "epoch": 0.78, + "learning_rate": 1.220346431317977e-07, + "loss": 0.697, + "step": 21340 + }, + { + "epoch": 0.78, + "learning_rate": 1.2203098106712564e-07, + "loss": 0.6961, + "step": 21341 + }, + { + "epoch": 0.78, + "learning_rate": 1.2202731900245357e-07, + "loss": 0.6961, + "step": 21342 + }, + { + "epoch": 0.78, + "learning_rate": 1.2202365693778152e-07, + "loss": 0.692, + "step": 21343 + }, + { + "epoch": 0.78, + "learning_rate": 1.2201999487310945e-07, + "loss": 0.6956, + "step": 21344 + }, + { + "epoch": 0.78, + "learning_rate": 1.220163328084374e-07, + "loss": 0.6903, + "step": 21345 + }, + { + "epoch": 0.78, + "learning_rate": 1.2201267074376533e-07, + "loss": 0.6909, + "step": 21346 + }, + { + "epoch": 0.78, + "learning_rate": 1.2200900867909328e-07, + "loss": 0.7042, + "step": 21347 + }, + { + "epoch": 0.78, + "learning_rate": 1.220053466144212e-07, + "loss": 0.6852, + "step": 21348 + }, + { + "epoch": 0.78, + "learning_rate": 1.2200168454974916e-07, + "loss": 0.6938, + "step": 21349 + }, + { + "epoch": 0.78, + "learning_rate": 1.2199802248507709e-07, + "loss": 0.6809, + "step": 21350 + }, + { + "epoch": 0.78, + "learning_rate": 1.2199436042040504e-07, + "loss": 0.7113, + "step": 21351 + }, + { + "epoch": 0.78, + "learning_rate": 1.2199069835573294e-07, + "loss": 0.7, + "step": 21352 + }, + { + "epoch": 0.78, + "learning_rate": 1.219870362910609e-07, + "loss": 0.7019, + "step": 21353 + }, + { + "epoch": 0.78, + "learning_rate": 1.2198337422638882e-07, + "loss": 0.69, + "step": 21354 + }, + { + "epoch": 0.78, + "learning_rate": 1.2197971216171677e-07, + "loss": 0.6869, + "step": 21355 + }, + { + "epoch": 0.78, + "learning_rate": 1.219760500970447e-07, + "loss": 0.6993, + "step": 21356 + }, + { + "epoch": 0.78, + "learning_rate": 1.2197238803237265e-07, + "loss": 0.6874, + "step": 21357 + }, + { + "epoch": 0.78, + "learning_rate": 1.2196872596770058e-07, + "loss": 0.6835, + "step": 21358 + }, + { + "epoch": 0.78, + "learning_rate": 1.2196506390302853e-07, + "loss": 0.6982, + "step": 21359 + }, + { + "epoch": 0.78, + "learning_rate": 1.2196140183835646e-07, + "loss": 0.6937, + "step": 21360 + }, + { + "epoch": 0.78, + "learning_rate": 1.2195773977368439e-07, + "loss": 0.7015, + "step": 21361 + }, + { + "epoch": 0.78, + "learning_rate": 1.2195407770901234e-07, + "loss": 0.6834, + "step": 21362 + }, + { + "epoch": 0.78, + "learning_rate": 1.2195041564434027e-07, + "loss": 0.6963, + "step": 21363 + }, + { + "epoch": 0.78, + "learning_rate": 1.2194675357966822e-07, + "loss": 0.686, + "step": 21364 + }, + { + "epoch": 0.78, + "learning_rate": 1.2194309151499615e-07, + "loss": 0.6928, + "step": 21365 + }, + { + "epoch": 0.78, + "learning_rate": 1.219394294503241e-07, + "loss": 0.6945, + "step": 21366 + }, + { + "epoch": 0.78, + "learning_rate": 1.2193576738565203e-07, + "loss": 0.6851, + "step": 21367 + }, + { + "epoch": 0.78, + "learning_rate": 1.2193210532097998e-07, + "loss": 0.6842, + "step": 21368 + }, + { + "epoch": 0.78, + "learning_rate": 1.219284432563079e-07, + "loss": 0.7026, + "step": 21369 + }, + { + "epoch": 0.78, + "learning_rate": 1.2192478119163583e-07, + "loss": 0.6845, + "step": 21370 + }, + { + "epoch": 0.78, + "learning_rate": 1.2192111912696376e-07, + "loss": 0.7053, + "step": 21371 + }, + { + "epoch": 0.78, + "learning_rate": 1.219174570622917e-07, + "loss": 0.686, + "step": 21372 + }, + { + "epoch": 0.78, + "learning_rate": 1.2191379499761964e-07, + "loss": 0.6954, + "step": 21373 + }, + { + "epoch": 0.78, + "learning_rate": 1.219101329329476e-07, + "loss": 0.6918, + "step": 21374 + }, + { + "epoch": 0.78, + "learning_rate": 1.2190647086827552e-07, + "loss": 0.6944, + "step": 21375 + }, + { + "epoch": 0.78, + "learning_rate": 1.2190280880360347e-07, + "loss": 0.6892, + "step": 21376 + }, + { + "epoch": 0.78, + "learning_rate": 1.218991467389314e-07, + "loss": 0.7029, + "step": 21377 + }, + { + "epoch": 0.78, + "learning_rate": 1.2189548467425935e-07, + "loss": 0.687, + "step": 21378 + }, + { + "epoch": 0.78, + "learning_rate": 1.2189182260958728e-07, + "loss": 0.707, + "step": 21379 + }, + { + "epoch": 0.78, + "learning_rate": 1.2188816054491523e-07, + "loss": 0.6937, + "step": 21380 + }, + { + "epoch": 0.78, + "learning_rate": 1.2188449848024316e-07, + "loss": 0.6882, + "step": 21381 + }, + { + "epoch": 0.78, + "learning_rate": 1.2188083641557108e-07, + "loss": 0.7021, + "step": 21382 + }, + { + "epoch": 0.78, + "learning_rate": 1.2187717435089904e-07, + "loss": 0.7036, + "step": 21383 + }, + { + "epoch": 0.78, + "learning_rate": 1.2187351228622696e-07, + "loss": 0.6812, + "step": 21384 + }, + { + "epoch": 0.78, + "learning_rate": 1.2186985022155492e-07, + "loss": 0.6953, + "step": 21385 + }, + { + "epoch": 0.78, + "learning_rate": 1.2186618815688284e-07, + "loss": 0.6804, + "step": 21386 + }, + { + "epoch": 0.78, + "learning_rate": 1.218625260922108e-07, + "loss": 0.6945, + "step": 21387 + }, + { + "epoch": 0.78, + "learning_rate": 1.218588640275387e-07, + "loss": 0.6884, + "step": 21388 + }, + { + "epoch": 0.78, + "learning_rate": 1.2185520196286665e-07, + "loss": 0.6868, + "step": 21389 + }, + { + "epoch": 0.78, + "learning_rate": 1.2185153989819458e-07, + "loss": 0.6912, + "step": 21390 + }, + { + "epoch": 0.78, + "learning_rate": 1.2184787783352253e-07, + "loss": 0.6917, + "step": 21391 + }, + { + "epoch": 0.78, + "learning_rate": 1.2184421576885046e-07, + "loss": 0.6856, + "step": 21392 + }, + { + "epoch": 0.78, + "learning_rate": 1.218405537041784e-07, + "loss": 0.6978, + "step": 21393 + }, + { + "epoch": 0.78, + "learning_rate": 1.2183689163950634e-07, + "loss": 0.702, + "step": 21394 + }, + { + "epoch": 0.78, + "learning_rate": 1.218332295748343e-07, + "loss": 0.6798, + "step": 21395 + }, + { + "epoch": 0.78, + "learning_rate": 1.2182956751016222e-07, + "loss": 0.7042, + "step": 21396 + }, + { + "epoch": 0.78, + "learning_rate": 1.2182590544549017e-07, + "loss": 0.6877, + "step": 21397 + }, + { + "epoch": 0.78, + "learning_rate": 1.218222433808181e-07, + "loss": 0.6953, + "step": 21398 + }, + { + "epoch": 0.78, + "learning_rate": 1.2181858131614605e-07, + "loss": 0.6951, + "step": 21399 + }, + { + "epoch": 0.78, + "learning_rate": 1.2181491925147398e-07, + "loss": 0.6925, + "step": 21400 + }, + { + "epoch": 0.78, + "learning_rate": 1.2181125718680193e-07, + "loss": 0.6945, + "step": 21401 + }, + { + "epoch": 0.78, + "learning_rate": 1.2180759512212986e-07, + "loss": 0.6924, + "step": 21402 + }, + { + "epoch": 0.78, + "learning_rate": 1.218039330574578e-07, + "loss": 0.6948, + "step": 21403 + }, + { + "epoch": 0.78, + "learning_rate": 1.2180027099278574e-07, + "loss": 0.6999, + "step": 21404 + }, + { + "epoch": 0.78, + "learning_rate": 1.2179660892811366e-07, + "loss": 0.6993, + "step": 21405 + }, + { + "epoch": 0.78, + "learning_rate": 1.217929468634416e-07, + "loss": 0.691, + "step": 21406 + }, + { + "epoch": 0.78, + "learning_rate": 1.2178928479876954e-07, + "loss": 0.6822, + "step": 21407 + }, + { + "epoch": 0.78, + "learning_rate": 1.2178562273409747e-07, + "loss": 0.6971, + "step": 21408 + }, + { + "epoch": 0.78, + "learning_rate": 1.217819606694254e-07, + "loss": 0.6955, + "step": 21409 + }, + { + "epoch": 0.78, + "learning_rate": 1.2177829860475335e-07, + "loss": 0.6983, + "step": 21410 + }, + { + "epoch": 0.78, + "learning_rate": 1.2177463654008128e-07, + "loss": 0.7078, + "step": 21411 + }, + { + "epoch": 0.78, + "learning_rate": 1.2177097447540923e-07, + "loss": 0.6946, + "step": 21412 + }, + { + "epoch": 0.78, + "learning_rate": 1.2176731241073716e-07, + "loss": 0.7072, + "step": 21413 + }, + { + "epoch": 0.78, + "learning_rate": 1.217636503460651e-07, + "loss": 0.7035, + "step": 21414 + }, + { + "epoch": 0.78, + "learning_rate": 1.2175998828139304e-07, + "loss": 0.6988, + "step": 21415 + }, + { + "epoch": 0.78, + "learning_rate": 1.21756326216721e-07, + "loss": 0.6932, + "step": 21416 + }, + { + "epoch": 0.78, + "learning_rate": 1.2175266415204892e-07, + "loss": 0.7019, + "step": 21417 + }, + { + "epoch": 0.78, + "learning_rate": 1.2174900208737687e-07, + "loss": 0.6938, + "step": 21418 + }, + { + "epoch": 0.78, + "learning_rate": 1.217453400227048e-07, + "loss": 0.6664, + "step": 21419 + }, + { + "epoch": 0.78, + "learning_rate": 1.2174167795803275e-07, + "loss": 0.7041, + "step": 21420 + }, + { + "epoch": 0.78, + "learning_rate": 1.2173801589336068e-07, + "loss": 0.6872, + "step": 21421 + }, + { + "epoch": 0.78, + "learning_rate": 1.2173435382868863e-07, + "loss": 0.6987, + "step": 21422 + }, + { + "epoch": 0.78, + "learning_rate": 1.2173069176401656e-07, + "loss": 0.7018, + "step": 21423 + }, + { + "epoch": 0.78, + "learning_rate": 1.2172702969934448e-07, + "loss": 0.6725, + "step": 21424 + }, + { + "epoch": 0.78, + "learning_rate": 1.217233676346724e-07, + "loss": 0.6893, + "step": 21425 + }, + { + "epoch": 0.78, + "learning_rate": 1.2171970557000036e-07, + "loss": 0.7062, + "step": 21426 + }, + { + "epoch": 0.78, + "learning_rate": 1.217160435053283e-07, + "loss": 0.6959, + "step": 21427 + }, + { + "epoch": 0.78, + "learning_rate": 1.2171238144065624e-07, + "loss": 0.6879, + "step": 21428 + }, + { + "epoch": 0.78, + "learning_rate": 1.2170871937598417e-07, + "loss": 0.6897, + "step": 21429 + }, + { + "epoch": 0.78, + "learning_rate": 1.2170505731131212e-07, + "loss": 0.6973, + "step": 21430 + }, + { + "epoch": 0.78, + "learning_rate": 1.2170139524664005e-07, + "loss": 0.6898, + "step": 21431 + }, + { + "epoch": 0.78, + "learning_rate": 1.2169773318196798e-07, + "loss": 0.6815, + "step": 21432 + }, + { + "epoch": 0.78, + "learning_rate": 1.2169407111729593e-07, + "loss": 0.6974, + "step": 21433 + }, + { + "epoch": 0.78, + "learning_rate": 1.2169040905262386e-07, + "loss": 0.6929, + "step": 21434 + }, + { + "epoch": 0.78, + "learning_rate": 1.216867469879518e-07, + "loss": 0.6974, + "step": 21435 + }, + { + "epoch": 0.78, + "learning_rate": 1.2168308492327974e-07, + "loss": 0.6924, + "step": 21436 + }, + { + "epoch": 0.78, + "learning_rate": 1.216794228586077e-07, + "loss": 0.7143, + "step": 21437 + }, + { + "epoch": 0.78, + "learning_rate": 1.2167576079393561e-07, + "loss": 0.6908, + "step": 21438 + }, + { + "epoch": 0.78, + "learning_rate": 1.2167209872926357e-07, + "loss": 0.6989, + "step": 21439 + }, + { + "epoch": 0.78, + "learning_rate": 1.216684366645915e-07, + "loss": 0.6932, + "step": 21440 + }, + { + "epoch": 0.78, + "learning_rate": 1.2166477459991945e-07, + "loss": 0.6828, + "step": 21441 + }, + { + "epoch": 0.78, + "learning_rate": 1.2166111253524735e-07, + "loss": 0.6938, + "step": 21442 + }, + { + "epoch": 0.78, + "learning_rate": 1.216574504705753e-07, + "loss": 0.6887, + "step": 21443 + }, + { + "epoch": 0.78, + "learning_rate": 1.2165378840590323e-07, + "loss": 0.6934, + "step": 21444 + }, + { + "epoch": 0.78, + "learning_rate": 1.2165012634123118e-07, + "loss": 0.7011, + "step": 21445 + }, + { + "epoch": 0.78, + "learning_rate": 1.216464642765591e-07, + "loss": 0.6975, + "step": 21446 + }, + { + "epoch": 0.78, + "learning_rate": 1.2164280221188706e-07, + "loss": 0.6858, + "step": 21447 + }, + { + "epoch": 0.78, + "learning_rate": 1.21639140147215e-07, + "loss": 0.6896, + "step": 21448 + }, + { + "epoch": 0.78, + "learning_rate": 1.2163547808254294e-07, + "loss": 0.6972, + "step": 21449 + }, + { + "epoch": 0.78, + "learning_rate": 1.2163181601787087e-07, + "loss": 0.6935, + "step": 21450 + }, + { + "epoch": 0.78, + "learning_rate": 1.2162815395319882e-07, + "loss": 0.6884, + "step": 21451 + }, + { + "epoch": 0.78, + "learning_rate": 1.2162449188852675e-07, + "loss": 0.6907, + "step": 21452 + }, + { + "epoch": 0.78, + "learning_rate": 1.2162082982385467e-07, + "loss": 0.6884, + "step": 21453 + }, + { + "epoch": 0.78, + "learning_rate": 1.2161716775918263e-07, + "loss": 0.6799, + "step": 21454 + }, + { + "epoch": 0.78, + "learning_rate": 1.2161350569451055e-07, + "loss": 0.6847, + "step": 21455 + }, + { + "epoch": 0.79, + "learning_rate": 1.216098436298385e-07, + "loss": 0.6866, + "step": 21456 + }, + { + "epoch": 0.79, + "learning_rate": 1.2160618156516643e-07, + "loss": 0.6961, + "step": 21457 + }, + { + "epoch": 0.79, + "learning_rate": 1.216025195004944e-07, + "loss": 0.7034, + "step": 21458 + }, + { + "epoch": 0.79, + "learning_rate": 1.2159885743582231e-07, + "loss": 0.6928, + "step": 21459 + }, + { + "epoch": 0.79, + "learning_rate": 1.2159519537115024e-07, + "loss": 0.6902, + "step": 21460 + }, + { + "epoch": 0.79, + "learning_rate": 1.2159153330647817e-07, + "loss": 0.6883, + "step": 21461 + }, + { + "epoch": 0.79, + "learning_rate": 1.2158787124180612e-07, + "loss": 0.7059, + "step": 21462 + }, + { + "epoch": 0.79, + "learning_rate": 1.2158420917713405e-07, + "loss": 0.6872, + "step": 21463 + }, + { + "epoch": 0.79, + "learning_rate": 1.21580547112462e-07, + "loss": 0.6893, + "step": 21464 + }, + { + "epoch": 0.79, + "learning_rate": 1.2157688504778993e-07, + "loss": 0.6981, + "step": 21465 + }, + { + "epoch": 0.79, + "learning_rate": 1.2157322298311788e-07, + "loss": 0.694, + "step": 21466 + }, + { + "epoch": 0.79, + "learning_rate": 1.215695609184458e-07, + "loss": 0.6852, + "step": 21467 + }, + { + "epoch": 0.79, + "learning_rate": 1.2156589885377376e-07, + "loss": 0.6952, + "step": 21468 + }, + { + "epoch": 0.79, + "learning_rate": 1.2156223678910169e-07, + "loss": 0.701, + "step": 21469 + }, + { + "epoch": 0.79, + "learning_rate": 1.2155857472442964e-07, + "loss": 0.6829, + "step": 21470 + }, + { + "epoch": 0.79, + "learning_rate": 1.2155491265975757e-07, + "loss": 0.6896, + "step": 21471 + }, + { + "epoch": 0.79, + "learning_rate": 1.2155125059508552e-07, + "loss": 0.7001, + "step": 21472 + }, + { + "epoch": 0.79, + "learning_rate": 1.2154758853041345e-07, + "loss": 0.714, + "step": 21473 + }, + { + "epoch": 0.79, + "learning_rate": 1.215439264657414e-07, + "loss": 0.7067, + "step": 21474 + }, + { + "epoch": 0.79, + "learning_rate": 1.2154026440106933e-07, + "loss": 0.6923, + "step": 21475 + }, + { + "epoch": 0.79, + "learning_rate": 1.2153660233639725e-07, + "loss": 0.7045, + "step": 21476 + }, + { + "epoch": 0.79, + "learning_rate": 1.215329402717252e-07, + "loss": 0.6987, + "step": 21477 + }, + { + "epoch": 0.79, + "learning_rate": 1.2152927820705313e-07, + "loss": 0.6993, + "step": 21478 + }, + { + "epoch": 0.79, + "learning_rate": 1.2152561614238106e-07, + "loss": 0.6905, + "step": 21479 + }, + { + "epoch": 0.79, + "learning_rate": 1.2152195407770899e-07, + "loss": 0.6787, + "step": 21480 + }, + { + "epoch": 0.79, + "learning_rate": 1.2151829201303694e-07, + "loss": 0.6978, + "step": 21481 + }, + { + "epoch": 0.79, + "learning_rate": 1.2151462994836487e-07, + "loss": 0.7042, + "step": 21482 + }, + { + "epoch": 0.79, + "learning_rate": 1.2151096788369282e-07, + "loss": 0.6737, + "step": 21483 + }, + { + "epoch": 0.79, + "learning_rate": 1.2150730581902075e-07, + "loss": 0.6911, + "step": 21484 + }, + { + "epoch": 0.79, + "learning_rate": 1.215036437543487e-07, + "loss": 0.6963, + "step": 21485 + }, + { + "epoch": 0.79, + "learning_rate": 1.2149998168967663e-07, + "loss": 0.693, + "step": 21486 + }, + { + "epoch": 0.79, + "learning_rate": 1.2149631962500458e-07, + "loss": 0.6923, + "step": 21487 + }, + { + "epoch": 0.79, + "learning_rate": 1.214926575603325e-07, + "loss": 0.6852, + "step": 21488 + }, + { + "epoch": 0.79, + "learning_rate": 1.2148899549566046e-07, + "loss": 0.6974, + "step": 21489 + }, + { + "epoch": 0.79, + "learning_rate": 1.2148533343098839e-07, + "loss": 0.6995, + "step": 21490 + }, + { + "epoch": 0.79, + "learning_rate": 1.2148167136631634e-07, + "loss": 0.6995, + "step": 21491 + }, + { + "epoch": 0.79, + "learning_rate": 1.2147800930164427e-07, + "loss": 0.6944, + "step": 21492 + }, + { + "epoch": 0.79, + "learning_rate": 1.2147434723697222e-07, + "loss": 0.6875, + "step": 21493 + }, + { + "epoch": 0.79, + "learning_rate": 1.2147068517230015e-07, + "loss": 0.6908, + "step": 21494 + }, + { + "epoch": 0.79, + "learning_rate": 1.214670231076281e-07, + "loss": 0.7102, + "step": 21495 + }, + { + "epoch": 0.79, + "learning_rate": 1.21463361042956e-07, + "loss": 0.6928, + "step": 21496 + }, + { + "epoch": 0.79, + "learning_rate": 1.2145969897828395e-07, + "loss": 0.6936, + "step": 21497 + }, + { + "epoch": 0.79, + "learning_rate": 1.2145603691361188e-07, + "loss": 0.7023, + "step": 21498 + }, + { + "epoch": 0.79, + "learning_rate": 1.2145237484893983e-07, + "loss": 0.6923, + "step": 21499 + }, + { + "epoch": 0.79, + "learning_rate": 1.2144871278426776e-07, + "loss": 0.7013, + "step": 21500 + }, + { + "epoch": 0.79, + "learning_rate": 1.214450507195957e-07, + "loss": 0.6938, + "step": 21501 + }, + { + "epoch": 0.79, + "learning_rate": 1.2144138865492364e-07, + "loss": 0.6918, + "step": 21502 + }, + { + "epoch": 0.79, + "learning_rate": 1.2143772659025156e-07, + "loss": 0.6968, + "step": 21503 + }, + { + "epoch": 0.79, + "learning_rate": 1.2143406452557952e-07, + "loss": 0.7051, + "step": 21504 + }, + { + "epoch": 0.79, + "learning_rate": 1.2143040246090744e-07, + "loss": 0.6842, + "step": 21505 + }, + { + "epoch": 0.79, + "learning_rate": 1.214267403962354e-07, + "loss": 0.6862, + "step": 21506 + }, + { + "epoch": 0.79, + "learning_rate": 1.2142307833156332e-07, + "loss": 0.6994, + "step": 21507 + }, + { + "epoch": 0.79, + "learning_rate": 1.2141941626689128e-07, + "loss": 0.6909, + "step": 21508 + }, + { + "epoch": 0.79, + "learning_rate": 1.214157542022192e-07, + "loss": 0.7026, + "step": 21509 + }, + { + "epoch": 0.79, + "learning_rate": 1.2141209213754716e-07, + "loss": 0.6862, + "step": 21510 + }, + { + "epoch": 0.79, + "learning_rate": 1.2140843007287508e-07, + "loss": 0.6911, + "step": 21511 + }, + { + "epoch": 0.79, + "learning_rate": 1.2140476800820304e-07, + "loss": 0.6934, + "step": 21512 + }, + { + "epoch": 0.79, + "learning_rate": 1.2140110594353096e-07, + "loss": 0.6964, + "step": 21513 + }, + { + "epoch": 0.79, + "learning_rate": 1.213974438788589e-07, + "loss": 0.7077, + "step": 21514 + }, + { + "epoch": 0.79, + "learning_rate": 1.2139378181418682e-07, + "loss": 0.7063, + "step": 21515 + }, + { + "epoch": 0.79, + "learning_rate": 1.2139011974951477e-07, + "loss": 0.7004, + "step": 21516 + }, + { + "epoch": 0.79, + "learning_rate": 1.213864576848427e-07, + "loss": 0.6972, + "step": 21517 + }, + { + "epoch": 0.79, + "learning_rate": 1.2138279562017065e-07, + "loss": 0.6859, + "step": 21518 + }, + { + "epoch": 0.79, + "learning_rate": 1.2137913355549858e-07, + "loss": 0.6789, + "step": 21519 + }, + { + "epoch": 0.79, + "learning_rate": 1.2137547149082653e-07, + "loss": 0.6969, + "step": 21520 + }, + { + "epoch": 0.79, + "learning_rate": 1.2137180942615446e-07, + "loss": 0.712, + "step": 21521 + }, + { + "epoch": 0.79, + "learning_rate": 1.213681473614824e-07, + "loss": 0.7029, + "step": 21522 + }, + { + "epoch": 0.79, + "learning_rate": 1.2136448529681034e-07, + "loss": 0.6886, + "step": 21523 + }, + { + "epoch": 0.79, + "learning_rate": 1.2136082323213826e-07, + "loss": 0.7015, + "step": 21524 + }, + { + "epoch": 0.79, + "learning_rate": 1.2135716116746622e-07, + "loss": 0.7204, + "step": 21525 + }, + { + "epoch": 0.79, + "learning_rate": 1.2135349910279414e-07, + "loss": 0.6955, + "step": 21526 + }, + { + "epoch": 0.79, + "learning_rate": 1.213498370381221e-07, + "loss": 0.692, + "step": 21527 + }, + { + "epoch": 0.79, + "learning_rate": 1.2134617497345002e-07, + "loss": 0.695, + "step": 21528 + }, + { + "epoch": 0.79, + "learning_rate": 1.2134251290877798e-07, + "loss": 0.6971, + "step": 21529 + }, + { + "epoch": 0.79, + "learning_rate": 1.213388508441059e-07, + "loss": 0.7055, + "step": 21530 + }, + { + "epoch": 0.79, + "learning_rate": 1.2133518877943386e-07, + "loss": 0.679, + "step": 21531 + }, + { + "epoch": 0.79, + "learning_rate": 1.2133152671476176e-07, + "loss": 0.7008, + "step": 21532 + }, + { + "epoch": 0.79, + "learning_rate": 1.213278646500897e-07, + "loss": 0.6964, + "step": 21533 + }, + { + "epoch": 0.79, + "learning_rate": 1.2132420258541764e-07, + "loss": 0.691, + "step": 21534 + }, + { + "epoch": 0.79, + "learning_rate": 1.213205405207456e-07, + "loss": 0.677, + "step": 21535 + }, + { + "epoch": 0.79, + "learning_rate": 1.2131687845607352e-07, + "loss": 0.697, + "step": 21536 + }, + { + "epoch": 0.79, + "learning_rate": 1.2131321639140147e-07, + "loss": 0.6939, + "step": 21537 + }, + { + "epoch": 0.79, + "learning_rate": 1.213095543267294e-07, + "loss": 0.6812, + "step": 21538 + }, + { + "epoch": 0.79, + "learning_rate": 1.2130589226205735e-07, + "loss": 0.6978, + "step": 21539 + }, + { + "epoch": 0.79, + "learning_rate": 1.2130223019738528e-07, + "loss": 0.7047, + "step": 21540 + }, + { + "epoch": 0.79, + "learning_rate": 1.2129856813271323e-07, + "loss": 0.692, + "step": 21541 + }, + { + "epoch": 0.79, + "learning_rate": 1.2129490606804116e-07, + "loss": 0.6922, + "step": 21542 + }, + { + "epoch": 0.79, + "learning_rate": 1.212912440033691e-07, + "loss": 0.684, + "step": 21543 + }, + { + "epoch": 0.79, + "learning_rate": 1.2128758193869704e-07, + "loss": 0.6951, + "step": 21544 + }, + { + "epoch": 0.79, + "learning_rate": 1.21283919874025e-07, + "loss": 0.6758, + "step": 21545 + }, + { + "epoch": 0.79, + "learning_rate": 1.2128025780935292e-07, + "loss": 0.6881, + "step": 21546 + }, + { + "epoch": 0.79, + "learning_rate": 1.2127659574468084e-07, + "loss": 0.7025, + "step": 21547 + }, + { + "epoch": 0.79, + "learning_rate": 1.212729336800088e-07, + "loss": 0.6925, + "step": 21548 + }, + { + "epoch": 0.79, + "learning_rate": 1.2126927161533672e-07, + "loss": 0.6803, + "step": 21549 + }, + { + "epoch": 0.79, + "learning_rate": 1.2126560955066465e-07, + "loss": 0.6829, + "step": 21550 + }, + { + "epoch": 0.79, + "learning_rate": 1.2126194748599258e-07, + "loss": 0.6824, + "step": 21551 + }, + { + "epoch": 0.79, + "learning_rate": 1.2125828542132053e-07, + "loss": 0.7002, + "step": 21552 + }, + { + "epoch": 0.79, + "learning_rate": 1.2125462335664846e-07, + "loss": 0.6911, + "step": 21553 + }, + { + "epoch": 0.79, + "learning_rate": 1.212509612919764e-07, + "loss": 0.684, + "step": 21554 + }, + { + "epoch": 0.79, + "learning_rate": 1.2124729922730434e-07, + "loss": 0.6838, + "step": 21555 + }, + { + "epoch": 0.79, + "learning_rate": 1.212436371626323e-07, + "loss": 0.7107, + "step": 21556 + }, + { + "epoch": 0.79, + "learning_rate": 1.2123997509796022e-07, + "loss": 0.6985, + "step": 21557 + }, + { + "epoch": 0.79, + "learning_rate": 1.2123631303328817e-07, + "loss": 0.6992, + "step": 21558 + }, + { + "epoch": 0.79, + "learning_rate": 1.212326509686161e-07, + "loss": 0.7104, + "step": 21559 + }, + { + "epoch": 0.79, + "learning_rate": 1.2122898890394405e-07, + "loss": 0.6857, + "step": 21560 + }, + { + "epoch": 0.79, + "learning_rate": 1.2122532683927197e-07, + "loss": 0.6995, + "step": 21561 + }, + { + "epoch": 0.79, + "learning_rate": 1.2122166477459993e-07, + "loss": 0.7135, + "step": 21562 + }, + { + "epoch": 0.79, + "learning_rate": 1.2121800270992785e-07, + "loss": 0.696, + "step": 21563 + }, + { + "epoch": 0.79, + "learning_rate": 1.212143406452558e-07, + "loss": 0.6893, + "step": 21564 + }, + { + "epoch": 0.79, + "learning_rate": 1.2121067858058373e-07, + "loss": 0.7002, + "step": 21565 + }, + { + "epoch": 0.79, + "learning_rate": 1.212070165159117e-07, + "loss": 0.6926, + "step": 21566 + }, + { + "epoch": 0.79, + "learning_rate": 1.2120335445123961e-07, + "loss": 0.6852, + "step": 21567 + }, + { + "epoch": 0.79, + "learning_rate": 1.2119969238656754e-07, + "loss": 0.6938, + "step": 21568 + }, + { + "epoch": 0.79, + "learning_rate": 1.2119603032189547e-07, + "loss": 0.6966, + "step": 21569 + }, + { + "epoch": 0.79, + "learning_rate": 1.2119236825722342e-07, + "loss": 0.6984, + "step": 21570 + }, + { + "epoch": 0.79, + "learning_rate": 1.2118870619255135e-07, + "loss": 0.689, + "step": 21571 + }, + { + "epoch": 0.79, + "learning_rate": 1.211850441278793e-07, + "loss": 0.6848, + "step": 21572 + }, + { + "epoch": 0.79, + "learning_rate": 1.2118138206320723e-07, + "loss": 0.6915, + "step": 21573 + }, + { + "epoch": 0.79, + "learning_rate": 1.2117771999853515e-07, + "loss": 0.6852, + "step": 21574 + }, + { + "epoch": 0.79, + "learning_rate": 1.211740579338631e-07, + "loss": 0.6911, + "step": 21575 + }, + { + "epoch": 0.79, + "learning_rate": 1.2117039586919103e-07, + "loss": 0.704, + "step": 21576 + }, + { + "epoch": 0.79, + "learning_rate": 1.21166733804519e-07, + "loss": 0.6918, + "step": 21577 + }, + { + "epoch": 0.79, + "learning_rate": 1.2116307173984691e-07, + "loss": 0.6873, + "step": 21578 + }, + { + "epoch": 0.79, + "learning_rate": 1.2115940967517487e-07, + "loss": 0.6982, + "step": 21579 + }, + { + "epoch": 0.79, + "learning_rate": 1.211557476105028e-07, + "loss": 0.6949, + "step": 21580 + }, + { + "epoch": 0.79, + "learning_rate": 1.2115208554583075e-07, + "loss": 0.692, + "step": 21581 + }, + { + "epoch": 0.79, + "learning_rate": 1.2114842348115867e-07, + "loss": 0.7001, + "step": 21582 + }, + { + "epoch": 0.79, + "learning_rate": 1.2114476141648663e-07, + "loss": 0.6993, + "step": 21583 + }, + { + "epoch": 0.79, + "learning_rate": 1.2114109935181455e-07, + "loss": 0.6923, + "step": 21584 + }, + { + "epoch": 0.79, + "learning_rate": 1.211374372871425e-07, + "loss": 0.6958, + "step": 21585 + }, + { + "epoch": 0.79, + "learning_rate": 1.211337752224704e-07, + "loss": 0.7058, + "step": 21586 + }, + { + "epoch": 0.79, + "learning_rate": 1.2113011315779836e-07, + "loss": 0.7028, + "step": 21587 + }, + { + "epoch": 0.79, + "learning_rate": 1.2112645109312629e-07, + "loss": 0.6872, + "step": 21588 + }, + { + "epoch": 0.79, + "learning_rate": 1.2112278902845424e-07, + "loss": 0.6831, + "step": 21589 + }, + { + "epoch": 0.79, + "learning_rate": 1.2111912696378217e-07, + "loss": 0.704, + "step": 21590 + }, + { + "epoch": 0.79, + "learning_rate": 1.2111546489911012e-07, + "loss": 0.6999, + "step": 21591 + }, + { + "epoch": 0.79, + "learning_rate": 1.2111180283443805e-07, + "loss": 0.6853, + "step": 21592 + }, + { + "epoch": 0.79, + "learning_rate": 1.21108140769766e-07, + "loss": 0.6941, + "step": 21593 + }, + { + "epoch": 0.79, + "learning_rate": 1.2110447870509393e-07, + "loss": 0.6988, + "step": 21594 + }, + { + "epoch": 0.79, + "learning_rate": 1.2110081664042185e-07, + "loss": 0.6935, + "step": 21595 + }, + { + "epoch": 0.79, + "learning_rate": 1.210971545757498e-07, + "loss": 0.6974, + "step": 21596 + }, + { + "epoch": 0.79, + "learning_rate": 1.2109349251107773e-07, + "loss": 0.6952, + "step": 21597 + }, + { + "epoch": 0.79, + "learning_rate": 1.2108983044640569e-07, + "loss": 0.6982, + "step": 21598 + }, + { + "epoch": 0.79, + "learning_rate": 1.210861683817336e-07, + "loss": 0.6983, + "step": 21599 + }, + { + "epoch": 0.79, + "learning_rate": 1.2108250631706157e-07, + "loss": 0.7023, + "step": 21600 + }, + { + "epoch": 0.79, + "learning_rate": 1.210788442523895e-07, + "loss": 0.6966, + "step": 21601 + }, + { + "epoch": 0.79, + "learning_rate": 1.2107518218771745e-07, + "loss": 0.6936, + "step": 21602 + }, + { + "epoch": 0.79, + "learning_rate": 1.2107152012304535e-07, + "loss": 0.7012, + "step": 21603 + }, + { + "epoch": 0.79, + "learning_rate": 1.210678580583733e-07, + "loss": 0.6935, + "step": 21604 + }, + { + "epoch": 0.79, + "learning_rate": 1.2106419599370123e-07, + "loss": 0.67, + "step": 21605 + }, + { + "epoch": 0.79, + "learning_rate": 1.2106053392902918e-07, + "loss": 0.6936, + "step": 21606 + }, + { + "epoch": 0.79, + "learning_rate": 1.210568718643571e-07, + "loss": 0.7007, + "step": 21607 + }, + { + "epoch": 0.79, + "learning_rate": 1.2105320979968506e-07, + "loss": 0.6827, + "step": 21608 + }, + { + "epoch": 0.79, + "learning_rate": 1.2104954773501299e-07, + "loss": 0.6965, + "step": 21609 + }, + { + "epoch": 0.79, + "learning_rate": 1.2104588567034094e-07, + "loss": 0.6987, + "step": 21610 + }, + { + "epoch": 0.79, + "learning_rate": 1.2104222360566887e-07, + "loss": 0.6872, + "step": 21611 + }, + { + "epoch": 0.79, + "learning_rate": 1.2103856154099682e-07, + "loss": 0.7059, + "step": 21612 + }, + { + "epoch": 0.79, + "learning_rate": 1.2103489947632475e-07, + "loss": 0.6952, + "step": 21613 + }, + { + "epoch": 0.79, + "learning_rate": 1.210312374116527e-07, + "loss": 0.6915, + "step": 21614 + }, + { + "epoch": 0.79, + "learning_rate": 1.2102757534698062e-07, + "loss": 0.6942, + "step": 21615 + }, + { + "epoch": 0.79, + "learning_rate": 1.2102391328230858e-07, + "loss": 0.6844, + "step": 21616 + }, + { + "epoch": 0.79, + "learning_rate": 1.210202512176365e-07, + "loss": 0.6901, + "step": 21617 + }, + { + "epoch": 0.79, + "learning_rate": 1.2101658915296443e-07, + "loss": 0.6751, + "step": 21618 + }, + { + "epoch": 0.79, + "learning_rate": 1.2101292708829238e-07, + "loss": 0.6896, + "step": 21619 + }, + { + "epoch": 0.79, + "learning_rate": 1.210092650236203e-07, + "loss": 0.6925, + "step": 21620 + }, + { + "epoch": 0.79, + "learning_rate": 1.2100560295894824e-07, + "loss": 0.6934, + "step": 21621 + }, + { + "epoch": 0.79, + "learning_rate": 1.2100194089427616e-07, + "loss": 0.6929, + "step": 21622 + }, + { + "epoch": 0.79, + "learning_rate": 1.2099827882960412e-07, + "loss": 0.6827, + "step": 21623 + }, + { + "epoch": 0.79, + "learning_rate": 1.2099461676493204e-07, + "loss": 0.6901, + "step": 21624 + }, + { + "epoch": 0.79, + "learning_rate": 1.2099095470026e-07, + "loss": 0.6919, + "step": 21625 + }, + { + "epoch": 0.79, + "learning_rate": 1.2098729263558792e-07, + "loss": 0.6956, + "step": 21626 + }, + { + "epoch": 0.79, + "learning_rate": 1.2098363057091588e-07, + "loss": 0.6926, + "step": 21627 + }, + { + "epoch": 0.79, + "learning_rate": 1.209799685062438e-07, + "loss": 0.6928, + "step": 21628 + }, + { + "epoch": 0.79, + "learning_rate": 1.2097630644157176e-07, + "loss": 0.7055, + "step": 21629 + }, + { + "epoch": 0.79, + "learning_rate": 1.2097264437689968e-07, + "loss": 0.6944, + "step": 21630 + }, + { + "epoch": 0.79, + "learning_rate": 1.2096898231222764e-07, + "loss": 0.7114, + "step": 21631 + }, + { + "epoch": 0.79, + "learning_rate": 1.2096532024755556e-07, + "loss": 0.6915, + "step": 21632 + }, + { + "epoch": 0.79, + "learning_rate": 1.2096165818288352e-07, + "loss": 0.6851, + "step": 21633 + }, + { + "epoch": 0.79, + "learning_rate": 1.2095799611821144e-07, + "loss": 0.6852, + "step": 21634 + }, + { + "epoch": 0.79, + "learning_rate": 1.209543340535394e-07, + "loss": 0.6975, + "step": 21635 + }, + { + "epoch": 0.79, + "learning_rate": 1.2095067198886732e-07, + "loss": 0.6866, + "step": 21636 + }, + { + "epoch": 0.79, + "learning_rate": 1.2094700992419528e-07, + "loss": 0.6784, + "step": 21637 + }, + { + "epoch": 0.79, + "learning_rate": 1.209433478595232e-07, + "loss": 0.694, + "step": 21638 + }, + { + "epoch": 0.79, + "learning_rate": 1.2093968579485113e-07, + "loss": 0.6866, + "step": 21639 + }, + { + "epoch": 0.79, + "learning_rate": 1.2093602373017906e-07, + "loss": 0.6845, + "step": 21640 + }, + { + "epoch": 0.79, + "learning_rate": 1.20932361665507e-07, + "loss": 0.6911, + "step": 21641 + }, + { + "epoch": 0.79, + "learning_rate": 1.2092869960083494e-07, + "loss": 0.6866, + "step": 21642 + }, + { + "epoch": 0.79, + "learning_rate": 1.209250375361629e-07, + "loss": 0.6896, + "step": 21643 + }, + { + "epoch": 0.79, + "learning_rate": 1.2092137547149082e-07, + "loss": 0.7008, + "step": 21644 + }, + { + "epoch": 0.79, + "learning_rate": 1.2091771340681874e-07, + "loss": 0.7017, + "step": 21645 + }, + { + "epoch": 0.79, + "learning_rate": 1.209140513421467e-07, + "loss": 0.6838, + "step": 21646 + }, + { + "epoch": 0.79, + "learning_rate": 1.2091038927747462e-07, + "loss": 0.6924, + "step": 21647 + }, + { + "epoch": 0.79, + "learning_rate": 1.2090672721280258e-07, + "loss": 0.7018, + "step": 21648 + }, + { + "epoch": 0.79, + "learning_rate": 1.209030651481305e-07, + "loss": 0.6959, + "step": 21649 + }, + { + "epoch": 0.79, + "learning_rate": 1.2089940308345846e-07, + "loss": 0.6924, + "step": 21650 + }, + { + "epoch": 0.79, + "learning_rate": 1.2089574101878638e-07, + "loss": 0.6965, + "step": 21651 + }, + { + "epoch": 0.79, + "learning_rate": 1.2089207895411434e-07, + "loss": 0.691, + "step": 21652 + }, + { + "epoch": 0.79, + "learning_rate": 1.2088841688944226e-07, + "loss": 0.7006, + "step": 21653 + }, + { + "epoch": 0.79, + "learning_rate": 1.2088475482477022e-07, + "loss": 0.6853, + "step": 21654 + }, + { + "epoch": 0.79, + "learning_rate": 1.2088109276009814e-07, + "loss": 0.6976, + "step": 21655 + }, + { + "epoch": 0.79, + "learning_rate": 1.208774306954261e-07, + "loss": 0.7125, + "step": 21656 + }, + { + "epoch": 0.79, + "learning_rate": 1.20873768630754e-07, + "loss": 0.7054, + "step": 21657 + }, + { + "epoch": 0.79, + "learning_rate": 1.2087010656608195e-07, + "loss": 0.7003, + "step": 21658 + }, + { + "epoch": 0.79, + "learning_rate": 1.2086644450140988e-07, + "loss": 0.6892, + "step": 21659 + }, + { + "epoch": 0.79, + "learning_rate": 1.2086278243673783e-07, + "loss": 0.687, + "step": 21660 + }, + { + "epoch": 0.79, + "learning_rate": 1.2085912037206576e-07, + "loss": 0.6955, + "step": 21661 + }, + { + "epoch": 0.79, + "learning_rate": 1.208554583073937e-07, + "loss": 0.6872, + "step": 21662 + }, + { + "epoch": 0.79, + "learning_rate": 1.2085179624272164e-07, + "loss": 0.7017, + "step": 21663 + }, + { + "epoch": 0.79, + "learning_rate": 1.208481341780496e-07, + "loss": 0.6827, + "step": 21664 + }, + { + "epoch": 0.79, + "learning_rate": 1.2084447211337752e-07, + "loss": 0.6956, + "step": 21665 + }, + { + "epoch": 0.79, + "learning_rate": 1.2084081004870544e-07, + "loss": 0.6973, + "step": 21666 + }, + { + "epoch": 0.79, + "learning_rate": 1.208371479840334e-07, + "loss": 0.6826, + "step": 21667 + }, + { + "epoch": 0.79, + "learning_rate": 1.2083348591936132e-07, + "loss": 0.7022, + "step": 21668 + }, + { + "epoch": 0.79, + "learning_rate": 1.2082982385468928e-07, + "loss": 0.6986, + "step": 21669 + }, + { + "epoch": 0.79, + "learning_rate": 1.208261617900172e-07, + "loss": 0.7091, + "step": 21670 + }, + { + "epoch": 0.79, + "learning_rate": 1.2082249972534516e-07, + "loss": 0.6838, + "step": 21671 + }, + { + "epoch": 0.79, + "learning_rate": 1.2081883766067308e-07, + "loss": 0.7041, + "step": 21672 + }, + { + "epoch": 0.79, + "learning_rate": 1.2081517559600103e-07, + "loss": 0.6884, + "step": 21673 + }, + { + "epoch": 0.79, + "learning_rate": 1.2081151353132896e-07, + "loss": 0.6926, + "step": 21674 + }, + { + "epoch": 0.79, + "learning_rate": 1.208078514666569e-07, + "loss": 0.6795, + "step": 21675 + }, + { + "epoch": 0.79, + "learning_rate": 1.2080418940198482e-07, + "loss": 0.6981, + "step": 21676 + }, + { + "epoch": 0.79, + "learning_rate": 1.2080052733731277e-07, + "loss": 0.674, + "step": 21677 + }, + { + "epoch": 0.79, + "learning_rate": 1.207968652726407e-07, + "loss": 0.6901, + "step": 21678 + }, + { + "epoch": 0.79, + "learning_rate": 1.2079320320796865e-07, + "loss": 0.701, + "step": 21679 + }, + { + "epoch": 0.79, + "learning_rate": 1.2078954114329657e-07, + "loss": 0.685, + "step": 21680 + }, + { + "epoch": 0.79, + "learning_rate": 1.2078587907862453e-07, + "loss": 0.6863, + "step": 21681 + }, + { + "epoch": 0.79, + "learning_rate": 1.2078221701395245e-07, + "loss": 0.6933, + "step": 21682 + }, + { + "epoch": 0.79, + "learning_rate": 1.207785549492804e-07, + "loss": 0.6854, + "step": 21683 + }, + { + "epoch": 0.79, + "learning_rate": 1.2077489288460833e-07, + "loss": 0.6942, + "step": 21684 + }, + { + "epoch": 0.79, + "learning_rate": 1.207712308199363e-07, + "loss": 0.6947, + "step": 21685 + }, + { + "epoch": 0.79, + "learning_rate": 1.2076756875526421e-07, + "loss": 0.6926, + "step": 21686 + }, + { + "epoch": 0.79, + "learning_rate": 1.2076390669059217e-07, + "loss": 0.6864, + "step": 21687 + }, + { + "epoch": 0.79, + "learning_rate": 1.207602446259201e-07, + "loss": 0.6931, + "step": 21688 + }, + { + "epoch": 0.79, + "learning_rate": 1.2075658256124802e-07, + "loss": 0.6934, + "step": 21689 + }, + { + "epoch": 0.79, + "learning_rate": 1.2075292049657597e-07, + "loss": 0.6878, + "step": 21690 + }, + { + "epoch": 0.79, + "learning_rate": 1.207492584319039e-07, + "loss": 0.6903, + "step": 21691 + }, + { + "epoch": 0.79, + "learning_rate": 1.2074559636723185e-07, + "loss": 0.6957, + "step": 21692 + }, + { + "epoch": 0.79, + "learning_rate": 1.2074193430255975e-07, + "loss": 0.6933, + "step": 21693 + }, + { + "epoch": 0.79, + "learning_rate": 1.207382722378877e-07, + "loss": 0.7003, + "step": 21694 + }, + { + "epoch": 0.79, + "learning_rate": 1.2073461017321563e-07, + "loss": 0.7021, + "step": 21695 + }, + { + "epoch": 0.79, + "learning_rate": 1.207309481085436e-07, + "loss": 0.6972, + "step": 21696 + }, + { + "epoch": 0.79, + "learning_rate": 1.2072728604387151e-07, + "loss": 0.6886, + "step": 21697 + }, + { + "epoch": 0.79, + "learning_rate": 1.2072362397919947e-07, + "loss": 0.7114, + "step": 21698 + }, + { + "epoch": 0.79, + "learning_rate": 1.207199619145274e-07, + "loss": 0.6943, + "step": 21699 + }, + { + "epoch": 0.79, + "learning_rate": 1.2071629984985535e-07, + "loss": 0.6893, + "step": 21700 + }, + { + "epoch": 0.79, + "learning_rate": 1.2071263778518327e-07, + "loss": 0.687, + "step": 21701 + }, + { + "epoch": 0.79, + "learning_rate": 1.2070897572051123e-07, + "loss": 0.693, + "step": 21702 + }, + { + "epoch": 0.79, + "learning_rate": 1.2070531365583915e-07, + "loss": 0.6944, + "step": 21703 + }, + { + "epoch": 0.79, + "learning_rate": 1.207016515911671e-07, + "loss": 0.686, + "step": 21704 + }, + { + "epoch": 0.79, + "learning_rate": 1.2069798952649503e-07, + "loss": 0.7075, + "step": 21705 + }, + { + "epoch": 0.79, + "learning_rate": 1.2069432746182299e-07, + "loss": 0.69, + "step": 21706 + }, + { + "epoch": 0.79, + "learning_rate": 1.206906653971509e-07, + "loss": 0.6857, + "step": 21707 + }, + { + "epoch": 0.79, + "learning_rate": 1.2068700333247887e-07, + "loss": 0.6803, + "step": 21708 + }, + { + "epoch": 0.79, + "learning_rate": 1.206833412678068e-07, + "loss": 0.6942, + "step": 21709 + }, + { + "epoch": 0.79, + "learning_rate": 1.2067967920313472e-07, + "loss": 0.696, + "step": 21710 + }, + { + "epoch": 0.79, + "learning_rate": 1.2067601713846265e-07, + "loss": 0.6985, + "step": 21711 + }, + { + "epoch": 0.79, + "learning_rate": 1.206723550737906e-07, + "loss": 0.6863, + "step": 21712 + }, + { + "epoch": 0.79, + "learning_rate": 1.2066869300911853e-07, + "loss": 0.7066, + "step": 21713 + }, + { + "epoch": 0.79, + "learning_rate": 1.2066503094444648e-07, + "loss": 0.68, + "step": 21714 + }, + { + "epoch": 0.79, + "learning_rate": 1.206613688797744e-07, + "loss": 0.6934, + "step": 21715 + }, + { + "epoch": 0.79, + "learning_rate": 1.2065770681510233e-07, + "loss": 0.7001, + "step": 21716 + }, + { + "epoch": 0.79, + "learning_rate": 1.2065404475043029e-07, + "loss": 0.7025, + "step": 21717 + }, + { + "epoch": 0.79, + "learning_rate": 1.206503826857582e-07, + "loss": 0.7133, + "step": 21718 + }, + { + "epoch": 0.79, + "learning_rate": 1.2064672062108617e-07, + "loss": 0.6998, + "step": 21719 + }, + { + "epoch": 0.79, + "learning_rate": 1.206430585564141e-07, + "loss": 0.6844, + "step": 21720 + }, + { + "epoch": 0.79, + "learning_rate": 1.2063939649174205e-07, + "loss": 0.6885, + "step": 21721 + }, + { + "epoch": 0.79, + "learning_rate": 1.2063573442706997e-07, + "loss": 0.678, + "step": 21722 + }, + { + "epoch": 0.79, + "learning_rate": 1.2063207236239793e-07, + "loss": 0.6926, + "step": 21723 + }, + { + "epoch": 0.79, + "learning_rate": 1.2062841029772585e-07, + "loss": 0.6919, + "step": 21724 + }, + { + "epoch": 0.79, + "learning_rate": 1.206247482330538e-07, + "loss": 0.6873, + "step": 21725 + }, + { + "epoch": 0.79, + "learning_rate": 1.2062108616838173e-07, + "loss": 0.6924, + "step": 21726 + }, + { + "epoch": 0.79, + "learning_rate": 1.2061742410370969e-07, + "loss": 0.6973, + "step": 21727 + }, + { + "epoch": 0.79, + "learning_rate": 1.206137620390376e-07, + "loss": 0.6877, + "step": 21728 + }, + { + "epoch": 0.8, + "learning_rate": 1.2061009997436554e-07, + "loss": 0.701, + "step": 21729 + }, + { + "epoch": 0.8, + "learning_rate": 1.2060643790969347e-07, + "loss": 0.6884, + "step": 21730 + }, + { + "epoch": 0.8, + "learning_rate": 1.2060277584502142e-07, + "loss": 0.6907, + "step": 21731 + }, + { + "epoch": 0.8, + "learning_rate": 1.2059911378034935e-07, + "loss": 0.6997, + "step": 21732 + }, + { + "epoch": 0.8, + "learning_rate": 1.205954517156773e-07, + "loss": 0.6932, + "step": 21733 + }, + { + "epoch": 0.8, + "learning_rate": 1.2059178965100523e-07, + "loss": 0.7005, + "step": 21734 + }, + { + "epoch": 0.8, + "learning_rate": 1.2058812758633318e-07, + "loss": 0.7055, + "step": 21735 + }, + { + "epoch": 0.8, + "learning_rate": 1.205844655216611e-07, + "loss": 0.6832, + "step": 21736 + }, + { + "epoch": 0.8, + "learning_rate": 1.2058080345698903e-07, + "loss": 0.6945, + "step": 21737 + }, + { + "epoch": 0.8, + "learning_rate": 1.2057714139231698e-07, + "loss": 0.6849, + "step": 21738 + }, + { + "epoch": 0.8, + "learning_rate": 1.205734793276449e-07, + "loss": 0.7003, + "step": 21739 + }, + { + "epoch": 0.8, + "learning_rate": 1.2056981726297286e-07, + "loss": 0.6994, + "step": 21740 + }, + { + "epoch": 0.8, + "learning_rate": 1.205661551983008e-07, + "loss": 0.6932, + "step": 21741 + }, + { + "epoch": 0.8, + "learning_rate": 1.2056249313362874e-07, + "loss": 0.6933, + "step": 21742 + }, + { + "epoch": 0.8, + "learning_rate": 1.2055883106895667e-07, + "loss": 0.6935, + "step": 21743 + }, + { + "epoch": 0.8, + "learning_rate": 1.2055516900428462e-07, + "loss": 0.6887, + "step": 21744 + }, + { + "epoch": 0.8, + "learning_rate": 1.2055150693961255e-07, + "loss": 0.6951, + "step": 21745 + }, + { + "epoch": 0.8, + "learning_rate": 1.205478448749405e-07, + "loss": 0.6908, + "step": 21746 + }, + { + "epoch": 0.8, + "learning_rate": 1.205441828102684e-07, + "loss": 0.7015, + "step": 21747 + }, + { + "epoch": 0.8, + "learning_rate": 1.2054052074559636e-07, + "loss": 0.6937, + "step": 21748 + }, + { + "epoch": 0.8, + "learning_rate": 1.2053685868092428e-07, + "loss": 0.6997, + "step": 21749 + }, + { + "epoch": 0.8, + "learning_rate": 1.2053319661625224e-07, + "loss": 0.7063, + "step": 21750 + }, + { + "epoch": 0.8, + "learning_rate": 1.2052953455158016e-07, + "loss": 0.6981, + "step": 21751 + }, + { + "epoch": 0.8, + "learning_rate": 1.2052587248690812e-07, + "loss": 0.6983, + "step": 21752 + }, + { + "epoch": 0.8, + "learning_rate": 1.2052221042223604e-07, + "loss": 0.6971, + "step": 21753 + }, + { + "epoch": 0.8, + "learning_rate": 1.20518548357564e-07, + "loss": 0.7069, + "step": 21754 + }, + { + "epoch": 0.8, + "learning_rate": 1.2051488629289192e-07, + "loss": 0.7011, + "step": 21755 + }, + { + "epoch": 0.8, + "learning_rate": 1.2051122422821988e-07, + "loss": 0.6952, + "step": 21756 + }, + { + "epoch": 0.8, + "learning_rate": 1.205075621635478e-07, + "loss": 0.6927, + "step": 21757 + }, + { + "epoch": 0.8, + "learning_rate": 1.2050390009887576e-07, + "loss": 0.6955, + "step": 21758 + }, + { + "epoch": 0.8, + "learning_rate": 1.2050023803420368e-07, + "loss": 0.6972, + "step": 21759 + }, + { + "epoch": 0.8, + "learning_rate": 1.204965759695316e-07, + "loss": 0.7001, + "step": 21760 + }, + { + "epoch": 0.8, + "learning_rate": 1.2049291390485956e-07, + "loss": 0.7047, + "step": 21761 + }, + { + "epoch": 0.8, + "learning_rate": 1.204892518401875e-07, + "loss": 0.7005, + "step": 21762 + }, + { + "epoch": 0.8, + "learning_rate": 1.2048558977551544e-07, + "loss": 0.6813, + "step": 21763 + }, + { + "epoch": 0.8, + "learning_rate": 1.2048192771084337e-07, + "loss": 0.6929, + "step": 21764 + }, + { + "epoch": 0.8, + "learning_rate": 1.204782656461713e-07, + "loss": 0.7198, + "step": 21765 + }, + { + "epoch": 0.8, + "learning_rate": 1.2047460358149922e-07, + "loss": 0.6913, + "step": 21766 + }, + { + "epoch": 0.8, + "learning_rate": 1.2047094151682718e-07, + "loss": 0.6989, + "step": 21767 + }, + { + "epoch": 0.8, + "learning_rate": 1.204672794521551e-07, + "loss": 0.6847, + "step": 21768 + }, + { + "epoch": 0.8, + "learning_rate": 1.2046361738748306e-07, + "loss": 0.6866, + "step": 21769 + }, + { + "epoch": 0.8, + "learning_rate": 1.2045995532281098e-07, + "loss": 0.695, + "step": 21770 + }, + { + "epoch": 0.8, + "learning_rate": 1.2045629325813894e-07, + "loss": 0.6942, + "step": 21771 + }, + { + "epoch": 0.8, + "learning_rate": 1.2045263119346686e-07, + "loss": 0.6949, + "step": 21772 + }, + { + "epoch": 0.8, + "learning_rate": 1.2044896912879482e-07, + "loss": 0.7067, + "step": 21773 + }, + { + "epoch": 0.8, + "learning_rate": 1.2044530706412274e-07, + "loss": 0.6895, + "step": 21774 + }, + { + "epoch": 0.8, + "learning_rate": 1.204416449994507e-07, + "loss": 0.6964, + "step": 21775 + }, + { + "epoch": 0.8, + "learning_rate": 1.2043798293477862e-07, + "loss": 0.6978, + "step": 21776 + }, + { + "epoch": 0.8, + "learning_rate": 1.2043432087010658e-07, + "loss": 0.6884, + "step": 21777 + }, + { + "epoch": 0.8, + "learning_rate": 1.204306588054345e-07, + "loss": 0.6995, + "step": 21778 + }, + { + "epoch": 0.8, + "learning_rate": 1.2042699674076246e-07, + "loss": 0.6898, + "step": 21779 + }, + { + "epoch": 0.8, + "learning_rate": 1.2042333467609038e-07, + "loss": 0.68, + "step": 21780 + }, + { + "epoch": 0.8, + "learning_rate": 1.204196726114183e-07, + "loss": 0.6824, + "step": 21781 + }, + { + "epoch": 0.8, + "learning_rate": 1.2041601054674626e-07, + "loss": 0.6901, + "step": 21782 + }, + { + "epoch": 0.8, + "learning_rate": 1.204123484820742e-07, + "loss": 0.7028, + "step": 21783 + }, + { + "epoch": 0.8, + "learning_rate": 1.2040868641740212e-07, + "loss": 0.7036, + "step": 21784 + }, + { + "epoch": 0.8, + "learning_rate": 1.2040502435273007e-07, + "loss": 0.6923, + "step": 21785 + }, + { + "epoch": 0.8, + "learning_rate": 1.20401362288058e-07, + "loss": 0.6793, + "step": 21786 + }, + { + "epoch": 0.8, + "learning_rate": 1.2039770022338592e-07, + "loss": 0.6912, + "step": 21787 + }, + { + "epoch": 0.8, + "learning_rate": 1.2039403815871388e-07, + "loss": 0.6792, + "step": 21788 + }, + { + "epoch": 0.8, + "learning_rate": 1.203903760940418e-07, + "loss": 0.6838, + "step": 21789 + }, + { + "epoch": 0.8, + "learning_rate": 1.2038671402936976e-07, + "loss": 0.6953, + "step": 21790 + }, + { + "epoch": 0.8, + "learning_rate": 1.2038305196469768e-07, + "loss": 0.7136, + "step": 21791 + }, + { + "epoch": 0.8, + "learning_rate": 1.2037938990002564e-07, + "loss": 0.6851, + "step": 21792 + }, + { + "epoch": 0.8, + "learning_rate": 1.2037572783535356e-07, + "loss": 0.7025, + "step": 21793 + }, + { + "epoch": 0.8, + "learning_rate": 1.2037206577068151e-07, + "loss": 0.6866, + "step": 21794 + }, + { + "epoch": 0.8, + "learning_rate": 1.2036840370600944e-07, + "loss": 0.6887, + "step": 21795 + }, + { + "epoch": 0.8, + "learning_rate": 1.203647416413374e-07, + "loss": 0.697, + "step": 21796 + }, + { + "epoch": 0.8, + "learning_rate": 1.2036107957666532e-07, + "loss": 0.6958, + "step": 21797 + }, + { + "epoch": 0.8, + "learning_rate": 1.2035741751199327e-07, + "loss": 0.7036, + "step": 21798 + }, + { + "epoch": 0.8, + "learning_rate": 1.203537554473212e-07, + "loss": 0.6922, + "step": 21799 + }, + { + "epoch": 0.8, + "learning_rate": 1.2035009338264915e-07, + "loss": 0.6985, + "step": 21800 + }, + { + "epoch": 0.8, + "learning_rate": 1.2034643131797705e-07, + "loss": 0.689, + "step": 21801 + }, + { + "epoch": 0.8, + "learning_rate": 1.20342769253305e-07, + "loss": 0.6986, + "step": 21802 + }, + { + "epoch": 0.8, + "learning_rate": 1.2033910718863293e-07, + "loss": 0.6996, + "step": 21803 + }, + { + "epoch": 0.8, + "learning_rate": 1.203354451239609e-07, + "loss": 0.6995, + "step": 21804 + }, + { + "epoch": 0.8, + "learning_rate": 1.2033178305928881e-07, + "loss": 0.6788, + "step": 21805 + }, + { + "epoch": 0.8, + "learning_rate": 1.2032812099461677e-07, + "loss": 0.7134, + "step": 21806 + }, + { + "epoch": 0.8, + "learning_rate": 1.203244589299447e-07, + "loss": 0.6959, + "step": 21807 + }, + { + "epoch": 0.8, + "learning_rate": 1.2032079686527262e-07, + "loss": 0.6933, + "step": 21808 + }, + { + "epoch": 0.8, + "learning_rate": 1.2031713480060057e-07, + "loss": 0.7032, + "step": 21809 + }, + { + "epoch": 0.8, + "learning_rate": 1.203134727359285e-07, + "loss": 0.7019, + "step": 21810 + }, + { + "epoch": 0.8, + "learning_rate": 1.2030981067125645e-07, + "loss": 0.6893, + "step": 21811 + }, + { + "epoch": 0.8, + "learning_rate": 1.2030614860658438e-07, + "loss": 0.6892, + "step": 21812 + }, + { + "epoch": 0.8, + "learning_rate": 1.2030248654191233e-07, + "loss": 0.6803, + "step": 21813 + }, + { + "epoch": 0.8, + "learning_rate": 1.2029882447724026e-07, + "loss": 0.6946, + "step": 21814 + }, + { + "epoch": 0.8, + "learning_rate": 1.2029516241256821e-07, + "loss": 0.6935, + "step": 21815 + }, + { + "epoch": 0.8, + "learning_rate": 1.2029150034789614e-07, + "loss": 0.6978, + "step": 21816 + }, + { + "epoch": 0.8, + "learning_rate": 1.202878382832241e-07, + "loss": 0.6879, + "step": 21817 + }, + { + "epoch": 0.8, + "learning_rate": 1.2028417621855202e-07, + "loss": 0.692, + "step": 21818 + }, + { + "epoch": 0.8, + "learning_rate": 1.2028051415387995e-07, + "loss": 0.6967, + "step": 21819 + }, + { + "epoch": 0.8, + "learning_rate": 1.2027685208920787e-07, + "loss": 0.6909, + "step": 21820 + }, + { + "epoch": 0.8, + "learning_rate": 1.2027319002453583e-07, + "loss": 0.6979, + "step": 21821 + }, + { + "epoch": 0.8, + "learning_rate": 1.2026952795986375e-07, + "loss": 0.6916, + "step": 21822 + }, + { + "epoch": 0.8, + "learning_rate": 1.202658658951917e-07, + "loss": 0.6933, + "step": 21823 + }, + { + "epoch": 0.8, + "learning_rate": 1.2026220383051963e-07, + "loss": 0.6975, + "step": 21824 + }, + { + "epoch": 0.8, + "learning_rate": 1.2025854176584759e-07, + "loss": 0.6868, + "step": 21825 + }, + { + "epoch": 0.8, + "learning_rate": 1.202548797011755e-07, + "loss": 0.6941, + "step": 21826 + }, + { + "epoch": 0.8, + "learning_rate": 1.2025121763650347e-07, + "loss": 0.697, + "step": 21827 + }, + { + "epoch": 0.8, + "learning_rate": 1.202475555718314e-07, + "loss": 0.6812, + "step": 21828 + }, + { + "epoch": 0.8, + "learning_rate": 1.2024389350715935e-07, + "loss": 0.6975, + "step": 21829 + }, + { + "epoch": 0.8, + "learning_rate": 1.2024023144248727e-07, + "loss": 0.692, + "step": 21830 + }, + { + "epoch": 0.8, + "learning_rate": 1.202365693778152e-07, + "loss": 0.6804, + "step": 21831 + }, + { + "epoch": 0.8, + "learning_rate": 1.2023290731314315e-07, + "loss": 0.6924, + "step": 21832 + }, + { + "epoch": 0.8, + "learning_rate": 1.2022924524847108e-07, + "loss": 0.6984, + "step": 21833 + }, + { + "epoch": 0.8, + "learning_rate": 1.2022558318379903e-07, + "loss": 0.6898, + "step": 21834 + }, + { + "epoch": 0.8, + "learning_rate": 1.2022192111912696e-07, + "loss": 0.6846, + "step": 21835 + }, + { + "epoch": 0.8, + "learning_rate": 1.202182590544549e-07, + "loss": 0.6935, + "step": 21836 + }, + { + "epoch": 0.8, + "learning_rate": 1.202145969897828e-07, + "loss": 0.6911, + "step": 21837 + }, + { + "epoch": 0.8, + "learning_rate": 1.2021093492511077e-07, + "loss": 0.691, + "step": 21838 + }, + { + "epoch": 0.8, + "learning_rate": 1.202072728604387e-07, + "loss": 0.6773, + "step": 21839 + }, + { + "epoch": 0.8, + "learning_rate": 1.2020361079576665e-07, + "loss": 0.6881, + "step": 21840 + }, + { + "epoch": 0.8, + "learning_rate": 1.2019994873109457e-07, + "loss": 0.6974, + "step": 21841 + }, + { + "epoch": 0.8, + "learning_rate": 1.2019628666642253e-07, + "loss": 0.6878, + "step": 21842 + }, + { + "epoch": 0.8, + "learning_rate": 1.2019262460175045e-07, + "loss": 0.6918, + "step": 21843 + }, + { + "epoch": 0.8, + "learning_rate": 1.201889625370784e-07, + "loss": 0.6907, + "step": 21844 + }, + { + "epoch": 0.8, + "learning_rate": 1.2018530047240633e-07, + "loss": 0.6859, + "step": 21845 + }, + { + "epoch": 0.8, + "learning_rate": 1.2018163840773429e-07, + "loss": 0.6898, + "step": 21846 + }, + { + "epoch": 0.8, + "learning_rate": 1.201779763430622e-07, + "loss": 0.6853, + "step": 21847 + }, + { + "epoch": 0.8, + "learning_rate": 1.2017431427839017e-07, + "loss": 0.7032, + "step": 21848 + }, + { + "epoch": 0.8, + "learning_rate": 1.201706522137181e-07, + "loss": 0.7094, + "step": 21849 + }, + { + "epoch": 0.8, + "learning_rate": 1.2016699014904604e-07, + "loss": 0.6871, + "step": 21850 + }, + { + "epoch": 0.8, + "learning_rate": 1.2016332808437397e-07, + "loss": 0.685, + "step": 21851 + }, + { + "epoch": 0.8, + "learning_rate": 1.201596660197019e-07, + "loss": 0.6976, + "step": 21852 + }, + { + "epoch": 0.8, + "learning_rate": 1.2015600395502985e-07, + "loss": 0.6936, + "step": 21853 + }, + { + "epoch": 0.8, + "learning_rate": 1.2015234189035778e-07, + "loss": 0.7097, + "step": 21854 + }, + { + "epoch": 0.8, + "learning_rate": 1.201486798256857e-07, + "loss": 0.6911, + "step": 21855 + }, + { + "epoch": 0.8, + "learning_rate": 1.2014501776101366e-07, + "loss": 0.6849, + "step": 21856 + }, + { + "epoch": 0.8, + "learning_rate": 1.2014135569634158e-07, + "loss": 0.6839, + "step": 21857 + }, + { + "epoch": 0.8, + "learning_rate": 1.201376936316695e-07, + "loss": 0.6906, + "step": 21858 + }, + { + "epoch": 0.8, + "learning_rate": 1.2013403156699746e-07, + "loss": 0.6913, + "step": 21859 + }, + { + "epoch": 0.8, + "learning_rate": 1.201303695023254e-07, + "loss": 0.6752, + "step": 21860 + }, + { + "epoch": 0.8, + "learning_rate": 1.2012670743765334e-07, + "loss": 0.7027, + "step": 21861 + }, + { + "epoch": 0.8, + "learning_rate": 1.2012304537298127e-07, + "loss": 0.7019, + "step": 21862 + }, + { + "epoch": 0.8, + "learning_rate": 1.2011938330830922e-07, + "loss": 0.7009, + "step": 21863 + }, + { + "epoch": 0.8, + "learning_rate": 1.2011572124363715e-07, + "loss": 0.691, + "step": 21864 + }, + { + "epoch": 0.8, + "learning_rate": 1.201120591789651e-07, + "loss": 0.6751, + "step": 21865 + }, + { + "epoch": 0.8, + "learning_rate": 1.2010839711429303e-07, + "loss": 0.6929, + "step": 21866 + }, + { + "epoch": 0.8, + "learning_rate": 1.2010473504962098e-07, + "loss": 0.6988, + "step": 21867 + }, + { + "epoch": 0.8, + "learning_rate": 1.201010729849489e-07, + "loss": 0.6912, + "step": 21868 + }, + { + "epoch": 0.8, + "learning_rate": 1.2009741092027686e-07, + "loss": 0.7145, + "step": 21869 + }, + { + "epoch": 0.8, + "learning_rate": 1.200937488556048e-07, + "loss": 0.7005, + "step": 21870 + }, + { + "epoch": 0.8, + "learning_rate": 1.2009008679093274e-07, + "loss": 0.6946, + "step": 21871 + }, + { + "epoch": 0.8, + "learning_rate": 1.2008642472626067e-07, + "loss": 0.6925, + "step": 21872 + }, + { + "epoch": 0.8, + "learning_rate": 1.200827626615886e-07, + "loss": 0.6977, + "step": 21873 + }, + { + "epoch": 0.8, + "learning_rate": 1.2007910059691652e-07, + "loss": 0.6986, + "step": 21874 + }, + { + "epoch": 0.8, + "learning_rate": 1.2007543853224448e-07, + "loss": 0.6769, + "step": 21875 + }, + { + "epoch": 0.8, + "learning_rate": 1.200717764675724e-07, + "loss": 0.6747, + "step": 21876 + }, + { + "epoch": 0.8, + "learning_rate": 1.2006811440290036e-07, + "loss": 0.7025, + "step": 21877 + }, + { + "epoch": 0.8, + "learning_rate": 1.2006445233822828e-07, + "loss": 0.6939, + "step": 21878 + }, + { + "epoch": 0.8, + "learning_rate": 1.200607902735562e-07, + "loss": 0.7125, + "step": 21879 + }, + { + "epoch": 0.8, + "learning_rate": 1.2005712820888416e-07, + "loss": 0.6866, + "step": 21880 + }, + { + "epoch": 0.8, + "learning_rate": 1.200534661442121e-07, + "loss": 0.6865, + "step": 21881 + }, + { + "epoch": 0.8, + "learning_rate": 1.2004980407954004e-07, + "loss": 0.6997, + "step": 21882 + }, + { + "epoch": 0.8, + "learning_rate": 1.2004614201486797e-07, + "loss": 0.6978, + "step": 21883 + }, + { + "epoch": 0.8, + "learning_rate": 1.2004247995019592e-07, + "loss": 0.6855, + "step": 21884 + }, + { + "epoch": 0.8, + "learning_rate": 1.2003881788552385e-07, + "loss": 0.6878, + "step": 21885 + }, + { + "epoch": 0.8, + "learning_rate": 1.200351558208518e-07, + "loss": 0.6985, + "step": 21886 + }, + { + "epoch": 0.8, + "learning_rate": 1.2003149375617973e-07, + "loss": 0.7023, + "step": 21887 + }, + { + "epoch": 0.8, + "learning_rate": 1.2002783169150768e-07, + "loss": 0.6855, + "step": 21888 + }, + { + "epoch": 0.8, + "learning_rate": 1.200241696268356e-07, + "loss": 0.6865, + "step": 21889 + }, + { + "epoch": 0.8, + "learning_rate": 1.2002050756216356e-07, + "loss": 0.6911, + "step": 21890 + }, + { + "epoch": 0.8, + "learning_rate": 1.2001684549749146e-07, + "loss": 0.6981, + "step": 21891 + }, + { + "epoch": 0.8, + "learning_rate": 1.2001318343281942e-07, + "loss": 0.691, + "step": 21892 + }, + { + "epoch": 0.8, + "learning_rate": 1.2000952136814734e-07, + "loss": 0.6943, + "step": 21893 + }, + { + "epoch": 0.8, + "learning_rate": 1.200058593034753e-07, + "loss": 0.6911, + "step": 21894 + }, + { + "epoch": 0.8, + "learning_rate": 1.2000219723880322e-07, + "loss": 0.6925, + "step": 21895 + }, + { + "epoch": 0.8, + "learning_rate": 1.1999853517413118e-07, + "loss": 0.6863, + "step": 21896 + }, + { + "epoch": 0.8, + "learning_rate": 1.199948731094591e-07, + "loss": 0.6937, + "step": 21897 + }, + { + "epoch": 0.8, + "learning_rate": 1.1999121104478706e-07, + "loss": 0.691, + "step": 21898 + }, + { + "epoch": 0.8, + "learning_rate": 1.1998754898011498e-07, + "loss": 0.6935, + "step": 21899 + }, + { + "epoch": 0.8, + "learning_rate": 1.199838869154429e-07, + "loss": 0.6961, + "step": 21900 + }, + { + "epoch": 0.8, + "learning_rate": 1.1998022485077086e-07, + "loss": 0.6776, + "step": 21901 + }, + { + "epoch": 0.8, + "learning_rate": 1.199765627860988e-07, + "loss": 0.694, + "step": 21902 + }, + { + "epoch": 0.8, + "learning_rate": 1.1997290072142674e-07, + "loss": 0.6923, + "step": 21903 + }, + { + "epoch": 0.8, + "learning_rate": 1.1996923865675467e-07, + "loss": 0.702, + "step": 21904 + }, + { + "epoch": 0.8, + "learning_rate": 1.1996557659208262e-07, + "loss": 0.6922, + "step": 21905 + }, + { + "epoch": 0.8, + "learning_rate": 1.1996191452741055e-07, + "loss": 0.7043, + "step": 21906 + }, + { + "epoch": 0.8, + "learning_rate": 1.199582524627385e-07, + "loss": 0.701, + "step": 21907 + }, + { + "epoch": 0.8, + "learning_rate": 1.1995459039806643e-07, + "loss": 0.6951, + "step": 21908 + }, + { + "epoch": 0.8, + "learning_rate": 1.1995092833339436e-07, + "loss": 0.6877, + "step": 21909 + }, + { + "epoch": 0.8, + "learning_rate": 1.1994726626872228e-07, + "loss": 0.6857, + "step": 21910 + }, + { + "epoch": 0.8, + "learning_rate": 1.1994360420405024e-07, + "loss": 0.6861, + "step": 21911 + }, + { + "epoch": 0.8, + "learning_rate": 1.1993994213937816e-07, + "loss": 0.6923, + "step": 21912 + }, + { + "epoch": 0.8, + "learning_rate": 1.1993628007470611e-07, + "loss": 0.6929, + "step": 21913 + }, + { + "epoch": 0.8, + "learning_rate": 1.1993261801003404e-07, + "loss": 0.6609, + "step": 21914 + }, + { + "epoch": 0.8, + "learning_rate": 1.19928955945362e-07, + "loss": 0.6814, + "step": 21915 + }, + { + "epoch": 0.8, + "learning_rate": 1.1992529388068992e-07, + "loss": 0.6847, + "step": 21916 + }, + { + "epoch": 0.8, + "learning_rate": 1.1992163181601787e-07, + "loss": 0.6904, + "step": 21917 + }, + { + "epoch": 0.8, + "learning_rate": 1.199179697513458e-07, + "loss": 0.6956, + "step": 21918 + }, + { + "epoch": 0.8, + "learning_rate": 1.1991430768667375e-07, + "loss": 0.6967, + "step": 21919 + }, + { + "epoch": 0.8, + "learning_rate": 1.1991064562200168e-07, + "loss": 0.6913, + "step": 21920 + }, + { + "epoch": 0.8, + "learning_rate": 1.1990698355732963e-07, + "loss": 0.6954, + "step": 21921 + }, + { + "epoch": 0.8, + "learning_rate": 1.1990332149265756e-07, + "loss": 0.7076, + "step": 21922 + }, + { + "epoch": 0.8, + "learning_rate": 1.198996594279855e-07, + "loss": 0.6893, + "step": 21923 + }, + { + "epoch": 0.8, + "learning_rate": 1.1989599736331344e-07, + "loss": 0.699, + "step": 21924 + }, + { + "epoch": 0.8, + "learning_rate": 1.1989233529864137e-07, + "loss": 0.6969, + "step": 21925 + }, + { + "epoch": 0.8, + "learning_rate": 1.1988867323396932e-07, + "loss": 0.7126, + "step": 21926 + }, + { + "epoch": 0.8, + "learning_rate": 1.1988501116929725e-07, + "loss": 0.6967, + "step": 21927 + }, + { + "epoch": 0.8, + "learning_rate": 1.1988134910462517e-07, + "loss": 0.6993, + "step": 21928 + }, + { + "epoch": 0.8, + "learning_rate": 1.198776870399531e-07, + "loss": 0.6876, + "step": 21929 + }, + { + "epoch": 0.8, + "learning_rate": 1.1987402497528105e-07, + "loss": 0.6803, + "step": 21930 + }, + { + "epoch": 0.8, + "learning_rate": 1.1987036291060898e-07, + "loss": 0.7025, + "step": 21931 + }, + { + "epoch": 0.8, + "learning_rate": 1.1986670084593693e-07, + "loss": 0.6925, + "step": 21932 + }, + { + "epoch": 0.8, + "learning_rate": 1.1986303878126486e-07, + "loss": 0.6937, + "step": 21933 + }, + { + "epoch": 0.8, + "learning_rate": 1.1985937671659281e-07, + "loss": 0.6919, + "step": 21934 + }, + { + "epoch": 0.8, + "learning_rate": 1.1985571465192074e-07, + "loss": 0.6948, + "step": 21935 + }, + { + "epoch": 0.8, + "learning_rate": 1.198520525872487e-07, + "loss": 0.7038, + "step": 21936 + }, + { + "epoch": 0.8, + "learning_rate": 1.1984839052257662e-07, + "loss": 0.6896, + "step": 21937 + }, + { + "epoch": 0.8, + "learning_rate": 1.1984472845790457e-07, + "loss": 0.6896, + "step": 21938 + }, + { + "epoch": 0.8, + "learning_rate": 1.198410663932325e-07, + "loss": 0.6901, + "step": 21939 + }, + { + "epoch": 0.8, + "learning_rate": 1.1983740432856045e-07, + "loss": 0.7017, + "step": 21940 + }, + { + "epoch": 0.8, + "learning_rate": 1.1983374226388838e-07, + "loss": 0.6888, + "step": 21941 + }, + { + "epoch": 0.8, + "learning_rate": 1.1983008019921633e-07, + "loss": 0.6835, + "step": 21942 + }, + { + "epoch": 0.8, + "learning_rate": 1.1982641813454426e-07, + "loss": 0.693, + "step": 21943 + }, + { + "epoch": 0.8, + "learning_rate": 1.1982275606987219e-07, + "loss": 0.7092, + "step": 21944 + }, + { + "epoch": 0.8, + "learning_rate": 1.198190940052001e-07, + "loss": 0.7072, + "step": 21945 + }, + { + "epoch": 0.8, + "learning_rate": 1.1981543194052807e-07, + "loss": 0.6955, + "step": 21946 + }, + { + "epoch": 0.8, + "learning_rate": 1.19811769875856e-07, + "loss": 0.6933, + "step": 21947 + }, + { + "epoch": 0.8, + "learning_rate": 1.1980810781118395e-07, + "loss": 0.7072, + "step": 21948 + }, + { + "epoch": 0.8, + "learning_rate": 1.1980444574651187e-07, + "loss": 0.6957, + "step": 21949 + }, + { + "epoch": 0.8, + "learning_rate": 1.198007836818398e-07, + "loss": 0.7014, + "step": 21950 + }, + { + "epoch": 0.8, + "learning_rate": 1.1979712161716775e-07, + "loss": 0.6847, + "step": 21951 + }, + { + "epoch": 0.8, + "learning_rate": 1.1979345955249568e-07, + "loss": 0.6885, + "step": 21952 + }, + { + "epoch": 0.8, + "learning_rate": 1.1978979748782363e-07, + "loss": 0.6841, + "step": 21953 + }, + { + "epoch": 0.8, + "learning_rate": 1.1978613542315156e-07, + "loss": 0.7008, + "step": 21954 + }, + { + "epoch": 0.8, + "learning_rate": 1.197824733584795e-07, + "loss": 0.6946, + "step": 21955 + }, + { + "epoch": 0.8, + "learning_rate": 1.1977881129380744e-07, + "loss": 0.6963, + "step": 21956 + }, + { + "epoch": 0.8, + "learning_rate": 1.197751492291354e-07, + "loss": 0.6937, + "step": 21957 + }, + { + "epoch": 0.8, + "learning_rate": 1.1977148716446332e-07, + "loss": 0.7081, + "step": 21958 + }, + { + "epoch": 0.8, + "learning_rate": 1.1976782509979127e-07, + "loss": 0.6912, + "step": 21959 + }, + { + "epoch": 0.8, + "learning_rate": 1.197641630351192e-07, + "loss": 0.6983, + "step": 21960 + }, + { + "epoch": 0.8, + "learning_rate": 1.1976050097044715e-07, + "loss": 0.7001, + "step": 21961 + }, + { + "epoch": 0.8, + "learning_rate": 1.1975683890577508e-07, + "loss": 0.708, + "step": 21962 + }, + { + "epoch": 0.8, + "learning_rate": 1.19753176841103e-07, + "loss": 0.6952, + "step": 21963 + }, + { + "epoch": 0.8, + "learning_rate": 1.1974951477643093e-07, + "loss": 0.6828, + "step": 21964 + }, + { + "epoch": 0.8, + "learning_rate": 1.1974585271175889e-07, + "loss": 0.6843, + "step": 21965 + }, + { + "epoch": 0.8, + "learning_rate": 1.197421906470868e-07, + "loss": 0.6862, + "step": 21966 + }, + { + "epoch": 0.8, + "learning_rate": 1.1973852858241477e-07, + "loss": 0.6899, + "step": 21967 + }, + { + "epoch": 0.8, + "learning_rate": 1.197348665177427e-07, + "loss": 0.6916, + "step": 21968 + }, + { + "epoch": 0.8, + "learning_rate": 1.1973120445307065e-07, + "loss": 0.6907, + "step": 21969 + }, + { + "epoch": 0.8, + "learning_rate": 1.1972754238839857e-07, + "loss": 0.6972, + "step": 21970 + }, + { + "epoch": 0.8, + "learning_rate": 1.197238803237265e-07, + "loss": 0.7003, + "step": 21971 + }, + { + "epoch": 0.8, + "learning_rate": 1.1972021825905445e-07, + "loss": 0.6981, + "step": 21972 + }, + { + "epoch": 0.8, + "learning_rate": 1.1971655619438238e-07, + "loss": 0.7038, + "step": 21973 + }, + { + "epoch": 0.8, + "learning_rate": 1.1971289412971033e-07, + "loss": 0.6844, + "step": 21974 + }, + { + "epoch": 0.8, + "learning_rate": 1.1970923206503826e-07, + "loss": 0.6891, + "step": 21975 + }, + { + "epoch": 0.8, + "learning_rate": 1.197055700003662e-07, + "loss": 0.6973, + "step": 21976 + }, + { + "epoch": 0.8, + "learning_rate": 1.1970190793569414e-07, + "loss": 0.6843, + "step": 21977 + }, + { + "epoch": 0.8, + "learning_rate": 1.196982458710221e-07, + "loss": 0.6932, + "step": 21978 + }, + { + "epoch": 0.8, + "learning_rate": 1.1969458380635002e-07, + "loss": 0.7011, + "step": 21979 + }, + { + "epoch": 0.8, + "learning_rate": 1.1969092174167794e-07, + "loss": 0.691, + "step": 21980 + }, + { + "epoch": 0.8, + "learning_rate": 1.1968725967700587e-07, + "loss": 0.6949, + "step": 21981 + }, + { + "epoch": 0.8, + "learning_rate": 1.1968359761233382e-07, + "loss": 0.6919, + "step": 21982 + }, + { + "epoch": 0.8, + "learning_rate": 1.1967993554766175e-07, + "loss": 0.6929, + "step": 21983 + }, + { + "epoch": 0.8, + "learning_rate": 1.196762734829897e-07, + "loss": 0.6837, + "step": 21984 + }, + { + "epoch": 0.8, + "learning_rate": 1.1967261141831763e-07, + "loss": 0.7013, + "step": 21985 + }, + { + "epoch": 0.8, + "learning_rate": 1.1966894935364558e-07, + "loss": 0.7024, + "step": 21986 + }, + { + "epoch": 0.8, + "learning_rate": 1.196652872889735e-07, + "loss": 0.6807, + "step": 21987 + }, + { + "epoch": 0.8, + "learning_rate": 1.1966162522430146e-07, + "loss": 0.6984, + "step": 21988 + }, + { + "epoch": 0.8, + "learning_rate": 1.196579631596294e-07, + "loss": 0.6964, + "step": 21989 + }, + { + "epoch": 0.8, + "learning_rate": 1.1965430109495734e-07, + "loss": 0.6989, + "step": 21990 + }, + { + "epoch": 0.8, + "learning_rate": 1.1965063903028527e-07, + "loss": 0.6965, + "step": 21991 + }, + { + "epoch": 0.8, + "learning_rate": 1.1964697696561322e-07, + "loss": 0.6906, + "step": 21992 + }, + { + "epoch": 0.8, + "learning_rate": 1.1964331490094115e-07, + "loss": 0.701, + "step": 21993 + }, + { + "epoch": 0.8, + "learning_rate": 1.1963965283626908e-07, + "loss": 0.7038, + "step": 21994 + }, + { + "epoch": 0.8, + "learning_rate": 1.1963599077159703e-07, + "loss": 0.6843, + "step": 21995 + }, + { + "epoch": 0.8, + "learning_rate": 1.1963232870692496e-07, + "loss": 0.7038, + "step": 21996 + }, + { + "epoch": 0.8, + "learning_rate": 1.196286666422529e-07, + "loss": 0.6874, + "step": 21997 + }, + { + "epoch": 0.8, + "learning_rate": 1.1962500457758084e-07, + "loss": 0.6975, + "step": 21998 + }, + { + "epoch": 0.8, + "learning_rate": 1.1962134251290876e-07, + "loss": 0.7206, + "step": 21999 + }, + { + "epoch": 0.8, + "learning_rate": 1.196176804482367e-07, + "loss": 0.7086, + "step": 22000 + }, + { + "epoch": 0.8, + "learning_rate": 1.1961401838356464e-07, + "loss": 0.6905, + "step": 22001 + }, + { + "epoch": 0.8, + "learning_rate": 1.1961035631889257e-07, + "loss": 0.6979, + "step": 22002 + }, + { + "epoch": 0.81, + "learning_rate": 1.1960669425422052e-07, + "loss": 0.6859, + "step": 22003 + }, + { + "epoch": 0.81, + "learning_rate": 1.1960303218954845e-07, + "loss": 0.6881, + "step": 22004 + }, + { + "epoch": 0.81, + "learning_rate": 1.195993701248764e-07, + "loss": 0.7041, + "step": 22005 + }, + { + "epoch": 0.81, + "learning_rate": 1.1959570806020433e-07, + "loss": 0.6802, + "step": 22006 + }, + { + "epoch": 0.81, + "learning_rate": 1.1959204599553228e-07, + "loss": 0.6976, + "step": 22007 + }, + { + "epoch": 0.81, + "learning_rate": 1.195883839308602e-07, + "loss": 0.692, + "step": 22008 + }, + { + "epoch": 0.81, + "learning_rate": 1.1958472186618816e-07, + "loss": 0.6882, + "step": 22009 + }, + { + "epoch": 0.81, + "learning_rate": 1.195810598015161e-07, + "loss": 0.6856, + "step": 22010 + }, + { + "epoch": 0.81, + "learning_rate": 1.1957739773684404e-07, + "loss": 0.6991, + "step": 22011 + }, + { + "epoch": 0.81, + "learning_rate": 1.1957373567217197e-07, + "loss": 0.6929, + "step": 22012 + }, + { + "epoch": 0.81, + "learning_rate": 1.1957007360749992e-07, + "loss": 0.6975, + "step": 22013 + }, + { + "epoch": 0.81, + "learning_rate": 1.1956641154282785e-07, + "loss": 0.696, + "step": 22014 + }, + { + "epoch": 0.81, + "learning_rate": 1.1956274947815578e-07, + "loss": 0.6958, + "step": 22015 + }, + { + "epoch": 0.81, + "learning_rate": 1.195590874134837e-07, + "loss": 0.6999, + "step": 22016 + }, + { + "epoch": 0.81, + "learning_rate": 1.1955542534881166e-07, + "loss": 0.6967, + "step": 22017 + }, + { + "epoch": 0.81, + "learning_rate": 1.1955176328413958e-07, + "loss": 0.6869, + "step": 22018 + }, + { + "epoch": 0.81, + "learning_rate": 1.1954810121946754e-07, + "loss": 0.7171, + "step": 22019 + }, + { + "epoch": 0.81, + "learning_rate": 1.1954443915479546e-07, + "loss": 0.6829, + "step": 22020 + }, + { + "epoch": 0.81, + "learning_rate": 1.195407770901234e-07, + "loss": 0.6922, + "step": 22021 + }, + { + "epoch": 0.81, + "learning_rate": 1.1953711502545134e-07, + "loss": 0.6923, + "step": 22022 + }, + { + "epoch": 0.81, + "learning_rate": 1.1953345296077927e-07, + "loss": 0.7021, + "step": 22023 + }, + { + "epoch": 0.81, + "learning_rate": 1.1952979089610722e-07, + "loss": 0.6945, + "step": 22024 + }, + { + "epoch": 0.81, + "learning_rate": 1.1952612883143515e-07, + "loss": 0.7021, + "step": 22025 + }, + { + "epoch": 0.81, + "learning_rate": 1.195224667667631e-07, + "loss": 0.6963, + "step": 22026 + }, + { + "epoch": 0.81, + "learning_rate": 1.1951880470209103e-07, + "loss": 0.6882, + "step": 22027 + }, + { + "epoch": 0.81, + "learning_rate": 1.1951514263741898e-07, + "loss": 0.6904, + "step": 22028 + }, + { + "epoch": 0.81, + "learning_rate": 1.195114805727469e-07, + "loss": 0.6925, + "step": 22029 + }, + { + "epoch": 0.81, + "learning_rate": 1.1950781850807486e-07, + "loss": 0.6831, + "step": 22030 + }, + { + "epoch": 0.81, + "learning_rate": 1.195041564434028e-07, + "loss": 0.6939, + "step": 22031 + }, + { + "epoch": 0.81, + "learning_rate": 1.1950049437873074e-07, + "loss": 0.7073, + "step": 22032 + }, + { + "epoch": 0.81, + "learning_rate": 1.1949683231405867e-07, + "loss": 0.6935, + "step": 22033 + }, + { + "epoch": 0.81, + "learning_rate": 1.194931702493866e-07, + "loss": 0.6891, + "step": 22034 + }, + { + "epoch": 0.81, + "learning_rate": 1.1948950818471452e-07, + "loss": 0.6844, + "step": 22035 + }, + { + "epoch": 0.81, + "learning_rate": 1.1948584612004247e-07, + "loss": 0.6971, + "step": 22036 + }, + { + "epoch": 0.81, + "learning_rate": 1.194821840553704e-07, + "loss": 0.6856, + "step": 22037 + }, + { + "epoch": 0.81, + "learning_rate": 1.1947852199069835e-07, + "loss": 0.6876, + "step": 22038 + }, + { + "epoch": 0.81, + "learning_rate": 1.1947485992602628e-07, + "loss": 0.6828, + "step": 22039 + }, + { + "epoch": 0.81, + "learning_rate": 1.1947119786135423e-07, + "loss": 0.688, + "step": 22040 + }, + { + "epoch": 0.81, + "learning_rate": 1.1946753579668216e-07, + "loss": 0.6869, + "step": 22041 + }, + { + "epoch": 0.81, + "learning_rate": 1.194638737320101e-07, + "loss": 0.702, + "step": 22042 + }, + { + "epoch": 0.81, + "learning_rate": 1.1946021166733804e-07, + "loss": 0.6879, + "step": 22043 + }, + { + "epoch": 0.81, + "learning_rate": 1.1945654960266597e-07, + "loss": 0.6966, + "step": 22044 + }, + { + "epoch": 0.81, + "learning_rate": 1.1945288753799392e-07, + "loss": 0.6915, + "step": 22045 + }, + { + "epoch": 0.81, + "learning_rate": 1.1944922547332185e-07, + "loss": 0.7026, + "step": 22046 + }, + { + "epoch": 0.81, + "learning_rate": 1.194455634086498e-07, + "loss": 0.6853, + "step": 22047 + }, + { + "epoch": 0.81, + "learning_rate": 1.1944190134397773e-07, + "loss": 0.6922, + "step": 22048 + }, + { + "epoch": 0.81, + "learning_rate": 1.1943823927930568e-07, + "loss": 0.6951, + "step": 22049 + }, + { + "epoch": 0.81, + "learning_rate": 1.194345772146336e-07, + "loss": 0.6931, + "step": 22050 + }, + { + "epoch": 0.81, + "learning_rate": 1.1943091514996156e-07, + "loss": 0.6904, + "step": 22051 + }, + { + "epoch": 0.81, + "learning_rate": 1.1942725308528946e-07, + "loss": 0.6888, + "step": 22052 + }, + { + "epoch": 0.81, + "learning_rate": 1.1942359102061741e-07, + "loss": 0.6924, + "step": 22053 + }, + { + "epoch": 0.81, + "learning_rate": 1.1941992895594534e-07, + "loss": 0.6887, + "step": 22054 + }, + { + "epoch": 0.81, + "learning_rate": 1.194162668912733e-07, + "loss": 0.6962, + "step": 22055 + }, + { + "epoch": 0.81, + "learning_rate": 1.1941260482660122e-07, + "loss": 0.6853, + "step": 22056 + }, + { + "epoch": 0.81, + "learning_rate": 1.1940894276192917e-07, + "loss": 0.6896, + "step": 22057 + }, + { + "epoch": 0.81, + "learning_rate": 1.194052806972571e-07, + "loss": 0.6818, + "step": 22058 + }, + { + "epoch": 0.81, + "learning_rate": 1.1940161863258505e-07, + "loss": 0.6902, + "step": 22059 + }, + { + "epoch": 0.81, + "learning_rate": 1.1939795656791298e-07, + "loss": 0.7057, + "step": 22060 + }, + { + "epoch": 0.81, + "learning_rate": 1.1939429450324093e-07, + "loss": 0.6985, + "step": 22061 + }, + { + "epoch": 0.81, + "learning_rate": 1.1939063243856886e-07, + "loss": 0.6928, + "step": 22062 + }, + { + "epoch": 0.81, + "learning_rate": 1.193869703738968e-07, + "loss": 0.6938, + "step": 22063 + }, + { + "epoch": 0.81, + "learning_rate": 1.1938330830922474e-07, + "loss": 0.7021, + "step": 22064 + }, + { + "epoch": 0.81, + "learning_rate": 1.1937964624455267e-07, + "loss": 0.6829, + "step": 22065 + }, + { + "epoch": 0.81, + "learning_rate": 1.1937598417988062e-07, + "loss": 0.7009, + "step": 22066 + }, + { + "epoch": 0.81, + "learning_rate": 1.1937232211520855e-07, + "loss": 0.6986, + "step": 22067 + }, + { + "epoch": 0.81, + "learning_rate": 1.193686600505365e-07, + "loss": 0.6866, + "step": 22068 + }, + { + "epoch": 0.81, + "learning_rate": 1.1936499798586443e-07, + "loss": 0.6875, + "step": 22069 + }, + { + "epoch": 0.81, + "learning_rate": 1.1936133592119235e-07, + "loss": 0.6976, + "step": 22070 + }, + { + "epoch": 0.81, + "learning_rate": 1.1935767385652028e-07, + "loss": 0.6926, + "step": 22071 + }, + { + "epoch": 0.81, + "learning_rate": 1.1935401179184823e-07, + "loss": 0.6932, + "step": 22072 + }, + { + "epoch": 0.81, + "learning_rate": 1.1935034972717616e-07, + "loss": 0.7003, + "step": 22073 + }, + { + "epoch": 0.81, + "learning_rate": 1.193466876625041e-07, + "loss": 0.6938, + "step": 22074 + }, + { + "epoch": 0.81, + "learning_rate": 1.1934302559783204e-07, + "loss": 0.6919, + "step": 22075 + }, + { + "epoch": 0.81, + "learning_rate": 1.1933936353316e-07, + "loss": 0.6793, + "step": 22076 + }, + { + "epoch": 0.81, + "learning_rate": 1.1933570146848792e-07, + "loss": 0.7063, + "step": 22077 + }, + { + "epoch": 0.81, + "learning_rate": 1.1933203940381587e-07, + "loss": 0.6835, + "step": 22078 + }, + { + "epoch": 0.81, + "learning_rate": 1.193283773391438e-07, + "loss": 0.6893, + "step": 22079 + }, + { + "epoch": 0.81, + "learning_rate": 1.1932471527447175e-07, + "loss": 0.7002, + "step": 22080 + }, + { + "epoch": 0.81, + "learning_rate": 1.1932105320979968e-07, + "loss": 0.697, + "step": 22081 + }, + { + "epoch": 0.81, + "learning_rate": 1.1931739114512763e-07, + "loss": 0.6903, + "step": 22082 + }, + { + "epoch": 0.81, + "learning_rate": 1.1931372908045556e-07, + "loss": 0.6995, + "step": 22083 + }, + { + "epoch": 0.81, + "learning_rate": 1.193100670157835e-07, + "loss": 0.6864, + "step": 22084 + }, + { + "epoch": 0.81, + "learning_rate": 1.1930640495111144e-07, + "loss": 0.6793, + "step": 22085 + }, + { + "epoch": 0.81, + "learning_rate": 1.1930274288643937e-07, + "loss": 0.6814, + "step": 22086 + }, + { + "epoch": 0.81, + "learning_rate": 1.1929908082176732e-07, + "loss": 0.6905, + "step": 22087 + }, + { + "epoch": 0.81, + "learning_rate": 1.1929541875709525e-07, + "loss": 0.7047, + "step": 22088 + }, + { + "epoch": 0.81, + "learning_rate": 1.1929175669242317e-07, + "loss": 0.7084, + "step": 22089 + }, + { + "epoch": 0.81, + "learning_rate": 1.1928809462775112e-07, + "loss": 0.6899, + "step": 22090 + }, + { + "epoch": 0.81, + "learning_rate": 1.1928443256307905e-07, + "loss": 0.6943, + "step": 22091 + }, + { + "epoch": 0.81, + "learning_rate": 1.1928077049840698e-07, + "loss": 0.6987, + "step": 22092 + }, + { + "epoch": 0.81, + "learning_rate": 1.1927710843373493e-07, + "loss": 0.6892, + "step": 22093 + }, + { + "epoch": 0.81, + "learning_rate": 1.1927344636906286e-07, + "loss": 0.6975, + "step": 22094 + }, + { + "epoch": 0.81, + "learning_rate": 1.192697843043908e-07, + "loss": 0.7019, + "step": 22095 + }, + { + "epoch": 0.81, + "learning_rate": 1.1926612223971874e-07, + "loss": 0.7024, + "step": 22096 + }, + { + "epoch": 0.81, + "learning_rate": 1.192624601750467e-07, + "loss": 0.7035, + "step": 22097 + }, + { + "epoch": 0.81, + "learning_rate": 1.1925879811037462e-07, + "loss": 0.682, + "step": 22098 + }, + { + "epoch": 0.81, + "learning_rate": 1.1925513604570257e-07, + "loss": 0.7014, + "step": 22099 + }, + { + "epoch": 0.81, + "learning_rate": 1.192514739810305e-07, + "loss": 0.6923, + "step": 22100 + }, + { + "epoch": 0.81, + "learning_rate": 1.1924781191635845e-07, + "loss": 0.7035, + "step": 22101 + }, + { + "epoch": 0.81, + "learning_rate": 1.1924414985168638e-07, + "loss": 0.6958, + "step": 22102 + }, + { + "epoch": 0.81, + "learning_rate": 1.1924048778701433e-07, + "loss": 0.698, + "step": 22103 + }, + { + "epoch": 0.81, + "learning_rate": 1.1923682572234226e-07, + "loss": 0.7056, + "step": 22104 + }, + { + "epoch": 0.81, + "learning_rate": 1.192331636576702e-07, + "loss": 0.6898, + "step": 22105 + }, + { + "epoch": 0.81, + "learning_rate": 1.192295015929981e-07, + "loss": 0.7017, + "step": 22106 + }, + { + "epoch": 0.81, + "learning_rate": 1.1922583952832606e-07, + "loss": 0.7046, + "step": 22107 + }, + { + "epoch": 0.81, + "learning_rate": 1.19222177463654e-07, + "loss": 0.6914, + "step": 22108 + }, + { + "epoch": 0.81, + "learning_rate": 1.1921851539898194e-07, + "loss": 0.6947, + "step": 22109 + }, + { + "epoch": 0.81, + "learning_rate": 1.1921485333430987e-07, + "loss": 0.6824, + "step": 22110 + }, + { + "epoch": 0.81, + "learning_rate": 1.1921119126963782e-07, + "loss": 0.6852, + "step": 22111 + }, + { + "epoch": 0.81, + "learning_rate": 1.1920752920496575e-07, + "loss": 0.7016, + "step": 22112 + }, + { + "epoch": 0.81, + "learning_rate": 1.1920386714029369e-07, + "loss": 0.6941, + "step": 22113 + }, + { + "epoch": 0.81, + "learning_rate": 1.1920020507562163e-07, + "loss": 0.687, + "step": 22114 + }, + { + "epoch": 0.81, + "learning_rate": 1.1919654301094957e-07, + "loss": 0.6981, + "step": 22115 + }, + { + "epoch": 0.81, + "learning_rate": 1.1919288094627751e-07, + "loss": 0.6798, + "step": 22116 + }, + { + "epoch": 0.81, + "learning_rate": 1.1918921888160545e-07, + "loss": 0.6921, + "step": 22117 + }, + { + "epoch": 0.81, + "learning_rate": 1.1918555681693339e-07, + "loss": 0.6885, + "step": 22118 + }, + { + "epoch": 0.81, + "learning_rate": 1.1918189475226133e-07, + "loss": 0.7035, + "step": 22119 + }, + { + "epoch": 0.81, + "learning_rate": 1.1917823268758927e-07, + "loss": 0.6815, + "step": 22120 + }, + { + "epoch": 0.81, + "learning_rate": 1.1917457062291721e-07, + "loss": 0.7017, + "step": 22121 + }, + { + "epoch": 0.81, + "learning_rate": 1.1917090855824514e-07, + "loss": 0.7043, + "step": 22122 + }, + { + "epoch": 0.81, + "learning_rate": 1.1916724649357308e-07, + "loss": 0.6861, + "step": 22123 + }, + { + "epoch": 0.81, + "learning_rate": 1.19163584428901e-07, + "loss": 0.6832, + "step": 22124 + }, + { + "epoch": 0.81, + "learning_rate": 1.1915992236422894e-07, + "loss": 0.6904, + "step": 22125 + }, + { + "epoch": 0.81, + "learning_rate": 1.1915626029955688e-07, + "loss": 0.6848, + "step": 22126 + }, + { + "epoch": 0.81, + "learning_rate": 1.1915259823488481e-07, + "loss": 0.6819, + "step": 22127 + }, + { + "epoch": 0.81, + "learning_rate": 1.1914893617021275e-07, + "loss": 0.6921, + "step": 22128 + }, + { + "epoch": 0.81, + "learning_rate": 1.1914527410554069e-07, + "loss": 0.6921, + "step": 22129 + }, + { + "epoch": 0.81, + "learning_rate": 1.1914161204086863e-07, + "loss": 0.6983, + "step": 22130 + }, + { + "epoch": 0.81, + "learning_rate": 1.1913794997619657e-07, + "loss": 0.6937, + "step": 22131 + }, + { + "epoch": 0.81, + "learning_rate": 1.1913428791152451e-07, + "loss": 0.7153, + "step": 22132 + }, + { + "epoch": 0.81, + "learning_rate": 1.1913062584685245e-07, + "loss": 0.6849, + "step": 22133 + }, + { + "epoch": 0.81, + "learning_rate": 1.1912696378218039e-07, + "loss": 0.6887, + "step": 22134 + }, + { + "epoch": 0.81, + "learning_rate": 1.1912330171750833e-07, + "loss": 0.6934, + "step": 22135 + }, + { + "epoch": 0.81, + "learning_rate": 1.1911963965283627e-07, + "loss": 0.698, + "step": 22136 + }, + { + "epoch": 0.81, + "learning_rate": 1.1911597758816421e-07, + "loss": 0.6768, + "step": 22137 + }, + { + "epoch": 0.81, + "learning_rate": 1.1911231552349215e-07, + "loss": 0.701, + "step": 22138 + }, + { + "epoch": 0.81, + "learning_rate": 1.1910865345882009e-07, + "loss": 0.6976, + "step": 22139 + }, + { + "epoch": 0.81, + "learning_rate": 1.1910499139414803e-07, + "loss": 0.6882, + "step": 22140 + }, + { + "epoch": 0.81, + "learning_rate": 1.1910132932947597e-07, + "loss": 0.6949, + "step": 22141 + }, + { + "epoch": 0.81, + "learning_rate": 1.1909766726480388e-07, + "loss": 0.6906, + "step": 22142 + }, + { + "epoch": 0.81, + "learning_rate": 1.1909400520013182e-07, + "loss": 0.6949, + "step": 22143 + }, + { + "epoch": 0.81, + "learning_rate": 1.1909034313545976e-07, + "loss": 0.7051, + "step": 22144 + }, + { + "epoch": 0.81, + "learning_rate": 1.190866810707877e-07, + "loss": 0.7129, + "step": 22145 + }, + { + "epoch": 0.81, + "learning_rate": 1.1908301900611564e-07, + "loss": 0.6859, + "step": 22146 + }, + { + "epoch": 0.81, + "learning_rate": 1.1907935694144358e-07, + "loss": 0.6957, + "step": 22147 + }, + { + "epoch": 0.81, + "learning_rate": 1.1907569487677152e-07, + "loss": 0.7037, + "step": 22148 + }, + { + "epoch": 0.81, + "learning_rate": 1.1907203281209945e-07, + "loss": 0.6958, + "step": 22149 + }, + { + "epoch": 0.81, + "learning_rate": 1.1906837074742739e-07, + "loss": 0.6958, + "step": 22150 + }, + { + "epoch": 0.81, + "learning_rate": 1.1906470868275533e-07, + "loss": 0.6846, + "step": 22151 + }, + { + "epoch": 0.81, + "learning_rate": 1.1906104661808327e-07, + "loss": 0.7022, + "step": 22152 + }, + { + "epoch": 0.81, + "learning_rate": 1.1905738455341121e-07, + "loss": 0.6914, + "step": 22153 + }, + { + "epoch": 0.81, + "learning_rate": 1.1905372248873915e-07, + "loss": 0.6958, + "step": 22154 + }, + { + "epoch": 0.81, + "learning_rate": 1.1905006042406709e-07, + "loss": 0.6871, + "step": 22155 + }, + { + "epoch": 0.81, + "learning_rate": 1.1904639835939503e-07, + "loss": 0.6874, + "step": 22156 + }, + { + "epoch": 0.81, + "learning_rate": 1.1904273629472297e-07, + "loss": 0.6876, + "step": 22157 + }, + { + "epoch": 0.81, + "learning_rate": 1.1903907423005091e-07, + "loss": 0.6895, + "step": 22158 + }, + { + "epoch": 0.81, + "learning_rate": 1.1903541216537885e-07, + "loss": 0.699, + "step": 22159 + }, + { + "epoch": 0.81, + "learning_rate": 1.1903175010070676e-07, + "loss": 0.6728, + "step": 22160 + }, + { + "epoch": 0.81, + "learning_rate": 1.190280880360347e-07, + "loss": 0.6854, + "step": 22161 + }, + { + "epoch": 0.81, + "learning_rate": 1.1902442597136264e-07, + "loss": 0.7012, + "step": 22162 + }, + { + "epoch": 0.81, + "learning_rate": 1.1902076390669058e-07, + "loss": 0.6908, + "step": 22163 + }, + { + "epoch": 0.81, + "learning_rate": 1.1901710184201852e-07, + "loss": 0.6906, + "step": 22164 + }, + { + "epoch": 0.81, + "learning_rate": 1.1901343977734646e-07, + "loss": 0.6941, + "step": 22165 + }, + { + "epoch": 0.81, + "learning_rate": 1.190097777126744e-07, + "loss": 0.7039, + "step": 22166 + }, + { + "epoch": 0.81, + "learning_rate": 1.1900611564800234e-07, + "loss": 0.693, + "step": 22167 + }, + { + "epoch": 0.81, + "learning_rate": 1.1900245358333028e-07, + "loss": 0.6871, + "step": 22168 + }, + { + "epoch": 0.81, + "learning_rate": 1.1899879151865822e-07, + "loss": 0.6927, + "step": 22169 + }, + { + "epoch": 0.81, + "learning_rate": 1.1899512945398616e-07, + "loss": 0.6978, + "step": 22170 + }, + { + "epoch": 0.81, + "learning_rate": 1.1899146738931409e-07, + "loss": 0.6932, + "step": 22171 + }, + { + "epoch": 0.81, + "learning_rate": 1.1898780532464203e-07, + "loss": 0.7062, + "step": 22172 + }, + { + "epoch": 0.81, + "learning_rate": 1.1898414325996997e-07, + "loss": 0.6885, + "step": 22173 + }, + { + "epoch": 0.81, + "learning_rate": 1.1898048119529791e-07, + "loss": 0.7144, + "step": 22174 + }, + { + "epoch": 0.81, + "learning_rate": 1.1897681913062585e-07, + "loss": 0.6999, + "step": 22175 + }, + { + "epoch": 0.81, + "learning_rate": 1.1897315706595379e-07, + "loss": 0.6921, + "step": 22176 + }, + { + "epoch": 0.81, + "learning_rate": 1.1896949500128173e-07, + "loss": 0.7005, + "step": 22177 + }, + { + "epoch": 0.81, + "learning_rate": 1.1896583293660964e-07, + "loss": 0.7004, + "step": 22178 + }, + { + "epoch": 0.81, + "learning_rate": 1.1896217087193758e-07, + "loss": 0.6812, + "step": 22179 + }, + { + "epoch": 0.81, + "learning_rate": 1.1895850880726552e-07, + "loss": 0.6965, + "step": 22180 + }, + { + "epoch": 0.81, + "learning_rate": 1.1895484674259346e-07, + "loss": 0.7028, + "step": 22181 + }, + { + "epoch": 0.81, + "learning_rate": 1.189511846779214e-07, + "loss": 0.6863, + "step": 22182 + }, + { + "epoch": 0.81, + "learning_rate": 1.1894752261324934e-07, + "loss": 0.6995, + "step": 22183 + }, + { + "epoch": 0.81, + "learning_rate": 1.1894386054857728e-07, + "loss": 0.6944, + "step": 22184 + }, + { + "epoch": 0.81, + "learning_rate": 1.1894019848390522e-07, + "loss": 0.6966, + "step": 22185 + }, + { + "epoch": 0.81, + "learning_rate": 1.1893653641923316e-07, + "loss": 0.7006, + "step": 22186 + }, + { + "epoch": 0.81, + "learning_rate": 1.189328743545611e-07, + "loss": 0.7024, + "step": 22187 + }, + { + "epoch": 0.81, + "learning_rate": 1.1892921228988904e-07, + "loss": 0.6881, + "step": 22188 + }, + { + "epoch": 0.81, + "learning_rate": 1.1892555022521698e-07, + "loss": 0.7003, + "step": 22189 + }, + { + "epoch": 0.81, + "learning_rate": 1.1892188816054492e-07, + "loss": 0.6992, + "step": 22190 + }, + { + "epoch": 0.81, + "learning_rate": 1.1891822609587286e-07, + "loss": 0.6962, + "step": 22191 + }, + { + "epoch": 0.81, + "learning_rate": 1.1891456403120079e-07, + "loss": 0.7015, + "step": 22192 + }, + { + "epoch": 0.81, + "learning_rate": 1.1891090196652873e-07, + "loss": 0.7025, + "step": 22193 + }, + { + "epoch": 0.81, + "learning_rate": 1.1890723990185667e-07, + "loss": 0.6944, + "step": 22194 + }, + { + "epoch": 0.81, + "learning_rate": 1.189035778371846e-07, + "loss": 0.7101, + "step": 22195 + }, + { + "epoch": 0.81, + "learning_rate": 1.1889991577251253e-07, + "loss": 0.7024, + "step": 22196 + }, + { + "epoch": 0.81, + "learning_rate": 1.1889625370784047e-07, + "loss": 0.693, + "step": 22197 + }, + { + "epoch": 0.81, + "learning_rate": 1.188925916431684e-07, + "loss": 0.6875, + "step": 22198 + }, + { + "epoch": 0.81, + "learning_rate": 1.1888892957849634e-07, + "loss": 0.6977, + "step": 22199 + }, + { + "epoch": 0.81, + "learning_rate": 1.1888526751382428e-07, + "loss": 0.6968, + "step": 22200 + }, + { + "epoch": 0.81, + "learning_rate": 1.1888160544915222e-07, + "loss": 0.6844, + "step": 22201 + }, + { + "epoch": 0.81, + "learning_rate": 1.1887794338448016e-07, + "loss": 0.6795, + "step": 22202 + }, + { + "epoch": 0.81, + "learning_rate": 1.188742813198081e-07, + "loss": 0.6972, + "step": 22203 + }, + { + "epoch": 0.81, + "learning_rate": 1.1887061925513604e-07, + "loss": 0.6881, + "step": 22204 + }, + { + "epoch": 0.81, + "learning_rate": 1.1886695719046398e-07, + "loss": 0.7055, + "step": 22205 + }, + { + "epoch": 0.81, + "learning_rate": 1.1886329512579192e-07, + "loss": 0.6992, + "step": 22206 + }, + { + "epoch": 0.81, + "learning_rate": 1.1885963306111986e-07, + "loss": 0.6757, + "step": 22207 + }, + { + "epoch": 0.81, + "learning_rate": 1.188559709964478e-07, + "loss": 0.6944, + "step": 22208 + }, + { + "epoch": 0.81, + "learning_rate": 1.1885230893177574e-07, + "loss": 0.6975, + "step": 22209 + }, + { + "epoch": 0.81, + "learning_rate": 1.1884864686710368e-07, + "loss": 0.6917, + "step": 22210 + }, + { + "epoch": 0.81, + "learning_rate": 1.1884498480243162e-07, + "loss": 0.6894, + "step": 22211 + }, + { + "epoch": 0.81, + "learning_rate": 1.1884132273775956e-07, + "loss": 0.7092, + "step": 22212 + }, + { + "epoch": 0.81, + "learning_rate": 1.188376606730875e-07, + "loss": 0.701, + "step": 22213 + }, + { + "epoch": 0.81, + "learning_rate": 1.1883399860841541e-07, + "loss": 0.6983, + "step": 22214 + }, + { + "epoch": 0.81, + "learning_rate": 1.1883033654374335e-07, + "loss": 0.6821, + "step": 22215 + }, + { + "epoch": 0.81, + "learning_rate": 1.1882667447907129e-07, + "loss": 0.7209, + "step": 22216 + }, + { + "epoch": 0.81, + "learning_rate": 1.1882301241439923e-07, + "loss": 0.693, + "step": 22217 + }, + { + "epoch": 0.81, + "learning_rate": 1.1881935034972717e-07, + "loss": 0.6908, + "step": 22218 + }, + { + "epoch": 0.81, + "learning_rate": 1.1881568828505511e-07, + "loss": 0.71, + "step": 22219 + }, + { + "epoch": 0.81, + "learning_rate": 1.1881202622038304e-07, + "loss": 0.6693, + "step": 22220 + }, + { + "epoch": 0.81, + "learning_rate": 1.1880836415571098e-07, + "loss": 0.698, + "step": 22221 + }, + { + "epoch": 0.81, + "learning_rate": 1.1880470209103892e-07, + "loss": 0.6961, + "step": 22222 + }, + { + "epoch": 0.81, + "learning_rate": 1.1880104002636686e-07, + "loss": 0.6841, + "step": 22223 + }, + { + "epoch": 0.81, + "learning_rate": 1.187973779616948e-07, + "loss": 0.6968, + "step": 22224 + }, + { + "epoch": 0.81, + "learning_rate": 1.1879371589702274e-07, + "loss": 0.6843, + "step": 22225 + }, + { + "epoch": 0.81, + "learning_rate": 1.1879005383235068e-07, + "loss": 0.6826, + "step": 22226 + }, + { + "epoch": 0.81, + "learning_rate": 1.1878639176767862e-07, + "loss": 0.6949, + "step": 22227 + }, + { + "epoch": 0.81, + "learning_rate": 1.1878272970300656e-07, + "loss": 0.7035, + "step": 22228 + }, + { + "epoch": 0.81, + "learning_rate": 1.187790676383345e-07, + "loss": 0.7009, + "step": 22229 + }, + { + "epoch": 0.81, + "learning_rate": 1.1877540557366244e-07, + "loss": 0.6931, + "step": 22230 + }, + { + "epoch": 0.81, + "learning_rate": 1.1877174350899038e-07, + "loss": 0.7033, + "step": 22231 + }, + { + "epoch": 0.81, + "learning_rate": 1.1876808144431829e-07, + "loss": 0.7051, + "step": 22232 + }, + { + "epoch": 0.81, + "learning_rate": 1.1876441937964623e-07, + "loss": 0.7006, + "step": 22233 + }, + { + "epoch": 0.81, + "learning_rate": 1.1876075731497417e-07, + "loss": 0.6968, + "step": 22234 + }, + { + "epoch": 0.81, + "learning_rate": 1.1875709525030211e-07, + "loss": 0.6892, + "step": 22235 + }, + { + "epoch": 0.81, + "learning_rate": 1.1875343318563005e-07, + "loss": 0.6833, + "step": 22236 + }, + { + "epoch": 0.81, + "learning_rate": 1.1874977112095799e-07, + "loss": 0.6886, + "step": 22237 + }, + { + "epoch": 0.81, + "learning_rate": 1.1874610905628593e-07, + "loss": 0.6895, + "step": 22238 + }, + { + "epoch": 0.81, + "learning_rate": 1.1874244699161387e-07, + "loss": 0.6891, + "step": 22239 + }, + { + "epoch": 0.81, + "learning_rate": 1.1873878492694181e-07, + "loss": 0.69, + "step": 22240 + }, + { + "epoch": 0.81, + "learning_rate": 1.1873512286226975e-07, + "loss": 0.6849, + "step": 22241 + }, + { + "epoch": 0.81, + "learning_rate": 1.1873146079759768e-07, + "loss": 0.6962, + "step": 22242 + }, + { + "epoch": 0.81, + "learning_rate": 1.1872779873292562e-07, + "loss": 0.6945, + "step": 22243 + }, + { + "epoch": 0.81, + "learning_rate": 1.1872413666825356e-07, + "loss": 0.6962, + "step": 22244 + }, + { + "epoch": 0.81, + "learning_rate": 1.187204746035815e-07, + "loss": 0.6993, + "step": 22245 + }, + { + "epoch": 0.81, + "learning_rate": 1.1871681253890944e-07, + "loss": 0.6887, + "step": 22246 + }, + { + "epoch": 0.81, + "learning_rate": 1.1871315047423738e-07, + "loss": 0.6909, + "step": 22247 + }, + { + "epoch": 0.81, + "learning_rate": 1.1870948840956532e-07, + "loss": 0.689, + "step": 22248 + }, + { + "epoch": 0.81, + "learning_rate": 1.1870582634489326e-07, + "loss": 0.6815, + "step": 22249 + }, + { + "epoch": 0.81, + "learning_rate": 1.1870216428022117e-07, + "loss": 0.6882, + "step": 22250 + }, + { + "epoch": 0.81, + "learning_rate": 1.1869850221554911e-07, + "loss": 0.7072, + "step": 22251 + }, + { + "epoch": 0.81, + "learning_rate": 1.1869484015087705e-07, + "loss": 0.6938, + "step": 22252 + }, + { + "epoch": 0.81, + "learning_rate": 1.1869117808620499e-07, + "loss": 0.7048, + "step": 22253 + }, + { + "epoch": 0.81, + "learning_rate": 1.1868751602153293e-07, + "loss": 0.6836, + "step": 22254 + }, + { + "epoch": 0.81, + "learning_rate": 1.1868385395686087e-07, + "loss": 0.6926, + "step": 22255 + }, + { + "epoch": 0.81, + "learning_rate": 1.1868019189218881e-07, + "loss": 0.69, + "step": 22256 + }, + { + "epoch": 0.81, + "learning_rate": 1.1867652982751675e-07, + "loss": 0.6991, + "step": 22257 + }, + { + "epoch": 0.81, + "learning_rate": 1.1867286776284469e-07, + "loss": 0.6866, + "step": 22258 + }, + { + "epoch": 0.81, + "learning_rate": 1.1866920569817263e-07, + "loss": 0.6921, + "step": 22259 + }, + { + "epoch": 0.81, + "learning_rate": 1.1866554363350057e-07, + "loss": 0.6844, + "step": 22260 + }, + { + "epoch": 0.81, + "learning_rate": 1.1866188156882851e-07, + "loss": 0.7028, + "step": 22261 + }, + { + "epoch": 0.81, + "learning_rate": 1.1865821950415645e-07, + "loss": 0.6836, + "step": 22262 + }, + { + "epoch": 0.81, + "learning_rate": 1.1865455743948438e-07, + "loss": 0.6962, + "step": 22263 + }, + { + "epoch": 0.81, + "learning_rate": 1.1865089537481232e-07, + "loss": 0.6926, + "step": 22264 + }, + { + "epoch": 0.81, + "learning_rate": 1.1864723331014026e-07, + "loss": 0.6996, + "step": 22265 + }, + { + "epoch": 0.81, + "learning_rate": 1.186435712454682e-07, + "loss": 0.6895, + "step": 22266 + }, + { + "epoch": 0.81, + "learning_rate": 1.1863990918079613e-07, + "loss": 0.7084, + "step": 22267 + }, + { + "epoch": 0.81, + "learning_rate": 1.1863624711612406e-07, + "loss": 0.7038, + "step": 22268 + }, + { + "epoch": 0.81, + "learning_rate": 1.1863258505145199e-07, + "loss": 0.6815, + "step": 22269 + }, + { + "epoch": 0.81, + "learning_rate": 1.1862892298677993e-07, + "loss": 0.7002, + "step": 22270 + }, + { + "epoch": 0.81, + "learning_rate": 1.1862526092210787e-07, + "loss": 0.6895, + "step": 22271 + }, + { + "epoch": 0.81, + "learning_rate": 1.1862159885743581e-07, + "loss": 0.7019, + "step": 22272 + }, + { + "epoch": 0.81, + "learning_rate": 1.1861793679276375e-07, + "loss": 0.6946, + "step": 22273 + }, + { + "epoch": 0.81, + "learning_rate": 1.1861427472809169e-07, + "loss": 0.7087, + "step": 22274 + }, + { + "epoch": 0.81, + "learning_rate": 1.1861061266341963e-07, + "loss": 0.6916, + "step": 22275 + }, + { + "epoch": 0.82, + "learning_rate": 1.1860695059874757e-07, + "loss": 0.6904, + "step": 22276 + }, + { + "epoch": 0.82, + "learning_rate": 1.1860328853407551e-07, + "loss": 0.7011, + "step": 22277 + }, + { + "epoch": 0.82, + "learning_rate": 1.1859962646940345e-07, + "loss": 0.6795, + "step": 22278 + }, + { + "epoch": 0.82, + "learning_rate": 1.1859596440473139e-07, + "loss": 0.6993, + "step": 22279 + }, + { + "epoch": 0.82, + "learning_rate": 1.1859230234005933e-07, + "loss": 0.6877, + "step": 22280 + }, + { + "epoch": 0.82, + "learning_rate": 1.1858864027538727e-07, + "loss": 0.7089, + "step": 22281 + }, + { + "epoch": 0.82, + "learning_rate": 1.1858497821071521e-07, + "loss": 0.6872, + "step": 22282 + }, + { + "epoch": 0.82, + "learning_rate": 1.1858131614604315e-07, + "loss": 0.695, + "step": 22283 + }, + { + "epoch": 0.82, + "learning_rate": 1.1857765408137109e-07, + "loss": 0.6933, + "step": 22284 + }, + { + "epoch": 0.82, + "learning_rate": 1.1857399201669901e-07, + "loss": 0.6937, + "step": 22285 + }, + { + "epoch": 0.82, + "learning_rate": 1.1857032995202694e-07, + "loss": 0.6838, + "step": 22286 + }, + { + "epoch": 0.82, + "learning_rate": 1.1856666788735488e-07, + "loss": 0.6881, + "step": 22287 + }, + { + "epoch": 0.82, + "learning_rate": 1.1856300582268282e-07, + "loss": 0.7026, + "step": 22288 + }, + { + "epoch": 0.82, + "learning_rate": 1.1855934375801076e-07, + "loss": 0.6818, + "step": 22289 + }, + { + "epoch": 0.82, + "learning_rate": 1.185556816933387e-07, + "loss": 0.6855, + "step": 22290 + }, + { + "epoch": 0.82, + "learning_rate": 1.1855201962866663e-07, + "loss": 0.701, + "step": 22291 + }, + { + "epoch": 0.82, + "learning_rate": 1.1854835756399457e-07, + "loss": 0.681, + "step": 22292 + }, + { + "epoch": 0.82, + "learning_rate": 1.1854469549932251e-07, + "loss": 0.6714, + "step": 22293 + }, + { + "epoch": 0.82, + "learning_rate": 1.1854103343465045e-07, + "loss": 0.6821, + "step": 22294 + }, + { + "epoch": 0.82, + "learning_rate": 1.1853737136997839e-07, + "loss": 0.6884, + "step": 22295 + }, + { + "epoch": 0.82, + "learning_rate": 1.1853370930530633e-07, + "loss": 0.6932, + "step": 22296 + }, + { + "epoch": 0.82, + "learning_rate": 1.1853004724063427e-07, + "loss": 0.686, + "step": 22297 + }, + { + "epoch": 0.82, + "learning_rate": 1.185263851759622e-07, + "loss": 0.6914, + "step": 22298 + }, + { + "epoch": 0.82, + "learning_rate": 1.1852272311129015e-07, + "loss": 0.7004, + "step": 22299 + }, + { + "epoch": 0.82, + "learning_rate": 1.1851906104661809e-07, + "loss": 0.6987, + "step": 22300 + }, + { + "epoch": 0.82, + "learning_rate": 1.1851539898194603e-07, + "loss": 0.6754, + "step": 22301 + }, + { + "epoch": 0.82, + "learning_rate": 1.1851173691727397e-07, + "loss": 0.6874, + "step": 22302 + }, + { + "epoch": 0.82, + "learning_rate": 1.185080748526019e-07, + "loss": 0.6897, + "step": 22303 + }, + { + "epoch": 0.82, + "learning_rate": 1.1850441278792982e-07, + "loss": 0.6871, + "step": 22304 + }, + { + "epoch": 0.82, + "learning_rate": 1.1850075072325776e-07, + "loss": 0.6954, + "step": 22305 + }, + { + "epoch": 0.82, + "learning_rate": 1.184970886585857e-07, + "loss": 0.6851, + "step": 22306 + }, + { + "epoch": 0.82, + "learning_rate": 1.1849342659391364e-07, + "loss": 0.7047, + "step": 22307 + }, + { + "epoch": 0.82, + "learning_rate": 1.1848976452924158e-07, + "loss": 0.7003, + "step": 22308 + }, + { + "epoch": 0.82, + "learning_rate": 1.1848610246456952e-07, + "loss": 0.7024, + "step": 22309 + }, + { + "epoch": 0.82, + "learning_rate": 1.1848244039989746e-07, + "loss": 0.7055, + "step": 22310 + }, + { + "epoch": 0.82, + "learning_rate": 1.184787783352254e-07, + "loss": 0.6952, + "step": 22311 + }, + { + "epoch": 0.82, + "learning_rate": 1.1847511627055333e-07, + "loss": 0.6888, + "step": 22312 + }, + { + "epoch": 0.82, + "learning_rate": 1.1847145420588127e-07, + "loss": 0.7056, + "step": 22313 + }, + { + "epoch": 0.82, + "learning_rate": 1.184677921412092e-07, + "loss": 0.7014, + "step": 22314 + }, + { + "epoch": 0.82, + "learning_rate": 1.1846413007653715e-07, + "loss": 0.6963, + "step": 22315 + }, + { + "epoch": 0.82, + "learning_rate": 1.1846046801186509e-07, + "loss": 0.6946, + "step": 22316 + }, + { + "epoch": 0.82, + "learning_rate": 1.1845680594719303e-07, + "loss": 0.6895, + "step": 22317 + }, + { + "epoch": 0.82, + "learning_rate": 1.1845314388252097e-07, + "loss": 0.6868, + "step": 22318 + }, + { + "epoch": 0.82, + "learning_rate": 1.184494818178489e-07, + "loss": 0.6985, + "step": 22319 + }, + { + "epoch": 0.82, + "learning_rate": 1.1844581975317685e-07, + "loss": 0.6825, + "step": 22320 + }, + { + "epoch": 0.82, + "learning_rate": 1.1844215768850479e-07, + "loss": 0.6968, + "step": 22321 + }, + { + "epoch": 0.82, + "learning_rate": 1.184384956238327e-07, + "loss": 0.6971, + "step": 22322 + }, + { + "epoch": 0.82, + "learning_rate": 1.1843483355916064e-07, + "loss": 0.6806, + "step": 22323 + }, + { + "epoch": 0.82, + "learning_rate": 1.1843117149448858e-07, + "loss": 0.7023, + "step": 22324 + }, + { + "epoch": 0.82, + "learning_rate": 1.1842750942981652e-07, + "loss": 0.7001, + "step": 22325 + }, + { + "epoch": 0.82, + "learning_rate": 1.1842384736514446e-07, + "loss": 0.6887, + "step": 22326 + }, + { + "epoch": 0.82, + "learning_rate": 1.184201853004724e-07, + "loss": 0.6981, + "step": 22327 + }, + { + "epoch": 0.82, + "learning_rate": 1.1841652323580034e-07, + "loss": 0.6827, + "step": 22328 + }, + { + "epoch": 0.82, + "learning_rate": 1.1841286117112828e-07, + "loss": 0.6956, + "step": 22329 + }, + { + "epoch": 0.82, + "learning_rate": 1.1840919910645622e-07, + "loss": 0.687, + "step": 22330 + }, + { + "epoch": 0.82, + "learning_rate": 1.1840553704178416e-07, + "loss": 0.7069, + "step": 22331 + }, + { + "epoch": 0.82, + "learning_rate": 1.184018749771121e-07, + "loss": 0.6926, + "step": 22332 + }, + { + "epoch": 0.82, + "learning_rate": 1.1839821291244004e-07, + "loss": 0.6873, + "step": 22333 + }, + { + "epoch": 0.82, + "learning_rate": 1.1839455084776796e-07, + "loss": 0.6827, + "step": 22334 + }, + { + "epoch": 0.82, + "learning_rate": 1.183908887830959e-07, + "loss": 0.6945, + "step": 22335 + }, + { + "epoch": 0.82, + "learning_rate": 1.1838722671842384e-07, + "loss": 0.6963, + "step": 22336 + }, + { + "epoch": 0.82, + "learning_rate": 1.1838356465375178e-07, + "loss": 0.6905, + "step": 22337 + }, + { + "epoch": 0.82, + "learning_rate": 1.1837990258907972e-07, + "loss": 0.691, + "step": 22338 + }, + { + "epoch": 0.82, + "learning_rate": 1.1837624052440766e-07, + "loss": 0.7022, + "step": 22339 + }, + { + "epoch": 0.82, + "learning_rate": 1.1837257845973558e-07, + "loss": 0.6936, + "step": 22340 + }, + { + "epoch": 0.82, + "learning_rate": 1.1836891639506352e-07, + "loss": 0.7019, + "step": 22341 + }, + { + "epoch": 0.82, + "learning_rate": 1.1836525433039146e-07, + "loss": 0.6876, + "step": 22342 + }, + { + "epoch": 0.82, + "learning_rate": 1.183615922657194e-07, + "loss": 0.6978, + "step": 22343 + }, + { + "epoch": 0.82, + "learning_rate": 1.1835793020104734e-07, + "loss": 0.6886, + "step": 22344 + }, + { + "epoch": 0.82, + "learning_rate": 1.1835426813637528e-07, + "loss": 0.7015, + "step": 22345 + }, + { + "epoch": 0.82, + "learning_rate": 1.1835060607170322e-07, + "loss": 0.7001, + "step": 22346 + }, + { + "epoch": 0.82, + "learning_rate": 1.1834694400703116e-07, + "loss": 0.7008, + "step": 22347 + }, + { + "epoch": 0.82, + "learning_rate": 1.183432819423591e-07, + "loss": 0.7015, + "step": 22348 + }, + { + "epoch": 0.82, + "learning_rate": 1.1833961987768704e-07, + "loss": 0.7008, + "step": 22349 + }, + { + "epoch": 0.82, + "learning_rate": 1.1833595781301498e-07, + "loss": 0.6872, + "step": 22350 + }, + { + "epoch": 0.82, + "learning_rate": 1.1833229574834292e-07, + "loss": 0.6874, + "step": 22351 + }, + { + "epoch": 0.82, + "learning_rate": 1.1832863368367086e-07, + "loss": 0.7037, + "step": 22352 + }, + { + "epoch": 0.82, + "learning_rate": 1.183249716189988e-07, + "loss": 0.7081, + "step": 22353 + }, + { + "epoch": 0.82, + "learning_rate": 1.1832130955432674e-07, + "loss": 0.6737, + "step": 22354 + }, + { + "epoch": 0.82, + "learning_rate": 1.1831764748965468e-07, + "loss": 0.67, + "step": 22355 + }, + { + "epoch": 0.82, + "learning_rate": 1.183139854249826e-07, + "loss": 0.6977, + "step": 22356 + }, + { + "epoch": 0.82, + "learning_rate": 1.1831032336031054e-07, + "loss": 0.6995, + "step": 22357 + }, + { + "epoch": 0.82, + "learning_rate": 1.1830666129563847e-07, + "loss": 0.6882, + "step": 22358 + }, + { + "epoch": 0.82, + "learning_rate": 1.1830299923096641e-07, + "loss": 0.6981, + "step": 22359 + }, + { + "epoch": 0.82, + "learning_rate": 1.1829933716629435e-07, + "loss": 0.6791, + "step": 22360 + }, + { + "epoch": 0.82, + "learning_rate": 1.1829567510162229e-07, + "loss": 0.709, + "step": 22361 + }, + { + "epoch": 0.82, + "learning_rate": 1.1829201303695022e-07, + "loss": 0.6886, + "step": 22362 + }, + { + "epoch": 0.82, + "learning_rate": 1.1828835097227816e-07, + "loss": 0.687, + "step": 22363 + }, + { + "epoch": 0.82, + "learning_rate": 1.182846889076061e-07, + "loss": 0.6866, + "step": 22364 + }, + { + "epoch": 0.82, + "learning_rate": 1.1828102684293404e-07, + "loss": 0.6828, + "step": 22365 + }, + { + "epoch": 0.82, + "learning_rate": 1.1827736477826198e-07, + "loss": 0.6968, + "step": 22366 + }, + { + "epoch": 0.82, + "learning_rate": 1.1827370271358992e-07, + "loss": 0.6959, + "step": 22367 + }, + { + "epoch": 0.82, + "learning_rate": 1.1827004064891786e-07, + "loss": 0.6861, + "step": 22368 + }, + { + "epoch": 0.82, + "learning_rate": 1.182663785842458e-07, + "loss": 0.6871, + "step": 22369 + }, + { + "epoch": 0.82, + "learning_rate": 1.1826271651957374e-07, + "loss": 0.6884, + "step": 22370 + }, + { + "epoch": 0.82, + "learning_rate": 1.1825905445490168e-07, + "loss": 0.6935, + "step": 22371 + }, + { + "epoch": 0.82, + "learning_rate": 1.1825539239022962e-07, + "loss": 0.688, + "step": 22372 + }, + { + "epoch": 0.82, + "learning_rate": 1.1825173032555756e-07, + "loss": 0.6877, + "step": 22373 + }, + { + "epoch": 0.82, + "learning_rate": 1.182480682608855e-07, + "loss": 0.6951, + "step": 22374 + }, + { + "epoch": 0.82, + "learning_rate": 1.1824440619621341e-07, + "loss": 0.6859, + "step": 22375 + }, + { + "epoch": 0.82, + "learning_rate": 1.1824074413154135e-07, + "loss": 0.702, + "step": 22376 + }, + { + "epoch": 0.82, + "learning_rate": 1.1823708206686929e-07, + "loss": 0.6954, + "step": 22377 + }, + { + "epoch": 0.82, + "learning_rate": 1.1823342000219723e-07, + "loss": 0.6961, + "step": 22378 + }, + { + "epoch": 0.82, + "learning_rate": 1.1822975793752517e-07, + "loss": 0.6959, + "step": 22379 + }, + { + "epoch": 0.82, + "learning_rate": 1.1822609587285311e-07, + "loss": 0.6851, + "step": 22380 + }, + { + "epoch": 0.82, + "learning_rate": 1.1822243380818105e-07, + "loss": 0.7024, + "step": 22381 + }, + { + "epoch": 0.82, + "learning_rate": 1.1821877174350899e-07, + "loss": 0.6889, + "step": 22382 + }, + { + "epoch": 0.82, + "learning_rate": 1.1821510967883692e-07, + "loss": 0.6859, + "step": 22383 + }, + { + "epoch": 0.82, + "learning_rate": 1.1821144761416486e-07, + "loss": 0.6854, + "step": 22384 + }, + { + "epoch": 0.82, + "learning_rate": 1.182077855494928e-07, + "loss": 0.6878, + "step": 22385 + }, + { + "epoch": 0.82, + "learning_rate": 1.1820412348482074e-07, + "loss": 0.6984, + "step": 22386 + }, + { + "epoch": 0.82, + "learning_rate": 1.1820046142014867e-07, + "loss": 0.6983, + "step": 22387 + }, + { + "epoch": 0.82, + "learning_rate": 1.1819679935547661e-07, + "loss": 0.6995, + "step": 22388 + }, + { + "epoch": 0.82, + "learning_rate": 1.1819313729080455e-07, + "loss": 0.691, + "step": 22389 + }, + { + "epoch": 0.82, + "learning_rate": 1.181894752261325e-07, + "loss": 0.6948, + "step": 22390 + }, + { + "epoch": 0.82, + "learning_rate": 1.1818581316146043e-07, + "loss": 0.6941, + "step": 22391 + }, + { + "epoch": 0.82, + "learning_rate": 1.1818215109678837e-07, + "loss": 0.6873, + "step": 22392 + }, + { + "epoch": 0.82, + "learning_rate": 1.1817848903211629e-07, + "loss": 0.6864, + "step": 22393 + }, + { + "epoch": 0.82, + "learning_rate": 1.1817482696744423e-07, + "loss": 0.693, + "step": 22394 + }, + { + "epoch": 0.82, + "learning_rate": 1.1817116490277217e-07, + "loss": 0.6931, + "step": 22395 + }, + { + "epoch": 0.82, + "learning_rate": 1.1816750283810011e-07, + "loss": 0.6997, + "step": 22396 + }, + { + "epoch": 0.82, + "learning_rate": 1.1816384077342805e-07, + "loss": 0.7008, + "step": 22397 + }, + { + "epoch": 0.82, + "learning_rate": 1.1816017870875599e-07, + "loss": 0.6901, + "step": 22398 + }, + { + "epoch": 0.82, + "learning_rate": 1.1815651664408393e-07, + "loss": 0.6944, + "step": 22399 + }, + { + "epoch": 0.82, + "learning_rate": 1.1815285457941187e-07, + "loss": 0.6942, + "step": 22400 + }, + { + "epoch": 0.82, + "learning_rate": 1.1814919251473981e-07, + "loss": 0.6936, + "step": 22401 + }, + { + "epoch": 0.82, + "learning_rate": 1.1814553045006775e-07, + "loss": 0.696, + "step": 22402 + }, + { + "epoch": 0.82, + "learning_rate": 1.1814186838539569e-07, + "loss": 0.6896, + "step": 22403 + }, + { + "epoch": 0.82, + "learning_rate": 1.1813820632072363e-07, + "loss": 0.6891, + "step": 22404 + }, + { + "epoch": 0.82, + "learning_rate": 1.1813454425605155e-07, + "loss": 0.681, + "step": 22405 + }, + { + "epoch": 0.82, + "learning_rate": 1.181308821913795e-07, + "loss": 0.6937, + "step": 22406 + }, + { + "epoch": 0.82, + "learning_rate": 1.1812722012670743e-07, + "loss": 0.693, + "step": 22407 + }, + { + "epoch": 0.82, + "learning_rate": 1.1812355806203537e-07, + "loss": 0.6975, + "step": 22408 + }, + { + "epoch": 0.82, + "learning_rate": 1.1811989599736331e-07, + "loss": 0.6998, + "step": 22409 + }, + { + "epoch": 0.82, + "learning_rate": 1.1811623393269125e-07, + "loss": 0.6898, + "step": 22410 + }, + { + "epoch": 0.82, + "learning_rate": 1.1811257186801917e-07, + "loss": 0.6893, + "step": 22411 + }, + { + "epoch": 0.82, + "learning_rate": 1.1810890980334711e-07, + "loss": 0.6972, + "step": 22412 + }, + { + "epoch": 0.82, + "learning_rate": 1.1810524773867505e-07, + "loss": 0.6891, + "step": 22413 + }, + { + "epoch": 0.82, + "learning_rate": 1.1810158567400299e-07, + "loss": 0.6984, + "step": 22414 + }, + { + "epoch": 0.82, + "learning_rate": 1.1809792360933093e-07, + "loss": 0.7064, + "step": 22415 + }, + { + "epoch": 0.82, + "learning_rate": 1.1809426154465887e-07, + "loss": 0.6987, + "step": 22416 + }, + { + "epoch": 0.82, + "learning_rate": 1.1809059947998681e-07, + "loss": 0.6914, + "step": 22417 + }, + { + "epoch": 0.82, + "learning_rate": 1.1808693741531475e-07, + "loss": 0.6927, + "step": 22418 + }, + { + "epoch": 0.82, + "learning_rate": 1.1808327535064269e-07, + "loss": 0.6947, + "step": 22419 + }, + { + "epoch": 0.82, + "learning_rate": 1.1807961328597063e-07, + "loss": 0.6959, + "step": 22420 + }, + { + "epoch": 0.82, + "learning_rate": 1.1807595122129857e-07, + "loss": 0.7104, + "step": 22421 + }, + { + "epoch": 0.82, + "learning_rate": 1.180722891566265e-07, + "loss": 0.6902, + "step": 22422 + }, + { + "epoch": 0.82, + "learning_rate": 1.1806862709195445e-07, + "loss": 0.6961, + "step": 22423 + }, + { + "epoch": 0.82, + "learning_rate": 1.1806496502728239e-07, + "loss": 0.6865, + "step": 22424 + }, + { + "epoch": 0.82, + "learning_rate": 1.1806130296261033e-07, + "loss": 0.7096, + "step": 22425 + }, + { + "epoch": 0.82, + "learning_rate": 1.1805764089793827e-07, + "loss": 0.6958, + "step": 22426 + }, + { + "epoch": 0.82, + "learning_rate": 1.1805397883326619e-07, + "loss": 0.6958, + "step": 22427 + }, + { + "epoch": 0.82, + "learning_rate": 1.1805031676859413e-07, + "loss": 0.6951, + "step": 22428 + }, + { + "epoch": 0.82, + "learning_rate": 1.1804665470392206e-07, + "loss": 0.6861, + "step": 22429 + }, + { + "epoch": 0.82, + "learning_rate": 1.1804299263925e-07, + "loss": 0.6838, + "step": 22430 + }, + { + "epoch": 0.82, + "learning_rate": 1.1803933057457794e-07, + "loss": 0.6961, + "step": 22431 + }, + { + "epoch": 0.82, + "learning_rate": 1.1803566850990587e-07, + "loss": 0.6809, + "step": 22432 + }, + { + "epoch": 0.82, + "learning_rate": 1.180320064452338e-07, + "loss": 0.7054, + "step": 22433 + }, + { + "epoch": 0.82, + "learning_rate": 1.1802834438056175e-07, + "loss": 0.7045, + "step": 22434 + }, + { + "epoch": 0.82, + "learning_rate": 1.1802468231588969e-07, + "loss": 0.6876, + "step": 22435 + }, + { + "epoch": 0.82, + "learning_rate": 1.1802102025121763e-07, + "loss": 0.6965, + "step": 22436 + }, + { + "epoch": 0.82, + "learning_rate": 1.1801735818654557e-07, + "loss": 0.7142, + "step": 22437 + }, + { + "epoch": 0.82, + "learning_rate": 1.180136961218735e-07, + "loss": 0.6954, + "step": 22438 + }, + { + "epoch": 0.82, + "learning_rate": 1.1801003405720145e-07, + "loss": 0.6834, + "step": 22439 + }, + { + "epoch": 0.82, + "learning_rate": 1.1800637199252939e-07, + "loss": 0.6915, + "step": 22440 + }, + { + "epoch": 0.82, + "learning_rate": 1.1800270992785733e-07, + "loss": 0.7012, + "step": 22441 + }, + { + "epoch": 0.82, + "learning_rate": 1.1799904786318527e-07, + "loss": 0.6972, + "step": 22442 + }, + { + "epoch": 0.82, + "learning_rate": 1.179953857985132e-07, + "loss": 0.6852, + "step": 22443 + }, + { + "epoch": 0.82, + "learning_rate": 1.1799172373384114e-07, + "loss": 0.6912, + "step": 22444 + }, + { + "epoch": 0.82, + "learning_rate": 1.1798806166916908e-07, + "loss": 0.7024, + "step": 22445 + }, + { + "epoch": 0.82, + "learning_rate": 1.1798439960449702e-07, + "loss": 0.6967, + "step": 22446 + }, + { + "epoch": 0.82, + "learning_rate": 1.1798073753982494e-07, + "loss": 0.6999, + "step": 22447 + }, + { + "epoch": 0.82, + "learning_rate": 1.1797707547515288e-07, + "loss": 0.6985, + "step": 22448 + }, + { + "epoch": 0.82, + "learning_rate": 1.1797341341048082e-07, + "loss": 0.6872, + "step": 22449 + }, + { + "epoch": 0.82, + "learning_rate": 1.1796975134580876e-07, + "loss": 0.6883, + "step": 22450 + }, + { + "epoch": 0.82, + "learning_rate": 1.179660892811367e-07, + "loss": 0.689, + "step": 22451 + }, + { + "epoch": 0.82, + "learning_rate": 1.1796242721646464e-07, + "loss": 0.6889, + "step": 22452 + }, + { + "epoch": 0.82, + "learning_rate": 1.1795876515179258e-07, + "loss": 0.697, + "step": 22453 + }, + { + "epoch": 0.82, + "learning_rate": 1.179551030871205e-07, + "loss": 0.6884, + "step": 22454 + }, + { + "epoch": 0.82, + "learning_rate": 1.1795144102244844e-07, + "loss": 0.6982, + "step": 22455 + }, + { + "epoch": 0.82, + "learning_rate": 1.1794777895777638e-07, + "loss": 0.6974, + "step": 22456 + }, + { + "epoch": 0.82, + "learning_rate": 1.1794411689310432e-07, + "loss": 0.6813, + "step": 22457 + }, + { + "epoch": 0.82, + "learning_rate": 1.1794045482843226e-07, + "loss": 0.7009, + "step": 22458 + }, + { + "epoch": 0.82, + "learning_rate": 1.179367927637602e-07, + "loss": 0.6925, + "step": 22459 + }, + { + "epoch": 0.82, + "learning_rate": 1.1793313069908814e-07, + "loss": 0.695, + "step": 22460 + }, + { + "epoch": 0.82, + "learning_rate": 1.1792946863441608e-07, + "loss": 0.6878, + "step": 22461 + }, + { + "epoch": 0.82, + "learning_rate": 1.1792580656974402e-07, + "loss": 0.6954, + "step": 22462 + }, + { + "epoch": 0.82, + "learning_rate": 1.1792214450507196e-07, + "loss": 0.6871, + "step": 22463 + }, + { + "epoch": 0.82, + "learning_rate": 1.179184824403999e-07, + "loss": 0.6971, + "step": 22464 + }, + { + "epoch": 0.82, + "learning_rate": 1.1791482037572782e-07, + "loss": 0.7088, + "step": 22465 + }, + { + "epoch": 0.82, + "learning_rate": 1.1791115831105576e-07, + "loss": 0.6831, + "step": 22466 + }, + { + "epoch": 0.82, + "learning_rate": 1.179074962463837e-07, + "loss": 0.7044, + "step": 22467 + }, + { + "epoch": 0.82, + "learning_rate": 1.1790383418171164e-07, + "loss": 0.6978, + "step": 22468 + }, + { + "epoch": 0.82, + "learning_rate": 1.1790017211703958e-07, + "loss": 0.6956, + "step": 22469 + }, + { + "epoch": 0.82, + "learning_rate": 1.1789651005236752e-07, + "loss": 0.6893, + "step": 22470 + }, + { + "epoch": 0.82, + "learning_rate": 1.1789284798769546e-07, + "loss": 0.6892, + "step": 22471 + }, + { + "epoch": 0.82, + "learning_rate": 1.178891859230234e-07, + "loss": 0.6939, + "step": 22472 + }, + { + "epoch": 0.82, + "learning_rate": 1.1788552385835134e-07, + "loss": 0.6901, + "step": 22473 + }, + { + "epoch": 0.82, + "learning_rate": 1.1788186179367928e-07, + "loss": 0.7011, + "step": 22474 + }, + { + "epoch": 0.82, + "learning_rate": 1.1787819972900722e-07, + "loss": 0.6865, + "step": 22475 + }, + { + "epoch": 0.82, + "learning_rate": 1.1787453766433514e-07, + "loss": 0.6942, + "step": 22476 + }, + { + "epoch": 0.82, + "learning_rate": 1.1787087559966308e-07, + "loss": 0.7067, + "step": 22477 + }, + { + "epoch": 0.82, + "learning_rate": 1.1786721353499102e-07, + "loss": 0.6898, + "step": 22478 + }, + { + "epoch": 0.82, + "learning_rate": 1.1786355147031896e-07, + "loss": 0.705, + "step": 22479 + }, + { + "epoch": 0.82, + "learning_rate": 1.178598894056469e-07, + "loss": 0.7065, + "step": 22480 + }, + { + "epoch": 0.82, + "learning_rate": 1.1785622734097484e-07, + "loss": 0.6878, + "step": 22481 + }, + { + "epoch": 0.82, + "learning_rate": 1.1785256527630278e-07, + "loss": 0.7088, + "step": 22482 + }, + { + "epoch": 0.82, + "learning_rate": 1.178489032116307e-07, + "loss": 0.6909, + "step": 22483 + }, + { + "epoch": 0.82, + "learning_rate": 1.1784524114695864e-07, + "loss": 0.6928, + "step": 22484 + }, + { + "epoch": 0.82, + "learning_rate": 1.1784157908228658e-07, + "loss": 0.6927, + "step": 22485 + }, + { + "epoch": 0.82, + "learning_rate": 1.1783791701761452e-07, + "loss": 0.6889, + "step": 22486 + }, + { + "epoch": 0.82, + "learning_rate": 1.1783425495294246e-07, + "loss": 0.6877, + "step": 22487 + }, + { + "epoch": 0.82, + "learning_rate": 1.178305928882704e-07, + "loss": 0.6993, + "step": 22488 + }, + { + "epoch": 0.82, + "learning_rate": 1.1782693082359834e-07, + "loss": 0.7054, + "step": 22489 + }, + { + "epoch": 0.82, + "learning_rate": 1.1782326875892628e-07, + "loss": 0.6917, + "step": 22490 + }, + { + "epoch": 0.82, + "learning_rate": 1.1781960669425422e-07, + "loss": 0.6883, + "step": 22491 + }, + { + "epoch": 0.82, + "learning_rate": 1.1781594462958216e-07, + "loss": 0.7047, + "step": 22492 + }, + { + "epoch": 0.82, + "learning_rate": 1.178122825649101e-07, + "loss": 0.6975, + "step": 22493 + }, + { + "epoch": 0.82, + "learning_rate": 1.1780862050023804e-07, + "loss": 0.6876, + "step": 22494 + }, + { + "epoch": 0.82, + "learning_rate": 1.1780495843556598e-07, + "loss": 0.6925, + "step": 22495 + }, + { + "epoch": 0.82, + "learning_rate": 1.1780129637089392e-07, + "loss": 0.6928, + "step": 22496 + }, + { + "epoch": 0.82, + "learning_rate": 1.1779763430622186e-07, + "loss": 0.7026, + "step": 22497 + }, + { + "epoch": 0.82, + "learning_rate": 1.1779397224154978e-07, + "loss": 0.6969, + "step": 22498 + }, + { + "epoch": 0.82, + "learning_rate": 1.1779031017687772e-07, + "loss": 0.6905, + "step": 22499 + }, + { + "epoch": 0.82, + "learning_rate": 1.1778664811220566e-07, + "loss": 0.6947, + "step": 22500 + }, + { + "epoch": 0.82, + "learning_rate": 1.1778298604753359e-07, + "loss": 0.6897, + "step": 22501 + }, + { + "epoch": 0.82, + "learning_rate": 1.1777932398286153e-07, + "loss": 0.6971, + "step": 22502 + }, + { + "epoch": 0.82, + "learning_rate": 1.1777566191818946e-07, + "loss": 0.6843, + "step": 22503 + }, + { + "epoch": 0.82, + "learning_rate": 1.177719998535174e-07, + "loss": 0.6959, + "step": 22504 + }, + { + "epoch": 0.82, + "learning_rate": 1.1776833778884534e-07, + "loss": 0.6857, + "step": 22505 + }, + { + "epoch": 0.82, + "learning_rate": 1.1776467572417327e-07, + "loss": 0.6853, + "step": 22506 + }, + { + "epoch": 0.82, + "learning_rate": 1.1776101365950121e-07, + "loss": 0.685, + "step": 22507 + }, + { + "epoch": 0.82, + "learning_rate": 1.1775735159482915e-07, + "loss": 0.6938, + "step": 22508 + }, + { + "epoch": 0.82, + "learning_rate": 1.177536895301571e-07, + "loss": 0.6976, + "step": 22509 + }, + { + "epoch": 0.82, + "learning_rate": 1.1775002746548503e-07, + "loss": 0.6915, + "step": 22510 + }, + { + "epoch": 0.82, + "learning_rate": 1.1774636540081297e-07, + "loss": 0.6956, + "step": 22511 + }, + { + "epoch": 0.82, + "learning_rate": 1.1774270333614091e-07, + "loss": 0.7035, + "step": 22512 + }, + { + "epoch": 0.82, + "learning_rate": 1.1773904127146885e-07, + "loss": 0.7026, + "step": 22513 + }, + { + "epoch": 0.82, + "learning_rate": 1.177353792067968e-07, + "loss": 0.6821, + "step": 22514 + }, + { + "epoch": 0.82, + "learning_rate": 1.1773171714212473e-07, + "loss": 0.701, + "step": 22515 + }, + { + "epoch": 0.82, + "learning_rate": 1.1772805507745267e-07, + "loss": 0.6994, + "step": 22516 + }, + { + "epoch": 0.82, + "learning_rate": 1.1772439301278061e-07, + "loss": 0.701, + "step": 22517 + }, + { + "epoch": 0.82, + "learning_rate": 1.1772073094810855e-07, + "loss": 0.6788, + "step": 22518 + }, + { + "epoch": 0.82, + "learning_rate": 1.1771706888343647e-07, + "loss": 0.6889, + "step": 22519 + }, + { + "epoch": 0.82, + "learning_rate": 1.1771340681876441e-07, + "loss": 0.6797, + "step": 22520 + }, + { + "epoch": 0.82, + "learning_rate": 1.1770974475409235e-07, + "loss": 0.6983, + "step": 22521 + }, + { + "epoch": 0.82, + "learning_rate": 1.1770608268942029e-07, + "loss": 0.6924, + "step": 22522 + }, + { + "epoch": 0.82, + "learning_rate": 1.1770242062474823e-07, + "loss": 0.685, + "step": 22523 + }, + { + "epoch": 0.82, + "learning_rate": 1.1769875856007617e-07, + "loss": 0.7023, + "step": 22524 + }, + { + "epoch": 0.82, + "learning_rate": 1.176950964954041e-07, + "loss": 0.6833, + "step": 22525 + }, + { + "epoch": 0.82, + "learning_rate": 1.1769143443073203e-07, + "loss": 0.6978, + "step": 22526 + }, + { + "epoch": 0.82, + "learning_rate": 1.1768777236605997e-07, + "loss": 0.6982, + "step": 22527 + }, + { + "epoch": 0.82, + "learning_rate": 1.1768411030138791e-07, + "loss": 0.6942, + "step": 22528 + }, + { + "epoch": 0.82, + "learning_rate": 1.1768044823671585e-07, + "loss": 0.7004, + "step": 22529 + }, + { + "epoch": 0.82, + "learning_rate": 1.176767861720438e-07, + "loss": 0.6998, + "step": 22530 + }, + { + "epoch": 0.82, + "learning_rate": 1.1767312410737173e-07, + "loss": 0.6697, + "step": 22531 + }, + { + "epoch": 0.82, + "learning_rate": 1.1766946204269967e-07, + "loss": 0.7022, + "step": 22532 + }, + { + "epoch": 0.82, + "learning_rate": 1.1766579997802761e-07, + "loss": 0.6851, + "step": 22533 + }, + { + "epoch": 0.82, + "learning_rate": 1.1766213791335555e-07, + "loss": 0.6861, + "step": 22534 + }, + { + "epoch": 0.82, + "learning_rate": 1.1765847584868349e-07, + "loss": 0.6937, + "step": 22535 + }, + { + "epoch": 0.82, + "learning_rate": 1.1765481378401143e-07, + "loss": 0.6937, + "step": 22536 + }, + { + "epoch": 0.82, + "learning_rate": 1.1765115171933935e-07, + "loss": 0.6911, + "step": 22537 + }, + { + "epoch": 0.82, + "learning_rate": 1.1764748965466729e-07, + "loss": 0.7055, + "step": 22538 + }, + { + "epoch": 0.82, + "learning_rate": 1.1764382758999523e-07, + "loss": 0.6879, + "step": 22539 + }, + { + "epoch": 0.82, + "learning_rate": 1.1764016552532317e-07, + "loss": 0.7042, + "step": 22540 + }, + { + "epoch": 0.82, + "learning_rate": 1.176365034606511e-07, + "loss": 0.7083, + "step": 22541 + }, + { + "epoch": 0.82, + "learning_rate": 1.1763284139597905e-07, + "loss": 0.7072, + "step": 22542 + }, + { + "epoch": 0.82, + "learning_rate": 1.1762917933130699e-07, + "loss": 0.6912, + "step": 22543 + }, + { + "epoch": 0.82, + "learning_rate": 1.1762551726663493e-07, + "loss": 0.6936, + "step": 22544 + }, + { + "epoch": 0.82, + "learning_rate": 1.1762185520196287e-07, + "loss": 0.6932, + "step": 22545 + }, + { + "epoch": 0.82, + "learning_rate": 1.176181931372908e-07, + "loss": 0.6745, + "step": 22546 + }, + { + "epoch": 0.82, + "learning_rate": 1.1761453107261873e-07, + "loss": 0.6892, + "step": 22547 + }, + { + "epoch": 0.82, + "learning_rate": 1.1761086900794667e-07, + "loss": 0.7013, + "step": 22548 + }, + { + "epoch": 0.83, + "learning_rate": 1.1760720694327461e-07, + "loss": 0.6984, + "step": 22549 + }, + { + "epoch": 0.83, + "learning_rate": 1.1760354487860255e-07, + "loss": 0.6918, + "step": 22550 + }, + { + "epoch": 0.83, + "learning_rate": 1.1759988281393049e-07, + "loss": 0.6939, + "step": 22551 + }, + { + "epoch": 0.83, + "learning_rate": 1.1759622074925843e-07, + "loss": 0.7063, + "step": 22552 + }, + { + "epoch": 0.83, + "learning_rate": 1.1759255868458637e-07, + "loss": 0.6943, + "step": 22553 + }, + { + "epoch": 0.83, + "learning_rate": 1.1758889661991431e-07, + "loss": 0.686, + "step": 22554 + }, + { + "epoch": 0.83, + "learning_rate": 1.1758523455524223e-07, + "loss": 0.6993, + "step": 22555 + }, + { + "epoch": 0.83, + "learning_rate": 1.1758157249057017e-07, + "loss": 0.6906, + "step": 22556 + }, + { + "epoch": 0.83, + "learning_rate": 1.175779104258981e-07, + "loss": 0.6912, + "step": 22557 + }, + { + "epoch": 0.83, + "learning_rate": 1.1757424836122605e-07, + "loss": 0.6912, + "step": 22558 + }, + { + "epoch": 0.83, + "learning_rate": 1.1757058629655399e-07, + "loss": 0.693, + "step": 22559 + }, + { + "epoch": 0.83, + "learning_rate": 1.1756692423188193e-07, + "loss": 0.6935, + "step": 22560 + }, + { + "epoch": 0.83, + "learning_rate": 1.1756326216720987e-07, + "loss": 0.69, + "step": 22561 + }, + { + "epoch": 0.83, + "learning_rate": 1.175596001025378e-07, + "loss": 0.684, + "step": 22562 + }, + { + "epoch": 0.83, + "learning_rate": 1.1755593803786575e-07, + "loss": 0.6863, + "step": 22563 + }, + { + "epoch": 0.83, + "learning_rate": 1.1755227597319368e-07, + "loss": 0.6964, + "step": 22564 + }, + { + "epoch": 0.83, + "learning_rate": 1.1754861390852162e-07, + "loss": 0.6951, + "step": 22565 + }, + { + "epoch": 0.83, + "learning_rate": 1.1754495184384956e-07, + "loss": 0.6915, + "step": 22566 + }, + { + "epoch": 0.83, + "learning_rate": 1.175412897791775e-07, + "loss": 0.6914, + "step": 22567 + }, + { + "epoch": 0.83, + "learning_rate": 1.1753762771450544e-07, + "loss": 0.6823, + "step": 22568 + }, + { + "epoch": 0.83, + "learning_rate": 1.1753396564983337e-07, + "loss": 0.7033, + "step": 22569 + }, + { + "epoch": 0.83, + "learning_rate": 1.1753030358516131e-07, + "loss": 0.7024, + "step": 22570 + }, + { + "epoch": 0.83, + "learning_rate": 1.1752664152048925e-07, + "loss": 0.6966, + "step": 22571 + }, + { + "epoch": 0.83, + "learning_rate": 1.1752297945581719e-07, + "loss": 0.7078, + "step": 22572 + }, + { + "epoch": 0.83, + "learning_rate": 1.1751931739114512e-07, + "loss": 0.6771, + "step": 22573 + }, + { + "epoch": 0.83, + "learning_rate": 1.1751565532647304e-07, + "loss": 0.6771, + "step": 22574 + }, + { + "epoch": 0.83, + "learning_rate": 1.1751199326180098e-07, + "loss": 0.7016, + "step": 22575 + }, + { + "epoch": 0.83, + "learning_rate": 1.1750833119712892e-07, + "loss": 0.6788, + "step": 22576 + }, + { + "epoch": 0.83, + "learning_rate": 1.1750466913245686e-07, + "loss": 0.7186, + "step": 22577 + }, + { + "epoch": 0.83, + "learning_rate": 1.175010070677848e-07, + "loss": 0.6925, + "step": 22578 + }, + { + "epoch": 0.83, + "learning_rate": 1.1749734500311274e-07, + "loss": 0.6832, + "step": 22579 + }, + { + "epoch": 0.83, + "learning_rate": 1.1749368293844068e-07, + "loss": 0.68, + "step": 22580 + }, + { + "epoch": 0.83, + "learning_rate": 1.1749002087376862e-07, + "loss": 0.6929, + "step": 22581 + }, + { + "epoch": 0.83, + "learning_rate": 1.1748635880909656e-07, + "loss": 0.7004, + "step": 22582 + }, + { + "epoch": 0.83, + "learning_rate": 1.174826967444245e-07, + "loss": 0.6942, + "step": 22583 + }, + { + "epoch": 0.83, + "learning_rate": 1.1747903467975244e-07, + "loss": 0.7025, + "step": 22584 + }, + { + "epoch": 0.83, + "learning_rate": 1.1747537261508038e-07, + "loss": 0.6903, + "step": 22585 + }, + { + "epoch": 0.83, + "learning_rate": 1.1747171055040832e-07, + "loss": 0.7053, + "step": 22586 + }, + { + "epoch": 0.83, + "learning_rate": 1.1746804848573626e-07, + "loss": 0.672, + "step": 22587 + }, + { + "epoch": 0.83, + "learning_rate": 1.174643864210642e-07, + "loss": 0.6962, + "step": 22588 + }, + { + "epoch": 0.83, + "learning_rate": 1.1746072435639214e-07, + "loss": 0.7142, + "step": 22589 + }, + { + "epoch": 0.83, + "learning_rate": 1.1745706229172008e-07, + "loss": 0.6971, + "step": 22590 + }, + { + "epoch": 0.83, + "learning_rate": 1.17453400227048e-07, + "loss": 0.6847, + "step": 22591 + }, + { + "epoch": 0.83, + "learning_rate": 1.1744973816237594e-07, + "loss": 0.6883, + "step": 22592 + }, + { + "epoch": 0.83, + "learning_rate": 1.1744607609770388e-07, + "loss": 0.6791, + "step": 22593 + }, + { + "epoch": 0.83, + "learning_rate": 1.1744241403303182e-07, + "loss": 0.6829, + "step": 22594 + }, + { + "epoch": 0.83, + "learning_rate": 1.1743875196835976e-07, + "loss": 0.6982, + "step": 22595 + }, + { + "epoch": 0.83, + "learning_rate": 1.1743508990368768e-07, + "loss": 0.6836, + "step": 22596 + }, + { + "epoch": 0.83, + "learning_rate": 1.1743142783901562e-07, + "loss": 0.6942, + "step": 22597 + }, + { + "epoch": 0.83, + "learning_rate": 1.1742776577434356e-07, + "loss": 0.7, + "step": 22598 + }, + { + "epoch": 0.83, + "learning_rate": 1.174241037096715e-07, + "loss": 0.695, + "step": 22599 + }, + { + "epoch": 0.83, + "learning_rate": 1.1742044164499944e-07, + "loss": 0.6939, + "step": 22600 + }, + { + "epoch": 0.83, + "learning_rate": 1.1741677958032738e-07, + "loss": 0.6846, + "step": 22601 + }, + { + "epoch": 0.83, + "learning_rate": 1.1741311751565532e-07, + "loss": 0.6933, + "step": 22602 + }, + { + "epoch": 0.83, + "learning_rate": 1.1740945545098326e-07, + "loss": 0.6855, + "step": 22603 + }, + { + "epoch": 0.83, + "learning_rate": 1.174057933863112e-07, + "loss": 0.6885, + "step": 22604 + }, + { + "epoch": 0.83, + "learning_rate": 1.1740213132163914e-07, + "loss": 0.6882, + "step": 22605 + }, + { + "epoch": 0.83, + "learning_rate": 1.1739846925696708e-07, + "loss": 0.687, + "step": 22606 + }, + { + "epoch": 0.83, + "learning_rate": 1.1739480719229502e-07, + "loss": 0.6997, + "step": 22607 + }, + { + "epoch": 0.83, + "learning_rate": 1.1739114512762296e-07, + "loss": 0.6843, + "step": 22608 + }, + { + "epoch": 0.83, + "learning_rate": 1.1738748306295088e-07, + "loss": 0.7086, + "step": 22609 + }, + { + "epoch": 0.83, + "learning_rate": 1.1738382099827882e-07, + "loss": 0.6918, + "step": 22610 + }, + { + "epoch": 0.83, + "learning_rate": 1.1738015893360676e-07, + "loss": 0.6989, + "step": 22611 + }, + { + "epoch": 0.83, + "learning_rate": 1.173764968689347e-07, + "loss": 0.7077, + "step": 22612 + }, + { + "epoch": 0.83, + "learning_rate": 1.1737283480426264e-07, + "loss": 0.7082, + "step": 22613 + }, + { + "epoch": 0.83, + "learning_rate": 1.1736917273959058e-07, + "loss": 0.6961, + "step": 22614 + }, + { + "epoch": 0.83, + "learning_rate": 1.1736551067491852e-07, + "loss": 0.6894, + "step": 22615 + }, + { + "epoch": 0.83, + "learning_rate": 1.1736184861024646e-07, + "loss": 0.6804, + "step": 22616 + }, + { + "epoch": 0.83, + "learning_rate": 1.173581865455744e-07, + "loss": 0.6841, + "step": 22617 + }, + { + "epoch": 0.83, + "learning_rate": 1.1735452448090232e-07, + "loss": 0.6968, + "step": 22618 + }, + { + "epoch": 0.83, + "learning_rate": 1.1735086241623026e-07, + "loss": 0.703, + "step": 22619 + }, + { + "epoch": 0.83, + "learning_rate": 1.173472003515582e-07, + "loss": 0.7042, + "step": 22620 + }, + { + "epoch": 0.83, + "learning_rate": 1.1734353828688614e-07, + "loss": 0.6974, + "step": 22621 + }, + { + "epoch": 0.83, + "learning_rate": 1.1733987622221408e-07, + "loss": 0.6906, + "step": 22622 + }, + { + "epoch": 0.83, + "learning_rate": 1.1733621415754202e-07, + "loss": 0.6917, + "step": 22623 + }, + { + "epoch": 0.83, + "learning_rate": 1.1733255209286996e-07, + "loss": 0.6901, + "step": 22624 + }, + { + "epoch": 0.83, + "learning_rate": 1.173288900281979e-07, + "loss": 0.7013, + "step": 22625 + }, + { + "epoch": 0.83, + "learning_rate": 1.1732522796352584e-07, + "loss": 0.695, + "step": 22626 + }, + { + "epoch": 0.83, + "learning_rate": 1.1732156589885375e-07, + "loss": 0.6879, + "step": 22627 + }, + { + "epoch": 0.83, + "learning_rate": 1.173179038341817e-07, + "loss": 0.6985, + "step": 22628 + }, + { + "epoch": 0.83, + "learning_rate": 1.1731424176950963e-07, + "loss": 0.6842, + "step": 22629 + }, + { + "epoch": 0.83, + "learning_rate": 1.1731057970483757e-07, + "loss": 0.6915, + "step": 22630 + }, + { + "epoch": 0.83, + "learning_rate": 1.1730691764016551e-07, + "loss": 0.6931, + "step": 22631 + }, + { + "epoch": 0.83, + "learning_rate": 1.1730325557549345e-07, + "loss": 0.6889, + "step": 22632 + }, + { + "epoch": 0.83, + "learning_rate": 1.172995935108214e-07, + "loss": 0.6895, + "step": 22633 + }, + { + "epoch": 0.83, + "learning_rate": 1.1729593144614933e-07, + "loss": 0.6825, + "step": 22634 + }, + { + "epoch": 0.83, + "learning_rate": 1.1729226938147727e-07, + "loss": 0.6954, + "step": 22635 + }, + { + "epoch": 0.83, + "learning_rate": 1.1728860731680521e-07, + "loss": 0.6998, + "step": 22636 + }, + { + "epoch": 0.83, + "learning_rate": 1.1728494525213315e-07, + "loss": 0.6886, + "step": 22637 + }, + { + "epoch": 0.83, + "learning_rate": 1.172812831874611e-07, + "loss": 0.7055, + "step": 22638 + }, + { + "epoch": 0.83, + "learning_rate": 1.1727762112278903e-07, + "loss": 0.7024, + "step": 22639 + }, + { + "epoch": 0.83, + "learning_rate": 1.1727395905811696e-07, + "loss": 0.6923, + "step": 22640 + }, + { + "epoch": 0.83, + "learning_rate": 1.172702969934449e-07, + "loss": 0.6893, + "step": 22641 + }, + { + "epoch": 0.83, + "learning_rate": 1.1726663492877284e-07, + "loss": 0.7041, + "step": 22642 + }, + { + "epoch": 0.83, + "learning_rate": 1.1726297286410078e-07, + "loss": 0.7041, + "step": 22643 + }, + { + "epoch": 0.83, + "learning_rate": 1.1725931079942872e-07, + "loss": 0.6752, + "step": 22644 + }, + { + "epoch": 0.83, + "learning_rate": 1.1725564873475663e-07, + "loss": 0.6933, + "step": 22645 + }, + { + "epoch": 0.83, + "learning_rate": 1.1725198667008457e-07, + "loss": 0.6931, + "step": 22646 + }, + { + "epoch": 0.83, + "learning_rate": 1.1724832460541251e-07, + "loss": 0.7008, + "step": 22647 + }, + { + "epoch": 0.83, + "learning_rate": 1.1724466254074045e-07, + "loss": 0.7011, + "step": 22648 + }, + { + "epoch": 0.83, + "learning_rate": 1.172410004760684e-07, + "loss": 0.6942, + "step": 22649 + }, + { + "epoch": 0.83, + "learning_rate": 1.1723733841139633e-07, + "loss": 0.7106, + "step": 22650 + }, + { + "epoch": 0.83, + "learning_rate": 1.1723367634672427e-07, + "loss": 0.685, + "step": 22651 + }, + { + "epoch": 0.83, + "learning_rate": 1.1723001428205221e-07, + "loss": 0.6972, + "step": 22652 + }, + { + "epoch": 0.83, + "learning_rate": 1.1722635221738015e-07, + "loss": 0.6955, + "step": 22653 + }, + { + "epoch": 0.83, + "learning_rate": 1.1722269015270809e-07, + "loss": 0.7095, + "step": 22654 + }, + { + "epoch": 0.83, + "learning_rate": 1.1721902808803603e-07, + "loss": 0.6986, + "step": 22655 + }, + { + "epoch": 0.83, + "learning_rate": 1.1721536602336397e-07, + "loss": 0.6922, + "step": 22656 + }, + { + "epoch": 0.83, + "learning_rate": 1.1721170395869191e-07, + "loss": 0.7027, + "step": 22657 + }, + { + "epoch": 0.83, + "learning_rate": 1.1720804189401985e-07, + "loss": 0.7006, + "step": 22658 + }, + { + "epoch": 0.83, + "learning_rate": 1.1720437982934779e-07, + "loss": 0.6845, + "step": 22659 + }, + { + "epoch": 0.83, + "learning_rate": 1.1720071776467573e-07, + "loss": 0.6803, + "step": 22660 + }, + { + "epoch": 0.83, + "learning_rate": 1.1719705570000367e-07, + "loss": 0.693, + "step": 22661 + }, + { + "epoch": 0.83, + "learning_rate": 1.171933936353316e-07, + "loss": 0.6922, + "step": 22662 + }, + { + "epoch": 0.83, + "learning_rate": 1.1718973157065953e-07, + "loss": 0.6878, + "step": 22663 + }, + { + "epoch": 0.83, + "learning_rate": 1.1718606950598747e-07, + "loss": 0.6559, + "step": 22664 + }, + { + "epoch": 0.83, + "learning_rate": 1.171824074413154e-07, + "loss": 0.6892, + "step": 22665 + }, + { + "epoch": 0.83, + "learning_rate": 1.1717874537664335e-07, + "loss": 0.7026, + "step": 22666 + }, + { + "epoch": 0.83, + "learning_rate": 1.1717508331197127e-07, + "loss": 0.6993, + "step": 22667 + }, + { + "epoch": 0.83, + "learning_rate": 1.1717142124729921e-07, + "loss": 0.6857, + "step": 22668 + }, + { + "epoch": 0.83, + "learning_rate": 1.1716775918262715e-07, + "loss": 0.6896, + "step": 22669 + }, + { + "epoch": 0.83, + "learning_rate": 1.1716409711795509e-07, + "loss": 0.6989, + "step": 22670 + }, + { + "epoch": 0.83, + "learning_rate": 1.1716043505328303e-07, + "loss": 0.695, + "step": 22671 + }, + { + "epoch": 0.83, + "learning_rate": 1.1715677298861097e-07, + "loss": 0.6704, + "step": 22672 + }, + { + "epoch": 0.83, + "learning_rate": 1.1715311092393891e-07, + "loss": 0.6905, + "step": 22673 + }, + { + "epoch": 0.83, + "learning_rate": 1.1714944885926685e-07, + "loss": 0.6876, + "step": 22674 + }, + { + "epoch": 0.83, + "learning_rate": 1.1714578679459479e-07, + "loss": 0.6996, + "step": 22675 + }, + { + "epoch": 0.83, + "learning_rate": 1.1714212472992273e-07, + "loss": 0.7162, + "step": 22676 + }, + { + "epoch": 0.83, + "learning_rate": 1.1713846266525067e-07, + "loss": 0.6896, + "step": 22677 + }, + { + "epoch": 0.83, + "learning_rate": 1.1713480060057861e-07, + "loss": 0.6966, + "step": 22678 + }, + { + "epoch": 0.83, + "learning_rate": 1.1713113853590655e-07, + "loss": 0.6944, + "step": 22679 + }, + { + "epoch": 0.83, + "learning_rate": 1.1712747647123449e-07, + "loss": 0.7065, + "step": 22680 + }, + { + "epoch": 0.83, + "learning_rate": 1.171238144065624e-07, + "loss": 0.7026, + "step": 22681 + }, + { + "epoch": 0.83, + "learning_rate": 1.1712015234189035e-07, + "loss": 0.6942, + "step": 22682 + }, + { + "epoch": 0.83, + "learning_rate": 1.1711649027721829e-07, + "loss": 0.6806, + "step": 22683 + }, + { + "epoch": 0.83, + "learning_rate": 1.1711282821254622e-07, + "loss": 0.7047, + "step": 22684 + }, + { + "epoch": 0.83, + "learning_rate": 1.1710916614787416e-07, + "loss": 0.7101, + "step": 22685 + }, + { + "epoch": 0.83, + "learning_rate": 1.171055040832021e-07, + "loss": 0.6852, + "step": 22686 + }, + { + "epoch": 0.83, + "learning_rate": 1.1710184201853004e-07, + "loss": 0.7013, + "step": 22687 + }, + { + "epoch": 0.83, + "learning_rate": 1.1709817995385798e-07, + "loss": 0.6804, + "step": 22688 + }, + { + "epoch": 0.83, + "learning_rate": 1.1709451788918591e-07, + "loss": 0.6952, + "step": 22689 + }, + { + "epoch": 0.83, + "learning_rate": 1.1709085582451385e-07, + "loss": 0.6883, + "step": 22690 + }, + { + "epoch": 0.83, + "learning_rate": 1.1708719375984179e-07, + "loss": 0.6936, + "step": 22691 + }, + { + "epoch": 0.83, + "learning_rate": 1.1708353169516973e-07, + "loss": 0.6875, + "step": 22692 + }, + { + "epoch": 0.83, + "learning_rate": 1.1707986963049767e-07, + "loss": 0.6928, + "step": 22693 + }, + { + "epoch": 0.83, + "learning_rate": 1.1707620756582561e-07, + "loss": 0.6997, + "step": 22694 + }, + { + "epoch": 0.83, + "learning_rate": 1.1707254550115355e-07, + "loss": 0.6854, + "step": 22695 + }, + { + "epoch": 0.83, + "learning_rate": 1.1706888343648149e-07, + "loss": 0.703, + "step": 22696 + }, + { + "epoch": 0.83, + "learning_rate": 1.1706522137180943e-07, + "loss": 0.7058, + "step": 22697 + }, + { + "epoch": 0.83, + "learning_rate": 1.1706155930713737e-07, + "loss": 0.6916, + "step": 22698 + }, + { + "epoch": 0.83, + "learning_rate": 1.1705789724246528e-07, + "loss": 0.6928, + "step": 22699 + }, + { + "epoch": 0.83, + "learning_rate": 1.1705423517779322e-07, + "loss": 0.6785, + "step": 22700 + }, + { + "epoch": 0.83, + "learning_rate": 1.1705057311312116e-07, + "loss": 0.6837, + "step": 22701 + }, + { + "epoch": 0.83, + "learning_rate": 1.170469110484491e-07, + "loss": 0.6944, + "step": 22702 + }, + { + "epoch": 0.83, + "learning_rate": 1.1704324898377704e-07, + "loss": 0.688, + "step": 22703 + }, + { + "epoch": 0.83, + "learning_rate": 1.1703958691910498e-07, + "loss": 0.6863, + "step": 22704 + }, + { + "epoch": 0.83, + "learning_rate": 1.1703592485443292e-07, + "loss": 0.6706, + "step": 22705 + }, + { + "epoch": 0.83, + "learning_rate": 1.1703226278976086e-07, + "loss": 0.6884, + "step": 22706 + }, + { + "epoch": 0.83, + "learning_rate": 1.170286007250888e-07, + "loss": 0.6986, + "step": 22707 + }, + { + "epoch": 0.83, + "learning_rate": 1.1702493866041674e-07, + "loss": 0.678, + "step": 22708 + }, + { + "epoch": 0.83, + "learning_rate": 1.1702127659574468e-07, + "loss": 0.6902, + "step": 22709 + }, + { + "epoch": 0.83, + "learning_rate": 1.1701761453107262e-07, + "loss": 0.6952, + "step": 22710 + }, + { + "epoch": 0.83, + "learning_rate": 1.1701395246640055e-07, + "loss": 0.7047, + "step": 22711 + }, + { + "epoch": 0.83, + "learning_rate": 1.1701029040172849e-07, + "loss": 0.698, + "step": 22712 + }, + { + "epoch": 0.83, + "learning_rate": 1.1700662833705643e-07, + "loss": 0.7007, + "step": 22713 + }, + { + "epoch": 0.83, + "learning_rate": 1.1700296627238437e-07, + "loss": 0.6888, + "step": 22714 + }, + { + "epoch": 0.83, + "learning_rate": 1.1699930420771231e-07, + "loss": 0.6947, + "step": 22715 + }, + { + "epoch": 0.83, + "learning_rate": 1.1699564214304025e-07, + "loss": 0.6797, + "step": 22716 + }, + { + "epoch": 0.83, + "learning_rate": 1.1699198007836816e-07, + "loss": 0.7107, + "step": 22717 + }, + { + "epoch": 0.83, + "learning_rate": 1.169883180136961e-07, + "loss": 0.6904, + "step": 22718 + }, + { + "epoch": 0.83, + "learning_rate": 1.1698465594902404e-07, + "loss": 0.6892, + "step": 22719 + }, + { + "epoch": 0.83, + "learning_rate": 1.1698099388435198e-07, + "loss": 0.6874, + "step": 22720 + }, + { + "epoch": 0.83, + "learning_rate": 1.1697733181967992e-07, + "loss": 0.6835, + "step": 22721 + }, + { + "epoch": 0.83, + "learning_rate": 1.1697366975500786e-07, + "loss": 0.7029, + "step": 22722 + }, + { + "epoch": 0.83, + "learning_rate": 1.169700076903358e-07, + "loss": 0.6896, + "step": 22723 + }, + { + "epoch": 0.83, + "learning_rate": 1.1696634562566374e-07, + "loss": 0.7043, + "step": 22724 + }, + { + "epoch": 0.83, + "learning_rate": 1.1696268356099168e-07, + "loss": 0.6934, + "step": 22725 + }, + { + "epoch": 0.83, + "learning_rate": 1.1695902149631962e-07, + "loss": 0.697, + "step": 22726 + }, + { + "epoch": 0.83, + "learning_rate": 1.1695535943164756e-07, + "loss": 0.6811, + "step": 22727 + }, + { + "epoch": 0.83, + "learning_rate": 1.169516973669755e-07, + "loss": 0.6982, + "step": 22728 + }, + { + "epoch": 0.83, + "learning_rate": 1.1694803530230344e-07, + "loss": 0.6836, + "step": 22729 + }, + { + "epoch": 0.83, + "learning_rate": 1.1694437323763138e-07, + "loss": 0.6871, + "step": 22730 + }, + { + "epoch": 0.83, + "learning_rate": 1.1694071117295932e-07, + "loss": 0.707, + "step": 22731 + }, + { + "epoch": 0.83, + "learning_rate": 1.1693704910828726e-07, + "loss": 0.6931, + "step": 22732 + }, + { + "epoch": 0.83, + "learning_rate": 1.1693338704361519e-07, + "loss": 0.6979, + "step": 22733 + }, + { + "epoch": 0.83, + "learning_rate": 1.1692972497894313e-07, + "loss": 0.6965, + "step": 22734 + }, + { + "epoch": 0.83, + "learning_rate": 1.1692606291427106e-07, + "loss": 0.7002, + "step": 22735 + }, + { + "epoch": 0.83, + "learning_rate": 1.16922400849599e-07, + "loss": 0.7093, + "step": 22736 + }, + { + "epoch": 0.83, + "learning_rate": 1.1691873878492694e-07, + "loss": 0.6837, + "step": 22737 + }, + { + "epoch": 0.83, + "learning_rate": 1.1691507672025486e-07, + "loss": 0.6846, + "step": 22738 + }, + { + "epoch": 0.83, + "learning_rate": 1.169114146555828e-07, + "loss": 0.6939, + "step": 22739 + }, + { + "epoch": 0.83, + "learning_rate": 1.1690775259091074e-07, + "loss": 0.6875, + "step": 22740 + }, + { + "epoch": 0.83, + "learning_rate": 1.1690409052623868e-07, + "loss": 0.709, + "step": 22741 + }, + { + "epoch": 0.83, + "learning_rate": 1.1690042846156662e-07, + "loss": 0.6954, + "step": 22742 + }, + { + "epoch": 0.83, + "learning_rate": 1.1689676639689456e-07, + "loss": 0.6886, + "step": 22743 + }, + { + "epoch": 0.83, + "learning_rate": 1.168931043322225e-07, + "loss": 0.6927, + "step": 22744 + }, + { + "epoch": 0.83, + "learning_rate": 1.1688944226755044e-07, + "loss": 0.7145, + "step": 22745 + }, + { + "epoch": 0.83, + "learning_rate": 1.1688578020287838e-07, + "loss": 0.6716, + "step": 22746 + }, + { + "epoch": 0.83, + "learning_rate": 1.1688211813820632e-07, + "loss": 0.6729, + "step": 22747 + }, + { + "epoch": 0.83, + "learning_rate": 1.1687845607353426e-07, + "loss": 0.6804, + "step": 22748 + }, + { + "epoch": 0.83, + "learning_rate": 1.168747940088622e-07, + "loss": 0.6882, + "step": 22749 + }, + { + "epoch": 0.83, + "learning_rate": 1.1687113194419014e-07, + "loss": 0.693, + "step": 22750 + }, + { + "epoch": 0.83, + "learning_rate": 1.1686746987951808e-07, + "loss": 0.6739, + "step": 22751 + }, + { + "epoch": 0.83, + "learning_rate": 1.16863807814846e-07, + "loss": 0.7072, + "step": 22752 + }, + { + "epoch": 0.83, + "learning_rate": 1.1686014575017393e-07, + "loss": 0.6817, + "step": 22753 + }, + { + "epoch": 0.83, + "learning_rate": 1.1685648368550187e-07, + "loss": 0.7005, + "step": 22754 + }, + { + "epoch": 0.83, + "learning_rate": 1.1685282162082981e-07, + "loss": 0.6843, + "step": 22755 + }, + { + "epoch": 0.83, + "learning_rate": 1.1684915955615775e-07, + "loss": 0.6987, + "step": 22756 + }, + { + "epoch": 0.83, + "learning_rate": 1.168454974914857e-07, + "loss": 0.6931, + "step": 22757 + }, + { + "epoch": 0.83, + "learning_rate": 1.1684183542681363e-07, + "loss": 0.698, + "step": 22758 + }, + { + "epoch": 0.83, + "learning_rate": 1.1683817336214157e-07, + "loss": 0.6811, + "step": 22759 + }, + { + "epoch": 0.83, + "learning_rate": 1.168345112974695e-07, + "loss": 0.7097, + "step": 22760 + }, + { + "epoch": 0.83, + "learning_rate": 1.1683084923279744e-07, + "loss": 0.7004, + "step": 22761 + }, + { + "epoch": 0.83, + "learning_rate": 1.1682718716812538e-07, + "loss": 0.6892, + "step": 22762 + }, + { + "epoch": 0.83, + "learning_rate": 1.1682352510345332e-07, + "loss": 0.6847, + "step": 22763 + }, + { + "epoch": 0.83, + "learning_rate": 1.1681986303878126e-07, + "loss": 0.7, + "step": 22764 + }, + { + "epoch": 0.83, + "learning_rate": 1.168162009741092e-07, + "loss": 0.6929, + "step": 22765 + }, + { + "epoch": 0.83, + "learning_rate": 1.1681253890943714e-07, + "loss": 0.6928, + "step": 22766 + }, + { + "epoch": 0.83, + "learning_rate": 1.1680887684476508e-07, + "loss": 0.6816, + "step": 22767 + }, + { + "epoch": 0.83, + "learning_rate": 1.1680521478009302e-07, + "loss": 0.7071, + "step": 22768 + }, + { + "epoch": 0.83, + "learning_rate": 1.1680155271542096e-07, + "loss": 0.7083, + "step": 22769 + }, + { + "epoch": 0.83, + "learning_rate": 1.1679789065074887e-07, + "loss": 0.6914, + "step": 22770 + }, + { + "epoch": 0.83, + "learning_rate": 1.1679422858607681e-07, + "loss": 0.7043, + "step": 22771 + }, + { + "epoch": 0.83, + "learning_rate": 1.1679056652140475e-07, + "loss": 0.6997, + "step": 22772 + }, + { + "epoch": 0.83, + "learning_rate": 1.1678690445673269e-07, + "loss": 0.6995, + "step": 22773 + }, + { + "epoch": 0.83, + "learning_rate": 1.1678324239206063e-07, + "loss": 0.697, + "step": 22774 + }, + { + "epoch": 0.83, + "learning_rate": 1.1677958032738857e-07, + "loss": 0.6875, + "step": 22775 + }, + { + "epoch": 0.83, + "learning_rate": 1.1677591826271651e-07, + "loss": 0.6888, + "step": 22776 + }, + { + "epoch": 0.83, + "learning_rate": 1.1677225619804445e-07, + "loss": 0.7092, + "step": 22777 + }, + { + "epoch": 0.83, + "learning_rate": 1.1676859413337239e-07, + "loss": 0.6821, + "step": 22778 + }, + { + "epoch": 0.83, + "learning_rate": 1.1676493206870033e-07, + "loss": 0.699, + "step": 22779 + }, + { + "epoch": 0.83, + "learning_rate": 1.1676127000402827e-07, + "loss": 0.6864, + "step": 22780 + }, + { + "epoch": 0.83, + "learning_rate": 1.1675760793935621e-07, + "loss": 0.708, + "step": 22781 + }, + { + "epoch": 0.83, + "learning_rate": 1.1675394587468414e-07, + "loss": 0.7094, + "step": 22782 + }, + { + "epoch": 0.83, + "learning_rate": 1.1675028381001208e-07, + "loss": 0.6922, + "step": 22783 + }, + { + "epoch": 0.83, + "learning_rate": 1.1674662174534002e-07, + "loss": 0.6667, + "step": 22784 + }, + { + "epoch": 0.83, + "learning_rate": 1.1674295968066796e-07, + "loss": 0.6841, + "step": 22785 + }, + { + "epoch": 0.83, + "learning_rate": 1.167392976159959e-07, + "loss": 0.6753, + "step": 22786 + }, + { + "epoch": 0.83, + "learning_rate": 1.1673563555132384e-07, + "loss": 0.681, + "step": 22787 + }, + { + "epoch": 0.83, + "learning_rate": 1.1673197348665175e-07, + "loss": 0.6944, + "step": 22788 + }, + { + "epoch": 0.83, + "learning_rate": 1.1672831142197969e-07, + "loss": 0.6883, + "step": 22789 + }, + { + "epoch": 0.83, + "learning_rate": 1.1672464935730763e-07, + "loss": 0.7082, + "step": 22790 + }, + { + "epoch": 0.83, + "learning_rate": 1.1672098729263557e-07, + "loss": 0.6899, + "step": 22791 + }, + { + "epoch": 0.83, + "learning_rate": 1.1671732522796351e-07, + "loss": 0.6975, + "step": 22792 + }, + { + "epoch": 0.83, + "learning_rate": 1.1671366316329145e-07, + "loss": 0.6879, + "step": 22793 + }, + { + "epoch": 0.83, + "learning_rate": 1.1671000109861939e-07, + "loss": 0.7069, + "step": 22794 + }, + { + "epoch": 0.83, + "learning_rate": 1.1670633903394733e-07, + "loss": 0.6706, + "step": 22795 + }, + { + "epoch": 0.83, + "learning_rate": 1.1670267696927527e-07, + "loss": 0.6859, + "step": 22796 + }, + { + "epoch": 0.83, + "learning_rate": 1.1669901490460321e-07, + "loss": 0.6844, + "step": 22797 + }, + { + "epoch": 0.83, + "learning_rate": 1.1669535283993115e-07, + "loss": 0.6888, + "step": 22798 + }, + { + "epoch": 0.83, + "learning_rate": 1.1669169077525909e-07, + "loss": 0.6949, + "step": 22799 + }, + { + "epoch": 0.83, + "learning_rate": 1.1668802871058703e-07, + "loss": 0.692, + "step": 22800 + }, + { + "epoch": 0.83, + "learning_rate": 1.1668436664591497e-07, + "loss": 0.698, + "step": 22801 + }, + { + "epoch": 0.83, + "learning_rate": 1.1668070458124291e-07, + "loss": 0.6859, + "step": 22802 + }, + { + "epoch": 0.83, + "learning_rate": 1.1667704251657085e-07, + "loss": 0.6984, + "step": 22803 + }, + { + "epoch": 0.83, + "learning_rate": 1.1667338045189878e-07, + "loss": 0.6982, + "step": 22804 + }, + { + "epoch": 0.83, + "learning_rate": 1.1666971838722672e-07, + "loss": 0.6764, + "step": 22805 + }, + { + "epoch": 0.83, + "learning_rate": 1.1666605632255464e-07, + "loss": 0.6946, + "step": 22806 + }, + { + "epoch": 0.83, + "learning_rate": 1.1666239425788258e-07, + "loss": 0.6713, + "step": 22807 + }, + { + "epoch": 0.83, + "learning_rate": 1.1665873219321052e-07, + "loss": 0.7016, + "step": 22808 + }, + { + "epoch": 0.83, + "learning_rate": 1.1665507012853845e-07, + "loss": 0.6991, + "step": 22809 + }, + { + "epoch": 0.83, + "learning_rate": 1.1665140806386639e-07, + "loss": 0.6953, + "step": 22810 + }, + { + "epoch": 0.83, + "learning_rate": 1.1664774599919433e-07, + "loss": 0.6788, + "step": 22811 + }, + { + "epoch": 0.83, + "learning_rate": 1.1664408393452227e-07, + "loss": 0.6975, + "step": 22812 + }, + { + "epoch": 0.83, + "learning_rate": 1.1664042186985021e-07, + "loss": 0.6738, + "step": 22813 + }, + { + "epoch": 0.83, + "learning_rate": 1.1663675980517815e-07, + "loss": 0.6866, + "step": 22814 + }, + { + "epoch": 0.83, + "learning_rate": 1.1663309774050609e-07, + "loss": 0.6929, + "step": 22815 + }, + { + "epoch": 0.83, + "learning_rate": 1.1662943567583403e-07, + "loss": 0.6968, + "step": 22816 + }, + { + "epoch": 0.83, + "learning_rate": 1.1662577361116197e-07, + "loss": 0.6865, + "step": 22817 + }, + { + "epoch": 0.83, + "learning_rate": 1.1662211154648991e-07, + "loss": 0.683, + "step": 22818 + }, + { + "epoch": 0.83, + "learning_rate": 1.1661844948181785e-07, + "loss": 0.6992, + "step": 22819 + }, + { + "epoch": 0.83, + "learning_rate": 1.1661478741714579e-07, + "loss": 0.7008, + "step": 22820 + }, + { + "epoch": 0.83, + "learning_rate": 1.1661112535247373e-07, + "loss": 0.7084, + "step": 22821 + }, + { + "epoch": 0.83, + "learning_rate": 1.1660746328780167e-07, + "loss": 0.7078, + "step": 22822 + }, + { + "epoch": 0.84, + "learning_rate": 1.1660380122312961e-07, + "loss": 0.6906, + "step": 22823 + }, + { + "epoch": 0.84, + "learning_rate": 1.1660013915845752e-07, + "loss": 0.677, + "step": 22824 + }, + { + "epoch": 0.84, + "learning_rate": 1.1659647709378546e-07, + "loss": 0.7072, + "step": 22825 + }, + { + "epoch": 0.84, + "learning_rate": 1.165928150291134e-07, + "loss": 0.679, + "step": 22826 + }, + { + "epoch": 0.84, + "learning_rate": 1.1658915296444134e-07, + "loss": 0.6897, + "step": 22827 + }, + { + "epoch": 0.84, + "learning_rate": 1.1658549089976928e-07, + "loss": 0.6878, + "step": 22828 + }, + { + "epoch": 0.84, + "learning_rate": 1.1658182883509722e-07, + "loss": 0.6973, + "step": 22829 + }, + { + "epoch": 0.84, + "learning_rate": 1.1657816677042516e-07, + "loss": 0.688, + "step": 22830 + }, + { + "epoch": 0.84, + "learning_rate": 1.1657450470575309e-07, + "loss": 0.685, + "step": 22831 + }, + { + "epoch": 0.84, + "learning_rate": 1.1657084264108103e-07, + "loss": 0.6866, + "step": 22832 + }, + { + "epoch": 0.84, + "learning_rate": 1.1656718057640897e-07, + "loss": 0.6972, + "step": 22833 + }, + { + "epoch": 0.84, + "learning_rate": 1.1656351851173691e-07, + "loss": 0.6934, + "step": 22834 + }, + { + "epoch": 0.84, + "learning_rate": 1.1655985644706485e-07, + "loss": 0.6856, + "step": 22835 + }, + { + "epoch": 0.84, + "learning_rate": 1.1655619438239279e-07, + "loss": 0.6881, + "step": 22836 + }, + { + "epoch": 0.84, + "learning_rate": 1.1655253231772073e-07, + "loss": 0.6896, + "step": 22837 + }, + { + "epoch": 0.84, + "learning_rate": 1.1654887025304867e-07, + "loss": 0.6988, + "step": 22838 + }, + { + "epoch": 0.84, + "learning_rate": 1.1654520818837661e-07, + "loss": 0.6928, + "step": 22839 + }, + { + "epoch": 0.84, + "learning_rate": 1.1654154612370455e-07, + "loss": 0.6945, + "step": 22840 + }, + { + "epoch": 0.84, + "learning_rate": 1.1653788405903249e-07, + "loss": 0.6722, + "step": 22841 + }, + { + "epoch": 0.84, + "learning_rate": 1.165342219943604e-07, + "loss": 0.6717, + "step": 22842 + }, + { + "epoch": 0.84, + "learning_rate": 1.1653055992968834e-07, + "loss": 0.6959, + "step": 22843 + }, + { + "epoch": 0.84, + "learning_rate": 1.1652689786501628e-07, + "loss": 0.7003, + "step": 22844 + }, + { + "epoch": 0.84, + "learning_rate": 1.1652323580034422e-07, + "loss": 0.691, + "step": 22845 + }, + { + "epoch": 0.84, + "learning_rate": 1.1651957373567216e-07, + "loss": 0.7032, + "step": 22846 + }, + { + "epoch": 0.84, + "learning_rate": 1.165159116710001e-07, + "loss": 0.7049, + "step": 22847 + }, + { + "epoch": 0.84, + "learning_rate": 1.1651224960632804e-07, + "loss": 0.6745, + "step": 22848 + }, + { + "epoch": 0.84, + "learning_rate": 1.1650858754165598e-07, + "loss": 0.6925, + "step": 22849 + }, + { + "epoch": 0.84, + "learning_rate": 1.1650492547698392e-07, + "loss": 0.6981, + "step": 22850 + }, + { + "epoch": 0.84, + "learning_rate": 1.1650126341231186e-07, + "loss": 0.6857, + "step": 22851 + }, + { + "epoch": 0.84, + "learning_rate": 1.164976013476398e-07, + "loss": 0.7022, + "step": 22852 + }, + { + "epoch": 0.84, + "learning_rate": 1.1649393928296773e-07, + "loss": 0.6955, + "step": 22853 + }, + { + "epoch": 0.84, + "learning_rate": 1.1649027721829567e-07, + "loss": 0.7101, + "step": 22854 + }, + { + "epoch": 0.84, + "learning_rate": 1.1648661515362361e-07, + "loss": 0.708, + "step": 22855 + }, + { + "epoch": 0.84, + "learning_rate": 1.1648295308895155e-07, + "loss": 0.6922, + "step": 22856 + }, + { + "epoch": 0.84, + "learning_rate": 1.1647929102427949e-07, + "loss": 0.6882, + "step": 22857 + }, + { + "epoch": 0.84, + "learning_rate": 1.1647562895960743e-07, + "loss": 0.6954, + "step": 22858 + }, + { + "epoch": 0.84, + "learning_rate": 1.1647196689493537e-07, + "loss": 0.6836, + "step": 22859 + }, + { + "epoch": 0.84, + "learning_rate": 1.1646830483026328e-07, + "loss": 0.688, + "step": 22860 + }, + { + "epoch": 0.84, + "learning_rate": 1.1646464276559122e-07, + "loss": 0.7098, + "step": 22861 + }, + { + "epoch": 0.84, + "learning_rate": 1.1646098070091916e-07, + "loss": 0.7094, + "step": 22862 + }, + { + "epoch": 0.84, + "learning_rate": 1.164573186362471e-07, + "loss": 0.6846, + "step": 22863 + }, + { + "epoch": 0.84, + "learning_rate": 1.1645365657157504e-07, + "loss": 0.7, + "step": 22864 + }, + { + "epoch": 0.84, + "learning_rate": 1.1644999450690298e-07, + "loss": 0.6881, + "step": 22865 + }, + { + "epoch": 0.84, + "learning_rate": 1.1644633244223092e-07, + "loss": 0.6886, + "step": 22866 + }, + { + "epoch": 0.84, + "learning_rate": 1.1644267037755886e-07, + "loss": 0.6915, + "step": 22867 + }, + { + "epoch": 0.84, + "learning_rate": 1.164390083128868e-07, + "loss": 0.6924, + "step": 22868 + }, + { + "epoch": 0.84, + "learning_rate": 1.1643534624821474e-07, + "loss": 0.6992, + "step": 22869 + }, + { + "epoch": 0.84, + "learning_rate": 1.1643168418354268e-07, + "loss": 0.7064, + "step": 22870 + }, + { + "epoch": 0.84, + "learning_rate": 1.1642802211887062e-07, + "loss": 0.707, + "step": 22871 + }, + { + "epoch": 0.84, + "learning_rate": 1.1642436005419856e-07, + "loss": 0.6798, + "step": 22872 + }, + { + "epoch": 0.84, + "learning_rate": 1.164206979895265e-07, + "loss": 0.7076, + "step": 22873 + }, + { + "epoch": 0.84, + "learning_rate": 1.1641703592485444e-07, + "loss": 0.6824, + "step": 22874 + }, + { + "epoch": 0.84, + "learning_rate": 1.1641337386018237e-07, + "loss": 0.6852, + "step": 22875 + }, + { + "epoch": 0.84, + "learning_rate": 1.1640971179551031e-07, + "loss": 0.6876, + "step": 22876 + }, + { + "epoch": 0.84, + "learning_rate": 1.1640604973083825e-07, + "loss": 0.6953, + "step": 22877 + }, + { + "epoch": 0.84, + "learning_rate": 1.1640238766616617e-07, + "loss": 0.6969, + "step": 22878 + }, + { + "epoch": 0.84, + "learning_rate": 1.1639872560149411e-07, + "loss": 0.7069, + "step": 22879 + }, + { + "epoch": 0.84, + "learning_rate": 1.1639506353682204e-07, + "loss": 0.6919, + "step": 22880 + }, + { + "epoch": 0.84, + "learning_rate": 1.1639140147214998e-07, + "loss": 0.7126, + "step": 22881 + }, + { + "epoch": 0.84, + "learning_rate": 1.1638773940747792e-07, + "loss": 0.7004, + "step": 22882 + }, + { + "epoch": 0.84, + "learning_rate": 1.1638407734280586e-07, + "loss": 0.6998, + "step": 22883 + }, + { + "epoch": 0.84, + "learning_rate": 1.163804152781338e-07, + "loss": 0.699, + "step": 22884 + }, + { + "epoch": 0.84, + "learning_rate": 1.1637675321346174e-07, + "loss": 0.6839, + "step": 22885 + }, + { + "epoch": 0.84, + "learning_rate": 1.1637309114878968e-07, + "loss": 0.7016, + "step": 22886 + }, + { + "epoch": 0.84, + "learning_rate": 1.1636942908411762e-07, + "loss": 0.6829, + "step": 22887 + }, + { + "epoch": 0.84, + "learning_rate": 1.1636576701944556e-07, + "loss": 0.6948, + "step": 22888 + }, + { + "epoch": 0.84, + "learning_rate": 1.163621049547735e-07, + "loss": 0.7004, + "step": 22889 + }, + { + "epoch": 0.84, + "learning_rate": 1.1635844289010144e-07, + "loss": 0.6821, + "step": 22890 + }, + { + "epoch": 0.84, + "learning_rate": 1.1635478082542938e-07, + "loss": 0.6979, + "step": 22891 + }, + { + "epoch": 0.84, + "learning_rate": 1.1635111876075732e-07, + "loss": 0.7162, + "step": 22892 + }, + { + "epoch": 0.84, + "learning_rate": 1.1634745669608526e-07, + "loss": 0.6908, + "step": 22893 + }, + { + "epoch": 0.84, + "learning_rate": 1.163437946314132e-07, + "loss": 0.6831, + "step": 22894 + }, + { + "epoch": 0.84, + "learning_rate": 1.1634013256674114e-07, + "loss": 0.6878, + "step": 22895 + }, + { + "epoch": 0.84, + "learning_rate": 1.1633647050206905e-07, + "loss": 0.6792, + "step": 22896 + }, + { + "epoch": 0.84, + "learning_rate": 1.1633280843739699e-07, + "loss": 0.6934, + "step": 22897 + }, + { + "epoch": 0.84, + "learning_rate": 1.1632914637272493e-07, + "loss": 0.6888, + "step": 22898 + }, + { + "epoch": 0.84, + "learning_rate": 1.1632548430805287e-07, + "loss": 0.6871, + "step": 22899 + }, + { + "epoch": 0.84, + "learning_rate": 1.1632182224338081e-07, + "loss": 0.7023, + "step": 22900 + }, + { + "epoch": 0.84, + "learning_rate": 1.1631816017870875e-07, + "loss": 0.7012, + "step": 22901 + }, + { + "epoch": 0.84, + "learning_rate": 1.1631449811403668e-07, + "loss": 0.6995, + "step": 22902 + }, + { + "epoch": 0.84, + "learning_rate": 1.1631083604936462e-07, + "loss": 0.6983, + "step": 22903 + }, + { + "epoch": 0.84, + "learning_rate": 1.1630717398469256e-07, + "loss": 0.6957, + "step": 22904 + }, + { + "epoch": 0.84, + "learning_rate": 1.163035119200205e-07, + "loss": 0.6882, + "step": 22905 + }, + { + "epoch": 0.84, + "learning_rate": 1.1629984985534844e-07, + "loss": 0.6884, + "step": 22906 + }, + { + "epoch": 0.84, + "learning_rate": 1.1629618779067638e-07, + "loss": 0.686, + "step": 22907 + }, + { + "epoch": 0.84, + "learning_rate": 1.1629252572600432e-07, + "loss": 0.7124, + "step": 22908 + }, + { + "epoch": 0.84, + "learning_rate": 1.1628886366133226e-07, + "loss": 0.7075, + "step": 22909 + }, + { + "epoch": 0.84, + "learning_rate": 1.162852015966602e-07, + "loss": 0.7014, + "step": 22910 + }, + { + "epoch": 0.84, + "learning_rate": 1.1628153953198814e-07, + "loss": 0.6924, + "step": 22911 + }, + { + "epoch": 0.84, + "learning_rate": 1.1627787746731608e-07, + "loss": 0.6664, + "step": 22912 + }, + { + "epoch": 0.84, + "learning_rate": 1.1627421540264402e-07, + "loss": 0.688, + "step": 22913 + }, + { + "epoch": 0.84, + "learning_rate": 1.1627055333797193e-07, + "loss": 0.6859, + "step": 22914 + }, + { + "epoch": 0.84, + "learning_rate": 1.1626689127329987e-07, + "loss": 0.6981, + "step": 22915 + }, + { + "epoch": 0.84, + "learning_rate": 1.1626322920862781e-07, + "loss": 0.6947, + "step": 22916 + }, + { + "epoch": 0.84, + "learning_rate": 1.1625956714395575e-07, + "loss": 0.6969, + "step": 22917 + }, + { + "epoch": 0.84, + "learning_rate": 1.1625590507928369e-07, + "loss": 0.6852, + "step": 22918 + }, + { + "epoch": 0.84, + "learning_rate": 1.1625224301461163e-07, + "loss": 0.6894, + "step": 22919 + }, + { + "epoch": 0.84, + "learning_rate": 1.1624858094993957e-07, + "loss": 0.6987, + "step": 22920 + }, + { + "epoch": 0.84, + "learning_rate": 1.1624491888526751e-07, + "loss": 0.6863, + "step": 22921 + }, + { + "epoch": 0.84, + "learning_rate": 1.1624125682059545e-07, + "loss": 0.6815, + "step": 22922 + }, + { + "epoch": 0.84, + "learning_rate": 1.1623759475592339e-07, + "loss": 0.6819, + "step": 22923 + }, + { + "epoch": 0.84, + "learning_rate": 1.1623393269125132e-07, + "loss": 0.6981, + "step": 22924 + }, + { + "epoch": 0.84, + "learning_rate": 1.1623027062657926e-07, + "loss": 0.691, + "step": 22925 + }, + { + "epoch": 0.84, + "learning_rate": 1.162266085619072e-07, + "loss": 0.6927, + "step": 22926 + }, + { + "epoch": 0.84, + "learning_rate": 1.1622294649723514e-07, + "loss": 0.698, + "step": 22927 + }, + { + "epoch": 0.84, + "learning_rate": 1.1621928443256308e-07, + "loss": 0.6899, + "step": 22928 + }, + { + "epoch": 0.84, + "learning_rate": 1.1621562236789102e-07, + "loss": 0.6913, + "step": 22929 + }, + { + "epoch": 0.84, + "learning_rate": 1.1621196030321896e-07, + "loss": 0.7038, + "step": 22930 + }, + { + "epoch": 0.84, + "learning_rate": 1.162082982385469e-07, + "loss": 0.6975, + "step": 22931 + }, + { + "epoch": 0.84, + "learning_rate": 1.1620463617387481e-07, + "loss": 0.6898, + "step": 22932 + }, + { + "epoch": 0.84, + "learning_rate": 1.1620097410920275e-07, + "loss": 0.6909, + "step": 22933 + }, + { + "epoch": 0.84, + "learning_rate": 1.1619731204453069e-07, + "loss": 0.6888, + "step": 22934 + }, + { + "epoch": 0.84, + "learning_rate": 1.1619364997985863e-07, + "loss": 0.6862, + "step": 22935 + }, + { + "epoch": 0.84, + "learning_rate": 1.1618998791518657e-07, + "loss": 0.7077, + "step": 22936 + }, + { + "epoch": 0.84, + "learning_rate": 1.1618632585051451e-07, + "loss": 0.7004, + "step": 22937 + }, + { + "epoch": 0.84, + "learning_rate": 1.1618266378584245e-07, + "loss": 0.6868, + "step": 22938 + }, + { + "epoch": 0.84, + "learning_rate": 1.1617900172117039e-07, + "loss": 0.716, + "step": 22939 + }, + { + "epoch": 0.84, + "learning_rate": 1.1617533965649833e-07, + "loss": 0.6788, + "step": 22940 + }, + { + "epoch": 0.84, + "learning_rate": 1.1617167759182627e-07, + "loss": 0.6879, + "step": 22941 + }, + { + "epoch": 0.84, + "learning_rate": 1.1616801552715421e-07, + "loss": 0.6931, + "step": 22942 + }, + { + "epoch": 0.84, + "learning_rate": 1.1616435346248215e-07, + "loss": 0.6971, + "step": 22943 + }, + { + "epoch": 0.84, + "learning_rate": 1.1616069139781009e-07, + "loss": 0.693, + "step": 22944 + }, + { + "epoch": 0.84, + "learning_rate": 1.1615702933313803e-07, + "loss": 0.6939, + "step": 22945 + }, + { + "epoch": 0.84, + "learning_rate": 1.1615336726846596e-07, + "loss": 0.6869, + "step": 22946 + }, + { + "epoch": 0.84, + "learning_rate": 1.161497052037939e-07, + "loss": 0.6913, + "step": 22947 + }, + { + "epoch": 0.84, + "learning_rate": 1.1614604313912184e-07, + "loss": 0.7081, + "step": 22948 + }, + { + "epoch": 0.84, + "learning_rate": 1.1614238107444978e-07, + "loss": 0.6878, + "step": 22949 + }, + { + "epoch": 0.84, + "learning_rate": 1.161387190097777e-07, + "loss": 0.6727, + "step": 22950 + }, + { + "epoch": 0.84, + "learning_rate": 1.1613505694510563e-07, + "loss": 0.6922, + "step": 22951 + }, + { + "epoch": 0.84, + "learning_rate": 1.1613139488043357e-07, + "loss": 0.6963, + "step": 22952 + }, + { + "epoch": 0.84, + "learning_rate": 1.1612773281576151e-07, + "loss": 0.6955, + "step": 22953 + }, + { + "epoch": 0.84, + "learning_rate": 1.1612407075108945e-07, + "loss": 0.7041, + "step": 22954 + }, + { + "epoch": 0.84, + "learning_rate": 1.1612040868641739e-07, + "loss": 0.7045, + "step": 22955 + }, + { + "epoch": 0.84, + "learning_rate": 1.1611674662174533e-07, + "loss": 0.6979, + "step": 22956 + }, + { + "epoch": 0.84, + "learning_rate": 1.1611308455707327e-07, + "loss": 0.7012, + "step": 22957 + }, + { + "epoch": 0.84, + "learning_rate": 1.1610942249240121e-07, + "loss": 0.6912, + "step": 22958 + }, + { + "epoch": 0.84, + "learning_rate": 1.1610576042772915e-07, + "loss": 0.7098, + "step": 22959 + }, + { + "epoch": 0.84, + "learning_rate": 1.1610209836305709e-07, + "loss": 0.695, + "step": 22960 + }, + { + "epoch": 0.84, + "learning_rate": 1.1609843629838503e-07, + "loss": 0.6926, + "step": 22961 + }, + { + "epoch": 0.84, + "learning_rate": 1.1609477423371297e-07, + "loss": 0.6951, + "step": 22962 + }, + { + "epoch": 0.84, + "learning_rate": 1.1609111216904091e-07, + "loss": 0.6899, + "step": 22963 + }, + { + "epoch": 0.84, + "learning_rate": 1.1608745010436885e-07, + "loss": 0.6998, + "step": 22964 + }, + { + "epoch": 0.84, + "learning_rate": 1.1608378803969679e-07, + "loss": 0.695, + "step": 22965 + }, + { + "epoch": 0.84, + "learning_rate": 1.1608012597502473e-07, + "loss": 0.6984, + "step": 22966 + }, + { + "epoch": 0.84, + "learning_rate": 1.1607646391035266e-07, + "loss": 0.6953, + "step": 22967 + }, + { + "epoch": 0.84, + "learning_rate": 1.1607280184568058e-07, + "loss": 0.7, + "step": 22968 + }, + { + "epoch": 0.84, + "learning_rate": 1.1606913978100852e-07, + "loss": 0.6674, + "step": 22969 + }, + { + "epoch": 0.84, + "learning_rate": 1.1606547771633646e-07, + "loss": 0.6903, + "step": 22970 + }, + { + "epoch": 0.84, + "learning_rate": 1.160618156516644e-07, + "loss": 0.6968, + "step": 22971 + }, + { + "epoch": 0.84, + "learning_rate": 1.1605815358699234e-07, + "loss": 0.7028, + "step": 22972 + }, + { + "epoch": 0.84, + "learning_rate": 1.1605449152232027e-07, + "loss": 0.7027, + "step": 22973 + }, + { + "epoch": 0.84, + "learning_rate": 1.1605082945764821e-07, + "loss": 0.6916, + "step": 22974 + }, + { + "epoch": 0.84, + "learning_rate": 1.1604716739297615e-07, + "loss": 0.7008, + "step": 22975 + }, + { + "epoch": 0.84, + "learning_rate": 1.1604350532830409e-07, + "loss": 0.6958, + "step": 22976 + }, + { + "epoch": 0.84, + "learning_rate": 1.1603984326363203e-07, + "loss": 0.6946, + "step": 22977 + }, + { + "epoch": 0.84, + "learning_rate": 1.1603618119895997e-07, + "loss": 0.6949, + "step": 22978 + }, + { + "epoch": 0.84, + "learning_rate": 1.1603251913428791e-07, + "loss": 0.6869, + "step": 22979 + }, + { + "epoch": 0.84, + "learning_rate": 1.1602885706961585e-07, + "loss": 0.6939, + "step": 22980 + }, + { + "epoch": 0.84, + "learning_rate": 1.1602519500494379e-07, + "loss": 0.6892, + "step": 22981 + }, + { + "epoch": 0.84, + "learning_rate": 1.1602153294027173e-07, + "loss": 0.6871, + "step": 22982 + }, + { + "epoch": 0.84, + "learning_rate": 1.1601787087559967e-07, + "loss": 0.7026, + "step": 22983 + }, + { + "epoch": 0.84, + "learning_rate": 1.1601420881092761e-07, + "loss": 0.6861, + "step": 22984 + }, + { + "epoch": 0.84, + "learning_rate": 1.1601054674625555e-07, + "loss": 0.6866, + "step": 22985 + }, + { + "epoch": 0.84, + "learning_rate": 1.1600688468158346e-07, + "loss": 0.6929, + "step": 22986 + }, + { + "epoch": 0.84, + "learning_rate": 1.160032226169114e-07, + "loss": 0.7012, + "step": 22987 + }, + { + "epoch": 0.84, + "learning_rate": 1.1599956055223934e-07, + "loss": 0.6856, + "step": 22988 + }, + { + "epoch": 0.84, + "learning_rate": 1.1599589848756728e-07, + "loss": 0.6902, + "step": 22989 + }, + { + "epoch": 0.84, + "learning_rate": 1.1599223642289522e-07, + "loss": 0.6876, + "step": 22990 + }, + { + "epoch": 0.84, + "learning_rate": 1.1598857435822316e-07, + "loss": 0.7012, + "step": 22991 + }, + { + "epoch": 0.84, + "learning_rate": 1.159849122935511e-07, + "loss": 0.693, + "step": 22992 + }, + { + "epoch": 0.84, + "learning_rate": 1.1598125022887904e-07, + "loss": 0.6993, + "step": 22993 + }, + { + "epoch": 0.84, + "learning_rate": 1.1597758816420698e-07, + "loss": 0.7002, + "step": 22994 + }, + { + "epoch": 0.84, + "learning_rate": 1.1597392609953491e-07, + "loss": 0.6947, + "step": 22995 + }, + { + "epoch": 0.84, + "learning_rate": 1.1597026403486285e-07, + "loss": 0.6966, + "step": 22996 + }, + { + "epoch": 0.84, + "learning_rate": 1.1596660197019079e-07, + "loss": 0.6913, + "step": 22997 + }, + { + "epoch": 0.84, + "learning_rate": 1.1596293990551873e-07, + "loss": 0.6909, + "step": 22998 + }, + { + "epoch": 0.84, + "learning_rate": 1.1595927784084667e-07, + "loss": 0.685, + "step": 22999 + }, + { + "epoch": 0.84, + "learning_rate": 1.1595561577617461e-07, + "loss": 0.6894, + "step": 23000 + }, + { + "epoch": 0.84, + "learning_rate": 1.1595195371150255e-07, + "loss": 0.6875, + "step": 23001 + }, + { + "epoch": 0.84, + "learning_rate": 1.1594829164683049e-07, + "loss": 0.6909, + "step": 23002 + }, + { + "epoch": 0.84, + "learning_rate": 1.1594462958215843e-07, + "loss": 0.6914, + "step": 23003 + }, + { + "epoch": 0.84, + "learning_rate": 1.1594096751748634e-07, + "loss": 0.6955, + "step": 23004 + }, + { + "epoch": 0.84, + "learning_rate": 1.1593730545281428e-07, + "loss": 0.6977, + "step": 23005 + }, + { + "epoch": 0.84, + "learning_rate": 1.1593364338814222e-07, + "loss": 0.7058, + "step": 23006 + }, + { + "epoch": 0.84, + "learning_rate": 1.1592998132347016e-07, + "loss": 0.6951, + "step": 23007 + }, + { + "epoch": 0.84, + "learning_rate": 1.159263192587981e-07, + "loss": 0.6866, + "step": 23008 + }, + { + "epoch": 0.84, + "learning_rate": 1.1592265719412604e-07, + "loss": 0.6988, + "step": 23009 + }, + { + "epoch": 0.84, + "learning_rate": 1.1591899512945398e-07, + "loss": 0.6934, + "step": 23010 + }, + { + "epoch": 0.84, + "learning_rate": 1.1591533306478192e-07, + "loss": 0.6898, + "step": 23011 + }, + { + "epoch": 0.84, + "learning_rate": 1.1591167100010986e-07, + "loss": 0.697, + "step": 23012 + }, + { + "epoch": 0.84, + "learning_rate": 1.159080089354378e-07, + "loss": 0.706, + "step": 23013 + }, + { + "epoch": 0.84, + "learning_rate": 1.1590434687076574e-07, + "loss": 0.6897, + "step": 23014 + }, + { + "epoch": 0.84, + "learning_rate": 1.1590068480609368e-07, + "loss": 0.6892, + "step": 23015 + }, + { + "epoch": 0.84, + "learning_rate": 1.158970227414216e-07, + "loss": 0.6948, + "step": 23016 + }, + { + "epoch": 0.84, + "learning_rate": 1.1589336067674955e-07, + "loss": 0.701, + "step": 23017 + }, + { + "epoch": 0.84, + "learning_rate": 1.1588969861207749e-07, + "loss": 0.6831, + "step": 23018 + }, + { + "epoch": 0.84, + "learning_rate": 1.1588603654740543e-07, + "loss": 0.7057, + "step": 23019 + }, + { + "epoch": 0.84, + "learning_rate": 1.1588237448273337e-07, + "loss": 0.6769, + "step": 23020 + }, + { + "epoch": 0.84, + "learning_rate": 1.158787124180613e-07, + "loss": 0.6895, + "step": 23021 + }, + { + "epoch": 0.84, + "learning_rate": 1.1587505035338922e-07, + "loss": 0.6916, + "step": 23022 + }, + { + "epoch": 0.84, + "learning_rate": 1.1587138828871716e-07, + "loss": 0.6953, + "step": 23023 + }, + { + "epoch": 0.84, + "learning_rate": 1.158677262240451e-07, + "loss": 0.6993, + "step": 23024 + }, + { + "epoch": 0.84, + "learning_rate": 1.1586406415937304e-07, + "loss": 0.6908, + "step": 23025 + }, + { + "epoch": 0.84, + "learning_rate": 1.1586040209470098e-07, + "loss": 0.6845, + "step": 23026 + }, + { + "epoch": 0.84, + "learning_rate": 1.1585674003002892e-07, + "loss": 0.6903, + "step": 23027 + }, + { + "epoch": 0.84, + "learning_rate": 1.1585307796535686e-07, + "loss": 0.6969, + "step": 23028 + }, + { + "epoch": 0.84, + "learning_rate": 1.158494159006848e-07, + "loss": 0.6849, + "step": 23029 + }, + { + "epoch": 0.84, + "learning_rate": 1.1584575383601274e-07, + "loss": 0.7033, + "step": 23030 + }, + { + "epoch": 0.84, + "learning_rate": 1.1584209177134068e-07, + "loss": 0.6849, + "step": 23031 + }, + { + "epoch": 0.84, + "learning_rate": 1.1583842970666862e-07, + "loss": 0.7103, + "step": 23032 + }, + { + "epoch": 0.84, + "learning_rate": 1.1583476764199656e-07, + "loss": 0.7111, + "step": 23033 + }, + { + "epoch": 0.84, + "learning_rate": 1.158311055773245e-07, + "loss": 0.6969, + "step": 23034 + }, + { + "epoch": 0.84, + "learning_rate": 1.1582744351265244e-07, + "loss": 0.6949, + "step": 23035 + }, + { + "epoch": 0.84, + "learning_rate": 1.1582378144798038e-07, + "loss": 0.7073, + "step": 23036 + }, + { + "epoch": 0.84, + "learning_rate": 1.1582011938330832e-07, + "loss": 0.6985, + "step": 23037 + }, + { + "epoch": 0.84, + "learning_rate": 1.1581645731863624e-07, + "loss": 0.6933, + "step": 23038 + }, + { + "epoch": 0.84, + "learning_rate": 1.1581279525396418e-07, + "loss": 0.6962, + "step": 23039 + }, + { + "epoch": 0.84, + "learning_rate": 1.1580913318929211e-07, + "loss": 0.6938, + "step": 23040 + }, + { + "epoch": 0.84, + "learning_rate": 1.1580547112462005e-07, + "loss": 0.6761, + "step": 23041 + }, + { + "epoch": 0.84, + "learning_rate": 1.1580180905994799e-07, + "loss": 0.6829, + "step": 23042 + }, + { + "epoch": 0.84, + "learning_rate": 1.1579814699527593e-07, + "loss": 0.6835, + "step": 23043 + }, + { + "epoch": 0.84, + "learning_rate": 1.1579448493060386e-07, + "loss": 0.6882, + "step": 23044 + }, + { + "epoch": 0.84, + "learning_rate": 1.157908228659318e-07, + "loss": 0.6874, + "step": 23045 + }, + { + "epoch": 0.84, + "learning_rate": 1.1578716080125974e-07, + "loss": 0.6849, + "step": 23046 + }, + { + "epoch": 0.84, + "learning_rate": 1.1578349873658768e-07, + "loss": 0.6991, + "step": 23047 + }, + { + "epoch": 0.84, + "learning_rate": 1.1577983667191562e-07, + "loss": 0.7019, + "step": 23048 + }, + { + "epoch": 0.84, + "learning_rate": 1.1577617460724356e-07, + "loss": 0.7022, + "step": 23049 + }, + { + "epoch": 0.84, + "learning_rate": 1.157725125425715e-07, + "loss": 0.7056, + "step": 23050 + }, + { + "epoch": 0.84, + "learning_rate": 1.1576885047789944e-07, + "loss": 0.6796, + "step": 23051 + }, + { + "epoch": 0.84, + "learning_rate": 1.1576518841322738e-07, + "loss": 0.6877, + "step": 23052 + }, + { + "epoch": 0.84, + "learning_rate": 1.1576152634855532e-07, + "loss": 0.6861, + "step": 23053 + }, + { + "epoch": 0.84, + "learning_rate": 1.1575786428388326e-07, + "loss": 0.706, + "step": 23054 + }, + { + "epoch": 0.84, + "learning_rate": 1.157542022192112e-07, + "loss": 0.6782, + "step": 23055 + }, + { + "epoch": 0.84, + "learning_rate": 1.1575054015453914e-07, + "loss": 0.7041, + "step": 23056 + }, + { + "epoch": 0.84, + "learning_rate": 1.1574687808986708e-07, + "loss": 0.7012, + "step": 23057 + }, + { + "epoch": 0.84, + "learning_rate": 1.1574321602519499e-07, + "loss": 0.7001, + "step": 23058 + }, + { + "epoch": 0.84, + "learning_rate": 1.1573955396052293e-07, + "loss": 0.7147, + "step": 23059 + }, + { + "epoch": 0.84, + "learning_rate": 1.1573589189585087e-07, + "loss": 0.706, + "step": 23060 + }, + { + "epoch": 0.84, + "learning_rate": 1.1573222983117881e-07, + "loss": 0.6972, + "step": 23061 + }, + { + "epoch": 0.84, + "learning_rate": 1.1572856776650675e-07, + "loss": 0.7043, + "step": 23062 + }, + { + "epoch": 0.84, + "learning_rate": 1.1572490570183469e-07, + "loss": 0.6943, + "step": 23063 + }, + { + "epoch": 0.84, + "learning_rate": 1.1572124363716263e-07, + "loss": 0.6985, + "step": 23064 + }, + { + "epoch": 0.84, + "learning_rate": 1.1571758157249057e-07, + "loss": 0.6956, + "step": 23065 + }, + { + "epoch": 0.84, + "learning_rate": 1.157139195078185e-07, + "loss": 0.7036, + "step": 23066 + }, + { + "epoch": 0.84, + "learning_rate": 1.1571025744314644e-07, + "loss": 0.6862, + "step": 23067 + }, + { + "epoch": 0.84, + "learning_rate": 1.1570659537847438e-07, + "loss": 0.7061, + "step": 23068 + }, + { + "epoch": 0.84, + "learning_rate": 1.1570293331380232e-07, + "loss": 0.698, + "step": 23069 + }, + { + "epoch": 0.84, + "learning_rate": 1.1569927124913026e-07, + "loss": 0.6933, + "step": 23070 + }, + { + "epoch": 0.84, + "learning_rate": 1.156956091844582e-07, + "loss": 0.6911, + "step": 23071 + }, + { + "epoch": 0.84, + "learning_rate": 1.1569194711978614e-07, + "loss": 0.6943, + "step": 23072 + }, + { + "epoch": 0.84, + "learning_rate": 1.1568828505511408e-07, + "loss": 0.7005, + "step": 23073 + }, + { + "epoch": 0.84, + "learning_rate": 1.1568462299044202e-07, + "loss": 0.7043, + "step": 23074 + }, + { + "epoch": 0.84, + "learning_rate": 1.1568096092576996e-07, + "loss": 0.689, + "step": 23075 + }, + { + "epoch": 0.84, + "learning_rate": 1.1567729886109787e-07, + "loss": 0.6964, + "step": 23076 + }, + { + "epoch": 0.84, + "learning_rate": 1.1567363679642581e-07, + "loss": 0.7025, + "step": 23077 + }, + { + "epoch": 0.84, + "learning_rate": 1.1566997473175375e-07, + "loss": 0.7003, + "step": 23078 + }, + { + "epoch": 0.84, + "learning_rate": 1.1566631266708169e-07, + "loss": 0.6822, + "step": 23079 + }, + { + "epoch": 0.84, + "learning_rate": 1.1566265060240963e-07, + "loss": 0.6883, + "step": 23080 + }, + { + "epoch": 0.84, + "learning_rate": 1.1565898853773757e-07, + "loss": 0.6961, + "step": 23081 + }, + { + "epoch": 0.84, + "learning_rate": 1.1565532647306551e-07, + "loss": 0.6877, + "step": 23082 + }, + { + "epoch": 0.84, + "learning_rate": 1.1565166440839345e-07, + "loss": 0.7132, + "step": 23083 + }, + { + "epoch": 0.84, + "learning_rate": 1.1564800234372139e-07, + "loss": 0.6938, + "step": 23084 + }, + { + "epoch": 0.84, + "learning_rate": 1.1564434027904933e-07, + "loss": 0.6991, + "step": 23085 + }, + { + "epoch": 0.84, + "learning_rate": 1.1564067821437727e-07, + "loss": 0.7028, + "step": 23086 + }, + { + "epoch": 0.84, + "learning_rate": 1.156370161497052e-07, + "loss": 0.6865, + "step": 23087 + }, + { + "epoch": 0.84, + "learning_rate": 1.1563335408503314e-07, + "loss": 0.7053, + "step": 23088 + }, + { + "epoch": 0.84, + "learning_rate": 1.1562969202036108e-07, + "loss": 0.6989, + "step": 23089 + }, + { + "epoch": 0.84, + "learning_rate": 1.1562602995568902e-07, + "loss": 0.6894, + "step": 23090 + }, + { + "epoch": 0.84, + "learning_rate": 1.1562236789101696e-07, + "loss": 0.6997, + "step": 23091 + }, + { + "epoch": 0.84, + "learning_rate": 1.156187058263449e-07, + "loss": 0.6978, + "step": 23092 + }, + { + "epoch": 0.84, + "learning_rate": 1.1561504376167284e-07, + "loss": 0.6783, + "step": 23093 + }, + { + "epoch": 0.84, + "learning_rate": 1.1561138169700075e-07, + "loss": 0.6881, + "step": 23094 + }, + { + "epoch": 0.84, + "learning_rate": 1.1560771963232869e-07, + "loss": 0.6988, + "step": 23095 + }, + { + "epoch": 0.85, + "learning_rate": 1.1560405756765663e-07, + "loss": 0.6767, + "step": 23096 + }, + { + "epoch": 0.85, + "learning_rate": 1.1560039550298457e-07, + "loss": 0.7099, + "step": 23097 + }, + { + "epoch": 0.85, + "learning_rate": 1.1559673343831251e-07, + "loss": 0.6968, + "step": 23098 + }, + { + "epoch": 0.85, + "learning_rate": 1.1559307137364045e-07, + "loss": 0.6848, + "step": 23099 + }, + { + "epoch": 0.85, + "learning_rate": 1.1558940930896839e-07, + "loss": 0.6927, + "step": 23100 + }, + { + "epoch": 0.85, + "learning_rate": 1.1558574724429633e-07, + "loss": 0.69, + "step": 23101 + }, + { + "epoch": 0.85, + "learning_rate": 1.1558208517962427e-07, + "loss": 0.6894, + "step": 23102 + }, + { + "epoch": 0.85, + "learning_rate": 1.1557842311495221e-07, + "loss": 0.696, + "step": 23103 + }, + { + "epoch": 0.85, + "learning_rate": 1.1557476105028015e-07, + "loss": 0.6987, + "step": 23104 + }, + { + "epoch": 0.85, + "learning_rate": 1.1557109898560809e-07, + "loss": 0.6912, + "step": 23105 + }, + { + "epoch": 0.85, + "learning_rate": 1.1556743692093603e-07, + "loss": 0.6983, + "step": 23106 + }, + { + "epoch": 0.85, + "learning_rate": 1.1556377485626397e-07, + "loss": 0.6941, + "step": 23107 + }, + { + "epoch": 0.85, + "learning_rate": 1.1556011279159191e-07, + "loss": 0.6885, + "step": 23108 + }, + { + "epoch": 0.85, + "learning_rate": 1.1555645072691983e-07, + "loss": 0.6886, + "step": 23109 + }, + { + "epoch": 0.85, + "learning_rate": 1.1555278866224777e-07, + "loss": 0.684, + "step": 23110 + }, + { + "epoch": 0.85, + "learning_rate": 1.1554912659757571e-07, + "loss": 0.7061, + "step": 23111 + }, + { + "epoch": 0.85, + "learning_rate": 1.1554546453290364e-07, + "loss": 0.6877, + "step": 23112 + }, + { + "epoch": 0.85, + "learning_rate": 1.1554180246823158e-07, + "loss": 0.7, + "step": 23113 + }, + { + "epoch": 0.85, + "learning_rate": 1.1553814040355952e-07, + "loss": 0.6936, + "step": 23114 + }, + { + "epoch": 0.85, + "learning_rate": 1.1553447833888745e-07, + "loss": 0.6867, + "step": 23115 + }, + { + "epoch": 0.85, + "learning_rate": 1.1553081627421539e-07, + "loss": 0.6893, + "step": 23116 + }, + { + "epoch": 0.85, + "learning_rate": 1.1552715420954333e-07, + "loss": 0.698, + "step": 23117 + }, + { + "epoch": 0.85, + "learning_rate": 1.1552349214487127e-07, + "loss": 0.6877, + "step": 23118 + }, + { + "epoch": 0.85, + "learning_rate": 1.1551983008019921e-07, + "loss": 0.6921, + "step": 23119 + }, + { + "epoch": 0.85, + "learning_rate": 1.1551616801552715e-07, + "loss": 0.678, + "step": 23120 + }, + { + "epoch": 0.85, + "learning_rate": 1.1551250595085509e-07, + "loss": 0.6952, + "step": 23121 + }, + { + "epoch": 0.85, + "learning_rate": 1.1550884388618303e-07, + "loss": 0.6863, + "step": 23122 + }, + { + "epoch": 0.85, + "learning_rate": 1.1550518182151097e-07, + "loss": 0.698, + "step": 23123 + }, + { + "epoch": 0.85, + "learning_rate": 1.1550151975683891e-07, + "loss": 0.683, + "step": 23124 + }, + { + "epoch": 0.85, + "learning_rate": 1.1549785769216685e-07, + "loss": 0.6942, + "step": 23125 + }, + { + "epoch": 0.85, + "learning_rate": 1.1549419562749479e-07, + "loss": 0.7008, + "step": 23126 + }, + { + "epoch": 0.85, + "learning_rate": 1.1549053356282273e-07, + "loss": 0.6973, + "step": 23127 + }, + { + "epoch": 0.85, + "learning_rate": 1.1548687149815067e-07, + "loss": 0.7048, + "step": 23128 + }, + { + "epoch": 0.85, + "learning_rate": 1.1548320943347858e-07, + "loss": 0.703, + "step": 23129 + }, + { + "epoch": 0.85, + "learning_rate": 1.1547954736880652e-07, + "loss": 0.6861, + "step": 23130 + }, + { + "epoch": 0.85, + "learning_rate": 1.1547588530413446e-07, + "loss": 0.6784, + "step": 23131 + }, + { + "epoch": 0.85, + "learning_rate": 1.154722232394624e-07, + "loss": 0.6903, + "step": 23132 + }, + { + "epoch": 0.85, + "learning_rate": 1.1546856117479034e-07, + "loss": 0.6918, + "step": 23133 + }, + { + "epoch": 0.85, + "learning_rate": 1.1546489911011828e-07, + "loss": 0.688, + "step": 23134 + }, + { + "epoch": 0.85, + "learning_rate": 1.1546123704544622e-07, + "loss": 0.6971, + "step": 23135 + }, + { + "epoch": 0.85, + "learning_rate": 1.1545757498077415e-07, + "loss": 0.6978, + "step": 23136 + }, + { + "epoch": 0.85, + "learning_rate": 1.1545391291610209e-07, + "loss": 0.688, + "step": 23137 + }, + { + "epoch": 0.85, + "learning_rate": 1.1545025085143003e-07, + "loss": 0.6852, + "step": 23138 + }, + { + "epoch": 0.85, + "learning_rate": 1.1544658878675797e-07, + "loss": 0.68, + "step": 23139 + }, + { + "epoch": 0.85, + "learning_rate": 1.154429267220859e-07, + "loss": 0.6932, + "step": 23140 + }, + { + "epoch": 0.85, + "learning_rate": 1.1543926465741385e-07, + "loss": 0.7045, + "step": 23141 + }, + { + "epoch": 0.85, + "learning_rate": 1.1543560259274179e-07, + "loss": 0.6762, + "step": 23142 + }, + { + "epoch": 0.85, + "learning_rate": 1.1543194052806973e-07, + "loss": 0.689, + "step": 23143 + }, + { + "epoch": 0.85, + "learning_rate": 1.1542827846339767e-07, + "loss": 0.6814, + "step": 23144 + }, + { + "epoch": 0.85, + "learning_rate": 1.154246163987256e-07, + "loss": 0.6855, + "step": 23145 + }, + { + "epoch": 0.85, + "learning_rate": 1.1542095433405355e-07, + "loss": 0.6905, + "step": 23146 + }, + { + "epoch": 0.85, + "learning_rate": 1.1541729226938146e-07, + "loss": 0.7107, + "step": 23147 + }, + { + "epoch": 0.85, + "learning_rate": 1.154136302047094e-07, + "loss": 0.698, + "step": 23148 + }, + { + "epoch": 0.85, + "learning_rate": 1.1540996814003734e-07, + "loss": 0.6892, + "step": 23149 + }, + { + "epoch": 0.85, + "learning_rate": 1.1540630607536528e-07, + "loss": 0.6776, + "step": 23150 + }, + { + "epoch": 0.85, + "learning_rate": 1.1540264401069322e-07, + "loss": 0.6693, + "step": 23151 + }, + { + "epoch": 0.85, + "learning_rate": 1.1539898194602116e-07, + "loss": 0.6937, + "step": 23152 + }, + { + "epoch": 0.85, + "learning_rate": 1.153953198813491e-07, + "loss": 0.6893, + "step": 23153 + }, + { + "epoch": 0.85, + "learning_rate": 1.1539165781667704e-07, + "loss": 0.6967, + "step": 23154 + }, + { + "epoch": 0.85, + "learning_rate": 1.1538799575200498e-07, + "loss": 0.698, + "step": 23155 + }, + { + "epoch": 0.85, + "learning_rate": 1.1538433368733292e-07, + "loss": 0.6858, + "step": 23156 + }, + { + "epoch": 0.85, + "learning_rate": 1.1538067162266086e-07, + "loss": 0.6933, + "step": 23157 + }, + { + "epoch": 0.85, + "learning_rate": 1.1537700955798878e-07, + "loss": 0.6956, + "step": 23158 + }, + { + "epoch": 0.85, + "learning_rate": 1.1537334749331672e-07, + "loss": 0.6961, + "step": 23159 + }, + { + "epoch": 0.85, + "learning_rate": 1.1536968542864466e-07, + "loss": 0.6899, + "step": 23160 + }, + { + "epoch": 0.85, + "learning_rate": 1.153660233639726e-07, + "loss": 0.6901, + "step": 23161 + }, + { + "epoch": 0.85, + "learning_rate": 1.1536236129930054e-07, + "loss": 0.7024, + "step": 23162 + }, + { + "epoch": 0.85, + "learning_rate": 1.1535869923462848e-07, + "loss": 0.6843, + "step": 23163 + }, + { + "epoch": 0.85, + "learning_rate": 1.1535503716995642e-07, + "loss": 0.6928, + "step": 23164 + }, + { + "epoch": 0.85, + "learning_rate": 1.1535137510528434e-07, + "loss": 0.6868, + "step": 23165 + }, + { + "epoch": 0.85, + "learning_rate": 1.1534771304061228e-07, + "loss": 0.6687, + "step": 23166 + }, + { + "epoch": 0.85, + "learning_rate": 1.1534405097594022e-07, + "loss": 0.6966, + "step": 23167 + }, + { + "epoch": 0.85, + "learning_rate": 1.1534038891126816e-07, + "loss": 0.6881, + "step": 23168 + }, + { + "epoch": 0.85, + "learning_rate": 1.153367268465961e-07, + "loss": 0.6817, + "step": 23169 + }, + { + "epoch": 0.85, + "learning_rate": 1.1533306478192404e-07, + "loss": 0.6789, + "step": 23170 + }, + { + "epoch": 0.85, + "learning_rate": 1.1532940271725198e-07, + "loss": 0.6921, + "step": 23171 + }, + { + "epoch": 0.85, + "learning_rate": 1.1532574065257992e-07, + "loss": 0.6884, + "step": 23172 + }, + { + "epoch": 0.85, + "learning_rate": 1.1532207858790786e-07, + "loss": 0.6946, + "step": 23173 + }, + { + "epoch": 0.85, + "learning_rate": 1.153184165232358e-07, + "loss": 0.6901, + "step": 23174 + }, + { + "epoch": 0.85, + "learning_rate": 1.1531475445856374e-07, + "loss": 0.6943, + "step": 23175 + }, + { + "epoch": 0.85, + "learning_rate": 1.1531109239389168e-07, + "loss": 0.6865, + "step": 23176 + }, + { + "epoch": 0.85, + "learning_rate": 1.1530743032921962e-07, + "loss": 0.6953, + "step": 23177 + }, + { + "epoch": 0.85, + "learning_rate": 1.1530376826454756e-07, + "loss": 0.6792, + "step": 23178 + }, + { + "epoch": 0.85, + "learning_rate": 1.153001061998755e-07, + "loss": 0.6895, + "step": 23179 + }, + { + "epoch": 0.85, + "learning_rate": 1.1529644413520342e-07, + "loss": 0.6969, + "step": 23180 + }, + { + "epoch": 0.85, + "learning_rate": 1.1529278207053136e-07, + "loss": 0.6933, + "step": 23181 + }, + { + "epoch": 0.85, + "learning_rate": 1.152891200058593e-07, + "loss": 0.6993, + "step": 23182 + }, + { + "epoch": 0.85, + "learning_rate": 1.1528545794118723e-07, + "loss": 0.6992, + "step": 23183 + }, + { + "epoch": 0.85, + "learning_rate": 1.1528179587651517e-07, + "loss": 0.6977, + "step": 23184 + }, + { + "epoch": 0.85, + "learning_rate": 1.1527813381184311e-07, + "loss": 0.6835, + "step": 23185 + }, + { + "epoch": 0.85, + "learning_rate": 1.1527447174717104e-07, + "loss": 0.6859, + "step": 23186 + }, + { + "epoch": 0.85, + "learning_rate": 1.1527080968249898e-07, + "loss": 0.6927, + "step": 23187 + }, + { + "epoch": 0.85, + "learning_rate": 1.1526714761782692e-07, + "loss": 0.6814, + "step": 23188 + }, + { + "epoch": 0.85, + "learning_rate": 1.1526348555315486e-07, + "loss": 0.7029, + "step": 23189 + }, + { + "epoch": 0.85, + "learning_rate": 1.152598234884828e-07, + "loss": 0.6908, + "step": 23190 + }, + { + "epoch": 0.85, + "learning_rate": 1.1525616142381074e-07, + "loss": 0.6897, + "step": 23191 + }, + { + "epoch": 0.85, + "learning_rate": 1.1525249935913868e-07, + "loss": 0.6855, + "step": 23192 + }, + { + "epoch": 0.85, + "learning_rate": 1.1524883729446662e-07, + "loss": 0.6992, + "step": 23193 + }, + { + "epoch": 0.85, + "learning_rate": 1.1524517522979456e-07, + "loss": 0.6901, + "step": 23194 + }, + { + "epoch": 0.85, + "learning_rate": 1.152415131651225e-07, + "loss": 0.7075, + "step": 23195 + }, + { + "epoch": 0.85, + "learning_rate": 1.1523785110045044e-07, + "loss": 0.6942, + "step": 23196 + }, + { + "epoch": 0.85, + "learning_rate": 1.1523418903577838e-07, + "loss": 0.6886, + "step": 23197 + }, + { + "epoch": 0.85, + "learning_rate": 1.1523052697110632e-07, + "loss": 0.6936, + "step": 23198 + }, + { + "epoch": 0.85, + "learning_rate": 1.1522686490643426e-07, + "loss": 0.6904, + "step": 23199 + }, + { + "epoch": 0.85, + "learning_rate": 1.152232028417622e-07, + "loss": 0.6875, + "step": 23200 + }, + { + "epoch": 0.85, + "learning_rate": 1.1521954077709011e-07, + "loss": 0.697, + "step": 23201 + }, + { + "epoch": 0.85, + "learning_rate": 1.1521587871241805e-07, + "loss": 0.7035, + "step": 23202 + }, + { + "epoch": 0.85, + "learning_rate": 1.1521221664774599e-07, + "loss": 0.6937, + "step": 23203 + }, + { + "epoch": 0.85, + "learning_rate": 1.1520855458307393e-07, + "loss": 0.6935, + "step": 23204 + }, + { + "epoch": 0.85, + "learning_rate": 1.1520489251840187e-07, + "loss": 0.679, + "step": 23205 + }, + { + "epoch": 0.85, + "learning_rate": 1.1520123045372981e-07, + "loss": 0.7012, + "step": 23206 + }, + { + "epoch": 0.85, + "learning_rate": 1.1519756838905774e-07, + "loss": 0.6968, + "step": 23207 + }, + { + "epoch": 0.85, + "learning_rate": 1.1519390632438568e-07, + "loss": 0.6973, + "step": 23208 + }, + { + "epoch": 0.85, + "learning_rate": 1.1519024425971362e-07, + "loss": 0.693, + "step": 23209 + }, + { + "epoch": 0.85, + "learning_rate": 1.1518658219504156e-07, + "loss": 0.6994, + "step": 23210 + }, + { + "epoch": 0.85, + "learning_rate": 1.151829201303695e-07, + "loss": 0.701, + "step": 23211 + }, + { + "epoch": 0.85, + "learning_rate": 1.1517925806569744e-07, + "loss": 0.6989, + "step": 23212 + }, + { + "epoch": 0.85, + "learning_rate": 1.1517559600102538e-07, + "loss": 0.6862, + "step": 23213 + }, + { + "epoch": 0.85, + "learning_rate": 1.1517193393635332e-07, + "loss": 0.6932, + "step": 23214 + }, + { + "epoch": 0.85, + "learning_rate": 1.1516827187168125e-07, + "loss": 0.7009, + "step": 23215 + }, + { + "epoch": 0.85, + "learning_rate": 1.151646098070092e-07, + "loss": 0.6946, + "step": 23216 + }, + { + "epoch": 0.85, + "learning_rate": 1.1516094774233713e-07, + "loss": 0.6922, + "step": 23217 + }, + { + "epoch": 0.85, + "learning_rate": 1.1515728567766507e-07, + "loss": 0.6978, + "step": 23218 + }, + { + "epoch": 0.85, + "learning_rate": 1.1515362361299299e-07, + "loss": 0.6749, + "step": 23219 + }, + { + "epoch": 0.85, + "learning_rate": 1.1514996154832093e-07, + "loss": 0.6973, + "step": 23220 + }, + { + "epoch": 0.85, + "learning_rate": 1.1514629948364887e-07, + "loss": 0.6961, + "step": 23221 + }, + { + "epoch": 0.85, + "learning_rate": 1.1514263741897681e-07, + "loss": 0.7007, + "step": 23222 + }, + { + "epoch": 0.85, + "learning_rate": 1.1513897535430475e-07, + "loss": 0.6942, + "step": 23223 + }, + { + "epoch": 0.85, + "learning_rate": 1.1513531328963269e-07, + "loss": 0.6926, + "step": 23224 + }, + { + "epoch": 0.85, + "learning_rate": 1.1513165122496063e-07, + "loss": 0.6959, + "step": 23225 + }, + { + "epoch": 0.85, + "learning_rate": 1.1512798916028857e-07, + "loss": 0.6979, + "step": 23226 + }, + { + "epoch": 0.85, + "learning_rate": 1.1512432709561651e-07, + "loss": 0.6945, + "step": 23227 + }, + { + "epoch": 0.85, + "learning_rate": 1.1512066503094445e-07, + "loss": 0.6871, + "step": 23228 + }, + { + "epoch": 0.85, + "learning_rate": 1.1511700296627237e-07, + "loss": 0.6934, + "step": 23229 + }, + { + "epoch": 0.85, + "learning_rate": 1.1511334090160031e-07, + "loss": 0.6901, + "step": 23230 + }, + { + "epoch": 0.85, + "learning_rate": 1.1510967883692825e-07, + "loss": 0.7044, + "step": 23231 + }, + { + "epoch": 0.85, + "learning_rate": 1.151060167722562e-07, + "loss": 0.6932, + "step": 23232 + }, + { + "epoch": 0.85, + "learning_rate": 1.1510235470758413e-07, + "loss": 0.6911, + "step": 23233 + }, + { + "epoch": 0.85, + "learning_rate": 1.1509869264291207e-07, + "loss": 0.6919, + "step": 23234 + }, + { + "epoch": 0.85, + "learning_rate": 1.1509503057824001e-07, + "loss": 0.6927, + "step": 23235 + }, + { + "epoch": 0.85, + "learning_rate": 1.1509136851356795e-07, + "loss": 0.6853, + "step": 23236 + }, + { + "epoch": 0.85, + "learning_rate": 1.1508770644889587e-07, + "loss": 0.6713, + "step": 23237 + }, + { + "epoch": 0.85, + "learning_rate": 1.1508404438422381e-07, + "loss": 0.6799, + "step": 23238 + }, + { + "epoch": 0.85, + "learning_rate": 1.1508038231955175e-07, + "loss": 0.6953, + "step": 23239 + }, + { + "epoch": 0.85, + "learning_rate": 1.1507672025487969e-07, + "loss": 0.7047, + "step": 23240 + }, + { + "epoch": 0.85, + "learning_rate": 1.1507305819020763e-07, + "loss": 0.685, + "step": 23241 + }, + { + "epoch": 0.85, + "learning_rate": 1.1506939612553557e-07, + "loss": 0.6921, + "step": 23242 + }, + { + "epoch": 0.85, + "learning_rate": 1.1506573406086351e-07, + "loss": 0.6865, + "step": 23243 + }, + { + "epoch": 0.85, + "learning_rate": 1.1506207199619145e-07, + "loss": 0.6954, + "step": 23244 + }, + { + "epoch": 0.85, + "learning_rate": 1.1505840993151939e-07, + "loss": 0.6924, + "step": 23245 + }, + { + "epoch": 0.85, + "learning_rate": 1.1505474786684733e-07, + "loss": 0.6977, + "step": 23246 + }, + { + "epoch": 0.85, + "learning_rate": 1.1505108580217527e-07, + "loss": 0.6909, + "step": 23247 + }, + { + "epoch": 0.85, + "learning_rate": 1.150474237375032e-07, + "loss": 0.6926, + "step": 23248 + }, + { + "epoch": 0.85, + "learning_rate": 1.1504376167283115e-07, + "loss": 0.6918, + "step": 23249 + }, + { + "epoch": 0.85, + "learning_rate": 1.1504009960815909e-07, + "loss": 0.6758, + "step": 23250 + }, + { + "epoch": 0.85, + "learning_rate": 1.1503643754348701e-07, + "loss": 0.6969, + "step": 23251 + }, + { + "epoch": 0.85, + "learning_rate": 1.1503277547881495e-07, + "loss": 0.6881, + "step": 23252 + }, + { + "epoch": 0.85, + "learning_rate": 1.1502911341414289e-07, + "loss": 0.7009, + "step": 23253 + }, + { + "epoch": 0.85, + "learning_rate": 1.1502545134947083e-07, + "loss": 0.6943, + "step": 23254 + }, + { + "epoch": 0.85, + "learning_rate": 1.1502178928479876e-07, + "loss": 0.6901, + "step": 23255 + }, + { + "epoch": 0.85, + "learning_rate": 1.1501812722012669e-07, + "loss": 0.6946, + "step": 23256 + }, + { + "epoch": 0.85, + "learning_rate": 1.1501446515545463e-07, + "loss": 0.6719, + "step": 23257 + }, + { + "epoch": 0.85, + "learning_rate": 1.1501080309078257e-07, + "loss": 0.6815, + "step": 23258 + }, + { + "epoch": 0.85, + "learning_rate": 1.150071410261105e-07, + "loss": 0.6837, + "step": 23259 + }, + { + "epoch": 0.85, + "learning_rate": 1.1500347896143845e-07, + "loss": 0.7136, + "step": 23260 + }, + { + "epoch": 0.85, + "learning_rate": 1.1499981689676639e-07, + "loss": 0.6942, + "step": 23261 + }, + { + "epoch": 0.85, + "learning_rate": 1.1499615483209433e-07, + "loss": 0.7044, + "step": 23262 + }, + { + "epoch": 0.85, + "learning_rate": 1.1499249276742227e-07, + "loss": 0.7044, + "step": 23263 + }, + { + "epoch": 0.85, + "learning_rate": 1.149888307027502e-07, + "loss": 0.7092, + "step": 23264 + }, + { + "epoch": 0.85, + "learning_rate": 1.1498516863807815e-07, + "loss": 0.6897, + "step": 23265 + }, + { + "epoch": 0.85, + "learning_rate": 1.1498150657340609e-07, + "loss": 0.6885, + "step": 23266 + }, + { + "epoch": 0.85, + "learning_rate": 1.1497784450873403e-07, + "loss": 0.6942, + "step": 23267 + }, + { + "epoch": 0.85, + "learning_rate": 1.1497418244406197e-07, + "loss": 0.6952, + "step": 23268 + }, + { + "epoch": 0.85, + "learning_rate": 1.149705203793899e-07, + "loss": 0.6837, + "step": 23269 + }, + { + "epoch": 0.85, + "learning_rate": 1.1496685831471785e-07, + "loss": 0.6957, + "step": 23270 + }, + { + "epoch": 0.85, + "learning_rate": 1.1496319625004579e-07, + "loss": 0.6917, + "step": 23271 + }, + { + "epoch": 0.85, + "learning_rate": 1.1495953418537373e-07, + "loss": 0.6832, + "step": 23272 + }, + { + "epoch": 0.85, + "learning_rate": 1.1495587212070164e-07, + "loss": 0.6919, + "step": 23273 + }, + { + "epoch": 0.85, + "learning_rate": 1.1495221005602958e-07, + "loss": 0.6954, + "step": 23274 + }, + { + "epoch": 0.85, + "learning_rate": 1.1494854799135752e-07, + "loss": 0.6856, + "step": 23275 + }, + { + "epoch": 0.85, + "learning_rate": 1.1494488592668546e-07, + "loss": 0.7015, + "step": 23276 + }, + { + "epoch": 0.85, + "learning_rate": 1.149412238620134e-07, + "loss": 0.6824, + "step": 23277 + }, + { + "epoch": 0.85, + "learning_rate": 1.1493756179734132e-07, + "loss": 0.6891, + "step": 23278 + }, + { + "epoch": 0.85, + "learning_rate": 1.1493389973266926e-07, + "loss": 0.701, + "step": 23279 + }, + { + "epoch": 0.85, + "learning_rate": 1.149302376679972e-07, + "loss": 0.6871, + "step": 23280 + }, + { + "epoch": 0.85, + "learning_rate": 1.1492657560332514e-07, + "loss": 0.6923, + "step": 23281 + }, + { + "epoch": 0.85, + "learning_rate": 1.1492291353865308e-07, + "loss": 0.6949, + "step": 23282 + }, + { + "epoch": 0.85, + "learning_rate": 1.1491925147398102e-07, + "loss": 0.6971, + "step": 23283 + }, + { + "epoch": 0.85, + "learning_rate": 1.1491558940930896e-07, + "loss": 0.6909, + "step": 23284 + }, + { + "epoch": 0.85, + "learning_rate": 1.149119273446369e-07, + "loss": 0.7022, + "step": 23285 + }, + { + "epoch": 0.85, + "learning_rate": 1.1490826527996484e-07, + "loss": 0.6916, + "step": 23286 + }, + { + "epoch": 0.85, + "learning_rate": 1.1490460321529278e-07, + "loss": 0.6792, + "step": 23287 + }, + { + "epoch": 0.85, + "learning_rate": 1.1490094115062072e-07, + "loss": 0.6922, + "step": 23288 + }, + { + "epoch": 0.85, + "learning_rate": 1.1489727908594866e-07, + "loss": 0.7026, + "step": 23289 + }, + { + "epoch": 0.85, + "learning_rate": 1.148936170212766e-07, + "loss": 0.6845, + "step": 23290 + }, + { + "epoch": 0.85, + "learning_rate": 1.1488995495660452e-07, + "loss": 0.682, + "step": 23291 + }, + { + "epoch": 0.85, + "learning_rate": 1.1488629289193246e-07, + "loss": 0.6824, + "step": 23292 + }, + { + "epoch": 0.85, + "learning_rate": 1.148826308272604e-07, + "loss": 0.6951, + "step": 23293 + }, + { + "epoch": 0.85, + "learning_rate": 1.1487896876258834e-07, + "loss": 0.6761, + "step": 23294 + }, + { + "epoch": 0.85, + "learning_rate": 1.1487530669791628e-07, + "loss": 0.7048, + "step": 23295 + }, + { + "epoch": 0.85, + "learning_rate": 1.1487164463324422e-07, + "loss": 0.6979, + "step": 23296 + }, + { + "epoch": 0.85, + "learning_rate": 1.1486798256857216e-07, + "loss": 0.6928, + "step": 23297 + }, + { + "epoch": 0.85, + "learning_rate": 1.148643205039001e-07, + "loss": 0.7016, + "step": 23298 + }, + { + "epoch": 0.85, + "learning_rate": 1.1486065843922804e-07, + "loss": 0.6968, + "step": 23299 + }, + { + "epoch": 0.85, + "learning_rate": 1.1485699637455596e-07, + "loss": 0.6908, + "step": 23300 + }, + { + "epoch": 0.85, + "learning_rate": 1.148533343098839e-07, + "loss": 0.6872, + "step": 23301 + }, + { + "epoch": 0.85, + "learning_rate": 1.1484967224521184e-07, + "loss": 0.6902, + "step": 23302 + }, + { + "epoch": 0.85, + "learning_rate": 1.1484601018053978e-07, + "loss": 0.6917, + "step": 23303 + }, + { + "epoch": 0.85, + "learning_rate": 1.1484234811586772e-07, + "loss": 0.7052, + "step": 23304 + }, + { + "epoch": 0.85, + "learning_rate": 1.1483868605119566e-07, + "loss": 0.6896, + "step": 23305 + }, + { + "epoch": 0.85, + "learning_rate": 1.148350239865236e-07, + "loss": 0.6998, + "step": 23306 + }, + { + "epoch": 0.85, + "learning_rate": 1.1483136192185154e-07, + "loss": 0.6872, + "step": 23307 + }, + { + "epoch": 0.85, + "learning_rate": 1.1482769985717948e-07, + "loss": 0.6893, + "step": 23308 + }, + { + "epoch": 0.85, + "learning_rate": 1.148240377925074e-07, + "loss": 0.6982, + "step": 23309 + }, + { + "epoch": 0.85, + "learning_rate": 1.1482037572783534e-07, + "loss": 0.689, + "step": 23310 + }, + { + "epoch": 0.85, + "learning_rate": 1.1481671366316328e-07, + "loss": 0.6844, + "step": 23311 + }, + { + "epoch": 0.85, + "learning_rate": 1.1481305159849122e-07, + "loss": 0.6883, + "step": 23312 + }, + { + "epoch": 0.85, + "learning_rate": 1.1480938953381916e-07, + "loss": 0.704, + "step": 23313 + }, + { + "epoch": 0.85, + "learning_rate": 1.148057274691471e-07, + "loss": 0.7151, + "step": 23314 + }, + { + "epoch": 0.85, + "learning_rate": 1.1480206540447504e-07, + "loss": 0.6947, + "step": 23315 + }, + { + "epoch": 0.85, + "learning_rate": 1.1479840333980298e-07, + "loss": 0.6985, + "step": 23316 + }, + { + "epoch": 0.85, + "learning_rate": 1.1479474127513092e-07, + "loss": 0.6944, + "step": 23317 + }, + { + "epoch": 0.85, + "learning_rate": 1.1479107921045886e-07, + "loss": 0.6915, + "step": 23318 + }, + { + "epoch": 0.85, + "learning_rate": 1.147874171457868e-07, + "loss": 0.6873, + "step": 23319 + }, + { + "epoch": 0.85, + "learning_rate": 1.1478375508111474e-07, + "loss": 0.7082, + "step": 23320 + }, + { + "epoch": 0.85, + "learning_rate": 1.1478009301644268e-07, + "loss": 0.6929, + "step": 23321 + }, + { + "epoch": 0.85, + "learning_rate": 1.147764309517706e-07, + "loss": 0.6776, + "step": 23322 + }, + { + "epoch": 0.85, + "learning_rate": 1.1477276888709854e-07, + "loss": 0.7049, + "step": 23323 + }, + { + "epoch": 0.85, + "learning_rate": 1.1476910682242648e-07, + "loss": 0.6934, + "step": 23324 + }, + { + "epoch": 0.85, + "learning_rate": 1.1476544475775442e-07, + "loss": 0.6985, + "step": 23325 + }, + { + "epoch": 0.85, + "learning_rate": 1.1476178269308236e-07, + "loss": 0.6854, + "step": 23326 + }, + { + "epoch": 0.85, + "learning_rate": 1.1475812062841028e-07, + "loss": 0.6885, + "step": 23327 + }, + { + "epoch": 0.85, + "learning_rate": 1.1475445856373822e-07, + "loss": 0.6844, + "step": 23328 + }, + { + "epoch": 0.85, + "learning_rate": 1.1475079649906616e-07, + "loss": 0.6965, + "step": 23329 + }, + { + "epoch": 0.85, + "learning_rate": 1.147471344343941e-07, + "loss": 0.6904, + "step": 23330 + }, + { + "epoch": 0.85, + "learning_rate": 1.1474347236972204e-07, + "loss": 0.6756, + "step": 23331 + }, + { + "epoch": 0.85, + "learning_rate": 1.1473981030504998e-07, + "loss": 0.6982, + "step": 23332 + }, + { + "epoch": 0.85, + "learning_rate": 1.1473614824037792e-07, + "loss": 0.7045, + "step": 23333 + }, + { + "epoch": 0.85, + "learning_rate": 1.1473248617570586e-07, + "loss": 0.684, + "step": 23334 + }, + { + "epoch": 0.85, + "learning_rate": 1.147288241110338e-07, + "loss": 0.7027, + "step": 23335 + }, + { + "epoch": 0.85, + "learning_rate": 1.1472516204636173e-07, + "loss": 0.6993, + "step": 23336 + }, + { + "epoch": 0.85, + "learning_rate": 1.1472149998168967e-07, + "loss": 0.6983, + "step": 23337 + }, + { + "epoch": 0.85, + "learning_rate": 1.1471783791701761e-07, + "loss": 0.6883, + "step": 23338 + }, + { + "epoch": 0.85, + "learning_rate": 1.1471417585234555e-07, + "loss": 0.6938, + "step": 23339 + }, + { + "epoch": 0.85, + "learning_rate": 1.147105137876735e-07, + "loss": 0.6909, + "step": 23340 + }, + { + "epoch": 0.85, + "learning_rate": 1.1470685172300143e-07, + "loss": 0.6905, + "step": 23341 + }, + { + "epoch": 0.85, + "learning_rate": 1.1470318965832937e-07, + "loss": 0.6887, + "step": 23342 + }, + { + "epoch": 0.85, + "learning_rate": 1.1469952759365731e-07, + "loss": 0.6989, + "step": 23343 + }, + { + "epoch": 0.85, + "learning_rate": 1.1469586552898524e-07, + "loss": 0.7, + "step": 23344 + }, + { + "epoch": 0.85, + "learning_rate": 1.1469220346431317e-07, + "loss": 0.6936, + "step": 23345 + }, + { + "epoch": 0.85, + "learning_rate": 1.1468854139964111e-07, + "loss": 0.6898, + "step": 23346 + }, + { + "epoch": 0.85, + "learning_rate": 1.1468487933496905e-07, + "loss": 0.6943, + "step": 23347 + }, + { + "epoch": 0.85, + "learning_rate": 1.1468121727029699e-07, + "loss": 0.6824, + "step": 23348 + }, + { + "epoch": 0.85, + "learning_rate": 1.1467755520562491e-07, + "loss": 0.7052, + "step": 23349 + }, + { + "epoch": 0.85, + "learning_rate": 1.1467389314095285e-07, + "loss": 0.6972, + "step": 23350 + }, + { + "epoch": 0.85, + "learning_rate": 1.146702310762808e-07, + "loss": 0.6891, + "step": 23351 + }, + { + "epoch": 0.85, + "learning_rate": 1.1466656901160873e-07, + "loss": 0.6898, + "step": 23352 + }, + { + "epoch": 0.85, + "learning_rate": 1.1466290694693667e-07, + "loss": 0.7057, + "step": 23353 + }, + { + "epoch": 0.85, + "learning_rate": 1.1465924488226461e-07, + "loss": 0.6945, + "step": 23354 + }, + { + "epoch": 0.85, + "learning_rate": 1.1465558281759255e-07, + "loss": 0.6974, + "step": 23355 + }, + { + "epoch": 0.85, + "learning_rate": 1.146519207529205e-07, + "loss": 0.7075, + "step": 23356 + }, + { + "epoch": 0.85, + "learning_rate": 1.1464825868824843e-07, + "loss": 0.696, + "step": 23357 + }, + { + "epoch": 0.85, + "learning_rate": 1.1464459662357637e-07, + "loss": 0.6924, + "step": 23358 + }, + { + "epoch": 0.85, + "learning_rate": 1.1464093455890431e-07, + "loss": 0.6974, + "step": 23359 + }, + { + "epoch": 0.85, + "learning_rate": 1.1463727249423225e-07, + "loss": 0.689, + "step": 23360 + }, + { + "epoch": 0.85, + "learning_rate": 1.146336104295602e-07, + "loss": 0.7039, + "step": 23361 + }, + { + "epoch": 0.85, + "learning_rate": 1.1462994836488813e-07, + "loss": 0.6908, + "step": 23362 + }, + { + "epoch": 0.85, + "learning_rate": 1.1462628630021605e-07, + "loss": 0.6949, + "step": 23363 + }, + { + "epoch": 0.85, + "learning_rate": 1.1462262423554399e-07, + "loss": 0.6921, + "step": 23364 + }, + { + "epoch": 0.85, + "learning_rate": 1.1461896217087193e-07, + "loss": 0.6988, + "step": 23365 + }, + { + "epoch": 0.85, + "learning_rate": 1.1461530010619987e-07, + "loss": 0.6894, + "step": 23366 + }, + { + "epoch": 0.85, + "learning_rate": 1.146116380415278e-07, + "loss": 0.695, + "step": 23367 + }, + { + "epoch": 0.85, + "learning_rate": 1.1460797597685575e-07, + "loss": 0.6865, + "step": 23368 + }, + { + "epoch": 0.86, + "learning_rate": 1.1460431391218369e-07, + "loss": 0.687, + "step": 23369 + }, + { + "epoch": 0.86, + "learning_rate": 1.1460065184751163e-07, + "loss": 0.6906, + "step": 23370 + }, + { + "epoch": 0.86, + "learning_rate": 1.1459698978283955e-07, + "loss": 0.6878, + "step": 23371 + }, + { + "epoch": 0.86, + "learning_rate": 1.1459332771816749e-07, + "loss": 0.6914, + "step": 23372 + }, + { + "epoch": 0.86, + "learning_rate": 1.1458966565349543e-07, + "loss": 0.6961, + "step": 23373 + }, + { + "epoch": 0.86, + "learning_rate": 1.1458600358882337e-07, + "loss": 0.6803, + "step": 23374 + }, + { + "epoch": 0.86, + "learning_rate": 1.1458234152415131e-07, + "loss": 0.7054, + "step": 23375 + }, + { + "epoch": 0.86, + "learning_rate": 1.1457867945947925e-07, + "loss": 0.6791, + "step": 23376 + }, + { + "epoch": 0.86, + "learning_rate": 1.1457501739480719e-07, + "loss": 0.7016, + "step": 23377 + }, + { + "epoch": 0.86, + "learning_rate": 1.1457135533013513e-07, + "loss": 0.7049, + "step": 23378 + }, + { + "epoch": 0.86, + "learning_rate": 1.1456769326546307e-07, + "loss": 0.7097, + "step": 23379 + }, + { + "epoch": 0.86, + "learning_rate": 1.1456403120079101e-07, + "loss": 0.6942, + "step": 23380 + }, + { + "epoch": 0.86, + "learning_rate": 1.1456036913611893e-07, + "loss": 0.6906, + "step": 23381 + }, + { + "epoch": 0.86, + "learning_rate": 1.1455670707144687e-07, + "loss": 0.7047, + "step": 23382 + }, + { + "epoch": 0.86, + "learning_rate": 1.145530450067748e-07, + "loss": 0.6824, + "step": 23383 + }, + { + "epoch": 0.86, + "learning_rate": 1.1454938294210275e-07, + "loss": 0.6912, + "step": 23384 + }, + { + "epoch": 0.86, + "learning_rate": 1.1454572087743069e-07, + "loss": 0.6987, + "step": 23385 + }, + { + "epoch": 0.86, + "learning_rate": 1.1454205881275863e-07, + "loss": 0.7004, + "step": 23386 + }, + { + "epoch": 0.86, + "learning_rate": 1.1453839674808657e-07, + "loss": 0.6939, + "step": 23387 + }, + { + "epoch": 0.86, + "learning_rate": 1.145347346834145e-07, + "loss": 0.6931, + "step": 23388 + }, + { + "epoch": 0.86, + "learning_rate": 1.1453107261874245e-07, + "loss": 0.6826, + "step": 23389 + }, + { + "epoch": 0.86, + "learning_rate": 1.1452741055407039e-07, + "loss": 0.7015, + "step": 23390 + }, + { + "epoch": 0.86, + "learning_rate": 1.1452374848939833e-07, + "loss": 0.7057, + "step": 23391 + }, + { + "epoch": 0.86, + "learning_rate": 1.1452008642472627e-07, + "loss": 0.6958, + "step": 23392 + }, + { + "epoch": 0.86, + "learning_rate": 1.1451642436005419e-07, + "loss": 0.7043, + "step": 23393 + }, + { + "epoch": 0.86, + "learning_rate": 1.1451276229538213e-07, + "loss": 0.6947, + "step": 23394 + }, + { + "epoch": 0.86, + "learning_rate": 1.1450910023071007e-07, + "loss": 0.6911, + "step": 23395 + }, + { + "epoch": 0.86, + "learning_rate": 1.1450543816603801e-07, + "loss": 0.6827, + "step": 23396 + }, + { + "epoch": 0.86, + "learning_rate": 1.1450177610136595e-07, + "loss": 0.6955, + "step": 23397 + }, + { + "epoch": 0.86, + "learning_rate": 1.1449811403669389e-07, + "loss": 0.7127, + "step": 23398 + }, + { + "epoch": 0.86, + "learning_rate": 1.144944519720218e-07, + "loss": 0.7012, + "step": 23399 + }, + { + "epoch": 0.86, + "learning_rate": 1.1449078990734974e-07, + "loss": 0.717, + "step": 23400 + }, + { + "epoch": 0.86, + "learning_rate": 1.1448712784267768e-07, + "loss": 0.689, + "step": 23401 + }, + { + "epoch": 0.86, + "learning_rate": 1.1448346577800562e-07, + "loss": 0.6888, + "step": 23402 + }, + { + "epoch": 0.86, + "learning_rate": 1.1447980371333356e-07, + "loss": 0.682, + "step": 23403 + }, + { + "epoch": 0.86, + "learning_rate": 1.144761416486615e-07, + "loss": 0.6978, + "step": 23404 + }, + { + "epoch": 0.86, + "learning_rate": 1.1447247958398944e-07, + "loss": 0.7067, + "step": 23405 + }, + { + "epoch": 0.86, + "learning_rate": 1.1446881751931738e-07, + "loss": 0.7095, + "step": 23406 + }, + { + "epoch": 0.86, + "learning_rate": 1.1446515545464532e-07, + "loss": 0.7014, + "step": 23407 + }, + { + "epoch": 0.86, + "learning_rate": 1.1446149338997326e-07, + "loss": 0.6963, + "step": 23408 + }, + { + "epoch": 0.86, + "learning_rate": 1.144578313253012e-07, + "loss": 0.7121, + "step": 23409 + }, + { + "epoch": 0.86, + "learning_rate": 1.1445416926062914e-07, + "loss": 0.7054, + "step": 23410 + }, + { + "epoch": 0.86, + "learning_rate": 1.1445050719595708e-07, + "loss": 0.6926, + "step": 23411 + }, + { + "epoch": 0.86, + "learning_rate": 1.1444684513128502e-07, + "loss": 0.698, + "step": 23412 + }, + { + "epoch": 0.86, + "learning_rate": 1.1444318306661296e-07, + "loss": 0.695, + "step": 23413 + }, + { + "epoch": 0.86, + "learning_rate": 1.144395210019409e-07, + "loss": 0.6878, + "step": 23414 + }, + { + "epoch": 0.86, + "learning_rate": 1.1443585893726883e-07, + "loss": 0.6939, + "step": 23415 + }, + { + "epoch": 0.86, + "learning_rate": 1.1443219687259677e-07, + "loss": 0.692, + "step": 23416 + }, + { + "epoch": 0.86, + "learning_rate": 1.144285348079247e-07, + "loss": 0.7045, + "step": 23417 + }, + { + "epoch": 0.86, + "learning_rate": 1.1442487274325264e-07, + "loss": 0.6925, + "step": 23418 + }, + { + "epoch": 0.86, + "learning_rate": 1.1442121067858058e-07, + "loss": 0.6833, + "step": 23419 + }, + { + "epoch": 0.86, + "learning_rate": 1.144175486139085e-07, + "loss": 0.6976, + "step": 23420 + }, + { + "epoch": 0.86, + "learning_rate": 1.1441388654923644e-07, + "loss": 0.6907, + "step": 23421 + }, + { + "epoch": 0.86, + "learning_rate": 1.1441022448456438e-07, + "loss": 0.7024, + "step": 23422 + }, + { + "epoch": 0.86, + "learning_rate": 1.1440656241989232e-07, + "loss": 0.699, + "step": 23423 + }, + { + "epoch": 0.86, + "learning_rate": 1.1440290035522026e-07, + "loss": 0.6951, + "step": 23424 + }, + { + "epoch": 0.86, + "learning_rate": 1.143992382905482e-07, + "loss": 0.6894, + "step": 23425 + }, + { + "epoch": 0.86, + "learning_rate": 1.1439557622587614e-07, + "loss": 0.6984, + "step": 23426 + }, + { + "epoch": 0.86, + "learning_rate": 1.1439191416120408e-07, + "loss": 0.684, + "step": 23427 + }, + { + "epoch": 0.86, + "learning_rate": 1.1438825209653202e-07, + "loss": 0.686, + "step": 23428 + }, + { + "epoch": 0.86, + "learning_rate": 1.1438459003185996e-07, + "loss": 0.6926, + "step": 23429 + }, + { + "epoch": 0.86, + "learning_rate": 1.143809279671879e-07, + "loss": 0.6889, + "step": 23430 + }, + { + "epoch": 0.86, + "learning_rate": 1.1437726590251584e-07, + "loss": 0.6993, + "step": 23431 + }, + { + "epoch": 0.86, + "learning_rate": 1.1437360383784378e-07, + "loss": 0.6915, + "step": 23432 + }, + { + "epoch": 0.86, + "learning_rate": 1.1436994177317172e-07, + "loss": 0.6979, + "step": 23433 + }, + { + "epoch": 0.86, + "learning_rate": 1.1436627970849966e-07, + "loss": 0.6906, + "step": 23434 + }, + { + "epoch": 0.86, + "learning_rate": 1.1436261764382758e-07, + "loss": 0.7047, + "step": 23435 + }, + { + "epoch": 0.86, + "learning_rate": 1.1435895557915552e-07, + "loss": 0.6874, + "step": 23436 + }, + { + "epoch": 0.86, + "learning_rate": 1.1435529351448346e-07, + "loss": 0.7158, + "step": 23437 + }, + { + "epoch": 0.86, + "learning_rate": 1.143516314498114e-07, + "loss": 0.6789, + "step": 23438 + }, + { + "epoch": 0.86, + "learning_rate": 1.1434796938513934e-07, + "loss": 0.6924, + "step": 23439 + }, + { + "epoch": 0.86, + "learning_rate": 1.1434430732046728e-07, + "loss": 0.6909, + "step": 23440 + }, + { + "epoch": 0.86, + "learning_rate": 1.1434064525579522e-07, + "loss": 0.6929, + "step": 23441 + }, + { + "epoch": 0.86, + "learning_rate": 1.1433698319112314e-07, + "loss": 0.6804, + "step": 23442 + }, + { + "epoch": 0.86, + "learning_rate": 1.1433332112645108e-07, + "loss": 0.6934, + "step": 23443 + }, + { + "epoch": 0.86, + "learning_rate": 1.1432965906177902e-07, + "loss": 0.7015, + "step": 23444 + }, + { + "epoch": 0.86, + "learning_rate": 1.1432599699710696e-07, + "loss": 0.6968, + "step": 23445 + }, + { + "epoch": 0.86, + "learning_rate": 1.143223349324349e-07, + "loss": 0.6835, + "step": 23446 + }, + { + "epoch": 0.86, + "learning_rate": 1.1431867286776284e-07, + "loss": 0.6881, + "step": 23447 + }, + { + "epoch": 0.86, + "learning_rate": 1.1431501080309078e-07, + "loss": 0.7072, + "step": 23448 + }, + { + "epoch": 0.86, + "learning_rate": 1.1431134873841872e-07, + "loss": 0.7051, + "step": 23449 + }, + { + "epoch": 0.86, + "learning_rate": 1.1430768667374666e-07, + "loss": 0.6788, + "step": 23450 + }, + { + "epoch": 0.86, + "learning_rate": 1.143040246090746e-07, + "loss": 0.6898, + "step": 23451 + }, + { + "epoch": 0.86, + "learning_rate": 1.1430036254440254e-07, + "loss": 0.6883, + "step": 23452 + }, + { + "epoch": 0.86, + "learning_rate": 1.1429670047973046e-07, + "loss": 0.7076, + "step": 23453 + }, + { + "epoch": 0.86, + "learning_rate": 1.142930384150584e-07, + "loss": 0.691, + "step": 23454 + }, + { + "epoch": 0.86, + "learning_rate": 1.1428937635038633e-07, + "loss": 0.6958, + "step": 23455 + }, + { + "epoch": 0.86, + "learning_rate": 1.1428571428571427e-07, + "loss": 0.6983, + "step": 23456 + }, + { + "epoch": 0.86, + "learning_rate": 1.1428205222104221e-07, + "loss": 0.6998, + "step": 23457 + }, + { + "epoch": 0.86, + "learning_rate": 1.1427839015637015e-07, + "loss": 0.687, + "step": 23458 + }, + { + "epoch": 0.86, + "learning_rate": 1.142747280916981e-07, + "loss": 0.6945, + "step": 23459 + }, + { + "epoch": 0.86, + "learning_rate": 1.1427106602702603e-07, + "loss": 0.6899, + "step": 23460 + }, + { + "epoch": 0.86, + "learning_rate": 1.1426740396235397e-07, + "loss": 0.6909, + "step": 23461 + }, + { + "epoch": 0.86, + "learning_rate": 1.1426374189768191e-07, + "loss": 0.6973, + "step": 23462 + }, + { + "epoch": 0.86, + "learning_rate": 1.1426007983300985e-07, + "loss": 0.687, + "step": 23463 + }, + { + "epoch": 0.86, + "learning_rate": 1.1425641776833778e-07, + "loss": 0.7031, + "step": 23464 + }, + { + "epoch": 0.86, + "learning_rate": 1.1425275570366572e-07, + "loss": 0.6951, + "step": 23465 + }, + { + "epoch": 0.86, + "learning_rate": 1.1424909363899366e-07, + "loss": 0.6919, + "step": 23466 + }, + { + "epoch": 0.86, + "learning_rate": 1.142454315743216e-07, + "loss": 0.6905, + "step": 23467 + }, + { + "epoch": 0.86, + "learning_rate": 1.1424176950964954e-07, + "loss": 0.6991, + "step": 23468 + }, + { + "epoch": 0.86, + "learning_rate": 1.1423810744497748e-07, + "loss": 0.6958, + "step": 23469 + }, + { + "epoch": 0.86, + "learning_rate": 1.1423444538030542e-07, + "loss": 0.6915, + "step": 23470 + }, + { + "epoch": 0.86, + "learning_rate": 1.1423078331563333e-07, + "loss": 0.6939, + "step": 23471 + }, + { + "epoch": 0.86, + "learning_rate": 1.1422712125096127e-07, + "loss": 0.6926, + "step": 23472 + }, + { + "epoch": 0.86, + "learning_rate": 1.1422345918628921e-07, + "loss": 0.689, + "step": 23473 + }, + { + "epoch": 0.86, + "learning_rate": 1.1421979712161715e-07, + "loss": 0.6937, + "step": 23474 + }, + { + "epoch": 0.86, + "learning_rate": 1.142161350569451e-07, + "loss": 0.6899, + "step": 23475 + }, + { + "epoch": 0.86, + "learning_rate": 1.1421247299227303e-07, + "loss": 0.7017, + "step": 23476 + }, + { + "epoch": 0.86, + "learning_rate": 1.1420881092760097e-07, + "loss": 0.7057, + "step": 23477 + }, + { + "epoch": 0.86, + "learning_rate": 1.1420514886292891e-07, + "loss": 0.6882, + "step": 23478 + }, + { + "epoch": 0.86, + "learning_rate": 1.1420148679825685e-07, + "loss": 0.7099, + "step": 23479 + }, + { + "epoch": 0.86, + "learning_rate": 1.141978247335848e-07, + "loss": 0.6932, + "step": 23480 + }, + { + "epoch": 0.86, + "learning_rate": 1.1419416266891273e-07, + "loss": 0.6978, + "step": 23481 + }, + { + "epoch": 0.86, + "learning_rate": 1.1419050060424067e-07, + "loss": 0.6998, + "step": 23482 + }, + { + "epoch": 0.86, + "learning_rate": 1.1418683853956861e-07, + "loss": 0.6965, + "step": 23483 + }, + { + "epoch": 0.86, + "learning_rate": 1.1418317647489655e-07, + "loss": 0.6954, + "step": 23484 + }, + { + "epoch": 0.86, + "learning_rate": 1.1417951441022449e-07, + "loss": 0.6815, + "step": 23485 + }, + { + "epoch": 0.86, + "learning_rate": 1.1417585234555242e-07, + "loss": 0.7008, + "step": 23486 + }, + { + "epoch": 0.86, + "learning_rate": 1.1417219028088036e-07, + "loss": 0.6899, + "step": 23487 + }, + { + "epoch": 0.86, + "learning_rate": 1.141685282162083e-07, + "loss": 0.6984, + "step": 23488 + }, + { + "epoch": 0.86, + "learning_rate": 1.1416486615153623e-07, + "loss": 0.6959, + "step": 23489 + }, + { + "epoch": 0.86, + "learning_rate": 1.1416120408686417e-07, + "loss": 0.6886, + "step": 23490 + }, + { + "epoch": 0.86, + "learning_rate": 1.1415754202219209e-07, + "loss": 0.6911, + "step": 23491 + }, + { + "epoch": 0.86, + "learning_rate": 1.1415387995752003e-07, + "loss": 0.6945, + "step": 23492 + }, + { + "epoch": 0.86, + "learning_rate": 1.1415021789284797e-07, + "loss": 0.6892, + "step": 23493 + }, + { + "epoch": 0.86, + "learning_rate": 1.1414655582817591e-07, + "loss": 0.6882, + "step": 23494 + }, + { + "epoch": 0.86, + "learning_rate": 1.1414289376350385e-07, + "loss": 0.6917, + "step": 23495 + }, + { + "epoch": 0.86, + "learning_rate": 1.1413923169883179e-07, + "loss": 0.6924, + "step": 23496 + }, + { + "epoch": 0.86, + "learning_rate": 1.1413556963415973e-07, + "loss": 0.6868, + "step": 23497 + }, + { + "epoch": 0.86, + "learning_rate": 1.1413190756948767e-07, + "loss": 0.689, + "step": 23498 + }, + { + "epoch": 0.86, + "learning_rate": 1.1412824550481561e-07, + "loss": 0.6839, + "step": 23499 + }, + { + "epoch": 0.86, + "learning_rate": 1.1412458344014355e-07, + "loss": 0.6995, + "step": 23500 + }, + { + "epoch": 0.86, + "learning_rate": 1.1412092137547149e-07, + "loss": 0.7033, + "step": 23501 + }, + { + "epoch": 0.86, + "learning_rate": 1.1411725931079943e-07, + "loss": 0.6891, + "step": 23502 + }, + { + "epoch": 0.86, + "learning_rate": 1.1411359724612737e-07, + "loss": 0.6889, + "step": 23503 + }, + { + "epoch": 0.86, + "learning_rate": 1.1410993518145531e-07, + "loss": 0.6944, + "step": 23504 + }, + { + "epoch": 0.86, + "learning_rate": 1.1410627311678325e-07, + "loss": 0.6878, + "step": 23505 + }, + { + "epoch": 0.86, + "learning_rate": 1.1410261105211119e-07, + "loss": 0.6887, + "step": 23506 + }, + { + "epoch": 0.86, + "learning_rate": 1.140989489874391e-07, + "loss": 0.6892, + "step": 23507 + }, + { + "epoch": 0.86, + "learning_rate": 1.1409528692276705e-07, + "loss": 0.7048, + "step": 23508 + }, + { + "epoch": 0.86, + "learning_rate": 1.1409162485809499e-07, + "loss": 0.6811, + "step": 23509 + }, + { + "epoch": 0.86, + "learning_rate": 1.1408796279342293e-07, + "loss": 0.6956, + "step": 23510 + }, + { + "epoch": 0.86, + "learning_rate": 1.1408430072875087e-07, + "loss": 0.6948, + "step": 23511 + }, + { + "epoch": 0.86, + "learning_rate": 1.140806386640788e-07, + "loss": 0.6896, + "step": 23512 + }, + { + "epoch": 0.86, + "learning_rate": 1.1407697659940673e-07, + "loss": 0.7004, + "step": 23513 + }, + { + "epoch": 0.86, + "learning_rate": 1.1407331453473467e-07, + "loss": 0.6879, + "step": 23514 + }, + { + "epoch": 0.86, + "learning_rate": 1.1406965247006261e-07, + "loss": 0.6955, + "step": 23515 + }, + { + "epoch": 0.86, + "learning_rate": 1.1406599040539055e-07, + "loss": 0.6872, + "step": 23516 + }, + { + "epoch": 0.86, + "learning_rate": 1.1406232834071849e-07, + "loss": 0.6888, + "step": 23517 + }, + { + "epoch": 0.86, + "learning_rate": 1.1405866627604643e-07, + "loss": 0.693, + "step": 23518 + }, + { + "epoch": 0.86, + "learning_rate": 1.1405500421137437e-07, + "loss": 0.6919, + "step": 23519 + }, + { + "epoch": 0.86, + "learning_rate": 1.1405134214670231e-07, + "loss": 0.7025, + "step": 23520 + }, + { + "epoch": 0.86, + "learning_rate": 1.1404768008203025e-07, + "loss": 0.6967, + "step": 23521 + }, + { + "epoch": 0.86, + "learning_rate": 1.1404401801735819e-07, + "loss": 0.7047, + "step": 23522 + }, + { + "epoch": 0.86, + "learning_rate": 1.1404035595268613e-07, + "loss": 0.693, + "step": 23523 + }, + { + "epoch": 0.86, + "learning_rate": 1.1403669388801404e-07, + "loss": 0.6969, + "step": 23524 + }, + { + "epoch": 0.86, + "learning_rate": 1.1403303182334198e-07, + "loss": 0.6904, + "step": 23525 + }, + { + "epoch": 0.86, + "learning_rate": 1.1402936975866992e-07, + "loss": 0.6955, + "step": 23526 + }, + { + "epoch": 0.86, + "learning_rate": 1.1402570769399786e-07, + "loss": 0.6885, + "step": 23527 + }, + { + "epoch": 0.86, + "learning_rate": 1.140220456293258e-07, + "loss": 0.6883, + "step": 23528 + }, + { + "epoch": 0.86, + "learning_rate": 1.1401838356465374e-07, + "loss": 0.6873, + "step": 23529 + }, + { + "epoch": 0.86, + "learning_rate": 1.1401472149998168e-07, + "loss": 0.6923, + "step": 23530 + }, + { + "epoch": 0.86, + "learning_rate": 1.1401105943530962e-07, + "loss": 0.695, + "step": 23531 + }, + { + "epoch": 0.86, + "learning_rate": 1.1400739737063756e-07, + "loss": 0.6923, + "step": 23532 + }, + { + "epoch": 0.86, + "learning_rate": 1.140037353059655e-07, + "loss": 0.688, + "step": 23533 + }, + { + "epoch": 0.86, + "learning_rate": 1.1400007324129344e-07, + "loss": 0.7, + "step": 23534 + }, + { + "epoch": 0.86, + "learning_rate": 1.1399641117662137e-07, + "loss": 0.7069, + "step": 23535 + }, + { + "epoch": 0.86, + "learning_rate": 1.1399274911194931e-07, + "loss": 0.6911, + "step": 23536 + }, + { + "epoch": 0.86, + "learning_rate": 1.1398908704727725e-07, + "loss": 0.7115, + "step": 23537 + }, + { + "epoch": 0.86, + "learning_rate": 1.1398542498260519e-07, + "loss": 0.6992, + "step": 23538 + }, + { + "epoch": 0.86, + "learning_rate": 1.1398176291793313e-07, + "loss": 0.6837, + "step": 23539 + }, + { + "epoch": 0.86, + "learning_rate": 1.1397810085326107e-07, + "loss": 0.698, + "step": 23540 + }, + { + "epoch": 0.86, + "learning_rate": 1.1397443878858901e-07, + "loss": 0.6941, + "step": 23541 + }, + { + "epoch": 0.86, + "learning_rate": 1.1397077672391692e-07, + "loss": 0.6918, + "step": 23542 + }, + { + "epoch": 0.86, + "learning_rate": 1.1396711465924486e-07, + "loss": 0.6905, + "step": 23543 + }, + { + "epoch": 0.86, + "learning_rate": 1.139634525945728e-07, + "loss": 0.6949, + "step": 23544 + }, + { + "epoch": 0.86, + "learning_rate": 1.1395979052990074e-07, + "loss": 0.6929, + "step": 23545 + }, + { + "epoch": 0.86, + "learning_rate": 1.1395612846522868e-07, + "loss": 0.6914, + "step": 23546 + }, + { + "epoch": 0.86, + "learning_rate": 1.1395246640055662e-07, + "loss": 0.6938, + "step": 23547 + }, + { + "epoch": 0.86, + "learning_rate": 1.1394880433588456e-07, + "loss": 0.6949, + "step": 23548 + }, + { + "epoch": 0.86, + "learning_rate": 1.139451422712125e-07, + "loss": 0.7, + "step": 23549 + }, + { + "epoch": 0.86, + "learning_rate": 1.1394148020654044e-07, + "loss": 0.6932, + "step": 23550 + }, + { + "epoch": 0.86, + "learning_rate": 1.1393781814186838e-07, + "loss": 0.6823, + "step": 23551 + }, + { + "epoch": 0.86, + "learning_rate": 1.1393415607719632e-07, + "loss": 0.6987, + "step": 23552 + }, + { + "epoch": 0.86, + "learning_rate": 1.1393049401252426e-07, + "loss": 0.7073, + "step": 23553 + }, + { + "epoch": 0.86, + "learning_rate": 1.139268319478522e-07, + "loss": 0.6901, + "step": 23554 + }, + { + "epoch": 0.86, + "learning_rate": 1.1392316988318014e-07, + "loss": 0.6872, + "step": 23555 + }, + { + "epoch": 0.86, + "learning_rate": 1.1391950781850808e-07, + "loss": 0.7101, + "step": 23556 + }, + { + "epoch": 0.86, + "learning_rate": 1.1391584575383601e-07, + "loss": 0.6746, + "step": 23557 + }, + { + "epoch": 0.86, + "learning_rate": 1.1391218368916395e-07, + "loss": 0.6836, + "step": 23558 + }, + { + "epoch": 0.86, + "learning_rate": 1.1390852162449189e-07, + "loss": 0.6835, + "step": 23559 + }, + { + "epoch": 0.86, + "learning_rate": 1.1390485955981982e-07, + "loss": 0.6975, + "step": 23560 + }, + { + "epoch": 0.86, + "learning_rate": 1.1390119749514776e-07, + "loss": 0.6962, + "step": 23561 + }, + { + "epoch": 0.86, + "learning_rate": 1.1389753543047568e-07, + "loss": 0.6971, + "step": 23562 + }, + { + "epoch": 0.86, + "learning_rate": 1.1389387336580362e-07, + "loss": 0.6932, + "step": 23563 + }, + { + "epoch": 0.86, + "learning_rate": 1.1389021130113156e-07, + "loss": 0.7038, + "step": 23564 + }, + { + "epoch": 0.86, + "learning_rate": 1.138865492364595e-07, + "loss": 0.6961, + "step": 23565 + }, + { + "epoch": 0.86, + "learning_rate": 1.1388288717178744e-07, + "loss": 0.6891, + "step": 23566 + }, + { + "epoch": 0.86, + "learning_rate": 1.1387922510711538e-07, + "loss": 0.6905, + "step": 23567 + }, + { + "epoch": 0.86, + "learning_rate": 1.1387556304244332e-07, + "loss": 0.6828, + "step": 23568 + }, + { + "epoch": 0.86, + "learning_rate": 1.1387190097777126e-07, + "loss": 0.6923, + "step": 23569 + }, + { + "epoch": 0.86, + "learning_rate": 1.138682389130992e-07, + "loss": 0.6953, + "step": 23570 + }, + { + "epoch": 0.86, + "learning_rate": 1.1386457684842714e-07, + "loss": 0.6906, + "step": 23571 + }, + { + "epoch": 0.86, + "learning_rate": 1.1386091478375508e-07, + "loss": 0.6915, + "step": 23572 + }, + { + "epoch": 0.86, + "learning_rate": 1.1385725271908302e-07, + "loss": 0.7083, + "step": 23573 + }, + { + "epoch": 0.86, + "learning_rate": 1.1385359065441096e-07, + "loss": 0.677, + "step": 23574 + }, + { + "epoch": 0.86, + "learning_rate": 1.138499285897389e-07, + "loss": 0.6904, + "step": 23575 + }, + { + "epoch": 0.86, + "learning_rate": 1.1384626652506684e-07, + "loss": 0.6927, + "step": 23576 + }, + { + "epoch": 0.86, + "learning_rate": 1.1384260446039478e-07, + "loss": 0.6954, + "step": 23577 + }, + { + "epoch": 0.86, + "learning_rate": 1.138389423957227e-07, + "loss": 0.6837, + "step": 23578 + }, + { + "epoch": 0.86, + "learning_rate": 1.1383528033105063e-07, + "loss": 0.7118, + "step": 23579 + }, + { + "epoch": 0.86, + "learning_rate": 1.1383161826637857e-07, + "loss": 0.6826, + "step": 23580 + }, + { + "epoch": 0.86, + "learning_rate": 1.1382795620170651e-07, + "loss": 0.674, + "step": 23581 + }, + { + "epoch": 0.86, + "learning_rate": 1.1382429413703445e-07, + "loss": 0.7052, + "step": 23582 + }, + { + "epoch": 0.86, + "learning_rate": 1.138206320723624e-07, + "loss": 0.6771, + "step": 23583 + }, + { + "epoch": 0.86, + "learning_rate": 1.1381697000769032e-07, + "loss": 0.6926, + "step": 23584 + }, + { + "epoch": 0.86, + "learning_rate": 1.1381330794301826e-07, + "loss": 0.6986, + "step": 23585 + }, + { + "epoch": 0.86, + "learning_rate": 1.138096458783462e-07, + "loss": 0.6881, + "step": 23586 + }, + { + "epoch": 0.86, + "learning_rate": 1.1380598381367414e-07, + "loss": 0.6783, + "step": 23587 + }, + { + "epoch": 0.86, + "learning_rate": 1.1380232174900208e-07, + "loss": 0.6883, + "step": 23588 + }, + { + "epoch": 0.86, + "learning_rate": 1.1379865968433002e-07, + "loss": 0.692, + "step": 23589 + }, + { + "epoch": 0.86, + "learning_rate": 1.1379499761965796e-07, + "loss": 0.7064, + "step": 23590 + }, + { + "epoch": 0.86, + "learning_rate": 1.137913355549859e-07, + "loss": 0.6808, + "step": 23591 + }, + { + "epoch": 0.86, + "learning_rate": 1.1378767349031384e-07, + "loss": 0.6939, + "step": 23592 + }, + { + "epoch": 0.86, + "learning_rate": 1.1378401142564178e-07, + "loss": 0.6755, + "step": 23593 + }, + { + "epoch": 0.86, + "learning_rate": 1.1378034936096972e-07, + "loss": 0.6973, + "step": 23594 + }, + { + "epoch": 0.86, + "learning_rate": 1.1377668729629766e-07, + "loss": 0.7025, + "step": 23595 + }, + { + "epoch": 0.86, + "learning_rate": 1.1377302523162557e-07, + "loss": 0.6962, + "step": 23596 + }, + { + "epoch": 0.86, + "learning_rate": 1.1376936316695351e-07, + "loss": 0.7031, + "step": 23597 + }, + { + "epoch": 0.86, + "learning_rate": 1.1376570110228145e-07, + "loss": 0.6929, + "step": 23598 + }, + { + "epoch": 0.86, + "learning_rate": 1.137620390376094e-07, + "loss": 0.6964, + "step": 23599 + }, + { + "epoch": 0.86, + "learning_rate": 1.1375837697293733e-07, + "loss": 0.6849, + "step": 23600 + }, + { + "epoch": 0.86, + "learning_rate": 1.1375471490826527e-07, + "loss": 0.692, + "step": 23601 + }, + { + "epoch": 0.86, + "learning_rate": 1.1375105284359321e-07, + "loss": 0.6951, + "step": 23602 + }, + { + "epoch": 0.86, + "learning_rate": 1.1374739077892115e-07, + "loss": 0.6796, + "step": 23603 + }, + { + "epoch": 0.86, + "learning_rate": 1.1374372871424909e-07, + "loss": 0.672, + "step": 23604 + }, + { + "epoch": 0.86, + "learning_rate": 1.1374006664957703e-07, + "loss": 0.6852, + "step": 23605 + }, + { + "epoch": 0.86, + "learning_rate": 1.1373640458490496e-07, + "loss": 0.6851, + "step": 23606 + }, + { + "epoch": 0.86, + "learning_rate": 1.137327425202329e-07, + "loss": 0.6865, + "step": 23607 + }, + { + "epoch": 0.86, + "learning_rate": 1.1372908045556084e-07, + "loss": 0.703, + "step": 23608 + }, + { + "epoch": 0.86, + "learning_rate": 1.1372541839088878e-07, + "loss": 0.7005, + "step": 23609 + }, + { + "epoch": 0.86, + "learning_rate": 1.1372175632621672e-07, + "loss": 0.6923, + "step": 23610 + }, + { + "epoch": 0.86, + "learning_rate": 1.1371809426154466e-07, + "loss": 0.7005, + "step": 23611 + }, + { + "epoch": 0.86, + "learning_rate": 1.137144321968726e-07, + "loss": 0.6997, + "step": 23612 + }, + { + "epoch": 0.86, + "learning_rate": 1.1371077013220054e-07, + "loss": 0.6987, + "step": 23613 + }, + { + "epoch": 0.86, + "learning_rate": 1.1370710806752845e-07, + "loss": 0.6892, + "step": 23614 + }, + { + "epoch": 0.86, + "learning_rate": 1.1370344600285639e-07, + "loss": 0.6873, + "step": 23615 + }, + { + "epoch": 0.86, + "learning_rate": 1.1369978393818433e-07, + "loss": 0.6902, + "step": 23616 + }, + { + "epoch": 0.86, + "learning_rate": 1.1369612187351227e-07, + "loss": 0.6849, + "step": 23617 + }, + { + "epoch": 0.86, + "learning_rate": 1.1369245980884021e-07, + "loss": 0.6822, + "step": 23618 + }, + { + "epoch": 0.86, + "learning_rate": 1.1368879774416815e-07, + "loss": 0.698, + "step": 23619 + }, + { + "epoch": 0.86, + "learning_rate": 1.1368513567949609e-07, + "loss": 0.7064, + "step": 23620 + }, + { + "epoch": 0.86, + "learning_rate": 1.1368147361482403e-07, + "loss": 0.6954, + "step": 23621 + }, + { + "epoch": 0.86, + "learning_rate": 1.1367781155015197e-07, + "loss": 0.7043, + "step": 23622 + }, + { + "epoch": 0.86, + "learning_rate": 1.1367414948547991e-07, + "loss": 0.6934, + "step": 23623 + }, + { + "epoch": 0.86, + "learning_rate": 1.1367048742080785e-07, + "loss": 0.6855, + "step": 23624 + }, + { + "epoch": 0.86, + "learning_rate": 1.1366682535613579e-07, + "loss": 0.6848, + "step": 23625 + }, + { + "epoch": 0.86, + "learning_rate": 1.1366316329146373e-07, + "loss": 0.6944, + "step": 23626 + }, + { + "epoch": 0.86, + "learning_rate": 1.1365950122679167e-07, + "loss": 0.6935, + "step": 23627 + }, + { + "epoch": 0.86, + "learning_rate": 1.136558391621196e-07, + "loss": 0.6911, + "step": 23628 + }, + { + "epoch": 0.86, + "learning_rate": 1.1365217709744754e-07, + "loss": 0.6975, + "step": 23629 + }, + { + "epoch": 0.86, + "learning_rate": 1.1364851503277548e-07, + "loss": 0.6898, + "step": 23630 + }, + { + "epoch": 0.86, + "learning_rate": 1.1364485296810342e-07, + "loss": 0.6909, + "step": 23631 + }, + { + "epoch": 0.86, + "learning_rate": 1.1364119090343134e-07, + "loss": 0.6925, + "step": 23632 + }, + { + "epoch": 0.86, + "learning_rate": 1.1363752883875927e-07, + "loss": 0.6997, + "step": 23633 + }, + { + "epoch": 0.86, + "learning_rate": 1.1363386677408721e-07, + "loss": 0.6831, + "step": 23634 + }, + { + "epoch": 0.86, + "learning_rate": 1.1363020470941515e-07, + "loss": 0.6916, + "step": 23635 + }, + { + "epoch": 0.86, + "learning_rate": 1.1362654264474309e-07, + "loss": 0.6952, + "step": 23636 + }, + { + "epoch": 0.86, + "learning_rate": 1.1362288058007103e-07, + "loss": 0.6838, + "step": 23637 + }, + { + "epoch": 0.86, + "learning_rate": 1.1361921851539897e-07, + "loss": 0.7088, + "step": 23638 + }, + { + "epoch": 0.86, + "learning_rate": 1.1361555645072691e-07, + "loss": 0.689, + "step": 23639 + }, + { + "epoch": 0.86, + "learning_rate": 1.1361189438605485e-07, + "loss": 0.7073, + "step": 23640 + }, + { + "epoch": 0.86, + "learning_rate": 1.1360823232138279e-07, + "loss": 0.699, + "step": 23641 + }, + { + "epoch": 0.86, + "learning_rate": 1.1360457025671073e-07, + "loss": 0.6929, + "step": 23642 + }, + { + "epoch": 0.87, + "learning_rate": 1.1360090819203867e-07, + "loss": 0.7082, + "step": 23643 + }, + { + "epoch": 0.87, + "learning_rate": 1.1359724612736661e-07, + "loss": 0.681, + "step": 23644 + }, + { + "epoch": 0.87, + "learning_rate": 1.1359358406269455e-07, + "loss": 0.7007, + "step": 23645 + }, + { + "epoch": 0.87, + "learning_rate": 1.1358992199802249e-07, + "loss": 0.693, + "step": 23646 + }, + { + "epoch": 0.87, + "learning_rate": 1.1358625993335043e-07, + "loss": 0.6986, + "step": 23647 + }, + { + "epoch": 0.87, + "learning_rate": 1.1358259786867837e-07, + "loss": 0.6858, + "step": 23648 + }, + { + "epoch": 0.87, + "learning_rate": 1.1357893580400631e-07, + "loss": 0.6831, + "step": 23649 + }, + { + "epoch": 0.87, + "learning_rate": 1.1357527373933422e-07, + "loss": 0.6939, + "step": 23650 + }, + { + "epoch": 0.87, + "learning_rate": 1.1357161167466216e-07, + "loss": 0.6967, + "step": 23651 + }, + { + "epoch": 0.87, + "learning_rate": 1.135679496099901e-07, + "loss": 0.6923, + "step": 23652 + }, + { + "epoch": 0.87, + "learning_rate": 1.1356428754531804e-07, + "loss": 0.7047, + "step": 23653 + }, + { + "epoch": 0.87, + "learning_rate": 1.1356062548064598e-07, + "loss": 0.6882, + "step": 23654 + }, + { + "epoch": 0.87, + "learning_rate": 1.1355696341597391e-07, + "loss": 0.688, + "step": 23655 + }, + { + "epoch": 0.87, + "learning_rate": 1.1355330135130185e-07, + "loss": 0.6812, + "step": 23656 + }, + { + "epoch": 0.87, + "learning_rate": 1.1354963928662979e-07, + "loss": 0.6965, + "step": 23657 + }, + { + "epoch": 0.87, + "learning_rate": 1.1354597722195773e-07, + "loss": 0.6987, + "step": 23658 + }, + { + "epoch": 0.87, + "learning_rate": 1.1354231515728567e-07, + "loss": 0.701, + "step": 23659 + }, + { + "epoch": 0.87, + "learning_rate": 1.1353865309261361e-07, + "loss": 0.6918, + "step": 23660 + }, + { + "epoch": 0.87, + "learning_rate": 1.1353499102794155e-07, + "loss": 0.6934, + "step": 23661 + }, + { + "epoch": 0.87, + "learning_rate": 1.1353132896326949e-07, + "loss": 0.698, + "step": 23662 + }, + { + "epoch": 0.87, + "learning_rate": 1.1352766689859743e-07, + "loss": 0.7, + "step": 23663 + }, + { + "epoch": 0.87, + "learning_rate": 1.1352400483392537e-07, + "loss": 0.6951, + "step": 23664 + }, + { + "epoch": 0.87, + "learning_rate": 1.1352034276925331e-07, + "loss": 0.6929, + "step": 23665 + }, + { + "epoch": 0.87, + "learning_rate": 1.1351668070458125e-07, + "loss": 0.7107, + "step": 23666 + }, + { + "epoch": 0.87, + "learning_rate": 1.1351301863990919e-07, + "loss": 0.697, + "step": 23667 + }, + { + "epoch": 0.87, + "learning_rate": 1.135093565752371e-07, + "loss": 0.6916, + "step": 23668 + }, + { + "epoch": 0.87, + "learning_rate": 1.1350569451056504e-07, + "loss": 0.6709, + "step": 23669 + }, + { + "epoch": 0.87, + "learning_rate": 1.1350203244589298e-07, + "loss": 0.6924, + "step": 23670 + }, + { + "epoch": 0.87, + "learning_rate": 1.1349837038122092e-07, + "loss": 0.6843, + "step": 23671 + }, + { + "epoch": 0.87, + "learning_rate": 1.1349470831654886e-07, + "loss": 0.7018, + "step": 23672 + }, + { + "epoch": 0.87, + "learning_rate": 1.134910462518768e-07, + "loss": 0.7012, + "step": 23673 + }, + { + "epoch": 0.87, + "learning_rate": 1.1348738418720474e-07, + "loss": 0.6981, + "step": 23674 + }, + { + "epoch": 0.87, + "learning_rate": 1.1348372212253268e-07, + "loss": 0.7003, + "step": 23675 + }, + { + "epoch": 0.87, + "learning_rate": 1.1348006005786062e-07, + "loss": 0.6923, + "step": 23676 + }, + { + "epoch": 0.87, + "learning_rate": 1.1347639799318855e-07, + "loss": 0.6971, + "step": 23677 + }, + { + "epoch": 0.87, + "learning_rate": 1.1347273592851649e-07, + "loss": 0.6939, + "step": 23678 + }, + { + "epoch": 0.87, + "learning_rate": 1.1346907386384443e-07, + "loss": 0.7102, + "step": 23679 + }, + { + "epoch": 0.87, + "learning_rate": 1.1346541179917237e-07, + "loss": 0.7012, + "step": 23680 + }, + { + "epoch": 0.87, + "learning_rate": 1.1346174973450031e-07, + "loss": 0.694, + "step": 23681 + }, + { + "epoch": 0.87, + "learning_rate": 1.1345808766982825e-07, + "loss": 0.6918, + "step": 23682 + }, + { + "epoch": 0.87, + "learning_rate": 1.1345442560515619e-07, + "loss": 0.7081, + "step": 23683 + }, + { + "epoch": 0.87, + "learning_rate": 1.1345076354048413e-07, + "loss": 0.6744, + "step": 23684 + }, + { + "epoch": 0.87, + "learning_rate": 1.1344710147581207e-07, + "loss": 0.7002, + "step": 23685 + }, + { + "epoch": 0.87, + "learning_rate": 1.1344343941113998e-07, + "loss": 0.6952, + "step": 23686 + }, + { + "epoch": 0.87, + "learning_rate": 1.1343977734646792e-07, + "loss": 0.7018, + "step": 23687 + }, + { + "epoch": 0.87, + "learning_rate": 1.1343611528179586e-07, + "loss": 0.7054, + "step": 23688 + }, + { + "epoch": 0.87, + "learning_rate": 1.134324532171238e-07, + "loss": 0.6889, + "step": 23689 + }, + { + "epoch": 0.87, + "learning_rate": 1.1342879115245174e-07, + "loss": 0.6797, + "step": 23690 + }, + { + "epoch": 0.87, + "learning_rate": 1.1342512908777968e-07, + "loss": 0.686, + "step": 23691 + }, + { + "epoch": 0.87, + "learning_rate": 1.1342146702310762e-07, + "loss": 0.6906, + "step": 23692 + }, + { + "epoch": 0.87, + "learning_rate": 1.1341780495843556e-07, + "loss": 0.6783, + "step": 23693 + }, + { + "epoch": 0.87, + "learning_rate": 1.134141428937635e-07, + "loss": 0.6957, + "step": 23694 + }, + { + "epoch": 0.87, + "learning_rate": 1.1341048082909144e-07, + "loss": 0.6861, + "step": 23695 + }, + { + "epoch": 0.87, + "learning_rate": 1.1340681876441938e-07, + "loss": 0.691, + "step": 23696 + }, + { + "epoch": 0.87, + "learning_rate": 1.1340315669974732e-07, + "loss": 0.7002, + "step": 23697 + }, + { + "epoch": 0.87, + "learning_rate": 1.1339949463507526e-07, + "loss": 0.7028, + "step": 23698 + }, + { + "epoch": 0.87, + "learning_rate": 1.1339583257040319e-07, + "loss": 0.7011, + "step": 23699 + }, + { + "epoch": 0.87, + "learning_rate": 1.1339217050573113e-07, + "loss": 0.6901, + "step": 23700 + }, + { + "epoch": 0.87, + "learning_rate": 1.1338850844105907e-07, + "loss": 0.6852, + "step": 23701 + }, + { + "epoch": 0.87, + "learning_rate": 1.1338484637638701e-07, + "loss": 0.6957, + "step": 23702 + }, + { + "epoch": 0.87, + "learning_rate": 1.1338118431171495e-07, + "loss": 0.6923, + "step": 23703 + }, + { + "epoch": 0.87, + "learning_rate": 1.1337752224704286e-07, + "loss": 0.699, + "step": 23704 + }, + { + "epoch": 0.87, + "learning_rate": 1.133738601823708e-07, + "loss": 0.6903, + "step": 23705 + }, + { + "epoch": 0.87, + "learning_rate": 1.1337019811769874e-07, + "loss": 0.6952, + "step": 23706 + }, + { + "epoch": 0.87, + "learning_rate": 1.1336653605302668e-07, + "loss": 0.6808, + "step": 23707 + }, + { + "epoch": 0.87, + "learning_rate": 1.1336287398835462e-07, + "loss": 0.7101, + "step": 23708 + }, + { + "epoch": 0.87, + "learning_rate": 1.1335921192368256e-07, + "loss": 0.7029, + "step": 23709 + }, + { + "epoch": 0.87, + "learning_rate": 1.133555498590105e-07, + "loss": 0.69, + "step": 23710 + }, + { + "epoch": 0.87, + "learning_rate": 1.1335188779433844e-07, + "loss": 0.6915, + "step": 23711 + }, + { + "epoch": 0.87, + "learning_rate": 1.1334822572966638e-07, + "loss": 0.705, + "step": 23712 + }, + { + "epoch": 0.87, + "learning_rate": 1.1334456366499432e-07, + "loss": 0.6851, + "step": 23713 + }, + { + "epoch": 0.87, + "learning_rate": 1.1334090160032226e-07, + "loss": 0.6831, + "step": 23714 + }, + { + "epoch": 0.87, + "learning_rate": 1.133372395356502e-07, + "loss": 0.6935, + "step": 23715 + }, + { + "epoch": 0.87, + "learning_rate": 1.1333357747097814e-07, + "loss": 0.7023, + "step": 23716 + }, + { + "epoch": 0.87, + "learning_rate": 1.1332991540630608e-07, + "loss": 0.6947, + "step": 23717 + }, + { + "epoch": 0.87, + "learning_rate": 1.1332625334163402e-07, + "loss": 0.7018, + "step": 23718 + }, + { + "epoch": 0.87, + "learning_rate": 1.1332259127696196e-07, + "loss": 0.6994, + "step": 23719 + }, + { + "epoch": 0.87, + "learning_rate": 1.1331892921228989e-07, + "loss": 0.6943, + "step": 23720 + }, + { + "epoch": 0.87, + "learning_rate": 1.1331526714761783e-07, + "loss": 0.6943, + "step": 23721 + }, + { + "epoch": 0.87, + "learning_rate": 1.1331160508294575e-07, + "loss": 0.7027, + "step": 23722 + }, + { + "epoch": 0.87, + "learning_rate": 1.1330794301827369e-07, + "loss": 0.7009, + "step": 23723 + }, + { + "epoch": 0.87, + "learning_rate": 1.1330428095360163e-07, + "loss": 0.6997, + "step": 23724 + }, + { + "epoch": 0.87, + "learning_rate": 1.1330061888892957e-07, + "loss": 0.6995, + "step": 23725 + }, + { + "epoch": 0.87, + "learning_rate": 1.132969568242575e-07, + "loss": 0.689, + "step": 23726 + }, + { + "epoch": 0.87, + "learning_rate": 1.1329329475958544e-07, + "loss": 0.6963, + "step": 23727 + }, + { + "epoch": 0.87, + "learning_rate": 1.1328963269491338e-07, + "loss": 0.7013, + "step": 23728 + }, + { + "epoch": 0.87, + "learning_rate": 1.1328597063024132e-07, + "loss": 0.693, + "step": 23729 + }, + { + "epoch": 0.87, + "learning_rate": 1.1328230856556926e-07, + "loss": 0.6978, + "step": 23730 + }, + { + "epoch": 0.87, + "learning_rate": 1.132786465008972e-07, + "loss": 0.6981, + "step": 23731 + }, + { + "epoch": 0.87, + "learning_rate": 1.1327498443622514e-07, + "loss": 0.6997, + "step": 23732 + }, + { + "epoch": 0.87, + "learning_rate": 1.1327132237155308e-07, + "loss": 0.6896, + "step": 23733 + }, + { + "epoch": 0.87, + "learning_rate": 1.1326766030688102e-07, + "loss": 0.6953, + "step": 23734 + }, + { + "epoch": 0.87, + "learning_rate": 1.1326399824220896e-07, + "loss": 0.6821, + "step": 23735 + }, + { + "epoch": 0.87, + "learning_rate": 1.132603361775369e-07, + "loss": 0.6927, + "step": 23736 + }, + { + "epoch": 0.87, + "learning_rate": 1.1325667411286484e-07, + "loss": 0.694, + "step": 23737 + }, + { + "epoch": 0.87, + "learning_rate": 1.1325301204819278e-07, + "loss": 0.6935, + "step": 23738 + }, + { + "epoch": 0.87, + "learning_rate": 1.1324934998352072e-07, + "loss": 0.6807, + "step": 23739 + }, + { + "epoch": 0.87, + "learning_rate": 1.1324568791884863e-07, + "loss": 0.6969, + "step": 23740 + }, + { + "epoch": 0.87, + "learning_rate": 1.1324202585417657e-07, + "loss": 0.6981, + "step": 23741 + }, + { + "epoch": 0.87, + "learning_rate": 1.1323836378950451e-07, + "loss": 0.6922, + "step": 23742 + }, + { + "epoch": 0.87, + "learning_rate": 1.1323470172483245e-07, + "loss": 0.6954, + "step": 23743 + }, + { + "epoch": 0.87, + "learning_rate": 1.1323103966016039e-07, + "loss": 0.7057, + "step": 23744 + }, + { + "epoch": 0.87, + "learning_rate": 1.1322737759548833e-07, + "loss": 0.6982, + "step": 23745 + }, + { + "epoch": 0.87, + "learning_rate": 1.1322371553081627e-07, + "loss": 0.6954, + "step": 23746 + }, + { + "epoch": 0.87, + "learning_rate": 1.1322005346614421e-07, + "loss": 0.6982, + "step": 23747 + }, + { + "epoch": 0.87, + "learning_rate": 1.1321639140147214e-07, + "loss": 0.7122, + "step": 23748 + }, + { + "epoch": 0.87, + "learning_rate": 1.1321272933680008e-07, + "loss": 0.6827, + "step": 23749 + }, + { + "epoch": 0.87, + "learning_rate": 1.1320906727212802e-07, + "loss": 0.6905, + "step": 23750 + }, + { + "epoch": 0.87, + "learning_rate": 1.1320540520745596e-07, + "loss": 0.6917, + "step": 23751 + }, + { + "epoch": 0.87, + "learning_rate": 1.132017431427839e-07, + "loss": 0.6876, + "step": 23752 + }, + { + "epoch": 0.87, + "learning_rate": 1.1319808107811184e-07, + "loss": 0.6895, + "step": 23753 + }, + { + "epoch": 0.87, + "learning_rate": 1.1319441901343978e-07, + "loss": 0.6888, + "step": 23754 + }, + { + "epoch": 0.87, + "learning_rate": 1.1319075694876772e-07, + "loss": 0.6914, + "step": 23755 + }, + { + "epoch": 0.87, + "learning_rate": 1.1318709488409566e-07, + "loss": 0.6751, + "step": 23756 + }, + { + "epoch": 0.87, + "learning_rate": 1.131834328194236e-07, + "loss": 0.6916, + "step": 23757 + }, + { + "epoch": 0.87, + "learning_rate": 1.1317977075475151e-07, + "loss": 0.6929, + "step": 23758 + }, + { + "epoch": 0.87, + "learning_rate": 1.1317610869007945e-07, + "loss": 0.6954, + "step": 23759 + }, + { + "epoch": 0.87, + "learning_rate": 1.1317244662540739e-07, + "loss": 0.689, + "step": 23760 + }, + { + "epoch": 0.87, + "learning_rate": 1.1316878456073533e-07, + "loss": 0.6922, + "step": 23761 + }, + { + "epoch": 0.87, + "learning_rate": 1.1316512249606327e-07, + "loss": 0.6865, + "step": 23762 + }, + { + "epoch": 0.87, + "learning_rate": 1.1316146043139121e-07, + "loss": 0.6994, + "step": 23763 + }, + { + "epoch": 0.87, + "learning_rate": 1.1315779836671915e-07, + "loss": 0.702, + "step": 23764 + }, + { + "epoch": 0.87, + "learning_rate": 1.1315413630204709e-07, + "loss": 0.6885, + "step": 23765 + }, + { + "epoch": 0.87, + "learning_rate": 1.1315047423737503e-07, + "loss": 0.7051, + "step": 23766 + }, + { + "epoch": 0.87, + "learning_rate": 1.1314681217270297e-07, + "loss": 0.6892, + "step": 23767 + }, + { + "epoch": 0.87, + "learning_rate": 1.1314315010803091e-07, + "loss": 0.6969, + "step": 23768 + }, + { + "epoch": 0.87, + "learning_rate": 1.1313948804335885e-07, + "loss": 0.7053, + "step": 23769 + }, + { + "epoch": 0.87, + "learning_rate": 1.1313582597868678e-07, + "loss": 0.7115, + "step": 23770 + }, + { + "epoch": 0.87, + "learning_rate": 1.1313216391401472e-07, + "loss": 0.689, + "step": 23771 + }, + { + "epoch": 0.87, + "learning_rate": 1.1312850184934266e-07, + "loss": 0.6768, + "step": 23772 + }, + { + "epoch": 0.87, + "learning_rate": 1.131248397846706e-07, + "loss": 0.6895, + "step": 23773 + }, + { + "epoch": 0.87, + "learning_rate": 1.1312117771999854e-07, + "loss": 0.6742, + "step": 23774 + }, + { + "epoch": 0.87, + "learning_rate": 1.1311751565532648e-07, + "loss": 0.6965, + "step": 23775 + }, + { + "epoch": 0.87, + "learning_rate": 1.1311385359065439e-07, + "loss": 0.702, + "step": 23776 + }, + { + "epoch": 0.87, + "learning_rate": 1.1311019152598233e-07, + "loss": 0.7052, + "step": 23777 + }, + { + "epoch": 0.87, + "learning_rate": 1.1310652946131027e-07, + "loss": 0.7031, + "step": 23778 + }, + { + "epoch": 0.87, + "learning_rate": 1.1310286739663821e-07, + "loss": 0.7024, + "step": 23779 + }, + { + "epoch": 0.87, + "learning_rate": 1.1309920533196615e-07, + "loss": 0.701, + "step": 23780 + }, + { + "epoch": 0.87, + "learning_rate": 1.1309554326729409e-07, + "loss": 0.6967, + "step": 23781 + }, + { + "epoch": 0.87, + "learning_rate": 1.1309188120262203e-07, + "loss": 0.6821, + "step": 23782 + }, + { + "epoch": 0.87, + "learning_rate": 1.1308821913794997e-07, + "loss": 0.6901, + "step": 23783 + }, + { + "epoch": 0.87, + "learning_rate": 1.1308455707327791e-07, + "loss": 0.6984, + "step": 23784 + }, + { + "epoch": 0.87, + "learning_rate": 1.1308089500860585e-07, + "loss": 0.6939, + "step": 23785 + }, + { + "epoch": 0.87, + "learning_rate": 1.1307723294393379e-07, + "loss": 0.7062, + "step": 23786 + }, + { + "epoch": 0.87, + "learning_rate": 1.1307357087926173e-07, + "loss": 0.6795, + "step": 23787 + }, + { + "epoch": 0.87, + "learning_rate": 1.1306990881458967e-07, + "loss": 0.6908, + "step": 23788 + }, + { + "epoch": 0.87, + "learning_rate": 1.1306624674991761e-07, + "loss": 0.6813, + "step": 23789 + }, + { + "epoch": 0.87, + "learning_rate": 1.1306258468524555e-07, + "loss": 0.6982, + "step": 23790 + }, + { + "epoch": 0.87, + "learning_rate": 1.1305892262057348e-07, + "loss": 0.6758, + "step": 23791 + }, + { + "epoch": 0.87, + "learning_rate": 1.1305526055590142e-07, + "loss": 0.697, + "step": 23792 + }, + { + "epoch": 0.87, + "learning_rate": 1.1305159849122936e-07, + "loss": 0.6985, + "step": 23793 + }, + { + "epoch": 0.87, + "learning_rate": 1.1304793642655728e-07, + "loss": 0.7017, + "step": 23794 + }, + { + "epoch": 0.87, + "learning_rate": 1.1304427436188522e-07, + "loss": 0.6871, + "step": 23795 + }, + { + "epoch": 0.87, + "learning_rate": 1.1304061229721316e-07, + "loss": 0.6927, + "step": 23796 + }, + { + "epoch": 0.87, + "learning_rate": 1.1303695023254109e-07, + "loss": 0.6972, + "step": 23797 + }, + { + "epoch": 0.87, + "learning_rate": 1.1303328816786903e-07, + "loss": 0.6871, + "step": 23798 + }, + { + "epoch": 0.87, + "learning_rate": 1.1302962610319697e-07, + "loss": 0.6913, + "step": 23799 + }, + { + "epoch": 0.87, + "learning_rate": 1.1302596403852491e-07, + "loss": 0.7049, + "step": 23800 + }, + { + "epoch": 0.87, + "learning_rate": 1.1302230197385285e-07, + "loss": 0.6942, + "step": 23801 + }, + { + "epoch": 0.87, + "learning_rate": 1.1301863990918079e-07, + "loss": 0.6935, + "step": 23802 + }, + { + "epoch": 0.87, + "learning_rate": 1.1301497784450873e-07, + "loss": 0.7039, + "step": 23803 + }, + { + "epoch": 0.87, + "learning_rate": 1.1301131577983667e-07, + "loss": 0.6897, + "step": 23804 + }, + { + "epoch": 0.87, + "learning_rate": 1.1300765371516461e-07, + "loss": 0.7, + "step": 23805 + }, + { + "epoch": 0.87, + "learning_rate": 1.1300399165049255e-07, + "loss": 0.6898, + "step": 23806 + }, + { + "epoch": 0.87, + "learning_rate": 1.1300032958582049e-07, + "loss": 0.691, + "step": 23807 + }, + { + "epoch": 0.87, + "learning_rate": 1.1299666752114843e-07, + "loss": 0.6944, + "step": 23808 + }, + { + "epoch": 0.87, + "learning_rate": 1.1299300545647637e-07, + "loss": 0.7038, + "step": 23809 + }, + { + "epoch": 0.87, + "learning_rate": 1.1298934339180431e-07, + "loss": 0.6813, + "step": 23810 + }, + { + "epoch": 0.87, + "learning_rate": 1.1298568132713225e-07, + "loss": 0.6862, + "step": 23811 + }, + { + "epoch": 0.87, + "learning_rate": 1.1298201926246016e-07, + "loss": 0.6836, + "step": 23812 + }, + { + "epoch": 0.87, + "learning_rate": 1.129783571977881e-07, + "loss": 0.6737, + "step": 23813 + }, + { + "epoch": 0.87, + "learning_rate": 1.1297469513311604e-07, + "loss": 0.6885, + "step": 23814 + }, + { + "epoch": 0.87, + "learning_rate": 1.1297103306844398e-07, + "loss": 0.6921, + "step": 23815 + }, + { + "epoch": 0.87, + "learning_rate": 1.1296737100377192e-07, + "loss": 0.6924, + "step": 23816 + }, + { + "epoch": 0.87, + "learning_rate": 1.1296370893909986e-07, + "loss": 0.6966, + "step": 23817 + }, + { + "epoch": 0.87, + "learning_rate": 1.129600468744278e-07, + "loss": 0.6927, + "step": 23818 + }, + { + "epoch": 0.87, + "learning_rate": 1.1295638480975573e-07, + "loss": 0.6969, + "step": 23819 + }, + { + "epoch": 0.87, + "learning_rate": 1.1295272274508367e-07, + "loss": 0.6937, + "step": 23820 + }, + { + "epoch": 0.87, + "learning_rate": 1.1294906068041161e-07, + "loss": 0.7001, + "step": 23821 + }, + { + "epoch": 0.87, + "learning_rate": 1.1294539861573955e-07, + "loss": 0.6978, + "step": 23822 + }, + { + "epoch": 0.87, + "learning_rate": 1.1294173655106749e-07, + "loss": 0.6943, + "step": 23823 + }, + { + "epoch": 0.87, + "learning_rate": 1.1293807448639543e-07, + "loss": 0.696, + "step": 23824 + }, + { + "epoch": 0.87, + "learning_rate": 1.1293441242172337e-07, + "loss": 0.6827, + "step": 23825 + }, + { + "epoch": 0.87, + "learning_rate": 1.1293075035705131e-07, + "loss": 0.6993, + "step": 23826 + }, + { + "epoch": 0.87, + "learning_rate": 1.1292708829237925e-07, + "loss": 0.7012, + "step": 23827 + }, + { + "epoch": 0.87, + "learning_rate": 1.1292342622770719e-07, + "loss": 0.7059, + "step": 23828 + }, + { + "epoch": 0.87, + "learning_rate": 1.1291976416303513e-07, + "loss": 0.6837, + "step": 23829 + }, + { + "epoch": 0.87, + "learning_rate": 1.1291610209836304e-07, + "loss": 0.6893, + "step": 23830 + }, + { + "epoch": 0.87, + "learning_rate": 1.1291244003369098e-07, + "loss": 0.6953, + "step": 23831 + }, + { + "epoch": 0.87, + "learning_rate": 1.1290877796901892e-07, + "loss": 0.6907, + "step": 23832 + }, + { + "epoch": 0.87, + "learning_rate": 1.1290511590434686e-07, + "loss": 0.6938, + "step": 23833 + }, + { + "epoch": 0.87, + "learning_rate": 1.129014538396748e-07, + "loss": 0.6918, + "step": 23834 + }, + { + "epoch": 0.87, + "learning_rate": 1.1289779177500274e-07, + "loss": 0.6895, + "step": 23835 + }, + { + "epoch": 0.87, + "learning_rate": 1.1289412971033068e-07, + "loss": 0.6917, + "step": 23836 + }, + { + "epoch": 0.87, + "learning_rate": 1.1289046764565862e-07, + "loss": 0.6781, + "step": 23837 + }, + { + "epoch": 0.87, + "learning_rate": 1.1288680558098656e-07, + "loss": 0.6949, + "step": 23838 + }, + { + "epoch": 0.87, + "learning_rate": 1.128831435163145e-07, + "loss": 0.6911, + "step": 23839 + }, + { + "epoch": 0.87, + "learning_rate": 1.1287948145164243e-07, + "loss": 0.6967, + "step": 23840 + }, + { + "epoch": 0.87, + "learning_rate": 1.1287581938697037e-07, + "loss": 0.6929, + "step": 23841 + }, + { + "epoch": 0.87, + "learning_rate": 1.128721573222983e-07, + "loss": 0.6914, + "step": 23842 + }, + { + "epoch": 0.87, + "learning_rate": 1.1286849525762625e-07, + "loss": 0.6999, + "step": 23843 + }, + { + "epoch": 0.87, + "learning_rate": 1.1286483319295419e-07, + "loss": 0.6989, + "step": 23844 + }, + { + "epoch": 0.87, + "learning_rate": 1.1286117112828213e-07, + "loss": 0.6982, + "step": 23845 + }, + { + "epoch": 0.87, + "learning_rate": 1.1285750906361007e-07, + "loss": 0.6943, + "step": 23846 + }, + { + "epoch": 0.87, + "learning_rate": 1.12853846998938e-07, + "loss": 0.7012, + "step": 23847 + }, + { + "epoch": 0.87, + "learning_rate": 1.1285018493426592e-07, + "loss": 0.6902, + "step": 23848 + }, + { + "epoch": 0.87, + "learning_rate": 1.1284652286959386e-07, + "loss": 0.6859, + "step": 23849 + }, + { + "epoch": 0.87, + "learning_rate": 1.128428608049218e-07, + "loss": 0.7018, + "step": 23850 + }, + { + "epoch": 0.87, + "learning_rate": 1.1283919874024974e-07, + "loss": 0.6996, + "step": 23851 + }, + { + "epoch": 0.87, + "learning_rate": 1.1283553667557768e-07, + "loss": 0.6912, + "step": 23852 + }, + { + "epoch": 0.87, + "learning_rate": 1.1283187461090562e-07, + "loss": 0.6734, + "step": 23853 + }, + { + "epoch": 0.87, + "learning_rate": 1.1282821254623356e-07, + "loss": 0.6905, + "step": 23854 + }, + { + "epoch": 0.87, + "learning_rate": 1.128245504815615e-07, + "loss": 0.7027, + "step": 23855 + }, + { + "epoch": 0.87, + "learning_rate": 1.1282088841688944e-07, + "loss": 0.7006, + "step": 23856 + }, + { + "epoch": 0.87, + "learning_rate": 1.1281722635221738e-07, + "loss": 0.7192, + "step": 23857 + }, + { + "epoch": 0.87, + "learning_rate": 1.1281356428754532e-07, + "loss": 0.7012, + "step": 23858 + }, + { + "epoch": 0.87, + "learning_rate": 1.1280990222287326e-07, + "loss": 0.7021, + "step": 23859 + }, + { + "epoch": 0.87, + "learning_rate": 1.128062401582012e-07, + "loss": 0.7122, + "step": 23860 + }, + { + "epoch": 0.87, + "learning_rate": 1.1280257809352914e-07, + "loss": 0.6951, + "step": 23861 + }, + { + "epoch": 0.87, + "learning_rate": 1.1279891602885707e-07, + "loss": 0.6902, + "step": 23862 + }, + { + "epoch": 0.87, + "learning_rate": 1.12795253964185e-07, + "loss": 0.7011, + "step": 23863 + }, + { + "epoch": 0.87, + "learning_rate": 1.1279159189951295e-07, + "loss": 0.7, + "step": 23864 + }, + { + "epoch": 0.87, + "learning_rate": 1.1278792983484089e-07, + "loss": 0.6913, + "step": 23865 + }, + { + "epoch": 0.87, + "learning_rate": 1.1278426777016881e-07, + "loss": 0.7058, + "step": 23866 + }, + { + "epoch": 0.87, + "learning_rate": 1.1278060570549675e-07, + "loss": 0.6965, + "step": 23867 + }, + { + "epoch": 0.87, + "learning_rate": 1.1277694364082468e-07, + "loss": 0.693, + "step": 23868 + }, + { + "epoch": 0.87, + "learning_rate": 1.1277328157615262e-07, + "loss": 0.6915, + "step": 23869 + }, + { + "epoch": 0.87, + "learning_rate": 1.1276961951148056e-07, + "loss": 0.6874, + "step": 23870 + }, + { + "epoch": 0.87, + "learning_rate": 1.127659574468085e-07, + "loss": 0.6888, + "step": 23871 + }, + { + "epoch": 0.87, + "learning_rate": 1.1276229538213644e-07, + "loss": 0.6908, + "step": 23872 + }, + { + "epoch": 0.87, + "learning_rate": 1.1275863331746438e-07, + "loss": 0.6914, + "step": 23873 + }, + { + "epoch": 0.87, + "learning_rate": 1.1275497125279232e-07, + "loss": 0.6904, + "step": 23874 + }, + { + "epoch": 0.87, + "learning_rate": 1.1275130918812026e-07, + "loss": 0.6872, + "step": 23875 + }, + { + "epoch": 0.87, + "learning_rate": 1.127476471234482e-07, + "loss": 0.6933, + "step": 23876 + }, + { + "epoch": 0.87, + "learning_rate": 1.1274398505877614e-07, + "loss": 0.7068, + "step": 23877 + }, + { + "epoch": 0.87, + "learning_rate": 1.1274032299410408e-07, + "loss": 0.7002, + "step": 23878 + }, + { + "epoch": 0.87, + "learning_rate": 1.1273666092943202e-07, + "loss": 0.691, + "step": 23879 + }, + { + "epoch": 0.87, + "learning_rate": 1.1273299886475996e-07, + "loss": 0.6831, + "step": 23880 + }, + { + "epoch": 0.87, + "learning_rate": 1.127293368000879e-07, + "loss": 0.6915, + "step": 23881 + }, + { + "epoch": 0.87, + "learning_rate": 1.1272567473541584e-07, + "loss": 0.6925, + "step": 23882 + }, + { + "epoch": 0.87, + "learning_rate": 1.1272201267074378e-07, + "loss": 0.6851, + "step": 23883 + }, + { + "epoch": 0.87, + "learning_rate": 1.1271835060607169e-07, + "loss": 0.6893, + "step": 23884 + }, + { + "epoch": 0.87, + "learning_rate": 1.1271468854139963e-07, + "loss": 0.6845, + "step": 23885 + }, + { + "epoch": 0.87, + "learning_rate": 1.1271102647672757e-07, + "loss": 0.6903, + "step": 23886 + }, + { + "epoch": 0.87, + "learning_rate": 1.1270736441205551e-07, + "loss": 0.6903, + "step": 23887 + }, + { + "epoch": 0.87, + "learning_rate": 1.1270370234738345e-07, + "loss": 0.6918, + "step": 23888 + }, + { + "epoch": 0.87, + "learning_rate": 1.1270004028271139e-07, + "loss": 0.7034, + "step": 23889 + }, + { + "epoch": 0.87, + "learning_rate": 1.1269637821803932e-07, + "loss": 0.686, + "step": 23890 + }, + { + "epoch": 0.87, + "learning_rate": 1.1269271615336726e-07, + "loss": 0.6874, + "step": 23891 + }, + { + "epoch": 0.87, + "learning_rate": 1.126890540886952e-07, + "loss": 0.7006, + "step": 23892 + }, + { + "epoch": 0.87, + "learning_rate": 1.1268539202402314e-07, + "loss": 0.6945, + "step": 23893 + }, + { + "epoch": 0.87, + "learning_rate": 1.1268172995935108e-07, + "loss": 0.6813, + "step": 23894 + }, + { + "epoch": 0.87, + "learning_rate": 1.1267806789467902e-07, + "loss": 0.7078, + "step": 23895 + }, + { + "epoch": 0.87, + "learning_rate": 1.1267440583000696e-07, + "loss": 0.6861, + "step": 23896 + }, + { + "epoch": 0.87, + "learning_rate": 1.126707437653349e-07, + "loss": 0.6916, + "step": 23897 + }, + { + "epoch": 0.87, + "learning_rate": 1.1266708170066284e-07, + "loss": 0.6909, + "step": 23898 + }, + { + "epoch": 0.87, + "learning_rate": 1.1266341963599078e-07, + "loss": 0.6932, + "step": 23899 + }, + { + "epoch": 0.87, + "learning_rate": 1.1265975757131872e-07, + "loss": 0.6919, + "step": 23900 + }, + { + "epoch": 0.87, + "learning_rate": 1.1265609550664663e-07, + "loss": 0.6977, + "step": 23901 + }, + { + "epoch": 0.87, + "learning_rate": 1.1265243344197457e-07, + "loss": 0.6882, + "step": 23902 + }, + { + "epoch": 0.87, + "learning_rate": 1.1264877137730251e-07, + "loss": 0.693, + "step": 23903 + }, + { + "epoch": 0.87, + "learning_rate": 1.1264510931263045e-07, + "loss": 0.6945, + "step": 23904 + }, + { + "epoch": 0.87, + "learning_rate": 1.1264144724795839e-07, + "loss": 0.7002, + "step": 23905 + }, + { + "epoch": 0.87, + "learning_rate": 1.1263778518328633e-07, + "loss": 0.6787, + "step": 23906 + }, + { + "epoch": 0.87, + "learning_rate": 1.1263412311861427e-07, + "loss": 0.6964, + "step": 23907 + }, + { + "epoch": 0.87, + "learning_rate": 1.1263046105394221e-07, + "loss": 0.6782, + "step": 23908 + }, + { + "epoch": 0.87, + "learning_rate": 1.1262679898927015e-07, + "loss": 0.687, + "step": 23909 + }, + { + "epoch": 0.87, + "learning_rate": 1.1262313692459809e-07, + "loss": 0.6804, + "step": 23910 + }, + { + "epoch": 0.87, + "learning_rate": 1.1261947485992602e-07, + "loss": 0.7061, + "step": 23911 + }, + { + "epoch": 0.87, + "learning_rate": 1.1261581279525396e-07, + "loss": 0.6971, + "step": 23912 + }, + { + "epoch": 0.87, + "learning_rate": 1.126121507305819e-07, + "loss": 0.6937, + "step": 23913 + }, + { + "epoch": 0.87, + "learning_rate": 1.1260848866590984e-07, + "loss": 0.6961, + "step": 23914 + }, + { + "epoch": 0.87, + "learning_rate": 1.1260482660123778e-07, + "loss": 0.7075, + "step": 23915 + }, + { + "epoch": 0.88, + "learning_rate": 1.1260116453656572e-07, + "loss": 0.6843, + "step": 23916 + }, + { + "epoch": 0.88, + "learning_rate": 1.1259750247189366e-07, + "loss": 0.6981, + "step": 23917 + }, + { + "epoch": 0.88, + "learning_rate": 1.125938404072216e-07, + "loss": 0.6912, + "step": 23918 + }, + { + "epoch": 0.88, + "learning_rate": 1.1259017834254951e-07, + "loss": 0.7039, + "step": 23919 + }, + { + "epoch": 0.88, + "learning_rate": 1.1258651627787745e-07, + "loss": 0.6817, + "step": 23920 + }, + { + "epoch": 0.88, + "learning_rate": 1.1258285421320539e-07, + "loss": 0.6978, + "step": 23921 + }, + { + "epoch": 0.88, + "learning_rate": 1.1257919214853333e-07, + "loss": 0.6821, + "step": 23922 + }, + { + "epoch": 0.88, + "learning_rate": 1.1257553008386127e-07, + "loss": 0.6879, + "step": 23923 + }, + { + "epoch": 0.88, + "learning_rate": 1.1257186801918921e-07, + "loss": 0.71, + "step": 23924 + }, + { + "epoch": 0.88, + "learning_rate": 1.1256820595451715e-07, + "loss": 0.6997, + "step": 23925 + }, + { + "epoch": 0.88, + "learning_rate": 1.1256454388984509e-07, + "loss": 0.6825, + "step": 23926 + }, + { + "epoch": 0.88, + "learning_rate": 1.1256088182517303e-07, + "loss": 0.6887, + "step": 23927 + }, + { + "epoch": 0.88, + "learning_rate": 1.1255721976050097e-07, + "loss": 0.6874, + "step": 23928 + }, + { + "epoch": 0.88, + "learning_rate": 1.1255355769582891e-07, + "loss": 0.6871, + "step": 23929 + }, + { + "epoch": 0.88, + "learning_rate": 1.1254989563115685e-07, + "loss": 0.677, + "step": 23930 + }, + { + "epoch": 0.88, + "learning_rate": 1.1254623356648479e-07, + "loss": 0.6996, + "step": 23931 + }, + { + "epoch": 0.88, + "learning_rate": 1.1254257150181273e-07, + "loss": 0.6888, + "step": 23932 + }, + { + "epoch": 0.88, + "learning_rate": 1.1253890943714065e-07, + "loss": 0.6941, + "step": 23933 + }, + { + "epoch": 0.88, + "learning_rate": 1.125352473724686e-07, + "loss": 0.7048, + "step": 23934 + }, + { + "epoch": 0.88, + "learning_rate": 1.1253158530779653e-07, + "loss": 0.6877, + "step": 23935 + }, + { + "epoch": 0.88, + "learning_rate": 1.1252792324312447e-07, + "loss": 0.6969, + "step": 23936 + }, + { + "epoch": 0.88, + "learning_rate": 1.125242611784524e-07, + "loss": 0.6894, + "step": 23937 + }, + { + "epoch": 0.88, + "learning_rate": 1.1252059911378034e-07, + "loss": 0.6952, + "step": 23938 + }, + { + "epoch": 0.88, + "learning_rate": 1.1251693704910827e-07, + "loss": 0.6859, + "step": 23939 + }, + { + "epoch": 0.88, + "learning_rate": 1.1251327498443621e-07, + "loss": 0.6974, + "step": 23940 + }, + { + "epoch": 0.88, + "learning_rate": 1.1250961291976415e-07, + "loss": 0.7037, + "step": 23941 + }, + { + "epoch": 0.88, + "learning_rate": 1.1250595085509209e-07, + "loss": 0.6835, + "step": 23942 + }, + { + "epoch": 0.88, + "learning_rate": 1.1250228879042003e-07, + "loss": 0.7065, + "step": 23943 + }, + { + "epoch": 0.88, + "learning_rate": 1.1249862672574797e-07, + "loss": 0.7035, + "step": 23944 + }, + { + "epoch": 0.88, + "learning_rate": 1.1249496466107591e-07, + "loss": 0.6989, + "step": 23945 + }, + { + "epoch": 0.88, + "learning_rate": 1.1249130259640385e-07, + "loss": 0.6983, + "step": 23946 + }, + { + "epoch": 0.88, + "learning_rate": 1.1248764053173179e-07, + "loss": 0.699, + "step": 23947 + }, + { + "epoch": 0.88, + "learning_rate": 1.1248397846705973e-07, + "loss": 0.6964, + "step": 23948 + }, + { + "epoch": 0.88, + "learning_rate": 1.1248031640238767e-07, + "loss": 0.7013, + "step": 23949 + }, + { + "epoch": 0.88, + "learning_rate": 1.1247665433771561e-07, + "loss": 0.6962, + "step": 23950 + }, + { + "epoch": 0.88, + "learning_rate": 1.1247299227304355e-07, + "loss": 0.6741, + "step": 23951 + }, + { + "epoch": 0.88, + "learning_rate": 1.1246933020837149e-07, + "loss": 0.6934, + "step": 23952 + }, + { + "epoch": 0.88, + "learning_rate": 1.1246566814369943e-07, + "loss": 0.6858, + "step": 23953 + }, + { + "epoch": 0.88, + "learning_rate": 1.1246200607902737e-07, + "loss": 0.683, + "step": 23954 + }, + { + "epoch": 0.88, + "learning_rate": 1.1245834401435528e-07, + "loss": 0.6775, + "step": 23955 + }, + { + "epoch": 0.88, + "learning_rate": 1.1245468194968322e-07, + "loss": 0.6924, + "step": 23956 + }, + { + "epoch": 0.88, + "learning_rate": 1.1245101988501116e-07, + "loss": 0.688, + "step": 23957 + }, + { + "epoch": 0.88, + "learning_rate": 1.124473578203391e-07, + "loss": 0.6738, + "step": 23958 + }, + { + "epoch": 0.88, + "learning_rate": 1.1244369575566704e-07, + "loss": 0.6873, + "step": 23959 + }, + { + "epoch": 0.88, + "learning_rate": 1.1244003369099497e-07, + "loss": 0.6982, + "step": 23960 + }, + { + "epoch": 0.88, + "learning_rate": 1.124363716263229e-07, + "loss": 0.6907, + "step": 23961 + }, + { + "epoch": 0.88, + "learning_rate": 1.1243270956165085e-07, + "loss": 0.6933, + "step": 23962 + }, + { + "epoch": 0.88, + "learning_rate": 1.1242904749697879e-07, + "loss": 0.6862, + "step": 23963 + }, + { + "epoch": 0.88, + "learning_rate": 1.1242538543230673e-07, + "loss": 0.6916, + "step": 23964 + }, + { + "epoch": 0.88, + "learning_rate": 1.1242172336763467e-07, + "loss": 0.6906, + "step": 23965 + }, + { + "epoch": 0.88, + "learning_rate": 1.124180613029626e-07, + "loss": 0.6968, + "step": 23966 + }, + { + "epoch": 0.88, + "learning_rate": 1.1241439923829055e-07, + "loss": 0.7074, + "step": 23967 + }, + { + "epoch": 0.88, + "learning_rate": 1.1241073717361849e-07, + "loss": 0.6875, + "step": 23968 + }, + { + "epoch": 0.88, + "learning_rate": 1.1240707510894643e-07, + "loss": 0.7021, + "step": 23969 + }, + { + "epoch": 0.88, + "learning_rate": 1.1240341304427437e-07, + "loss": 0.6865, + "step": 23970 + }, + { + "epoch": 0.88, + "learning_rate": 1.123997509796023e-07, + "loss": 0.6979, + "step": 23971 + }, + { + "epoch": 0.88, + "learning_rate": 1.1239608891493025e-07, + "loss": 0.6964, + "step": 23972 + }, + { + "epoch": 0.88, + "learning_rate": 1.1239242685025816e-07, + "loss": 0.6972, + "step": 23973 + }, + { + "epoch": 0.88, + "learning_rate": 1.123887647855861e-07, + "loss": 0.6913, + "step": 23974 + }, + { + "epoch": 0.88, + "learning_rate": 1.1238510272091404e-07, + "loss": 0.6911, + "step": 23975 + }, + { + "epoch": 0.88, + "learning_rate": 1.1238144065624198e-07, + "loss": 0.694, + "step": 23976 + }, + { + "epoch": 0.88, + "learning_rate": 1.1237777859156992e-07, + "loss": 0.6874, + "step": 23977 + }, + { + "epoch": 0.88, + "learning_rate": 1.1237411652689786e-07, + "loss": 0.6934, + "step": 23978 + }, + { + "epoch": 0.88, + "learning_rate": 1.123704544622258e-07, + "loss": 0.6927, + "step": 23979 + }, + { + "epoch": 0.88, + "learning_rate": 1.1236679239755374e-07, + "loss": 0.6994, + "step": 23980 + }, + { + "epoch": 0.88, + "learning_rate": 1.1236313033288168e-07, + "loss": 0.7029, + "step": 23981 + }, + { + "epoch": 0.88, + "learning_rate": 1.123594682682096e-07, + "loss": 0.6866, + "step": 23982 + }, + { + "epoch": 0.88, + "learning_rate": 1.1235580620353755e-07, + "loss": 0.6954, + "step": 23983 + }, + { + "epoch": 0.88, + "learning_rate": 1.1235214413886549e-07, + "loss": 0.6864, + "step": 23984 + }, + { + "epoch": 0.88, + "learning_rate": 1.1234848207419343e-07, + "loss": 0.695, + "step": 23985 + }, + { + "epoch": 0.88, + "learning_rate": 1.1234482000952137e-07, + "loss": 0.6886, + "step": 23986 + }, + { + "epoch": 0.88, + "learning_rate": 1.123411579448493e-07, + "loss": 0.6913, + "step": 23987 + }, + { + "epoch": 0.88, + "learning_rate": 1.1233749588017724e-07, + "loss": 0.6882, + "step": 23988 + }, + { + "epoch": 0.88, + "learning_rate": 1.1233383381550518e-07, + "loss": 0.6895, + "step": 23989 + }, + { + "epoch": 0.88, + "learning_rate": 1.1233017175083312e-07, + "loss": 0.6969, + "step": 23990 + }, + { + "epoch": 0.88, + "learning_rate": 1.1232650968616104e-07, + "loss": 0.7089, + "step": 23991 + }, + { + "epoch": 0.88, + "learning_rate": 1.1232284762148898e-07, + "loss": 0.6906, + "step": 23992 + }, + { + "epoch": 0.88, + "learning_rate": 1.1231918555681692e-07, + "loss": 0.6948, + "step": 23993 + }, + { + "epoch": 0.88, + "learning_rate": 1.1231552349214486e-07, + "loss": 0.6912, + "step": 23994 + }, + { + "epoch": 0.88, + "learning_rate": 1.123118614274728e-07, + "loss": 0.6888, + "step": 23995 + }, + { + "epoch": 0.88, + "learning_rate": 1.1230819936280074e-07, + "loss": 0.6994, + "step": 23996 + }, + { + "epoch": 0.88, + "learning_rate": 1.1230453729812868e-07, + "loss": 0.7008, + "step": 23997 + }, + { + "epoch": 0.88, + "learning_rate": 1.1230087523345662e-07, + "loss": 0.7164, + "step": 23998 + }, + { + "epoch": 0.88, + "learning_rate": 1.1229721316878456e-07, + "loss": 0.6977, + "step": 23999 + }, + { + "epoch": 0.88, + "learning_rate": 1.122935511041125e-07, + "loss": 0.6865, + "step": 24000 + }, + { + "epoch": 0.88, + "learning_rate": 1.1228988903944044e-07, + "loss": 0.6934, + "step": 24001 + }, + { + "epoch": 0.88, + "learning_rate": 1.1228622697476838e-07, + "loss": 0.6882, + "step": 24002 + }, + { + "epoch": 0.88, + "learning_rate": 1.1228256491009632e-07, + "loss": 0.6909, + "step": 24003 + }, + { + "epoch": 0.88, + "learning_rate": 1.1227890284542424e-07, + "loss": 0.6978, + "step": 24004 + }, + { + "epoch": 0.88, + "learning_rate": 1.1227524078075218e-07, + "loss": 0.6883, + "step": 24005 + }, + { + "epoch": 0.88, + "learning_rate": 1.1227157871608012e-07, + "loss": 0.6935, + "step": 24006 + }, + { + "epoch": 0.88, + "learning_rate": 1.1226791665140806e-07, + "loss": 0.6806, + "step": 24007 + }, + { + "epoch": 0.88, + "learning_rate": 1.12264254586736e-07, + "loss": 0.7, + "step": 24008 + }, + { + "epoch": 0.88, + "learning_rate": 1.1226059252206393e-07, + "loss": 0.6971, + "step": 24009 + }, + { + "epoch": 0.88, + "learning_rate": 1.1225693045739186e-07, + "loss": 0.7024, + "step": 24010 + }, + { + "epoch": 0.88, + "learning_rate": 1.122532683927198e-07, + "loss": 0.7008, + "step": 24011 + }, + { + "epoch": 0.88, + "learning_rate": 1.1224960632804774e-07, + "loss": 0.7038, + "step": 24012 + }, + { + "epoch": 0.88, + "learning_rate": 1.1224594426337568e-07, + "loss": 0.6847, + "step": 24013 + }, + { + "epoch": 0.88, + "learning_rate": 1.1224228219870362e-07, + "loss": 0.6802, + "step": 24014 + }, + { + "epoch": 0.88, + "learning_rate": 1.1223862013403156e-07, + "loss": 0.6829, + "step": 24015 + }, + { + "epoch": 0.88, + "learning_rate": 1.122349580693595e-07, + "loss": 0.6951, + "step": 24016 + }, + { + "epoch": 0.88, + "learning_rate": 1.1223129600468744e-07, + "loss": 0.6888, + "step": 24017 + }, + { + "epoch": 0.88, + "learning_rate": 1.1222763394001538e-07, + "loss": 0.6817, + "step": 24018 + }, + { + "epoch": 0.88, + "learning_rate": 1.1222397187534332e-07, + "loss": 0.7023, + "step": 24019 + }, + { + "epoch": 0.88, + "learning_rate": 1.1222030981067126e-07, + "loss": 0.7096, + "step": 24020 + }, + { + "epoch": 0.88, + "learning_rate": 1.122166477459992e-07, + "loss": 0.7033, + "step": 24021 + }, + { + "epoch": 0.88, + "learning_rate": 1.1221298568132714e-07, + "loss": 0.6766, + "step": 24022 + }, + { + "epoch": 0.88, + "learning_rate": 1.1220932361665508e-07, + "loss": 0.7067, + "step": 24023 + }, + { + "epoch": 0.88, + "learning_rate": 1.1220566155198302e-07, + "loss": 0.687, + "step": 24024 + }, + { + "epoch": 0.88, + "learning_rate": 1.1220199948731096e-07, + "loss": 0.6862, + "step": 24025 + }, + { + "epoch": 0.88, + "learning_rate": 1.1219833742263888e-07, + "loss": 0.7043, + "step": 24026 + }, + { + "epoch": 0.88, + "learning_rate": 1.1219467535796681e-07, + "loss": 0.698, + "step": 24027 + }, + { + "epoch": 0.88, + "learning_rate": 1.1219101329329475e-07, + "loss": 0.6852, + "step": 24028 + }, + { + "epoch": 0.88, + "learning_rate": 1.1218735122862269e-07, + "loss": 0.7035, + "step": 24029 + }, + { + "epoch": 0.88, + "learning_rate": 1.1218368916395063e-07, + "loss": 0.6898, + "step": 24030 + }, + { + "epoch": 0.88, + "learning_rate": 1.1218002709927856e-07, + "loss": 0.6977, + "step": 24031 + }, + { + "epoch": 0.88, + "learning_rate": 1.121763650346065e-07, + "loss": 0.7074, + "step": 24032 + }, + { + "epoch": 0.88, + "learning_rate": 1.1217270296993444e-07, + "loss": 0.6888, + "step": 24033 + }, + { + "epoch": 0.88, + "learning_rate": 1.1216904090526238e-07, + "loss": 0.6963, + "step": 24034 + }, + { + "epoch": 0.88, + "learning_rate": 1.1216537884059032e-07, + "loss": 0.679, + "step": 24035 + }, + { + "epoch": 0.88, + "learning_rate": 1.1216171677591826e-07, + "loss": 0.679, + "step": 24036 + }, + { + "epoch": 0.88, + "learning_rate": 1.121580547112462e-07, + "loss": 0.6815, + "step": 24037 + }, + { + "epoch": 0.88, + "learning_rate": 1.1215439264657414e-07, + "loss": 0.6977, + "step": 24038 + }, + { + "epoch": 0.88, + "learning_rate": 1.1215073058190208e-07, + "loss": 0.6862, + "step": 24039 + }, + { + "epoch": 0.88, + "learning_rate": 1.1214706851723002e-07, + "loss": 0.6872, + "step": 24040 + }, + { + "epoch": 0.88, + "learning_rate": 1.1214340645255796e-07, + "loss": 0.6933, + "step": 24041 + }, + { + "epoch": 0.88, + "learning_rate": 1.121397443878859e-07, + "loss": 0.6903, + "step": 24042 + }, + { + "epoch": 0.88, + "learning_rate": 1.1213608232321384e-07, + "loss": 0.6933, + "step": 24043 + }, + { + "epoch": 0.88, + "learning_rate": 1.1213242025854177e-07, + "loss": 0.7007, + "step": 24044 + }, + { + "epoch": 0.88, + "learning_rate": 1.1212875819386969e-07, + "loss": 0.694, + "step": 24045 + }, + { + "epoch": 0.88, + "learning_rate": 1.1212509612919763e-07, + "loss": 0.6988, + "step": 24046 + }, + { + "epoch": 0.88, + "learning_rate": 1.1212143406452557e-07, + "loss": 0.6908, + "step": 24047 + }, + { + "epoch": 0.88, + "learning_rate": 1.1211777199985351e-07, + "loss": 0.6964, + "step": 24048 + }, + { + "epoch": 0.88, + "learning_rate": 1.1211410993518145e-07, + "loss": 0.6895, + "step": 24049 + }, + { + "epoch": 0.88, + "learning_rate": 1.1211044787050939e-07, + "loss": 0.7033, + "step": 24050 + }, + { + "epoch": 0.88, + "learning_rate": 1.1210678580583733e-07, + "loss": 0.6965, + "step": 24051 + }, + { + "epoch": 0.88, + "learning_rate": 1.1210312374116527e-07, + "loss": 0.6927, + "step": 24052 + }, + { + "epoch": 0.88, + "learning_rate": 1.120994616764932e-07, + "loss": 0.711, + "step": 24053 + }, + { + "epoch": 0.88, + "learning_rate": 1.1209579961182113e-07, + "loss": 0.6939, + "step": 24054 + }, + { + "epoch": 0.88, + "learning_rate": 1.1209213754714907e-07, + "loss": 0.6866, + "step": 24055 + }, + { + "epoch": 0.88, + "learning_rate": 1.1208847548247701e-07, + "loss": 0.6937, + "step": 24056 + }, + { + "epoch": 0.88, + "learning_rate": 1.1208481341780495e-07, + "loss": 0.6958, + "step": 24057 + }, + { + "epoch": 0.88, + "learning_rate": 1.120811513531329e-07, + "loss": 0.6907, + "step": 24058 + }, + { + "epoch": 0.88, + "learning_rate": 1.1207748928846083e-07, + "loss": 0.7, + "step": 24059 + }, + { + "epoch": 0.88, + "learning_rate": 1.1207382722378877e-07, + "loss": 0.705, + "step": 24060 + }, + { + "epoch": 0.88, + "learning_rate": 1.1207016515911671e-07, + "loss": 0.6937, + "step": 24061 + }, + { + "epoch": 0.88, + "learning_rate": 1.1206650309444465e-07, + "loss": 0.6986, + "step": 24062 + }, + { + "epoch": 0.88, + "learning_rate": 1.1206284102977257e-07, + "loss": 0.7057, + "step": 24063 + }, + { + "epoch": 0.88, + "learning_rate": 1.1205917896510051e-07, + "loss": 0.6942, + "step": 24064 + }, + { + "epoch": 0.88, + "learning_rate": 1.1205551690042845e-07, + "loss": 0.6946, + "step": 24065 + }, + { + "epoch": 0.88, + "learning_rate": 1.1205185483575639e-07, + "loss": 0.6994, + "step": 24066 + }, + { + "epoch": 0.88, + "learning_rate": 1.1204819277108433e-07, + "loss": 0.6993, + "step": 24067 + }, + { + "epoch": 0.88, + "learning_rate": 1.1204453070641227e-07, + "loss": 0.7157, + "step": 24068 + }, + { + "epoch": 0.88, + "learning_rate": 1.1204086864174021e-07, + "loss": 0.7048, + "step": 24069 + }, + { + "epoch": 0.88, + "learning_rate": 1.1203720657706815e-07, + "loss": 0.7035, + "step": 24070 + }, + { + "epoch": 0.88, + "learning_rate": 1.1203354451239609e-07, + "loss": 0.6889, + "step": 24071 + }, + { + "epoch": 0.88, + "learning_rate": 1.1202988244772403e-07, + "loss": 0.689, + "step": 24072 + }, + { + "epoch": 0.88, + "learning_rate": 1.1202622038305197e-07, + "loss": 0.6929, + "step": 24073 + }, + { + "epoch": 0.88, + "learning_rate": 1.1202255831837991e-07, + "loss": 0.6839, + "step": 24074 + }, + { + "epoch": 0.88, + "learning_rate": 1.1201889625370783e-07, + "loss": 0.6863, + "step": 24075 + }, + { + "epoch": 0.88, + "learning_rate": 1.1201523418903577e-07, + "loss": 0.6863, + "step": 24076 + }, + { + "epoch": 0.88, + "learning_rate": 1.1201157212436371e-07, + "loss": 0.6952, + "step": 24077 + }, + { + "epoch": 0.88, + "learning_rate": 1.1200791005969165e-07, + "loss": 0.7006, + "step": 24078 + }, + { + "epoch": 0.88, + "learning_rate": 1.1200424799501959e-07, + "loss": 0.7102, + "step": 24079 + }, + { + "epoch": 0.88, + "learning_rate": 1.1200058593034753e-07, + "loss": 0.6943, + "step": 24080 + }, + { + "epoch": 0.88, + "learning_rate": 1.1199692386567545e-07, + "loss": 0.6971, + "step": 24081 + }, + { + "epoch": 0.88, + "learning_rate": 1.1199326180100339e-07, + "loss": 0.69, + "step": 24082 + }, + { + "epoch": 0.88, + "learning_rate": 1.1198959973633133e-07, + "loss": 0.7076, + "step": 24083 + }, + { + "epoch": 0.88, + "learning_rate": 1.1198593767165927e-07, + "loss": 0.6919, + "step": 24084 + }, + { + "epoch": 0.88, + "learning_rate": 1.119822756069872e-07, + "loss": 0.6975, + "step": 24085 + }, + { + "epoch": 0.88, + "learning_rate": 1.1197861354231515e-07, + "loss": 0.6901, + "step": 24086 + }, + { + "epoch": 0.88, + "learning_rate": 1.1197495147764309e-07, + "loss": 0.7028, + "step": 24087 + }, + { + "epoch": 0.88, + "learning_rate": 1.1197128941297103e-07, + "loss": 0.6948, + "step": 24088 + }, + { + "epoch": 0.88, + "learning_rate": 1.1196762734829897e-07, + "loss": 0.6923, + "step": 24089 + }, + { + "epoch": 0.88, + "learning_rate": 1.119639652836269e-07, + "loss": 0.6935, + "step": 24090 + }, + { + "epoch": 0.88, + "learning_rate": 1.1196030321895485e-07, + "loss": 0.6789, + "step": 24091 + }, + { + "epoch": 0.88, + "learning_rate": 1.1195664115428279e-07, + "loss": 0.7139, + "step": 24092 + }, + { + "epoch": 0.88, + "learning_rate": 1.1195297908961073e-07, + "loss": 0.6806, + "step": 24093 + }, + { + "epoch": 0.88, + "learning_rate": 1.1194931702493867e-07, + "loss": 0.693, + "step": 24094 + }, + { + "epoch": 0.88, + "learning_rate": 1.119456549602666e-07, + "loss": 0.6976, + "step": 24095 + }, + { + "epoch": 0.88, + "learning_rate": 1.1194199289559455e-07, + "loss": 0.6767, + "step": 24096 + }, + { + "epoch": 0.88, + "learning_rate": 1.1193833083092247e-07, + "loss": 0.7018, + "step": 24097 + }, + { + "epoch": 0.88, + "learning_rate": 1.1193466876625041e-07, + "loss": 0.677, + "step": 24098 + }, + { + "epoch": 0.88, + "learning_rate": 1.1193100670157834e-07, + "loss": 0.7011, + "step": 24099 + }, + { + "epoch": 0.88, + "learning_rate": 1.1192734463690628e-07, + "loss": 0.6942, + "step": 24100 + }, + { + "epoch": 0.88, + "learning_rate": 1.1192368257223422e-07, + "loss": 0.6953, + "step": 24101 + }, + { + "epoch": 0.88, + "learning_rate": 1.1192002050756215e-07, + "loss": 0.6739, + "step": 24102 + }, + { + "epoch": 0.88, + "learning_rate": 1.1191635844289009e-07, + "loss": 0.711, + "step": 24103 + }, + { + "epoch": 0.88, + "learning_rate": 1.1191269637821803e-07, + "loss": 0.6904, + "step": 24104 + }, + { + "epoch": 0.88, + "learning_rate": 1.1190903431354597e-07, + "loss": 0.6759, + "step": 24105 + }, + { + "epoch": 0.88, + "learning_rate": 1.119053722488739e-07, + "loss": 0.6992, + "step": 24106 + }, + { + "epoch": 0.88, + "learning_rate": 1.1190171018420184e-07, + "loss": 0.6867, + "step": 24107 + }, + { + "epoch": 0.88, + "learning_rate": 1.1189804811952978e-07, + "loss": 0.693, + "step": 24108 + }, + { + "epoch": 0.88, + "learning_rate": 1.1189438605485772e-07, + "loss": 0.6906, + "step": 24109 + }, + { + "epoch": 0.88, + "learning_rate": 1.1189072399018566e-07, + "loss": 0.7113, + "step": 24110 + }, + { + "epoch": 0.88, + "learning_rate": 1.118870619255136e-07, + "loss": 0.6992, + "step": 24111 + }, + { + "epoch": 0.88, + "learning_rate": 1.1188339986084154e-07, + "loss": 0.6906, + "step": 24112 + }, + { + "epoch": 0.88, + "learning_rate": 1.1187973779616948e-07, + "loss": 0.6961, + "step": 24113 + }, + { + "epoch": 0.88, + "learning_rate": 1.1187607573149742e-07, + "loss": 0.6906, + "step": 24114 + }, + { + "epoch": 0.88, + "learning_rate": 1.1187241366682536e-07, + "loss": 0.6944, + "step": 24115 + }, + { + "epoch": 0.88, + "learning_rate": 1.118687516021533e-07, + "loss": 0.7024, + "step": 24116 + }, + { + "epoch": 0.88, + "learning_rate": 1.1186508953748122e-07, + "loss": 0.6893, + "step": 24117 + }, + { + "epoch": 0.88, + "learning_rate": 1.1186142747280916e-07, + "loss": 0.6841, + "step": 24118 + }, + { + "epoch": 0.88, + "learning_rate": 1.118577654081371e-07, + "loss": 0.696, + "step": 24119 + }, + { + "epoch": 0.88, + "learning_rate": 1.1185410334346504e-07, + "loss": 0.6889, + "step": 24120 + }, + { + "epoch": 0.88, + "learning_rate": 1.1185044127879298e-07, + "loss": 0.6869, + "step": 24121 + }, + { + "epoch": 0.88, + "learning_rate": 1.1184677921412092e-07, + "loss": 0.6764, + "step": 24122 + }, + { + "epoch": 0.88, + "learning_rate": 1.1184311714944886e-07, + "loss": 0.6875, + "step": 24123 + }, + { + "epoch": 0.88, + "learning_rate": 1.1183945508477678e-07, + "loss": 0.7023, + "step": 24124 + }, + { + "epoch": 0.88, + "learning_rate": 1.1183579302010472e-07, + "loss": 0.6948, + "step": 24125 + }, + { + "epoch": 0.88, + "learning_rate": 1.1183213095543266e-07, + "loss": 0.6925, + "step": 24126 + }, + { + "epoch": 0.88, + "learning_rate": 1.118284688907606e-07, + "loss": 0.6933, + "step": 24127 + }, + { + "epoch": 0.88, + "learning_rate": 1.1182480682608854e-07, + "loss": 0.6988, + "step": 24128 + }, + { + "epoch": 0.88, + "learning_rate": 1.1182114476141648e-07, + "loss": 0.6994, + "step": 24129 + }, + { + "epoch": 0.88, + "learning_rate": 1.1181748269674442e-07, + "loss": 0.6845, + "step": 24130 + }, + { + "epoch": 0.88, + "learning_rate": 1.1181382063207236e-07, + "loss": 0.7038, + "step": 24131 + }, + { + "epoch": 0.88, + "learning_rate": 1.118101585674003e-07, + "loss": 0.6925, + "step": 24132 + }, + { + "epoch": 0.88, + "learning_rate": 1.1180649650272824e-07, + "loss": 0.7003, + "step": 24133 + }, + { + "epoch": 0.88, + "learning_rate": 1.1180283443805618e-07, + "loss": 0.7006, + "step": 24134 + }, + { + "epoch": 0.88, + "learning_rate": 1.117991723733841e-07, + "loss": 0.7097, + "step": 24135 + }, + { + "epoch": 0.88, + "learning_rate": 1.1179551030871204e-07, + "loss": 0.7001, + "step": 24136 + }, + { + "epoch": 0.88, + "learning_rate": 1.1179184824403998e-07, + "loss": 0.701, + "step": 24137 + }, + { + "epoch": 0.88, + "learning_rate": 1.1178818617936792e-07, + "loss": 0.688, + "step": 24138 + }, + { + "epoch": 0.88, + "learning_rate": 1.1178452411469586e-07, + "loss": 0.6923, + "step": 24139 + }, + { + "epoch": 0.88, + "learning_rate": 1.117808620500238e-07, + "loss": 0.6863, + "step": 24140 + }, + { + "epoch": 0.88, + "learning_rate": 1.1177719998535174e-07, + "loss": 0.6885, + "step": 24141 + }, + { + "epoch": 0.88, + "learning_rate": 1.1177353792067968e-07, + "loss": 0.6847, + "step": 24142 + }, + { + "epoch": 0.88, + "learning_rate": 1.1176987585600762e-07, + "loss": 0.686, + "step": 24143 + }, + { + "epoch": 0.88, + "learning_rate": 1.1176621379133556e-07, + "loss": 0.702, + "step": 24144 + }, + { + "epoch": 0.88, + "learning_rate": 1.117625517266635e-07, + "loss": 0.683, + "step": 24145 + }, + { + "epoch": 0.88, + "learning_rate": 1.1175888966199142e-07, + "loss": 0.7157, + "step": 24146 + }, + { + "epoch": 0.88, + "learning_rate": 1.1175522759731936e-07, + "loss": 0.6896, + "step": 24147 + }, + { + "epoch": 0.88, + "learning_rate": 1.117515655326473e-07, + "loss": 0.6848, + "step": 24148 + }, + { + "epoch": 0.88, + "learning_rate": 1.1174790346797524e-07, + "loss": 0.699, + "step": 24149 + }, + { + "epoch": 0.88, + "learning_rate": 1.1174424140330318e-07, + "loss": 0.6936, + "step": 24150 + }, + { + "epoch": 0.88, + "learning_rate": 1.1174057933863112e-07, + "loss": 0.6991, + "step": 24151 + }, + { + "epoch": 0.88, + "learning_rate": 1.1173691727395906e-07, + "loss": 0.6905, + "step": 24152 + }, + { + "epoch": 0.88, + "learning_rate": 1.1173325520928698e-07, + "loss": 0.7042, + "step": 24153 + }, + { + "epoch": 0.88, + "learning_rate": 1.1172959314461492e-07, + "loss": 0.6828, + "step": 24154 + }, + { + "epoch": 0.88, + "learning_rate": 1.1172593107994286e-07, + "loss": 0.6871, + "step": 24155 + }, + { + "epoch": 0.88, + "learning_rate": 1.117222690152708e-07, + "loss": 0.7035, + "step": 24156 + }, + { + "epoch": 0.88, + "learning_rate": 1.1171860695059874e-07, + "loss": 0.686, + "step": 24157 + }, + { + "epoch": 0.88, + "learning_rate": 1.1171494488592668e-07, + "loss": 0.6968, + "step": 24158 + }, + { + "epoch": 0.88, + "learning_rate": 1.1171128282125462e-07, + "loss": 0.6873, + "step": 24159 + }, + { + "epoch": 0.88, + "learning_rate": 1.1170762075658256e-07, + "loss": 0.6998, + "step": 24160 + }, + { + "epoch": 0.88, + "learning_rate": 1.117039586919105e-07, + "loss": 0.6863, + "step": 24161 + }, + { + "epoch": 0.88, + "learning_rate": 1.1170029662723844e-07, + "loss": 0.6946, + "step": 24162 + }, + { + "epoch": 0.88, + "learning_rate": 1.1169663456256638e-07, + "loss": 0.6801, + "step": 24163 + }, + { + "epoch": 0.88, + "learning_rate": 1.1169297249789431e-07, + "loss": 0.6959, + "step": 24164 + }, + { + "epoch": 0.88, + "learning_rate": 1.1168931043322225e-07, + "loss": 0.7016, + "step": 24165 + }, + { + "epoch": 0.88, + "learning_rate": 1.116856483685502e-07, + "loss": 0.6939, + "step": 24166 + }, + { + "epoch": 0.88, + "learning_rate": 1.1168198630387813e-07, + "loss": 0.685, + "step": 24167 + }, + { + "epoch": 0.88, + "learning_rate": 1.1167832423920606e-07, + "loss": 0.6856, + "step": 24168 + }, + { + "epoch": 0.88, + "learning_rate": 1.11674662174534e-07, + "loss": 0.6984, + "step": 24169 + }, + { + "epoch": 0.88, + "learning_rate": 1.1167100010986194e-07, + "loss": 0.693, + "step": 24170 + }, + { + "epoch": 0.88, + "learning_rate": 1.1166733804518987e-07, + "loss": 0.6958, + "step": 24171 + }, + { + "epoch": 0.88, + "learning_rate": 1.1166367598051781e-07, + "loss": 0.6884, + "step": 24172 + }, + { + "epoch": 0.88, + "learning_rate": 1.1166001391584573e-07, + "loss": 0.6878, + "step": 24173 + }, + { + "epoch": 0.88, + "learning_rate": 1.1165635185117367e-07, + "loss": 0.6971, + "step": 24174 + }, + { + "epoch": 0.88, + "learning_rate": 1.1165268978650161e-07, + "loss": 0.6825, + "step": 24175 + }, + { + "epoch": 0.88, + "learning_rate": 1.1164902772182955e-07, + "loss": 0.677, + "step": 24176 + }, + { + "epoch": 0.88, + "learning_rate": 1.116453656571575e-07, + "loss": 0.6994, + "step": 24177 + }, + { + "epoch": 0.88, + "learning_rate": 1.1164170359248543e-07, + "loss": 0.7065, + "step": 24178 + }, + { + "epoch": 0.88, + "learning_rate": 1.1163804152781337e-07, + "loss": 0.6992, + "step": 24179 + }, + { + "epoch": 0.88, + "learning_rate": 1.1163437946314131e-07, + "loss": 0.6996, + "step": 24180 + }, + { + "epoch": 0.88, + "learning_rate": 1.1163071739846925e-07, + "loss": 0.6956, + "step": 24181 + }, + { + "epoch": 0.88, + "learning_rate": 1.116270553337972e-07, + "loss": 0.7028, + "step": 24182 + }, + { + "epoch": 0.88, + "learning_rate": 1.1162339326912513e-07, + "loss": 0.6984, + "step": 24183 + }, + { + "epoch": 0.88, + "learning_rate": 1.1161973120445307e-07, + "loss": 0.6942, + "step": 24184 + }, + { + "epoch": 0.88, + "learning_rate": 1.1161606913978101e-07, + "loss": 0.6971, + "step": 24185 + }, + { + "epoch": 0.88, + "learning_rate": 1.1161240707510895e-07, + "loss": 0.6869, + "step": 24186 + }, + { + "epoch": 0.88, + "learning_rate": 1.116087450104369e-07, + "loss": 0.6944, + "step": 24187 + }, + { + "epoch": 0.88, + "learning_rate": 1.1160508294576483e-07, + "loss": 0.6926, + "step": 24188 + }, + { + "epoch": 0.89, + "learning_rate": 1.1160142088109275e-07, + "loss": 0.6908, + "step": 24189 + }, + { + "epoch": 0.89, + "learning_rate": 1.1159775881642069e-07, + "loss": 0.6945, + "step": 24190 + }, + { + "epoch": 0.89, + "learning_rate": 1.1159409675174863e-07, + "loss": 0.6842, + "step": 24191 + }, + { + "epoch": 0.89, + "learning_rate": 1.1159043468707657e-07, + "loss": 0.6967, + "step": 24192 + }, + { + "epoch": 0.89, + "learning_rate": 1.1158677262240451e-07, + "loss": 0.6947, + "step": 24193 + }, + { + "epoch": 0.89, + "learning_rate": 1.1158311055773245e-07, + "loss": 0.6909, + "step": 24194 + }, + { + "epoch": 0.89, + "learning_rate": 1.1157944849306037e-07, + "loss": 0.6929, + "step": 24195 + }, + { + "epoch": 0.89, + "learning_rate": 1.1157578642838831e-07, + "loss": 0.7019, + "step": 24196 + }, + { + "epoch": 0.89, + "learning_rate": 1.1157212436371625e-07, + "loss": 0.6946, + "step": 24197 + }, + { + "epoch": 0.89, + "learning_rate": 1.1156846229904419e-07, + "loss": 0.6821, + "step": 24198 + }, + { + "epoch": 0.89, + "learning_rate": 1.1156480023437213e-07, + "loss": 0.6866, + "step": 24199 + }, + { + "epoch": 0.89, + "learning_rate": 1.1156113816970007e-07, + "loss": 0.6956, + "step": 24200 + }, + { + "epoch": 0.89, + "learning_rate": 1.1155747610502801e-07, + "loss": 0.6779, + "step": 24201 + }, + { + "epoch": 0.89, + "learning_rate": 1.1155381404035595e-07, + "loss": 0.6825, + "step": 24202 + }, + { + "epoch": 0.89, + "learning_rate": 1.1155015197568389e-07, + "loss": 0.6937, + "step": 24203 + }, + { + "epoch": 0.89, + "learning_rate": 1.1154648991101183e-07, + "loss": 0.6868, + "step": 24204 + }, + { + "epoch": 0.89, + "learning_rate": 1.1154282784633977e-07, + "loss": 0.6949, + "step": 24205 + }, + { + "epoch": 0.89, + "learning_rate": 1.1153916578166771e-07, + "loss": 0.6885, + "step": 24206 + }, + { + "epoch": 0.89, + "learning_rate": 1.1153550371699563e-07, + "loss": 0.6983, + "step": 24207 + }, + { + "epoch": 0.89, + "learning_rate": 1.1153184165232357e-07, + "loss": 0.6984, + "step": 24208 + }, + { + "epoch": 0.89, + "learning_rate": 1.115281795876515e-07, + "loss": 0.6845, + "step": 24209 + }, + { + "epoch": 0.89, + "learning_rate": 1.1152451752297945e-07, + "loss": 0.6932, + "step": 24210 + }, + { + "epoch": 0.89, + "learning_rate": 1.1152085545830739e-07, + "loss": 0.7011, + "step": 24211 + }, + { + "epoch": 0.89, + "learning_rate": 1.1151719339363533e-07, + "loss": 0.6931, + "step": 24212 + }, + { + "epoch": 0.89, + "learning_rate": 1.1151353132896327e-07, + "loss": 0.6852, + "step": 24213 + }, + { + "epoch": 0.89, + "learning_rate": 1.115098692642912e-07, + "loss": 0.6863, + "step": 24214 + }, + { + "epoch": 0.89, + "learning_rate": 1.1150620719961915e-07, + "loss": 0.7087, + "step": 24215 + }, + { + "epoch": 0.89, + "learning_rate": 1.1150254513494709e-07, + "loss": 0.6948, + "step": 24216 + }, + { + "epoch": 0.89, + "learning_rate": 1.1149888307027501e-07, + "loss": 0.6868, + "step": 24217 + }, + { + "epoch": 0.89, + "learning_rate": 1.1149522100560295e-07, + "loss": 0.693, + "step": 24218 + }, + { + "epoch": 0.89, + "learning_rate": 1.1149155894093089e-07, + "loss": 0.6939, + "step": 24219 + }, + { + "epoch": 0.89, + "learning_rate": 1.1148789687625883e-07, + "loss": 0.6872, + "step": 24220 + }, + { + "epoch": 0.89, + "learning_rate": 1.1148423481158677e-07, + "loss": 0.703, + "step": 24221 + }, + { + "epoch": 0.89, + "learning_rate": 1.1148057274691471e-07, + "loss": 0.692, + "step": 24222 + }, + { + "epoch": 0.89, + "learning_rate": 1.1147691068224265e-07, + "loss": 0.6885, + "step": 24223 + }, + { + "epoch": 0.89, + "learning_rate": 1.1147324861757059e-07, + "loss": 0.705, + "step": 24224 + }, + { + "epoch": 0.89, + "learning_rate": 1.114695865528985e-07, + "loss": 0.6923, + "step": 24225 + }, + { + "epoch": 0.89, + "learning_rate": 1.1146592448822644e-07, + "loss": 0.691, + "step": 24226 + }, + { + "epoch": 0.89, + "learning_rate": 1.1146226242355438e-07, + "loss": 0.6907, + "step": 24227 + }, + { + "epoch": 0.89, + "learning_rate": 1.1145860035888232e-07, + "loss": 0.6856, + "step": 24228 + }, + { + "epoch": 0.89, + "learning_rate": 1.1145493829421026e-07, + "loss": 0.692, + "step": 24229 + }, + { + "epoch": 0.89, + "learning_rate": 1.114512762295382e-07, + "loss": 0.6923, + "step": 24230 + }, + { + "epoch": 0.89, + "learning_rate": 1.1144761416486614e-07, + "loss": 0.6818, + "step": 24231 + }, + { + "epoch": 0.89, + "learning_rate": 1.1144395210019408e-07, + "loss": 0.6953, + "step": 24232 + }, + { + "epoch": 0.89, + "learning_rate": 1.1144029003552202e-07, + "loss": 0.6899, + "step": 24233 + }, + { + "epoch": 0.89, + "learning_rate": 1.1143662797084996e-07, + "loss": 0.6983, + "step": 24234 + }, + { + "epoch": 0.89, + "learning_rate": 1.114329659061779e-07, + "loss": 0.6975, + "step": 24235 + }, + { + "epoch": 0.89, + "learning_rate": 1.1142930384150584e-07, + "loss": 0.6936, + "step": 24236 + }, + { + "epoch": 0.89, + "learning_rate": 1.1142564177683378e-07, + "loss": 0.71, + "step": 24237 + }, + { + "epoch": 0.89, + "learning_rate": 1.1142197971216172e-07, + "loss": 0.6759, + "step": 24238 + }, + { + "epoch": 0.89, + "learning_rate": 1.1141831764748965e-07, + "loss": 0.6909, + "step": 24239 + }, + { + "epoch": 0.89, + "learning_rate": 1.1141465558281759e-07, + "loss": 0.6778, + "step": 24240 + }, + { + "epoch": 0.89, + "learning_rate": 1.1141099351814553e-07, + "loss": 0.6905, + "step": 24241 + }, + { + "epoch": 0.89, + "learning_rate": 1.1140733145347347e-07, + "loss": 0.6885, + "step": 24242 + }, + { + "epoch": 0.89, + "learning_rate": 1.114036693888014e-07, + "loss": 0.7041, + "step": 24243 + }, + { + "epoch": 0.89, + "learning_rate": 1.1140000732412932e-07, + "loss": 0.6863, + "step": 24244 + }, + { + "epoch": 0.89, + "learning_rate": 1.1139634525945726e-07, + "loss": 0.7075, + "step": 24245 + }, + { + "epoch": 0.89, + "learning_rate": 1.113926831947852e-07, + "loss": 0.7021, + "step": 24246 + }, + { + "epoch": 0.89, + "learning_rate": 1.1138902113011314e-07, + "loss": 0.6818, + "step": 24247 + }, + { + "epoch": 0.89, + "learning_rate": 1.1138535906544108e-07, + "loss": 0.709, + "step": 24248 + }, + { + "epoch": 0.89, + "learning_rate": 1.1138169700076902e-07, + "loss": 0.6948, + "step": 24249 + }, + { + "epoch": 0.89, + "learning_rate": 1.1137803493609696e-07, + "loss": 0.6978, + "step": 24250 + }, + { + "epoch": 0.89, + "learning_rate": 1.113743728714249e-07, + "loss": 0.6955, + "step": 24251 + }, + { + "epoch": 0.89, + "learning_rate": 1.1137071080675284e-07, + "loss": 0.7016, + "step": 24252 + }, + { + "epoch": 0.89, + "learning_rate": 1.1136704874208078e-07, + "loss": 0.6865, + "step": 24253 + }, + { + "epoch": 0.89, + "learning_rate": 1.1136338667740872e-07, + "loss": 0.6993, + "step": 24254 + }, + { + "epoch": 0.89, + "learning_rate": 1.1135972461273666e-07, + "loss": 0.7103, + "step": 24255 + }, + { + "epoch": 0.89, + "learning_rate": 1.113560625480646e-07, + "loss": 0.6951, + "step": 24256 + }, + { + "epoch": 0.89, + "learning_rate": 1.1135240048339254e-07, + "loss": 0.6823, + "step": 24257 + }, + { + "epoch": 0.89, + "learning_rate": 1.1134873841872048e-07, + "loss": 0.7048, + "step": 24258 + }, + { + "epoch": 0.89, + "learning_rate": 1.1134507635404842e-07, + "loss": 0.6948, + "step": 24259 + }, + { + "epoch": 0.89, + "learning_rate": 1.1134141428937636e-07, + "loss": 0.6969, + "step": 24260 + }, + { + "epoch": 0.89, + "learning_rate": 1.1133775222470428e-07, + "loss": 0.6938, + "step": 24261 + }, + { + "epoch": 0.89, + "learning_rate": 1.1133409016003222e-07, + "loss": 0.7066, + "step": 24262 + }, + { + "epoch": 0.89, + "learning_rate": 1.1133042809536016e-07, + "loss": 0.7095, + "step": 24263 + }, + { + "epoch": 0.89, + "learning_rate": 1.113267660306881e-07, + "loss": 0.69, + "step": 24264 + }, + { + "epoch": 0.89, + "learning_rate": 1.1132310396601604e-07, + "loss": 0.6993, + "step": 24265 + }, + { + "epoch": 0.89, + "learning_rate": 1.1131944190134396e-07, + "loss": 0.6909, + "step": 24266 + }, + { + "epoch": 0.89, + "learning_rate": 1.113157798366719e-07, + "loss": 0.6907, + "step": 24267 + }, + { + "epoch": 0.89, + "learning_rate": 1.1131211777199984e-07, + "loss": 0.6913, + "step": 24268 + }, + { + "epoch": 0.89, + "learning_rate": 1.1130845570732778e-07, + "loss": 0.6899, + "step": 24269 + }, + { + "epoch": 0.89, + "learning_rate": 1.1130479364265572e-07, + "loss": 0.6964, + "step": 24270 + }, + { + "epoch": 0.89, + "learning_rate": 1.1130113157798366e-07, + "loss": 0.6962, + "step": 24271 + }, + { + "epoch": 0.89, + "learning_rate": 1.112974695133116e-07, + "loss": 0.692, + "step": 24272 + }, + { + "epoch": 0.89, + "learning_rate": 1.1129380744863954e-07, + "loss": 0.6997, + "step": 24273 + }, + { + "epoch": 0.89, + "learning_rate": 1.1129014538396748e-07, + "loss": 0.6986, + "step": 24274 + }, + { + "epoch": 0.89, + "learning_rate": 1.1128648331929542e-07, + "loss": 0.6918, + "step": 24275 + }, + { + "epoch": 0.89, + "learning_rate": 1.1128282125462336e-07, + "loss": 0.6948, + "step": 24276 + }, + { + "epoch": 0.89, + "learning_rate": 1.112791591899513e-07, + "loss": 0.6983, + "step": 24277 + }, + { + "epoch": 0.89, + "learning_rate": 1.1127549712527924e-07, + "loss": 0.6886, + "step": 24278 + }, + { + "epoch": 0.89, + "learning_rate": 1.1127183506060716e-07, + "loss": 0.6872, + "step": 24279 + }, + { + "epoch": 0.89, + "learning_rate": 1.112681729959351e-07, + "loss": 0.6928, + "step": 24280 + }, + { + "epoch": 0.89, + "learning_rate": 1.1126451093126304e-07, + "loss": 0.6948, + "step": 24281 + }, + { + "epoch": 0.89, + "learning_rate": 1.1126084886659098e-07, + "loss": 0.6744, + "step": 24282 + }, + { + "epoch": 0.89, + "learning_rate": 1.1125718680191892e-07, + "loss": 0.6772, + "step": 24283 + }, + { + "epoch": 0.89, + "learning_rate": 1.1125352473724685e-07, + "loss": 0.6993, + "step": 24284 + }, + { + "epoch": 0.89, + "learning_rate": 1.112498626725748e-07, + "loss": 0.6921, + "step": 24285 + }, + { + "epoch": 0.89, + "learning_rate": 1.1124620060790273e-07, + "loss": 0.6931, + "step": 24286 + }, + { + "epoch": 0.89, + "learning_rate": 1.1124253854323067e-07, + "loss": 0.706, + "step": 24287 + }, + { + "epoch": 0.89, + "learning_rate": 1.112388764785586e-07, + "loss": 0.7031, + "step": 24288 + }, + { + "epoch": 0.89, + "learning_rate": 1.1123521441388654e-07, + "loss": 0.7022, + "step": 24289 + }, + { + "epoch": 0.89, + "learning_rate": 1.1123155234921448e-07, + "loss": 0.7118, + "step": 24290 + }, + { + "epoch": 0.89, + "learning_rate": 1.1122789028454242e-07, + "loss": 0.6973, + "step": 24291 + }, + { + "epoch": 0.89, + "learning_rate": 1.1122422821987036e-07, + "loss": 0.6874, + "step": 24292 + }, + { + "epoch": 0.89, + "learning_rate": 1.112205661551983e-07, + "loss": 0.6994, + "step": 24293 + }, + { + "epoch": 0.89, + "learning_rate": 1.1121690409052624e-07, + "loss": 0.7046, + "step": 24294 + }, + { + "epoch": 0.89, + "learning_rate": 1.1121324202585418e-07, + "loss": 0.7054, + "step": 24295 + }, + { + "epoch": 0.89, + "learning_rate": 1.112095799611821e-07, + "loss": 0.6981, + "step": 24296 + }, + { + "epoch": 0.89, + "learning_rate": 1.1120591789651003e-07, + "loss": 0.6788, + "step": 24297 + }, + { + "epoch": 0.89, + "learning_rate": 1.1120225583183797e-07, + "loss": 0.6965, + "step": 24298 + }, + { + "epoch": 0.89, + "learning_rate": 1.1119859376716591e-07, + "loss": 0.6912, + "step": 24299 + }, + { + "epoch": 0.89, + "learning_rate": 1.1119493170249385e-07, + "loss": 0.7062, + "step": 24300 + }, + { + "epoch": 0.89, + "learning_rate": 1.111912696378218e-07, + "loss": 0.6861, + "step": 24301 + }, + { + "epoch": 0.89, + "learning_rate": 1.1118760757314973e-07, + "loss": 0.6892, + "step": 24302 + }, + { + "epoch": 0.89, + "learning_rate": 1.1118394550847767e-07, + "loss": 0.6858, + "step": 24303 + }, + { + "epoch": 0.89, + "learning_rate": 1.1118028344380561e-07, + "loss": 0.6866, + "step": 24304 + }, + { + "epoch": 0.89, + "learning_rate": 1.1117662137913355e-07, + "loss": 0.6965, + "step": 24305 + }, + { + "epoch": 0.89, + "learning_rate": 1.111729593144615e-07, + "loss": 0.6943, + "step": 24306 + }, + { + "epoch": 0.89, + "learning_rate": 1.1116929724978943e-07, + "loss": 0.6997, + "step": 24307 + }, + { + "epoch": 0.89, + "learning_rate": 1.1116563518511737e-07, + "loss": 0.6971, + "step": 24308 + }, + { + "epoch": 0.89, + "learning_rate": 1.1116197312044531e-07, + "loss": 0.6949, + "step": 24309 + }, + { + "epoch": 0.89, + "learning_rate": 1.1115831105577324e-07, + "loss": 0.6838, + "step": 24310 + }, + { + "epoch": 0.89, + "learning_rate": 1.1115464899110118e-07, + "loss": 0.7116, + "step": 24311 + }, + { + "epoch": 0.89, + "learning_rate": 1.1115098692642912e-07, + "loss": 0.6881, + "step": 24312 + }, + { + "epoch": 0.89, + "learning_rate": 1.1114732486175706e-07, + "loss": 0.6905, + "step": 24313 + }, + { + "epoch": 0.89, + "learning_rate": 1.1114366279708499e-07, + "loss": 0.6877, + "step": 24314 + }, + { + "epoch": 0.89, + "learning_rate": 1.1114000073241291e-07, + "loss": 0.705, + "step": 24315 + }, + { + "epoch": 0.89, + "learning_rate": 1.1113633866774085e-07, + "loss": 0.6754, + "step": 24316 + }, + { + "epoch": 0.89, + "learning_rate": 1.1113267660306879e-07, + "loss": 0.7015, + "step": 24317 + }, + { + "epoch": 0.89, + "learning_rate": 1.1112901453839673e-07, + "loss": 0.6923, + "step": 24318 + }, + { + "epoch": 0.89, + "learning_rate": 1.1112535247372467e-07, + "loss": 0.6985, + "step": 24319 + }, + { + "epoch": 0.89, + "learning_rate": 1.1112169040905261e-07, + "loss": 0.6916, + "step": 24320 + }, + { + "epoch": 0.89, + "learning_rate": 1.1111802834438055e-07, + "loss": 0.6915, + "step": 24321 + }, + { + "epoch": 0.89, + "learning_rate": 1.1111436627970849e-07, + "loss": 0.6994, + "step": 24322 + }, + { + "epoch": 0.89, + "learning_rate": 1.1111070421503643e-07, + "loss": 0.6842, + "step": 24323 + }, + { + "epoch": 0.89, + "learning_rate": 1.1110704215036437e-07, + "loss": 0.6939, + "step": 24324 + }, + { + "epoch": 0.89, + "learning_rate": 1.1110338008569231e-07, + "loss": 0.6846, + "step": 24325 + }, + { + "epoch": 0.89, + "learning_rate": 1.1109971802102025e-07, + "loss": 0.6798, + "step": 24326 + }, + { + "epoch": 0.89, + "learning_rate": 1.1109605595634819e-07, + "loss": 0.6821, + "step": 24327 + }, + { + "epoch": 0.89, + "learning_rate": 1.1109239389167613e-07, + "loss": 0.698, + "step": 24328 + }, + { + "epoch": 0.89, + "learning_rate": 1.1108873182700407e-07, + "loss": 0.6963, + "step": 24329 + }, + { + "epoch": 0.89, + "learning_rate": 1.1108506976233201e-07, + "loss": 0.7024, + "step": 24330 + }, + { + "epoch": 0.89, + "learning_rate": 1.1108140769765995e-07, + "loss": 0.691, + "step": 24331 + }, + { + "epoch": 0.89, + "learning_rate": 1.1107774563298787e-07, + "loss": 0.689, + "step": 24332 + }, + { + "epoch": 0.89, + "learning_rate": 1.110740835683158e-07, + "loss": 0.6922, + "step": 24333 + }, + { + "epoch": 0.89, + "learning_rate": 1.1107042150364375e-07, + "loss": 0.7006, + "step": 24334 + }, + { + "epoch": 0.89, + "learning_rate": 1.1106675943897169e-07, + "loss": 0.6972, + "step": 24335 + }, + { + "epoch": 0.89, + "learning_rate": 1.1106309737429963e-07, + "loss": 0.6935, + "step": 24336 + }, + { + "epoch": 0.89, + "learning_rate": 1.1105943530962755e-07, + "loss": 0.6956, + "step": 24337 + }, + { + "epoch": 0.89, + "learning_rate": 1.1105577324495549e-07, + "loss": 0.7037, + "step": 24338 + }, + { + "epoch": 0.89, + "learning_rate": 1.1105211118028343e-07, + "loss": 0.6949, + "step": 24339 + }, + { + "epoch": 0.89, + "learning_rate": 1.1104844911561137e-07, + "loss": 0.685, + "step": 24340 + }, + { + "epoch": 0.89, + "learning_rate": 1.1104478705093931e-07, + "loss": 0.6845, + "step": 24341 + }, + { + "epoch": 0.89, + "learning_rate": 1.1104112498626725e-07, + "loss": 0.6985, + "step": 24342 + }, + { + "epoch": 0.89, + "learning_rate": 1.1103746292159519e-07, + "loss": 0.7142, + "step": 24343 + }, + { + "epoch": 0.89, + "learning_rate": 1.1103380085692313e-07, + "loss": 0.6855, + "step": 24344 + }, + { + "epoch": 0.89, + "learning_rate": 1.1103013879225107e-07, + "loss": 0.6975, + "step": 24345 + }, + { + "epoch": 0.89, + "learning_rate": 1.1102647672757901e-07, + "loss": 0.7116, + "step": 24346 + }, + { + "epoch": 0.89, + "learning_rate": 1.1102281466290695e-07, + "loss": 0.6824, + "step": 24347 + }, + { + "epoch": 0.89, + "learning_rate": 1.1101915259823489e-07, + "loss": 0.6776, + "step": 24348 + }, + { + "epoch": 0.89, + "learning_rate": 1.1101549053356283e-07, + "loss": 0.6996, + "step": 24349 + }, + { + "epoch": 0.89, + "learning_rate": 1.1101182846889074e-07, + "loss": 0.6974, + "step": 24350 + }, + { + "epoch": 0.89, + "learning_rate": 1.1100816640421868e-07, + "loss": 0.711, + "step": 24351 + }, + { + "epoch": 0.89, + "learning_rate": 1.1100450433954662e-07, + "loss": 0.6886, + "step": 24352 + }, + { + "epoch": 0.89, + "learning_rate": 1.1100084227487456e-07, + "loss": 0.6919, + "step": 24353 + }, + { + "epoch": 0.89, + "learning_rate": 1.109971802102025e-07, + "loss": 0.6926, + "step": 24354 + }, + { + "epoch": 0.89, + "learning_rate": 1.1099351814553044e-07, + "loss": 0.6923, + "step": 24355 + }, + { + "epoch": 0.89, + "learning_rate": 1.1098985608085838e-07, + "loss": 0.6973, + "step": 24356 + }, + { + "epoch": 0.89, + "learning_rate": 1.1098619401618632e-07, + "loss": 0.6811, + "step": 24357 + }, + { + "epoch": 0.89, + "learning_rate": 1.1098253195151426e-07, + "loss": 0.7001, + "step": 24358 + }, + { + "epoch": 0.89, + "learning_rate": 1.1097886988684219e-07, + "loss": 0.6993, + "step": 24359 + }, + { + "epoch": 0.89, + "learning_rate": 1.1097520782217013e-07, + "loss": 0.6883, + "step": 24360 + }, + { + "epoch": 0.89, + "learning_rate": 1.1097154575749807e-07, + "loss": 0.6823, + "step": 24361 + }, + { + "epoch": 0.89, + "learning_rate": 1.1096788369282601e-07, + "loss": 0.683, + "step": 24362 + }, + { + "epoch": 0.89, + "learning_rate": 1.1096422162815395e-07, + "loss": 0.6988, + "step": 24363 + }, + { + "epoch": 0.89, + "learning_rate": 1.1096055956348189e-07, + "loss": 0.6911, + "step": 24364 + }, + { + "epoch": 0.89, + "learning_rate": 1.1095689749880983e-07, + "loss": 0.6926, + "step": 24365 + }, + { + "epoch": 0.89, + "learning_rate": 1.1095323543413777e-07, + "loss": 0.6873, + "step": 24366 + }, + { + "epoch": 0.89, + "learning_rate": 1.1094957336946571e-07, + "loss": 0.6734, + "step": 24367 + }, + { + "epoch": 0.89, + "learning_rate": 1.1094591130479362e-07, + "loss": 0.6815, + "step": 24368 + }, + { + "epoch": 0.89, + "learning_rate": 1.1094224924012156e-07, + "loss": 0.6887, + "step": 24369 + }, + { + "epoch": 0.89, + "learning_rate": 1.109385871754495e-07, + "loss": 0.6946, + "step": 24370 + }, + { + "epoch": 0.89, + "learning_rate": 1.1093492511077744e-07, + "loss": 0.6925, + "step": 24371 + }, + { + "epoch": 0.89, + "learning_rate": 1.1093126304610538e-07, + "loss": 0.689, + "step": 24372 + }, + { + "epoch": 0.89, + "learning_rate": 1.1092760098143332e-07, + "loss": 0.6926, + "step": 24373 + }, + { + "epoch": 0.89, + "learning_rate": 1.1092393891676126e-07, + "loss": 0.6947, + "step": 24374 + }, + { + "epoch": 0.89, + "learning_rate": 1.109202768520892e-07, + "loss": 0.6908, + "step": 24375 + }, + { + "epoch": 0.89, + "learning_rate": 1.1091661478741714e-07, + "loss": 0.696, + "step": 24376 + }, + { + "epoch": 0.89, + "learning_rate": 1.1091295272274508e-07, + "loss": 0.7022, + "step": 24377 + }, + { + "epoch": 0.89, + "learning_rate": 1.1090929065807302e-07, + "loss": 0.6943, + "step": 24378 + }, + { + "epoch": 0.89, + "learning_rate": 1.1090562859340096e-07, + "loss": 0.6974, + "step": 24379 + }, + { + "epoch": 0.89, + "learning_rate": 1.109019665287289e-07, + "loss": 0.703, + "step": 24380 + }, + { + "epoch": 0.89, + "learning_rate": 1.1089830446405683e-07, + "loss": 0.6904, + "step": 24381 + }, + { + "epoch": 0.89, + "learning_rate": 1.1089464239938477e-07, + "loss": 0.7078, + "step": 24382 + }, + { + "epoch": 0.89, + "learning_rate": 1.1089098033471271e-07, + "loss": 0.6936, + "step": 24383 + }, + { + "epoch": 0.89, + "learning_rate": 1.1088731827004065e-07, + "loss": 0.6893, + "step": 24384 + }, + { + "epoch": 0.89, + "learning_rate": 1.1088365620536859e-07, + "loss": 0.6897, + "step": 24385 + }, + { + "epoch": 0.89, + "learning_rate": 1.108799941406965e-07, + "loss": 0.6847, + "step": 24386 + }, + { + "epoch": 0.89, + "learning_rate": 1.1087633207602444e-07, + "loss": 0.6855, + "step": 24387 + }, + { + "epoch": 0.89, + "learning_rate": 1.1087267001135238e-07, + "loss": 0.699, + "step": 24388 + }, + { + "epoch": 0.89, + "learning_rate": 1.1086900794668032e-07, + "loss": 0.6979, + "step": 24389 + }, + { + "epoch": 0.89, + "learning_rate": 1.1086534588200826e-07, + "loss": 0.6949, + "step": 24390 + }, + { + "epoch": 0.89, + "learning_rate": 1.108616838173362e-07, + "loss": 0.6799, + "step": 24391 + }, + { + "epoch": 0.89, + "learning_rate": 1.1085802175266414e-07, + "loss": 0.6958, + "step": 24392 + }, + { + "epoch": 0.89, + "learning_rate": 1.1085435968799208e-07, + "loss": 0.6948, + "step": 24393 + }, + { + "epoch": 0.89, + "learning_rate": 1.1085069762332002e-07, + "loss": 0.6996, + "step": 24394 + }, + { + "epoch": 0.89, + "learning_rate": 1.1084703555864796e-07, + "loss": 0.7012, + "step": 24395 + }, + { + "epoch": 0.89, + "learning_rate": 1.108433734939759e-07, + "loss": 0.7005, + "step": 24396 + }, + { + "epoch": 0.89, + "learning_rate": 1.1083971142930384e-07, + "loss": 0.69, + "step": 24397 + }, + { + "epoch": 0.89, + "learning_rate": 1.1083604936463178e-07, + "loss": 0.7063, + "step": 24398 + }, + { + "epoch": 0.89, + "learning_rate": 1.1083238729995972e-07, + "loss": 0.6842, + "step": 24399 + }, + { + "epoch": 0.89, + "learning_rate": 1.1082872523528766e-07, + "loss": 0.7007, + "step": 24400 + }, + { + "epoch": 0.89, + "learning_rate": 1.108250631706156e-07, + "loss": 0.6989, + "step": 24401 + }, + { + "epoch": 0.89, + "learning_rate": 1.1082140110594354e-07, + "loss": 0.6997, + "step": 24402 + }, + { + "epoch": 0.89, + "learning_rate": 1.1081773904127147e-07, + "loss": 0.6829, + "step": 24403 + }, + { + "epoch": 0.89, + "learning_rate": 1.108140769765994e-07, + "loss": 0.6835, + "step": 24404 + }, + { + "epoch": 0.89, + "learning_rate": 1.1081041491192733e-07, + "loss": 0.6911, + "step": 24405 + }, + { + "epoch": 0.89, + "learning_rate": 1.1080675284725527e-07, + "loss": 0.7011, + "step": 24406 + }, + { + "epoch": 0.89, + "learning_rate": 1.1080309078258321e-07, + "loss": 0.6893, + "step": 24407 + }, + { + "epoch": 0.89, + "learning_rate": 1.1079942871791114e-07, + "loss": 0.6786, + "step": 24408 + }, + { + "epoch": 0.89, + "learning_rate": 1.1079576665323908e-07, + "loss": 0.697, + "step": 24409 + }, + { + "epoch": 0.89, + "learning_rate": 1.1079210458856702e-07, + "loss": 0.6915, + "step": 24410 + }, + { + "epoch": 0.89, + "learning_rate": 1.1078844252389496e-07, + "loss": 0.6925, + "step": 24411 + }, + { + "epoch": 0.89, + "learning_rate": 1.107847804592229e-07, + "loss": 0.6908, + "step": 24412 + }, + { + "epoch": 0.89, + "learning_rate": 1.1078111839455084e-07, + "loss": 0.7138, + "step": 24413 + }, + { + "epoch": 0.89, + "learning_rate": 1.1077745632987878e-07, + "loss": 0.6951, + "step": 24414 + }, + { + "epoch": 0.89, + "learning_rate": 1.1077379426520672e-07, + "loss": 0.6855, + "step": 24415 + }, + { + "epoch": 0.89, + "learning_rate": 1.1077013220053466e-07, + "loss": 0.6956, + "step": 24416 + }, + { + "epoch": 0.89, + "learning_rate": 1.107664701358626e-07, + "loss": 0.6904, + "step": 24417 + }, + { + "epoch": 0.89, + "learning_rate": 1.1076280807119054e-07, + "loss": 0.697, + "step": 24418 + }, + { + "epoch": 0.89, + "learning_rate": 1.1075914600651848e-07, + "loss": 0.6908, + "step": 24419 + }, + { + "epoch": 0.89, + "learning_rate": 1.1075548394184642e-07, + "loss": 0.6882, + "step": 24420 + }, + { + "epoch": 0.89, + "learning_rate": 1.1075182187717436e-07, + "loss": 0.6931, + "step": 24421 + }, + { + "epoch": 0.89, + "learning_rate": 1.1074815981250227e-07, + "loss": 0.6981, + "step": 24422 + }, + { + "epoch": 0.89, + "learning_rate": 1.1074449774783021e-07, + "loss": 0.6984, + "step": 24423 + }, + { + "epoch": 0.89, + "learning_rate": 1.1074083568315815e-07, + "loss": 0.686, + "step": 24424 + }, + { + "epoch": 0.89, + "learning_rate": 1.107371736184861e-07, + "loss": 0.7062, + "step": 24425 + }, + { + "epoch": 0.89, + "learning_rate": 1.1073351155381403e-07, + "loss": 0.6864, + "step": 24426 + }, + { + "epoch": 0.89, + "learning_rate": 1.1072984948914197e-07, + "loss": 0.6929, + "step": 24427 + }, + { + "epoch": 0.89, + "learning_rate": 1.1072618742446991e-07, + "loss": 0.6976, + "step": 24428 + }, + { + "epoch": 0.89, + "learning_rate": 1.1072252535979785e-07, + "loss": 0.6855, + "step": 24429 + }, + { + "epoch": 0.89, + "learning_rate": 1.1071886329512578e-07, + "loss": 0.7023, + "step": 24430 + }, + { + "epoch": 0.89, + "learning_rate": 1.1071520123045372e-07, + "loss": 0.7034, + "step": 24431 + }, + { + "epoch": 0.89, + "learning_rate": 1.1071153916578166e-07, + "loss": 0.697, + "step": 24432 + }, + { + "epoch": 0.89, + "learning_rate": 1.107078771011096e-07, + "loss": 0.6871, + "step": 24433 + }, + { + "epoch": 0.89, + "learning_rate": 1.1070421503643754e-07, + "loss": 0.6983, + "step": 24434 + }, + { + "epoch": 0.89, + "learning_rate": 1.1070055297176548e-07, + "loss": 0.6829, + "step": 24435 + }, + { + "epoch": 0.89, + "learning_rate": 1.1069689090709342e-07, + "loss": 0.6855, + "step": 24436 + }, + { + "epoch": 0.89, + "learning_rate": 1.1069322884242136e-07, + "loss": 0.6944, + "step": 24437 + }, + { + "epoch": 0.89, + "learning_rate": 1.106895667777493e-07, + "loss": 0.693, + "step": 24438 + }, + { + "epoch": 0.89, + "learning_rate": 1.1068590471307724e-07, + "loss": 0.6858, + "step": 24439 + }, + { + "epoch": 0.89, + "learning_rate": 1.1068224264840515e-07, + "loss": 0.6755, + "step": 24440 + }, + { + "epoch": 0.89, + "learning_rate": 1.1067858058373309e-07, + "loss": 0.702, + "step": 24441 + }, + { + "epoch": 0.89, + "learning_rate": 1.1067491851906103e-07, + "loss": 0.6841, + "step": 24442 + }, + { + "epoch": 0.89, + "learning_rate": 1.1067125645438897e-07, + "loss": 0.7057, + "step": 24443 + }, + { + "epoch": 0.89, + "learning_rate": 1.1066759438971691e-07, + "loss": 0.7162, + "step": 24444 + }, + { + "epoch": 0.89, + "learning_rate": 1.1066393232504485e-07, + "loss": 0.6886, + "step": 24445 + }, + { + "epoch": 0.89, + "learning_rate": 1.1066027026037279e-07, + "loss": 0.6933, + "step": 24446 + }, + { + "epoch": 0.89, + "learning_rate": 1.1065660819570073e-07, + "loss": 0.6981, + "step": 24447 + }, + { + "epoch": 0.89, + "learning_rate": 1.1065294613102867e-07, + "loss": 0.6886, + "step": 24448 + }, + { + "epoch": 0.89, + "learning_rate": 1.1064928406635661e-07, + "loss": 0.6884, + "step": 24449 + }, + { + "epoch": 0.89, + "learning_rate": 1.1064562200168455e-07, + "loss": 0.7069, + "step": 24450 + }, + { + "epoch": 0.89, + "learning_rate": 1.1064195993701249e-07, + "loss": 0.7003, + "step": 24451 + }, + { + "epoch": 0.89, + "learning_rate": 1.1063829787234042e-07, + "loss": 0.6837, + "step": 24452 + }, + { + "epoch": 0.89, + "learning_rate": 1.1063463580766836e-07, + "loss": 0.6911, + "step": 24453 + }, + { + "epoch": 0.89, + "learning_rate": 1.106309737429963e-07, + "loss": 0.6977, + "step": 24454 + }, + { + "epoch": 0.89, + "learning_rate": 1.1062731167832424e-07, + "loss": 0.6868, + "step": 24455 + }, + { + "epoch": 0.89, + "learning_rate": 1.1062364961365218e-07, + "loss": 0.7134, + "step": 24456 + }, + { + "epoch": 0.89, + "learning_rate": 1.1061998754898012e-07, + "loss": 0.684, + "step": 24457 + }, + { + "epoch": 0.89, + "learning_rate": 1.1061632548430803e-07, + "loss": 0.6843, + "step": 24458 + }, + { + "epoch": 0.89, + "learning_rate": 1.1061266341963597e-07, + "loss": 0.7023, + "step": 24459 + }, + { + "epoch": 0.89, + "learning_rate": 1.1060900135496391e-07, + "loss": 0.6924, + "step": 24460 + }, + { + "epoch": 0.89, + "learning_rate": 1.1060533929029185e-07, + "loss": 0.7087, + "step": 24461 + }, + { + "epoch": 0.89, + "learning_rate": 1.1060167722561979e-07, + "loss": 0.6865, + "step": 24462 + }, + { + "epoch": 0.9, + "learning_rate": 1.1059801516094773e-07, + "loss": 0.687, + "step": 24463 + }, + { + "epoch": 0.9, + "learning_rate": 1.1059435309627567e-07, + "loss": 0.6989, + "step": 24464 + }, + { + "epoch": 0.9, + "learning_rate": 1.1059069103160361e-07, + "loss": 0.6826, + "step": 24465 + }, + { + "epoch": 0.9, + "learning_rate": 1.1058702896693155e-07, + "loss": 0.6953, + "step": 24466 + }, + { + "epoch": 0.9, + "learning_rate": 1.1058336690225949e-07, + "loss": 0.7047, + "step": 24467 + }, + { + "epoch": 0.9, + "learning_rate": 1.1057970483758743e-07, + "loss": 0.6935, + "step": 24468 + }, + { + "epoch": 0.9, + "learning_rate": 1.1057604277291537e-07, + "loss": 0.6874, + "step": 24469 + }, + { + "epoch": 0.9, + "learning_rate": 1.1057238070824331e-07, + "loss": 0.6968, + "step": 24470 + }, + { + "epoch": 0.9, + "learning_rate": 1.1056871864357125e-07, + "loss": 0.697, + "step": 24471 + }, + { + "epoch": 0.9, + "learning_rate": 1.1056505657889919e-07, + "loss": 0.6925, + "step": 24472 + }, + { + "epoch": 0.9, + "learning_rate": 1.1056139451422713e-07, + "loss": 0.6822, + "step": 24473 + }, + { + "epoch": 0.9, + "learning_rate": 1.1055773244955506e-07, + "loss": 0.691, + "step": 24474 + }, + { + "epoch": 0.9, + "learning_rate": 1.10554070384883e-07, + "loss": 0.696, + "step": 24475 + }, + { + "epoch": 0.9, + "learning_rate": 1.1055040832021092e-07, + "loss": 0.6951, + "step": 24476 + }, + { + "epoch": 0.9, + "learning_rate": 1.1054674625553886e-07, + "loss": 0.6853, + "step": 24477 + }, + { + "epoch": 0.9, + "learning_rate": 1.105430841908668e-07, + "loss": 0.6935, + "step": 24478 + }, + { + "epoch": 0.9, + "learning_rate": 1.1053942212619473e-07, + "loss": 0.7, + "step": 24479 + }, + { + "epoch": 0.9, + "learning_rate": 1.1053576006152267e-07, + "loss": 0.6954, + "step": 24480 + }, + { + "epoch": 0.9, + "learning_rate": 1.1053209799685061e-07, + "loss": 0.6895, + "step": 24481 + }, + { + "epoch": 0.9, + "learning_rate": 1.1052843593217855e-07, + "loss": 0.6829, + "step": 24482 + }, + { + "epoch": 0.9, + "learning_rate": 1.1052477386750649e-07, + "loss": 0.7089, + "step": 24483 + }, + { + "epoch": 0.9, + "learning_rate": 1.1052111180283443e-07, + "loss": 0.6833, + "step": 24484 + }, + { + "epoch": 0.9, + "learning_rate": 1.1051744973816237e-07, + "loss": 0.7071, + "step": 24485 + }, + { + "epoch": 0.9, + "learning_rate": 1.1051378767349031e-07, + "loss": 0.6887, + "step": 24486 + }, + { + "epoch": 0.9, + "learning_rate": 1.1051012560881825e-07, + "loss": 0.6775, + "step": 24487 + }, + { + "epoch": 0.9, + "learning_rate": 1.1050646354414619e-07, + "loss": 0.6971, + "step": 24488 + }, + { + "epoch": 0.9, + "learning_rate": 1.1050280147947413e-07, + "loss": 0.7016, + "step": 24489 + }, + { + "epoch": 0.9, + "learning_rate": 1.1049913941480207e-07, + "loss": 0.6944, + "step": 24490 + }, + { + "epoch": 0.9, + "learning_rate": 1.1049547735013001e-07, + "loss": 0.6839, + "step": 24491 + }, + { + "epoch": 0.9, + "learning_rate": 1.1049181528545795e-07, + "loss": 0.6904, + "step": 24492 + }, + { + "epoch": 0.9, + "learning_rate": 1.1048815322078589e-07, + "loss": 0.6895, + "step": 24493 + }, + { + "epoch": 0.9, + "learning_rate": 1.104844911561138e-07, + "loss": 0.6969, + "step": 24494 + }, + { + "epoch": 0.9, + "learning_rate": 1.1048082909144174e-07, + "loss": 0.701, + "step": 24495 + }, + { + "epoch": 0.9, + "learning_rate": 1.1047716702676968e-07, + "loss": 0.6891, + "step": 24496 + }, + { + "epoch": 0.9, + "learning_rate": 1.1047350496209762e-07, + "loss": 0.6861, + "step": 24497 + }, + { + "epoch": 0.9, + "learning_rate": 1.1046984289742556e-07, + "loss": 0.7015, + "step": 24498 + }, + { + "epoch": 0.9, + "learning_rate": 1.104661808327535e-07, + "loss": 0.7053, + "step": 24499 + }, + { + "epoch": 0.9, + "learning_rate": 1.1046251876808144e-07, + "loss": 0.6935, + "step": 24500 + }, + { + "epoch": 0.9, + "learning_rate": 1.1045885670340937e-07, + "loss": 0.6935, + "step": 24501 + }, + { + "epoch": 0.9, + "learning_rate": 1.1045519463873731e-07, + "loss": 0.6971, + "step": 24502 + }, + { + "epoch": 0.9, + "learning_rate": 1.1045153257406525e-07, + "loss": 0.6892, + "step": 24503 + }, + { + "epoch": 0.9, + "learning_rate": 1.1044787050939319e-07, + "loss": 0.6931, + "step": 24504 + }, + { + "epoch": 0.9, + "learning_rate": 1.1044420844472113e-07, + "loss": 0.6837, + "step": 24505 + }, + { + "epoch": 0.9, + "learning_rate": 1.1044054638004907e-07, + "loss": 0.6923, + "step": 24506 + }, + { + "epoch": 0.9, + "learning_rate": 1.1043688431537701e-07, + "loss": 0.7011, + "step": 24507 + }, + { + "epoch": 0.9, + "learning_rate": 1.1043322225070495e-07, + "loss": 0.6795, + "step": 24508 + }, + { + "epoch": 0.9, + "learning_rate": 1.1042956018603289e-07, + "loss": 0.692, + "step": 24509 + }, + { + "epoch": 0.9, + "learning_rate": 1.1042589812136083e-07, + "loss": 0.6831, + "step": 24510 + }, + { + "epoch": 0.9, + "learning_rate": 1.1042223605668877e-07, + "loss": 0.7109, + "step": 24511 + }, + { + "epoch": 0.9, + "learning_rate": 1.1041857399201668e-07, + "loss": 0.6868, + "step": 24512 + }, + { + "epoch": 0.9, + "learning_rate": 1.1041491192734462e-07, + "loss": 0.6895, + "step": 24513 + }, + { + "epoch": 0.9, + "learning_rate": 1.1041124986267256e-07, + "loss": 0.6966, + "step": 24514 + }, + { + "epoch": 0.9, + "learning_rate": 1.104075877980005e-07, + "loss": 0.6833, + "step": 24515 + }, + { + "epoch": 0.9, + "learning_rate": 1.1040392573332844e-07, + "loss": 0.6987, + "step": 24516 + }, + { + "epoch": 0.9, + "learning_rate": 1.1040026366865638e-07, + "loss": 0.6953, + "step": 24517 + }, + { + "epoch": 0.9, + "learning_rate": 1.1039660160398432e-07, + "loss": 0.6829, + "step": 24518 + }, + { + "epoch": 0.9, + "learning_rate": 1.1039293953931226e-07, + "loss": 0.6876, + "step": 24519 + }, + { + "epoch": 0.9, + "learning_rate": 1.103892774746402e-07, + "loss": 0.6885, + "step": 24520 + }, + { + "epoch": 0.9, + "learning_rate": 1.1038561540996814e-07, + "loss": 0.69, + "step": 24521 + }, + { + "epoch": 0.9, + "learning_rate": 1.1038195334529608e-07, + "loss": 0.6874, + "step": 24522 + }, + { + "epoch": 0.9, + "learning_rate": 1.1037829128062401e-07, + "loss": 0.6828, + "step": 24523 + }, + { + "epoch": 0.9, + "learning_rate": 1.1037462921595195e-07, + "loss": 0.6924, + "step": 24524 + }, + { + "epoch": 0.9, + "learning_rate": 1.1037096715127989e-07, + "loss": 0.6922, + "step": 24525 + }, + { + "epoch": 0.9, + "learning_rate": 1.1036730508660783e-07, + "loss": 0.6974, + "step": 24526 + }, + { + "epoch": 0.9, + "learning_rate": 1.1036364302193577e-07, + "loss": 0.7041, + "step": 24527 + }, + { + "epoch": 0.9, + "learning_rate": 1.1035998095726371e-07, + "loss": 0.6949, + "step": 24528 + }, + { + "epoch": 0.9, + "learning_rate": 1.1035631889259165e-07, + "loss": 0.6878, + "step": 24529 + }, + { + "epoch": 0.9, + "learning_rate": 1.1035265682791956e-07, + "loss": 0.6912, + "step": 24530 + }, + { + "epoch": 0.9, + "learning_rate": 1.103489947632475e-07, + "loss": 0.6914, + "step": 24531 + }, + { + "epoch": 0.9, + "learning_rate": 1.1034533269857544e-07, + "loss": 0.6985, + "step": 24532 + }, + { + "epoch": 0.9, + "learning_rate": 1.1034167063390338e-07, + "loss": 0.6959, + "step": 24533 + }, + { + "epoch": 0.9, + "learning_rate": 1.1033800856923132e-07, + "loss": 0.6975, + "step": 24534 + }, + { + "epoch": 0.9, + "learning_rate": 1.1033434650455926e-07, + "loss": 0.6892, + "step": 24535 + }, + { + "epoch": 0.9, + "learning_rate": 1.103306844398872e-07, + "loss": 0.6907, + "step": 24536 + }, + { + "epoch": 0.9, + "learning_rate": 1.1032702237521514e-07, + "loss": 0.6833, + "step": 24537 + }, + { + "epoch": 0.9, + "learning_rate": 1.1032336031054308e-07, + "loss": 0.7014, + "step": 24538 + }, + { + "epoch": 0.9, + "learning_rate": 1.1031969824587102e-07, + "loss": 0.6833, + "step": 24539 + }, + { + "epoch": 0.9, + "learning_rate": 1.1031603618119896e-07, + "loss": 0.688, + "step": 24540 + }, + { + "epoch": 0.9, + "learning_rate": 1.103123741165269e-07, + "loss": 0.6932, + "step": 24541 + }, + { + "epoch": 0.9, + "learning_rate": 1.1030871205185484e-07, + "loss": 0.6907, + "step": 24542 + }, + { + "epoch": 0.9, + "learning_rate": 1.1030504998718278e-07, + "loss": 0.6993, + "step": 24543 + }, + { + "epoch": 0.9, + "learning_rate": 1.1030138792251071e-07, + "loss": 0.692, + "step": 24544 + }, + { + "epoch": 0.9, + "learning_rate": 1.1029772585783865e-07, + "loss": 0.6896, + "step": 24545 + }, + { + "epoch": 0.9, + "learning_rate": 1.1029406379316659e-07, + "loss": 0.6842, + "step": 24546 + }, + { + "epoch": 0.9, + "learning_rate": 1.1029040172849453e-07, + "loss": 0.6911, + "step": 24547 + }, + { + "epoch": 0.9, + "learning_rate": 1.1028673966382245e-07, + "loss": 0.6879, + "step": 24548 + }, + { + "epoch": 0.9, + "learning_rate": 1.102830775991504e-07, + "loss": 0.6944, + "step": 24549 + }, + { + "epoch": 0.9, + "learning_rate": 1.1027941553447832e-07, + "loss": 0.6963, + "step": 24550 + }, + { + "epoch": 0.9, + "learning_rate": 1.1027575346980626e-07, + "loss": 0.6913, + "step": 24551 + }, + { + "epoch": 0.9, + "learning_rate": 1.102720914051342e-07, + "loss": 0.6931, + "step": 24552 + }, + { + "epoch": 0.9, + "learning_rate": 1.1026842934046214e-07, + "loss": 0.7023, + "step": 24553 + }, + { + "epoch": 0.9, + "learning_rate": 1.1026476727579008e-07, + "loss": 0.6954, + "step": 24554 + }, + { + "epoch": 0.9, + "learning_rate": 1.1026110521111802e-07, + "loss": 0.6926, + "step": 24555 + }, + { + "epoch": 0.9, + "learning_rate": 1.1025744314644596e-07, + "loss": 0.6918, + "step": 24556 + }, + { + "epoch": 0.9, + "learning_rate": 1.102537810817739e-07, + "loss": 0.6919, + "step": 24557 + }, + { + "epoch": 0.9, + "learning_rate": 1.1025011901710184e-07, + "loss": 0.6912, + "step": 24558 + }, + { + "epoch": 0.9, + "learning_rate": 1.1024645695242978e-07, + "loss": 0.6856, + "step": 24559 + }, + { + "epoch": 0.9, + "learning_rate": 1.1024279488775772e-07, + "loss": 0.6931, + "step": 24560 + }, + { + "epoch": 0.9, + "learning_rate": 1.1023913282308566e-07, + "loss": 0.6832, + "step": 24561 + }, + { + "epoch": 0.9, + "learning_rate": 1.102354707584136e-07, + "loss": 0.6896, + "step": 24562 + }, + { + "epoch": 0.9, + "learning_rate": 1.1023180869374154e-07, + "loss": 0.6939, + "step": 24563 + }, + { + "epoch": 0.9, + "learning_rate": 1.1022814662906948e-07, + "loss": 0.6903, + "step": 24564 + }, + { + "epoch": 0.9, + "learning_rate": 1.1022448456439742e-07, + "loss": 0.6784, + "step": 24565 + }, + { + "epoch": 0.9, + "learning_rate": 1.1022082249972533e-07, + "loss": 0.6929, + "step": 24566 + }, + { + "epoch": 0.9, + "learning_rate": 1.1021716043505327e-07, + "loss": 0.6917, + "step": 24567 + }, + { + "epoch": 0.9, + "learning_rate": 1.1021349837038121e-07, + "loss": 0.6946, + "step": 24568 + }, + { + "epoch": 0.9, + "learning_rate": 1.1020983630570915e-07, + "loss": 0.7096, + "step": 24569 + }, + { + "epoch": 0.9, + "learning_rate": 1.1020617424103709e-07, + "loss": 0.6868, + "step": 24570 + }, + { + "epoch": 0.9, + "learning_rate": 1.1020251217636503e-07, + "loss": 0.6907, + "step": 24571 + }, + { + "epoch": 0.9, + "learning_rate": 1.1019885011169296e-07, + "loss": 0.6906, + "step": 24572 + }, + { + "epoch": 0.9, + "learning_rate": 1.101951880470209e-07, + "loss": 0.6895, + "step": 24573 + }, + { + "epoch": 0.9, + "learning_rate": 1.1019152598234884e-07, + "loss": 0.6983, + "step": 24574 + }, + { + "epoch": 0.9, + "learning_rate": 1.1018786391767678e-07, + "loss": 0.6787, + "step": 24575 + }, + { + "epoch": 0.9, + "learning_rate": 1.1018420185300472e-07, + "loss": 0.6905, + "step": 24576 + }, + { + "epoch": 0.9, + "learning_rate": 1.1018053978833266e-07, + "loss": 0.6992, + "step": 24577 + }, + { + "epoch": 0.9, + "learning_rate": 1.101768777236606e-07, + "loss": 0.686, + "step": 24578 + }, + { + "epoch": 0.9, + "learning_rate": 1.1017321565898854e-07, + "loss": 0.6917, + "step": 24579 + }, + { + "epoch": 0.9, + "learning_rate": 1.1016955359431648e-07, + "loss": 0.6949, + "step": 24580 + }, + { + "epoch": 0.9, + "learning_rate": 1.1016589152964442e-07, + "loss": 0.6905, + "step": 24581 + }, + { + "epoch": 0.9, + "learning_rate": 1.1016222946497236e-07, + "loss": 0.689, + "step": 24582 + }, + { + "epoch": 0.9, + "learning_rate": 1.101585674003003e-07, + "loss": 0.6869, + "step": 24583 + }, + { + "epoch": 0.9, + "learning_rate": 1.1015490533562821e-07, + "loss": 0.6966, + "step": 24584 + }, + { + "epoch": 0.9, + "learning_rate": 1.1015124327095615e-07, + "loss": 0.7064, + "step": 24585 + }, + { + "epoch": 0.9, + "learning_rate": 1.1014758120628409e-07, + "loss": 0.6737, + "step": 24586 + }, + { + "epoch": 0.9, + "learning_rate": 1.1014391914161203e-07, + "loss": 0.7008, + "step": 24587 + }, + { + "epoch": 0.9, + "learning_rate": 1.1014025707693997e-07, + "loss": 0.6897, + "step": 24588 + }, + { + "epoch": 0.9, + "learning_rate": 1.1013659501226791e-07, + "loss": 0.6814, + "step": 24589 + }, + { + "epoch": 0.9, + "learning_rate": 1.1013293294759585e-07, + "loss": 0.6951, + "step": 24590 + }, + { + "epoch": 0.9, + "learning_rate": 1.1012927088292379e-07, + "loss": 0.7027, + "step": 24591 + }, + { + "epoch": 0.9, + "learning_rate": 1.1012560881825173e-07, + "loss": 0.7038, + "step": 24592 + }, + { + "epoch": 0.9, + "learning_rate": 1.1012194675357967e-07, + "loss": 0.6803, + "step": 24593 + }, + { + "epoch": 0.9, + "learning_rate": 1.101182846889076e-07, + "loss": 0.6885, + "step": 24594 + }, + { + "epoch": 0.9, + "learning_rate": 1.1011462262423554e-07, + "loss": 0.6874, + "step": 24595 + }, + { + "epoch": 0.9, + "learning_rate": 1.1011096055956348e-07, + "loss": 0.6922, + "step": 24596 + }, + { + "epoch": 0.9, + "learning_rate": 1.1010729849489142e-07, + "loss": 0.6818, + "step": 24597 + }, + { + "epoch": 0.9, + "learning_rate": 1.1010363643021936e-07, + "loss": 0.6767, + "step": 24598 + }, + { + "epoch": 0.9, + "learning_rate": 1.100999743655473e-07, + "loss": 0.6994, + "step": 24599 + }, + { + "epoch": 0.9, + "learning_rate": 1.1009631230087524e-07, + "loss": 0.6974, + "step": 24600 + }, + { + "epoch": 0.9, + "learning_rate": 1.1009265023620318e-07, + "loss": 0.6938, + "step": 24601 + }, + { + "epoch": 0.9, + "learning_rate": 1.1008898817153109e-07, + "loss": 0.6886, + "step": 24602 + }, + { + "epoch": 0.9, + "learning_rate": 1.1008532610685903e-07, + "loss": 0.6865, + "step": 24603 + }, + { + "epoch": 0.9, + "learning_rate": 1.1008166404218697e-07, + "loss": 0.688, + "step": 24604 + }, + { + "epoch": 0.9, + "learning_rate": 1.1007800197751491e-07, + "loss": 0.6867, + "step": 24605 + }, + { + "epoch": 0.9, + "learning_rate": 1.1007433991284285e-07, + "loss": 0.6969, + "step": 24606 + }, + { + "epoch": 0.9, + "learning_rate": 1.1007067784817079e-07, + "loss": 0.6945, + "step": 24607 + }, + { + "epoch": 0.9, + "learning_rate": 1.1006701578349873e-07, + "loss": 0.7114, + "step": 24608 + }, + { + "epoch": 0.9, + "learning_rate": 1.1006335371882667e-07, + "loss": 0.6989, + "step": 24609 + }, + { + "epoch": 0.9, + "learning_rate": 1.1005969165415461e-07, + "loss": 0.7065, + "step": 24610 + }, + { + "epoch": 0.9, + "learning_rate": 1.1005602958948255e-07, + "loss": 0.6917, + "step": 24611 + }, + { + "epoch": 0.9, + "learning_rate": 1.1005236752481049e-07, + "loss": 0.6909, + "step": 24612 + }, + { + "epoch": 0.9, + "learning_rate": 1.1004870546013843e-07, + "loss": 0.682, + "step": 24613 + }, + { + "epoch": 0.9, + "learning_rate": 1.1004504339546637e-07, + "loss": 0.6967, + "step": 24614 + }, + { + "epoch": 0.9, + "learning_rate": 1.100413813307943e-07, + "loss": 0.6997, + "step": 24615 + }, + { + "epoch": 0.9, + "learning_rate": 1.1003771926612224e-07, + "loss": 0.7004, + "step": 24616 + }, + { + "epoch": 0.9, + "learning_rate": 1.1003405720145018e-07, + "loss": 0.6878, + "step": 24617 + }, + { + "epoch": 0.9, + "learning_rate": 1.1003039513677812e-07, + "loss": 0.6994, + "step": 24618 + }, + { + "epoch": 0.9, + "learning_rate": 1.1002673307210606e-07, + "loss": 0.6974, + "step": 24619 + }, + { + "epoch": 0.9, + "learning_rate": 1.1002307100743398e-07, + "loss": 0.7006, + "step": 24620 + }, + { + "epoch": 0.9, + "learning_rate": 1.1001940894276191e-07, + "loss": 0.6805, + "step": 24621 + }, + { + "epoch": 0.9, + "learning_rate": 1.1001574687808985e-07, + "loss": 0.6884, + "step": 24622 + }, + { + "epoch": 0.9, + "learning_rate": 1.1001208481341779e-07, + "loss": 0.7, + "step": 24623 + }, + { + "epoch": 0.9, + "learning_rate": 1.1000842274874573e-07, + "loss": 0.701, + "step": 24624 + }, + { + "epoch": 0.9, + "learning_rate": 1.1000476068407367e-07, + "loss": 0.6796, + "step": 24625 + }, + { + "epoch": 0.9, + "learning_rate": 1.1000109861940161e-07, + "loss": 0.6937, + "step": 24626 + }, + { + "epoch": 0.9, + "learning_rate": 1.0999743655472955e-07, + "loss": 0.6864, + "step": 24627 + }, + { + "epoch": 0.9, + "learning_rate": 1.0999377449005749e-07, + "loss": 0.6926, + "step": 24628 + }, + { + "epoch": 0.9, + "learning_rate": 1.0999011242538543e-07, + "loss": 0.6859, + "step": 24629 + }, + { + "epoch": 0.9, + "learning_rate": 1.0998645036071337e-07, + "loss": 0.6996, + "step": 24630 + }, + { + "epoch": 0.9, + "learning_rate": 1.0998278829604131e-07, + "loss": 0.6883, + "step": 24631 + }, + { + "epoch": 0.9, + "learning_rate": 1.0997912623136925e-07, + "loss": 0.687, + "step": 24632 + }, + { + "epoch": 0.9, + "learning_rate": 1.0997546416669719e-07, + "loss": 0.6816, + "step": 24633 + }, + { + "epoch": 0.9, + "learning_rate": 1.0997180210202513e-07, + "loss": 0.6971, + "step": 24634 + }, + { + "epoch": 0.9, + "learning_rate": 1.0996814003735307e-07, + "loss": 0.7025, + "step": 24635 + }, + { + "epoch": 0.9, + "learning_rate": 1.0996447797268101e-07, + "loss": 0.6925, + "step": 24636 + }, + { + "epoch": 0.9, + "learning_rate": 1.0996081590800894e-07, + "loss": 0.6879, + "step": 24637 + }, + { + "epoch": 0.9, + "learning_rate": 1.0995715384333686e-07, + "loss": 0.6858, + "step": 24638 + }, + { + "epoch": 0.9, + "learning_rate": 1.099534917786648e-07, + "loss": 0.6871, + "step": 24639 + }, + { + "epoch": 0.9, + "learning_rate": 1.0994982971399274e-07, + "loss": 0.7022, + "step": 24640 + }, + { + "epoch": 0.9, + "learning_rate": 1.0994616764932068e-07, + "loss": 0.7017, + "step": 24641 + }, + { + "epoch": 0.9, + "learning_rate": 1.0994250558464862e-07, + "loss": 0.6984, + "step": 24642 + }, + { + "epoch": 0.9, + "learning_rate": 1.0993884351997655e-07, + "loss": 0.7058, + "step": 24643 + }, + { + "epoch": 0.9, + "learning_rate": 1.0993518145530449e-07, + "loss": 0.6973, + "step": 24644 + }, + { + "epoch": 0.9, + "learning_rate": 1.0993151939063243e-07, + "loss": 0.6916, + "step": 24645 + }, + { + "epoch": 0.9, + "learning_rate": 1.0992785732596037e-07, + "loss": 0.6861, + "step": 24646 + }, + { + "epoch": 0.9, + "learning_rate": 1.0992419526128831e-07, + "loss": 0.6905, + "step": 24647 + }, + { + "epoch": 0.9, + "learning_rate": 1.0992053319661625e-07, + "loss": 0.6906, + "step": 24648 + }, + { + "epoch": 0.9, + "learning_rate": 1.0991687113194419e-07, + "loss": 0.698, + "step": 24649 + }, + { + "epoch": 0.9, + "learning_rate": 1.0991320906727213e-07, + "loss": 0.6948, + "step": 24650 + }, + { + "epoch": 0.9, + "learning_rate": 1.0990954700260007e-07, + "loss": 0.7037, + "step": 24651 + }, + { + "epoch": 0.9, + "learning_rate": 1.0990588493792801e-07, + "loss": 0.6855, + "step": 24652 + }, + { + "epoch": 0.9, + "learning_rate": 1.0990222287325595e-07, + "loss": 0.6717, + "step": 24653 + }, + { + "epoch": 0.9, + "learning_rate": 1.0989856080858389e-07, + "loss": 0.6932, + "step": 24654 + }, + { + "epoch": 0.9, + "learning_rate": 1.0989489874391183e-07, + "loss": 0.6976, + "step": 24655 + }, + { + "epoch": 0.9, + "learning_rate": 1.0989123667923974e-07, + "loss": 0.6951, + "step": 24656 + }, + { + "epoch": 0.9, + "learning_rate": 1.0988757461456768e-07, + "loss": 0.6961, + "step": 24657 + }, + { + "epoch": 0.9, + "learning_rate": 1.0988391254989562e-07, + "loss": 0.6991, + "step": 24658 + }, + { + "epoch": 0.9, + "learning_rate": 1.0988025048522356e-07, + "loss": 0.6995, + "step": 24659 + }, + { + "epoch": 0.9, + "learning_rate": 1.098765884205515e-07, + "loss": 0.6975, + "step": 24660 + }, + { + "epoch": 0.9, + "learning_rate": 1.0987292635587944e-07, + "loss": 0.6849, + "step": 24661 + }, + { + "epoch": 0.9, + "learning_rate": 1.0986926429120738e-07, + "loss": 0.6904, + "step": 24662 + }, + { + "epoch": 0.9, + "learning_rate": 1.0986560222653532e-07, + "loss": 0.6808, + "step": 24663 + }, + { + "epoch": 0.9, + "learning_rate": 1.0986194016186325e-07, + "loss": 0.6879, + "step": 24664 + }, + { + "epoch": 0.9, + "learning_rate": 1.0985827809719119e-07, + "loss": 0.6874, + "step": 24665 + }, + { + "epoch": 0.9, + "learning_rate": 1.0985461603251913e-07, + "loss": 0.6927, + "step": 24666 + }, + { + "epoch": 0.9, + "learning_rate": 1.0985095396784707e-07, + "loss": 0.6879, + "step": 24667 + }, + { + "epoch": 0.9, + "learning_rate": 1.0984729190317501e-07, + "loss": 0.6847, + "step": 24668 + }, + { + "epoch": 0.9, + "learning_rate": 1.0984362983850295e-07, + "loss": 0.6981, + "step": 24669 + }, + { + "epoch": 0.9, + "learning_rate": 1.0983996777383089e-07, + "loss": 0.702, + "step": 24670 + }, + { + "epoch": 0.9, + "learning_rate": 1.0983630570915883e-07, + "loss": 0.6881, + "step": 24671 + }, + { + "epoch": 0.9, + "learning_rate": 1.0983264364448677e-07, + "loss": 0.7048, + "step": 24672 + }, + { + "epoch": 0.9, + "learning_rate": 1.0982898157981468e-07, + "loss": 0.6954, + "step": 24673 + }, + { + "epoch": 0.9, + "learning_rate": 1.0982531951514262e-07, + "loss": 0.6857, + "step": 24674 + }, + { + "epoch": 0.9, + "learning_rate": 1.0982165745047056e-07, + "loss": 0.6985, + "step": 24675 + }, + { + "epoch": 0.9, + "learning_rate": 1.098179953857985e-07, + "loss": 0.6903, + "step": 24676 + }, + { + "epoch": 0.9, + "learning_rate": 1.0981433332112644e-07, + "loss": 0.7007, + "step": 24677 + }, + { + "epoch": 0.9, + "learning_rate": 1.0981067125645438e-07, + "loss": 0.6969, + "step": 24678 + }, + { + "epoch": 0.9, + "learning_rate": 1.0980700919178232e-07, + "loss": 0.6866, + "step": 24679 + }, + { + "epoch": 0.9, + "learning_rate": 1.0980334712711026e-07, + "loss": 0.6947, + "step": 24680 + }, + { + "epoch": 0.9, + "learning_rate": 1.097996850624382e-07, + "loss": 0.6981, + "step": 24681 + }, + { + "epoch": 0.9, + "learning_rate": 1.0979602299776614e-07, + "loss": 0.6857, + "step": 24682 + }, + { + "epoch": 0.9, + "learning_rate": 1.0979236093309408e-07, + "loss": 0.6875, + "step": 24683 + }, + { + "epoch": 0.9, + "learning_rate": 1.0978869886842202e-07, + "loss": 0.6879, + "step": 24684 + }, + { + "epoch": 0.9, + "learning_rate": 1.0978503680374996e-07, + "loss": 0.6948, + "step": 24685 + }, + { + "epoch": 0.9, + "learning_rate": 1.0978137473907789e-07, + "loss": 0.7031, + "step": 24686 + }, + { + "epoch": 0.9, + "learning_rate": 1.0977771267440583e-07, + "loss": 0.6933, + "step": 24687 + }, + { + "epoch": 0.9, + "learning_rate": 1.0977405060973377e-07, + "loss": 0.6987, + "step": 24688 + }, + { + "epoch": 0.9, + "learning_rate": 1.097703885450617e-07, + "loss": 0.7022, + "step": 24689 + }, + { + "epoch": 0.9, + "learning_rate": 1.0976672648038965e-07, + "loss": 0.6896, + "step": 24690 + }, + { + "epoch": 0.9, + "learning_rate": 1.0976306441571757e-07, + "loss": 0.6972, + "step": 24691 + }, + { + "epoch": 0.9, + "learning_rate": 1.097594023510455e-07, + "loss": 0.6954, + "step": 24692 + }, + { + "epoch": 0.9, + "learning_rate": 1.0975574028637344e-07, + "loss": 0.69, + "step": 24693 + }, + { + "epoch": 0.9, + "learning_rate": 1.0975207822170138e-07, + "loss": 0.6873, + "step": 24694 + }, + { + "epoch": 0.9, + "learning_rate": 1.0974841615702932e-07, + "loss": 0.684, + "step": 24695 + }, + { + "epoch": 0.9, + "learning_rate": 1.0974475409235726e-07, + "loss": 0.6955, + "step": 24696 + }, + { + "epoch": 0.9, + "learning_rate": 1.097410920276852e-07, + "loss": 0.7062, + "step": 24697 + }, + { + "epoch": 0.9, + "learning_rate": 1.0973742996301314e-07, + "loss": 0.704, + "step": 24698 + }, + { + "epoch": 0.9, + "learning_rate": 1.0973376789834108e-07, + "loss": 0.6822, + "step": 24699 + }, + { + "epoch": 0.9, + "learning_rate": 1.0973010583366902e-07, + "loss": 0.7011, + "step": 24700 + }, + { + "epoch": 0.9, + "learning_rate": 1.0972644376899696e-07, + "loss": 0.6958, + "step": 24701 + }, + { + "epoch": 0.9, + "learning_rate": 1.097227817043249e-07, + "loss": 0.7158, + "step": 24702 + }, + { + "epoch": 0.9, + "learning_rate": 1.0971911963965284e-07, + "loss": 0.6905, + "step": 24703 + }, + { + "epoch": 0.9, + "learning_rate": 1.0971545757498078e-07, + "loss": 0.6913, + "step": 24704 + }, + { + "epoch": 0.9, + "learning_rate": 1.0971179551030872e-07, + "loss": 0.6973, + "step": 24705 + }, + { + "epoch": 0.9, + "learning_rate": 1.0970813344563666e-07, + "loss": 0.688, + "step": 24706 + }, + { + "epoch": 0.9, + "learning_rate": 1.097044713809646e-07, + "loss": 0.6875, + "step": 24707 + }, + { + "epoch": 0.9, + "learning_rate": 1.0970080931629252e-07, + "loss": 0.6896, + "step": 24708 + }, + { + "epoch": 0.9, + "learning_rate": 1.0969714725162045e-07, + "loss": 0.6899, + "step": 24709 + }, + { + "epoch": 0.9, + "learning_rate": 1.0969348518694839e-07, + "loss": 0.6914, + "step": 24710 + }, + { + "epoch": 0.9, + "learning_rate": 1.0968982312227633e-07, + "loss": 0.6855, + "step": 24711 + }, + { + "epoch": 0.9, + "learning_rate": 1.0968616105760427e-07, + "loss": 0.6785, + "step": 24712 + }, + { + "epoch": 0.9, + "learning_rate": 1.0968249899293221e-07, + "loss": 0.6939, + "step": 24713 + }, + { + "epoch": 0.9, + "learning_rate": 1.0967883692826014e-07, + "loss": 0.71, + "step": 24714 + }, + { + "epoch": 0.9, + "learning_rate": 1.0967517486358808e-07, + "loss": 0.7046, + "step": 24715 + }, + { + "epoch": 0.9, + "learning_rate": 1.0967151279891602e-07, + "loss": 0.6828, + "step": 24716 + }, + { + "epoch": 0.9, + "learning_rate": 1.0966785073424396e-07, + "loss": 0.705, + "step": 24717 + }, + { + "epoch": 0.9, + "learning_rate": 1.096641886695719e-07, + "loss": 0.6897, + "step": 24718 + }, + { + "epoch": 0.9, + "learning_rate": 1.0966052660489984e-07, + "loss": 0.7115, + "step": 24719 + }, + { + "epoch": 0.9, + "learning_rate": 1.0965686454022778e-07, + "loss": 0.6943, + "step": 24720 + }, + { + "epoch": 0.9, + "learning_rate": 1.0965320247555572e-07, + "loss": 0.6962, + "step": 24721 + }, + { + "epoch": 0.9, + "learning_rate": 1.0964954041088366e-07, + "loss": 0.6876, + "step": 24722 + }, + { + "epoch": 0.9, + "learning_rate": 1.096458783462116e-07, + "loss": 0.6855, + "step": 24723 + }, + { + "epoch": 0.9, + "learning_rate": 1.0964221628153954e-07, + "loss": 0.6923, + "step": 24724 + }, + { + "epoch": 0.9, + "learning_rate": 1.0963855421686748e-07, + "loss": 0.6959, + "step": 24725 + }, + { + "epoch": 0.9, + "learning_rate": 1.0963489215219542e-07, + "loss": 0.6909, + "step": 24726 + }, + { + "epoch": 0.9, + "learning_rate": 1.0963123008752333e-07, + "loss": 0.6906, + "step": 24727 + }, + { + "epoch": 0.9, + "learning_rate": 1.0962756802285127e-07, + "loss": 0.7116, + "step": 24728 + }, + { + "epoch": 0.9, + "learning_rate": 1.0962390595817921e-07, + "loss": 0.6975, + "step": 24729 + }, + { + "epoch": 0.9, + "learning_rate": 1.0962024389350715e-07, + "loss": 0.6773, + "step": 24730 + }, + { + "epoch": 0.9, + "learning_rate": 1.0961658182883509e-07, + "loss": 0.6897, + "step": 24731 + }, + { + "epoch": 0.9, + "learning_rate": 1.0961291976416303e-07, + "loss": 0.6901, + "step": 24732 + }, + { + "epoch": 0.9, + "learning_rate": 1.0960925769949097e-07, + "loss": 0.6986, + "step": 24733 + }, + { + "epoch": 0.9, + "learning_rate": 1.0960559563481891e-07, + "loss": 0.6921, + "step": 24734 + }, + { + "epoch": 0.9, + "learning_rate": 1.0960193357014684e-07, + "loss": 0.6929, + "step": 24735 + }, + { + "epoch": 0.91, + "learning_rate": 1.0959827150547478e-07, + "loss": 0.6986, + "step": 24736 + }, + { + "epoch": 0.91, + "learning_rate": 1.0959460944080272e-07, + "loss": 0.6906, + "step": 24737 + }, + { + "epoch": 0.91, + "learning_rate": 1.0959094737613066e-07, + "loss": 0.7149, + "step": 24738 + }, + { + "epoch": 0.91, + "learning_rate": 1.095872853114586e-07, + "loss": 0.7029, + "step": 24739 + }, + { + "epoch": 0.91, + "learning_rate": 1.0958362324678654e-07, + "loss": 0.6981, + "step": 24740 + }, + { + "epoch": 0.91, + "learning_rate": 1.0957996118211448e-07, + "loss": 0.6959, + "step": 24741 + }, + { + "epoch": 0.91, + "learning_rate": 1.0957629911744242e-07, + "loss": 0.6953, + "step": 24742 + }, + { + "epoch": 0.91, + "learning_rate": 1.0957263705277036e-07, + "loss": 0.6894, + "step": 24743 + }, + { + "epoch": 0.91, + "learning_rate": 1.095689749880983e-07, + "loss": 0.6998, + "step": 24744 + }, + { + "epoch": 0.91, + "learning_rate": 1.0956531292342621e-07, + "loss": 0.6932, + "step": 24745 + }, + { + "epoch": 0.91, + "learning_rate": 1.0956165085875415e-07, + "loss": 0.6943, + "step": 24746 + }, + { + "epoch": 0.91, + "learning_rate": 1.0955798879408209e-07, + "loss": 0.6873, + "step": 24747 + }, + { + "epoch": 0.91, + "learning_rate": 1.0955432672941003e-07, + "loss": 0.6997, + "step": 24748 + }, + { + "epoch": 0.91, + "learning_rate": 1.0955066466473797e-07, + "loss": 0.6895, + "step": 24749 + }, + { + "epoch": 0.91, + "learning_rate": 1.0954700260006591e-07, + "loss": 0.7081, + "step": 24750 + }, + { + "epoch": 0.91, + "learning_rate": 1.0954334053539385e-07, + "loss": 0.6937, + "step": 24751 + }, + { + "epoch": 0.91, + "learning_rate": 1.0953967847072179e-07, + "loss": 0.6941, + "step": 24752 + }, + { + "epoch": 0.91, + "learning_rate": 1.0953601640604973e-07, + "loss": 0.6826, + "step": 24753 + }, + { + "epoch": 0.91, + "learning_rate": 1.0953235434137767e-07, + "loss": 0.7062, + "step": 24754 + }, + { + "epoch": 0.91, + "learning_rate": 1.0952869227670561e-07, + "loss": 0.6857, + "step": 24755 + }, + { + "epoch": 0.91, + "learning_rate": 1.0952503021203355e-07, + "loss": 0.7036, + "step": 24756 + }, + { + "epoch": 0.91, + "learning_rate": 1.0952136814736148e-07, + "loss": 0.7066, + "step": 24757 + }, + { + "epoch": 0.91, + "learning_rate": 1.0951770608268941e-07, + "loss": 0.696, + "step": 24758 + }, + { + "epoch": 0.91, + "learning_rate": 1.0951404401801735e-07, + "loss": 0.6971, + "step": 24759 + }, + { + "epoch": 0.91, + "learning_rate": 1.095103819533453e-07, + "loss": 0.6878, + "step": 24760 + }, + { + "epoch": 0.91, + "learning_rate": 1.0950671988867323e-07, + "loss": 0.6854, + "step": 24761 + }, + { + "epoch": 0.91, + "learning_rate": 1.0950305782400117e-07, + "loss": 0.6937, + "step": 24762 + }, + { + "epoch": 0.91, + "learning_rate": 1.0949939575932909e-07, + "loss": 0.6913, + "step": 24763 + }, + { + "epoch": 0.91, + "learning_rate": 1.0949573369465703e-07, + "loss": 0.7104, + "step": 24764 + }, + { + "epoch": 0.91, + "learning_rate": 1.0949207162998497e-07, + "loss": 0.6992, + "step": 24765 + }, + { + "epoch": 0.91, + "learning_rate": 1.0948840956531291e-07, + "loss": 0.6922, + "step": 24766 + }, + { + "epoch": 0.91, + "learning_rate": 1.0948474750064085e-07, + "loss": 0.7124, + "step": 24767 + }, + { + "epoch": 0.91, + "learning_rate": 1.0948108543596879e-07, + "loss": 0.6932, + "step": 24768 + }, + { + "epoch": 0.91, + "learning_rate": 1.0947742337129673e-07, + "loss": 0.7001, + "step": 24769 + }, + { + "epoch": 0.91, + "learning_rate": 1.0947376130662467e-07, + "loss": 0.6921, + "step": 24770 + }, + { + "epoch": 0.91, + "learning_rate": 1.0947009924195261e-07, + "loss": 0.6894, + "step": 24771 + }, + { + "epoch": 0.91, + "learning_rate": 1.0946643717728055e-07, + "loss": 0.6836, + "step": 24772 + }, + { + "epoch": 0.91, + "learning_rate": 1.0946277511260849e-07, + "loss": 0.6959, + "step": 24773 + }, + { + "epoch": 0.91, + "learning_rate": 1.0945911304793643e-07, + "loss": 0.6975, + "step": 24774 + }, + { + "epoch": 0.91, + "learning_rate": 1.0945545098326437e-07, + "loss": 0.7079, + "step": 24775 + }, + { + "epoch": 0.91, + "learning_rate": 1.0945178891859231e-07, + "loss": 0.6917, + "step": 24776 + }, + { + "epoch": 0.91, + "learning_rate": 1.0944812685392025e-07, + "loss": 0.6886, + "step": 24777 + }, + { + "epoch": 0.91, + "learning_rate": 1.0944446478924819e-07, + "loss": 0.6964, + "step": 24778 + }, + { + "epoch": 0.91, + "learning_rate": 1.0944080272457611e-07, + "loss": 0.6876, + "step": 24779 + }, + { + "epoch": 0.91, + "learning_rate": 1.0943714065990405e-07, + "loss": 0.6899, + "step": 24780 + }, + { + "epoch": 0.91, + "learning_rate": 1.0943347859523198e-07, + "loss": 0.6996, + "step": 24781 + }, + { + "epoch": 0.91, + "learning_rate": 1.0942981653055992e-07, + "loss": 0.6927, + "step": 24782 + }, + { + "epoch": 0.91, + "learning_rate": 1.0942615446588786e-07, + "loss": 0.6906, + "step": 24783 + }, + { + "epoch": 0.91, + "learning_rate": 1.0942249240121579e-07, + "loss": 0.7045, + "step": 24784 + }, + { + "epoch": 0.91, + "learning_rate": 1.0941883033654373e-07, + "loss": 0.6968, + "step": 24785 + }, + { + "epoch": 0.91, + "learning_rate": 1.0941516827187167e-07, + "loss": 0.6854, + "step": 24786 + }, + { + "epoch": 0.91, + "learning_rate": 1.0941150620719961e-07, + "loss": 0.6928, + "step": 24787 + }, + { + "epoch": 0.91, + "learning_rate": 1.0940784414252755e-07, + "loss": 0.6888, + "step": 24788 + }, + { + "epoch": 0.91, + "learning_rate": 1.0940418207785549e-07, + "loss": 0.6924, + "step": 24789 + }, + { + "epoch": 0.91, + "learning_rate": 1.0940052001318343e-07, + "loss": 0.6889, + "step": 24790 + }, + { + "epoch": 0.91, + "learning_rate": 1.0939685794851137e-07, + "loss": 0.6907, + "step": 24791 + }, + { + "epoch": 0.91, + "learning_rate": 1.093931958838393e-07, + "loss": 0.6937, + "step": 24792 + }, + { + "epoch": 0.91, + "learning_rate": 1.0938953381916725e-07, + "loss": 0.6975, + "step": 24793 + }, + { + "epoch": 0.91, + "learning_rate": 1.0938587175449519e-07, + "loss": 0.6697, + "step": 24794 + }, + { + "epoch": 0.91, + "learning_rate": 1.0938220968982313e-07, + "loss": 0.6924, + "step": 24795 + }, + { + "epoch": 0.91, + "learning_rate": 1.0937854762515107e-07, + "loss": 0.6958, + "step": 24796 + }, + { + "epoch": 0.91, + "learning_rate": 1.09374885560479e-07, + "loss": 0.6864, + "step": 24797 + }, + { + "epoch": 0.91, + "learning_rate": 1.0937122349580695e-07, + "loss": 0.7046, + "step": 24798 + }, + { + "epoch": 0.91, + "learning_rate": 1.0936756143113486e-07, + "loss": 0.6986, + "step": 24799 + }, + { + "epoch": 0.91, + "learning_rate": 1.093638993664628e-07, + "loss": 0.6907, + "step": 24800 + }, + { + "epoch": 0.91, + "learning_rate": 1.0936023730179074e-07, + "loss": 0.6854, + "step": 24801 + }, + { + "epoch": 0.91, + "learning_rate": 1.0935657523711868e-07, + "loss": 0.6975, + "step": 24802 + }, + { + "epoch": 0.91, + "learning_rate": 1.0935291317244662e-07, + "loss": 0.7051, + "step": 24803 + }, + { + "epoch": 0.91, + "learning_rate": 1.0934925110777456e-07, + "loss": 0.6853, + "step": 24804 + }, + { + "epoch": 0.91, + "learning_rate": 1.093455890431025e-07, + "loss": 0.7048, + "step": 24805 + }, + { + "epoch": 0.91, + "learning_rate": 1.0934192697843043e-07, + "loss": 0.6918, + "step": 24806 + }, + { + "epoch": 0.91, + "learning_rate": 1.0933826491375837e-07, + "loss": 0.6974, + "step": 24807 + }, + { + "epoch": 0.91, + "learning_rate": 1.093346028490863e-07, + "loss": 0.689, + "step": 24808 + }, + { + "epoch": 0.91, + "learning_rate": 1.0933094078441425e-07, + "loss": 0.6935, + "step": 24809 + }, + { + "epoch": 0.91, + "learning_rate": 1.0932727871974219e-07, + "loss": 0.6984, + "step": 24810 + }, + { + "epoch": 0.91, + "learning_rate": 1.0932361665507013e-07, + "loss": 0.702, + "step": 24811 + }, + { + "epoch": 0.91, + "learning_rate": 1.0931995459039807e-07, + "loss": 0.6901, + "step": 24812 + }, + { + "epoch": 0.91, + "learning_rate": 1.09316292525726e-07, + "loss": 0.7031, + "step": 24813 + }, + { + "epoch": 0.91, + "learning_rate": 1.0931263046105395e-07, + "loss": 0.692, + "step": 24814 + }, + { + "epoch": 0.91, + "learning_rate": 1.0930896839638189e-07, + "loss": 0.6895, + "step": 24815 + }, + { + "epoch": 0.91, + "learning_rate": 1.0930530633170982e-07, + "loss": 0.7013, + "step": 24816 + }, + { + "epoch": 0.91, + "learning_rate": 1.0930164426703774e-07, + "loss": 0.6947, + "step": 24817 + }, + { + "epoch": 0.91, + "learning_rate": 1.0929798220236568e-07, + "loss": 0.6993, + "step": 24818 + }, + { + "epoch": 0.91, + "learning_rate": 1.0929432013769362e-07, + "loss": 0.6911, + "step": 24819 + }, + { + "epoch": 0.91, + "learning_rate": 1.0929065807302156e-07, + "loss": 0.6798, + "step": 24820 + }, + { + "epoch": 0.91, + "learning_rate": 1.092869960083495e-07, + "loss": 0.6956, + "step": 24821 + }, + { + "epoch": 0.91, + "learning_rate": 1.0928333394367744e-07, + "loss": 0.7037, + "step": 24822 + }, + { + "epoch": 0.91, + "learning_rate": 1.0927967187900538e-07, + "loss": 0.6864, + "step": 24823 + }, + { + "epoch": 0.91, + "learning_rate": 1.0927600981433332e-07, + "loss": 0.6843, + "step": 24824 + }, + { + "epoch": 0.91, + "learning_rate": 1.0927234774966126e-07, + "loss": 0.6723, + "step": 24825 + }, + { + "epoch": 0.91, + "learning_rate": 1.092686856849892e-07, + "loss": 0.7075, + "step": 24826 + }, + { + "epoch": 0.91, + "learning_rate": 1.0926502362031714e-07, + "loss": 0.6796, + "step": 24827 + }, + { + "epoch": 0.91, + "learning_rate": 1.0926136155564506e-07, + "loss": 0.6872, + "step": 24828 + }, + { + "epoch": 0.91, + "learning_rate": 1.09257699490973e-07, + "loss": 0.6887, + "step": 24829 + }, + { + "epoch": 0.91, + "learning_rate": 1.0925403742630094e-07, + "loss": 0.6884, + "step": 24830 + }, + { + "epoch": 0.91, + "learning_rate": 1.0925037536162888e-07, + "loss": 0.7039, + "step": 24831 + }, + { + "epoch": 0.91, + "learning_rate": 1.0924671329695682e-07, + "loss": 0.6976, + "step": 24832 + }, + { + "epoch": 0.91, + "learning_rate": 1.0924305123228476e-07, + "loss": 0.6923, + "step": 24833 + }, + { + "epoch": 0.91, + "learning_rate": 1.092393891676127e-07, + "loss": 0.6956, + "step": 24834 + }, + { + "epoch": 0.91, + "learning_rate": 1.0923572710294062e-07, + "loss": 0.6876, + "step": 24835 + }, + { + "epoch": 0.91, + "learning_rate": 1.0923206503826856e-07, + "loss": 0.7017, + "step": 24836 + }, + { + "epoch": 0.91, + "learning_rate": 1.092284029735965e-07, + "loss": 0.6916, + "step": 24837 + }, + { + "epoch": 0.91, + "learning_rate": 1.0922474090892444e-07, + "loss": 0.6943, + "step": 24838 + }, + { + "epoch": 0.91, + "learning_rate": 1.0922107884425238e-07, + "loss": 0.7003, + "step": 24839 + }, + { + "epoch": 0.91, + "learning_rate": 1.0921741677958032e-07, + "loss": 0.7034, + "step": 24840 + }, + { + "epoch": 0.91, + "learning_rate": 1.0921375471490826e-07, + "loss": 0.6898, + "step": 24841 + }, + { + "epoch": 0.91, + "learning_rate": 1.092100926502362e-07, + "loss": 0.6953, + "step": 24842 + }, + { + "epoch": 0.91, + "learning_rate": 1.0920643058556414e-07, + "loss": 0.6932, + "step": 24843 + }, + { + "epoch": 0.91, + "learning_rate": 1.0920276852089208e-07, + "loss": 0.6963, + "step": 24844 + }, + { + "epoch": 0.91, + "learning_rate": 1.0919910645622002e-07, + "loss": 0.7195, + "step": 24845 + }, + { + "epoch": 0.91, + "learning_rate": 1.0919544439154796e-07, + "loss": 0.6804, + "step": 24846 + }, + { + "epoch": 0.91, + "learning_rate": 1.091917823268759e-07, + "loss": 0.6794, + "step": 24847 + }, + { + "epoch": 0.91, + "learning_rate": 1.0918812026220384e-07, + "loss": 0.691, + "step": 24848 + }, + { + "epoch": 0.91, + "learning_rate": 1.0918445819753178e-07, + "loss": 0.6996, + "step": 24849 + }, + { + "epoch": 0.91, + "learning_rate": 1.091807961328597e-07, + "loss": 0.6877, + "step": 24850 + }, + { + "epoch": 0.91, + "learning_rate": 1.0917713406818764e-07, + "loss": 0.6905, + "step": 24851 + }, + { + "epoch": 0.91, + "learning_rate": 1.0917347200351558e-07, + "loss": 0.6968, + "step": 24852 + }, + { + "epoch": 0.91, + "learning_rate": 1.0916980993884351e-07, + "loss": 0.6774, + "step": 24853 + }, + { + "epoch": 0.91, + "learning_rate": 1.0916614787417145e-07, + "loss": 0.6898, + "step": 24854 + }, + { + "epoch": 0.91, + "learning_rate": 1.0916248580949938e-07, + "loss": 0.6982, + "step": 24855 + }, + { + "epoch": 0.91, + "learning_rate": 1.0915882374482732e-07, + "loss": 0.6763, + "step": 24856 + }, + { + "epoch": 0.91, + "learning_rate": 1.0915516168015526e-07, + "loss": 0.7053, + "step": 24857 + }, + { + "epoch": 0.91, + "learning_rate": 1.091514996154832e-07, + "loss": 0.6828, + "step": 24858 + }, + { + "epoch": 0.91, + "learning_rate": 1.0914783755081114e-07, + "loss": 0.6973, + "step": 24859 + }, + { + "epoch": 0.91, + "learning_rate": 1.0914417548613908e-07, + "loss": 0.6908, + "step": 24860 + }, + { + "epoch": 0.91, + "learning_rate": 1.0914051342146702e-07, + "loss": 0.6971, + "step": 24861 + }, + { + "epoch": 0.91, + "learning_rate": 1.0913685135679496e-07, + "loss": 0.6962, + "step": 24862 + }, + { + "epoch": 0.91, + "learning_rate": 1.091331892921229e-07, + "loss": 0.6896, + "step": 24863 + }, + { + "epoch": 0.91, + "learning_rate": 1.0912952722745084e-07, + "loss": 0.6975, + "step": 24864 + }, + { + "epoch": 0.91, + "learning_rate": 1.0912586516277878e-07, + "loss": 0.6901, + "step": 24865 + }, + { + "epoch": 0.91, + "learning_rate": 1.0912220309810672e-07, + "loss": 0.7079, + "step": 24866 + }, + { + "epoch": 0.91, + "learning_rate": 1.0911854103343466e-07, + "loss": 0.688, + "step": 24867 + }, + { + "epoch": 0.91, + "learning_rate": 1.091148789687626e-07, + "loss": 0.6853, + "step": 24868 + }, + { + "epoch": 0.91, + "learning_rate": 1.0911121690409054e-07, + "loss": 0.6985, + "step": 24869 + }, + { + "epoch": 0.91, + "learning_rate": 1.0910755483941848e-07, + "loss": 0.6932, + "step": 24870 + }, + { + "epoch": 0.91, + "learning_rate": 1.0910389277474639e-07, + "loss": 0.6979, + "step": 24871 + }, + { + "epoch": 0.91, + "learning_rate": 1.0910023071007433e-07, + "loss": 0.6964, + "step": 24872 + }, + { + "epoch": 0.91, + "learning_rate": 1.0909656864540227e-07, + "loss": 0.6943, + "step": 24873 + }, + { + "epoch": 0.91, + "learning_rate": 1.0909290658073021e-07, + "loss": 0.6964, + "step": 24874 + }, + { + "epoch": 0.91, + "learning_rate": 1.0908924451605815e-07, + "loss": 0.6956, + "step": 24875 + }, + { + "epoch": 0.91, + "learning_rate": 1.0908558245138609e-07, + "loss": 0.6877, + "step": 24876 + }, + { + "epoch": 0.91, + "learning_rate": 1.0908192038671401e-07, + "loss": 0.7007, + "step": 24877 + }, + { + "epoch": 0.91, + "learning_rate": 1.0907825832204195e-07, + "loss": 0.6899, + "step": 24878 + }, + { + "epoch": 0.91, + "learning_rate": 1.090745962573699e-07, + "loss": 0.6942, + "step": 24879 + }, + { + "epoch": 0.91, + "learning_rate": 1.0907093419269783e-07, + "loss": 0.6958, + "step": 24880 + }, + { + "epoch": 0.91, + "learning_rate": 1.0906727212802577e-07, + "loss": 0.7025, + "step": 24881 + }, + { + "epoch": 0.91, + "learning_rate": 1.0906361006335371e-07, + "loss": 0.6879, + "step": 24882 + }, + { + "epoch": 0.91, + "learning_rate": 1.0905994799868165e-07, + "loss": 0.6895, + "step": 24883 + }, + { + "epoch": 0.91, + "learning_rate": 1.090562859340096e-07, + "loss": 0.6951, + "step": 24884 + }, + { + "epoch": 0.91, + "learning_rate": 1.0905262386933753e-07, + "loss": 0.6889, + "step": 24885 + }, + { + "epoch": 0.91, + "learning_rate": 1.0904896180466547e-07, + "loss": 0.6901, + "step": 24886 + }, + { + "epoch": 0.91, + "learning_rate": 1.0904529973999341e-07, + "loss": 0.6947, + "step": 24887 + }, + { + "epoch": 0.91, + "learning_rate": 1.0904163767532135e-07, + "loss": 0.7013, + "step": 24888 + }, + { + "epoch": 0.91, + "learning_rate": 1.0903797561064927e-07, + "loss": 0.7026, + "step": 24889 + }, + { + "epoch": 0.91, + "learning_rate": 1.0903431354597721e-07, + "loss": 0.7017, + "step": 24890 + }, + { + "epoch": 0.91, + "learning_rate": 1.0903065148130515e-07, + "loss": 0.6932, + "step": 24891 + }, + { + "epoch": 0.91, + "learning_rate": 1.0902698941663309e-07, + "loss": 0.6926, + "step": 24892 + }, + { + "epoch": 0.91, + "learning_rate": 1.0902332735196103e-07, + "loss": 0.6884, + "step": 24893 + }, + { + "epoch": 0.91, + "learning_rate": 1.0901966528728897e-07, + "loss": 0.6913, + "step": 24894 + }, + { + "epoch": 0.91, + "learning_rate": 1.0901600322261691e-07, + "loss": 0.6892, + "step": 24895 + }, + { + "epoch": 0.91, + "learning_rate": 1.0901234115794485e-07, + "loss": 0.6923, + "step": 24896 + }, + { + "epoch": 0.91, + "learning_rate": 1.0900867909327279e-07, + "loss": 0.6836, + "step": 24897 + }, + { + "epoch": 0.91, + "learning_rate": 1.0900501702860073e-07, + "loss": 0.6875, + "step": 24898 + }, + { + "epoch": 0.91, + "learning_rate": 1.0900135496392865e-07, + "loss": 0.6966, + "step": 24899 + }, + { + "epoch": 0.91, + "learning_rate": 1.089976928992566e-07, + "loss": 0.6996, + "step": 24900 + }, + { + "epoch": 0.91, + "learning_rate": 1.0899403083458453e-07, + "loss": 0.6957, + "step": 24901 + }, + { + "epoch": 0.91, + "learning_rate": 1.0899036876991247e-07, + "loss": 0.6958, + "step": 24902 + }, + { + "epoch": 0.91, + "learning_rate": 1.0898670670524041e-07, + "loss": 0.682, + "step": 24903 + }, + { + "epoch": 0.91, + "learning_rate": 1.0898304464056835e-07, + "loss": 0.6982, + "step": 24904 + }, + { + "epoch": 0.91, + "learning_rate": 1.0897938257589629e-07, + "loss": 0.69, + "step": 24905 + }, + { + "epoch": 0.91, + "learning_rate": 1.0897572051122423e-07, + "loss": 0.6911, + "step": 24906 + }, + { + "epoch": 0.91, + "learning_rate": 1.0897205844655215e-07, + "loss": 0.6865, + "step": 24907 + }, + { + "epoch": 0.91, + "learning_rate": 1.0896839638188009e-07, + "loss": 0.6954, + "step": 24908 + }, + { + "epoch": 0.91, + "learning_rate": 1.0896473431720803e-07, + "loss": 0.6873, + "step": 24909 + }, + { + "epoch": 0.91, + "learning_rate": 1.0896107225253597e-07, + "loss": 0.7004, + "step": 24910 + }, + { + "epoch": 0.91, + "learning_rate": 1.089574101878639e-07, + "loss": 0.6933, + "step": 24911 + }, + { + "epoch": 0.91, + "learning_rate": 1.0895374812319185e-07, + "loss": 0.6907, + "step": 24912 + }, + { + "epoch": 0.91, + "learning_rate": 1.0895008605851979e-07, + "loss": 0.6863, + "step": 24913 + }, + { + "epoch": 0.91, + "learning_rate": 1.0894642399384773e-07, + "loss": 0.6753, + "step": 24914 + }, + { + "epoch": 0.91, + "learning_rate": 1.0894276192917567e-07, + "loss": 0.6969, + "step": 24915 + }, + { + "epoch": 0.91, + "learning_rate": 1.089390998645036e-07, + "loss": 0.6766, + "step": 24916 + }, + { + "epoch": 0.91, + "learning_rate": 1.0893543779983155e-07, + "loss": 0.6934, + "step": 24917 + }, + { + "epoch": 0.91, + "learning_rate": 1.0893177573515949e-07, + "loss": 0.6937, + "step": 24918 + }, + { + "epoch": 0.91, + "learning_rate": 1.0892811367048743e-07, + "loss": 0.7032, + "step": 24919 + }, + { + "epoch": 0.91, + "learning_rate": 1.0892445160581537e-07, + "loss": 0.6925, + "step": 24920 + }, + { + "epoch": 0.91, + "learning_rate": 1.0892078954114329e-07, + "loss": 0.6925, + "step": 24921 + }, + { + "epoch": 0.91, + "learning_rate": 1.0891712747647123e-07, + "loss": 0.7044, + "step": 24922 + }, + { + "epoch": 0.91, + "learning_rate": 1.0891346541179917e-07, + "loss": 0.6969, + "step": 24923 + }, + { + "epoch": 0.91, + "learning_rate": 1.0890980334712711e-07, + "loss": 0.6883, + "step": 24924 + }, + { + "epoch": 0.91, + "learning_rate": 1.0890614128245504e-07, + "loss": 0.7027, + "step": 24925 + }, + { + "epoch": 0.91, + "learning_rate": 1.0890247921778297e-07, + "loss": 0.7079, + "step": 24926 + }, + { + "epoch": 0.91, + "learning_rate": 1.088988171531109e-07, + "loss": 0.7027, + "step": 24927 + }, + { + "epoch": 0.91, + "learning_rate": 1.0889515508843885e-07, + "loss": 0.689, + "step": 24928 + }, + { + "epoch": 0.91, + "learning_rate": 1.0889149302376679e-07, + "loss": 0.7016, + "step": 24929 + }, + { + "epoch": 0.91, + "learning_rate": 1.0888783095909473e-07, + "loss": 0.6919, + "step": 24930 + }, + { + "epoch": 0.91, + "learning_rate": 1.0888416889442267e-07, + "loss": 0.6863, + "step": 24931 + }, + { + "epoch": 0.91, + "learning_rate": 1.088805068297506e-07, + "loss": 0.6875, + "step": 24932 + }, + { + "epoch": 0.91, + "learning_rate": 1.0887684476507855e-07, + "loss": 0.6843, + "step": 24933 + }, + { + "epoch": 0.91, + "learning_rate": 1.0887318270040649e-07, + "loss": 0.7014, + "step": 24934 + }, + { + "epoch": 0.91, + "learning_rate": 1.0886952063573442e-07, + "loss": 0.7077, + "step": 24935 + }, + { + "epoch": 0.91, + "learning_rate": 1.0886585857106236e-07, + "loss": 0.6937, + "step": 24936 + }, + { + "epoch": 0.91, + "learning_rate": 1.088621965063903e-07, + "loss": 0.6875, + "step": 24937 + }, + { + "epoch": 0.91, + "learning_rate": 1.0885853444171824e-07, + "loss": 0.6941, + "step": 24938 + }, + { + "epoch": 0.91, + "learning_rate": 1.0885487237704618e-07, + "loss": 0.7082, + "step": 24939 + }, + { + "epoch": 0.91, + "learning_rate": 1.0885121031237412e-07, + "loss": 0.6932, + "step": 24940 + }, + { + "epoch": 0.91, + "learning_rate": 1.0884754824770206e-07, + "loss": 0.6981, + "step": 24941 + }, + { + "epoch": 0.91, + "learning_rate": 1.0884388618303e-07, + "loss": 0.6943, + "step": 24942 + }, + { + "epoch": 0.91, + "learning_rate": 1.0884022411835792e-07, + "loss": 0.6805, + "step": 24943 + }, + { + "epoch": 0.91, + "learning_rate": 1.0883656205368586e-07, + "loss": 0.678, + "step": 24944 + }, + { + "epoch": 0.91, + "learning_rate": 1.088328999890138e-07, + "loss": 0.6921, + "step": 24945 + }, + { + "epoch": 0.91, + "learning_rate": 1.0882923792434174e-07, + "loss": 0.6979, + "step": 24946 + }, + { + "epoch": 0.91, + "learning_rate": 1.0882557585966968e-07, + "loss": 0.6893, + "step": 24947 + }, + { + "epoch": 0.91, + "learning_rate": 1.088219137949976e-07, + "loss": 0.6877, + "step": 24948 + }, + { + "epoch": 0.91, + "learning_rate": 1.0881825173032554e-07, + "loss": 0.6912, + "step": 24949 + }, + { + "epoch": 0.91, + "learning_rate": 1.0881458966565348e-07, + "loss": 0.6915, + "step": 24950 + }, + { + "epoch": 0.91, + "learning_rate": 1.0881092760098142e-07, + "loss": 0.6929, + "step": 24951 + }, + { + "epoch": 0.91, + "learning_rate": 1.0880726553630936e-07, + "loss": 0.6976, + "step": 24952 + }, + { + "epoch": 0.91, + "learning_rate": 1.088036034716373e-07, + "loss": 0.6991, + "step": 24953 + }, + { + "epoch": 0.91, + "learning_rate": 1.0879994140696524e-07, + "loss": 0.693, + "step": 24954 + }, + { + "epoch": 0.91, + "learning_rate": 1.0879627934229318e-07, + "loss": 0.6869, + "step": 24955 + }, + { + "epoch": 0.91, + "learning_rate": 1.0879261727762112e-07, + "loss": 0.6943, + "step": 24956 + }, + { + "epoch": 0.91, + "learning_rate": 1.0878895521294906e-07, + "loss": 0.6979, + "step": 24957 + }, + { + "epoch": 0.91, + "learning_rate": 1.08785293148277e-07, + "loss": 0.6943, + "step": 24958 + }, + { + "epoch": 0.91, + "learning_rate": 1.0878163108360494e-07, + "loss": 0.6924, + "step": 24959 + }, + { + "epoch": 0.91, + "learning_rate": 1.0877796901893288e-07, + "loss": 0.7004, + "step": 24960 + }, + { + "epoch": 0.91, + "learning_rate": 1.087743069542608e-07, + "loss": 0.684, + "step": 24961 + }, + { + "epoch": 0.91, + "learning_rate": 1.0877064488958874e-07, + "loss": 0.7002, + "step": 24962 + }, + { + "epoch": 0.91, + "learning_rate": 1.0876698282491668e-07, + "loss": 0.6998, + "step": 24963 + }, + { + "epoch": 0.91, + "learning_rate": 1.0876332076024462e-07, + "loss": 0.6888, + "step": 24964 + }, + { + "epoch": 0.91, + "learning_rate": 1.0875965869557256e-07, + "loss": 0.6966, + "step": 24965 + }, + { + "epoch": 0.91, + "learning_rate": 1.087559966309005e-07, + "loss": 0.7032, + "step": 24966 + }, + { + "epoch": 0.91, + "learning_rate": 1.0875233456622844e-07, + "loss": 0.6859, + "step": 24967 + }, + { + "epoch": 0.91, + "learning_rate": 1.0874867250155638e-07, + "loss": 0.695, + "step": 24968 + }, + { + "epoch": 0.91, + "learning_rate": 1.0874501043688432e-07, + "loss": 0.6908, + "step": 24969 + }, + { + "epoch": 0.91, + "learning_rate": 1.0874134837221224e-07, + "loss": 0.6812, + "step": 24970 + }, + { + "epoch": 0.91, + "learning_rate": 1.0873768630754018e-07, + "loss": 0.7034, + "step": 24971 + }, + { + "epoch": 0.91, + "learning_rate": 1.0873402424286812e-07, + "loss": 0.6757, + "step": 24972 + }, + { + "epoch": 0.91, + "learning_rate": 1.0873036217819606e-07, + "loss": 0.69, + "step": 24973 + }, + { + "epoch": 0.91, + "learning_rate": 1.08726700113524e-07, + "loss": 0.6953, + "step": 24974 + }, + { + "epoch": 0.91, + "learning_rate": 1.0872303804885194e-07, + "loss": 0.7027, + "step": 24975 + }, + { + "epoch": 0.91, + "learning_rate": 1.0871937598417988e-07, + "loss": 0.6979, + "step": 24976 + }, + { + "epoch": 0.91, + "learning_rate": 1.0871571391950782e-07, + "loss": 0.6894, + "step": 24977 + }, + { + "epoch": 0.91, + "learning_rate": 1.0871205185483576e-07, + "loss": 0.6993, + "step": 24978 + }, + { + "epoch": 0.91, + "learning_rate": 1.0870838979016368e-07, + "loss": 0.6941, + "step": 24979 + }, + { + "epoch": 0.91, + "learning_rate": 1.0870472772549162e-07, + "loss": 0.6909, + "step": 24980 + }, + { + "epoch": 0.91, + "learning_rate": 1.0870106566081956e-07, + "loss": 0.6952, + "step": 24981 + }, + { + "epoch": 0.91, + "learning_rate": 1.086974035961475e-07, + "loss": 0.683, + "step": 24982 + }, + { + "epoch": 0.91, + "learning_rate": 1.0869374153147544e-07, + "loss": 0.701, + "step": 24983 + }, + { + "epoch": 0.91, + "learning_rate": 1.0869007946680338e-07, + "loss": 0.6958, + "step": 24984 + }, + { + "epoch": 0.91, + "learning_rate": 1.0868641740213132e-07, + "loss": 0.6993, + "step": 24985 + }, + { + "epoch": 0.91, + "learning_rate": 1.0868275533745926e-07, + "loss": 0.6943, + "step": 24986 + }, + { + "epoch": 0.91, + "learning_rate": 1.086790932727872e-07, + "loss": 0.6958, + "step": 24987 + }, + { + "epoch": 0.91, + "learning_rate": 1.0867543120811514e-07, + "loss": 0.6908, + "step": 24988 + }, + { + "epoch": 0.91, + "learning_rate": 1.0867176914344308e-07, + "loss": 0.6898, + "step": 24989 + }, + { + "epoch": 0.91, + "learning_rate": 1.0866810707877102e-07, + "loss": 0.6843, + "step": 24990 + }, + { + "epoch": 0.91, + "learning_rate": 1.0866444501409896e-07, + "loss": 0.7059, + "step": 24991 + }, + { + "epoch": 0.91, + "learning_rate": 1.0866078294942688e-07, + "loss": 0.6939, + "step": 24992 + }, + { + "epoch": 0.91, + "learning_rate": 1.0865712088475482e-07, + "loss": 0.6899, + "step": 24993 + }, + { + "epoch": 0.91, + "learning_rate": 1.0865345882008276e-07, + "loss": 0.6966, + "step": 24994 + }, + { + "epoch": 0.91, + "learning_rate": 1.086497967554107e-07, + "loss": 0.6909, + "step": 24995 + }, + { + "epoch": 0.91, + "learning_rate": 1.0864613469073864e-07, + "loss": 0.6872, + "step": 24996 + }, + { + "epoch": 0.91, + "learning_rate": 1.0864247262606655e-07, + "loss": 0.6847, + "step": 24997 + }, + { + "epoch": 0.91, + "learning_rate": 1.086388105613945e-07, + "loss": 0.6959, + "step": 24998 + }, + { + "epoch": 0.91, + "learning_rate": 1.0863514849672243e-07, + "loss": 0.6991, + "step": 24999 + }, + { + "epoch": 0.91, + "learning_rate": 1.0863148643205037e-07, + "loss": 0.6901, + "step": 25000 + }, + { + "epoch": 0.91, + "learning_rate": 1.0862782436737831e-07, + "loss": 0.7154, + "step": 25001 + }, + { + "epoch": 0.91, + "learning_rate": 1.0862416230270625e-07, + "loss": 0.6852, + "step": 25002 + }, + { + "epoch": 0.91, + "learning_rate": 1.086205002380342e-07, + "loss": 0.701, + "step": 25003 + }, + { + "epoch": 0.91, + "learning_rate": 1.0861683817336213e-07, + "loss": 0.6911, + "step": 25004 + }, + { + "epoch": 0.91, + "learning_rate": 1.0861317610869007e-07, + "loss": 0.6925, + "step": 25005 + }, + { + "epoch": 0.91, + "learning_rate": 1.0860951404401801e-07, + "loss": 0.6929, + "step": 25006 + }, + { + "epoch": 0.91, + "learning_rate": 1.0860585197934595e-07, + "loss": 0.6922, + "step": 25007 + }, + { + "epoch": 0.91, + "learning_rate": 1.086021899146739e-07, + "loss": 0.7054, + "step": 25008 + }, + { + "epoch": 0.92, + "learning_rate": 1.0859852785000183e-07, + "loss": 0.6973, + "step": 25009 + }, + { + "epoch": 0.92, + "learning_rate": 1.0859486578532977e-07, + "loss": 0.7045, + "step": 25010 + }, + { + "epoch": 0.92, + "learning_rate": 1.0859120372065771e-07, + "loss": 0.6901, + "step": 25011 + }, + { + "epoch": 0.92, + "learning_rate": 1.0858754165598565e-07, + "loss": 0.7092, + "step": 25012 + }, + { + "epoch": 0.92, + "learning_rate": 1.085838795913136e-07, + "loss": 0.6924, + "step": 25013 + }, + { + "epoch": 0.92, + "learning_rate": 1.0858021752664152e-07, + "loss": 0.688, + "step": 25014 + }, + { + "epoch": 0.92, + "learning_rate": 1.0857655546196945e-07, + "loss": 0.6835, + "step": 25015 + }, + { + "epoch": 0.92, + "learning_rate": 1.0857289339729739e-07, + "loss": 0.6842, + "step": 25016 + }, + { + "epoch": 0.92, + "learning_rate": 1.0856923133262533e-07, + "loss": 0.6977, + "step": 25017 + }, + { + "epoch": 0.92, + "learning_rate": 1.0856556926795327e-07, + "loss": 0.6987, + "step": 25018 + }, + { + "epoch": 0.92, + "learning_rate": 1.085619072032812e-07, + "loss": 0.6885, + "step": 25019 + }, + { + "epoch": 0.92, + "learning_rate": 1.0855824513860913e-07, + "loss": 0.7049, + "step": 25020 + }, + { + "epoch": 0.92, + "learning_rate": 1.0855458307393707e-07, + "loss": 0.6973, + "step": 25021 + }, + { + "epoch": 0.92, + "learning_rate": 1.0855092100926501e-07, + "loss": 0.6898, + "step": 25022 + }, + { + "epoch": 0.92, + "learning_rate": 1.0854725894459295e-07, + "loss": 0.6937, + "step": 25023 + }, + { + "epoch": 0.92, + "learning_rate": 1.0854359687992089e-07, + "loss": 0.6936, + "step": 25024 + }, + { + "epoch": 0.92, + "learning_rate": 1.0853993481524883e-07, + "loss": 0.6846, + "step": 25025 + }, + { + "epoch": 0.92, + "learning_rate": 1.0853627275057677e-07, + "loss": 0.708, + "step": 25026 + }, + { + "epoch": 0.92, + "learning_rate": 1.0853261068590471e-07, + "loss": 0.686, + "step": 25027 + }, + { + "epoch": 0.92, + "learning_rate": 1.0852894862123265e-07, + "loss": 0.7101, + "step": 25028 + }, + { + "epoch": 0.92, + "learning_rate": 1.0852528655656059e-07, + "loss": 0.6845, + "step": 25029 + }, + { + "epoch": 0.92, + "learning_rate": 1.0852162449188853e-07, + "loss": 0.6907, + "step": 25030 + }, + { + "epoch": 0.92, + "learning_rate": 1.0851796242721647e-07, + "loss": 0.6773, + "step": 25031 + }, + { + "epoch": 0.92, + "learning_rate": 1.0851430036254441e-07, + "loss": 0.6866, + "step": 25032 + }, + { + "epoch": 0.92, + "learning_rate": 1.0851063829787233e-07, + "loss": 0.6896, + "step": 25033 + }, + { + "epoch": 0.92, + "learning_rate": 1.0850697623320027e-07, + "loss": 0.6888, + "step": 25034 + }, + { + "epoch": 0.92, + "learning_rate": 1.085033141685282e-07, + "loss": 0.6891, + "step": 25035 + }, + { + "epoch": 0.92, + "learning_rate": 1.0849965210385615e-07, + "loss": 0.6893, + "step": 25036 + }, + { + "epoch": 0.92, + "learning_rate": 1.0849599003918409e-07, + "loss": 0.6877, + "step": 25037 + }, + { + "epoch": 0.92, + "learning_rate": 1.0849232797451203e-07, + "loss": 0.6955, + "step": 25038 + }, + { + "epoch": 0.92, + "learning_rate": 1.0848866590983997e-07, + "loss": 0.7072, + "step": 25039 + }, + { + "epoch": 0.92, + "learning_rate": 1.084850038451679e-07, + "loss": 0.6959, + "step": 25040 + }, + { + "epoch": 0.92, + "learning_rate": 1.0848134178049583e-07, + "loss": 0.6785, + "step": 25041 + }, + { + "epoch": 0.92, + "learning_rate": 1.0847767971582377e-07, + "loss": 0.695, + "step": 25042 + }, + { + "epoch": 0.92, + "learning_rate": 1.0847401765115171e-07, + "loss": 0.6931, + "step": 25043 + }, + { + "epoch": 0.92, + "learning_rate": 1.0847035558647965e-07, + "loss": 0.6919, + "step": 25044 + }, + { + "epoch": 0.92, + "learning_rate": 1.0846669352180759e-07, + "loss": 0.6998, + "step": 25045 + }, + { + "epoch": 0.92, + "learning_rate": 1.0846303145713553e-07, + "loss": 0.7069, + "step": 25046 + }, + { + "epoch": 0.92, + "learning_rate": 1.0845936939246347e-07, + "loss": 0.6905, + "step": 25047 + }, + { + "epoch": 0.92, + "learning_rate": 1.0845570732779141e-07, + "loss": 0.6964, + "step": 25048 + }, + { + "epoch": 0.92, + "learning_rate": 1.0845204526311935e-07, + "loss": 0.6987, + "step": 25049 + }, + { + "epoch": 0.92, + "learning_rate": 1.0844838319844727e-07, + "loss": 0.6964, + "step": 25050 + }, + { + "epoch": 0.92, + "learning_rate": 1.084447211337752e-07, + "loss": 0.6935, + "step": 25051 + }, + { + "epoch": 0.92, + "learning_rate": 1.0844105906910315e-07, + "loss": 0.692, + "step": 25052 + }, + { + "epoch": 0.92, + "learning_rate": 1.0843739700443109e-07, + "loss": 0.6857, + "step": 25053 + }, + { + "epoch": 0.92, + "learning_rate": 1.0843373493975903e-07, + "loss": 0.6925, + "step": 25054 + }, + { + "epoch": 0.92, + "learning_rate": 1.0843007287508696e-07, + "loss": 0.696, + "step": 25055 + }, + { + "epoch": 0.92, + "learning_rate": 1.084264108104149e-07, + "loss": 0.6848, + "step": 25056 + }, + { + "epoch": 0.92, + "learning_rate": 1.0842274874574284e-07, + "loss": 0.6955, + "step": 25057 + }, + { + "epoch": 0.92, + "learning_rate": 1.0841908668107078e-07, + "loss": 0.6843, + "step": 25058 + }, + { + "epoch": 0.92, + "learning_rate": 1.0841542461639872e-07, + "loss": 0.6926, + "step": 25059 + }, + { + "epoch": 0.92, + "learning_rate": 1.0841176255172666e-07, + "loss": 0.6964, + "step": 25060 + }, + { + "epoch": 0.92, + "learning_rate": 1.084081004870546e-07, + "loss": 0.695, + "step": 25061 + }, + { + "epoch": 0.92, + "learning_rate": 1.0840443842238254e-07, + "loss": 0.6964, + "step": 25062 + }, + { + "epoch": 0.92, + "learning_rate": 1.0840077635771047e-07, + "loss": 0.6959, + "step": 25063 + }, + { + "epoch": 0.92, + "learning_rate": 1.0839711429303841e-07, + "loss": 0.6956, + "step": 25064 + }, + { + "epoch": 0.92, + "learning_rate": 1.0839345222836635e-07, + "loss": 0.6919, + "step": 25065 + }, + { + "epoch": 0.92, + "learning_rate": 1.0838979016369429e-07, + "loss": 0.6906, + "step": 25066 + }, + { + "epoch": 0.92, + "learning_rate": 1.0838612809902223e-07, + "loss": 0.6927, + "step": 25067 + }, + { + "epoch": 0.92, + "learning_rate": 1.0838246603435014e-07, + "loss": 0.7054, + "step": 25068 + }, + { + "epoch": 0.92, + "learning_rate": 1.0837880396967808e-07, + "loss": 0.6959, + "step": 25069 + }, + { + "epoch": 0.92, + "learning_rate": 1.0837514190500602e-07, + "loss": 0.6996, + "step": 25070 + }, + { + "epoch": 0.92, + "learning_rate": 1.0837147984033396e-07, + "loss": 0.7065, + "step": 25071 + }, + { + "epoch": 0.92, + "learning_rate": 1.083678177756619e-07, + "loss": 0.6975, + "step": 25072 + }, + { + "epoch": 0.92, + "learning_rate": 1.0836415571098984e-07, + "loss": 0.6985, + "step": 25073 + }, + { + "epoch": 0.92, + "learning_rate": 1.0836049364631778e-07, + "loss": 0.7007, + "step": 25074 + }, + { + "epoch": 0.92, + "learning_rate": 1.0835683158164572e-07, + "loss": 0.6982, + "step": 25075 + }, + { + "epoch": 0.92, + "learning_rate": 1.0835316951697366e-07, + "loss": 0.6894, + "step": 25076 + }, + { + "epoch": 0.92, + "learning_rate": 1.083495074523016e-07, + "loss": 0.6923, + "step": 25077 + }, + { + "epoch": 0.92, + "learning_rate": 1.0834584538762954e-07, + "loss": 0.6962, + "step": 25078 + }, + { + "epoch": 0.92, + "learning_rate": 1.0834218332295748e-07, + "loss": 0.7058, + "step": 25079 + }, + { + "epoch": 0.92, + "learning_rate": 1.0833852125828542e-07, + "loss": 0.6975, + "step": 25080 + }, + { + "epoch": 0.92, + "learning_rate": 1.0833485919361336e-07, + "loss": 0.6972, + "step": 25081 + }, + { + "epoch": 0.92, + "learning_rate": 1.083311971289413e-07, + "loss": 0.6865, + "step": 25082 + }, + { + "epoch": 0.92, + "learning_rate": 1.0832753506426924e-07, + "loss": 0.6987, + "step": 25083 + }, + { + "epoch": 0.92, + "learning_rate": 1.0832387299959718e-07, + "loss": 0.6854, + "step": 25084 + }, + { + "epoch": 0.92, + "learning_rate": 1.0832021093492511e-07, + "loss": 0.6877, + "step": 25085 + }, + { + "epoch": 0.92, + "learning_rate": 1.0831654887025304e-07, + "loss": 0.6916, + "step": 25086 + }, + { + "epoch": 0.92, + "learning_rate": 1.0831288680558098e-07, + "loss": 0.7025, + "step": 25087 + }, + { + "epoch": 0.92, + "learning_rate": 1.0830922474090892e-07, + "loss": 0.6969, + "step": 25088 + }, + { + "epoch": 0.92, + "learning_rate": 1.0830556267623686e-07, + "loss": 0.6963, + "step": 25089 + }, + { + "epoch": 0.92, + "learning_rate": 1.0830190061156478e-07, + "loss": 0.6951, + "step": 25090 + }, + { + "epoch": 0.92, + "learning_rate": 1.0829823854689272e-07, + "loss": 0.686, + "step": 25091 + }, + { + "epoch": 0.92, + "learning_rate": 1.0829457648222066e-07, + "loss": 0.701, + "step": 25092 + }, + { + "epoch": 0.92, + "learning_rate": 1.082909144175486e-07, + "loss": 0.7015, + "step": 25093 + }, + { + "epoch": 0.92, + "learning_rate": 1.0828725235287654e-07, + "loss": 0.6995, + "step": 25094 + }, + { + "epoch": 0.92, + "learning_rate": 1.0828359028820448e-07, + "loss": 0.7031, + "step": 25095 + }, + { + "epoch": 0.92, + "learning_rate": 1.0827992822353242e-07, + "loss": 0.7013, + "step": 25096 + }, + { + "epoch": 0.92, + "learning_rate": 1.0827626615886036e-07, + "loss": 0.6889, + "step": 25097 + }, + { + "epoch": 0.92, + "learning_rate": 1.082726040941883e-07, + "loss": 0.6916, + "step": 25098 + }, + { + "epoch": 0.92, + "learning_rate": 1.0826894202951624e-07, + "loss": 0.6893, + "step": 25099 + }, + { + "epoch": 0.92, + "learning_rate": 1.0826527996484418e-07, + "loss": 0.6852, + "step": 25100 + }, + { + "epoch": 0.92, + "learning_rate": 1.0826161790017212e-07, + "loss": 0.6939, + "step": 25101 + }, + { + "epoch": 0.92, + "learning_rate": 1.0825795583550006e-07, + "loss": 0.6926, + "step": 25102 + }, + { + "epoch": 0.92, + "learning_rate": 1.08254293770828e-07, + "loss": 0.6964, + "step": 25103 + }, + { + "epoch": 0.92, + "learning_rate": 1.0825063170615592e-07, + "loss": 0.7036, + "step": 25104 + }, + { + "epoch": 0.92, + "learning_rate": 1.0824696964148386e-07, + "loss": 0.6983, + "step": 25105 + }, + { + "epoch": 0.92, + "learning_rate": 1.082433075768118e-07, + "loss": 0.6939, + "step": 25106 + }, + { + "epoch": 0.92, + "learning_rate": 1.0823964551213974e-07, + "loss": 0.6931, + "step": 25107 + }, + { + "epoch": 0.92, + "learning_rate": 1.0823598344746768e-07, + "loss": 0.7003, + "step": 25108 + }, + { + "epoch": 0.92, + "learning_rate": 1.0823232138279562e-07, + "loss": 0.6998, + "step": 25109 + }, + { + "epoch": 0.92, + "learning_rate": 1.0822865931812356e-07, + "loss": 0.6914, + "step": 25110 + }, + { + "epoch": 0.92, + "learning_rate": 1.082249972534515e-07, + "loss": 0.7083, + "step": 25111 + }, + { + "epoch": 0.92, + "learning_rate": 1.0822133518877942e-07, + "loss": 0.688, + "step": 25112 + }, + { + "epoch": 0.92, + "learning_rate": 1.0821767312410736e-07, + "loss": 0.7013, + "step": 25113 + }, + { + "epoch": 0.92, + "learning_rate": 1.082140110594353e-07, + "loss": 0.7051, + "step": 25114 + }, + { + "epoch": 0.92, + "learning_rate": 1.0821034899476324e-07, + "loss": 0.6964, + "step": 25115 + }, + { + "epoch": 0.92, + "learning_rate": 1.0820668693009118e-07, + "loss": 0.6967, + "step": 25116 + }, + { + "epoch": 0.92, + "learning_rate": 1.0820302486541912e-07, + "loss": 0.6944, + "step": 25117 + }, + { + "epoch": 0.92, + "learning_rate": 1.0819936280074706e-07, + "loss": 0.6932, + "step": 25118 + }, + { + "epoch": 0.92, + "learning_rate": 1.08195700736075e-07, + "loss": 0.6917, + "step": 25119 + }, + { + "epoch": 0.92, + "learning_rate": 1.0819203867140294e-07, + "loss": 0.6952, + "step": 25120 + }, + { + "epoch": 0.92, + "learning_rate": 1.0818837660673088e-07, + "loss": 0.6951, + "step": 25121 + }, + { + "epoch": 0.92, + "learning_rate": 1.081847145420588e-07, + "loss": 0.6958, + "step": 25122 + }, + { + "epoch": 0.92, + "learning_rate": 1.0818105247738673e-07, + "loss": 0.6926, + "step": 25123 + }, + { + "epoch": 0.92, + "learning_rate": 1.0817739041271467e-07, + "loss": 0.6972, + "step": 25124 + }, + { + "epoch": 0.92, + "learning_rate": 1.0817372834804261e-07, + "loss": 0.6916, + "step": 25125 + }, + { + "epoch": 0.92, + "learning_rate": 1.0817006628337055e-07, + "loss": 0.6898, + "step": 25126 + }, + { + "epoch": 0.92, + "learning_rate": 1.081664042186985e-07, + "loss": 0.7067, + "step": 25127 + }, + { + "epoch": 0.92, + "learning_rate": 1.0816274215402643e-07, + "loss": 0.7037, + "step": 25128 + }, + { + "epoch": 0.92, + "learning_rate": 1.0815908008935437e-07, + "loss": 0.6873, + "step": 25129 + }, + { + "epoch": 0.92, + "learning_rate": 1.0815541802468231e-07, + "loss": 0.6987, + "step": 25130 + }, + { + "epoch": 0.92, + "learning_rate": 1.0815175596001025e-07, + "loss": 0.6911, + "step": 25131 + }, + { + "epoch": 0.92, + "learning_rate": 1.081480938953382e-07, + "loss": 0.6938, + "step": 25132 + }, + { + "epoch": 0.92, + "learning_rate": 1.0814443183066613e-07, + "loss": 0.6917, + "step": 25133 + }, + { + "epoch": 0.92, + "learning_rate": 1.0814076976599406e-07, + "loss": 0.6894, + "step": 25134 + }, + { + "epoch": 0.92, + "learning_rate": 1.08137107701322e-07, + "loss": 0.6938, + "step": 25135 + }, + { + "epoch": 0.92, + "learning_rate": 1.0813344563664994e-07, + "loss": 0.692, + "step": 25136 + }, + { + "epoch": 0.92, + "learning_rate": 1.0812978357197788e-07, + "loss": 0.6848, + "step": 25137 + }, + { + "epoch": 0.92, + "learning_rate": 1.0812612150730582e-07, + "loss": 0.6937, + "step": 25138 + }, + { + "epoch": 0.92, + "learning_rate": 1.0812245944263376e-07, + "loss": 0.7021, + "step": 25139 + }, + { + "epoch": 0.92, + "learning_rate": 1.0811879737796167e-07, + "loss": 0.6993, + "step": 25140 + }, + { + "epoch": 0.92, + "learning_rate": 1.0811513531328961e-07, + "loss": 0.6936, + "step": 25141 + }, + { + "epoch": 0.92, + "learning_rate": 1.0811147324861755e-07, + "loss": 0.694, + "step": 25142 + }, + { + "epoch": 0.92, + "learning_rate": 1.0810781118394549e-07, + "loss": 0.698, + "step": 25143 + }, + { + "epoch": 0.92, + "learning_rate": 1.0810414911927343e-07, + "loss": 0.695, + "step": 25144 + }, + { + "epoch": 0.92, + "learning_rate": 1.0810048705460137e-07, + "loss": 0.6928, + "step": 25145 + }, + { + "epoch": 0.92, + "learning_rate": 1.0809682498992931e-07, + "loss": 0.6877, + "step": 25146 + }, + { + "epoch": 0.92, + "learning_rate": 1.0809316292525725e-07, + "loss": 0.6943, + "step": 25147 + }, + { + "epoch": 0.92, + "learning_rate": 1.0808950086058519e-07, + "loss": 0.7015, + "step": 25148 + }, + { + "epoch": 0.92, + "learning_rate": 1.0808583879591313e-07, + "loss": 0.6931, + "step": 25149 + }, + { + "epoch": 0.92, + "learning_rate": 1.0808217673124107e-07, + "loss": 0.6985, + "step": 25150 + }, + { + "epoch": 0.92, + "learning_rate": 1.0807851466656901e-07, + "loss": 0.7011, + "step": 25151 + }, + { + "epoch": 0.92, + "learning_rate": 1.0807485260189695e-07, + "loss": 0.6962, + "step": 25152 + }, + { + "epoch": 0.92, + "learning_rate": 1.0807119053722489e-07, + "loss": 0.7005, + "step": 25153 + }, + { + "epoch": 0.92, + "learning_rate": 1.0806752847255283e-07, + "loss": 0.7082, + "step": 25154 + }, + { + "epoch": 0.92, + "learning_rate": 1.0806386640788077e-07, + "loss": 0.6996, + "step": 25155 + }, + { + "epoch": 0.92, + "learning_rate": 1.080602043432087e-07, + "loss": 0.6915, + "step": 25156 + }, + { + "epoch": 0.92, + "learning_rate": 1.0805654227853664e-07, + "loss": 0.6906, + "step": 25157 + }, + { + "epoch": 0.92, + "learning_rate": 1.0805288021386457e-07, + "loss": 0.6913, + "step": 25158 + }, + { + "epoch": 0.92, + "learning_rate": 1.080492181491925e-07, + "loss": 0.6928, + "step": 25159 + }, + { + "epoch": 0.92, + "learning_rate": 1.0804555608452045e-07, + "loss": 0.6881, + "step": 25160 + }, + { + "epoch": 0.92, + "learning_rate": 1.0804189401984837e-07, + "loss": 0.6932, + "step": 25161 + }, + { + "epoch": 0.92, + "learning_rate": 1.0803823195517631e-07, + "loss": 0.6808, + "step": 25162 + }, + { + "epoch": 0.92, + "learning_rate": 1.0803456989050425e-07, + "loss": 0.7109, + "step": 25163 + }, + { + "epoch": 0.92, + "learning_rate": 1.0803090782583219e-07, + "loss": 0.6928, + "step": 25164 + }, + { + "epoch": 0.92, + "learning_rate": 1.0802724576116013e-07, + "loss": 0.6898, + "step": 25165 + }, + { + "epoch": 0.92, + "learning_rate": 1.0802358369648807e-07, + "loss": 0.6861, + "step": 25166 + }, + { + "epoch": 0.92, + "learning_rate": 1.0801992163181601e-07, + "loss": 0.6877, + "step": 25167 + }, + { + "epoch": 0.92, + "learning_rate": 1.0801625956714395e-07, + "loss": 0.6893, + "step": 25168 + }, + { + "epoch": 0.92, + "learning_rate": 1.0801259750247189e-07, + "loss": 0.6889, + "step": 25169 + }, + { + "epoch": 0.92, + "learning_rate": 1.0800893543779983e-07, + "loss": 0.7037, + "step": 25170 + }, + { + "epoch": 0.92, + "learning_rate": 1.0800527337312777e-07, + "loss": 0.6954, + "step": 25171 + }, + { + "epoch": 0.92, + "learning_rate": 1.0800161130845571e-07, + "loss": 0.6784, + "step": 25172 + }, + { + "epoch": 0.92, + "learning_rate": 1.0799794924378365e-07, + "loss": 0.6908, + "step": 25173 + }, + { + "epoch": 0.92, + "learning_rate": 1.0799428717911159e-07, + "loss": 0.6843, + "step": 25174 + }, + { + "epoch": 0.92, + "learning_rate": 1.0799062511443953e-07, + "loss": 0.6947, + "step": 25175 + }, + { + "epoch": 0.92, + "learning_rate": 1.0798696304976744e-07, + "loss": 0.6957, + "step": 25176 + }, + { + "epoch": 0.92, + "learning_rate": 1.0798330098509538e-07, + "loss": 0.6863, + "step": 25177 + }, + { + "epoch": 0.92, + "learning_rate": 1.0797963892042332e-07, + "loss": 0.6826, + "step": 25178 + }, + { + "epoch": 0.92, + "learning_rate": 1.0797597685575126e-07, + "loss": 0.7035, + "step": 25179 + }, + { + "epoch": 0.92, + "learning_rate": 1.079723147910792e-07, + "loss": 0.6789, + "step": 25180 + }, + { + "epoch": 0.92, + "learning_rate": 1.0796865272640714e-07, + "loss": 0.6928, + "step": 25181 + }, + { + "epoch": 0.92, + "learning_rate": 1.0796499066173508e-07, + "loss": 0.7068, + "step": 25182 + }, + { + "epoch": 0.92, + "learning_rate": 1.0796132859706301e-07, + "loss": 0.6986, + "step": 25183 + }, + { + "epoch": 0.92, + "learning_rate": 1.0795766653239095e-07, + "loss": 0.707, + "step": 25184 + }, + { + "epoch": 0.92, + "learning_rate": 1.0795400446771889e-07, + "loss": 0.6946, + "step": 25185 + }, + { + "epoch": 0.92, + "learning_rate": 1.0795034240304683e-07, + "loss": 0.6939, + "step": 25186 + }, + { + "epoch": 0.92, + "learning_rate": 1.0794668033837477e-07, + "loss": 0.6823, + "step": 25187 + }, + { + "epoch": 0.92, + "learning_rate": 1.0794301827370271e-07, + "loss": 0.694, + "step": 25188 + }, + { + "epoch": 0.92, + "learning_rate": 1.0793935620903065e-07, + "loss": 0.6893, + "step": 25189 + }, + { + "epoch": 0.92, + "learning_rate": 1.0793569414435859e-07, + "loss": 0.6955, + "step": 25190 + }, + { + "epoch": 0.92, + "learning_rate": 1.0793203207968653e-07, + "loss": 0.6928, + "step": 25191 + }, + { + "epoch": 0.92, + "learning_rate": 1.0792837001501447e-07, + "loss": 0.6913, + "step": 25192 + }, + { + "epoch": 0.92, + "learning_rate": 1.0792470795034241e-07, + "loss": 0.6946, + "step": 25193 + }, + { + "epoch": 0.92, + "learning_rate": 1.0792104588567032e-07, + "loss": 0.6945, + "step": 25194 + }, + { + "epoch": 0.92, + "learning_rate": 1.0791738382099826e-07, + "loss": 0.7007, + "step": 25195 + }, + { + "epoch": 0.92, + "learning_rate": 1.079137217563262e-07, + "loss": 0.6876, + "step": 25196 + }, + { + "epoch": 0.92, + "learning_rate": 1.0791005969165414e-07, + "loss": 0.6863, + "step": 25197 + }, + { + "epoch": 0.92, + "learning_rate": 1.0790639762698208e-07, + "loss": 0.7072, + "step": 25198 + }, + { + "epoch": 0.92, + "learning_rate": 1.0790273556231002e-07, + "loss": 0.6951, + "step": 25199 + }, + { + "epoch": 0.92, + "learning_rate": 1.0789907349763796e-07, + "loss": 0.6887, + "step": 25200 + }, + { + "epoch": 0.92, + "learning_rate": 1.078954114329659e-07, + "loss": 0.694, + "step": 25201 + }, + { + "epoch": 0.92, + "learning_rate": 1.0789174936829384e-07, + "loss": 0.6881, + "step": 25202 + }, + { + "epoch": 0.92, + "learning_rate": 1.0788808730362178e-07, + "loss": 0.6839, + "step": 25203 + }, + { + "epoch": 0.92, + "learning_rate": 1.0788442523894972e-07, + "loss": 0.6844, + "step": 25204 + }, + { + "epoch": 0.92, + "learning_rate": 1.0788076317427765e-07, + "loss": 0.6873, + "step": 25205 + }, + { + "epoch": 0.92, + "learning_rate": 1.0787710110960559e-07, + "loss": 0.6919, + "step": 25206 + }, + { + "epoch": 0.92, + "learning_rate": 1.0787343904493353e-07, + "loss": 0.6883, + "step": 25207 + }, + { + "epoch": 0.92, + "learning_rate": 1.0786977698026147e-07, + "loss": 0.695, + "step": 25208 + }, + { + "epoch": 0.92, + "learning_rate": 1.0786611491558941e-07, + "loss": 0.6835, + "step": 25209 + }, + { + "epoch": 0.92, + "learning_rate": 1.0786245285091735e-07, + "loss": 0.6896, + "step": 25210 + }, + { + "epoch": 0.92, + "learning_rate": 1.0785879078624529e-07, + "loss": 0.6933, + "step": 25211 + }, + { + "epoch": 0.92, + "learning_rate": 1.078551287215732e-07, + "loss": 0.6791, + "step": 25212 + }, + { + "epoch": 0.92, + "learning_rate": 1.0785146665690114e-07, + "loss": 0.6889, + "step": 25213 + }, + { + "epoch": 0.92, + "learning_rate": 1.0784780459222908e-07, + "loss": 0.6891, + "step": 25214 + }, + { + "epoch": 0.92, + "learning_rate": 1.0784414252755702e-07, + "loss": 0.6903, + "step": 25215 + }, + { + "epoch": 0.92, + "learning_rate": 1.0784048046288496e-07, + "loss": 0.6863, + "step": 25216 + }, + { + "epoch": 0.92, + "learning_rate": 1.078368183982129e-07, + "loss": 0.6763, + "step": 25217 + }, + { + "epoch": 0.92, + "learning_rate": 1.0783315633354084e-07, + "loss": 0.6872, + "step": 25218 + }, + { + "epoch": 0.92, + "learning_rate": 1.0782949426886878e-07, + "loss": 0.6968, + "step": 25219 + }, + { + "epoch": 0.92, + "learning_rate": 1.0782583220419672e-07, + "loss": 0.6968, + "step": 25220 + }, + { + "epoch": 0.92, + "learning_rate": 1.0782217013952466e-07, + "loss": 0.6989, + "step": 25221 + }, + { + "epoch": 0.92, + "learning_rate": 1.078185080748526e-07, + "loss": 0.6945, + "step": 25222 + }, + { + "epoch": 0.92, + "learning_rate": 1.0781484601018054e-07, + "loss": 0.7014, + "step": 25223 + }, + { + "epoch": 0.92, + "learning_rate": 1.0781118394550848e-07, + "loss": 0.6873, + "step": 25224 + }, + { + "epoch": 0.92, + "learning_rate": 1.0780752188083642e-07, + "loss": 0.69, + "step": 25225 + }, + { + "epoch": 0.92, + "learning_rate": 1.0780385981616436e-07, + "loss": 0.6966, + "step": 25226 + }, + { + "epoch": 0.92, + "learning_rate": 1.0780019775149229e-07, + "loss": 0.6936, + "step": 25227 + }, + { + "epoch": 0.92, + "learning_rate": 1.0779653568682023e-07, + "loss": 0.6829, + "step": 25228 + }, + { + "epoch": 0.92, + "learning_rate": 1.0779287362214817e-07, + "loss": 0.6947, + "step": 25229 + }, + { + "epoch": 0.92, + "learning_rate": 1.077892115574761e-07, + "loss": 0.6931, + "step": 25230 + }, + { + "epoch": 0.92, + "learning_rate": 1.0778554949280404e-07, + "loss": 0.6925, + "step": 25231 + }, + { + "epoch": 0.92, + "learning_rate": 1.0778188742813196e-07, + "loss": 0.6938, + "step": 25232 + }, + { + "epoch": 0.92, + "learning_rate": 1.077782253634599e-07, + "loss": 0.6926, + "step": 25233 + }, + { + "epoch": 0.92, + "learning_rate": 1.0777456329878784e-07, + "loss": 0.6876, + "step": 25234 + }, + { + "epoch": 0.92, + "learning_rate": 1.0777090123411578e-07, + "loss": 0.6989, + "step": 25235 + }, + { + "epoch": 0.92, + "learning_rate": 1.0776723916944372e-07, + "loss": 0.6958, + "step": 25236 + }, + { + "epoch": 0.92, + "learning_rate": 1.0776357710477166e-07, + "loss": 0.6921, + "step": 25237 + }, + { + "epoch": 0.92, + "learning_rate": 1.077599150400996e-07, + "loss": 0.677, + "step": 25238 + }, + { + "epoch": 0.92, + "learning_rate": 1.0775625297542754e-07, + "loss": 0.7027, + "step": 25239 + }, + { + "epoch": 0.92, + "learning_rate": 1.0775259091075548e-07, + "loss": 0.6934, + "step": 25240 + }, + { + "epoch": 0.92, + "learning_rate": 1.0774892884608342e-07, + "loss": 0.6866, + "step": 25241 + }, + { + "epoch": 0.92, + "learning_rate": 1.0774526678141136e-07, + "loss": 0.6941, + "step": 25242 + }, + { + "epoch": 0.92, + "learning_rate": 1.077416047167393e-07, + "loss": 0.69, + "step": 25243 + }, + { + "epoch": 0.92, + "learning_rate": 1.0773794265206724e-07, + "loss": 0.6928, + "step": 25244 + }, + { + "epoch": 0.92, + "learning_rate": 1.0773428058739518e-07, + "loss": 0.6995, + "step": 25245 + }, + { + "epoch": 0.92, + "learning_rate": 1.0773061852272312e-07, + "loss": 0.6909, + "step": 25246 + }, + { + "epoch": 0.92, + "learning_rate": 1.0772695645805106e-07, + "loss": 0.698, + "step": 25247 + }, + { + "epoch": 0.92, + "learning_rate": 1.0772329439337897e-07, + "loss": 0.6922, + "step": 25248 + }, + { + "epoch": 0.92, + "learning_rate": 1.0771963232870691e-07, + "loss": 0.6845, + "step": 25249 + }, + { + "epoch": 0.92, + "learning_rate": 1.0771597026403485e-07, + "loss": 0.7084, + "step": 25250 + }, + { + "epoch": 0.92, + "learning_rate": 1.077123081993628e-07, + "loss": 0.6959, + "step": 25251 + }, + { + "epoch": 0.92, + "learning_rate": 1.0770864613469073e-07, + "loss": 0.7114, + "step": 25252 + }, + { + "epoch": 0.92, + "learning_rate": 1.0770498407001867e-07, + "loss": 0.6912, + "step": 25253 + }, + { + "epoch": 0.92, + "learning_rate": 1.077013220053466e-07, + "loss": 0.6861, + "step": 25254 + }, + { + "epoch": 0.92, + "learning_rate": 1.0769765994067454e-07, + "loss": 0.6974, + "step": 25255 + }, + { + "epoch": 0.92, + "learning_rate": 1.0769399787600248e-07, + "loss": 0.6819, + "step": 25256 + }, + { + "epoch": 0.92, + "learning_rate": 1.0769033581133042e-07, + "loss": 0.6985, + "step": 25257 + }, + { + "epoch": 0.92, + "learning_rate": 1.0768667374665836e-07, + "loss": 0.6899, + "step": 25258 + }, + { + "epoch": 0.92, + "learning_rate": 1.076830116819863e-07, + "loss": 0.6915, + "step": 25259 + }, + { + "epoch": 0.92, + "learning_rate": 1.0767934961731424e-07, + "loss": 0.6973, + "step": 25260 + }, + { + "epoch": 0.92, + "learning_rate": 1.0767568755264218e-07, + "loss": 0.6849, + "step": 25261 + }, + { + "epoch": 0.92, + "learning_rate": 1.0767202548797012e-07, + "loss": 0.6919, + "step": 25262 + }, + { + "epoch": 0.92, + "learning_rate": 1.0766836342329806e-07, + "loss": 0.6892, + "step": 25263 + }, + { + "epoch": 0.92, + "learning_rate": 1.07664701358626e-07, + "loss": 0.7118, + "step": 25264 + }, + { + "epoch": 0.92, + "learning_rate": 1.0766103929395394e-07, + "loss": 0.6969, + "step": 25265 + }, + { + "epoch": 0.92, + "learning_rate": 1.0765737722928185e-07, + "loss": 0.6816, + "step": 25266 + }, + { + "epoch": 0.92, + "learning_rate": 1.0765371516460979e-07, + "loss": 0.6952, + "step": 25267 + }, + { + "epoch": 0.92, + "learning_rate": 1.0765005309993773e-07, + "loss": 0.7003, + "step": 25268 + }, + { + "epoch": 0.92, + "learning_rate": 1.0764639103526567e-07, + "loss": 0.6907, + "step": 25269 + }, + { + "epoch": 0.92, + "learning_rate": 1.0764272897059361e-07, + "loss": 0.6831, + "step": 25270 + }, + { + "epoch": 0.92, + "learning_rate": 1.0763906690592155e-07, + "loss": 0.6954, + "step": 25271 + }, + { + "epoch": 0.92, + "learning_rate": 1.0763540484124949e-07, + "loss": 0.697, + "step": 25272 + }, + { + "epoch": 0.92, + "learning_rate": 1.0763174277657743e-07, + "loss": 0.7092, + "step": 25273 + }, + { + "epoch": 0.92, + "learning_rate": 1.0762808071190537e-07, + "loss": 0.6935, + "step": 25274 + }, + { + "epoch": 0.92, + "learning_rate": 1.0762441864723331e-07, + "loss": 0.6961, + "step": 25275 + }, + { + "epoch": 0.92, + "learning_rate": 1.0762075658256124e-07, + "loss": 0.6978, + "step": 25276 + }, + { + "epoch": 0.92, + "learning_rate": 1.0761709451788918e-07, + "loss": 0.6995, + "step": 25277 + }, + { + "epoch": 0.92, + "learning_rate": 1.0761343245321712e-07, + "loss": 0.6863, + "step": 25278 + }, + { + "epoch": 0.92, + "learning_rate": 1.0760977038854506e-07, + "loss": 0.6981, + "step": 25279 + }, + { + "epoch": 0.92, + "learning_rate": 1.07606108323873e-07, + "loss": 0.6946, + "step": 25280 + }, + { + "epoch": 0.92, + "learning_rate": 1.0760244625920094e-07, + "loss": 0.6833, + "step": 25281 + }, + { + "epoch": 0.92, + "learning_rate": 1.0759878419452888e-07, + "loss": 0.6955, + "step": 25282 + }, + { + "epoch": 0.93, + "learning_rate": 1.0759512212985682e-07, + "loss": 0.6814, + "step": 25283 + }, + { + "epoch": 0.93, + "learning_rate": 1.0759146006518473e-07, + "loss": 0.6961, + "step": 25284 + }, + { + "epoch": 0.93, + "learning_rate": 1.0758779800051267e-07, + "loss": 0.6919, + "step": 25285 + }, + { + "epoch": 0.93, + "learning_rate": 1.0758413593584061e-07, + "loss": 0.6961, + "step": 25286 + }, + { + "epoch": 0.93, + "learning_rate": 1.0758047387116855e-07, + "loss": 0.6962, + "step": 25287 + }, + { + "epoch": 0.93, + "learning_rate": 1.0757681180649649e-07, + "loss": 0.6937, + "step": 25288 + }, + { + "epoch": 0.93, + "learning_rate": 1.0757314974182443e-07, + "loss": 0.6878, + "step": 25289 + }, + { + "epoch": 0.93, + "learning_rate": 1.0756948767715237e-07, + "loss": 0.6886, + "step": 25290 + }, + { + "epoch": 0.93, + "learning_rate": 1.0756582561248031e-07, + "loss": 0.6862, + "step": 25291 + }, + { + "epoch": 0.93, + "learning_rate": 1.0756216354780825e-07, + "loss": 0.6881, + "step": 25292 + }, + { + "epoch": 0.93, + "learning_rate": 1.0755850148313619e-07, + "loss": 0.6934, + "step": 25293 + }, + { + "epoch": 0.93, + "learning_rate": 1.0755483941846413e-07, + "loss": 0.7028, + "step": 25294 + }, + { + "epoch": 0.93, + "learning_rate": 1.0755117735379207e-07, + "loss": 0.6845, + "step": 25295 + }, + { + "epoch": 0.93, + "learning_rate": 1.0754751528912001e-07, + "loss": 0.6926, + "step": 25296 + }, + { + "epoch": 0.93, + "learning_rate": 1.0754385322444795e-07, + "loss": 0.7008, + "step": 25297 + }, + { + "epoch": 0.93, + "learning_rate": 1.0754019115977588e-07, + "loss": 0.701, + "step": 25298 + }, + { + "epoch": 0.93, + "learning_rate": 1.0753652909510382e-07, + "loss": 0.6977, + "step": 25299 + }, + { + "epoch": 0.93, + "learning_rate": 1.0753286703043176e-07, + "loss": 0.6956, + "step": 25300 + }, + { + "epoch": 0.93, + "learning_rate": 1.075292049657597e-07, + "loss": 0.7007, + "step": 25301 + }, + { + "epoch": 0.93, + "learning_rate": 1.0752554290108762e-07, + "loss": 0.6939, + "step": 25302 + }, + { + "epoch": 0.93, + "learning_rate": 1.0752188083641555e-07, + "loss": 0.7006, + "step": 25303 + }, + { + "epoch": 0.93, + "learning_rate": 1.0751821877174349e-07, + "loss": 0.6829, + "step": 25304 + }, + { + "epoch": 0.93, + "learning_rate": 1.0751455670707143e-07, + "loss": 0.677, + "step": 25305 + }, + { + "epoch": 0.93, + "learning_rate": 1.0751089464239937e-07, + "loss": 0.7083, + "step": 25306 + }, + { + "epoch": 0.93, + "learning_rate": 1.0750723257772731e-07, + "loss": 0.6987, + "step": 25307 + }, + { + "epoch": 0.93, + "learning_rate": 1.0750357051305525e-07, + "loss": 0.6835, + "step": 25308 + }, + { + "epoch": 0.93, + "learning_rate": 1.0749990844838319e-07, + "loss": 0.6957, + "step": 25309 + }, + { + "epoch": 0.93, + "learning_rate": 1.0749624638371113e-07, + "loss": 0.6885, + "step": 25310 + }, + { + "epoch": 0.93, + "learning_rate": 1.0749258431903907e-07, + "loss": 0.7014, + "step": 25311 + }, + { + "epoch": 0.93, + "learning_rate": 1.0748892225436701e-07, + "loss": 0.6814, + "step": 25312 + }, + { + "epoch": 0.93, + "learning_rate": 1.0748526018969495e-07, + "loss": 0.6985, + "step": 25313 + }, + { + "epoch": 0.93, + "learning_rate": 1.0748159812502289e-07, + "loss": 0.7008, + "step": 25314 + }, + { + "epoch": 0.93, + "learning_rate": 1.0747793606035083e-07, + "loss": 0.6863, + "step": 25315 + }, + { + "epoch": 0.93, + "learning_rate": 1.0747427399567877e-07, + "loss": 0.6896, + "step": 25316 + }, + { + "epoch": 0.93, + "learning_rate": 1.0747061193100671e-07, + "loss": 0.6931, + "step": 25317 + }, + { + "epoch": 0.93, + "learning_rate": 1.0746694986633465e-07, + "loss": 0.6917, + "step": 25318 + }, + { + "epoch": 0.93, + "learning_rate": 1.0746328780166258e-07, + "loss": 0.6978, + "step": 25319 + }, + { + "epoch": 0.93, + "learning_rate": 1.074596257369905e-07, + "loss": 0.6815, + "step": 25320 + }, + { + "epoch": 0.93, + "learning_rate": 1.0745596367231844e-07, + "loss": 0.6894, + "step": 25321 + }, + { + "epoch": 0.93, + "learning_rate": 1.0745230160764638e-07, + "loss": 0.6949, + "step": 25322 + }, + { + "epoch": 0.93, + "learning_rate": 1.0744863954297432e-07, + "loss": 0.6995, + "step": 25323 + }, + { + "epoch": 0.93, + "learning_rate": 1.0744497747830226e-07, + "loss": 0.6933, + "step": 25324 + }, + { + "epoch": 0.93, + "learning_rate": 1.0744131541363019e-07, + "loss": 0.6907, + "step": 25325 + }, + { + "epoch": 0.93, + "learning_rate": 1.0743765334895813e-07, + "loss": 0.6952, + "step": 25326 + }, + { + "epoch": 0.93, + "learning_rate": 1.0743399128428607e-07, + "loss": 0.7002, + "step": 25327 + }, + { + "epoch": 0.93, + "learning_rate": 1.0743032921961401e-07, + "loss": 0.6954, + "step": 25328 + }, + { + "epoch": 0.93, + "learning_rate": 1.0742666715494195e-07, + "loss": 0.6936, + "step": 25329 + }, + { + "epoch": 0.93, + "learning_rate": 1.0742300509026989e-07, + "loss": 0.6913, + "step": 25330 + }, + { + "epoch": 0.93, + "learning_rate": 1.0741934302559783e-07, + "loss": 0.7108, + "step": 25331 + }, + { + "epoch": 0.93, + "learning_rate": 1.0741568096092577e-07, + "loss": 0.7002, + "step": 25332 + }, + { + "epoch": 0.93, + "learning_rate": 1.0741201889625371e-07, + "loss": 0.6763, + "step": 25333 + }, + { + "epoch": 0.93, + "learning_rate": 1.0740835683158165e-07, + "loss": 0.7004, + "step": 25334 + }, + { + "epoch": 0.93, + "learning_rate": 1.0740469476690959e-07, + "loss": 0.696, + "step": 25335 + }, + { + "epoch": 0.93, + "learning_rate": 1.0740103270223753e-07, + "loss": 0.6876, + "step": 25336 + }, + { + "epoch": 0.93, + "learning_rate": 1.0739737063756547e-07, + "loss": 0.6921, + "step": 25337 + }, + { + "epoch": 0.93, + "learning_rate": 1.0739370857289338e-07, + "loss": 0.7009, + "step": 25338 + }, + { + "epoch": 0.93, + "learning_rate": 1.0739004650822132e-07, + "loss": 0.6873, + "step": 25339 + }, + { + "epoch": 0.93, + "learning_rate": 1.0738638444354926e-07, + "loss": 0.676, + "step": 25340 + }, + { + "epoch": 0.93, + "learning_rate": 1.073827223788772e-07, + "loss": 0.6957, + "step": 25341 + }, + { + "epoch": 0.93, + "learning_rate": 1.0737906031420514e-07, + "loss": 0.7012, + "step": 25342 + }, + { + "epoch": 0.93, + "learning_rate": 1.0737539824953308e-07, + "loss": 0.6891, + "step": 25343 + }, + { + "epoch": 0.93, + "learning_rate": 1.0737173618486102e-07, + "loss": 0.6909, + "step": 25344 + }, + { + "epoch": 0.93, + "learning_rate": 1.0736807412018896e-07, + "loss": 0.7039, + "step": 25345 + }, + { + "epoch": 0.93, + "learning_rate": 1.073644120555169e-07, + "loss": 0.6873, + "step": 25346 + }, + { + "epoch": 0.93, + "learning_rate": 1.0736074999084483e-07, + "loss": 0.6979, + "step": 25347 + }, + { + "epoch": 0.93, + "learning_rate": 1.0735708792617277e-07, + "loss": 0.6935, + "step": 25348 + }, + { + "epoch": 0.93, + "learning_rate": 1.0735342586150071e-07, + "loss": 0.6884, + "step": 25349 + }, + { + "epoch": 0.93, + "learning_rate": 1.0734976379682865e-07, + "loss": 0.6942, + "step": 25350 + }, + { + "epoch": 0.93, + "learning_rate": 1.0734610173215659e-07, + "loss": 0.684, + "step": 25351 + }, + { + "epoch": 0.93, + "learning_rate": 1.0734243966748453e-07, + "loss": 0.702, + "step": 25352 + }, + { + "epoch": 0.93, + "learning_rate": 1.0733877760281247e-07, + "loss": 0.692, + "step": 25353 + }, + { + "epoch": 0.93, + "learning_rate": 1.0733511553814041e-07, + "loss": 0.6854, + "step": 25354 + }, + { + "epoch": 0.93, + "learning_rate": 1.0733145347346835e-07, + "loss": 0.696, + "step": 25355 + }, + { + "epoch": 0.93, + "learning_rate": 1.0732779140879626e-07, + "loss": 0.6933, + "step": 25356 + }, + { + "epoch": 0.93, + "learning_rate": 1.073241293441242e-07, + "loss": 0.693, + "step": 25357 + }, + { + "epoch": 0.93, + "learning_rate": 1.0732046727945214e-07, + "loss": 0.7026, + "step": 25358 + }, + { + "epoch": 0.93, + "learning_rate": 1.0731680521478008e-07, + "loss": 0.6807, + "step": 25359 + }, + { + "epoch": 0.93, + "learning_rate": 1.0731314315010802e-07, + "loss": 0.6925, + "step": 25360 + }, + { + "epoch": 0.93, + "learning_rate": 1.0730948108543596e-07, + "loss": 0.6957, + "step": 25361 + }, + { + "epoch": 0.93, + "learning_rate": 1.073058190207639e-07, + "loss": 0.6826, + "step": 25362 + }, + { + "epoch": 0.93, + "learning_rate": 1.0730215695609184e-07, + "loss": 0.6847, + "step": 25363 + }, + { + "epoch": 0.93, + "learning_rate": 1.0729849489141978e-07, + "loss": 0.6966, + "step": 25364 + }, + { + "epoch": 0.93, + "learning_rate": 1.0729483282674772e-07, + "loss": 0.6985, + "step": 25365 + }, + { + "epoch": 0.93, + "learning_rate": 1.0729117076207566e-07, + "loss": 0.6863, + "step": 25366 + }, + { + "epoch": 0.93, + "learning_rate": 1.072875086974036e-07, + "loss": 0.6922, + "step": 25367 + }, + { + "epoch": 0.93, + "learning_rate": 1.0728384663273153e-07, + "loss": 0.6964, + "step": 25368 + }, + { + "epoch": 0.93, + "learning_rate": 1.0728018456805947e-07, + "loss": 0.6791, + "step": 25369 + }, + { + "epoch": 0.93, + "learning_rate": 1.0727652250338741e-07, + "loss": 0.7071, + "step": 25370 + }, + { + "epoch": 0.93, + "learning_rate": 1.0727286043871535e-07, + "loss": 0.6898, + "step": 25371 + }, + { + "epoch": 0.93, + "learning_rate": 1.0726919837404329e-07, + "loss": 0.6814, + "step": 25372 + }, + { + "epoch": 0.93, + "learning_rate": 1.0726553630937123e-07, + "loss": 0.6967, + "step": 25373 + }, + { + "epoch": 0.93, + "learning_rate": 1.0726187424469914e-07, + "loss": 0.6981, + "step": 25374 + }, + { + "epoch": 0.93, + "learning_rate": 1.0725821218002708e-07, + "loss": 0.6951, + "step": 25375 + }, + { + "epoch": 0.93, + "learning_rate": 1.0725455011535502e-07, + "loss": 0.6933, + "step": 25376 + }, + { + "epoch": 0.93, + "learning_rate": 1.0725088805068296e-07, + "loss": 0.6819, + "step": 25377 + }, + { + "epoch": 0.93, + "learning_rate": 1.072472259860109e-07, + "loss": 0.703, + "step": 25378 + }, + { + "epoch": 0.93, + "learning_rate": 1.0724356392133884e-07, + "loss": 0.6965, + "step": 25379 + }, + { + "epoch": 0.93, + "learning_rate": 1.0723990185666678e-07, + "loss": 0.6974, + "step": 25380 + }, + { + "epoch": 0.93, + "learning_rate": 1.0723623979199472e-07, + "loss": 0.7033, + "step": 25381 + }, + { + "epoch": 0.93, + "learning_rate": 1.0723257772732266e-07, + "loss": 0.6821, + "step": 25382 + }, + { + "epoch": 0.93, + "learning_rate": 1.072289156626506e-07, + "loss": 0.6923, + "step": 25383 + }, + { + "epoch": 0.93, + "learning_rate": 1.0722525359797854e-07, + "loss": 0.6992, + "step": 25384 + }, + { + "epoch": 0.93, + "learning_rate": 1.0722159153330648e-07, + "loss": 0.7061, + "step": 25385 + }, + { + "epoch": 0.93, + "learning_rate": 1.0721792946863442e-07, + "loss": 0.6967, + "step": 25386 + }, + { + "epoch": 0.93, + "learning_rate": 1.0721426740396236e-07, + "loss": 0.6952, + "step": 25387 + }, + { + "epoch": 0.93, + "learning_rate": 1.072106053392903e-07, + "loss": 0.6942, + "step": 25388 + }, + { + "epoch": 0.93, + "learning_rate": 1.0720694327461824e-07, + "loss": 0.6937, + "step": 25389 + }, + { + "epoch": 0.93, + "learning_rate": 1.0720328120994617e-07, + "loss": 0.6928, + "step": 25390 + }, + { + "epoch": 0.93, + "learning_rate": 1.071996191452741e-07, + "loss": 0.6928, + "step": 25391 + }, + { + "epoch": 0.93, + "learning_rate": 1.0719595708060203e-07, + "loss": 0.7007, + "step": 25392 + }, + { + "epoch": 0.93, + "learning_rate": 1.0719229501592997e-07, + "loss": 0.6987, + "step": 25393 + }, + { + "epoch": 0.93, + "learning_rate": 1.0718863295125791e-07, + "loss": 0.6922, + "step": 25394 + }, + { + "epoch": 0.93, + "learning_rate": 1.0718497088658585e-07, + "loss": 0.6994, + "step": 25395 + }, + { + "epoch": 0.93, + "learning_rate": 1.0718130882191378e-07, + "loss": 0.7009, + "step": 25396 + }, + { + "epoch": 0.93, + "learning_rate": 1.0717764675724172e-07, + "loss": 0.6914, + "step": 25397 + }, + { + "epoch": 0.93, + "learning_rate": 1.0717398469256966e-07, + "loss": 0.681, + "step": 25398 + }, + { + "epoch": 0.93, + "learning_rate": 1.071703226278976e-07, + "loss": 0.6904, + "step": 25399 + }, + { + "epoch": 0.93, + "learning_rate": 1.0716666056322554e-07, + "loss": 0.6774, + "step": 25400 + }, + { + "epoch": 0.93, + "learning_rate": 1.0716299849855348e-07, + "loss": 0.7026, + "step": 25401 + }, + { + "epoch": 0.93, + "learning_rate": 1.0715933643388142e-07, + "loss": 0.6851, + "step": 25402 + }, + { + "epoch": 0.93, + "learning_rate": 1.0715567436920936e-07, + "loss": 0.6931, + "step": 25403 + }, + { + "epoch": 0.93, + "learning_rate": 1.071520123045373e-07, + "loss": 0.6876, + "step": 25404 + }, + { + "epoch": 0.93, + "learning_rate": 1.0714835023986524e-07, + "loss": 0.6944, + "step": 25405 + }, + { + "epoch": 0.93, + "learning_rate": 1.0714468817519318e-07, + "loss": 0.6775, + "step": 25406 + }, + { + "epoch": 0.93, + "learning_rate": 1.0714102611052112e-07, + "loss": 0.6995, + "step": 25407 + }, + { + "epoch": 0.93, + "learning_rate": 1.0713736404584906e-07, + "loss": 0.6905, + "step": 25408 + }, + { + "epoch": 0.93, + "learning_rate": 1.07133701981177e-07, + "loss": 0.6948, + "step": 25409 + }, + { + "epoch": 0.93, + "learning_rate": 1.0713003991650491e-07, + "loss": 0.6786, + "step": 25410 + }, + { + "epoch": 0.93, + "learning_rate": 1.0712637785183285e-07, + "loss": 0.699, + "step": 25411 + }, + { + "epoch": 0.93, + "learning_rate": 1.0712271578716079e-07, + "loss": 0.6991, + "step": 25412 + }, + { + "epoch": 0.93, + "learning_rate": 1.0711905372248873e-07, + "loss": 0.7007, + "step": 25413 + }, + { + "epoch": 0.93, + "learning_rate": 1.0711539165781667e-07, + "loss": 0.7009, + "step": 25414 + }, + { + "epoch": 0.93, + "learning_rate": 1.0711172959314461e-07, + "loss": 0.6936, + "step": 25415 + }, + { + "epoch": 0.93, + "learning_rate": 1.0710806752847255e-07, + "loss": 0.6949, + "step": 25416 + }, + { + "epoch": 0.93, + "learning_rate": 1.0710440546380049e-07, + "loss": 0.6942, + "step": 25417 + }, + { + "epoch": 0.93, + "learning_rate": 1.0710074339912842e-07, + "loss": 0.6916, + "step": 25418 + }, + { + "epoch": 0.93, + "learning_rate": 1.0709708133445636e-07, + "loss": 0.6847, + "step": 25419 + }, + { + "epoch": 0.93, + "learning_rate": 1.070934192697843e-07, + "loss": 0.6864, + "step": 25420 + }, + { + "epoch": 0.93, + "learning_rate": 1.0708975720511224e-07, + "loss": 0.693, + "step": 25421 + }, + { + "epoch": 0.93, + "learning_rate": 1.0708609514044018e-07, + "loss": 0.6951, + "step": 25422 + }, + { + "epoch": 0.93, + "learning_rate": 1.0708243307576812e-07, + "loss": 0.7137, + "step": 25423 + }, + { + "epoch": 0.93, + "learning_rate": 1.0707877101109606e-07, + "loss": 0.6972, + "step": 25424 + }, + { + "epoch": 0.93, + "learning_rate": 1.07075108946424e-07, + "loss": 0.6903, + "step": 25425 + }, + { + "epoch": 0.93, + "learning_rate": 1.0707144688175194e-07, + "loss": 0.6915, + "step": 25426 + }, + { + "epoch": 0.93, + "learning_rate": 1.0706778481707988e-07, + "loss": 0.6959, + "step": 25427 + }, + { + "epoch": 0.93, + "learning_rate": 1.0706412275240779e-07, + "loss": 0.6938, + "step": 25428 + }, + { + "epoch": 0.93, + "learning_rate": 1.0706046068773573e-07, + "loss": 0.6971, + "step": 25429 + }, + { + "epoch": 0.93, + "learning_rate": 1.0705679862306367e-07, + "loss": 0.7043, + "step": 25430 + }, + { + "epoch": 0.93, + "learning_rate": 1.0705313655839161e-07, + "loss": 0.6893, + "step": 25431 + }, + { + "epoch": 0.93, + "learning_rate": 1.0704947449371955e-07, + "loss": 0.6931, + "step": 25432 + }, + { + "epoch": 0.93, + "learning_rate": 1.0704581242904749e-07, + "loss": 0.6972, + "step": 25433 + }, + { + "epoch": 0.93, + "learning_rate": 1.0704215036437543e-07, + "loss": 0.7084, + "step": 25434 + }, + { + "epoch": 0.93, + "learning_rate": 1.0703848829970337e-07, + "loss": 0.687, + "step": 25435 + }, + { + "epoch": 0.93, + "learning_rate": 1.0703482623503131e-07, + "loss": 0.6883, + "step": 25436 + }, + { + "epoch": 0.93, + "learning_rate": 1.0703116417035925e-07, + "loss": 0.6918, + "step": 25437 + }, + { + "epoch": 0.93, + "learning_rate": 1.0702750210568719e-07, + "loss": 0.6909, + "step": 25438 + }, + { + "epoch": 0.93, + "learning_rate": 1.0702384004101512e-07, + "loss": 0.6893, + "step": 25439 + }, + { + "epoch": 0.93, + "learning_rate": 1.0702017797634306e-07, + "loss": 0.7011, + "step": 25440 + }, + { + "epoch": 0.93, + "learning_rate": 1.07016515911671e-07, + "loss": 0.6933, + "step": 25441 + }, + { + "epoch": 0.93, + "learning_rate": 1.0701285384699894e-07, + "loss": 0.6975, + "step": 25442 + }, + { + "epoch": 0.93, + "learning_rate": 1.0700919178232688e-07, + "loss": 0.6928, + "step": 25443 + }, + { + "epoch": 0.93, + "learning_rate": 1.0700552971765482e-07, + "loss": 0.6983, + "step": 25444 + }, + { + "epoch": 0.93, + "learning_rate": 1.0700186765298273e-07, + "loss": 0.7001, + "step": 25445 + }, + { + "epoch": 0.93, + "learning_rate": 1.0699820558831067e-07, + "loss": 0.7012, + "step": 25446 + }, + { + "epoch": 0.93, + "learning_rate": 1.0699454352363861e-07, + "loss": 0.6728, + "step": 25447 + }, + { + "epoch": 0.93, + "learning_rate": 1.0699088145896655e-07, + "loss": 0.7024, + "step": 25448 + }, + { + "epoch": 0.93, + "learning_rate": 1.0698721939429449e-07, + "loss": 0.6913, + "step": 25449 + }, + { + "epoch": 0.93, + "learning_rate": 1.0698355732962243e-07, + "loss": 0.694, + "step": 25450 + }, + { + "epoch": 0.93, + "learning_rate": 1.0697989526495037e-07, + "loss": 0.7121, + "step": 25451 + }, + { + "epoch": 0.93, + "learning_rate": 1.0697623320027831e-07, + "loss": 0.6999, + "step": 25452 + }, + { + "epoch": 0.93, + "learning_rate": 1.0697257113560625e-07, + "loss": 0.7033, + "step": 25453 + }, + { + "epoch": 0.93, + "learning_rate": 1.0696890907093419e-07, + "loss": 0.694, + "step": 25454 + }, + { + "epoch": 0.93, + "learning_rate": 1.0696524700626213e-07, + "loss": 0.7084, + "step": 25455 + }, + { + "epoch": 0.93, + "learning_rate": 1.0696158494159007e-07, + "loss": 0.7022, + "step": 25456 + }, + { + "epoch": 0.93, + "learning_rate": 1.0695792287691801e-07, + "loss": 0.6962, + "step": 25457 + }, + { + "epoch": 0.93, + "learning_rate": 1.0695426081224595e-07, + "loss": 0.6906, + "step": 25458 + }, + { + "epoch": 0.93, + "learning_rate": 1.0695059874757389e-07, + "loss": 0.6921, + "step": 25459 + }, + { + "epoch": 0.93, + "learning_rate": 1.0694693668290183e-07, + "loss": 0.6958, + "step": 25460 + }, + { + "epoch": 0.93, + "learning_rate": 1.0694327461822976e-07, + "loss": 0.7005, + "step": 25461 + }, + { + "epoch": 0.93, + "learning_rate": 1.069396125535577e-07, + "loss": 0.6887, + "step": 25462 + }, + { + "epoch": 0.93, + "learning_rate": 1.0693595048888562e-07, + "loss": 0.7099, + "step": 25463 + }, + { + "epoch": 0.93, + "learning_rate": 1.0693228842421356e-07, + "loss": 0.6772, + "step": 25464 + }, + { + "epoch": 0.93, + "learning_rate": 1.069286263595415e-07, + "loss": 0.6974, + "step": 25465 + }, + { + "epoch": 0.93, + "learning_rate": 1.0692496429486944e-07, + "loss": 0.685, + "step": 25466 + }, + { + "epoch": 0.93, + "learning_rate": 1.0692130223019737e-07, + "loss": 0.6983, + "step": 25467 + }, + { + "epoch": 0.93, + "learning_rate": 1.0691764016552531e-07, + "loss": 0.6932, + "step": 25468 + }, + { + "epoch": 0.93, + "learning_rate": 1.0691397810085325e-07, + "loss": 0.6928, + "step": 25469 + }, + { + "epoch": 0.93, + "learning_rate": 1.0691031603618119e-07, + "loss": 0.6906, + "step": 25470 + }, + { + "epoch": 0.93, + "learning_rate": 1.0690665397150913e-07, + "loss": 0.6918, + "step": 25471 + }, + { + "epoch": 0.93, + "learning_rate": 1.0690299190683707e-07, + "loss": 0.6904, + "step": 25472 + }, + { + "epoch": 0.93, + "learning_rate": 1.0689932984216501e-07, + "loss": 0.7093, + "step": 25473 + }, + { + "epoch": 0.93, + "learning_rate": 1.0689566777749295e-07, + "loss": 0.6889, + "step": 25474 + }, + { + "epoch": 0.93, + "learning_rate": 1.0689200571282089e-07, + "loss": 0.6914, + "step": 25475 + }, + { + "epoch": 0.93, + "learning_rate": 1.0688834364814883e-07, + "loss": 0.6936, + "step": 25476 + }, + { + "epoch": 0.93, + "learning_rate": 1.0688468158347677e-07, + "loss": 0.7025, + "step": 25477 + }, + { + "epoch": 0.93, + "learning_rate": 1.0688101951880471e-07, + "loss": 0.6934, + "step": 25478 + }, + { + "epoch": 0.93, + "learning_rate": 1.0687735745413265e-07, + "loss": 0.6965, + "step": 25479 + }, + { + "epoch": 0.93, + "learning_rate": 1.0687369538946059e-07, + "loss": 0.6757, + "step": 25480 + }, + { + "epoch": 0.93, + "learning_rate": 1.068700333247885e-07, + "loss": 0.6942, + "step": 25481 + }, + { + "epoch": 0.93, + "learning_rate": 1.0686637126011644e-07, + "loss": 0.6894, + "step": 25482 + }, + { + "epoch": 0.93, + "learning_rate": 1.0686270919544438e-07, + "loss": 0.6902, + "step": 25483 + }, + { + "epoch": 0.93, + "learning_rate": 1.0685904713077232e-07, + "loss": 0.7012, + "step": 25484 + }, + { + "epoch": 0.93, + "learning_rate": 1.0685538506610026e-07, + "loss": 0.7002, + "step": 25485 + }, + { + "epoch": 0.93, + "learning_rate": 1.068517230014282e-07, + "loss": 0.6915, + "step": 25486 + }, + { + "epoch": 0.93, + "learning_rate": 1.0684806093675614e-07, + "loss": 0.7013, + "step": 25487 + }, + { + "epoch": 0.93, + "learning_rate": 1.0684439887208407e-07, + "loss": 0.6886, + "step": 25488 + }, + { + "epoch": 0.93, + "learning_rate": 1.0684073680741201e-07, + "loss": 0.6923, + "step": 25489 + }, + { + "epoch": 0.93, + "learning_rate": 1.0683707474273995e-07, + "loss": 0.6924, + "step": 25490 + }, + { + "epoch": 0.93, + "learning_rate": 1.0683341267806789e-07, + "loss": 0.6857, + "step": 25491 + }, + { + "epoch": 0.93, + "learning_rate": 1.0682975061339583e-07, + "loss": 0.6964, + "step": 25492 + }, + { + "epoch": 0.93, + "learning_rate": 1.0682608854872377e-07, + "loss": 0.6932, + "step": 25493 + }, + { + "epoch": 0.93, + "learning_rate": 1.0682242648405171e-07, + "loss": 0.7065, + "step": 25494 + }, + { + "epoch": 0.93, + "learning_rate": 1.0681876441937965e-07, + "loss": 0.693, + "step": 25495 + }, + { + "epoch": 0.93, + "learning_rate": 1.0681510235470759e-07, + "loss": 0.6875, + "step": 25496 + }, + { + "epoch": 0.93, + "learning_rate": 1.0681144029003553e-07, + "loss": 0.6786, + "step": 25497 + }, + { + "epoch": 0.93, + "learning_rate": 1.0680777822536347e-07, + "loss": 0.6829, + "step": 25498 + }, + { + "epoch": 0.93, + "learning_rate": 1.0680411616069138e-07, + "loss": 0.7028, + "step": 25499 + }, + { + "epoch": 0.93, + "learning_rate": 1.0680045409601932e-07, + "loss": 0.689, + "step": 25500 + }, + { + "epoch": 0.93, + "learning_rate": 1.0679679203134726e-07, + "loss": 0.6831, + "step": 25501 + }, + { + "epoch": 0.93, + "learning_rate": 1.067931299666752e-07, + "loss": 0.6832, + "step": 25502 + }, + { + "epoch": 0.93, + "learning_rate": 1.0678946790200314e-07, + "loss": 0.688, + "step": 25503 + }, + { + "epoch": 0.93, + "learning_rate": 1.0678580583733108e-07, + "loss": 0.6889, + "step": 25504 + }, + { + "epoch": 0.93, + "learning_rate": 1.0678214377265902e-07, + "loss": 0.6897, + "step": 25505 + }, + { + "epoch": 0.93, + "learning_rate": 1.0677848170798696e-07, + "loss": 0.6668, + "step": 25506 + }, + { + "epoch": 0.93, + "learning_rate": 1.067748196433149e-07, + "loss": 0.6906, + "step": 25507 + }, + { + "epoch": 0.93, + "learning_rate": 1.0677115757864284e-07, + "loss": 0.682, + "step": 25508 + }, + { + "epoch": 0.93, + "learning_rate": 1.0676749551397078e-07, + "loss": 0.6985, + "step": 25509 + }, + { + "epoch": 0.93, + "learning_rate": 1.067638334492987e-07, + "loss": 0.6946, + "step": 25510 + }, + { + "epoch": 0.93, + "learning_rate": 1.0676017138462665e-07, + "loss": 0.6952, + "step": 25511 + }, + { + "epoch": 0.93, + "learning_rate": 1.0675650931995459e-07, + "loss": 0.6956, + "step": 25512 + }, + { + "epoch": 0.93, + "learning_rate": 1.0675284725528253e-07, + "loss": 0.6943, + "step": 25513 + }, + { + "epoch": 0.93, + "learning_rate": 1.0674918519061047e-07, + "loss": 0.697, + "step": 25514 + }, + { + "epoch": 0.93, + "learning_rate": 1.067455231259384e-07, + "loss": 0.6925, + "step": 25515 + }, + { + "epoch": 0.93, + "learning_rate": 1.0674186106126635e-07, + "loss": 0.6932, + "step": 25516 + }, + { + "epoch": 0.93, + "learning_rate": 1.0673819899659426e-07, + "loss": 0.6997, + "step": 25517 + }, + { + "epoch": 0.93, + "learning_rate": 1.067345369319222e-07, + "loss": 0.6997, + "step": 25518 + }, + { + "epoch": 0.93, + "learning_rate": 1.0673087486725014e-07, + "loss": 0.6965, + "step": 25519 + }, + { + "epoch": 0.93, + "learning_rate": 1.0672721280257808e-07, + "loss": 0.6961, + "step": 25520 + }, + { + "epoch": 0.93, + "learning_rate": 1.0672355073790602e-07, + "loss": 0.6866, + "step": 25521 + }, + { + "epoch": 0.93, + "learning_rate": 1.0671988867323396e-07, + "loss": 0.6926, + "step": 25522 + }, + { + "epoch": 0.93, + "learning_rate": 1.067162266085619e-07, + "loss": 0.7015, + "step": 25523 + }, + { + "epoch": 0.93, + "learning_rate": 1.0671256454388984e-07, + "loss": 0.7014, + "step": 25524 + }, + { + "epoch": 0.93, + "learning_rate": 1.0670890247921778e-07, + "loss": 0.6949, + "step": 25525 + }, + { + "epoch": 0.93, + "learning_rate": 1.0670524041454572e-07, + "loss": 0.6983, + "step": 25526 + }, + { + "epoch": 0.93, + "learning_rate": 1.0670157834987366e-07, + "loss": 0.6868, + "step": 25527 + }, + { + "epoch": 0.93, + "learning_rate": 1.066979162852016e-07, + "loss": 0.6817, + "step": 25528 + }, + { + "epoch": 0.93, + "learning_rate": 1.0669425422052954e-07, + "loss": 0.7, + "step": 25529 + }, + { + "epoch": 0.93, + "learning_rate": 1.0669059215585748e-07, + "loss": 0.6934, + "step": 25530 + }, + { + "epoch": 0.93, + "learning_rate": 1.0668693009118542e-07, + "loss": 0.7028, + "step": 25531 + }, + { + "epoch": 0.93, + "learning_rate": 1.0668326802651334e-07, + "loss": 0.69, + "step": 25532 + }, + { + "epoch": 0.93, + "learning_rate": 1.0667960596184128e-07, + "loss": 0.6982, + "step": 25533 + }, + { + "epoch": 0.93, + "learning_rate": 1.0667594389716922e-07, + "loss": 0.6895, + "step": 25534 + }, + { + "epoch": 0.93, + "learning_rate": 1.0667228183249715e-07, + "loss": 0.7043, + "step": 25535 + }, + { + "epoch": 0.93, + "learning_rate": 1.0666861976782509e-07, + "loss": 0.69, + "step": 25536 + }, + { + "epoch": 0.93, + "learning_rate": 1.0666495770315303e-07, + "loss": 0.697, + "step": 25537 + }, + { + "epoch": 0.93, + "learning_rate": 1.0666129563848096e-07, + "loss": 0.6804, + "step": 25538 + }, + { + "epoch": 0.93, + "learning_rate": 1.066576335738089e-07, + "loss": 0.6999, + "step": 25539 + }, + { + "epoch": 0.93, + "learning_rate": 1.0665397150913684e-07, + "loss": 0.6869, + "step": 25540 + }, + { + "epoch": 0.93, + "learning_rate": 1.0665030944446478e-07, + "loss": 0.7094, + "step": 25541 + }, + { + "epoch": 0.93, + "learning_rate": 1.0664664737979272e-07, + "loss": 0.6949, + "step": 25542 + }, + { + "epoch": 0.93, + "learning_rate": 1.0664298531512066e-07, + "loss": 0.7004, + "step": 25543 + }, + { + "epoch": 0.93, + "learning_rate": 1.066393232504486e-07, + "loss": 0.7018, + "step": 25544 + }, + { + "epoch": 0.93, + "learning_rate": 1.0663566118577654e-07, + "loss": 0.6849, + "step": 25545 + }, + { + "epoch": 0.93, + "learning_rate": 1.0663199912110448e-07, + "loss": 0.7007, + "step": 25546 + }, + { + "epoch": 0.93, + "learning_rate": 1.0662833705643242e-07, + "loss": 0.7, + "step": 25547 + }, + { + "epoch": 0.93, + "learning_rate": 1.0662467499176036e-07, + "loss": 0.6846, + "step": 25548 + }, + { + "epoch": 0.93, + "learning_rate": 1.066210129270883e-07, + "loss": 0.6927, + "step": 25549 + }, + { + "epoch": 0.93, + "learning_rate": 1.0661735086241624e-07, + "loss": 0.6823, + "step": 25550 + }, + { + "epoch": 0.93, + "learning_rate": 1.0661368879774418e-07, + "loss": 0.6948, + "step": 25551 + }, + { + "epoch": 0.93, + "learning_rate": 1.0661002673307212e-07, + "loss": 0.6919, + "step": 25552 + }, + { + "epoch": 0.93, + "learning_rate": 1.0660636466840003e-07, + "loss": 0.6673, + "step": 25553 + }, + { + "epoch": 0.93, + "learning_rate": 1.0660270260372797e-07, + "loss": 0.6888, + "step": 25554 + }, + { + "epoch": 0.93, + "learning_rate": 1.0659904053905591e-07, + "loss": 0.6925, + "step": 25555 + }, + { + "epoch": 0.94, + "learning_rate": 1.0659537847438385e-07, + "loss": 0.6954, + "step": 25556 + }, + { + "epoch": 0.94, + "learning_rate": 1.0659171640971179e-07, + "loss": 0.6891, + "step": 25557 + }, + { + "epoch": 0.94, + "learning_rate": 1.0658805434503973e-07, + "loss": 0.6878, + "step": 25558 + }, + { + "epoch": 0.94, + "learning_rate": 1.0658439228036766e-07, + "loss": 0.6874, + "step": 25559 + }, + { + "epoch": 0.94, + "learning_rate": 1.065807302156956e-07, + "loss": 0.6911, + "step": 25560 + }, + { + "epoch": 0.94, + "learning_rate": 1.0657706815102354e-07, + "loss": 0.6948, + "step": 25561 + }, + { + "epoch": 0.94, + "learning_rate": 1.0657340608635148e-07, + "loss": 0.6867, + "step": 25562 + }, + { + "epoch": 0.94, + "learning_rate": 1.0656974402167942e-07, + "loss": 0.6879, + "step": 25563 + }, + { + "epoch": 0.94, + "learning_rate": 1.0656608195700736e-07, + "loss": 0.7017, + "step": 25564 + }, + { + "epoch": 0.94, + "learning_rate": 1.065624198923353e-07, + "loss": 0.6859, + "step": 25565 + }, + { + "epoch": 0.94, + "learning_rate": 1.0655875782766324e-07, + "loss": 0.6758, + "step": 25566 + }, + { + "epoch": 0.94, + "learning_rate": 1.0655509576299118e-07, + "loss": 0.705, + "step": 25567 + }, + { + "epoch": 0.94, + "learning_rate": 1.0655143369831912e-07, + "loss": 0.6928, + "step": 25568 + }, + { + "epoch": 0.94, + "learning_rate": 1.0654777163364706e-07, + "loss": 0.7019, + "step": 25569 + }, + { + "epoch": 0.94, + "learning_rate": 1.06544109568975e-07, + "loss": 0.7002, + "step": 25570 + }, + { + "epoch": 0.94, + "learning_rate": 1.0654044750430291e-07, + "loss": 0.6926, + "step": 25571 + }, + { + "epoch": 0.94, + "learning_rate": 1.0653678543963085e-07, + "loss": 0.7165, + "step": 25572 + }, + { + "epoch": 0.94, + "learning_rate": 1.0653312337495879e-07, + "loss": 0.6887, + "step": 25573 + }, + { + "epoch": 0.94, + "learning_rate": 1.0652946131028673e-07, + "loss": 0.6888, + "step": 25574 + }, + { + "epoch": 0.94, + "learning_rate": 1.0652579924561467e-07, + "loss": 0.6937, + "step": 25575 + }, + { + "epoch": 0.94, + "learning_rate": 1.0652213718094261e-07, + "loss": 0.6988, + "step": 25576 + }, + { + "epoch": 0.94, + "learning_rate": 1.0651847511627055e-07, + "loss": 0.7003, + "step": 25577 + }, + { + "epoch": 0.94, + "learning_rate": 1.0651481305159849e-07, + "loss": 0.686, + "step": 25578 + }, + { + "epoch": 0.94, + "learning_rate": 1.0651115098692643e-07, + "loss": 0.7089, + "step": 25579 + }, + { + "epoch": 0.94, + "learning_rate": 1.0650748892225437e-07, + "loss": 0.6847, + "step": 25580 + }, + { + "epoch": 0.94, + "learning_rate": 1.065038268575823e-07, + "loss": 0.687, + "step": 25581 + }, + { + "epoch": 0.94, + "learning_rate": 1.0650016479291024e-07, + "loss": 0.6906, + "step": 25582 + }, + { + "epoch": 0.94, + "learning_rate": 1.0649650272823818e-07, + "loss": 0.6964, + "step": 25583 + }, + { + "epoch": 0.94, + "learning_rate": 1.0649284066356612e-07, + "loss": 0.6939, + "step": 25584 + }, + { + "epoch": 0.94, + "learning_rate": 1.0648917859889406e-07, + "loss": 0.6954, + "step": 25585 + }, + { + "epoch": 0.94, + "learning_rate": 1.06485516534222e-07, + "loss": 0.6942, + "step": 25586 + }, + { + "epoch": 0.94, + "learning_rate": 1.0648185446954993e-07, + "loss": 0.703, + "step": 25587 + }, + { + "epoch": 0.94, + "learning_rate": 1.0647819240487787e-07, + "loss": 0.7089, + "step": 25588 + }, + { + "epoch": 0.94, + "learning_rate": 1.0647453034020579e-07, + "loss": 0.7001, + "step": 25589 + }, + { + "epoch": 0.94, + "learning_rate": 1.0647086827553373e-07, + "loss": 0.6807, + "step": 25590 + }, + { + "epoch": 0.94, + "learning_rate": 1.0646720621086167e-07, + "loss": 0.6806, + "step": 25591 + }, + { + "epoch": 0.94, + "learning_rate": 1.0646354414618961e-07, + "loss": 0.6852, + "step": 25592 + }, + { + "epoch": 0.94, + "learning_rate": 1.0645988208151755e-07, + "loss": 0.7006, + "step": 25593 + }, + { + "epoch": 0.94, + "learning_rate": 1.0645622001684549e-07, + "loss": 0.6945, + "step": 25594 + }, + { + "epoch": 0.94, + "learning_rate": 1.0645255795217343e-07, + "loss": 0.6943, + "step": 25595 + }, + { + "epoch": 0.94, + "learning_rate": 1.0644889588750137e-07, + "loss": 0.6955, + "step": 25596 + }, + { + "epoch": 0.94, + "learning_rate": 1.0644523382282931e-07, + "loss": 0.6854, + "step": 25597 + }, + { + "epoch": 0.94, + "learning_rate": 1.0644157175815725e-07, + "loss": 0.6875, + "step": 25598 + }, + { + "epoch": 0.94, + "learning_rate": 1.0643790969348519e-07, + "loss": 0.699, + "step": 25599 + }, + { + "epoch": 0.94, + "learning_rate": 1.0643424762881313e-07, + "loss": 0.6967, + "step": 25600 + }, + { + "epoch": 0.94, + "learning_rate": 1.0643058556414107e-07, + "loss": 0.6945, + "step": 25601 + }, + { + "epoch": 0.94, + "learning_rate": 1.0642692349946901e-07, + "loss": 0.6986, + "step": 25602 + }, + { + "epoch": 0.94, + "learning_rate": 1.0642326143479693e-07, + "loss": 0.6927, + "step": 25603 + }, + { + "epoch": 0.94, + "learning_rate": 1.0641959937012487e-07, + "loss": 0.6976, + "step": 25604 + }, + { + "epoch": 0.94, + "learning_rate": 1.0641593730545281e-07, + "loss": 0.6875, + "step": 25605 + }, + { + "epoch": 0.94, + "learning_rate": 1.0641227524078075e-07, + "loss": 0.6937, + "step": 25606 + }, + { + "epoch": 0.94, + "learning_rate": 1.0640861317610868e-07, + "loss": 0.6867, + "step": 25607 + }, + { + "epoch": 0.94, + "learning_rate": 1.0640495111143661e-07, + "loss": 0.694, + "step": 25608 + }, + { + "epoch": 0.94, + "learning_rate": 1.0640128904676455e-07, + "loss": 0.6952, + "step": 25609 + }, + { + "epoch": 0.94, + "learning_rate": 1.0639762698209249e-07, + "loss": 0.6891, + "step": 25610 + }, + { + "epoch": 0.94, + "learning_rate": 1.0639396491742043e-07, + "loss": 0.6858, + "step": 25611 + }, + { + "epoch": 0.94, + "learning_rate": 1.0639030285274837e-07, + "loss": 0.7011, + "step": 25612 + }, + { + "epoch": 0.94, + "learning_rate": 1.0638664078807631e-07, + "loss": 0.6916, + "step": 25613 + }, + { + "epoch": 0.94, + "learning_rate": 1.0638297872340425e-07, + "loss": 0.7015, + "step": 25614 + }, + { + "epoch": 0.94, + "learning_rate": 1.0637931665873219e-07, + "loss": 0.6979, + "step": 25615 + }, + { + "epoch": 0.94, + "learning_rate": 1.0637565459406013e-07, + "loss": 0.6905, + "step": 25616 + }, + { + "epoch": 0.94, + "learning_rate": 1.0637199252938807e-07, + "loss": 0.6973, + "step": 25617 + }, + { + "epoch": 0.94, + "learning_rate": 1.06368330464716e-07, + "loss": 0.6885, + "step": 25618 + }, + { + "epoch": 0.94, + "learning_rate": 1.0636466840004395e-07, + "loss": 0.6929, + "step": 25619 + }, + { + "epoch": 0.94, + "learning_rate": 1.0636100633537189e-07, + "loss": 0.6884, + "step": 25620 + }, + { + "epoch": 0.94, + "learning_rate": 1.0635734427069983e-07, + "loss": 0.6873, + "step": 25621 + }, + { + "epoch": 0.94, + "learning_rate": 1.0635368220602777e-07, + "loss": 0.6913, + "step": 25622 + }, + { + "epoch": 0.94, + "learning_rate": 1.063500201413557e-07, + "loss": 0.677, + "step": 25623 + }, + { + "epoch": 0.94, + "learning_rate": 1.0634635807668365e-07, + "loss": 0.6949, + "step": 25624 + }, + { + "epoch": 0.94, + "learning_rate": 1.0634269601201156e-07, + "loss": 0.6745, + "step": 25625 + }, + { + "epoch": 0.94, + "learning_rate": 1.063390339473395e-07, + "loss": 0.7103, + "step": 25626 + }, + { + "epoch": 0.94, + "learning_rate": 1.0633537188266744e-07, + "loss": 0.7039, + "step": 25627 + }, + { + "epoch": 0.94, + "learning_rate": 1.0633170981799538e-07, + "loss": 0.6932, + "step": 25628 + }, + { + "epoch": 0.94, + "learning_rate": 1.0632804775332332e-07, + "loss": 0.702, + "step": 25629 + }, + { + "epoch": 0.94, + "learning_rate": 1.0632438568865125e-07, + "loss": 0.6993, + "step": 25630 + }, + { + "epoch": 0.94, + "learning_rate": 1.0632072362397919e-07, + "loss": 0.6967, + "step": 25631 + }, + { + "epoch": 0.94, + "learning_rate": 1.0631706155930713e-07, + "loss": 0.682, + "step": 25632 + }, + { + "epoch": 0.94, + "learning_rate": 1.0631339949463507e-07, + "loss": 0.7045, + "step": 25633 + }, + { + "epoch": 0.94, + "learning_rate": 1.06309737429963e-07, + "loss": 0.6917, + "step": 25634 + }, + { + "epoch": 0.94, + "learning_rate": 1.0630607536529095e-07, + "loss": 0.6926, + "step": 25635 + }, + { + "epoch": 0.94, + "learning_rate": 1.0630241330061889e-07, + "loss": 0.7005, + "step": 25636 + }, + { + "epoch": 0.94, + "learning_rate": 1.0629875123594683e-07, + "loss": 0.6865, + "step": 25637 + }, + { + "epoch": 0.94, + "learning_rate": 1.0629508917127477e-07, + "loss": 0.7015, + "step": 25638 + }, + { + "epoch": 0.94, + "learning_rate": 1.062914271066027e-07, + "loss": 0.7, + "step": 25639 + }, + { + "epoch": 0.94, + "learning_rate": 1.0628776504193065e-07, + "loss": 0.689, + "step": 25640 + }, + { + "epoch": 0.94, + "learning_rate": 1.0628410297725859e-07, + "loss": 0.6889, + "step": 25641 + }, + { + "epoch": 0.94, + "learning_rate": 1.0628044091258653e-07, + "loss": 0.6923, + "step": 25642 + }, + { + "epoch": 0.94, + "learning_rate": 1.0627677884791444e-07, + "loss": 0.6834, + "step": 25643 + }, + { + "epoch": 0.94, + "learning_rate": 1.0627311678324238e-07, + "loss": 0.6921, + "step": 25644 + }, + { + "epoch": 0.94, + "learning_rate": 1.0626945471857032e-07, + "loss": 0.6942, + "step": 25645 + }, + { + "epoch": 0.94, + "learning_rate": 1.0626579265389826e-07, + "loss": 0.6993, + "step": 25646 + }, + { + "epoch": 0.94, + "learning_rate": 1.062621305892262e-07, + "loss": 0.6968, + "step": 25647 + }, + { + "epoch": 0.94, + "learning_rate": 1.0625846852455414e-07, + "loss": 0.6978, + "step": 25648 + }, + { + "epoch": 0.94, + "learning_rate": 1.0625480645988208e-07, + "loss": 0.7041, + "step": 25649 + }, + { + "epoch": 0.94, + "learning_rate": 1.0625114439521002e-07, + "loss": 0.7092, + "step": 25650 + }, + { + "epoch": 0.94, + "learning_rate": 1.0624748233053796e-07, + "loss": 0.6864, + "step": 25651 + }, + { + "epoch": 0.94, + "learning_rate": 1.0624382026586588e-07, + "loss": 0.6877, + "step": 25652 + }, + { + "epoch": 0.94, + "learning_rate": 1.0624015820119382e-07, + "loss": 0.6966, + "step": 25653 + }, + { + "epoch": 0.94, + "learning_rate": 1.0623649613652176e-07, + "loss": 0.6908, + "step": 25654 + }, + { + "epoch": 0.94, + "learning_rate": 1.062328340718497e-07, + "loss": 0.6858, + "step": 25655 + }, + { + "epoch": 0.94, + "learning_rate": 1.0622917200717764e-07, + "loss": 0.69, + "step": 25656 + }, + { + "epoch": 0.94, + "learning_rate": 1.0622550994250558e-07, + "loss": 0.697, + "step": 25657 + }, + { + "epoch": 0.94, + "learning_rate": 1.0622184787783352e-07, + "loss": 0.6981, + "step": 25658 + }, + { + "epoch": 0.94, + "learning_rate": 1.0621818581316146e-07, + "loss": 0.6995, + "step": 25659 + }, + { + "epoch": 0.94, + "learning_rate": 1.062145237484894e-07, + "loss": 0.6881, + "step": 25660 + }, + { + "epoch": 0.94, + "learning_rate": 1.0621086168381732e-07, + "loss": 0.7085, + "step": 25661 + }, + { + "epoch": 0.94, + "learning_rate": 1.0620719961914526e-07, + "loss": 0.6868, + "step": 25662 + }, + { + "epoch": 0.94, + "learning_rate": 1.062035375544732e-07, + "loss": 0.693, + "step": 25663 + }, + { + "epoch": 0.94, + "learning_rate": 1.0619987548980114e-07, + "loss": 0.6746, + "step": 25664 + }, + { + "epoch": 0.94, + "learning_rate": 1.0619621342512908e-07, + "loss": 0.6897, + "step": 25665 + }, + { + "epoch": 0.94, + "learning_rate": 1.0619255136045702e-07, + "loss": 0.6929, + "step": 25666 + }, + { + "epoch": 0.94, + "learning_rate": 1.0618888929578496e-07, + "loss": 0.6936, + "step": 25667 + }, + { + "epoch": 0.94, + "learning_rate": 1.061852272311129e-07, + "loss": 0.6942, + "step": 25668 + }, + { + "epoch": 0.94, + "learning_rate": 1.0618156516644084e-07, + "loss": 0.6953, + "step": 25669 + }, + { + "epoch": 0.94, + "learning_rate": 1.0617790310176878e-07, + "loss": 0.6965, + "step": 25670 + }, + { + "epoch": 0.94, + "learning_rate": 1.0617424103709672e-07, + "loss": 0.6887, + "step": 25671 + }, + { + "epoch": 0.94, + "learning_rate": 1.0617057897242466e-07, + "loss": 0.712, + "step": 25672 + }, + { + "epoch": 0.94, + "learning_rate": 1.061669169077526e-07, + "loss": 0.6838, + "step": 25673 + }, + { + "epoch": 0.94, + "learning_rate": 1.0616325484308052e-07, + "loss": 0.6952, + "step": 25674 + }, + { + "epoch": 0.94, + "learning_rate": 1.0615959277840846e-07, + "loss": 0.7044, + "step": 25675 + }, + { + "epoch": 0.94, + "learning_rate": 1.061559307137364e-07, + "loss": 0.6876, + "step": 25676 + }, + { + "epoch": 0.94, + "learning_rate": 1.0615226864906434e-07, + "loss": 0.7005, + "step": 25677 + }, + { + "epoch": 0.94, + "learning_rate": 1.0614860658439228e-07, + "loss": 0.702, + "step": 25678 + }, + { + "epoch": 0.94, + "learning_rate": 1.061449445197202e-07, + "loss": 0.6918, + "step": 25679 + }, + { + "epoch": 0.94, + "learning_rate": 1.0614128245504814e-07, + "loss": 0.6948, + "step": 25680 + }, + { + "epoch": 0.94, + "learning_rate": 1.0613762039037608e-07, + "loss": 0.6917, + "step": 25681 + }, + { + "epoch": 0.94, + "learning_rate": 1.0613395832570402e-07, + "loss": 0.6849, + "step": 25682 + }, + { + "epoch": 0.94, + "learning_rate": 1.0613029626103196e-07, + "loss": 0.6901, + "step": 25683 + }, + { + "epoch": 0.94, + "learning_rate": 1.061266341963599e-07, + "loss": 0.6906, + "step": 25684 + }, + { + "epoch": 0.94, + "learning_rate": 1.0612297213168784e-07, + "loss": 0.6945, + "step": 25685 + }, + { + "epoch": 0.94, + "learning_rate": 1.0611931006701578e-07, + "loss": 0.7027, + "step": 25686 + }, + { + "epoch": 0.94, + "learning_rate": 1.0611564800234372e-07, + "loss": 0.6879, + "step": 25687 + }, + { + "epoch": 0.94, + "learning_rate": 1.0611198593767166e-07, + "loss": 0.6995, + "step": 25688 + }, + { + "epoch": 0.94, + "learning_rate": 1.061083238729996e-07, + "loss": 0.6942, + "step": 25689 + }, + { + "epoch": 0.94, + "learning_rate": 1.0610466180832754e-07, + "loss": 0.6896, + "step": 25690 + }, + { + "epoch": 0.94, + "learning_rate": 1.0610099974365548e-07, + "loss": 0.7041, + "step": 25691 + }, + { + "epoch": 0.94, + "learning_rate": 1.0609733767898342e-07, + "loss": 0.6937, + "step": 25692 + }, + { + "epoch": 0.94, + "learning_rate": 1.0609367561431136e-07, + "loss": 0.6902, + "step": 25693 + }, + { + "epoch": 0.94, + "learning_rate": 1.060900135496393e-07, + "loss": 0.694, + "step": 25694 + }, + { + "epoch": 0.94, + "learning_rate": 1.0608635148496724e-07, + "loss": 0.6828, + "step": 25695 + }, + { + "epoch": 0.94, + "learning_rate": 1.0608268942029516e-07, + "loss": 0.6912, + "step": 25696 + }, + { + "epoch": 0.94, + "learning_rate": 1.0607902735562309e-07, + "loss": 0.6873, + "step": 25697 + }, + { + "epoch": 0.94, + "learning_rate": 1.0607536529095103e-07, + "loss": 0.704, + "step": 25698 + }, + { + "epoch": 0.94, + "learning_rate": 1.0607170322627897e-07, + "loss": 0.6973, + "step": 25699 + }, + { + "epoch": 0.94, + "learning_rate": 1.0606804116160691e-07, + "loss": 0.6949, + "step": 25700 + }, + { + "epoch": 0.94, + "learning_rate": 1.0606437909693484e-07, + "loss": 0.6884, + "step": 25701 + }, + { + "epoch": 0.94, + "learning_rate": 1.0606071703226278e-07, + "loss": 0.6909, + "step": 25702 + }, + { + "epoch": 0.94, + "learning_rate": 1.0605705496759072e-07, + "loss": 0.6742, + "step": 25703 + }, + { + "epoch": 0.94, + "learning_rate": 1.0605339290291866e-07, + "loss": 0.7009, + "step": 25704 + }, + { + "epoch": 0.94, + "learning_rate": 1.060497308382466e-07, + "loss": 0.6855, + "step": 25705 + }, + { + "epoch": 0.94, + "learning_rate": 1.0604606877357453e-07, + "loss": 0.6874, + "step": 25706 + }, + { + "epoch": 0.94, + "learning_rate": 1.0604240670890247e-07, + "loss": 0.6895, + "step": 25707 + }, + { + "epoch": 0.94, + "learning_rate": 1.0603874464423041e-07, + "loss": 0.6903, + "step": 25708 + }, + { + "epoch": 0.94, + "learning_rate": 1.0603508257955835e-07, + "loss": 0.688, + "step": 25709 + }, + { + "epoch": 0.94, + "learning_rate": 1.060314205148863e-07, + "loss": 0.698, + "step": 25710 + }, + { + "epoch": 0.94, + "learning_rate": 1.0602775845021423e-07, + "loss": 0.7034, + "step": 25711 + }, + { + "epoch": 0.94, + "learning_rate": 1.0602409638554217e-07, + "loss": 0.6864, + "step": 25712 + }, + { + "epoch": 0.94, + "learning_rate": 1.0602043432087011e-07, + "loss": 0.6868, + "step": 25713 + }, + { + "epoch": 0.94, + "learning_rate": 1.0601677225619805e-07, + "loss": 0.6945, + "step": 25714 + }, + { + "epoch": 0.94, + "learning_rate": 1.0601311019152597e-07, + "loss": 0.6768, + "step": 25715 + }, + { + "epoch": 0.94, + "learning_rate": 1.0600944812685391e-07, + "loss": 0.6953, + "step": 25716 + }, + { + "epoch": 0.94, + "learning_rate": 1.0600578606218185e-07, + "loss": 0.6997, + "step": 25717 + }, + { + "epoch": 0.94, + "learning_rate": 1.0600212399750979e-07, + "loss": 0.6998, + "step": 25718 + }, + { + "epoch": 0.94, + "learning_rate": 1.0599846193283773e-07, + "loss": 0.6855, + "step": 25719 + }, + { + "epoch": 0.94, + "learning_rate": 1.0599479986816567e-07, + "loss": 0.6884, + "step": 25720 + }, + { + "epoch": 0.94, + "learning_rate": 1.0599113780349361e-07, + "loss": 0.6919, + "step": 25721 + }, + { + "epoch": 0.94, + "learning_rate": 1.0598747573882155e-07, + "loss": 0.7059, + "step": 25722 + }, + { + "epoch": 0.94, + "learning_rate": 1.0598381367414947e-07, + "loss": 0.6879, + "step": 25723 + }, + { + "epoch": 0.94, + "learning_rate": 1.0598015160947741e-07, + "loss": 0.7078, + "step": 25724 + }, + { + "epoch": 0.94, + "learning_rate": 1.0597648954480535e-07, + "loss": 0.7044, + "step": 25725 + }, + { + "epoch": 0.94, + "learning_rate": 1.059728274801333e-07, + "loss": 0.6932, + "step": 25726 + }, + { + "epoch": 0.94, + "learning_rate": 1.0596916541546123e-07, + "loss": 0.7038, + "step": 25727 + }, + { + "epoch": 0.94, + "learning_rate": 1.0596550335078917e-07, + "loss": 0.6927, + "step": 25728 + }, + { + "epoch": 0.94, + "learning_rate": 1.0596184128611711e-07, + "loss": 0.7008, + "step": 25729 + }, + { + "epoch": 0.94, + "learning_rate": 1.0595817922144505e-07, + "loss": 0.6898, + "step": 25730 + }, + { + "epoch": 0.94, + "learning_rate": 1.05954517156773e-07, + "loss": 0.6846, + "step": 25731 + }, + { + "epoch": 0.94, + "learning_rate": 1.0595085509210093e-07, + "loss": 0.6938, + "step": 25732 + }, + { + "epoch": 0.94, + "learning_rate": 1.0594719302742885e-07, + "loss": 0.688, + "step": 25733 + }, + { + "epoch": 0.94, + "learning_rate": 1.0594353096275679e-07, + "loss": 0.6755, + "step": 25734 + }, + { + "epoch": 0.94, + "learning_rate": 1.0593986889808473e-07, + "loss": 0.6932, + "step": 25735 + }, + { + "epoch": 0.94, + "learning_rate": 1.0593620683341267e-07, + "loss": 0.6949, + "step": 25736 + }, + { + "epoch": 0.94, + "learning_rate": 1.059325447687406e-07, + "loss": 0.6931, + "step": 25737 + }, + { + "epoch": 0.94, + "learning_rate": 1.0592888270406855e-07, + "loss": 0.6936, + "step": 25738 + }, + { + "epoch": 0.94, + "learning_rate": 1.0592522063939649e-07, + "loss": 0.6981, + "step": 25739 + }, + { + "epoch": 0.94, + "learning_rate": 1.0592155857472443e-07, + "loss": 0.6916, + "step": 25740 + }, + { + "epoch": 0.94, + "learning_rate": 1.0591789651005237e-07, + "loss": 0.7076, + "step": 25741 + }, + { + "epoch": 0.94, + "learning_rate": 1.059142344453803e-07, + "loss": 0.6917, + "step": 25742 + }, + { + "epoch": 0.94, + "learning_rate": 1.0591057238070825e-07, + "loss": 0.6993, + "step": 25743 + }, + { + "epoch": 0.94, + "learning_rate": 1.0590691031603619e-07, + "loss": 0.6797, + "step": 25744 + }, + { + "epoch": 0.94, + "learning_rate": 1.0590324825136411e-07, + "loss": 0.6822, + "step": 25745 + }, + { + "epoch": 0.94, + "learning_rate": 1.0589958618669205e-07, + "loss": 0.6987, + "step": 25746 + }, + { + "epoch": 0.94, + "learning_rate": 1.0589592412201999e-07, + "loss": 0.6861, + "step": 25747 + }, + { + "epoch": 0.94, + "learning_rate": 1.0589226205734793e-07, + "loss": 0.703, + "step": 25748 + }, + { + "epoch": 0.94, + "learning_rate": 1.0588859999267587e-07, + "loss": 0.6932, + "step": 25749 + }, + { + "epoch": 0.94, + "learning_rate": 1.0588493792800381e-07, + "loss": 0.681, + "step": 25750 + }, + { + "epoch": 0.94, + "learning_rate": 1.0588127586333173e-07, + "loss": 0.6862, + "step": 25751 + }, + { + "epoch": 0.94, + "learning_rate": 1.0587761379865967e-07, + "loss": 0.6942, + "step": 25752 + }, + { + "epoch": 0.94, + "learning_rate": 1.058739517339876e-07, + "loss": 0.6891, + "step": 25753 + }, + { + "epoch": 0.94, + "learning_rate": 1.0587028966931555e-07, + "loss": 0.6919, + "step": 25754 + }, + { + "epoch": 0.94, + "learning_rate": 1.0586662760464349e-07, + "loss": 0.694, + "step": 25755 + }, + { + "epoch": 0.94, + "learning_rate": 1.0586296553997143e-07, + "loss": 0.6816, + "step": 25756 + }, + { + "epoch": 0.94, + "learning_rate": 1.0585930347529937e-07, + "loss": 0.6822, + "step": 25757 + }, + { + "epoch": 0.94, + "learning_rate": 1.058556414106273e-07, + "loss": 0.7041, + "step": 25758 + }, + { + "epoch": 0.94, + "learning_rate": 1.0585197934595525e-07, + "loss": 0.6827, + "step": 25759 + }, + { + "epoch": 0.94, + "learning_rate": 1.0584831728128319e-07, + "loss": 0.6987, + "step": 25760 + }, + { + "epoch": 0.94, + "learning_rate": 1.0584465521661113e-07, + "loss": 0.7005, + "step": 25761 + }, + { + "epoch": 0.94, + "learning_rate": 1.0584099315193907e-07, + "loss": 0.6953, + "step": 25762 + }, + { + "epoch": 0.94, + "learning_rate": 1.05837331087267e-07, + "loss": 0.6904, + "step": 25763 + }, + { + "epoch": 0.94, + "learning_rate": 1.0583366902259494e-07, + "loss": 0.6799, + "step": 25764 + }, + { + "epoch": 0.94, + "learning_rate": 1.0583000695792288e-07, + "loss": 0.7013, + "step": 25765 + }, + { + "epoch": 0.94, + "learning_rate": 1.0582634489325082e-07, + "loss": 0.7041, + "step": 25766 + }, + { + "epoch": 0.94, + "learning_rate": 1.0582268282857875e-07, + "loss": 0.6849, + "step": 25767 + }, + { + "epoch": 0.94, + "learning_rate": 1.0581902076390669e-07, + "loss": 0.6957, + "step": 25768 + }, + { + "epoch": 0.94, + "learning_rate": 1.0581535869923462e-07, + "loss": 0.6938, + "step": 25769 + }, + { + "epoch": 0.94, + "learning_rate": 1.0581169663456256e-07, + "loss": 0.6918, + "step": 25770 + }, + { + "epoch": 0.94, + "learning_rate": 1.058080345698905e-07, + "loss": 0.6849, + "step": 25771 + }, + { + "epoch": 0.94, + "learning_rate": 1.0580437250521842e-07, + "loss": 0.6838, + "step": 25772 + }, + { + "epoch": 0.94, + "learning_rate": 1.0580071044054636e-07, + "loss": 0.6953, + "step": 25773 + }, + { + "epoch": 0.94, + "learning_rate": 1.057970483758743e-07, + "loss": 0.6902, + "step": 25774 + }, + { + "epoch": 0.94, + "learning_rate": 1.0579338631120224e-07, + "loss": 0.6968, + "step": 25775 + }, + { + "epoch": 0.94, + "learning_rate": 1.0578972424653018e-07, + "loss": 0.6886, + "step": 25776 + }, + { + "epoch": 0.94, + "learning_rate": 1.0578606218185812e-07, + "loss": 0.7059, + "step": 25777 + }, + { + "epoch": 0.94, + "learning_rate": 1.0578240011718606e-07, + "loss": 0.6839, + "step": 25778 + }, + { + "epoch": 0.94, + "learning_rate": 1.05778738052514e-07, + "loss": 0.6969, + "step": 25779 + }, + { + "epoch": 0.94, + "learning_rate": 1.0577507598784194e-07, + "loss": 0.698, + "step": 25780 + }, + { + "epoch": 0.94, + "learning_rate": 1.0577141392316988e-07, + "loss": 0.6895, + "step": 25781 + }, + { + "epoch": 0.94, + "learning_rate": 1.0576775185849782e-07, + "loss": 0.6895, + "step": 25782 + }, + { + "epoch": 0.94, + "learning_rate": 1.0576408979382576e-07, + "loss": 0.6836, + "step": 25783 + }, + { + "epoch": 0.94, + "learning_rate": 1.057604277291537e-07, + "loss": 0.6896, + "step": 25784 + }, + { + "epoch": 0.94, + "learning_rate": 1.0575676566448164e-07, + "loss": 0.6889, + "step": 25785 + }, + { + "epoch": 0.94, + "learning_rate": 1.0575310359980958e-07, + "loss": 0.6851, + "step": 25786 + }, + { + "epoch": 0.94, + "learning_rate": 1.057494415351375e-07, + "loss": 0.6928, + "step": 25787 + }, + { + "epoch": 0.94, + "learning_rate": 1.0574577947046544e-07, + "loss": 0.6856, + "step": 25788 + }, + { + "epoch": 0.94, + "learning_rate": 1.0574211740579338e-07, + "loss": 0.6974, + "step": 25789 + }, + { + "epoch": 0.94, + "learning_rate": 1.0573845534112132e-07, + "loss": 0.694, + "step": 25790 + }, + { + "epoch": 0.94, + "learning_rate": 1.0573479327644926e-07, + "loss": 0.6903, + "step": 25791 + }, + { + "epoch": 0.94, + "learning_rate": 1.057311312117772e-07, + "loss": 0.6822, + "step": 25792 + }, + { + "epoch": 0.94, + "learning_rate": 1.0572746914710514e-07, + "loss": 0.6982, + "step": 25793 + }, + { + "epoch": 0.94, + "learning_rate": 1.0572380708243306e-07, + "loss": 0.6828, + "step": 25794 + }, + { + "epoch": 0.94, + "learning_rate": 1.05720145017761e-07, + "loss": 0.695, + "step": 25795 + }, + { + "epoch": 0.94, + "learning_rate": 1.0571648295308894e-07, + "loss": 0.6902, + "step": 25796 + }, + { + "epoch": 0.94, + "learning_rate": 1.0571282088841688e-07, + "loss": 0.6855, + "step": 25797 + }, + { + "epoch": 0.94, + "learning_rate": 1.0570915882374482e-07, + "loss": 0.7058, + "step": 25798 + }, + { + "epoch": 0.94, + "learning_rate": 1.0570549675907276e-07, + "loss": 0.6838, + "step": 25799 + }, + { + "epoch": 0.94, + "learning_rate": 1.057018346944007e-07, + "loss": 0.6967, + "step": 25800 + }, + { + "epoch": 0.94, + "learning_rate": 1.0569817262972864e-07, + "loss": 0.6932, + "step": 25801 + }, + { + "epoch": 0.94, + "learning_rate": 1.0569451056505658e-07, + "loss": 0.6981, + "step": 25802 + }, + { + "epoch": 0.94, + "learning_rate": 1.0569084850038452e-07, + "loss": 0.6844, + "step": 25803 + }, + { + "epoch": 0.94, + "learning_rate": 1.0568718643571246e-07, + "loss": 0.6833, + "step": 25804 + }, + { + "epoch": 0.94, + "learning_rate": 1.0568352437104038e-07, + "loss": 0.687, + "step": 25805 + }, + { + "epoch": 0.94, + "learning_rate": 1.0567986230636832e-07, + "loss": 0.6852, + "step": 25806 + }, + { + "epoch": 0.94, + "learning_rate": 1.0567620024169626e-07, + "loss": 0.6986, + "step": 25807 + }, + { + "epoch": 0.94, + "learning_rate": 1.056725381770242e-07, + "loss": 0.6968, + "step": 25808 + }, + { + "epoch": 0.94, + "learning_rate": 1.0566887611235214e-07, + "loss": 0.6957, + "step": 25809 + }, + { + "epoch": 0.94, + "learning_rate": 1.0566521404768008e-07, + "loss": 0.6923, + "step": 25810 + }, + { + "epoch": 0.94, + "learning_rate": 1.0566155198300802e-07, + "loss": 0.6973, + "step": 25811 + }, + { + "epoch": 0.94, + "learning_rate": 1.0565788991833596e-07, + "loss": 0.7066, + "step": 25812 + }, + { + "epoch": 0.94, + "learning_rate": 1.056542278536639e-07, + "loss": 0.6916, + "step": 25813 + }, + { + "epoch": 0.94, + "learning_rate": 1.0565056578899184e-07, + "loss": 0.6862, + "step": 25814 + }, + { + "epoch": 0.94, + "learning_rate": 1.0564690372431978e-07, + "loss": 0.6929, + "step": 25815 + }, + { + "epoch": 0.94, + "learning_rate": 1.056432416596477e-07, + "loss": 0.6925, + "step": 25816 + }, + { + "epoch": 0.94, + "learning_rate": 1.0563957959497564e-07, + "loss": 0.6892, + "step": 25817 + }, + { + "epoch": 0.94, + "learning_rate": 1.0563591753030358e-07, + "loss": 0.6916, + "step": 25818 + }, + { + "epoch": 0.94, + "learning_rate": 1.0563225546563152e-07, + "loss": 0.6848, + "step": 25819 + }, + { + "epoch": 0.94, + "learning_rate": 1.0562859340095946e-07, + "loss": 0.6967, + "step": 25820 + }, + { + "epoch": 0.94, + "learning_rate": 1.056249313362874e-07, + "loss": 0.6947, + "step": 25821 + }, + { + "epoch": 0.94, + "learning_rate": 1.0562126927161532e-07, + "loss": 0.6824, + "step": 25822 + }, + { + "epoch": 0.94, + "learning_rate": 1.0561760720694326e-07, + "loss": 0.7005, + "step": 25823 + }, + { + "epoch": 0.94, + "learning_rate": 1.056139451422712e-07, + "loss": 0.6939, + "step": 25824 + }, + { + "epoch": 0.94, + "learning_rate": 1.0561028307759914e-07, + "loss": 0.6985, + "step": 25825 + }, + { + "epoch": 0.94, + "learning_rate": 1.0560662101292707e-07, + "loss": 0.7032, + "step": 25826 + }, + { + "epoch": 0.94, + "learning_rate": 1.0560295894825501e-07, + "loss": 0.6993, + "step": 25827 + }, + { + "epoch": 0.94, + "learning_rate": 1.0559929688358295e-07, + "loss": 0.6918, + "step": 25828 + }, + { + "epoch": 0.95, + "learning_rate": 1.055956348189109e-07, + "loss": 0.7073, + "step": 25829 + }, + { + "epoch": 0.95, + "learning_rate": 1.0559197275423883e-07, + "loss": 0.6932, + "step": 25830 + }, + { + "epoch": 0.95, + "learning_rate": 1.0558831068956677e-07, + "loss": 0.6924, + "step": 25831 + }, + { + "epoch": 0.95, + "learning_rate": 1.0558464862489471e-07, + "loss": 0.6963, + "step": 25832 + }, + { + "epoch": 0.95, + "learning_rate": 1.0558098656022265e-07, + "loss": 0.6949, + "step": 25833 + }, + { + "epoch": 0.95, + "learning_rate": 1.055773244955506e-07, + "loss": 0.6963, + "step": 25834 + }, + { + "epoch": 0.95, + "learning_rate": 1.0557366243087853e-07, + "loss": 0.6951, + "step": 25835 + }, + { + "epoch": 0.95, + "learning_rate": 1.0557000036620647e-07, + "loss": 0.7007, + "step": 25836 + }, + { + "epoch": 0.95, + "learning_rate": 1.0556633830153441e-07, + "loss": 0.6826, + "step": 25837 + }, + { + "epoch": 0.95, + "learning_rate": 1.0556267623686234e-07, + "loss": 0.7041, + "step": 25838 + }, + { + "epoch": 0.95, + "learning_rate": 1.0555901417219028e-07, + "loss": 0.6894, + "step": 25839 + }, + { + "epoch": 0.95, + "learning_rate": 1.0555535210751821e-07, + "loss": 0.6933, + "step": 25840 + }, + { + "epoch": 0.95, + "learning_rate": 1.0555169004284615e-07, + "loss": 0.6949, + "step": 25841 + }, + { + "epoch": 0.95, + "learning_rate": 1.0554802797817409e-07, + "loss": 0.6984, + "step": 25842 + }, + { + "epoch": 0.95, + "learning_rate": 1.0554436591350201e-07, + "loss": 0.6886, + "step": 25843 + }, + { + "epoch": 0.95, + "learning_rate": 1.0554070384882995e-07, + "loss": 0.6855, + "step": 25844 + }, + { + "epoch": 0.95, + "learning_rate": 1.055370417841579e-07, + "loss": 0.6855, + "step": 25845 + }, + { + "epoch": 0.95, + "learning_rate": 1.0553337971948583e-07, + "loss": 0.6932, + "step": 25846 + }, + { + "epoch": 0.95, + "learning_rate": 1.0552971765481377e-07, + "loss": 0.6987, + "step": 25847 + }, + { + "epoch": 0.95, + "learning_rate": 1.0552605559014171e-07, + "loss": 0.6937, + "step": 25848 + }, + { + "epoch": 0.95, + "learning_rate": 1.0552239352546965e-07, + "loss": 0.6909, + "step": 25849 + }, + { + "epoch": 0.95, + "learning_rate": 1.055187314607976e-07, + "loss": 0.6967, + "step": 25850 + }, + { + "epoch": 0.95, + "learning_rate": 1.0551506939612553e-07, + "loss": 0.6959, + "step": 25851 + }, + { + "epoch": 0.95, + "learning_rate": 1.0551140733145347e-07, + "loss": 0.6921, + "step": 25852 + }, + { + "epoch": 0.95, + "learning_rate": 1.0550774526678141e-07, + "loss": 0.6936, + "step": 25853 + }, + { + "epoch": 0.95, + "learning_rate": 1.0550408320210935e-07, + "loss": 0.6887, + "step": 25854 + }, + { + "epoch": 0.95, + "learning_rate": 1.0550042113743729e-07, + "loss": 0.7045, + "step": 25855 + }, + { + "epoch": 0.95, + "learning_rate": 1.0549675907276523e-07, + "loss": 0.7014, + "step": 25856 + }, + { + "epoch": 0.95, + "learning_rate": 1.0549309700809317e-07, + "loss": 0.688, + "step": 25857 + }, + { + "epoch": 0.95, + "learning_rate": 1.0548943494342109e-07, + "loss": 0.6889, + "step": 25858 + }, + { + "epoch": 0.95, + "learning_rate": 1.0548577287874903e-07, + "loss": 0.6793, + "step": 25859 + }, + { + "epoch": 0.95, + "learning_rate": 1.0548211081407697e-07, + "loss": 0.6933, + "step": 25860 + }, + { + "epoch": 0.95, + "learning_rate": 1.054784487494049e-07, + "loss": 0.6871, + "step": 25861 + }, + { + "epoch": 0.95, + "learning_rate": 1.0547478668473285e-07, + "loss": 0.6969, + "step": 25862 + }, + { + "epoch": 0.95, + "learning_rate": 1.0547112462006079e-07, + "loss": 0.6835, + "step": 25863 + }, + { + "epoch": 0.95, + "learning_rate": 1.0546746255538873e-07, + "loss": 0.6941, + "step": 25864 + }, + { + "epoch": 0.95, + "learning_rate": 1.0546380049071665e-07, + "loss": 0.7051, + "step": 25865 + }, + { + "epoch": 0.95, + "learning_rate": 1.0546013842604459e-07, + "loss": 0.7022, + "step": 25866 + }, + { + "epoch": 0.95, + "learning_rate": 1.0545647636137253e-07, + "loss": 0.691, + "step": 25867 + }, + { + "epoch": 0.95, + "learning_rate": 1.0545281429670047e-07, + "loss": 0.6859, + "step": 25868 + }, + { + "epoch": 0.95, + "learning_rate": 1.0544915223202841e-07, + "loss": 0.705, + "step": 25869 + }, + { + "epoch": 0.95, + "learning_rate": 1.0544549016735635e-07, + "loss": 0.6871, + "step": 25870 + }, + { + "epoch": 0.95, + "learning_rate": 1.0544182810268429e-07, + "loss": 0.6785, + "step": 25871 + }, + { + "epoch": 0.95, + "learning_rate": 1.0543816603801223e-07, + "loss": 0.6913, + "step": 25872 + }, + { + "epoch": 0.95, + "learning_rate": 1.0543450397334017e-07, + "loss": 0.6961, + "step": 25873 + }, + { + "epoch": 0.95, + "learning_rate": 1.0543084190866811e-07, + "loss": 0.6905, + "step": 25874 + }, + { + "epoch": 0.95, + "learning_rate": 1.0542717984399605e-07, + "loss": 0.711, + "step": 25875 + }, + { + "epoch": 0.95, + "learning_rate": 1.0542351777932397e-07, + "loss": 0.6983, + "step": 25876 + }, + { + "epoch": 0.95, + "learning_rate": 1.054198557146519e-07, + "loss": 0.6935, + "step": 25877 + }, + { + "epoch": 0.95, + "learning_rate": 1.0541619364997985e-07, + "loss": 0.702, + "step": 25878 + }, + { + "epoch": 0.95, + "learning_rate": 1.0541253158530779e-07, + "loss": 0.6912, + "step": 25879 + }, + { + "epoch": 0.95, + "learning_rate": 1.0540886952063573e-07, + "loss": 0.68, + "step": 25880 + }, + { + "epoch": 0.95, + "learning_rate": 1.0540520745596367e-07, + "loss": 0.7031, + "step": 25881 + }, + { + "epoch": 0.95, + "learning_rate": 1.054015453912916e-07, + "loss": 0.7017, + "step": 25882 + }, + { + "epoch": 0.95, + "learning_rate": 1.0539788332661955e-07, + "loss": 0.6878, + "step": 25883 + }, + { + "epoch": 0.95, + "learning_rate": 1.0539422126194748e-07, + "loss": 0.6927, + "step": 25884 + }, + { + "epoch": 0.95, + "learning_rate": 1.0539055919727542e-07, + "loss": 0.6965, + "step": 25885 + }, + { + "epoch": 0.95, + "learning_rate": 1.0538689713260336e-07, + "loss": 0.6764, + "step": 25886 + }, + { + "epoch": 0.95, + "learning_rate": 1.0538323506793129e-07, + "loss": 0.7012, + "step": 25887 + }, + { + "epoch": 0.95, + "learning_rate": 1.0537957300325923e-07, + "loss": 0.697, + "step": 25888 + }, + { + "epoch": 0.95, + "learning_rate": 1.0537591093858717e-07, + "loss": 0.7072, + "step": 25889 + }, + { + "epoch": 0.95, + "learning_rate": 1.0537224887391511e-07, + "loss": 0.6936, + "step": 25890 + }, + { + "epoch": 0.95, + "learning_rate": 1.0536858680924305e-07, + "loss": 0.6973, + "step": 25891 + }, + { + "epoch": 0.95, + "learning_rate": 1.0536492474457099e-07, + "loss": 0.708, + "step": 25892 + }, + { + "epoch": 0.95, + "learning_rate": 1.0536126267989893e-07, + "loss": 0.6895, + "step": 25893 + }, + { + "epoch": 0.95, + "learning_rate": 1.0535760061522684e-07, + "loss": 0.6902, + "step": 25894 + }, + { + "epoch": 0.95, + "learning_rate": 1.0535393855055478e-07, + "loss": 0.6892, + "step": 25895 + }, + { + "epoch": 0.95, + "learning_rate": 1.0535027648588272e-07, + "loss": 0.6972, + "step": 25896 + }, + { + "epoch": 0.95, + "learning_rate": 1.0534661442121066e-07, + "loss": 0.6873, + "step": 25897 + }, + { + "epoch": 0.95, + "learning_rate": 1.053429523565386e-07, + "loss": 0.6976, + "step": 25898 + }, + { + "epoch": 0.95, + "learning_rate": 1.0533929029186654e-07, + "loss": 0.7016, + "step": 25899 + }, + { + "epoch": 0.95, + "learning_rate": 1.0533562822719448e-07, + "loss": 0.6909, + "step": 25900 + }, + { + "epoch": 0.95, + "learning_rate": 1.0533196616252242e-07, + "loss": 0.6908, + "step": 25901 + }, + { + "epoch": 0.95, + "learning_rate": 1.0532830409785036e-07, + "loss": 0.6855, + "step": 25902 + }, + { + "epoch": 0.95, + "learning_rate": 1.053246420331783e-07, + "loss": 0.6823, + "step": 25903 + }, + { + "epoch": 0.95, + "learning_rate": 1.0532097996850624e-07, + "loss": 0.7007, + "step": 25904 + }, + { + "epoch": 0.95, + "learning_rate": 1.0531731790383418e-07, + "loss": 0.7017, + "step": 25905 + }, + { + "epoch": 0.95, + "learning_rate": 1.0531365583916212e-07, + "loss": 0.688, + "step": 25906 + }, + { + "epoch": 0.95, + "learning_rate": 1.0530999377449006e-07, + "loss": 0.6986, + "step": 25907 + }, + { + "epoch": 0.95, + "learning_rate": 1.05306331709818e-07, + "loss": 0.6969, + "step": 25908 + }, + { + "epoch": 0.95, + "learning_rate": 1.0530266964514593e-07, + "loss": 0.6976, + "step": 25909 + }, + { + "epoch": 0.95, + "learning_rate": 1.0529900758047387e-07, + "loss": 0.7074, + "step": 25910 + }, + { + "epoch": 0.95, + "learning_rate": 1.0529534551580181e-07, + "loss": 0.6972, + "step": 25911 + }, + { + "epoch": 0.95, + "learning_rate": 1.0529168345112974e-07, + "loss": 0.6878, + "step": 25912 + }, + { + "epoch": 0.95, + "learning_rate": 1.0528802138645768e-07, + "loss": 0.6862, + "step": 25913 + }, + { + "epoch": 0.95, + "learning_rate": 1.052843593217856e-07, + "loss": 0.7034, + "step": 25914 + }, + { + "epoch": 0.95, + "learning_rate": 1.0528069725711354e-07, + "loss": 0.7014, + "step": 25915 + }, + { + "epoch": 0.95, + "learning_rate": 1.0527703519244148e-07, + "loss": 0.7059, + "step": 25916 + }, + { + "epoch": 0.95, + "learning_rate": 1.0527337312776942e-07, + "loss": 0.6856, + "step": 25917 + }, + { + "epoch": 0.95, + "learning_rate": 1.0526971106309736e-07, + "loss": 0.6963, + "step": 25918 + }, + { + "epoch": 0.95, + "learning_rate": 1.052660489984253e-07, + "loss": 0.7002, + "step": 25919 + }, + { + "epoch": 0.95, + "learning_rate": 1.0526238693375324e-07, + "loss": 0.6896, + "step": 25920 + }, + { + "epoch": 0.95, + "learning_rate": 1.0525872486908118e-07, + "loss": 0.6914, + "step": 25921 + }, + { + "epoch": 0.95, + "learning_rate": 1.0525506280440912e-07, + "loss": 0.7031, + "step": 25922 + }, + { + "epoch": 0.95, + "learning_rate": 1.0525140073973706e-07, + "loss": 0.6995, + "step": 25923 + }, + { + "epoch": 0.95, + "learning_rate": 1.05247738675065e-07, + "loss": 0.6987, + "step": 25924 + }, + { + "epoch": 0.95, + "learning_rate": 1.0524407661039294e-07, + "loss": 0.6863, + "step": 25925 + }, + { + "epoch": 0.95, + "learning_rate": 1.0524041454572088e-07, + "loss": 0.6924, + "step": 25926 + }, + { + "epoch": 0.95, + "learning_rate": 1.0523675248104882e-07, + "loss": 0.6824, + "step": 25927 + }, + { + "epoch": 0.95, + "learning_rate": 1.0523309041637676e-07, + "loss": 0.6998, + "step": 25928 + }, + { + "epoch": 0.95, + "learning_rate": 1.052294283517047e-07, + "loss": 0.6967, + "step": 25929 + }, + { + "epoch": 0.95, + "learning_rate": 1.0522576628703262e-07, + "loss": 0.7012, + "step": 25930 + }, + { + "epoch": 0.95, + "learning_rate": 1.0522210422236056e-07, + "loss": 0.699, + "step": 25931 + }, + { + "epoch": 0.95, + "learning_rate": 1.052184421576885e-07, + "loss": 0.6991, + "step": 25932 + }, + { + "epoch": 0.95, + "learning_rate": 1.0521478009301644e-07, + "loss": 0.6972, + "step": 25933 + }, + { + "epoch": 0.95, + "learning_rate": 1.0521111802834438e-07, + "loss": 0.6945, + "step": 25934 + }, + { + "epoch": 0.95, + "learning_rate": 1.0520745596367232e-07, + "loss": 0.691, + "step": 25935 + }, + { + "epoch": 0.95, + "learning_rate": 1.0520379389900024e-07, + "loss": 0.6955, + "step": 25936 + }, + { + "epoch": 0.95, + "learning_rate": 1.0520013183432818e-07, + "loss": 0.6806, + "step": 25937 + }, + { + "epoch": 0.95, + "learning_rate": 1.0519646976965612e-07, + "loss": 0.6865, + "step": 25938 + }, + { + "epoch": 0.95, + "learning_rate": 1.0519280770498406e-07, + "loss": 0.6907, + "step": 25939 + }, + { + "epoch": 0.95, + "learning_rate": 1.05189145640312e-07, + "loss": 0.6899, + "step": 25940 + }, + { + "epoch": 0.95, + "learning_rate": 1.0518548357563994e-07, + "loss": 0.6883, + "step": 25941 + }, + { + "epoch": 0.95, + "learning_rate": 1.0518182151096788e-07, + "loss": 0.6942, + "step": 25942 + }, + { + "epoch": 0.95, + "learning_rate": 1.0517815944629582e-07, + "loss": 0.7041, + "step": 25943 + }, + { + "epoch": 0.95, + "learning_rate": 1.0517449738162376e-07, + "loss": 0.6901, + "step": 25944 + }, + { + "epoch": 0.95, + "learning_rate": 1.051708353169517e-07, + "loss": 0.6972, + "step": 25945 + }, + { + "epoch": 0.95, + "learning_rate": 1.0516717325227964e-07, + "loss": 0.6933, + "step": 25946 + }, + { + "epoch": 0.95, + "learning_rate": 1.0516351118760758e-07, + "loss": 0.7023, + "step": 25947 + }, + { + "epoch": 0.95, + "learning_rate": 1.051598491229355e-07, + "loss": 0.6843, + "step": 25948 + }, + { + "epoch": 0.95, + "learning_rate": 1.0515618705826343e-07, + "loss": 0.6863, + "step": 25949 + }, + { + "epoch": 0.95, + "learning_rate": 1.0515252499359137e-07, + "loss": 0.6911, + "step": 25950 + }, + { + "epoch": 0.95, + "learning_rate": 1.0514886292891931e-07, + "loss": 0.691, + "step": 25951 + }, + { + "epoch": 0.95, + "learning_rate": 1.0514520086424725e-07, + "loss": 0.6874, + "step": 25952 + }, + { + "epoch": 0.95, + "learning_rate": 1.051415387995752e-07, + "loss": 0.6899, + "step": 25953 + }, + { + "epoch": 0.95, + "learning_rate": 1.0513787673490313e-07, + "loss": 0.6859, + "step": 25954 + }, + { + "epoch": 0.95, + "learning_rate": 1.0513421467023107e-07, + "loss": 0.6881, + "step": 25955 + }, + { + "epoch": 0.95, + "learning_rate": 1.0513055260555901e-07, + "loss": 0.6824, + "step": 25956 + }, + { + "epoch": 0.95, + "learning_rate": 1.0512689054088695e-07, + "loss": 0.6863, + "step": 25957 + }, + { + "epoch": 0.95, + "learning_rate": 1.0512322847621488e-07, + "loss": 0.6827, + "step": 25958 + }, + { + "epoch": 0.95, + "learning_rate": 1.0511956641154282e-07, + "loss": 0.7019, + "step": 25959 + }, + { + "epoch": 0.95, + "learning_rate": 1.0511590434687076e-07, + "loss": 0.6924, + "step": 25960 + }, + { + "epoch": 0.95, + "learning_rate": 1.051122422821987e-07, + "loss": 0.6884, + "step": 25961 + }, + { + "epoch": 0.95, + "learning_rate": 1.0510858021752664e-07, + "loss": 0.6884, + "step": 25962 + }, + { + "epoch": 0.95, + "learning_rate": 1.0510491815285458e-07, + "loss": 0.6964, + "step": 25963 + }, + { + "epoch": 0.95, + "learning_rate": 1.0510125608818252e-07, + "loss": 0.6908, + "step": 25964 + }, + { + "epoch": 0.95, + "learning_rate": 1.0509759402351046e-07, + "loss": 0.6927, + "step": 25965 + }, + { + "epoch": 0.95, + "learning_rate": 1.0509393195883837e-07, + "loss": 0.6948, + "step": 25966 + }, + { + "epoch": 0.95, + "learning_rate": 1.0509026989416631e-07, + "loss": 0.6896, + "step": 25967 + }, + { + "epoch": 0.95, + "learning_rate": 1.0508660782949425e-07, + "loss": 0.692, + "step": 25968 + }, + { + "epoch": 0.95, + "learning_rate": 1.050829457648222e-07, + "loss": 0.6994, + "step": 25969 + }, + { + "epoch": 0.95, + "learning_rate": 1.0507928370015013e-07, + "loss": 0.6893, + "step": 25970 + }, + { + "epoch": 0.95, + "learning_rate": 1.0507562163547807e-07, + "loss": 0.7003, + "step": 25971 + }, + { + "epoch": 0.95, + "learning_rate": 1.0507195957080601e-07, + "loss": 0.6934, + "step": 25972 + }, + { + "epoch": 0.95, + "learning_rate": 1.0506829750613395e-07, + "loss": 0.6986, + "step": 25973 + }, + { + "epoch": 0.95, + "learning_rate": 1.0506463544146189e-07, + "loss": 0.6938, + "step": 25974 + }, + { + "epoch": 0.95, + "learning_rate": 1.0506097337678983e-07, + "loss": 0.7013, + "step": 25975 + }, + { + "epoch": 0.95, + "learning_rate": 1.0505731131211777e-07, + "loss": 0.6916, + "step": 25976 + }, + { + "epoch": 0.95, + "learning_rate": 1.0505364924744571e-07, + "loss": 0.7002, + "step": 25977 + }, + { + "epoch": 0.95, + "learning_rate": 1.0504998718277365e-07, + "loss": 0.6855, + "step": 25978 + }, + { + "epoch": 0.95, + "learning_rate": 1.0504632511810159e-07, + "loss": 0.6954, + "step": 25979 + }, + { + "epoch": 0.95, + "learning_rate": 1.0504266305342952e-07, + "loss": 0.688, + "step": 25980 + }, + { + "epoch": 0.95, + "learning_rate": 1.0503900098875746e-07, + "loss": 0.6778, + "step": 25981 + }, + { + "epoch": 0.95, + "learning_rate": 1.050353389240854e-07, + "loss": 0.6892, + "step": 25982 + }, + { + "epoch": 0.95, + "learning_rate": 1.0503167685941334e-07, + "loss": 0.6914, + "step": 25983 + }, + { + "epoch": 0.95, + "learning_rate": 1.0502801479474127e-07, + "loss": 0.681, + "step": 25984 + }, + { + "epoch": 0.95, + "learning_rate": 1.0502435273006919e-07, + "loss": 0.6864, + "step": 25985 + }, + { + "epoch": 0.95, + "learning_rate": 1.0502069066539713e-07, + "loss": 0.7031, + "step": 25986 + }, + { + "epoch": 0.95, + "learning_rate": 1.0501702860072507e-07, + "loss": 0.6763, + "step": 25987 + }, + { + "epoch": 0.95, + "learning_rate": 1.0501336653605301e-07, + "loss": 0.6821, + "step": 25988 + }, + { + "epoch": 0.95, + "learning_rate": 1.0500970447138095e-07, + "loss": 0.7035, + "step": 25989 + }, + { + "epoch": 0.95, + "learning_rate": 1.0500604240670889e-07, + "loss": 0.6914, + "step": 25990 + }, + { + "epoch": 0.95, + "learning_rate": 1.0500238034203683e-07, + "loss": 0.6986, + "step": 25991 + }, + { + "epoch": 0.95, + "learning_rate": 1.0499871827736477e-07, + "loss": 0.6856, + "step": 25992 + }, + { + "epoch": 0.95, + "learning_rate": 1.0499505621269271e-07, + "loss": 0.6974, + "step": 25993 + }, + { + "epoch": 0.95, + "learning_rate": 1.0499139414802065e-07, + "loss": 0.7024, + "step": 25994 + }, + { + "epoch": 0.95, + "learning_rate": 1.0498773208334859e-07, + "loss": 0.6963, + "step": 25995 + }, + { + "epoch": 0.95, + "learning_rate": 1.0498407001867653e-07, + "loss": 0.7037, + "step": 25996 + }, + { + "epoch": 0.95, + "learning_rate": 1.0498040795400447e-07, + "loss": 0.6834, + "step": 25997 + }, + { + "epoch": 0.95, + "learning_rate": 1.0497674588933241e-07, + "loss": 0.6903, + "step": 25998 + }, + { + "epoch": 0.95, + "learning_rate": 1.0497308382466035e-07, + "loss": 0.7011, + "step": 25999 + }, + { + "epoch": 0.95, + "learning_rate": 1.0496942175998829e-07, + "loss": 0.6952, + "step": 26000 + }, + { + "epoch": 0.95, + "learning_rate": 1.0496575969531623e-07, + "loss": 0.7049, + "step": 26001 + }, + { + "epoch": 0.95, + "learning_rate": 1.0496209763064415e-07, + "loss": 0.7025, + "step": 26002 + }, + { + "epoch": 0.95, + "learning_rate": 1.0495843556597208e-07, + "loss": 0.6866, + "step": 26003 + }, + { + "epoch": 0.95, + "learning_rate": 1.0495477350130002e-07, + "loss": 0.7029, + "step": 26004 + }, + { + "epoch": 0.95, + "learning_rate": 1.0495111143662796e-07, + "loss": 0.6846, + "step": 26005 + }, + { + "epoch": 0.95, + "learning_rate": 1.049474493719559e-07, + "loss": 0.6879, + "step": 26006 + }, + { + "epoch": 0.95, + "learning_rate": 1.0494378730728383e-07, + "loss": 0.691, + "step": 26007 + }, + { + "epoch": 0.95, + "learning_rate": 1.0494012524261177e-07, + "loss": 0.6903, + "step": 26008 + }, + { + "epoch": 0.95, + "learning_rate": 1.0493646317793971e-07, + "loss": 0.6858, + "step": 26009 + }, + { + "epoch": 0.95, + "learning_rate": 1.0493280111326765e-07, + "loss": 0.6783, + "step": 26010 + }, + { + "epoch": 0.95, + "learning_rate": 1.0492913904859559e-07, + "loss": 0.7212, + "step": 26011 + }, + { + "epoch": 0.95, + "learning_rate": 1.0492547698392353e-07, + "loss": 0.6982, + "step": 26012 + }, + { + "epoch": 0.95, + "learning_rate": 1.0492181491925147e-07, + "loss": 0.7002, + "step": 26013 + }, + { + "epoch": 0.95, + "learning_rate": 1.0491815285457941e-07, + "loss": 0.6854, + "step": 26014 + }, + { + "epoch": 0.95, + "learning_rate": 1.0491449078990735e-07, + "loss": 0.694, + "step": 26015 + }, + { + "epoch": 0.95, + "learning_rate": 1.0491082872523529e-07, + "loss": 0.7046, + "step": 26016 + }, + { + "epoch": 0.95, + "learning_rate": 1.0490716666056323e-07, + "loss": 0.6798, + "step": 26017 + }, + { + "epoch": 0.95, + "learning_rate": 1.0490350459589117e-07, + "loss": 0.6882, + "step": 26018 + }, + { + "epoch": 0.95, + "learning_rate": 1.0489984253121911e-07, + "loss": 0.6944, + "step": 26019 + }, + { + "epoch": 0.95, + "learning_rate": 1.0489618046654702e-07, + "loss": 0.6982, + "step": 26020 + }, + { + "epoch": 0.95, + "learning_rate": 1.0489251840187496e-07, + "loss": 0.6995, + "step": 26021 + }, + { + "epoch": 0.95, + "learning_rate": 1.048888563372029e-07, + "loss": 0.6891, + "step": 26022 + }, + { + "epoch": 0.95, + "learning_rate": 1.0488519427253084e-07, + "loss": 0.7011, + "step": 26023 + }, + { + "epoch": 0.95, + "learning_rate": 1.0488153220785878e-07, + "loss": 0.6876, + "step": 26024 + }, + { + "epoch": 0.95, + "learning_rate": 1.0487787014318672e-07, + "loss": 0.6858, + "step": 26025 + }, + { + "epoch": 0.95, + "learning_rate": 1.0487420807851466e-07, + "loss": 0.7024, + "step": 26026 + }, + { + "epoch": 0.95, + "learning_rate": 1.048705460138426e-07, + "loss": 0.7013, + "step": 26027 + }, + { + "epoch": 0.95, + "learning_rate": 1.0486688394917054e-07, + "loss": 0.694, + "step": 26028 + }, + { + "epoch": 0.95, + "learning_rate": 1.0486322188449847e-07, + "loss": 0.7031, + "step": 26029 + }, + { + "epoch": 0.95, + "learning_rate": 1.0485955981982641e-07, + "loss": 0.6931, + "step": 26030 + }, + { + "epoch": 0.95, + "learning_rate": 1.0485589775515435e-07, + "loss": 0.6953, + "step": 26031 + }, + { + "epoch": 0.95, + "learning_rate": 1.0485223569048229e-07, + "loss": 0.6995, + "step": 26032 + }, + { + "epoch": 0.95, + "learning_rate": 1.0484857362581023e-07, + "loss": 0.6976, + "step": 26033 + }, + { + "epoch": 0.95, + "learning_rate": 1.0484491156113817e-07, + "loss": 0.6876, + "step": 26034 + }, + { + "epoch": 0.95, + "learning_rate": 1.0484124949646611e-07, + "loss": 0.6957, + "step": 26035 + }, + { + "epoch": 0.95, + "learning_rate": 1.0483758743179405e-07, + "loss": 0.6901, + "step": 26036 + }, + { + "epoch": 0.95, + "learning_rate": 1.0483392536712199e-07, + "loss": 0.6956, + "step": 26037 + }, + { + "epoch": 0.95, + "learning_rate": 1.048302633024499e-07, + "loss": 0.6959, + "step": 26038 + }, + { + "epoch": 0.95, + "learning_rate": 1.0482660123777784e-07, + "loss": 0.6956, + "step": 26039 + }, + { + "epoch": 0.95, + "learning_rate": 1.0482293917310578e-07, + "loss": 0.6962, + "step": 26040 + }, + { + "epoch": 0.95, + "learning_rate": 1.0481927710843372e-07, + "loss": 0.6856, + "step": 26041 + }, + { + "epoch": 0.95, + "learning_rate": 1.0481561504376166e-07, + "loss": 0.6973, + "step": 26042 + }, + { + "epoch": 0.95, + "learning_rate": 1.048119529790896e-07, + "loss": 0.6889, + "step": 26043 + }, + { + "epoch": 0.95, + "learning_rate": 1.0480829091441754e-07, + "loss": 0.6806, + "step": 26044 + }, + { + "epoch": 0.95, + "learning_rate": 1.0480462884974548e-07, + "loss": 0.7043, + "step": 26045 + }, + { + "epoch": 0.95, + "learning_rate": 1.0480096678507342e-07, + "loss": 0.6973, + "step": 26046 + }, + { + "epoch": 0.95, + "learning_rate": 1.0479730472040136e-07, + "loss": 0.6928, + "step": 26047 + }, + { + "epoch": 0.95, + "learning_rate": 1.047936426557293e-07, + "loss": 0.6973, + "step": 26048 + }, + { + "epoch": 0.95, + "learning_rate": 1.0478998059105724e-07, + "loss": 0.6969, + "step": 26049 + }, + { + "epoch": 0.95, + "learning_rate": 1.0478631852638518e-07, + "loss": 0.6943, + "step": 26050 + }, + { + "epoch": 0.95, + "learning_rate": 1.0478265646171311e-07, + "loss": 0.6901, + "step": 26051 + }, + { + "epoch": 0.95, + "learning_rate": 1.0477899439704105e-07, + "loss": 0.692, + "step": 26052 + }, + { + "epoch": 0.95, + "learning_rate": 1.0477533233236899e-07, + "loss": 0.6951, + "step": 26053 + }, + { + "epoch": 0.95, + "learning_rate": 1.0477167026769693e-07, + "loss": 0.6936, + "step": 26054 + }, + { + "epoch": 0.95, + "learning_rate": 1.0476800820302487e-07, + "loss": 0.692, + "step": 26055 + }, + { + "epoch": 0.95, + "learning_rate": 1.0476434613835278e-07, + "loss": 0.696, + "step": 26056 + }, + { + "epoch": 0.95, + "learning_rate": 1.0476068407368072e-07, + "loss": 0.6932, + "step": 26057 + }, + { + "epoch": 0.95, + "learning_rate": 1.0475702200900866e-07, + "loss": 0.6845, + "step": 26058 + }, + { + "epoch": 0.95, + "learning_rate": 1.047533599443366e-07, + "loss": 0.7087, + "step": 26059 + }, + { + "epoch": 0.95, + "learning_rate": 1.0474969787966454e-07, + "loss": 0.6878, + "step": 26060 + }, + { + "epoch": 0.95, + "learning_rate": 1.0474603581499248e-07, + "loss": 0.687, + "step": 26061 + }, + { + "epoch": 0.95, + "learning_rate": 1.0474237375032042e-07, + "loss": 0.6871, + "step": 26062 + }, + { + "epoch": 0.95, + "learning_rate": 1.0473871168564836e-07, + "loss": 0.6965, + "step": 26063 + }, + { + "epoch": 0.95, + "learning_rate": 1.047350496209763e-07, + "loss": 0.6977, + "step": 26064 + }, + { + "epoch": 0.95, + "learning_rate": 1.0473138755630424e-07, + "loss": 0.6868, + "step": 26065 + }, + { + "epoch": 0.95, + "learning_rate": 1.0472772549163218e-07, + "loss": 0.7023, + "step": 26066 + }, + { + "epoch": 0.95, + "learning_rate": 1.0472406342696012e-07, + "loss": 0.7011, + "step": 26067 + }, + { + "epoch": 0.95, + "learning_rate": 1.0472040136228806e-07, + "loss": 0.6855, + "step": 26068 + }, + { + "epoch": 0.95, + "learning_rate": 1.04716739297616e-07, + "loss": 0.6941, + "step": 26069 + }, + { + "epoch": 0.95, + "learning_rate": 1.0471307723294394e-07, + "loss": 0.6938, + "step": 26070 + }, + { + "epoch": 0.95, + "learning_rate": 1.0470941516827188e-07, + "loss": 0.692, + "step": 26071 + }, + { + "epoch": 0.95, + "learning_rate": 1.0470575310359981e-07, + "loss": 0.7058, + "step": 26072 + }, + { + "epoch": 0.95, + "learning_rate": 1.0470209103892775e-07, + "loss": 0.6913, + "step": 26073 + }, + { + "epoch": 0.95, + "learning_rate": 1.0469842897425567e-07, + "loss": 0.6928, + "step": 26074 + }, + { + "epoch": 0.95, + "learning_rate": 1.0469476690958361e-07, + "loss": 0.695, + "step": 26075 + }, + { + "epoch": 0.95, + "learning_rate": 1.0469110484491155e-07, + "loss": 0.689, + "step": 26076 + }, + { + "epoch": 0.95, + "learning_rate": 1.046874427802395e-07, + "loss": 0.6931, + "step": 26077 + }, + { + "epoch": 0.95, + "learning_rate": 1.0468378071556742e-07, + "loss": 0.6987, + "step": 26078 + }, + { + "epoch": 0.95, + "learning_rate": 1.0468011865089536e-07, + "loss": 0.7074, + "step": 26079 + }, + { + "epoch": 0.95, + "learning_rate": 1.046764565862233e-07, + "loss": 0.6888, + "step": 26080 + }, + { + "epoch": 0.95, + "learning_rate": 1.0467279452155124e-07, + "loss": 0.6969, + "step": 26081 + }, + { + "epoch": 0.95, + "learning_rate": 1.0466913245687918e-07, + "loss": 0.7008, + "step": 26082 + }, + { + "epoch": 0.95, + "learning_rate": 1.0466547039220712e-07, + "loss": 0.698, + "step": 26083 + }, + { + "epoch": 0.95, + "learning_rate": 1.0466180832753506e-07, + "loss": 0.6924, + "step": 26084 + }, + { + "epoch": 0.95, + "learning_rate": 1.04658146262863e-07, + "loss": 0.7061, + "step": 26085 + }, + { + "epoch": 0.95, + "learning_rate": 1.0465448419819094e-07, + "loss": 0.6891, + "step": 26086 + }, + { + "epoch": 0.95, + "learning_rate": 1.0465082213351888e-07, + "loss": 0.7056, + "step": 26087 + }, + { + "epoch": 0.95, + "learning_rate": 1.0464716006884682e-07, + "loss": 0.6828, + "step": 26088 + }, + { + "epoch": 0.95, + "learning_rate": 1.0464349800417476e-07, + "loss": 0.6818, + "step": 26089 + }, + { + "epoch": 0.95, + "learning_rate": 1.046398359395027e-07, + "loss": 0.7065, + "step": 26090 + }, + { + "epoch": 0.95, + "learning_rate": 1.0463617387483064e-07, + "loss": 0.6948, + "step": 26091 + }, + { + "epoch": 0.95, + "learning_rate": 1.0463251181015855e-07, + "loss": 0.6975, + "step": 26092 + }, + { + "epoch": 0.95, + "learning_rate": 1.0462884974548649e-07, + "loss": 0.6884, + "step": 26093 + }, + { + "epoch": 0.95, + "learning_rate": 1.0462518768081443e-07, + "loss": 0.6933, + "step": 26094 + }, + { + "epoch": 0.95, + "learning_rate": 1.0462152561614237e-07, + "loss": 0.691, + "step": 26095 + }, + { + "epoch": 0.95, + "learning_rate": 1.0461786355147031e-07, + "loss": 0.6983, + "step": 26096 + }, + { + "epoch": 0.95, + "learning_rate": 1.0461420148679825e-07, + "loss": 0.6908, + "step": 26097 + }, + { + "epoch": 0.95, + "learning_rate": 1.0461053942212619e-07, + "loss": 0.6863, + "step": 26098 + }, + { + "epoch": 0.95, + "learning_rate": 1.0460687735745413e-07, + "loss": 0.6943, + "step": 26099 + }, + { + "epoch": 0.95, + "learning_rate": 1.0460321529278206e-07, + "loss": 0.6853, + "step": 26100 + }, + { + "epoch": 0.95, + "learning_rate": 1.0459955322811e-07, + "loss": 0.6894, + "step": 26101 + }, + { + "epoch": 0.95, + "learning_rate": 1.0459589116343794e-07, + "loss": 0.6788, + "step": 26102 + }, + { + "epoch": 0.96, + "learning_rate": 1.0459222909876588e-07, + "loss": 0.6993, + "step": 26103 + }, + { + "epoch": 0.96, + "learning_rate": 1.0458856703409382e-07, + "loss": 0.701, + "step": 26104 + }, + { + "epoch": 0.96, + "learning_rate": 1.0458490496942176e-07, + "loss": 0.6938, + "step": 26105 + }, + { + "epoch": 0.96, + "learning_rate": 1.045812429047497e-07, + "loss": 0.6976, + "step": 26106 + }, + { + "epoch": 0.96, + "learning_rate": 1.0457758084007764e-07, + "loss": 0.6841, + "step": 26107 + }, + { + "epoch": 0.96, + "learning_rate": 1.0457391877540558e-07, + "loss": 0.6846, + "step": 26108 + }, + { + "epoch": 0.96, + "learning_rate": 1.0457025671073352e-07, + "loss": 0.6867, + "step": 26109 + }, + { + "epoch": 0.96, + "learning_rate": 1.0456659464606143e-07, + "loss": 0.685, + "step": 26110 + }, + { + "epoch": 0.96, + "learning_rate": 1.0456293258138937e-07, + "loss": 0.6957, + "step": 26111 + }, + { + "epoch": 0.96, + "learning_rate": 1.0455927051671731e-07, + "loss": 0.6963, + "step": 26112 + }, + { + "epoch": 0.96, + "learning_rate": 1.0455560845204525e-07, + "loss": 0.6921, + "step": 26113 + }, + { + "epoch": 0.96, + "learning_rate": 1.0455194638737319e-07, + "loss": 0.696, + "step": 26114 + }, + { + "epoch": 0.96, + "learning_rate": 1.0454828432270113e-07, + "loss": 0.7121, + "step": 26115 + }, + { + "epoch": 0.96, + "learning_rate": 1.0454462225802907e-07, + "loss": 0.6909, + "step": 26116 + }, + { + "epoch": 0.96, + "learning_rate": 1.0454096019335701e-07, + "loss": 0.6939, + "step": 26117 + }, + { + "epoch": 0.96, + "learning_rate": 1.0453729812868495e-07, + "loss": 0.7013, + "step": 26118 + }, + { + "epoch": 0.96, + "learning_rate": 1.0453363606401289e-07, + "loss": 0.6914, + "step": 26119 + }, + { + "epoch": 0.96, + "learning_rate": 1.0452997399934083e-07, + "loss": 0.695, + "step": 26120 + }, + { + "epoch": 0.96, + "learning_rate": 1.0452631193466877e-07, + "loss": 0.6884, + "step": 26121 + }, + { + "epoch": 0.96, + "learning_rate": 1.045226498699967e-07, + "loss": 0.7008, + "step": 26122 + }, + { + "epoch": 0.96, + "learning_rate": 1.0451898780532464e-07, + "loss": 0.6941, + "step": 26123 + }, + { + "epoch": 0.96, + "learning_rate": 1.0451532574065258e-07, + "loss": 0.6978, + "step": 26124 + }, + { + "epoch": 0.96, + "learning_rate": 1.0451166367598052e-07, + "loss": 0.6996, + "step": 26125 + }, + { + "epoch": 0.96, + "learning_rate": 1.0450800161130846e-07, + "loss": 0.6999, + "step": 26126 + }, + { + "epoch": 0.96, + "learning_rate": 1.045043395466364e-07, + "loss": 0.6901, + "step": 26127 + }, + { + "epoch": 0.96, + "learning_rate": 1.0450067748196431e-07, + "loss": 0.6898, + "step": 26128 + }, + { + "epoch": 0.96, + "learning_rate": 1.0449701541729225e-07, + "loss": 0.6945, + "step": 26129 + }, + { + "epoch": 0.96, + "learning_rate": 1.0449335335262019e-07, + "loss": 0.6918, + "step": 26130 + }, + { + "epoch": 0.96, + "learning_rate": 1.0448969128794813e-07, + "loss": 0.6924, + "step": 26131 + }, + { + "epoch": 0.96, + "learning_rate": 1.0448602922327607e-07, + "loss": 0.6917, + "step": 26132 + }, + { + "epoch": 0.96, + "learning_rate": 1.0448236715860401e-07, + "loss": 0.6959, + "step": 26133 + }, + { + "epoch": 0.96, + "learning_rate": 1.0447870509393195e-07, + "loss": 0.6851, + "step": 26134 + }, + { + "epoch": 0.96, + "learning_rate": 1.0447504302925989e-07, + "loss": 0.6892, + "step": 26135 + }, + { + "epoch": 0.96, + "learning_rate": 1.0447138096458783e-07, + "loss": 0.698, + "step": 26136 + }, + { + "epoch": 0.96, + "learning_rate": 1.0446771889991577e-07, + "loss": 0.6749, + "step": 26137 + }, + { + "epoch": 0.96, + "learning_rate": 1.0446405683524371e-07, + "loss": 0.7011, + "step": 26138 + }, + { + "epoch": 0.96, + "learning_rate": 1.0446039477057165e-07, + "loss": 0.7011, + "step": 26139 + }, + { + "epoch": 0.96, + "learning_rate": 1.0445673270589959e-07, + "loss": 0.6994, + "step": 26140 + }, + { + "epoch": 0.96, + "learning_rate": 1.0445307064122753e-07, + "loss": 0.7019, + "step": 26141 + }, + { + "epoch": 0.96, + "learning_rate": 1.0444940857655547e-07, + "loss": 0.6955, + "step": 26142 + }, + { + "epoch": 0.96, + "learning_rate": 1.044457465118834e-07, + "loss": 0.6973, + "step": 26143 + }, + { + "epoch": 0.96, + "learning_rate": 1.0444208444721134e-07, + "loss": 0.7014, + "step": 26144 + }, + { + "epoch": 0.96, + "learning_rate": 1.0443842238253928e-07, + "loss": 0.6982, + "step": 26145 + }, + { + "epoch": 0.96, + "learning_rate": 1.044347603178672e-07, + "loss": 0.6904, + "step": 26146 + }, + { + "epoch": 0.96, + "learning_rate": 1.0443109825319514e-07, + "loss": 0.6908, + "step": 26147 + }, + { + "epoch": 0.96, + "learning_rate": 1.0442743618852308e-07, + "loss": 0.6984, + "step": 26148 + }, + { + "epoch": 0.96, + "learning_rate": 1.0442377412385101e-07, + "loss": 0.7024, + "step": 26149 + }, + { + "epoch": 0.96, + "learning_rate": 1.0442011205917895e-07, + "loss": 0.6934, + "step": 26150 + }, + { + "epoch": 0.96, + "learning_rate": 1.0441644999450689e-07, + "loss": 0.6911, + "step": 26151 + }, + { + "epoch": 0.96, + "learning_rate": 1.0441278792983483e-07, + "loss": 0.6991, + "step": 26152 + }, + { + "epoch": 0.96, + "learning_rate": 1.0440912586516277e-07, + "loss": 0.689, + "step": 26153 + }, + { + "epoch": 0.96, + "learning_rate": 1.0440546380049071e-07, + "loss": 0.6883, + "step": 26154 + }, + { + "epoch": 0.96, + "learning_rate": 1.0440180173581865e-07, + "loss": 0.6871, + "step": 26155 + }, + { + "epoch": 0.96, + "learning_rate": 1.0439813967114659e-07, + "loss": 0.6875, + "step": 26156 + }, + { + "epoch": 0.96, + "learning_rate": 1.0439447760647453e-07, + "loss": 0.6909, + "step": 26157 + }, + { + "epoch": 0.96, + "learning_rate": 1.0439081554180247e-07, + "loss": 0.6799, + "step": 26158 + }, + { + "epoch": 0.96, + "learning_rate": 1.0438715347713041e-07, + "loss": 0.6977, + "step": 26159 + }, + { + "epoch": 0.96, + "learning_rate": 1.0438349141245835e-07, + "loss": 0.6983, + "step": 26160 + }, + { + "epoch": 0.96, + "learning_rate": 1.0437982934778629e-07, + "loss": 0.6961, + "step": 26161 + }, + { + "epoch": 0.96, + "learning_rate": 1.0437616728311423e-07, + "loss": 0.6973, + "step": 26162 + }, + { + "epoch": 0.96, + "learning_rate": 1.0437250521844217e-07, + "loss": 0.6922, + "step": 26163 + }, + { + "epoch": 0.96, + "learning_rate": 1.0436884315377008e-07, + "loss": 0.6905, + "step": 26164 + }, + { + "epoch": 0.96, + "learning_rate": 1.0436518108909802e-07, + "loss": 0.7048, + "step": 26165 + }, + { + "epoch": 0.96, + "learning_rate": 1.0436151902442596e-07, + "loss": 0.6924, + "step": 26166 + }, + { + "epoch": 0.96, + "learning_rate": 1.043578569597539e-07, + "loss": 0.691, + "step": 26167 + }, + { + "epoch": 0.96, + "learning_rate": 1.0435419489508184e-07, + "loss": 0.6994, + "step": 26168 + }, + { + "epoch": 0.96, + "learning_rate": 1.0435053283040978e-07, + "loss": 0.6837, + "step": 26169 + }, + { + "epoch": 0.96, + "learning_rate": 1.0434687076573772e-07, + "loss": 0.6928, + "step": 26170 + }, + { + "epoch": 0.96, + "learning_rate": 1.0434320870106565e-07, + "loss": 0.6935, + "step": 26171 + }, + { + "epoch": 0.96, + "learning_rate": 1.0433954663639359e-07, + "loss": 0.6897, + "step": 26172 + }, + { + "epoch": 0.96, + "learning_rate": 1.0433588457172153e-07, + "loss": 0.684, + "step": 26173 + }, + { + "epoch": 0.96, + "learning_rate": 1.0433222250704947e-07, + "loss": 0.6869, + "step": 26174 + }, + { + "epoch": 0.96, + "learning_rate": 1.0432856044237741e-07, + "loss": 0.6773, + "step": 26175 + }, + { + "epoch": 0.96, + "learning_rate": 1.0432489837770535e-07, + "loss": 0.7035, + "step": 26176 + }, + { + "epoch": 0.96, + "learning_rate": 1.0432123631303329e-07, + "loss": 0.6752, + "step": 26177 + }, + { + "epoch": 0.96, + "learning_rate": 1.0431757424836123e-07, + "loss": 0.7018, + "step": 26178 + }, + { + "epoch": 0.96, + "learning_rate": 1.0431391218368917e-07, + "loss": 0.6924, + "step": 26179 + }, + { + "epoch": 0.96, + "learning_rate": 1.0431025011901711e-07, + "loss": 0.6871, + "step": 26180 + }, + { + "epoch": 0.96, + "learning_rate": 1.0430658805434505e-07, + "loss": 0.6898, + "step": 26181 + }, + { + "epoch": 0.96, + "learning_rate": 1.0430292598967296e-07, + "loss": 0.6952, + "step": 26182 + }, + { + "epoch": 0.96, + "learning_rate": 1.042992639250009e-07, + "loss": 0.6796, + "step": 26183 + }, + { + "epoch": 0.96, + "learning_rate": 1.0429560186032884e-07, + "loss": 0.6964, + "step": 26184 + }, + { + "epoch": 0.96, + "learning_rate": 1.0429193979565678e-07, + "loss": 0.6937, + "step": 26185 + }, + { + "epoch": 0.96, + "learning_rate": 1.0428827773098472e-07, + "loss": 0.6904, + "step": 26186 + }, + { + "epoch": 0.96, + "learning_rate": 1.0428461566631266e-07, + "loss": 0.6954, + "step": 26187 + }, + { + "epoch": 0.96, + "learning_rate": 1.042809536016406e-07, + "loss": 0.6942, + "step": 26188 + }, + { + "epoch": 0.96, + "learning_rate": 1.0427729153696854e-07, + "loss": 0.6899, + "step": 26189 + }, + { + "epoch": 0.96, + "learning_rate": 1.0427362947229648e-07, + "loss": 0.7044, + "step": 26190 + }, + { + "epoch": 0.96, + "learning_rate": 1.0426996740762442e-07, + "loss": 0.6999, + "step": 26191 + }, + { + "epoch": 0.96, + "learning_rate": 1.0426630534295235e-07, + "loss": 0.6963, + "step": 26192 + }, + { + "epoch": 0.96, + "learning_rate": 1.0426264327828029e-07, + "loss": 0.6926, + "step": 26193 + }, + { + "epoch": 0.96, + "learning_rate": 1.0425898121360823e-07, + "loss": 0.6915, + "step": 26194 + }, + { + "epoch": 0.96, + "learning_rate": 1.0425531914893617e-07, + "loss": 0.6865, + "step": 26195 + }, + { + "epoch": 0.96, + "learning_rate": 1.0425165708426411e-07, + "loss": 0.6942, + "step": 26196 + }, + { + "epoch": 0.96, + "learning_rate": 1.0424799501959205e-07, + "loss": 0.6926, + "step": 26197 + }, + { + "epoch": 0.96, + "learning_rate": 1.0424433295491999e-07, + "loss": 0.6955, + "step": 26198 + }, + { + "epoch": 0.96, + "learning_rate": 1.0424067089024793e-07, + "loss": 0.6928, + "step": 26199 + }, + { + "epoch": 0.96, + "learning_rate": 1.0423700882557584e-07, + "loss": 0.6945, + "step": 26200 + }, + { + "epoch": 0.96, + "learning_rate": 1.0423334676090378e-07, + "loss": 0.6991, + "step": 26201 + }, + { + "epoch": 0.96, + "learning_rate": 1.0422968469623172e-07, + "loss": 0.6912, + "step": 26202 + }, + { + "epoch": 0.96, + "learning_rate": 1.0422602263155966e-07, + "loss": 0.6963, + "step": 26203 + }, + { + "epoch": 0.96, + "learning_rate": 1.042223605668876e-07, + "loss": 0.6948, + "step": 26204 + }, + { + "epoch": 0.96, + "learning_rate": 1.0421869850221554e-07, + "loss": 0.6975, + "step": 26205 + }, + { + "epoch": 0.96, + "learning_rate": 1.0421503643754348e-07, + "loss": 0.6949, + "step": 26206 + }, + { + "epoch": 0.96, + "learning_rate": 1.0421137437287142e-07, + "loss": 0.6966, + "step": 26207 + }, + { + "epoch": 0.96, + "learning_rate": 1.0420771230819936e-07, + "loss": 0.6892, + "step": 26208 + }, + { + "epoch": 0.96, + "learning_rate": 1.042040502435273e-07, + "loss": 0.6995, + "step": 26209 + }, + { + "epoch": 0.96, + "learning_rate": 1.0420038817885524e-07, + "loss": 0.7083, + "step": 26210 + }, + { + "epoch": 0.96, + "learning_rate": 1.0419672611418318e-07, + "loss": 0.6939, + "step": 26211 + }, + { + "epoch": 0.96, + "learning_rate": 1.0419306404951112e-07, + "loss": 0.6871, + "step": 26212 + }, + { + "epoch": 0.96, + "learning_rate": 1.0418940198483906e-07, + "loss": 0.6881, + "step": 26213 + }, + { + "epoch": 0.96, + "learning_rate": 1.0418573992016699e-07, + "loss": 0.6851, + "step": 26214 + }, + { + "epoch": 0.96, + "learning_rate": 1.0418207785549493e-07, + "loss": 0.6966, + "step": 26215 + }, + { + "epoch": 0.96, + "learning_rate": 1.0417841579082287e-07, + "loss": 0.6896, + "step": 26216 + }, + { + "epoch": 0.96, + "learning_rate": 1.0417475372615079e-07, + "loss": 0.7002, + "step": 26217 + }, + { + "epoch": 0.96, + "learning_rate": 1.0417109166147873e-07, + "loss": 0.7099, + "step": 26218 + }, + { + "epoch": 0.96, + "learning_rate": 1.0416742959680667e-07, + "loss": 0.6739, + "step": 26219 + }, + { + "epoch": 0.96, + "learning_rate": 1.041637675321346e-07, + "loss": 0.6895, + "step": 26220 + }, + { + "epoch": 0.96, + "learning_rate": 1.0416010546746254e-07, + "loss": 0.6883, + "step": 26221 + }, + { + "epoch": 0.96, + "learning_rate": 1.0415644340279048e-07, + "loss": 0.6949, + "step": 26222 + }, + { + "epoch": 0.96, + "learning_rate": 1.0415278133811842e-07, + "loss": 0.6943, + "step": 26223 + }, + { + "epoch": 0.96, + "learning_rate": 1.0414911927344636e-07, + "loss": 0.6801, + "step": 26224 + }, + { + "epoch": 0.96, + "learning_rate": 1.041454572087743e-07, + "loss": 0.6914, + "step": 26225 + }, + { + "epoch": 0.96, + "learning_rate": 1.0414179514410224e-07, + "loss": 0.6982, + "step": 26226 + }, + { + "epoch": 0.96, + "learning_rate": 1.0413813307943018e-07, + "loss": 0.6991, + "step": 26227 + }, + { + "epoch": 0.96, + "learning_rate": 1.0413447101475812e-07, + "loss": 0.6892, + "step": 26228 + }, + { + "epoch": 0.96, + "learning_rate": 1.0413080895008606e-07, + "loss": 0.6902, + "step": 26229 + }, + { + "epoch": 0.96, + "learning_rate": 1.04127146885414e-07, + "loss": 0.6954, + "step": 26230 + }, + { + "epoch": 0.96, + "learning_rate": 1.0412348482074194e-07, + "loss": 0.6969, + "step": 26231 + }, + { + "epoch": 0.96, + "learning_rate": 1.0411982275606988e-07, + "loss": 0.6936, + "step": 26232 + }, + { + "epoch": 0.96, + "learning_rate": 1.0411616069139782e-07, + "loss": 0.7046, + "step": 26233 + }, + { + "epoch": 0.96, + "learning_rate": 1.0411249862672576e-07, + "loss": 0.6828, + "step": 26234 + }, + { + "epoch": 0.96, + "learning_rate": 1.0410883656205367e-07, + "loss": 0.6912, + "step": 26235 + }, + { + "epoch": 0.96, + "learning_rate": 1.0410517449738161e-07, + "loss": 0.6846, + "step": 26236 + }, + { + "epoch": 0.96, + "learning_rate": 1.0410151243270955e-07, + "loss": 0.6956, + "step": 26237 + }, + { + "epoch": 0.96, + "learning_rate": 1.0409785036803749e-07, + "loss": 0.7106, + "step": 26238 + }, + { + "epoch": 0.96, + "learning_rate": 1.0409418830336543e-07, + "loss": 0.6822, + "step": 26239 + }, + { + "epoch": 0.96, + "learning_rate": 1.0409052623869337e-07, + "loss": 0.6926, + "step": 26240 + }, + { + "epoch": 0.96, + "learning_rate": 1.0408686417402131e-07, + "loss": 0.7017, + "step": 26241 + }, + { + "epoch": 0.96, + "learning_rate": 1.0408320210934924e-07, + "loss": 0.6974, + "step": 26242 + }, + { + "epoch": 0.96, + "learning_rate": 1.0407954004467718e-07, + "loss": 0.6924, + "step": 26243 + }, + { + "epoch": 0.96, + "learning_rate": 1.0407587798000512e-07, + "loss": 0.6944, + "step": 26244 + }, + { + "epoch": 0.96, + "learning_rate": 1.0407221591533306e-07, + "loss": 0.6862, + "step": 26245 + }, + { + "epoch": 0.96, + "learning_rate": 1.04068553850661e-07, + "loss": 0.6852, + "step": 26246 + }, + { + "epoch": 0.96, + "learning_rate": 1.0406489178598894e-07, + "loss": 0.7011, + "step": 26247 + }, + { + "epoch": 0.96, + "learning_rate": 1.0406122972131688e-07, + "loss": 0.688, + "step": 26248 + }, + { + "epoch": 0.96, + "learning_rate": 1.0405756765664482e-07, + "loss": 0.7014, + "step": 26249 + }, + { + "epoch": 0.96, + "learning_rate": 1.0405390559197276e-07, + "loss": 0.6867, + "step": 26250 + }, + { + "epoch": 0.96, + "learning_rate": 1.040502435273007e-07, + "loss": 0.6856, + "step": 26251 + }, + { + "epoch": 0.96, + "learning_rate": 1.0404658146262864e-07, + "loss": 0.6935, + "step": 26252 + }, + { + "epoch": 0.96, + "learning_rate": 1.0404291939795655e-07, + "loss": 0.6939, + "step": 26253 + }, + { + "epoch": 0.96, + "learning_rate": 1.0403925733328449e-07, + "loss": 0.6891, + "step": 26254 + }, + { + "epoch": 0.96, + "learning_rate": 1.0403559526861243e-07, + "loss": 0.7028, + "step": 26255 + }, + { + "epoch": 0.96, + "learning_rate": 1.0403193320394037e-07, + "loss": 0.6974, + "step": 26256 + }, + { + "epoch": 0.96, + "learning_rate": 1.0402827113926831e-07, + "loss": 0.6928, + "step": 26257 + }, + { + "epoch": 0.96, + "learning_rate": 1.0402460907459625e-07, + "loss": 0.6979, + "step": 26258 + }, + { + "epoch": 0.96, + "learning_rate": 1.0402094700992419e-07, + "loss": 0.6815, + "step": 26259 + }, + { + "epoch": 0.96, + "learning_rate": 1.0401728494525213e-07, + "loss": 0.6896, + "step": 26260 + }, + { + "epoch": 0.96, + "learning_rate": 1.0401362288058007e-07, + "loss": 0.6961, + "step": 26261 + }, + { + "epoch": 0.96, + "learning_rate": 1.0400996081590801e-07, + "loss": 0.6959, + "step": 26262 + }, + { + "epoch": 0.96, + "learning_rate": 1.0400629875123594e-07, + "loss": 0.6973, + "step": 26263 + }, + { + "epoch": 0.96, + "learning_rate": 1.0400263668656388e-07, + "loss": 0.6983, + "step": 26264 + }, + { + "epoch": 0.96, + "learning_rate": 1.0399897462189182e-07, + "loss": 0.6965, + "step": 26265 + }, + { + "epoch": 0.96, + "learning_rate": 1.0399531255721976e-07, + "loss": 0.7003, + "step": 26266 + }, + { + "epoch": 0.96, + "learning_rate": 1.039916504925477e-07, + "loss": 0.6887, + "step": 26267 + }, + { + "epoch": 0.96, + "learning_rate": 1.0398798842787564e-07, + "loss": 0.6846, + "step": 26268 + }, + { + "epoch": 0.96, + "learning_rate": 1.0398432636320358e-07, + "loss": 0.6917, + "step": 26269 + }, + { + "epoch": 0.96, + "learning_rate": 1.0398066429853152e-07, + "loss": 0.697, + "step": 26270 + }, + { + "epoch": 0.96, + "learning_rate": 1.0397700223385943e-07, + "loss": 0.6865, + "step": 26271 + }, + { + "epoch": 0.96, + "learning_rate": 1.0397334016918737e-07, + "loss": 0.7018, + "step": 26272 + }, + { + "epoch": 0.96, + "learning_rate": 1.0396967810451531e-07, + "loss": 0.692, + "step": 26273 + }, + { + "epoch": 0.96, + "learning_rate": 1.0396601603984325e-07, + "loss": 0.7, + "step": 26274 + }, + { + "epoch": 0.96, + "learning_rate": 1.0396235397517119e-07, + "loss": 0.6972, + "step": 26275 + }, + { + "epoch": 0.96, + "learning_rate": 1.0395869191049913e-07, + "loss": 0.6947, + "step": 26276 + }, + { + "epoch": 0.96, + "learning_rate": 1.0395502984582707e-07, + "loss": 0.6942, + "step": 26277 + }, + { + "epoch": 0.96, + "learning_rate": 1.0395136778115501e-07, + "loss": 0.6917, + "step": 26278 + }, + { + "epoch": 0.96, + "learning_rate": 1.0394770571648295e-07, + "loss": 0.6783, + "step": 26279 + }, + { + "epoch": 0.96, + "learning_rate": 1.0394404365181089e-07, + "loss": 0.6918, + "step": 26280 + }, + { + "epoch": 0.96, + "learning_rate": 1.0394038158713883e-07, + "loss": 0.6939, + "step": 26281 + }, + { + "epoch": 0.96, + "learning_rate": 1.0393671952246677e-07, + "loss": 0.6945, + "step": 26282 + }, + { + "epoch": 0.96, + "learning_rate": 1.0393305745779471e-07, + "loss": 0.702, + "step": 26283 + }, + { + "epoch": 0.96, + "learning_rate": 1.0392939539312265e-07, + "loss": 0.6849, + "step": 26284 + }, + { + "epoch": 0.96, + "learning_rate": 1.0392573332845058e-07, + "loss": 0.6889, + "step": 26285 + }, + { + "epoch": 0.96, + "learning_rate": 1.0392207126377852e-07, + "loss": 0.7029, + "step": 26286 + }, + { + "epoch": 0.96, + "learning_rate": 1.0391840919910646e-07, + "loss": 0.6932, + "step": 26287 + }, + { + "epoch": 0.96, + "learning_rate": 1.039147471344344e-07, + "loss": 0.6933, + "step": 26288 + }, + { + "epoch": 0.96, + "learning_rate": 1.0391108506976232e-07, + "loss": 0.6867, + "step": 26289 + }, + { + "epoch": 0.96, + "learning_rate": 1.0390742300509026e-07, + "loss": 0.6929, + "step": 26290 + }, + { + "epoch": 0.96, + "learning_rate": 1.0390376094041819e-07, + "loss": 0.6839, + "step": 26291 + }, + { + "epoch": 0.96, + "learning_rate": 1.0390009887574613e-07, + "loss": 0.6973, + "step": 26292 + }, + { + "epoch": 0.96, + "learning_rate": 1.0389643681107407e-07, + "loss": 0.7151, + "step": 26293 + }, + { + "epoch": 0.96, + "learning_rate": 1.0389277474640201e-07, + "loss": 0.7012, + "step": 26294 + }, + { + "epoch": 0.96, + "learning_rate": 1.0388911268172995e-07, + "loss": 0.6966, + "step": 26295 + }, + { + "epoch": 0.96, + "learning_rate": 1.0388545061705789e-07, + "loss": 0.6963, + "step": 26296 + }, + { + "epoch": 0.96, + "learning_rate": 1.0388178855238583e-07, + "loss": 0.7016, + "step": 26297 + }, + { + "epoch": 0.96, + "learning_rate": 1.0387812648771377e-07, + "loss": 0.7061, + "step": 26298 + }, + { + "epoch": 0.96, + "learning_rate": 1.0387446442304171e-07, + "loss": 0.7071, + "step": 26299 + }, + { + "epoch": 0.96, + "learning_rate": 1.0387080235836965e-07, + "loss": 0.7004, + "step": 26300 + }, + { + "epoch": 0.96, + "learning_rate": 1.0386714029369759e-07, + "loss": 0.6954, + "step": 26301 + }, + { + "epoch": 0.96, + "learning_rate": 1.0386347822902553e-07, + "loss": 0.6964, + "step": 26302 + }, + { + "epoch": 0.96, + "learning_rate": 1.0385981616435347e-07, + "loss": 0.702, + "step": 26303 + }, + { + "epoch": 0.96, + "learning_rate": 1.0385615409968141e-07, + "loss": 0.6803, + "step": 26304 + }, + { + "epoch": 0.96, + "learning_rate": 1.0385249203500935e-07, + "loss": 0.6855, + "step": 26305 + }, + { + "epoch": 0.96, + "learning_rate": 1.0384882997033729e-07, + "loss": 0.694, + "step": 26306 + }, + { + "epoch": 0.96, + "learning_rate": 1.038451679056652e-07, + "loss": 0.7084, + "step": 26307 + }, + { + "epoch": 0.96, + "learning_rate": 1.0384150584099314e-07, + "loss": 0.6779, + "step": 26308 + }, + { + "epoch": 0.96, + "learning_rate": 1.0383784377632108e-07, + "loss": 0.6805, + "step": 26309 + }, + { + "epoch": 0.96, + "learning_rate": 1.0383418171164902e-07, + "loss": 0.6844, + "step": 26310 + }, + { + "epoch": 0.96, + "learning_rate": 1.0383051964697696e-07, + "loss": 0.6898, + "step": 26311 + }, + { + "epoch": 0.96, + "learning_rate": 1.0382685758230489e-07, + "loss": 0.6949, + "step": 26312 + }, + { + "epoch": 0.96, + "learning_rate": 1.0382319551763283e-07, + "loss": 0.6962, + "step": 26313 + }, + { + "epoch": 0.96, + "learning_rate": 1.0381953345296077e-07, + "loss": 0.6829, + "step": 26314 + }, + { + "epoch": 0.96, + "learning_rate": 1.0381587138828871e-07, + "loss": 0.6912, + "step": 26315 + }, + { + "epoch": 0.96, + "learning_rate": 1.0381220932361665e-07, + "loss": 0.6988, + "step": 26316 + }, + { + "epoch": 0.96, + "learning_rate": 1.0380854725894459e-07, + "loss": 0.6859, + "step": 26317 + }, + { + "epoch": 0.96, + "learning_rate": 1.0380488519427253e-07, + "loss": 0.6921, + "step": 26318 + }, + { + "epoch": 0.96, + "learning_rate": 1.0380122312960047e-07, + "loss": 0.6897, + "step": 26319 + }, + { + "epoch": 0.96, + "learning_rate": 1.0379756106492841e-07, + "loss": 0.696, + "step": 26320 + }, + { + "epoch": 0.96, + "learning_rate": 1.0379389900025635e-07, + "loss": 0.6921, + "step": 26321 + }, + { + "epoch": 0.96, + "learning_rate": 1.0379023693558429e-07, + "loss": 0.6965, + "step": 26322 + }, + { + "epoch": 0.96, + "learning_rate": 1.0378657487091223e-07, + "loss": 0.7012, + "step": 26323 + }, + { + "epoch": 0.96, + "learning_rate": 1.0378291280624017e-07, + "loss": 0.6789, + "step": 26324 + }, + { + "epoch": 0.96, + "learning_rate": 1.0377925074156808e-07, + "loss": 0.6967, + "step": 26325 + }, + { + "epoch": 0.96, + "learning_rate": 1.0377558867689602e-07, + "loss": 0.6995, + "step": 26326 + }, + { + "epoch": 0.96, + "learning_rate": 1.0377192661222396e-07, + "loss": 0.7032, + "step": 26327 + }, + { + "epoch": 0.96, + "learning_rate": 1.037682645475519e-07, + "loss": 0.698, + "step": 26328 + }, + { + "epoch": 0.96, + "learning_rate": 1.0376460248287984e-07, + "loss": 0.7182, + "step": 26329 + }, + { + "epoch": 0.96, + "learning_rate": 1.0376094041820778e-07, + "loss": 0.6942, + "step": 26330 + }, + { + "epoch": 0.96, + "learning_rate": 1.0375727835353572e-07, + "loss": 0.6911, + "step": 26331 + }, + { + "epoch": 0.96, + "learning_rate": 1.0375361628886366e-07, + "loss": 0.6881, + "step": 26332 + }, + { + "epoch": 0.96, + "learning_rate": 1.037499542241916e-07, + "loss": 0.697, + "step": 26333 + }, + { + "epoch": 0.96, + "learning_rate": 1.0374629215951953e-07, + "loss": 0.69, + "step": 26334 + }, + { + "epoch": 0.96, + "learning_rate": 1.0374263009484747e-07, + "loss": 0.7017, + "step": 26335 + }, + { + "epoch": 0.96, + "learning_rate": 1.037389680301754e-07, + "loss": 0.6927, + "step": 26336 + }, + { + "epoch": 0.96, + "learning_rate": 1.0373530596550335e-07, + "loss": 0.6996, + "step": 26337 + }, + { + "epoch": 0.96, + "learning_rate": 1.0373164390083129e-07, + "loss": 0.6957, + "step": 26338 + }, + { + "epoch": 0.96, + "learning_rate": 1.0372798183615923e-07, + "loss": 0.6916, + "step": 26339 + }, + { + "epoch": 0.96, + "learning_rate": 1.0372431977148717e-07, + "loss": 0.7025, + "step": 26340 + }, + { + "epoch": 0.96, + "learning_rate": 1.037206577068151e-07, + "loss": 0.6855, + "step": 26341 + }, + { + "epoch": 0.96, + "learning_rate": 1.0371699564214305e-07, + "loss": 0.6857, + "step": 26342 + }, + { + "epoch": 0.96, + "learning_rate": 1.0371333357747096e-07, + "loss": 0.686, + "step": 26343 + }, + { + "epoch": 0.96, + "learning_rate": 1.037096715127989e-07, + "loss": 0.6873, + "step": 26344 + }, + { + "epoch": 0.96, + "learning_rate": 1.0370600944812684e-07, + "loss": 0.7019, + "step": 26345 + }, + { + "epoch": 0.96, + "learning_rate": 1.0370234738345478e-07, + "loss": 0.6922, + "step": 26346 + }, + { + "epoch": 0.96, + "learning_rate": 1.0369868531878272e-07, + "loss": 0.6991, + "step": 26347 + }, + { + "epoch": 0.96, + "learning_rate": 1.0369502325411066e-07, + "loss": 0.6932, + "step": 26348 + }, + { + "epoch": 0.96, + "learning_rate": 1.036913611894386e-07, + "loss": 0.681, + "step": 26349 + }, + { + "epoch": 0.96, + "learning_rate": 1.0368769912476654e-07, + "loss": 0.6916, + "step": 26350 + }, + { + "epoch": 0.96, + "learning_rate": 1.0368403706009448e-07, + "loss": 0.6929, + "step": 26351 + }, + { + "epoch": 0.96, + "learning_rate": 1.0368037499542242e-07, + "loss": 0.6903, + "step": 26352 + }, + { + "epoch": 0.96, + "learning_rate": 1.0367671293075036e-07, + "loss": 0.6874, + "step": 26353 + }, + { + "epoch": 0.96, + "learning_rate": 1.036730508660783e-07, + "loss": 0.7043, + "step": 26354 + }, + { + "epoch": 0.96, + "learning_rate": 1.0366938880140624e-07, + "loss": 0.6986, + "step": 26355 + }, + { + "epoch": 0.96, + "learning_rate": 1.0366572673673417e-07, + "loss": 0.6884, + "step": 26356 + }, + { + "epoch": 0.96, + "learning_rate": 1.036620646720621e-07, + "loss": 0.684, + "step": 26357 + }, + { + "epoch": 0.96, + "learning_rate": 1.0365840260739004e-07, + "loss": 0.6972, + "step": 26358 + }, + { + "epoch": 0.96, + "learning_rate": 1.0365474054271798e-07, + "loss": 0.6976, + "step": 26359 + }, + { + "epoch": 0.96, + "learning_rate": 1.0365107847804592e-07, + "loss": 0.697, + "step": 26360 + }, + { + "epoch": 0.96, + "learning_rate": 1.0364741641337385e-07, + "loss": 0.7012, + "step": 26361 + }, + { + "epoch": 0.96, + "learning_rate": 1.0364375434870178e-07, + "loss": 0.6995, + "step": 26362 + }, + { + "epoch": 0.96, + "learning_rate": 1.0364009228402972e-07, + "loss": 0.6982, + "step": 26363 + }, + { + "epoch": 0.96, + "learning_rate": 1.0363643021935766e-07, + "loss": 0.6946, + "step": 26364 + }, + { + "epoch": 0.96, + "learning_rate": 1.036327681546856e-07, + "loss": 0.6998, + "step": 26365 + }, + { + "epoch": 0.96, + "learning_rate": 1.0362910609001354e-07, + "loss": 0.7011, + "step": 26366 + }, + { + "epoch": 0.96, + "learning_rate": 1.0362544402534148e-07, + "loss": 0.6938, + "step": 26367 + }, + { + "epoch": 0.96, + "learning_rate": 1.0362178196066942e-07, + "loss": 0.7072, + "step": 26368 + }, + { + "epoch": 0.96, + "learning_rate": 1.0361811989599736e-07, + "loss": 0.6986, + "step": 26369 + }, + { + "epoch": 0.96, + "learning_rate": 1.036144578313253e-07, + "loss": 0.6838, + "step": 26370 + }, + { + "epoch": 0.96, + "learning_rate": 1.0361079576665324e-07, + "loss": 0.6985, + "step": 26371 + }, + { + "epoch": 0.96, + "learning_rate": 1.0360713370198118e-07, + "loss": 0.6885, + "step": 26372 + }, + { + "epoch": 0.96, + "learning_rate": 1.0360347163730912e-07, + "loss": 0.6804, + "step": 26373 + }, + { + "epoch": 0.96, + "learning_rate": 1.0359980957263706e-07, + "loss": 0.6976, + "step": 26374 + }, + { + "epoch": 0.96, + "learning_rate": 1.03596147507965e-07, + "loss": 0.6998, + "step": 26375 + }, + { + "epoch": 0.97, + "learning_rate": 1.0359248544329294e-07, + "loss": 0.6972, + "step": 26376 + }, + { + "epoch": 0.97, + "learning_rate": 1.0358882337862088e-07, + "loss": 0.6993, + "step": 26377 + }, + { + "epoch": 0.97, + "learning_rate": 1.035851613139488e-07, + "loss": 0.7008, + "step": 26378 + }, + { + "epoch": 0.97, + "learning_rate": 1.0358149924927673e-07, + "loss": 0.6982, + "step": 26379 + }, + { + "epoch": 0.97, + "learning_rate": 1.0357783718460467e-07, + "loss": 0.694, + "step": 26380 + }, + { + "epoch": 0.97, + "learning_rate": 1.0357417511993261e-07, + "loss": 0.694, + "step": 26381 + }, + { + "epoch": 0.97, + "learning_rate": 1.0357051305526055e-07, + "loss": 0.6938, + "step": 26382 + }, + { + "epoch": 0.97, + "learning_rate": 1.0356685099058848e-07, + "loss": 0.7025, + "step": 26383 + }, + { + "epoch": 0.97, + "learning_rate": 1.0356318892591642e-07, + "loss": 0.6967, + "step": 26384 + }, + { + "epoch": 0.97, + "learning_rate": 1.0355952686124436e-07, + "loss": 0.6912, + "step": 26385 + }, + { + "epoch": 0.97, + "learning_rate": 1.035558647965723e-07, + "loss": 0.6914, + "step": 26386 + }, + { + "epoch": 0.97, + "learning_rate": 1.0355220273190024e-07, + "loss": 0.6902, + "step": 26387 + }, + { + "epoch": 0.97, + "learning_rate": 1.0354854066722818e-07, + "loss": 0.6942, + "step": 26388 + }, + { + "epoch": 0.97, + "learning_rate": 1.0354487860255612e-07, + "loss": 0.6873, + "step": 26389 + }, + { + "epoch": 0.97, + "learning_rate": 1.0354121653788406e-07, + "loss": 0.6899, + "step": 26390 + }, + { + "epoch": 0.97, + "learning_rate": 1.03537554473212e-07, + "loss": 0.6977, + "step": 26391 + }, + { + "epoch": 0.97, + "learning_rate": 1.0353389240853994e-07, + "loss": 0.694, + "step": 26392 + }, + { + "epoch": 0.97, + "learning_rate": 1.0353023034386788e-07, + "loss": 0.6952, + "step": 26393 + }, + { + "epoch": 0.97, + "learning_rate": 1.0352656827919582e-07, + "loss": 0.698, + "step": 26394 + }, + { + "epoch": 0.97, + "learning_rate": 1.0352290621452376e-07, + "loss": 0.6904, + "step": 26395 + }, + { + "epoch": 0.97, + "learning_rate": 1.035192441498517e-07, + "loss": 0.685, + "step": 26396 + }, + { + "epoch": 0.97, + "learning_rate": 1.0351558208517961e-07, + "loss": 0.6976, + "step": 26397 + }, + { + "epoch": 0.97, + "learning_rate": 1.0351192002050755e-07, + "loss": 0.6867, + "step": 26398 + }, + { + "epoch": 0.97, + "learning_rate": 1.0350825795583549e-07, + "loss": 0.6857, + "step": 26399 + }, + { + "epoch": 0.97, + "learning_rate": 1.0350459589116343e-07, + "loss": 0.6966, + "step": 26400 + }, + { + "epoch": 0.97, + "learning_rate": 1.0350093382649137e-07, + "loss": 0.6872, + "step": 26401 + }, + { + "epoch": 0.97, + "learning_rate": 1.0349727176181931e-07, + "loss": 0.6976, + "step": 26402 + }, + { + "epoch": 0.97, + "learning_rate": 1.0349360969714725e-07, + "loss": 0.6877, + "step": 26403 + }, + { + "epoch": 0.97, + "learning_rate": 1.0348994763247519e-07, + "loss": 0.6933, + "step": 26404 + }, + { + "epoch": 0.97, + "learning_rate": 1.0348628556780312e-07, + "loss": 0.6681, + "step": 26405 + }, + { + "epoch": 0.97, + "learning_rate": 1.0348262350313106e-07, + "loss": 0.6999, + "step": 26406 + }, + { + "epoch": 0.97, + "learning_rate": 1.03478961438459e-07, + "loss": 0.7109, + "step": 26407 + }, + { + "epoch": 0.97, + "learning_rate": 1.0347529937378694e-07, + "loss": 0.6895, + "step": 26408 + }, + { + "epoch": 0.97, + "learning_rate": 1.0347163730911488e-07, + "loss": 0.6904, + "step": 26409 + }, + { + "epoch": 0.97, + "learning_rate": 1.0346797524444282e-07, + "loss": 0.6922, + "step": 26410 + }, + { + "epoch": 0.97, + "learning_rate": 1.0346431317977076e-07, + "loss": 0.6849, + "step": 26411 + }, + { + "epoch": 0.97, + "learning_rate": 1.034606511150987e-07, + "loss": 0.6862, + "step": 26412 + }, + { + "epoch": 0.97, + "learning_rate": 1.0345698905042664e-07, + "loss": 0.6973, + "step": 26413 + }, + { + "epoch": 0.97, + "learning_rate": 1.0345332698575458e-07, + "loss": 0.7049, + "step": 26414 + }, + { + "epoch": 0.97, + "learning_rate": 1.0344966492108249e-07, + "loss": 0.6925, + "step": 26415 + }, + { + "epoch": 0.97, + "learning_rate": 1.0344600285641043e-07, + "loss": 0.6887, + "step": 26416 + }, + { + "epoch": 0.97, + "learning_rate": 1.0344234079173837e-07, + "loss": 0.6857, + "step": 26417 + }, + { + "epoch": 0.97, + "learning_rate": 1.0343867872706631e-07, + "loss": 0.703, + "step": 26418 + }, + { + "epoch": 0.97, + "learning_rate": 1.0343501666239425e-07, + "loss": 0.6908, + "step": 26419 + }, + { + "epoch": 0.97, + "learning_rate": 1.0343135459772219e-07, + "loss": 0.688, + "step": 26420 + }, + { + "epoch": 0.97, + "learning_rate": 1.0342769253305013e-07, + "loss": 0.6937, + "step": 26421 + }, + { + "epoch": 0.97, + "learning_rate": 1.0342403046837807e-07, + "loss": 0.7006, + "step": 26422 + }, + { + "epoch": 0.97, + "learning_rate": 1.0342036840370601e-07, + "loss": 0.6872, + "step": 26423 + }, + { + "epoch": 0.97, + "learning_rate": 1.0341670633903395e-07, + "loss": 0.7025, + "step": 26424 + }, + { + "epoch": 0.97, + "learning_rate": 1.0341304427436189e-07, + "loss": 0.7006, + "step": 26425 + }, + { + "epoch": 0.97, + "learning_rate": 1.0340938220968983e-07, + "loss": 0.6918, + "step": 26426 + }, + { + "epoch": 0.97, + "learning_rate": 1.0340572014501775e-07, + "loss": 0.7038, + "step": 26427 + }, + { + "epoch": 0.97, + "learning_rate": 1.034020580803457e-07, + "loss": 0.681, + "step": 26428 + }, + { + "epoch": 0.97, + "learning_rate": 1.0339839601567363e-07, + "loss": 0.6869, + "step": 26429 + }, + { + "epoch": 0.97, + "learning_rate": 1.0339473395100157e-07, + "loss": 0.6828, + "step": 26430 + }, + { + "epoch": 0.97, + "learning_rate": 1.0339107188632951e-07, + "loss": 0.694, + "step": 26431 + }, + { + "epoch": 0.97, + "learning_rate": 1.0338740982165745e-07, + "loss": 0.693, + "step": 26432 + }, + { + "epoch": 0.97, + "learning_rate": 1.0338374775698537e-07, + "loss": 0.6786, + "step": 26433 + }, + { + "epoch": 0.97, + "learning_rate": 1.0338008569231331e-07, + "loss": 0.6999, + "step": 26434 + }, + { + "epoch": 0.97, + "learning_rate": 1.0337642362764125e-07, + "loss": 0.6802, + "step": 26435 + }, + { + "epoch": 0.97, + "learning_rate": 1.0337276156296919e-07, + "loss": 0.6905, + "step": 26436 + }, + { + "epoch": 0.97, + "learning_rate": 1.0336909949829713e-07, + "loss": 0.6901, + "step": 26437 + }, + { + "epoch": 0.97, + "learning_rate": 1.0336543743362507e-07, + "loss": 0.6999, + "step": 26438 + }, + { + "epoch": 0.97, + "learning_rate": 1.0336177536895301e-07, + "loss": 0.7017, + "step": 26439 + }, + { + "epoch": 0.97, + "learning_rate": 1.0335811330428095e-07, + "loss": 0.7022, + "step": 26440 + }, + { + "epoch": 0.97, + "learning_rate": 1.0335445123960889e-07, + "loss": 0.6776, + "step": 26441 + }, + { + "epoch": 0.97, + "learning_rate": 1.0335078917493683e-07, + "loss": 0.6949, + "step": 26442 + }, + { + "epoch": 0.97, + "learning_rate": 1.0334712711026477e-07, + "loss": 0.6904, + "step": 26443 + }, + { + "epoch": 0.97, + "learning_rate": 1.0334346504559271e-07, + "loss": 0.6957, + "step": 26444 + }, + { + "epoch": 0.97, + "learning_rate": 1.0333980298092065e-07, + "loss": 0.6929, + "step": 26445 + }, + { + "epoch": 0.97, + "learning_rate": 1.0333614091624859e-07, + "loss": 0.6939, + "step": 26446 + }, + { + "epoch": 0.97, + "learning_rate": 1.0333247885157653e-07, + "loss": 0.6917, + "step": 26447 + }, + { + "epoch": 0.97, + "learning_rate": 1.0332881678690447e-07, + "loss": 0.6972, + "step": 26448 + }, + { + "epoch": 0.97, + "learning_rate": 1.0332515472223239e-07, + "loss": 0.6922, + "step": 26449 + }, + { + "epoch": 0.97, + "learning_rate": 1.0332149265756033e-07, + "loss": 0.7118, + "step": 26450 + }, + { + "epoch": 0.97, + "learning_rate": 1.0331783059288826e-07, + "loss": 0.6945, + "step": 26451 + }, + { + "epoch": 0.97, + "learning_rate": 1.033141685282162e-07, + "loss": 0.7115, + "step": 26452 + }, + { + "epoch": 0.97, + "learning_rate": 1.0331050646354414e-07, + "loss": 0.6918, + "step": 26453 + }, + { + "epoch": 0.97, + "learning_rate": 1.0330684439887207e-07, + "loss": 0.6869, + "step": 26454 + }, + { + "epoch": 0.97, + "learning_rate": 1.033031823342e-07, + "loss": 0.6821, + "step": 26455 + }, + { + "epoch": 0.97, + "learning_rate": 1.0329952026952795e-07, + "loss": 0.6981, + "step": 26456 + }, + { + "epoch": 0.97, + "learning_rate": 1.0329585820485589e-07, + "loss": 0.6862, + "step": 26457 + }, + { + "epoch": 0.97, + "learning_rate": 1.0329219614018383e-07, + "loss": 0.6909, + "step": 26458 + }, + { + "epoch": 0.97, + "learning_rate": 1.0328853407551177e-07, + "loss": 0.7077, + "step": 26459 + }, + { + "epoch": 0.97, + "learning_rate": 1.032848720108397e-07, + "loss": 0.6899, + "step": 26460 + }, + { + "epoch": 0.97, + "learning_rate": 1.0328120994616765e-07, + "loss": 0.6978, + "step": 26461 + }, + { + "epoch": 0.97, + "learning_rate": 1.0327754788149559e-07, + "loss": 0.6942, + "step": 26462 + }, + { + "epoch": 0.97, + "learning_rate": 1.0327388581682353e-07, + "loss": 0.6738, + "step": 26463 + }, + { + "epoch": 0.97, + "learning_rate": 1.0327022375215147e-07, + "loss": 0.6993, + "step": 26464 + }, + { + "epoch": 0.97, + "learning_rate": 1.032665616874794e-07, + "loss": 0.6779, + "step": 26465 + }, + { + "epoch": 0.97, + "learning_rate": 1.0326289962280735e-07, + "loss": 0.7047, + "step": 26466 + }, + { + "epoch": 0.97, + "learning_rate": 1.0325923755813529e-07, + "loss": 0.7019, + "step": 26467 + }, + { + "epoch": 0.97, + "learning_rate": 1.0325557549346323e-07, + "loss": 0.6946, + "step": 26468 + }, + { + "epoch": 0.97, + "learning_rate": 1.0325191342879114e-07, + "loss": 0.6835, + "step": 26469 + }, + { + "epoch": 0.97, + "learning_rate": 1.0324825136411908e-07, + "loss": 0.6819, + "step": 26470 + }, + { + "epoch": 0.97, + "learning_rate": 1.0324458929944702e-07, + "loss": 0.707, + "step": 26471 + }, + { + "epoch": 0.97, + "learning_rate": 1.0324092723477496e-07, + "loss": 0.7082, + "step": 26472 + }, + { + "epoch": 0.97, + "learning_rate": 1.032372651701029e-07, + "loss": 0.6946, + "step": 26473 + }, + { + "epoch": 0.97, + "learning_rate": 1.0323360310543084e-07, + "loss": 0.6728, + "step": 26474 + }, + { + "epoch": 0.97, + "learning_rate": 1.0322994104075878e-07, + "loss": 0.6963, + "step": 26475 + }, + { + "epoch": 0.97, + "learning_rate": 1.032262789760867e-07, + "loss": 0.6873, + "step": 26476 + }, + { + "epoch": 0.97, + "learning_rate": 1.0322261691141465e-07, + "loss": 0.6901, + "step": 26477 + }, + { + "epoch": 0.97, + "learning_rate": 1.0321895484674258e-07, + "loss": 0.6976, + "step": 26478 + }, + { + "epoch": 0.97, + "learning_rate": 1.0321529278207052e-07, + "loss": 0.6973, + "step": 26479 + }, + { + "epoch": 0.97, + "learning_rate": 1.0321163071739846e-07, + "loss": 0.7042, + "step": 26480 + }, + { + "epoch": 0.97, + "learning_rate": 1.032079686527264e-07, + "loss": 0.6788, + "step": 26481 + }, + { + "epoch": 0.97, + "learning_rate": 1.0320430658805434e-07, + "loss": 0.6909, + "step": 26482 + }, + { + "epoch": 0.97, + "learning_rate": 1.0320064452338228e-07, + "loss": 0.7042, + "step": 26483 + }, + { + "epoch": 0.97, + "learning_rate": 1.0319698245871022e-07, + "loss": 0.7016, + "step": 26484 + }, + { + "epoch": 0.97, + "learning_rate": 1.0319332039403816e-07, + "loss": 0.6874, + "step": 26485 + }, + { + "epoch": 0.97, + "learning_rate": 1.031896583293661e-07, + "loss": 0.6859, + "step": 26486 + }, + { + "epoch": 0.97, + "learning_rate": 1.0318599626469402e-07, + "loss": 0.6978, + "step": 26487 + }, + { + "epoch": 0.97, + "learning_rate": 1.0318233420002196e-07, + "loss": 0.6966, + "step": 26488 + }, + { + "epoch": 0.97, + "learning_rate": 1.031786721353499e-07, + "loss": 0.6942, + "step": 26489 + }, + { + "epoch": 0.97, + "learning_rate": 1.0317501007067784e-07, + "loss": 0.696, + "step": 26490 + }, + { + "epoch": 0.97, + "learning_rate": 1.0317134800600578e-07, + "loss": 0.6835, + "step": 26491 + }, + { + "epoch": 0.97, + "learning_rate": 1.0316768594133372e-07, + "loss": 0.6758, + "step": 26492 + }, + { + "epoch": 0.97, + "learning_rate": 1.0316402387666166e-07, + "loss": 0.6891, + "step": 26493 + }, + { + "epoch": 0.97, + "learning_rate": 1.031603618119896e-07, + "loss": 0.7057, + "step": 26494 + }, + { + "epoch": 0.97, + "learning_rate": 1.0315669974731754e-07, + "loss": 0.704, + "step": 26495 + }, + { + "epoch": 0.97, + "learning_rate": 1.0315303768264548e-07, + "loss": 0.6909, + "step": 26496 + }, + { + "epoch": 0.97, + "learning_rate": 1.0314937561797342e-07, + "loss": 0.6882, + "step": 26497 + }, + { + "epoch": 0.97, + "learning_rate": 1.0314571355330134e-07, + "loss": 0.6959, + "step": 26498 + }, + { + "epoch": 0.97, + "learning_rate": 1.0314205148862928e-07, + "loss": 0.6839, + "step": 26499 + }, + { + "epoch": 0.97, + "learning_rate": 1.0313838942395722e-07, + "loss": 0.7012, + "step": 26500 + }, + { + "epoch": 0.97, + "learning_rate": 1.0313472735928516e-07, + "loss": 0.6878, + "step": 26501 + }, + { + "epoch": 0.97, + "learning_rate": 1.031310652946131e-07, + "loss": 0.6793, + "step": 26502 + }, + { + "epoch": 0.97, + "learning_rate": 1.0312740322994104e-07, + "loss": 0.7043, + "step": 26503 + }, + { + "epoch": 0.97, + "learning_rate": 1.0312374116526898e-07, + "loss": 0.7003, + "step": 26504 + }, + { + "epoch": 0.97, + "learning_rate": 1.031200791005969e-07, + "loss": 0.684, + "step": 26505 + }, + { + "epoch": 0.97, + "learning_rate": 1.0311641703592484e-07, + "loss": 0.6961, + "step": 26506 + }, + { + "epoch": 0.97, + "learning_rate": 1.0311275497125278e-07, + "loss": 0.6881, + "step": 26507 + }, + { + "epoch": 0.97, + "learning_rate": 1.0310909290658072e-07, + "loss": 0.6867, + "step": 26508 + }, + { + "epoch": 0.97, + "learning_rate": 1.0310543084190866e-07, + "loss": 0.6948, + "step": 26509 + }, + { + "epoch": 0.97, + "learning_rate": 1.031017687772366e-07, + "loss": 0.6744, + "step": 26510 + }, + { + "epoch": 0.97, + "learning_rate": 1.0309810671256454e-07, + "loss": 0.6715, + "step": 26511 + }, + { + "epoch": 0.97, + "learning_rate": 1.0309444464789248e-07, + "loss": 0.6977, + "step": 26512 + }, + { + "epoch": 0.97, + "learning_rate": 1.0309078258322042e-07, + "loss": 0.6833, + "step": 26513 + }, + { + "epoch": 0.97, + "learning_rate": 1.0308712051854836e-07, + "loss": 0.7046, + "step": 26514 + }, + { + "epoch": 0.97, + "learning_rate": 1.030834584538763e-07, + "loss": 0.6965, + "step": 26515 + }, + { + "epoch": 0.97, + "learning_rate": 1.0307979638920424e-07, + "loss": 0.6844, + "step": 26516 + }, + { + "epoch": 0.97, + "learning_rate": 1.0307613432453218e-07, + "loss": 0.685, + "step": 26517 + }, + { + "epoch": 0.97, + "learning_rate": 1.0307247225986012e-07, + "loss": 0.7044, + "step": 26518 + }, + { + "epoch": 0.97, + "learning_rate": 1.0306881019518806e-07, + "loss": 0.6952, + "step": 26519 + }, + { + "epoch": 0.97, + "learning_rate": 1.0306514813051598e-07, + "loss": 0.6871, + "step": 26520 + }, + { + "epoch": 0.97, + "learning_rate": 1.0306148606584392e-07, + "loss": 0.7036, + "step": 26521 + }, + { + "epoch": 0.97, + "learning_rate": 1.0305782400117186e-07, + "loss": 0.6942, + "step": 26522 + }, + { + "epoch": 0.97, + "learning_rate": 1.0305416193649979e-07, + "loss": 0.6827, + "step": 26523 + }, + { + "epoch": 0.97, + "learning_rate": 1.0305049987182773e-07, + "loss": 0.706, + "step": 26524 + }, + { + "epoch": 0.97, + "learning_rate": 1.0304683780715566e-07, + "loss": 0.7005, + "step": 26525 + }, + { + "epoch": 0.97, + "learning_rate": 1.030431757424836e-07, + "loss": 0.7064, + "step": 26526 + }, + { + "epoch": 0.97, + "learning_rate": 1.0303951367781154e-07, + "loss": 0.6902, + "step": 26527 + }, + { + "epoch": 0.97, + "learning_rate": 1.0303585161313948e-07, + "loss": 0.6582, + "step": 26528 + }, + { + "epoch": 0.97, + "learning_rate": 1.0303218954846742e-07, + "loss": 0.6988, + "step": 26529 + }, + { + "epoch": 0.97, + "learning_rate": 1.0302852748379536e-07, + "loss": 0.6897, + "step": 26530 + }, + { + "epoch": 0.97, + "learning_rate": 1.030248654191233e-07, + "loss": 0.6876, + "step": 26531 + }, + { + "epoch": 0.97, + "learning_rate": 1.0302120335445124e-07, + "loss": 0.6889, + "step": 26532 + }, + { + "epoch": 0.97, + "learning_rate": 1.0301754128977918e-07, + "loss": 0.7022, + "step": 26533 + }, + { + "epoch": 0.97, + "learning_rate": 1.0301387922510712e-07, + "loss": 0.6942, + "step": 26534 + }, + { + "epoch": 0.97, + "learning_rate": 1.0301021716043505e-07, + "loss": 0.6918, + "step": 26535 + }, + { + "epoch": 0.97, + "learning_rate": 1.03006555095763e-07, + "loss": 0.6937, + "step": 26536 + }, + { + "epoch": 0.97, + "learning_rate": 1.0300289303109093e-07, + "loss": 0.7038, + "step": 26537 + }, + { + "epoch": 0.97, + "learning_rate": 1.0299923096641887e-07, + "loss": 0.6885, + "step": 26538 + }, + { + "epoch": 0.97, + "learning_rate": 1.0299556890174681e-07, + "loss": 0.6866, + "step": 26539 + }, + { + "epoch": 0.97, + "learning_rate": 1.0299190683707475e-07, + "loss": 0.6889, + "step": 26540 + }, + { + "epoch": 0.97, + "learning_rate": 1.0298824477240267e-07, + "loss": 0.6928, + "step": 26541 + }, + { + "epoch": 0.97, + "learning_rate": 1.0298458270773061e-07, + "loss": 0.7034, + "step": 26542 + }, + { + "epoch": 0.97, + "learning_rate": 1.0298092064305855e-07, + "loss": 0.6881, + "step": 26543 + }, + { + "epoch": 0.97, + "learning_rate": 1.0297725857838649e-07, + "loss": 0.6806, + "step": 26544 + }, + { + "epoch": 0.97, + "learning_rate": 1.0297359651371443e-07, + "loss": 0.7013, + "step": 26545 + }, + { + "epoch": 0.97, + "learning_rate": 1.0296993444904237e-07, + "loss": 0.69, + "step": 26546 + }, + { + "epoch": 0.97, + "learning_rate": 1.029662723843703e-07, + "loss": 0.6957, + "step": 26547 + }, + { + "epoch": 0.97, + "learning_rate": 1.0296261031969823e-07, + "loss": 0.7008, + "step": 26548 + }, + { + "epoch": 0.97, + "learning_rate": 1.0295894825502617e-07, + "loss": 0.6965, + "step": 26549 + }, + { + "epoch": 0.97, + "learning_rate": 1.0295528619035411e-07, + "loss": 0.6834, + "step": 26550 + }, + { + "epoch": 0.97, + "learning_rate": 1.0295162412568205e-07, + "loss": 0.6933, + "step": 26551 + }, + { + "epoch": 0.97, + "learning_rate": 1.0294796206101e-07, + "loss": 0.6955, + "step": 26552 + }, + { + "epoch": 0.97, + "learning_rate": 1.0294429999633793e-07, + "loss": 0.6786, + "step": 26553 + }, + { + "epoch": 0.97, + "learning_rate": 1.0294063793166587e-07, + "loss": 0.6925, + "step": 26554 + }, + { + "epoch": 0.97, + "learning_rate": 1.0293697586699381e-07, + "loss": 0.6898, + "step": 26555 + }, + { + "epoch": 0.97, + "learning_rate": 1.0293331380232175e-07, + "loss": 0.6956, + "step": 26556 + }, + { + "epoch": 0.97, + "learning_rate": 1.029296517376497e-07, + "loss": 0.6938, + "step": 26557 + }, + { + "epoch": 0.97, + "learning_rate": 1.0292598967297763e-07, + "loss": 0.6956, + "step": 26558 + }, + { + "epoch": 0.97, + "learning_rate": 1.0292232760830555e-07, + "loss": 0.6937, + "step": 26559 + }, + { + "epoch": 0.97, + "learning_rate": 1.0291866554363349e-07, + "loss": 0.7141, + "step": 26560 + }, + { + "epoch": 0.97, + "learning_rate": 1.0291500347896143e-07, + "loss": 0.6765, + "step": 26561 + }, + { + "epoch": 0.97, + "learning_rate": 1.0291134141428937e-07, + "loss": 0.6838, + "step": 26562 + }, + { + "epoch": 0.97, + "learning_rate": 1.0290767934961731e-07, + "loss": 0.6882, + "step": 26563 + }, + { + "epoch": 0.97, + "learning_rate": 1.0290401728494525e-07, + "loss": 0.6934, + "step": 26564 + }, + { + "epoch": 0.97, + "learning_rate": 1.0290035522027319e-07, + "loss": 0.6835, + "step": 26565 + }, + { + "epoch": 0.97, + "learning_rate": 1.0289669315560113e-07, + "loss": 0.6825, + "step": 26566 + }, + { + "epoch": 0.97, + "learning_rate": 1.0289303109092907e-07, + "loss": 0.7001, + "step": 26567 + }, + { + "epoch": 0.97, + "learning_rate": 1.02889369026257e-07, + "loss": 0.6782, + "step": 26568 + }, + { + "epoch": 0.97, + "learning_rate": 1.0288570696158493e-07, + "loss": 0.6949, + "step": 26569 + }, + { + "epoch": 0.97, + "learning_rate": 1.0288204489691287e-07, + "loss": 0.6868, + "step": 26570 + }, + { + "epoch": 0.97, + "learning_rate": 1.0287838283224081e-07, + "loss": 0.6888, + "step": 26571 + }, + { + "epoch": 0.97, + "learning_rate": 1.0287472076756875e-07, + "loss": 0.6852, + "step": 26572 + }, + { + "epoch": 0.97, + "learning_rate": 1.0287105870289669e-07, + "loss": 0.6938, + "step": 26573 + }, + { + "epoch": 0.97, + "learning_rate": 1.0286739663822463e-07, + "loss": 0.6958, + "step": 26574 + }, + { + "epoch": 0.97, + "learning_rate": 1.0286373457355257e-07, + "loss": 0.6814, + "step": 26575 + }, + { + "epoch": 0.97, + "learning_rate": 1.0286007250888051e-07, + "loss": 0.6884, + "step": 26576 + }, + { + "epoch": 0.97, + "learning_rate": 1.0285641044420843e-07, + "loss": 0.68, + "step": 26577 + }, + { + "epoch": 0.97, + "learning_rate": 1.0285274837953637e-07, + "loss": 0.6891, + "step": 26578 + }, + { + "epoch": 0.97, + "learning_rate": 1.028490863148643e-07, + "loss": 0.7112, + "step": 26579 + }, + { + "epoch": 0.97, + "learning_rate": 1.0284542425019225e-07, + "loss": 0.6962, + "step": 26580 + }, + { + "epoch": 0.97, + "learning_rate": 1.0284176218552019e-07, + "loss": 0.7039, + "step": 26581 + }, + { + "epoch": 0.97, + "learning_rate": 1.0283810012084813e-07, + "loss": 0.6996, + "step": 26582 + }, + { + "epoch": 0.97, + "learning_rate": 1.0283443805617607e-07, + "loss": 0.6894, + "step": 26583 + }, + { + "epoch": 0.97, + "learning_rate": 1.02830775991504e-07, + "loss": 0.6857, + "step": 26584 + }, + { + "epoch": 0.97, + "learning_rate": 1.0282711392683195e-07, + "loss": 0.6823, + "step": 26585 + }, + { + "epoch": 0.97, + "learning_rate": 1.0282345186215989e-07, + "loss": 0.6954, + "step": 26586 + }, + { + "epoch": 0.97, + "learning_rate": 1.0281978979748783e-07, + "loss": 0.6937, + "step": 26587 + }, + { + "epoch": 0.97, + "learning_rate": 1.0281612773281577e-07, + "loss": 0.6938, + "step": 26588 + }, + { + "epoch": 0.97, + "learning_rate": 1.028124656681437e-07, + "loss": 0.6765, + "step": 26589 + }, + { + "epoch": 0.97, + "learning_rate": 1.0280880360347165e-07, + "loss": 0.6947, + "step": 26590 + }, + { + "epoch": 0.97, + "learning_rate": 1.0280514153879957e-07, + "loss": 0.6884, + "step": 26591 + }, + { + "epoch": 0.97, + "learning_rate": 1.0280147947412751e-07, + "loss": 0.6895, + "step": 26592 + }, + { + "epoch": 0.97, + "learning_rate": 1.0279781740945545e-07, + "loss": 0.6964, + "step": 26593 + }, + { + "epoch": 0.97, + "learning_rate": 1.0279415534478338e-07, + "loss": 0.6867, + "step": 26594 + }, + { + "epoch": 0.97, + "learning_rate": 1.0279049328011132e-07, + "loss": 0.7012, + "step": 26595 + }, + { + "epoch": 0.97, + "learning_rate": 1.0278683121543925e-07, + "loss": 0.6891, + "step": 26596 + }, + { + "epoch": 0.97, + "learning_rate": 1.0278316915076718e-07, + "loss": 0.6828, + "step": 26597 + }, + { + "epoch": 0.97, + "learning_rate": 1.0277950708609512e-07, + "loss": 0.6856, + "step": 26598 + }, + { + "epoch": 0.97, + "learning_rate": 1.0277584502142306e-07, + "loss": 0.7012, + "step": 26599 + }, + { + "epoch": 0.97, + "learning_rate": 1.02772182956751e-07, + "loss": 0.6851, + "step": 26600 + }, + { + "epoch": 0.97, + "learning_rate": 1.0276852089207894e-07, + "loss": 0.6982, + "step": 26601 + }, + { + "epoch": 0.97, + "learning_rate": 1.0276485882740688e-07, + "loss": 0.7057, + "step": 26602 + }, + { + "epoch": 0.97, + "learning_rate": 1.0276119676273482e-07, + "loss": 0.6935, + "step": 26603 + }, + { + "epoch": 0.97, + "learning_rate": 1.0275753469806276e-07, + "loss": 0.6885, + "step": 26604 + }, + { + "epoch": 0.97, + "learning_rate": 1.027538726333907e-07, + "loss": 0.6912, + "step": 26605 + }, + { + "epoch": 0.97, + "learning_rate": 1.0275021056871864e-07, + "loss": 0.7046, + "step": 26606 + }, + { + "epoch": 0.97, + "learning_rate": 1.0274654850404658e-07, + "loss": 0.6984, + "step": 26607 + }, + { + "epoch": 0.97, + "learning_rate": 1.0274288643937452e-07, + "loss": 0.6893, + "step": 26608 + }, + { + "epoch": 0.97, + "learning_rate": 1.0273922437470246e-07, + "loss": 0.7065, + "step": 26609 + }, + { + "epoch": 0.97, + "learning_rate": 1.027355623100304e-07, + "loss": 0.6979, + "step": 26610 + }, + { + "epoch": 0.97, + "learning_rate": 1.0273190024535834e-07, + "loss": 0.7082, + "step": 26611 + }, + { + "epoch": 0.97, + "learning_rate": 1.0272823818068626e-07, + "loss": 0.6881, + "step": 26612 + }, + { + "epoch": 0.97, + "learning_rate": 1.027245761160142e-07, + "loss": 0.7222, + "step": 26613 + }, + { + "epoch": 0.97, + "learning_rate": 1.0272091405134214e-07, + "loss": 0.694, + "step": 26614 + }, + { + "epoch": 0.97, + "learning_rate": 1.0271725198667008e-07, + "loss": 0.7028, + "step": 26615 + }, + { + "epoch": 0.97, + "learning_rate": 1.0271358992199802e-07, + "loss": 0.6971, + "step": 26616 + }, + { + "epoch": 0.97, + "learning_rate": 1.0270992785732596e-07, + "loss": 0.6797, + "step": 26617 + }, + { + "epoch": 0.97, + "learning_rate": 1.0270626579265388e-07, + "loss": 0.6903, + "step": 26618 + }, + { + "epoch": 0.97, + "learning_rate": 1.0270260372798182e-07, + "loss": 0.6918, + "step": 26619 + }, + { + "epoch": 0.97, + "learning_rate": 1.0269894166330976e-07, + "loss": 0.7009, + "step": 26620 + }, + { + "epoch": 0.97, + "learning_rate": 1.026952795986377e-07, + "loss": 0.6878, + "step": 26621 + }, + { + "epoch": 0.97, + "learning_rate": 1.0269161753396564e-07, + "loss": 0.6963, + "step": 26622 + }, + { + "epoch": 0.97, + "learning_rate": 1.0268795546929358e-07, + "loss": 0.7216, + "step": 26623 + }, + { + "epoch": 0.97, + "learning_rate": 1.0268429340462152e-07, + "loss": 0.7137, + "step": 26624 + }, + { + "epoch": 0.97, + "learning_rate": 1.0268063133994946e-07, + "loss": 0.6738, + "step": 26625 + }, + { + "epoch": 0.97, + "learning_rate": 1.026769692752774e-07, + "loss": 0.6944, + "step": 26626 + }, + { + "epoch": 0.97, + "learning_rate": 1.0267330721060534e-07, + "loss": 0.7052, + "step": 26627 + }, + { + "epoch": 0.97, + "learning_rate": 1.0266964514593328e-07, + "loss": 0.675, + "step": 26628 + }, + { + "epoch": 0.97, + "learning_rate": 1.0266598308126122e-07, + "loss": 0.6902, + "step": 26629 + }, + { + "epoch": 0.97, + "learning_rate": 1.0266232101658914e-07, + "loss": 0.7043, + "step": 26630 + }, + { + "epoch": 0.97, + "learning_rate": 1.0265865895191708e-07, + "loss": 0.6922, + "step": 26631 + }, + { + "epoch": 0.97, + "learning_rate": 1.0265499688724502e-07, + "loss": 0.6982, + "step": 26632 + }, + { + "epoch": 0.97, + "learning_rate": 1.0265133482257296e-07, + "loss": 0.6932, + "step": 26633 + }, + { + "epoch": 0.97, + "learning_rate": 1.026476727579009e-07, + "loss": 0.6803, + "step": 26634 + }, + { + "epoch": 0.97, + "learning_rate": 1.0264401069322884e-07, + "loss": 0.701, + "step": 26635 + }, + { + "epoch": 0.97, + "learning_rate": 1.0264034862855678e-07, + "loss": 0.6859, + "step": 26636 + }, + { + "epoch": 0.97, + "learning_rate": 1.0263668656388472e-07, + "loss": 0.6952, + "step": 26637 + }, + { + "epoch": 0.97, + "learning_rate": 1.0263302449921266e-07, + "loss": 0.6881, + "step": 26638 + }, + { + "epoch": 0.97, + "learning_rate": 1.026293624345406e-07, + "loss": 0.6856, + "step": 26639 + }, + { + "epoch": 0.97, + "learning_rate": 1.0262570036986852e-07, + "loss": 0.7017, + "step": 26640 + }, + { + "epoch": 0.97, + "learning_rate": 1.0262203830519646e-07, + "loss": 0.6797, + "step": 26641 + }, + { + "epoch": 0.97, + "learning_rate": 1.026183762405244e-07, + "loss": 0.6768, + "step": 26642 + }, + { + "epoch": 0.97, + "learning_rate": 1.0261471417585234e-07, + "loss": 0.7085, + "step": 26643 + }, + { + "epoch": 0.97, + "learning_rate": 1.0261105211118028e-07, + "loss": 0.6971, + "step": 26644 + }, + { + "epoch": 0.97, + "learning_rate": 1.0260739004650822e-07, + "loss": 0.6908, + "step": 26645 + }, + { + "epoch": 0.97, + "learning_rate": 1.0260372798183616e-07, + "loss": 0.7014, + "step": 26646 + }, + { + "epoch": 0.97, + "learning_rate": 1.026000659171641e-07, + "loss": 0.6992, + "step": 26647 + }, + { + "epoch": 0.97, + "learning_rate": 1.0259640385249202e-07, + "loss": 0.6949, + "step": 26648 + }, + { + "epoch": 0.98, + "learning_rate": 1.0259274178781996e-07, + "loss": 0.6907, + "step": 26649 + }, + { + "epoch": 0.98, + "learning_rate": 1.025890797231479e-07, + "loss": 0.6834, + "step": 26650 + }, + { + "epoch": 0.98, + "learning_rate": 1.0258541765847584e-07, + "loss": 0.7038, + "step": 26651 + }, + { + "epoch": 0.98, + "learning_rate": 1.0258175559380378e-07, + "loss": 0.6998, + "step": 26652 + }, + { + "epoch": 0.98, + "learning_rate": 1.0257809352913172e-07, + "loss": 0.6798, + "step": 26653 + }, + { + "epoch": 0.98, + "learning_rate": 1.0257443146445966e-07, + "loss": 0.6887, + "step": 26654 + }, + { + "epoch": 0.98, + "learning_rate": 1.025707693997876e-07, + "loss": 0.7072, + "step": 26655 + }, + { + "epoch": 0.98, + "learning_rate": 1.0256710733511553e-07, + "loss": 0.6954, + "step": 26656 + }, + { + "epoch": 0.98, + "learning_rate": 1.0256344527044347e-07, + "loss": 0.6915, + "step": 26657 + }, + { + "epoch": 0.98, + "learning_rate": 1.0255978320577141e-07, + "loss": 0.6885, + "step": 26658 + }, + { + "epoch": 0.98, + "learning_rate": 1.0255612114109935e-07, + "loss": 0.6891, + "step": 26659 + }, + { + "epoch": 0.98, + "learning_rate": 1.025524590764273e-07, + "loss": 0.6813, + "step": 26660 + }, + { + "epoch": 0.98, + "learning_rate": 1.0254879701175523e-07, + "loss": 0.7045, + "step": 26661 + }, + { + "epoch": 0.98, + "learning_rate": 1.0254513494708316e-07, + "loss": 0.6959, + "step": 26662 + }, + { + "epoch": 0.98, + "learning_rate": 1.025414728824111e-07, + "loss": 0.6938, + "step": 26663 + }, + { + "epoch": 0.98, + "learning_rate": 1.0253781081773904e-07, + "loss": 0.6831, + "step": 26664 + }, + { + "epoch": 0.98, + "learning_rate": 1.0253414875306698e-07, + "loss": 0.6904, + "step": 26665 + }, + { + "epoch": 0.98, + "learning_rate": 1.0253048668839491e-07, + "loss": 0.6882, + "step": 26666 + }, + { + "epoch": 0.98, + "learning_rate": 1.0252682462372283e-07, + "loss": 0.6963, + "step": 26667 + }, + { + "epoch": 0.98, + "learning_rate": 1.0252316255905077e-07, + "loss": 0.6911, + "step": 26668 + }, + { + "epoch": 0.98, + "learning_rate": 1.0251950049437871e-07, + "loss": 0.6955, + "step": 26669 + }, + { + "epoch": 0.98, + "learning_rate": 1.0251583842970665e-07, + "loss": 0.699, + "step": 26670 + }, + { + "epoch": 0.98, + "learning_rate": 1.025121763650346e-07, + "loss": 0.7037, + "step": 26671 + }, + { + "epoch": 0.98, + "learning_rate": 1.0250851430036253e-07, + "loss": 0.6946, + "step": 26672 + }, + { + "epoch": 0.98, + "learning_rate": 1.0250485223569047e-07, + "loss": 0.6813, + "step": 26673 + }, + { + "epoch": 0.98, + "learning_rate": 1.0250119017101841e-07, + "loss": 0.6889, + "step": 26674 + }, + { + "epoch": 0.98, + "learning_rate": 1.0249752810634635e-07, + "loss": 0.6983, + "step": 26675 + }, + { + "epoch": 0.98, + "learning_rate": 1.024938660416743e-07, + "loss": 0.6935, + "step": 26676 + }, + { + "epoch": 0.98, + "learning_rate": 1.0249020397700223e-07, + "loss": 0.7049, + "step": 26677 + }, + { + "epoch": 0.98, + "learning_rate": 1.0248654191233017e-07, + "loss": 0.6914, + "step": 26678 + }, + { + "epoch": 0.98, + "learning_rate": 1.0248287984765811e-07, + "loss": 0.7014, + "step": 26679 + }, + { + "epoch": 0.98, + "learning_rate": 1.0247921778298605e-07, + "loss": 0.7036, + "step": 26680 + }, + { + "epoch": 0.98, + "learning_rate": 1.02475555718314e-07, + "loss": 0.6882, + "step": 26681 + }, + { + "epoch": 0.98, + "learning_rate": 1.0247189365364193e-07, + "loss": 0.6923, + "step": 26682 + }, + { + "epoch": 0.98, + "learning_rate": 1.0246823158896987e-07, + "loss": 0.6911, + "step": 26683 + }, + { + "epoch": 0.98, + "learning_rate": 1.0246456952429779e-07, + "loss": 0.6992, + "step": 26684 + }, + { + "epoch": 0.98, + "learning_rate": 1.0246090745962573e-07, + "loss": 0.6965, + "step": 26685 + }, + { + "epoch": 0.98, + "learning_rate": 1.0245724539495367e-07, + "loss": 0.6875, + "step": 26686 + }, + { + "epoch": 0.98, + "learning_rate": 1.024535833302816e-07, + "loss": 0.6983, + "step": 26687 + }, + { + "epoch": 0.98, + "learning_rate": 1.0244992126560955e-07, + "loss": 0.6938, + "step": 26688 + }, + { + "epoch": 0.98, + "learning_rate": 1.0244625920093747e-07, + "loss": 0.704, + "step": 26689 + }, + { + "epoch": 0.98, + "learning_rate": 1.0244259713626541e-07, + "loss": 0.6998, + "step": 26690 + }, + { + "epoch": 0.98, + "learning_rate": 1.0243893507159335e-07, + "loss": 0.6941, + "step": 26691 + }, + { + "epoch": 0.98, + "learning_rate": 1.0243527300692129e-07, + "loss": 0.6879, + "step": 26692 + }, + { + "epoch": 0.98, + "learning_rate": 1.0243161094224923e-07, + "loss": 0.6944, + "step": 26693 + }, + { + "epoch": 0.98, + "learning_rate": 1.0242794887757717e-07, + "loss": 0.6939, + "step": 26694 + }, + { + "epoch": 0.98, + "learning_rate": 1.0242428681290511e-07, + "loss": 0.6999, + "step": 26695 + }, + { + "epoch": 0.98, + "learning_rate": 1.0242062474823305e-07, + "loss": 0.6713, + "step": 26696 + }, + { + "epoch": 0.98, + "learning_rate": 1.0241696268356099e-07, + "loss": 0.6923, + "step": 26697 + }, + { + "epoch": 0.98, + "learning_rate": 1.0241330061888893e-07, + "loss": 0.6859, + "step": 26698 + }, + { + "epoch": 0.98, + "learning_rate": 1.0240963855421687e-07, + "loss": 0.6968, + "step": 26699 + }, + { + "epoch": 0.98, + "learning_rate": 1.0240597648954481e-07, + "loss": 0.6958, + "step": 26700 + }, + { + "epoch": 0.98, + "learning_rate": 1.0240231442487275e-07, + "loss": 0.6894, + "step": 26701 + }, + { + "epoch": 0.98, + "learning_rate": 1.0239865236020067e-07, + "loss": 0.6908, + "step": 26702 + }, + { + "epoch": 0.98, + "learning_rate": 1.023949902955286e-07, + "loss": 0.6877, + "step": 26703 + }, + { + "epoch": 0.98, + "learning_rate": 1.0239132823085655e-07, + "loss": 0.7086, + "step": 26704 + }, + { + "epoch": 0.98, + "learning_rate": 1.0238766616618449e-07, + "loss": 0.6948, + "step": 26705 + }, + { + "epoch": 0.98, + "learning_rate": 1.0238400410151243e-07, + "loss": 0.6839, + "step": 26706 + }, + { + "epoch": 0.98, + "learning_rate": 1.0238034203684037e-07, + "loss": 0.692, + "step": 26707 + }, + { + "epoch": 0.98, + "learning_rate": 1.023766799721683e-07, + "loss": 0.6951, + "step": 26708 + }, + { + "epoch": 0.98, + "learning_rate": 1.0237301790749625e-07, + "loss": 0.6999, + "step": 26709 + }, + { + "epoch": 0.98, + "learning_rate": 1.0236935584282419e-07, + "loss": 0.6895, + "step": 26710 + }, + { + "epoch": 0.98, + "learning_rate": 1.0236569377815211e-07, + "loss": 0.7078, + "step": 26711 + }, + { + "epoch": 0.98, + "learning_rate": 1.0236203171348005e-07, + "loss": 0.693, + "step": 26712 + }, + { + "epoch": 0.98, + "learning_rate": 1.0235836964880799e-07, + "loss": 0.6909, + "step": 26713 + }, + { + "epoch": 0.98, + "learning_rate": 1.0235470758413593e-07, + "loss": 0.6715, + "step": 26714 + }, + { + "epoch": 0.98, + "learning_rate": 1.0235104551946387e-07, + "loss": 0.6912, + "step": 26715 + }, + { + "epoch": 0.98, + "learning_rate": 1.0234738345479181e-07, + "loss": 0.6951, + "step": 26716 + }, + { + "epoch": 0.98, + "learning_rate": 1.0234372139011975e-07, + "loss": 0.6925, + "step": 26717 + }, + { + "epoch": 0.98, + "learning_rate": 1.0234005932544769e-07, + "loss": 0.6951, + "step": 26718 + }, + { + "epoch": 0.98, + "learning_rate": 1.0233639726077563e-07, + "loss": 0.6812, + "step": 26719 + }, + { + "epoch": 0.98, + "learning_rate": 1.0233273519610354e-07, + "loss": 0.6919, + "step": 26720 + }, + { + "epoch": 0.98, + "learning_rate": 1.0232907313143148e-07, + "loss": 0.6993, + "step": 26721 + }, + { + "epoch": 0.98, + "learning_rate": 1.0232541106675942e-07, + "loss": 0.6836, + "step": 26722 + }, + { + "epoch": 0.98, + "learning_rate": 1.0232174900208736e-07, + "loss": 0.6937, + "step": 26723 + }, + { + "epoch": 0.98, + "learning_rate": 1.023180869374153e-07, + "loss": 0.6888, + "step": 26724 + }, + { + "epoch": 0.98, + "learning_rate": 1.0231442487274324e-07, + "loss": 0.7018, + "step": 26725 + }, + { + "epoch": 0.98, + "learning_rate": 1.0231076280807118e-07, + "loss": 0.6896, + "step": 26726 + }, + { + "epoch": 0.98, + "learning_rate": 1.0230710074339912e-07, + "loss": 0.6856, + "step": 26727 + }, + { + "epoch": 0.98, + "learning_rate": 1.0230343867872706e-07, + "loss": 0.6718, + "step": 26728 + }, + { + "epoch": 0.98, + "learning_rate": 1.02299776614055e-07, + "loss": 0.6973, + "step": 26729 + }, + { + "epoch": 0.98, + "learning_rate": 1.0229611454938294e-07, + "loss": 0.6859, + "step": 26730 + }, + { + "epoch": 0.98, + "learning_rate": 1.0229245248471088e-07, + "loss": 0.702, + "step": 26731 + }, + { + "epoch": 0.98, + "learning_rate": 1.0228879042003882e-07, + "loss": 0.6877, + "step": 26732 + }, + { + "epoch": 0.98, + "learning_rate": 1.0228512835536675e-07, + "loss": 0.6929, + "step": 26733 + }, + { + "epoch": 0.98, + "learning_rate": 1.0228146629069469e-07, + "loss": 0.6959, + "step": 26734 + }, + { + "epoch": 0.98, + "learning_rate": 1.0227780422602263e-07, + "loss": 0.689, + "step": 26735 + }, + { + "epoch": 0.98, + "learning_rate": 1.0227414216135057e-07, + "loss": 0.6896, + "step": 26736 + }, + { + "epoch": 0.98, + "learning_rate": 1.0227048009667851e-07, + "loss": 0.6768, + "step": 26737 + }, + { + "epoch": 0.98, + "learning_rate": 1.0226681803200642e-07, + "loss": 0.6861, + "step": 26738 + }, + { + "epoch": 0.98, + "learning_rate": 1.0226315596733436e-07, + "loss": 0.6925, + "step": 26739 + }, + { + "epoch": 0.98, + "learning_rate": 1.022594939026623e-07, + "loss": 0.6979, + "step": 26740 + }, + { + "epoch": 0.98, + "learning_rate": 1.0225583183799024e-07, + "loss": 0.6985, + "step": 26741 + }, + { + "epoch": 0.98, + "learning_rate": 1.0225216977331818e-07, + "loss": 0.6939, + "step": 26742 + }, + { + "epoch": 0.98, + "learning_rate": 1.0224850770864612e-07, + "loss": 0.7094, + "step": 26743 + }, + { + "epoch": 0.98, + "learning_rate": 1.0224484564397406e-07, + "loss": 0.6939, + "step": 26744 + }, + { + "epoch": 0.98, + "learning_rate": 1.02241183579302e-07, + "loss": 0.669, + "step": 26745 + }, + { + "epoch": 0.98, + "learning_rate": 1.0223752151462994e-07, + "loss": 0.6955, + "step": 26746 + }, + { + "epoch": 0.98, + "learning_rate": 1.0223385944995788e-07, + "loss": 0.6962, + "step": 26747 + }, + { + "epoch": 0.98, + "learning_rate": 1.0223019738528582e-07, + "loss": 0.6942, + "step": 26748 + }, + { + "epoch": 0.98, + "learning_rate": 1.0222653532061376e-07, + "loss": 0.6939, + "step": 26749 + }, + { + "epoch": 0.98, + "learning_rate": 1.022228732559417e-07, + "loss": 0.7015, + "step": 26750 + }, + { + "epoch": 0.98, + "learning_rate": 1.0221921119126964e-07, + "loss": 0.6937, + "step": 26751 + }, + { + "epoch": 0.98, + "learning_rate": 1.0221554912659758e-07, + "loss": 0.7004, + "step": 26752 + }, + { + "epoch": 0.98, + "learning_rate": 1.0221188706192552e-07, + "loss": 0.6893, + "step": 26753 + }, + { + "epoch": 0.98, + "learning_rate": 1.0220822499725346e-07, + "loss": 0.695, + "step": 26754 + }, + { + "epoch": 0.98, + "learning_rate": 1.0220456293258139e-07, + "loss": 0.6956, + "step": 26755 + }, + { + "epoch": 0.98, + "learning_rate": 1.0220090086790932e-07, + "loss": 0.7083, + "step": 26756 + }, + { + "epoch": 0.98, + "learning_rate": 1.0219723880323726e-07, + "loss": 0.6843, + "step": 26757 + }, + { + "epoch": 0.98, + "learning_rate": 1.021935767385652e-07, + "loss": 0.6996, + "step": 26758 + }, + { + "epoch": 0.98, + "learning_rate": 1.0218991467389314e-07, + "loss": 0.6873, + "step": 26759 + }, + { + "epoch": 0.98, + "learning_rate": 1.0218625260922106e-07, + "loss": 0.6831, + "step": 26760 + }, + { + "epoch": 0.98, + "learning_rate": 1.02182590544549e-07, + "loss": 0.6804, + "step": 26761 + }, + { + "epoch": 0.98, + "learning_rate": 1.0217892847987694e-07, + "loss": 0.6914, + "step": 26762 + }, + { + "epoch": 0.98, + "learning_rate": 1.0217526641520488e-07, + "loss": 0.6899, + "step": 26763 + }, + { + "epoch": 0.98, + "learning_rate": 1.0217160435053282e-07, + "loss": 0.6903, + "step": 26764 + }, + { + "epoch": 0.98, + "learning_rate": 1.0216794228586076e-07, + "loss": 0.6797, + "step": 26765 + }, + { + "epoch": 0.98, + "learning_rate": 1.021642802211887e-07, + "loss": 0.6907, + "step": 26766 + }, + { + "epoch": 0.98, + "learning_rate": 1.0216061815651664e-07, + "loss": 0.6991, + "step": 26767 + }, + { + "epoch": 0.98, + "learning_rate": 1.0215695609184458e-07, + "loss": 0.6971, + "step": 26768 + }, + { + "epoch": 0.98, + "learning_rate": 1.0215329402717252e-07, + "loss": 0.6928, + "step": 26769 + }, + { + "epoch": 0.98, + "learning_rate": 1.0214963196250046e-07, + "loss": 0.6905, + "step": 26770 + }, + { + "epoch": 0.98, + "learning_rate": 1.021459698978284e-07, + "loss": 0.6946, + "step": 26771 + }, + { + "epoch": 0.98, + "learning_rate": 1.0214230783315634e-07, + "loss": 0.688, + "step": 26772 + }, + { + "epoch": 0.98, + "learning_rate": 1.0213864576848428e-07, + "loss": 0.7001, + "step": 26773 + }, + { + "epoch": 0.98, + "learning_rate": 1.021349837038122e-07, + "loss": 0.6906, + "step": 26774 + }, + { + "epoch": 0.98, + "learning_rate": 1.0213132163914013e-07, + "loss": 0.6867, + "step": 26775 + }, + { + "epoch": 0.98, + "learning_rate": 1.0212765957446807e-07, + "loss": 0.6942, + "step": 26776 + }, + { + "epoch": 0.98, + "learning_rate": 1.0212399750979601e-07, + "loss": 0.6896, + "step": 26777 + }, + { + "epoch": 0.98, + "learning_rate": 1.0212033544512395e-07, + "loss": 0.7068, + "step": 26778 + }, + { + "epoch": 0.98, + "learning_rate": 1.021166733804519e-07, + "loss": 0.6895, + "step": 26779 + }, + { + "epoch": 0.98, + "learning_rate": 1.0211301131577983e-07, + "loss": 0.6956, + "step": 26780 + }, + { + "epoch": 0.98, + "learning_rate": 1.0210934925110777e-07, + "loss": 0.7071, + "step": 26781 + }, + { + "epoch": 0.98, + "learning_rate": 1.021056871864357e-07, + "loss": 0.6963, + "step": 26782 + }, + { + "epoch": 0.98, + "learning_rate": 1.0210202512176364e-07, + "loss": 0.7014, + "step": 26783 + }, + { + "epoch": 0.98, + "learning_rate": 1.0209836305709158e-07, + "loss": 0.6936, + "step": 26784 + }, + { + "epoch": 0.98, + "learning_rate": 1.0209470099241952e-07, + "loss": 0.6946, + "step": 26785 + }, + { + "epoch": 0.98, + "learning_rate": 1.0209103892774746e-07, + "loss": 0.6909, + "step": 26786 + }, + { + "epoch": 0.98, + "learning_rate": 1.020873768630754e-07, + "loss": 0.6899, + "step": 26787 + }, + { + "epoch": 0.98, + "learning_rate": 1.0208371479840334e-07, + "loss": 0.688, + "step": 26788 + }, + { + "epoch": 0.98, + "learning_rate": 1.0208005273373128e-07, + "loss": 0.6881, + "step": 26789 + }, + { + "epoch": 0.98, + "learning_rate": 1.0207639066905922e-07, + "loss": 0.6889, + "step": 26790 + }, + { + "epoch": 0.98, + "learning_rate": 1.0207272860438716e-07, + "loss": 0.6937, + "step": 26791 + }, + { + "epoch": 0.98, + "learning_rate": 1.0206906653971507e-07, + "loss": 0.7052, + "step": 26792 + }, + { + "epoch": 0.98, + "learning_rate": 1.0206540447504301e-07, + "loss": 0.6904, + "step": 26793 + }, + { + "epoch": 0.98, + "learning_rate": 1.0206174241037095e-07, + "loss": 0.6789, + "step": 26794 + }, + { + "epoch": 0.98, + "learning_rate": 1.020580803456989e-07, + "loss": 0.7009, + "step": 26795 + }, + { + "epoch": 0.98, + "learning_rate": 1.0205441828102683e-07, + "loss": 0.7074, + "step": 26796 + }, + { + "epoch": 0.98, + "learning_rate": 1.0205075621635477e-07, + "loss": 0.6913, + "step": 26797 + }, + { + "epoch": 0.98, + "learning_rate": 1.0204709415168271e-07, + "loss": 0.6844, + "step": 26798 + }, + { + "epoch": 0.98, + "learning_rate": 1.0204343208701065e-07, + "loss": 0.6906, + "step": 26799 + }, + { + "epoch": 0.98, + "learning_rate": 1.020397700223386e-07, + "loss": 0.6888, + "step": 26800 + }, + { + "epoch": 0.98, + "learning_rate": 1.0203610795766653e-07, + "loss": 0.6974, + "step": 26801 + }, + { + "epoch": 0.98, + "learning_rate": 1.0203244589299447e-07, + "loss": 0.7022, + "step": 26802 + }, + { + "epoch": 0.98, + "learning_rate": 1.0202878382832241e-07, + "loss": 0.7013, + "step": 26803 + }, + { + "epoch": 0.98, + "learning_rate": 1.0202512176365034e-07, + "loss": 0.6931, + "step": 26804 + }, + { + "epoch": 0.98, + "learning_rate": 1.0202145969897828e-07, + "loss": 0.6881, + "step": 26805 + }, + { + "epoch": 0.98, + "learning_rate": 1.0201779763430622e-07, + "loss": 0.6947, + "step": 26806 + }, + { + "epoch": 0.98, + "learning_rate": 1.0201413556963416e-07, + "loss": 0.6938, + "step": 26807 + }, + { + "epoch": 0.98, + "learning_rate": 1.020104735049621e-07, + "loss": 0.7065, + "step": 26808 + }, + { + "epoch": 0.98, + "learning_rate": 1.0200681144029004e-07, + "loss": 0.6916, + "step": 26809 + }, + { + "epoch": 0.98, + "learning_rate": 1.0200314937561795e-07, + "loss": 0.695, + "step": 26810 + }, + { + "epoch": 0.98, + "learning_rate": 1.0199948731094589e-07, + "loss": 0.695, + "step": 26811 + }, + { + "epoch": 0.98, + "learning_rate": 1.0199582524627383e-07, + "loss": 0.691, + "step": 26812 + }, + { + "epoch": 0.98, + "learning_rate": 1.0199216318160177e-07, + "loss": 0.6969, + "step": 26813 + }, + { + "epoch": 0.98, + "learning_rate": 1.0198850111692971e-07, + "loss": 0.6951, + "step": 26814 + }, + { + "epoch": 0.98, + "learning_rate": 1.0198483905225765e-07, + "loss": 0.6844, + "step": 26815 + }, + { + "epoch": 0.98, + "learning_rate": 1.0198117698758559e-07, + "loss": 0.6925, + "step": 26816 + }, + { + "epoch": 0.98, + "learning_rate": 1.0197751492291353e-07, + "loss": 0.6904, + "step": 26817 + }, + { + "epoch": 0.98, + "learning_rate": 1.0197385285824147e-07, + "loss": 0.7058, + "step": 26818 + }, + { + "epoch": 0.98, + "learning_rate": 1.0197019079356941e-07, + "loss": 0.6964, + "step": 26819 + }, + { + "epoch": 0.98, + "learning_rate": 1.0196652872889735e-07, + "loss": 0.6958, + "step": 26820 + }, + { + "epoch": 0.98, + "learning_rate": 1.0196286666422529e-07, + "loss": 0.6955, + "step": 26821 + }, + { + "epoch": 0.98, + "learning_rate": 1.0195920459955323e-07, + "loss": 0.6959, + "step": 26822 + }, + { + "epoch": 0.98, + "learning_rate": 1.0195554253488117e-07, + "loss": 0.6972, + "step": 26823 + }, + { + "epoch": 0.98, + "learning_rate": 1.0195188047020911e-07, + "loss": 0.6877, + "step": 26824 + }, + { + "epoch": 0.98, + "learning_rate": 1.0194821840553705e-07, + "loss": 0.6995, + "step": 26825 + }, + { + "epoch": 0.98, + "learning_rate": 1.0194455634086498e-07, + "loss": 0.699, + "step": 26826 + }, + { + "epoch": 0.98, + "learning_rate": 1.0194089427619292e-07, + "loss": 0.6913, + "step": 26827 + }, + { + "epoch": 0.98, + "learning_rate": 1.0193723221152085e-07, + "loss": 0.6942, + "step": 26828 + }, + { + "epoch": 0.98, + "learning_rate": 1.0193357014684879e-07, + "loss": 0.6917, + "step": 26829 + }, + { + "epoch": 0.98, + "learning_rate": 1.0192990808217673e-07, + "loss": 0.6914, + "step": 26830 + }, + { + "epoch": 0.98, + "learning_rate": 1.0192624601750465e-07, + "loss": 0.7001, + "step": 26831 + }, + { + "epoch": 0.98, + "learning_rate": 1.0192258395283259e-07, + "loss": 0.7077, + "step": 26832 + }, + { + "epoch": 0.98, + "learning_rate": 1.0191892188816053e-07, + "loss": 0.6868, + "step": 26833 + }, + { + "epoch": 0.98, + "learning_rate": 1.0191525982348847e-07, + "loss": 0.6942, + "step": 26834 + }, + { + "epoch": 0.98, + "learning_rate": 1.0191159775881641e-07, + "loss": 0.6957, + "step": 26835 + }, + { + "epoch": 0.98, + "learning_rate": 1.0190793569414435e-07, + "loss": 0.6994, + "step": 26836 + }, + { + "epoch": 0.98, + "learning_rate": 1.0190427362947229e-07, + "loss": 0.6958, + "step": 26837 + }, + { + "epoch": 0.98, + "learning_rate": 1.0190061156480023e-07, + "loss": 0.6999, + "step": 26838 + }, + { + "epoch": 0.98, + "learning_rate": 1.0189694950012817e-07, + "loss": 0.6909, + "step": 26839 + }, + { + "epoch": 0.98, + "learning_rate": 1.0189328743545611e-07, + "loss": 0.7011, + "step": 26840 + }, + { + "epoch": 0.98, + "learning_rate": 1.0188962537078405e-07, + "loss": 0.7018, + "step": 26841 + }, + { + "epoch": 0.98, + "learning_rate": 1.0188596330611199e-07, + "loss": 0.6922, + "step": 26842 + }, + { + "epoch": 0.98, + "learning_rate": 1.0188230124143993e-07, + "loss": 0.6853, + "step": 26843 + }, + { + "epoch": 0.98, + "learning_rate": 1.0187863917676787e-07, + "loss": 0.68, + "step": 26844 + }, + { + "epoch": 0.98, + "learning_rate": 1.0187497711209581e-07, + "loss": 0.6773, + "step": 26845 + }, + { + "epoch": 0.98, + "learning_rate": 1.0187131504742372e-07, + "loss": 0.708, + "step": 26846 + }, + { + "epoch": 0.98, + "learning_rate": 1.0186765298275166e-07, + "loss": 0.6891, + "step": 26847 + }, + { + "epoch": 0.98, + "learning_rate": 1.018639909180796e-07, + "loss": 0.6722, + "step": 26848 + }, + { + "epoch": 0.98, + "learning_rate": 1.0186032885340754e-07, + "loss": 0.6806, + "step": 26849 + }, + { + "epoch": 0.98, + "learning_rate": 1.0185666678873548e-07, + "loss": 0.6848, + "step": 26850 + }, + { + "epoch": 0.98, + "learning_rate": 1.0185300472406342e-07, + "loss": 0.708, + "step": 26851 + }, + { + "epoch": 0.98, + "learning_rate": 1.0184934265939136e-07, + "loss": 0.7003, + "step": 26852 + }, + { + "epoch": 0.98, + "learning_rate": 1.0184568059471929e-07, + "loss": 0.6936, + "step": 26853 + }, + { + "epoch": 0.98, + "learning_rate": 1.0184201853004723e-07, + "loss": 0.6809, + "step": 26854 + }, + { + "epoch": 0.98, + "learning_rate": 1.0183835646537517e-07, + "loss": 0.6857, + "step": 26855 + }, + { + "epoch": 0.98, + "learning_rate": 1.0183469440070311e-07, + "loss": 0.6998, + "step": 26856 + }, + { + "epoch": 0.98, + "learning_rate": 1.0183103233603105e-07, + "loss": 0.6968, + "step": 26857 + }, + { + "epoch": 0.98, + "learning_rate": 1.0182737027135899e-07, + "loss": 0.6898, + "step": 26858 + }, + { + "epoch": 0.98, + "learning_rate": 1.0182370820668693e-07, + "loss": 0.6906, + "step": 26859 + }, + { + "epoch": 0.98, + "learning_rate": 1.0182004614201487e-07, + "loss": 0.6814, + "step": 26860 + }, + { + "epoch": 0.98, + "learning_rate": 1.0181638407734281e-07, + "loss": 0.6865, + "step": 26861 + }, + { + "epoch": 0.98, + "learning_rate": 1.0181272201267075e-07, + "loss": 0.6894, + "step": 26862 + }, + { + "epoch": 0.98, + "learning_rate": 1.0180905994799869e-07, + "loss": 0.6907, + "step": 26863 + }, + { + "epoch": 0.98, + "learning_rate": 1.018053978833266e-07, + "loss": 0.7107, + "step": 26864 + }, + { + "epoch": 0.98, + "learning_rate": 1.0180173581865454e-07, + "loss": 0.6984, + "step": 26865 + }, + { + "epoch": 0.98, + "learning_rate": 1.0179807375398248e-07, + "loss": 0.6967, + "step": 26866 + }, + { + "epoch": 0.98, + "learning_rate": 1.0179441168931042e-07, + "loss": 0.6949, + "step": 26867 + }, + { + "epoch": 0.98, + "learning_rate": 1.0179074962463836e-07, + "loss": 0.6874, + "step": 26868 + }, + { + "epoch": 0.98, + "learning_rate": 1.017870875599663e-07, + "loss": 0.6948, + "step": 26869 + }, + { + "epoch": 0.98, + "learning_rate": 1.0178342549529424e-07, + "loss": 0.6885, + "step": 26870 + }, + { + "epoch": 0.98, + "learning_rate": 1.0177976343062218e-07, + "loss": 0.6906, + "step": 26871 + }, + { + "epoch": 0.98, + "learning_rate": 1.0177610136595012e-07, + "loss": 0.6937, + "step": 26872 + }, + { + "epoch": 0.98, + "learning_rate": 1.0177243930127806e-07, + "loss": 0.7039, + "step": 26873 + }, + { + "epoch": 0.98, + "learning_rate": 1.01768777236606e-07, + "loss": 0.7031, + "step": 26874 + }, + { + "epoch": 0.98, + "learning_rate": 1.0176511517193393e-07, + "loss": 0.7001, + "step": 26875 + }, + { + "epoch": 0.98, + "learning_rate": 1.0176145310726187e-07, + "loss": 0.6821, + "step": 26876 + }, + { + "epoch": 0.98, + "learning_rate": 1.0175779104258981e-07, + "loss": 0.6904, + "step": 26877 + }, + { + "epoch": 0.98, + "learning_rate": 1.0175412897791775e-07, + "loss": 0.6987, + "step": 26878 + }, + { + "epoch": 0.98, + "learning_rate": 1.0175046691324569e-07, + "loss": 0.6876, + "step": 26879 + }, + { + "epoch": 0.98, + "learning_rate": 1.0174680484857363e-07, + "loss": 0.6937, + "step": 26880 + }, + { + "epoch": 0.98, + "learning_rate": 1.0174314278390157e-07, + "loss": 0.6909, + "step": 26881 + }, + { + "epoch": 0.98, + "learning_rate": 1.0173948071922948e-07, + "loss": 0.6911, + "step": 26882 + }, + { + "epoch": 0.98, + "learning_rate": 1.0173581865455742e-07, + "loss": 0.6908, + "step": 26883 + }, + { + "epoch": 0.98, + "learning_rate": 1.0173215658988536e-07, + "loss": 0.6826, + "step": 26884 + }, + { + "epoch": 0.98, + "learning_rate": 1.017284945252133e-07, + "loss": 0.6944, + "step": 26885 + }, + { + "epoch": 0.98, + "learning_rate": 1.0172483246054124e-07, + "loss": 0.6954, + "step": 26886 + }, + { + "epoch": 0.98, + "learning_rate": 1.0172117039586918e-07, + "loss": 0.6937, + "step": 26887 + }, + { + "epoch": 0.98, + "learning_rate": 1.0171750833119712e-07, + "loss": 0.6903, + "step": 26888 + }, + { + "epoch": 0.98, + "learning_rate": 1.0171384626652506e-07, + "loss": 0.6998, + "step": 26889 + }, + { + "epoch": 0.98, + "learning_rate": 1.01710184201853e-07, + "loss": 0.6807, + "step": 26890 + }, + { + "epoch": 0.98, + "learning_rate": 1.0170652213718094e-07, + "loss": 0.6959, + "step": 26891 + }, + { + "epoch": 0.98, + "learning_rate": 1.0170286007250888e-07, + "loss": 0.7061, + "step": 26892 + }, + { + "epoch": 0.98, + "learning_rate": 1.0169919800783682e-07, + "loss": 0.6896, + "step": 26893 + }, + { + "epoch": 0.98, + "learning_rate": 1.0169553594316476e-07, + "loss": 0.7008, + "step": 26894 + }, + { + "epoch": 0.98, + "learning_rate": 1.016918738784927e-07, + "loss": 0.6916, + "step": 26895 + }, + { + "epoch": 0.98, + "learning_rate": 1.0168821181382063e-07, + "loss": 0.6922, + "step": 26896 + }, + { + "epoch": 0.98, + "learning_rate": 1.0168454974914857e-07, + "loss": 0.6996, + "step": 26897 + }, + { + "epoch": 0.98, + "learning_rate": 1.0168088768447651e-07, + "loss": 0.681, + "step": 26898 + }, + { + "epoch": 0.98, + "learning_rate": 1.0167722561980445e-07, + "loss": 0.6955, + "step": 26899 + }, + { + "epoch": 0.98, + "learning_rate": 1.0167356355513237e-07, + "loss": 0.6838, + "step": 26900 + }, + { + "epoch": 0.98, + "learning_rate": 1.0166990149046031e-07, + "loss": 0.692, + "step": 26901 + }, + { + "epoch": 0.98, + "learning_rate": 1.0166623942578824e-07, + "loss": 0.6912, + "step": 26902 + }, + { + "epoch": 0.98, + "learning_rate": 1.0166257736111618e-07, + "loss": 0.6999, + "step": 26903 + }, + { + "epoch": 0.98, + "learning_rate": 1.0165891529644412e-07, + "loss": 0.7016, + "step": 26904 + }, + { + "epoch": 0.98, + "learning_rate": 1.0165525323177206e-07, + "loss": 0.6917, + "step": 26905 + }, + { + "epoch": 0.98, + "learning_rate": 1.016515911671e-07, + "loss": 0.6966, + "step": 26906 + }, + { + "epoch": 0.98, + "learning_rate": 1.0164792910242794e-07, + "loss": 0.6839, + "step": 26907 + }, + { + "epoch": 0.98, + "learning_rate": 1.0164426703775588e-07, + "loss": 0.702, + "step": 26908 + }, + { + "epoch": 0.98, + "learning_rate": 1.0164060497308382e-07, + "loss": 0.6723, + "step": 26909 + }, + { + "epoch": 0.98, + "learning_rate": 1.0163694290841176e-07, + "loss": 0.704, + "step": 26910 + }, + { + "epoch": 0.98, + "learning_rate": 1.016332808437397e-07, + "loss": 0.6817, + "step": 26911 + }, + { + "epoch": 0.98, + "learning_rate": 1.0162961877906764e-07, + "loss": 0.6982, + "step": 26912 + }, + { + "epoch": 0.98, + "learning_rate": 1.0162595671439558e-07, + "loss": 0.6905, + "step": 26913 + }, + { + "epoch": 0.98, + "learning_rate": 1.0162229464972352e-07, + "loss": 0.7021, + "step": 26914 + }, + { + "epoch": 0.98, + "learning_rate": 1.0161863258505146e-07, + "loss": 0.7016, + "step": 26915 + }, + { + "epoch": 0.98, + "learning_rate": 1.016149705203794e-07, + "loss": 0.6897, + "step": 26916 + }, + { + "epoch": 0.98, + "learning_rate": 1.0161130845570734e-07, + "loss": 0.6994, + "step": 26917 + }, + { + "epoch": 0.98, + "learning_rate": 1.0160764639103525e-07, + "loss": 0.694, + "step": 26918 + }, + { + "epoch": 0.98, + "learning_rate": 1.016039843263632e-07, + "loss": 0.6883, + "step": 26919 + }, + { + "epoch": 0.98, + "learning_rate": 1.0160032226169113e-07, + "loss": 0.6878, + "step": 26920 + }, + { + "epoch": 0.98, + "learning_rate": 1.0159666019701907e-07, + "loss": 0.6941, + "step": 26921 + }, + { + "epoch": 0.98, + "learning_rate": 1.0159299813234701e-07, + "loss": 0.6943, + "step": 26922 + }, + { + "epoch": 0.99, + "learning_rate": 1.0158933606767495e-07, + "loss": 0.6913, + "step": 26923 + }, + { + "epoch": 0.99, + "learning_rate": 1.0158567400300288e-07, + "loss": 0.6735, + "step": 26924 + }, + { + "epoch": 0.99, + "learning_rate": 1.0158201193833082e-07, + "loss": 0.6821, + "step": 26925 + }, + { + "epoch": 0.99, + "learning_rate": 1.0157834987365876e-07, + "loss": 0.6966, + "step": 26926 + }, + { + "epoch": 0.99, + "learning_rate": 1.015746878089867e-07, + "loss": 0.6893, + "step": 26927 + }, + { + "epoch": 0.99, + "learning_rate": 1.0157102574431464e-07, + "loss": 0.7067, + "step": 26928 + }, + { + "epoch": 0.99, + "learning_rate": 1.0156736367964258e-07, + "loss": 0.693, + "step": 26929 + }, + { + "epoch": 0.99, + "learning_rate": 1.0156370161497052e-07, + "loss": 0.6854, + "step": 26930 + }, + { + "epoch": 0.99, + "learning_rate": 1.0156003955029846e-07, + "loss": 0.689, + "step": 26931 + }, + { + "epoch": 0.99, + "learning_rate": 1.015563774856264e-07, + "loss": 0.692, + "step": 26932 + }, + { + "epoch": 0.99, + "learning_rate": 1.0155271542095434e-07, + "loss": 0.6912, + "step": 26933 + }, + { + "epoch": 0.99, + "learning_rate": 1.0154905335628228e-07, + "loss": 0.6932, + "step": 26934 + }, + { + "epoch": 0.99, + "learning_rate": 1.0154539129161022e-07, + "loss": 0.6819, + "step": 26935 + }, + { + "epoch": 0.99, + "learning_rate": 1.0154172922693813e-07, + "loss": 0.6986, + "step": 26936 + }, + { + "epoch": 0.99, + "learning_rate": 1.0153806716226607e-07, + "loss": 0.6903, + "step": 26937 + }, + { + "epoch": 0.99, + "learning_rate": 1.0153440509759401e-07, + "loss": 0.6953, + "step": 26938 + }, + { + "epoch": 0.99, + "learning_rate": 1.0153074303292195e-07, + "loss": 0.685, + "step": 26939 + }, + { + "epoch": 0.99, + "learning_rate": 1.0152708096824989e-07, + "loss": 0.6893, + "step": 26940 + }, + { + "epoch": 0.99, + "learning_rate": 1.0152341890357783e-07, + "loss": 0.686, + "step": 26941 + }, + { + "epoch": 0.99, + "learning_rate": 1.0151975683890577e-07, + "loss": 0.6934, + "step": 26942 + }, + { + "epoch": 0.99, + "learning_rate": 1.0151609477423371e-07, + "loss": 0.6894, + "step": 26943 + }, + { + "epoch": 0.99, + "learning_rate": 1.0151243270956165e-07, + "loss": 0.6933, + "step": 26944 + }, + { + "epoch": 0.99, + "learning_rate": 1.0150877064488959e-07, + "loss": 0.6906, + "step": 26945 + }, + { + "epoch": 0.99, + "learning_rate": 1.0150510858021752e-07, + "loss": 0.6976, + "step": 26946 + }, + { + "epoch": 0.99, + "learning_rate": 1.0150144651554546e-07, + "loss": 0.7022, + "step": 26947 + }, + { + "epoch": 0.99, + "learning_rate": 1.014977844508734e-07, + "loss": 0.6891, + "step": 26948 + }, + { + "epoch": 0.99, + "learning_rate": 1.0149412238620134e-07, + "loss": 0.6767, + "step": 26949 + }, + { + "epoch": 0.99, + "learning_rate": 1.0149046032152928e-07, + "loss": 0.6845, + "step": 26950 + }, + { + "epoch": 0.99, + "learning_rate": 1.0148679825685722e-07, + "loss": 0.6967, + "step": 26951 + }, + { + "epoch": 0.99, + "learning_rate": 1.0148313619218516e-07, + "loss": 0.6943, + "step": 26952 + }, + { + "epoch": 0.99, + "learning_rate": 1.014794741275131e-07, + "loss": 0.6916, + "step": 26953 + }, + { + "epoch": 0.99, + "learning_rate": 1.0147581206284101e-07, + "loss": 0.689, + "step": 26954 + }, + { + "epoch": 0.99, + "learning_rate": 1.0147214999816895e-07, + "loss": 0.6975, + "step": 26955 + }, + { + "epoch": 0.99, + "learning_rate": 1.0146848793349689e-07, + "loss": 0.6861, + "step": 26956 + }, + { + "epoch": 0.99, + "learning_rate": 1.0146482586882483e-07, + "loss": 0.7045, + "step": 26957 + }, + { + "epoch": 0.99, + "learning_rate": 1.0146116380415277e-07, + "loss": 0.704, + "step": 26958 + }, + { + "epoch": 0.99, + "learning_rate": 1.0145750173948071e-07, + "loss": 0.6897, + "step": 26959 + }, + { + "epoch": 0.99, + "learning_rate": 1.0145383967480865e-07, + "loss": 0.6922, + "step": 26960 + }, + { + "epoch": 0.99, + "learning_rate": 1.0145017761013659e-07, + "loss": 0.6877, + "step": 26961 + }, + { + "epoch": 0.99, + "learning_rate": 1.0144651554546453e-07, + "loss": 0.6796, + "step": 26962 + }, + { + "epoch": 0.99, + "learning_rate": 1.0144285348079247e-07, + "loss": 0.6833, + "step": 26963 + }, + { + "epoch": 0.99, + "learning_rate": 1.0143919141612041e-07, + "loss": 0.6955, + "step": 26964 + }, + { + "epoch": 0.99, + "learning_rate": 1.0143552935144835e-07, + "loss": 0.6917, + "step": 26965 + }, + { + "epoch": 0.99, + "learning_rate": 1.0143186728677629e-07, + "loss": 0.6943, + "step": 26966 + }, + { + "epoch": 0.99, + "learning_rate": 1.0142820522210422e-07, + "loss": 0.6854, + "step": 26967 + }, + { + "epoch": 0.99, + "learning_rate": 1.0142454315743216e-07, + "loss": 0.698, + "step": 26968 + }, + { + "epoch": 0.99, + "learning_rate": 1.014208810927601e-07, + "loss": 0.6861, + "step": 26969 + }, + { + "epoch": 0.99, + "learning_rate": 1.0141721902808804e-07, + "loss": 0.6772, + "step": 26970 + }, + { + "epoch": 0.99, + "learning_rate": 1.0141355696341598e-07, + "loss": 0.6821, + "step": 26971 + }, + { + "epoch": 0.99, + "learning_rate": 1.014098948987439e-07, + "loss": 0.6956, + "step": 26972 + }, + { + "epoch": 0.99, + "learning_rate": 1.0140623283407183e-07, + "loss": 0.6743, + "step": 26973 + }, + { + "epoch": 0.99, + "learning_rate": 1.0140257076939977e-07, + "loss": 0.7018, + "step": 26974 + }, + { + "epoch": 0.99, + "learning_rate": 1.0139890870472771e-07, + "loss": 0.6942, + "step": 26975 + }, + { + "epoch": 0.99, + "learning_rate": 1.0139524664005565e-07, + "loss": 0.7025, + "step": 26976 + }, + { + "epoch": 0.99, + "learning_rate": 1.0139158457538359e-07, + "loss": 0.7093, + "step": 26977 + }, + { + "epoch": 0.99, + "learning_rate": 1.0138792251071153e-07, + "loss": 0.6982, + "step": 26978 + }, + { + "epoch": 0.99, + "learning_rate": 1.0138426044603947e-07, + "loss": 0.6981, + "step": 26979 + }, + { + "epoch": 0.99, + "learning_rate": 1.0138059838136741e-07, + "loss": 0.6905, + "step": 26980 + }, + { + "epoch": 0.99, + "learning_rate": 1.0137693631669535e-07, + "loss": 0.7096, + "step": 26981 + }, + { + "epoch": 0.99, + "learning_rate": 1.0137327425202329e-07, + "loss": 0.6973, + "step": 26982 + }, + { + "epoch": 0.99, + "learning_rate": 1.0136961218735123e-07, + "loss": 0.7063, + "step": 26983 + }, + { + "epoch": 0.99, + "learning_rate": 1.0136595012267917e-07, + "loss": 0.6899, + "step": 26984 + }, + { + "epoch": 0.99, + "learning_rate": 1.0136228805800711e-07, + "loss": 0.6858, + "step": 26985 + }, + { + "epoch": 0.99, + "learning_rate": 1.0135862599333505e-07, + "loss": 0.6852, + "step": 26986 + }, + { + "epoch": 0.99, + "learning_rate": 1.0135496392866299e-07, + "loss": 0.6989, + "step": 26987 + }, + { + "epoch": 0.99, + "learning_rate": 1.0135130186399093e-07, + "loss": 0.6759, + "step": 26988 + }, + { + "epoch": 0.99, + "learning_rate": 1.0134763979931884e-07, + "loss": 0.6965, + "step": 26989 + }, + { + "epoch": 0.99, + "learning_rate": 1.0134397773464678e-07, + "loss": 0.6774, + "step": 26990 + }, + { + "epoch": 0.99, + "learning_rate": 1.0134031566997472e-07, + "loss": 0.7005, + "step": 26991 + }, + { + "epoch": 0.99, + "learning_rate": 1.0133665360530266e-07, + "loss": 0.699, + "step": 26992 + }, + { + "epoch": 0.99, + "learning_rate": 1.013329915406306e-07, + "loss": 0.6851, + "step": 26993 + }, + { + "epoch": 0.99, + "learning_rate": 1.0132932947595854e-07, + "loss": 0.6826, + "step": 26994 + }, + { + "epoch": 0.99, + "learning_rate": 1.0132566741128647e-07, + "loss": 0.6933, + "step": 26995 + }, + { + "epoch": 0.99, + "learning_rate": 1.0132200534661441e-07, + "loss": 0.6894, + "step": 26996 + }, + { + "epoch": 0.99, + "learning_rate": 1.0131834328194235e-07, + "loss": 0.6992, + "step": 26997 + }, + { + "epoch": 0.99, + "learning_rate": 1.0131468121727029e-07, + "loss": 0.6905, + "step": 26998 + }, + { + "epoch": 0.99, + "learning_rate": 1.0131101915259823e-07, + "loss": 0.69, + "step": 26999 + }, + { + "epoch": 0.99, + "learning_rate": 1.0130735708792617e-07, + "loss": 0.6735, + "step": 27000 + }, + { + "epoch": 0.99, + "learning_rate": 1.0130369502325411e-07, + "loss": 0.6833, + "step": 27001 + }, + { + "epoch": 0.99, + "learning_rate": 1.0130003295858205e-07, + "loss": 0.7158, + "step": 27002 + }, + { + "epoch": 0.99, + "learning_rate": 1.0129637089390999e-07, + "loss": 0.6837, + "step": 27003 + }, + { + "epoch": 0.99, + "learning_rate": 1.0129270882923793e-07, + "loss": 0.7015, + "step": 27004 + }, + { + "epoch": 0.99, + "learning_rate": 1.0128904676456587e-07, + "loss": 0.6956, + "step": 27005 + }, + { + "epoch": 0.99, + "learning_rate": 1.0128538469989381e-07, + "loss": 0.6967, + "step": 27006 + }, + { + "epoch": 0.99, + "learning_rate": 1.0128172263522172e-07, + "loss": 0.6901, + "step": 27007 + }, + { + "epoch": 0.99, + "learning_rate": 1.0127806057054966e-07, + "loss": 0.6888, + "step": 27008 + }, + { + "epoch": 0.99, + "learning_rate": 1.012743985058776e-07, + "loss": 0.6832, + "step": 27009 + }, + { + "epoch": 0.99, + "learning_rate": 1.0127073644120554e-07, + "loss": 0.7062, + "step": 27010 + }, + { + "epoch": 0.99, + "learning_rate": 1.0126707437653348e-07, + "loss": 0.6855, + "step": 27011 + }, + { + "epoch": 0.99, + "learning_rate": 1.0126341231186142e-07, + "loss": 0.6906, + "step": 27012 + }, + { + "epoch": 0.99, + "learning_rate": 1.0125975024718936e-07, + "loss": 0.6873, + "step": 27013 + }, + { + "epoch": 0.99, + "learning_rate": 1.012560881825173e-07, + "loss": 0.6979, + "step": 27014 + }, + { + "epoch": 0.99, + "learning_rate": 1.0125242611784524e-07, + "loss": 0.6879, + "step": 27015 + }, + { + "epoch": 0.99, + "learning_rate": 1.0124876405317317e-07, + "loss": 0.7076, + "step": 27016 + }, + { + "epoch": 0.99, + "learning_rate": 1.0124510198850111e-07, + "loss": 0.6782, + "step": 27017 + }, + { + "epoch": 0.99, + "learning_rate": 1.0124143992382905e-07, + "loss": 0.6835, + "step": 27018 + }, + { + "epoch": 0.99, + "learning_rate": 1.0123777785915699e-07, + "loss": 0.6906, + "step": 27019 + }, + { + "epoch": 0.99, + "learning_rate": 1.0123411579448493e-07, + "loss": 0.6789, + "step": 27020 + }, + { + "epoch": 0.99, + "learning_rate": 1.0123045372981287e-07, + "loss": 0.7164, + "step": 27021 + }, + { + "epoch": 0.99, + "learning_rate": 1.0122679166514081e-07, + "loss": 0.6946, + "step": 27022 + }, + { + "epoch": 0.99, + "learning_rate": 1.0122312960046875e-07, + "loss": 0.7031, + "step": 27023 + }, + { + "epoch": 0.99, + "learning_rate": 1.0121946753579669e-07, + "loss": 0.699, + "step": 27024 + }, + { + "epoch": 0.99, + "learning_rate": 1.012158054711246e-07, + "loss": 0.6925, + "step": 27025 + }, + { + "epoch": 0.99, + "learning_rate": 1.0121214340645254e-07, + "loss": 0.6975, + "step": 27026 + }, + { + "epoch": 0.99, + "learning_rate": 1.0120848134178048e-07, + "loss": 0.7069, + "step": 27027 + }, + { + "epoch": 0.99, + "learning_rate": 1.0120481927710842e-07, + "loss": 0.6901, + "step": 27028 + }, + { + "epoch": 0.99, + "learning_rate": 1.0120115721243636e-07, + "loss": 0.6919, + "step": 27029 + }, + { + "epoch": 0.99, + "learning_rate": 1.011974951477643e-07, + "loss": 0.6983, + "step": 27030 + }, + { + "epoch": 0.99, + "learning_rate": 1.0119383308309224e-07, + "loss": 0.6896, + "step": 27031 + }, + { + "epoch": 0.99, + "learning_rate": 1.0119017101842018e-07, + "loss": 0.6994, + "step": 27032 + }, + { + "epoch": 0.99, + "learning_rate": 1.0118650895374812e-07, + "loss": 0.6984, + "step": 27033 + }, + { + "epoch": 0.99, + "learning_rate": 1.0118284688907606e-07, + "loss": 0.6919, + "step": 27034 + }, + { + "epoch": 0.99, + "learning_rate": 1.01179184824404e-07, + "loss": 0.7032, + "step": 27035 + }, + { + "epoch": 0.99, + "learning_rate": 1.0117552275973194e-07, + "loss": 0.6938, + "step": 27036 + }, + { + "epoch": 0.99, + "learning_rate": 1.0117186069505988e-07, + "loss": 0.6902, + "step": 27037 + }, + { + "epoch": 0.99, + "learning_rate": 1.0116819863038781e-07, + "loss": 0.7029, + "step": 27038 + }, + { + "epoch": 0.99, + "learning_rate": 1.0116453656571575e-07, + "loss": 0.6904, + "step": 27039 + }, + { + "epoch": 0.99, + "learning_rate": 1.0116087450104369e-07, + "loss": 0.6915, + "step": 27040 + }, + { + "epoch": 0.99, + "learning_rate": 1.0115721243637163e-07, + "loss": 0.6931, + "step": 27041 + }, + { + "epoch": 0.99, + "learning_rate": 1.0115355037169957e-07, + "loss": 0.6929, + "step": 27042 + }, + { + "epoch": 0.99, + "learning_rate": 1.0114988830702749e-07, + "loss": 0.7036, + "step": 27043 + }, + { + "epoch": 0.99, + "learning_rate": 1.0114622624235542e-07, + "loss": 0.6678, + "step": 27044 + }, + { + "epoch": 0.99, + "learning_rate": 1.0114256417768336e-07, + "loss": 0.6917, + "step": 27045 + }, + { + "epoch": 0.99, + "learning_rate": 1.011389021130113e-07, + "loss": 0.6903, + "step": 27046 + }, + { + "epoch": 0.99, + "learning_rate": 1.0113524004833924e-07, + "loss": 0.6898, + "step": 27047 + }, + { + "epoch": 0.99, + "learning_rate": 1.0113157798366718e-07, + "loss": 0.6875, + "step": 27048 + }, + { + "epoch": 0.99, + "learning_rate": 1.0112791591899512e-07, + "loss": 0.6782, + "step": 27049 + }, + { + "epoch": 0.99, + "learning_rate": 1.0112425385432306e-07, + "loss": 0.6844, + "step": 27050 + }, + { + "epoch": 0.99, + "learning_rate": 1.01120591789651e-07, + "loss": 0.6945, + "step": 27051 + }, + { + "epoch": 0.99, + "learning_rate": 1.0111692972497894e-07, + "loss": 0.6967, + "step": 27052 + }, + { + "epoch": 0.99, + "learning_rate": 1.0111326766030688e-07, + "loss": 0.6953, + "step": 27053 + }, + { + "epoch": 0.99, + "learning_rate": 1.0110960559563482e-07, + "loss": 0.702, + "step": 27054 + }, + { + "epoch": 0.99, + "learning_rate": 1.0110594353096276e-07, + "loss": 0.7006, + "step": 27055 + }, + { + "epoch": 0.99, + "learning_rate": 1.011022814662907e-07, + "loss": 0.6792, + "step": 27056 + }, + { + "epoch": 0.99, + "learning_rate": 1.0109861940161864e-07, + "loss": 0.6972, + "step": 27057 + }, + { + "epoch": 0.99, + "learning_rate": 1.0109495733694658e-07, + "loss": 0.6868, + "step": 27058 + }, + { + "epoch": 0.99, + "learning_rate": 1.0109129527227452e-07, + "loss": 0.7109, + "step": 27059 + }, + { + "epoch": 0.99, + "learning_rate": 1.0108763320760245e-07, + "loss": 0.7138, + "step": 27060 + }, + { + "epoch": 0.99, + "learning_rate": 1.0108397114293037e-07, + "loss": 0.7038, + "step": 27061 + }, + { + "epoch": 0.99, + "learning_rate": 1.0108030907825831e-07, + "loss": 0.6993, + "step": 27062 + }, + { + "epoch": 0.99, + "learning_rate": 1.0107664701358625e-07, + "loss": 0.6791, + "step": 27063 + }, + { + "epoch": 0.99, + "learning_rate": 1.0107298494891419e-07, + "loss": 0.6931, + "step": 27064 + }, + { + "epoch": 0.99, + "learning_rate": 1.0106932288424213e-07, + "loss": 0.7011, + "step": 27065 + }, + { + "epoch": 0.99, + "learning_rate": 1.0106566081957006e-07, + "loss": 0.6937, + "step": 27066 + }, + { + "epoch": 0.99, + "learning_rate": 1.01061998754898e-07, + "loss": 0.6866, + "step": 27067 + }, + { + "epoch": 0.99, + "learning_rate": 1.0105833669022594e-07, + "loss": 0.6917, + "step": 27068 + }, + { + "epoch": 0.99, + "learning_rate": 1.0105467462555388e-07, + "loss": 0.7033, + "step": 27069 + }, + { + "epoch": 0.99, + "learning_rate": 1.0105101256088182e-07, + "loss": 0.6987, + "step": 27070 + }, + { + "epoch": 0.99, + "learning_rate": 1.0104735049620976e-07, + "loss": 0.6963, + "step": 27071 + }, + { + "epoch": 0.99, + "learning_rate": 1.010436884315377e-07, + "loss": 0.6939, + "step": 27072 + }, + { + "epoch": 0.99, + "learning_rate": 1.0104002636686564e-07, + "loss": 0.6901, + "step": 27073 + }, + { + "epoch": 0.99, + "learning_rate": 1.0103636430219358e-07, + "loss": 0.6915, + "step": 27074 + }, + { + "epoch": 0.99, + "learning_rate": 1.0103270223752152e-07, + "loss": 0.7034, + "step": 27075 + }, + { + "epoch": 0.99, + "learning_rate": 1.0102904017284946e-07, + "loss": 0.7147, + "step": 27076 + }, + { + "epoch": 0.99, + "learning_rate": 1.010253781081774e-07, + "loss": 0.6955, + "step": 27077 + }, + { + "epoch": 0.99, + "learning_rate": 1.0102171604350534e-07, + "loss": 0.6981, + "step": 27078 + }, + { + "epoch": 0.99, + "learning_rate": 1.0101805397883325e-07, + "loss": 0.7036, + "step": 27079 + }, + { + "epoch": 0.99, + "learning_rate": 1.0101439191416119e-07, + "loss": 0.6768, + "step": 27080 + }, + { + "epoch": 0.99, + "learning_rate": 1.0101072984948913e-07, + "loss": 0.6754, + "step": 27081 + }, + { + "epoch": 0.99, + "learning_rate": 1.0100706778481707e-07, + "loss": 0.6954, + "step": 27082 + }, + { + "epoch": 0.99, + "learning_rate": 1.0100340572014501e-07, + "loss": 0.6993, + "step": 27083 + }, + { + "epoch": 0.99, + "learning_rate": 1.0099974365547295e-07, + "loss": 0.6913, + "step": 27084 + }, + { + "epoch": 0.99, + "learning_rate": 1.0099608159080089e-07, + "loss": 0.6948, + "step": 27085 + }, + { + "epoch": 0.99, + "learning_rate": 1.0099241952612883e-07, + "loss": 0.6899, + "step": 27086 + }, + { + "epoch": 0.99, + "learning_rate": 1.0098875746145676e-07, + "loss": 0.6858, + "step": 27087 + }, + { + "epoch": 0.99, + "learning_rate": 1.009850953967847e-07, + "loss": 0.691, + "step": 27088 + }, + { + "epoch": 0.99, + "learning_rate": 1.0098143333211264e-07, + "loss": 0.6869, + "step": 27089 + }, + { + "epoch": 0.99, + "learning_rate": 1.0097777126744058e-07, + "loss": 0.6857, + "step": 27090 + }, + { + "epoch": 0.99, + "learning_rate": 1.0097410920276852e-07, + "loss": 0.6817, + "step": 27091 + }, + { + "epoch": 0.99, + "learning_rate": 1.0097044713809646e-07, + "loss": 0.7053, + "step": 27092 + }, + { + "epoch": 0.99, + "learning_rate": 1.009667850734244e-07, + "loss": 0.707, + "step": 27093 + }, + { + "epoch": 0.99, + "learning_rate": 1.0096312300875234e-07, + "loss": 0.6915, + "step": 27094 + }, + { + "epoch": 0.99, + "learning_rate": 1.0095946094408028e-07, + "loss": 0.6844, + "step": 27095 + }, + { + "epoch": 0.99, + "learning_rate": 1.0095579887940822e-07, + "loss": 0.697, + "step": 27096 + }, + { + "epoch": 0.99, + "learning_rate": 1.0095213681473613e-07, + "loss": 0.7016, + "step": 27097 + }, + { + "epoch": 0.99, + "learning_rate": 1.0094847475006407e-07, + "loss": 0.7017, + "step": 27098 + }, + { + "epoch": 0.99, + "learning_rate": 1.0094481268539201e-07, + "loss": 0.6705, + "step": 27099 + }, + { + "epoch": 0.99, + "learning_rate": 1.0094115062071995e-07, + "loss": 0.6989, + "step": 27100 + }, + { + "epoch": 0.99, + "learning_rate": 1.0093748855604789e-07, + "loss": 0.6868, + "step": 27101 + }, + { + "epoch": 0.99, + "learning_rate": 1.0093382649137583e-07, + "loss": 0.6908, + "step": 27102 + }, + { + "epoch": 0.99, + "learning_rate": 1.0093016442670377e-07, + "loss": 0.7011, + "step": 27103 + }, + { + "epoch": 0.99, + "learning_rate": 1.0092650236203171e-07, + "loss": 0.6975, + "step": 27104 + }, + { + "epoch": 0.99, + "learning_rate": 1.0092284029735965e-07, + "loss": 0.6902, + "step": 27105 + }, + { + "epoch": 0.99, + "learning_rate": 1.0091917823268759e-07, + "loss": 0.7081, + "step": 27106 + }, + { + "epoch": 0.99, + "learning_rate": 1.0091551616801553e-07, + "loss": 0.688, + "step": 27107 + }, + { + "epoch": 0.99, + "learning_rate": 1.0091185410334347e-07, + "loss": 0.6929, + "step": 27108 + }, + { + "epoch": 0.99, + "learning_rate": 1.009081920386714e-07, + "loss": 0.6875, + "step": 27109 + }, + { + "epoch": 0.99, + "learning_rate": 1.0090452997399934e-07, + "loss": 0.6926, + "step": 27110 + }, + { + "epoch": 0.99, + "learning_rate": 1.0090086790932728e-07, + "loss": 0.699, + "step": 27111 + }, + { + "epoch": 0.99, + "learning_rate": 1.0089720584465522e-07, + "loss": 0.7051, + "step": 27112 + }, + { + "epoch": 0.99, + "learning_rate": 1.0089354377998316e-07, + "loss": 0.6989, + "step": 27113 + }, + { + "epoch": 0.99, + "learning_rate": 1.008898817153111e-07, + "loss": 0.6968, + "step": 27114 + }, + { + "epoch": 0.99, + "learning_rate": 1.0088621965063901e-07, + "loss": 0.6882, + "step": 27115 + }, + { + "epoch": 0.99, + "learning_rate": 1.0088255758596695e-07, + "loss": 0.6743, + "step": 27116 + }, + { + "epoch": 0.99, + "learning_rate": 1.0087889552129489e-07, + "loss": 0.7059, + "step": 27117 + }, + { + "epoch": 0.99, + "learning_rate": 1.0087523345662283e-07, + "loss": 0.7093, + "step": 27118 + }, + { + "epoch": 0.99, + "learning_rate": 1.0087157139195077e-07, + "loss": 0.6809, + "step": 27119 + }, + { + "epoch": 0.99, + "learning_rate": 1.0086790932727871e-07, + "loss": 0.6921, + "step": 27120 + }, + { + "epoch": 0.99, + "learning_rate": 1.0086424726260665e-07, + "loss": 0.6977, + "step": 27121 + }, + { + "epoch": 0.99, + "learning_rate": 1.0086058519793459e-07, + "loss": 0.686, + "step": 27122 + }, + { + "epoch": 0.99, + "learning_rate": 1.0085692313326253e-07, + "loss": 0.6928, + "step": 27123 + }, + { + "epoch": 0.99, + "learning_rate": 1.0085326106859047e-07, + "loss": 0.6951, + "step": 27124 + }, + { + "epoch": 0.99, + "learning_rate": 1.0084959900391841e-07, + "loss": 0.6871, + "step": 27125 + }, + { + "epoch": 0.99, + "learning_rate": 1.0084593693924635e-07, + "loss": 0.6809, + "step": 27126 + }, + { + "epoch": 0.99, + "learning_rate": 1.0084227487457429e-07, + "loss": 0.6978, + "step": 27127 + }, + { + "epoch": 0.99, + "learning_rate": 1.0083861280990223e-07, + "loss": 0.691, + "step": 27128 + }, + { + "epoch": 0.99, + "learning_rate": 1.0083495074523017e-07, + "loss": 0.6796, + "step": 27129 + }, + { + "epoch": 0.99, + "learning_rate": 1.0083128868055811e-07, + "loss": 0.6935, + "step": 27130 + }, + { + "epoch": 0.99, + "learning_rate": 1.0082762661588603e-07, + "loss": 0.7003, + "step": 27131 + }, + { + "epoch": 0.99, + "learning_rate": 1.0082396455121397e-07, + "loss": 0.6951, + "step": 27132 + }, + { + "epoch": 0.99, + "learning_rate": 1.008203024865419e-07, + "loss": 0.7006, + "step": 27133 + }, + { + "epoch": 0.99, + "learning_rate": 1.0081664042186984e-07, + "loss": 0.6823, + "step": 27134 + }, + { + "epoch": 0.99, + "learning_rate": 1.0081297835719778e-07, + "loss": 0.7021, + "step": 27135 + }, + { + "epoch": 0.99, + "learning_rate": 1.0080931629252571e-07, + "loss": 0.6959, + "step": 27136 + }, + { + "epoch": 0.99, + "learning_rate": 1.0080565422785365e-07, + "loss": 0.7071, + "step": 27137 + }, + { + "epoch": 0.99, + "learning_rate": 1.0080199216318159e-07, + "loss": 0.6916, + "step": 27138 + }, + { + "epoch": 0.99, + "learning_rate": 1.0079833009850953e-07, + "loss": 0.7019, + "step": 27139 + }, + { + "epoch": 0.99, + "learning_rate": 1.0079466803383747e-07, + "loss": 0.6948, + "step": 27140 + }, + { + "epoch": 0.99, + "learning_rate": 1.0079100596916541e-07, + "loss": 0.6861, + "step": 27141 + }, + { + "epoch": 0.99, + "learning_rate": 1.0078734390449335e-07, + "loss": 0.7073, + "step": 27142 + }, + { + "epoch": 0.99, + "learning_rate": 1.0078368183982129e-07, + "loss": 0.6997, + "step": 27143 + }, + { + "epoch": 0.99, + "learning_rate": 1.0078001977514923e-07, + "loss": 0.6937, + "step": 27144 + }, + { + "epoch": 0.99, + "learning_rate": 1.0077635771047717e-07, + "loss": 0.6877, + "step": 27145 + }, + { + "epoch": 0.99, + "learning_rate": 1.0077269564580511e-07, + "loss": 0.6873, + "step": 27146 + }, + { + "epoch": 0.99, + "learning_rate": 1.0076903358113305e-07, + "loss": 0.6815, + "step": 27147 + }, + { + "epoch": 0.99, + "learning_rate": 1.0076537151646099e-07, + "loss": 0.7015, + "step": 27148 + }, + { + "epoch": 0.99, + "learning_rate": 1.0076170945178893e-07, + "loss": 0.6904, + "step": 27149 + }, + { + "epoch": 0.99, + "learning_rate": 1.0075804738711687e-07, + "loss": 0.6917, + "step": 27150 + }, + { + "epoch": 0.99, + "learning_rate": 1.0075438532244478e-07, + "loss": 0.6722, + "step": 27151 + }, + { + "epoch": 0.99, + "learning_rate": 1.0075072325777272e-07, + "loss": 0.6722, + "step": 27152 + }, + { + "epoch": 0.99, + "learning_rate": 1.0074706119310066e-07, + "loss": 0.6961, + "step": 27153 + }, + { + "epoch": 0.99, + "learning_rate": 1.007433991284286e-07, + "loss": 0.7039, + "step": 27154 + }, + { + "epoch": 0.99, + "learning_rate": 1.0073973706375654e-07, + "loss": 0.6987, + "step": 27155 + }, + { + "epoch": 0.99, + "learning_rate": 1.0073607499908448e-07, + "loss": 0.6889, + "step": 27156 + }, + { + "epoch": 0.99, + "learning_rate": 1.0073241293441242e-07, + "loss": 0.6901, + "step": 27157 + }, + { + "epoch": 0.99, + "learning_rate": 1.0072875086974035e-07, + "loss": 0.6865, + "step": 27158 + }, + { + "epoch": 0.99, + "learning_rate": 1.0072508880506829e-07, + "loss": 0.7011, + "step": 27159 + }, + { + "epoch": 0.99, + "learning_rate": 1.0072142674039623e-07, + "loss": 0.7093, + "step": 27160 + }, + { + "epoch": 0.99, + "learning_rate": 1.0071776467572417e-07, + "loss": 0.6903, + "step": 27161 + }, + { + "epoch": 0.99, + "learning_rate": 1.007141026110521e-07, + "loss": 0.6875, + "step": 27162 + }, + { + "epoch": 0.99, + "learning_rate": 1.0071044054638005e-07, + "loss": 0.693, + "step": 27163 + }, + { + "epoch": 0.99, + "learning_rate": 1.0070677848170799e-07, + "loss": 0.7139, + "step": 27164 + }, + { + "epoch": 0.99, + "learning_rate": 1.0070311641703593e-07, + "loss": 0.6899, + "step": 27165 + }, + { + "epoch": 0.99, + "learning_rate": 1.0069945435236387e-07, + "loss": 0.6901, + "step": 27166 + }, + { + "epoch": 0.99, + "learning_rate": 1.006957922876918e-07, + "loss": 0.6972, + "step": 27167 + }, + { + "epoch": 0.99, + "learning_rate": 1.0069213022301975e-07, + "loss": 0.6892, + "step": 27168 + }, + { + "epoch": 0.99, + "learning_rate": 1.0068846815834766e-07, + "loss": 0.6794, + "step": 27169 + }, + { + "epoch": 0.99, + "learning_rate": 1.006848060936756e-07, + "loss": 0.6977, + "step": 27170 + }, + { + "epoch": 0.99, + "learning_rate": 1.0068114402900354e-07, + "loss": 0.6924, + "step": 27171 + }, + { + "epoch": 0.99, + "learning_rate": 1.0067748196433148e-07, + "loss": 0.6943, + "step": 27172 + }, + { + "epoch": 0.99, + "learning_rate": 1.0067381989965942e-07, + "loss": 0.6939, + "step": 27173 + }, + { + "epoch": 0.99, + "learning_rate": 1.0067015783498736e-07, + "loss": 0.6955, + "step": 27174 + }, + { + "epoch": 0.99, + "learning_rate": 1.006664957703153e-07, + "loss": 0.6959, + "step": 27175 + }, + { + "epoch": 0.99, + "learning_rate": 1.0066283370564324e-07, + "loss": 0.6856, + "step": 27176 + }, + { + "epoch": 0.99, + "learning_rate": 1.0065917164097118e-07, + "loss": 0.711, + "step": 27177 + }, + { + "epoch": 0.99, + "learning_rate": 1.0065550957629912e-07, + "loss": 0.697, + "step": 27178 + }, + { + "epoch": 0.99, + "learning_rate": 1.0065184751162706e-07, + "loss": 0.6931, + "step": 27179 + }, + { + "epoch": 0.99, + "learning_rate": 1.0064818544695499e-07, + "loss": 0.6746, + "step": 27180 + }, + { + "epoch": 0.99, + "learning_rate": 1.0064452338228293e-07, + "loss": 0.7059, + "step": 27181 + }, + { + "epoch": 0.99, + "learning_rate": 1.0064086131761087e-07, + "loss": 0.6868, + "step": 27182 + }, + { + "epoch": 0.99, + "learning_rate": 1.006371992529388e-07, + "loss": 0.7023, + "step": 27183 + }, + { + "epoch": 0.99, + "learning_rate": 1.0063353718826675e-07, + "loss": 0.6902, + "step": 27184 + }, + { + "epoch": 0.99, + "learning_rate": 1.0062987512359469e-07, + "loss": 0.681, + "step": 27185 + }, + { + "epoch": 0.99, + "learning_rate": 1.0062621305892263e-07, + "loss": 0.6868, + "step": 27186 + }, + { + "epoch": 0.99, + "learning_rate": 1.0062255099425054e-07, + "loss": 0.6972, + "step": 27187 + }, + { + "epoch": 0.99, + "learning_rate": 1.0061888892957848e-07, + "loss": 0.7012, + "step": 27188 + }, + { + "epoch": 0.99, + "learning_rate": 1.0061522686490642e-07, + "loss": 0.6897, + "step": 27189 + }, + { + "epoch": 0.99, + "learning_rate": 1.0061156480023436e-07, + "loss": 0.7003, + "step": 27190 + }, + { + "epoch": 0.99, + "learning_rate": 1.006079027355623e-07, + "loss": 0.687, + "step": 27191 + }, + { + "epoch": 0.99, + "learning_rate": 1.0060424067089024e-07, + "loss": 0.6878, + "step": 27192 + }, + { + "epoch": 0.99, + "learning_rate": 1.0060057860621818e-07, + "loss": 0.696, + "step": 27193 + }, + { + "epoch": 0.99, + "learning_rate": 1.0059691654154612e-07, + "loss": 0.6872, + "step": 27194 + }, + { + "epoch": 0.99, + "learning_rate": 1.0059325447687406e-07, + "loss": 0.6964, + "step": 27195 + }, + { + "epoch": 1.0, + "learning_rate": 1.00589592412202e-07, + "loss": 0.6836, + "step": 27196 + }, + { + "epoch": 1.0, + "learning_rate": 1.0058593034752994e-07, + "loss": 0.6798, + "step": 27197 + }, + { + "epoch": 1.0, + "learning_rate": 1.0058226828285788e-07, + "loss": 0.6875, + "step": 27198 + }, + { + "epoch": 1.0, + "learning_rate": 1.0057860621818582e-07, + "loss": 0.6656, + "step": 27199 + }, + { + "epoch": 1.0, + "learning_rate": 1.0057494415351376e-07, + "loss": 0.6901, + "step": 27200 + }, + { + "epoch": 1.0, + "learning_rate": 1.005712820888417e-07, + "loss": 0.7047, + "step": 27201 + }, + { + "epoch": 1.0, + "learning_rate": 1.0056762002416962e-07, + "loss": 0.6765, + "step": 27202 + }, + { + "epoch": 1.0, + "learning_rate": 1.0056395795949756e-07, + "loss": 0.7024, + "step": 27203 + }, + { + "epoch": 1.0, + "learning_rate": 1.005602958948255e-07, + "loss": 0.6963, + "step": 27204 + }, + { + "epoch": 1.0, + "learning_rate": 1.0055663383015343e-07, + "loss": 0.6903, + "step": 27205 + }, + { + "epoch": 1.0, + "learning_rate": 1.0055297176548137e-07, + "loss": 0.6756, + "step": 27206 + }, + { + "epoch": 1.0, + "learning_rate": 1.005493097008093e-07, + "loss": 0.6916, + "step": 27207 + }, + { + "epoch": 1.0, + "learning_rate": 1.0054564763613724e-07, + "loss": 0.6835, + "step": 27208 + }, + { + "epoch": 1.0, + "learning_rate": 1.0054198557146518e-07, + "loss": 0.6845, + "step": 27209 + }, + { + "epoch": 1.0, + "learning_rate": 1.0053832350679312e-07, + "loss": 0.6983, + "step": 27210 + }, + { + "epoch": 1.0, + "learning_rate": 1.0053466144212106e-07, + "loss": 0.6915, + "step": 27211 + }, + { + "epoch": 1.0, + "learning_rate": 1.00530999377449e-07, + "loss": 0.6914, + "step": 27212 + }, + { + "epoch": 1.0, + "learning_rate": 1.0052733731277694e-07, + "loss": 0.6898, + "step": 27213 + }, + { + "epoch": 1.0, + "learning_rate": 1.0052367524810488e-07, + "loss": 0.6915, + "step": 27214 + }, + { + "epoch": 1.0, + "learning_rate": 1.0052001318343282e-07, + "loss": 0.6847, + "step": 27215 + }, + { + "epoch": 1.0, + "learning_rate": 1.0051635111876076e-07, + "loss": 0.7007, + "step": 27216 + }, + { + "epoch": 1.0, + "learning_rate": 1.005126890540887e-07, + "loss": 0.6884, + "step": 27217 + }, + { + "epoch": 1.0, + "learning_rate": 1.0050902698941664e-07, + "loss": 0.6901, + "step": 27218 + }, + { + "epoch": 1.0, + "learning_rate": 1.0050536492474458e-07, + "loss": 0.6944, + "step": 27219 + }, + { + "epoch": 1.0, + "learning_rate": 1.0050170286007252e-07, + "loss": 0.6991, + "step": 27220 + }, + { + "epoch": 1.0, + "learning_rate": 1.0049804079540046e-07, + "loss": 0.6958, + "step": 27221 + }, + { + "epoch": 1.0, + "learning_rate": 1.004943787307284e-07, + "loss": 0.6906, + "step": 27222 + }, + { + "epoch": 1.0, + "learning_rate": 1.0049071666605631e-07, + "loss": 0.6973, + "step": 27223 + }, + { + "epoch": 1.0, + "learning_rate": 1.0048705460138425e-07, + "loss": 0.6855, + "step": 27224 + }, + { + "epoch": 1.0, + "learning_rate": 1.0048339253671219e-07, + "loss": 0.6951, + "step": 27225 + }, + { + "epoch": 1.0, + "learning_rate": 1.0047973047204013e-07, + "loss": 0.6863, + "step": 27226 + }, + { + "epoch": 1.0, + "learning_rate": 1.0047606840736807e-07, + "loss": 0.6992, + "step": 27227 + }, + { + "epoch": 1.0, + "learning_rate": 1.0047240634269601e-07, + "loss": 0.7029, + "step": 27228 + }, + { + "epoch": 1.0, + "learning_rate": 1.0046874427802394e-07, + "loss": 0.7012, + "step": 27229 + }, + { + "epoch": 1.0, + "learning_rate": 1.0046508221335188e-07, + "loss": 0.7061, + "step": 27230 + }, + { + "epoch": 1.0, + "learning_rate": 1.0046142014867982e-07, + "loss": 0.6861, + "step": 27231 + }, + { + "epoch": 1.0, + "learning_rate": 1.0045775808400776e-07, + "loss": 0.687, + "step": 27232 + }, + { + "epoch": 1.0, + "learning_rate": 1.004540960193357e-07, + "loss": 0.6951, + "step": 27233 + }, + { + "epoch": 1.0, + "learning_rate": 1.0045043395466364e-07, + "loss": 0.6882, + "step": 27234 + }, + { + "epoch": 1.0, + "learning_rate": 1.0044677188999158e-07, + "loss": 0.6877, + "step": 27235 + }, + { + "epoch": 1.0, + "learning_rate": 1.0044310982531952e-07, + "loss": 0.7022, + "step": 27236 + }, + { + "epoch": 1.0, + "learning_rate": 1.0043944776064746e-07, + "loss": 0.6835, + "step": 27237 + }, + { + "epoch": 1.0, + "learning_rate": 1.004357856959754e-07, + "loss": 0.6952, + "step": 27238 + }, + { + "epoch": 1.0, + "learning_rate": 1.0043212363130334e-07, + "loss": 0.6921, + "step": 27239 + }, + { + "epoch": 1.0, + "learning_rate": 1.0042846156663128e-07, + "loss": 0.6908, + "step": 27240 + }, + { + "epoch": 1.0, + "learning_rate": 1.0042479950195919e-07, + "loss": 0.7075, + "step": 27241 + }, + { + "epoch": 1.0, + "learning_rate": 1.0042113743728713e-07, + "loss": 0.6838, + "step": 27242 + }, + { + "epoch": 1.0, + "learning_rate": 1.0041747537261507e-07, + "loss": 0.6841, + "step": 27243 + }, + { + "epoch": 1.0, + "learning_rate": 1.0041381330794301e-07, + "loss": 0.7026, + "step": 27244 + }, + { + "epoch": 1.0, + "learning_rate": 1.0041015124327095e-07, + "loss": 0.6907, + "step": 27245 + }, + { + "epoch": 1.0, + "learning_rate": 1.0040648917859889e-07, + "loss": 0.6924, + "step": 27246 + }, + { + "epoch": 1.0, + "learning_rate": 1.0040282711392683e-07, + "loss": 0.6925, + "step": 27247 + }, + { + "epoch": 1.0, + "learning_rate": 1.0039916504925477e-07, + "loss": 0.6975, + "step": 27248 + }, + { + "epoch": 1.0, + "learning_rate": 1.0039550298458271e-07, + "loss": 0.6899, + "step": 27249 + }, + { + "epoch": 1.0, + "learning_rate": 1.0039184091991065e-07, + "loss": 0.6956, + "step": 27250 + }, + { + "epoch": 1.0, + "learning_rate": 1.0038817885523857e-07, + "loss": 0.6931, + "step": 27251 + }, + { + "epoch": 1.0, + "learning_rate": 1.0038451679056651e-07, + "loss": 0.6893, + "step": 27252 + }, + { + "epoch": 1.0, + "learning_rate": 1.0038085472589445e-07, + "loss": 0.6949, + "step": 27253 + }, + { + "epoch": 1.0, + "learning_rate": 1.003771926612224e-07, + "loss": 0.6871, + "step": 27254 + }, + { + "epoch": 1.0, + "learning_rate": 1.0037353059655033e-07, + "loss": 0.6767, + "step": 27255 + }, + { + "epoch": 1.0, + "learning_rate": 1.0036986853187827e-07, + "loss": 0.6892, + "step": 27256 + }, + { + "epoch": 1.0, + "learning_rate": 1.0036620646720621e-07, + "loss": 0.6926, + "step": 27257 + }, + { + "epoch": 1.0, + "learning_rate": 1.0036254440253415e-07, + "loss": 0.697, + "step": 27258 + }, + { + "epoch": 1.0, + "learning_rate": 1.0035888233786207e-07, + "loss": 0.6882, + "step": 27259 + }, + { + "epoch": 1.0, + "learning_rate": 1.0035522027319001e-07, + "loss": 0.6984, + "step": 27260 + }, + { + "epoch": 1.0, + "learning_rate": 1.0035155820851795e-07, + "loss": 0.6889, + "step": 27261 + }, + { + "epoch": 1.0, + "learning_rate": 1.0034789614384589e-07, + "loss": 0.6955, + "step": 27262 + }, + { + "epoch": 1.0, + "learning_rate": 1.0034423407917383e-07, + "loss": 0.6813, + "step": 27263 + }, + { + "epoch": 1.0, + "learning_rate": 1.0034057201450177e-07, + "loss": 0.6908, + "step": 27264 + }, + { + "epoch": 1.0, + "learning_rate": 1.0033690994982971e-07, + "loss": 0.7044, + "step": 27265 + }, + { + "epoch": 1.0, + "learning_rate": 1.0033324788515765e-07, + "loss": 0.6971, + "step": 27266 + }, + { + "epoch": 1.0, + "learning_rate": 1.0032958582048559e-07, + "loss": 0.683, + "step": 27267 + }, + { + "epoch": 1.0, + "learning_rate": 1.0032592375581353e-07, + "loss": 0.7013, + "step": 27268 + }, + { + "epoch": 1.0, + "learning_rate": 1.0032226169114147e-07, + "loss": 0.6958, + "step": 27269 + }, + { + "epoch": 1.0, + "learning_rate": 1.0031859962646941e-07, + "loss": 0.682, + "step": 27270 + }, + { + "epoch": 1.0, + "learning_rate": 1.0031493756179735e-07, + "loss": 0.6837, + "step": 27271 + }, + { + "epoch": 1.0, + "learning_rate": 1.0031127549712529e-07, + "loss": 0.6867, + "step": 27272 + }, + { + "epoch": 1.0, + "learning_rate": 1.0030761343245321e-07, + "loss": 0.6898, + "step": 27273 + }, + { + "epoch": 1.0, + "learning_rate": 1.0030395136778115e-07, + "loss": 0.6943, + "step": 27274 + }, + { + "epoch": 1.0, + "learning_rate": 1.0030028930310909e-07, + "loss": 0.7012, + "step": 27275 + }, + { + "epoch": 1.0, + "learning_rate": 1.0029662723843703e-07, + "loss": 0.697, + "step": 27276 + }, + { + "epoch": 1.0, + "learning_rate": 1.0029296517376496e-07, + "loss": 0.68, + "step": 27277 + }, + { + "epoch": 1.0, + "learning_rate": 1.0028930310909289e-07, + "loss": 0.6892, + "step": 27278 + }, + { + "epoch": 1.0, + "learning_rate": 1.0028564104442083e-07, + "loss": 0.6869, + "step": 27279 + }, + { + "epoch": 1.0, + "learning_rate": 1.0028197897974877e-07, + "loss": 0.6902, + "step": 27280 + }, + { + "epoch": 1.0, + "learning_rate": 1.002783169150767e-07, + "loss": 0.6875, + "step": 27281 + }, + { + "epoch": 1.0, + "learning_rate": 1.0027465485040465e-07, + "loss": 0.6962, + "step": 27282 + }, + { + "epoch": 1.0, + "learning_rate": 1.0027099278573259e-07, + "loss": 0.6936, + "step": 27283 + }, + { + "epoch": 1.0, + "learning_rate": 1.0026733072106053e-07, + "loss": 0.702, + "step": 27284 + }, + { + "epoch": 1.0, + "learning_rate": 1.0026366865638847e-07, + "loss": 0.6934, + "step": 27285 + }, + { + "epoch": 1.0, + "learning_rate": 1.002600065917164e-07, + "loss": 0.6883, + "step": 27286 + }, + { + "epoch": 1.0, + "learning_rate": 1.0025634452704435e-07, + "loss": 0.6907, + "step": 27287 + }, + { + "epoch": 1.0, + "learning_rate": 1.0025268246237229e-07, + "loss": 0.6984, + "step": 27288 + }, + { + "epoch": 1.0, + "learning_rate": 1.0024902039770023e-07, + "loss": 0.6924, + "step": 27289 + }, + { + "epoch": 1.0, + "learning_rate": 1.0024535833302817e-07, + "loss": 0.6909, + "step": 27290 + }, + { + "epoch": 1.0, + "learning_rate": 1.002416962683561e-07, + "loss": 0.6971, + "step": 27291 + }, + { + "epoch": 1.0, + "learning_rate": 1.0023803420368405e-07, + "loss": 0.702, + "step": 27292 + }, + { + "epoch": 1.0, + "learning_rate": 1.0023437213901199e-07, + "loss": 0.6749, + "step": 27293 + }, + { + "epoch": 1.0, + "learning_rate": 1.0023071007433993e-07, + "loss": 0.6888, + "step": 27294 + }, + { + "epoch": 1.0, + "learning_rate": 1.0022704800966784e-07, + "loss": 0.6961, + "step": 27295 + }, + { + "epoch": 1.0, + "learning_rate": 1.0022338594499578e-07, + "loss": 0.6986, + "step": 27296 + }, + { + "epoch": 1.0, + "learning_rate": 1.0021972388032372e-07, + "loss": 0.695, + "step": 27297 + }, + { + "epoch": 1.0, + "learning_rate": 1.0021606181565166e-07, + "loss": 0.6955, + "step": 27298 + }, + { + "epoch": 1.0, + "learning_rate": 1.002123997509796e-07, + "loss": 0.6947, + "step": 27299 + }, + { + "epoch": 1.0, + "learning_rate": 1.0020873768630753e-07, + "loss": 0.6988, + "step": 27300 + }, + { + "epoch": 1.0, + "learning_rate": 1.0020507562163547e-07, + "loss": 0.6875, + "step": 27301 + }, + { + "epoch": 1.0, + "learning_rate": 1.002014135569634e-07, + "loss": 0.7046, + "step": 27302 + }, + { + "epoch": 1.0, + "learning_rate": 1.0019775149229135e-07, + "loss": 0.692, + "step": 27303 + }, + { + "epoch": 1.0, + "learning_rate": 1.0019408942761929e-07, + "loss": 0.6917, + "step": 27304 + }, + { + "epoch": 1.0, + "learning_rate": 1.0019042736294723e-07, + "loss": 0.6934, + "step": 27305 + }, + { + "epoch": 1.0, + "learning_rate": 1.0018676529827516e-07, + "loss": 0.6951, + "step": 27306 + }, + { + "epoch": 1.0, + "learning_rate": 1.001831032336031e-07, + "loss": 0.6996, + "step": 27307 + }, + { + "epoch": 1.0, + "learning_rate": 1.0017944116893104e-07, + "loss": 0.6969, + "step": 27308 + }, + { + "epoch": 1.0, + "learning_rate": 1.0017577910425898e-07, + "loss": 0.7022, + "step": 27309 + }, + { + "epoch": 1.0, + "learning_rate": 1.0017211703958692e-07, + "loss": 0.6902, + "step": 27310 + }, + { + "epoch": 1.0, + "learning_rate": 1.0016845497491486e-07, + "loss": 0.6813, + "step": 27311 + }, + { + "epoch": 1.0, + "learning_rate": 1.001647929102428e-07, + "loss": 0.6952, + "step": 27312 + }, + { + "epoch": 1.0, + "learning_rate": 1.0016113084557072e-07, + "loss": 0.7022, + "step": 27313 + }, + { + "epoch": 1.0, + "learning_rate": 1.0015746878089866e-07, + "loss": 0.6985, + "step": 27314 + }, + { + "epoch": 1.0, + "learning_rate": 1.001538067162266e-07, + "loss": 0.6976, + "step": 27315 + }, + { + "epoch": 1.0, + "learning_rate": 1.0015014465155454e-07, + "loss": 0.7046, + "step": 27316 + }, + { + "epoch": 1.0, + "learning_rate": 1.0014648258688248e-07, + "loss": 0.6962, + "step": 27317 + }, + { + "epoch": 1.0, + "learning_rate": 1.0014282052221042e-07, + "loss": 0.6789, + "step": 27318 + }, + { + "epoch": 1.0, + "learning_rate": 1.0013915845753836e-07, + "loss": 0.7025, + "step": 27319 + }, + { + "epoch": 1.0, + "learning_rate": 1.001354963928663e-07, + "loss": 0.6902, + "step": 27320 + }, + { + "epoch": 1.0, + "learning_rate": 1.0013183432819424e-07, + "loss": 0.6742, + "step": 27321 + }, + { + "epoch": 1.0, + "learning_rate": 1.0012817226352216e-07, + "loss": 0.6976, + "step": 27322 + }, + { + "epoch": 1.0, + "learning_rate": 1.001245101988501e-07, + "loss": 0.6925, + "step": 27323 + }, + { + "epoch": 1.0, + "learning_rate": 1.0012084813417804e-07, + "loss": 0.6777, + "step": 27324 + }, + { + "epoch": 1.0, + "learning_rate": 1.0011718606950598e-07, + "loss": 0.6789, + "step": 27325 + }, + { + "epoch": 1.0, + "learning_rate": 1.0011352400483392e-07, + "loss": 0.6916, + "step": 27326 + }, + { + "epoch": 1.0, + "learning_rate": 1.0010986194016186e-07, + "loss": 0.7016, + "step": 27327 + }, + { + "epoch": 1.0, + "learning_rate": 1.001061998754898e-07, + "loss": 0.6952, + "step": 27328 + }, + { + "epoch": 1.0, + "learning_rate": 1.0010253781081774e-07, + "loss": 0.7003, + "step": 27329 + }, + { + "epoch": 1.0, + "learning_rate": 1.0009887574614568e-07, + "loss": 0.6834, + "step": 27330 + }, + { + "epoch": 1.0, + "learning_rate": 1.000952136814736e-07, + "loss": 0.7048, + "step": 27331 + }, + { + "epoch": 1.0, + "learning_rate": 1.0009155161680154e-07, + "loss": 0.6679, + "step": 27332 + }, + { + "epoch": 1.0, + "eval_accuracy": 0.6418143899895725, + "eval_loss": 0.6932150721549988, + "eval_runtime": 6.4321, + "eval_samples_per_second": 3578.316, + "eval_steps_per_second": 27.985, + "step": 27332 + }, + { + "epoch": 1.0, + "learning_rate": 1.0008788955212948e-07, + "loss": 0.6725, + "step": 27333 + }, + { + "epoch": 1.0, + "learning_rate": 1.0008422748745742e-07, + "loss": 0.6888, + "step": 27334 + }, + { + "epoch": 1.0, + "learning_rate": 1.0008056542278536e-07, + "loss": 0.6907, + "step": 27335 + }, + { + "epoch": 1.0, + "learning_rate": 1.000769033581133e-07, + "loss": 0.6908, + "step": 27336 + }, + { + "epoch": 1.0, + "learning_rate": 1.0007324129344124e-07, + "loss": 0.7007, + "step": 27337 + }, + { + "epoch": 1.0, + "learning_rate": 1.0006957922876918e-07, + "loss": 0.6908, + "step": 27338 + }, + { + "epoch": 1.0, + "learning_rate": 1.0006591716409712e-07, + "loss": 0.7037, + "step": 27339 + }, + { + "epoch": 1.0, + "learning_rate": 1.0006225509942506e-07, + "loss": 0.6853, + "step": 27340 + }, + { + "epoch": 1.0, + "learning_rate": 1.00058593034753e-07, + "loss": 0.6951, + "step": 27341 + }, + { + "epoch": 1.0, + "learning_rate": 1.0005493097008094e-07, + "loss": 0.7088, + "step": 27342 + }, + { + "epoch": 1.0, + "learning_rate": 1.0005126890540888e-07, + "loss": 0.6978, + "step": 27343 + }, + { + "epoch": 1.0, + "learning_rate": 1.000476068407368e-07, + "loss": 0.7053, + "step": 27344 + }, + { + "epoch": 1.0, + "learning_rate": 1.0004394477606474e-07, + "loss": 0.6923, + "step": 27345 + }, + { + "epoch": 1.0, + "learning_rate": 1.0004028271139268e-07, + "loss": 0.684, + "step": 27346 + }, + { + "epoch": 1.0, + "learning_rate": 1.0003662064672062e-07, + "loss": 0.6883, + "step": 27347 + }, + { + "epoch": 1.0, + "learning_rate": 1.0003295858204856e-07, + "loss": 0.689, + "step": 27348 + }, + { + "epoch": 1.0, + "learning_rate": 1.0002929651737648e-07, + "loss": 0.6859, + "step": 27349 + }, + { + "epoch": 1.0, + "learning_rate": 1.0002563445270442e-07, + "loss": 0.6751, + "step": 27350 + }, + { + "epoch": 1.0, + "learning_rate": 1.0002197238803236e-07, + "loss": 0.6882, + "step": 27351 + }, + { + "epoch": 1.0, + "learning_rate": 1.000183103233603e-07, + "loss": 0.6878, + "step": 27352 + }, + { + "epoch": 1.0, + "learning_rate": 1.0001464825868824e-07, + "loss": 0.6907, + "step": 27353 + }, + { + "epoch": 1.0, + "learning_rate": 1.0001098619401618e-07, + "loss": 0.6849, + "step": 27354 + }, + { + "epoch": 1.0, + "learning_rate": 1.0000732412934412e-07, + "loss": 0.6881, + "step": 27355 + }, + { + "epoch": 1.0, + "learning_rate": 1.0000366206467206e-07, + "loss": 0.6896, + "step": 27356 + }, + { + "epoch": 1.0, + "learning_rate": 1e-07, + "loss": 0.7028, + "step": 27357 + }, + { + "epoch": 1.0, + "learning_rate": 9.999633793532794e-08, + "loss": 0.6879, + "step": 27358 + }, + { + "epoch": 1.0, + "learning_rate": 9.999267587065588e-08, + "loss": 0.6798, + "step": 27359 + }, + { + "epoch": 1.0, + "learning_rate": 9.998901380598382e-08, + "loss": 0.704, + "step": 27360 + }, + { + "epoch": 1.0, + "learning_rate": 9.998535174131176e-08, + "loss": 0.669, + "step": 27361 + }, + { + "epoch": 1.0, + "learning_rate": 9.998168967663968e-08, + "loss": 0.6988, + "step": 27362 + }, + { + "epoch": 1.0, + "learning_rate": 9.997802761196762e-08, + "loss": 0.6907, + "step": 27363 + }, + { + "epoch": 1.0, + "learning_rate": 9.997436554729556e-08, + "loss": 0.6847, + "step": 27364 + }, + { + "epoch": 1.0, + "learning_rate": 9.99707034826235e-08, + "loss": 0.6903, + "step": 27365 + }, + { + "epoch": 1.0, + "learning_rate": 9.996704141795144e-08, + "loss": 0.6851, + "step": 27366 + }, + { + "epoch": 1.0, + "learning_rate": 9.996337935327938e-08, + "loss": 0.6872, + "step": 27367 + }, + { + "epoch": 1.0, + "learning_rate": 9.995971728860731e-08, + "loss": 0.6954, + "step": 27368 + }, + { + "epoch": 1.0, + "learning_rate": 9.995605522393525e-08, + "loss": 0.687, + "step": 27369 + }, + { + "epoch": 1.0, + "learning_rate": 9.995239315926319e-08, + "loss": 0.6815, + "step": 27370 + }, + { + "epoch": 1.0, + "learning_rate": 9.994873109459111e-08, + "loss": 0.7079, + "step": 27371 + }, + { + "epoch": 1.0, + "learning_rate": 9.994506902991905e-08, + "loss": 0.6993, + "step": 27372 + }, + { + "epoch": 1.0, + "learning_rate": 9.9941406965247e-08, + "loss": 0.6846, + "step": 27373 + }, + { + "epoch": 1.0, + "learning_rate": 9.993774490057493e-08, + "loss": 0.6824, + "step": 27374 + }, + { + "epoch": 1.0, + "learning_rate": 9.993408283590287e-08, + "loss": 0.6961, + "step": 27375 + }, + { + "epoch": 1.0, + "learning_rate": 9.993042077123081e-08, + "loss": 0.7068, + "step": 27376 + }, + { + "epoch": 1.0, + "learning_rate": 9.992675870655875e-08, + "loss": 0.702, + "step": 27377 + }, + { + "epoch": 1.0, + "learning_rate": 9.99230966418867e-08, + "loss": 0.698, + "step": 27378 + }, + { + "epoch": 1.0, + "learning_rate": 9.991943457721463e-08, + "loss": 0.6811, + "step": 27379 + }, + { + "epoch": 1.0, + "learning_rate": 9.991577251254256e-08, + "loss": 0.7022, + "step": 27380 + }, + { + "epoch": 1.0, + "learning_rate": 9.99121104478705e-08, + "loss": 0.6964, + "step": 27381 + }, + { + "epoch": 1.0, + "learning_rate": 9.990844838319844e-08, + "loss": 0.6954, + "step": 27382 + }, + { + "epoch": 1.0, + "learning_rate": 9.990478631852638e-08, + "loss": 0.6891, + "step": 27383 + }, + { + "epoch": 1.0, + "learning_rate": 9.990112425385432e-08, + "loss": 0.691, + "step": 27384 + }, + { + "epoch": 1.0, + "learning_rate": 9.989746218918226e-08, + "loss": 0.6986, + "step": 27385 + }, + { + "epoch": 1.0, + "learning_rate": 9.98938001245102e-08, + "loss": 0.6978, + "step": 27386 + }, + { + "epoch": 1.0, + "learning_rate": 9.989013805983814e-08, + "loss": 0.6859, + "step": 27387 + }, + { + "epoch": 1.0, + "learning_rate": 9.988647599516608e-08, + "loss": 0.6946, + "step": 27388 + }, + { + "epoch": 1.0, + "learning_rate": 9.988281393049401e-08, + "loss": 0.6984, + "step": 27389 + }, + { + "epoch": 1.0, + "learning_rate": 9.987915186582195e-08, + "loss": 0.6877, + "step": 27390 + }, + { + "epoch": 1.0, + "learning_rate": 9.987548980114989e-08, + "loss": 0.6986, + "step": 27391 + }, + { + "epoch": 1.0, + "learning_rate": 9.987182773647783e-08, + "loss": 0.6933, + "step": 27392 + }, + { + "epoch": 1.0, + "learning_rate": 9.986816567180575e-08, + "loss": 0.6847, + "step": 27393 + }, + { + "epoch": 1.0, + "learning_rate": 9.98645036071337e-08, + "loss": 0.7026, + "step": 27394 + }, + { + "epoch": 1.0, + "learning_rate": 9.986084154246163e-08, + "loss": 0.6933, + "step": 27395 + }, + { + "epoch": 1.0, + "learning_rate": 9.985717947778957e-08, + "loss": 0.7046, + "step": 27396 + }, + { + "epoch": 1.0, + "learning_rate": 9.985351741311751e-08, + "loss": 0.6977, + "step": 27397 + }, + { + "epoch": 1.0, + "learning_rate": 9.984985534844544e-08, + "loss": 0.6885, + "step": 27398 + }, + { + "epoch": 1.0, + "learning_rate": 9.984619328377338e-08, + "loss": 0.6916, + "step": 27399 + }, + { + "epoch": 1.0, + "learning_rate": 9.984253121910132e-08, + "loss": 0.6852, + "step": 27400 + }, + { + "epoch": 1.0, + "learning_rate": 9.983886915442926e-08, + "loss": 0.6811, + "step": 27401 + }, + { + "epoch": 1.0, + "learning_rate": 9.98352070897572e-08, + "loss": 0.6834, + "step": 27402 + }, + { + "epoch": 1.0, + "learning_rate": 9.983154502508514e-08, + "loss": 0.6977, + "step": 27403 + }, + { + "epoch": 1.0, + "learning_rate": 9.982788296041308e-08, + "loss": 0.6789, + "step": 27404 + }, + { + "epoch": 1.0, + "learning_rate": 9.982422089574102e-08, + "loss": 0.6955, + "step": 27405 + }, + { + "epoch": 1.0, + "learning_rate": 9.982055883106896e-08, + "loss": 0.7061, + "step": 27406 + }, + { + "epoch": 1.0, + "learning_rate": 9.981689676639689e-08, + "loss": 0.695, + "step": 27407 + }, + { + "epoch": 1.0, + "learning_rate": 9.981323470172483e-08, + "loss": 0.7105, + "step": 27408 + }, + { + "epoch": 1.0, + "learning_rate": 9.980957263705277e-08, + "loss": 0.6874, + "step": 27409 + }, + { + "epoch": 1.0, + "learning_rate": 9.98059105723807e-08, + "loss": 0.6828, + "step": 27410 + }, + { + "epoch": 1.0, + "learning_rate": 9.980224850770865e-08, + "loss": 0.6906, + "step": 27411 + }, + { + "epoch": 1.0, + "learning_rate": 9.979858644303659e-08, + "loss": 0.7027, + "step": 27412 + }, + { + "epoch": 1.0, + "learning_rate": 9.979492437836453e-08, + "loss": 0.6959, + "step": 27413 + }, + { + "epoch": 1.0, + "learning_rate": 9.979126231369247e-08, + "loss": 0.6923, + "step": 27414 + }, + { + "epoch": 1.0, + "learning_rate": 9.978760024902039e-08, + "loss": 0.69, + "step": 27415 + }, + { + "epoch": 1.0, + "learning_rate": 9.978393818434833e-08, + "loss": 0.6965, + "step": 27416 + }, + { + "epoch": 1.0, + "learning_rate": 9.978027611967626e-08, + "loss": 0.6955, + "step": 27417 + }, + { + "epoch": 1.0, + "learning_rate": 9.97766140550042e-08, + "loss": 0.6994, + "step": 27418 + }, + { + "epoch": 1.0, + "learning_rate": 9.977295199033214e-08, + "loss": 0.6978, + "step": 27419 + }, + { + "epoch": 1.0, + "learning_rate": 9.976928992566008e-08, + "loss": 0.6971, + "step": 27420 + }, + { + "epoch": 1.0, + "learning_rate": 9.976562786098802e-08, + "loss": 0.6962, + "step": 27421 + }, + { + "epoch": 1.0, + "learning_rate": 9.976196579631596e-08, + "loss": 0.7042, + "step": 27422 + }, + { + "epoch": 1.0, + "learning_rate": 9.97583037316439e-08, + "loss": 0.6906, + "step": 27423 + }, + { + "epoch": 1.0, + "learning_rate": 9.975464166697184e-08, + "loss": 0.6839, + "step": 27424 + }, + { + "epoch": 1.0, + "learning_rate": 9.975097960229977e-08, + "loss": 0.6955, + "step": 27425 + }, + { + "epoch": 1.0, + "learning_rate": 9.97473175376277e-08, + "loss": 0.6943, + "step": 27426 + }, + { + "epoch": 1.0, + "learning_rate": 9.974365547295564e-08, + "loss": 0.6828, + "step": 27427 + }, + { + "epoch": 1.0, + "learning_rate": 9.973999340828358e-08, + "loss": 0.6913, + "step": 27428 + }, + { + "epoch": 1.0, + "learning_rate": 9.973633134361152e-08, + "loss": 0.6997, + "step": 27429 + }, + { + "epoch": 1.0, + "learning_rate": 9.973266927893946e-08, + "loss": 0.7054, + "step": 27430 + }, + { + "epoch": 1.0, + "learning_rate": 9.97290072142674e-08, + "loss": 0.6829, + "step": 27431 + }, + { + "epoch": 1.0, + "learning_rate": 9.972534514959534e-08, + "loss": 0.691, + "step": 27432 + }, + { + "epoch": 1.0, + "learning_rate": 9.972168308492328e-08, + "loss": 0.682, + "step": 27433 + }, + { + "epoch": 1.0, + "learning_rate": 9.971802102025121e-08, + "loss": 0.6963, + "step": 27434 + }, + { + "epoch": 1.0, + "learning_rate": 9.971435895557915e-08, + "loss": 0.6842, + "step": 27435 + }, + { + "epoch": 1.0, + "learning_rate": 9.971069689090709e-08, + "loss": 0.6972, + "step": 27436 + }, + { + "epoch": 1.0, + "learning_rate": 9.970703482623503e-08, + "loss": 0.6899, + "step": 27437 + }, + { + "epoch": 1.0, + "learning_rate": 9.970337276156297e-08, + "loss": 0.7026, + "step": 27438 + }, + { + "epoch": 1.0, + "learning_rate": 9.96997106968909e-08, + "loss": 0.6913, + "step": 27439 + }, + { + "epoch": 1.0, + "learning_rate": 9.969604863221884e-08, + "loss": 0.7012, + "step": 27440 + }, + { + "epoch": 1.0, + "learning_rate": 9.969238656754678e-08, + "loss": 0.6891, + "step": 27441 + }, + { + "epoch": 1.0, + "learning_rate": 9.968872450287472e-08, + "loss": 0.6746, + "step": 27442 + }, + { + "epoch": 1.0, + "learning_rate": 9.968506243820264e-08, + "loss": 0.7003, + "step": 27443 + }, + { + "epoch": 1.0, + "learning_rate": 9.968140037353058e-08, + "loss": 0.704, + "step": 27444 + }, + { + "epoch": 1.0, + "learning_rate": 9.967773830885852e-08, + "loss": 0.6988, + "step": 27445 + }, + { + "epoch": 1.0, + "learning_rate": 9.967407624418646e-08, + "loss": 0.6907, + "step": 27446 + }, + { + "epoch": 1.0, + "learning_rate": 9.96704141795144e-08, + "loss": 0.6789, + "step": 27447 + }, + { + "epoch": 1.0, + "learning_rate": 9.966675211484234e-08, + "loss": 0.6986, + "step": 27448 + }, + { + "epoch": 1.0, + "learning_rate": 9.966309005017028e-08, + "loss": 0.6963, + "step": 27449 + }, + { + "epoch": 1.0, + "learning_rate": 9.965942798549822e-08, + "loss": 0.7041, + "step": 27450 + }, + { + "epoch": 1.0, + "learning_rate": 9.965576592082616e-08, + "loss": 0.6831, + "step": 27451 + }, + { + "epoch": 1.0, + "learning_rate": 9.965210385615409e-08, + "loss": 0.6812, + "step": 27452 + }, + { + "epoch": 1.0, + "learning_rate": 9.964844179148203e-08, + "loss": 0.6973, + "step": 27453 + }, + { + "epoch": 1.0, + "learning_rate": 9.964477972680997e-08, + "loss": 0.6926, + "step": 27454 + }, + { + "epoch": 1.0, + "learning_rate": 9.964111766213791e-08, + "loss": 0.6973, + "step": 27455 + }, + { + "epoch": 1.0, + "learning_rate": 9.963745559746585e-08, + "loss": 0.6878, + "step": 27456 + }, + { + "epoch": 1.0, + "learning_rate": 9.963379353279379e-08, + "loss": 0.6946, + "step": 27457 + }, + { + "epoch": 1.0, + "learning_rate": 9.963013146812173e-08, + "loss": 0.6837, + "step": 27458 + }, + { + "epoch": 1.0, + "learning_rate": 9.962646940344967e-08, + "loss": 0.7009, + "step": 27459 + }, + { + "epoch": 1.0, + "learning_rate": 9.962280733877761e-08, + "loss": 0.6924, + "step": 27460 + }, + { + "epoch": 1.0, + "learning_rate": 9.961914527410554e-08, + "loss": 0.7122, + "step": 27461 + }, + { + "epoch": 1.0, + "learning_rate": 9.961548320943348e-08, + "loss": 0.7033, + "step": 27462 + }, + { + "epoch": 1.0, + "learning_rate": 9.961182114476142e-08, + "loss": 0.6838, + "step": 27463 + }, + { + "epoch": 1.0, + "learning_rate": 9.960815908008934e-08, + "loss": 0.7018, + "step": 27464 + }, + { + "epoch": 1.0, + "learning_rate": 9.960449701541728e-08, + "loss": 0.696, + "step": 27465 + }, + { + "epoch": 1.0, + "learning_rate": 9.960083495074522e-08, + "loss": 0.6831, + "step": 27466 + }, + { + "epoch": 1.0, + "learning_rate": 9.959717288607316e-08, + "loss": 0.6914, + "step": 27467 + }, + { + "epoch": 1.0, + "learning_rate": 9.95935108214011e-08, + "loss": 0.6842, + "step": 27468 + }, + { + "epoch": 1.01, + "learning_rate": 9.958984875672904e-08, + "loss": 0.6987, + "step": 27469 + }, + { + "epoch": 1.01, + "learning_rate": 9.958618669205697e-08, + "loss": 0.7041, + "step": 27470 + }, + { + "epoch": 1.01, + "learning_rate": 9.958252462738491e-08, + "loss": 0.6844, + "step": 27471 + }, + { + "epoch": 1.01, + "learning_rate": 9.957886256271285e-08, + "loss": 0.6993, + "step": 27472 + }, + { + "epoch": 1.01, + "learning_rate": 9.957520049804079e-08, + "loss": 0.6913, + "step": 27473 + }, + { + "epoch": 1.01, + "learning_rate": 9.957153843336873e-08, + "loss": 0.7011, + "step": 27474 + }, + { + "epoch": 1.01, + "learning_rate": 9.956787636869667e-08, + "loss": 0.69, + "step": 27475 + }, + { + "epoch": 1.01, + "learning_rate": 9.956421430402461e-08, + "loss": 0.6906, + "step": 27476 + }, + { + "epoch": 1.01, + "learning_rate": 9.956055223935255e-08, + "loss": 0.6912, + "step": 27477 + }, + { + "epoch": 1.01, + "learning_rate": 9.955689017468049e-08, + "loss": 0.6943, + "step": 27478 + }, + { + "epoch": 1.01, + "learning_rate": 9.955322811000842e-08, + "loss": 0.6834, + "step": 27479 + }, + { + "epoch": 1.01, + "learning_rate": 9.954956604533636e-08, + "loss": 0.6864, + "step": 27480 + }, + { + "epoch": 1.01, + "learning_rate": 9.95459039806643e-08, + "loss": 0.6963, + "step": 27481 + }, + { + "epoch": 1.01, + "learning_rate": 9.954224191599224e-08, + "loss": 0.6929, + "step": 27482 + }, + { + "epoch": 1.01, + "learning_rate": 9.953857985132018e-08, + "loss": 0.6992, + "step": 27483 + }, + { + "epoch": 1.01, + "learning_rate": 9.953491778664811e-08, + "loss": 0.6861, + "step": 27484 + }, + { + "epoch": 1.01, + "learning_rate": 9.953125572197605e-08, + "loss": 0.6853, + "step": 27485 + }, + { + "epoch": 1.01, + "learning_rate": 9.952759365730398e-08, + "loss": 0.7057, + "step": 27486 + }, + { + "epoch": 1.01, + "learning_rate": 9.952393159263192e-08, + "loss": 0.6898, + "step": 27487 + }, + { + "epoch": 1.01, + "learning_rate": 9.952026952795985e-08, + "loss": 0.6999, + "step": 27488 + }, + { + "epoch": 1.01, + "learning_rate": 9.951660746328779e-08, + "loss": 0.6935, + "step": 27489 + }, + { + "epoch": 1.01, + "learning_rate": 9.951294539861573e-08, + "loss": 0.6938, + "step": 27490 + }, + { + "epoch": 1.01, + "learning_rate": 9.950928333394367e-08, + "loss": 0.6927, + "step": 27491 + }, + { + "epoch": 1.01, + "learning_rate": 9.950562126927161e-08, + "loss": 0.6691, + "step": 27492 + }, + { + "epoch": 1.01, + "learning_rate": 9.950195920459955e-08, + "loss": 0.6849, + "step": 27493 + }, + { + "epoch": 1.01, + "learning_rate": 9.949829713992749e-08, + "loss": 0.6977, + "step": 27494 + }, + { + "epoch": 1.01, + "learning_rate": 9.949463507525543e-08, + "loss": 0.6851, + "step": 27495 + }, + { + "epoch": 1.01, + "learning_rate": 9.949097301058337e-08, + "loss": 0.6763, + "step": 27496 + }, + { + "epoch": 1.01, + "learning_rate": 9.94873109459113e-08, + "loss": 0.6866, + "step": 27497 + }, + { + "epoch": 1.01, + "learning_rate": 9.948364888123923e-08, + "loss": 0.6972, + "step": 27498 + }, + { + "epoch": 1.01, + "learning_rate": 9.947998681656717e-08, + "loss": 0.7026, + "step": 27499 + }, + { + "epoch": 1.01, + "learning_rate": 9.947632475189511e-08, + "loss": 0.6942, + "step": 27500 + }, + { + "epoch": 1.01, + "learning_rate": 9.947266268722305e-08, + "loss": 0.6883, + "step": 27501 + }, + { + "epoch": 1.01, + "learning_rate": 9.9469000622551e-08, + "loss": 0.68, + "step": 27502 + }, + { + "epoch": 1.01, + "learning_rate": 9.946533855787893e-08, + "loss": 0.6972, + "step": 27503 + }, + { + "epoch": 1.01, + "learning_rate": 9.946167649320687e-08, + "loss": 0.6882, + "step": 27504 + }, + { + "epoch": 1.01, + "learning_rate": 9.945801442853481e-08, + "loss": 0.7056, + "step": 27505 + }, + { + "epoch": 1.01, + "learning_rate": 9.945435236386274e-08, + "loss": 0.7014, + "step": 27506 + }, + { + "epoch": 1.01, + "learning_rate": 9.945069029919068e-08, + "loss": 0.6943, + "step": 27507 + }, + { + "epoch": 1.01, + "learning_rate": 9.944702823451862e-08, + "loss": 0.6986, + "step": 27508 + }, + { + "epoch": 1.01, + "learning_rate": 9.944336616984656e-08, + "loss": 0.6938, + "step": 27509 + }, + { + "epoch": 1.01, + "learning_rate": 9.943970410517449e-08, + "loss": 0.6966, + "step": 27510 + }, + { + "epoch": 1.01, + "learning_rate": 9.943604204050243e-08, + "loss": 0.6881, + "step": 27511 + }, + { + "epoch": 1.01, + "learning_rate": 9.943237997583037e-08, + "loss": 0.691, + "step": 27512 + }, + { + "epoch": 1.01, + "learning_rate": 9.942871791115831e-08, + "loss": 0.7128, + "step": 27513 + }, + { + "epoch": 1.01, + "learning_rate": 9.942505584648625e-08, + "loss": 0.6993, + "step": 27514 + }, + { + "epoch": 1.01, + "learning_rate": 9.942139378181417e-08, + "loss": 0.6875, + "step": 27515 + }, + { + "epoch": 1.01, + "learning_rate": 9.941773171714211e-08, + "loss": 0.7017, + "step": 27516 + }, + { + "epoch": 1.01, + "learning_rate": 9.941406965247005e-08, + "loss": 0.7034, + "step": 27517 + }, + { + "epoch": 1.01, + "learning_rate": 9.941040758779799e-08, + "loss": 0.6823, + "step": 27518 + }, + { + "epoch": 1.01, + "learning_rate": 9.940674552312593e-08, + "loss": 0.6967, + "step": 27519 + }, + { + "epoch": 1.01, + "learning_rate": 9.940308345845387e-08, + "loss": 0.6976, + "step": 27520 + }, + { + "epoch": 1.01, + "learning_rate": 9.939942139378181e-08, + "loss": 0.6907, + "step": 27521 + }, + { + "epoch": 1.01, + "learning_rate": 9.939575932910975e-08, + "loss": 0.7078, + "step": 27522 + }, + { + "epoch": 1.01, + "learning_rate": 9.939209726443769e-08, + "loss": 0.6977, + "step": 27523 + }, + { + "epoch": 1.01, + "learning_rate": 9.938843519976562e-08, + "loss": 0.7014, + "step": 27524 + }, + { + "epoch": 1.01, + "learning_rate": 9.938477313509356e-08, + "loss": 0.6962, + "step": 27525 + }, + { + "epoch": 1.01, + "learning_rate": 9.93811110704215e-08, + "loss": 0.6858, + "step": 27526 + }, + { + "epoch": 1.01, + "learning_rate": 9.937744900574944e-08, + "loss": 0.6709, + "step": 27527 + }, + { + "epoch": 1.01, + "learning_rate": 9.937378694107738e-08, + "loss": 0.6863, + "step": 27528 + }, + { + "epoch": 1.01, + "learning_rate": 9.937012487640532e-08, + "loss": 0.7, + "step": 27529 + }, + { + "epoch": 1.01, + "learning_rate": 9.936646281173326e-08, + "loss": 0.6882, + "step": 27530 + }, + { + "epoch": 1.01, + "learning_rate": 9.93628007470612e-08, + "loss": 0.6939, + "step": 27531 + }, + { + "epoch": 1.01, + "learning_rate": 9.935913868238913e-08, + "loss": 0.6838, + "step": 27532 + }, + { + "epoch": 1.01, + "learning_rate": 9.935547661771707e-08, + "loss": 0.6802, + "step": 27533 + }, + { + "epoch": 1.01, + "learning_rate": 9.9351814553045e-08, + "loss": 0.7017, + "step": 27534 + }, + { + "epoch": 1.01, + "learning_rate": 9.934815248837293e-08, + "loss": 0.7056, + "step": 27535 + }, + { + "epoch": 1.01, + "learning_rate": 9.934449042370087e-08, + "loss": 0.6936, + "step": 27536 + }, + { + "epoch": 1.01, + "learning_rate": 9.934082835902881e-08, + "loss": 0.675, + "step": 27537 + }, + { + "epoch": 1.01, + "learning_rate": 9.933716629435675e-08, + "loss": 0.6927, + "step": 27538 + }, + { + "epoch": 1.01, + "learning_rate": 9.933350422968469e-08, + "loss": 0.7081, + "step": 27539 + }, + { + "epoch": 1.01, + "learning_rate": 9.932984216501263e-08, + "loss": 0.6896, + "step": 27540 + }, + { + "epoch": 1.01, + "learning_rate": 9.932618010034057e-08, + "loss": 0.6941, + "step": 27541 + }, + { + "epoch": 1.01, + "learning_rate": 9.93225180356685e-08, + "loss": 0.6889, + "step": 27542 + }, + { + "epoch": 1.01, + "learning_rate": 9.931885597099644e-08, + "loss": 0.6947, + "step": 27543 + }, + { + "epoch": 1.01, + "learning_rate": 9.931519390632438e-08, + "loss": 0.6839, + "step": 27544 + }, + { + "epoch": 1.01, + "learning_rate": 9.931153184165232e-08, + "loss": 0.694, + "step": 27545 + }, + { + "epoch": 1.01, + "learning_rate": 9.930786977698026e-08, + "loss": 0.7086, + "step": 27546 + }, + { + "epoch": 1.01, + "learning_rate": 9.93042077123082e-08, + "loss": 0.6743, + "step": 27547 + }, + { + "epoch": 1.01, + "learning_rate": 9.930054564763614e-08, + "loss": 0.6848, + "step": 27548 + }, + { + "epoch": 1.01, + "learning_rate": 9.929688358296408e-08, + "loss": 0.6893, + "step": 27549 + }, + { + "epoch": 1.01, + "learning_rate": 9.9293221518292e-08, + "loss": 0.7014, + "step": 27550 + }, + { + "epoch": 1.01, + "learning_rate": 9.928955945361994e-08, + "loss": 0.6865, + "step": 27551 + }, + { + "epoch": 1.01, + "learning_rate": 9.928589738894788e-08, + "loss": 0.6863, + "step": 27552 + }, + { + "epoch": 1.01, + "learning_rate": 9.928223532427582e-08, + "loss": 0.6884, + "step": 27553 + }, + { + "epoch": 1.01, + "learning_rate": 9.927857325960376e-08, + "loss": 0.6723, + "step": 27554 + }, + { + "epoch": 1.01, + "learning_rate": 9.92749111949317e-08, + "loss": 0.6823, + "step": 27555 + }, + { + "epoch": 1.01, + "learning_rate": 9.927124913025964e-08, + "loss": 0.7023, + "step": 27556 + }, + { + "epoch": 1.01, + "learning_rate": 9.926758706558757e-08, + "loss": 0.6974, + "step": 27557 + }, + { + "epoch": 1.01, + "learning_rate": 9.926392500091551e-08, + "loss": 0.6917, + "step": 27558 + }, + { + "epoch": 1.01, + "learning_rate": 9.926026293624344e-08, + "loss": 0.6864, + "step": 27559 + }, + { + "epoch": 1.01, + "learning_rate": 9.925660087157138e-08, + "loss": 0.699, + "step": 27560 + }, + { + "epoch": 1.01, + "learning_rate": 9.925293880689932e-08, + "loss": 0.6925, + "step": 27561 + }, + { + "epoch": 1.01, + "learning_rate": 9.924927674222726e-08, + "loss": 0.6725, + "step": 27562 + }, + { + "epoch": 1.01, + "learning_rate": 9.92456146775552e-08, + "loss": 0.6989, + "step": 27563 + }, + { + "epoch": 1.01, + "learning_rate": 9.924195261288314e-08, + "loss": 0.6832, + "step": 27564 + }, + { + "epoch": 1.01, + "learning_rate": 9.923829054821108e-08, + "loss": 0.7006, + "step": 27565 + }, + { + "epoch": 1.01, + "learning_rate": 9.923462848353902e-08, + "loss": 0.688, + "step": 27566 + }, + { + "epoch": 1.01, + "learning_rate": 9.923096641886696e-08, + "loss": 0.7037, + "step": 27567 + }, + { + "epoch": 1.01, + "learning_rate": 9.922730435419488e-08, + "loss": 0.6982, + "step": 27568 + }, + { + "epoch": 1.01, + "learning_rate": 9.922364228952282e-08, + "loss": 0.6929, + "step": 27569 + }, + { + "epoch": 1.01, + "learning_rate": 9.921998022485076e-08, + "loss": 0.699, + "step": 27570 + }, + { + "epoch": 1.01, + "learning_rate": 9.92163181601787e-08, + "loss": 0.6904, + "step": 27571 + }, + { + "epoch": 1.01, + "learning_rate": 9.921265609550664e-08, + "loss": 0.6959, + "step": 27572 + }, + { + "epoch": 1.01, + "learning_rate": 9.920899403083458e-08, + "loss": 0.7112, + "step": 27573 + }, + { + "epoch": 1.01, + "learning_rate": 9.920533196616252e-08, + "loss": 0.6987, + "step": 27574 + }, + { + "epoch": 1.01, + "learning_rate": 9.920166990149046e-08, + "loss": 0.6906, + "step": 27575 + }, + { + "epoch": 1.01, + "learning_rate": 9.91980078368184e-08, + "loss": 0.6872, + "step": 27576 + }, + { + "epoch": 1.01, + "learning_rate": 9.919434577214633e-08, + "loss": 0.6865, + "step": 27577 + }, + { + "epoch": 1.01, + "learning_rate": 9.919068370747427e-08, + "loss": 0.6934, + "step": 27578 + }, + { + "epoch": 1.01, + "learning_rate": 9.918702164280221e-08, + "loss": 0.6912, + "step": 27579 + }, + { + "epoch": 1.01, + "learning_rate": 9.918335957813015e-08, + "loss": 0.6932, + "step": 27580 + }, + { + "epoch": 1.01, + "learning_rate": 9.917969751345808e-08, + "loss": 0.7028, + "step": 27581 + }, + { + "epoch": 1.01, + "learning_rate": 9.917603544878602e-08, + "loss": 0.6842, + "step": 27582 + }, + { + "epoch": 1.01, + "learning_rate": 9.917237338411396e-08, + "loss": 0.697, + "step": 27583 + }, + { + "epoch": 1.01, + "learning_rate": 9.91687113194419e-08, + "loss": 0.7004, + "step": 27584 + }, + { + "epoch": 1.01, + "learning_rate": 9.916504925476984e-08, + "loss": 0.7003, + "step": 27585 + }, + { + "epoch": 1.01, + "learning_rate": 9.916138719009776e-08, + "loss": 0.7072, + "step": 27586 + }, + { + "epoch": 1.01, + "learning_rate": 9.91577251254257e-08, + "loss": 0.6894, + "step": 27587 + }, + { + "epoch": 1.01, + "learning_rate": 9.915406306075364e-08, + "loss": 0.6935, + "step": 27588 + }, + { + "epoch": 1.01, + "learning_rate": 9.915040099608158e-08, + "loss": 0.7006, + "step": 27589 + }, + { + "epoch": 1.01, + "learning_rate": 9.914673893140952e-08, + "loss": 0.69, + "step": 27590 + }, + { + "epoch": 1.01, + "learning_rate": 9.914307686673746e-08, + "loss": 0.6951, + "step": 27591 + }, + { + "epoch": 1.01, + "learning_rate": 9.91394148020654e-08, + "loss": 0.6895, + "step": 27592 + }, + { + "epoch": 1.01, + "learning_rate": 9.913575273739334e-08, + "loss": 0.6932, + "step": 27593 + }, + { + "epoch": 1.01, + "learning_rate": 9.913209067272128e-08, + "loss": 0.6881, + "step": 27594 + }, + { + "epoch": 1.01, + "learning_rate": 9.912842860804921e-08, + "loss": 0.6911, + "step": 27595 + }, + { + "epoch": 1.01, + "learning_rate": 9.912476654337715e-08, + "loss": 0.6918, + "step": 27596 + }, + { + "epoch": 1.01, + "learning_rate": 9.912110447870509e-08, + "loss": 0.6848, + "step": 27597 + }, + { + "epoch": 1.01, + "learning_rate": 9.911744241403303e-08, + "loss": 0.6928, + "step": 27598 + }, + { + "epoch": 1.01, + "learning_rate": 9.911378034936097e-08, + "loss": 0.7008, + "step": 27599 + }, + { + "epoch": 1.01, + "learning_rate": 9.911011828468891e-08, + "loss": 0.68, + "step": 27600 + }, + { + "epoch": 1.01, + "learning_rate": 9.910645622001685e-08, + "loss": 0.6822, + "step": 27601 + }, + { + "epoch": 1.01, + "learning_rate": 9.910279415534479e-08, + "loss": 0.696, + "step": 27602 + }, + { + "epoch": 1.01, + "learning_rate": 9.909913209067271e-08, + "loss": 0.7127, + "step": 27603 + }, + { + "epoch": 1.01, + "learning_rate": 9.909547002600065e-08, + "loss": 0.6988, + "step": 27604 + }, + { + "epoch": 1.01, + "learning_rate": 9.90918079613286e-08, + "loss": 0.684, + "step": 27605 + }, + { + "epoch": 1.01, + "learning_rate": 9.908814589665652e-08, + "loss": 0.6819, + "step": 27606 + }, + { + "epoch": 1.01, + "learning_rate": 9.908448383198446e-08, + "loss": 0.6862, + "step": 27607 + }, + { + "epoch": 1.01, + "learning_rate": 9.90808217673124e-08, + "loss": 0.7035, + "step": 27608 + }, + { + "epoch": 1.01, + "learning_rate": 9.907715970264034e-08, + "loss": 0.6922, + "step": 27609 + }, + { + "epoch": 1.01, + "learning_rate": 9.907349763796828e-08, + "loss": 0.7072, + "step": 27610 + }, + { + "epoch": 1.01, + "learning_rate": 9.906983557329622e-08, + "loss": 0.6857, + "step": 27611 + }, + { + "epoch": 1.01, + "learning_rate": 9.906617350862416e-08, + "loss": 0.6962, + "step": 27612 + }, + { + "epoch": 1.01, + "learning_rate": 9.906251144395209e-08, + "loss": 0.6951, + "step": 27613 + }, + { + "epoch": 1.01, + "learning_rate": 9.905884937928003e-08, + "loss": 0.6893, + "step": 27614 + }, + { + "epoch": 1.01, + "learning_rate": 9.905518731460797e-08, + "loss": 0.69, + "step": 27615 + }, + { + "epoch": 1.01, + "learning_rate": 9.905152524993591e-08, + "loss": 0.6891, + "step": 27616 + }, + { + "epoch": 1.01, + "learning_rate": 9.904786318526385e-08, + "loss": 0.6918, + "step": 27617 + }, + { + "epoch": 1.01, + "learning_rate": 9.904420112059179e-08, + "loss": 0.6949, + "step": 27618 + }, + { + "epoch": 1.01, + "learning_rate": 9.904053905591973e-08, + "loss": 0.6815, + "step": 27619 + }, + { + "epoch": 1.01, + "learning_rate": 9.903687699124767e-08, + "loss": 0.6916, + "step": 27620 + }, + { + "epoch": 1.01, + "learning_rate": 9.903321492657561e-08, + "loss": 0.6966, + "step": 27621 + }, + { + "epoch": 1.01, + "learning_rate": 9.902955286190353e-08, + "loss": 0.6958, + "step": 27622 + }, + { + "epoch": 1.01, + "learning_rate": 9.902589079723147e-08, + "loss": 0.6967, + "step": 27623 + }, + { + "epoch": 1.01, + "learning_rate": 9.902222873255941e-08, + "loss": 0.6908, + "step": 27624 + }, + { + "epoch": 1.01, + "learning_rate": 9.901856666788735e-08, + "loss": 0.6949, + "step": 27625 + }, + { + "epoch": 1.01, + "learning_rate": 9.90149046032153e-08, + "loss": 0.6977, + "step": 27626 + }, + { + "epoch": 1.01, + "learning_rate": 9.901124253854323e-08, + "loss": 0.6832, + "step": 27627 + }, + { + "epoch": 1.01, + "learning_rate": 9.900758047387116e-08, + "loss": 0.6952, + "step": 27628 + }, + { + "epoch": 1.01, + "learning_rate": 9.90039184091991e-08, + "loss": 0.6804, + "step": 27629 + }, + { + "epoch": 1.01, + "learning_rate": 9.900025634452704e-08, + "loss": 0.6879, + "step": 27630 + }, + { + "epoch": 1.01, + "learning_rate": 9.899659427985497e-08, + "loss": 0.6865, + "step": 27631 + }, + { + "epoch": 1.01, + "learning_rate": 9.899293221518291e-08, + "loss": 0.6841, + "step": 27632 + }, + { + "epoch": 1.01, + "learning_rate": 9.898927015051085e-08, + "loss": 0.7002, + "step": 27633 + }, + { + "epoch": 1.01, + "learning_rate": 9.898560808583879e-08, + "loss": 0.6935, + "step": 27634 + }, + { + "epoch": 1.01, + "learning_rate": 9.898194602116673e-08, + "loss": 0.6836, + "step": 27635 + }, + { + "epoch": 1.01, + "learning_rate": 9.897828395649467e-08, + "loss": 0.6838, + "step": 27636 + }, + { + "epoch": 1.01, + "learning_rate": 9.89746218918226e-08, + "loss": 0.693, + "step": 27637 + }, + { + "epoch": 1.01, + "learning_rate": 9.897095982715055e-08, + "loss": 0.6902, + "step": 27638 + }, + { + "epoch": 1.01, + "learning_rate": 9.896729776247849e-08, + "loss": 0.6979, + "step": 27639 + }, + { + "epoch": 1.01, + "learning_rate": 9.896363569780641e-08, + "loss": 0.6945, + "step": 27640 + }, + { + "epoch": 1.01, + "learning_rate": 9.895997363313435e-08, + "loss": 0.6919, + "step": 27641 + }, + { + "epoch": 1.01, + "learning_rate": 9.895631156846229e-08, + "loss": 0.6879, + "step": 27642 + }, + { + "epoch": 1.01, + "learning_rate": 9.895264950379023e-08, + "loss": 0.7028, + "step": 27643 + }, + { + "epoch": 1.01, + "learning_rate": 9.894898743911817e-08, + "loss": 0.6704, + "step": 27644 + }, + { + "epoch": 1.01, + "learning_rate": 9.894532537444611e-08, + "loss": 0.6764, + "step": 27645 + }, + { + "epoch": 1.01, + "learning_rate": 9.894166330977405e-08, + "loss": 0.6892, + "step": 27646 + }, + { + "epoch": 1.01, + "learning_rate": 9.893800124510199e-08, + "loss": 0.7002, + "step": 27647 + }, + { + "epoch": 1.01, + "learning_rate": 9.893433918042993e-08, + "loss": 0.6944, + "step": 27648 + }, + { + "epoch": 1.01, + "learning_rate": 9.893067711575786e-08, + "loss": 0.6894, + "step": 27649 + }, + { + "epoch": 1.01, + "learning_rate": 9.89270150510858e-08, + "loss": 0.6976, + "step": 27650 + }, + { + "epoch": 1.01, + "learning_rate": 9.892335298641374e-08, + "loss": 0.6932, + "step": 27651 + }, + { + "epoch": 1.01, + "learning_rate": 9.891969092174167e-08, + "loss": 0.6959, + "step": 27652 + }, + { + "epoch": 1.01, + "learning_rate": 9.89160288570696e-08, + "loss": 0.7026, + "step": 27653 + }, + { + "epoch": 1.01, + "learning_rate": 9.891236679239755e-08, + "loss": 0.6956, + "step": 27654 + }, + { + "epoch": 1.01, + "learning_rate": 9.890870472772549e-08, + "loss": 0.7034, + "step": 27655 + }, + { + "epoch": 1.01, + "learning_rate": 9.890504266305343e-08, + "loss": 0.6851, + "step": 27656 + }, + { + "epoch": 1.01, + "learning_rate": 9.890138059838137e-08, + "loss": 0.6935, + "step": 27657 + }, + { + "epoch": 1.01, + "learning_rate": 9.889771853370929e-08, + "loss": 0.694, + "step": 27658 + }, + { + "epoch": 1.01, + "learning_rate": 9.889405646903723e-08, + "loss": 0.6996, + "step": 27659 + }, + { + "epoch": 1.01, + "learning_rate": 9.889039440436517e-08, + "loss": 0.6831, + "step": 27660 + }, + { + "epoch": 1.01, + "learning_rate": 9.888673233969311e-08, + "loss": 0.7007, + "step": 27661 + }, + { + "epoch": 1.01, + "learning_rate": 9.888307027502105e-08, + "loss": 0.6886, + "step": 27662 + }, + { + "epoch": 1.01, + "learning_rate": 9.887940821034899e-08, + "loss": 0.6907, + "step": 27663 + }, + { + "epoch": 1.01, + "learning_rate": 9.887574614567693e-08, + "loss": 0.6787, + "step": 27664 + }, + { + "epoch": 1.01, + "learning_rate": 9.887208408100487e-08, + "loss": 0.684, + "step": 27665 + }, + { + "epoch": 1.01, + "learning_rate": 9.886842201633281e-08, + "loss": 0.6965, + "step": 27666 + }, + { + "epoch": 1.01, + "learning_rate": 9.886475995166074e-08, + "loss": 0.6899, + "step": 27667 + }, + { + "epoch": 1.01, + "learning_rate": 9.886109788698868e-08, + "loss": 0.6938, + "step": 27668 + }, + { + "epoch": 1.01, + "learning_rate": 9.885743582231662e-08, + "loss": 0.6849, + "step": 27669 + }, + { + "epoch": 1.01, + "learning_rate": 9.885377375764456e-08, + "loss": 0.7015, + "step": 27670 + }, + { + "epoch": 1.01, + "learning_rate": 9.88501116929725e-08, + "loss": 0.6974, + "step": 27671 + }, + { + "epoch": 1.01, + "learning_rate": 9.884644962830044e-08, + "loss": 0.6889, + "step": 27672 + }, + { + "epoch": 1.01, + "learning_rate": 9.884278756362838e-08, + "loss": 0.6914, + "step": 27673 + }, + { + "epoch": 1.01, + "learning_rate": 9.88391254989563e-08, + "loss": 0.6967, + "step": 27674 + }, + { + "epoch": 1.01, + "learning_rate": 9.883546343428424e-08, + "loss": 0.6833, + "step": 27675 + }, + { + "epoch": 1.01, + "learning_rate": 9.883180136961218e-08, + "loss": 0.6999, + "step": 27676 + }, + { + "epoch": 1.01, + "learning_rate": 9.882813930494011e-08, + "loss": 0.6953, + "step": 27677 + }, + { + "epoch": 1.01, + "learning_rate": 9.882447724026805e-08, + "loss": 0.6833, + "step": 27678 + }, + { + "epoch": 1.01, + "learning_rate": 9.882081517559599e-08, + "loss": 0.7038, + "step": 27679 + }, + { + "epoch": 1.01, + "learning_rate": 9.881715311092393e-08, + "loss": 0.6967, + "step": 27680 + }, + { + "epoch": 1.01, + "learning_rate": 9.881349104625187e-08, + "loss": 0.6998, + "step": 27681 + }, + { + "epoch": 1.01, + "learning_rate": 9.880982898157981e-08, + "loss": 0.6951, + "step": 27682 + }, + { + "epoch": 1.01, + "learning_rate": 9.880616691690775e-08, + "loss": 0.6846, + "step": 27683 + }, + { + "epoch": 1.01, + "learning_rate": 9.880250485223569e-08, + "loss": 0.699, + "step": 27684 + }, + { + "epoch": 1.01, + "learning_rate": 9.879884278756362e-08, + "loss": 0.6819, + "step": 27685 + }, + { + "epoch": 1.01, + "learning_rate": 9.879518072289156e-08, + "loss": 0.6947, + "step": 27686 + }, + { + "epoch": 1.01, + "learning_rate": 9.87915186582195e-08, + "loss": 0.6917, + "step": 27687 + }, + { + "epoch": 1.01, + "learning_rate": 9.878785659354744e-08, + "loss": 0.7037, + "step": 27688 + }, + { + "epoch": 1.01, + "learning_rate": 9.878419452887538e-08, + "loss": 0.6884, + "step": 27689 + }, + { + "epoch": 1.01, + "learning_rate": 9.878053246420332e-08, + "loss": 0.6786, + "step": 27690 + }, + { + "epoch": 1.01, + "learning_rate": 9.877687039953126e-08, + "loss": 0.7099, + "step": 27691 + }, + { + "epoch": 1.01, + "learning_rate": 9.87732083348592e-08, + "loss": 0.6887, + "step": 27692 + }, + { + "epoch": 1.01, + "learning_rate": 9.876954627018714e-08, + "loss": 0.689, + "step": 27693 + }, + { + "epoch": 1.01, + "learning_rate": 9.876588420551506e-08, + "loss": 0.6872, + "step": 27694 + }, + { + "epoch": 1.01, + "learning_rate": 9.8762222140843e-08, + "loss": 0.6869, + "step": 27695 + }, + { + "epoch": 1.01, + "learning_rate": 9.875856007617094e-08, + "loss": 0.6999, + "step": 27696 + }, + { + "epoch": 1.01, + "learning_rate": 9.875489801149888e-08, + "loss": 0.6927, + "step": 27697 + }, + { + "epoch": 1.01, + "learning_rate": 9.875123594682682e-08, + "loss": 0.6975, + "step": 27698 + }, + { + "epoch": 1.01, + "learning_rate": 9.874757388215475e-08, + "loss": 0.6797, + "step": 27699 + }, + { + "epoch": 1.01, + "learning_rate": 9.874391181748269e-08, + "loss": 0.7074, + "step": 27700 + }, + { + "epoch": 1.01, + "learning_rate": 9.874024975281063e-08, + "loss": 0.6865, + "step": 27701 + }, + { + "epoch": 1.01, + "learning_rate": 9.873658768813857e-08, + "loss": 0.6927, + "step": 27702 + }, + { + "epoch": 1.01, + "learning_rate": 9.87329256234665e-08, + "loss": 0.6926, + "step": 27703 + }, + { + "epoch": 1.01, + "learning_rate": 9.872926355879444e-08, + "loss": 0.6807, + "step": 27704 + }, + { + "epoch": 1.01, + "learning_rate": 9.872560149412238e-08, + "loss": 0.7046, + "step": 27705 + }, + { + "epoch": 1.01, + "learning_rate": 9.872193942945032e-08, + "loss": 0.6923, + "step": 27706 + }, + { + "epoch": 1.01, + "learning_rate": 9.871827736477826e-08, + "loss": 0.7005, + "step": 27707 + }, + { + "epoch": 1.01, + "learning_rate": 9.87146153001062e-08, + "loss": 0.69, + "step": 27708 + }, + { + "epoch": 1.01, + "learning_rate": 9.871095323543414e-08, + "loss": 0.6827, + "step": 27709 + }, + { + "epoch": 1.01, + "learning_rate": 9.870729117076208e-08, + "loss": 0.6932, + "step": 27710 + }, + { + "epoch": 1.01, + "learning_rate": 9.870362910609002e-08, + "loss": 0.7115, + "step": 27711 + }, + { + "epoch": 1.01, + "learning_rate": 9.869996704141794e-08, + "loss": 0.6846, + "step": 27712 + }, + { + "epoch": 1.01, + "learning_rate": 9.869630497674588e-08, + "loss": 0.6836, + "step": 27713 + }, + { + "epoch": 1.01, + "learning_rate": 9.869264291207382e-08, + "loss": 0.7016, + "step": 27714 + }, + { + "epoch": 1.01, + "learning_rate": 9.868898084740176e-08, + "loss": 0.6845, + "step": 27715 + }, + { + "epoch": 1.01, + "learning_rate": 9.86853187827297e-08, + "loss": 0.7019, + "step": 27716 + }, + { + "epoch": 1.01, + "learning_rate": 9.868165671805764e-08, + "loss": 0.6935, + "step": 27717 + }, + { + "epoch": 1.01, + "learning_rate": 9.867799465338558e-08, + "loss": 0.7, + "step": 27718 + }, + { + "epoch": 1.01, + "learning_rate": 9.867433258871352e-08, + "loss": 0.704, + "step": 27719 + }, + { + "epoch": 1.01, + "learning_rate": 9.867067052404145e-08, + "loss": 0.6997, + "step": 27720 + }, + { + "epoch": 1.01, + "learning_rate": 9.866700845936939e-08, + "loss": 0.6946, + "step": 27721 + }, + { + "epoch": 1.01, + "learning_rate": 9.866334639469733e-08, + "loss": 0.6881, + "step": 27722 + }, + { + "epoch": 1.01, + "learning_rate": 9.865968433002525e-08, + "loss": 0.6834, + "step": 27723 + }, + { + "epoch": 1.01, + "learning_rate": 9.86560222653532e-08, + "loss": 0.6914, + "step": 27724 + }, + { + "epoch": 1.01, + "learning_rate": 9.865236020068113e-08, + "loss": 0.6959, + "step": 27725 + }, + { + "epoch": 1.01, + "learning_rate": 9.864869813600907e-08, + "loss": 0.6964, + "step": 27726 + }, + { + "epoch": 1.01, + "learning_rate": 9.864503607133701e-08, + "loss": 0.6866, + "step": 27727 + }, + { + "epoch": 1.01, + "learning_rate": 9.864137400666495e-08, + "loss": 0.6916, + "step": 27728 + }, + { + "epoch": 1.01, + "learning_rate": 9.86377119419929e-08, + "loss": 0.6837, + "step": 27729 + }, + { + "epoch": 1.01, + "learning_rate": 9.863404987732082e-08, + "loss": 0.6902, + "step": 27730 + }, + { + "epoch": 1.01, + "learning_rate": 9.863038781264876e-08, + "loss": 0.6942, + "step": 27731 + }, + { + "epoch": 1.01, + "learning_rate": 9.86267257479767e-08, + "loss": 0.683, + "step": 27732 + }, + { + "epoch": 1.01, + "learning_rate": 9.862306368330464e-08, + "loss": 0.6848, + "step": 27733 + }, + { + "epoch": 1.01, + "learning_rate": 9.861940161863258e-08, + "loss": 0.6991, + "step": 27734 + }, + { + "epoch": 1.01, + "learning_rate": 9.861573955396052e-08, + "loss": 0.6717, + "step": 27735 + }, + { + "epoch": 1.01, + "learning_rate": 9.861207748928846e-08, + "loss": 0.698, + "step": 27736 + }, + { + "epoch": 1.01, + "learning_rate": 9.86084154246164e-08, + "loss": 0.7031, + "step": 27737 + }, + { + "epoch": 1.01, + "learning_rate": 9.860475335994434e-08, + "loss": 0.6865, + "step": 27738 + }, + { + "epoch": 1.01, + "learning_rate": 9.860109129527227e-08, + "loss": 0.7113, + "step": 27739 + }, + { + "epoch": 1.01, + "learning_rate": 9.859742923060021e-08, + "loss": 0.7059, + "step": 27740 + }, + { + "epoch": 1.01, + "learning_rate": 9.859376716592815e-08, + "loss": 0.6947, + "step": 27741 + }, + { + "epoch": 1.02, + "learning_rate": 9.859010510125609e-08, + "loss": 0.6982, + "step": 27742 + }, + { + "epoch": 1.02, + "learning_rate": 9.858644303658403e-08, + "loss": 0.6963, + "step": 27743 + }, + { + "epoch": 1.02, + "learning_rate": 9.858278097191197e-08, + "loss": 0.6931, + "step": 27744 + }, + { + "epoch": 1.02, + "learning_rate": 9.85791189072399e-08, + "loss": 0.6849, + "step": 27745 + }, + { + "epoch": 1.02, + "learning_rate": 9.857545684256783e-08, + "loss": 0.6926, + "step": 27746 + }, + { + "epoch": 1.02, + "learning_rate": 9.857179477789577e-08, + "loss": 0.704, + "step": 27747 + }, + { + "epoch": 1.02, + "learning_rate": 9.85681327132237e-08, + "loss": 0.6992, + "step": 27748 + }, + { + "epoch": 1.02, + "learning_rate": 9.856447064855164e-08, + "loss": 0.7044, + "step": 27749 + }, + { + "epoch": 1.02, + "learning_rate": 9.856080858387958e-08, + "loss": 0.6988, + "step": 27750 + }, + { + "epoch": 1.02, + "learning_rate": 9.855714651920752e-08, + "loss": 0.6839, + "step": 27751 + }, + { + "epoch": 1.02, + "learning_rate": 9.855348445453546e-08, + "loss": 0.6941, + "step": 27752 + }, + { + "epoch": 1.02, + "learning_rate": 9.85498223898634e-08, + "loss": 0.6934, + "step": 27753 + }, + { + "epoch": 1.02, + "learning_rate": 9.854616032519134e-08, + "loss": 0.7006, + "step": 27754 + }, + { + "epoch": 1.02, + "learning_rate": 9.854249826051928e-08, + "loss": 0.6961, + "step": 27755 + }, + { + "epoch": 1.02, + "learning_rate": 9.853883619584722e-08, + "loss": 0.6883, + "step": 27756 + }, + { + "epoch": 1.02, + "learning_rate": 9.853517413117515e-08, + "loss": 0.7046, + "step": 27757 + }, + { + "epoch": 1.02, + "learning_rate": 9.853151206650309e-08, + "loss": 0.7105, + "step": 27758 + }, + { + "epoch": 1.02, + "learning_rate": 9.852785000183103e-08, + "loss": 0.6902, + "step": 27759 + }, + { + "epoch": 1.02, + "learning_rate": 9.852418793715897e-08, + "loss": 0.692, + "step": 27760 + }, + { + "epoch": 1.02, + "learning_rate": 9.85205258724869e-08, + "loss": 0.7084, + "step": 27761 + }, + { + "epoch": 1.02, + "learning_rate": 9.851686380781485e-08, + "loss": 0.7018, + "step": 27762 + }, + { + "epoch": 1.02, + "learning_rate": 9.851320174314279e-08, + "loss": 0.6985, + "step": 27763 + }, + { + "epoch": 1.02, + "learning_rate": 9.850953967847073e-08, + "loss": 0.6964, + "step": 27764 + }, + { + "epoch": 1.02, + "learning_rate": 9.850587761379867e-08, + "loss": 0.696, + "step": 27765 + }, + { + "epoch": 1.02, + "learning_rate": 9.850221554912659e-08, + "loss": 0.6979, + "step": 27766 + }, + { + "epoch": 1.02, + "learning_rate": 9.849855348445453e-08, + "loss": 0.6982, + "step": 27767 + }, + { + "epoch": 1.02, + "learning_rate": 9.849489141978247e-08, + "loss": 0.6947, + "step": 27768 + }, + { + "epoch": 1.02, + "learning_rate": 9.84912293551104e-08, + "loss": 0.7082, + "step": 27769 + }, + { + "epoch": 1.02, + "learning_rate": 9.848756729043834e-08, + "loss": 0.7048, + "step": 27770 + }, + { + "epoch": 1.02, + "learning_rate": 9.848390522576628e-08, + "loss": 0.6964, + "step": 27771 + }, + { + "epoch": 1.02, + "learning_rate": 9.848024316109422e-08, + "loss": 0.6903, + "step": 27772 + }, + { + "epoch": 1.02, + "learning_rate": 9.847658109642216e-08, + "loss": 0.699, + "step": 27773 + }, + { + "epoch": 1.02, + "learning_rate": 9.84729190317501e-08, + "loss": 0.6895, + "step": 27774 + }, + { + "epoch": 1.02, + "learning_rate": 9.846925696707803e-08, + "loss": 0.6938, + "step": 27775 + }, + { + "epoch": 1.02, + "learning_rate": 9.846559490240597e-08, + "loss": 0.6911, + "step": 27776 + }, + { + "epoch": 1.02, + "learning_rate": 9.84619328377339e-08, + "loss": 0.693, + "step": 27777 + }, + { + "epoch": 1.02, + "learning_rate": 9.845827077306185e-08, + "loss": 0.6858, + "step": 27778 + }, + { + "epoch": 1.02, + "learning_rate": 9.845460870838979e-08, + "loss": 0.6945, + "step": 27779 + }, + { + "epoch": 1.02, + "learning_rate": 9.845094664371773e-08, + "loss": 0.679, + "step": 27780 + }, + { + "epoch": 1.02, + "learning_rate": 9.844728457904566e-08, + "loss": 0.7019, + "step": 27781 + }, + { + "epoch": 1.02, + "learning_rate": 9.84436225143736e-08, + "loss": 0.6975, + "step": 27782 + }, + { + "epoch": 1.02, + "learning_rate": 9.843996044970154e-08, + "loss": 0.6851, + "step": 27783 + }, + { + "epoch": 1.02, + "learning_rate": 9.843629838502947e-08, + "loss": 0.7059, + "step": 27784 + }, + { + "epoch": 1.02, + "learning_rate": 9.843263632035741e-08, + "loss": 0.7114, + "step": 27785 + }, + { + "epoch": 1.02, + "learning_rate": 9.842897425568535e-08, + "loss": 0.7, + "step": 27786 + }, + { + "epoch": 1.02, + "learning_rate": 9.842531219101329e-08, + "loss": 0.7084, + "step": 27787 + }, + { + "epoch": 1.02, + "learning_rate": 9.842165012634123e-08, + "loss": 0.68, + "step": 27788 + }, + { + "epoch": 1.02, + "learning_rate": 9.841798806166917e-08, + "loss": 0.6862, + "step": 27789 + }, + { + "epoch": 1.02, + "learning_rate": 9.841432599699711e-08, + "loss": 0.6857, + "step": 27790 + }, + { + "epoch": 1.02, + "learning_rate": 9.841066393232504e-08, + "loss": 0.694, + "step": 27791 + }, + { + "epoch": 1.02, + "learning_rate": 9.840700186765298e-08, + "loss": 0.69, + "step": 27792 + }, + { + "epoch": 1.02, + "learning_rate": 9.840333980298092e-08, + "loss": 0.6816, + "step": 27793 + }, + { + "epoch": 1.02, + "learning_rate": 9.839967773830884e-08, + "loss": 0.6981, + "step": 27794 + }, + { + "epoch": 1.02, + "learning_rate": 9.839601567363678e-08, + "loss": 0.6886, + "step": 27795 + }, + { + "epoch": 1.02, + "learning_rate": 9.839235360896472e-08, + "loss": 0.7004, + "step": 27796 + }, + { + "epoch": 1.02, + "learning_rate": 9.838869154429266e-08, + "loss": 0.6888, + "step": 27797 + }, + { + "epoch": 1.02, + "learning_rate": 9.83850294796206e-08, + "loss": 0.6916, + "step": 27798 + }, + { + "epoch": 1.02, + "learning_rate": 9.838136741494854e-08, + "loss": 0.6895, + "step": 27799 + }, + { + "epoch": 1.02, + "learning_rate": 9.837770535027648e-08, + "loss": 0.7039, + "step": 27800 + }, + { + "epoch": 1.02, + "learning_rate": 9.837404328560442e-08, + "loss": 0.6796, + "step": 27801 + }, + { + "epoch": 1.02, + "learning_rate": 9.837038122093235e-08, + "loss": 0.6871, + "step": 27802 + }, + { + "epoch": 1.02, + "learning_rate": 9.836671915626029e-08, + "loss": 0.6907, + "step": 27803 + }, + { + "epoch": 1.02, + "learning_rate": 9.836305709158823e-08, + "loss": 0.6819, + "step": 27804 + }, + { + "epoch": 1.02, + "learning_rate": 9.835939502691617e-08, + "loss": 0.6913, + "step": 27805 + }, + { + "epoch": 1.02, + "learning_rate": 9.835573296224411e-08, + "loss": 0.6837, + "step": 27806 + }, + { + "epoch": 1.02, + "learning_rate": 9.835207089757205e-08, + "loss": 0.6958, + "step": 27807 + }, + { + "epoch": 1.02, + "learning_rate": 9.834840883289999e-08, + "loss": 0.6956, + "step": 27808 + }, + { + "epoch": 1.02, + "learning_rate": 9.834474676822793e-08, + "loss": 0.6874, + "step": 27809 + }, + { + "epoch": 1.02, + "learning_rate": 9.834108470355587e-08, + "loss": 0.6954, + "step": 27810 + }, + { + "epoch": 1.02, + "learning_rate": 9.83374226388838e-08, + "loss": 0.6958, + "step": 27811 + }, + { + "epoch": 1.02, + "learning_rate": 9.833376057421174e-08, + "loss": 0.6939, + "step": 27812 + }, + { + "epoch": 1.02, + "learning_rate": 9.833009850953968e-08, + "loss": 0.6978, + "step": 27813 + }, + { + "epoch": 1.02, + "learning_rate": 9.832643644486762e-08, + "loss": 0.6986, + "step": 27814 + }, + { + "epoch": 1.02, + "learning_rate": 9.832277438019556e-08, + "loss": 0.6957, + "step": 27815 + }, + { + "epoch": 1.02, + "learning_rate": 9.831911231552348e-08, + "loss": 0.6952, + "step": 27816 + }, + { + "epoch": 1.02, + "learning_rate": 9.831545025085142e-08, + "loss": 0.6968, + "step": 27817 + }, + { + "epoch": 1.02, + "learning_rate": 9.831178818617936e-08, + "loss": 0.6869, + "step": 27818 + }, + { + "epoch": 1.02, + "learning_rate": 9.83081261215073e-08, + "loss": 0.6878, + "step": 27819 + }, + { + "epoch": 1.02, + "learning_rate": 9.830446405683523e-08, + "loss": 0.6869, + "step": 27820 + }, + { + "epoch": 1.02, + "learning_rate": 9.830080199216317e-08, + "loss": 0.6984, + "step": 27821 + }, + { + "epoch": 1.02, + "learning_rate": 9.829713992749111e-08, + "loss": 0.6924, + "step": 27822 + }, + { + "epoch": 1.02, + "learning_rate": 9.829347786281905e-08, + "loss": 0.6929, + "step": 27823 + }, + { + "epoch": 1.02, + "learning_rate": 9.828981579814699e-08, + "loss": 0.6899, + "step": 27824 + }, + { + "epoch": 1.02, + "learning_rate": 9.828615373347493e-08, + "loss": 0.6888, + "step": 27825 + }, + { + "epoch": 1.02, + "learning_rate": 9.828249166880287e-08, + "loss": 0.6952, + "step": 27826 + }, + { + "epoch": 1.02, + "learning_rate": 9.827882960413081e-08, + "loss": 0.6967, + "step": 27827 + }, + { + "epoch": 1.02, + "learning_rate": 9.827516753945875e-08, + "loss": 0.7062, + "step": 27828 + }, + { + "epoch": 1.02, + "learning_rate": 9.827150547478668e-08, + "loss": 0.6896, + "step": 27829 + }, + { + "epoch": 1.02, + "learning_rate": 9.826784341011462e-08, + "loss": 0.6938, + "step": 27830 + }, + { + "epoch": 1.02, + "learning_rate": 9.826418134544256e-08, + "loss": 0.7032, + "step": 27831 + }, + { + "epoch": 1.02, + "learning_rate": 9.82605192807705e-08, + "loss": 0.7023, + "step": 27832 + }, + { + "epoch": 1.02, + "learning_rate": 9.825685721609844e-08, + "loss": 0.6899, + "step": 27833 + }, + { + "epoch": 1.02, + "learning_rate": 9.825319515142638e-08, + "loss": 0.695, + "step": 27834 + }, + { + "epoch": 1.02, + "learning_rate": 9.824953308675432e-08, + "loss": 0.6939, + "step": 27835 + }, + { + "epoch": 1.02, + "learning_rate": 9.824587102208226e-08, + "loss": 0.6994, + "step": 27836 + }, + { + "epoch": 1.02, + "learning_rate": 9.82422089574102e-08, + "loss": 0.6805, + "step": 27837 + }, + { + "epoch": 1.02, + "learning_rate": 9.823854689273812e-08, + "loss": 0.6909, + "step": 27838 + }, + { + "epoch": 1.02, + "learning_rate": 9.823488482806606e-08, + "loss": 0.7002, + "step": 27839 + }, + { + "epoch": 1.02, + "learning_rate": 9.823122276339399e-08, + "loss": 0.6936, + "step": 27840 + }, + { + "epoch": 1.02, + "learning_rate": 9.822756069872193e-08, + "loss": 0.6923, + "step": 27841 + }, + { + "epoch": 1.02, + "learning_rate": 9.822389863404987e-08, + "loss": 0.69, + "step": 27842 + }, + { + "epoch": 1.02, + "learning_rate": 9.822023656937781e-08, + "loss": 0.6876, + "step": 27843 + }, + { + "epoch": 1.02, + "learning_rate": 9.821657450470575e-08, + "loss": 0.6927, + "step": 27844 + }, + { + "epoch": 1.02, + "learning_rate": 9.821291244003369e-08, + "loss": 0.7285, + "step": 27845 + }, + { + "epoch": 1.02, + "learning_rate": 9.820925037536163e-08, + "loss": 0.696, + "step": 27846 + }, + { + "epoch": 1.02, + "learning_rate": 9.820558831068955e-08, + "loss": 0.6851, + "step": 27847 + }, + { + "epoch": 1.02, + "learning_rate": 9.82019262460175e-08, + "loss": 0.7016, + "step": 27848 + }, + { + "epoch": 1.02, + "learning_rate": 9.819826418134543e-08, + "loss": 0.6918, + "step": 27849 + }, + { + "epoch": 1.02, + "learning_rate": 9.819460211667337e-08, + "loss": 0.7022, + "step": 27850 + }, + { + "epoch": 1.02, + "learning_rate": 9.819094005200131e-08, + "loss": 0.6946, + "step": 27851 + }, + { + "epoch": 1.02, + "learning_rate": 9.818727798732925e-08, + "loss": 0.6872, + "step": 27852 + }, + { + "epoch": 1.02, + "learning_rate": 9.81836159226572e-08, + "loss": 0.7088, + "step": 27853 + }, + { + "epoch": 1.02, + "learning_rate": 9.817995385798513e-08, + "loss": 0.6921, + "step": 27854 + }, + { + "epoch": 1.02, + "learning_rate": 9.817629179331307e-08, + "loss": 0.6993, + "step": 27855 + }, + { + "epoch": 1.02, + "learning_rate": 9.8172629728641e-08, + "loss": 0.6933, + "step": 27856 + }, + { + "epoch": 1.02, + "learning_rate": 9.816896766396894e-08, + "loss": 0.699, + "step": 27857 + }, + { + "epoch": 1.02, + "learning_rate": 9.816530559929688e-08, + "loss": 0.6876, + "step": 27858 + }, + { + "epoch": 1.02, + "learning_rate": 9.816164353462482e-08, + "loss": 0.6998, + "step": 27859 + }, + { + "epoch": 1.02, + "learning_rate": 9.815798146995276e-08, + "loss": 0.683, + "step": 27860 + }, + { + "epoch": 1.02, + "learning_rate": 9.81543194052807e-08, + "loss": 0.6924, + "step": 27861 + }, + { + "epoch": 1.02, + "learning_rate": 9.815065734060863e-08, + "loss": 0.6933, + "step": 27862 + }, + { + "epoch": 1.02, + "learning_rate": 9.814699527593657e-08, + "loss": 0.6957, + "step": 27863 + }, + { + "epoch": 1.02, + "learning_rate": 9.814333321126451e-08, + "loss": 0.689, + "step": 27864 + }, + { + "epoch": 1.02, + "learning_rate": 9.813967114659243e-08, + "loss": 0.6993, + "step": 27865 + }, + { + "epoch": 1.02, + "learning_rate": 9.813600908192037e-08, + "loss": 0.6981, + "step": 27866 + }, + { + "epoch": 1.02, + "learning_rate": 9.813234701724831e-08, + "loss": 0.6883, + "step": 27867 + }, + { + "epoch": 1.02, + "learning_rate": 9.812868495257625e-08, + "loss": 0.7047, + "step": 27868 + }, + { + "epoch": 1.02, + "learning_rate": 9.812502288790419e-08, + "loss": 0.7012, + "step": 27869 + }, + { + "epoch": 1.02, + "learning_rate": 9.812136082323213e-08, + "loss": 0.691, + "step": 27870 + }, + { + "epoch": 1.02, + "learning_rate": 9.811769875856007e-08, + "loss": 0.7069, + "step": 27871 + }, + { + "epoch": 1.02, + "learning_rate": 9.811403669388801e-08, + "loss": 0.6926, + "step": 27872 + }, + { + "epoch": 1.02, + "learning_rate": 9.811037462921595e-08, + "loss": 0.6931, + "step": 27873 + }, + { + "epoch": 1.02, + "learning_rate": 9.810671256454388e-08, + "loss": 0.6889, + "step": 27874 + }, + { + "epoch": 1.02, + "learning_rate": 9.810305049987182e-08, + "loss": 0.6813, + "step": 27875 + }, + { + "epoch": 1.02, + "learning_rate": 9.809938843519976e-08, + "loss": 0.6946, + "step": 27876 + }, + { + "epoch": 1.02, + "learning_rate": 9.80957263705277e-08, + "loss": 0.6998, + "step": 27877 + }, + { + "epoch": 1.02, + "learning_rate": 9.809206430585564e-08, + "loss": 0.698, + "step": 27878 + }, + { + "epoch": 1.02, + "learning_rate": 9.808840224118358e-08, + "loss": 0.6765, + "step": 27879 + }, + { + "epoch": 1.02, + "learning_rate": 9.808474017651152e-08, + "loss": 0.6855, + "step": 27880 + }, + { + "epoch": 1.02, + "learning_rate": 9.808107811183946e-08, + "loss": 0.7016, + "step": 27881 + }, + { + "epoch": 1.02, + "learning_rate": 9.80774160471674e-08, + "loss": 0.6942, + "step": 27882 + }, + { + "epoch": 1.02, + "learning_rate": 9.807375398249533e-08, + "loss": 0.6848, + "step": 27883 + }, + { + "epoch": 1.02, + "learning_rate": 9.807009191782327e-08, + "loss": 0.6837, + "step": 27884 + }, + { + "epoch": 1.02, + "learning_rate": 9.80664298531512e-08, + "loss": 0.689, + "step": 27885 + }, + { + "epoch": 1.02, + "learning_rate": 9.806276778847915e-08, + "loss": 0.6994, + "step": 27886 + }, + { + "epoch": 1.02, + "learning_rate": 9.805910572380707e-08, + "loss": 0.6848, + "step": 27887 + }, + { + "epoch": 1.02, + "learning_rate": 9.805544365913501e-08, + "loss": 0.691, + "step": 27888 + }, + { + "epoch": 1.02, + "learning_rate": 9.805178159446295e-08, + "loss": 0.6945, + "step": 27889 + }, + { + "epoch": 1.02, + "learning_rate": 9.804811952979089e-08, + "loss": 0.7007, + "step": 27890 + }, + { + "epoch": 1.02, + "learning_rate": 9.804445746511883e-08, + "loss": 0.6908, + "step": 27891 + }, + { + "epoch": 1.02, + "learning_rate": 9.804079540044676e-08, + "loss": 0.7036, + "step": 27892 + }, + { + "epoch": 1.02, + "learning_rate": 9.80371333357747e-08, + "loss": 0.6785, + "step": 27893 + }, + { + "epoch": 1.02, + "learning_rate": 9.803347127110264e-08, + "loss": 0.6881, + "step": 27894 + }, + { + "epoch": 1.02, + "learning_rate": 9.802980920643058e-08, + "loss": 0.6859, + "step": 27895 + }, + { + "epoch": 1.02, + "learning_rate": 9.802614714175852e-08, + "loss": 0.6806, + "step": 27896 + }, + { + "epoch": 1.02, + "learning_rate": 9.802248507708646e-08, + "loss": 0.7065, + "step": 27897 + }, + { + "epoch": 1.02, + "learning_rate": 9.80188230124144e-08, + "loss": 0.6902, + "step": 27898 + }, + { + "epoch": 1.02, + "learning_rate": 9.801516094774234e-08, + "loss": 0.6895, + "step": 27899 + }, + { + "epoch": 1.02, + "learning_rate": 9.801149888307028e-08, + "loss": 0.6994, + "step": 27900 + }, + { + "epoch": 1.02, + "learning_rate": 9.80078368183982e-08, + "loss": 0.6898, + "step": 27901 + }, + { + "epoch": 1.02, + "learning_rate": 9.800417475372614e-08, + "loss": 0.6936, + "step": 27902 + }, + { + "epoch": 1.02, + "learning_rate": 9.800051268905408e-08, + "loss": 0.6823, + "step": 27903 + }, + { + "epoch": 1.02, + "learning_rate": 9.799685062438202e-08, + "loss": 0.6848, + "step": 27904 + }, + { + "epoch": 1.02, + "learning_rate": 9.799318855970996e-08, + "loss": 0.699, + "step": 27905 + }, + { + "epoch": 1.02, + "learning_rate": 9.79895264950379e-08, + "loss": 0.6832, + "step": 27906 + }, + { + "epoch": 1.02, + "learning_rate": 9.798586443036584e-08, + "loss": 0.6956, + "step": 27907 + }, + { + "epoch": 1.02, + "learning_rate": 9.798220236569378e-08, + "loss": 0.6788, + "step": 27908 + }, + { + "epoch": 1.02, + "learning_rate": 9.797854030102171e-08, + "loss": 0.6887, + "step": 27909 + }, + { + "epoch": 1.02, + "learning_rate": 9.797487823634965e-08, + "loss": 0.6924, + "step": 27910 + }, + { + "epoch": 1.02, + "learning_rate": 9.797121617167758e-08, + "loss": 0.6862, + "step": 27911 + }, + { + "epoch": 1.02, + "learning_rate": 9.796755410700552e-08, + "loss": 0.6872, + "step": 27912 + }, + { + "epoch": 1.02, + "learning_rate": 9.796389204233346e-08, + "loss": 0.6776, + "step": 27913 + }, + { + "epoch": 1.02, + "learning_rate": 9.79602299776614e-08, + "loss": 0.6953, + "step": 27914 + }, + { + "epoch": 1.02, + "learning_rate": 9.795656791298934e-08, + "loss": 0.7025, + "step": 27915 + }, + { + "epoch": 1.02, + "learning_rate": 9.795290584831728e-08, + "loss": 0.6878, + "step": 27916 + }, + { + "epoch": 1.02, + "learning_rate": 9.794924378364522e-08, + "loss": 0.7018, + "step": 27917 + }, + { + "epoch": 1.02, + "learning_rate": 9.794558171897316e-08, + "loss": 0.6925, + "step": 27918 + }, + { + "epoch": 1.02, + "learning_rate": 9.794191965430108e-08, + "loss": 0.6842, + "step": 27919 + }, + { + "epoch": 1.02, + "learning_rate": 9.793825758962902e-08, + "loss": 0.6931, + "step": 27920 + }, + { + "epoch": 1.02, + "learning_rate": 9.793459552495696e-08, + "loss": 0.685, + "step": 27921 + }, + { + "epoch": 1.02, + "learning_rate": 9.79309334602849e-08, + "loss": 0.7, + "step": 27922 + }, + { + "epoch": 1.02, + "learning_rate": 9.792727139561284e-08, + "loss": 0.6941, + "step": 27923 + }, + { + "epoch": 1.02, + "learning_rate": 9.792360933094078e-08, + "loss": 0.7018, + "step": 27924 + }, + { + "epoch": 1.02, + "learning_rate": 9.791994726626872e-08, + "loss": 0.6867, + "step": 27925 + }, + { + "epoch": 1.02, + "learning_rate": 9.791628520159666e-08, + "loss": 0.6821, + "step": 27926 + }, + { + "epoch": 1.02, + "learning_rate": 9.79126231369246e-08, + "loss": 0.6864, + "step": 27927 + }, + { + "epoch": 1.02, + "learning_rate": 9.790896107225253e-08, + "loss": 0.6893, + "step": 27928 + }, + { + "epoch": 1.02, + "learning_rate": 9.790529900758047e-08, + "loss": 0.6925, + "step": 27929 + }, + { + "epoch": 1.02, + "learning_rate": 9.790163694290841e-08, + "loss": 0.6967, + "step": 27930 + }, + { + "epoch": 1.02, + "learning_rate": 9.789797487823635e-08, + "loss": 0.679, + "step": 27931 + }, + { + "epoch": 1.02, + "learning_rate": 9.789431281356429e-08, + "loss": 0.6926, + "step": 27932 + }, + { + "epoch": 1.02, + "learning_rate": 9.789065074889222e-08, + "loss": 0.6981, + "step": 27933 + }, + { + "epoch": 1.02, + "learning_rate": 9.788698868422016e-08, + "loss": 0.6986, + "step": 27934 + }, + { + "epoch": 1.02, + "learning_rate": 9.78833266195481e-08, + "loss": 0.6946, + "step": 27935 + }, + { + "epoch": 1.02, + "learning_rate": 9.787966455487602e-08, + "loss": 0.6884, + "step": 27936 + }, + { + "epoch": 1.02, + "learning_rate": 9.787600249020396e-08, + "loss": 0.6899, + "step": 27937 + }, + { + "epoch": 1.02, + "learning_rate": 9.78723404255319e-08, + "loss": 0.691, + "step": 27938 + }, + { + "epoch": 1.02, + "learning_rate": 9.786867836085984e-08, + "loss": 0.6986, + "step": 27939 + }, + { + "epoch": 1.02, + "learning_rate": 9.786501629618778e-08, + "loss": 0.6851, + "step": 27940 + }, + { + "epoch": 1.02, + "learning_rate": 9.786135423151572e-08, + "loss": 0.6871, + "step": 27941 + }, + { + "epoch": 1.02, + "learning_rate": 9.785769216684366e-08, + "loss": 0.6847, + "step": 27942 + }, + { + "epoch": 1.02, + "learning_rate": 9.78540301021716e-08, + "loss": 0.7017, + "step": 27943 + }, + { + "epoch": 1.02, + "learning_rate": 9.785036803749954e-08, + "loss": 0.695, + "step": 27944 + }, + { + "epoch": 1.02, + "learning_rate": 9.784670597282747e-08, + "loss": 0.6973, + "step": 27945 + }, + { + "epoch": 1.02, + "learning_rate": 9.784304390815541e-08, + "loss": 0.69, + "step": 27946 + }, + { + "epoch": 1.02, + "learning_rate": 9.783938184348335e-08, + "loss": 0.6962, + "step": 27947 + }, + { + "epoch": 1.02, + "learning_rate": 9.783571977881129e-08, + "loss": 0.6927, + "step": 27948 + }, + { + "epoch": 1.02, + "learning_rate": 9.783205771413923e-08, + "loss": 0.7076, + "step": 27949 + }, + { + "epoch": 1.02, + "learning_rate": 9.782839564946717e-08, + "loss": 0.6862, + "step": 27950 + }, + { + "epoch": 1.02, + "learning_rate": 9.782473358479511e-08, + "loss": 0.6922, + "step": 27951 + }, + { + "epoch": 1.02, + "learning_rate": 9.782107152012305e-08, + "loss": 0.6929, + "step": 27952 + }, + { + "epoch": 1.02, + "learning_rate": 9.781740945545099e-08, + "loss": 0.7078, + "step": 27953 + }, + { + "epoch": 1.02, + "learning_rate": 9.781374739077892e-08, + "loss": 0.6912, + "step": 27954 + }, + { + "epoch": 1.02, + "learning_rate": 9.781008532610686e-08, + "loss": 0.6838, + "step": 27955 + }, + { + "epoch": 1.02, + "learning_rate": 9.78064232614348e-08, + "loss": 0.6907, + "step": 27956 + }, + { + "epoch": 1.02, + "learning_rate": 9.780276119676274e-08, + "loss": 0.695, + "step": 27957 + }, + { + "epoch": 1.02, + "learning_rate": 9.779909913209066e-08, + "loss": 0.6907, + "step": 27958 + }, + { + "epoch": 1.02, + "learning_rate": 9.77954370674186e-08, + "loss": 0.6973, + "step": 27959 + }, + { + "epoch": 1.02, + "learning_rate": 9.779177500274654e-08, + "loss": 0.6976, + "step": 27960 + }, + { + "epoch": 1.02, + "learning_rate": 9.778811293807448e-08, + "loss": 0.6931, + "step": 27961 + }, + { + "epoch": 1.02, + "learning_rate": 9.778445087340242e-08, + "loss": 0.6884, + "step": 27962 + }, + { + "epoch": 1.02, + "learning_rate": 9.778078880873035e-08, + "loss": 0.6921, + "step": 27963 + }, + { + "epoch": 1.02, + "learning_rate": 9.777712674405829e-08, + "loss": 0.6942, + "step": 27964 + }, + { + "epoch": 1.02, + "learning_rate": 9.777346467938623e-08, + "loss": 0.6972, + "step": 27965 + }, + { + "epoch": 1.02, + "learning_rate": 9.776980261471417e-08, + "loss": 0.6895, + "step": 27966 + }, + { + "epoch": 1.02, + "learning_rate": 9.776614055004211e-08, + "loss": 0.6913, + "step": 27967 + }, + { + "epoch": 1.02, + "learning_rate": 9.776247848537005e-08, + "loss": 0.6942, + "step": 27968 + }, + { + "epoch": 1.02, + "learning_rate": 9.775881642069799e-08, + "loss": 0.6955, + "step": 27969 + }, + { + "epoch": 1.02, + "learning_rate": 9.775515435602593e-08, + "loss": 0.7081, + "step": 27970 + }, + { + "epoch": 1.02, + "learning_rate": 9.775149229135387e-08, + "loss": 0.7031, + "step": 27971 + }, + { + "epoch": 1.02, + "learning_rate": 9.77478302266818e-08, + "loss": 0.7029, + "step": 27972 + }, + { + "epoch": 1.02, + "learning_rate": 9.774416816200973e-08, + "loss": 0.6962, + "step": 27973 + }, + { + "epoch": 1.02, + "learning_rate": 9.774050609733767e-08, + "loss": 0.6954, + "step": 27974 + }, + { + "epoch": 1.02, + "learning_rate": 9.773684403266561e-08, + "loss": 0.6887, + "step": 27975 + }, + { + "epoch": 1.02, + "learning_rate": 9.773318196799355e-08, + "loss": 0.6822, + "step": 27976 + }, + { + "epoch": 1.02, + "learning_rate": 9.77295199033215e-08, + "loss": 0.6859, + "step": 27977 + }, + { + "epoch": 1.02, + "learning_rate": 9.772585783864943e-08, + "loss": 0.7047, + "step": 27978 + }, + { + "epoch": 1.02, + "learning_rate": 9.772219577397737e-08, + "loss": 0.7019, + "step": 27979 + }, + { + "epoch": 1.02, + "learning_rate": 9.77185337093053e-08, + "loss": 0.6982, + "step": 27980 + }, + { + "epoch": 1.02, + "learning_rate": 9.771487164463324e-08, + "loss": 0.6992, + "step": 27981 + }, + { + "epoch": 1.02, + "learning_rate": 9.771120957996117e-08, + "loss": 0.6929, + "step": 27982 + }, + { + "epoch": 1.02, + "learning_rate": 9.770754751528911e-08, + "loss": 0.6832, + "step": 27983 + }, + { + "epoch": 1.02, + "learning_rate": 9.770388545061705e-08, + "loss": 0.6937, + "step": 27984 + }, + { + "epoch": 1.02, + "learning_rate": 9.770022338594499e-08, + "loss": 0.6976, + "step": 27985 + }, + { + "epoch": 1.02, + "learning_rate": 9.769656132127293e-08, + "loss": 0.6907, + "step": 27986 + }, + { + "epoch": 1.02, + "learning_rate": 9.769289925660087e-08, + "loss": 0.7064, + "step": 27987 + }, + { + "epoch": 1.02, + "learning_rate": 9.768923719192881e-08, + "loss": 0.6951, + "step": 27988 + }, + { + "epoch": 1.02, + "learning_rate": 9.768557512725675e-08, + "loss": 0.7017, + "step": 27989 + }, + { + "epoch": 1.02, + "learning_rate": 9.768191306258467e-08, + "loss": 0.6972, + "step": 27990 + }, + { + "epoch": 1.02, + "learning_rate": 9.767825099791261e-08, + "loss": 0.696, + "step": 27991 + }, + { + "epoch": 1.02, + "learning_rate": 9.767458893324055e-08, + "loss": 0.6945, + "step": 27992 + }, + { + "epoch": 1.02, + "learning_rate": 9.767092686856849e-08, + "loss": 0.698, + "step": 27993 + }, + { + "epoch": 1.02, + "learning_rate": 9.766726480389643e-08, + "loss": 0.7016, + "step": 27994 + }, + { + "epoch": 1.02, + "learning_rate": 9.766360273922437e-08, + "loss": 0.6915, + "step": 27995 + }, + { + "epoch": 1.02, + "learning_rate": 9.765994067455231e-08, + "loss": 0.6933, + "step": 27996 + }, + { + "epoch": 1.02, + "learning_rate": 9.765627860988025e-08, + "loss": 0.6935, + "step": 27997 + }, + { + "epoch": 1.02, + "learning_rate": 9.765261654520819e-08, + "loss": 0.6936, + "step": 27998 + }, + { + "epoch": 1.02, + "learning_rate": 9.764895448053612e-08, + "loss": 0.696, + "step": 27999 + }, + { + "epoch": 1.02, + "learning_rate": 9.764529241586406e-08, + "loss": 0.6983, + "step": 28000 + }, + { + "epoch": 1.02, + "learning_rate": 9.7641630351192e-08, + "loss": 0.6948, + "step": 28001 + }, + { + "epoch": 1.02, + "learning_rate": 9.763796828651994e-08, + "loss": 0.7063, + "step": 28002 + }, + { + "epoch": 1.02, + "learning_rate": 9.763430622184788e-08, + "loss": 0.7008, + "step": 28003 + }, + { + "epoch": 1.02, + "learning_rate": 9.76306441571758e-08, + "loss": 0.6942, + "step": 28004 + }, + { + "epoch": 1.02, + "learning_rate": 9.762698209250375e-08, + "loss": 0.6977, + "step": 28005 + }, + { + "epoch": 1.02, + "learning_rate": 9.762332002783169e-08, + "loss": 0.6888, + "step": 28006 + }, + { + "epoch": 1.02, + "learning_rate": 9.761965796315963e-08, + "loss": 0.6839, + "step": 28007 + }, + { + "epoch": 1.02, + "learning_rate": 9.761599589848755e-08, + "loss": 0.6869, + "step": 28008 + }, + { + "epoch": 1.02, + "learning_rate": 9.761233383381549e-08, + "loss": 0.7032, + "step": 28009 + }, + { + "epoch": 1.02, + "learning_rate": 9.760867176914343e-08, + "loss": 0.6883, + "step": 28010 + }, + { + "epoch": 1.02, + "learning_rate": 9.760500970447137e-08, + "loss": 0.6798, + "step": 28011 + }, + { + "epoch": 1.02, + "learning_rate": 9.760134763979931e-08, + "loss": 0.6875, + "step": 28012 + }, + { + "epoch": 1.02, + "learning_rate": 9.759768557512725e-08, + "loss": 0.6772, + "step": 28013 + }, + { + "epoch": 1.02, + "learning_rate": 9.759402351045519e-08, + "loss": 0.6955, + "step": 28014 + }, + { + "epoch": 1.02, + "learning_rate": 9.759036144578313e-08, + "loss": 0.6906, + "step": 28015 + }, + { + "epoch": 1.03, + "learning_rate": 9.758669938111107e-08, + "loss": 0.6879, + "step": 28016 + }, + { + "epoch": 1.03, + "learning_rate": 9.7583037316439e-08, + "loss": 0.6911, + "step": 28017 + }, + { + "epoch": 1.03, + "learning_rate": 9.757937525176694e-08, + "loss": 0.6894, + "step": 28018 + }, + { + "epoch": 1.03, + "learning_rate": 9.757571318709488e-08, + "loss": 0.7011, + "step": 28019 + }, + { + "epoch": 1.03, + "learning_rate": 9.757205112242282e-08, + "loss": 0.6902, + "step": 28020 + }, + { + "epoch": 1.03, + "learning_rate": 9.756838905775076e-08, + "loss": 0.684, + "step": 28021 + }, + { + "epoch": 1.03, + "learning_rate": 9.75647269930787e-08, + "loss": 0.6926, + "step": 28022 + }, + { + "epoch": 1.03, + "learning_rate": 9.756106492840664e-08, + "loss": 0.6853, + "step": 28023 + }, + { + "epoch": 1.03, + "learning_rate": 9.755740286373458e-08, + "loss": 0.6819, + "step": 28024 + }, + { + "epoch": 1.03, + "learning_rate": 9.755374079906252e-08, + "loss": 0.6968, + "step": 28025 + }, + { + "epoch": 1.03, + "learning_rate": 9.755007873439044e-08, + "loss": 0.6911, + "step": 28026 + }, + { + "epoch": 1.03, + "learning_rate": 9.754641666971838e-08, + "loss": 0.7039, + "step": 28027 + }, + { + "epoch": 1.03, + "learning_rate": 9.754275460504632e-08, + "loss": 0.7067, + "step": 28028 + }, + { + "epoch": 1.03, + "learning_rate": 9.753909254037425e-08, + "loss": 0.6927, + "step": 28029 + }, + { + "epoch": 1.03, + "learning_rate": 9.753543047570219e-08, + "loss": 0.6985, + "step": 28030 + }, + { + "epoch": 1.03, + "learning_rate": 9.753176841103013e-08, + "loss": 0.7035, + "step": 28031 + }, + { + "epoch": 1.03, + "learning_rate": 9.752810634635807e-08, + "loss": 0.6902, + "step": 28032 + }, + { + "epoch": 1.03, + "learning_rate": 9.752444428168601e-08, + "loss": 0.6958, + "step": 28033 + }, + { + "epoch": 1.03, + "learning_rate": 9.752078221701395e-08, + "loss": 0.6893, + "step": 28034 + }, + { + "epoch": 1.03, + "learning_rate": 9.751712015234188e-08, + "loss": 0.693, + "step": 28035 + }, + { + "epoch": 1.03, + "learning_rate": 9.751345808766982e-08, + "loss": 0.6875, + "step": 28036 + }, + { + "epoch": 1.03, + "learning_rate": 9.750979602299776e-08, + "loss": 0.6875, + "step": 28037 + }, + { + "epoch": 1.03, + "learning_rate": 9.75061339583257e-08, + "loss": 0.6929, + "step": 28038 + }, + { + "epoch": 1.03, + "learning_rate": 9.750247189365364e-08, + "loss": 0.6913, + "step": 28039 + }, + { + "epoch": 1.03, + "learning_rate": 9.749880982898158e-08, + "loss": 0.698, + "step": 28040 + }, + { + "epoch": 1.03, + "learning_rate": 9.749514776430952e-08, + "loss": 0.6925, + "step": 28041 + }, + { + "epoch": 1.03, + "learning_rate": 9.749148569963746e-08, + "loss": 0.697, + "step": 28042 + }, + { + "epoch": 1.03, + "learning_rate": 9.74878236349654e-08, + "loss": 0.6971, + "step": 28043 + }, + { + "epoch": 1.03, + "learning_rate": 9.748416157029332e-08, + "loss": 0.6921, + "step": 28044 + }, + { + "epoch": 1.03, + "learning_rate": 9.748049950562126e-08, + "loss": 0.6941, + "step": 28045 + }, + { + "epoch": 1.03, + "learning_rate": 9.74768374409492e-08, + "loss": 0.6931, + "step": 28046 + }, + { + "epoch": 1.03, + "learning_rate": 9.747317537627714e-08, + "loss": 0.6976, + "step": 28047 + }, + { + "epoch": 1.03, + "learning_rate": 9.746951331160508e-08, + "loss": 0.6976, + "step": 28048 + }, + { + "epoch": 1.03, + "learning_rate": 9.746585124693302e-08, + "loss": 0.6991, + "step": 28049 + }, + { + "epoch": 1.03, + "learning_rate": 9.746218918226096e-08, + "loss": 0.6765, + "step": 28050 + }, + { + "epoch": 1.03, + "learning_rate": 9.745852711758889e-08, + "loss": 0.692, + "step": 28051 + }, + { + "epoch": 1.03, + "learning_rate": 9.745486505291683e-08, + "loss": 0.6991, + "step": 28052 + }, + { + "epoch": 1.03, + "learning_rate": 9.745120298824476e-08, + "loss": 0.6898, + "step": 28053 + }, + { + "epoch": 1.03, + "learning_rate": 9.74475409235727e-08, + "loss": 0.6839, + "step": 28054 + }, + { + "epoch": 1.03, + "learning_rate": 9.744387885890064e-08, + "loss": 0.7062, + "step": 28055 + }, + { + "epoch": 1.03, + "learning_rate": 9.744021679422858e-08, + "loss": 0.6779, + "step": 28056 + }, + { + "epoch": 1.03, + "learning_rate": 9.743655472955652e-08, + "loss": 0.6981, + "step": 28057 + }, + { + "epoch": 1.03, + "learning_rate": 9.743289266488446e-08, + "loss": 0.6915, + "step": 28058 + }, + { + "epoch": 1.03, + "learning_rate": 9.74292306002124e-08, + "loss": 0.702, + "step": 28059 + }, + { + "epoch": 1.03, + "learning_rate": 9.742556853554034e-08, + "loss": 0.684, + "step": 28060 + }, + { + "epoch": 1.03, + "learning_rate": 9.742190647086828e-08, + "loss": 0.7045, + "step": 28061 + }, + { + "epoch": 1.03, + "learning_rate": 9.74182444061962e-08, + "loss": 0.6988, + "step": 28062 + }, + { + "epoch": 1.03, + "learning_rate": 9.741458234152414e-08, + "loss": 0.6974, + "step": 28063 + }, + { + "epoch": 1.03, + "learning_rate": 9.741092027685208e-08, + "loss": 0.6848, + "step": 28064 + }, + { + "epoch": 1.03, + "learning_rate": 9.740725821218002e-08, + "loss": 0.6957, + "step": 28065 + }, + { + "epoch": 1.03, + "learning_rate": 9.740359614750796e-08, + "loss": 0.6932, + "step": 28066 + }, + { + "epoch": 1.03, + "learning_rate": 9.73999340828359e-08, + "loss": 0.6987, + "step": 28067 + }, + { + "epoch": 1.03, + "learning_rate": 9.739627201816384e-08, + "loss": 0.696, + "step": 28068 + }, + { + "epoch": 1.03, + "learning_rate": 9.739260995349178e-08, + "loss": 0.6833, + "step": 28069 + }, + { + "epoch": 1.03, + "learning_rate": 9.738894788881972e-08, + "loss": 0.6858, + "step": 28070 + }, + { + "epoch": 1.03, + "learning_rate": 9.738528582414765e-08, + "loss": 0.6986, + "step": 28071 + }, + { + "epoch": 1.03, + "learning_rate": 9.738162375947559e-08, + "loss": 0.6976, + "step": 28072 + }, + { + "epoch": 1.03, + "learning_rate": 9.737796169480353e-08, + "loss": 0.696, + "step": 28073 + }, + { + "epoch": 1.03, + "learning_rate": 9.737429963013147e-08, + "loss": 0.6862, + "step": 28074 + }, + { + "epoch": 1.03, + "learning_rate": 9.73706375654594e-08, + "loss": 0.7037, + "step": 28075 + }, + { + "epoch": 1.03, + "learning_rate": 9.736697550078734e-08, + "loss": 0.6936, + "step": 28076 + }, + { + "epoch": 1.03, + "learning_rate": 9.736331343611528e-08, + "loss": 0.6692, + "step": 28077 + }, + { + "epoch": 1.03, + "learning_rate": 9.735965137144321e-08, + "loss": 0.6952, + "step": 28078 + }, + { + "epoch": 1.03, + "learning_rate": 9.735598930677115e-08, + "loss": 0.6924, + "step": 28079 + }, + { + "epoch": 1.03, + "learning_rate": 9.735232724209908e-08, + "loss": 0.6971, + "step": 28080 + }, + { + "epoch": 1.03, + "learning_rate": 9.734866517742702e-08, + "loss": 0.6944, + "step": 28081 + }, + { + "epoch": 1.03, + "learning_rate": 9.734500311275496e-08, + "loss": 0.6924, + "step": 28082 + }, + { + "epoch": 1.03, + "learning_rate": 9.73413410480829e-08, + "loss": 0.6958, + "step": 28083 + }, + { + "epoch": 1.03, + "learning_rate": 9.733767898341084e-08, + "loss": 0.6882, + "step": 28084 + }, + { + "epoch": 1.03, + "learning_rate": 9.733401691873878e-08, + "loss": 0.6905, + "step": 28085 + }, + { + "epoch": 1.03, + "learning_rate": 9.733035485406672e-08, + "loss": 0.6868, + "step": 28086 + }, + { + "epoch": 1.03, + "learning_rate": 9.732669278939466e-08, + "loss": 0.6973, + "step": 28087 + }, + { + "epoch": 1.03, + "learning_rate": 9.73230307247226e-08, + "loss": 0.6836, + "step": 28088 + }, + { + "epoch": 1.03, + "learning_rate": 9.731936866005053e-08, + "loss": 0.6949, + "step": 28089 + }, + { + "epoch": 1.03, + "learning_rate": 9.731570659537847e-08, + "loss": 0.7021, + "step": 28090 + }, + { + "epoch": 1.03, + "learning_rate": 9.731204453070641e-08, + "loss": 0.6939, + "step": 28091 + }, + { + "epoch": 1.03, + "learning_rate": 9.730838246603435e-08, + "loss": 0.6966, + "step": 28092 + }, + { + "epoch": 1.03, + "learning_rate": 9.730472040136229e-08, + "loss": 0.686, + "step": 28093 + }, + { + "epoch": 1.03, + "learning_rate": 9.730105833669023e-08, + "loss": 0.6893, + "step": 28094 + }, + { + "epoch": 1.03, + "learning_rate": 9.729739627201817e-08, + "loss": 0.6851, + "step": 28095 + }, + { + "epoch": 1.03, + "learning_rate": 9.729373420734611e-08, + "loss": 0.7215, + "step": 28096 + }, + { + "epoch": 1.03, + "learning_rate": 9.729007214267403e-08, + "loss": 0.6979, + "step": 28097 + }, + { + "epoch": 1.03, + "learning_rate": 9.728641007800197e-08, + "loss": 0.7008, + "step": 28098 + }, + { + "epoch": 1.03, + "learning_rate": 9.728274801332991e-08, + "loss": 0.696, + "step": 28099 + }, + { + "epoch": 1.03, + "learning_rate": 9.727908594865784e-08, + "loss": 0.7082, + "step": 28100 + }, + { + "epoch": 1.03, + "learning_rate": 9.727542388398578e-08, + "loss": 0.6897, + "step": 28101 + }, + { + "epoch": 1.03, + "learning_rate": 9.727176181931372e-08, + "loss": 0.6825, + "step": 28102 + }, + { + "epoch": 1.03, + "learning_rate": 9.726809975464166e-08, + "loss": 0.699, + "step": 28103 + }, + { + "epoch": 1.03, + "learning_rate": 9.72644376899696e-08, + "loss": 0.6894, + "step": 28104 + }, + { + "epoch": 1.03, + "learning_rate": 9.726077562529754e-08, + "loss": 0.6874, + "step": 28105 + }, + { + "epoch": 1.03, + "learning_rate": 9.725711356062548e-08, + "loss": 0.6903, + "step": 28106 + }, + { + "epoch": 1.03, + "learning_rate": 9.725345149595341e-08, + "loss": 0.7017, + "step": 28107 + }, + { + "epoch": 1.03, + "learning_rate": 9.724978943128135e-08, + "loss": 0.6934, + "step": 28108 + }, + { + "epoch": 1.03, + "learning_rate": 9.724612736660929e-08, + "loss": 0.7046, + "step": 28109 + }, + { + "epoch": 1.03, + "learning_rate": 9.724246530193723e-08, + "loss": 0.7028, + "step": 28110 + }, + { + "epoch": 1.03, + "learning_rate": 9.723880323726517e-08, + "loss": 0.6768, + "step": 28111 + }, + { + "epoch": 1.03, + "learning_rate": 9.72351411725931e-08, + "loss": 0.6838, + "step": 28112 + }, + { + "epoch": 1.03, + "learning_rate": 9.723147910792105e-08, + "loss": 0.6865, + "step": 28113 + }, + { + "epoch": 1.03, + "learning_rate": 9.722781704324899e-08, + "loss": 0.6879, + "step": 28114 + }, + { + "epoch": 1.03, + "learning_rate": 9.722415497857693e-08, + "loss": 0.6871, + "step": 28115 + }, + { + "epoch": 1.03, + "learning_rate": 9.722049291390485e-08, + "loss": 0.7021, + "step": 28116 + }, + { + "epoch": 1.03, + "learning_rate": 9.721683084923279e-08, + "loss": 0.7014, + "step": 28117 + }, + { + "epoch": 1.03, + "learning_rate": 9.721316878456073e-08, + "loss": 0.6879, + "step": 28118 + }, + { + "epoch": 1.03, + "learning_rate": 9.720950671988867e-08, + "loss": 0.6977, + "step": 28119 + }, + { + "epoch": 1.03, + "learning_rate": 9.720584465521661e-08, + "loss": 0.6916, + "step": 28120 + }, + { + "epoch": 1.03, + "learning_rate": 9.720218259054454e-08, + "loss": 0.6977, + "step": 28121 + }, + { + "epoch": 1.03, + "learning_rate": 9.719852052587248e-08, + "loss": 0.6994, + "step": 28122 + }, + { + "epoch": 1.03, + "learning_rate": 9.719485846120042e-08, + "loss": 0.6897, + "step": 28123 + }, + { + "epoch": 1.03, + "learning_rate": 9.719119639652836e-08, + "loss": 0.6794, + "step": 28124 + }, + { + "epoch": 1.03, + "learning_rate": 9.718753433185629e-08, + "loss": 0.6852, + "step": 28125 + }, + { + "epoch": 1.03, + "learning_rate": 9.718387226718423e-08, + "loss": 0.6909, + "step": 28126 + }, + { + "epoch": 1.03, + "learning_rate": 9.718021020251217e-08, + "loss": 0.6895, + "step": 28127 + }, + { + "epoch": 1.03, + "learning_rate": 9.71765481378401e-08, + "loss": 0.6983, + "step": 28128 + }, + { + "epoch": 1.03, + "learning_rate": 9.717288607316805e-08, + "loss": 0.6838, + "step": 28129 + }, + { + "epoch": 1.03, + "learning_rate": 9.716922400849599e-08, + "loss": 0.6958, + "step": 28130 + }, + { + "epoch": 1.03, + "learning_rate": 9.716556194382393e-08, + "loss": 0.6963, + "step": 28131 + }, + { + "epoch": 1.03, + "learning_rate": 9.716189987915187e-08, + "loss": 0.6995, + "step": 28132 + }, + { + "epoch": 1.03, + "learning_rate": 9.71582378144798e-08, + "loss": 0.7042, + "step": 28133 + }, + { + "epoch": 1.03, + "learning_rate": 9.715457574980773e-08, + "loss": 0.7031, + "step": 28134 + }, + { + "epoch": 1.03, + "learning_rate": 9.715091368513567e-08, + "loss": 0.6959, + "step": 28135 + }, + { + "epoch": 1.03, + "learning_rate": 9.714725162046361e-08, + "loss": 0.6987, + "step": 28136 + }, + { + "epoch": 1.03, + "learning_rate": 9.714358955579155e-08, + "loss": 0.6927, + "step": 28137 + }, + { + "epoch": 1.03, + "learning_rate": 9.713992749111949e-08, + "loss": 0.6838, + "step": 28138 + }, + { + "epoch": 1.03, + "learning_rate": 9.713626542644743e-08, + "loss": 0.7045, + "step": 28139 + }, + { + "epoch": 1.03, + "learning_rate": 9.713260336177537e-08, + "loss": 0.7013, + "step": 28140 + }, + { + "epoch": 1.03, + "learning_rate": 9.712894129710331e-08, + "loss": 0.6831, + "step": 28141 + }, + { + "epoch": 1.03, + "learning_rate": 9.712527923243125e-08, + "loss": 0.6966, + "step": 28142 + }, + { + "epoch": 1.03, + "learning_rate": 9.712161716775918e-08, + "loss": 0.6881, + "step": 28143 + }, + { + "epoch": 1.03, + "learning_rate": 9.711795510308712e-08, + "loss": 0.6871, + "step": 28144 + }, + { + "epoch": 1.03, + "learning_rate": 9.711429303841506e-08, + "loss": 0.6878, + "step": 28145 + }, + { + "epoch": 1.03, + "learning_rate": 9.711063097374298e-08, + "loss": 0.6897, + "step": 28146 + }, + { + "epoch": 1.03, + "learning_rate": 9.710696890907092e-08, + "loss": 0.6929, + "step": 28147 + }, + { + "epoch": 1.03, + "learning_rate": 9.710330684439886e-08, + "loss": 0.6928, + "step": 28148 + }, + { + "epoch": 1.03, + "learning_rate": 9.70996447797268e-08, + "loss": 0.692, + "step": 28149 + }, + { + "epoch": 1.03, + "learning_rate": 9.709598271505474e-08, + "loss": 0.6961, + "step": 28150 + }, + { + "epoch": 1.03, + "learning_rate": 9.709232065038268e-08, + "loss": 0.6976, + "step": 28151 + }, + { + "epoch": 1.03, + "learning_rate": 9.708865858571061e-08, + "loss": 0.6929, + "step": 28152 + }, + { + "epoch": 1.03, + "learning_rate": 9.708499652103855e-08, + "loss": 0.6897, + "step": 28153 + }, + { + "epoch": 1.03, + "learning_rate": 9.708133445636649e-08, + "loss": 0.6865, + "step": 28154 + }, + { + "epoch": 1.03, + "learning_rate": 9.707767239169443e-08, + "loss": 0.6987, + "step": 28155 + }, + { + "epoch": 1.03, + "learning_rate": 9.707401032702237e-08, + "loss": 0.7053, + "step": 28156 + }, + { + "epoch": 1.03, + "learning_rate": 9.707034826235031e-08, + "loss": 0.6754, + "step": 28157 + }, + { + "epoch": 1.03, + "learning_rate": 9.706668619767825e-08, + "loss": 0.6943, + "step": 28158 + }, + { + "epoch": 1.03, + "learning_rate": 9.706302413300619e-08, + "loss": 0.6805, + "step": 28159 + }, + { + "epoch": 1.03, + "learning_rate": 9.705936206833413e-08, + "loss": 0.7024, + "step": 28160 + }, + { + "epoch": 1.03, + "learning_rate": 9.705570000366206e-08, + "loss": 0.7001, + "step": 28161 + }, + { + "epoch": 1.03, + "learning_rate": 9.705203793899e-08, + "loss": 0.7006, + "step": 28162 + }, + { + "epoch": 1.03, + "learning_rate": 9.704837587431794e-08, + "loss": 0.6977, + "step": 28163 + }, + { + "epoch": 1.03, + "learning_rate": 9.704471380964588e-08, + "loss": 0.693, + "step": 28164 + }, + { + "epoch": 1.03, + "learning_rate": 9.704105174497382e-08, + "loss": 0.6979, + "step": 28165 + }, + { + "epoch": 1.03, + "learning_rate": 9.703738968030176e-08, + "loss": 0.6825, + "step": 28166 + }, + { + "epoch": 1.03, + "learning_rate": 9.70337276156297e-08, + "loss": 0.6916, + "step": 28167 + }, + { + "epoch": 1.03, + "learning_rate": 9.703006555095762e-08, + "loss": 0.684, + "step": 28168 + }, + { + "epoch": 1.03, + "learning_rate": 9.702640348628556e-08, + "loss": 0.7019, + "step": 28169 + }, + { + "epoch": 1.03, + "learning_rate": 9.70227414216135e-08, + "loss": 0.7034, + "step": 28170 + }, + { + "epoch": 1.03, + "learning_rate": 9.701907935694143e-08, + "loss": 0.6864, + "step": 28171 + }, + { + "epoch": 1.03, + "learning_rate": 9.701541729226937e-08, + "loss": 0.6819, + "step": 28172 + }, + { + "epoch": 1.03, + "learning_rate": 9.701175522759731e-08, + "loss": 0.6913, + "step": 28173 + }, + { + "epoch": 1.03, + "learning_rate": 9.700809316292525e-08, + "loss": 0.6932, + "step": 28174 + }, + { + "epoch": 1.03, + "learning_rate": 9.700443109825319e-08, + "loss": 0.693, + "step": 28175 + }, + { + "epoch": 1.03, + "learning_rate": 9.700076903358113e-08, + "loss": 0.6916, + "step": 28176 + }, + { + "epoch": 1.03, + "learning_rate": 9.699710696890907e-08, + "loss": 0.6797, + "step": 28177 + }, + { + "epoch": 1.03, + "learning_rate": 9.699344490423701e-08, + "loss": 0.6975, + "step": 28178 + }, + { + "epoch": 1.03, + "learning_rate": 9.698978283956494e-08, + "loss": 0.6963, + "step": 28179 + }, + { + "epoch": 1.03, + "learning_rate": 9.698612077489288e-08, + "loss": 0.6887, + "step": 28180 + }, + { + "epoch": 1.03, + "learning_rate": 9.698245871022082e-08, + "loss": 0.6899, + "step": 28181 + }, + { + "epoch": 1.03, + "learning_rate": 9.697879664554876e-08, + "loss": 0.6907, + "step": 28182 + }, + { + "epoch": 1.03, + "learning_rate": 9.69751345808767e-08, + "loss": 0.6874, + "step": 28183 + }, + { + "epoch": 1.03, + "learning_rate": 9.697147251620464e-08, + "loss": 0.6706, + "step": 28184 + }, + { + "epoch": 1.03, + "learning_rate": 9.696781045153258e-08, + "loss": 0.6937, + "step": 28185 + }, + { + "epoch": 1.03, + "learning_rate": 9.696414838686052e-08, + "loss": 0.696, + "step": 28186 + }, + { + "epoch": 1.03, + "learning_rate": 9.696048632218846e-08, + "loss": 0.6945, + "step": 28187 + }, + { + "epoch": 1.03, + "learning_rate": 9.695682425751638e-08, + "loss": 0.6952, + "step": 28188 + }, + { + "epoch": 1.03, + "learning_rate": 9.695316219284432e-08, + "loss": 0.7055, + "step": 28189 + }, + { + "epoch": 1.03, + "learning_rate": 9.694950012817226e-08, + "loss": 0.7007, + "step": 28190 + }, + { + "epoch": 1.03, + "learning_rate": 9.69458380635002e-08, + "loss": 0.6934, + "step": 28191 + }, + { + "epoch": 1.03, + "learning_rate": 9.694217599882813e-08, + "loss": 0.6863, + "step": 28192 + }, + { + "epoch": 1.03, + "learning_rate": 9.693851393415607e-08, + "loss": 0.701, + "step": 28193 + }, + { + "epoch": 1.03, + "learning_rate": 9.693485186948401e-08, + "loss": 0.7016, + "step": 28194 + }, + { + "epoch": 1.03, + "learning_rate": 9.693118980481195e-08, + "loss": 0.6994, + "step": 28195 + }, + { + "epoch": 1.03, + "learning_rate": 9.692752774013989e-08, + "loss": 0.6899, + "step": 28196 + }, + { + "epoch": 1.03, + "learning_rate": 9.692386567546781e-08, + "loss": 0.6924, + "step": 28197 + }, + { + "epoch": 1.03, + "learning_rate": 9.692020361079575e-08, + "loss": 0.6891, + "step": 28198 + }, + { + "epoch": 1.03, + "learning_rate": 9.69165415461237e-08, + "loss": 0.6875, + "step": 28199 + }, + { + "epoch": 1.03, + "learning_rate": 9.691287948145163e-08, + "loss": 0.6849, + "step": 28200 + }, + { + "epoch": 1.03, + "learning_rate": 9.690921741677957e-08, + "loss": 0.6991, + "step": 28201 + }, + { + "epoch": 1.03, + "learning_rate": 9.690555535210751e-08, + "loss": 0.6962, + "step": 28202 + }, + { + "epoch": 1.03, + "learning_rate": 9.690189328743545e-08, + "loss": 0.6931, + "step": 28203 + }, + { + "epoch": 1.03, + "learning_rate": 9.68982312227634e-08, + "loss": 0.6975, + "step": 28204 + }, + { + "epoch": 1.03, + "learning_rate": 9.689456915809133e-08, + "loss": 0.6888, + "step": 28205 + }, + { + "epoch": 1.03, + "learning_rate": 9.689090709341926e-08, + "loss": 0.7168, + "step": 28206 + }, + { + "epoch": 1.03, + "learning_rate": 9.68872450287472e-08, + "loss": 0.6872, + "step": 28207 + }, + { + "epoch": 1.03, + "learning_rate": 9.688358296407514e-08, + "loss": 0.6916, + "step": 28208 + }, + { + "epoch": 1.03, + "learning_rate": 9.687992089940308e-08, + "loss": 0.6964, + "step": 28209 + }, + { + "epoch": 1.03, + "learning_rate": 9.687625883473102e-08, + "loss": 0.6957, + "step": 28210 + }, + { + "epoch": 1.03, + "learning_rate": 9.687259677005896e-08, + "loss": 0.6766, + "step": 28211 + }, + { + "epoch": 1.03, + "learning_rate": 9.68689347053869e-08, + "loss": 0.7036, + "step": 28212 + }, + { + "epoch": 1.03, + "learning_rate": 9.686527264071484e-08, + "loss": 0.6942, + "step": 28213 + }, + { + "epoch": 1.03, + "learning_rate": 9.686161057604277e-08, + "loss": 0.6977, + "step": 28214 + }, + { + "epoch": 1.03, + "learning_rate": 9.685794851137071e-08, + "loss": 0.6983, + "step": 28215 + }, + { + "epoch": 1.03, + "learning_rate": 9.685428644669865e-08, + "loss": 0.6959, + "step": 28216 + }, + { + "epoch": 1.03, + "learning_rate": 9.685062438202657e-08, + "loss": 0.687, + "step": 28217 + }, + { + "epoch": 1.03, + "learning_rate": 9.684696231735451e-08, + "loss": 0.6896, + "step": 28218 + }, + { + "epoch": 1.03, + "learning_rate": 9.684330025268245e-08, + "loss": 0.694, + "step": 28219 + }, + { + "epoch": 1.03, + "learning_rate": 9.68396381880104e-08, + "loss": 0.7072, + "step": 28220 + }, + { + "epoch": 1.03, + "learning_rate": 9.683597612333833e-08, + "loss": 0.6961, + "step": 28221 + }, + { + "epoch": 1.03, + "learning_rate": 9.683231405866627e-08, + "loss": 0.69, + "step": 28222 + }, + { + "epoch": 1.03, + "learning_rate": 9.682865199399421e-08, + "loss": 0.6929, + "step": 28223 + }, + { + "epoch": 1.03, + "learning_rate": 9.682498992932214e-08, + "loss": 0.6814, + "step": 28224 + }, + { + "epoch": 1.03, + "learning_rate": 9.682132786465008e-08, + "loss": 0.6837, + "step": 28225 + }, + { + "epoch": 1.03, + "learning_rate": 9.681766579997802e-08, + "loss": 0.7088, + "step": 28226 + }, + { + "epoch": 1.03, + "learning_rate": 9.681400373530596e-08, + "loss": 0.699, + "step": 28227 + }, + { + "epoch": 1.03, + "learning_rate": 9.68103416706339e-08, + "loss": 0.7003, + "step": 28228 + }, + { + "epoch": 1.03, + "learning_rate": 9.680667960596184e-08, + "loss": 0.7059, + "step": 28229 + }, + { + "epoch": 1.03, + "learning_rate": 9.680301754128978e-08, + "loss": 0.6677, + "step": 28230 + }, + { + "epoch": 1.03, + "learning_rate": 9.679935547661772e-08, + "loss": 0.6938, + "step": 28231 + }, + { + "epoch": 1.03, + "learning_rate": 9.679569341194566e-08, + "loss": 0.6899, + "step": 28232 + }, + { + "epoch": 1.03, + "learning_rate": 9.679203134727359e-08, + "loss": 0.6984, + "step": 28233 + }, + { + "epoch": 1.03, + "learning_rate": 9.678836928260153e-08, + "loss": 0.6964, + "step": 28234 + }, + { + "epoch": 1.03, + "learning_rate": 9.678470721792947e-08, + "loss": 0.6924, + "step": 28235 + }, + { + "epoch": 1.03, + "learning_rate": 9.67810451532574e-08, + "loss": 0.7031, + "step": 28236 + }, + { + "epoch": 1.03, + "learning_rate": 9.677738308858535e-08, + "loss": 0.6979, + "step": 28237 + }, + { + "epoch": 1.03, + "learning_rate": 9.677372102391329e-08, + "loss": 0.6994, + "step": 28238 + }, + { + "epoch": 1.03, + "learning_rate": 9.677005895924121e-08, + "loss": 0.6925, + "step": 28239 + }, + { + "epoch": 1.03, + "learning_rate": 9.676639689456915e-08, + "loss": 0.6941, + "step": 28240 + }, + { + "epoch": 1.03, + "learning_rate": 9.676273482989709e-08, + "loss": 0.6833, + "step": 28241 + }, + { + "epoch": 1.03, + "learning_rate": 9.675907276522502e-08, + "loss": 0.7085, + "step": 28242 + }, + { + "epoch": 1.03, + "learning_rate": 9.675541070055296e-08, + "loss": 0.6929, + "step": 28243 + }, + { + "epoch": 1.03, + "learning_rate": 9.67517486358809e-08, + "loss": 0.692, + "step": 28244 + }, + { + "epoch": 1.03, + "learning_rate": 9.674808657120884e-08, + "loss": 0.6927, + "step": 28245 + }, + { + "epoch": 1.03, + "learning_rate": 9.674442450653678e-08, + "loss": 0.697, + "step": 28246 + }, + { + "epoch": 1.03, + "learning_rate": 9.674076244186472e-08, + "loss": 0.6941, + "step": 28247 + }, + { + "epoch": 1.03, + "learning_rate": 9.673710037719266e-08, + "loss": 0.7043, + "step": 28248 + }, + { + "epoch": 1.03, + "learning_rate": 9.67334383125206e-08, + "loss": 0.6903, + "step": 28249 + }, + { + "epoch": 1.03, + "learning_rate": 9.672977624784854e-08, + "loss": 0.691, + "step": 28250 + }, + { + "epoch": 1.03, + "learning_rate": 9.672611418317647e-08, + "loss": 0.695, + "step": 28251 + }, + { + "epoch": 1.03, + "learning_rate": 9.67224521185044e-08, + "loss": 0.6883, + "step": 28252 + }, + { + "epoch": 1.03, + "learning_rate": 9.671879005383235e-08, + "loss": 0.6983, + "step": 28253 + }, + { + "epoch": 1.03, + "learning_rate": 9.671512798916029e-08, + "loss": 0.6877, + "step": 28254 + }, + { + "epoch": 1.03, + "learning_rate": 9.671146592448822e-08, + "loss": 0.7002, + "step": 28255 + }, + { + "epoch": 1.03, + "learning_rate": 9.670780385981616e-08, + "loss": 0.6883, + "step": 28256 + }, + { + "epoch": 1.03, + "learning_rate": 9.67041417951441e-08, + "loss": 0.692, + "step": 28257 + }, + { + "epoch": 1.03, + "learning_rate": 9.670047973047204e-08, + "loss": 0.6945, + "step": 28258 + }, + { + "epoch": 1.03, + "learning_rate": 9.669681766579998e-08, + "loss": 0.6863, + "step": 28259 + }, + { + "epoch": 1.03, + "learning_rate": 9.669315560112791e-08, + "loss": 0.7007, + "step": 28260 + }, + { + "epoch": 1.03, + "learning_rate": 9.668949353645585e-08, + "loss": 0.6887, + "step": 28261 + }, + { + "epoch": 1.03, + "learning_rate": 9.668583147178379e-08, + "loss": 0.6896, + "step": 28262 + }, + { + "epoch": 1.03, + "learning_rate": 9.668216940711172e-08, + "loss": 0.6957, + "step": 28263 + }, + { + "epoch": 1.03, + "learning_rate": 9.667850734243966e-08, + "loss": 0.7022, + "step": 28264 + }, + { + "epoch": 1.03, + "learning_rate": 9.66748452777676e-08, + "loss": 0.6994, + "step": 28265 + }, + { + "epoch": 1.03, + "learning_rate": 9.667118321309554e-08, + "loss": 0.694, + "step": 28266 + }, + { + "epoch": 1.03, + "learning_rate": 9.666752114842348e-08, + "loss": 0.7022, + "step": 28267 + }, + { + "epoch": 1.03, + "learning_rate": 9.666385908375142e-08, + "loss": 0.6952, + "step": 28268 + }, + { + "epoch": 1.03, + "learning_rate": 9.666019701907934e-08, + "loss": 0.6915, + "step": 28269 + }, + { + "epoch": 1.03, + "learning_rate": 9.665653495440728e-08, + "loss": 0.6984, + "step": 28270 + }, + { + "epoch": 1.03, + "learning_rate": 9.665287288973522e-08, + "loss": 0.6909, + "step": 28271 + }, + { + "epoch": 1.03, + "learning_rate": 9.664921082506316e-08, + "loss": 0.6984, + "step": 28272 + }, + { + "epoch": 1.03, + "learning_rate": 9.66455487603911e-08, + "loss": 0.6882, + "step": 28273 + }, + { + "epoch": 1.03, + "learning_rate": 9.664188669571904e-08, + "loss": 0.697, + "step": 28274 + }, + { + "epoch": 1.03, + "learning_rate": 9.663822463104698e-08, + "loss": 0.7082, + "step": 28275 + }, + { + "epoch": 1.03, + "learning_rate": 9.663456256637492e-08, + "loss": 0.6951, + "step": 28276 + }, + { + "epoch": 1.03, + "learning_rate": 9.663090050170286e-08, + "loss": 0.6988, + "step": 28277 + }, + { + "epoch": 1.03, + "learning_rate": 9.662723843703079e-08, + "loss": 0.6866, + "step": 28278 + }, + { + "epoch": 1.03, + "learning_rate": 9.662357637235873e-08, + "loss": 0.6972, + "step": 28279 + }, + { + "epoch": 1.03, + "learning_rate": 9.661991430768667e-08, + "loss": 0.6988, + "step": 28280 + }, + { + "epoch": 1.03, + "learning_rate": 9.661625224301461e-08, + "loss": 0.6859, + "step": 28281 + }, + { + "epoch": 1.03, + "learning_rate": 9.661259017834255e-08, + "loss": 0.6845, + "step": 28282 + }, + { + "epoch": 1.03, + "learning_rate": 9.660892811367049e-08, + "loss": 0.6928, + "step": 28283 + }, + { + "epoch": 1.03, + "learning_rate": 9.660526604899843e-08, + "loss": 0.6906, + "step": 28284 + }, + { + "epoch": 1.03, + "learning_rate": 9.660160398432636e-08, + "loss": 0.7032, + "step": 28285 + }, + { + "epoch": 1.03, + "learning_rate": 9.65979419196543e-08, + "loss": 0.7049, + "step": 28286 + }, + { + "epoch": 1.03, + "learning_rate": 9.659427985498224e-08, + "loss": 0.6856, + "step": 28287 + }, + { + "epoch": 1.03, + "learning_rate": 9.659061779031016e-08, + "loss": 0.6988, + "step": 28288 + }, + { + "epoch": 1.04, + "learning_rate": 9.65869557256381e-08, + "loss": 0.6922, + "step": 28289 + }, + { + "epoch": 1.04, + "learning_rate": 9.658329366096604e-08, + "loss": 0.6952, + "step": 28290 + }, + { + "epoch": 1.04, + "learning_rate": 9.657963159629398e-08, + "loss": 0.6914, + "step": 28291 + }, + { + "epoch": 1.04, + "learning_rate": 9.657596953162192e-08, + "loss": 0.7029, + "step": 28292 + }, + { + "epoch": 1.04, + "learning_rate": 9.657230746694986e-08, + "loss": 0.6865, + "step": 28293 + }, + { + "epoch": 1.04, + "learning_rate": 9.65686454022778e-08, + "loss": 0.6902, + "step": 28294 + }, + { + "epoch": 1.04, + "learning_rate": 9.656498333760574e-08, + "loss": 0.6805, + "step": 28295 + }, + { + "epoch": 1.04, + "learning_rate": 9.656132127293367e-08, + "loss": 0.679, + "step": 28296 + }, + { + "epoch": 1.04, + "learning_rate": 9.655765920826161e-08, + "loss": 0.7059, + "step": 28297 + }, + { + "epoch": 1.04, + "learning_rate": 9.655399714358955e-08, + "loss": 0.699, + "step": 28298 + }, + { + "epoch": 1.04, + "learning_rate": 9.655033507891749e-08, + "loss": 0.693, + "step": 28299 + }, + { + "epoch": 1.04, + "learning_rate": 9.654667301424543e-08, + "loss": 0.7068, + "step": 28300 + }, + { + "epoch": 1.04, + "learning_rate": 9.654301094957337e-08, + "loss": 0.6894, + "step": 28301 + }, + { + "epoch": 1.04, + "learning_rate": 9.653934888490131e-08, + "loss": 0.6866, + "step": 28302 + }, + { + "epoch": 1.04, + "learning_rate": 9.653568682022925e-08, + "loss": 0.6842, + "step": 28303 + }, + { + "epoch": 1.04, + "learning_rate": 9.653202475555719e-08, + "loss": 0.6951, + "step": 28304 + }, + { + "epoch": 1.04, + "learning_rate": 9.652836269088512e-08, + "loss": 0.7011, + "step": 28305 + }, + { + "epoch": 1.04, + "learning_rate": 9.652470062621306e-08, + "loss": 0.6912, + "step": 28306 + }, + { + "epoch": 1.04, + "learning_rate": 9.6521038561541e-08, + "loss": 0.69, + "step": 28307 + }, + { + "epoch": 1.04, + "learning_rate": 9.651737649686894e-08, + "loss": 0.6867, + "step": 28308 + }, + { + "epoch": 1.04, + "learning_rate": 9.651371443219688e-08, + "loss": 0.7004, + "step": 28309 + }, + { + "epoch": 1.04, + "learning_rate": 9.65100523675248e-08, + "loss": 0.6868, + "step": 28310 + }, + { + "epoch": 1.04, + "learning_rate": 9.650639030285274e-08, + "loss": 0.7088, + "step": 28311 + }, + { + "epoch": 1.04, + "learning_rate": 9.650272823818068e-08, + "loss": 0.7046, + "step": 28312 + }, + { + "epoch": 1.04, + "learning_rate": 9.649906617350862e-08, + "loss": 0.6912, + "step": 28313 + }, + { + "epoch": 1.04, + "learning_rate": 9.649540410883655e-08, + "loss": 0.6892, + "step": 28314 + }, + { + "epoch": 1.04, + "learning_rate": 9.649174204416449e-08, + "loss": 0.7001, + "step": 28315 + }, + { + "epoch": 1.04, + "learning_rate": 9.648807997949243e-08, + "loss": 0.6886, + "step": 28316 + }, + { + "epoch": 1.04, + "learning_rate": 9.648441791482037e-08, + "loss": 0.6925, + "step": 28317 + }, + { + "epoch": 1.04, + "learning_rate": 9.648075585014831e-08, + "loss": 0.6885, + "step": 28318 + }, + { + "epoch": 1.04, + "learning_rate": 9.647709378547625e-08, + "loss": 0.6852, + "step": 28319 + }, + { + "epoch": 1.04, + "learning_rate": 9.647343172080419e-08, + "loss": 0.6816, + "step": 28320 + }, + { + "epoch": 1.04, + "learning_rate": 9.646976965613213e-08, + "loss": 0.692, + "step": 28321 + }, + { + "epoch": 1.04, + "learning_rate": 9.646610759146005e-08, + "loss": 0.686, + "step": 28322 + }, + { + "epoch": 1.04, + "learning_rate": 9.6462445526788e-08, + "loss": 0.692, + "step": 28323 + }, + { + "epoch": 1.04, + "learning_rate": 9.645878346211593e-08, + "loss": 0.6833, + "step": 28324 + }, + { + "epoch": 1.04, + "learning_rate": 9.645512139744387e-08, + "loss": 0.7039, + "step": 28325 + }, + { + "epoch": 1.04, + "learning_rate": 9.645145933277181e-08, + "loss": 0.6883, + "step": 28326 + }, + { + "epoch": 1.04, + "learning_rate": 9.644779726809975e-08, + "loss": 0.6896, + "step": 28327 + }, + { + "epoch": 1.04, + "learning_rate": 9.64441352034277e-08, + "loss": 0.7107, + "step": 28328 + }, + { + "epoch": 1.04, + "learning_rate": 9.644047313875563e-08, + "loss": 0.6916, + "step": 28329 + }, + { + "epoch": 1.04, + "learning_rate": 9.643681107408357e-08, + "loss": 0.6985, + "step": 28330 + }, + { + "epoch": 1.04, + "learning_rate": 9.64331490094115e-08, + "loss": 0.6833, + "step": 28331 + }, + { + "epoch": 1.04, + "learning_rate": 9.642948694473944e-08, + "loss": 0.7062, + "step": 28332 + }, + { + "epoch": 1.04, + "learning_rate": 9.642582488006738e-08, + "loss": 0.6984, + "step": 28333 + }, + { + "epoch": 1.04, + "learning_rate": 9.642216281539531e-08, + "loss": 0.6999, + "step": 28334 + }, + { + "epoch": 1.04, + "learning_rate": 9.641850075072325e-08, + "loss": 0.6883, + "step": 28335 + }, + { + "epoch": 1.04, + "learning_rate": 9.641483868605119e-08, + "loss": 0.6851, + "step": 28336 + }, + { + "epoch": 1.04, + "learning_rate": 9.641117662137913e-08, + "loss": 0.678, + "step": 28337 + }, + { + "epoch": 1.04, + "learning_rate": 9.640751455670707e-08, + "loss": 0.7032, + "step": 28338 + }, + { + "epoch": 1.04, + "learning_rate": 9.640385249203501e-08, + "loss": 0.6925, + "step": 28339 + }, + { + "epoch": 1.04, + "learning_rate": 9.640019042736293e-08, + "loss": 0.7045, + "step": 28340 + }, + { + "epoch": 1.04, + "learning_rate": 9.639652836269087e-08, + "loss": 0.6939, + "step": 28341 + }, + { + "epoch": 1.04, + "learning_rate": 9.639286629801881e-08, + "loss": 0.7084, + "step": 28342 + }, + { + "epoch": 1.04, + "learning_rate": 9.638920423334675e-08, + "loss": 0.6756, + "step": 28343 + }, + { + "epoch": 1.04, + "learning_rate": 9.638554216867469e-08, + "loss": 0.7012, + "step": 28344 + }, + { + "epoch": 1.04, + "learning_rate": 9.638188010400263e-08, + "loss": 0.6872, + "step": 28345 + }, + { + "epoch": 1.04, + "learning_rate": 9.637821803933057e-08, + "loss": 0.6921, + "step": 28346 + }, + { + "epoch": 1.04, + "learning_rate": 9.637455597465851e-08, + "loss": 0.6953, + "step": 28347 + }, + { + "epoch": 1.04, + "learning_rate": 9.637089390998645e-08, + "loss": 0.7021, + "step": 28348 + }, + { + "epoch": 1.04, + "learning_rate": 9.636723184531438e-08, + "loss": 0.6996, + "step": 28349 + }, + { + "epoch": 1.04, + "learning_rate": 9.636356978064232e-08, + "loss": 0.6906, + "step": 28350 + }, + { + "epoch": 1.04, + "learning_rate": 9.635990771597026e-08, + "loss": 0.6893, + "step": 28351 + }, + { + "epoch": 1.04, + "learning_rate": 9.63562456512982e-08, + "loss": 0.6862, + "step": 28352 + }, + { + "epoch": 1.04, + "learning_rate": 9.635258358662614e-08, + "loss": 0.6949, + "step": 28353 + }, + { + "epoch": 1.04, + "learning_rate": 9.634892152195408e-08, + "loss": 0.6953, + "step": 28354 + }, + { + "epoch": 1.04, + "learning_rate": 9.634525945728202e-08, + "loss": 0.6885, + "step": 28355 + }, + { + "epoch": 1.04, + "learning_rate": 9.634159739260995e-08, + "loss": 0.6933, + "step": 28356 + }, + { + "epoch": 1.04, + "learning_rate": 9.633793532793789e-08, + "loss": 0.6941, + "step": 28357 + }, + { + "epoch": 1.04, + "learning_rate": 9.633427326326583e-08, + "loss": 0.6973, + "step": 28358 + }, + { + "epoch": 1.04, + "learning_rate": 9.633061119859375e-08, + "loss": 0.6867, + "step": 28359 + }, + { + "epoch": 1.04, + "learning_rate": 9.632694913392169e-08, + "loss": 0.698, + "step": 28360 + }, + { + "epoch": 1.04, + "learning_rate": 9.632328706924963e-08, + "loss": 0.6846, + "step": 28361 + }, + { + "epoch": 1.04, + "learning_rate": 9.631962500457757e-08, + "loss": 0.6897, + "step": 28362 + }, + { + "epoch": 1.04, + "learning_rate": 9.631596293990551e-08, + "loss": 0.6954, + "step": 28363 + }, + { + "epoch": 1.04, + "learning_rate": 9.631230087523345e-08, + "loss": 0.6853, + "step": 28364 + }, + { + "epoch": 1.04, + "learning_rate": 9.630863881056139e-08, + "loss": 0.6913, + "step": 28365 + }, + { + "epoch": 1.04, + "learning_rate": 9.630497674588933e-08, + "loss": 0.6932, + "step": 28366 + }, + { + "epoch": 1.04, + "learning_rate": 9.630131468121726e-08, + "loss": 0.7074, + "step": 28367 + }, + { + "epoch": 1.04, + "learning_rate": 9.62976526165452e-08, + "loss": 0.6851, + "step": 28368 + }, + { + "epoch": 1.04, + "learning_rate": 9.629399055187314e-08, + "loss": 0.6984, + "step": 28369 + }, + { + "epoch": 1.04, + "learning_rate": 9.629032848720108e-08, + "loss": 0.688, + "step": 28370 + }, + { + "epoch": 1.04, + "learning_rate": 9.628666642252902e-08, + "loss": 0.7017, + "step": 28371 + }, + { + "epoch": 1.04, + "learning_rate": 9.628300435785696e-08, + "loss": 0.6891, + "step": 28372 + }, + { + "epoch": 1.04, + "learning_rate": 9.62793422931849e-08, + "loss": 0.692, + "step": 28373 + }, + { + "epoch": 1.04, + "learning_rate": 9.627568022851284e-08, + "loss": 0.6869, + "step": 28374 + }, + { + "epoch": 1.04, + "learning_rate": 9.627201816384078e-08, + "loss": 0.7025, + "step": 28375 + }, + { + "epoch": 1.04, + "learning_rate": 9.62683560991687e-08, + "loss": 0.7076, + "step": 28376 + }, + { + "epoch": 1.04, + "learning_rate": 9.626469403449664e-08, + "loss": 0.7021, + "step": 28377 + }, + { + "epoch": 1.04, + "learning_rate": 9.626103196982458e-08, + "loss": 0.6932, + "step": 28378 + }, + { + "epoch": 1.04, + "learning_rate": 9.625736990515252e-08, + "loss": 0.7012, + "step": 28379 + }, + { + "epoch": 1.04, + "learning_rate": 9.625370784048046e-08, + "loss": 0.6969, + "step": 28380 + }, + { + "epoch": 1.04, + "learning_rate": 9.625004577580839e-08, + "loss": 0.7019, + "step": 28381 + }, + { + "epoch": 1.04, + "learning_rate": 9.624638371113633e-08, + "loss": 0.6945, + "step": 28382 + }, + { + "epoch": 1.04, + "learning_rate": 9.624272164646427e-08, + "loss": 0.7041, + "step": 28383 + }, + { + "epoch": 1.04, + "learning_rate": 9.623905958179221e-08, + "loss": 0.6834, + "step": 28384 + }, + { + "epoch": 1.04, + "learning_rate": 9.623539751712014e-08, + "loss": 0.6987, + "step": 28385 + }, + { + "epoch": 1.04, + "learning_rate": 9.623173545244808e-08, + "loss": 0.6947, + "step": 28386 + }, + { + "epoch": 1.04, + "learning_rate": 9.622807338777602e-08, + "loss": 0.693, + "step": 28387 + }, + { + "epoch": 1.04, + "learning_rate": 9.622441132310396e-08, + "loss": 0.685, + "step": 28388 + }, + { + "epoch": 1.04, + "learning_rate": 9.62207492584319e-08, + "loss": 0.6944, + "step": 28389 + }, + { + "epoch": 1.04, + "learning_rate": 9.621708719375984e-08, + "loss": 0.6964, + "step": 28390 + }, + { + "epoch": 1.04, + "learning_rate": 9.621342512908778e-08, + "loss": 0.6962, + "step": 28391 + }, + { + "epoch": 1.04, + "learning_rate": 9.620976306441572e-08, + "loss": 0.6939, + "step": 28392 + }, + { + "epoch": 1.04, + "learning_rate": 9.620610099974366e-08, + "loss": 0.6947, + "step": 28393 + }, + { + "epoch": 1.04, + "learning_rate": 9.620243893507158e-08, + "loss": 0.6977, + "step": 28394 + }, + { + "epoch": 1.04, + "learning_rate": 9.619877687039952e-08, + "loss": 0.6774, + "step": 28395 + }, + { + "epoch": 1.04, + "learning_rate": 9.619511480572746e-08, + "loss": 0.705, + "step": 28396 + }, + { + "epoch": 1.04, + "learning_rate": 9.61914527410554e-08, + "loss": 0.7005, + "step": 28397 + }, + { + "epoch": 1.04, + "learning_rate": 9.618779067638334e-08, + "loss": 0.6885, + "step": 28398 + }, + { + "epoch": 1.04, + "learning_rate": 9.618412861171128e-08, + "loss": 0.6953, + "step": 28399 + }, + { + "epoch": 1.04, + "learning_rate": 9.618046654703922e-08, + "loss": 0.6894, + "step": 28400 + }, + { + "epoch": 1.04, + "learning_rate": 9.617680448236716e-08, + "loss": 0.6897, + "step": 28401 + }, + { + "epoch": 1.04, + "learning_rate": 9.61731424176951e-08, + "loss": 0.6912, + "step": 28402 + }, + { + "epoch": 1.04, + "learning_rate": 9.616948035302303e-08, + "loss": 0.6898, + "step": 28403 + }, + { + "epoch": 1.04, + "learning_rate": 9.616581828835097e-08, + "loss": 0.6975, + "step": 28404 + }, + { + "epoch": 1.04, + "learning_rate": 9.61621562236789e-08, + "loss": 0.6853, + "step": 28405 + }, + { + "epoch": 1.04, + "learning_rate": 9.615849415900684e-08, + "loss": 0.6934, + "step": 28406 + }, + { + "epoch": 1.04, + "learning_rate": 9.615483209433478e-08, + "loss": 0.6814, + "step": 28407 + }, + { + "epoch": 1.04, + "learning_rate": 9.615117002966272e-08, + "loss": 0.6927, + "step": 28408 + }, + { + "epoch": 1.04, + "learning_rate": 9.614750796499066e-08, + "loss": 0.6956, + "step": 28409 + }, + { + "epoch": 1.04, + "learning_rate": 9.61438459003186e-08, + "loss": 0.6998, + "step": 28410 + }, + { + "epoch": 1.04, + "learning_rate": 9.614018383564654e-08, + "loss": 0.6927, + "step": 28411 + }, + { + "epoch": 1.04, + "learning_rate": 9.613652177097446e-08, + "loss": 0.6948, + "step": 28412 + }, + { + "epoch": 1.04, + "learning_rate": 9.61328597063024e-08, + "loss": 0.6969, + "step": 28413 + }, + { + "epoch": 1.04, + "learning_rate": 9.612919764163034e-08, + "loss": 0.7065, + "step": 28414 + }, + { + "epoch": 1.04, + "learning_rate": 9.612553557695828e-08, + "loss": 0.6906, + "step": 28415 + }, + { + "epoch": 1.04, + "learning_rate": 9.612187351228622e-08, + "loss": 0.6984, + "step": 28416 + }, + { + "epoch": 1.04, + "learning_rate": 9.611821144761416e-08, + "loss": 0.6901, + "step": 28417 + }, + { + "epoch": 1.04, + "learning_rate": 9.61145493829421e-08, + "loss": 0.6851, + "step": 28418 + }, + { + "epoch": 1.04, + "learning_rate": 9.611088731827004e-08, + "loss": 0.7031, + "step": 28419 + }, + { + "epoch": 1.04, + "learning_rate": 9.610722525359798e-08, + "loss": 0.6897, + "step": 28420 + }, + { + "epoch": 1.04, + "learning_rate": 9.610356318892591e-08, + "loss": 0.6898, + "step": 28421 + }, + { + "epoch": 1.04, + "learning_rate": 9.609990112425385e-08, + "loss": 0.7136, + "step": 28422 + }, + { + "epoch": 1.04, + "learning_rate": 9.609623905958179e-08, + "loss": 0.6905, + "step": 28423 + }, + { + "epoch": 1.04, + "learning_rate": 9.609257699490973e-08, + "loss": 0.695, + "step": 28424 + }, + { + "epoch": 1.04, + "learning_rate": 9.608891493023767e-08, + "loss": 0.7078, + "step": 28425 + }, + { + "epoch": 1.04, + "learning_rate": 9.608525286556561e-08, + "loss": 0.6904, + "step": 28426 + }, + { + "epoch": 1.04, + "learning_rate": 9.608159080089354e-08, + "loss": 0.6887, + "step": 28427 + }, + { + "epoch": 1.04, + "learning_rate": 9.607792873622148e-08, + "loss": 0.6954, + "step": 28428 + }, + { + "epoch": 1.04, + "learning_rate": 9.607426667154942e-08, + "loss": 0.6994, + "step": 28429 + }, + { + "epoch": 1.04, + "learning_rate": 9.607060460687734e-08, + "loss": 0.6865, + "step": 28430 + }, + { + "epoch": 1.04, + "learning_rate": 9.606694254220528e-08, + "loss": 0.6906, + "step": 28431 + }, + { + "epoch": 1.04, + "learning_rate": 9.606328047753322e-08, + "loss": 0.6973, + "step": 28432 + }, + { + "epoch": 1.04, + "learning_rate": 9.605961841286116e-08, + "loss": 0.6987, + "step": 28433 + }, + { + "epoch": 1.04, + "learning_rate": 9.60559563481891e-08, + "loss": 0.6851, + "step": 28434 + }, + { + "epoch": 1.04, + "learning_rate": 9.605229428351704e-08, + "loss": 0.6947, + "step": 28435 + }, + { + "epoch": 1.04, + "learning_rate": 9.604863221884498e-08, + "loss": 0.6913, + "step": 28436 + }, + { + "epoch": 1.04, + "learning_rate": 9.604497015417292e-08, + "loss": 0.6835, + "step": 28437 + }, + { + "epoch": 1.04, + "learning_rate": 9.604130808950086e-08, + "loss": 0.6934, + "step": 28438 + }, + { + "epoch": 1.04, + "learning_rate": 9.603764602482879e-08, + "loss": 0.7053, + "step": 28439 + }, + { + "epoch": 1.04, + "learning_rate": 9.603398396015673e-08, + "loss": 0.6898, + "step": 28440 + }, + { + "epoch": 1.04, + "learning_rate": 9.603032189548467e-08, + "loss": 0.6931, + "step": 28441 + }, + { + "epoch": 1.04, + "learning_rate": 9.602665983081261e-08, + "loss": 0.6851, + "step": 28442 + }, + { + "epoch": 1.04, + "learning_rate": 9.602299776614055e-08, + "loss": 0.6873, + "step": 28443 + }, + { + "epoch": 1.04, + "learning_rate": 9.601933570146849e-08, + "loss": 0.7117, + "step": 28444 + }, + { + "epoch": 1.04, + "learning_rate": 9.601567363679643e-08, + "loss": 0.6877, + "step": 28445 + }, + { + "epoch": 1.04, + "learning_rate": 9.601201157212437e-08, + "loss": 0.6866, + "step": 28446 + }, + { + "epoch": 1.04, + "learning_rate": 9.600834950745231e-08, + "loss": 0.6939, + "step": 28447 + }, + { + "epoch": 1.04, + "learning_rate": 9.600468744278023e-08, + "loss": 0.6997, + "step": 28448 + }, + { + "epoch": 1.04, + "learning_rate": 9.600102537810817e-08, + "loss": 0.7006, + "step": 28449 + }, + { + "epoch": 1.04, + "learning_rate": 9.599736331343611e-08, + "loss": 0.6956, + "step": 28450 + }, + { + "epoch": 1.04, + "learning_rate": 9.599370124876405e-08, + "loss": 0.6906, + "step": 28451 + }, + { + "epoch": 1.04, + "learning_rate": 9.599003918409198e-08, + "loss": 0.6844, + "step": 28452 + }, + { + "epoch": 1.04, + "learning_rate": 9.598637711941992e-08, + "loss": 0.6955, + "step": 28453 + }, + { + "epoch": 1.04, + "learning_rate": 9.598271505474786e-08, + "loss": 0.694, + "step": 28454 + }, + { + "epoch": 1.04, + "learning_rate": 9.59790529900758e-08, + "loss": 0.6897, + "step": 28455 + }, + { + "epoch": 1.04, + "learning_rate": 9.597539092540374e-08, + "loss": 0.6954, + "step": 28456 + }, + { + "epoch": 1.04, + "learning_rate": 9.597172886073167e-08, + "loss": 0.6992, + "step": 28457 + }, + { + "epoch": 1.04, + "learning_rate": 9.596806679605961e-08, + "loss": 0.6925, + "step": 28458 + }, + { + "epoch": 1.04, + "learning_rate": 9.596440473138755e-08, + "loss": 0.6904, + "step": 28459 + }, + { + "epoch": 1.04, + "learning_rate": 9.596074266671549e-08, + "loss": 0.6868, + "step": 28460 + }, + { + "epoch": 1.04, + "learning_rate": 9.595708060204343e-08, + "loss": 0.6989, + "step": 28461 + }, + { + "epoch": 1.04, + "learning_rate": 9.595341853737137e-08, + "loss": 0.7008, + "step": 28462 + }, + { + "epoch": 1.04, + "learning_rate": 9.59497564726993e-08, + "loss": 0.6899, + "step": 28463 + }, + { + "epoch": 1.04, + "learning_rate": 9.594609440802725e-08, + "loss": 0.6995, + "step": 28464 + }, + { + "epoch": 1.04, + "learning_rate": 9.594243234335519e-08, + "loss": 0.6933, + "step": 28465 + }, + { + "epoch": 1.04, + "learning_rate": 9.593877027868311e-08, + "loss": 0.6895, + "step": 28466 + }, + { + "epoch": 1.04, + "learning_rate": 9.593510821401105e-08, + "loss": 0.6888, + "step": 28467 + }, + { + "epoch": 1.04, + "learning_rate": 9.593144614933899e-08, + "loss": 0.6863, + "step": 28468 + }, + { + "epoch": 1.04, + "learning_rate": 9.592778408466693e-08, + "loss": 0.6974, + "step": 28469 + }, + { + "epoch": 1.04, + "learning_rate": 9.592412201999487e-08, + "loss": 0.6912, + "step": 28470 + }, + { + "epoch": 1.04, + "learning_rate": 9.592045995532281e-08, + "loss": 0.7025, + "step": 28471 + }, + { + "epoch": 1.04, + "learning_rate": 9.591679789065075e-08, + "loss": 0.7014, + "step": 28472 + }, + { + "epoch": 1.04, + "learning_rate": 9.591313582597868e-08, + "loss": 0.6756, + "step": 28473 + }, + { + "epoch": 1.04, + "learning_rate": 9.590947376130662e-08, + "loss": 0.6975, + "step": 28474 + }, + { + "epoch": 1.04, + "learning_rate": 9.590581169663456e-08, + "loss": 0.6893, + "step": 28475 + }, + { + "epoch": 1.04, + "learning_rate": 9.590214963196249e-08, + "loss": 0.6826, + "step": 28476 + }, + { + "epoch": 1.04, + "learning_rate": 9.589848756729043e-08, + "loss": 0.6929, + "step": 28477 + }, + { + "epoch": 1.04, + "learning_rate": 9.589482550261837e-08, + "loss": 0.6947, + "step": 28478 + }, + { + "epoch": 1.04, + "learning_rate": 9.58911634379463e-08, + "loss": 0.679, + "step": 28479 + }, + { + "epoch": 1.04, + "learning_rate": 9.588750137327425e-08, + "loss": 0.6986, + "step": 28480 + }, + { + "epoch": 1.04, + "learning_rate": 9.588383930860219e-08, + "loss": 0.6816, + "step": 28481 + }, + { + "epoch": 1.04, + "learning_rate": 9.588017724393013e-08, + "loss": 0.6945, + "step": 28482 + }, + { + "epoch": 1.04, + "learning_rate": 9.587651517925807e-08, + "loss": 0.6925, + "step": 28483 + }, + { + "epoch": 1.04, + "learning_rate": 9.587285311458599e-08, + "loss": 0.7014, + "step": 28484 + }, + { + "epoch": 1.04, + "learning_rate": 9.586919104991393e-08, + "loss": 0.6962, + "step": 28485 + }, + { + "epoch": 1.04, + "learning_rate": 9.586552898524187e-08, + "loss": 0.6995, + "step": 28486 + }, + { + "epoch": 1.04, + "learning_rate": 9.586186692056981e-08, + "loss": 0.7039, + "step": 28487 + }, + { + "epoch": 1.04, + "learning_rate": 9.585820485589775e-08, + "loss": 0.6825, + "step": 28488 + }, + { + "epoch": 1.04, + "learning_rate": 9.585454279122569e-08, + "loss": 0.692, + "step": 28489 + }, + { + "epoch": 1.04, + "learning_rate": 9.585088072655363e-08, + "loss": 0.7083, + "step": 28490 + }, + { + "epoch": 1.04, + "learning_rate": 9.584721866188157e-08, + "loss": 0.679, + "step": 28491 + }, + { + "epoch": 1.04, + "learning_rate": 9.584355659720951e-08, + "loss": 0.6979, + "step": 28492 + }, + { + "epoch": 1.04, + "learning_rate": 9.583989453253744e-08, + "loss": 0.6936, + "step": 28493 + }, + { + "epoch": 1.04, + "learning_rate": 9.583623246786538e-08, + "loss": 0.7067, + "step": 28494 + }, + { + "epoch": 1.04, + "learning_rate": 9.583257040319332e-08, + "loss": 0.7074, + "step": 28495 + }, + { + "epoch": 1.04, + "learning_rate": 9.582890833852126e-08, + "loss": 0.6889, + "step": 28496 + }, + { + "epoch": 1.04, + "learning_rate": 9.58252462738492e-08, + "loss": 0.6813, + "step": 28497 + }, + { + "epoch": 1.04, + "learning_rate": 9.582158420917712e-08, + "loss": 0.6994, + "step": 28498 + }, + { + "epoch": 1.04, + "learning_rate": 9.581792214450506e-08, + "loss": 0.6936, + "step": 28499 + }, + { + "epoch": 1.04, + "learning_rate": 9.5814260079833e-08, + "loss": 0.6921, + "step": 28500 + }, + { + "epoch": 1.04, + "learning_rate": 9.581059801516094e-08, + "loss": 0.6967, + "step": 28501 + }, + { + "epoch": 1.04, + "learning_rate": 9.580693595048887e-08, + "loss": 0.6932, + "step": 28502 + }, + { + "epoch": 1.04, + "learning_rate": 9.580327388581681e-08, + "loss": 0.6972, + "step": 28503 + }, + { + "epoch": 1.04, + "learning_rate": 9.579961182114475e-08, + "loss": 0.6906, + "step": 28504 + }, + { + "epoch": 1.04, + "learning_rate": 9.579594975647269e-08, + "loss": 0.7013, + "step": 28505 + }, + { + "epoch": 1.04, + "learning_rate": 9.579228769180063e-08, + "loss": 0.6948, + "step": 28506 + }, + { + "epoch": 1.04, + "learning_rate": 9.578862562712857e-08, + "loss": 0.7029, + "step": 28507 + }, + { + "epoch": 1.04, + "learning_rate": 9.578496356245651e-08, + "loss": 0.6979, + "step": 28508 + }, + { + "epoch": 1.04, + "learning_rate": 9.578130149778445e-08, + "loss": 0.6913, + "step": 28509 + }, + { + "epoch": 1.04, + "learning_rate": 9.577763943311239e-08, + "loss": 0.6932, + "step": 28510 + }, + { + "epoch": 1.04, + "learning_rate": 9.577397736844032e-08, + "loss": 0.7125, + "step": 28511 + }, + { + "epoch": 1.04, + "learning_rate": 9.577031530376826e-08, + "loss": 0.6958, + "step": 28512 + }, + { + "epoch": 1.04, + "learning_rate": 9.57666532390962e-08, + "loss": 0.6957, + "step": 28513 + }, + { + "epoch": 1.04, + "learning_rate": 9.576299117442414e-08, + "loss": 0.6868, + "step": 28514 + }, + { + "epoch": 1.04, + "learning_rate": 9.575932910975208e-08, + "loss": 0.6911, + "step": 28515 + }, + { + "epoch": 1.04, + "learning_rate": 9.575566704508002e-08, + "loss": 0.69, + "step": 28516 + }, + { + "epoch": 1.04, + "learning_rate": 9.575200498040796e-08, + "loss": 0.689, + "step": 28517 + }, + { + "epoch": 1.04, + "learning_rate": 9.57483429157359e-08, + "loss": 0.6968, + "step": 28518 + }, + { + "epoch": 1.04, + "learning_rate": 9.574468085106384e-08, + "loss": 0.7045, + "step": 28519 + }, + { + "epoch": 1.04, + "learning_rate": 9.574101878639176e-08, + "loss": 0.6921, + "step": 28520 + }, + { + "epoch": 1.04, + "learning_rate": 9.57373567217197e-08, + "loss": 0.6864, + "step": 28521 + }, + { + "epoch": 1.04, + "learning_rate": 9.573369465704764e-08, + "loss": 0.6934, + "step": 28522 + }, + { + "epoch": 1.04, + "learning_rate": 9.573003259237557e-08, + "loss": 0.6757, + "step": 28523 + }, + { + "epoch": 1.04, + "learning_rate": 9.572637052770351e-08, + "loss": 0.6852, + "step": 28524 + }, + { + "epoch": 1.04, + "learning_rate": 9.572270846303145e-08, + "loss": 0.6916, + "step": 28525 + }, + { + "epoch": 1.04, + "learning_rate": 9.571904639835939e-08, + "loss": 0.7043, + "step": 28526 + }, + { + "epoch": 1.04, + "learning_rate": 9.571538433368733e-08, + "loss": 0.6945, + "step": 28527 + }, + { + "epoch": 1.04, + "learning_rate": 9.571172226901527e-08, + "loss": 0.6949, + "step": 28528 + }, + { + "epoch": 1.04, + "learning_rate": 9.57080602043432e-08, + "loss": 0.6948, + "step": 28529 + }, + { + "epoch": 1.04, + "learning_rate": 9.570439813967114e-08, + "loss": 0.6783, + "step": 28530 + }, + { + "epoch": 1.04, + "learning_rate": 9.570073607499908e-08, + "loss": 0.7035, + "step": 28531 + }, + { + "epoch": 1.04, + "learning_rate": 9.569707401032702e-08, + "loss": 0.6778, + "step": 28532 + }, + { + "epoch": 1.04, + "learning_rate": 9.569341194565496e-08, + "loss": 0.6926, + "step": 28533 + }, + { + "epoch": 1.04, + "learning_rate": 9.56897498809829e-08, + "loss": 0.6926, + "step": 28534 + }, + { + "epoch": 1.04, + "learning_rate": 9.568608781631084e-08, + "loss": 0.6933, + "step": 28535 + }, + { + "epoch": 1.04, + "learning_rate": 9.568242575163878e-08, + "loss": 0.7186, + "step": 28536 + }, + { + "epoch": 1.04, + "learning_rate": 9.567876368696672e-08, + "loss": 0.6942, + "step": 28537 + }, + { + "epoch": 1.04, + "learning_rate": 9.567510162229464e-08, + "loss": 0.6933, + "step": 28538 + }, + { + "epoch": 1.04, + "learning_rate": 9.567143955762258e-08, + "loss": 0.7012, + "step": 28539 + }, + { + "epoch": 1.04, + "learning_rate": 9.566777749295052e-08, + "loss": 0.7009, + "step": 28540 + }, + { + "epoch": 1.04, + "learning_rate": 9.566411542827846e-08, + "loss": 0.6949, + "step": 28541 + }, + { + "epoch": 1.04, + "learning_rate": 9.56604533636064e-08, + "loss": 0.6957, + "step": 28542 + }, + { + "epoch": 1.04, + "learning_rate": 9.565679129893434e-08, + "loss": 0.7022, + "step": 28543 + }, + { + "epoch": 1.04, + "learning_rate": 9.565312923426227e-08, + "loss": 0.6998, + "step": 28544 + }, + { + "epoch": 1.04, + "learning_rate": 9.564946716959021e-08, + "loss": 0.6966, + "step": 28545 + }, + { + "epoch": 1.04, + "learning_rate": 9.564580510491815e-08, + "loss": 0.6928, + "step": 28546 + }, + { + "epoch": 1.04, + "learning_rate": 9.564214304024608e-08, + "loss": 0.6856, + "step": 28547 + }, + { + "epoch": 1.04, + "learning_rate": 9.563848097557402e-08, + "loss": 0.6912, + "step": 28548 + }, + { + "epoch": 1.04, + "learning_rate": 9.563481891090196e-08, + "loss": 0.6824, + "step": 28549 + }, + { + "epoch": 1.04, + "learning_rate": 9.56311568462299e-08, + "loss": 0.6936, + "step": 28550 + }, + { + "epoch": 1.04, + "learning_rate": 9.562749478155784e-08, + "loss": 0.6851, + "step": 28551 + }, + { + "epoch": 1.04, + "learning_rate": 9.562383271688577e-08, + "loss": 0.701, + "step": 28552 + }, + { + "epoch": 1.04, + "learning_rate": 9.562017065221371e-08, + "loss": 0.6858, + "step": 28553 + }, + { + "epoch": 1.04, + "learning_rate": 9.561650858754165e-08, + "loss": 0.6863, + "step": 28554 + }, + { + "epoch": 1.04, + "learning_rate": 9.56128465228696e-08, + "loss": 0.6934, + "step": 28555 + }, + { + "epoch": 1.04, + "learning_rate": 9.560918445819752e-08, + "loss": 0.7007, + "step": 28556 + }, + { + "epoch": 1.04, + "learning_rate": 9.560552239352546e-08, + "loss": 0.7006, + "step": 28557 + }, + { + "epoch": 1.04, + "learning_rate": 9.56018603288534e-08, + "loss": 0.6924, + "step": 28558 + }, + { + "epoch": 1.04, + "learning_rate": 9.559819826418134e-08, + "loss": 0.6835, + "step": 28559 + }, + { + "epoch": 1.04, + "learning_rate": 9.559453619950928e-08, + "loss": 0.7031, + "step": 28560 + }, + { + "epoch": 1.04, + "learning_rate": 9.559087413483722e-08, + "loss": 0.7026, + "step": 28561 + }, + { + "epoch": 1.05, + "learning_rate": 9.558721207016516e-08, + "loss": 0.6929, + "step": 28562 + }, + { + "epoch": 1.05, + "learning_rate": 9.55835500054931e-08, + "loss": 0.6806, + "step": 28563 + }, + { + "epoch": 1.05, + "learning_rate": 9.557988794082104e-08, + "loss": 0.6909, + "step": 28564 + }, + { + "epoch": 1.05, + "learning_rate": 9.557622587614897e-08, + "loss": 0.7057, + "step": 28565 + }, + { + "epoch": 1.05, + "learning_rate": 9.557256381147691e-08, + "loss": 0.7029, + "step": 28566 + }, + { + "epoch": 1.05, + "learning_rate": 9.556890174680485e-08, + "loss": 0.699, + "step": 28567 + }, + { + "epoch": 1.05, + "learning_rate": 9.556523968213279e-08, + "loss": 0.695, + "step": 28568 + }, + { + "epoch": 1.05, + "learning_rate": 9.556157761746071e-08, + "loss": 0.6897, + "step": 28569 + }, + { + "epoch": 1.05, + "learning_rate": 9.555791555278865e-08, + "loss": 0.6945, + "step": 28570 + }, + { + "epoch": 1.05, + "learning_rate": 9.55542534881166e-08, + "loss": 0.6964, + "step": 28571 + }, + { + "epoch": 1.05, + "learning_rate": 9.555059142344453e-08, + "loss": 0.6971, + "step": 28572 + }, + { + "epoch": 1.05, + "learning_rate": 9.554692935877247e-08, + "loss": 0.6852, + "step": 28573 + }, + { + "epoch": 1.05, + "learning_rate": 9.55432672941004e-08, + "loss": 0.696, + "step": 28574 + }, + { + "epoch": 1.05, + "learning_rate": 9.553960522942834e-08, + "loss": 0.6983, + "step": 28575 + }, + { + "epoch": 1.05, + "learning_rate": 9.553594316475628e-08, + "loss": 0.6813, + "step": 28576 + }, + { + "epoch": 1.05, + "learning_rate": 9.553228110008422e-08, + "loss": 0.6928, + "step": 28577 + }, + { + "epoch": 1.05, + "learning_rate": 9.552861903541216e-08, + "loss": 0.6922, + "step": 28578 + }, + { + "epoch": 1.05, + "learning_rate": 9.55249569707401e-08, + "loss": 0.6937, + "step": 28579 + }, + { + "epoch": 1.05, + "learning_rate": 9.552129490606804e-08, + "loss": 0.6898, + "step": 28580 + }, + { + "epoch": 1.05, + "learning_rate": 9.551763284139598e-08, + "loss": 0.7014, + "step": 28581 + }, + { + "epoch": 1.05, + "learning_rate": 9.551397077672392e-08, + "loss": 0.6973, + "step": 28582 + }, + { + "epoch": 1.05, + "learning_rate": 9.551030871205185e-08, + "loss": 0.7019, + "step": 28583 + }, + { + "epoch": 1.05, + "learning_rate": 9.550664664737979e-08, + "loss": 0.6939, + "step": 28584 + }, + { + "epoch": 1.05, + "learning_rate": 9.550298458270773e-08, + "loss": 0.6989, + "step": 28585 + }, + { + "epoch": 1.05, + "learning_rate": 9.549932251803567e-08, + "loss": 0.695, + "step": 28586 + }, + { + "epoch": 1.05, + "learning_rate": 9.54956604533636e-08, + "loss": 0.6998, + "step": 28587 + }, + { + "epoch": 1.05, + "learning_rate": 9.549199838869155e-08, + "loss": 0.7027, + "step": 28588 + }, + { + "epoch": 1.05, + "learning_rate": 9.548833632401949e-08, + "loss": 0.6817, + "step": 28589 + }, + { + "epoch": 1.05, + "learning_rate": 9.548467425934743e-08, + "loss": 0.6942, + "step": 28590 + }, + { + "epoch": 1.05, + "learning_rate": 9.548101219467535e-08, + "loss": 0.6888, + "step": 28591 + }, + { + "epoch": 1.05, + "learning_rate": 9.547735013000329e-08, + "loss": 0.689, + "step": 28592 + }, + { + "epoch": 1.05, + "learning_rate": 9.547368806533122e-08, + "loss": 0.6839, + "step": 28593 + }, + { + "epoch": 1.05, + "learning_rate": 9.547002600065916e-08, + "loss": 0.7089, + "step": 28594 + }, + { + "epoch": 1.05, + "learning_rate": 9.54663639359871e-08, + "loss": 0.6878, + "step": 28595 + }, + { + "epoch": 1.05, + "learning_rate": 9.546270187131504e-08, + "loss": 0.6957, + "step": 28596 + }, + { + "epoch": 1.05, + "learning_rate": 9.545903980664298e-08, + "loss": 0.6933, + "step": 28597 + }, + { + "epoch": 1.05, + "learning_rate": 9.545537774197092e-08, + "loss": 0.683, + "step": 28598 + }, + { + "epoch": 1.05, + "learning_rate": 9.545171567729886e-08, + "loss": 0.6923, + "step": 28599 + }, + { + "epoch": 1.05, + "learning_rate": 9.54480536126268e-08, + "loss": 0.6942, + "step": 28600 + }, + { + "epoch": 1.05, + "learning_rate": 9.544439154795473e-08, + "loss": 0.6913, + "step": 28601 + }, + { + "epoch": 1.05, + "learning_rate": 9.544072948328267e-08, + "loss": 0.6926, + "step": 28602 + }, + { + "epoch": 1.05, + "learning_rate": 9.54370674186106e-08, + "loss": 0.6906, + "step": 28603 + }, + { + "epoch": 1.05, + "learning_rate": 9.543340535393855e-08, + "loss": 0.6916, + "step": 28604 + }, + { + "epoch": 1.05, + "learning_rate": 9.542974328926649e-08, + "loss": 0.693, + "step": 28605 + }, + { + "epoch": 1.05, + "learning_rate": 9.542608122459443e-08, + "loss": 0.6966, + "step": 28606 + }, + { + "epoch": 1.05, + "learning_rate": 9.542241915992237e-08, + "loss": 0.6988, + "step": 28607 + }, + { + "epoch": 1.05, + "learning_rate": 9.54187570952503e-08, + "loss": 0.6931, + "step": 28608 + }, + { + "epoch": 1.05, + "learning_rate": 9.541509503057825e-08, + "loss": 0.7, + "step": 28609 + }, + { + "epoch": 1.05, + "learning_rate": 9.541143296590617e-08, + "loss": 0.6962, + "step": 28610 + }, + { + "epoch": 1.05, + "learning_rate": 9.540777090123411e-08, + "loss": 0.6925, + "step": 28611 + }, + { + "epoch": 1.05, + "learning_rate": 9.540410883656205e-08, + "loss": 0.6907, + "step": 28612 + }, + { + "epoch": 1.05, + "learning_rate": 9.540044677188999e-08, + "loss": 0.6817, + "step": 28613 + }, + { + "epoch": 1.05, + "learning_rate": 9.539678470721793e-08, + "loss": 0.6931, + "step": 28614 + }, + { + "epoch": 1.05, + "learning_rate": 9.539312264254586e-08, + "loss": 0.6854, + "step": 28615 + }, + { + "epoch": 1.05, + "learning_rate": 9.53894605778738e-08, + "loss": 0.6934, + "step": 28616 + }, + { + "epoch": 1.05, + "learning_rate": 9.538579851320174e-08, + "loss": 0.6983, + "step": 28617 + }, + { + "epoch": 1.05, + "learning_rate": 9.538213644852968e-08, + "loss": 0.6894, + "step": 28618 + }, + { + "epoch": 1.05, + "learning_rate": 9.53784743838576e-08, + "loss": 0.69, + "step": 28619 + }, + { + "epoch": 1.05, + "learning_rate": 9.537481231918554e-08, + "loss": 0.7149, + "step": 28620 + }, + { + "epoch": 1.05, + "learning_rate": 9.537115025451348e-08, + "loss": 0.6862, + "step": 28621 + }, + { + "epoch": 1.05, + "learning_rate": 9.536748818984142e-08, + "loss": 0.6961, + "step": 28622 + }, + { + "epoch": 1.05, + "learning_rate": 9.536382612516936e-08, + "loss": 0.6992, + "step": 28623 + }, + { + "epoch": 1.05, + "learning_rate": 9.53601640604973e-08, + "loss": 0.6939, + "step": 28624 + }, + { + "epoch": 1.05, + "learning_rate": 9.535650199582524e-08, + "loss": 0.6914, + "step": 28625 + }, + { + "epoch": 1.05, + "learning_rate": 9.535283993115318e-08, + "loss": 0.6918, + "step": 28626 + }, + { + "epoch": 1.05, + "learning_rate": 9.534917786648112e-08, + "loss": 0.6963, + "step": 28627 + }, + { + "epoch": 1.05, + "learning_rate": 9.534551580180905e-08, + "loss": 0.6902, + "step": 28628 + }, + { + "epoch": 1.05, + "learning_rate": 9.534185373713699e-08, + "loss": 0.6851, + "step": 28629 + }, + { + "epoch": 1.05, + "learning_rate": 9.533819167246493e-08, + "loss": 0.6866, + "step": 28630 + }, + { + "epoch": 1.05, + "learning_rate": 9.533452960779287e-08, + "loss": 0.696, + "step": 28631 + }, + { + "epoch": 1.05, + "learning_rate": 9.533086754312081e-08, + "loss": 0.6957, + "step": 28632 + }, + { + "epoch": 1.05, + "learning_rate": 9.532720547844875e-08, + "loss": 0.6923, + "step": 28633 + }, + { + "epoch": 1.05, + "learning_rate": 9.532354341377669e-08, + "loss": 0.698, + "step": 28634 + }, + { + "epoch": 1.05, + "learning_rate": 9.531988134910463e-08, + "loss": 0.6793, + "step": 28635 + }, + { + "epoch": 1.05, + "learning_rate": 9.531621928443257e-08, + "loss": 0.69, + "step": 28636 + }, + { + "epoch": 1.05, + "learning_rate": 9.53125572197605e-08, + "loss": 0.6882, + "step": 28637 + }, + { + "epoch": 1.05, + "learning_rate": 9.530889515508844e-08, + "loss": 0.6886, + "step": 28638 + }, + { + "epoch": 1.05, + "learning_rate": 9.530523309041638e-08, + "loss": 0.6926, + "step": 28639 + }, + { + "epoch": 1.05, + "learning_rate": 9.53015710257443e-08, + "loss": 0.6904, + "step": 28640 + }, + { + "epoch": 1.05, + "learning_rate": 9.529790896107224e-08, + "loss": 0.6841, + "step": 28641 + }, + { + "epoch": 1.05, + "learning_rate": 9.529424689640018e-08, + "loss": 0.6961, + "step": 28642 + }, + { + "epoch": 1.05, + "learning_rate": 9.529058483172812e-08, + "loss": 0.6949, + "step": 28643 + }, + { + "epoch": 1.05, + "learning_rate": 9.528692276705606e-08, + "loss": 0.6914, + "step": 28644 + }, + { + "epoch": 1.05, + "learning_rate": 9.5283260702384e-08, + "loss": 0.6855, + "step": 28645 + }, + { + "epoch": 1.05, + "learning_rate": 9.527959863771193e-08, + "loss": 0.6925, + "step": 28646 + }, + { + "epoch": 1.05, + "learning_rate": 9.527593657303987e-08, + "loss": 0.7003, + "step": 28647 + }, + { + "epoch": 1.05, + "learning_rate": 9.527227450836781e-08, + "loss": 0.6952, + "step": 28648 + }, + { + "epoch": 1.05, + "learning_rate": 9.526861244369575e-08, + "loss": 0.7026, + "step": 28649 + }, + { + "epoch": 1.05, + "learning_rate": 9.526495037902369e-08, + "loss": 0.6894, + "step": 28650 + }, + { + "epoch": 1.05, + "learning_rate": 9.526128831435163e-08, + "loss": 0.6891, + "step": 28651 + }, + { + "epoch": 1.05, + "learning_rate": 9.525762624967957e-08, + "loss": 0.6921, + "step": 28652 + }, + { + "epoch": 1.05, + "learning_rate": 9.525396418500751e-08, + "loss": 0.6815, + "step": 28653 + }, + { + "epoch": 1.05, + "learning_rate": 9.525030212033545e-08, + "loss": 0.6995, + "step": 28654 + }, + { + "epoch": 1.05, + "learning_rate": 9.524664005566338e-08, + "loss": 0.6901, + "step": 28655 + }, + { + "epoch": 1.05, + "learning_rate": 9.524297799099132e-08, + "loss": 0.6785, + "step": 28656 + }, + { + "epoch": 1.05, + "learning_rate": 9.523931592631926e-08, + "loss": 0.6961, + "step": 28657 + }, + { + "epoch": 1.05, + "learning_rate": 9.52356538616472e-08, + "loss": 0.7032, + "step": 28658 + }, + { + "epoch": 1.05, + "learning_rate": 9.523199179697514e-08, + "loss": 0.6895, + "step": 28659 + }, + { + "epoch": 1.05, + "learning_rate": 9.522832973230308e-08, + "loss": 0.7, + "step": 28660 + }, + { + "epoch": 1.05, + "learning_rate": 9.522466766763102e-08, + "loss": 0.689, + "step": 28661 + }, + { + "epoch": 1.05, + "learning_rate": 9.522100560295894e-08, + "loss": 0.6902, + "step": 28662 + }, + { + "epoch": 1.05, + "learning_rate": 9.521734353828688e-08, + "loss": 0.6911, + "step": 28663 + }, + { + "epoch": 1.05, + "learning_rate": 9.521368147361481e-08, + "loss": 0.6938, + "step": 28664 + }, + { + "epoch": 1.05, + "learning_rate": 9.521001940894275e-08, + "loss": 0.6902, + "step": 28665 + }, + { + "epoch": 1.05, + "learning_rate": 9.520635734427069e-08, + "loss": 0.6987, + "step": 28666 + }, + { + "epoch": 1.05, + "learning_rate": 9.520269527959863e-08, + "loss": 0.6953, + "step": 28667 + }, + { + "epoch": 1.05, + "learning_rate": 9.519903321492657e-08, + "loss": 0.6905, + "step": 28668 + }, + { + "epoch": 1.05, + "learning_rate": 9.519537115025451e-08, + "loss": 0.6845, + "step": 28669 + }, + { + "epoch": 1.05, + "learning_rate": 9.519170908558245e-08, + "loss": 0.6924, + "step": 28670 + }, + { + "epoch": 1.05, + "learning_rate": 9.518804702091039e-08, + "loss": 0.6832, + "step": 28671 + }, + { + "epoch": 1.05, + "learning_rate": 9.518438495623833e-08, + "loss": 0.7002, + "step": 28672 + }, + { + "epoch": 1.05, + "learning_rate": 9.518072289156625e-08, + "loss": 0.6905, + "step": 28673 + }, + { + "epoch": 1.05, + "learning_rate": 9.51770608268942e-08, + "loss": 0.696, + "step": 28674 + }, + { + "epoch": 1.05, + "learning_rate": 9.517339876222213e-08, + "loss": 0.6864, + "step": 28675 + }, + { + "epoch": 1.05, + "learning_rate": 9.516973669755007e-08, + "loss": 0.7005, + "step": 28676 + }, + { + "epoch": 1.05, + "learning_rate": 9.516607463287801e-08, + "loss": 0.7079, + "step": 28677 + }, + { + "epoch": 1.05, + "learning_rate": 9.516241256820595e-08, + "loss": 0.6891, + "step": 28678 + }, + { + "epoch": 1.05, + "learning_rate": 9.51587505035339e-08, + "loss": 0.6962, + "step": 28679 + }, + { + "epoch": 1.05, + "learning_rate": 9.515508843886183e-08, + "loss": 0.6867, + "step": 28680 + }, + { + "epoch": 1.05, + "learning_rate": 9.515142637418977e-08, + "loss": 0.6901, + "step": 28681 + }, + { + "epoch": 1.05, + "learning_rate": 9.51477643095177e-08, + "loss": 0.6843, + "step": 28682 + }, + { + "epoch": 1.05, + "learning_rate": 9.514410224484564e-08, + "loss": 0.7011, + "step": 28683 + }, + { + "epoch": 1.05, + "learning_rate": 9.514044018017358e-08, + "loss": 0.6762, + "step": 28684 + }, + { + "epoch": 1.05, + "learning_rate": 9.513677811550152e-08, + "loss": 0.6926, + "step": 28685 + }, + { + "epoch": 1.05, + "learning_rate": 9.513311605082945e-08, + "loss": 0.6964, + "step": 28686 + }, + { + "epoch": 1.05, + "learning_rate": 9.512945398615739e-08, + "loss": 0.6938, + "step": 28687 + }, + { + "epoch": 1.05, + "learning_rate": 9.512579192148533e-08, + "loss": 0.7007, + "step": 28688 + }, + { + "epoch": 1.05, + "learning_rate": 9.512212985681327e-08, + "loss": 0.6972, + "step": 28689 + }, + { + "epoch": 1.05, + "learning_rate": 9.511846779214121e-08, + "loss": 0.69, + "step": 28690 + }, + { + "epoch": 1.05, + "learning_rate": 9.511480572746913e-08, + "loss": 0.6828, + "step": 28691 + }, + { + "epoch": 1.05, + "learning_rate": 9.511114366279707e-08, + "loss": 0.6973, + "step": 28692 + }, + { + "epoch": 1.05, + "learning_rate": 9.510748159812501e-08, + "loss": 0.697, + "step": 28693 + }, + { + "epoch": 1.05, + "learning_rate": 9.510381953345295e-08, + "loss": 0.6925, + "step": 28694 + }, + { + "epoch": 1.05, + "learning_rate": 9.51001574687809e-08, + "loss": 0.6735, + "step": 28695 + }, + { + "epoch": 1.05, + "learning_rate": 9.509649540410883e-08, + "loss": 0.7011, + "step": 28696 + }, + { + "epoch": 1.05, + "learning_rate": 9.509283333943677e-08, + "loss": 0.6996, + "step": 28697 + }, + { + "epoch": 1.05, + "learning_rate": 9.508917127476471e-08, + "loss": 0.6878, + "step": 28698 + }, + { + "epoch": 1.05, + "learning_rate": 9.508550921009265e-08, + "loss": 0.6957, + "step": 28699 + }, + { + "epoch": 1.05, + "learning_rate": 9.508184714542058e-08, + "loss": 0.6949, + "step": 28700 + }, + { + "epoch": 1.05, + "learning_rate": 9.507818508074852e-08, + "loss": 0.6915, + "step": 28701 + }, + { + "epoch": 1.05, + "learning_rate": 9.507452301607646e-08, + "loss": 0.6896, + "step": 28702 + }, + { + "epoch": 1.05, + "learning_rate": 9.50708609514044e-08, + "loss": 0.6933, + "step": 28703 + }, + { + "epoch": 1.05, + "learning_rate": 9.506719888673234e-08, + "loss": 0.6939, + "step": 28704 + }, + { + "epoch": 1.05, + "learning_rate": 9.506353682206028e-08, + "loss": 0.6901, + "step": 28705 + }, + { + "epoch": 1.05, + "learning_rate": 9.505987475738822e-08, + "loss": 0.6936, + "step": 28706 + }, + { + "epoch": 1.05, + "learning_rate": 9.505621269271616e-08, + "loss": 0.7046, + "step": 28707 + }, + { + "epoch": 1.05, + "learning_rate": 9.505255062804409e-08, + "loss": 0.695, + "step": 28708 + }, + { + "epoch": 1.05, + "learning_rate": 9.504888856337203e-08, + "loss": 0.7072, + "step": 28709 + }, + { + "epoch": 1.05, + "learning_rate": 9.504522649869997e-08, + "loss": 0.6979, + "step": 28710 + }, + { + "epoch": 1.05, + "learning_rate": 9.504156443402789e-08, + "loss": 0.6914, + "step": 28711 + }, + { + "epoch": 1.05, + "learning_rate": 9.503790236935583e-08, + "loss": 0.6908, + "step": 28712 + }, + { + "epoch": 1.05, + "learning_rate": 9.503424030468377e-08, + "loss": 0.6848, + "step": 28713 + }, + { + "epoch": 1.05, + "learning_rate": 9.503057824001171e-08, + "loss": 0.6886, + "step": 28714 + }, + { + "epoch": 1.05, + "learning_rate": 9.502691617533965e-08, + "loss": 0.6984, + "step": 28715 + }, + { + "epoch": 1.05, + "learning_rate": 9.502325411066759e-08, + "loss": 0.6861, + "step": 28716 + }, + { + "epoch": 1.05, + "learning_rate": 9.501959204599552e-08, + "loss": 0.6956, + "step": 28717 + }, + { + "epoch": 1.05, + "learning_rate": 9.501592998132346e-08, + "loss": 0.6918, + "step": 28718 + }, + { + "epoch": 1.05, + "learning_rate": 9.50122679166514e-08, + "loss": 0.6797, + "step": 28719 + }, + { + "epoch": 1.05, + "learning_rate": 9.500860585197934e-08, + "loss": 0.7022, + "step": 28720 + }, + { + "epoch": 1.05, + "learning_rate": 9.500494378730728e-08, + "loss": 0.6934, + "step": 28721 + }, + { + "epoch": 1.05, + "learning_rate": 9.500128172263522e-08, + "loss": 0.6918, + "step": 28722 + }, + { + "epoch": 1.05, + "learning_rate": 9.499761965796316e-08, + "loss": 0.6876, + "step": 28723 + }, + { + "epoch": 1.05, + "learning_rate": 9.49939575932911e-08, + "loss": 0.6918, + "step": 28724 + }, + { + "epoch": 1.05, + "learning_rate": 9.499029552861904e-08, + "loss": 0.6868, + "step": 28725 + }, + { + "epoch": 1.05, + "learning_rate": 9.498663346394697e-08, + "loss": 0.6933, + "step": 28726 + }, + { + "epoch": 1.05, + "learning_rate": 9.49829713992749e-08, + "loss": 0.7016, + "step": 28727 + }, + { + "epoch": 1.05, + "learning_rate": 9.497930933460285e-08, + "loss": 0.6834, + "step": 28728 + }, + { + "epoch": 1.05, + "learning_rate": 9.497564726993078e-08, + "loss": 0.6946, + "step": 28729 + }, + { + "epoch": 1.05, + "learning_rate": 9.497198520525872e-08, + "loss": 0.6916, + "step": 28730 + }, + { + "epoch": 1.05, + "learning_rate": 9.496832314058666e-08, + "loss": 0.6884, + "step": 28731 + }, + { + "epoch": 1.05, + "learning_rate": 9.49646610759146e-08, + "loss": 0.6857, + "step": 28732 + }, + { + "epoch": 1.05, + "learning_rate": 9.496099901124253e-08, + "loss": 0.6928, + "step": 28733 + }, + { + "epoch": 1.05, + "learning_rate": 9.495733694657047e-08, + "loss": 0.6927, + "step": 28734 + }, + { + "epoch": 1.05, + "learning_rate": 9.49536748818984e-08, + "loss": 0.6994, + "step": 28735 + }, + { + "epoch": 1.05, + "learning_rate": 9.495001281722634e-08, + "loss": 0.6965, + "step": 28736 + }, + { + "epoch": 1.05, + "learning_rate": 9.494635075255428e-08, + "loss": 0.7065, + "step": 28737 + }, + { + "epoch": 1.05, + "learning_rate": 9.494268868788222e-08, + "loss": 0.6969, + "step": 28738 + }, + { + "epoch": 1.05, + "learning_rate": 9.493902662321016e-08, + "loss": 0.6949, + "step": 28739 + }, + { + "epoch": 1.05, + "learning_rate": 9.49353645585381e-08, + "loss": 0.6891, + "step": 28740 + }, + { + "epoch": 1.05, + "learning_rate": 9.493170249386604e-08, + "loss": 0.6931, + "step": 28741 + }, + { + "epoch": 1.05, + "learning_rate": 9.492804042919398e-08, + "loss": 0.7066, + "step": 28742 + }, + { + "epoch": 1.05, + "learning_rate": 9.492437836452192e-08, + "loss": 0.692, + "step": 28743 + }, + { + "epoch": 1.05, + "learning_rate": 9.492071629984984e-08, + "loss": 0.6925, + "step": 28744 + }, + { + "epoch": 1.05, + "learning_rate": 9.491705423517778e-08, + "loss": 0.7052, + "step": 28745 + }, + { + "epoch": 1.05, + "learning_rate": 9.491339217050572e-08, + "loss": 0.6842, + "step": 28746 + }, + { + "epoch": 1.05, + "learning_rate": 9.490973010583366e-08, + "loss": 0.6921, + "step": 28747 + }, + { + "epoch": 1.05, + "learning_rate": 9.49060680411616e-08, + "loss": 0.6996, + "step": 28748 + }, + { + "epoch": 1.05, + "learning_rate": 9.490240597648954e-08, + "loss": 0.6879, + "step": 28749 + }, + { + "epoch": 1.05, + "learning_rate": 9.489874391181748e-08, + "loss": 0.6939, + "step": 28750 + }, + { + "epoch": 1.05, + "learning_rate": 9.489508184714542e-08, + "loss": 0.7062, + "step": 28751 + }, + { + "epoch": 1.05, + "learning_rate": 9.489141978247336e-08, + "loss": 0.6936, + "step": 28752 + }, + { + "epoch": 1.05, + "learning_rate": 9.488775771780129e-08, + "loss": 0.683, + "step": 28753 + }, + { + "epoch": 1.05, + "learning_rate": 9.488409565312923e-08, + "loss": 0.689, + "step": 28754 + }, + { + "epoch": 1.05, + "learning_rate": 9.488043358845717e-08, + "loss": 0.6879, + "step": 28755 + }, + { + "epoch": 1.05, + "learning_rate": 9.487677152378511e-08, + "loss": 0.6992, + "step": 28756 + }, + { + "epoch": 1.05, + "learning_rate": 9.487310945911304e-08, + "loss": 0.6956, + "step": 28757 + }, + { + "epoch": 1.05, + "learning_rate": 9.486944739444098e-08, + "loss": 0.6888, + "step": 28758 + }, + { + "epoch": 1.05, + "learning_rate": 9.486578532976892e-08, + "loss": 0.6905, + "step": 28759 + }, + { + "epoch": 1.05, + "learning_rate": 9.486212326509686e-08, + "loss": 0.6949, + "step": 28760 + }, + { + "epoch": 1.05, + "learning_rate": 9.48584612004248e-08, + "loss": 0.7211, + "step": 28761 + }, + { + "epoch": 1.05, + "learning_rate": 9.485479913575272e-08, + "loss": 0.7037, + "step": 28762 + }, + { + "epoch": 1.05, + "learning_rate": 9.485113707108066e-08, + "loss": 0.6952, + "step": 28763 + }, + { + "epoch": 1.05, + "learning_rate": 9.48474750064086e-08, + "loss": 0.6985, + "step": 28764 + }, + { + "epoch": 1.05, + "learning_rate": 9.484381294173654e-08, + "loss": 0.6921, + "step": 28765 + }, + { + "epoch": 1.05, + "learning_rate": 9.484015087706448e-08, + "loss": 0.6916, + "step": 28766 + }, + { + "epoch": 1.05, + "learning_rate": 9.483648881239242e-08, + "loss": 0.6828, + "step": 28767 + }, + { + "epoch": 1.05, + "learning_rate": 9.483282674772036e-08, + "loss": 0.6927, + "step": 28768 + }, + { + "epoch": 1.05, + "learning_rate": 9.48291646830483e-08, + "loss": 0.6783, + "step": 28769 + }, + { + "epoch": 1.05, + "learning_rate": 9.482550261837624e-08, + "loss": 0.6727, + "step": 28770 + }, + { + "epoch": 1.05, + "learning_rate": 9.482184055370417e-08, + "loss": 0.6816, + "step": 28771 + }, + { + "epoch": 1.05, + "learning_rate": 9.481817848903211e-08, + "loss": 0.6901, + "step": 28772 + }, + { + "epoch": 1.05, + "learning_rate": 9.481451642436005e-08, + "loss": 0.6854, + "step": 28773 + }, + { + "epoch": 1.05, + "learning_rate": 9.481085435968799e-08, + "loss": 0.6947, + "step": 28774 + }, + { + "epoch": 1.05, + "learning_rate": 9.480719229501593e-08, + "loss": 0.6932, + "step": 28775 + }, + { + "epoch": 1.05, + "learning_rate": 9.480353023034387e-08, + "loss": 0.6989, + "step": 28776 + }, + { + "epoch": 1.05, + "learning_rate": 9.479986816567181e-08, + "loss": 0.7034, + "step": 28777 + }, + { + "epoch": 1.05, + "learning_rate": 9.479620610099975e-08, + "loss": 0.6992, + "step": 28778 + }, + { + "epoch": 1.05, + "learning_rate": 9.479254403632768e-08, + "loss": 0.694, + "step": 28779 + }, + { + "epoch": 1.05, + "learning_rate": 9.478888197165562e-08, + "loss": 0.6948, + "step": 28780 + }, + { + "epoch": 1.05, + "learning_rate": 9.478521990698356e-08, + "loss": 0.6833, + "step": 28781 + }, + { + "epoch": 1.05, + "learning_rate": 9.478155784231148e-08, + "loss": 0.6916, + "step": 28782 + }, + { + "epoch": 1.05, + "learning_rate": 9.477789577763942e-08, + "loss": 0.6783, + "step": 28783 + }, + { + "epoch": 1.05, + "learning_rate": 9.477423371296736e-08, + "loss": 0.695, + "step": 28784 + }, + { + "epoch": 1.05, + "learning_rate": 9.47705716482953e-08, + "loss": 0.6861, + "step": 28785 + }, + { + "epoch": 1.05, + "learning_rate": 9.476690958362324e-08, + "loss": 0.6968, + "step": 28786 + }, + { + "epoch": 1.05, + "learning_rate": 9.476324751895118e-08, + "loss": 0.7011, + "step": 28787 + }, + { + "epoch": 1.05, + "learning_rate": 9.475958545427912e-08, + "loss": 0.6813, + "step": 28788 + }, + { + "epoch": 1.05, + "learning_rate": 9.475592338960705e-08, + "loss": 0.7066, + "step": 28789 + }, + { + "epoch": 1.05, + "learning_rate": 9.475226132493499e-08, + "loss": 0.6718, + "step": 28790 + }, + { + "epoch": 1.05, + "learning_rate": 9.474859926026293e-08, + "loss": 0.7018, + "step": 28791 + }, + { + "epoch": 1.05, + "learning_rate": 9.474493719559087e-08, + "loss": 0.6848, + "step": 28792 + }, + { + "epoch": 1.05, + "learning_rate": 9.474127513091881e-08, + "loss": 0.6944, + "step": 28793 + }, + { + "epoch": 1.05, + "learning_rate": 9.473761306624675e-08, + "loss": 0.6867, + "step": 28794 + }, + { + "epoch": 1.05, + "learning_rate": 9.473395100157469e-08, + "loss": 0.6908, + "step": 28795 + }, + { + "epoch": 1.05, + "learning_rate": 9.473028893690263e-08, + "loss": 0.6978, + "step": 28796 + }, + { + "epoch": 1.05, + "learning_rate": 9.472662687223057e-08, + "loss": 0.7005, + "step": 28797 + }, + { + "epoch": 1.05, + "learning_rate": 9.47229648075585e-08, + "loss": 0.6825, + "step": 28798 + }, + { + "epoch": 1.05, + "learning_rate": 9.471930274288643e-08, + "loss": 0.6969, + "step": 28799 + }, + { + "epoch": 1.05, + "learning_rate": 9.471564067821437e-08, + "loss": 0.7028, + "step": 28800 + }, + { + "epoch": 1.05, + "learning_rate": 9.471197861354231e-08, + "loss": 0.6795, + "step": 28801 + }, + { + "epoch": 1.05, + "learning_rate": 9.470831654887025e-08, + "loss": 0.7025, + "step": 28802 + }, + { + "epoch": 1.05, + "learning_rate": 9.47046544841982e-08, + "loss": 0.6937, + "step": 28803 + }, + { + "epoch": 1.05, + "learning_rate": 9.470099241952612e-08, + "loss": 0.687, + "step": 28804 + }, + { + "epoch": 1.05, + "learning_rate": 9.469733035485406e-08, + "loss": 0.6817, + "step": 28805 + }, + { + "epoch": 1.05, + "learning_rate": 9.4693668290182e-08, + "loss": 0.6882, + "step": 28806 + }, + { + "epoch": 1.05, + "learning_rate": 9.469000622550993e-08, + "loss": 0.6835, + "step": 28807 + }, + { + "epoch": 1.05, + "learning_rate": 9.468634416083787e-08, + "loss": 0.6739, + "step": 28808 + }, + { + "epoch": 1.05, + "learning_rate": 9.468268209616581e-08, + "loss": 0.6901, + "step": 28809 + }, + { + "epoch": 1.05, + "learning_rate": 9.467902003149375e-08, + "loss": 0.6906, + "step": 28810 + }, + { + "epoch": 1.05, + "learning_rate": 9.467535796682169e-08, + "loss": 0.687, + "step": 28811 + }, + { + "epoch": 1.05, + "learning_rate": 9.467169590214963e-08, + "loss": 0.6998, + "step": 28812 + }, + { + "epoch": 1.05, + "learning_rate": 9.466803383747757e-08, + "loss": 0.7106, + "step": 28813 + }, + { + "epoch": 1.05, + "learning_rate": 9.466437177280551e-08, + "loss": 0.6769, + "step": 28814 + }, + { + "epoch": 1.05, + "learning_rate": 9.466070970813345e-08, + "loss": 0.6946, + "step": 28815 + }, + { + "epoch": 1.05, + "learning_rate": 9.465704764346137e-08, + "loss": 0.6899, + "step": 28816 + }, + { + "epoch": 1.05, + "learning_rate": 9.465338557878931e-08, + "loss": 0.697, + "step": 28817 + }, + { + "epoch": 1.05, + "learning_rate": 9.464972351411725e-08, + "loss": 0.696, + "step": 28818 + }, + { + "epoch": 1.05, + "learning_rate": 9.464606144944519e-08, + "loss": 0.705, + "step": 28819 + }, + { + "epoch": 1.05, + "learning_rate": 9.464239938477313e-08, + "loss": 0.6841, + "step": 28820 + }, + { + "epoch": 1.05, + "learning_rate": 9.463873732010107e-08, + "loss": 0.697, + "step": 28821 + }, + { + "epoch": 1.05, + "learning_rate": 9.463507525542901e-08, + "loss": 0.701, + "step": 28822 + }, + { + "epoch": 1.05, + "learning_rate": 9.463141319075695e-08, + "loss": 0.6874, + "step": 28823 + }, + { + "epoch": 1.05, + "learning_rate": 9.462775112608489e-08, + "loss": 0.6885, + "step": 28824 + }, + { + "epoch": 1.05, + "learning_rate": 9.462408906141282e-08, + "loss": 0.6976, + "step": 28825 + }, + { + "epoch": 1.05, + "learning_rate": 9.462042699674076e-08, + "loss": 0.6905, + "step": 28826 + }, + { + "epoch": 1.05, + "learning_rate": 9.46167649320687e-08, + "loss": 0.6974, + "step": 28827 + }, + { + "epoch": 1.05, + "learning_rate": 9.461310286739663e-08, + "loss": 0.7088, + "step": 28828 + }, + { + "epoch": 1.05, + "learning_rate": 9.460944080272457e-08, + "loss": 0.6894, + "step": 28829 + }, + { + "epoch": 1.05, + "learning_rate": 9.46057787380525e-08, + "loss": 0.6985, + "step": 28830 + }, + { + "epoch": 1.05, + "learning_rate": 9.460211667338045e-08, + "loss": 0.7054, + "step": 28831 + }, + { + "epoch": 1.05, + "learning_rate": 9.459845460870839e-08, + "loss": 0.6956, + "step": 28832 + }, + { + "epoch": 1.05, + "learning_rate": 9.459479254403633e-08, + "loss": 0.6917, + "step": 28833 + }, + { + "epoch": 1.05, + "learning_rate": 9.459113047936425e-08, + "loss": 0.6787, + "step": 28834 + }, + { + "epoch": 1.05, + "learning_rate": 9.458746841469219e-08, + "loss": 0.6831, + "step": 28835 + }, + { + "epoch": 1.06, + "learning_rate": 9.458380635002013e-08, + "loss": 0.6928, + "step": 28836 + }, + { + "epoch": 1.06, + "learning_rate": 9.458014428534807e-08, + "loss": 0.6992, + "step": 28837 + }, + { + "epoch": 1.06, + "learning_rate": 9.457648222067601e-08, + "loss": 0.693, + "step": 28838 + }, + { + "epoch": 1.06, + "learning_rate": 9.457282015600395e-08, + "loss": 0.7096, + "step": 28839 + }, + { + "epoch": 1.06, + "learning_rate": 9.456915809133189e-08, + "loss": 0.6995, + "step": 28840 + }, + { + "epoch": 1.06, + "learning_rate": 9.456549602665983e-08, + "loss": 0.6875, + "step": 28841 + }, + { + "epoch": 1.06, + "learning_rate": 9.456183396198777e-08, + "loss": 0.6894, + "step": 28842 + }, + { + "epoch": 1.06, + "learning_rate": 9.45581718973157e-08, + "loss": 0.694, + "step": 28843 + }, + { + "epoch": 1.06, + "learning_rate": 9.455450983264364e-08, + "loss": 0.6807, + "step": 28844 + }, + { + "epoch": 1.06, + "learning_rate": 9.455084776797158e-08, + "loss": 0.6829, + "step": 28845 + }, + { + "epoch": 1.06, + "learning_rate": 9.454718570329952e-08, + "loss": 0.7093, + "step": 28846 + }, + { + "epoch": 1.06, + "learning_rate": 9.454352363862746e-08, + "loss": 0.6973, + "step": 28847 + }, + { + "epoch": 1.06, + "learning_rate": 9.45398615739554e-08, + "loss": 0.6996, + "step": 28848 + }, + { + "epoch": 1.06, + "learning_rate": 9.453619950928334e-08, + "loss": 0.6952, + "step": 28849 + }, + { + "epoch": 1.06, + "learning_rate": 9.453253744461126e-08, + "loss": 0.6975, + "step": 28850 + }, + { + "epoch": 1.06, + "learning_rate": 9.45288753799392e-08, + "loss": 0.6946, + "step": 28851 + }, + { + "epoch": 1.06, + "learning_rate": 9.452521331526714e-08, + "loss": 0.6975, + "step": 28852 + }, + { + "epoch": 1.06, + "learning_rate": 9.452155125059507e-08, + "loss": 0.6956, + "step": 28853 + }, + { + "epoch": 1.06, + "learning_rate": 9.451788918592301e-08, + "loss": 0.6988, + "step": 28854 + }, + { + "epoch": 1.06, + "learning_rate": 9.451422712125095e-08, + "loss": 0.6953, + "step": 28855 + }, + { + "epoch": 1.06, + "learning_rate": 9.451056505657889e-08, + "loss": 0.7055, + "step": 28856 + }, + { + "epoch": 1.06, + "learning_rate": 9.450690299190683e-08, + "loss": 0.687, + "step": 28857 + }, + { + "epoch": 1.06, + "learning_rate": 9.450324092723477e-08, + "loss": 0.6973, + "step": 28858 + }, + { + "epoch": 1.06, + "learning_rate": 9.449957886256271e-08, + "loss": 0.6935, + "step": 28859 + }, + { + "epoch": 1.06, + "learning_rate": 9.449591679789065e-08, + "loss": 0.7005, + "step": 28860 + }, + { + "epoch": 1.06, + "learning_rate": 9.449225473321858e-08, + "loss": 0.7159, + "step": 28861 + }, + { + "epoch": 1.06, + "learning_rate": 9.448859266854652e-08, + "loss": 0.6815, + "step": 28862 + }, + { + "epoch": 1.06, + "learning_rate": 9.448493060387446e-08, + "loss": 0.6935, + "step": 28863 + }, + { + "epoch": 1.06, + "learning_rate": 9.44812685392024e-08, + "loss": 0.7006, + "step": 28864 + }, + { + "epoch": 1.06, + "learning_rate": 9.447760647453034e-08, + "loss": 0.7036, + "step": 28865 + }, + { + "epoch": 1.06, + "learning_rate": 9.447394440985828e-08, + "loss": 0.6998, + "step": 28866 + }, + { + "epoch": 1.06, + "learning_rate": 9.447028234518622e-08, + "loss": 0.7103, + "step": 28867 + }, + { + "epoch": 1.06, + "learning_rate": 9.446662028051416e-08, + "loss": 0.6945, + "step": 28868 + }, + { + "epoch": 1.06, + "learning_rate": 9.44629582158421e-08, + "loss": 0.6962, + "step": 28869 + }, + { + "epoch": 1.06, + "learning_rate": 9.445929615117002e-08, + "loss": 0.6926, + "step": 28870 + }, + { + "epoch": 1.06, + "learning_rate": 9.445563408649796e-08, + "loss": 0.7009, + "step": 28871 + }, + { + "epoch": 1.06, + "learning_rate": 9.44519720218259e-08, + "loss": 0.706, + "step": 28872 + }, + { + "epoch": 1.06, + "learning_rate": 9.444830995715384e-08, + "loss": 0.6784, + "step": 28873 + }, + { + "epoch": 1.06, + "learning_rate": 9.444464789248178e-08, + "loss": 0.6917, + "step": 28874 + }, + { + "epoch": 1.06, + "learning_rate": 9.444098582780971e-08, + "loss": 0.6968, + "step": 28875 + }, + { + "epoch": 1.06, + "learning_rate": 9.443732376313765e-08, + "loss": 0.7017, + "step": 28876 + }, + { + "epoch": 1.06, + "learning_rate": 9.443366169846559e-08, + "loss": 0.6895, + "step": 28877 + }, + { + "epoch": 1.06, + "learning_rate": 9.442999963379353e-08, + "loss": 0.692, + "step": 28878 + }, + { + "epoch": 1.06, + "learning_rate": 9.442633756912146e-08, + "loss": 0.6866, + "step": 28879 + }, + { + "epoch": 1.06, + "learning_rate": 9.44226755044494e-08, + "loss": 0.7021, + "step": 28880 + }, + { + "epoch": 1.06, + "learning_rate": 9.441901343977734e-08, + "loss": 0.6884, + "step": 28881 + }, + { + "epoch": 1.06, + "learning_rate": 9.441535137510528e-08, + "loss": 0.6972, + "step": 28882 + }, + { + "epoch": 1.06, + "learning_rate": 9.441168931043322e-08, + "loss": 0.6939, + "step": 28883 + }, + { + "epoch": 1.06, + "learning_rate": 9.440802724576116e-08, + "loss": 0.6942, + "step": 28884 + }, + { + "epoch": 1.06, + "learning_rate": 9.44043651810891e-08, + "loss": 0.691, + "step": 28885 + }, + { + "epoch": 1.06, + "learning_rate": 9.440070311641704e-08, + "loss": 0.684, + "step": 28886 + }, + { + "epoch": 1.06, + "learning_rate": 9.439704105174498e-08, + "loss": 0.6862, + "step": 28887 + }, + { + "epoch": 1.06, + "learning_rate": 9.43933789870729e-08, + "loss": 0.6946, + "step": 28888 + }, + { + "epoch": 1.06, + "learning_rate": 9.438971692240084e-08, + "loss": 0.696, + "step": 28889 + }, + { + "epoch": 1.06, + "learning_rate": 9.438605485772878e-08, + "loss": 0.6922, + "step": 28890 + }, + { + "epoch": 1.06, + "learning_rate": 9.438239279305672e-08, + "loss": 0.6914, + "step": 28891 + }, + { + "epoch": 1.06, + "learning_rate": 9.437873072838466e-08, + "loss": 0.6952, + "step": 28892 + }, + { + "epoch": 1.06, + "learning_rate": 9.43750686637126e-08, + "loss": 0.696, + "step": 28893 + }, + { + "epoch": 1.06, + "learning_rate": 9.437140659904054e-08, + "loss": 0.6986, + "step": 28894 + }, + { + "epoch": 1.06, + "learning_rate": 9.436774453436848e-08, + "loss": 0.686, + "step": 28895 + }, + { + "epoch": 1.06, + "learning_rate": 9.436408246969641e-08, + "loss": 0.6996, + "step": 28896 + }, + { + "epoch": 1.06, + "learning_rate": 9.436042040502435e-08, + "loss": 0.6932, + "step": 28897 + }, + { + "epoch": 1.06, + "learning_rate": 9.435675834035229e-08, + "loss": 0.6871, + "step": 28898 + }, + { + "epoch": 1.06, + "learning_rate": 9.435309627568022e-08, + "loss": 0.689, + "step": 28899 + }, + { + "epoch": 1.06, + "learning_rate": 9.434943421100816e-08, + "loss": 0.6914, + "step": 28900 + }, + { + "epoch": 1.06, + "learning_rate": 9.43457721463361e-08, + "loss": 0.7088, + "step": 28901 + }, + { + "epoch": 1.06, + "learning_rate": 9.434211008166404e-08, + "loss": 0.6898, + "step": 28902 + }, + { + "epoch": 1.06, + "learning_rate": 9.433844801699198e-08, + "loss": 0.7037, + "step": 28903 + }, + { + "epoch": 1.06, + "learning_rate": 9.433478595231992e-08, + "loss": 0.6953, + "step": 28904 + }, + { + "epoch": 1.06, + "learning_rate": 9.433112388764786e-08, + "loss": 0.7045, + "step": 28905 + }, + { + "epoch": 1.06, + "learning_rate": 9.432746182297578e-08, + "loss": 0.687, + "step": 28906 + }, + { + "epoch": 1.06, + "learning_rate": 9.432379975830372e-08, + "loss": 0.6901, + "step": 28907 + }, + { + "epoch": 1.06, + "learning_rate": 9.432013769363166e-08, + "loss": 0.6854, + "step": 28908 + }, + { + "epoch": 1.06, + "learning_rate": 9.43164756289596e-08, + "loss": 0.7001, + "step": 28909 + }, + { + "epoch": 1.06, + "learning_rate": 9.431281356428754e-08, + "loss": 0.6998, + "step": 28910 + }, + { + "epoch": 1.06, + "learning_rate": 9.430915149961548e-08, + "loss": 0.7077, + "step": 28911 + }, + { + "epoch": 1.06, + "learning_rate": 9.430548943494342e-08, + "loss": 0.7005, + "step": 28912 + }, + { + "epoch": 1.06, + "learning_rate": 9.430182737027136e-08, + "loss": 0.6965, + "step": 28913 + }, + { + "epoch": 1.06, + "learning_rate": 9.42981653055993e-08, + "loss": 0.703, + "step": 28914 + }, + { + "epoch": 1.06, + "learning_rate": 9.429450324092723e-08, + "loss": 0.6958, + "step": 28915 + }, + { + "epoch": 1.06, + "learning_rate": 9.429084117625517e-08, + "loss": 0.6688, + "step": 28916 + }, + { + "epoch": 1.06, + "learning_rate": 9.428717911158311e-08, + "loss": 0.6983, + "step": 28917 + }, + { + "epoch": 1.06, + "learning_rate": 9.428351704691105e-08, + "loss": 0.7053, + "step": 28918 + }, + { + "epoch": 1.06, + "learning_rate": 9.427985498223899e-08, + "loss": 0.6958, + "step": 28919 + }, + { + "epoch": 1.06, + "learning_rate": 9.427619291756693e-08, + "loss": 0.6943, + "step": 28920 + }, + { + "epoch": 1.06, + "learning_rate": 9.427253085289485e-08, + "loss": 0.6928, + "step": 28921 + }, + { + "epoch": 1.06, + "learning_rate": 9.42688687882228e-08, + "loss": 0.6978, + "step": 28922 + }, + { + "epoch": 1.06, + "learning_rate": 9.426520672355073e-08, + "loss": 0.6885, + "step": 28923 + }, + { + "epoch": 1.06, + "learning_rate": 9.426154465887866e-08, + "loss": 0.6828, + "step": 28924 + }, + { + "epoch": 1.06, + "learning_rate": 9.42578825942066e-08, + "loss": 0.6982, + "step": 28925 + }, + { + "epoch": 1.06, + "learning_rate": 9.425422052953454e-08, + "loss": 0.6995, + "step": 28926 + }, + { + "epoch": 1.06, + "learning_rate": 9.425055846486248e-08, + "loss": 0.6951, + "step": 28927 + }, + { + "epoch": 1.06, + "learning_rate": 9.424689640019042e-08, + "loss": 0.6821, + "step": 28928 + }, + { + "epoch": 1.06, + "learning_rate": 9.424323433551836e-08, + "loss": 0.6858, + "step": 28929 + }, + { + "epoch": 1.06, + "learning_rate": 9.42395722708463e-08, + "loss": 0.6844, + "step": 28930 + }, + { + "epoch": 1.06, + "learning_rate": 9.423591020617424e-08, + "loss": 0.7152, + "step": 28931 + }, + { + "epoch": 1.06, + "learning_rate": 9.423224814150218e-08, + "loss": 0.6714, + "step": 28932 + }, + { + "epoch": 1.06, + "learning_rate": 9.422858607683011e-08, + "loss": 0.6837, + "step": 28933 + }, + { + "epoch": 1.06, + "learning_rate": 9.422492401215805e-08, + "loss": 0.6896, + "step": 28934 + }, + { + "epoch": 1.06, + "learning_rate": 9.422126194748599e-08, + "loss": 0.7021, + "step": 28935 + }, + { + "epoch": 1.06, + "learning_rate": 9.421759988281393e-08, + "loss": 0.6904, + "step": 28936 + }, + { + "epoch": 1.06, + "learning_rate": 9.421393781814187e-08, + "loss": 0.6979, + "step": 28937 + }, + { + "epoch": 1.06, + "learning_rate": 9.42102757534698e-08, + "loss": 0.681, + "step": 28938 + }, + { + "epoch": 1.06, + "learning_rate": 9.420661368879775e-08, + "loss": 0.6921, + "step": 28939 + }, + { + "epoch": 1.06, + "learning_rate": 9.420295162412569e-08, + "loss": 0.6878, + "step": 28940 + }, + { + "epoch": 1.06, + "learning_rate": 9.419928955945363e-08, + "loss": 0.6968, + "step": 28941 + }, + { + "epoch": 1.06, + "learning_rate": 9.419562749478155e-08, + "loss": 0.6857, + "step": 28942 + }, + { + "epoch": 1.06, + "learning_rate": 9.419196543010949e-08, + "loss": 0.6939, + "step": 28943 + }, + { + "epoch": 1.06, + "learning_rate": 9.418830336543743e-08, + "loss": 0.7001, + "step": 28944 + }, + { + "epoch": 1.06, + "learning_rate": 9.418464130076536e-08, + "loss": 0.7053, + "step": 28945 + }, + { + "epoch": 1.06, + "learning_rate": 9.41809792360933e-08, + "loss": 0.6983, + "step": 28946 + }, + { + "epoch": 1.06, + "learning_rate": 9.417731717142124e-08, + "loss": 0.7018, + "step": 28947 + }, + { + "epoch": 1.06, + "learning_rate": 9.417365510674918e-08, + "loss": 0.6908, + "step": 28948 + }, + { + "epoch": 1.06, + "learning_rate": 9.416999304207712e-08, + "loss": 0.687, + "step": 28949 + }, + { + "epoch": 1.06, + "learning_rate": 9.416633097740506e-08, + "loss": 0.6916, + "step": 28950 + }, + { + "epoch": 1.06, + "learning_rate": 9.416266891273299e-08, + "loss": 0.6972, + "step": 28951 + }, + { + "epoch": 1.06, + "learning_rate": 9.415900684806093e-08, + "loss": 0.703, + "step": 28952 + }, + { + "epoch": 1.06, + "learning_rate": 9.415534478338887e-08, + "loss": 0.685, + "step": 28953 + }, + { + "epoch": 1.06, + "learning_rate": 9.41516827187168e-08, + "loss": 0.6877, + "step": 28954 + }, + { + "epoch": 1.06, + "learning_rate": 9.414802065404475e-08, + "loss": 0.7008, + "step": 28955 + }, + { + "epoch": 1.06, + "learning_rate": 9.414435858937269e-08, + "loss": 0.7026, + "step": 28956 + }, + { + "epoch": 1.06, + "learning_rate": 9.414069652470063e-08, + "loss": 0.6974, + "step": 28957 + }, + { + "epoch": 1.06, + "learning_rate": 9.413703446002857e-08, + "loss": 0.7061, + "step": 28958 + }, + { + "epoch": 1.06, + "learning_rate": 9.41333723953565e-08, + "loss": 0.6887, + "step": 28959 + }, + { + "epoch": 1.06, + "learning_rate": 9.412971033068443e-08, + "loss": 0.6996, + "step": 28960 + }, + { + "epoch": 1.06, + "learning_rate": 9.412604826601237e-08, + "loss": 0.6982, + "step": 28961 + }, + { + "epoch": 1.06, + "learning_rate": 9.412238620134031e-08, + "loss": 0.6855, + "step": 28962 + }, + { + "epoch": 1.06, + "learning_rate": 9.411872413666825e-08, + "loss": 0.6931, + "step": 28963 + }, + { + "epoch": 1.06, + "learning_rate": 9.411506207199619e-08, + "loss": 0.6904, + "step": 28964 + }, + { + "epoch": 1.06, + "learning_rate": 9.411140000732413e-08, + "loss": 0.6978, + "step": 28965 + }, + { + "epoch": 1.06, + "learning_rate": 9.410773794265207e-08, + "loss": 0.6907, + "step": 28966 + }, + { + "epoch": 1.06, + "learning_rate": 9.410407587798e-08, + "loss": 0.6888, + "step": 28967 + }, + { + "epoch": 1.06, + "learning_rate": 9.410041381330794e-08, + "loss": 0.7022, + "step": 28968 + }, + { + "epoch": 1.06, + "learning_rate": 9.409675174863588e-08, + "loss": 0.6976, + "step": 28969 + }, + { + "epoch": 1.06, + "learning_rate": 9.40930896839638e-08, + "loss": 0.6941, + "step": 28970 + }, + { + "epoch": 1.06, + "learning_rate": 9.408942761929174e-08, + "loss": 0.6935, + "step": 28971 + }, + { + "epoch": 1.06, + "learning_rate": 9.408576555461968e-08, + "loss": 0.6993, + "step": 28972 + }, + { + "epoch": 1.06, + "learning_rate": 9.408210348994762e-08, + "loss": 0.7005, + "step": 28973 + }, + { + "epoch": 1.06, + "learning_rate": 9.407844142527556e-08, + "loss": 0.6871, + "step": 28974 + }, + { + "epoch": 1.06, + "learning_rate": 9.40747793606035e-08, + "loss": 0.7036, + "step": 28975 + }, + { + "epoch": 1.06, + "learning_rate": 9.407111729593144e-08, + "loss": 0.6793, + "step": 28976 + }, + { + "epoch": 1.06, + "learning_rate": 9.406745523125938e-08, + "loss": 0.6995, + "step": 28977 + }, + { + "epoch": 1.06, + "learning_rate": 9.406379316658731e-08, + "loss": 0.686, + "step": 28978 + }, + { + "epoch": 1.06, + "learning_rate": 9.406013110191525e-08, + "loss": 0.7045, + "step": 28979 + }, + { + "epoch": 1.06, + "learning_rate": 9.405646903724319e-08, + "loss": 0.6858, + "step": 28980 + }, + { + "epoch": 1.06, + "learning_rate": 9.405280697257113e-08, + "loss": 0.6967, + "step": 28981 + }, + { + "epoch": 1.06, + "learning_rate": 9.404914490789907e-08, + "loss": 0.6976, + "step": 28982 + }, + { + "epoch": 1.06, + "learning_rate": 9.404548284322701e-08, + "loss": 0.6953, + "step": 28983 + }, + { + "epoch": 1.06, + "learning_rate": 9.404182077855495e-08, + "loss": 0.6896, + "step": 28984 + }, + { + "epoch": 1.06, + "learning_rate": 9.403815871388289e-08, + "loss": 0.6878, + "step": 28985 + }, + { + "epoch": 1.06, + "learning_rate": 9.403449664921083e-08, + "loss": 0.6996, + "step": 28986 + }, + { + "epoch": 1.06, + "learning_rate": 9.403083458453876e-08, + "loss": 0.7174, + "step": 28987 + }, + { + "epoch": 1.06, + "learning_rate": 9.40271725198667e-08, + "loss": 0.6872, + "step": 28988 + }, + { + "epoch": 1.06, + "learning_rate": 9.402351045519464e-08, + "loss": 0.6956, + "step": 28989 + }, + { + "epoch": 1.06, + "learning_rate": 9.401984839052258e-08, + "loss": 0.6814, + "step": 28990 + }, + { + "epoch": 1.06, + "learning_rate": 9.401618632585052e-08, + "loss": 0.7026, + "step": 28991 + }, + { + "epoch": 1.06, + "learning_rate": 9.401252426117844e-08, + "loss": 0.6973, + "step": 28992 + }, + { + "epoch": 1.06, + "learning_rate": 9.400886219650638e-08, + "loss": 0.7028, + "step": 28993 + }, + { + "epoch": 1.06, + "learning_rate": 9.400520013183432e-08, + "loss": 0.6916, + "step": 28994 + }, + { + "epoch": 1.06, + "learning_rate": 9.400153806716226e-08, + "loss": 0.6973, + "step": 28995 + }, + { + "epoch": 1.06, + "learning_rate": 9.399787600249019e-08, + "loss": 0.7055, + "step": 28996 + }, + { + "epoch": 1.06, + "learning_rate": 9.399421393781813e-08, + "loss": 0.6995, + "step": 28997 + }, + { + "epoch": 1.06, + "learning_rate": 9.399055187314607e-08, + "loss": 0.6978, + "step": 28998 + }, + { + "epoch": 1.06, + "learning_rate": 9.398688980847401e-08, + "loss": 0.6974, + "step": 28999 + }, + { + "epoch": 1.06, + "learning_rate": 9.398322774380195e-08, + "loss": 0.6979, + "step": 29000 + }, + { + "epoch": 1.06, + "learning_rate": 9.397956567912989e-08, + "loss": 0.6889, + "step": 29001 + }, + { + "epoch": 1.06, + "learning_rate": 9.397590361445783e-08, + "loss": 0.6976, + "step": 29002 + }, + { + "epoch": 1.06, + "learning_rate": 9.397224154978577e-08, + "loss": 0.6985, + "step": 29003 + }, + { + "epoch": 1.06, + "learning_rate": 9.396857948511371e-08, + "loss": 0.7078, + "step": 29004 + }, + { + "epoch": 1.06, + "learning_rate": 9.396491742044164e-08, + "loss": 0.6958, + "step": 29005 + }, + { + "epoch": 1.06, + "learning_rate": 9.396125535576958e-08, + "loss": 0.6896, + "step": 29006 + }, + { + "epoch": 1.06, + "learning_rate": 9.395759329109752e-08, + "loss": 0.6862, + "step": 29007 + }, + { + "epoch": 1.06, + "learning_rate": 9.395393122642546e-08, + "loss": 0.7014, + "step": 29008 + }, + { + "epoch": 1.06, + "learning_rate": 9.39502691617534e-08, + "loss": 0.6927, + "step": 29009 + }, + { + "epoch": 1.06, + "learning_rate": 9.394660709708134e-08, + "loss": 0.6981, + "step": 29010 + }, + { + "epoch": 1.06, + "learning_rate": 9.394294503240928e-08, + "loss": 0.7054, + "step": 29011 + }, + { + "epoch": 1.06, + "learning_rate": 9.393928296773722e-08, + "loss": 0.69, + "step": 29012 + }, + { + "epoch": 1.06, + "learning_rate": 9.393562090306516e-08, + "loss": 0.692, + "step": 29013 + }, + { + "epoch": 1.06, + "learning_rate": 9.393195883839308e-08, + "loss": 0.6863, + "step": 29014 + }, + { + "epoch": 1.06, + "learning_rate": 9.392829677372102e-08, + "loss": 0.6922, + "step": 29015 + }, + { + "epoch": 1.06, + "learning_rate": 9.392463470904895e-08, + "loss": 0.7115, + "step": 29016 + }, + { + "epoch": 1.06, + "learning_rate": 9.392097264437689e-08, + "loss": 0.6998, + "step": 29017 + }, + { + "epoch": 1.06, + "learning_rate": 9.391731057970483e-08, + "loss": 0.6945, + "step": 29018 + }, + { + "epoch": 1.06, + "learning_rate": 9.391364851503277e-08, + "loss": 0.6934, + "step": 29019 + }, + { + "epoch": 1.06, + "learning_rate": 9.390998645036071e-08, + "loss": 0.7009, + "step": 29020 + }, + { + "epoch": 1.06, + "learning_rate": 9.390632438568865e-08, + "loss": 0.6947, + "step": 29021 + }, + { + "epoch": 1.06, + "learning_rate": 9.390266232101659e-08, + "loss": 0.7119, + "step": 29022 + }, + { + "epoch": 1.06, + "learning_rate": 9.389900025634452e-08, + "loss": 0.6938, + "step": 29023 + }, + { + "epoch": 1.06, + "learning_rate": 9.389533819167246e-08, + "loss": 0.6918, + "step": 29024 + }, + { + "epoch": 1.06, + "learning_rate": 9.38916761270004e-08, + "loss": 0.7002, + "step": 29025 + }, + { + "epoch": 1.06, + "learning_rate": 9.388801406232833e-08, + "loss": 0.6975, + "step": 29026 + }, + { + "epoch": 1.06, + "learning_rate": 9.388435199765627e-08, + "loss": 0.6821, + "step": 29027 + }, + { + "epoch": 1.06, + "learning_rate": 9.388068993298421e-08, + "loss": 0.6972, + "step": 29028 + }, + { + "epoch": 1.06, + "learning_rate": 9.387702786831215e-08, + "loss": 0.6906, + "step": 29029 + }, + { + "epoch": 1.06, + "learning_rate": 9.38733658036401e-08, + "loss": 0.6882, + "step": 29030 + }, + { + "epoch": 1.06, + "learning_rate": 9.386970373896803e-08, + "loss": 0.6973, + "step": 29031 + }, + { + "epoch": 1.06, + "learning_rate": 9.386604167429596e-08, + "loss": 0.6988, + "step": 29032 + }, + { + "epoch": 1.06, + "learning_rate": 9.38623796096239e-08, + "loss": 0.6753, + "step": 29033 + }, + { + "epoch": 1.06, + "learning_rate": 9.385871754495184e-08, + "loss": 0.6936, + "step": 29034 + }, + { + "epoch": 1.06, + "learning_rate": 9.385505548027978e-08, + "loss": 0.6912, + "step": 29035 + }, + { + "epoch": 1.06, + "learning_rate": 9.385139341560772e-08, + "loss": 0.6801, + "step": 29036 + }, + { + "epoch": 1.06, + "learning_rate": 9.384773135093566e-08, + "loss": 0.6868, + "step": 29037 + }, + { + "epoch": 1.06, + "learning_rate": 9.384406928626359e-08, + "loss": 0.6863, + "step": 29038 + }, + { + "epoch": 1.06, + "learning_rate": 9.384040722159153e-08, + "loss": 0.6939, + "step": 29039 + }, + { + "epoch": 1.06, + "learning_rate": 9.383674515691947e-08, + "loss": 0.6956, + "step": 29040 + }, + { + "epoch": 1.06, + "learning_rate": 9.38330830922474e-08, + "loss": 0.6842, + "step": 29041 + }, + { + "epoch": 1.06, + "learning_rate": 9.382942102757533e-08, + "loss": 0.6901, + "step": 29042 + }, + { + "epoch": 1.06, + "learning_rate": 9.382575896290327e-08, + "loss": 0.6828, + "step": 29043 + }, + { + "epoch": 1.06, + "learning_rate": 9.382209689823121e-08, + "loss": 0.6845, + "step": 29044 + }, + { + "epoch": 1.06, + "learning_rate": 9.381843483355915e-08, + "loss": 0.6826, + "step": 29045 + }, + { + "epoch": 1.06, + "learning_rate": 9.38147727688871e-08, + "loss": 0.6951, + "step": 29046 + }, + { + "epoch": 1.06, + "learning_rate": 9.381111070421503e-08, + "loss": 0.6983, + "step": 29047 + }, + { + "epoch": 1.06, + "learning_rate": 9.380744863954297e-08, + "loss": 0.6874, + "step": 29048 + }, + { + "epoch": 1.06, + "learning_rate": 9.380378657487091e-08, + "loss": 0.7053, + "step": 29049 + }, + { + "epoch": 1.06, + "learning_rate": 9.380012451019884e-08, + "loss": 0.6883, + "step": 29050 + }, + { + "epoch": 1.06, + "learning_rate": 9.379646244552678e-08, + "loss": 0.6974, + "step": 29051 + }, + { + "epoch": 1.06, + "learning_rate": 9.379280038085472e-08, + "loss": 0.6997, + "step": 29052 + }, + { + "epoch": 1.06, + "learning_rate": 9.378913831618266e-08, + "loss": 0.694, + "step": 29053 + }, + { + "epoch": 1.06, + "learning_rate": 9.37854762515106e-08, + "loss": 0.6871, + "step": 29054 + }, + { + "epoch": 1.06, + "learning_rate": 9.378181418683854e-08, + "loss": 0.6892, + "step": 29055 + }, + { + "epoch": 1.06, + "learning_rate": 9.377815212216648e-08, + "loss": 0.696, + "step": 29056 + }, + { + "epoch": 1.06, + "learning_rate": 9.377449005749442e-08, + "loss": 0.688, + "step": 29057 + }, + { + "epoch": 1.06, + "learning_rate": 9.377082799282236e-08, + "loss": 0.6972, + "step": 29058 + }, + { + "epoch": 1.06, + "learning_rate": 9.376716592815029e-08, + "loss": 0.6879, + "step": 29059 + }, + { + "epoch": 1.06, + "learning_rate": 9.376350386347823e-08, + "loss": 0.6921, + "step": 29060 + }, + { + "epoch": 1.06, + "learning_rate": 9.375984179880617e-08, + "loss": 0.6908, + "step": 29061 + }, + { + "epoch": 1.06, + "learning_rate": 9.37561797341341e-08, + "loss": 0.709, + "step": 29062 + }, + { + "epoch": 1.06, + "learning_rate": 9.375251766946203e-08, + "loss": 0.6951, + "step": 29063 + }, + { + "epoch": 1.06, + "learning_rate": 9.374885560478997e-08, + "loss": 0.679, + "step": 29064 + }, + { + "epoch": 1.06, + "learning_rate": 9.374519354011791e-08, + "loss": 0.6893, + "step": 29065 + }, + { + "epoch": 1.06, + "learning_rate": 9.374153147544585e-08, + "loss": 0.6872, + "step": 29066 + }, + { + "epoch": 1.06, + "learning_rate": 9.373786941077379e-08, + "loss": 0.6928, + "step": 29067 + }, + { + "epoch": 1.06, + "learning_rate": 9.373420734610172e-08, + "loss": 0.6868, + "step": 29068 + }, + { + "epoch": 1.06, + "learning_rate": 9.373054528142966e-08, + "loss": 0.7029, + "step": 29069 + }, + { + "epoch": 1.06, + "learning_rate": 9.37268832167576e-08, + "loss": 0.7168, + "step": 29070 + }, + { + "epoch": 1.06, + "learning_rate": 9.372322115208554e-08, + "loss": 0.6879, + "step": 29071 + }, + { + "epoch": 1.06, + "learning_rate": 9.371955908741348e-08, + "loss": 0.7093, + "step": 29072 + }, + { + "epoch": 1.06, + "learning_rate": 9.371589702274142e-08, + "loss": 0.6943, + "step": 29073 + }, + { + "epoch": 1.06, + "learning_rate": 9.371223495806936e-08, + "loss": 0.6922, + "step": 29074 + }, + { + "epoch": 1.06, + "learning_rate": 9.37085728933973e-08, + "loss": 0.6783, + "step": 29075 + }, + { + "epoch": 1.06, + "learning_rate": 9.370491082872524e-08, + "loss": 0.6803, + "step": 29076 + }, + { + "epoch": 1.06, + "learning_rate": 9.370124876405317e-08, + "loss": 0.69, + "step": 29077 + }, + { + "epoch": 1.06, + "learning_rate": 9.36975866993811e-08, + "loss": 0.6923, + "step": 29078 + }, + { + "epoch": 1.06, + "learning_rate": 9.369392463470905e-08, + "loss": 0.6874, + "step": 29079 + }, + { + "epoch": 1.06, + "learning_rate": 9.369026257003699e-08, + "loss": 0.7062, + "step": 29080 + }, + { + "epoch": 1.06, + "learning_rate": 9.368660050536493e-08, + "loss": 0.6917, + "step": 29081 + }, + { + "epoch": 1.06, + "learning_rate": 9.368293844069287e-08, + "loss": 0.6908, + "step": 29082 + }, + { + "epoch": 1.06, + "learning_rate": 9.36792763760208e-08, + "loss": 0.7018, + "step": 29083 + }, + { + "epoch": 1.06, + "learning_rate": 9.367561431134874e-08, + "loss": 0.7068, + "step": 29084 + }, + { + "epoch": 1.06, + "learning_rate": 9.367195224667667e-08, + "loss": 0.6976, + "step": 29085 + }, + { + "epoch": 1.06, + "learning_rate": 9.366829018200461e-08, + "loss": 0.697, + "step": 29086 + }, + { + "epoch": 1.06, + "learning_rate": 9.366462811733254e-08, + "loss": 0.6908, + "step": 29087 + }, + { + "epoch": 1.06, + "learning_rate": 9.366096605266048e-08, + "loss": 0.6987, + "step": 29088 + }, + { + "epoch": 1.06, + "learning_rate": 9.365730398798842e-08, + "loss": 0.6934, + "step": 29089 + }, + { + "epoch": 1.06, + "learning_rate": 9.365364192331636e-08, + "loss": 0.6924, + "step": 29090 + }, + { + "epoch": 1.06, + "learning_rate": 9.36499798586443e-08, + "loss": 0.701, + "step": 29091 + }, + { + "epoch": 1.06, + "learning_rate": 9.364631779397224e-08, + "loss": 0.7004, + "step": 29092 + }, + { + "epoch": 1.06, + "learning_rate": 9.364265572930018e-08, + "loss": 0.7007, + "step": 29093 + }, + { + "epoch": 1.06, + "learning_rate": 9.36389936646281e-08, + "loss": 0.7006, + "step": 29094 + }, + { + "epoch": 1.06, + "learning_rate": 9.363533159995604e-08, + "loss": 0.6962, + "step": 29095 + }, + { + "epoch": 1.06, + "learning_rate": 9.363166953528398e-08, + "loss": 0.6946, + "step": 29096 + }, + { + "epoch": 1.06, + "learning_rate": 9.362800747061192e-08, + "loss": 0.6876, + "step": 29097 + }, + { + "epoch": 1.06, + "learning_rate": 9.362434540593986e-08, + "loss": 0.6857, + "step": 29098 + }, + { + "epoch": 1.06, + "learning_rate": 9.36206833412678e-08, + "loss": 0.6987, + "step": 29099 + }, + { + "epoch": 1.06, + "learning_rate": 9.361702127659574e-08, + "loss": 0.6981, + "step": 29100 + }, + { + "epoch": 1.06, + "learning_rate": 9.361335921192368e-08, + "loss": 0.7099, + "step": 29101 + }, + { + "epoch": 1.06, + "learning_rate": 9.360969714725162e-08, + "loss": 0.6915, + "step": 29102 + }, + { + "epoch": 1.06, + "learning_rate": 9.360603508257955e-08, + "loss": 0.6964, + "step": 29103 + }, + { + "epoch": 1.06, + "learning_rate": 9.360237301790749e-08, + "loss": 0.6803, + "step": 29104 + }, + { + "epoch": 1.06, + "learning_rate": 9.359871095323543e-08, + "loss": 0.6883, + "step": 29105 + }, + { + "epoch": 1.06, + "learning_rate": 9.359504888856337e-08, + "loss": 0.6865, + "step": 29106 + }, + { + "epoch": 1.06, + "learning_rate": 9.359138682389131e-08, + "loss": 0.6994, + "step": 29107 + }, + { + "epoch": 1.06, + "learning_rate": 9.358772475921925e-08, + "loss": 0.6949, + "step": 29108 + }, + { + "epoch": 1.07, + "learning_rate": 9.358406269454718e-08, + "loss": 0.6953, + "step": 29109 + }, + { + "epoch": 1.07, + "learning_rate": 9.358040062987512e-08, + "loss": 0.7008, + "step": 29110 + }, + { + "epoch": 1.07, + "learning_rate": 9.357673856520306e-08, + "loss": 0.6998, + "step": 29111 + }, + { + "epoch": 1.07, + "learning_rate": 9.357307650053098e-08, + "loss": 0.6896, + "step": 29112 + }, + { + "epoch": 1.07, + "learning_rate": 9.356941443585892e-08, + "loss": 0.681, + "step": 29113 + }, + { + "epoch": 1.07, + "learning_rate": 9.356575237118686e-08, + "loss": 0.6861, + "step": 29114 + }, + { + "epoch": 1.07, + "learning_rate": 9.35620903065148e-08, + "loss": 0.7066, + "step": 29115 + }, + { + "epoch": 1.07, + "learning_rate": 9.355842824184274e-08, + "loss": 0.7031, + "step": 29116 + }, + { + "epoch": 1.07, + "learning_rate": 9.355476617717068e-08, + "loss": 0.6895, + "step": 29117 + }, + { + "epoch": 1.07, + "learning_rate": 9.355110411249862e-08, + "loss": 0.6874, + "step": 29118 + }, + { + "epoch": 1.07, + "learning_rate": 9.354744204782656e-08, + "loss": 0.6821, + "step": 29119 + }, + { + "epoch": 1.07, + "learning_rate": 9.35437799831545e-08, + "loss": 0.6961, + "step": 29120 + }, + { + "epoch": 1.07, + "learning_rate": 9.354011791848243e-08, + "loss": 0.6839, + "step": 29121 + }, + { + "epoch": 1.07, + "learning_rate": 9.353645585381037e-08, + "loss": 0.6891, + "step": 29122 + }, + { + "epoch": 1.07, + "learning_rate": 9.353279378913831e-08, + "loss": 0.6941, + "step": 29123 + }, + { + "epoch": 1.07, + "learning_rate": 9.352913172446625e-08, + "loss": 0.6998, + "step": 29124 + }, + { + "epoch": 1.07, + "learning_rate": 9.352546965979419e-08, + "loss": 0.6951, + "step": 29125 + }, + { + "epoch": 1.07, + "learning_rate": 9.352180759512213e-08, + "loss": 0.6823, + "step": 29126 + }, + { + "epoch": 1.07, + "learning_rate": 9.351814553045007e-08, + "loss": 0.692, + "step": 29127 + }, + { + "epoch": 1.07, + "learning_rate": 9.351448346577801e-08, + "loss": 0.6906, + "step": 29128 + }, + { + "epoch": 1.07, + "learning_rate": 9.351082140110595e-08, + "loss": 0.6926, + "step": 29129 + }, + { + "epoch": 1.07, + "learning_rate": 9.350715933643388e-08, + "loss": 0.6893, + "step": 29130 + }, + { + "epoch": 1.07, + "learning_rate": 9.350349727176182e-08, + "loss": 0.6924, + "step": 29131 + }, + { + "epoch": 1.07, + "learning_rate": 9.349983520708976e-08, + "loss": 0.6971, + "step": 29132 + }, + { + "epoch": 1.07, + "learning_rate": 9.34961731424177e-08, + "loss": 0.6922, + "step": 29133 + }, + { + "epoch": 1.07, + "learning_rate": 9.349251107774562e-08, + "loss": 0.6902, + "step": 29134 + }, + { + "epoch": 1.07, + "learning_rate": 9.348884901307356e-08, + "loss": 0.7024, + "step": 29135 + }, + { + "epoch": 1.07, + "learning_rate": 9.34851869484015e-08, + "loss": 0.6879, + "step": 29136 + }, + { + "epoch": 1.07, + "learning_rate": 9.348152488372944e-08, + "loss": 0.6789, + "step": 29137 + }, + { + "epoch": 1.07, + "learning_rate": 9.347786281905738e-08, + "loss": 0.7073, + "step": 29138 + }, + { + "epoch": 1.07, + "learning_rate": 9.347420075438531e-08, + "loss": 0.7021, + "step": 29139 + }, + { + "epoch": 1.07, + "learning_rate": 9.347053868971325e-08, + "loss": 0.6814, + "step": 29140 + }, + { + "epoch": 1.07, + "learning_rate": 9.346687662504119e-08, + "loss": 0.6911, + "step": 29141 + }, + { + "epoch": 1.07, + "learning_rate": 9.346321456036913e-08, + "loss": 0.6946, + "step": 29142 + }, + { + "epoch": 1.07, + "learning_rate": 9.345955249569707e-08, + "loss": 0.6815, + "step": 29143 + }, + { + "epoch": 1.07, + "learning_rate": 9.345589043102501e-08, + "loss": 0.6927, + "step": 29144 + }, + { + "epoch": 1.07, + "learning_rate": 9.345222836635295e-08, + "loss": 0.6854, + "step": 29145 + }, + { + "epoch": 1.07, + "learning_rate": 9.344856630168089e-08, + "loss": 0.7077, + "step": 29146 + }, + { + "epoch": 1.07, + "learning_rate": 9.344490423700883e-08, + "loss": 0.6958, + "step": 29147 + }, + { + "epoch": 1.07, + "learning_rate": 9.344124217233675e-08, + "loss": 0.6919, + "step": 29148 + }, + { + "epoch": 1.07, + "learning_rate": 9.34375801076647e-08, + "loss": 0.6906, + "step": 29149 + }, + { + "epoch": 1.07, + "learning_rate": 9.343391804299263e-08, + "loss": 0.6875, + "step": 29150 + }, + { + "epoch": 1.07, + "learning_rate": 9.343025597832057e-08, + "loss": 0.7075, + "step": 29151 + }, + { + "epoch": 1.07, + "learning_rate": 9.342659391364851e-08, + "loss": 0.6844, + "step": 29152 + }, + { + "epoch": 1.07, + "learning_rate": 9.342293184897645e-08, + "loss": 0.6944, + "step": 29153 + }, + { + "epoch": 1.07, + "learning_rate": 9.34192697843044e-08, + "loss": 0.6941, + "step": 29154 + }, + { + "epoch": 1.07, + "learning_rate": 9.341560771963233e-08, + "loss": 0.6966, + "step": 29155 + }, + { + "epoch": 1.07, + "learning_rate": 9.341194565496026e-08, + "loss": 0.7054, + "step": 29156 + }, + { + "epoch": 1.07, + "learning_rate": 9.34082835902882e-08, + "loss": 0.6875, + "step": 29157 + }, + { + "epoch": 1.07, + "learning_rate": 9.340462152561613e-08, + "loss": 0.697, + "step": 29158 + }, + { + "epoch": 1.07, + "learning_rate": 9.340095946094407e-08, + "loss": 0.6822, + "step": 29159 + }, + { + "epoch": 1.07, + "learning_rate": 9.339729739627201e-08, + "loss": 0.6857, + "step": 29160 + }, + { + "epoch": 1.07, + "learning_rate": 9.339363533159995e-08, + "loss": 0.6957, + "step": 29161 + }, + { + "epoch": 1.07, + "learning_rate": 9.338997326692789e-08, + "loss": 0.6984, + "step": 29162 + }, + { + "epoch": 1.07, + "learning_rate": 9.338631120225583e-08, + "loss": 0.6862, + "step": 29163 + }, + { + "epoch": 1.07, + "learning_rate": 9.338264913758377e-08, + "loss": 0.7034, + "step": 29164 + }, + { + "epoch": 1.07, + "learning_rate": 9.337898707291171e-08, + "loss": 0.6813, + "step": 29165 + }, + { + "epoch": 1.07, + "learning_rate": 9.337532500823963e-08, + "loss": 0.6773, + "step": 29166 + }, + { + "epoch": 1.07, + "learning_rate": 9.337166294356757e-08, + "loss": 0.6784, + "step": 29167 + }, + { + "epoch": 1.07, + "learning_rate": 9.336800087889551e-08, + "loss": 0.6745, + "step": 29168 + }, + { + "epoch": 1.07, + "learning_rate": 9.336433881422345e-08, + "loss": 0.6958, + "step": 29169 + }, + { + "epoch": 1.07, + "learning_rate": 9.33606767495514e-08, + "loss": 0.697, + "step": 29170 + }, + { + "epoch": 1.07, + "learning_rate": 9.335701468487933e-08, + "loss": 0.6929, + "step": 29171 + }, + { + "epoch": 1.07, + "learning_rate": 9.335335262020727e-08, + "loss": 0.6964, + "step": 29172 + }, + { + "epoch": 1.07, + "learning_rate": 9.334969055553521e-08, + "loss": 0.693, + "step": 29173 + }, + { + "epoch": 1.07, + "learning_rate": 9.334602849086315e-08, + "loss": 0.6994, + "step": 29174 + }, + { + "epoch": 1.07, + "learning_rate": 9.334236642619108e-08, + "loss": 0.6747, + "step": 29175 + }, + { + "epoch": 1.07, + "learning_rate": 9.333870436151902e-08, + "loss": 0.7004, + "step": 29176 + }, + { + "epoch": 1.07, + "learning_rate": 9.333504229684696e-08, + "loss": 0.6996, + "step": 29177 + }, + { + "epoch": 1.07, + "learning_rate": 9.33313802321749e-08, + "loss": 0.7007, + "step": 29178 + }, + { + "epoch": 1.07, + "learning_rate": 9.332771816750284e-08, + "loss": 0.6973, + "step": 29179 + }, + { + "epoch": 1.07, + "learning_rate": 9.332405610283077e-08, + "loss": 0.711, + "step": 29180 + }, + { + "epoch": 1.07, + "learning_rate": 9.33203940381587e-08, + "loss": 0.687, + "step": 29181 + }, + { + "epoch": 1.07, + "learning_rate": 9.331673197348665e-08, + "loss": 0.699, + "step": 29182 + }, + { + "epoch": 1.07, + "learning_rate": 9.331306990881459e-08, + "loss": 0.6949, + "step": 29183 + }, + { + "epoch": 1.07, + "learning_rate": 9.330940784414251e-08, + "loss": 0.6986, + "step": 29184 + }, + { + "epoch": 1.07, + "learning_rate": 9.330574577947045e-08, + "loss": 0.6913, + "step": 29185 + }, + { + "epoch": 1.07, + "learning_rate": 9.330208371479839e-08, + "loss": 0.6953, + "step": 29186 + }, + { + "epoch": 1.07, + "learning_rate": 9.329842165012633e-08, + "loss": 0.6861, + "step": 29187 + }, + { + "epoch": 1.07, + "learning_rate": 9.329475958545427e-08, + "loss": 0.6952, + "step": 29188 + }, + { + "epoch": 1.07, + "learning_rate": 9.329109752078221e-08, + "loss": 0.6906, + "step": 29189 + }, + { + "epoch": 1.07, + "learning_rate": 9.328743545611015e-08, + "loss": 0.6869, + "step": 29190 + }, + { + "epoch": 1.07, + "learning_rate": 9.328377339143809e-08, + "loss": 0.6939, + "step": 29191 + }, + { + "epoch": 1.07, + "learning_rate": 9.328011132676603e-08, + "loss": 0.6821, + "step": 29192 + }, + { + "epoch": 1.07, + "learning_rate": 9.327644926209396e-08, + "loss": 0.6937, + "step": 29193 + }, + { + "epoch": 1.07, + "learning_rate": 9.32727871974219e-08, + "loss": 0.6885, + "step": 29194 + }, + { + "epoch": 1.07, + "learning_rate": 9.326912513274984e-08, + "loss": 0.6883, + "step": 29195 + }, + { + "epoch": 1.07, + "learning_rate": 9.326546306807778e-08, + "loss": 0.689, + "step": 29196 + }, + { + "epoch": 1.07, + "learning_rate": 9.326180100340572e-08, + "loss": 0.697, + "step": 29197 + }, + { + "epoch": 1.07, + "learning_rate": 9.325813893873366e-08, + "loss": 0.6989, + "step": 29198 + }, + { + "epoch": 1.07, + "learning_rate": 9.32544768740616e-08, + "loss": 0.7044, + "step": 29199 + }, + { + "epoch": 1.07, + "learning_rate": 9.325081480938954e-08, + "loss": 0.6808, + "step": 29200 + }, + { + "epoch": 1.07, + "learning_rate": 9.324715274471748e-08, + "loss": 0.6946, + "step": 29201 + }, + { + "epoch": 1.07, + "learning_rate": 9.32434906800454e-08, + "loss": 0.6915, + "step": 29202 + }, + { + "epoch": 1.07, + "learning_rate": 9.323982861537334e-08, + "loss": 0.6892, + "step": 29203 + }, + { + "epoch": 1.07, + "learning_rate": 9.323616655070128e-08, + "loss": 0.6926, + "step": 29204 + }, + { + "epoch": 1.07, + "learning_rate": 9.323250448602921e-08, + "loss": 0.6948, + "step": 29205 + }, + { + "epoch": 1.07, + "learning_rate": 9.322884242135715e-08, + "loss": 0.6899, + "step": 29206 + }, + { + "epoch": 1.07, + "learning_rate": 9.322518035668509e-08, + "loss": 0.6863, + "step": 29207 + }, + { + "epoch": 1.07, + "learning_rate": 9.322151829201303e-08, + "loss": 0.6917, + "step": 29208 + }, + { + "epoch": 1.07, + "learning_rate": 9.321785622734097e-08, + "loss": 0.7025, + "step": 29209 + }, + { + "epoch": 1.07, + "learning_rate": 9.321419416266891e-08, + "loss": 0.6996, + "step": 29210 + }, + { + "epoch": 1.07, + "learning_rate": 9.321053209799684e-08, + "loss": 0.697, + "step": 29211 + }, + { + "epoch": 1.07, + "learning_rate": 9.320687003332478e-08, + "loss": 0.7052, + "step": 29212 + }, + { + "epoch": 1.07, + "learning_rate": 9.320320796865272e-08, + "loss": 0.6903, + "step": 29213 + }, + { + "epoch": 1.07, + "learning_rate": 9.319954590398066e-08, + "loss": 0.7004, + "step": 29214 + }, + { + "epoch": 1.07, + "learning_rate": 9.31958838393086e-08, + "loss": 0.6982, + "step": 29215 + }, + { + "epoch": 1.07, + "learning_rate": 9.319222177463654e-08, + "loss": 0.685, + "step": 29216 + }, + { + "epoch": 1.07, + "learning_rate": 9.318855970996448e-08, + "loss": 0.6943, + "step": 29217 + }, + { + "epoch": 1.07, + "learning_rate": 9.318489764529242e-08, + "loss": 0.6973, + "step": 29218 + }, + { + "epoch": 1.07, + "learning_rate": 9.318123558062036e-08, + "loss": 0.6919, + "step": 29219 + }, + { + "epoch": 1.07, + "learning_rate": 9.317757351594828e-08, + "loss": 0.6975, + "step": 29220 + }, + { + "epoch": 1.07, + "learning_rate": 9.317391145127622e-08, + "loss": 0.6918, + "step": 29221 + }, + { + "epoch": 1.07, + "learning_rate": 9.317024938660416e-08, + "loss": 0.6896, + "step": 29222 + }, + { + "epoch": 1.07, + "learning_rate": 9.31665873219321e-08, + "loss": 0.7013, + "step": 29223 + }, + { + "epoch": 1.07, + "learning_rate": 9.316292525726004e-08, + "loss": 0.7017, + "step": 29224 + }, + { + "epoch": 1.07, + "learning_rate": 9.315926319258798e-08, + "loss": 0.6914, + "step": 29225 + }, + { + "epoch": 1.07, + "learning_rate": 9.315560112791592e-08, + "loss": 0.694, + "step": 29226 + }, + { + "epoch": 1.07, + "learning_rate": 9.315193906324385e-08, + "loss": 0.6935, + "step": 29227 + }, + { + "epoch": 1.07, + "learning_rate": 9.314827699857179e-08, + "loss": 0.701, + "step": 29228 + }, + { + "epoch": 1.07, + "learning_rate": 9.314461493389972e-08, + "loss": 0.7021, + "step": 29229 + }, + { + "epoch": 1.07, + "learning_rate": 9.314095286922766e-08, + "loss": 0.6908, + "step": 29230 + }, + { + "epoch": 1.07, + "learning_rate": 9.31372908045556e-08, + "loss": 0.7021, + "step": 29231 + }, + { + "epoch": 1.07, + "learning_rate": 9.313362873988354e-08, + "loss": 0.7019, + "step": 29232 + }, + { + "epoch": 1.07, + "learning_rate": 9.312996667521148e-08, + "loss": 0.6865, + "step": 29233 + }, + { + "epoch": 1.07, + "learning_rate": 9.312630461053942e-08, + "loss": 0.6976, + "step": 29234 + }, + { + "epoch": 1.07, + "learning_rate": 9.312264254586736e-08, + "loss": 0.6897, + "step": 29235 + }, + { + "epoch": 1.07, + "learning_rate": 9.31189804811953e-08, + "loss": 0.6976, + "step": 29236 + }, + { + "epoch": 1.07, + "learning_rate": 9.311531841652324e-08, + "loss": 0.6981, + "step": 29237 + }, + { + "epoch": 1.07, + "learning_rate": 9.311165635185116e-08, + "loss": 0.6892, + "step": 29238 + }, + { + "epoch": 1.07, + "learning_rate": 9.31079942871791e-08, + "loss": 0.7031, + "step": 29239 + }, + { + "epoch": 1.07, + "learning_rate": 9.310433222250704e-08, + "loss": 0.703, + "step": 29240 + }, + { + "epoch": 1.07, + "learning_rate": 9.310067015783498e-08, + "loss": 0.6967, + "step": 29241 + }, + { + "epoch": 1.07, + "learning_rate": 9.309700809316292e-08, + "loss": 0.6901, + "step": 29242 + }, + { + "epoch": 1.07, + "learning_rate": 9.309334602849086e-08, + "loss": 0.6797, + "step": 29243 + }, + { + "epoch": 1.07, + "learning_rate": 9.30896839638188e-08, + "loss": 0.6884, + "step": 29244 + }, + { + "epoch": 1.07, + "learning_rate": 9.308602189914674e-08, + "loss": 0.693, + "step": 29245 + }, + { + "epoch": 1.07, + "learning_rate": 9.308235983447468e-08, + "loss": 0.6975, + "step": 29246 + }, + { + "epoch": 1.07, + "learning_rate": 9.307869776980261e-08, + "loss": 0.6856, + "step": 29247 + }, + { + "epoch": 1.07, + "learning_rate": 9.307503570513055e-08, + "loss": 0.6884, + "step": 29248 + }, + { + "epoch": 1.07, + "learning_rate": 9.307137364045849e-08, + "loss": 0.6902, + "step": 29249 + }, + { + "epoch": 1.07, + "learning_rate": 9.306771157578643e-08, + "loss": 0.7062, + "step": 29250 + }, + { + "epoch": 1.07, + "learning_rate": 9.306404951111436e-08, + "loss": 0.692, + "step": 29251 + }, + { + "epoch": 1.07, + "learning_rate": 9.30603874464423e-08, + "loss": 0.6833, + "step": 29252 + }, + { + "epoch": 1.07, + "learning_rate": 9.305672538177024e-08, + "loss": 0.6989, + "step": 29253 + }, + { + "epoch": 1.07, + "learning_rate": 9.305306331709818e-08, + "loss": 0.7087, + "step": 29254 + }, + { + "epoch": 1.07, + "learning_rate": 9.304940125242612e-08, + "loss": 0.6913, + "step": 29255 + }, + { + "epoch": 1.07, + "learning_rate": 9.304573918775404e-08, + "loss": 0.6935, + "step": 29256 + }, + { + "epoch": 1.07, + "learning_rate": 9.304207712308198e-08, + "loss": 0.6996, + "step": 29257 + }, + { + "epoch": 1.07, + "learning_rate": 9.303841505840992e-08, + "loss": 0.703, + "step": 29258 + }, + { + "epoch": 1.07, + "learning_rate": 9.303475299373786e-08, + "loss": 0.6967, + "step": 29259 + }, + { + "epoch": 1.07, + "learning_rate": 9.30310909290658e-08, + "loss": 0.6982, + "step": 29260 + }, + { + "epoch": 1.07, + "learning_rate": 9.302742886439374e-08, + "loss": 0.692, + "step": 29261 + }, + { + "epoch": 1.07, + "learning_rate": 9.302376679972168e-08, + "loss": 0.7038, + "step": 29262 + }, + { + "epoch": 1.07, + "learning_rate": 9.302010473504962e-08, + "loss": 0.6872, + "step": 29263 + }, + { + "epoch": 1.07, + "learning_rate": 9.301644267037756e-08, + "loss": 0.6994, + "step": 29264 + }, + { + "epoch": 1.07, + "learning_rate": 9.301278060570549e-08, + "loss": 0.7017, + "step": 29265 + }, + { + "epoch": 1.07, + "learning_rate": 9.300911854103343e-08, + "loss": 0.6913, + "step": 29266 + }, + { + "epoch": 1.07, + "learning_rate": 9.300545647636137e-08, + "loss": 0.6961, + "step": 29267 + }, + { + "epoch": 1.07, + "learning_rate": 9.300179441168931e-08, + "loss": 0.6798, + "step": 29268 + }, + { + "epoch": 1.07, + "learning_rate": 9.299813234701725e-08, + "loss": 0.6689, + "step": 29269 + }, + { + "epoch": 1.07, + "learning_rate": 9.299447028234519e-08, + "loss": 0.6829, + "step": 29270 + }, + { + "epoch": 1.07, + "learning_rate": 9.299080821767313e-08, + "loss": 0.6939, + "step": 29271 + }, + { + "epoch": 1.07, + "learning_rate": 9.298714615300107e-08, + "loss": 0.6841, + "step": 29272 + }, + { + "epoch": 1.07, + "learning_rate": 9.2983484088329e-08, + "loss": 0.6861, + "step": 29273 + }, + { + "epoch": 1.07, + "learning_rate": 9.297982202365693e-08, + "loss": 0.6977, + "step": 29274 + }, + { + "epoch": 1.07, + "learning_rate": 9.297615995898487e-08, + "loss": 0.6844, + "step": 29275 + }, + { + "epoch": 1.07, + "learning_rate": 9.29724978943128e-08, + "loss": 0.7087, + "step": 29276 + }, + { + "epoch": 1.07, + "learning_rate": 9.296883582964074e-08, + "loss": 0.6936, + "step": 29277 + }, + { + "epoch": 1.07, + "learning_rate": 9.296517376496868e-08, + "loss": 0.6891, + "step": 29278 + }, + { + "epoch": 1.07, + "learning_rate": 9.296151170029662e-08, + "loss": 0.6858, + "step": 29279 + }, + { + "epoch": 1.07, + "learning_rate": 9.295784963562456e-08, + "loss": 0.6847, + "step": 29280 + }, + { + "epoch": 1.07, + "learning_rate": 9.29541875709525e-08, + "loss": 0.6967, + "step": 29281 + }, + { + "epoch": 1.07, + "learning_rate": 9.295052550628044e-08, + "loss": 0.6963, + "step": 29282 + }, + { + "epoch": 1.07, + "learning_rate": 9.294686344160837e-08, + "loss": 0.6934, + "step": 29283 + }, + { + "epoch": 1.07, + "learning_rate": 9.294320137693631e-08, + "loss": 0.7037, + "step": 29284 + }, + { + "epoch": 1.07, + "learning_rate": 9.293953931226425e-08, + "loss": 0.688, + "step": 29285 + }, + { + "epoch": 1.07, + "learning_rate": 9.293587724759219e-08, + "loss": 0.7072, + "step": 29286 + }, + { + "epoch": 1.07, + "learning_rate": 9.293221518292013e-08, + "loss": 0.7033, + "step": 29287 + }, + { + "epoch": 1.07, + "learning_rate": 9.292855311824807e-08, + "loss": 0.6929, + "step": 29288 + }, + { + "epoch": 1.07, + "learning_rate": 9.292489105357601e-08, + "loss": 0.7114, + "step": 29289 + }, + { + "epoch": 1.07, + "learning_rate": 9.292122898890395e-08, + "loss": 0.7001, + "step": 29290 + }, + { + "epoch": 1.07, + "learning_rate": 9.291756692423189e-08, + "loss": 0.6954, + "step": 29291 + }, + { + "epoch": 1.07, + "learning_rate": 9.291390485955981e-08, + "loss": 0.696, + "step": 29292 + }, + { + "epoch": 1.07, + "learning_rate": 9.291024279488775e-08, + "loss": 0.7047, + "step": 29293 + }, + { + "epoch": 1.07, + "learning_rate": 9.290658073021569e-08, + "loss": 0.693, + "step": 29294 + }, + { + "epoch": 1.07, + "learning_rate": 9.290291866554363e-08, + "loss": 0.6879, + "step": 29295 + }, + { + "epoch": 1.07, + "learning_rate": 9.289925660087157e-08, + "loss": 0.6945, + "step": 29296 + }, + { + "epoch": 1.07, + "learning_rate": 9.28955945361995e-08, + "loss": 0.6844, + "step": 29297 + }, + { + "epoch": 1.07, + "learning_rate": 9.289193247152744e-08, + "loss": 0.6958, + "step": 29298 + }, + { + "epoch": 1.07, + "learning_rate": 9.288827040685538e-08, + "loss": 0.6932, + "step": 29299 + }, + { + "epoch": 1.07, + "learning_rate": 9.288460834218332e-08, + "loss": 0.6973, + "step": 29300 + }, + { + "epoch": 1.07, + "learning_rate": 9.288094627751125e-08, + "loss": 0.6956, + "step": 29301 + }, + { + "epoch": 1.07, + "learning_rate": 9.287728421283919e-08, + "loss": 0.6908, + "step": 29302 + }, + { + "epoch": 1.07, + "learning_rate": 9.287362214816713e-08, + "loss": 0.6956, + "step": 29303 + }, + { + "epoch": 1.07, + "learning_rate": 9.286996008349507e-08, + "loss": 0.6918, + "step": 29304 + }, + { + "epoch": 1.07, + "learning_rate": 9.2866298018823e-08, + "loss": 0.6955, + "step": 29305 + }, + { + "epoch": 1.07, + "learning_rate": 9.286263595415095e-08, + "loss": 0.6838, + "step": 29306 + }, + { + "epoch": 1.07, + "learning_rate": 9.285897388947889e-08, + "loss": 0.699, + "step": 29307 + }, + { + "epoch": 1.07, + "learning_rate": 9.285531182480683e-08, + "loss": 0.6949, + "step": 29308 + }, + { + "epoch": 1.07, + "learning_rate": 9.285164976013477e-08, + "loss": 0.6871, + "step": 29309 + }, + { + "epoch": 1.07, + "learning_rate": 9.284798769546269e-08, + "loss": 0.6937, + "step": 29310 + }, + { + "epoch": 1.07, + "learning_rate": 9.284432563079063e-08, + "loss": 0.6929, + "step": 29311 + }, + { + "epoch": 1.07, + "learning_rate": 9.284066356611857e-08, + "loss": 0.697, + "step": 29312 + }, + { + "epoch": 1.07, + "learning_rate": 9.283700150144651e-08, + "loss": 0.7115, + "step": 29313 + }, + { + "epoch": 1.07, + "learning_rate": 9.283333943677445e-08, + "loss": 0.6856, + "step": 29314 + }, + { + "epoch": 1.07, + "learning_rate": 9.282967737210239e-08, + "loss": 0.6986, + "step": 29315 + }, + { + "epoch": 1.07, + "learning_rate": 9.282601530743033e-08, + "loss": 0.6981, + "step": 29316 + }, + { + "epoch": 1.07, + "learning_rate": 9.282235324275827e-08, + "loss": 0.6949, + "step": 29317 + }, + { + "epoch": 1.07, + "learning_rate": 9.281869117808621e-08, + "loss": 0.6902, + "step": 29318 + }, + { + "epoch": 1.07, + "learning_rate": 9.281502911341414e-08, + "loss": 0.7046, + "step": 29319 + }, + { + "epoch": 1.07, + "learning_rate": 9.281136704874208e-08, + "loss": 0.7047, + "step": 29320 + }, + { + "epoch": 1.07, + "learning_rate": 9.280770498407002e-08, + "loss": 0.7005, + "step": 29321 + }, + { + "epoch": 1.07, + "learning_rate": 9.280404291939795e-08, + "loss": 0.6939, + "step": 29322 + }, + { + "epoch": 1.07, + "learning_rate": 9.280038085472588e-08, + "loss": 0.6729, + "step": 29323 + }, + { + "epoch": 1.07, + "learning_rate": 9.279671879005382e-08, + "loss": 0.6896, + "step": 29324 + }, + { + "epoch": 1.07, + "learning_rate": 9.279305672538176e-08, + "loss": 0.7014, + "step": 29325 + }, + { + "epoch": 1.07, + "learning_rate": 9.27893946607097e-08, + "loss": 0.6875, + "step": 29326 + }, + { + "epoch": 1.07, + "learning_rate": 9.278573259603764e-08, + "loss": 0.696, + "step": 29327 + }, + { + "epoch": 1.07, + "learning_rate": 9.278207053136557e-08, + "loss": 0.7137, + "step": 29328 + }, + { + "epoch": 1.07, + "learning_rate": 9.277840846669351e-08, + "loss": 0.6978, + "step": 29329 + }, + { + "epoch": 1.07, + "learning_rate": 9.277474640202145e-08, + "loss": 0.6884, + "step": 29330 + }, + { + "epoch": 1.07, + "learning_rate": 9.277108433734939e-08, + "loss": 0.6856, + "step": 29331 + }, + { + "epoch": 1.07, + "learning_rate": 9.276742227267733e-08, + "loss": 0.6934, + "step": 29332 + }, + { + "epoch": 1.07, + "learning_rate": 9.276376020800527e-08, + "loss": 0.7036, + "step": 29333 + }, + { + "epoch": 1.07, + "learning_rate": 9.276009814333321e-08, + "loss": 0.69, + "step": 29334 + }, + { + "epoch": 1.07, + "learning_rate": 9.275643607866115e-08, + "loss": 0.6811, + "step": 29335 + }, + { + "epoch": 1.07, + "learning_rate": 9.275277401398909e-08, + "loss": 0.6963, + "step": 29336 + }, + { + "epoch": 1.07, + "learning_rate": 9.274911194931702e-08, + "loss": 0.6885, + "step": 29337 + }, + { + "epoch": 1.07, + "learning_rate": 9.274544988464496e-08, + "loss": 0.6953, + "step": 29338 + }, + { + "epoch": 1.07, + "learning_rate": 9.27417878199729e-08, + "loss": 0.6905, + "step": 29339 + }, + { + "epoch": 1.07, + "learning_rate": 9.273812575530084e-08, + "loss": 0.7045, + "step": 29340 + }, + { + "epoch": 1.07, + "learning_rate": 9.273446369062878e-08, + "loss": 0.6822, + "step": 29341 + }, + { + "epoch": 1.07, + "learning_rate": 9.273080162595672e-08, + "loss": 0.6806, + "step": 29342 + }, + { + "epoch": 1.07, + "learning_rate": 9.272713956128466e-08, + "loss": 0.6809, + "step": 29343 + }, + { + "epoch": 1.07, + "learning_rate": 9.272347749661258e-08, + "loss": 0.6933, + "step": 29344 + }, + { + "epoch": 1.07, + "learning_rate": 9.271981543194052e-08, + "loss": 0.6975, + "step": 29345 + }, + { + "epoch": 1.07, + "learning_rate": 9.271615336726846e-08, + "loss": 0.696, + "step": 29346 + }, + { + "epoch": 1.07, + "learning_rate": 9.271249130259639e-08, + "loss": 0.6998, + "step": 29347 + }, + { + "epoch": 1.07, + "learning_rate": 9.270882923792433e-08, + "loss": 0.6952, + "step": 29348 + }, + { + "epoch": 1.07, + "learning_rate": 9.270516717325227e-08, + "loss": 0.6943, + "step": 29349 + }, + { + "epoch": 1.07, + "learning_rate": 9.270150510858021e-08, + "loss": 0.7087, + "step": 29350 + }, + { + "epoch": 1.07, + "learning_rate": 9.269784304390815e-08, + "loss": 0.7001, + "step": 29351 + }, + { + "epoch": 1.07, + "learning_rate": 9.269418097923609e-08, + "loss": 0.6863, + "step": 29352 + }, + { + "epoch": 1.07, + "learning_rate": 9.269051891456403e-08, + "loss": 0.6991, + "step": 29353 + }, + { + "epoch": 1.07, + "learning_rate": 9.268685684989197e-08, + "loss": 0.6998, + "step": 29354 + }, + { + "epoch": 1.07, + "learning_rate": 9.26831947852199e-08, + "loss": 0.6945, + "step": 29355 + }, + { + "epoch": 1.07, + "learning_rate": 9.267953272054784e-08, + "loss": 0.6878, + "step": 29356 + }, + { + "epoch": 1.07, + "learning_rate": 9.267587065587578e-08, + "loss": 0.6915, + "step": 29357 + }, + { + "epoch": 1.07, + "learning_rate": 9.267220859120372e-08, + "loss": 0.6936, + "step": 29358 + }, + { + "epoch": 1.07, + "learning_rate": 9.266854652653166e-08, + "loss": 0.6826, + "step": 29359 + }, + { + "epoch": 1.07, + "learning_rate": 9.26648844618596e-08, + "loss": 0.6969, + "step": 29360 + }, + { + "epoch": 1.07, + "learning_rate": 9.266122239718754e-08, + "loss": 0.6779, + "step": 29361 + }, + { + "epoch": 1.07, + "learning_rate": 9.265756033251548e-08, + "loss": 0.6888, + "step": 29362 + }, + { + "epoch": 1.07, + "learning_rate": 9.265389826784342e-08, + "loss": 0.688, + "step": 29363 + }, + { + "epoch": 1.07, + "learning_rate": 9.265023620317134e-08, + "loss": 0.7058, + "step": 29364 + }, + { + "epoch": 1.07, + "learning_rate": 9.264657413849928e-08, + "loss": 0.7088, + "step": 29365 + }, + { + "epoch": 1.07, + "learning_rate": 9.264291207382722e-08, + "loss": 0.6881, + "step": 29366 + }, + { + "epoch": 1.07, + "learning_rate": 9.263925000915516e-08, + "loss": 0.6855, + "step": 29367 + }, + { + "epoch": 1.07, + "learning_rate": 9.263558794448309e-08, + "loss": 0.6848, + "step": 29368 + }, + { + "epoch": 1.07, + "learning_rate": 9.263192587981103e-08, + "loss": 0.6951, + "step": 29369 + }, + { + "epoch": 1.07, + "learning_rate": 9.262826381513897e-08, + "loss": 0.7031, + "step": 29370 + }, + { + "epoch": 1.07, + "learning_rate": 9.262460175046691e-08, + "loss": 0.6929, + "step": 29371 + }, + { + "epoch": 1.07, + "learning_rate": 9.262093968579485e-08, + "loss": 0.6935, + "step": 29372 + }, + { + "epoch": 1.07, + "learning_rate": 9.261727762112278e-08, + "loss": 0.7009, + "step": 29373 + }, + { + "epoch": 1.07, + "learning_rate": 9.261361555645072e-08, + "loss": 0.6999, + "step": 29374 + }, + { + "epoch": 1.07, + "learning_rate": 9.260995349177866e-08, + "loss": 0.6906, + "step": 29375 + }, + { + "epoch": 1.07, + "learning_rate": 9.26062914271066e-08, + "loss": 0.698, + "step": 29376 + }, + { + "epoch": 1.07, + "learning_rate": 9.260262936243454e-08, + "loss": 0.7033, + "step": 29377 + }, + { + "epoch": 1.07, + "learning_rate": 9.259896729776248e-08, + "loss": 0.7026, + "step": 29378 + }, + { + "epoch": 1.07, + "learning_rate": 9.259530523309042e-08, + "loss": 0.6872, + "step": 29379 + }, + { + "epoch": 1.07, + "learning_rate": 9.259164316841836e-08, + "loss": 0.7013, + "step": 29380 + }, + { + "epoch": 1.07, + "learning_rate": 9.25879811037463e-08, + "loss": 0.6879, + "step": 29381 + }, + { + "epoch": 1.08, + "learning_rate": 9.258431903907422e-08, + "loss": 0.6847, + "step": 29382 + }, + { + "epoch": 1.08, + "learning_rate": 9.258065697440216e-08, + "loss": 0.6943, + "step": 29383 + }, + { + "epoch": 1.08, + "learning_rate": 9.25769949097301e-08, + "loss": 0.7032, + "step": 29384 + }, + { + "epoch": 1.08, + "learning_rate": 9.257333284505804e-08, + "loss": 0.6958, + "step": 29385 + }, + { + "epoch": 1.08, + "learning_rate": 9.256967078038598e-08, + "loss": 0.677, + "step": 29386 + }, + { + "epoch": 1.08, + "learning_rate": 9.256600871571392e-08, + "loss": 0.7077, + "step": 29387 + }, + { + "epoch": 1.08, + "learning_rate": 9.256234665104186e-08, + "loss": 0.7095, + "step": 29388 + }, + { + "epoch": 1.08, + "learning_rate": 9.25586845863698e-08, + "loss": 0.6982, + "step": 29389 + }, + { + "epoch": 1.08, + "learning_rate": 9.255502252169773e-08, + "loss": 0.7119, + "step": 29390 + }, + { + "epoch": 1.08, + "learning_rate": 9.255136045702567e-08, + "loss": 0.6895, + "step": 29391 + }, + { + "epoch": 1.08, + "learning_rate": 9.254769839235361e-08, + "loss": 0.6767, + "step": 29392 + }, + { + "epoch": 1.08, + "learning_rate": 9.254403632768153e-08, + "loss": 0.6948, + "step": 29393 + }, + { + "epoch": 1.08, + "learning_rate": 9.254037426300947e-08, + "loss": 0.693, + "step": 29394 + }, + { + "epoch": 1.08, + "learning_rate": 9.253671219833741e-08, + "loss": 0.6965, + "step": 29395 + }, + { + "epoch": 1.08, + "learning_rate": 9.253305013366535e-08, + "loss": 0.692, + "step": 29396 + }, + { + "epoch": 1.08, + "learning_rate": 9.25293880689933e-08, + "loss": 0.694, + "step": 29397 + }, + { + "epoch": 1.08, + "learning_rate": 9.252572600432123e-08, + "loss": 0.6984, + "step": 29398 + }, + { + "epoch": 1.08, + "learning_rate": 9.252206393964917e-08, + "loss": 0.6851, + "step": 29399 + }, + { + "epoch": 1.08, + "learning_rate": 9.25184018749771e-08, + "loss": 0.6878, + "step": 29400 + }, + { + "epoch": 1.08, + "learning_rate": 9.251473981030504e-08, + "loss": 0.6949, + "step": 29401 + }, + { + "epoch": 1.08, + "learning_rate": 9.251107774563298e-08, + "loss": 0.6823, + "step": 29402 + }, + { + "epoch": 1.08, + "learning_rate": 9.250741568096092e-08, + "loss": 0.6876, + "step": 29403 + }, + { + "epoch": 1.08, + "learning_rate": 9.250375361628886e-08, + "loss": 0.6955, + "step": 29404 + }, + { + "epoch": 1.08, + "learning_rate": 9.25000915516168e-08, + "loss": 0.6947, + "step": 29405 + }, + { + "epoch": 1.08, + "learning_rate": 9.249642948694474e-08, + "loss": 0.6931, + "step": 29406 + }, + { + "epoch": 1.08, + "learning_rate": 9.249276742227268e-08, + "loss": 0.6855, + "step": 29407 + }, + { + "epoch": 1.08, + "learning_rate": 9.248910535760062e-08, + "loss": 0.6935, + "step": 29408 + }, + { + "epoch": 1.08, + "learning_rate": 9.248544329292855e-08, + "loss": 0.697, + "step": 29409 + }, + { + "epoch": 1.08, + "learning_rate": 9.248178122825649e-08, + "loss": 0.6957, + "step": 29410 + }, + { + "epoch": 1.08, + "learning_rate": 9.247811916358443e-08, + "loss": 0.697, + "step": 29411 + }, + { + "epoch": 1.08, + "learning_rate": 9.247445709891237e-08, + "loss": 0.6961, + "step": 29412 + }, + { + "epoch": 1.08, + "learning_rate": 9.24707950342403e-08, + "loss": 0.6826, + "step": 29413 + }, + { + "epoch": 1.08, + "learning_rate": 9.246713296956825e-08, + "loss": 0.6873, + "step": 29414 + }, + { + "epoch": 1.08, + "learning_rate": 9.246347090489617e-08, + "loss": 0.6943, + "step": 29415 + }, + { + "epoch": 1.08, + "learning_rate": 9.245980884022411e-08, + "loss": 0.6827, + "step": 29416 + }, + { + "epoch": 1.08, + "learning_rate": 9.245614677555205e-08, + "loss": 0.6839, + "step": 29417 + }, + { + "epoch": 1.08, + "learning_rate": 9.245248471087998e-08, + "loss": 0.6765, + "step": 29418 + }, + { + "epoch": 1.08, + "learning_rate": 9.244882264620792e-08, + "loss": 0.6971, + "step": 29419 + }, + { + "epoch": 1.08, + "learning_rate": 9.244516058153586e-08, + "loss": 0.6851, + "step": 29420 + }, + { + "epoch": 1.08, + "learning_rate": 9.24414985168638e-08, + "loss": 0.6845, + "step": 29421 + }, + { + "epoch": 1.08, + "learning_rate": 9.243783645219174e-08, + "loss": 0.7031, + "step": 29422 + }, + { + "epoch": 1.08, + "learning_rate": 9.243417438751968e-08, + "loss": 0.6986, + "step": 29423 + }, + { + "epoch": 1.08, + "learning_rate": 9.243051232284762e-08, + "loss": 0.6847, + "step": 29424 + }, + { + "epoch": 1.08, + "learning_rate": 9.242685025817556e-08, + "loss": 0.6912, + "step": 29425 + }, + { + "epoch": 1.08, + "learning_rate": 9.24231881935035e-08, + "loss": 0.6891, + "step": 29426 + }, + { + "epoch": 1.08, + "learning_rate": 9.241952612883143e-08, + "loss": 0.6938, + "step": 29427 + }, + { + "epoch": 1.08, + "learning_rate": 9.241586406415937e-08, + "loss": 0.706, + "step": 29428 + }, + { + "epoch": 1.08, + "learning_rate": 9.24122019994873e-08, + "loss": 0.7006, + "step": 29429 + }, + { + "epoch": 1.08, + "learning_rate": 9.240853993481525e-08, + "loss": 0.7008, + "step": 29430 + }, + { + "epoch": 1.08, + "learning_rate": 9.240487787014319e-08, + "loss": 0.6859, + "step": 29431 + }, + { + "epoch": 1.08, + "learning_rate": 9.240121580547113e-08, + "loss": 0.6891, + "step": 29432 + }, + { + "epoch": 1.08, + "learning_rate": 9.239755374079907e-08, + "loss": 0.6946, + "step": 29433 + }, + { + "epoch": 1.08, + "learning_rate": 9.2393891676127e-08, + "loss": 0.6865, + "step": 29434 + }, + { + "epoch": 1.08, + "learning_rate": 9.239022961145495e-08, + "loss": 0.6912, + "step": 29435 + }, + { + "epoch": 1.08, + "learning_rate": 9.238656754678287e-08, + "loss": 0.69, + "step": 29436 + }, + { + "epoch": 1.08, + "learning_rate": 9.238290548211081e-08, + "loss": 0.6801, + "step": 29437 + }, + { + "epoch": 1.08, + "learning_rate": 9.237924341743875e-08, + "loss": 0.6944, + "step": 29438 + }, + { + "epoch": 1.08, + "learning_rate": 9.237558135276668e-08, + "loss": 0.6852, + "step": 29439 + }, + { + "epoch": 1.08, + "learning_rate": 9.237191928809462e-08, + "loss": 0.69, + "step": 29440 + }, + { + "epoch": 1.08, + "learning_rate": 9.236825722342256e-08, + "loss": 0.6869, + "step": 29441 + }, + { + "epoch": 1.08, + "learning_rate": 9.23645951587505e-08, + "loss": 0.7001, + "step": 29442 + }, + { + "epoch": 1.08, + "learning_rate": 9.236093309407844e-08, + "loss": 0.7021, + "step": 29443 + }, + { + "epoch": 1.08, + "learning_rate": 9.235727102940638e-08, + "loss": 0.6991, + "step": 29444 + }, + { + "epoch": 1.08, + "learning_rate": 9.23536089647343e-08, + "loss": 0.6993, + "step": 29445 + }, + { + "epoch": 1.08, + "learning_rate": 9.234994690006224e-08, + "loss": 0.6864, + "step": 29446 + }, + { + "epoch": 1.08, + "learning_rate": 9.234628483539018e-08, + "loss": 0.7062, + "step": 29447 + }, + { + "epoch": 1.08, + "learning_rate": 9.234262277071812e-08, + "loss": 0.6952, + "step": 29448 + }, + { + "epoch": 1.08, + "learning_rate": 9.233896070604606e-08, + "loss": 0.6835, + "step": 29449 + }, + { + "epoch": 1.08, + "learning_rate": 9.2335298641374e-08, + "loss": 0.697, + "step": 29450 + }, + { + "epoch": 1.08, + "learning_rate": 9.233163657670194e-08, + "loss": 0.6909, + "step": 29451 + }, + { + "epoch": 1.08, + "learning_rate": 9.232797451202988e-08, + "loss": 0.7023, + "step": 29452 + }, + { + "epoch": 1.08, + "learning_rate": 9.232431244735782e-08, + "loss": 0.6892, + "step": 29453 + }, + { + "epoch": 1.08, + "learning_rate": 9.232065038268575e-08, + "loss": 0.6918, + "step": 29454 + }, + { + "epoch": 1.08, + "learning_rate": 9.231698831801369e-08, + "loss": 0.6936, + "step": 29455 + }, + { + "epoch": 1.08, + "learning_rate": 9.231332625334163e-08, + "loss": 0.6977, + "step": 29456 + }, + { + "epoch": 1.08, + "learning_rate": 9.230966418866957e-08, + "loss": 0.688, + "step": 29457 + }, + { + "epoch": 1.08, + "learning_rate": 9.230600212399751e-08, + "loss": 0.6934, + "step": 29458 + }, + { + "epoch": 1.08, + "learning_rate": 9.230234005932545e-08, + "loss": 0.6952, + "step": 29459 + }, + { + "epoch": 1.08, + "learning_rate": 9.229867799465339e-08, + "loss": 0.7096, + "step": 29460 + }, + { + "epoch": 1.08, + "learning_rate": 9.229501592998132e-08, + "loss": 0.6888, + "step": 29461 + }, + { + "epoch": 1.08, + "learning_rate": 9.229135386530926e-08, + "loss": 0.6849, + "step": 29462 + }, + { + "epoch": 1.08, + "learning_rate": 9.22876918006372e-08, + "loss": 0.6979, + "step": 29463 + }, + { + "epoch": 1.08, + "learning_rate": 9.228402973596512e-08, + "loss": 0.6873, + "step": 29464 + }, + { + "epoch": 1.08, + "learning_rate": 9.228036767129306e-08, + "loss": 0.6939, + "step": 29465 + }, + { + "epoch": 1.08, + "learning_rate": 9.2276705606621e-08, + "loss": 0.696, + "step": 29466 + }, + { + "epoch": 1.08, + "learning_rate": 9.227304354194894e-08, + "loss": 0.6989, + "step": 29467 + }, + { + "epoch": 1.08, + "learning_rate": 9.226938147727688e-08, + "loss": 0.6924, + "step": 29468 + }, + { + "epoch": 1.08, + "learning_rate": 9.226571941260482e-08, + "loss": 0.6992, + "step": 29469 + }, + { + "epoch": 1.08, + "learning_rate": 9.226205734793276e-08, + "loss": 0.7025, + "step": 29470 + }, + { + "epoch": 1.08, + "learning_rate": 9.22583952832607e-08, + "loss": 0.6936, + "step": 29471 + }, + { + "epoch": 1.08, + "learning_rate": 9.225473321858863e-08, + "loss": 0.6975, + "step": 29472 + }, + { + "epoch": 1.08, + "learning_rate": 9.225107115391657e-08, + "loss": 0.6968, + "step": 29473 + }, + { + "epoch": 1.08, + "learning_rate": 9.224740908924451e-08, + "loss": 0.6839, + "step": 29474 + }, + { + "epoch": 1.08, + "learning_rate": 9.224374702457245e-08, + "loss": 0.6998, + "step": 29475 + }, + { + "epoch": 1.08, + "learning_rate": 9.224008495990039e-08, + "loss": 0.6984, + "step": 29476 + }, + { + "epoch": 1.08, + "learning_rate": 9.223642289522833e-08, + "loss": 0.6923, + "step": 29477 + }, + { + "epoch": 1.08, + "learning_rate": 9.223276083055627e-08, + "loss": 0.7052, + "step": 29478 + }, + { + "epoch": 1.08, + "learning_rate": 9.222909876588421e-08, + "loss": 0.6929, + "step": 29479 + }, + { + "epoch": 1.08, + "learning_rate": 9.222543670121214e-08, + "loss": 0.7057, + "step": 29480 + }, + { + "epoch": 1.08, + "learning_rate": 9.222177463654008e-08, + "loss": 0.6795, + "step": 29481 + }, + { + "epoch": 1.08, + "learning_rate": 9.221811257186802e-08, + "loss": 0.6936, + "step": 29482 + }, + { + "epoch": 1.08, + "learning_rate": 9.221445050719596e-08, + "loss": 0.694, + "step": 29483 + }, + { + "epoch": 1.08, + "learning_rate": 9.22107884425239e-08, + "loss": 0.7036, + "step": 29484 + }, + { + "epoch": 1.08, + "learning_rate": 9.220712637785184e-08, + "loss": 0.6896, + "step": 29485 + }, + { + "epoch": 1.08, + "learning_rate": 9.220346431317976e-08, + "loss": 0.6991, + "step": 29486 + }, + { + "epoch": 1.08, + "learning_rate": 9.21998022485077e-08, + "loss": 0.6907, + "step": 29487 + }, + { + "epoch": 1.08, + "learning_rate": 9.219614018383564e-08, + "loss": 0.6895, + "step": 29488 + }, + { + "epoch": 1.08, + "learning_rate": 9.219247811916357e-08, + "loss": 0.6925, + "step": 29489 + }, + { + "epoch": 1.08, + "learning_rate": 9.218881605449151e-08, + "loss": 0.6788, + "step": 29490 + }, + { + "epoch": 1.08, + "learning_rate": 9.218515398981945e-08, + "loss": 0.6977, + "step": 29491 + }, + { + "epoch": 1.08, + "learning_rate": 9.218149192514739e-08, + "loss": 0.7011, + "step": 29492 + }, + { + "epoch": 1.08, + "learning_rate": 9.217782986047533e-08, + "loss": 0.684, + "step": 29493 + }, + { + "epoch": 1.08, + "learning_rate": 9.217416779580327e-08, + "loss": 0.6937, + "step": 29494 + }, + { + "epoch": 1.08, + "learning_rate": 9.217050573113121e-08, + "loss": 0.6802, + "step": 29495 + }, + { + "epoch": 1.08, + "learning_rate": 9.216684366645915e-08, + "loss": 0.6974, + "step": 29496 + }, + { + "epoch": 1.08, + "learning_rate": 9.216318160178709e-08, + "loss": 0.69, + "step": 29497 + }, + { + "epoch": 1.08, + "learning_rate": 9.215951953711502e-08, + "loss": 0.6904, + "step": 29498 + }, + { + "epoch": 1.08, + "learning_rate": 9.215585747244296e-08, + "loss": 0.6883, + "step": 29499 + }, + { + "epoch": 1.08, + "learning_rate": 9.21521954077709e-08, + "loss": 0.6803, + "step": 29500 + }, + { + "epoch": 1.08, + "learning_rate": 9.214853334309883e-08, + "loss": 0.6895, + "step": 29501 + }, + { + "epoch": 1.08, + "learning_rate": 9.214487127842677e-08, + "loss": 0.6841, + "step": 29502 + }, + { + "epoch": 1.08, + "learning_rate": 9.214120921375471e-08, + "loss": 0.6955, + "step": 29503 + }, + { + "epoch": 1.08, + "learning_rate": 9.213754714908265e-08, + "loss": 0.6922, + "step": 29504 + }, + { + "epoch": 1.08, + "learning_rate": 9.21338850844106e-08, + "loss": 0.6884, + "step": 29505 + }, + { + "epoch": 1.08, + "learning_rate": 9.213022301973853e-08, + "loss": 0.6953, + "step": 29506 + }, + { + "epoch": 1.08, + "learning_rate": 9.212656095506646e-08, + "loss": 0.7004, + "step": 29507 + }, + { + "epoch": 1.08, + "learning_rate": 9.21228988903944e-08, + "loss": 0.6875, + "step": 29508 + }, + { + "epoch": 1.08, + "learning_rate": 9.211923682572234e-08, + "loss": 0.7036, + "step": 29509 + }, + { + "epoch": 1.08, + "learning_rate": 9.211557476105027e-08, + "loss": 0.6824, + "step": 29510 + }, + { + "epoch": 1.08, + "learning_rate": 9.211191269637821e-08, + "loss": 0.6839, + "step": 29511 + }, + { + "epoch": 1.08, + "learning_rate": 9.210825063170615e-08, + "loss": 0.689, + "step": 29512 + }, + { + "epoch": 1.08, + "learning_rate": 9.210458856703409e-08, + "loss": 0.6939, + "step": 29513 + }, + { + "epoch": 1.08, + "learning_rate": 9.210092650236203e-08, + "loss": 0.6941, + "step": 29514 + }, + { + "epoch": 1.08, + "learning_rate": 9.209726443768997e-08, + "loss": 0.7069, + "step": 29515 + }, + { + "epoch": 1.08, + "learning_rate": 9.20936023730179e-08, + "loss": 0.689, + "step": 29516 + }, + { + "epoch": 1.08, + "learning_rate": 9.208994030834583e-08, + "loss": 0.6903, + "step": 29517 + }, + { + "epoch": 1.08, + "learning_rate": 9.208627824367377e-08, + "loss": 0.6971, + "step": 29518 + }, + { + "epoch": 1.08, + "learning_rate": 9.208261617900171e-08, + "loss": 0.6835, + "step": 29519 + }, + { + "epoch": 1.08, + "learning_rate": 9.207895411432965e-08, + "loss": 0.6924, + "step": 29520 + }, + { + "epoch": 1.08, + "learning_rate": 9.20752920496576e-08, + "loss": 0.704, + "step": 29521 + }, + { + "epoch": 1.08, + "learning_rate": 9.207162998498553e-08, + "loss": 0.6896, + "step": 29522 + }, + { + "epoch": 1.08, + "learning_rate": 9.206796792031347e-08, + "loss": 0.7029, + "step": 29523 + }, + { + "epoch": 1.08, + "learning_rate": 9.206430585564141e-08, + "loss": 0.6984, + "step": 29524 + }, + { + "epoch": 1.08, + "learning_rate": 9.206064379096934e-08, + "loss": 0.6916, + "step": 29525 + }, + { + "epoch": 1.08, + "learning_rate": 9.205698172629728e-08, + "loss": 0.7081, + "step": 29526 + }, + { + "epoch": 1.08, + "learning_rate": 9.205331966162522e-08, + "loss": 0.6998, + "step": 29527 + }, + { + "epoch": 1.08, + "learning_rate": 9.204965759695316e-08, + "loss": 0.689, + "step": 29528 + }, + { + "epoch": 1.08, + "learning_rate": 9.20459955322811e-08, + "loss": 0.6963, + "step": 29529 + }, + { + "epoch": 1.08, + "learning_rate": 9.204233346760904e-08, + "loss": 0.6902, + "step": 29530 + }, + { + "epoch": 1.08, + "learning_rate": 9.203867140293698e-08, + "loss": 0.691, + "step": 29531 + }, + { + "epoch": 1.08, + "learning_rate": 9.20350093382649e-08, + "loss": 0.6989, + "step": 29532 + }, + { + "epoch": 1.08, + "learning_rate": 9.203134727359285e-08, + "loss": 0.6947, + "step": 29533 + }, + { + "epoch": 1.08, + "learning_rate": 9.202768520892079e-08, + "loss": 0.6852, + "step": 29534 + }, + { + "epoch": 1.08, + "learning_rate": 9.202402314424871e-08, + "loss": 0.6989, + "step": 29535 + }, + { + "epoch": 1.08, + "learning_rate": 9.202036107957665e-08, + "loss": 0.6947, + "step": 29536 + }, + { + "epoch": 1.08, + "learning_rate": 9.201669901490459e-08, + "loss": 0.6939, + "step": 29537 + }, + { + "epoch": 1.08, + "learning_rate": 9.201303695023253e-08, + "loss": 0.6983, + "step": 29538 + }, + { + "epoch": 1.08, + "learning_rate": 9.200937488556047e-08, + "loss": 0.683, + "step": 29539 + }, + { + "epoch": 1.08, + "learning_rate": 9.200571282088841e-08, + "loss": 0.6876, + "step": 29540 + }, + { + "epoch": 1.08, + "learning_rate": 9.200205075621635e-08, + "loss": 0.7007, + "step": 29541 + }, + { + "epoch": 1.08, + "learning_rate": 9.199838869154429e-08, + "loss": 0.6885, + "step": 29542 + }, + { + "epoch": 1.08, + "learning_rate": 9.199472662687222e-08, + "loss": 0.6931, + "step": 29543 + }, + { + "epoch": 1.08, + "learning_rate": 9.199106456220016e-08, + "loss": 0.6915, + "step": 29544 + }, + { + "epoch": 1.08, + "learning_rate": 9.19874024975281e-08, + "loss": 0.7025, + "step": 29545 + }, + { + "epoch": 1.08, + "learning_rate": 9.198374043285604e-08, + "loss": 0.6909, + "step": 29546 + }, + { + "epoch": 1.08, + "learning_rate": 9.198007836818398e-08, + "loss": 0.7007, + "step": 29547 + }, + { + "epoch": 1.08, + "learning_rate": 9.197641630351192e-08, + "loss": 0.6912, + "step": 29548 + }, + { + "epoch": 1.08, + "learning_rate": 9.197275423883986e-08, + "loss": 0.6863, + "step": 29549 + }, + { + "epoch": 1.08, + "learning_rate": 9.19690921741678e-08, + "loss": 0.6843, + "step": 29550 + }, + { + "epoch": 1.08, + "learning_rate": 9.196543010949574e-08, + "loss": 0.6848, + "step": 29551 + }, + { + "epoch": 1.08, + "learning_rate": 9.196176804482367e-08, + "loss": 0.6989, + "step": 29552 + }, + { + "epoch": 1.08, + "learning_rate": 9.19581059801516e-08, + "loss": 0.6944, + "step": 29553 + }, + { + "epoch": 1.08, + "learning_rate": 9.195444391547955e-08, + "loss": 0.6956, + "step": 29554 + }, + { + "epoch": 1.08, + "learning_rate": 9.195078185080749e-08, + "loss": 0.6993, + "step": 29555 + }, + { + "epoch": 1.08, + "learning_rate": 9.194711978613543e-08, + "loss": 0.6997, + "step": 29556 + }, + { + "epoch": 1.08, + "learning_rate": 9.194345772146335e-08, + "loss": 0.6968, + "step": 29557 + }, + { + "epoch": 1.08, + "learning_rate": 9.193979565679129e-08, + "loss": 0.7025, + "step": 29558 + }, + { + "epoch": 1.08, + "learning_rate": 9.193613359211923e-08, + "loss": 0.679, + "step": 29559 + }, + { + "epoch": 1.08, + "learning_rate": 9.193247152744717e-08, + "loss": 0.6969, + "step": 29560 + }, + { + "epoch": 1.08, + "learning_rate": 9.19288094627751e-08, + "loss": 0.7017, + "step": 29561 + }, + { + "epoch": 1.08, + "learning_rate": 9.192514739810304e-08, + "loss": 0.6909, + "step": 29562 + }, + { + "epoch": 1.08, + "learning_rate": 9.192148533343098e-08, + "loss": 0.6867, + "step": 29563 + }, + { + "epoch": 1.08, + "learning_rate": 9.191782326875892e-08, + "loss": 0.7089, + "step": 29564 + }, + { + "epoch": 1.08, + "learning_rate": 9.191416120408686e-08, + "loss": 0.6957, + "step": 29565 + }, + { + "epoch": 1.08, + "learning_rate": 9.19104991394148e-08, + "loss": 0.6871, + "step": 29566 + }, + { + "epoch": 1.08, + "learning_rate": 9.190683707474274e-08, + "loss": 0.6924, + "step": 29567 + }, + { + "epoch": 1.08, + "learning_rate": 9.190317501007068e-08, + "loss": 0.6778, + "step": 29568 + }, + { + "epoch": 1.08, + "learning_rate": 9.189951294539862e-08, + "loss": 0.6832, + "step": 29569 + }, + { + "epoch": 1.08, + "learning_rate": 9.189585088072654e-08, + "loss": 0.689, + "step": 29570 + }, + { + "epoch": 1.08, + "learning_rate": 9.189218881605448e-08, + "loss": 0.6837, + "step": 29571 + }, + { + "epoch": 1.08, + "learning_rate": 9.188852675138242e-08, + "loss": 0.6864, + "step": 29572 + }, + { + "epoch": 1.08, + "learning_rate": 9.188486468671036e-08, + "loss": 0.6902, + "step": 29573 + }, + { + "epoch": 1.08, + "learning_rate": 9.18812026220383e-08, + "loss": 0.6964, + "step": 29574 + }, + { + "epoch": 1.08, + "learning_rate": 9.187754055736624e-08, + "loss": 0.7002, + "step": 29575 + }, + { + "epoch": 1.08, + "learning_rate": 9.187387849269418e-08, + "loss": 0.6959, + "step": 29576 + }, + { + "epoch": 1.08, + "learning_rate": 9.187021642802212e-08, + "loss": 0.6914, + "step": 29577 + }, + { + "epoch": 1.08, + "learning_rate": 9.186655436335006e-08, + "loss": 0.6999, + "step": 29578 + }, + { + "epoch": 1.08, + "learning_rate": 9.186289229867799e-08, + "loss": 0.6905, + "step": 29579 + }, + { + "epoch": 1.08, + "learning_rate": 9.185923023400593e-08, + "loss": 0.7006, + "step": 29580 + }, + { + "epoch": 1.08, + "learning_rate": 9.185556816933386e-08, + "loss": 0.6987, + "step": 29581 + }, + { + "epoch": 1.08, + "learning_rate": 9.18519061046618e-08, + "loss": 0.6982, + "step": 29582 + }, + { + "epoch": 1.08, + "learning_rate": 9.184824403998974e-08, + "loss": 0.694, + "step": 29583 + }, + { + "epoch": 1.08, + "learning_rate": 9.184458197531768e-08, + "loss": 0.7064, + "step": 29584 + }, + { + "epoch": 1.08, + "learning_rate": 9.184091991064562e-08, + "loss": 0.7133, + "step": 29585 + }, + { + "epoch": 1.08, + "learning_rate": 9.183725784597356e-08, + "loss": 0.6935, + "step": 29586 + }, + { + "epoch": 1.08, + "learning_rate": 9.18335957813015e-08, + "loss": 0.6882, + "step": 29587 + }, + { + "epoch": 1.08, + "learning_rate": 9.182993371662942e-08, + "loss": 0.6883, + "step": 29588 + }, + { + "epoch": 1.08, + "learning_rate": 9.182627165195736e-08, + "loss": 0.6935, + "step": 29589 + }, + { + "epoch": 1.08, + "learning_rate": 9.18226095872853e-08, + "loss": 0.6918, + "step": 29590 + }, + { + "epoch": 1.08, + "learning_rate": 9.181894752261324e-08, + "loss": 0.7, + "step": 29591 + }, + { + "epoch": 1.08, + "learning_rate": 9.181528545794118e-08, + "loss": 0.7088, + "step": 29592 + }, + { + "epoch": 1.08, + "learning_rate": 9.181162339326912e-08, + "loss": 0.6969, + "step": 29593 + }, + { + "epoch": 1.08, + "learning_rate": 9.180796132859706e-08, + "loss": 0.6877, + "step": 29594 + }, + { + "epoch": 1.08, + "learning_rate": 9.1804299263925e-08, + "loss": 0.6893, + "step": 29595 + }, + { + "epoch": 1.08, + "learning_rate": 9.180063719925294e-08, + "loss": 0.6914, + "step": 29596 + }, + { + "epoch": 1.08, + "learning_rate": 9.179697513458087e-08, + "loss": 0.6994, + "step": 29597 + }, + { + "epoch": 1.08, + "learning_rate": 9.179331306990881e-08, + "loss": 0.6927, + "step": 29598 + }, + { + "epoch": 1.08, + "learning_rate": 9.178965100523675e-08, + "loss": 0.6778, + "step": 29599 + }, + { + "epoch": 1.08, + "learning_rate": 9.178598894056469e-08, + "loss": 0.6872, + "step": 29600 + }, + { + "epoch": 1.08, + "learning_rate": 9.178232687589263e-08, + "loss": 0.6899, + "step": 29601 + }, + { + "epoch": 1.08, + "learning_rate": 9.177866481122057e-08, + "loss": 0.6907, + "step": 29602 + }, + { + "epoch": 1.08, + "learning_rate": 9.17750027465485e-08, + "loss": 0.6958, + "step": 29603 + }, + { + "epoch": 1.08, + "learning_rate": 9.177134068187644e-08, + "loss": 0.6922, + "step": 29604 + }, + { + "epoch": 1.08, + "learning_rate": 9.176767861720438e-08, + "loss": 0.6872, + "step": 29605 + }, + { + "epoch": 1.08, + "learning_rate": 9.17640165525323e-08, + "loss": 0.7032, + "step": 29606 + }, + { + "epoch": 1.08, + "learning_rate": 9.176035448786024e-08, + "loss": 0.6772, + "step": 29607 + }, + { + "epoch": 1.08, + "learning_rate": 9.175669242318818e-08, + "loss": 0.6954, + "step": 29608 + }, + { + "epoch": 1.08, + "learning_rate": 9.175303035851612e-08, + "loss": 0.6974, + "step": 29609 + }, + { + "epoch": 1.08, + "learning_rate": 9.174936829384406e-08, + "loss": 0.6948, + "step": 29610 + }, + { + "epoch": 1.08, + "learning_rate": 9.1745706229172e-08, + "loss": 0.6915, + "step": 29611 + }, + { + "epoch": 1.08, + "learning_rate": 9.174204416449994e-08, + "loss": 0.6995, + "step": 29612 + }, + { + "epoch": 1.08, + "learning_rate": 9.173838209982788e-08, + "loss": 0.6852, + "step": 29613 + }, + { + "epoch": 1.08, + "learning_rate": 9.173472003515582e-08, + "loss": 0.6913, + "step": 29614 + }, + { + "epoch": 1.08, + "learning_rate": 9.173105797048375e-08, + "loss": 0.702, + "step": 29615 + }, + { + "epoch": 1.08, + "learning_rate": 9.172739590581169e-08, + "loss": 0.6921, + "step": 29616 + }, + { + "epoch": 1.08, + "learning_rate": 9.172373384113963e-08, + "loss": 0.6985, + "step": 29617 + }, + { + "epoch": 1.08, + "learning_rate": 9.172007177646757e-08, + "loss": 0.6967, + "step": 29618 + }, + { + "epoch": 1.08, + "learning_rate": 9.171640971179551e-08, + "loss": 0.7051, + "step": 29619 + }, + { + "epoch": 1.08, + "learning_rate": 9.171274764712345e-08, + "loss": 0.6951, + "step": 29620 + }, + { + "epoch": 1.08, + "learning_rate": 9.170908558245139e-08, + "loss": 0.6951, + "step": 29621 + }, + { + "epoch": 1.08, + "learning_rate": 9.170542351777933e-08, + "loss": 0.6887, + "step": 29622 + }, + { + "epoch": 1.08, + "learning_rate": 9.170176145310727e-08, + "loss": 0.6708, + "step": 29623 + }, + { + "epoch": 1.08, + "learning_rate": 9.16980993884352e-08, + "loss": 0.6943, + "step": 29624 + }, + { + "epoch": 1.08, + "learning_rate": 9.169443732376313e-08, + "loss": 0.7004, + "step": 29625 + }, + { + "epoch": 1.08, + "learning_rate": 9.169077525909107e-08, + "loss": 0.7002, + "step": 29626 + }, + { + "epoch": 1.08, + "learning_rate": 9.168711319441901e-08, + "loss": 0.6901, + "step": 29627 + }, + { + "epoch": 1.08, + "learning_rate": 9.168345112974694e-08, + "loss": 0.6994, + "step": 29628 + }, + { + "epoch": 1.08, + "learning_rate": 9.167978906507488e-08, + "loss": 0.703, + "step": 29629 + }, + { + "epoch": 1.08, + "learning_rate": 9.167612700040282e-08, + "loss": 0.683, + "step": 29630 + }, + { + "epoch": 1.08, + "learning_rate": 9.167246493573076e-08, + "loss": 0.6932, + "step": 29631 + }, + { + "epoch": 1.08, + "learning_rate": 9.16688028710587e-08, + "loss": 0.7063, + "step": 29632 + }, + { + "epoch": 1.08, + "learning_rate": 9.166514080638663e-08, + "loss": 0.7032, + "step": 29633 + }, + { + "epoch": 1.08, + "learning_rate": 9.166147874171457e-08, + "loss": 0.6813, + "step": 29634 + }, + { + "epoch": 1.08, + "learning_rate": 9.165781667704251e-08, + "loss": 0.695, + "step": 29635 + }, + { + "epoch": 1.08, + "learning_rate": 9.165415461237045e-08, + "loss": 0.702, + "step": 29636 + }, + { + "epoch": 1.08, + "learning_rate": 9.165049254769839e-08, + "loss": 0.6881, + "step": 29637 + }, + { + "epoch": 1.08, + "learning_rate": 9.164683048302633e-08, + "loss": 0.7105, + "step": 29638 + }, + { + "epoch": 1.08, + "learning_rate": 9.164316841835427e-08, + "loss": 0.6963, + "step": 29639 + }, + { + "epoch": 1.08, + "learning_rate": 9.163950635368221e-08, + "loss": 0.6957, + "step": 29640 + }, + { + "epoch": 1.08, + "learning_rate": 9.163584428901015e-08, + "loss": 0.7017, + "step": 29641 + }, + { + "epoch": 1.08, + "learning_rate": 9.163218222433807e-08, + "loss": 0.6885, + "step": 29642 + }, + { + "epoch": 1.08, + "learning_rate": 9.162852015966601e-08, + "loss": 0.6924, + "step": 29643 + }, + { + "epoch": 1.08, + "learning_rate": 9.162485809499395e-08, + "loss": 0.7003, + "step": 29644 + }, + { + "epoch": 1.08, + "learning_rate": 9.16211960303219e-08, + "loss": 0.6876, + "step": 29645 + }, + { + "epoch": 1.08, + "learning_rate": 9.161753396564983e-08, + "loss": 0.6744, + "step": 29646 + }, + { + "epoch": 1.08, + "learning_rate": 9.161387190097777e-08, + "loss": 0.6925, + "step": 29647 + }, + { + "epoch": 1.08, + "learning_rate": 9.161020983630571e-08, + "loss": 0.693, + "step": 29648 + }, + { + "epoch": 1.08, + "learning_rate": 9.160654777163364e-08, + "loss": 0.7072, + "step": 29649 + }, + { + "epoch": 1.08, + "learning_rate": 9.160288570696158e-08, + "loss": 0.6864, + "step": 29650 + }, + { + "epoch": 1.08, + "learning_rate": 9.159922364228952e-08, + "loss": 0.6827, + "step": 29651 + }, + { + "epoch": 1.08, + "learning_rate": 9.159556157761745e-08, + "loss": 0.6897, + "step": 29652 + }, + { + "epoch": 1.08, + "learning_rate": 9.159189951294539e-08, + "loss": 0.6838, + "step": 29653 + }, + { + "epoch": 1.08, + "learning_rate": 9.158823744827333e-08, + "loss": 0.7091, + "step": 29654 + }, + { + "epoch": 1.08, + "learning_rate": 9.158457538360127e-08, + "loss": 0.7076, + "step": 29655 + }, + { + "epoch": 1.09, + "learning_rate": 9.15809133189292e-08, + "loss": 0.6902, + "step": 29656 + }, + { + "epoch": 1.09, + "learning_rate": 9.157725125425715e-08, + "loss": 0.6964, + "step": 29657 + }, + { + "epoch": 1.09, + "learning_rate": 9.157358918958509e-08, + "loss": 0.6953, + "step": 29658 + }, + { + "epoch": 1.09, + "learning_rate": 9.156992712491303e-08, + "loss": 0.6908, + "step": 29659 + }, + { + "epoch": 1.09, + "learning_rate": 9.156626506024095e-08, + "loss": 0.6834, + "step": 29660 + }, + { + "epoch": 1.09, + "learning_rate": 9.156260299556889e-08, + "loss": 0.6943, + "step": 29661 + }, + { + "epoch": 1.09, + "learning_rate": 9.155894093089683e-08, + "loss": 0.681, + "step": 29662 + }, + { + "epoch": 1.09, + "learning_rate": 9.155527886622477e-08, + "loss": 0.6926, + "step": 29663 + }, + { + "epoch": 1.09, + "learning_rate": 9.155161680155271e-08, + "loss": 0.6992, + "step": 29664 + }, + { + "epoch": 1.09, + "learning_rate": 9.154795473688065e-08, + "loss": 0.6951, + "step": 29665 + }, + { + "epoch": 1.09, + "learning_rate": 9.154429267220859e-08, + "loss": 0.6931, + "step": 29666 + }, + { + "epoch": 1.09, + "learning_rate": 9.154063060753653e-08, + "loss": 0.6843, + "step": 29667 + }, + { + "epoch": 1.09, + "learning_rate": 9.153696854286447e-08, + "loss": 0.6952, + "step": 29668 + }, + { + "epoch": 1.09, + "learning_rate": 9.15333064781924e-08, + "loss": 0.6962, + "step": 29669 + }, + { + "epoch": 1.09, + "learning_rate": 9.152964441352034e-08, + "loss": 0.6874, + "step": 29670 + }, + { + "epoch": 1.09, + "learning_rate": 9.152598234884828e-08, + "loss": 0.7019, + "step": 29671 + }, + { + "epoch": 1.09, + "learning_rate": 9.152232028417622e-08, + "loss": 0.7001, + "step": 29672 + }, + { + "epoch": 1.09, + "learning_rate": 9.151865821950416e-08, + "loss": 0.6885, + "step": 29673 + }, + { + "epoch": 1.09, + "learning_rate": 9.151499615483209e-08, + "loss": 0.6946, + "step": 29674 + }, + { + "epoch": 1.09, + "learning_rate": 9.151133409016003e-08, + "loss": 0.6965, + "step": 29675 + }, + { + "epoch": 1.09, + "learning_rate": 9.150767202548797e-08, + "loss": 0.6851, + "step": 29676 + }, + { + "epoch": 1.09, + "learning_rate": 9.15040099608159e-08, + "loss": 0.6886, + "step": 29677 + }, + { + "epoch": 1.09, + "learning_rate": 9.150034789614383e-08, + "loss": 0.6915, + "step": 29678 + }, + { + "epoch": 1.09, + "learning_rate": 9.149668583147177e-08, + "loss": 0.7004, + "step": 29679 + }, + { + "epoch": 1.09, + "learning_rate": 9.149302376679971e-08, + "loss": 0.6898, + "step": 29680 + }, + { + "epoch": 1.09, + "learning_rate": 9.148936170212765e-08, + "loss": 0.6923, + "step": 29681 + }, + { + "epoch": 1.09, + "learning_rate": 9.148569963745559e-08, + "loss": 0.6934, + "step": 29682 + }, + { + "epoch": 1.09, + "learning_rate": 9.148203757278353e-08, + "loss": 0.7049, + "step": 29683 + }, + { + "epoch": 1.09, + "learning_rate": 9.147837550811147e-08, + "loss": 0.6999, + "step": 29684 + }, + { + "epoch": 1.09, + "learning_rate": 9.147471344343941e-08, + "loss": 0.69, + "step": 29685 + }, + { + "epoch": 1.09, + "learning_rate": 9.147105137876735e-08, + "loss": 0.6997, + "step": 29686 + }, + { + "epoch": 1.09, + "learning_rate": 9.146738931409528e-08, + "loss": 0.6793, + "step": 29687 + }, + { + "epoch": 1.09, + "learning_rate": 9.146372724942322e-08, + "loss": 0.696, + "step": 29688 + }, + { + "epoch": 1.09, + "learning_rate": 9.146006518475116e-08, + "loss": 0.6924, + "step": 29689 + }, + { + "epoch": 1.09, + "learning_rate": 9.14564031200791e-08, + "loss": 0.6861, + "step": 29690 + }, + { + "epoch": 1.09, + "learning_rate": 9.145274105540704e-08, + "loss": 0.6845, + "step": 29691 + }, + { + "epoch": 1.09, + "learning_rate": 9.144907899073498e-08, + "loss": 0.6918, + "step": 29692 + }, + { + "epoch": 1.09, + "learning_rate": 9.144541692606292e-08, + "loss": 0.686, + "step": 29693 + }, + { + "epoch": 1.09, + "learning_rate": 9.144175486139086e-08, + "loss": 0.6944, + "step": 29694 + }, + { + "epoch": 1.09, + "learning_rate": 9.14380927967188e-08, + "loss": 0.6968, + "step": 29695 + }, + { + "epoch": 1.09, + "learning_rate": 9.143443073204672e-08, + "loss": 0.7131, + "step": 29696 + }, + { + "epoch": 1.09, + "learning_rate": 9.143076866737466e-08, + "loss": 0.6841, + "step": 29697 + }, + { + "epoch": 1.09, + "learning_rate": 9.14271066027026e-08, + "loss": 0.6913, + "step": 29698 + }, + { + "epoch": 1.09, + "learning_rate": 9.142344453803053e-08, + "loss": 0.6883, + "step": 29699 + }, + { + "epoch": 1.09, + "learning_rate": 9.141978247335847e-08, + "loss": 0.687, + "step": 29700 + }, + { + "epoch": 1.09, + "learning_rate": 9.141612040868641e-08, + "loss": 0.6985, + "step": 29701 + }, + { + "epoch": 1.09, + "learning_rate": 9.141245834401435e-08, + "loss": 0.6958, + "step": 29702 + }, + { + "epoch": 1.09, + "learning_rate": 9.140879627934229e-08, + "loss": 0.706, + "step": 29703 + }, + { + "epoch": 1.09, + "learning_rate": 9.140513421467023e-08, + "loss": 0.7035, + "step": 29704 + }, + { + "epoch": 1.09, + "learning_rate": 9.140147214999816e-08, + "loss": 0.6915, + "step": 29705 + }, + { + "epoch": 1.09, + "learning_rate": 9.13978100853261e-08, + "loss": 0.7067, + "step": 29706 + }, + { + "epoch": 1.09, + "learning_rate": 9.139414802065404e-08, + "loss": 0.6911, + "step": 29707 + }, + { + "epoch": 1.09, + "learning_rate": 9.139048595598198e-08, + "loss": 0.6864, + "step": 29708 + }, + { + "epoch": 1.09, + "learning_rate": 9.138682389130992e-08, + "loss": 0.7038, + "step": 29709 + }, + { + "epoch": 1.09, + "learning_rate": 9.138316182663786e-08, + "loss": 0.697, + "step": 29710 + }, + { + "epoch": 1.09, + "learning_rate": 9.13794997619658e-08, + "loss": 0.6929, + "step": 29711 + }, + { + "epoch": 1.09, + "learning_rate": 9.137583769729374e-08, + "loss": 0.6991, + "step": 29712 + }, + { + "epoch": 1.09, + "learning_rate": 9.137217563262168e-08, + "loss": 0.6952, + "step": 29713 + }, + { + "epoch": 1.09, + "learning_rate": 9.13685135679496e-08, + "loss": 0.6959, + "step": 29714 + }, + { + "epoch": 1.09, + "learning_rate": 9.136485150327754e-08, + "loss": 0.6946, + "step": 29715 + }, + { + "epoch": 1.09, + "learning_rate": 9.136118943860548e-08, + "loss": 0.6972, + "step": 29716 + }, + { + "epoch": 1.09, + "learning_rate": 9.135752737393342e-08, + "loss": 0.6983, + "step": 29717 + }, + { + "epoch": 1.09, + "learning_rate": 9.135386530926136e-08, + "loss": 0.6974, + "step": 29718 + }, + { + "epoch": 1.09, + "learning_rate": 9.13502032445893e-08, + "loss": 0.6861, + "step": 29719 + }, + { + "epoch": 1.09, + "learning_rate": 9.134654117991723e-08, + "loss": 0.6991, + "step": 29720 + }, + { + "epoch": 1.09, + "learning_rate": 9.134287911524517e-08, + "loss": 0.6902, + "step": 29721 + }, + { + "epoch": 1.09, + "learning_rate": 9.133921705057311e-08, + "loss": 0.6857, + "step": 29722 + }, + { + "epoch": 1.09, + "learning_rate": 9.133555498590104e-08, + "loss": 0.7091, + "step": 29723 + }, + { + "epoch": 1.09, + "learning_rate": 9.133189292122898e-08, + "loss": 0.709, + "step": 29724 + }, + { + "epoch": 1.09, + "learning_rate": 9.132823085655692e-08, + "loss": 0.6836, + "step": 29725 + }, + { + "epoch": 1.09, + "learning_rate": 9.132456879188486e-08, + "loss": 0.6969, + "step": 29726 + }, + { + "epoch": 1.09, + "learning_rate": 9.13209067272128e-08, + "loss": 0.6997, + "step": 29727 + }, + { + "epoch": 1.09, + "learning_rate": 9.131724466254074e-08, + "loss": 0.6891, + "step": 29728 + }, + { + "epoch": 1.09, + "learning_rate": 9.131358259786868e-08, + "loss": 0.688, + "step": 29729 + }, + { + "epoch": 1.09, + "learning_rate": 9.130992053319662e-08, + "loss": 0.6975, + "step": 29730 + }, + { + "epoch": 1.09, + "learning_rate": 9.130625846852456e-08, + "loss": 0.6939, + "step": 29731 + }, + { + "epoch": 1.09, + "learning_rate": 9.130259640385248e-08, + "loss": 0.7011, + "step": 29732 + }, + { + "epoch": 1.09, + "learning_rate": 9.129893433918042e-08, + "loss": 0.7068, + "step": 29733 + }, + { + "epoch": 1.09, + "learning_rate": 9.129527227450836e-08, + "loss": 0.7, + "step": 29734 + }, + { + "epoch": 1.09, + "learning_rate": 9.12916102098363e-08, + "loss": 0.6885, + "step": 29735 + }, + { + "epoch": 1.09, + "learning_rate": 9.128794814516424e-08, + "loss": 0.701, + "step": 29736 + }, + { + "epoch": 1.09, + "learning_rate": 9.128428608049218e-08, + "loss": 0.6922, + "step": 29737 + }, + { + "epoch": 1.09, + "learning_rate": 9.128062401582012e-08, + "loss": 0.6904, + "step": 29738 + }, + { + "epoch": 1.09, + "learning_rate": 9.127696195114806e-08, + "loss": 0.7013, + "step": 29739 + }, + { + "epoch": 1.09, + "learning_rate": 9.1273299886476e-08, + "loss": 0.6781, + "step": 29740 + }, + { + "epoch": 1.09, + "learning_rate": 9.126963782180393e-08, + "loss": 0.6913, + "step": 29741 + }, + { + "epoch": 1.09, + "learning_rate": 9.126597575713187e-08, + "loss": 0.6887, + "step": 29742 + }, + { + "epoch": 1.09, + "learning_rate": 9.126231369245981e-08, + "loss": 0.6888, + "step": 29743 + }, + { + "epoch": 1.09, + "learning_rate": 9.125865162778775e-08, + "loss": 0.6869, + "step": 29744 + }, + { + "epoch": 1.09, + "learning_rate": 9.125498956311567e-08, + "loss": 0.6929, + "step": 29745 + }, + { + "epoch": 1.09, + "learning_rate": 9.125132749844361e-08, + "loss": 0.6881, + "step": 29746 + }, + { + "epoch": 1.09, + "learning_rate": 9.124766543377155e-08, + "loss": 0.6834, + "step": 29747 + }, + { + "epoch": 1.09, + "learning_rate": 9.12440033690995e-08, + "loss": 0.6991, + "step": 29748 + }, + { + "epoch": 1.09, + "learning_rate": 9.124034130442743e-08, + "loss": 0.68, + "step": 29749 + }, + { + "epoch": 1.09, + "learning_rate": 9.123667923975536e-08, + "loss": 0.695, + "step": 29750 + }, + { + "epoch": 1.09, + "learning_rate": 9.12330171750833e-08, + "loss": 0.6963, + "step": 29751 + }, + { + "epoch": 1.09, + "learning_rate": 9.122935511041124e-08, + "loss": 0.6958, + "step": 29752 + }, + { + "epoch": 1.09, + "learning_rate": 9.122569304573918e-08, + "loss": 0.6917, + "step": 29753 + }, + { + "epoch": 1.09, + "learning_rate": 9.122203098106712e-08, + "loss": 0.6964, + "step": 29754 + }, + { + "epoch": 1.09, + "learning_rate": 9.121836891639506e-08, + "loss": 0.691, + "step": 29755 + }, + { + "epoch": 1.09, + "learning_rate": 9.1214706851723e-08, + "loss": 0.708, + "step": 29756 + }, + { + "epoch": 1.09, + "learning_rate": 9.121104478705094e-08, + "loss": 0.6934, + "step": 29757 + }, + { + "epoch": 1.09, + "learning_rate": 9.120738272237888e-08, + "loss": 0.6945, + "step": 29758 + }, + { + "epoch": 1.09, + "learning_rate": 9.120372065770681e-08, + "loss": 0.6936, + "step": 29759 + }, + { + "epoch": 1.09, + "learning_rate": 9.120005859303475e-08, + "loss": 0.6846, + "step": 29760 + }, + { + "epoch": 1.09, + "learning_rate": 9.119639652836269e-08, + "loss": 0.6911, + "step": 29761 + }, + { + "epoch": 1.09, + "learning_rate": 9.119273446369063e-08, + "loss": 0.6881, + "step": 29762 + }, + { + "epoch": 1.09, + "learning_rate": 9.118907239901857e-08, + "loss": 0.6948, + "step": 29763 + }, + { + "epoch": 1.09, + "learning_rate": 9.118541033434651e-08, + "loss": 0.6818, + "step": 29764 + }, + { + "epoch": 1.09, + "learning_rate": 9.118174826967445e-08, + "loss": 0.6993, + "step": 29765 + }, + { + "epoch": 1.09, + "learning_rate": 9.117808620500239e-08, + "loss": 0.6993, + "step": 29766 + }, + { + "epoch": 1.09, + "learning_rate": 9.117442414033031e-08, + "loss": 0.6894, + "step": 29767 + }, + { + "epoch": 1.09, + "learning_rate": 9.117076207565825e-08, + "loss": 0.6934, + "step": 29768 + }, + { + "epoch": 1.09, + "learning_rate": 9.116710001098618e-08, + "loss": 0.7088, + "step": 29769 + }, + { + "epoch": 1.09, + "learning_rate": 9.116343794631412e-08, + "loss": 0.6888, + "step": 29770 + }, + { + "epoch": 1.09, + "learning_rate": 9.115977588164206e-08, + "loss": 0.6926, + "step": 29771 + }, + { + "epoch": 1.09, + "learning_rate": 9.115611381697e-08, + "loss": 0.693, + "step": 29772 + }, + { + "epoch": 1.09, + "learning_rate": 9.115245175229794e-08, + "loss": 0.6961, + "step": 29773 + }, + { + "epoch": 1.09, + "learning_rate": 9.114878968762588e-08, + "loss": 0.6983, + "step": 29774 + }, + { + "epoch": 1.09, + "learning_rate": 9.114512762295382e-08, + "loss": 0.6985, + "step": 29775 + }, + { + "epoch": 1.09, + "learning_rate": 9.114146555828176e-08, + "loss": 0.6979, + "step": 29776 + }, + { + "epoch": 1.09, + "learning_rate": 9.113780349360969e-08, + "loss": 0.6857, + "step": 29777 + }, + { + "epoch": 1.09, + "learning_rate": 9.113414142893763e-08, + "loss": 0.7037, + "step": 29778 + }, + { + "epoch": 1.09, + "learning_rate": 9.113047936426557e-08, + "loss": 0.6943, + "step": 29779 + }, + { + "epoch": 1.09, + "learning_rate": 9.11268172995935e-08, + "loss": 0.7028, + "step": 29780 + }, + { + "epoch": 1.09, + "learning_rate": 9.112315523492145e-08, + "loss": 0.7064, + "step": 29781 + }, + { + "epoch": 1.09, + "learning_rate": 9.111949317024939e-08, + "loss": 0.6928, + "step": 29782 + }, + { + "epoch": 1.09, + "learning_rate": 9.111583110557733e-08, + "loss": 0.6808, + "step": 29783 + }, + { + "epoch": 1.09, + "learning_rate": 9.111216904090527e-08, + "loss": 0.6905, + "step": 29784 + }, + { + "epoch": 1.09, + "learning_rate": 9.11085069762332e-08, + "loss": 0.6907, + "step": 29785 + }, + { + "epoch": 1.09, + "learning_rate": 9.110484491156113e-08, + "loss": 0.6956, + "step": 29786 + }, + { + "epoch": 1.09, + "learning_rate": 9.110118284688907e-08, + "loss": 0.696, + "step": 29787 + }, + { + "epoch": 1.09, + "learning_rate": 9.109752078221701e-08, + "loss": 0.6981, + "step": 29788 + }, + { + "epoch": 1.09, + "learning_rate": 9.109385871754495e-08, + "loss": 0.6931, + "step": 29789 + }, + { + "epoch": 1.09, + "learning_rate": 9.109019665287289e-08, + "loss": 0.69, + "step": 29790 + }, + { + "epoch": 1.09, + "learning_rate": 9.108653458820082e-08, + "loss": 0.6762, + "step": 29791 + }, + { + "epoch": 1.09, + "learning_rate": 9.108287252352876e-08, + "loss": 0.6999, + "step": 29792 + }, + { + "epoch": 1.09, + "learning_rate": 9.10792104588567e-08, + "loss": 0.6915, + "step": 29793 + }, + { + "epoch": 1.09, + "learning_rate": 9.107554839418464e-08, + "loss": 0.7023, + "step": 29794 + }, + { + "epoch": 1.09, + "learning_rate": 9.107188632951257e-08, + "loss": 0.6997, + "step": 29795 + }, + { + "epoch": 1.09, + "learning_rate": 9.10682242648405e-08, + "loss": 0.6955, + "step": 29796 + }, + { + "epoch": 1.09, + "learning_rate": 9.106456220016844e-08, + "loss": 0.6906, + "step": 29797 + }, + { + "epoch": 1.09, + "learning_rate": 9.106090013549638e-08, + "loss": 0.6927, + "step": 29798 + }, + { + "epoch": 1.09, + "learning_rate": 9.105723807082432e-08, + "loss": 0.7084, + "step": 29799 + }, + { + "epoch": 1.09, + "learning_rate": 9.105357600615226e-08, + "loss": 0.6796, + "step": 29800 + }, + { + "epoch": 1.09, + "learning_rate": 9.10499139414802e-08, + "loss": 0.686, + "step": 29801 + }, + { + "epoch": 1.09, + "learning_rate": 9.104625187680814e-08, + "loss": 0.7041, + "step": 29802 + }, + { + "epoch": 1.09, + "learning_rate": 9.104258981213608e-08, + "loss": 0.6984, + "step": 29803 + }, + { + "epoch": 1.09, + "learning_rate": 9.103892774746401e-08, + "loss": 0.6917, + "step": 29804 + }, + { + "epoch": 1.09, + "learning_rate": 9.103526568279195e-08, + "loss": 0.691, + "step": 29805 + }, + { + "epoch": 1.09, + "learning_rate": 9.103160361811989e-08, + "loss": 0.695, + "step": 29806 + }, + { + "epoch": 1.09, + "learning_rate": 9.102794155344783e-08, + "loss": 0.7051, + "step": 29807 + }, + { + "epoch": 1.09, + "learning_rate": 9.102427948877577e-08, + "loss": 0.6911, + "step": 29808 + }, + { + "epoch": 1.09, + "learning_rate": 9.102061742410371e-08, + "loss": 0.6925, + "step": 29809 + }, + { + "epoch": 1.09, + "learning_rate": 9.101695535943165e-08, + "loss": 0.6852, + "step": 29810 + }, + { + "epoch": 1.09, + "learning_rate": 9.101329329475959e-08, + "loss": 0.6893, + "step": 29811 + }, + { + "epoch": 1.09, + "learning_rate": 9.100963123008753e-08, + "loss": 0.6895, + "step": 29812 + }, + { + "epoch": 1.09, + "learning_rate": 9.100596916541546e-08, + "loss": 0.6822, + "step": 29813 + }, + { + "epoch": 1.09, + "learning_rate": 9.10023071007434e-08, + "loss": 0.689, + "step": 29814 + }, + { + "epoch": 1.09, + "learning_rate": 9.099864503607134e-08, + "loss": 0.6937, + "step": 29815 + }, + { + "epoch": 1.09, + "learning_rate": 9.099498297139926e-08, + "loss": 0.6907, + "step": 29816 + }, + { + "epoch": 1.09, + "learning_rate": 9.09913209067272e-08, + "loss": 0.6909, + "step": 29817 + }, + { + "epoch": 1.09, + "learning_rate": 9.098765884205514e-08, + "loss": 0.6973, + "step": 29818 + }, + { + "epoch": 1.09, + "learning_rate": 9.098399677738308e-08, + "loss": 0.68, + "step": 29819 + }, + { + "epoch": 1.09, + "learning_rate": 9.098033471271102e-08, + "loss": 0.6879, + "step": 29820 + }, + { + "epoch": 1.09, + "learning_rate": 9.097667264803896e-08, + "loss": 0.6961, + "step": 29821 + }, + { + "epoch": 1.09, + "learning_rate": 9.097301058336689e-08, + "loss": 0.6911, + "step": 29822 + }, + { + "epoch": 1.09, + "learning_rate": 9.096934851869483e-08, + "loss": 0.6975, + "step": 29823 + }, + { + "epoch": 1.09, + "learning_rate": 9.096568645402277e-08, + "loss": 0.6955, + "step": 29824 + }, + { + "epoch": 1.09, + "learning_rate": 9.096202438935071e-08, + "loss": 0.6908, + "step": 29825 + }, + { + "epoch": 1.09, + "learning_rate": 9.095836232467865e-08, + "loss": 0.7047, + "step": 29826 + }, + { + "epoch": 1.09, + "learning_rate": 9.095470026000659e-08, + "loss": 0.7009, + "step": 29827 + }, + { + "epoch": 1.09, + "learning_rate": 9.095103819533453e-08, + "loss": 0.6776, + "step": 29828 + }, + { + "epoch": 1.09, + "learning_rate": 9.094737613066247e-08, + "loss": 0.6977, + "step": 29829 + }, + { + "epoch": 1.09, + "learning_rate": 9.094371406599041e-08, + "loss": 0.6881, + "step": 29830 + }, + { + "epoch": 1.09, + "learning_rate": 9.094005200131834e-08, + "loss": 0.6908, + "step": 29831 + }, + { + "epoch": 1.09, + "learning_rate": 9.093638993664628e-08, + "loss": 0.6936, + "step": 29832 + }, + { + "epoch": 1.09, + "learning_rate": 9.093272787197422e-08, + "loss": 0.692, + "step": 29833 + }, + { + "epoch": 1.09, + "learning_rate": 9.092906580730216e-08, + "loss": 0.7005, + "step": 29834 + }, + { + "epoch": 1.09, + "learning_rate": 9.09254037426301e-08, + "loss": 0.6872, + "step": 29835 + }, + { + "epoch": 1.09, + "learning_rate": 9.092174167795804e-08, + "loss": 0.6874, + "step": 29836 + }, + { + "epoch": 1.09, + "learning_rate": 9.091807961328598e-08, + "loss": 0.6912, + "step": 29837 + }, + { + "epoch": 1.09, + "learning_rate": 9.09144175486139e-08, + "loss": 0.7028, + "step": 29838 + }, + { + "epoch": 1.09, + "learning_rate": 9.091075548394184e-08, + "loss": 0.7043, + "step": 29839 + }, + { + "epoch": 1.09, + "learning_rate": 9.090709341926977e-08, + "loss": 0.6922, + "step": 29840 + }, + { + "epoch": 1.09, + "learning_rate": 9.090343135459771e-08, + "loss": 0.6844, + "step": 29841 + }, + { + "epoch": 1.09, + "learning_rate": 9.089976928992565e-08, + "loss": 0.6977, + "step": 29842 + }, + { + "epoch": 1.09, + "learning_rate": 9.089610722525359e-08, + "loss": 0.6967, + "step": 29843 + }, + { + "epoch": 1.09, + "learning_rate": 9.089244516058153e-08, + "loss": 0.6909, + "step": 29844 + }, + { + "epoch": 1.09, + "learning_rate": 9.088878309590947e-08, + "loss": 0.6837, + "step": 29845 + }, + { + "epoch": 1.09, + "learning_rate": 9.088512103123741e-08, + "loss": 0.6912, + "step": 29846 + }, + { + "epoch": 1.09, + "learning_rate": 9.088145896656535e-08, + "loss": 0.7025, + "step": 29847 + }, + { + "epoch": 1.09, + "learning_rate": 9.087779690189329e-08, + "loss": 0.7018, + "step": 29848 + }, + { + "epoch": 1.09, + "learning_rate": 9.087413483722122e-08, + "loss": 0.6933, + "step": 29849 + }, + { + "epoch": 1.09, + "learning_rate": 9.087047277254916e-08, + "loss": 0.6868, + "step": 29850 + }, + { + "epoch": 1.09, + "learning_rate": 9.08668107078771e-08, + "loss": 0.6999, + "step": 29851 + }, + { + "epoch": 1.09, + "learning_rate": 9.086314864320504e-08, + "loss": 0.6966, + "step": 29852 + }, + { + "epoch": 1.09, + "learning_rate": 9.085948657853298e-08, + "loss": 0.6954, + "step": 29853 + }, + { + "epoch": 1.09, + "learning_rate": 9.085582451386092e-08, + "loss": 0.6921, + "step": 29854 + }, + { + "epoch": 1.09, + "learning_rate": 9.085216244918885e-08, + "loss": 0.6794, + "step": 29855 + }, + { + "epoch": 1.09, + "learning_rate": 9.08485003845168e-08, + "loss": 0.6904, + "step": 29856 + }, + { + "epoch": 1.09, + "learning_rate": 9.084483831984472e-08, + "loss": 0.6857, + "step": 29857 + }, + { + "epoch": 1.09, + "learning_rate": 9.084117625517266e-08, + "loss": 0.6913, + "step": 29858 + }, + { + "epoch": 1.09, + "learning_rate": 9.08375141905006e-08, + "loss": 0.6987, + "step": 29859 + }, + { + "epoch": 1.09, + "learning_rate": 9.083385212582854e-08, + "loss": 0.6809, + "step": 29860 + }, + { + "epoch": 1.09, + "learning_rate": 9.083019006115648e-08, + "loss": 0.6923, + "step": 29861 + }, + { + "epoch": 1.09, + "learning_rate": 9.082652799648441e-08, + "loss": 0.6991, + "step": 29862 + }, + { + "epoch": 1.09, + "learning_rate": 9.082286593181235e-08, + "loss": 0.6902, + "step": 29863 + }, + { + "epoch": 1.09, + "learning_rate": 9.081920386714029e-08, + "loss": 0.6985, + "step": 29864 + }, + { + "epoch": 1.09, + "learning_rate": 9.081554180246823e-08, + "loss": 0.6956, + "step": 29865 + }, + { + "epoch": 1.09, + "learning_rate": 9.081187973779615e-08, + "loss": 0.6967, + "step": 29866 + }, + { + "epoch": 1.09, + "learning_rate": 9.08082176731241e-08, + "loss": 0.6901, + "step": 29867 + }, + { + "epoch": 1.09, + "learning_rate": 9.080455560845203e-08, + "loss": 0.7003, + "step": 29868 + }, + { + "epoch": 1.09, + "learning_rate": 9.080089354377997e-08, + "loss": 0.6898, + "step": 29869 + }, + { + "epoch": 1.09, + "learning_rate": 9.079723147910791e-08, + "loss": 0.6816, + "step": 29870 + }, + { + "epoch": 1.09, + "learning_rate": 9.079356941443585e-08, + "loss": 0.6842, + "step": 29871 + }, + { + "epoch": 1.09, + "learning_rate": 9.07899073497638e-08, + "loss": 0.6902, + "step": 29872 + }, + { + "epoch": 1.09, + "learning_rate": 9.078624528509173e-08, + "loss": 0.6947, + "step": 29873 + }, + { + "epoch": 1.09, + "learning_rate": 9.078258322041967e-08, + "loss": 0.7038, + "step": 29874 + }, + { + "epoch": 1.09, + "learning_rate": 9.07789211557476e-08, + "loss": 0.6965, + "step": 29875 + }, + { + "epoch": 1.09, + "learning_rate": 9.077525909107554e-08, + "loss": 0.7061, + "step": 29876 + }, + { + "epoch": 1.09, + "learning_rate": 9.077159702640348e-08, + "loss": 0.7087, + "step": 29877 + }, + { + "epoch": 1.09, + "learning_rate": 9.076793496173142e-08, + "loss": 0.6862, + "step": 29878 + }, + { + "epoch": 1.09, + "learning_rate": 9.076427289705936e-08, + "loss": 0.6933, + "step": 29879 + }, + { + "epoch": 1.09, + "learning_rate": 9.07606108323873e-08, + "loss": 0.6917, + "step": 29880 + }, + { + "epoch": 1.09, + "learning_rate": 9.075694876771524e-08, + "loss": 0.6909, + "step": 29881 + }, + { + "epoch": 1.09, + "learning_rate": 9.075328670304318e-08, + "loss": 0.6837, + "step": 29882 + }, + { + "epoch": 1.09, + "learning_rate": 9.074962463837112e-08, + "loss": 0.6941, + "step": 29883 + }, + { + "epoch": 1.09, + "learning_rate": 9.074596257369905e-08, + "loss": 0.7029, + "step": 29884 + }, + { + "epoch": 1.09, + "learning_rate": 9.074230050902699e-08, + "loss": 0.686, + "step": 29885 + }, + { + "epoch": 1.09, + "learning_rate": 9.073863844435493e-08, + "loss": 0.6911, + "step": 29886 + }, + { + "epoch": 1.09, + "learning_rate": 9.073497637968285e-08, + "loss": 0.6848, + "step": 29887 + }, + { + "epoch": 1.09, + "learning_rate": 9.073131431501079e-08, + "loss": 0.6808, + "step": 29888 + }, + { + "epoch": 1.09, + "learning_rate": 9.072765225033873e-08, + "loss": 0.6932, + "step": 29889 + }, + { + "epoch": 1.09, + "learning_rate": 9.072399018566667e-08, + "loss": 0.6892, + "step": 29890 + }, + { + "epoch": 1.09, + "learning_rate": 9.072032812099461e-08, + "loss": 0.6924, + "step": 29891 + }, + { + "epoch": 1.09, + "learning_rate": 9.071666605632255e-08, + "loss": 0.6997, + "step": 29892 + }, + { + "epoch": 1.09, + "learning_rate": 9.071300399165048e-08, + "loss": 0.6931, + "step": 29893 + }, + { + "epoch": 1.09, + "learning_rate": 9.070934192697842e-08, + "loss": 0.6834, + "step": 29894 + }, + { + "epoch": 1.09, + "learning_rate": 9.070567986230636e-08, + "loss": 0.6943, + "step": 29895 + }, + { + "epoch": 1.09, + "learning_rate": 9.07020177976343e-08, + "loss": 0.6868, + "step": 29896 + }, + { + "epoch": 1.09, + "learning_rate": 9.069835573296224e-08, + "loss": 0.6912, + "step": 29897 + }, + { + "epoch": 1.09, + "learning_rate": 9.069469366829018e-08, + "loss": 0.6998, + "step": 29898 + }, + { + "epoch": 1.09, + "learning_rate": 9.069103160361812e-08, + "loss": 0.7091, + "step": 29899 + }, + { + "epoch": 1.09, + "learning_rate": 9.068736953894606e-08, + "loss": 0.6963, + "step": 29900 + }, + { + "epoch": 1.09, + "learning_rate": 9.0683707474274e-08, + "loss": 0.6934, + "step": 29901 + }, + { + "epoch": 1.09, + "learning_rate": 9.068004540960193e-08, + "loss": 0.6971, + "step": 29902 + }, + { + "epoch": 1.09, + "learning_rate": 9.067638334492987e-08, + "loss": 0.7081, + "step": 29903 + }, + { + "epoch": 1.09, + "learning_rate": 9.06727212802578e-08, + "loss": 0.6805, + "step": 29904 + }, + { + "epoch": 1.09, + "learning_rate": 9.066905921558575e-08, + "loss": 0.6939, + "step": 29905 + }, + { + "epoch": 1.09, + "learning_rate": 9.066539715091369e-08, + "loss": 0.6968, + "step": 29906 + }, + { + "epoch": 1.09, + "learning_rate": 9.066173508624163e-08, + "loss": 0.6972, + "step": 29907 + }, + { + "epoch": 1.09, + "learning_rate": 9.065807302156957e-08, + "loss": 0.6979, + "step": 29908 + }, + { + "epoch": 1.09, + "learning_rate": 9.065441095689749e-08, + "loss": 0.6976, + "step": 29909 + }, + { + "epoch": 1.09, + "learning_rate": 9.065074889222543e-08, + "loss": 0.6935, + "step": 29910 + }, + { + "epoch": 1.09, + "learning_rate": 9.064708682755336e-08, + "loss": 0.6882, + "step": 29911 + }, + { + "epoch": 1.09, + "learning_rate": 9.06434247628813e-08, + "loss": 0.6994, + "step": 29912 + }, + { + "epoch": 1.09, + "learning_rate": 9.063976269820924e-08, + "loss": 0.6874, + "step": 29913 + }, + { + "epoch": 1.09, + "learning_rate": 9.063610063353718e-08, + "loss": 0.6832, + "step": 29914 + }, + { + "epoch": 1.09, + "learning_rate": 9.063243856886512e-08, + "loss": 0.6835, + "step": 29915 + }, + { + "epoch": 1.09, + "learning_rate": 9.062877650419306e-08, + "loss": 0.6976, + "step": 29916 + }, + { + "epoch": 1.09, + "learning_rate": 9.0625114439521e-08, + "loss": 0.6971, + "step": 29917 + }, + { + "epoch": 1.09, + "learning_rate": 9.062145237484894e-08, + "loss": 0.6899, + "step": 29918 + }, + { + "epoch": 1.09, + "learning_rate": 9.061779031017688e-08, + "loss": 0.708, + "step": 29919 + }, + { + "epoch": 1.09, + "learning_rate": 9.06141282455048e-08, + "loss": 0.7043, + "step": 29920 + }, + { + "epoch": 1.09, + "learning_rate": 9.061046618083274e-08, + "loss": 0.6949, + "step": 29921 + }, + { + "epoch": 1.09, + "learning_rate": 9.060680411616068e-08, + "loss": 0.6957, + "step": 29922 + }, + { + "epoch": 1.09, + "learning_rate": 9.060314205148862e-08, + "loss": 0.6976, + "step": 29923 + }, + { + "epoch": 1.09, + "learning_rate": 9.059947998681656e-08, + "loss": 0.7059, + "step": 29924 + }, + { + "epoch": 1.09, + "learning_rate": 9.05958179221445e-08, + "loss": 0.6979, + "step": 29925 + }, + { + "epoch": 1.09, + "learning_rate": 9.059215585747244e-08, + "loss": 0.6869, + "step": 29926 + }, + { + "epoch": 1.09, + "learning_rate": 9.058849379280038e-08, + "loss": 0.6923, + "step": 29927 + }, + { + "epoch": 1.09, + "learning_rate": 9.058483172812832e-08, + "loss": 0.7128, + "step": 29928 + }, + { + "epoch": 1.1, + "learning_rate": 9.058116966345625e-08, + "loss": 0.6888, + "step": 29929 + }, + { + "epoch": 1.1, + "learning_rate": 9.057750759878419e-08, + "loss": 0.6998, + "step": 29930 + }, + { + "epoch": 1.1, + "learning_rate": 9.057384553411213e-08, + "loss": 0.6942, + "step": 29931 + }, + { + "epoch": 1.1, + "learning_rate": 9.057018346944007e-08, + "loss": 0.6937, + "step": 29932 + }, + { + "epoch": 1.1, + "learning_rate": 9.0566521404768e-08, + "loss": 0.686, + "step": 29933 + }, + { + "epoch": 1.1, + "learning_rate": 9.056285934009594e-08, + "loss": 0.6867, + "step": 29934 + }, + { + "epoch": 1.1, + "learning_rate": 9.055919727542388e-08, + "loss": 0.6946, + "step": 29935 + }, + { + "epoch": 1.1, + "learning_rate": 9.055553521075182e-08, + "loss": 0.6952, + "step": 29936 + }, + { + "epoch": 1.1, + "learning_rate": 9.055187314607976e-08, + "loss": 0.7087, + "step": 29937 + }, + { + "epoch": 1.1, + "learning_rate": 9.054821108140768e-08, + "loss": 0.6907, + "step": 29938 + }, + { + "epoch": 1.1, + "learning_rate": 9.054454901673562e-08, + "loss": 0.6848, + "step": 29939 + }, + { + "epoch": 1.1, + "learning_rate": 9.054088695206356e-08, + "loss": 0.6949, + "step": 29940 + }, + { + "epoch": 1.1, + "learning_rate": 9.05372248873915e-08, + "loss": 0.6787, + "step": 29941 + }, + { + "epoch": 1.1, + "learning_rate": 9.053356282271944e-08, + "loss": 0.6826, + "step": 29942 + }, + { + "epoch": 1.1, + "learning_rate": 9.052990075804738e-08, + "loss": 0.6933, + "step": 29943 + }, + { + "epoch": 1.1, + "learning_rate": 9.052623869337532e-08, + "loss": 0.7014, + "step": 29944 + }, + { + "epoch": 1.1, + "learning_rate": 9.052257662870326e-08, + "loss": 0.6923, + "step": 29945 + }, + { + "epoch": 1.1, + "learning_rate": 9.05189145640312e-08, + "loss": 0.6937, + "step": 29946 + }, + { + "epoch": 1.1, + "learning_rate": 9.051525249935913e-08, + "loss": 0.6973, + "step": 29947 + }, + { + "epoch": 1.1, + "learning_rate": 9.051159043468707e-08, + "loss": 0.7052, + "step": 29948 + }, + { + "epoch": 1.1, + "learning_rate": 9.050792837001501e-08, + "loss": 0.689, + "step": 29949 + }, + { + "epoch": 1.1, + "learning_rate": 9.050426630534295e-08, + "loss": 0.6918, + "step": 29950 + }, + { + "epoch": 1.1, + "learning_rate": 9.050060424067089e-08, + "loss": 0.6957, + "step": 29951 + }, + { + "epoch": 1.1, + "learning_rate": 9.049694217599883e-08, + "loss": 0.6997, + "step": 29952 + }, + { + "epoch": 1.1, + "learning_rate": 9.049328011132677e-08, + "loss": 0.6878, + "step": 29953 + }, + { + "epoch": 1.1, + "learning_rate": 9.048961804665471e-08, + "loss": 0.6845, + "step": 29954 + }, + { + "epoch": 1.1, + "learning_rate": 9.048595598198264e-08, + "loss": 0.6837, + "step": 29955 + }, + { + "epoch": 1.1, + "learning_rate": 9.048229391731058e-08, + "loss": 0.6951, + "step": 29956 + }, + { + "epoch": 1.1, + "learning_rate": 9.047863185263852e-08, + "loss": 0.6991, + "step": 29957 + }, + { + "epoch": 1.1, + "learning_rate": 9.047496978796644e-08, + "loss": 0.6912, + "step": 29958 + }, + { + "epoch": 1.1, + "learning_rate": 9.047130772329438e-08, + "loss": 0.6957, + "step": 29959 + }, + { + "epoch": 1.1, + "learning_rate": 9.046764565862232e-08, + "loss": 0.6831, + "step": 29960 + }, + { + "epoch": 1.1, + "learning_rate": 9.046398359395026e-08, + "loss": 0.6928, + "step": 29961 + }, + { + "epoch": 1.1, + "learning_rate": 9.04603215292782e-08, + "loss": 0.6897, + "step": 29962 + }, + { + "epoch": 1.1, + "learning_rate": 9.045665946460614e-08, + "loss": 0.704, + "step": 29963 + }, + { + "epoch": 1.1, + "learning_rate": 9.045299739993408e-08, + "loss": 0.6988, + "step": 29964 + }, + { + "epoch": 1.1, + "learning_rate": 9.044933533526201e-08, + "loss": 0.7008, + "step": 29965 + }, + { + "epoch": 1.1, + "learning_rate": 9.044567327058995e-08, + "loss": 0.694, + "step": 29966 + }, + { + "epoch": 1.1, + "learning_rate": 9.044201120591789e-08, + "loss": 0.7081, + "step": 29967 + }, + { + "epoch": 1.1, + "learning_rate": 9.043834914124583e-08, + "loss": 0.7031, + "step": 29968 + }, + { + "epoch": 1.1, + "learning_rate": 9.043468707657377e-08, + "loss": 0.6867, + "step": 29969 + }, + { + "epoch": 1.1, + "learning_rate": 9.043102501190171e-08, + "loss": 0.6814, + "step": 29970 + }, + { + "epoch": 1.1, + "learning_rate": 9.042736294722965e-08, + "loss": 0.7147, + "step": 29971 + }, + { + "epoch": 1.1, + "learning_rate": 9.042370088255759e-08, + "loss": 0.6886, + "step": 29972 + }, + { + "epoch": 1.1, + "learning_rate": 9.042003881788553e-08, + "loss": 0.7023, + "step": 29973 + }, + { + "epoch": 1.1, + "learning_rate": 9.041637675321346e-08, + "loss": 0.6753, + "step": 29974 + }, + { + "epoch": 1.1, + "learning_rate": 9.04127146885414e-08, + "loss": 0.7017, + "step": 29975 + }, + { + "epoch": 1.1, + "learning_rate": 9.040905262386933e-08, + "loss": 0.6936, + "step": 29976 + }, + { + "epoch": 1.1, + "learning_rate": 9.040539055919727e-08, + "loss": 0.685, + "step": 29977 + }, + { + "epoch": 1.1, + "learning_rate": 9.040172849452521e-08, + "loss": 0.6982, + "step": 29978 + }, + { + "epoch": 1.1, + "learning_rate": 9.039806642985315e-08, + "loss": 0.701, + "step": 29979 + }, + { + "epoch": 1.1, + "learning_rate": 9.039440436518108e-08, + "loss": 0.6819, + "step": 29980 + }, + { + "epoch": 1.1, + "learning_rate": 9.039074230050902e-08, + "loss": 0.6854, + "step": 29981 + }, + { + "epoch": 1.1, + "learning_rate": 9.038708023583696e-08, + "loss": 0.6893, + "step": 29982 + }, + { + "epoch": 1.1, + "learning_rate": 9.038341817116489e-08, + "loss": 0.6768, + "step": 29983 + }, + { + "epoch": 1.1, + "learning_rate": 9.037975610649283e-08, + "loss": 0.6907, + "step": 29984 + }, + { + "epoch": 1.1, + "learning_rate": 9.037609404182077e-08, + "loss": 0.6797, + "step": 29985 + }, + { + "epoch": 1.1, + "learning_rate": 9.037243197714871e-08, + "loss": 0.6917, + "step": 29986 + }, + { + "epoch": 1.1, + "learning_rate": 9.036876991247665e-08, + "loss": 0.7029, + "step": 29987 + }, + { + "epoch": 1.1, + "learning_rate": 9.036510784780459e-08, + "loss": 0.6923, + "step": 29988 + }, + { + "epoch": 1.1, + "learning_rate": 9.036144578313253e-08, + "loss": 0.6941, + "step": 29989 + }, + { + "epoch": 1.1, + "learning_rate": 9.035778371846047e-08, + "loss": 0.696, + "step": 29990 + }, + { + "epoch": 1.1, + "learning_rate": 9.035412165378841e-08, + "loss": 0.6932, + "step": 29991 + }, + { + "epoch": 1.1, + "learning_rate": 9.035045958911633e-08, + "loss": 0.689, + "step": 29992 + }, + { + "epoch": 1.1, + "learning_rate": 9.034679752444427e-08, + "loss": 0.6914, + "step": 29993 + }, + { + "epoch": 1.1, + "learning_rate": 9.034313545977221e-08, + "loss": 0.7058, + "step": 29994 + }, + { + "epoch": 1.1, + "learning_rate": 9.033947339510015e-08, + "loss": 0.6874, + "step": 29995 + }, + { + "epoch": 1.1, + "learning_rate": 9.03358113304281e-08, + "loss": 0.6829, + "step": 29996 + }, + { + "epoch": 1.1, + "learning_rate": 9.033214926575603e-08, + "loss": 0.6831, + "step": 29997 + }, + { + "epoch": 1.1, + "learning_rate": 9.032848720108397e-08, + "loss": 0.708, + "step": 29998 + }, + { + "epoch": 1.1, + "learning_rate": 9.032482513641191e-08, + "loss": 0.6984, + "step": 29999 + }, + { + "epoch": 1.1, + "learning_rate": 9.032116307173985e-08, + "loss": 0.7041, + "step": 30000 + }, + { + "epoch": 1.1, + "learning_rate": 9.031750100706778e-08, + "loss": 0.6805, + "step": 30001 + }, + { + "epoch": 1.1, + "learning_rate": 9.031383894239572e-08, + "loss": 0.6922, + "step": 30002 + }, + { + "epoch": 1.1, + "learning_rate": 9.031017687772366e-08, + "loss": 0.7, + "step": 30003 + }, + { + "epoch": 1.1, + "learning_rate": 9.030651481305159e-08, + "loss": 0.6879, + "step": 30004 + }, + { + "epoch": 1.1, + "learning_rate": 9.030285274837953e-08, + "loss": 0.7004, + "step": 30005 + }, + { + "epoch": 1.1, + "learning_rate": 9.029919068370747e-08, + "loss": 0.6865, + "step": 30006 + }, + { + "epoch": 1.1, + "learning_rate": 9.02955286190354e-08, + "loss": 0.7092, + "step": 30007 + }, + { + "epoch": 1.1, + "learning_rate": 9.029186655436335e-08, + "loss": 0.6943, + "step": 30008 + }, + { + "epoch": 1.1, + "learning_rate": 9.028820448969129e-08, + "loss": 0.7035, + "step": 30009 + }, + { + "epoch": 1.1, + "learning_rate": 9.028454242501921e-08, + "loss": 0.6891, + "step": 30010 + }, + { + "epoch": 1.1, + "learning_rate": 9.028088036034715e-08, + "loss": 0.6939, + "step": 30011 + }, + { + "epoch": 1.1, + "learning_rate": 9.027721829567509e-08, + "loss": 0.6855, + "step": 30012 + }, + { + "epoch": 1.1, + "learning_rate": 9.027355623100303e-08, + "loss": 0.6852, + "step": 30013 + }, + { + "epoch": 1.1, + "learning_rate": 9.026989416633097e-08, + "loss": 0.6841, + "step": 30014 + }, + { + "epoch": 1.1, + "learning_rate": 9.026623210165891e-08, + "loss": 0.6941, + "step": 30015 + }, + { + "epoch": 1.1, + "learning_rate": 9.026257003698685e-08, + "loss": 0.6907, + "step": 30016 + }, + { + "epoch": 1.1, + "learning_rate": 9.025890797231479e-08, + "loss": 0.7013, + "step": 30017 + }, + { + "epoch": 1.1, + "learning_rate": 9.025524590764273e-08, + "loss": 0.6958, + "step": 30018 + }, + { + "epoch": 1.1, + "learning_rate": 9.025158384297066e-08, + "loss": 0.6874, + "step": 30019 + }, + { + "epoch": 1.1, + "learning_rate": 9.02479217782986e-08, + "loss": 0.6894, + "step": 30020 + }, + { + "epoch": 1.1, + "learning_rate": 9.024425971362654e-08, + "loss": 0.7055, + "step": 30021 + }, + { + "epoch": 1.1, + "learning_rate": 9.024059764895448e-08, + "loss": 0.7081, + "step": 30022 + }, + { + "epoch": 1.1, + "learning_rate": 9.023693558428242e-08, + "loss": 0.698, + "step": 30023 + }, + { + "epoch": 1.1, + "learning_rate": 9.023327351961036e-08, + "loss": 0.6991, + "step": 30024 + }, + { + "epoch": 1.1, + "learning_rate": 9.02296114549383e-08, + "loss": 0.6832, + "step": 30025 + }, + { + "epoch": 1.1, + "learning_rate": 9.022594939026623e-08, + "loss": 0.6912, + "step": 30026 + }, + { + "epoch": 1.1, + "learning_rate": 9.022228732559417e-08, + "loss": 0.6896, + "step": 30027 + }, + { + "epoch": 1.1, + "learning_rate": 9.02186252609221e-08, + "loss": 0.7011, + "step": 30028 + }, + { + "epoch": 1.1, + "learning_rate": 9.021496319625003e-08, + "loss": 0.6922, + "step": 30029 + }, + { + "epoch": 1.1, + "learning_rate": 9.021130113157797e-08, + "loss": 0.692, + "step": 30030 + }, + { + "epoch": 1.1, + "learning_rate": 9.020763906690591e-08, + "loss": 0.6997, + "step": 30031 + }, + { + "epoch": 1.1, + "learning_rate": 9.020397700223385e-08, + "loss": 0.6915, + "step": 30032 + }, + { + "epoch": 1.1, + "learning_rate": 9.020031493756179e-08, + "loss": 0.7004, + "step": 30033 + }, + { + "epoch": 1.1, + "learning_rate": 9.019665287288973e-08, + "loss": 0.6942, + "step": 30034 + }, + { + "epoch": 1.1, + "learning_rate": 9.019299080821767e-08, + "loss": 0.6953, + "step": 30035 + }, + { + "epoch": 1.1, + "learning_rate": 9.018932874354561e-08, + "loss": 0.6999, + "step": 30036 + }, + { + "epoch": 1.1, + "learning_rate": 9.018566667887354e-08, + "loss": 0.6915, + "step": 30037 + }, + { + "epoch": 1.1, + "learning_rate": 9.018200461420148e-08, + "loss": 0.6895, + "step": 30038 + }, + { + "epoch": 1.1, + "learning_rate": 9.017834254952942e-08, + "loss": 0.6952, + "step": 30039 + }, + { + "epoch": 1.1, + "learning_rate": 9.017468048485736e-08, + "loss": 0.6902, + "step": 30040 + }, + { + "epoch": 1.1, + "learning_rate": 9.01710184201853e-08, + "loss": 0.6954, + "step": 30041 + }, + { + "epoch": 1.1, + "learning_rate": 9.016735635551324e-08, + "loss": 0.6939, + "step": 30042 + }, + { + "epoch": 1.1, + "learning_rate": 9.016369429084118e-08, + "loss": 0.6995, + "step": 30043 + }, + { + "epoch": 1.1, + "learning_rate": 9.016003222616912e-08, + "loss": 0.6934, + "step": 30044 + }, + { + "epoch": 1.1, + "learning_rate": 9.015637016149706e-08, + "loss": 0.6856, + "step": 30045 + }, + { + "epoch": 1.1, + "learning_rate": 9.015270809682498e-08, + "loss": 0.6903, + "step": 30046 + }, + { + "epoch": 1.1, + "learning_rate": 9.014904603215292e-08, + "loss": 0.6916, + "step": 30047 + }, + { + "epoch": 1.1, + "learning_rate": 9.014538396748086e-08, + "loss": 0.6887, + "step": 30048 + }, + { + "epoch": 1.1, + "learning_rate": 9.01417219028088e-08, + "loss": 0.6993, + "step": 30049 + }, + { + "epoch": 1.1, + "learning_rate": 9.013805983813674e-08, + "loss": 0.6991, + "step": 30050 + }, + { + "epoch": 1.1, + "learning_rate": 9.013439777346467e-08, + "loss": 0.6753, + "step": 30051 + }, + { + "epoch": 1.1, + "learning_rate": 9.013073570879261e-08, + "loss": 0.7097, + "step": 30052 + }, + { + "epoch": 1.1, + "learning_rate": 9.012707364412055e-08, + "loss": 0.694, + "step": 30053 + }, + { + "epoch": 1.1, + "learning_rate": 9.012341157944849e-08, + "loss": 0.6835, + "step": 30054 + }, + { + "epoch": 1.1, + "learning_rate": 9.011974951477642e-08, + "loss": 0.6874, + "step": 30055 + }, + { + "epoch": 1.1, + "learning_rate": 9.011608745010436e-08, + "loss": 0.6928, + "step": 30056 + }, + { + "epoch": 1.1, + "learning_rate": 9.01124253854323e-08, + "loss": 0.6976, + "step": 30057 + }, + { + "epoch": 1.1, + "learning_rate": 9.010876332076024e-08, + "loss": 0.6877, + "step": 30058 + }, + { + "epoch": 1.1, + "learning_rate": 9.010510125608818e-08, + "loss": 0.6837, + "step": 30059 + }, + { + "epoch": 1.1, + "learning_rate": 9.010143919141612e-08, + "loss": 0.7032, + "step": 30060 + }, + { + "epoch": 1.1, + "learning_rate": 9.009777712674406e-08, + "loss": 0.6918, + "step": 30061 + }, + { + "epoch": 1.1, + "learning_rate": 9.0094115062072e-08, + "loss": 0.7005, + "step": 30062 + }, + { + "epoch": 1.1, + "learning_rate": 9.009045299739994e-08, + "loss": 0.6904, + "step": 30063 + }, + { + "epoch": 1.1, + "learning_rate": 9.008679093272786e-08, + "loss": 0.7001, + "step": 30064 + }, + { + "epoch": 1.1, + "learning_rate": 9.00831288680558e-08, + "loss": 0.6841, + "step": 30065 + }, + { + "epoch": 1.1, + "learning_rate": 9.007946680338374e-08, + "loss": 0.6867, + "step": 30066 + }, + { + "epoch": 1.1, + "learning_rate": 9.007580473871168e-08, + "loss": 0.6971, + "step": 30067 + }, + { + "epoch": 1.1, + "learning_rate": 9.007214267403962e-08, + "loss": 0.6913, + "step": 30068 + }, + { + "epoch": 1.1, + "learning_rate": 9.006848060936756e-08, + "loss": 0.6904, + "step": 30069 + }, + { + "epoch": 1.1, + "learning_rate": 9.00648185446955e-08, + "loss": 0.6948, + "step": 30070 + }, + { + "epoch": 1.1, + "learning_rate": 9.006115648002344e-08, + "loss": 0.7075, + "step": 30071 + }, + { + "epoch": 1.1, + "learning_rate": 9.005749441535137e-08, + "loss": 0.6909, + "step": 30072 + }, + { + "epoch": 1.1, + "learning_rate": 9.005383235067931e-08, + "loss": 0.7071, + "step": 30073 + }, + { + "epoch": 1.1, + "learning_rate": 9.005017028600725e-08, + "loss": 0.7021, + "step": 30074 + }, + { + "epoch": 1.1, + "learning_rate": 9.004650822133518e-08, + "loss": 0.6927, + "step": 30075 + }, + { + "epoch": 1.1, + "learning_rate": 9.004284615666312e-08, + "loss": 0.6839, + "step": 30076 + }, + { + "epoch": 1.1, + "learning_rate": 9.003918409199106e-08, + "loss": 0.6891, + "step": 30077 + }, + { + "epoch": 1.1, + "learning_rate": 9.0035522027319e-08, + "loss": 0.6812, + "step": 30078 + }, + { + "epoch": 1.1, + "learning_rate": 9.003185996264694e-08, + "loss": 0.6921, + "step": 30079 + }, + { + "epoch": 1.1, + "learning_rate": 9.002819789797488e-08, + "loss": 0.6963, + "step": 30080 + }, + { + "epoch": 1.1, + "learning_rate": 9.002453583330282e-08, + "loss": 0.6973, + "step": 30081 + }, + { + "epoch": 1.1, + "learning_rate": 9.002087376863074e-08, + "loss": 0.7029, + "step": 30082 + }, + { + "epoch": 1.1, + "learning_rate": 9.001721170395868e-08, + "loss": 0.6923, + "step": 30083 + }, + { + "epoch": 1.1, + "learning_rate": 9.001354963928662e-08, + "loss": 0.6998, + "step": 30084 + }, + { + "epoch": 1.1, + "learning_rate": 9.000988757461456e-08, + "loss": 0.6831, + "step": 30085 + }, + { + "epoch": 1.1, + "learning_rate": 9.00062255099425e-08, + "loss": 0.6826, + "step": 30086 + }, + { + "epoch": 1.1, + "learning_rate": 9.000256344527044e-08, + "loss": 0.6945, + "step": 30087 + }, + { + "epoch": 1.1, + "learning_rate": 8.999890138059838e-08, + "loss": 0.6978, + "step": 30088 + }, + { + "epoch": 1.1, + "learning_rate": 8.999523931592632e-08, + "loss": 0.6939, + "step": 30089 + }, + { + "epoch": 1.1, + "learning_rate": 8.999157725125426e-08, + "loss": 0.6984, + "step": 30090 + }, + { + "epoch": 1.1, + "learning_rate": 8.998791518658219e-08, + "loss": 0.6964, + "step": 30091 + }, + { + "epoch": 1.1, + "learning_rate": 8.998425312191013e-08, + "loss": 0.6974, + "step": 30092 + }, + { + "epoch": 1.1, + "learning_rate": 8.998059105723807e-08, + "loss": 0.6872, + "step": 30093 + }, + { + "epoch": 1.1, + "learning_rate": 8.997692899256601e-08, + "loss": 0.6982, + "step": 30094 + }, + { + "epoch": 1.1, + "learning_rate": 8.997326692789395e-08, + "loss": 0.6895, + "step": 30095 + }, + { + "epoch": 1.1, + "learning_rate": 8.996960486322189e-08, + "loss": 0.6918, + "step": 30096 + }, + { + "epoch": 1.1, + "learning_rate": 8.996594279854981e-08, + "loss": 0.6953, + "step": 30097 + }, + { + "epoch": 1.1, + "learning_rate": 8.996228073387775e-08, + "loss": 0.6773, + "step": 30098 + }, + { + "epoch": 1.1, + "learning_rate": 8.99586186692057e-08, + "loss": 0.698, + "step": 30099 + }, + { + "epoch": 1.1, + "learning_rate": 8.995495660453362e-08, + "loss": 0.6888, + "step": 30100 + }, + { + "epoch": 1.1, + "learning_rate": 8.995129453986156e-08, + "loss": 0.7045, + "step": 30101 + }, + { + "epoch": 1.1, + "learning_rate": 8.99476324751895e-08, + "loss": 0.6936, + "step": 30102 + }, + { + "epoch": 1.1, + "learning_rate": 8.994397041051744e-08, + "loss": 0.6932, + "step": 30103 + }, + { + "epoch": 1.1, + "learning_rate": 8.994030834584538e-08, + "loss": 0.7005, + "step": 30104 + }, + { + "epoch": 1.1, + "learning_rate": 8.993664628117332e-08, + "loss": 0.6851, + "step": 30105 + }, + { + "epoch": 1.1, + "learning_rate": 8.993298421650126e-08, + "loss": 0.6932, + "step": 30106 + }, + { + "epoch": 1.1, + "learning_rate": 8.99293221518292e-08, + "loss": 0.6987, + "step": 30107 + }, + { + "epoch": 1.1, + "learning_rate": 8.992566008715714e-08, + "loss": 0.6958, + "step": 30108 + }, + { + "epoch": 1.1, + "learning_rate": 8.992199802248507e-08, + "loss": 0.6824, + "step": 30109 + }, + { + "epoch": 1.1, + "learning_rate": 8.991833595781301e-08, + "loss": 0.7117, + "step": 30110 + }, + { + "epoch": 1.1, + "learning_rate": 8.991467389314095e-08, + "loss": 0.6943, + "step": 30111 + }, + { + "epoch": 1.1, + "learning_rate": 8.991101182846889e-08, + "loss": 0.6963, + "step": 30112 + }, + { + "epoch": 1.1, + "learning_rate": 8.990734976379683e-08, + "loss": 0.7025, + "step": 30113 + }, + { + "epoch": 1.1, + "learning_rate": 8.990368769912477e-08, + "loss": 0.6877, + "step": 30114 + }, + { + "epoch": 1.1, + "learning_rate": 8.990002563445271e-08, + "loss": 0.6977, + "step": 30115 + }, + { + "epoch": 1.1, + "learning_rate": 8.989636356978065e-08, + "loss": 0.6908, + "step": 30116 + }, + { + "epoch": 1.1, + "learning_rate": 8.989270150510859e-08, + "loss": 0.7079, + "step": 30117 + }, + { + "epoch": 1.1, + "learning_rate": 8.988903944043651e-08, + "loss": 0.6861, + "step": 30118 + }, + { + "epoch": 1.1, + "learning_rate": 8.988537737576445e-08, + "loss": 0.7032, + "step": 30119 + }, + { + "epoch": 1.1, + "learning_rate": 8.98817153110924e-08, + "loss": 0.6827, + "step": 30120 + }, + { + "epoch": 1.1, + "learning_rate": 8.987805324642032e-08, + "loss": 0.6938, + "step": 30121 + }, + { + "epoch": 1.1, + "learning_rate": 8.987439118174826e-08, + "loss": 0.6836, + "step": 30122 + }, + { + "epoch": 1.1, + "learning_rate": 8.98707291170762e-08, + "loss": 0.6909, + "step": 30123 + }, + { + "epoch": 1.1, + "learning_rate": 8.986706705240414e-08, + "loss": 0.6882, + "step": 30124 + }, + { + "epoch": 1.1, + "learning_rate": 8.986340498773208e-08, + "loss": 0.6958, + "step": 30125 + }, + { + "epoch": 1.1, + "learning_rate": 8.985974292306002e-08, + "loss": 0.6918, + "step": 30126 + }, + { + "epoch": 1.1, + "learning_rate": 8.985608085838795e-08, + "loss": 0.6886, + "step": 30127 + }, + { + "epoch": 1.1, + "learning_rate": 8.985241879371589e-08, + "loss": 0.6939, + "step": 30128 + }, + { + "epoch": 1.1, + "learning_rate": 8.984875672904383e-08, + "loss": 0.688, + "step": 30129 + }, + { + "epoch": 1.1, + "learning_rate": 8.984509466437177e-08, + "loss": 0.6898, + "step": 30130 + }, + { + "epoch": 1.1, + "learning_rate": 8.98414325996997e-08, + "loss": 0.6826, + "step": 30131 + }, + { + "epoch": 1.1, + "learning_rate": 8.983777053502765e-08, + "loss": 0.6978, + "step": 30132 + }, + { + "epoch": 1.1, + "learning_rate": 8.983410847035559e-08, + "loss": 0.6952, + "step": 30133 + }, + { + "epoch": 1.1, + "learning_rate": 8.983044640568353e-08, + "loss": 0.7, + "step": 30134 + }, + { + "epoch": 1.1, + "learning_rate": 8.982678434101147e-08, + "loss": 0.6987, + "step": 30135 + }, + { + "epoch": 1.1, + "learning_rate": 8.982312227633939e-08, + "loss": 0.7, + "step": 30136 + }, + { + "epoch": 1.1, + "learning_rate": 8.981946021166733e-08, + "loss": 0.6976, + "step": 30137 + }, + { + "epoch": 1.1, + "learning_rate": 8.981579814699527e-08, + "loss": 0.6905, + "step": 30138 + }, + { + "epoch": 1.1, + "learning_rate": 8.981213608232321e-08, + "loss": 0.6973, + "step": 30139 + }, + { + "epoch": 1.1, + "learning_rate": 8.980847401765115e-08, + "loss": 0.6817, + "step": 30140 + }, + { + "epoch": 1.1, + "learning_rate": 8.980481195297909e-08, + "loss": 0.6839, + "step": 30141 + }, + { + "epoch": 1.1, + "learning_rate": 8.980114988830703e-08, + "loss": 0.6785, + "step": 30142 + }, + { + "epoch": 1.1, + "learning_rate": 8.979748782363496e-08, + "loss": 0.6921, + "step": 30143 + }, + { + "epoch": 1.1, + "learning_rate": 8.97938257589629e-08, + "loss": 0.6909, + "step": 30144 + }, + { + "epoch": 1.1, + "learning_rate": 8.979016369429084e-08, + "loss": 0.6958, + "step": 30145 + }, + { + "epoch": 1.1, + "learning_rate": 8.978650162961877e-08, + "loss": 0.6899, + "step": 30146 + }, + { + "epoch": 1.1, + "learning_rate": 8.97828395649467e-08, + "loss": 0.6935, + "step": 30147 + }, + { + "epoch": 1.1, + "learning_rate": 8.977917750027465e-08, + "loss": 0.6944, + "step": 30148 + }, + { + "epoch": 1.1, + "learning_rate": 8.977551543560259e-08, + "loss": 0.6977, + "step": 30149 + }, + { + "epoch": 1.1, + "learning_rate": 8.977185337093053e-08, + "loss": 0.7073, + "step": 30150 + }, + { + "epoch": 1.1, + "learning_rate": 8.976819130625847e-08, + "loss": 0.7013, + "step": 30151 + }, + { + "epoch": 1.1, + "learning_rate": 8.97645292415864e-08, + "loss": 0.6945, + "step": 30152 + }, + { + "epoch": 1.1, + "learning_rate": 8.976086717691434e-08, + "loss": 0.6897, + "step": 30153 + }, + { + "epoch": 1.1, + "learning_rate": 8.975720511224227e-08, + "loss": 0.6925, + "step": 30154 + }, + { + "epoch": 1.1, + "learning_rate": 8.975354304757021e-08, + "loss": 0.6922, + "step": 30155 + }, + { + "epoch": 1.1, + "learning_rate": 8.974988098289815e-08, + "loss": 0.6872, + "step": 30156 + }, + { + "epoch": 1.1, + "learning_rate": 8.974621891822609e-08, + "loss": 0.6825, + "step": 30157 + }, + { + "epoch": 1.1, + "learning_rate": 8.974255685355403e-08, + "loss": 0.6878, + "step": 30158 + }, + { + "epoch": 1.1, + "learning_rate": 8.973889478888197e-08, + "loss": 0.6985, + "step": 30159 + }, + { + "epoch": 1.1, + "learning_rate": 8.973523272420991e-08, + "loss": 0.6828, + "step": 30160 + }, + { + "epoch": 1.1, + "learning_rate": 8.973157065953785e-08, + "loss": 0.6977, + "step": 30161 + }, + { + "epoch": 1.1, + "learning_rate": 8.972790859486579e-08, + "loss": 0.687, + "step": 30162 + }, + { + "epoch": 1.1, + "learning_rate": 8.972424653019372e-08, + "loss": 0.6962, + "step": 30163 + }, + { + "epoch": 1.1, + "learning_rate": 8.972058446552166e-08, + "loss": 0.6954, + "step": 30164 + }, + { + "epoch": 1.1, + "learning_rate": 8.97169224008496e-08, + "loss": 0.6817, + "step": 30165 + }, + { + "epoch": 1.1, + "learning_rate": 8.971326033617754e-08, + "loss": 0.6883, + "step": 30166 + }, + { + "epoch": 1.1, + "learning_rate": 8.970959827150548e-08, + "loss": 0.6897, + "step": 30167 + }, + { + "epoch": 1.1, + "learning_rate": 8.97059362068334e-08, + "loss": 0.6846, + "step": 30168 + }, + { + "epoch": 1.1, + "learning_rate": 8.970227414216134e-08, + "loss": 0.6878, + "step": 30169 + }, + { + "epoch": 1.1, + "learning_rate": 8.969861207748928e-08, + "loss": 0.686, + "step": 30170 + }, + { + "epoch": 1.1, + "learning_rate": 8.969495001281722e-08, + "loss": 0.6939, + "step": 30171 + }, + { + "epoch": 1.1, + "learning_rate": 8.969128794814515e-08, + "loss": 0.6941, + "step": 30172 + }, + { + "epoch": 1.1, + "learning_rate": 8.968762588347309e-08, + "loss": 0.6843, + "step": 30173 + }, + { + "epoch": 1.1, + "learning_rate": 8.968396381880103e-08, + "loss": 0.6927, + "step": 30174 + }, + { + "epoch": 1.1, + "learning_rate": 8.968030175412897e-08, + "loss": 0.6785, + "step": 30175 + }, + { + "epoch": 1.1, + "learning_rate": 8.967663968945691e-08, + "loss": 0.6963, + "step": 30176 + }, + { + "epoch": 1.1, + "learning_rate": 8.967297762478485e-08, + "loss": 0.6943, + "step": 30177 + }, + { + "epoch": 1.1, + "learning_rate": 8.966931556011279e-08, + "loss": 0.6904, + "step": 30178 + }, + { + "epoch": 1.1, + "learning_rate": 8.966565349544073e-08, + "loss": 0.6813, + "step": 30179 + }, + { + "epoch": 1.1, + "learning_rate": 8.966199143076867e-08, + "loss": 0.6991, + "step": 30180 + }, + { + "epoch": 1.1, + "learning_rate": 8.96583293660966e-08, + "loss": 0.7008, + "step": 30181 + }, + { + "epoch": 1.1, + "learning_rate": 8.965466730142454e-08, + "loss": 0.7054, + "step": 30182 + }, + { + "epoch": 1.1, + "learning_rate": 8.965100523675248e-08, + "loss": 0.6938, + "step": 30183 + }, + { + "epoch": 1.1, + "learning_rate": 8.964734317208042e-08, + "loss": 0.7045, + "step": 30184 + }, + { + "epoch": 1.1, + "learning_rate": 8.964368110740836e-08, + "loss": 0.6876, + "step": 30185 + }, + { + "epoch": 1.1, + "learning_rate": 8.96400190427363e-08, + "loss": 0.6958, + "step": 30186 + }, + { + "epoch": 1.1, + "learning_rate": 8.963635697806424e-08, + "loss": 0.7039, + "step": 30187 + }, + { + "epoch": 1.1, + "learning_rate": 8.963269491339218e-08, + "loss": 0.7025, + "step": 30188 + }, + { + "epoch": 1.1, + "learning_rate": 8.962903284872012e-08, + "loss": 0.683, + "step": 30189 + }, + { + "epoch": 1.1, + "learning_rate": 8.962537078404804e-08, + "loss": 0.6836, + "step": 30190 + }, + { + "epoch": 1.1, + "learning_rate": 8.962170871937598e-08, + "loss": 0.6977, + "step": 30191 + }, + { + "epoch": 1.1, + "learning_rate": 8.961804665470391e-08, + "loss": 0.6935, + "step": 30192 + }, + { + "epoch": 1.1, + "learning_rate": 8.961438459003185e-08, + "loss": 0.7024, + "step": 30193 + }, + { + "epoch": 1.1, + "learning_rate": 8.961072252535979e-08, + "loss": 0.7059, + "step": 30194 + }, + { + "epoch": 1.1, + "learning_rate": 8.960706046068773e-08, + "loss": 0.699, + "step": 30195 + }, + { + "epoch": 1.1, + "learning_rate": 8.960339839601567e-08, + "loss": 0.6875, + "step": 30196 + }, + { + "epoch": 1.1, + "learning_rate": 8.959973633134361e-08, + "loss": 0.6927, + "step": 30197 + }, + { + "epoch": 1.1, + "learning_rate": 8.959607426667155e-08, + "loss": 0.6914, + "step": 30198 + }, + { + "epoch": 1.1, + "learning_rate": 8.959241220199948e-08, + "loss": 0.6935, + "step": 30199 + }, + { + "epoch": 1.1, + "learning_rate": 8.958875013732742e-08, + "loss": 0.6943, + "step": 30200 + }, + { + "epoch": 1.1, + "learning_rate": 8.958508807265536e-08, + "loss": 0.698, + "step": 30201 + }, + { + "epoch": 1.11, + "learning_rate": 8.95814260079833e-08, + "loss": 0.7056, + "step": 30202 + }, + { + "epoch": 1.11, + "learning_rate": 8.957776394331124e-08, + "loss": 0.6925, + "step": 30203 + }, + { + "epoch": 1.11, + "learning_rate": 8.957410187863918e-08, + "loss": 0.6594, + "step": 30204 + }, + { + "epoch": 1.11, + "learning_rate": 8.957043981396712e-08, + "loss": 0.7013, + "step": 30205 + }, + { + "epoch": 1.11, + "learning_rate": 8.956677774929506e-08, + "loss": 0.6911, + "step": 30206 + }, + { + "epoch": 1.11, + "learning_rate": 8.9563115684623e-08, + "loss": 0.6968, + "step": 30207 + }, + { + "epoch": 1.11, + "learning_rate": 8.955945361995092e-08, + "loss": 0.6862, + "step": 30208 + }, + { + "epoch": 1.11, + "learning_rate": 8.955579155527886e-08, + "loss": 0.7013, + "step": 30209 + }, + { + "epoch": 1.11, + "learning_rate": 8.95521294906068e-08, + "loss": 0.6876, + "step": 30210 + }, + { + "epoch": 1.11, + "learning_rate": 8.954846742593474e-08, + "loss": 0.6856, + "step": 30211 + }, + { + "epoch": 1.11, + "learning_rate": 8.954480536126268e-08, + "loss": 0.6911, + "step": 30212 + }, + { + "epoch": 1.11, + "learning_rate": 8.954114329659062e-08, + "loss": 0.6921, + "step": 30213 + }, + { + "epoch": 1.11, + "learning_rate": 8.953748123191855e-08, + "loss": 0.6947, + "step": 30214 + }, + { + "epoch": 1.11, + "learning_rate": 8.953381916724649e-08, + "loss": 0.6913, + "step": 30215 + }, + { + "epoch": 1.11, + "learning_rate": 8.953015710257443e-08, + "loss": 0.6911, + "step": 30216 + }, + { + "epoch": 1.11, + "learning_rate": 8.952649503790235e-08, + "loss": 0.7103, + "step": 30217 + }, + { + "epoch": 1.11, + "learning_rate": 8.95228329732303e-08, + "loss": 0.6919, + "step": 30218 + }, + { + "epoch": 1.11, + "learning_rate": 8.951917090855823e-08, + "loss": 0.7, + "step": 30219 + }, + { + "epoch": 1.11, + "learning_rate": 8.951550884388617e-08, + "loss": 0.6768, + "step": 30220 + }, + { + "epoch": 1.11, + "learning_rate": 8.951184677921411e-08, + "loss": 0.6916, + "step": 30221 + }, + { + "epoch": 1.11, + "learning_rate": 8.950818471454205e-08, + "loss": 0.6909, + "step": 30222 + }, + { + "epoch": 1.11, + "learning_rate": 8.950452264987e-08, + "loss": 0.6898, + "step": 30223 + }, + { + "epoch": 1.11, + "learning_rate": 8.950086058519793e-08, + "loss": 0.6782, + "step": 30224 + }, + { + "epoch": 1.11, + "learning_rate": 8.949719852052587e-08, + "loss": 0.7035, + "step": 30225 + }, + { + "epoch": 1.11, + "learning_rate": 8.94935364558538e-08, + "loss": 0.6938, + "step": 30226 + }, + { + "epoch": 1.11, + "learning_rate": 8.948987439118174e-08, + "loss": 0.6913, + "step": 30227 + }, + { + "epoch": 1.11, + "learning_rate": 8.948621232650968e-08, + "loss": 0.6928, + "step": 30228 + }, + { + "epoch": 1.11, + "learning_rate": 8.948255026183762e-08, + "loss": 0.6947, + "step": 30229 + }, + { + "epoch": 1.11, + "learning_rate": 8.947888819716556e-08, + "loss": 0.674, + "step": 30230 + }, + { + "epoch": 1.11, + "learning_rate": 8.94752261324935e-08, + "loss": 0.6889, + "step": 30231 + }, + { + "epoch": 1.11, + "learning_rate": 8.947156406782144e-08, + "loss": 0.6831, + "step": 30232 + }, + { + "epoch": 1.11, + "learning_rate": 8.946790200314938e-08, + "loss": 0.6776, + "step": 30233 + }, + { + "epoch": 1.11, + "learning_rate": 8.946423993847732e-08, + "loss": 0.7107, + "step": 30234 + }, + { + "epoch": 1.11, + "learning_rate": 8.946057787380525e-08, + "loss": 0.7033, + "step": 30235 + }, + { + "epoch": 1.11, + "learning_rate": 8.945691580913319e-08, + "loss": 0.7137, + "step": 30236 + }, + { + "epoch": 1.11, + "learning_rate": 8.945325374446113e-08, + "loss": 0.6982, + "step": 30237 + }, + { + "epoch": 1.11, + "learning_rate": 8.944959167978907e-08, + "loss": 0.6817, + "step": 30238 + }, + { + "epoch": 1.11, + "learning_rate": 8.9445929615117e-08, + "loss": 0.6849, + "step": 30239 + }, + { + "epoch": 1.11, + "learning_rate": 8.944226755044493e-08, + "loss": 0.686, + "step": 30240 + }, + { + "epoch": 1.11, + "learning_rate": 8.943860548577287e-08, + "loss": 0.698, + "step": 30241 + }, + { + "epoch": 1.11, + "learning_rate": 8.943494342110081e-08, + "loss": 0.6981, + "step": 30242 + }, + { + "epoch": 1.11, + "learning_rate": 8.943128135642874e-08, + "loss": 0.7013, + "step": 30243 + }, + { + "epoch": 1.11, + "learning_rate": 8.942761929175668e-08, + "loss": 0.6918, + "step": 30244 + }, + { + "epoch": 1.11, + "learning_rate": 8.942395722708462e-08, + "loss": 0.7017, + "step": 30245 + }, + { + "epoch": 1.11, + "learning_rate": 8.942029516241256e-08, + "loss": 0.6903, + "step": 30246 + }, + { + "epoch": 1.11, + "learning_rate": 8.94166330977405e-08, + "loss": 0.7066, + "step": 30247 + }, + { + "epoch": 1.11, + "learning_rate": 8.941297103306844e-08, + "loss": 0.7033, + "step": 30248 + }, + { + "epoch": 1.11, + "learning_rate": 8.940930896839638e-08, + "loss": 0.6906, + "step": 30249 + }, + { + "epoch": 1.11, + "learning_rate": 8.940564690372432e-08, + "loss": 0.6956, + "step": 30250 + }, + { + "epoch": 1.11, + "learning_rate": 8.940198483905226e-08, + "loss": 0.6819, + "step": 30251 + }, + { + "epoch": 1.11, + "learning_rate": 8.939832277438019e-08, + "loss": 0.6921, + "step": 30252 + }, + { + "epoch": 1.11, + "learning_rate": 8.939466070970813e-08, + "loss": 0.6921, + "step": 30253 + }, + { + "epoch": 1.11, + "learning_rate": 8.939099864503607e-08, + "loss": 0.7018, + "step": 30254 + }, + { + "epoch": 1.11, + "learning_rate": 8.9387336580364e-08, + "loss": 0.6881, + "step": 30255 + }, + { + "epoch": 1.11, + "learning_rate": 8.938367451569195e-08, + "loss": 0.7172, + "step": 30256 + }, + { + "epoch": 1.11, + "learning_rate": 8.938001245101989e-08, + "loss": 0.7012, + "step": 30257 + }, + { + "epoch": 1.11, + "learning_rate": 8.937635038634783e-08, + "loss": 0.6953, + "step": 30258 + }, + { + "epoch": 1.11, + "learning_rate": 8.937268832167577e-08, + "loss": 0.6885, + "step": 30259 + }, + { + "epoch": 1.11, + "learning_rate": 8.93690262570037e-08, + "loss": 0.7077, + "step": 30260 + }, + { + "epoch": 1.11, + "learning_rate": 8.936536419233163e-08, + "loss": 0.6947, + "step": 30261 + }, + { + "epoch": 1.11, + "learning_rate": 8.936170212765957e-08, + "loss": 0.6821, + "step": 30262 + }, + { + "epoch": 1.11, + "learning_rate": 8.93580400629875e-08, + "loss": 0.6982, + "step": 30263 + }, + { + "epoch": 1.11, + "learning_rate": 8.935437799831544e-08, + "loss": 0.6928, + "step": 30264 + }, + { + "epoch": 1.11, + "learning_rate": 8.935071593364338e-08, + "loss": 0.6879, + "step": 30265 + }, + { + "epoch": 1.11, + "learning_rate": 8.934705386897132e-08, + "loss": 0.6823, + "step": 30266 + }, + { + "epoch": 1.11, + "learning_rate": 8.934339180429926e-08, + "loss": 0.6982, + "step": 30267 + }, + { + "epoch": 1.11, + "learning_rate": 8.93397297396272e-08, + "loss": 0.6918, + "step": 30268 + }, + { + "epoch": 1.11, + "learning_rate": 8.933606767495514e-08, + "loss": 0.6906, + "step": 30269 + }, + { + "epoch": 1.11, + "learning_rate": 8.933240561028307e-08, + "loss": 0.6917, + "step": 30270 + }, + { + "epoch": 1.11, + "learning_rate": 8.9328743545611e-08, + "loss": 0.6933, + "step": 30271 + }, + { + "epoch": 1.11, + "learning_rate": 8.932508148093894e-08, + "loss": 0.6853, + "step": 30272 + }, + { + "epoch": 1.11, + "learning_rate": 8.932141941626688e-08, + "loss": 0.6817, + "step": 30273 + }, + { + "epoch": 1.11, + "learning_rate": 8.931775735159482e-08, + "loss": 0.69, + "step": 30274 + }, + { + "epoch": 1.11, + "learning_rate": 8.931409528692276e-08, + "loss": 0.7009, + "step": 30275 + }, + { + "epoch": 1.11, + "learning_rate": 8.93104332222507e-08, + "loss": 0.6929, + "step": 30276 + }, + { + "epoch": 1.11, + "learning_rate": 8.930677115757864e-08, + "loss": 0.7028, + "step": 30277 + }, + { + "epoch": 1.11, + "learning_rate": 8.930310909290658e-08, + "loss": 0.6922, + "step": 30278 + }, + { + "epoch": 1.11, + "learning_rate": 8.929944702823451e-08, + "loss": 0.7, + "step": 30279 + }, + { + "epoch": 1.11, + "learning_rate": 8.929578496356245e-08, + "loss": 0.695, + "step": 30280 + }, + { + "epoch": 1.11, + "learning_rate": 8.929212289889039e-08, + "loss": 0.6833, + "step": 30281 + }, + { + "epoch": 1.11, + "learning_rate": 8.928846083421833e-08, + "loss": 0.6887, + "step": 30282 + }, + { + "epoch": 1.11, + "learning_rate": 8.928479876954627e-08, + "loss": 0.6967, + "step": 30283 + }, + { + "epoch": 1.11, + "learning_rate": 8.928113670487421e-08, + "loss": 0.7028, + "step": 30284 + }, + { + "epoch": 1.11, + "learning_rate": 8.927747464020214e-08, + "loss": 0.6931, + "step": 30285 + }, + { + "epoch": 1.11, + "learning_rate": 8.927381257553008e-08, + "loss": 0.6832, + "step": 30286 + }, + { + "epoch": 1.11, + "learning_rate": 8.927015051085802e-08, + "loss": 0.6867, + "step": 30287 + }, + { + "epoch": 1.11, + "learning_rate": 8.926648844618594e-08, + "loss": 0.6858, + "step": 30288 + }, + { + "epoch": 1.11, + "learning_rate": 8.926282638151388e-08, + "loss": 0.7162, + "step": 30289 + }, + { + "epoch": 1.11, + "learning_rate": 8.925916431684182e-08, + "loss": 0.6952, + "step": 30290 + }, + { + "epoch": 1.11, + "learning_rate": 8.925550225216976e-08, + "loss": 0.6919, + "step": 30291 + }, + { + "epoch": 1.11, + "learning_rate": 8.92518401874977e-08, + "loss": 0.6931, + "step": 30292 + }, + { + "epoch": 1.11, + "learning_rate": 8.924817812282564e-08, + "loss": 0.6892, + "step": 30293 + }, + { + "epoch": 1.11, + "learning_rate": 8.924451605815358e-08, + "loss": 0.6902, + "step": 30294 + }, + { + "epoch": 1.11, + "learning_rate": 8.924085399348152e-08, + "loss": 0.6947, + "step": 30295 + }, + { + "epoch": 1.11, + "learning_rate": 8.923719192880946e-08, + "loss": 0.6963, + "step": 30296 + }, + { + "epoch": 1.11, + "learning_rate": 8.923352986413739e-08, + "loss": 0.6871, + "step": 30297 + }, + { + "epoch": 1.11, + "learning_rate": 8.922986779946533e-08, + "loss": 0.6925, + "step": 30298 + }, + { + "epoch": 1.11, + "learning_rate": 8.922620573479327e-08, + "loss": 0.7155, + "step": 30299 + }, + { + "epoch": 1.11, + "learning_rate": 8.922254367012121e-08, + "loss": 0.6882, + "step": 30300 + }, + { + "epoch": 1.11, + "learning_rate": 8.921888160544915e-08, + "loss": 0.6919, + "step": 30301 + }, + { + "epoch": 1.11, + "learning_rate": 8.921521954077709e-08, + "loss": 0.6792, + "step": 30302 + }, + { + "epoch": 1.11, + "learning_rate": 8.921155747610503e-08, + "loss": 0.696, + "step": 30303 + }, + { + "epoch": 1.11, + "learning_rate": 8.920789541143297e-08, + "loss": 0.6958, + "step": 30304 + }, + { + "epoch": 1.11, + "learning_rate": 8.920423334676091e-08, + "loss": 0.6977, + "step": 30305 + }, + { + "epoch": 1.11, + "learning_rate": 8.920057128208884e-08, + "loss": 0.7061, + "step": 30306 + }, + { + "epoch": 1.11, + "learning_rate": 8.919690921741678e-08, + "loss": 0.7059, + "step": 30307 + }, + { + "epoch": 1.11, + "learning_rate": 8.919324715274472e-08, + "loss": 0.7099, + "step": 30308 + }, + { + "epoch": 1.11, + "learning_rate": 8.918958508807266e-08, + "loss": 0.6892, + "step": 30309 + }, + { + "epoch": 1.11, + "learning_rate": 8.918592302340058e-08, + "loss": 0.6939, + "step": 30310 + }, + { + "epoch": 1.11, + "learning_rate": 8.918226095872852e-08, + "loss": 0.6932, + "step": 30311 + }, + { + "epoch": 1.11, + "learning_rate": 8.917859889405646e-08, + "loss": 0.7034, + "step": 30312 + }, + { + "epoch": 1.11, + "learning_rate": 8.91749368293844e-08, + "loss": 0.7014, + "step": 30313 + }, + { + "epoch": 1.11, + "learning_rate": 8.917127476471234e-08, + "loss": 0.6925, + "step": 30314 + }, + { + "epoch": 1.11, + "learning_rate": 8.916761270004027e-08, + "loss": 0.6922, + "step": 30315 + }, + { + "epoch": 1.11, + "learning_rate": 8.916395063536821e-08, + "loss": 0.6917, + "step": 30316 + }, + { + "epoch": 1.11, + "learning_rate": 8.916028857069615e-08, + "loss": 0.6868, + "step": 30317 + }, + { + "epoch": 1.11, + "learning_rate": 8.915662650602409e-08, + "loss": 0.7121, + "step": 30318 + }, + { + "epoch": 1.11, + "learning_rate": 8.915296444135203e-08, + "loss": 0.6888, + "step": 30319 + }, + { + "epoch": 1.11, + "learning_rate": 8.914930237667997e-08, + "loss": 0.6922, + "step": 30320 + }, + { + "epoch": 1.11, + "learning_rate": 8.914564031200791e-08, + "loss": 0.6848, + "step": 30321 + }, + { + "epoch": 1.11, + "learning_rate": 8.914197824733585e-08, + "loss": 0.6926, + "step": 30322 + }, + { + "epoch": 1.11, + "learning_rate": 8.913831618266379e-08, + "loss": 0.6907, + "step": 30323 + }, + { + "epoch": 1.11, + "learning_rate": 8.913465411799172e-08, + "loss": 0.6821, + "step": 30324 + }, + { + "epoch": 1.11, + "learning_rate": 8.913099205331966e-08, + "loss": 0.6923, + "step": 30325 + }, + { + "epoch": 1.11, + "learning_rate": 8.91273299886476e-08, + "loss": 0.6989, + "step": 30326 + }, + { + "epoch": 1.11, + "learning_rate": 8.912366792397554e-08, + "loss": 0.6899, + "step": 30327 + }, + { + "epoch": 1.11, + "learning_rate": 8.912000585930348e-08, + "loss": 0.6858, + "step": 30328 + }, + { + "epoch": 1.11, + "learning_rate": 8.911634379463141e-08, + "loss": 0.6864, + "step": 30329 + }, + { + "epoch": 1.11, + "learning_rate": 8.911268172995935e-08, + "loss": 0.6944, + "step": 30330 + }, + { + "epoch": 1.11, + "learning_rate": 8.91090196652873e-08, + "loss": 0.6951, + "step": 30331 + }, + { + "epoch": 1.11, + "learning_rate": 8.910535760061522e-08, + "loss": 0.6977, + "step": 30332 + }, + { + "epoch": 1.11, + "learning_rate": 8.910169553594316e-08, + "loss": 0.6863, + "step": 30333 + }, + { + "epoch": 1.11, + "learning_rate": 8.909803347127109e-08, + "loss": 0.6862, + "step": 30334 + }, + { + "epoch": 1.11, + "learning_rate": 8.909437140659903e-08, + "loss": 0.7068, + "step": 30335 + }, + { + "epoch": 1.11, + "learning_rate": 8.909070934192697e-08, + "loss": 0.6827, + "step": 30336 + }, + { + "epoch": 1.11, + "learning_rate": 8.908704727725491e-08, + "loss": 0.6978, + "step": 30337 + }, + { + "epoch": 1.11, + "learning_rate": 8.908338521258285e-08, + "loss": 0.7136, + "step": 30338 + }, + { + "epoch": 1.11, + "learning_rate": 8.907972314791079e-08, + "loss": 0.696, + "step": 30339 + }, + { + "epoch": 1.11, + "learning_rate": 8.907606108323873e-08, + "loss": 0.6823, + "step": 30340 + }, + { + "epoch": 1.11, + "learning_rate": 8.907239901856667e-08, + "loss": 0.7004, + "step": 30341 + }, + { + "epoch": 1.11, + "learning_rate": 8.90687369538946e-08, + "loss": 0.6797, + "step": 30342 + }, + { + "epoch": 1.11, + "learning_rate": 8.906507488922253e-08, + "loss": 0.6823, + "step": 30343 + }, + { + "epoch": 1.11, + "learning_rate": 8.906141282455047e-08, + "loss": 0.6933, + "step": 30344 + }, + { + "epoch": 1.11, + "learning_rate": 8.905775075987841e-08, + "loss": 0.6857, + "step": 30345 + }, + { + "epoch": 1.11, + "learning_rate": 8.905408869520635e-08, + "loss": 0.6897, + "step": 30346 + }, + { + "epoch": 1.11, + "learning_rate": 8.90504266305343e-08, + "loss": 0.6801, + "step": 30347 + }, + { + "epoch": 1.11, + "learning_rate": 8.904676456586223e-08, + "loss": 0.6997, + "step": 30348 + }, + { + "epoch": 1.11, + "learning_rate": 8.904310250119017e-08, + "loss": 0.6999, + "step": 30349 + }, + { + "epoch": 1.11, + "learning_rate": 8.903944043651811e-08, + "loss": 0.6946, + "step": 30350 + }, + { + "epoch": 1.11, + "learning_rate": 8.903577837184604e-08, + "loss": 0.6778, + "step": 30351 + }, + { + "epoch": 1.11, + "learning_rate": 8.903211630717398e-08, + "loss": 0.702, + "step": 30352 + }, + { + "epoch": 1.11, + "learning_rate": 8.902845424250192e-08, + "loss": 0.6889, + "step": 30353 + }, + { + "epoch": 1.11, + "learning_rate": 8.902479217782986e-08, + "loss": 0.6926, + "step": 30354 + }, + { + "epoch": 1.11, + "learning_rate": 8.90211301131578e-08, + "loss": 0.6853, + "step": 30355 + }, + { + "epoch": 1.11, + "learning_rate": 8.901746804848573e-08, + "loss": 0.6976, + "step": 30356 + }, + { + "epoch": 1.11, + "learning_rate": 8.901380598381367e-08, + "loss": 0.6963, + "step": 30357 + }, + { + "epoch": 1.11, + "learning_rate": 8.901014391914161e-08, + "loss": 0.6961, + "step": 30358 + }, + { + "epoch": 1.11, + "learning_rate": 8.900648185446955e-08, + "loss": 0.7146, + "step": 30359 + }, + { + "epoch": 1.11, + "learning_rate": 8.900281978979747e-08, + "loss": 0.7135, + "step": 30360 + }, + { + "epoch": 1.11, + "learning_rate": 8.899915772512541e-08, + "loss": 0.6971, + "step": 30361 + }, + { + "epoch": 1.11, + "learning_rate": 8.899549566045335e-08, + "loss": 0.7003, + "step": 30362 + }, + { + "epoch": 1.11, + "learning_rate": 8.899183359578129e-08, + "loss": 0.6946, + "step": 30363 + }, + { + "epoch": 1.11, + "learning_rate": 8.898817153110923e-08, + "loss": 0.7031, + "step": 30364 + }, + { + "epoch": 1.11, + "learning_rate": 8.898450946643717e-08, + "loss": 0.6892, + "step": 30365 + }, + { + "epoch": 1.11, + "learning_rate": 8.898084740176511e-08, + "loss": 0.688, + "step": 30366 + }, + { + "epoch": 1.11, + "learning_rate": 8.897718533709305e-08, + "loss": 0.6889, + "step": 30367 + }, + { + "epoch": 1.11, + "learning_rate": 8.897352327242099e-08, + "loss": 0.6948, + "step": 30368 + }, + { + "epoch": 1.11, + "learning_rate": 8.896986120774892e-08, + "loss": 0.7088, + "step": 30369 + }, + { + "epoch": 1.11, + "learning_rate": 8.896619914307686e-08, + "loss": 0.6946, + "step": 30370 + }, + { + "epoch": 1.11, + "learning_rate": 8.89625370784048e-08, + "loss": 0.7017, + "step": 30371 + }, + { + "epoch": 1.11, + "learning_rate": 8.895887501373274e-08, + "loss": 0.6929, + "step": 30372 + }, + { + "epoch": 1.11, + "learning_rate": 8.895521294906068e-08, + "loss": 0.6916, + "step": 30373 + }, + { + "epoch": 1.11, + "learning_rate": 8.895155088438862e-08, + "loss": 0.6909, + "step": 30374 + }, + { + "epoch": 1.11, + "learning_rate": 8.894788881971656e-08, + "loss": 0.6944, + "step": 30375 + }, + { + "epoch": 1.11, + "learning_rate": 8.89442267550445e-08, + "loss": 0.6825, + "step": 30376 + }, + { + "epoch": 1.11, + "learning_rate": 8.894056469037244e-08, + "loss": 0.6905, + "step": 30377 + }, + { + "epoch": 1.11, + "learning_rate": 8.893690262570037e-08, + "loss": 0.674, + "step": 30378 + }, + { + "epoch": 1.11, + "learning_rate": 8.89332405610283e-08, + "loss": 0.6971, + "step": 30379 + }, + { + "epoch": 1.11, + "learning_rate": 8.892957849635625e-08, + "loss": 0.6936, + "step": 30380 + }, + { + "epoch": 1.11, + "learning_rate": 8.892591643168417e-08, + "loss": 0.6927, + "step": 30381 + }, + { + "epoch": 1.11, + "learning_rate": 8.892225436701211e-08, + "loss": 0.6959, + "step": 30382 + }, + { + "epoch": 1.11, + "learning_rate": 8.891859230234005e-08, + "loss": 0.7021, + "step": 30383 + }, + { + "epoch": 1.11, + "learning_rate": 8.891493023766799e-08, + "loss": 0.6929, + "step": 30384 + }, + { + "epoch": 1.11, + "learning_rate": 8.891126817299593e-08, + "loss": 0.7003, + "step": 30385 + }, + { + "epoch": 1.11, + "learning_rate": 8.890760610832387e-08, + "loss": 0.6808, + "step": 30386 + }, + { + "epoch": 1.11, + "learning_rate": 8.89039440436518e-08, + "loss": 0.7023, + "step": 30387 + }, + { + "epoch": 1.11, + "learning_rate": 8.890028197897974e-08, + "loss": 0.699, + "step": 30388 + }, + { + "epoch": 1.11, + "learning_rate": 8.889661991430768e-08, + "loss": 0.6995, + "step": 30389 + }, + { + "epoch": 1.11, + "learning_rate": 8.889295784963562e-08, + "loss": 0.6912, + "step": 30390 + }, + { + "epoch": 1.11, + "learning_rate": 8.888929578496356e-08, + "loss": 0.7016, + "step": 30391 + }, + { + "epoch": 1.11, + "learning_rate": 8.88856337202915e-08, + "loss": 0.6918, + "step": 30392 + }, + { + "epoch": 1.11, + "learning_rate": 8.888197165561944e-08, + "loss": 0.6906, + "step": 30393 + }, + { + "epoch": 1.11, + "learning_rate": 8.887830959094738e-08, + "loss": 0.6899, + "step": 30394 + }, + { + "epoch": 1.11, + "learning_rate": 8.887464752627532e-08, + "loss": 0.6901, + "step": 30395 + }, + { + "epoch": 1.11, + "learning_rate": 8.887098546160324e-08, + "loss": 0.7025, + "step": 30396 + }, + { + "epoch": 1.11, + "learning_rate": 8.886732339693118e-08, + "loss": 0.6978, + "step": 30397 + }, + { + "epoch": 1.11, + "learning_rate": 8.886366133225912e-08, + "loss": 0.6935, + "step": 30398 + }, + { + "epoch": 1.11, + "learning_rate": 8.885999926758706e-08, + "loss": 0.699, + "step": 30399 + }, + { + "epoch": 1.11, + "learning_rate": 8.8856337202915e-08, + "loss": 0.6867, + "step": 30400 + }, + { + "epoch": 1.11, + "learning_rate": 8.885267513824294e-08, + "loss": 0.6966, + "step": 30401 + }, + { + "epoch": 1.11, + "learning_rate": 8.884901307357088e-08, + "loss": 0.691, + "step": 30402 + }, + { + "epoch": 1.11, + "learning_rate": 8.884535100889881e-08, + "loss": 0.6876, + "step": 30403 + }, + { + "epoch": 1.11, + "learning_rate": 8.884168894422675e-08, + "loss": 0.6902, + "step": 30404 + }, + { + "epoch": 1.11, + "learning_rate": 8.883802687955468e-08, + "loss": 0.7036, + "step": 30405 + }, + { + "epoch": 1.11, + "learning_rate": 8.883436481488262e-08, + "loss": 0.6924, + "step": 30406 + }, + { + "epoch": 1.11, + "learning_rate": 8.883070275021056e-08, + "loss": 0.6798, + "step": 30407 + }, + { + "epoch": 1.11, + "learning_rate": 8.88270406855385e-08, + "loss": 0.6909, + "step": 30408 + }, + { + "epoch": 1.11, + "learning_rate": 8.882337862086644e-08, + "loss": 0.7029, + "step": 30409 + }, + { + "epoch": 1.11, + "learning_rate": 8.881971655619438e-08, + "loss": 0.704, + "step": 30410 + }, + { + "epoch": 1.11, + "learning_rate": 8.881605449152232e-08, + "loss": 0.6908, + "step": 30411 + }, + { + "epoch": 1.11, + "learning_rate": 8.881239242685026e-08, + "loss": 0.679, + "step": 30412 + }, + { + "epoch": 1.11, + "learning_rate": 8.88087303621782e-08, + "loss": 0.6917, + "step": 30413 + }, + { + "epoch": 1.11, + "learning_rate": 8.880506829750612e-08, + "loss": 0.6935, + "step": 30414 + }, + { + "epoch": 1.11, + "learning_rate": 8.880140623283406e-08, + "loss": 0.6839, + "step": 30415 + }, + { + "epoch": 1.11, + "learning_rate": 8.8797744168162e-08, + "loss": 0.7063, + "step": 30416 + }, + { + "epoch": 1.11, + "learning_rate": 8.879408210348994e-08, + "loss": 0.6801, + "step": 30417 + }, + { + "epoch": 1.11, + "learning_rate": 8.879042003881788e-08, + "loss": 0.6915, + "step": 30418 + }, + { + "epoch": 1.11, + "learning_rate": 8.878675797414582e-08, + "loss": 0.705, + "step": 30419 + }, + { + "epoch": 1.11, + "learning_rate": 8.878309590947376e-08, + "loss": 0.6982, + "step": 30420 + }, + { + "epoch": 1.11, + "learning_rate": 8.87794338448017e-08, + "loss": 0.6843, + "step": 30421 + }, + { + "epoch": 1.11, + "learning_rate": 8.877577178012964e-08, + "loss": 0.7053, + "step": 30422 + }, + { + "epoch": 1.11, + "learning_rate": 8.877210971545757e-08, + "loss": 0.6885, + "step": 30423 + }, + { + "epoch": 1.11, + "learning_rate": 8.876844765078551e-08, + "loss": 0.6994, + "step": 30424 + }, + { + "epoch": 1.11, + "learning_rate": 8.876478558611345e-08, + "loss": 0.7077, + "step": 30425 + }, + { + "epoch": 1.11, + "learning_rate": 8.876112352144139e-08, + "loss": 0.6859, + "step": 30426 + }, + { + "epoch": 1.11, + "learning_rate": 8.875746145676932e-08, + "loss": 0.6907, + "step": 30427 + }, + { + "epoch": 1.11, + "learning_rate": 8.875379939209726e-08, + "loss": 0.6915, + "step": 30428 + }, + { + "epoch": 1.11, + "learning_rate": 8.87501373274252e-08, + "loss": 0.6913, + "step": 30429 + }, + { + "epoch": 1.11, + "learning_rate": 8.874647526275314e-08, + "loss": 0.6945, + "step": 30430 + }, + { + "epoch": 1.11, + "learning_rate": 8.874281319808108e-08, + "loss": 0.6821, + "step": 30431 + }, + { + "epoch": 1.11, + "learning_rate": 8.8739151133409e-08, + "loss": 0.6966, + "step": 30432 + }, + { + "epoch": 1.11, + "learning_rate": 8.873548906873694e-08, + "loss": 0.6935, + "step": 30433 + }, + { + "epoch": 1.11, + "learning_rate": 8.873182700406488e-08, + "loss": 0.7056, + "step": 30434 + }, + { + "epoch": 1.11, + "learning_rate": 8.872816493939282e-08, + "loss": 0.6934, + "step": 30435 + }, + { + "epoch": 1.11, + "learning_rate": 8.872450287472076e-08, + "loss": 0.6874, + "step": 30436 + }, + { + "epoch": 1.11, + "learning_rate": 8.87208408100487e-08, + "loss": 0.6976, + "step": 30437 + }, + { + "epoch": 1.11, + "learning_rate": 8.871717874537664e-08, + "loss": 0.6928, + "step": 30438 + }, + { + "epoch": 1.11, + "learning_rate": 8.871351668070458e-08, + "loss": 0.701, + "step": 30439 + }, + { + "epoch": 1.11, + "learning_rate": 8.870985461603252e-08, + "loss": 0.6992, + "step": 30440 + }, + { + "epoch": 1.11, + "learning_rate": 8.870619255136045e-08, + "loss": 0.6902, + "step": 30441 + }, + { + "epoch": 1.11, + "learning_rate": 8.870253048668839e-08, + "loss": 0.6942, + "step": 30442 + }, + { + "epoch": 1.11, + "learning_rate": 8.869886842201633e-08, + "loss": 0.7005, + "step": 30443 + }, + { + "epoch": 1.11, + "learning_rate": 8.869520635734427e-08, + "loss": 0.691, + "step": 30444 + }, + { + "epoch": 1.11, + "learning_rate": 8.869154429267221e-08, + "loss": 0.698, + "step": 30445 + }, + { + "epoch": 1.11, + "learning_rate": 8.868788222800015e-08, + "loss": 0.6974, + "step": 30446 + }, + { + "epoch": 1.11, + "learning_rate": 8.868422016332809e-08, + "loss": 0.6977, + "step": 30447 + }, + { + "epoch": 1.11, + "learning_rate": 8.868055809865603e-08, + "loss": 0.7094, + "step": 30448 + }, + { + "epoch": 1.11, + "learning_rate": 8.867689603398395e-08, + "loss": 0.7001, + "step": 30449 + }, + { + "epoch": 1.11, + "learning_rate": 8.86732339693119e-08, + "loss": 0.6996, + "step": 30450 + }, + { + "epoch": 1.11, + "learning_rate": 8.866957190463983e-08, + "loss": 0.6967, + "step": 30451 + }, + { + "epoch": 1.11, + "learning_rate": 8.866590983996776e-08, + "loss": 0.694, + "step": 30452 + }, + { + "epoch": 1.11, + "learning_rate": 8.86622477752957e-08, + "loss": 0.6993, + "step": 30453 + }, + { + "epoch": 1.11, + "learning_rate": 8.865858571062364e-08, + "loss": 0.6849, + "step": 30454 + }, + { + "epoch": 1.11, + "learning_rate": 8.865492364595158e-08, + "loss": 0.6967, + "step": 30455 + }, + { + "epoch": 1.11, + "learning_rate": 8.865126158127952e-08, + "loss": 0.6923, + "step": 30456 + }, + { + "epoch": 1.11, + "learning_rate": 8.864759951660746e-08, + "loss": 0.6806, + "step": 30457 + }, + { + "epoch": 1.11, + "learning_rate": 8.86439374519354e-08, + "loss": 0.6896, + "step": 30458 + }, + { + "epoch": 1.11, + "learning_rate": 8.864027538726333e-08, + "loss": 0.7038, + "step": 30459 + }, + { + "epoch": 1.11, + "learning_rate": 8.863661332259127e-08, + "loss": 0.6763, + "step": 30460 + }, + { + "epoch": 1.11, + "learning_rate": 8.863295125791921e-08, + "loss": 0.6978, + "step": 30461 + }, + { + "epoch": 1.11, + "learning_rate": 8.862928919324715e-08, + "loss": 0.6925, + "step": 30462 + }, + { + "epoch": 1.11, + "learning_rate": 8.862562712857509e-08, + "loss": 0.693, + "step": 30463 + }, + { + "epoch": 1.11, + "learning_rate": 8.862196506390303e-08, + "loss": 0.6822, + "step": 30464 + }, + { + "epoch": 1.11, + "learning_rate": 8.861830299923097e-08, + "loss": 0.6848, + "step": 30465 + }, + { + "epoch": 1.11, + "learning_rate": 8.861464093455891e-08, + "loss": 0.6891, + "step": 30466 + }, + { + "epoch": 1.11, + "learning_rate": 8.861097886988685e-08, + "loss": 0.6915, + "step": 30467 + }, + { + "epoch": 1.11, + "learning_rate": 8.860731680521477e-08, + "loss": 0.7078, + "step": 30468 + }, + { + "epoch": 1.11, + "learning_rate": 8.860365474054271e-08, + "loss": 0.7001, + "step": 30469 + }, + { + "epoch": 1.11, + "learning_rate": 8.859999267587065e-08, + "loss": 0.6954, + "step": 30470 + }, + { + "epoch": 1.11, + "learning_rate": 8.85963306111986e-08, + "loss": 0.6852, + "step": 30471 + }, + { + "epoch": 1.11, + "learning_rate": 8.859266854652653e-08, + "loss": 0.6763, + "step": 30472 + }, + { + "epoch": 1.11, + "learning_rate": 8.858900648185446e-08, + "loss": 0.6923, + "step": 30473 + }, + { + "epoch": 1.11, + "learning_rate": 8.85853444171824e-08, + "loss": 0.6844, + "step": 30474 + }, + { + "epoch": 1.11, + "learning_rate": 8.858168235251034e-08, + "loss": 0.6986, + "step": 30475 + }, + { + "epoch": 1.12, + "learning_rate": 8.857802028783828e-08, + "loss": 0.6864, + "step": 30476 + }, + { + "epoch": 1.12, + "learning_rate": 8.857435822316621e-08, + "loss": 0.7013, + "step": 30477 + }, + { + "epoch": 1.12, + "learning_rate": 8.857069615849415e-08, + "loss": 0.6922, + "step": 30478 + }, + { + "epoch": 1.12, + "learning_rate": 8.856703409382209e-08, + "loss": 0.6926, + "step": 30479 + }, + { + "epoch": 1.12, + "learning_rate": 8.856337202915003e-08, + "loss": 0.6918, + "step": 30480 + }, + { + "epoch": 1.12, + "learning_rate": 8.855970996447797e-08, + "loss": 0.6919, + "step": 30481 + }, + { + "epoch": 1.12, + "learning_rate": 8.85560478998059e-08, + "loss": 0.6927, + "step": 30482 + }, + { + "epoch": 1.12, + "learning_rate": 8.855238583513385e-08, + "loss": 0.6954, + "step": 30483 + }, + { + "epoch": 1.12, + "learning_rate": 8.854872377046179e-08, + "loss": 0.7058, + "step": 30484 + }, + { + "epoch": 1.12, + "learning_rate": 8.854506170578973e-08, + "loss": 0.6909, + "step": 30485 + }, + { + "epoch": 1.12, + "learning_rate": 8.854139964111765e-08, + "loss": 0.6952, + "step": 30486 + }, + { + "epoch": 1.12, + "learning_rate": 8.853773757644559e-08, + "loss": 0.6905, + "step": 30487 + }, + { + "epoch": 1.12, + "learning_rate": 8.853407551177353e-08, + "loss": 0.6966, + "step": 30488 + }, + { + "epoch": 1.12, + "learning_rate": 8.853041344710147e-08, + "loss": 0.6939, + "step": 30489 + }, + { + "epoch": 1.12, + "learning_rate": 8.852675138242941e-08, + "loss": 0.7029, + "step": 30490 + }, + { + "epoch": 1.12, + "learning_rate": 8.852308931775735e-08, + "loss": 0.6786, + "step": 30491 + }, + { + "epoch": 1.12, + "learning_rate": 8.851942725308529e-08, + "loss": 0.6992, + "step": 30492 + }, + { + "epoch": 1.12, + "learning_rate": 8.851576518841323e-08, + "loss": 0.6964, + "step": 30493 + }, + { + "epoch": 1.12, + "learning_rate": 8.851210312374117e-08, + "loss": 0.6977, + "step": 30494 + }, + { + "epoch": 1.12, + "learning_rate": 8.85084410590691e-08, + "loss": 0.6928, + "step": 30495 + }, + { + "epoch": 1.12, + "learning_rate": 8.850477899439704e-08, + "loss": 0.6904, + "step": 30496 + }, + { + "epoch": 1.12, + "learning_rate": 8.850111692972498e-08, + "loss": 0.6861, + "step": 30497 + }, + { + "epoch": 1.12, + "learning_rate": 8.84974548650529e-08, + "loss": 0.6994, + "step": 30498 + }, + { + "epoch": 1.12, + "learning_rate": 8.849379280038085e-08, + "loss": 0.684, + "step": 30499 + }, + { + "epoch": 1.12, + "learning_rate": 8.849013073570879e-08, + "loss": 0.702, + "step": 30500 + }, + { + "epoch": 1.12, + "learning_rate": 8.848646867103673e-08, + "loss": 0.7017, + "step": 30501 + }, + { + "epoch": 1.12, + "learning_rate": 8.848280660636467e-08, + "loss": 0.6971, + "step": 30502 + }, + { + "epoch": 1.12, + "learning_rate": 8.84791445416926e-08, + "loss": 0.6977, + "step": 30503 + }, + { + "epoch": 1.12, + "learning_rate": 8.847548247702053e-08, + "loss": 0.6903, + "step": 30504 + }, + { + "epoch": 1.12, + "learning_rate": 8.847182041234847e-08, + "loss": 0.6992, + "step": 30505 + }, + { + "epoch": 1.12, + "learning_rate": 8.846815834767641e-08, + "loss": 0.709, + "step": 30506 + }, + { + "epoch": 1.12, + "learning_rate": 8.846449628300435e-08, + "loss": 0.6925, + "step": 30507 + }, + { + "epoch": 1.12, + "learning_rate": 8.846083421833229e-08, + "loss": 0.6923, + "step": 30508 + }, + { + "epoch": 1.12, + "learning_rate": 8.845717215366023e-08, + "loss": 0.6957, + "step": 30509 + }, + { + "epoch": 1.12, + "learning_rate": 8.845351008898817e-08, + "loss": 0.6952, + "step": 30510 + }, + { + "epoch": 1.12, + "learning_rate": 8.844984802431611e-08, + "loss": 0.6914, + "step": 30511 + }, + { + "epoch": 1.12, + "learning_rate": 8.844618595964405e-08, + "loss": 0.6935, + "step": 30512 + }, + { + "epoch": 1.12, + "learning_rate": 8.844252389497198e-08, + "loss": 0.6844, + "step": 30513 + }, + { + "epoch": 1.12, + "learning_rate": 8.843886183029992e-08, + "loss": 0.7063, + "step": 30514 + }, + { + "epoch": 1.12, + "learning_rate": 8.843519976562786e-08, + "loss": 0.6914, + "step": 30515 + }, + { + "epoch": 1.12, + "learning_rate": 8.84315377009558e-08, + "loss": 0.6961, + "step": 30516 + }, + { + "epoch": 1.12, + "learning_rate": 8.842787563628374e-08, + "loss": 0.6835, + "step": 30517 + }, + { + "epoch": 1.12, + "learning_rate": 8.842421357161168e-08, + "loss": 0.6885, + "step": 30518 + }, + { + "epoch": 1.12, + "learning_rate": 8.842055150693962e-08, + "loss": 0.6873, + "step": 30519 + }, + { + "epoch": 1.12, + "learning_rate": 8.841688944226754e-08, + "loss": 0.6859, + "step": 30520 + }, + { + "epoch": 1.12, + "learning_rate": 8.841322737759548e-08, + "loss": 0.6959, + "step": 30521 + }, + { + "epoch": 1.12, + "learning_rate": 8.840956531292342e-08, + "loss": 0.6713, + "step": 30522 + }, + { + "epoch": 1.12, + "learning_rate": 8.840590324825135e-08, + "loss": 0.6895, + "step": 30523 + }, + { + "epoch": 1.12, + "learning_rate": 8.840224118357929e-08, + "loss": 0.6947, + "step": 30524 + }, + { + "epoch": 1.12, + "learning_rate": 8.839857911890723e-08, + "loss": 0.6915, + "step": 30525 + }, + { + "epoch": 1.12, + "learning_rate": 8.839491705423517e-08, + "loss": 0.6959, + "step": 30526 + }, + { + "epoch": 1.12, + "learning_rate": 8.839125498956311e-08, + "loss": 0.7155, + "step": 30527 + }, + { + "epoch": 1.12, + "learning_rate": 8.838759292489105e-08, + "loss": 0.6898, + "step": 30528 + }, + { + "epoch": 1.12, + "learning_rate": 8.838393086021899e-08, + "loss": 0.6952, + "step": 30529 + }, + { + "epoch": 1.12, + "learning_rate": 8.838026879554693e-08, + "loss": 0.6953, + "step": 30530 + }, + { + "epoch": 1.12, + "learning_rate": 8.837660673087486e-08, + "loss": 0.6946, + "step": 30531 + }, + { + "epoch": 1.12, + "learning_rate": 8.83729446662028e-08, + "loss": 0.7054, + "step": 30532 + }, + { + "epoch": 1.12, + "learning_rate": 8.836928260153074e-08, + "loss": 0.6852, + "step": 30533 + }, + { + "epoch": 1.12, + "learning_rate": 8.836562053685868e-08, + "loss": 0.6914, + "step": 30534 + }, + { + "epoch": 1.12, + "learning_rate": 8.836195847218662e-08, + "loss": 0.682, + "step": 30535 + }, + { + "epoch": 1.12, + "learning_rate": 8.835829640751456e-08, + "loss": 0.6987, + "step": 30536 + }, + { + "epoch": 1.12, + "learning_rate": 8.83546343428425e-08, + "loss": 0.693, + "step": 30537 + }, + { + "epoch": 1.12, + "learning_rate": 8.835097227817044e-08, + "loss": 0.6976, + "step": 30538 + }, + { + "epoch": 1.12, + "learning_rate": 8.834731021349838e-08, + "loss": 0.6909, + "step": 30539 + }, + { + "epoch": 1.12, + "learning_rate": 8.83436481488263e-08, + "loss": 0.69, + "step": 30540 + }, + { + "epoch": 1.12, + "learning_rate": 8.833998608415424e-08, + "loss": 0.6907, + "step": 30541 + }, + { + "epoch": 1.12, + "learning_rate": 8.833632401948218e-08, + "loss": 0.7078, + "step": 30542 + }, + { + "epoch": 1.12, + "learning_rate": 8.833266195481012e-08, + "loss": 0.688, + "step": 30543 + }, + { + "epoch": 1.12, + "learning_rate": 8.832899989013805e-08, + "loss": 0.6917, + "step": 30544 + }, + { + "epoch": 1.12, + "learning_rate": 8.832533782546599e-08, + "loss": 0.6834, + "step": 30545 + }, + { + "epoch": 1.12, + "learning_rate": 8.832167576079393e-08, + "loss": 0.7013, + "step": 30546 + }, + { + "epoch": 1.12, + "learning_rate": 8.831801369612187e-08, + "loss": 0.6941, + "step": 30547 + }, + { + "epoch": 1.12, + "learning_rate": 8.831435163144981e-08, + "loss": 0.704, + "step": 30548 + }, + { + "epoch": 1.12, + "learning_rate": 8.831068956677774e-08, + "loss": 0.6795, + "step": 30549 + }, + { + "epoch": 1.12, + "learning_rate": 8.830702750210568e-08, + "loss": 0.6867, + "step": 30550 + }, + { + "epoch": 1.12, + "learning_rate": 8.830336543743362e-08, + "loss": 0.6863, + "step": 30551 + }, + { + "epoch": 1.12, + "learning_rate": 8.829970337276156e-08, + "loss": 0.6956, + "step": 30552 + }, + { + "epoch": 1.12, + "learning_rate": 8.82960413080895e-08, + "loss": 0.6962, + "step": 30553 + }, + { + "epoch": 1.12, + "learning_rate": 8.829237924341744e-08, + "loss": 0.6798, + "step": 30554 + }, + { + "epoch": 1.12, + "learning_rate": 8.828871717874538e-08, + "loss": 0.6863, + "step": 30555 + }, + { + "epoch": 1.12, + "learning_rate": 8.828505511407332e-08, + "loss": 0.6939, + "step": 30556 + }, + { + "epoch": 1.12, + "learning_rate": 8.828139304940126e-08, + "loss": 0.6946, + "step": 30557 + }, + { + "epoch": 1.12, + "learning_rate": 8.827773098472918e-08, + "loss": 0.696, + "step": 30558 + }, + { + "epoch": 1.12, + "learning_rate": 8.827406892005712e-08, + "loss": 0.6893, + "step": 30559 + }, + { + "epoch": 1.12, + "learning_rate": 8.827040685538506e-08, + "loss": 0.6942, + "step": 30560 + }, + { + "epoch": 1.12, + "learning_rate": 8.8266744790713e-08, + "loss": 0.6967, + "step": 30561 + }, + { + "epoch": 1.12, + "learning_rate": 8.826308272604094e-08, + "loss": 0.6873, + "step": 30562 + }, + { + "epoch": 1.12, + "learning_rate": 8.825942066136888e-08, + "loss": 0.6872, + "step": 30563 + }, + { + "epoch": 1.12, + "learning_rate": 8.825575859669682e-08, + "loss": 0.6898, + "step": 30564 + }, + { + "epoch": 1.12, + "learning_rate": 8.825209653202476e-08, + "loss": 0.6882, + "step": 30565 + }, + { + "epoch": 1.12, + "learning_rate": 8.824843446735269e-08, + "loss": 0.6829, + "step": 30566 + }, + { + "epoch": 1.12, + "learning_rate": 8.824477240268063e-08, + "loss": 0.6845, + "step": 30567 + }, + { + "epoch": 1.12, + "learning_rate": 8.824111033800857e-08, + "loss": 0.6984, + "step": 30568 + }, + { + "epoch": 1.12, + "learning_rate": 8.82374482733365e-08, + "loss": 0.6934, + "step": 30569 + }, + { + "epoch": 1.12, + "learning_rate": 8.823378620866443e-08, + "loss": 0.7067, + "step": 30570 + }, + { + "epoch": 1.12, + "learning_rate": 8.823012414399237e-08, + "loss": 0.6955, + "step": 30571 + }, + { + "epoch": 1.12, + "learning_rate": 8.822646207932031e-08, + "loss": 0.6935, + "step": 30572 + }, + { + "epoch": 1.12, + "learning_rate": 8.822280001464825e-08, + "loss": 0.6926, + "step": 30573 + }, + { + "epoch": 1.12, + "learning_rate": 8.82191379499762e-08, + "loss": 0.6981, + "step": 30574 + }, + { + "epoch": 1.12, + "learning_rate": 8.821547588530413e-08, + "loss": 0.698, + "step": 30575 + }, + { + "epoch": 1.12, + "learning_rate": 8.821181382063206e-08, + "loss": 0.6921, + "step": 30576 + }, + { + "epoch": 1.12, + "learning_rate": 8.820815175596e-08, + "loss": 0.6944, + "step": 30577 + }, + { + "epoch": 1.12, + "learning_rate": 8.820448969128794e-08, + "loss": 0.6895, + "step": 30578 + }, + { + "epoch": 1.12, + "learning_rate": 8.820082762661588e-08, + "loss": 0.6826, + "step": 30579 + }, + { + "epoch": 1.12, + "learning_rate": 8.819716556194382e-08, + "loss": 0.6901, + "step": 30580 + }, + { + "epoch": 1.12, + "learning_rate": 8.819350349727176e-08, + "loss": 0.6823, + "step": 30581 + }, + { + "epoch": 1.12, + "learning_rate": 8.81898414325997e-08, + "loss": 0.6913, + "step": 30582 + }, + { + "epoch": 1.12, + "learning_rate": 8.818617936792764e-08, + "loss": 0.6867, + "step": 30583 + }, + { + "epoch": 1.12, + "learning_rate": 8.818251730325558e-08, + "loss": 0.6894, + "step": 30584 + }, + { + "epoch": 1.12, + "learning_rate": 8.817885523858351e-08, + "loss": 0.6904, + "step": 30585 + }, + { + "epoch": 1.12, + "learning_rate": 8.817519317391145e-08, + "loss": 0.69, + "step": 30586 + }, + { + "epoch": 1.12, + "learning_rate": 8.817153110923939e-08, + "loss": 0.695, + "step": 30587 + }, + { + "epoch": 1.12, + "learning_rate": 8.816786904456733e-08, + "loss": 0.6936, + "step": 30588 + }, + { + "epoch": 1.12, + "learning_rate": 8.816420697989527e-08, + "loss": 0.6913, + "step": 30589 + }, + { + "epoch": 1.12, + "learning_rate": 8.816054491522321e-08, + "loss": 0.6833, + "step": 30590 + }, + { + "epoch": 1.12, + "learning_rate": 8.815688285055113e-08, + "loss": 0.6895, + "step": 30591 + }, + { + "epoch": 1.12, + "learning_rate": 8.815322078587907e-08, + "loss": 0.6896, + "step": 30592 + }, + { + "epoch": 1.12, + "learning_rate": 8.814955872120701e-08, + "loss": 0.6871, + "step": 30593 + }, + { + "epoch": 1.12, + "learning_rate": 8.814589665653494e-08, + "loss": 0.6869, + "step": 30594 + }, + { + "epoch": 1.12, + "learning_rate": 8.814223459186288e-08, + "loss": 0.6954, + "step": 30595 + }, + { + "epoch": 1.12, + "learning_rate": 8.813857252719082e-08, + "loss": 0.7028, + "step": 30596 + }, + { + "epoch": 1.12, + "learning_rate": 8.813491046251876e-08, + "loss": 0.7079, + "step": 30597 + }, + { + "epoch": 1.12, + "learning_rate": 8.81312483978467e-08, + "loss": 0.6926, + "step": 30598 + }, + { + "epoch": 1.12, + "learning_rate": 8.812758633317464e-08, + "loss": 0.6908, + "step": 30599 + }, + { + "epoch": 1.12, + "learning_rate": 8.812392426850258e-08, + "loss": 0.6945, + "step": 30600 + }, + { + "epoch": 1.12, + "learning_rate": 8.812026220383052e-08, + "loss": 0.6936, + "step": 30601 + }, + { + "epoch": 1.12, + "learning_rate": 8.811660013915846e-08, + "loss": 0.6902, + "step": 30602 + }, + { + "epoch": 1.12, + "learning_rate": 8.811293807448639e-08, + "loss": 0.6926, + "step": 30603 + }, + { + "epoch": 1.12, + "learning_rate": 8.810927600981433e-08, + "loss": 0.6912, + "step": 30604 + }, + { + "epoch": 1.12, + "learning_rate": 8.810561394514227e-08, + "loss": 0.6873, + "step": 30605 + }, + { + "epoch": 1.12, + "learning_rate": 8.81019518804702e-08, + "loss": 0.6897, + "step": 30606 + }, + { + "epoch": 1.12, + "learning_rate": 8.809828981579815e-08, + "loss": 0.6953, + "step": 30607 + }, + { + "epoch": 1.12, + "learning_rate": 8.809462775112609e-08, + "loss": 0.6948, + "step": 30608 + }, + { + "epoch": 1.12, + "learning_rate": 8.809096568645403e-08, + "loss": 0.701, + "step": 30609 + }, + { + "epoch": 1.12, + "learning_rate": 8.808730362178197e-08, + "loss": 0.7014, + "step": 30610 + }, + { + "epoch": 1.12, + "learning_rate": 8.80836415571099e-08, + "loss": 0.6997, + "step": 30611 + }, + { + "epoch": 1.12, + "learning_rate": 8.807997949243783e-08, + "loss": 0.6982, + "step": 30612 + }, + { + "epoch": 1.12, + "learning_rate": 8.807631742776577e-08, + "loss": 0.6921, + "step": 30613 + }, + { + "epoch": 1.12, + "learning_rate": 8.807265536309371e-08, + "loss": 0.6951, + "step": 30614 + }, + { + "epoch": 1.12, + "learning_rate": 8.806899329842164e-08, + "loss": 0.6872, + "step": 30615 + }, + { + "epoch": 1.12, + "learning_rate": 8.806533123374958e-08, + "loss": 0.6808, + "step": 30616 + }, + { + "epoch": 1.12, + "learning_rate": 8.806166916907752e-08, + "loss": 0.6972, + "step": 30617 + }, + { + "epoch": 1.12, + "learning_rate": 8.805800710440546e-08, + "loss": 0.695, + "step": 30618 + }, + { + "epoch": 1.12, + "learning_rate": 8.80543450397334e-08, + "loss": 0.6798, + "step": 30619 + }, + { + "epoch": 1.12, + "learning_rate": 8.805068297506134e-08, + "loss": 0.7035, + "step": 30620 + }, + { + "epoch": 1.12, + "learning_rate": 8.804702091038927e-08, + "loss": 0.6889, + "step": 30621 + }, + { + "epoch": 1.12, + "learning_rate": 8.80433588457172e-08, + "loss": 0.6811, + "step": 30622 + }, + { + "epoch": 1.12, + "learning_rate": 8.803969678104515e-08, + "loss": 0.7118, + "step": 30623 + }, + { + "epoch": 1.12, + "learning_rate": 8.803603471637309e-08, + "loss": 0.6876, + "step": 30624 + }, + { + "epoch": 1.12, + "learning_rate": 8.803237265170103e-08, + "loss": 0.7038, + "step": 30625 + }, + { + "epoch": 1.12, + "learning_rate": 8.802871058702896e-08, + "loss": 0.6894, + "step": 30626 + }, + { + "epoch": 1.12, + "learning_rate": 8.80250485223569e-08, + "loss": 0.6991, + "step": 30627 + }, + { + "epoch": 1.12, + "learning_rate": 8.802138645768484e-08, + "loss": 0.6878, + "step": 30628 + }, + { + "epoch": 1.12, + "learning_rate": 8.801772439301277e-08, + "loss": 0.6905, + "step": 30629 + }, + { + "epoch": 1.12, + "learning_rate": 8.801406232834071e-08, + "loss": 0.704, + "step": 30630 + }, + { + "epoch": 1.12, + "learning_rate": 8.801040026366865e-08, + "loss": 0.6835, + "step": 30631 + }, + { + "epoch": 1.12, + "learning_rate": 8.800673819899659e-08, + "loss": 0.6837, + "step": 30632 + }, + { + "epoch": 1.12, + "learning_rate": 8.800307613432453e-08, + "loss": 0.6918, + "step": 30633 + }, + { + "epoch": 1.12, + "learning_rate": 8.799941406965247e-08, + "loss": 0.6919, + "step": 30634 + }, + { + "epoch": 1.12, + "learning_rate": 8.799575200498041e-08, + "loss": 0.6932, + "step": 30635 + }, + { + "epoch": 1.12, + "learning_rate": 8.799208994030835e-08, + "loss": 0.6918, + "step": 30636 + }, + { + "epoch": 1.12, + "learning_rate": 8.798842787563628e-08, + "loss": 0.6851, + "step": 30637 + }, + { + "epoch": 1.12, + "learning_rate": 8.798476581096422e-08, + "loss": 0.6986, + "step": 30638 + }, + { + "epoch": 1.12, + "learning_rate": 8.798110374629216e-08, + "loss": 0.7056, + "step": 30639 + }, + { + "epoch": 1.12, + "learning_rate": 8.797744168162008e-08, + "loss": 0.7098, + "step": 30640 + }, + { + "epoch": 1.12, + "learning_rate": 8.797377961694802e-08, + "loss": 0.6858, + "step": 30641 + }, + { + "epoch": 1.12, + "learning_rate": 8.797011755227596e-08, + "loss": 0.6938, + "step": 30642 + }, + { + "epoch": 1.12, + "learning_rate": 8.79664554876039e-08, + "loss": 0.6889, + "step": 30643 + }, + { + "epoch": 1.12, + "learning_rate": 8.796279342293184e-08, + "loss": 0.6972, + "step": 30644 + }, + { + "epoch": 1.12, + "learning_rate": 8.795913135825978e-08, + "loss": 0.6912, + "step": 30645 + }, + { + "epoch": 1.12, + "learning_rate": 8.795546929358772e-08, + "loss": 0.6815, + "step": 30646 + }, + { + "epoch": 1.12, + "learning_rate": 8.795180722891565e-08, + "loss": 0.7069, + "step": 30647 + }, + { + "epoch": 1.12, + "learning_rate": 8.794814516424359e-08, + "loss": 0.6865, + "step": 30648 + }, + { + "epoch": 1.12, + "learning_rate": 8.794448309957153e-08, + "loss": 0.6852, + "step": 30649 + }, + { + "epoch": 1.12, + "learning_rate": 8.794082103489947e-08, + "loss": 0.7073, + "step": 30650 + }, + { + "epoch": 1.12, + "learning_rate": 8.793715897022741e-08, + "loss": 0.6961, + "step": 30651 + }, + { + "epoch": 1.12, + "learning_rate": 8.793349690555535e-08, + "loss": 0.6891, + "step": 30652 + }, + { + "epoch": 1.12, + "learning_rate": 8.792983484088329e-08, + "loss": 0.705, + "step": 30653 + }, + { + "epoch": 1.12, + "learning_rate": 8.792617277621123e-08, + "loss": 0.6976, + "step": 30654 + }, + { + "epoch": 1.12, + "learning_rate": 8.792251071153917e-08, + "loss": 0.6973, + "step": 30655 + }, + { + "epoch": 1.12, + "learning_rate": 8.79188486468671e-08, + "loss": 0.7147, + "step": 30656 + }, + { + "epoch": 1.12, + "learning_rate": 8.791518658219504e-08, + "loss": 0.6926, + "step": 30657 + }, + { + "epoch": 1.12, + "learning_rate": 8.791152451752298e-08, + "loss": 0.699, + "step": 30658 + }, + { + "epoch": 1.12, + "learning_rate": 8.790786245285092e-08, + "loss": 0.6988, + "step": 30659 + }, + { + "epoch": 1.12, + "learning_rate": 8.790420038817886e-08, + "loss": 0.696, + "step": 30660 + }, + { + "epoch": 1.12, + "learning_rate": 8.79005383235068e-08, + "loss": 0.6982, + "step": 30661 + }, + { + "epoch": 1.12, + "learning_rate": 8.789687625883472e-08, + "loss": 0.6865, + "step": 30662 + }, + { + "epoch": 1.12, + "learning_rate": 8.789321419416266e-08, + "loss": 0.714, + "step": 30663 + }, + { + "epoch": 1.12, + "learning_rate": 8.78895521294906e-08, + "loss": 0.6981, + "step": 30664 + }, + { + "epoch": 1.12, + "learning_rate": 8.788589006481853e-08, + "loss": 0.6763, + "step": 30665 + }, + { + "epoch": 1.12, + "learning_rate": 8.788222800014647e-08, + "loss": 0.6911, + "step": 30666 + }, + { + "epoch": 1.12, + "learning_rate": 8.787856593547441e-08, + "loss": 0.6879, + "step": 30667 + }, + { + "epoch": 1.12, + "learning_rate": 8.787490387080235e-08, + "loss": 0.6894, + "step": 30668 + }, + { + "epoch": 1.12, + "learning_rate": 8.787124180613029e-08, + "loss": 0.6914, + "step": 30669 + }, + { + "epoch": 1.12, + "learning_rate": 8.786757974145823e-08, + "loss": 0.6967, + "step": 30670 + }, + { + "epoch": 1.12, + "learning_rate": 8.786391767678617e-08, + "loss": 0.6939, + "step": 30671 + }, + { + "epoch": 1.12, + "learning_rate": 8.786025561211411e-08, + "loss": 0.6886, + "step": 30672 + }, + { + "epoch": 1.12, + "learning_rate": 8.785659354744205e-08, + "loss": 0.6893, + "step": 30673 + }, + { + "epoch": 1.12, + "learning_rate": 8.785293148276998e-08, + "loss": 0.6858, + "step": 30674 + }, + { + "epoch": 1.12, + "learning_rate": 8.784926941809792e-08, + "loss": 0.69, + "step": 30675 + }, + { + "epoch": 1.12, + "learning_rate": 8.784560735342586e-08, + "loss": 0.6836, + "step": 30676 + }, + { + "epoch": 1.12, + "learning_rate": 8.78419452887538e-08, + "loss": 0.7024, + "step": 30677 + }, + { + "epoch": 1.12, + "learning_rate": 8.783828322408174e-08, + "loss": 0.6993, + "step": 30678 + }, + { + "epoch": 1.12, + "learning_rate": 8.783462115940968e-08, + "loss": 0.6858, + "step": 30679 + }, + { + "epoch": 1.12, + "learning_rate": 8.783095909473762e-08, + "loss": 0.6875, + "step": 30680 + }, + { + "epoch": 1.12, + "learning_rate": 8.782729703006556e-08, + "loss": 0.6863, + "step": 30681 + }, + { + "epoch": 1.12, + "learning_rate": 8.78236349653935e-08, + "loss": 0.678, + "step": 30682 + }, + { + "epoch": 1.12, + "learning_rate": 8.781997290072142e-08, + "loss": 0.6987, + "step": 30683 + }, + { + "epoch": 1.12, + "learning_rate": 8.781631083604936e-08, + "loss": 0.693, + "step": 30684 + }, + { + "epoch": 1.12, + "learning_rate": 8.78126487713773e-08, + "loss": 0.6847, + "step": 30685 + }, + { + "epoch": 1.12, + "learning_rate": 8.780898670670523e-08, + "loss": 0.6999, + "step": 30686 + }, + { + "epoch": 1.12, + "learning_rate": 8.780532464203317e-08, + "loss": 0.684, + "step": 30687 + }, + { + "epoch": 1.12, + "learning_rate": 8.780166257736111e-08, + "loss": 0.7002, + "step": 30688 + }, + { + "epoch": 1.12, + "learning_rate": 8.779800051268905e-08, + "loss": 0.6915, + "step": 30689 + }, + { + "epoch": 1.12, + "learning_rate": 8.779433844801699e-08, + "loss": 0.6947, + "step": 30690 + }, + { + "epoch": 1.12, + "learning_rate": 8.779067638334493e-08, + "loss": 0.6879, + "step": 30691 + }, + { + "epoch": 1.12, + "learning_rate": 8.778701431867285e-08, + "loss": 0.6805, + "step": 30692 + }, + { + "epoch": 1.12, + "learning_rate": 8.77833522540008e-08, + "loss": 0.6982, + "step": 30693 + }, + { + "epoch": 1.12, + "learning_rate": 8.777969018932873e-08, + "loss": 0.6975, + "step": 30694 + }, + { + "epoch": 1.12, + "learning_rate": 8.777602812465667e-08, + "loss": 0.6885, + "step": 30695 + }, + { + "epoch": 1.12, + "learning_rate": 8.777236605998461e-08, + "loss": 0.6855, + "step": 30696 + }, + { + "epoch": 1.12, + "learning_rate": 8.776870399531255e-08, + "loss": 0.676, + "step": 30697 + }, + { + "epoch": 1.12, + "learning_rate": 8.77650419306405e-08, + "loss": 0.7012, + "step": 30698 + }, + { + "epoch": 1.12, + "learning_rate": 8.776137986596843e-08, + "loss": 0.6945, + "step": 30699 + }, + { + "epoch": 1.12, + "learning_rate": 8.775771780129637e-08, + "loss": 0.6932, + "step": 30700 + }, + { + "epoch": 1.12, + "learning_rate": 8.77540557366243e-08, + "loss": 0.697, + "step": 30701 + }, + { + "epoch": 1.12, + "learning_rate": 8.775039367195224e-08, + "loss": 0.6893, + "step": 30702 + }, + { + "epoch": 1.12, + "learning_rate": 8.774673160728018e-08, + "loss": 0.6966, + "step": 30703 + }, + { + "epoch": 1.12, + "learning_rate": 8.774306954260812e-08, + "loss": 0.682, + "step": 30704 + }, + { + "epoch": 1.12, + "learning_rate": 8.773940747793606e-08, + "loss": 0.7069, + "step": 30705 + }, + { + "epoch": 1.12, + "learning_rate": 8.7735745413264e-08, + "loss": 0.6898, + "step": 30706 + }, + { + "epoch": 1.12, + "learning_rate": 8.773208334859194e-08, + "loss": 0.6934, + "step": 30707 + }, + { + "epoch": 1.12, + "learning_rate": 8.772842128391987e-08, + "loss": 0.6951, + "step": 30708 + }, + { + "epoch": 1.12, + "learning_rate": 8.772475921924781e-08, + "loss": 0.6834, + "step": 30709 + }, + { + "epoch": 1.12, + "learning_rate": 8.772109715457575e-08, + "loss": 0.6895, + "step": 30710 + }, + { + "epoch": 1.12, + "learning_rate": 8.771743508990367e-08, + "loss": 0.6848, + "step": 30711 + }, + { + "epoch": 1.12, + "learning_rate": 8.771377302523161e-08, + "loss": 0.6867, + "step": 30712 + }, + { + "epoch": 1.12, + "learning_rate": 8.771011096055955e-08, + "loss": 0.6973, + "step": 30713 + }, + { + "epoch": 1.12, + "learning_rate": 8.77064488958875e-08, + "loss": 0.6913, + "step": 30714 + }, + { + "epoch": 1.12, + "learning_rate": 8.770278683121543e-08, + "loss": 0.6924, + "step": 30715 + }, + { + "epoch": 1.12, + "learning_rate": 8.769912476654337e-08, + "loss": 0.6873, + "step": 30716 + }, + { + "epoch": 1.12, + "learning_rate": 8.769546270187131e-08, + "loss": 0.7011, + "step": 30717 + }, + { + "epoch": 1.12, + "learning_rate": 8.769180063719925e-08, + "loss": 0.7054, + "step": 30718 + }, + { + "epoch": 1.12, + "learning_rate": 8.768813857252718e-08, + "loss": 0.6906, + "step": 30719 + }, + { + "epoch": 1.12, + "learning_rate": 8.768447650785512e-08, + "loss": 0.711, + "step": 30720 + }, + { + "epoch": 1.12, + "learning_rate": 8.768081444318306e-08, + "loss": 0.6907, + "step": 30721 + }, + { + "epoch": 1.12, + "learning_rate": 8.7677152378511e-08, + "loss": 0.706, + "step": 30722 + }, + { + "epoch": 1.12, + "learning_rate": 8.767349031383894e-08, + "loss": 0.6912, + "step": 30723 + }, + { + "epoch": 1.12, + "learning_rate": 8.766982824916688e-08, + "loss": 0.6873, + "step": 30724 + }, + { + "epoch": 1.12, + "learning_rate": 8.766616618449482e-08, + "loss": 0.685, + "step": 30725 + }, + { + "epoch": 1.12, + "learning_rate": 8.766250411982276e-08, + "loss": 0.6831, + "step": 30726 + }, + { + "epoch": 1.12, + "learning_rate": 8.76588420551507e-08, + "loss": 0.7035, + "step": 30727 + }, + { + "epoch": 1.12, + "learning_rate": 8.765517999047863e-08, + "loss": 0.6958, + "step": 30728 + }, + { + "epoch": 1.12, + "learning_rate": 8.765151792580657e-08, + "loss": 0.6929, + "step": 30729 + }, + { + "epoch": 1.12, + "learning_rate": 8.76478558611345e-08, + "loss": 0.7148, + "step": 30730 + }, + { + "epoch": 1.12, + "learning_rate": 8.764419379646245e-08, + "loss": 0.6913, + "step": 30731 + }, + { + "epoch": 1.12, + "learning_rate": 8.764053173179039e-08, + "loss": 0.701, + "step": 30732 + }, + { + "epoch": 1.12, + "learning_rate": 8.763686966711831e-08, + "loss": 0.6959, + "step": 30733 + }, + { + "epoch": 1.12, + "learning_rate": 8.763320760244625e-08, + "loss": 0.7041, + "step": 30734 + }, + { + "epoch": 1.12, + "learning_rate": 8.762954553777419e-08, + "loss": 0.6847, + "step": 30735 + }, + { + "epoch": 1.12, + "learning_rate": 8.762588347310213e-08, + "loss": 0.686, + "step": 30736 + }, + { + "epoch": 1.12, + "learning_rate": 8.762222140843006e-08, + "loss": 0.6972, + "step": 30737 + }, + { + "epoch": 1.12, + "learning_rate": 8.7618559343758e-08, + "loss": 0.6967, + "step": 30738 + }, + { + "epoch": 1.12, + "learning_rate": 8.761489727908594e-08, + "loss": 0.6925, + "step": 30739 + }, + { + "epoch": 1.12, + "learning_rate": 8.761123521441388e-08, + "loss": 0.7017, + "step": 30740 + }, + { + "epoch": 1.12, + "learning_rate": 8.760757314974182e-08, + "loss": 0.6954, + "step": 30741 + }, + { + "epoch": 1.12, + "learning_rate": 8.760391108506976e-08, + "loss": 0.6907, + "step": 30742 + }, + { + "epoch": 1.12, + "learning_rate": 8.76002490203977e-08, + "loss": 0.7007, + "step": 30743 + }, + { + "epoch": 1.12, + "learning_rate": 8.759658695572564e-08, + "loss": 0.7, + "step": 30744 + }, + { + "epoch": 1.12, + "learning_rate": 8.759292489105358e-08, + "loss": 0.6958, + "step": 30745 + }, + { + "epoch": 1.12, + "learning_rate": 8.75892628263815e-08, + "loss": 0.7006, + "step": 30746 + }, + { + "epoch": 1.12, + "learning_rate": 8.758560076170944e-08, + "loss": 0.6829, + "step": 30747 + }, + { + "epoch": 1.12, + "learning_rate": 8.758193869703738e-08, + "loss": 0.7014, + "step": 30748 + }, + { + "epoch": 1.13, + "learning_rate": 8.757827663236532e-08, + "loss": 0.7027, + "step": 30749 + }, + { + "epoch": 1.13, + "learning_rate": 8.757461456769326e-08, + "loss": 0.6943, + "step": 30750 + }, + { + "epoch": 1.13, + "learning_rate": 8.75709525030212e-08, + "loss": 0.697, + "step": 30751 + }, + { + "epoch": 1.13, + "learning_rate": 8.756729043834914e-08, + "loss": 0.6888, + "step": 30752 + }, + { + "epoch": 1.13, + "learning_rate": 8.756362837367708e-08, + "loss": 0.7062, + "step": 30753 + }, + { + "epoch": 1.13, + "learning_rate": 8.755996630900502e-08, + "loss": 0.6955, + "step": 30754 + }, + { + "epoch": 1.13, + "learning_rate": 8.755630424433295e-08, + "loss": 0.69, + "step": 30755 + }, + { + "epoch": 1.13, + "learning_rate": 8.755264217966089e-08, + "loss": 0.7008, + "step": 30756 + }, + { + "epoch": 1.13, + "learning_rate": 8.754898011498882e-08, + "loss": 0.7064, + "step": 30757 + }, + { + "epoch": 1.13, + "learning_rate": 8.754531805031676e-08, + "loss": 0.6971, + "step": 30758 + }, + { + "epoch": 1.13, + "learning_rate": 8.75416559856447e-08, + "loss": 0.6857, + "step": 30759 + }, + { + "epoch": 1.13, + "learning_rate": 8.753799392097264e-08, + "loss": 0.6891, + "step": 30760 + }, + { + "epoch": 1.13, + "learning_rate": 8.753433185630058e-08, + "loss": 0.6968, + "step": 30761 + }, + { + "epoch": 1.13, + "learning_rate": 8.753066979162852e-08, + "loss": 0.687, + "step": 30762 + }, + { + "epoch": 1.13, + "learning_rate": 8.752700772695646e-08, + "loss": 0.6928, + "step": 30763 + }, + { + "epoch": 1.13, + "learning_rate": 8.752334566228438e-08, + "loss": 0.7044, + "step": 30764 + }, + { + "epoch": 1.13, + "learning_rate": 8.751968359761232e-08, + "loss": 0.7115, + "step": 30765 + }, + { + "epoch": 1.13, + "learning_rate": 8.751602153294026e-08, + "loss": 0.7034, + "step": 30766 + }, + { + "epoch": 1.13, + "learning_rate": 8.75123594682682e-08, + "loss": 0.6934, + "step": 30767 + }, + { + "epoch": 1.13, + "learning_rate": 8.750869740359614e-08, + "loss": 0.7024, + "step": 30768 + }, + { + "epoch": 1.13, + "learning_rate": 8.750503533892408e-08, + "loss": 0.7, + "step": 30769 + }, + { + "epoch": 1.13, + "learning_rate": 8.750137327425202e-08, + "loss": 0.7013, + "step": 30770 + }, + { + "epoch": 1.13, + "learning_rate": 8.749771120957996e-08, + "loss": 0.7088, + "step": 30771 + }, + { + "epoch": 1.13, + "learning_rate": 8.74940491449079e-08, + "loss": 0.6959, + "step": 30772 + }, + { + "epoch": 1.13, + "learning_rate": 8.749038708023583e-08, + "loss": 0.6966, + "step": 30773 + }, + { + "epoch": 1.13, + "learning_rate": 8.748672501556377e-08, + "loss": 0.6983, + "step": 30774 + }, + { + "epoch": 1.13, + "learning_rate": 8.748306295089171e-08, + "loss": 0.6813, + "step": 30775 + }, + { + "epoch": 1.13, + "learning_rate": 8.747940088621965e-08, + "loss": 0.6867, + "step": 30776 + }, + { + "epoch": 1.13, + "learning_rate": 8.747573882154759e-08, + "loss": 0.7006, + "step": 30777 + }, + { + "epoch": 1.13, + "learning_rate": 8.747207675687553e-08, + "loss": 0.6988, + "step": 30778 + }, + { + "epoch": 1.13, + "learning_rate": 8.746841469220346e-08, + "loss": 0.6903, + "step": 30779 + }, + { + "epoch": 1.13, + "learning_rate": 8.74647526275314e-08, + "loss": 0.6936, + "step": 30780 + }, + { + "epoch": 1.13, + "learning_rate": 8.746109056285934e-08, + "loss": 0.7016, + "step": 30781 + }, + { + "epoch": 1.13, + "learning_rate": 8.745742849818726e-08, + "loss": 0.6958, + "step": 30782 + }, + { + "epoch": 1.13, + "learning_rate": 8.74537664335152e-08, + "loss": 0.6907, + "step": 30783 + }, + { + "epoch": 1.13, + "learning_rate": 8.745010436884314e-08, + "loss": 0.6904, + "step": 30784 + }, + { + "epoch": 1.13, + "learning_rate": 8.744644230417108e-08, + "loss": 0.6861, + "step": 30785 + }, + { + "epoch": 1.13, + "learning_rate": 8.744278023949902e-08, + "loss": 0.7001, + "step": 30786 + }, + { + "epoch": 1.13, + "learning_rate": 8.743911817482696e-08, + "loss": 0.692, + "step": 30787 + }, + { + "epoch": 1.13, + "learning_rate": 8.74354561101549e-08, + "loss": 0.6852, + "step": 30788 + }, + { + "epoch": 1.13, + "learning_rate": 8.743179404548284e-08, + "loss": 0.6924, + "step": 30789 + }, + { + "epoch": 1.13, + "learning_rate": 8.742813198081078e-08, + "loss": 0.7086, + "step": 30790 + }, + { + "epoch": 1.13, + "learning_rate": 8.742446991613871e-08, + "loss": 0.6882, + "step": 30791 + }, + { + "epoch": 1.13, + "learning_rate": 8.742080785146665e-08, + "loss": 0.7005, + "step": 30792 + }, + { + "epoch": 1.13, + "learning_rate": 8.741714578679459e-08, + "loss": 0.6925, + "step": 30793 + }, + { + "epoch": 1.13, + "learning_rate": 8.741348372212253e-08, + "loss": 0.6871, + "step": 30794 + }, + { + "epoch": 1.13, + "learning_rate": 8.740982165745047e-08, + "loss": 0.7008, + "step": 30795 + }, + { + "epoch": 1.13, + "learning_rate": 8.740615959277841e-08, + "loss": 0.6931, + "step": 30796 + }, + { + "epoch": 1.13, + "learning_rate": 8.740249752810635e-08, + "loss": 0.6885, + "step": 30797 + }, + { + "epoch": 1.13, + "learning_rate": 8.739883546343429e-08, + "loss": 0.6899, + "step": 30798 + }, + { + "epoch": 1.13, + "learning_rate": 8.739517339876223e-08, + "loss": 0.6887, + "step": 30799 + }, + { + "epoch": 1.13, + "learning_rate": 8.739151133409016e-08, + "loss": 0.6763, + "step": 30800 + }, + { + "epoch": 1.13, + "learning_rate": 8.73878492694181e-08, + "loss": 0.6965, + "step": 30801 + }, + { + "epoch": 1.13, + "learning_rate": 8.738418720474604e-08, + "loss": 0.6819, + "step": 30802 + }, + { + "epoch": 1.13, + "learning_rate": 8.738052514007398e-08, + "loss": 0.6954, + "step": 30803 + }, + { + "epoch": 1.13, + "learning_rate": 8.73768630754019e-08, + "loss": 0.7026, + "step": 30804 + }, + { + "epoch": 1.13, + "learning_rate": 8.737320101072984e-08, + "loss": 0.6922, + "step": 30805 + }, + { + "epoch": 1.13, + "learning_rate": 8.736953894605778e-08, + "loss": 0.6847, + "step": 30806 + }, + { + "epoch": 1.13, + "learning_rate": 8.736587688138572e-08, + "loss": 0.6885, + "step": 30807 + }, + { + "epoch": 1.13, + "learning_rate": 8.736221481671366e-08, + "loss": 0.6849, + "step": 30808 + }, + { + "epoch": 1.13, + "learning_rate": 8.735855275204159e-08, + "loss": 0.6927, + "step": 30809 + }, + { + "epoch": 1.13, + "learning_rate": 8.735489068736953e-08, + "loss": 0.6944, + "step": 30810 + }, + { + "epoch": 1.13, + "learning_rate": 8.735122862269747e-08, + "loss": 0.6772, + "step": 30811 + }, + { + "epoch": 1.13, + "learning_rate": 8.734756655802541e-08, + "loss": 0.6849, + "step": 30812 + }, + { + "epoch": 1.13, + "learning_rate": 8.734390449335335e-08, + "loss": 0.6956, + "step": 30813 + }, + { + "epoch": 1.13, + "learning_rate": 8.734024242868129e-08, + "loss": 0.6852, + "step": 30814 + }, + { + "epoch": 1.13, + "learning_rate": 8.733658036400923e-08, + "loss": 0.6803, + "step": 30815 + }, + { + "epoch": 1.13, + "learning_rate": 8.733291829933717e-08, + "loss": 0.7019, + "step": 30816 + }, + { + "epoch": 1.13, + "learning_rate": 8.732925623466511e-08, + "loss": 0.7026, + "step": 30817 + }, + { + "epoch": 1.13, + "learning_rate": 8.732559416999303e-08, + "loss": 0.6933, + "step": 30818 + }, + { + "epoch": 1.13, + "learning_rate": 8.732193210532097e-08, + "loss": 0.698, + "step": 30819 + }, + { + "epoch": 1.13, + "learning_rate": 8.731827004064891e-08, + "loss": 0.6932, + "step": 30820 + }, + { + "epoch": 1.13, + "learning_rate": 8.731460797597685e-08, + "loss": 0.675, + "step": 30821 + }, + { + "epoch": 1.13, + "learning_rate": 8.73109459113048e-08, + "loss": 0.6985, + "step": 30822 + }, + { + "epoch": 1.13, + "learning_rate": 8.730728384663273e-08, + "loss": 0.7051, + "step": 30823 + }, + { + "epoch": 1.13, + "learning_rate": 8.730362178196067e-08, + "loss": 0.6994, + "step": 30824 + }, + { + "epoch": 1.13, + "learning_rate": 8.72999597172886e-08, + "loss": 0.6919, + "step": 30825 + }, + { + "epoch": 1.13, + "learning_rate": 8.729629765261654e-08, + "loss": 0.6846, + "step": 30826 + }, + { + "epoch": 1.13, + "learning_rate": 8.729263558794448e-08, + "loss": 0.6948, + "step": 30827 + }, + { + "epoch": 1.13, + "learning_rate": 8.728897352327241e-08, + "loss": 0.6858, + "step": 30828 + }, + { + "epoch": 1.13, + "learning_rate": 8.728531145860035e-08, + "loss": 0.6955, + "step": 30829 + }, + { + "epoch": 1.13, + "learning_rate": 8.728164939392829e-08, + "loss": 0.6816, + "step": 30830 + }, + { + "epoch": 1.13, + "learning_rate": 8.727798732925623e-08, + "loss": 0.6835, + "step": 30831 + }, + { + "epoch": 1.13, + "learning_rate": 8.727432526458417e-08, + "loss": 0.6924, + "step": 30832 + }, + { + "epoch": 1.13, + "learning_rate": 8.727066319991211e-08, + "loss": 0.6972, + "step": 30833 + }, + { + "epoch": 1.13, + "learning_rate": 8.726700113524005e-08, + "loss": 0.6968, + "step": 30834 + }, + { + "epoch": 1.13, + "learning_rate": 8.726333907056799e-08, + "loss": 0.6793, + "step": 30835 + }, + { + "epoch": 1.13, + "learning_rate": 8.725967700589591e-08, + "loss": 0.6937, + "step": 30836 + }, + { + "epoch": 1.13, + "learning_rate": 8.725601494122385e-08, + "loss": 0.6903, + "step": 30837 + }, + { + "epoch": 1.13, + "learning_rate": 8.725235287655179e-08, + "loss": 0.6884, + "step": 30838 + }, + { + "epoch": 1.13, + "learning_rate": 8.724869081187973e-08, + "loss": 0.6986, + "step": 30839 + }, + { + "epoch": 1.13, + "learning_rate": 8.724502874720767e-08, + "loss": 0.6936, + "step": 30840 + }, + { + "epoch": 1.13, + "learning_rate": 8.724136668253561e-08, + "loss": 0.7079, + "step": 30841 + }, + { + "epoch": 1.13, + "learning_rate": 8.723770461786355e-08, + "loss": 0.681, + "step": 30842 + }, + { + "epoch": 1.13, + "learning_rate": 8.723404255319149e-08, + "loss": 0.6873, + "step": 30843 + }, + { + "epoch": 1.13, + "learning_rate": 8.723038048851943e-08, + "loss": 0.6941, + "step": 30844 + }, + { + "epoch": 1.13, + "learning_rate": 8.722671842384736e-08, + "loss": 0.6916, + "step": 30845 + }, + { + "epoch": 1.13, + "learning_rate": 8.72230563591753e-08, + "loss": 0.6956, + "step": 30846 + }, + { + "epoch": 1.13, + "learning_rate": 8.721939429450324e-08, + "loss": 0.7033, + "step": 30847 + }, + { + "epoch": 1.13, + "learning_rate": 8.721573222983118e-08, + "loss": 0.7057, + "step": 30848 + }, + { + "epoch": 1.13, + "learning_rate": 8.721207016515912e-08, + "loss": 0.6999, + "step": 30849 + }, + { + "epoch": 1.13, + "learning_rate": 8.720840810048705e-08, + "loss": 0.6791, + "step": 30850 + }, + { + "epoch": 1.13, + "learning_rate": 8.720474603581499e-08, + "loss": 0.698, + "step": 30851 + }, + { + "epoch": 1.13, + "learning_rate": 8.720108397114293e-08, + "loss": 0.698, + "step": 30852 + }, + { + "epoch": 1.13, + "learning_rate": 8.719742190647087e-08, + "loss": 0.6896, + "step": 30853 + }, + { + "epoch": 1.13, + "learning_rate": 8.719375984179879e-08, + "loss": 0.6936, + "step": 30854 + }, + { + "epoch": 1.13, + "learning_rate": 8.719009777712673e-08, + "loss": 0.6951, + "step": 30855 + }, + { + "epoch": 1.13, + "learning_rate": 8.718643571245467e-08, + "loss": 0.6916, + "step": 30856 + }, + { + "epoch": 1.13, + "learning_rate": 8.718277364778261e-08, + "loss": 0.6929, + "step": 30857 + }, + { + "epoch": 1.13, + "learning_rate": 8.717911158311055e-08, + "loss": 0.6931, + "step": 30858 + }, + { + "epoch": 1.13, + "learning_rate": 8.717544951843849e-08, + "loss": 0.6968, + "step": 30859 + }, + { + "epoch": 1.13, + "learning_rate": 8.717178745376643e-08, + "loss": 0.6953, + "step": 30860 + }, + { + "epoch": 1.13, + "learning_rate": 8.716812538909437e-08, + "loss": 0.7001, + "step": 30861 + }, + { + "epoch": 1.13, + "learning_rate": 8.716446332442231e-08, + "loss": 0.6995, + "step": 30862 + }, + { + "epoch": 1.13, + "learning_rate": 8.716080125975024e-08, + "loss": 0.6876, + "step": 30863 + }, + { + "epoch": 1.13, + "learning_rate": 8.715713919507818e-08, + "loss": 0.6921, + "step": 30864 + }, + { + "epoch": 1.13, + "learning_rate": 8.715347713040612e-08, + "loss": 0.6943, + "step": 30865 + }, + { + "epoch": 1.13, + "learning_rate": 8.714981506573406e-08, + "loss": 0.6847, + "step": 30866 + }, + { + "epoch": 1.13, + "learning_rate": 8.7146153001062e-08, + "loss": 0.6919, + "step": 30867 + }, + { + "epoch": 1.13, + "learning_rate": 8.714249093638994e-08, + "loss": 0.6773, + "step": 30868 + }, + { + "epoch": 1.13, + "learning_rate": 8.713882887171788e-08, + "loss": 0.6977, + "step": 30869 + }, + { + "epoch": 1.13, + "learning_rate": 8.713516680704582e-08, + "loss": 0.7019, + "step": 30870 + }, + { + "epoch": 1.13, + "learning_rate": 8.713150474237376e-08, + "loss": 0.6901, + "step": 30871 + }, + { + "epoch": 1.13, + "learning_rate": 8.712784267770168e-08, + "loss": 0.6978, + "step": 30872 + }, + { + "epoch": 1.13, + "learning_rate": 8.712418061302962e-08, + "loss": 0.6932, + "step": 30873 + }, + { + "epoch": 1.13, + "learning_rate": 8.712051854835756e-08, + "loss": 0.6928, + "step": 30874 + }, + { + "epoch": 1.13, + "learning_rate": 8.711685648368549e-08, + "loss": 0.6902, + "step": 30875 + }, + { + "epoch": 1.13, + "learning_rate": 8.711319441901343e-08, + "loss": 0.7031, + "step": 30876 + }, + { + "epoch": 1.13, + "learning_rate": 8.710953235434137e-08, + "loss": 0.6922, + "step": 30877 + }, + { + "epoch": 1.13, + "learning_rate": 8.710587028966931e-08, + "loss": 0.6878, + "step": 30878 + }, + { + "epoch": 1.13, + "learning_rate": 8.710220822499725e-08, + "loss": 0.6961, + "step": 30879 + }, + { + "epoch": 1.13, + "learning_rate": 8.709854616032519e-08, + "loss": 0.6935, + "step": 30880 + }, + { + "epoch": 1.13, + "learning_rate": 8.709488409565312e-08, + "loss": 0.6968, + "step": 30881 + }, + { + "epoch": 1.13, + "learning_rate": 8.709122203098106e-08, + "loss": 0.679, + "step": 30882 + }, + { + "epoch": 1.13, + "learning_rate": 8.7087559966309e-08, + "loss": 0.7037, + "step": 30883 + }, + { + "epoch": 1.13, + "learning_rate": 8.708389790163694e-08, + "loss": 0.6969, + "step": 30884 + }, + { + "epoch": 1.13, + "learning_rate": 8.708023583696488e-08, + "loss": 0.6927, + "step": 30885 + }, + { + "epoch": 1.13, + "learning_rate": 8.707657377229282e-08, + "loss": 0.7025, + "step": 30886 + }, + { + "epoch": 1.13, + "learning_rate": 8.707291170762076e-08, + "loss": 0.6761, + "step": 30887 + }, + { + "epoch": 1.13, + "learning_rate": 8.70692496429487e-08, + "loss": 0.6887, + "step": 30888 + }, + { + "epoch": 1.13, + "learning_rate": 8.706558757827664e-08, + "loss": 0.6846, + "step": 30889 + }, + { + "epoch": 1.13, + "learning_rate": 8.706192551360456e-08, + "loss": 0.6899, + "step": 30890 + }, + { + "epoch": 1.13, + "learning_rate": 8.70582634489325e-08, + "loss": 0.6965, + "step": 30891 + }, + { + "epoch": 1.13, + "learning_rate": 8.705460138426044e-08, + "loss": 0.6977, + "step": 30892 + }, + { + "epoch": 1.13, + "learning_rate": 8.705093931958838e-08, + "loss": 0.6986, + "step": 30893 + }, + { + "epoch": 1.13, + "learning_rate": 8.704727725491632e-08, + "loss": 0.6868, + "step": 30894 + }, + { + "epoch": 1.13, + "learning_rate": 8.704361519024426e-08, + "loss": 0.6875, + "step": 30895 + }, + { + "epoch": 1.13, + "learning_rate": 8.703995312557219e-08, + "loss": 0.6911, + "step": 30896 + }, + { + "epoch": 1.13, + "learning_rate": 8.703629106090013e-08, + "loss": 0.7016, + "step": 30897 + }, + { + "epoch": 1.13, + "learning_rate": 8.703262899622807e-08, + "loss": 0.6927, + "step": 30898 + }, + { + "epoch": 1.13, + "learning_rate": 8.7028966931556e-08, + "loss": 0.6825, + "step": 30899 + }, + { + "epoch": 1.13, + "learning_rate": 8.702530486688394e-08, + "loss": 0.7089, + "step": 30900 + }, + { + "epoch": 1.13, + "learning_rate": 8.702164280221188e-08, + "loss": 0.7051, + "step": 30901 + }, + { + "epoch": 1.13, + "learning_rate": 8.701798073753982e-08, + "loss": 0.7006, + "step": 30902 + }, + { + "epoch": 1.13, + "learning_rate": 8.701431867286776e-08, + "loss": 0.6897, + "step": 30903 + }, + { + "epoch": 1.13, + "learning_rate": 8.70106566081957e-08, + "loss": 0.6729, + "step": 30904 + }, + { + "epoch": 1.13, + "learning_rate": 8.700699454352364e-08, + "loss": 0.6931, + "step": 30905 + }, + { + "epoch": 1.13, + "learning_rate": 8.700333247885158e-08, + "loss": 0.6851, + "step": 30906 + }, + { + "epoch": 1.13, + "learning_rate": 8.699967041417952e-08, + "loss": 0.6874, + "step": 30907 + }, + { + "epoch": 1.13, + "learning_rate": 8.699600834950744e-08, + "loss": 0.6861, + "step": 30908 + }, + { + "epoch": 1.13, + "learning_rate": 8.699234628483538e-08, + "loss": 0.7122, + "step": 30909 + }, + { + "epoch": 1.13, + "learning_rate": 8.698868422016332e-08, + "loss": 0.6959, + "step": 30910 + }, + { + "epoch": 1.13, + "learning_rate": 8.698502215549126e-08, + "loss": 0.6838, + "step": 30911 + }, + { + "epoch": 1.13, + "learning_rate": 8.69813600908192e-08, + "loss": 0.6951, + "step": 30912 + }, + { + "epoch": 1.13, + "learning_rate": 8.697769802614714e-08, + "loss": 0.6869, + "step": 30913 + }, + { + "epoch": 1.13, + "learning_rate": 8.697403596147508e-08, + "loss": 0.6904, + "step": 30914 + }, + { + "epoch": 1.13, + "learning_rate": 8.697037389680302e-08, + "loss": 0.6999, + "step": 30915 + }, + { + "epoch": 1.13, + "learning_rate": 8.696671183213096e-08, + "loss": 0.6997, + "step": 30916 + }, + { + "epoch": 1.13, + "learning_rate": 8.696304976745889e-08, + "loss": 0.6886, + "step": 30917 + }, + { + "epoch": 1.13, + "learning_rate": 8.695938770278683e-08, + "loss": 0.6824, + "step": 30918 + }, + { + "epoch": 1.13, + "learning_rate": 8.695572563811477e-08, + "loss": 0.6821, + "step": 30919 + }, + { + "epoch": 1.13, + "learning_rate": 8.695206357344271e-08, + "loss": 0.6889, + "step": 30920 + }, + { + "epoch": 1.13, + "learning_rate": 8.694840150877064e-08, + "loss": 0.6881, + "step": 30921 + }, + { + "epoch": 1.13, + "learning_rate": 8.694473944409858e-08, + "loss": 0.687, + "step": 30922 + }, + { + "epoch": 1.13, + "learning_rate": 8.694107737942651e-08, + "loss": 0.6919, + "step": 30923 + }, + { + "epoch": 1.13, + "learning_rate": 8.693741531475445e-08, + "loss": 0.6968, + "step": 30924 + }, + { + "epoch": 1.13, + "learning_rate": 8.69337532500824e-08, + "loss": 0.6954, + "step": 30925 + }, + { + "epoch": 1.13, + "learning_rate": 8.693009118541032e-08, + "loss": 0.6922, + "step": 30926 + }, + { + "epoch": 1.13, + "learning_rate": 8.692642912073826e-08, + "loss": 0.6925, + "step": 30927 + }, + { + "epoch": 1.13, + "learning_rate": 8.69227670560662e-08, + "loss": 0.6899, + "step": 30928 + }, + { + "epoch": 1.13, + "learning_rate": 8.691910499139414e-08, + "loss": 0.6976, + "step": 30929 + }, + { + "epoch": 1.13, + "learning_rate": 8.691544292672208e-08, + "loss": 0.6873, + "step": 30930 + }, + { + "epoch": 1.13, + "learning_rate": 8.691178086205002e-08, + "loss": 0.695, + "step": 30931 + }, + { + "epoch": 1.13, + "learning_rate": 8.690811879737796e-08, + "loss": 0.6938, + "step": 30932 + }, + { + "epoch": 1.13, + "learning_rate": 8.69044567327059e-08, + "loss": 0.6976, + "step": 30933 + }, + { + "epoch": 1.13, + "learning_rate": 8.690079466803384e-08, + "loss": 0.6989, + "step": 30934 + }, + { + "epoch": 1.13, + "learning_rate": 8.689713260336177e-08, + "loss": 0.6853, + "step": 30935 + }, + { + "epoch": 1.13, + "learning_rate": 8.689347053868971e-08, + "loss": 0.6955, + "step": 30936 + }, + { + "epoch": 1.13, + "learning_rate": 8.688980847401765e-08, + "loss": 0.6884, + "step": 30937 + }, + { + "epoch": 1.13, + "learning_rate": 8.688614640934559e-08, + "loss": 0.6849, + "step": 30938 + }, + { + "epoch": 1.13, + "learning_rate": 8.688248434467353e-08, + "loss": 0.6839, + "step": 30939 + }, + { + "epoch": 1.13, + "learning_rate": 8.687882228000147e-08, + "loss": 0.7029, + "step": 30940 + }, + { + "epoch": 1.13, + "learning_rate": 8.687516021532941e-08, + "loss": 0.6781, + "step": 30941 + }, + { + "epoch": 1.13, + "learning_rate": 8.687149815065735e-08, + "loss": 0.6909, + "step": 30942 + }, + { + "epoch": 1.13, + "learning_rate": 8.686783608598527e-08, + "loss": 0.6996, + "step": 30943 + }, + { + "epoch": 1.13, + "learning_rate": 8.686417402131321e-08, + "loss": 0.6985, + "step": 30944 + }, + { + "epoch": 1.13, + "learning_rate": 8.686051195664114e-08, + "loss": 0.6929, + "step": 30945 + }, + { + "epoch": 1.13, + "learning_rate": 8.685684989196908e-08, + "loss": 0.679, + "step": 30946 + }, + { + "epoch": 1.13, + "learning_rate": 8.685318782729702e-08, + "loss": 0.682, + "step": 30947 + }, + { + "epoch": 1.13, + "learning_rate": 8.684952576262496e-08, + "loss": 0.6968, + "step": 30948 + }, + { + "epoch": 1.13, + "learning_rate": 8.68458636979529e-08, + "loss": 0.6879, + "step": 30949 + }, + { + "epoch": 1.13, + "learning_rate": 8.684220163328084e-08, + "loss": 0.7104, + "step": 30950 + }, + { + "epoch": 1.13, + "learning_rate": 8.683853956860878e-08, + "loss": 0.6891, + "step": 30951 + }, + { + "epoch": 1.13, + "learning_rate": 8.683487750393672e-08, + "loss": 0.6943, + "step": 30952 + }, + { + "epoch": 1.13, + "learning_rate": 8.683121543926465e-08, + "loss": 0.6921, + "step": 30953 + }, + { + "epoch": 1.13, + "learning_rate": 8.682755337459259e-08, + "loss": 0.6948, + "step": 30954 + }, + { + "epoch": 1.13, + "learning_rate": 8.682389130992053e-08, + "loss": 0.6983, + "step": 30955 + }, + { + "epoch": 1.13, + "learning_rate": 8.682022924524847e-08, + "loss": 0.6836, + "step": 30956 + }, + { + "epoch": 1.13, + "learning_rate": 8.68165671805764e-08, + "loss": 0.6996, + "step": 30957 + }, + { + "epoch": 1.13, + "learning_rate": 8.681290511590435e-08, + "loss": 0.6906, + "step": 30958 + }, + { + "epoch": 1.13, + "learning_rate": 8.680924305123229e-08, + "loss": 0.6889, + "step": 30959 + }, + { + "epoch": 1.13, + "learning_rate": 8.680558098656023e-08, + "loss": 0.6939, + "step": 30960 + }, + { + "epoch": 1.13, + "learning_rate": 8.680191892188817e-08, + "loss": 0.6907, + "step": 30961 + }, + { + "epoch": 1.13, + "learning_rate": 8.679825685721609e-08, + "loss": 0.6952, + "step": 30962 + }, + { + "epoch": 1.13, + "learning_rate": 8.679459479254403e-08, + "loss": 0.6933, + "step": 30963 + }, + { + "epoch": 1.13, + "learning_rate": 8.679093272787197e-08, + "loss": 0.6971, + "step": 30964 + }, + { + "epoch": 1.13, + "learning_rate": 8.678727066319991e-08, + "loss": 0.6766, + "step": 30965 + }, + { + "epoch": 1.13, + "learning_rate": 8.678360859852785e-08, + "loss": 0.6959, + "step": 30966 + }, + { + "epoch": 1.13, + "learning_rate": 8.677994653385578e-08, + "loss": 0.6991, + "step": 30967 + }, + { + "epoch": 1.13, + "learning_rate": 8.677628446918372e-08, + "loss": 0.7046, + "step": 30968 + }, + { + "epoch": 1.13, + "learning_rate": 8.677262240451166e-08, + "loss": 0.69, + "step": 30969 + }, + { + "epoch": 1.13, + "learning_rate": 8.67689603398396e-08, + "loss": 0.6995, + "step": 30970 + }, + { + "epoch": 1.13, + "learning_rate": 8.676529827516753e-08, + "loss": 0.6945, + "step": 30971 + }, + { + "epoch": 1.13, + "learning_rate": 8.676163621049547e-08, + "loss": 0.705, + "step": 30972 + }, + { + "epoch": 1.13, + "learning_rate": 8.67579741458234e-08, + "loss": 0.7012, + "step": 30973 + }, + { + "epoch": 1.13, + "learning_rate": 8.675431208115135e-08, + "loss": 0.6942, + "step": 30974 + }, + { + "epoch": 1.13, + "learning_rate": 8.675065001647929e-08, + "loss": 0.6937, + "step": 30975 + }, + { + "epoch": 1.13, + "learning_rate": 8.674698795180723e-08, + "loss": 0.6944, + "step": 30976 + }, + { + "epoch": 1.13, + "learning_rate": 8.674332588713517e-08, + "loss": 0.6958, + "step": 30977 + }, + { + "epoch": 1.13, + "learning_rate": 8.67396638224631e-08, + "loss": 0.7022, + "step": 30978 + }, + { + "epoch": 1.13, + "learning_rate": 8.673600175779105e-08, + "loss": 0.6844, + "step": 30979 + }, + { + "epoch": 1.13, + "learning_rate": 8.673233969311897e-08, + "loss": 0.685, + "step": 30980 + }, + { + "epoch": 1.13, + "learning_rate": 8.672867762844691e-08, + "loss": 0.6952, + "step": 30981 + }, + { + "epoch": 1.13, + "learning_rate": 8.672501556377485e-08, + "loss": 0.7037, + "step": 30982 + }, + { + "epoch": 1.13, + "learning_rate": 8.672135349910279e-08, + "loss": 0.6989, + "step": 30983 + }, + { + "epoch": 1.13, + "learning_rate": 8.671769143443073e-08, + "loss": 0.6974, + "step": 30984 + }, + { + "epoch": 1.13, + "learning_rate": 8.671402936975867e-08, + "loss": 0.7016, + "step": 30985 + }, + { + "epoch": 1.13, + "learning_rate": 8.671036730508661e-08, + "loss": 0.6934, + "step": 30986 + }, + { + "epoch": 1.13, + "learning_rate": 8.670670524041455e-08, + "loss": 0.7055, + "step": 30987 + }, + { + "epoch": 1.13, + "learning_rate": 8.670304317574249e-08, + "loss": 0.6931, + "step": 30988 + }, + { + "epoch": 1.13, + "learning_rate": 8.669938111107042e-08, + "loss": 0.6933, + "step": 30989 + }, + { + "epoch": 1.13, + "learning_rate": 8.669571904639836e-08, + "loss": 0.7052, + "step": 30990 + }, + { + "epoch": 1.13, + "learning_rate": 8.66920569817263e-08, + "loss": 0.6991, + "step": 30991 + }, + { + "epoch": 1.13, + "learning_rate": 8.668839491705422e-08, + "loss": 0.6999, + "step": 30992 + }, + { + "epoch": 1.13, + "learning_rate": 8.668473285238216e-08, + "loss": 0.6901, + "step": 30993 + }, + { + "epoch": 1.13, + "learning_rate": 8.66810707877101e-08, + "loss": 0.7021, + "step": 30994 + }, + { + "epoch": 1.13, + "learning_rate": 8.667740872303804e-08, + "loss": 0.6945, + "step": 30995 + }, + { + "epoch": 1.13, + "learning_rate": 8.667374665836598e-08, + "loss": 0.6841, + "step": 30996 + }, + { + "epoch": 1.13, + "learning_rate": 8.667008459369392e-08, + "loss": 0.69, + "step": 30997 + }, + { + "epoch": 1.13, + "learning_rate": 8.666642252902185e-08, + "loss": 0.6918, + "step": 30998 + }, + { + "epoch": 1.13, + "learning_rate": 8.666276046434979e-08, + "loss": 0.6992, + "step": 30999 + }, + { + "epoch": 1.13, + "learning_rate": 8.665909839967773e-08, + "loss": 0.6896, + "step": 31000 + }, + { + "epoch": 1.13, + "learning_rate": 8.665543633500567e-08, + "loss": 0.6924, + "step": 31001 + }, + { + "epoch": 1.13, + "learning_rate": 8.665177427033361e-08, + "loss": 0.703, + "step": 31002 + }, + { + "epoch": 1.13, + "learning_rate": 8.664811220566155e-08, + "loss": 0.6929, + "step": 31003 + }, + { + "epoch": 1.13, + "learning_rate": 8.664445014098949e-08, + "loss": 0.6927, + "step": 31004 + }, + { + "epoch": 1.13, + "learning_rate": 8.664078807631743e-08, + "loss": 0.7029, + "step": 31005 + }, + { + "epoch": 1.13, + "learning_rate": 8.663712601164537e-08, + "loss": 0.6898, + "step": 31006 + }, + { + "epoch": 1.13, + "learning_rate": 8.66334639469733e-08, + "loss": 0.702, + "step": 31007 + }, + { + "epoch": 1.13, + "learning_rate": 8.662980188230124e-08, + "loss": 0.695, + "step": 31008 + }, + { + "epoch": 1.13, + "learning_rate": 8.662613981762918e-08, + "loss": 0.6866, + "step": 31009 + }, + { + "epoch": 1.13, + "learning_rate": 8.662247775295712e-08, + "loss": 0.6851, + "step": 31010 + }, + { + "epoch": 1.13, + "learning_rate": 8.661881568828506e-08, + "loss": 0.6865, + "step": 31011 + }, + { + "epoch": 1.13, + "learning_rate": 8.6615153623613e-08, + "loss": 0.7008, + "step": 31012 + }, + { + "epoch": 1.13, + "learning_rate": 8.661149155894094e-08, + "loss": 0.6912, + "step": 31013 + }, + { + "epoch": 1.13, + "learning_rate": 8.660782949426886e-08, + "loss": 0.6844, + "step": 31014 + }, + { + "epoch": 1.13, + "learning_rate": 8.66041674295968e-08, + "loss": 0.693, + "step": 31015 + }, + { + "epoch": 1.13, + "learning_rate": 8.660050536492473e-08, + "loss": 0.7011, + "step": 31016 + }, + { + "epoch": 1.13, + "learning_rate": 8.659684330025267e-08, + "loss": 0.6969, + "step": 31017 + }, + { + "epoch": 1.13, + "learning_rate": 8.659318123558061e-08, + "loss": 0.6999, + "step": 31018 + }, + { + "epoch": 1.13, + "learning_rate": 8.658951917090855e-08, + "loss": 0.681, + "step": 31019 + }, + { + "epoch": 1.13, + "learning_rate": 8.658585710623649e-08, + "loss": 0.6959, + "step": 31020 + }, + { + "epoch": 1.13, + "learning_rate": 8.658219504156443e-08, + "loss": 0.6991, + "step": 31021 + }, + { + "epoch": 1.14, + "learning_rate": 8.657853297689237e-08, + "loss": 0.7016, + "step": 31022 + }, + { + "epoch": 1.14, + "learning_rate": 8.657487091222031e-08, + "loss": 0.6831, + "step": 31023 + }, + { + "epoch": 1.14, + "learning_rate": 8.657120884754824e-08, + "loss": 0.7062, + "step": 31024 + }, + { + "epoch": 1.14, + "learning_rate": 8.656754678287618e-08, + "loss": 0.7, + "step": 31025 + }, + { + "epoch": 1.14, + "learning_rate": 8.656388471820412e-08, + "loss": 0.6977, + "step": 31026 + }, + { + "epoch": 1.14, + "learning_rate": 8.656022265353206e-08, + "loss": 0.6706, + "step": 31027 + }, + { + "epoch": 1.14, + "learning_rate": 8.655656058886e-08, + "loss": 0.6861, + "step": 31028 + }, + { + "epoch": 1.14, + "learning_rate": 8.655289852418794e-08, + "loss": 0.6858, + "step": 31029 + }, + { + "epoch": 1.14, + "learning_rate": 8.654923645951588e-08, + "loss": 0.7003, + "step": 31030 + }, + { + "epoch": 1.14, + "learning_rate": 8.654557439484382e-08, + "loss": 0.6951, + "step": 31031 + }, + { + "epoch": 1.14, + "learning_rate": 8.654191233017176e-08, + "loss": 0.6816, + "step": 31032 + }, + { + "epoch": 1.14, + "learning_rate": 8.653825026549968e-08, + "loss": 0.6817, + "step": 31033 + }, + { + "epoch": 1.14, + "learning_rate": 8.653458820082762e-08, + "loss": 0.6896, + "step": 31034 + }, + { + "epoch": 1.14, + "learning_rate": 8.653092613615556e-08, + "loss": 0.6955, + "step": 31035 + }, + { + "epoch": 1.14, + "learning_rate": 8.65272640714835e-08, + "loss": 0.6915, + "step": 31036 + }, + { + "epoch": 1.14, + "learning_rate": 8.652360200681144e-08, + "loss": 0.6806, + "step": 31037 + }, + { + "epoch": 1.14, + "learning_rate": 8.651993994213937e-08, + "loss": 0.6932, + "step": 31038 + }, + { + "epoch": 1.14, + "learning_rate": 8.651627787746731e-08, + "loss": 0.6915, + "step": 31039 + }, + { + "epoch": 1.14, + "learning_rate": 8.651261581279525e-08, + "loss": 0.689, + "step": 31040 + }, + { + "epoch": 1.14, + "learning_rate": 8.650895374812319e-08, + "loss": 0.6943, + "step": 31041 + }, + { + "epoch": 1.14, + "learning_rate": 8.650529168345112e-08, + "loss": 0.7085, + "step": 31042 + }, + { + "epoch": 1.14, + "learning_rate": 8.650162961877905e-08, + "loss": 0.6897, + "step": 31043 + }, + { + "epoch": 1.14, + "learning_rate": 8.6497967554107e-08, + "loss": 0.6998, + "step": 31044 + }, + { + "epoch": 1.14, + "learning_rate": 8.649430548943493e-08, + "loss": 0.6956, + "step": 31045 + }, + { + "epoch": 1.14, + "learning_rate": 8.649064342476287e-08, + "loss": 0.689, + "step": 31046 + }, + { + "epoch": 1.14, + "learning_rate": 8.648698136009081e-08, + "loss": 0.6906, + "step": 31047 + }, + { + "epoch": 1.14, + "learning_rate": 8.648331929541875e-08, + "loss": 0.6918, + "step": 31048 + }, + { + "epoch": 1.14, + "learning_rate": 8.64796572307467e-08, + "loss": 0.6864, + "step": 31049 + }, + { + "epoch": 1.14, + "learning_rate": 8.647599516607463e-08, + "loss": 0.6869, + "step": 31050 + }, + { + "epoch": 1.14, + "learning_rate": 8.647233310140256e-08, + "loss": 0.6892, + "step": 31051 + }, + { + "epoch": 1.14, + "learning_rate": 8.64686710367305e-08, + "loss": 0.7098, + "step": 31052 + }, + { + "epoch": 1.14, + "learning_rate": 8.646500897205844e-08, + "loss": 0.6909, + "step": 31053 + }, + { + "epoch": 1.14, + "learning_rate": 8.646134690738638e-08, + "loss": 0.6869, + "step": 31054 + }, + { + "epoch": 1.14, + "learning_rate": 8.645768484271432e-08, + "loss": 0.7025, + "step": 31055 + }, + { + "epoch": 1.14, + "learning_rate": 8.645402277804226e-08, + "loss": 0.6982, + "step": 31056 + }, + { + "epoch": 1.14, + "learning_rate": 8.64503607133702e-08, + "loss": 0.699, + "step": 31057 + }, + { + "epoch": 1.14, + "learning_rate": 8.644669864869814e-08, + "loss": 0.6845, + "step": 31058 + }, + { + "epoch": 1.14, + "learning_rate": 8.644303658402608e-08, + "loss": 0.6919, + "step": 31059 + }, + { + "epoch": 1.14, + "learning_rate": 8.643937451935401e-08, + "loss": 0.6898, + "step": 31060 + }, + { + "epoch": 1.14, + "learning_rate": 8.643571245468195e-08, + "loss": 0.7001, + "step": 31061 + }, + { + "epoch": 1.14, + "learning_rate": 8.643205039000989e-08, + "loss": 0.6946, + "step": 31062 + }, + { + "epoch": 1.14, + "learning_rate": 8.642838832533781e-08, + "loss": 0.6924, + "step": 31063 + }, + { + "epoch": 1.14, + "learning_rate": 8.642472626066575e-08, + "loss": 0.6901, + "step": 31064 + }, + { + "epoch": 1.14, + "learning_rate": 8.64210641959937e-08, + "loss": 0.6884, + "step": 31065 + }, + { + "epoch": 1.14, + "learning_rate": 8.641740213132163e-08, + "loss": 0.6991, + "step": 31066 + }, + { + "epoch": 1.14, + "learning_rate": 8.641374006664957e-08, + "loss": 0.6888, + "step": 31067 + }, + { + "epoch": 1.14, + "learning_rate": 8.641007800197751e-08, + "loss": 0.6945, + "step": 31068 + }, + { + "epoch": 1.14, + "learning_rate": 8.640641593730544e-08, + "loss": 0.7039, + "step": 31069 + }, + { + "epoch": 1.14, + "learning_rate": 8.640275387263338e-08, + "loss": 0.6972, + "step": 31070 + }, + { + "epoch": 1.14, + "learning_rate": 8.639909180796132e-08, + "loss": 0.6848, + "step": 31071 + }, + { + "epoch": 1.14, + "learning_rate": 8.639542974328926e-08, + "loss": 0.6945, + "step": 31072 + }, + { + "epoch": 1.14, + "learning_rate": 8.63917676786172e-08, + "loss": 0.6934, + "step": 31073 + }, + { + "epoch": 1.14, + "learning_rate": 8.638810561394514e-08, + "loss": 0.6833, + "step": 31074 + }, + { + "epoch": 1.14, + "learning_rate": 8.638444354927308e-08, + "loss": 0.6829, + "step": 31075 + }, + { + "epoch": 1.14, + "learning_rate": 8.638078148460102e-08, + "loss": 0.6875, + "step": 31076 + }, + { + "epoch": 1.14, + "learning_rate": 8.637711941992896e-08, + "loss": 0.6911, + "step": 31077 + }, + { + "epoch": 1.14, + "learning_rate": 8.637345735525689e-08, + "loss": 0.6905, + "step": 31078 + }, + { + "epoch": 1.14, + "learning_rate": 8.636979529058483e-08, + "loss": 0.6991, + "step": 31079 + }, + { + "epoch": 1.14, + "learning_rate": 8.636613322591277e-08, + "loss": 0.7064, + "step": 31080 + }, + { + "epoch": 1.14, + "learning_rate": 8.63624711612407e-08, + "loss": 0.691, + "step": 31081 + }, + { + "epoch": 1.14, + "learning_rate": 8.635880909656865e-08, + "loss": 0.6906, + "step": 31082 + }, + { + "epoch": 1.14, + "learning_rate": 8.635514703189659e-08, + "loss": 0.6993, + "step": 31083 + }, + { + "epoch": 1.14, + "learning_rate": 8.635148496722453e-08, + "loss": 0.6938, + "step": 31084 + }, + { + "epoch": 1.14, + "learning_rate": 8.634782290255245e-08, + "loss": 0.6891, + "step": 31085 + }, + { + "epoch": 1.14, + "learning_rate": 8.634416083788039e-08, + "loss": 0.6952, + "step": 31086 + }, + { + "epoch": 1.14, + "learning_rate": 8.634049877320832e-08, + "loss": 0.6886, + "step": 31087 + }, + { + "epoch": 1.14, + "learning_rate": 8.633683670853626e-08, + "loss": 0.7151, + "step": 31088 + }, + { + "epoch": 1.14, + "learning_rate": 8.63331746438642e-08, + "loss": 0.6829, + "step": 31089 + }, + { + "epoch": 1.14, + "learning_rate": 8.632951257919214e-08, + "loss": 0.7048, + "step": 31090 + }, + { + "epoch": 1.14, + "learning_rate": 8.632585051452008e-08, + "loss": 0.6905, + "step": 31091 + }, + { + "epoch": 1.14, + "learning_rate": 8.632218844984802e-08, + "loss": 0.6952, + "step": 31092 + }, + { + "epoch": 1.14, + "learning_rate": 8.631852638517596e-08, + "loss": 0.6968, + "step": 31093 + }, + { + "epoch": 1.14, + "learning_rate": 8.63148643205039e-08, + "loss": 0.6996, + "step": 31094 + }, + { + "epoch": 1.14, + "learning_rate": 8.631120225583184e-08, + "loss": 0.694, + "step": 31095 + }, + { + "epoch": 1.14, + "learning_rate": 8.630754019115977e-08, + "loss": 0.6946, + "step": 31096 + }, + { + "epoch": 1.14, + "learning_rate": 8.63038781264877e-08, + "loss": 0.6856, + "step": 31097 + }, + { + "epoch": 1.14, + "learning_rate": 8.630021606181565e-08, + "loss": 0.7059, + "step": 31098 + }, + { + "epoch": 1.14, + "learning_rate": 8.629655399714359e-08, + "loss": 0.7002, + "step": 31099 + }, + { + "epoch": 1.14, + "learning_rate": 8.629289193247152e-08, + "loss": 0.7097, + "step": 31100 + }, + { + "epoch": 1.14, + "learning_rate": 8.628922986779946e-08, + "loss": 0.6918, + "step": 31101 + }, + { + "epoch": 1.14, + "learning_rate": 8.62855678031274e-08, + "loss": 0.6973, + "step": 31102 + }, + { + "epoch": 1.14, + "learning_rate": 8.628190573845534e-08, + "loss": 0.669, + "step": 31103 + }, + { + "epoch": 1.14, + "learning_rate": 8.627824367378328e-08, + "loss": 0.6907, + "step": 31104 + }, + { + "epoch": 1.14, + "learning_rate": 8.627458160911121e-08, + "loss": 0.7069, + "step": 31105 + }, + { + "epoch": 1.14, + "learning_rate": 8.627091954443915e-08, + "loss": 0.6951, + "step": 31106 + }, + { + "epoch": 1.14, + "learning_rate": 8.626725747976709e-08, + "loss": 0.694, + "step": 31107 + }, + { + "epoch": 1.14, + "learning_rate": 8.626359541509503e-08, + "loss": 0.6853, + "step": 31108 + }, + { + "epoch": 1.14, + "learning_rate": 8.625993335042296e-08, + "loss": 0.6883, + "step": 31109 + }, + { + "epoch": 1.14, + "learning_rate": 8.62562712857509e-08, + "loss": 0.6822, + "step": 31110 + }, + { + "epoch": 1.14, + "learning_rate": 8.625260922107884e-08, + "loss": 0.7002, + "step": 31111 + }, + { + "epoch": 1.14, + "learning_rate": 8.624894715640678e-08, + "loss": 0.6989, + "step": 31112 + }, + { + "epoch": 1.14, + "learning_rate": 8.624528509173472e-08, + "loss": 0.6901, + "step": 31113 + }, + { + "epoch": 1.14, + "learning_rate": 8.624162302706264e-08, + "loss": 0.6921, + "step": 31114 + }, + { + "epoch": 1.14, + "learning_rate": 8.623796096239058e-08, + "loss": 0.686, + "step": 31115 + }, + { + "epoch": 1.14, + "learning_rate": 8.623429889771852e-08, + "loss": 0.6884, + "step": 31116 + }, + { + "epoch": 1.14, + "learning_rate": 8.623063683304646e-08, + "loss": 0.6953, + "step": 31117 + }, + { + "epoch": 1.14, + "learning_rate": 8.62269747683744e-08, + "loss": 0.6959, + "step": 31118 + }, + { + "epoch": 1.14, + "learning_rate": 8.622331270370234e-08, + "loss": 0.6935, + "step": 31119 + }, + { + "epoch": 1.14, + "learning_rate": 8.621965063903028e-08, + "loss": 0.7007, + "step": 31120 + }, + { + "epoch": 1.14, + "learning_rate": 8.621598857435822e-08, + "loss": 0.7093, + "step": 31121 + }, + { + "epoch": 1.14, + "learning_rate": 8.621232650968616e-08, + "loss": 0.6903, + "step": 31122 + }, + { + "epoch": 1.14, + "learning_rate": 8.620866444501409e-08, + "loss": 0.7022, + "step": 31123 + }, + { + "epoch": 1.14, + "learning_rate": 8.620500238034203e-08, + "loss": 0.6875, + "step": 31124 + }, + { + "epoch": 1.14, + "learning_rate": 8.620134031566997e-08, + "loss": 0.7046, + "step": 31125 + }, + { + "epoch": 1.14, + "learning_rate": 8.619767825099791e-08, + "loss": 0.692, + "step": 31126 + }, + { + "epoch": 1.14, + "learning_rate": 8.619401618632585e-08, + "loss": 0.6906, + "step": 31127 + }, + { + "epoch": 1.14, + "learning_rate": 8.619035412165379e-08, + "loss": 0.6962, + "step": 31128 + }, + { + "epoch": 1.14, + "learning_rate": 8.618669205698173e-08, + "loss": 0.6968, + "step": 31129 + }, + { + "epoch": 1.14, + "learning_rate": 8.618302999230967e-08, + "loss": 0.6866, + "step": 31130 + }, + { + "epoch": 1.14, + "learning_rate": 8.61793679276376e-08, + "loss": 0.6932, + "step": 31131 + }, + { + "epoch": 1.14, + "learning_rate": 8.617570586296554e-08, + "loss": 0.6974, + "step": 31132 + }, + { + "epoch": 1.14, + "learning_rate": 8.617204379829348e-08, + "loss": 0.691, + "step": 31133 + }, + { + "epoch": 1.14, + "learning_rate": 8.61683817336214e-08, + "loss": 0.697, + "step": 31134 + }, + { + "epoch": 1.14, + "learning_rate": 8.616471966894934e-08, + "loss": 0.6892, + "step": 31135 + }, + { + "epoch": 1.14, + "learning_rate": 8.616105760427728e-08, + "loss": 0.6961, + "step": 31136 + }, + { + "epoch": 1.14, + "learning_rate": 8.615739553960522e-08, + "loss": 0.6875, + "step": 31137 + }, + { + "epoch": 1.14, + "learning_rate": 8.615373347493316e-08, + "loss": 0.6981, + "step": 31138 + }, + { + "epoch": 1.14, + "learning_rate": 8.61500714102611e-08, + "loss": 0.692, + "step": 31139 + }, + { + "epoch": 1.14, + "learning_rate": 8.614640934558904e-08, + "loss": 0.6937, + "step": 31140 + }, + { + "epoch": 1.14, + "learning_rate": 8.614274728091697e-08, + "loss": 0.6882, + "step": 31141 + }, + { + "epoch": 1.14, + "learning_rate": 8.613908521624491e-08, + "loss": 0.6864, + "step": 31142 + }, + { + "epoch": 1.14, + "learning_rate": 8.613542315157285e-08, + "loss": 0.7036, + "step": 31143 + }, + { + "epoch": 1.14, + "learning_rate": 8.613176108690079e-08, + "loss": 0.6933, + "step": 31144 + }, + { + "epoch": 1.14, + "learning_rate": 8.612809902222873e-08, + "loss": 0.6941, + "step": 31145 + }, + { + "epoch": 1.14, + "learning_rate": 8.612443695755667e-08, + "loss": 0.6919, + "step": 31146 + }, + { + "epoch": 1.14, + "learning_rate": 8.612077489288461e-08, + "loss": 0.6949, + "step": 31147 + }, + { + "epoch": 1.14, + "learning_rate": 8.611711282821255e-08, + "loss": 0.6856, + "step": 31148 + }, + { + "epoch": 1.14, + "learning_rate": 8.611345076354049e-08, + "loss": 0.6883, + "step": 31149 + }, + { + "epoch": 1.14, + "learning_rate": 8.610978869886842e-08, + "loss": 0.6912, + "step": 31150 + }, + { + "epoch": 1.14, + "learning_rate": 8.610612663419636e-08, + "loss": 0.7051, + "step": 31151 + }, + { + "epoch": 1.14, + "learning_rate": 8.61024645695243e-08, + "loss": 0.6954, + "step": 31152 + }, + { + "epoch": 1.14, + "learning_rate": 8.609880250485224e-08, + "loss": 0.6968, + "step": 31153 + }, + { + "epoch": 1.14, + "learning_rate": 8.609514044018018e-08, + "loss": 0.6807, + "step": 31154 + }, + { + "epoch": 1.14, + "learning_rate": 8.609147837550812e-08, + "loss": 0.6897, + "step": 31155 + }, + { + "epoch": 1.14, + "learning_rate": 8.608781631083604e-08, + "loss": 0.6838, + "step": 31156 + }, + { + "epoch": 1.14, + "learning_rate": 8.608415424616398e-08, + "loss": 0.6869, + "step": 31157 + }, + { + "epoch": 1.14, + "learning_rate": 8.608049218149192e-08, + "loss": 0.6926, + "step": 31158 + }, + { + "epoch": 1.14, + "learning_rate": 8.607683011681985e-08, + "loss": 0.6936, + "step": 31159 + }, + { + "epoch": 1.14, + "learning_rate": 8.607316805214779e-08, + "loss": 0.6952, + "step": 31160 + }, + { + "epoch": 1.14, + "learning_rate": 8.606950598747573e-08, + "loss": 0.6906, + "step": 31161 + }, + { + "epoch": 1.14, + "learning_rate": 8.606584392280367e-08, + "loss": 0.6967, + "step": 31162 + }, + { + "epoch": 1.14, + "learning_rate": 8.606218185813161e-08, + "loss": 0.6928, + "step": 31163 + }, + { + "epoch": 1.14, + "learning_rate": 8.605851979345955e-08, + "loss": 0.691, + "step": 31164 + }, + { + "epoch": 1.14, + "learning_rate": 8.605485772878749e-08, + "loss": 0.6812, + "step": 31165 + }, + { + "epoch": 1.14, + "learning_rate": 8.605119566411543e-08, + "loss": 0.6917, + "step": 31166 + }, + { + "epoch": 1.14, + "learning_rate": 8.604753359944337e-08, + "loss": 0.6973, + "step": 31167 + }, + { + "epoch": 1.14, + "learning_rate": 8.60438715347713e-08, + "loss": 0.702, + "step": 31168 + }, + { + "epoch": 1.14, + "learning_rate": 8.604020947009923e-08, + "loss": 0.6947, + "step": 31169 + }, + { + "epoch": 1.14, + "learning_rate": 8.603654740542717e-08, + "loss": 0.6862, + "step": 31170 + }, + { + "epoch": 1.14, + "learning_rate": 8.603288534075511e-08, + "loss": 0.7001, + "step": 31171 + }, + { + "epoch": 1.14, + "learning_rate": 8.602922327608305e-08, + "loss": 0.6933, + "step": 31172 + }, + { + "epoch": 1.14, + "learning_rate": 8.6025561211411e-08, + "loss": 0.6874, + "step": 31173 + }, + { + "epoch": 1.14, + "learning_rate": 8.602189914673893e-08, + "loss": 0.6953, + "step": 31174 + }, + { + "epoch": 1.14, + "learning_rate": 8.601823708206687e-08, + "loss": 0.7031, + "step": 31175 + }, + { + "epoch": 1.14, + "learning_rate": 8.601457501739481e-08, + "loss": 0.691, + "step": 31176 + }, + { + "epoch": 1.14, + "learning_rate": 8.601091295272274e-08, + "loss": 0.699, + "step": 31177 + }, + { + "epoch": 1.14, + "learning_rate": 8.600725088805068e-08, + "loss": 0.7003, + "step": 31178 + }, + { + "epoch": 1.14, + "learning_rate": 8.600358882337862e-08, + "loss": 0.6969, + "step": 31179 + }, + { + "epoch": 1.14, + "learning_rate": 8.599992675870655e-08, + "loss": 0.6923, + "step": 31180 + }, + { + "epoch": 1.14, + "learning_rate": 8.599626469403449e-08, + "loss": 0.6932, + "step": 31181 + }, + { + "epoch": 1.14, + "learning_rate": 8.599260262936243e-08, + "loss": 0.6886, + "step": 31182 + }, + { + "epoch": 1.14, + "learning_rate": 8.598894056469037e-08, + "loss": 0.6882, + "step": 31183 + }, + { + "epoch": 1.14, + "learning_rate": 8.598527850001831e-08, + "loss": 0.6819, + "step": 31184 + }, + { + "epoch": 1.14, + "learning_rate": 8.598161643534625e-08, + "loss": 0.6806, + "step": 31185 + }, + { + "epoch": 1.14, + "learning_rate": 8.597795437067417e-08, + "loss": 0.6874, + "step": 31186 + }, + { + "epoch": 1.14, + "learning_rate": 8.597429230600211e-08, + "loss": 0.6983, + "step": 31187 + }, + { + "epoch": 1.14, + "learning_rate": 8.597063024133005e-08, + "loss": 0.6888, + "step": 31188 + }, + { + "epoch": 1.14, + "learning_rate": 8.596696817665799e-08, + "loss": 0.6969, + "step": 31189 + }, + { + "epoch": 1.14, + "learning_rate": 8.596330611198593e-08, + "loss": 0.6981, + "step": 31190 + }, + { + "epoch": 1.14, + "learning_rate": 8.595964404731387e-08, + "loss": 0.6844, + "step": 31191 + }, + { + "epoch": 1.14, + "learning_rate": 8.595598198264181e-08, + "loss": 0.6901, + "step": 31192 + }, + { + "epoch": 1.14, + "learning_rate": 8.595231991796975e-08, + "loss": 0.689, + "step": 31193 + }, + { + "epoch": 1.14, + "learning_rate": 8.594865785329769e-08, + "loss": 0.6878, + "step": 31194 + }, + { + "epoch": 1.14, + "learning_rate": 8.594499578862562e-08, + "loss": 0.6876, + "step": 31195 + }, + { + "epoch": 1.14, + "learning_rate": 8.594133372395356e-08, + "loss": 0.6959, + "step": 31196 + }, + { + "epoch": 1.14, + "learning_rate": 8.59376716592815e-08, + "loss": 0.6975, + "step": 31197 + }, + { + "epoch": 1.14, + "learning_rate": 8.593400959460944e-08, + "loss": 0.6885, + "step": 31198 + }, + { + "epoch": 1.14, + "learning_rate": 8.593034752993738e-08, + "loss": 0.697, + "step": 31199 + }, + { + "epoch": 1.14, + "learning_rate": 8.592668546526532e-08, + "loss": 0.6875, + "step": 31200 + }, + { + "epoch": 1.14, + "learning_rate": 8.592302340059326e-08, + "loss": 0.694, + "step": 31201 + }, + { + "epoch": 1.14, + "learning_rate": 8.591936133592119e-08, + "loss": 0.6895, + "step": 31202 + }, + { + "epoch": 1.14, + "learning_rate": 8.591569927124913e-08, + "loss": 0.6904, + "step": 31203 + }, + { + "epoch": 1.14, + "learning_rate": 8.591203720657707e-08, + "loss": 0.691, + "step": 31204 + }, + { + "epoch": 1.14, + "learning_rate": 8.590837514190499e-08, + "loss": 0.7011, + "step": 31205 + }, + { + "epoch": 1.14, + "learning_rate": 8.590471307723293e-08, + "loss": 0.7017, + "step": 31206 + }, + { + "epoch": 1.14, + "learning_rate": 8.590105101256087e-08, + "loss": 0.6902, + "step": 31207 + }, + { + "epoch": 1.14, + "learning_rate": 8.589738894788881e-08, + "loss": 0.6916, + "step": 31208 + }, + { + "epoch": 1.14, + "learning_rate": 8.589372688321675e-08, + "loss": 0.7023, + "step": 31209 + }, + { + "epoch": 1.14, + "learning_rate": 8.589006481854469e-08, + "loss": 0.6994, + "step": 31210 + }, + { + "epoch": 1.14, + "learning_rate": 8.588640275387263e-08, + "loss": 0.6857, + "step": 31211 + }, + { + "epoch": 1.14, + "learning_rate": 8.588274068920057e-08, + "loss": 0.7224, + "step": 31212 + }, + { + "epoch": 1.14, + "learning_rate": 8.58790786245285e-08, + "loss": 0.7046, + "step": 31213 + }, + { + "epoch": 1.14, + "learning_rate": 8.587541655985644e-08, + "loss": 0.7057, + "step": 31214 + }, + { + "epoch": 1.14, + "learning_rate": 8.587175449518438e-08, + "loss": 0.6931, + "step": 31215 + }, + { + "epoch": 1.14, + "learning_rate": 8.586809243051232e-08, + "loss": 0.6951, + "step": 31216 + }, + { + "epoch": 1.14, + "learning_rate": 8.586443036584026e-08, + "loss": 0.6851, + "step": 31217 + }, + { + "epoch": 1.14, + "learning_rate": 8.58607683011682e-08, + "loss": 0.6941, + "step": 31218 + }, + { + "epoch": 1.14, + "learning_rate": 8.585710623649614e-08, + "loss": 0.6879, + "step": 31219 + }, + { + "epoch": 1.14, + "learning_rate": 8.585344417182408e-08, + "loss": 0.6895, + "step": 31220 + }, + { + "epoch": 1.14, + "learning_rate": 8.584978210715202e-08, + "loss": 0.6984, + "step": 31221 + }, + { + "epoch": 1.14, + "learning_rate": 8.584612004247994e-08, + "loss": 0.6787, + "step": 31222 + }, + { + "epoch": 1.14, + "learning_rate": 8.584245797780788e-08, + "loss": 0.684, + "step": 31223 + }, + { + "epoch": 1.14, + "learning_rate": 8.583879591313582e-08, + "loss": 0.6893, + "step": 31224 + }, + { + "epoch": 1.14, + "learning_rate": 8.583513384846376e-08, + "loss": 0.6958, + "step": 31225 + }, + { + "epoch": 1.14, + "learning_rate": 8.58314717837917e-08, + "loss": 0.6884, + "step": 31226 + }, + { + "epoch": 1.14, + "learning_rate": 8.582780971911963e-08, + "loss": 0.6997, + "step": 31227 + }, + { + "epoch": 1.14, + "learning_rate": 8.582414765444757e-08, + "loss": 0.693, + "step": 31228 + }, + { + "epoch": 1.14, + "learning_rate": 8.582048558977551e-08, + "loss": 0.6875, + "step": 31229 + }, + { + "epoch": 1.14, + "learning_rate": 8.581682352510345e-08, + "loss": 0.6996, + "step": 31230 + }, + { + "epoch": 1.14, + "learning_rate": 8.581316146043138e-08, + "loss": 0.6962, + "step": 31231 + }, + { + "epoch": 1.14, + "learning_rate": 8.580949939575932e-08, + "loss": 0.695, + "step": 31232 + }, + { + "epoch": 1.14, + "learning_rate": 8.580583733108726e-08, + "loss": 0.6961, + "step": 31233 + }, + { + "epoch": 1.14, + "learning_rate": 8.58021752664152e-08, + "loss": 0.6852, + "step": 31234 + }, + { + "epoch": 1.14, + "learning_rate": 8.579851320174314e-08, + "loss": 0.7002, + "step": 31235 + }, + { + "epoch": 1.14, + "learning_rate": 8.579485113707108e-08, + "loss": 0.7003, + "step": 31236 + }, + { + "epoch": 1.14, + "learning_rate": 8.579118907239902e-08, + "loss": 0.6918, + "step": 31237 + }, + { + "epoch": 1.14, + "learning_rate": 8.578752700772696e-08, + "loss": 0.7016, + "step": 31238 + }, + { + "epoch": 1.14, + "learning_rate": 8.57838649430549e-08, + "loss": 0.6919, + "step": 31239 + }, + { + "epoch": 1.14, + "learning_rate": 8.578020287838282e-08, + "loss": 0.6917, + "step": 31240 + }, + { + "epoch": 1.14, + "learning_rate": 8.577654081371076e-08, + "loss": 0.7, + "step": 31241 + }, + { + "epoch": 1.14, + "learning_rate": 8.57728787490387e-08, + "loss": 0.692, + "step": 31242 + }, + { + "epoch": 1.14, + "learning_rate": 8.576921668436664e-08, + "loss": 0.6918, + "step": 31243 + }, + { + "epoch": 1.14, + "learning_rate": 8.576555461969458e-08, + "loss": 0.6943, + "step": 31244 + }, + { + "epoch": 1.14, + "learning_rate": 8.576189255502252e-08, + "loss": 0.6939, + "step": 31245 + }, + { + "epoch": 1.14, + "learning_rate": 8.575823049035046e-08, + "loss": 0.6845, + "step": 31246 + }, + { + "epoch": 1.14, + "learning_rate": 8.57545684256784e-08, + "loss": 0.6971, + "step": 31247 + }, + { + "epoch": 1.14, + "learning_rate": 8.575090636100633e-08, + "loss": 0.695, + "step": 31248 + }, + { + "epoch": 1.14, + "learning_rate": 8.574724429633427e-08, + "loss": 0.6866, + "step": 31249 + }, + { + "epoch": 1.14, + "learning_rate": 8.574358223166221e-08, + "loss": 0.7138, + "step": 31250 + }, + { + "epoch": 1.14, + "learning_rate": 8.573992016699014e-08, + "loss": 0.6978, + "step": 31251 + }, + { + "epoch": 1.14, + "learning_rate": 8.573625810231808e-08, + "loss": 0.6819, + "step": 31252 + }, + { + "epoch": 1.14, + "learning_rate": 8.573259603764602e-08, + "loss": 0.683, + "step": 31253 + }, + { + "epoch": 1.14, + "learning_rate": 8.572893397297396e-08, + "loss": 0.6881, + "step": 31254 + }, + { + "epoch": 1.14, + "learning_rate": 8.57252719083019e-08, + "loss": 0.6996, + "step": 31255 + }, + { + "epoch": 1.14, + "learning_rate": 8.572160984362984e-08, + "loss": 0.6828, + "step": 31256 + }, + { + "epoch": 1.14, + "learning_rate": 8.571794777895778e-08, + "loss": 0.6849, + "step": 31257 + }, + { + "epoch": 1.14, + "learning_rate": 8.57142857142857e-08, + "loss": 0.7069, + "step": 31258 + }, + { + "epoch": 1.14, + "learning_rate": 8.571062364961364e-08, + "loss": 0.6882, + "step": 31259 + }, + { + "epoch": 1.14, + "learning_rate": 8.570696158494158e-08, + "loss": 0.7081, + "step": 31260 + }, + { + "epoch": 1.14, + "learning_rate": 8.570329952026952e-08, + "loss": 0.7019, + "step": 31261 + }, + { + "epoch": 1.14, + "learning_rate": 8.569963745559746e-08, + "loss": 0.6894, + "step": 31262 + }, + { + "epoch": 1.14, + "learning_rate": 8.56959753909254e-08, + "loss": 0.703, + "step": 31263 + }, + { + "epoch": 1.14, + "learning_rate": 8.569231332625334e-08, + "loss": 0.6904, + "step": 31264 + }, + { + "epoch": 1.14, + "learning_rate": 8.568865126158128e-08, + "loss": 0.688, + "step": 31265 + }, + { + "epoch": 1.14, + "learning_rate": 8.568498919690922e-08, + "loss": 0.6819, + "step": 31266 + }, + { + "epoch": 1.14, + "learning_rate": 8.568132713223715e-08, + "loss": 0.6981, + "step": 31267 + }, + { + "epoch": 1.14, + "learning_rate": 8.567766506756509e-08, + "loss": 0.6926, + "step": 31268 + }, + { + "epoch": 1.14, + "learning_rate": 8.567400300289303e-08, + "loss": 0.7031, + "step": 31269 + }, + { + "epoch": 1.14, + "learning_rate": 8.567034093822097e-08, + "loss": 0.705, + "step": 31270 + }, + { + "epoch": 1.14, + "learning_rate": 8.566667887354891e-08, + "loss": 0.6949, + "step": 31271 + }, + { + "epoch": 1.14, + "learning_rate": 8.566301680887685e-08, + "loss": 0.6986, + "step": 31272 + }, + { + "epoch": 1.14, + "learning_rate": 8.565935474420478e-08, + "loss": 0.6901, + "step": 31273 + }, + { + "epoch": 1.14, + "learning_rate": 8.565569267953272e-08, + "loss": 0.6942, + "step": 31274 + }, + { + "epoch": 1.14, + "learning_rate": 8.565203061486066e-08, + "loss": 0.6898, + "step": 31275 + }, + { + "epoch": 1.14, + "learning_rate": 8.564836855018858e-08, + "loss": 0.6977, + "step": 31276 + }, + { + "epoch": 1.14, + "learning_rate": 8.564470648551652e-08, + "loss": 0.7077, + "step": 31277 + }, + { + "epoch": 1.14, + "learning_rate": 8.564104442084446e-08, + "loss": 0.6903, + "step": 31278 + }, + { + "epoch": 1.14, + "learning_rate": 8.56373823561724e-08, + "loss": 0.6832, + "step": 31279 + }, + { + "epoch": 1.14, + "learning_rate": 8.563372029150034e-08, + "loss": 0.6794, + "step": 31280 + }, + { + "epoch": 1.14, + "learning_rate": 8.563005822682828e-08, + "loss": 0.689, + "step": 31281 + }, + { + "epoch": 1.14, + "learning_rate": 8.562639616215622e-08, + "loss": 0.693, + "step": 31282 + }, + { + "epoch": 1.14, + "learning_rate": 8.562273409748416e-08, + "loss": 0.6961, + "step": 31283 + }, + { + "epoch": 1.14, + "learning_rate": 8.56190720328121e-08, + "loss": 0.6888, + "step": 31284 + }, + { + "epoch": 1.14, + "learning_rate": 8.561540996814003e-08, + "loss": 0.6927, + "step": 31285 + }, + { + "epoch": 1.14, + "learning_rate": 8.561174790346797e-08, + "loss": 0.6989, + "step": 31286 + }, + { + "epoch": 1.14, + "learning_rate": 8.560808583879591e-08, + "loss": 0.6832, + "step": 31287 + }, + { + "epoch": 1.14, + "learning_rate": 8.560442377412385e-08, + "loss": 0.6921, + "step": 31288 + }, + { + "epoch": 1.14, + "learning_rate": 8.560076170945179e-08, + "loss": 0.6808, + "step": 31289 + }, + { + "epoch": 1.14, + "learning_rate": 8.559709964477973e-08, + "loss": 0.7123, + "step": 31290 + }, + { + "epoch": 1.14, + "learning_rate": 8.559343758010767e-08, + "loss": 0.691, + "step": 31291 + }, + { + "epoch": 1.14, + "learning_rate": 8.558977551543561e-08, + "loss": 0.6794, + "step": 31292 + }, + { + "epoch": 1.14, + "learning_rate": 8.558611345076355e-08, + "loss": 0.6914, + "step": 31293 + }, + { + "epoch": 1.14, + "learning_rate": 8.558245138609147e-08, + "loss": 0.6848, + "step": 31294 + }, + { + "epoch": 1.14, + "learning_rate": 8.557878932141941e-08, + "loss": 0.6981, + "step": 31295 + }, + { + "epoch": 1.15, + "learning_rate": 8.557512725674735e-08, + "loss": 0.6987, + "step": 31296 + }, + { + "epoch": 1.15, + "learning_rate": 8.557146519207528e-08, + "loss": 0.6962, + "step": 31297 + }, + { + "epoch": 1.15, + "learning_rate": 8.556780312740322e-08, + "loss": 0.6986, + "step": 31298 + }, + { + "epoch": 1.15, + "learning_rate": 8.556414106273116e-08, + "loss": 0.6988, + "step": 31299 + }, + { + "epoch": 1.15, + "learning_rate": 8.55604789980591e-08, + "loss": 0.7032, + "step": 31300 + }, + { + "epoch": 1.15, + "learning_rate": 8.555681693338704e-08, + "loss": 0.6989, + "step": 31301 + }, + { + "epoch": 1.15, + "learning_rate": 8.555315486871498e-08, + "loss": 0.6945, + "step": 31302 + }, + { + "epoch": 1.15, + "learning_rate": 8.554949280404291e-08, + "loss": 0.6886, + "step": 31303 + }, + { + "epoch": 1.15, + "learning_rate": 8.554583073937085e-08, + "loss": 0.6992, + "step": 31304 + }, + { + "epoch": 1.15, + "learning_rate": 8.554216867469879e-08, + "loss": 0.6978, + "step": 31305 + }, + { + "epoch": 1.15, + "learning_rate": 8.553850661002673e-08, + "loss": 0.6941, + "step": 31306 + }, + { + "epoch": 1.15, + "learning_rate": 8.553484454535467e-08, + "loss": 0.6958, + "step": 31307 + }, + { + "epoch": 1.15, + "learning_rate": 8.553118248068261e-08, + "loss": 0.6903, + "step": 31308 + }, + { + "epoch": 1.15, + "learning_rate": 8.552752041601055e-08, + "loss": 0.6741, + "step": 31309 + }, + { + "epoch": 1.15, + "learning_rate": 8.552385835133849e-08, + "loss": 0.6859, + "step": 31310 + }, + { + "epoch": 1.15, + "learning_rate": 8.552019628666643e-08, + "loss": 0.6896, + "step": 31311 + }, + { + "epoch": 1.15, + "learning_rate": 8.551653422199435e-08, + "loss": 0.6897, + "step": 31312 + }, + { + "epoch": 1.15, + "learning_rate": 8.551287215732229e-08, + "loss": 0.6967, + "step": 31313 + }, + { + "epoch": 1.15, + "learning_rate": 8.550921009265023e-08, + "loss": 0.6835, + "step": 31314 + }, + { + "epoch": 1.15, + "learning_rate": 8.550554802797817e-08, + "loss": 0.6898, + "step": 31315 + }, + { + "epoch": 1.15, + "learning_rate": 8.550188596330611e-08, + "loss": 0.6862, + "step": 31316 + }, + { + "epoch": 1.15, + "learning_rate": 8.549822389863405e-08, + "loss": 0.7009, + "step": 31317 + }, + { + "epoch": 1.15, + "learning_rate": 8.549456183396199e-08, + "loss": 0.7051, + "step": 31318 + }, + { + "epoch": 1.15, + "learning_rate": 8.549089976928992e-08, + "loss": 0.6861, + "step": 31319 + }, + { + "epoch": 1.15, + "learning_rate": 8.548723770461786e-08, + "loss": 0.6972, + "step": 31320 + }, + { + "epoch": 1.15, + "learning_rate": 8.54835756399458e-08, + "loss": 0.6785, + "step": 31321 + }, + { + "epoch": 1.15, + "learning_rate": 8.547991357527373e-08, + "loss": 0.6932, + "step": 31322 + }, + { + "epoch": 1.15, + "learning_rate": 8.547625151060167e-08, + "loss": 0.6948, + "step": 31323 + }, + { + "epoch": 1.15, + "learning_rate": 8.54725894459296e-08, + "loss": 0.6883, + "step": 31324 + }, + { + "epoch": 1.15, + "learning_rate": 8.546892738125755e-08, + "loss": 0.6873, + "step": 31325 + }, + { + "epoch": 1.15, + "learning_rate": 8.546526531658549e-08, + "loss": 0.6881, + "step": 31326 + }, + { + "epoch": 1.15, + "learning_rate": 8.546160325191343e-08, + "loss": 0.6973, + "step": 31327 + }, + { + "epoch": 1.15, + "learning_rate": 8.545794118724137e-08, + "loss": 0.691, + "step": 31328 + }, + { + "epoch": 1.15, + "learning_rate": 8.54542791225693e-08, + "loss": 0.6923, + "step": 31329 + }, + { + "epoch": 1.15, + "learning_rate": 8.545061705789723e-08, + "loss": 0.6861, + "step": 31330 + }, + { + "epoch": 1.15, + "learning_rate": 8.544695499322517e-08, + "loss": 0.6974, + "step": 31331 + }, + { + "epoch": 1.15, + "learning_rate": 8.544329292855311e-08, + "loss": 0.7021, + "step": 31332 + }, + { + "epoch": 1.15, + "learning_rate": 8.543963086388105e-08, + "loss": 0.682, + "step": 31333 + }, + { + "epoch": 1.15, + "learning_rate": 8.543596879920899e-08, + "loss": 0.6924, + "step": 31334 + }, + { + "epoch": 1.15, + "learning_rate": 8.543230673453693e-08, + "loss": 0.6925, + "step": 31335 + }, + { + "epoch": 1.15, + "learning_rate": 8.542864466986487e-08, + "loss": 0.6895, + "step": 31336 + }, + { + "epoch": 1.15, + "learning_rate": 8.542498260519281e-08, + "loss": 0.6903, + "step": 31337 + }, + { + "epoch": 1.15, + "learning_rate": 8.542132054052075e-08, + "loss": 0.7009, + "step": 31338 + }, + { + "epoch": 1.15, + "learning_rate": 8.541765847584868e-08, + "loss": 0.6856, + "step": 31339 + }, + { + "epoch": 1.15, + "learning_rate": 8.541399641117662e-08, + "loss": 0.6971, + "step": 31340 + }, + { + "epoch": 1.15, + "learning_rate": 8.541033434650456e-08, + "loss": 0.6996, + "step": 31341 + }, + { + "epoch": 1.15, + "learning_rate": 8.54066722818325e-08, + "loss": 0.694, + "step": 31342 + }, + { + "epoch": 1.15, + "learning_rate": 8.540301021716044e-08, + "loss": 0.7031, + "step": 31343 + }, + { + "epoch": 1.15, + "learning_rate": 8.539934815248836e-08, + "loss": 0.6894, + "step": 31344 + }, + { + "epoch": 1.15, + "learning_rate": 8.53956860878163e-08, + "loss": 0.6896, + "step": 31345 + }, + { + "epoch": 1.15, + "learning_rate": 8.539202402314424e-08, + "loss": 0.6962, + "step": 31346 + }, + { + "epoch": 1.15, + "learning_rate": 8.538836195847218e-08, + "loss": 0.6922, + "step": 31347 + }, + { + "epoch": 1.15, + "learning_rate": 8.538469989380011e-08, + "loss": 0.6766, + "step": 31348 + }, + { + "epoch": 1.15, + "learning_rate": 8.538103782912805e-08, + "loss": 0.6966, + "step": 31349 + }, + { + "epoch": 1.15, + "learning_rate": 8.537737576445599e-08, + "loss": 0.6895, + "step": 31350 + }, + { + "epoch": 1.15, + "learning_rate": 8.537371369978393e-08, + "loss": 0.6957, + "step": 31351 + }, + { + "epoch": 1.15, + "learning_rate": 8.537005163511187e-08, + "loss": 0.6849, + "step": 31352 + }, + { + "epoch": 1.15, + "learning_rate": 8.536638957043981e-08, + "loss": 0.6922, + "step": 31353 + }, + { + "epoch": 1.15, + "learning_rate": 8.536272750576775e-08, + "loss": 0.6997, + "step": 31354 + }, + { + "epoch": 1.15, + "learning_rate": 8.535906544109569e-08, + "loss": 0.68, + "step": 31355 + }, + { + "epoch": 1.15, + "learning_rate": 8.535540337642363e-08, + "loss": 0.7011, + "step": 31356 + }, + { + "epoch": 1.15, + "learning_rate": 8.535174131175156e-08, + "loss": 0.6979, + "step": 31357 + }, + { + "epoch": 1.15, + "learning_rate": 8.53480792470795e-08, + "loss": 0.7007, + "step": 31358 + }, + { + "epoch": 1.15, + "learning_rate": 8.534441718240744e-08, + "loss": 0.7024, + "step": 31359 + }, + { + "epoch": 1.15, + "learning_rate": 8.534075511773538e-08, + "loss": 0.6923, + "step": 31360 + }, + { + "epoch": 1.15, + "learning_rate": 8.533709305306332e-08, + "loss": 0.6799, + "step": 31361 + }, + { + "epoch": 1.15, + "learning_rate": 8.533343098839126e-08, + "loss": 0.7022, + "step": 31362 + }, + { + "epoch": 1.15, + "learning_rate": 8.53297689237192e-08, + "loss": 0.695, + "step": 31363 + }, + { + "epoch": 1.15, + "learning_rate": 8.532610685904714e-08, + "loss": 0.6936, + "step": 31364 + }, + { + "epoch": 1.15, + "learning_rate": 8.532244479437508e-08, + "loss": 0.6978, + "step": 31365 + }, + { + "epoch": 1.15, + "learning_rate": 8.5318782729703e-08, + "loss": 0.7022, + "step": 31366 + }, + { + "epoch": 1.15, + "learning_rate": 8.531512066503094e-08, + "loss": 0.7064, + "step": 31367 + }, + { + "epoch": 1.15, + "learning_rate": 8.531145860035887e-08, + "loss": 0.6917, + "step": 31368 + }, + { + "epoch": 1.15, + "learning_rate": 8.530779653568681e-08, + "loss": 0.6892, + "step": 31369 + }, + { + "epoch": 1.15, + "learning_rate": 8.530413447101475e-08, + "loss": 0.6882, + "step": 31370 + }, + { + "epoch": 1.15, + "learning_rate": 8.530047240634269e-08, + "loss": 0.7026, + "step": 31371 + }, + { + "epoch": 1.15, + "learning_rate": 8.529681034167063e-08, + "loss": 0.6858, + "step": 31372 + }, + { + "epoch": 1.15, + "learning_rate": 8.529314827699857e-08, + "loss": 0.6807, + "step": 31373 + }, + { + "epoch": 1.15, + "learning_rate": 8.528948621232651e-08, + "loss": 0.6998, + "step": 31374 + }, + { + "epoch": 1.15, + "learning_rate": 8.528582414765444e-08, + "loss": 0.7144, + "step": 31375 + }, + { + "epoch": 1.15, + "learning_rate": 8.528216208298238e-08, + "loss": 0.6974, + "step": 31376 + }, + { + "epoch": 1.15, + "learning_rate": 8.527850001831032e-08, + "loss": 0.6841, + "step": 31377 + }, + { + "epoch": 1.15, + "learning_rate": 8.527483795363826e-08, + "loss": 0.6877, + "step": 31378 + }, + { + "epoch": 1.15, + "learning_rate": 8.52711758889662e-08, + "loss": 0.7027, + "step": 31379 + }, + { + "epoch": 1.15, + "learning_rate": 8.526751382429414e-08, + "loss": 0.6918, + "step": 31380 + }, + { + "epoch": 1.15, + "learning_rate": 8.526385175962208e-08, + "loss": 0.7012, + "step": 31381 + }, + { + "epoch": 1.15, + "learning_rate": 8.526018969495002e-08, + "loss": 0.687, + "step": 31382 + }, + { + "epoch": 1.15, + "learning_rate": 8.525652763027796e-08, + "loss": 0.694, + "step": 31383 + }, + { + "epoch": 1.15, + "learning_rate": 8.525286556560588e-08, + "loss": 0.6957, + "step": 31384 + }, + { + "epoch": 1.15, + "learning_rate": 8.524920350093382e-08, + "loss": 0.6952, + "step": 31385 + }, + { + "epoch": 1.15, + "learning_rate": 8.524554143626176e-08, + "loss": 0.6931, + "step": 31386 + }, + { + "epoch": 1.15, + "learning_rate": 8.52418793715897e-08, + "loss": 0.6943, + "step": 31387 + }, + { + "epoch": 1.15, + "learning_rate": 8.523821730691764e-08, + "loss": 0.6896, + "step": 31388 + }, + { + "epoch": 1.15, + "learning_rate": 8.523455524224558e-08, + "loss": 0.6945, + "step": 31389 + }, + { + "epoch": 1.15, + "learning_rate": 8.523089317757351e-08, + "loss": 0.7006, + "step": 31390 + }, + { + "epoch": 1.15, + "learning_rate": 8.522723111290145e-08, + "loss": 0.7018, + "step": 31391 + }, + { + "epoch": 1.15, + "learning_rate": 8.522356904822939e-08, + "loss": 0.6832, + "step": 31392 + }, + { + "epoch": 1.15, + "learning_rate": 8.521990698355732e-08, + "loss": 0.6896, + "step": 31393 + }, + { + "epoch": 1.15, + "learning_rate": 8.521624491888526e-08, + "loss": 0.6919, + "step": 31394 + }, + { + "epoch": 1.15, + "learning_rate": 8.52125828542132e-08, + "loss": 0.6962, + "step": 31395 + }, + { + "epoch": 1.15, + "learning_rate": 8.520892078954114e-08, + "loss": 0.6946, + "step": 31396 + }, + { + "epoch": 1.15, + "learning_rate": 8.520525872486907e-08, + "loss": 0.6926, + "step": 31397 + }, + { + "epoch": 1.15, + "learning_rate": 8.520159666019701e-08, + "loss": 0.7001, + "step": 31398 + }, + { + "epoch": 1.15, + "learning_rate": 8.519793459552495e-08, + "loss": 0.6919, + "step": 31399 + }, + { + "epoch": 1.15, + "learning_rate": 8.51942725308529e-08, + "loss": 0.7056, + "step": 31400 + }, + { + "epoch": 1.15, + "learning_rate": 8.519061046618082e-08, + "loss": 0.6952, + "step": 31401 + }, + { + "epoch": 1.15, + "learning_rate": 8.518694840150876e-08, + "loss": 0.6988, + "step": 31402 + }, + { + "epoch": 1.15, + "learning_rate": 8.51832863368367e-08, + "loss": 0.6915, + "step": 31403 + }, + { + "epoch": 1.15, + "learning_rate": 8.517962427216464e-08, + "loss": 0.6906, + "step": 31404 + }, + { + "epoch": 1.15, + "learning_rate": 8.517596220749258e-08, + "loss": 0.7025, + "step": 31405 + }, + { + "epoch": 1.15, + "learning_rate": 8.517230014282052e-08, + "loss": 0.6905, + "step": 31406 + }, + { + "epoch": 1.15, + "learning_rate": 8.516863807814846e-08, + "loss": 0.6993, + "step": 31407 + }, + { + "epoch": 1.15, + "learning_rate": 8.51649760134764e-08, + "loss": 0.6967, + "step": 31408 + }, + { + "epoch": 1.15, + "learning_rate": 8.516131394880434e-08, + "loss": 0.6858, + "step": 31409 + }, + { + "epoch": 1.15, + "learning_rate": 8.515765188413227e-08, + "loss": 0.6815, + "step": 31410 + }, + { + "epoch": 1.15, + "learning_rate": 8.515398981946021e-08, + "loss": 0.6973, + "step": 31411 + }, + { + "epoch": 1.15, + "learning_rate": 8.515032775478815e-08, + "loss": 0.6971, + "step": 31412 + }, + { + "epoch": 1.15, + "learning_rate": 8.514666569011609e-08, + "loss": 0.6829, + "step": 31413 + }, + { + "epoch": 1.15, + "learning_rate": 8.514300362544403e-08, + "loss": 0.6963, + "step": 31414 + }, + { + "epoch": 1.15, + "learning_rate": 8.513934156077195e-08, + "loss": 0.7002, + "step": 31415 + }, + { + "epoch": 1.15, + "learning_rate": 8.51356794960999e-08, + "loss": 0.6828, + "step": 31416 + }, + { + "epoch": 1.15, + "learning_rate": 8.513201743142783e-08, + "loss": 0.6785, + "step": 31417 + }, + { + "epoch": 1.15, + "learning_rate": 8.512835536675577e-08, + "loss": 0.6899, + "step": 31418 + }, + { + "epoch": 1.15, + "learning_rate": 8.51246933020837e-08, + "loss": 0.6816, + "step": 31419 + }, + { + "epoch": 1.15, + "learning_rate": 8.512103123741164e-08, + "loss": 0.6836, + "step": 31420 + }, + { + "epoch": 1.15, + "learning_rate": 8.511736917273958e-08, + "loss": 0.694, + "step": 31421 + }, + { + "epoch": 1.15, + "learning_rate": 8.511370710806752e-08, + "loss": 0.6994, + "step": 31422 + }, + { + "epoch": 1.15, + "learning_rate": 8.511004504339546e-08, + "loss": 0.6963, + "step": 31423 + }, + { + "epoch": 1.15, + "learning_rate": 8.51063829787234e-08, + "loss": 0.6922, + "step": 31424 + }, + { + "epoch": 1.15, + "learning_rate": 8.510272091405134e-08, + "loss": 0.6735, + "step": 31425 + }, + { + "epoch": 1.15, + "learning_rate": 8.509905884937928e-08, + "loss": 0.6957, + "step": 31426 + }, + { + "epoch": 1.15, + "learning_rate": 8.509539678470722e-08, + "loss": 0.6938, + "step": 31427 + }, + { + "epoch": 1.15, + "learning_rate": 8.509173472003515e-08, + "loss": 0.6998, + "step": 31428 + }, + { + "epoch": 1.15, + "learning_rate": 8.508807265536309e-08, + "loss": 0.6963, + "step": 31429 + }, + { + "epoch": 1.15, + "learning_rate": 8.508441059069103e-08, + "loss": 0.6943, + "step": 31430 + }, + { + "epoch": 1.15, + "learning_rate": 8.508074852601897e-08, + "loss": 0.6883, + "step": 31431 + }, + { + "epoch": 1.15, + "learning_rate": 8.50770864613469e-08, + "loss": 0.7044, + "step": 31432 + }, + { + "epoch": 1.15, + "learning_rate": 8.507342439667485e-08, + "loss": 0.6924, + "step": 31433 + }, + { + "epoch": 1.15, + "learning_rate": 8.506976233200279e-08, + "loss": 0.6947, + "step": 31434 + }, + { + "epoch": 1.15, + "learning_rate": 8.506610026733073e-08, + "loss": 0.6864, + "step": 31435 + }, + { + "epoch": 1.15, + "learning_rate": 8.506243820265867e-08, + "loss": 0.6932, + "step": 31436 + }, + { + "epoch": 1.15, + "learning_rate": 8.505877613798659e-08, + "loss": 0.6752, + "step": 31437 + }, + { + "epoch": 1.15, + "learning_rate": 8.505511407331453e-08, + "loss": 0.6903, + "step": 31438 + }, + { + "epoch": 1.15, + "learning_rate": 8.505145200864246e-08, + "loss": 0.6902, + "step": 31439 + }, + { + "epoch": 1.15, + "learning_rate": 8.50477899439704e-08, + "loss": 0.6991, + "step": 31440 + }, + { + "epoch": 1.15, + "learning_rate": 8.504412787929834e-08, + "loss": 0.6915, + "step": 31441 + }, + { + "epoch": 1.15, + "learning_rate": 8.504046581462628e-08, + "loss": 0.6985, + "step": 31442 + }, + { + "epoch": 1.15, + "learning_rate": 8.503680374995422e-08, + "loss": 0.7032, + "step": 31443 + }, + { + "epoch": 1.15, + "learning_rate": 8.503314168528216e-08, + "loss": 0.7029, + "step": 31444 + }, + { + "epoch": 1.15, + "learning_rate": 8.50294796206101e-08, + "loss": 0.6936, + "step": 31445 + }, + { + "epoch": 1.15, + "learning_rate": 8.502581755593803e-08, + "loss": 0.691, + "step": 31446 + }, + { + "epoch": 1.15, + "learning_rate": 8.502215549126597e-08, + "loss": 0.6838, + "step": 31447 + }, + { + "epoch": 1.15, + "learning_rate": 8.50184934265939e-08, + "loss": 0.6861, + "step": 31448 + }, + { + "epoch": 1.15, + "learning_rate": 8.501483136192185e-08, + "loss": 0.6959, + "step": 31449 + }, + { + "epoch": 1.15, + "learning_rate": 8.501116929724979e-08, + "loss": 0.6907, + "step": 31450 + }, + { + "epoch": 1.15, + "learning_rate": 8.500750723257773e-08, + "loss": 0.6884, + "step": 31451 + }, + { + "epoch": 1.15, + "learning_rate": 8.500384516790567e-08, + "loss": 0.7031, + "step": 31452 + }, + { + "epoch": 1.15, + "learning_rate": 8.50001831032336e-08, + "loss": 0.6887, + "step": 31453 + }, + { + "epoch": 1.15, + "learning_rate": 8.499652103856155e-08, + "loss": 0.7184, + "step": 31454 + }, + { + "epoch": 1.15, + "learning_rate": 8.499285897388947e-08, + "loss": 0.6819, + "step": 31455 + }, + { + "epoch": 1.15, + "learning_rate": 8.498919690921741e-08, + "loss": 0.6931, + "step": 31456 + }, + { + "epoch": 1.15, + "learning_rate": 8.498553484454535e-08, + "loss": 0.7017, + "step": 31457 + }, + { + "epoch": 1.15, + "learning_rate": 8.498187277987329e-08, + "loss": 0.6917, + "step": 31458 + }, + { + "epoch": 1.15, + "learning_rate": 8.497821071520123e-08, + "loss": 0.6873, + "step": 31459 + }, + { + "epoch": 1.15, + "learning_rate": 8.497454865052917e-08, + "loss": 0.7062, + "step": 31460 + }, + { + "epoch": 1.15, + "learning_rate": 8.49708865858571e-08, + "loss": 0.6936, + "step": 31461 + }, + { + "epoch": 1.15, + "learning_rate": 8.496722452118504e-08, + "loss": 0.6959, + "step": 31462 + }, + { + "epoch": 1.15, + "learning_rate": 8.496356245651298e-08, + "loss": 0.6864, + "step": 31463 + }, + { + "epoch": 1.15, + "learning_rate": 8.49599003918409e-08, + "loss": 0.7112, + "step": 31464 + }, + { + "epoch": 1.15, + "learning_rate": 8.495623832716884e-08, + "loss": 0.6934, + "step": 31465 + }, + { + "epoch": 1.15, + "learning_rate": 8.495257626249678e-08, + "loss": 0.6929, + "step": 31466 + }, + { + "epoch": 1.15, + "learning_rate": 8.494891419782472e-08, + "loss": 0.6934, + "step": 31467 + }, + { + "epoch": 1.15, + "learning_rate": 8.494525213315266e-08, + "loss": 0.6912, + "step": 31468 + }, + { + "epoch": 1.15, + "learning_rate": 8.49415900684806e-08, + "loss": 0.6855, + "step": 31469 + }, + { + "epoch": 1.15, + "learning_rate": 8.493792800380854e-08, + "loss": 0.6943, + "step": 31470 + }, + { + "epoch": 1.15, + "learning_rate": 8.493426593913648e-08, + "loss": 0.6924, + "step": 31471 + }, + { + "epoch": 1.15, + "learning_rate": 8.493060387446442e-08, + "loss": 0.6953, + "step": 31472 + }, + { + "epoch": 1.15, + "learning_rate": 8.492694180979235e-08, + "loss": 0.6989, + "step": 31473 + }, + { + "epoch": 1.15, + "learning_rate": 8.492327974512029e-08, + "loss": 0.7051, + "step": 31474 + }, + { + "epoch": 1.15, + "learning_rate": 8.491961768044823e-08, + "loss": 0.7021, + "step": 31475 + }, + { + "epoch": 1.15, + "learning_rate": 8.491595561577617e-08, + "loss": 0.687, + "step": 31476 + }, + { + "epoch": 1.15, + "learning_rate": 8.491229355110411e-08, + "loss": 0.6915, + "step": 31477 + }, + { + "epoch": 1.15, + "learning_rate": 8.490863148643205e-08, + "loss": 0.6967, + "step": 31478 + }, + { + "epoch": 1.15, + "learning_rate": 8.490496942175999e-08, + "loss": 0.6745, + "step": 31479 + }, + { + "epoch": 1.15, + "learning_rate": 8.490130735708793e-08, + "loss": 0.7144, + "step": 31480 + }, + { + "epoch": 1.15, + "learning_rate": 8.489764529241587e-08, + "loss": 0.6985, + "step": 31481 + }, + { + "epoch": 1.15, + "learning_rate": 8.48939832277438e-08, + "loss": 0.6988, + "step": 31482 + }, + { + "epoch": 1.15, + "learning_rate": 8.489032116307174e-08, + "loss": 0.6996, + "step": 31483 + }, + { + "epoch": 1.15, + "learning_rate": 8.488665909839968e-08, + "loss": 0.6722, + "step": 31484 + }, + { + "epoch": 1.15, + "learning_rate": 8.488299703372762e-08, + "loss": 0.6859, + "step": 31485 + }, + { + "epoch": 1.15, + "learning_rate": 8.487933496905554e-08, + "loss": 0.6879, + "step": 31486 + }, + { + "epoch": 1.15, + "learning_rate": 8.487567290438348e-08, + "loss": 0.6917, + "step": 31487 + }, + { + "epoch": 1.15, + "learning_rate": 8.487201083971142e-08, + "loss": 0.6898, + "step": 31488 + }, + { + "epoch": 1.15, + "learning_rate": 8.486834877503936e-08, + "loss": 0.674, + "step": 31489 + }, + { + "epoch": 1.15, + "learning_rate": 8.48646867103673e-08, + "loss": 0.6988, + "step": 31490 + }, + { + "epoch": 1.15, + "learning_rate": 8.486102464569523e-08, + "loss": 0.6996, + "step": 31491 + }, + { + "epoch": 1.15, + "learning_rate": 8.485736258102317e-08, + "loss": 0.6999, + "step": 31492 + }, + { + "epoch": 1.15, + "learning_rate": 8.485370051635111e-08, + "loss": 0.6942, + "step": 31493 + }, + { + "epoch": 1.15, + "learning_rate": 8.485003845167905e-08, + "loss": 0.6999, + "step": 31494 + }, + { + "epoch": 1.15, + "learning_rate": 8.484637638700699e-08, + "loss": 0.703, + "step": 31495 + }, + { + "epoch": 1.15, + "learning_rate": 8.484271432233493e-08, + "loss": 0.696, + "step": 31496 + }, + { + "epoch": 1.15, + "learning_rate": 8.483905225766287e-08, + "loss": 0.6913, + "step": 31497 + }, + { + "epoch": 1.15, + "learning_rate": 8.483539019299081e-08, + "loss": 0.7046, + "step": 31498 + }, + { + "epoch": 1.15, + "learning_rate": 8.483172812831875e-08, + "loss": 0.6988, + "step": 31499 + }, + { + "epoch": 1.15, + "learning_rate": 8.482806606364668e-08, + "loss": 0.7033, + "step": 31500 + }, + { + "epoch": 1.15, + "learning_rate": 8.482440399897462e-08, + "loss": 0.6923, + "step": 31501 + }, + { + "epoch": 1.15, + "learning_rate": 8.482074193430256e-08, + "loss": 0.6981, + "step": 31502 + }, + { + "epoch": 1.15, + "learning_rate": 8.48170798696305e-08, + "loss": 0.6851, + "step": 31503 + }, + { + "epoch": 1.15, + "learning_rate": 8.481341780495844e-08, + "loss": 0.6819, + "step": 31504 + }, + { + "epoch": 1.15, + "learning_rate": 8.480975574028638e-08, + "loss": 0.6984, + "step": 31505 + }, + { + "epoch": 1.15, + "learning_rate": 8.480609367561432e-08, + "loss": 0.6963, + "step": 31506 + }, + { + "epoch": 1.15, + "learning_rate": 8.480243161094226e-08, + "loss": 0.6852, + "step": 31507 + }, + { + "epoch": 1.15, + "learning_rate": 8.479876954627018e-08, + "loss": 0.7135, + "step": 31508 + }, + { + "epoch": 1.15, + "learning_rate": 8.479510748159812e-08, + "loss": 0.7028, + "step": 31509 + }, + { + "epoch": 1.15, + "learning_rate": 8.479144541692605e-08, + "loss": 0.686, + "step": 31510 + }, + { + "epoch": 1.15, + "learning_rate": 8.478778335225399e-08, + "loss": 0.6862, + "step": 31511 + }, + { + "epoch": 1.15, + "learning_rate": 8.478412128758193e-08, + "loss": 0.6896, + "step": 31512 + }, + { + "epoch": 1.15, + "learning_rate": 8.478045922290987e-08, + "loss": 0.7089, + "step": 31513 + }, + { + "epoch": 1.15, + "learning_rate": 8.477679715823781e-08, + "loss": 0.7099, + "step": 31514 + }, + { + "epoch": 1.15, + "learning_rate": 8.477313509356575e-08, + "loss": 0.6799, + "step": 31515 + }, + { + "epoch": 1.15, + "learning_rate": 8.476947302889369e-08, + "loss": 0.6916, + "step": 31516 + }, + { + "epoch": 1.15, + "learning_rate": 8.476581096422163e-08, + "loss": 0.6941, + "step": 31517 + }, + { + "epoch": 1.15, + "learning_rate": 8.476214889954955e-08, + "loss": 0.6879, + "step": 31518 + }, + { + "epoch": 1.15, + "learning_rate": 8.47584868348775e-08, + "loss": 0.6857, + "step": 31519 + }, + { + "epoch": 1.15, + "learning_rate": 8.475482477020543e-08, + "loss": 0.6853, + "step": 31520 + }, + { + "epoch": 1.15, + "learning_rate": 8.475116270553337e-08, + "loss": 0.6961, + "step": 31521 + }, + { + "epoch": 1.15, + "learning_rate": 8.474750064086131e-08, + "loss": 0.6919, + "step": 31522 + }, + { + "epoch": 1.15, + "learning_rate": 8.474383857618925e-08, + "loss": 0.7063, + "step": 31523 + }, + { + "epoch": 1.15, + "learning_rate": 8.47401765115172e-08, + "loss": 0.6921, + "step": 31524 + }, + { + "epoch": 1.15, + "learning_rate": 8.473651444684513e-08, + "loss": 0.6892, + "step": 31525 + }, + { + "epoch": 1.15, + "learning_rate": 8.473285238217307e-08, + "loss": 0.6873, + "step": 31526 + }, + { + "epoch": 1.15, + "learning_rate": 8.4729190317501e-08, + "loss": 0.6945, + "step": 31527 + }, + { + "epoch": 1.15, + "learning_rate": 8.472552825282894e-08, + "loss": 0.6965, + "step": 31528 + }, + { + "epoch": 1.15, + "learning_rate": 8.472186618815688e-08, + "loss": 0.6978, + "step": 31529 + }, + { + "epoch": 1.15, + "learning_rate": 8.471820412348482e-08, + "loss": 0.7059, + "step": 31530 + }, + { + "epoch": 1.15, + "learning_rate": 8.471454205881276e-08, + "loss": 0.6976, + "step": 31531 + }, + { + "epoch": 1.15, + "learning_rate": 8.471087999414069e-08, + "loss": 0.7018, + "step": 31532 + }, + { + "epoch": 1.15, + "learning_rate": 8.470721792946863e-08, + "loss": 0.692, + "step": 31533 + }, + { + "epoch": 1.15, + "learning_rate": 8.470355586479657e-08, + "loss": 0.6934, + "step": 31534 + }, + { + "epoch": 1.15, + "learning_rate": 8.469989380012451e-08, + "loss": 0.6793, + "step": 31535 + }, + { + "epoch": 1.15, + "learning_rate": 8.469623173545243e-08, + "loss": 0.693, + "step": 31536 + }, + { + "epoch": 1.15, + "learning_rate": 8.469256967078037e-08, + "loss": 0.6984, + "step": 31537 + }, + { + "epoch": 1.15, + "learning_rate": 8.468890760610831e-08, + "loss": 0.6882, + "step": 31538 + }, + { + "epoch": 1.15, + "learning_rate": 8.468524554143625e-08, + "loss": 0.6921, + "step": 31539 + }, + { + "epoch": 1.15, + "learning_rate": 8.46815834767642e-08, + "loss": 0.6986, + "step": 31540 + }, + { + "epoch": 1.15, + "learning_rate": 8.467792141209213e-08, + "loss": 0.7005, + "step": 31541 + }, + { + "epoch": 1.15, + "learning_rate": 8.467425934742007e-08, + "loss": 0.6994, + "step": 31542 + }, + { + "epoch": 1.15, + "learning_rate": 8.467059728274801e-08, + "loss": 0.6908, + "step": 31543 + }, + { + "epoch": 1.15, + "learning_rate": 8.466693521807595e-08, + "loss": 0.6999, + "step": 31544 + }, + { + "epoch": 1.15, + "learning_rate": 8.466327315340388e-08, + "loss": 0.6969, + "step": 31545 + }, + { + "epoch": 1.15, + "learning_rate": 8.465961108873182e-08, + "loss": 0.6912, + "step": 31546 + }, + { + "epoch": 1.15, + "learning_rate": 8.465594902405976e-08, + "loss": 0.6914, + "step": 31547 + }, + { + "epoch": 1.15, + "learning_rate": 8.46522869593877e-08, + "loss": 0.6851, + "step": 31548 + }, + { + "epoch": 1.15, + "learning_rate": 8.464862489471564e-08, + "loss": 0.7017, + "step": 31549 + }, + { + "epoch": 1.15, + "learning_rate": 8.464496283004358e-08, + "loss": 0.6899, + "step": 31550 + }, + { + "epoch": 1.15, + "learning_rate": 8.464130076537152e-08, + "loss": 0.6812, + "step": 31551 + }, + { + "epoch": 1.15, + "learning_rate": 8.463763870069946e-08, + "loss": 0.6765, + "step": 31552 + }, + { + "epoch": 1.15, + "learning_rate": 8.46339766360274e-08, + "loss": 0.6887, + "step": 31553 + }, + { + "epoch": 1.15, + "learning_rate": 8.463031457135533e-08, + "loss": 0.6914, + "step": 31554 + }, + { + "epoch": 1.15, + "learning_rate": 8.462665250668327e-08, + "loss": 0.7063, + "step": 31555 + }, + { + "epoch": 1.15, + "learning_rate": 8.46229904420112e-08, + "loss": 0.6948, + "step": 31556 + }, + { + "epoch": 1.15, + "learning_rate": 8.461932837733913e-08, + "loss": 0.6892, + "step": 31557 + }, + { + "epoch": 1.15, + "learning_rate": 8.461566631266707e-08, + "loss": 0.6885, + "step": 31558 + }, + { + "epoch": 1.15, + "learning_rate": 8.461200424799501e-08, + "loss": 0.6924, + "step": 31559 + }, + { + "epoch": 1.15, + "learning_rate": 8.460834218332295e-08, + "loss": 0.6888, + "step": 31560 + }, + { + "epoch": 1.15, + "learning_rate": 8.460468011865089e-08, + "loss": 0.6955, + "step": 31561 + }, + { + "epoch": 1.15, + "learning_rate": 8.460101805397883e-08, + "loss": 0.6658, + "step": 31562 + }, + { + "epoch": 1.15, + "learning_rate": 8.459735598930676e-08, + "loss": 0.6956, + "step": 31563 + }, + { + "epoch": 1.15, + "learning_rate": 8.45936939246347e-08, + "loss": 0.7031, + "step": 31564 + }, + { + "epoch": 1.15, + "learning_rate": 8.459003185996264e-08, + "loss": 0.6991, + "step": 31565 + }, + { + "epoch": 1.15, + "learning_rate": 8.458636979529058e-08, + "loss": 0.6874, + "step": 31566 + }, + { + "epoch": 1.15, + "learning_rate": 8.458270773061852e-08, + "loss": 0.6923, + "step": 31567 + }, + { + "epoch": 1.15, + "learning_rate": 8.457904566594646e-08, + "loss": 0.6979, + "step": 31568 + }, + { + "epoch": 1.16, + "learning_rate": 8.45753836012744e-08, + "loss": 0.6799, + "step": 31569 + }, + { + "epoch": 1.16, + "learning_rate": 8.457172153660234e-08, + "loss": 0.6854, + "step": 31570 + }, + { + "epoch": 1.16, + "learning_rate": 8.456805947193028e-08, + "loss": 0.6929, + "step": 31571 + }, + { + "epoch": 1.16, + "learning_rate": 8.45643974072582e-08, + "loss": 0.6927, + "step": 31572 + }, + { + "epoch": 1.16, + "learning_rate": 8.456073534258615e-08, + "loss": 0.6979, + "step": 31573 + }, + { + "epoch": 1.16, + "learning_rate": 8.455707327791409e-08, + "loss": 0.6969, + "step": 31574 + }, + { + "epoch": 1.16, + "learning_rate": 8.455341121324202e-08, + "loss": 0.7004, + "step": 31575 + }, + { + "epoch": 1.16, + "learning_rate": 8.454974914856996e-08, + "loss": 0.692, + "step": 31576 + }, + { + "epoch": 1.16, + "learning_rate": 8.45460870838979e-08, + "loss": 0.6938, + "step": 31577 + }, + { + "epoch": 1.16, + "learning_rate": 8.454242501922584e-08, + "loss": 0.6989, + "step": 31578 + }, + { + "epoch": 1.16, + "learning_rate": 8.453876295455377e-08, + "loss": 0.6937, + "step": 31579 + }, + { + "epoch": 1.16, + "learning_rate": 8.453510088988171e-08, + "loss": 0.6889, + "step": 31580 + }, + { + "epoch": 1.16, + "learning_rate": 8.453143882520964e-08, + "loss": 0.6979, + "step": 31581 + }, + { + "epoch": 1.16, + "learning_rate": 8.452777676053758e-08, + "loss": 0.6918, + "step": 31582 + }, + { + "epoch": 1.16, + "learning_rate": 8.452411469586552e-08, + "loss": 0.6968, + "step": 31583 + }, + { + "epoch": 1.16, + "learning_rate": 8.452045263119346e-08, + "loss": 0.7064, + "step": 31584 + }, + { + "epoch": 1.16, + "learning_rate": 8.45167905665214e-08, + "loss": 0.6808, + "step": 31585 + }, + { + "epoch": 1.16, + "learning_rate": 8.451312850184934e-08, + "loss": 0.6991, + "step": 31586 + }, + { + "epoch": 1.16, + "learning_rate": 8.450946643717728e-08, + "loss": 0.6777, + "step": 31587 + }, + { + "epoch": 1.16, + "learning_rate": 8.450580437250522e-08, + "loss": 0.6929, + "step": 31588 + }, + { + "epoch": 1.16, + "learning_rate": 8.450214230783316e-08, + "loss": 0.6914, + "step": 31589 + }, + { + "epoch": 1.16, + "learning_rate": 8.449848024316108e-08, + "loss": 0.6913, + "step": 31590 + }, + { + "epoch": 1.16, + "learning_rate": 8.449481817848902e-08, + "loss": 0.6776, + "step": 31591 + }, + { + "epoch": 1.16, + "learning_rate": 8.449115611381696e-08, + "loss": 0.6937, + "step": 31592 + }, + { + "epoch": 1.16, + "learning_rate": 8.44874940491449e-08, + "loss": 0.6914, + "step": 31593 + }, + { + "epoch": 1.16, + "learning_rate": 8.448383198447284e-08, + "loss": 0.6827, + "step": 31594 + }, + { + "epoch": 1.16, + "learning_rate": 8.448016991980078e-08, + "loss": 0.7049, + "step": 31595 + }, + { + "epoch": 1.16, + "learning_rate": 8.447650785512872e-08, + "loss": 0.6973, + "step": 31596 + }, + { + "epoch": 1.16, + "learning_rate": 8.447284579045666e-08, + "loss": 0.6867, + "step": 31597 + }, + { + "epoch": 1.16, + "learning_rate": 8.44691837257846e-08, + "loss": 0.6867, + "step": 31598 + }, + { + "epoch": 1.16, + "learning_rate": 8.446552166111253e-08, + "loss": 0.6806, + "step": 31599 + }, + { + "epoch": 1.16, + "learning_rate": 8.446185959644047e-08, + "loss": 0.6889, + "step": 31600 + }, + { + "epoch": 1.16, + "learning_rate": 8.445819753176841e-08, + "loss": 0.6885, + "step": 31601 + }, + { + "epoch": 1.16, + "learning_rate": 8.445453546709635e-08, + "loss": 0.6958, + "step": 31602 + }, + { + "epoch": 1.16, + "learning_rate": 8.445087340242428e-08, + "loss": 0.6937, + "step": 31603 + }, + { + "epoch": 1.16, + "learning_rate": 8.444721133775222e-08, + "loss": 0.6924, + "step": 31604 + }, + { + "epoch": 1.16, + "learning_rate": 8.444354927308016e-08, + "loss": 0.6954, + "step": 31605 + }, + { + "epoch": 1.16, + "learning_rate": 8.44398872084081e-08, + "loss": 0.6964, + "step": 31606 + }, + { + "epoch": 1.16, + "learning_rate": 8.443622514373604e-08, + "loss": 0.7003, + "step": 31607 + }, + { + "epoch": 1.16, + "learning_rate": 8.443256307906396e-08, + "loss": 0.6865, + "step": 31608 + }, + { + "epoch": 1.16, + "learning_rate": 8.44289010143919e-08, + "loss": 0.6923, + "step": 31609 + }, + { + "epoch": 1.16, + "learning_rate": 8.442523894971984e-08, + "loss": 0.698, + "step": 31610 + }, + { + "epoch": 1.16, + "learning_rate": 8.442157688504778e-08, + "loss": 0.6763, + "step": 31611 + }, + { + "epoch": 1.16, + "learning_rate": 8.441791482037572e-08, + "loss": 0.6943, + "step": 31612 + }, + { + "epoch": 1.16, + "learning_rate": 8.441425275570366e-08, + "loss": 0.6792, + "step": 31613 + }, + { + "epoch": 1.16, + "learning_rate": 8.44105906910316e-08, + "loss": 0.696, + "step": 31614 + }, + { + "epoch": 1.16, + "learning_rate": 8.440692862635954e-08, + "loss": 0.6884, + "step": 31615 + }, + { + "epoch": 1.16, + "learning_rate": 8.440326656168748e-08, + "loss": 0.6889, + "step": 31616 + }, + { + "epoch": 1.16, + "learning_rate": 8.439960449701541e-08, + "loss": 0.6922, + "step": 31617 + }, + { + "epoch": 1.16, + "learning_rate": 8.439594243234335e-08, + "loss": 0.6972, + "step": 31618 + }, + { + "epoch": 1.16, + "learning_rate": 8.439228036767129e-08, + "loss": 0.6918, + "step": 31619 + }, + { + "epoch": 1.16, + "learning_rate": 8.438861830299923e-08, + "loss": 0.6942, + "step": 31620 + }, + { + "epoch": 1.16, + "learning_rate": 8.438495623832717e-08, + "loss": 0.6899, + "step": 31621 + }, + { + "epoch": 1.16, + "learning_rate": 8.438129417365511e-08, + "loss": 0.6923, + "step": 31622 + }, + { + "epoch": 1.16, + "learning_rate": 8.437763210898305e-08, + "loss": 0.6906, + "step": 31623 + }, + { + "epoch": 1.16, + "learning_rate": 8.437397004431099e-08, + "loss": 0.7027, + "step": 31624 + }, + { + "epoch": 1.16, + "learning_rate": 8.437030797963892e-08, + "loss": 0.7017, + "step": 31625 + }, + { + "epoch": 1.16, + "learning_rate": 8.436664591496686e-08, + "loss": 0.6711, + "step": 31626 + }, + { + "epoch": 1.16, + "learning_rate": 8.43629838502948e-08, + "loss": 0.6953, + "step": 31627 + }, + { + "epoch": 1.16, + "learning_rate": 8.435932178562272e-08, + "loss": 0.6932, + "step": 31628 + }, + { + "epoch": 1.16, + "learning_rate": 8.435565972095066e-08, + "loss": 0.6744, + "step": 31629 + }, + { + "epoch": 1.16, + "learning_rate": 8.43519976562786e-08, + "loss": 0.6831, + "step": 31630 + }, + { + "epoch": 1.16, + "learning_rate": 8.434833559160654e-08, + "loss": 0.7124, + "step": 31631 + }, + { + "epoch": 1.16, + "learning_rate": 8.434467352693448e-08, + "loss": 0.6941, + "step": 31632 + }, + { + "epoch": 1.16, + "learning_rate": 8.434101146226242e-08, + "loss": 0.6867, + "step": 31633 + }, + { + "epoch": 1.16, + "learning_rate": 8.433734939759036e-08, + "loss": 0.6792, + "step": 31634 + }, + { + "epoch": 1.16, + "learning_rate": 8.433368733291829e-08, + "loss": 0.6888, + "step": 31635 + }, + { + "epoch": 1.16, + "learning_rate": 8.433002526824623e-08, + "loss": 0.6913, + "step": 31636 + }, + { + "epoch": 1.16, + "learning_rate": 8.432636320357417e-08, + "loss": 0.6944, + "step": 31637 + }, + { + "epoch": 1.16, + "learning_rate": 8.432270113890211e-08, + "loss": 0.7013, + "step": 31638 + }, + { + "epoch": 1.16, + "learning_rate": 8.431903907423005e-08, + "loss": 0.6902, + "step": 31639 + }, + { + "epoch": 1.16, + "learning_rate": 8.431537700955799e-08, + "loss": 0.6985, + "step": 31640 + }, + { + "epoch": 1.16, + "learning_rate": 8.431171494488593e-08, + "loss": 0.6951, + "step": 31641 + }, + { + "epoch": 1.16, + "learning_rate": 8.430805288021387e-08, + "loss": 0.7029, + "step": 31642 + }, + { + "epoch": 1.16, + "learning_rate": 8.430439081554181e-08, + "loss": 0.6976, + "step": 31643 + }, + { + "epoch": 1.16, + "learning_rate": 8.430072875086973e-08, + "loss": 0.6893, + "step": 31644 + }, + { + "epoch": 1.16, + "learning_rate": 8.429706668619767e-08, + "loss": 0.6879, + "step": 31645 + }, + { + "epoch": 1.16, + "learning_rate": 8.429340462152561e-08, + "loss": 0.6892, + "step": 31646 + }, + { + "epoch": 1.16, + "learning_rate": 8.428974255685355e-08, + "loss": 0.6969, + "step": 31647 + }, + { + "epoch": 1.16, + "learning_rate": 8.42860804921815e-08, + "loss": 0.7073, + "step": 31648 + }, + { + "epoch": 1.16, + "learning_rate": 8.428241842750942e-08, + "loss": 0.7034, + "step": 31649 + }, + { + "epoch": 1.16, + "learning_rate": 8.427875636283736e-08, + "loss": 0.6981, + "step": 31650 + }, + { + "epoch": 1.16, + "learning_rate": 8.42750942981653e-08, + "loss": 0.7016, + "step": 31651 + }, + { + "epoch": 1.16, + "learning_rate": 8.427143223349324e-08, + "loss": 0.684, + "step": 31652 + }, + { + "epoch": 1.16, + "learning_rate": 8.426777016882117e-08, + "loss": 0.7035, + "step": 31653 + }, + { + "epoch": 1.16, + "learning_rate": 8.426410810414911e-08, + "loss": 0.6947, + "step": 31654 + }, + { + "epoch": 1.16, + "learning_rate": 8.426044603947705e-08, + "loss": 0.6929, + "step": 31655 + }, + { + "epoch": 1.16, + "learning_rate": 8.425678397480499e-08, + "loss": 0.6899, + "step": 31656 + }, + { + "epoch": 1.16, + "learning_rate": 8.425312191013293e-08, + "loss": 0.6883, + "step": 31657 + }, + { + "epoch": 1.16, + "learning_rate": 8.424945984546087e-08, + "loss": 0.7007, + "step": 31658 + }, + { + "epoch": 1.16, + "learning_rate": 8.424579778078881e-08, + "loss": 0.6991, + "step": 31659 + }, + { + "epoch": 1.16, + "learning_rate": 8.424213571611675e-08, + "loss": 0.6938, + "step": 31660 + }, + { + "epoch": 1.16, + "learning_rate": 8.423847365144469e-08, + "loss": 0.697, + "step": 31661 + }, + { + "epoch": 1.16, + "learning_rate": 8.423481158677261e-08, + "loss": 0.6884, + "step": 31662 + }, + { + "epoch": 1.16, + "learning_rate": 8.423114952210055e-08, + "loss": 0.6812, + "step": 31663 + }, + { + "epoch": 1.16, + "learning_rate": 8.422748745742849e-08, + "loss": 0.69, + "step": 31664 + }, + { + "epoch": 1.16, + "learning_rate": 8.422382539275643e-08, + "loss": 0.6801, + "step": 31665 + }, + { + "epoch": 1.16, + "learning_rate": 8.422016332808437e-08, + "loss": 0.6996, + "step": 31666 + }, + { + "epoch": 1.16, + "learning_rate": 8.421650126341231e-08, + "loss": 0.7117, + "step": 31667 + }, + { + "epoch": 1.16, + "learning_rate": 8.421283919874025e-08, + "loss": 0.6879, + "step": 31668 + }, + { + "epoch": 1.16, + "learning_rate": 8.420917713406819e-08, + "loss": 0.6935, + "step": 31669 + }, + { + "epoch": 1.16, + "learning_rate": 8.420551506939613e-08, + "loss": 0.6942, + "step": 31670 + }, + { + "epoch": 1.16, + "learning_rate": 8.420185300472406e-08, + "loss": 0.6739, + "step": 31671 + }, + { + "epoch": 1.16, + "learning_rate": 8.4198190940052e-08, + "loss": 0.7015, + "step": 31672 + }, + { + "epoch": 1.16, + "learning_rate": 8.419452887537994e-08, + "loss": 0.6888, + "step": 31673 + }, + { + "epoch": 1.16, + "learning_rate": 8.419086681070787e-08, + "loss": 0.6756, + "step": 31674 + }, + { + "epoch": 1.16, + "learning_rate": 8.41872047460358e-08, + "loss": 0.6886, + "step": 31675 + }, + { + "epoch": 1.16, + "learning_rate": 8.418354268136375e-08, + "loss": 0.703, + "step": 31676 + }, + { + "epoch": 1.16, + "learning_rate": 8.417988061669169e-08, + "loss": 0.6913, + "step": 31677 + }, + { + "epoch": 1.16, + "learning_rate": 8.417621855201963e-08, + "loss": 0.6969, + "step": 31678 + }, + { + "epoch": 1.16, + "learning_rate": 8.417255648734757e-08, + "loss": 0.686, + "step": 31679 + }, + { + "epoch": 1.16, + "learning_rate": 8.416889442267549e-08, + "loss": 0.6947, + "step": 31680 + }, + { + "epoch": 1.16, + "learning_rate": 8.416523235800343e-08, + "loss": 0.6854, + "step": 31681 + }, + { + "epoch": 1.16, + "learning_rate": 8.416157029333137e-08, + "loss": 0.6952, + "step": 31682 + }, + { + "epoch": 1.16, + "learning_rate": 8.415790822865931e-08, + "loss": 0.6914, + "step": 31683 + }, + { + "epoch": 1.16, + "learning_rate": 8.415424616398725e-08, + "loss": 0.6903, + "step": 31684 + }, + { + "epoch": 1.16, + "learning_rate": 8.415058409931519e-08, + "loss": 0.6912, + "step": 31685 + }, + { + "epoch": 1.16, + "learning_rate": 8.414692203464313e-08, + "loss": 0.6972, + "step": 31686 + }, + { + "epoch": 1.16, + "learning_rate": 8.414325996997107e-08, + "loss": 0.6896, + "step": 31687 + }, + { + "epoch": 1.16, + "learning_rate": 8.413959790529901e-08, + "loss": 0.6885, + "step": 31688 + }, + { + "epoch": 1.16, + "learning_rate": 8.413593584062694e-08, + "loss": 0.6918, + "step": 31689 + }, + { + "epoch": 1.16, + "learning_rate": 8.413227377595488e-08, + "loss": 0.6891, + "step": 31690 + }, + { + "epoch": 1.16, + "learning_rate": 8.412861171128282e-08, + "loss": 0.6749, + "step": 31691 + }, + { + "epoch": 1.16, + "learning_rate": 8.412494964661076e-08, + "loss": 0.6996, + "step": 31692 + }, + { + "epoch": 1.16, + "learning_rate": 8.41212875819387e-08, + "loss": 0.686, + "step": 31693 + }, + { + "epoch": 1.16, + "learning_rate": 8.411762551726664e-08, + "loss": 0.6903, + "step": 31694 + }, + { + "epoch": 1.16, + "learning_rate": 8.411396345259458e-08, + "loss": 0.6898, + "step": 31695 + }, + { + "epoch": 1.16, + "learning_rate": 8.41103013879225e-08, + "loss": 0.6968, + "step": 31696 + }, + { + "epoch": 1.16, + "learning_rate": 8.410663932325044e-08, + "loss": 0.6964, + "step": 31697 + }, + { + "epoch": 1.16, + "learning_rate": 8.410297725857838e-08, + "loss": 0.6988, + "step": 31698 + }, + { + "epoch": 1.16, + "learning_rate": 8.409931519390631e-08, + "loss": 0.6966, + "step": 31699 + }, + { + "epoch": 1.16, + "learning_rate": 8.409565312923425e-08, + "loss": 0.707, + "step": 31700 + }, + { + "epoch": 1.16, + "learning_rate": 8.409199106456219e-08, + "loss": 0.6935, + "step": 31701 + }, + { + "epoch": 1.16, + "learning_rate": 8.408832899989013e-08, + "loss": 0.6974, + "step": 31702 + }, + { + "epoch": 1.16, + "learning_rate": 8.408466693521807e-08, + "loss": 0.6795, + "step": 31703 + }, + { + "epoch": 1.16, + "learning_rate": 8.408100487054601e-08, + "loss": 0.6829, + "step": 31704 + }, + { + "epoch": 1.16, + "learning_rate": 8.407734280587395e-08, + "loss": 0.6985, + "step": 31705 + }, + { + "epoch": 1.16, + "learning_rate": 8.407368074120189e-08, + "loss": 0.6935, + "step": 31706 + }, + { + "epoch": 1.16, + "learning_rate": 8.407001867652982e-08, + "loss": 0.7031, + "step": 31707 + }, + { + "epoch": 1.16, + "learning_rate": 8.406635661185776e-08, + "loss": 0.684, + "step": 31708 + }, + { + "epoch": 1.16, + "learning_rate": 8.40626945471857e-08, + "loss": 0.6908, + "step": 31709 + }, + { + "epoch": 1.16, + "learning_rate": 8.405903248251364e-08, + "loss": 0.6998, + "step": 31710 + }, + { + "epoch": 1.16, + "learning_rate": 8.405537041784158e-08, + "loss": 0.6925, + "step": 31711 + }, + { + "epoch": 1.16, + "learning_rate": 8.405170835316952e-08, + "loss": 0.7048, + "step": 31712 + }, + { + "epoch": 1.16, + "learning_rate": 8.404804628849746e-08, + "loss": 0.6737, + "step": 31713 + }, + { + "epoch": 1.16, + "learning_rate": 8.40443842238254e-08, + "loss": 0.6984, + "step": 31714 + }, + { + "epoch": 1.16, + "learning_rate": 8.404072215915334e-08, + "loss": 0.6985, + "step": 31715 + }, + { + "epoch": 1.16, + "learning_rate": 8.403706009448126e-08, + "loss": 0.6892, + "step": 31716 + }, + { + "epoch": 1.16, + "learning_rate": 8.40333980298092e-08, + "loss": 0.6941, + "step": 31717 + }, + { + "epoch": 1.16, + "learning_rate": 8.402973596513714e-08, + "loss": 0.6854, + "step": 31718 + }, + { + "epoch": 1.16, + "learning_rate": 8.402607390046508e-08, + "loss": 0.6949, + "step": 31719 + }, + { + "epoch": 1.16, + "learning_rate": 8.402241183579301e-08, + "loss": 0.6939, + "step": 31720 + }, + { + "epoch": 1.16, + "learning_rate": 8.401874977112095e-08, + "loss": 0.6934, + "step": 31721 + }, + { + "epoch": 1.16, + "learning_rate": 8.401508770644889e-08, + "loss": 0.6943, + "step": 31722 + }, + { + "epoch": 1.16, + "learning_rate": 8.401142564177683e-08, + "loss": 0.701, + "step": 31723 + }, + { + "epoch": 1.16, + "learning_rate": 8.400776357710477e-08, + "loss": 0.6945, + "step": 31724 + }, + { + "epoch": 1.16, + "learning_rate": 8.40041015124327e-08, + "loss": 0.6981, + "step": 31725 + }, + { + "epoch": 1.16, + "learning_rate": 8.400043944776064e-08, + "loss": 0.6958, + "step": 31726 + }, + { + "epoch": 1.16, + "learning_rate": 8.399677738308858e-08, + "loss": 0.6923, + "step": 31727 + }, + { + "epoch": 1.16, + "learning_rate": 8.399311531841652e-08, + "loss": 0.6884, + "step": 31728 + }, + { + "epoch": 1.16, + "learning_rate": 8.398945325374446e-08, + "loss": 0.6902, + "step": 31729 + }, + { + "epoch": 1.16, + "learning_rate": 8.39857911890724e-08, + "loss": 0.6964, + "step": 31730 + }, + { + "epoch": 1.16, + "learning_rate": 8.398212912440034e-08, + "loss": 0.6933, + "step": 31731 + }, + { + "epoch": 1.16, + "learning_rate": 8.397846705972828e-08, + "loss": 0.6898, + "step": 31732 + }, + { + "epoch": 1.16, + "learning_rate": 8.397480499505622e-08, + "loss": 0.677, + "step": 31733 + }, + { + "epoch": 1.16, + "learning_rate": 8.397114293038414e-08, + "loss": 0.7064, + "step": 31734 + }, + { + "epoch": 1.16, + "learning_rate": 8.396748086571208e-08, + "loss": 0.6682, + "step": 31735 + }, + { + "epoch": 1.16, + "learning_rate": 8.396381880104002e-08, + "loss": 0.7033, + "step": 31736 + }, + { + "epoch": 1.16, + "learning_rate": 8.396015673636796e-08, + "loss": 0.7064, + "step": 31737 + }, + { + "epoch": 1.16, + "learning_rate": 8.39564946716959e-08, + "loss": 0.7012, + "step": 31738 + }, + { + "epoch": 1.16, + "learning_rate": 8.395283260702384e-08, + "loss": 0.6944, + "step": 31739 + }, + { + "epoch": 1.16, + "learning_rate": 8.394917054235178e-08, + "loss": 0.6902, + "step": 31740 + }, + { + "epoch": 1.16, + "learning_rate": 8.394550847767972e-08, + "loss": 0.6879, + "step": 31741 + }, + { + "epoch": 1.16, + "learning_rate": 8.394184641300765e-08, + "loss": 0.6963, + "step": 31742 + }, + { + "epoch": 1.16, + "learning_rate": 8.393818434833559e-08, + "loss": 0.692, + "step": 31743 + }, + { + "epoch": 1.16, + "learning_rate": 8.393452228366353e-08, + "loss": 0.6783, + "step": 31744 + }, + { + "epoch": 1.16, + "learning_rate": 8.393086021899146e-08, + "loss": 0.6878, + "step": 31745 + }, + { + "epoch": 1.16, + "learning_rate": 8.39271981543194e-08, + "loss": 0.6844, + "step": 31746 + }, + { + "epoch": 1.16, + "learning_rate": 8.392353608964734e-08, + "loss": 0.6923, + "step": 31747 + }, + { + "epoch": 1.16, + "learning_rate": 8.391987402497528e-08, + "loss": 0.6824, + "step": 31748 + }, + { + "epoch": 1.16, + "learning_rate": 8.391621196030322e-08, + "loss": 0.6842, + "step": 31749 + }, + { + "epoch": 1.16, + "learning_rate": 8.391254989563116e-08, + "loss": 0.6894, + "step": 31750 + }, + { + "epoch": 1.16, + "learning_rate": 8.39088878309591e-08, + "loss": 0.6946, + "step": 31751 + }, + { + "epoch": 1.16, + "learning_rate": 8.390522576628702e-08, + "loss": 0.6978, + "step": 31752 + }, + { + "epoch": 1.16, + "learning_rate": 8.390156370161496e-08, + "loss": 0.6864, + "step": 31753 + }, + { + "epoch": 1.16, + "learning_rate": 8.38979016369429e-08, + "loss": 0.6947, + "step": 31754 + }, + { + "epoch": 1.16, + "learning_rate": 8.389423957227084e-08, + "loss": 0.6977, + "step": 31755 + }, + { + "epoch": 1.16, + "learning_rate": 8.389057750759878e-08, + "loss": 0.7011, + "step": 31756 + }, + { + "epoch": 1.16, + "learning_rate": 8.388691544292672e-08, + "loss": 0.7064, + "step": 31757 + }, + { + "epoch": 1.16, + "learning_rate": 8.388325337825466e-08, + "loss": 0.6898, + "step": 31758 + }, + { + "epoch": 1.16, + "learning_rate": 8.38795913135826e-08, + "loss": 0.6851, + "step": 31759 + }, + { + "epoch": 1.16, + "learning_rate": 8.387592924891054e-08, + "loss": 0.6939, + "step": 31760 + }, + { + "epoch": 1.16, + "learning_rate": 8.387226718423847e-08, + "loss": 0.6883, + "step": 31761 + }, + { + "epoch": 1.16, + "learning_rate": 8.386860511956641e-08, + "loss": 0.6865, + "step": 31762 + }, + { + "epoch": 1.16, + "learning_rate": 8.386494305489435e-08, + "loss": 0.6901, + "step": 31763 + }, + { + "epoch": 1.16, + "learning_rate": 8.386128099022229e-08, + "loss": 0.6922, + "step": 31764 + }, + { + "epoch": 1.16, + "learning_rate": 8.385761892555023e-08, + "loss": 0.6948, + "step": 31765 + }, + { + "epoch": 1.16, + "learning_rate": 8.385395686087817e-08, + "loss": 0.7049, + "step": 31766 + }, + { + "epoch": 1.16, + "learning_rate": 8.38502947962061e-08, + "loss": 0.6878, + "step": 31767 + }, + { + "epoch": 1.16, + "learning_rate": 8.384663273153403e-08, + "loss": 0.7061, + "step": 31768 + }, + { + "epoch": 1.16, + "learning_rate": 8.384297066686197e-08, + "loss": 0.6883, + "step": 31769 + }, + { + "epoch": 1.16, + "learning_rate": 8.38393086021899e-08, + "loss": 0.6801, + "step": 31770 + }, + { + "epoch": 1.16, + "learning_rate": 8.383564653751784e-08, + "loss": 0.7016, + "step": 31771 + }, + { + "epoch": 1.16, + "learning_rate": 8.383198447284578e-08, + "loss": 0.6967, + "step": 31772 + }, + { + "epoch": 1.16, + "learning_rate": 8.382832240817372e-08, + "loss": 0.6975, + "step": 31773 + }, + { + "epoch": 1.16, + "learning_rate": 8.382466034350166e-08, + "loss": 0.6978, + "step": 31774 + }, + { + "epoch": 1.16, + "learning_rate": 8.38209982788296e-08, + "loss": 0.6789, + "step": 31775 + }, + { + "epoch": 1.16, + "learning_rate": 8.381733621415754e-08, + "loss": 0.695, + "step": 31776 + }, + { + "epoch": 1.16, + "learning_rate": 8.381367414948548e-08, + "loss": 0.6876, + "step": 31777 + }, + { + "epoch": 1.16, + "learning_rate": 8.381001208481342e-08, + "loss": 0.687, + "step": 31778 + }, + { + "epoch": 1.16, + "learning_rate": 8.380635002014135e-08, + "loss": 0.7001, + "step": 31779 + }, + { + "epoch": 1.16, + "learning_rate": 8.380268795546929e-08, + "loss": 0.69, + "step": 31780 + }, + { + "epoch": 1.16, + "learning_rate": 8.379902589079723e-08, + "loss": 0.7041, + "step": 31781 + }, + { + "epoch": 1.16, + "learning_rate": 8.379536382612517e-08, + "loss": 0.696, + "step": 31782 + }, + { + "epoch": 1.16, + "learning_rate": 8.37917017614531e-08, + "loss": 0.6972, + "step": 31783 + }, + { + "epoch": 1.16, + "learning_rate": 8.378803969678105e-08, + "loss": 0.6963, + "step": 31784 + }, + { + "epoch": 1.16, + "learning_rate": 8.378437763210899e-08, + "loss": 0.7053, + "step": 31785 + }, + { + "epoch": 1.16, + "learning_rate": 8.378071556743693e-08, + "loss": 0.7017, + "step": 31786 + }, + { + "epoch": 1.16, + "learning_rate": 8.377705350276485e-08, + "loss": 0.6881, + "step": 31787 + }, + { + "epoch": 1.16, + "learning_rate": 8.377339143809279e-08, + "loss": 0.6882, + "step": 31788 + }, + { + "epoch": 1.16, + "learning_rate": 8.376972937342073e-08, + "loss": 0.6788, + "step": 31789 + }, + { + "epoch": 1.16, + "learning_rate": 8.376606730874867e-08, + "loss": 0.6873, + "step": 31790 + }, + { + "epoch": 1.16, + "learning_rate": 8.37624052440766e-08, + "loss": 0.701, + "step": 31791 + }, + { + "epoch": 1.16, + "learning_rate": 8.375874317940454e-08, + "loss": 0.694, + "step": 31792 + }, + { + "epoch": 1.16, + "learning_rate": 8.375508111473248e-08, + "loss": 0.6945, + "step": 31793 + }, + { + "epoch": 1.16, + "learning_rate": 8.375141905006042e-08, + "loss": 0.6877, + "step": 31794 + }, + { + "epoch": 1.16, + "learning_rate": 8.374775698538836e-08, + "loss": 0.6927, + "step": 31795 + }, + { + "epoch": 1.16, + "learning_rate": 8.374409492071629e-08, + "loss": 0.6997, + "step": 31796 + }, + { + "epoch": 1.16, + "learning_rate": 8.374043285604423e-08, + "loss": 0.6909, + "step": 31797 + }, + { + "epoch": 1.16, + "learning_rate": 8.373677079137217e-08, + "loss": 0.7001, + "step": 31798 + }, + { + "epoch": 1.16, + "learning_rate": 8.37331087267001e-08, + "loss": 0.694, + "step": 31799 + }, + { + "epoch": 1.16, + "learning_rate": 8.372944666202805e-08, + "loss": 0.6904, + "step": 31800 + }, + { + "epoch": 1.16, + "learning_rate": 8.372578459735599e-08, + "loss": 0.6911, + "step": 31801 + }, + { + "epoch": 1.16, + "learning_rate": 8.372212253268393e-08, + "loss": 0.6943, + "step": 31802 + }, + { + "epoch": 1.16, + "learning_rate": 8.371846046801187e-08, + "loss": 0.6925, + "step": 31803 + }, + { + "epoch": 1.16, + "learning_rate": 8.37147984033398e-08, + "loss": 0.6853, + "step": 31804 + }, + { + "epoch": 1.16, + "learning_rate": 8.371113633866773e-08, + "loss": 0.6964, + "step": 31805 + }, + { + "epoch": 1.16, + "learning_rate": 8.370747427399567e-08, + "loss": 0.681, + "step": 31806 + }, + { + "epoch": 1.16, + "learning_rate": 8.370381220932361e-08, + "loss": 0.691, + "step": 31807 + }, + { + "epoch": 1.16, + "learning_rate": 8.370015014465155e-08, + "loss": 0.6811, + "step": 31808 + }, + { + "epoch": 1.16, + "learning_rate": 8.369648807997949e-08, + "loss": 0.696, + "step": 31809 + }, + { + "epoch": 1.16, + "learning_rate": 8.369282601530743e-08, + "loss": 0.6914, + "step": 31810 + }, + { + "epoch": 1.16, + "learning_rate": 8.368916395063537e-08, + "loss": 0.6939, + "step": 31811 + }, + { + "epoch": 1.16, + "learning_rate": 8.368550188596331e-08, + "loss": 0.6846, + "step": 31812 + }, + { + "epoch": 1.16, + "learning_rate": 8.368183982129124e-08, + "loss": 0.6787, + "step": 31813 + }, + { + "epoch": 1.16, + "learning_rate": 8.367817775661918e-08, + "loss": 0.6812, + "step": 31814 + }, + { + "epoch": 1.16, + "learning_rate": 8.367451569194712e-08, + "loss": 0.6919, + "step": 31815 + }, + { + "epoch": 1.16, + "learning_rate": 8.367085362727504e-08, + "loss": 0.682, + "step": 31816 + }, + { + "epoch": 1.16, + "learning_rate": 8.366719156260298e-08, + "loss": 0.7052, + "step": 31817 + }, + { + "epoch": 1.16, + "learning_rate": 8.366352949793092e-08, + "loss": 0.6909, + "step": 31818 + }, + { + "epoch": 1.16, + "learning_rate": 8.365986743325886e-08, + "loss": 0.701, + "step": 31819 + }, + { + "epoch": 1.16, + "learning_rate": 8.36562053685868e-08, + "loss": 0.7116, + "step": 31820 + }, + { + "epoch": 1.16, + "learning_rate": 8.365254330391474e-08, + "loss": 0.7065, + "step": 31821 + }, + { + "epoch": 1.16, + "learning_rate": 8.364888123924268e-08, + "loss": 0.6868, + "step": 31822 + }, + { + "epoch": 1.16, + "learning_rate": 8.364521917457061e-08, + "loss": 0.6864, + "step": 31823 + }, + { + "epoch": 1.16, + "learning_rate": 8.364155710989855e-08, + "loss": 0.6938, + "step": 31824 + }, + { + "epoch": 1.16, + "learning_rate": 8.363789504522649e-08, + "loss": 0.6857, + "step": 31825 + }, + { + "epoch": 1.16, + "learning_rate": 8.363423298055443e-08, + "loss": 0.6995, + "step": 31826 + }, + { + "epoch": 1.16, + "learning_rate": 8.363057091588237e-08, + "loss": 0.6964, + "step": 31827 + }, + { + "epoch": 1.16, + "learning_rate": 8.362690885121031e-08, + "loss": 0.6953, + "step": 31828 + }, + { + "epoch": 1.16, + "learning_rate": 8.362324678653825e-08, + "loss": 0.6865, + "step": 31829 + }, + { + "epoch": 1.16, + "learning_rate": 8.361958472186619e-08, + "loss": 0.6919, + "step": 31830 + }, + { + "epoch": 1.16, + "learning_rate": 8.361592265719413e-08, + "loss": 0.7029, + "step": 31831 + }, + { + "epoch": 1.16, + "learning_rate": 8.361226059252206e-08, + "loss": 0.6926, + "step": 31832 + }, + { + "epoch": 1.16, + "learning_rate": 8.360859852785e-08, + "loss": 0.6948, + "step": 31833 + }, + { + "epoch": 1.16, + "learning_rate": 8.360493646317794e-08, + "loss": 0.6931, + "step": 31834 + }, + { + "epoch": 1.16, + "learning_rate": 8.360127439850588e-08, + "loss": 0.6866, + "step": 31835 + }, + { + "epoch": 1.16, + "learning_rate": 8.359761233383382e-08, + "loss": 0.6928, + "step": 31836 + }, + { + "epoch": 1.16, + "learning_rate": 8.359395026916176e-08, + "loss": 0.6981, + "step": 31837 + }, + { + "epoch": 1.16, + "learning_rate": 8.359028820448968e-08, + "loss": 0.6993, + "step": 31838 + }, + { + "epoch": 1.16, + "learning_rate": 8.358662613981762e-08, + "loss": 0.6938, + "step": 31839 + }, + { + "epoch": 1.16, + "learning_rate": 8.358296407514556e-08, + "loss": 0.6902, + "step": 31840 + }, + { + "epoch": 1.16, + "learning_rate": 8.357930201047349e-08, + "loss": 0.6971, + "step": 31841 + }, + { + "epoch": 1.17, + "learning_rate": 8.357563994580143e-08, + "loss": 0.7015, + "step": 31842 + }, + { + "epoch": 1.17, + "learning_rate": 8.357197788112937e-08, + "loss": 0.6926, + "step": 31843 + }, + { + "epoch": 1.17, + "learning_rate": 8.356831581645731e-08, + "loss": 0.6997, + "step": 31844 + }, + { + "epoch": 1.17, + "learning_rate": 8.356465375178525e-08, + "loss": 0.692, + "step": 31845 + }, + { + "epoch": 1.17, + "learning_rate": 8.356099168711319e-08, + "loss": 0.69, + "step": 31846 + }, + { + "epoch": 1.17, + "learning_rate": 8.355732962244113e-08, + "loss": 0.6882, + "step": 31847 + }, + { + "epoch": 1.17, + "learning_rate": 8.355366755776907e-08, + "loss": 0.6865, + "step": 31848 + }, + { + "epoch": 1.17, + "learning_rate": 8.355000549309701e-08, + "loss": 0.7092, + "step": 31849 + }, + { + "epoch": 1.17, + "learning_rate": 8.354634342842494e-08, + "loss": 0.686, + "step": 31850 + }, + { + "epoch": 1.17, + "learning_rate": 8.354268136375288e-08, + "loss": 0.7043, + "step": 31851 + }, + { + "epoch": 1.17, + "learning_rate": 8.353901929908082e-08, + "loss": 0.7007, + "step": 31852 + }, + { + "epoch": 1.17, + "learning_rate": 8.353535723440876e-08, + "loss": 0.6894, + "step": 31853 + }, + { + "epoch": 1.17, + "learning_rate": 8.35316951697367e-08, + "loss": 0.7049, + "step": 31854 + }, + { + "epoch": 1.17, + "learning_rate": 8.352803310506464e-08, + "loss": 0.6969, + "step": 31855 + }, + { + "epoch": 1.17, + "learning_rate": 8.352437104039258e-08, + "loss": 0.7007, + "step": 31856 + }, + { + "epoch": 1.17, + "learning_rate": 8.352070897572052e-08, + "loss": 0.7005, + "step": 31857 + }, + { + "epoch": 1.17, + "learning_rate": 8.351704691104846e-08, + "loss": 0.6972, + "step": 31858 + }, + { + "epoch": 1.17, + "learning_rate": 8.351338484637638e-08, + "loss": 0.6705, + "step": 31859 + }, + { + "epoch": 1.17, + "learning_rate": 8.350972278170432e-08, + "loss": 0.6985, + "step": 31860 + }, + { + "epoch": 1.17, + "learning_rate": 8.350606071703226e-08, + "loss": 0.6807, + "step": 31861 + }, + { + "epoch": 1.17, + "learning_rate": 8.350239865236019e-08, + "loss": 0.6973, + "step": 31862 + }, + { + "epoch": 1.17, + "learning_rate": 8.349873658768813e-08, + "loss": 0.689, + "step": 31863 + }, + { + "epoch": 1.17, + "learning_rate": 8.349507452301607e-08, + "loss": 0.7079, + "step": 31864 + }, + { + "epoch": 1.17, + "learning_rate": 8.349141245834401e-08, + "loss": 0.6714, + "step": 31865 + }, + { + "epoch": 1.17, + "learning_rate": 8.348775039367195e-08, + "loss": 0.707, + "step": 31866 + }, + { + "epoch": 1.17, + "learning_rate": 8.348408832899989e-08, + "loss": 0.687, + "step": 31867 + }, + { + "epoch": 1.17, + "learning_rate": 8.348042626432782e-08, + "loss": 0.6966, + "step": 31868 + }, + { + "epoch": 1.17, + "learning_rate": 8.347676419965576e-08, + "loss": 0.7023, + "step": 31869 + }, + { + "epoch": 1.17, + "learning_rate": 8.34731021349837e-08, + "loss": 0.7027, + "step": 31870 + }, + { + "epoch": 1.17, + "learning_rate": 8.346944007031164e-08, + "loss": 0.6947, + "step": 31871 + }, + { + "epoch": 1.17, + "learning_rate": 8.346577800563957e-08, + "loss": 0.6921, + "step": 31872 + }, + { + "epoch": 1.17, + "learning_rate": 8.346211594096751e-08, + "loss": 0.691, + "step": 31873 + }, + { + "epoch": 1.17, + "learning_rate": 8.345845387629545e-08, + "loss": 0.7057, + "step": 31874 + }, + { + "epoch": 1.17, + "learning_rate": 8.34547918116234e-08, + "loss": 0.6806, + "step": 31875 + }, + { + "epoch": 1.17, + "learning_rate": 8.345112974695133e-08, + "loss": 0.6769, + "step": 31876 + }, + { + "epoch": 1.17, + "learning_rate": 8.344746768227926e-08, + "loss": 0.6915, + "step": 31877 + }, + { + "epoch": 1.17, + "learning_rate": 8.34438056176072e-08, + "loss": 0.6909, + "step": 31878 + }, + { + "epoch": 1.17, + "learning_rate": 8.344014355293514e-08, + "loss": 0.6921, + "step": 31879 + }, + { + "epoch": 1.17, + "learning_rate": 8.343648148826308e-08, + "loss": 0.6899, + "step": 31880 + }, + { + "epoch": 1.17, + "learning_rate": 8.343281942359102e-08, + "loss": 0.692, + "step": 31881 + }, + { + "epoch": 1.17, + "learning_rate": 8.342915735891896e-08, + "loss": 0.6968, + "step": 31882 + }, + { + "epoch": 1.17, + "learning_rate": 8.34254952942469e-08, + "loss": 0.6833, + "step": 31883 + }, + { + "epoch": 1.17, + "learning_rate": 8.342183322957483e-08, + "loss": 0.6909, + "step": 31884 + }, + { + "epoch": 1.17, + "learning_rate": 8.341817116490277e-08, + "loss": 0.6961, + "step": 31885 + }, + { + "epoch": 1.17, + "learning_rate": 8.341450910023071e-08, + "loss": 0.7025, + "step": 31886 + }, + { + "epoch": 1.17, + "learning_rate": 8.341084703555863e-08, + "loss": 0.6879, + "step": 31887 + }, + { + "epoch": 1.17, + "learning_rate": 8.340718497088657e-08, + "loss": 0.6791, + "step": 31888 + }, + { + "epoch": 1.17, + "learning_rate": 8.340352290621451e-08, + "loss": 0.6697, + "step": 31889 + }, + { + "epoch": 1.17, + "learning_rate": 8.339986084154245e-08, + "loss": 0.6934, + "step": 31890 + }, + { + "epoch": 1.17, + "learning_rate": 8.33961987768704e-08, + "loss": 0.6958, + "step": 31891 + }, + { + "epoch": 1.17, + "learning_rate": 8.339253671219833e-08, + "loss": 0.6911, + "step": 31892 + }, + { + "epoch": 1.17, + "learning_rate": 8.338887464752627e-08, + "loss": 0.7063, + "step": 31893 + }, + { + "epoch": 1.17, + "learning_rate": 8.338521258285421e-08, + "loss": 0.6925, + "step": 31894 + }, + { + "epoch": 1.17, + "learning_rate": 8.338155051818214e-08, + "loss": 0.6935, + "step": 31895 + }, + { + "epoch": 1.17, + "learning_rate": 8.337788845351008e-08, + "loss": 0.6843, + "step": 31896 + }, + { + "epoch": 1.17, + "learning_rate": 8.337422638883802e-08, + "loss": 0.6935, + "step": 31897 + }, + { + "epoch": 1.17, + "learning_rate": 8.337056432416596e-08, + "loss": 0.682, + "step": 31898 + }, + { + "epoch": 1.17, + "learning_rate": 8.33669022594939e-08, + "loss": 0.6878, + "step": 31899 + }, + { + "epoch": 1.17, + "learning_rate": 8.336324019482184e-08, + "loss": 0.6996, + "step": 31900 + }, + { + "epoch": 1.17, + "learning_rate": 8.335957813014978e-08, + "loss": 0.7046, + "step": 31901 + }, + { + "epoch": 1.17, + "learning_rate": 8.335591606547772e-08, + "loss": 0.6853, + "step": 31902 + }, + { + "epoch": 1.17, + "learning_rate": 8.335225400080566e-08, + "loss": 0.7036, + "step": 31903 + }, + { + "epoch": 1.17, + "learning_rate": 8.334859193613359e-08, + "loss": 0.6966, + "step": 31904 + }, + { + "epoch": 1.17, + "learning_rate": 8.334492987146153e-08, + "loss": 0.7013, + "step": 31905 + }, + { + "epoch": 1.17, + "learning_rate": 8.334126780678947e-08, + "loss": 0.7071, + "step": 31906 + }, + { + "epoch": 1.17, + "learning_rate": 8.33376057421174e-08, + "loss": 0.6853, + "step": 31907 + }, + { + "epoch": 1.17, + "learning_rate": 8.333394367744535e-08, + "loss": 0.6963, + "step": 31908 + }, + { + "epoch": 1.17, + "learning_rate": 8.333028161277327e-08, + "loss": 0.7026, + "step": 31909 + }, + { + "epoch": 1.17, + "learning_rate": 8.332661954810121e-08, + "loss": 0.6911, + "step": 31910 + }, + { + "epoch": 1.17, + "learning_rate": 8.332295748342915e-08, + "loss": 0.6886, + "step": 31911 + }, + { + "epoch": 1.17, + "learning_rate": 8.331929541875709e-08, + "loss": 0.6929, + "step": 31912 + }, + { + "epoch": 1.17, + "learning_rate": 8.331563335408502e-08, + "loss": 0.6844, + "step": 31913 + }, + { + "epoch": 1.17, + "learning_rate": 8.331197128941296e-08, + "loss": 0.6896, + "step": 31914 + }, + { + "epoch": 1.17, + "learning_rate": 8.33083092247409e-08, + "loss": 0.6857, + "step": 31915 + }, + { + "epoch": 1.17, + "learning_rate": 8.330464716006884e-08, + "loss": 0.7053, + "step": 31916 + }, + { + "epoch": 1.17, + "learning_rate": 8.330098509539678e-08, + "loss": 0.6851, + "step": 31917 + }, + { + "epoch": 1.17, + "learning_rate": 8.329732303072472e-08, + "loss": 0.6879, + "step": 31918 + }, + { + "epoch": 1.17, + "learning_rate": 8.329366096605266e-08, + "loss": 0.7156, + "step": 31919 + }, + { + "epoch": 1.17, + "learning_rate": 8.32899989013806e-08, + "loss": 0.7129, + "step": 31920 + }, + { + "epoch": 1.17, + "learning_rate": 8.328633683670854e-08, + "loss": 0.6737, + "step": 31921 + }, + { + "epoch": 1.17, + "learning_rate": 8.328267477203647e-08, + "loss": 0.6943, + "step": 31922 + }, + { + "epoch": 1.17, + "learning_rate": 8.32790127073644e-08, + "loss": 0.6998, + "step": 31923 + }, + { + "epoch": 1.17, + "learning_rate": 8.327535064269235e-08, + "loss": 0.6971, + "step": 31924 + }, + { + "epoch": 1.17, + "learning_rate": 8.327168857802029e-08, + "loss": 0.6933, + "step": 31925 + }, + { + "epoch": 1.17, + "learning_rate": 8.326802651334823e-08, + "loss": 0.6977, + "step": 31926 + }, + { + "epoch": 1.17, + "learning_rate": 8.326436444867617e-08, + "loss": 0.6797, + "step": 31927 + }, + { + "epoch": 1.17, + "learning_rate": 8.32607023840041e-08, + "loss": 0.685, + "step": 31928 + }, + { + "epoch": 1.17, + "learning_rate": 8.325704031933204e-08, + "loss": 0.6959, + "step": 31929 + }, + { + "epoch": 1.17, + "learning_rate": 8.325337825465998e-08, + "loss": 0.7028, + "step": 31930 + }, + { + "epoch": 1.17, + "learning_rate": 8.324971618998791e-08, + "loss": 0.7068, + "step": 31931 + }, + { + "epoch": 1.17, + "learning_rate": 8.324605412531585e-08, + "loss": 0.695, + "step": 31932 + }, + { + "epoch": 1.17, + "learning_rate": 8.324239206064378e-08, + "loss": 0.7025, + "step": 31933 + }, + { + "epoch": 1.17, + "learning_rate": 8.323872999597172e-08, + "loss": 0.6913, + "step": 31934 + }, + { + "epoch": 1.17, + "learning_rate": 8.323506793129966e-08, + "loss": 0.6983, + "step": 31935 + }, + { + "epoch": 1.17, + "learning_rate": 8.32314058666276e-08, + "loss": 0.695, + "step": 31936 + }, + { + "epoch": 1.17, + "learning_rate": 8.322774380195554e-08, + "loss": 0.6894, + "step": 31937 + }, + { + "epoch": 1.17, + "learning_rate": 8.322408173728348e-08, + "loss": 0.7003, + "step": 31938 + }, + { + "epoch": 1.17, + "learning_rate": 8.322041967261142e-08, + "loss": 0.7062, + "step": 31939 + }, + { + "epoch": 1.17, + "learning_rate": 8.321675760793934e-08, + "loss": 0.6979, + "step": 31940 + }, + { + "epoch": 1.17, + "learning_rate": 8.321309554326728e-08, + "loss": 0.6945, + "step": 31941 + }, + { + "epoch": 1.17, + "learning_rate": 8.320943347859522e-08, + "loss": 0.6909, + "step": 31942 + }, + { + "epoch": 1.17, + "learning_rate": 8.320577141392316e-08, + "loss": 0.6833, + "step": 31943 + }, + { + "epoch": 1.17, + "learning_rate": 8.32021093492511e-08, + "loss": 0.6943, + "step": 31944 + }, + { + "epoch": 1.17, + "learning_rate": 8.319844728457904e-08, + "loss": 0.6976, + "step": 31945 + }, + { + "epoch": 1.17, + "learning_rate": 8.319478521990698e-08, + "loss": 0.6816, + "step": 31946 + }, + { + "epoch": 1.17, + "learning_rate": 8.319112315523492e-08, + "loss": 0.7043, + "step": 31947 + }, + { + "epoch": 1.17, + "learning_rate": 8.318746109056286e-08, + "loss": 0.6752, + "step": 31948 + }, + { + "epoch": 1.17, + "learning_rate": 8.318379902589079e-08, + "loss": 0.6967, + "step": 31949 + }, + { + "epoch": 1.17, + "learning_rate": 8.318013696121873e-08, + "loss": 0.6909, + "step": 31950 + }, + { + "epoch": 1.17, + "learning_rate": 8.317647489654667e-08, + "loss": 0.7117, + "step": 31951 + }, + { + "epoch": 1.17, + "learning_rate": 8.317281283187461e-08, + "loss": 0.6959, + "step": 31952 + }, + { + "epoch": 1.17, + "learning_rate": 8.316915076720255e-08, + "loss": 0.6974, + "step": 31953 + }, + { + "epoch": 1.17, + "learning_rate": 8.316548870253049e-08, + "loss": 0.7152, + "step": 31954 + }, + { + "epoch": 1.17, + "learning_rate": 8.316182663785842e-08, + "loss": 0.6854, + "step": 31955 + }, + { + "epoch": 1.17, + "learning_rate": 8.315816457318636e-08, + "loss": 0.6953, + "step": 31956 + }, + { + "epoch": 1.17, + "learning_rate": 8.31545025085143e-08, + "loss": 0.6832, + "step": 31957 + }, + { + "epoch": 1.17, + "learning_rate": 8.315084044384222e-08, + "loss": 0.6917, + "step": 31958 + }, + { + "epoch": 1.17, + "learning_rate": 8.314717837917016e-08, + "loss": 0.6958, + "step": 31959 + }, + { + "epoch": 1.17, + "learning_rate": 8.31435163144981e-08, + "loss": 0.6884, + "step": 31960 + }, + { + "epoch": 1.17, + "learning_rate": 8.313985424982604e-08, + "loss": 0.6896, + "step": 31961 + }, + { + "epoch": 1.17, + "learning_rate": 8.313619218515398e-08, + "loss": 0.7086, + "step": 31962 + }, + { + "epoch": 1.17, + "learning_rate": 8.313253012048192e-08, + "loss": 0.6877, + "step": 31963 + }, + { + "epoch": 1.17, + "learning_rate": 8.312886805580986e-08, + "loss": 0.6999, + "step": 31964 + }, + { + "epoch": 1.17, + "learning_rate": 8.31252059911378e-08, + "loss": 0.6955, + "step": 31965 + }, + { + "epoch": 1.17, + "learning_rate": 8.312154392646574e-08, + "loss": 0.6967, + "step": 31966 + }, + { + "epoch": 1.17, + "learning_rate": 8.311788186179367e-08, + "loss": 0.694, + "step": 31967 + }, + { + "epoch": 1.17, + "learning_rate": 8.311421979712161e-08, + "loss": 0.6905, + "step": 31968 + }, + { + "epoch": 1.17, + "learning_rate": 8.311055773244955e-08, + "loss": 0.6987, + "step": 31969 + }, + { + "epoch": 1.17, + "learning_rate": 8.310689566777749e-08, + "loss": 0.6933, + "step": 31970 + }, + { + "epoch": 1.17, + "learning_rate": 8.310323360310543e-08, + "loss": 0.6903, + "step": 31971 + }, + { + "epoch": 1.17, + "learning_rate": 8.309957153843337e-08, + "loss": 0.7015, + "step": 31972 + }, + { + "epoch": 1.17, + "learning_rate": 8.309590947376131e-08, + "loss": 0.6853, + "step": 31973 + }, + { + "epoch": 1.17, + "learning_rate": 8.309224740908925e-08, + "loss": 0.7052, + "step": 31974 + }, + { + "epoch": 1.17, + "learning_rate": 8.308858534441719e-08, + "loss": 0.7003, + "step": 31975 + }, + { + "epoch": 1.17, + "learning_rate": 8.308492327974512e-08, + "loss": 0.6747, + "step": 31976 + }, + { + "epoch": 1.17, + "learning_rate": 8.308126121507306e-08, + "loss": 0.6819, + "step": 31977 + }, + { + "epoch": 1.17, + "learning_rate": 8.3077599150401e-08, + "loss": 0.6943, + "step": 31978 + }, + { + "epoch": 1.17, + "learning_rate": 8.307393708572894e-08, + "loss": 0.6911, + "step": 31979 + }, + { + "epoch": 1.17, + "learning_rate": 8.307027502105686e-08, + "loss": 0.6911, + "step": 31980 + }, + { + "epoch": 1.17, + "learning_rate": 8.30666129563848e-08, + "loss": 0.716, + "step": 31981 + }, + { + "epoch": 1.17, + "learning_rate": 8.306295089171274e-08, + "loss": 0.6905, + "step": 31982 + }, + { + "epoch": 1.17, + "learning_rate": 8.305928882704068e-08, + "loss": 0.6945, + "step": 31983 + }, + { + "epoch": 1.17, + "learning_rate": 8.305562676236862e-08, + "loss": 0.6942, + "step": 31984 + }, + { + "epoch": 1.17, + "learning_rate": 8.305196469769655e-08, + "loss": 0.6911, + "step": 31985 + }, + { + "epoch": 1.17, + "learning_rate": 8.304830263302449e-08, + "loss": 0.6838, + "step": 31986 + }, + { + "epoch": 1.17, + "learning_rate": 8.304464056835243e-08, + "loss": 0.6916, + "step": 31987 + }, + { + "epoch": 1.17, + "learning_rate": 8.304097850368037e-08, + "loss": 0.6849, + "step": 31988 + }, + { + "epoch": 1.17, + "learning_rate": 8.303731643900831e-08, + "loss": 0.6868, + "step": 31989 + }, + { + "epoch": 1.17, + "learning_rate": 8.303365437433625e-08, + "loss": 0.691, + "step": 31990 + }, + { + "epoch": 1.17, + "learning_rate": 8.302999230966419e-08, + "loss": 0.7011, + "step": 31991 + }, + { + "epoch": 1.17, + "learning_rate": 8.302633024499213e-08, + "loss": 0.7023, + "step": 31992 + }, + { + "epoch": 1.17, + "learning_rate": 8.302266818032007e-08, + "loss": 0.6985, + "step": 31993 + }, + { + "epoch": 1.17, + "learning_rate": 8.3019006115648e-08, + "loss": 0.6871, + "step": 31994 + }, + { + "epoch": 1.17, + "learning_rate": 8.301534405097593e-08, + "loss": 0.6635, + "step": 31995 + }, + { + "epoch": 1.17, + "learning_rate": 8.301168198630387e-08, + "loss": 0.6889, + "step": 31996 + }, + { + "epoch": 1.17, + "learning_rate": 8.300801992163181e-08, + "loss": 0.7076, + "step": 31997 + }, + { + "epoch": 1.17, + "learning_rate": 8.300435785695975e-08, + "loss": 0.6913, + "step": 31998 + }, + { + "epoch": 1.17, + "learning_rate": 8.30006957922877e-08, + "loss": 0.6817, + "step": 31999 + }, + { + "epoch": 1.17, + "learning_rate": 8.299703372761563e-08, + "loss": 0.7044, + "step": 32000 + }, + { + "epoch": 1.17, + "learning_rate": 8.299337166294356e-08, + "loss": 0.6903, + "step": 32001 + }, + { + "epoch": 1.17, + "learning_rate": 8.29897095982715e-08, + "loss": 0.6881, + "step": 32002 + }, + { + "epoch": 1.17, + "learning_rate": 8.298604753359944e-08, + "loss": 0.6971, + "step": 32003 + }, + { + "epoch": 1.17, + "learning_rate": 8.298238546892737e-08, + "loss": 0.6908, + "step": 32004 + }, + { + "epoch": 1.17, + "learning_rate": 8.297872340425531e-08, + "loss": 0.7131, + "step": 32005 + }, + { + "epoch": 1.17, + "learning_rate": 8.297506133958325e-08, + "loss": 0.6792, + "step": 32006 + }, + { + "epoch": 1.17, + "learning_rate": 8.297139927491119e-08, + "loss": 0.6942, + "step": 32007 + }, + { + "epoch": 1.17, + "learning_rate": 8.296773721023913e-08, + "loss": 0.698, + "step": 32008 + }, + { + "epoch": 1.17, + "learning_rate": 8.296407514556707e-08, + "loss": 0.6964, + "step": 32009 + }, + { + "epoch": 1.17, + "learning_rate": 8.296041308089501e-08, + "loss": 0.6881, + "step": 32010 + }, + { + "epoch": 1.17, + "learning_rate": 8.295675101622295e-08, + "loss": 0.6947, + "step": 32011 + }, + { + "epoch": 1.17, + "learning_rate": 8.295308895155087e-08, + "loss": 0.7089, + "step": 32012 + }, + { + "epoch": 1.17, + "learning_rate": 8.294942688687881e-08, + "loss": 0.6858, + "step": 32013 + }, + { + "epoch": 1.17, + "learning_rate": 8.294576482220675e-08, + "loss": 0.6894, + "step": 32014 + }, + { + "epoch": 1.17, + "learning_rate": 8.29421027575347e-08, + "loss": 0.7041, + "step": 32015 + }, + { + "epoch": 1.17, + "learning_rate": 8.293844069286263e-08, + "loss": 0.7036, + "step": 32016 + }, + { + "epoch": 1.17, + "learning_rate": 8.293477862819057e-08, + "loss": 0.696, + "step": 32017 + }, + { + "epoch": 1.17, + "learning_rate": 8.293111656351851e-08, + "loss": 0.7053, + "step": 32018 + }, + { + "epoch": 1.17, + "learning_rate": 8.292745449884645e-08, + "loss": 0.6889, + "step": 32019 + }, + { + "epoch": 1.17, + "learning_rate": 8.292379243417439e-08, + "loss": 0.6895, + "step": 32020 + }, + { + "epoch": 1.17, + "learning_rate": 8.292013036950232e-08, + "loss": 0.7105, + "step": 32021 + }, + { + "epoch": 1.17, + "learning_rate": 8.291646830483026e-08, + "loss": 0.696, + "step": 32022 + }, + { + "epoch": 1.17, + "learning_rate": 8.29128062401582e-08, + "loss": 0.6973, + "step": 32023 + }, + { + "epoch": 1.17, + "learning_rate": 8.290914417548614e-08, + "loss": 0.6832, + "step": 32024 + }, + { + "epoch": 1.17, + "learning_rate": 8.290548211081408e-08, + "loss": 0.7112, + "step": 32025 + }, + { + "epoch": 1.17, + "learning_rate": 8.2901820046142e-08, + "loss": 0.6837, + "step": 32026 + }, + { + "epoch": 1.17, + "learning_rate": 8.289815798146995e-08, + "loss": 0.7029, + "step": 32027 + }, + { + "epoch": 1.17, + "learning_rate": 8.289449591679789e-08, + "loss": 0.7027, + "step": 32028 + }, + { + "epoch": 1.17, + "learning_rate": 8.289083385212583e-08, + "loss": 0.6745, + "step": 32029 + }, + { + "epoch": 1.17, + "learning_rate": 8.288717178745375e-08, + "loss": 0.6978, + "step": 32030 + }, + { + "epoch": 1.17, + "learning_rate": 8.288350972278169e-08, + "loss": 0.6956, + "step": 32031 + }, + { + "epoch": 1.17, + "learning_rate": 8.287984765810963e-08, + "loss": 0.699, + "step": 32032 + }, + { + "epoch": 1.17, + "learning_rate": 8.287618559343757e-08, + "loss": 0.688, + "step": 32033 + }, + { + "epoch": 1.17, + "learning_rate": 8.287252352876551e-08, + "loss": 0.6956, + "step": 32034 + }, + { + "epoch": 1.17, + "learning_rate": 8.286886146409345e-08, + "loss": 0.6886, + "step": 32035 + }, + { + "epoch": 1.17, + "learning_rate": 8.286519939942139e-08, + "loss": 0.7001, + "step": 32036 + }, + { + "epoch": 1.17, + "learning_rate": 8.286153733474933e-08, + "loss": 0.7134, + "step": 32037 + }, + { + "epoch": 1.17, + "learning_rate": 8.285787527007727e-08, + "loss": 0.6871, + "step": 32038 + }, + { + "epoch": 1.17, + "learning_rate": 8.28542132054052e-08, + "loss": 0.6963, + "step": 32039 + }, + { + "epoch": 1.17, + "learning_rate": 8.285055114073314e-08, + "loss": 0.6704, + "step": 32040 + }, + { + "epoch": 1.17, + "learning_rate": 8.284688907606108e-08, + "loss": 0.6943, + "step": 32041 + }, + { + "epoch": 1.17, + "learning_rate": 8.284322701138902e-08, + "loss": 0.6996, + "step": 32042 + }, + { + "epoch": 1.17, + "learning_rate": 8.283956494671696e-08, + "loss": 0.6904, + "step": 32043 + }, + { + "epoch": 1.17, + "learning_rate": 8.28359028820449e-08, + "loss": 0.6874, + "step": 32044 + }, + { + "epoch": 1.17, + "learning_rate": 8.283224081737284e-08, + "loss": 0.7056, + "step": 32045 + }, + { + "epoch": 1.17, + "learning_rate": 8.282857875270078e-08, + "loss": 0.6895, + "step": 32046 + }, + { + "epoch": 1.17, + "learning_rate": 8.282491668802872e-08, + "loss": 0.69, + "step": 32047 + }, + { + "epoch": 1.17, + "learning_rate": 8.282125462335665e-08, + "loss": 0.6859, + "step": 32048 + }, + { + "epoch": 1.17, + "learning_rate": 8.281759255868458e-08, + "loss": 0.6994, + "step": 32049 + }, + { + "epoch": 1.17, + "learning_rate": 8.281393049401252e-08, + "loss": 0.6944, + "step": 32050 + }, + { + "epoch": 1.17, + "learning_rate": 8.281026842934045e-08, + "loss": 0.6846, + "step": 32051 + }, + { + "epoch": 1.17, + "learning_rate": 8.280660636466839e-08, + "loss": 0.6873, + "step": 32052 + }, + { + "epoch": 1.17, + "learning_rate": 8.280294429999633e-08, + "loss": 0.6894, + "step": 32053 + }, + { + "epoch": 1.17, + "learning_rate": 8.279928223532427e-08, + "loss": 0.6957, + "step": 32054 + }, + { + "epoch": 1.17, + "learning_rate": 8.279562017065221e-08, + "loss": 0.6863, + "step": 32055 + }, + { + "epoch": 1.17, + "learning_rate": 8.279195810598015e-08, + "loss": 0.6973, + "step": 32056 + }, + { + "epoch": 1.17, + "learning_rate": 8.278829604130808e-08, + "loss": 0.691, + "step": 32057 + }, + { + "epoch": 1.17, + "learning_rate": 8.278463397663602e-08, + "loss": 0.6964, + "step": 32058 + }, + { + "epoch": 1.17, + "learning_rate": 8.278097191196396e-08, + "loss": 0.699, + "step": 32059 + }, + { + "epoch": 1.17, + "learning_rate": 8.27773098472919e-08, + "loss": 0.6942, + "step": 32060 + }, + { + "epoch": 1.17, + "learning_rate": 8.277364778261984e-08, + "loss": 0.7024, + "step": 32061 + }, + { + "epoch": 1.17, + "learning_rate": 8.276998571794778e-08, + "loss": 0.6943, + "step": 32062 + }, + { + "epoch": 1.17, + "learning_rate": 8.276632365327572e-08, + "loss": 0.6961, + "step": 32063 + }, + { + "epoch": 1.17, + "learning_rate": 8.276266158860366e-08, + "loss": 0.6915, + "step": 32064 + }, + { + "epoch": 1.17, + "learning_rate": 8.27589995239316e-08, + "loss": 0.6791, + "step": 32065 + }, + { + "epoch": 1.17, + "learning_rate": 8.275533745925952e-08, + "loss": 0.6865, + "step": 32066 + }, + { + "epoch": 1.17, + "learning_rate": 8.275167539458746e-08, + "loss": 0.6897, + "step": 32067 + }, + { + "epoch": 1.17, + "learning_rate": 8.27480133299154e-08, + "loss": 0.6893, + "step": 32068 + }, + { + "epoch": 1.17, + "learning_rate": 8.274435126524334e-08, + "loss": 0.6869, + "step": 32069 + }, + { + "epoch": 1.17, + "learning_rate": 8.274068920057128e-08, + "loss": 0.6959, + "step": 32070 + }, + { + "epoch": 1.17, + "learning_rate": 8.273702713589922e-08, + "loss": 0.6921, + "step": 32071 + }, + { + "epoch": 1.17, + "learning_rate": 8.273336507122715e-08, + "loss": 0.6969, + "step": 32072 + }, + { + "epoch": 1.17, + "learning_rate": 8.272970300655509e-08, + "loss": 0.6924, + "step": 32073 + }, + { + "epoch": 1.17, + "learning_rate": 8.272604094188303e-08, + "loss": 0.6918, + "step": 32074 + }, + { + "epoch": 1.17, + "learning_rate": 8.272237887721096e-08, + "loss": 0.6983, + "step": 32075 + }, + { + "epoch": 1.17, + "learning_rate": 8.27187168125389e-08, + "loss": 0.6898, + "step": 32076 + }, + { + "epoch": 1.17, + "learning_rate": 8.271505474786684e-08, + "loss": 0.6745, + "step": 32077 + }, + { + "epoch": 1.17, + "learning_rate": 8.271139268319478e-08, + "loss": 0.6855, + "step": 32078 + }, + { + "epoch": 1.17, + "learning_rate": 8.270773061852272e-08, + "loss": 0.6918, + "step": 32079 + }, + { + "epoch": 1.17, + "learning_rate": 8.270406855385066e-08, + "loss": 0.6903, + "step": 32080 + }, + { + "epoch": 1.17, + "learning_rate": 8.27004064891786e-08, + "loss": 0.6902, + "step": 32081 + }, + { + "epoch": 1.17, + "learning_rate": 8.269674442450654e-08, + "loss": 0.6939, + "step": 32082 + }, + { + "epoch": 1.17, + "learning_rate": 8.269308235983448e-08, + "loss": 0.698, + "step": 32083 + }, + { + "epoch": 1.17, + "learning_rate": 8.26894202951624e-08, + "loss": 0.6815, + "step": 32084 + }, + { + "epoch": 1.17, + "learning_rate": 8.268575823049034e-08, + "loss": 0.6807, + "step": 32085 + }, + { + "epoch": 1.17, + "learning_rate": 8.268209616581828e-08, + "loss": 0.688, + "step": 32086 + }, + { + "epoch": 1.17, + "learning_rate": 8.267843410114622e-08, + "loss": 0.7102, + "step": 32087 + }, + { + "epoch": 1.17, + "learning_rate": 8.267477203647416e-08, + "loss": 0.6806, + "step": 32088 + }, + { + "epoch": 1.17, + "learning_rate": 8.26711099718021e-08, + "loss": 0.7048, + "step": 32089 + }, + { + "epoch": 1.17, + "learning_rate": 8.266744790713004e-08, + "loss": 0.6917, + "step": 32090 + }, + { + "epoch": 1.17, + "learning_rate": 8.266378584245798e-08, + "loss": 0.7005, + "step": 32091 + }, + { + "epoch": 1.17, + "learning_rate": 8.266012377778592e-08, + "loss": 0.6823, + "step": 32092 + }, + { + "epoch": 1.17, + "learning_rate": 8.265646171311385e-08, + "loss": 0.6949, + "step": 32093 + }, + { + "epoch": 1.17, + "learning_rate": 8.265279964844179e-08, + "loss": 0.6889, + "step": 32094 + }, + { + "epoch": 1.17, + "learning_rate": 8.264913758376973e-08, + "loss": 0.6949, + "step": 32095 + }, + { + "epoch": 1.17, + "learning_rate": 8.264547551909767e-08, + "loss": 0.7025, + "step": 32096 + }, + { + "epoch": 1.17, + "learning_rate": 8.26418134544256e-08, + "loss": 0.6895, + "step": 32097 + }, + { + "epoch": 1.17, + "learning_rate": 8.263815138975354e-08, + "loss": 0.6923, + "step": 32098 + }, + { + "epoch": 1.17, + "learning_rate": 8.263448932508148e-08, + "loss": 0.693, + "step": 32099 + }, + { + "epoch": 1.17, + "learning_rate": 8.263082726040942e-08, + "loss": 0.6931, + "step": 32100 + }, + { + "epoch": 1.17, + "learning_rate": 8.262716519573736e-08, + "loss": 0.6915, + "step": 32101 + }, + { + "epoch": 1.17, + "learning_rate": 8.262350313106528e-08, + "loss": 0.7019, + "step": 32102 + }, + { + "epoch": 1.17, + "learning_rate": 8.261984106639322e-08, + "loss": 0.7139, + "step": 32103 + }, + { + "epoch": 1.17, + "learning_rate": 8.261617900172116e-08, + "loss": 0.7093, + "step": 32104 + }, + { + "epoch": 1.17, + "learning_rate": 8.26125169370491e-08, + "loss": 0.6929, + "step": 32105 + }, + { + "epoch": 1.17, + "learning_rate": 8.260885487237704e-08, + "loss": 0.7112, + "step": 32106 + }, + { + "epoch": 1.17, + "learning_rate": 8.260519280770498e-08, + "loss": 0.6886, + "step": 32107 + }, + { + "epoch": 1.17, + "learning_rate": 8.260153074303292e-08, + "loss": 0.6827, + "step": 32108 + }, + { + "epoch": 1.17, + "learning_rate": 8.259786867836086e-08, + "loss": 0.6943, + "step": 32109 + }, + { + "epoch": 1.17, + "learning_rate": 8.25942066136888e-08, + "loss": 0.6898, + "step": 32110 + }, + { + "epoch": 1.17, + "learning_rate": 8.259054454901673e-08, + "loss": 0.6871, + "step": 32111 + }, + { + "epoch": 1.17, + "learning_rate": 8.258688248434467e-08, + "loss": 0.7095, + "step": 32112 + }, + { + "epoch": 1.17, + "learning_rate": 8.258322041967261e-08, + "loss": 0.6888, + "step": 32113 + }, + { + "epoch": 1.17, + "learning_rate": 8.257955835500055e-08, + "loss": 0.6846, + "step": 32114 + }, + { + "epoch": 1.17, + "learning_rate": 8.257589629032849e-08, + "loss": 0.6996, + "step": 32115 + }, + { + "epoch": 1.18, + "learning_rate": 8.257223422565643e-08, + "loss": 0.6969, + "step": 32116 + }, + { + "epoch": 1.18, + "learning_rate": 8.256857216098437e-08, + "loss": 0.696, + "step": 32117 + }, + { + "epoch": 1.18, + "learning_rate": 8.256491009631231e-08, + "loss": 0.7028, + "step": 32118 + }, + { + "epoch": 1.18, + "learning_rate": 8.256124803164023e-08, + "loss": 0.7025, + "step": 32119 + }, + { + "epoch": 1.18, + "learning_rate": 8.255758596696817e-08, + "loss": 0.6905, + "step": 32120 + }, + { + "epoch": 1.18, + "learning_rate": 8.25539239022961e-08, + "loss": 0.6929, + "step": 32121 + }, + { + "epoch": 1.18, + "learning_rate": 8.255026183762404e-08, + "loss": 0.6855, + "step": 32122 + }, + { + "epoch": 1.18, + "learning_rate": 8.254659977295198e-08, + "loss": 0.7019, + "step": 32123 + }, + { + "epoch": 1.18, + "learning_rate": 8.254293770827992e-08, + "loss": 0.6973, + "step": 32124 + }, + { + "epoch": 1.18, + "learning_rate": 8.253927564360786e-08, + "loss": 0.698, + "step": 32125 + }, + { + "epoch": 1.18, + "learning_rate": 8.25356135789358e-08, + "loss": 0.6954, + "step": 32126 + }, + { + "epoch": 1.18, + "learning_rate": 8.253195151426374e-08, + "loss": 0.6946, + "step": 32127 + }, + { + "epoch": 1.18, + "learning_rate": 8.252828944959168e-08, + "loss": 0.6904, + "step": 32128 + }, + { + "epoch": 1.18, + "learning_rate": 8.252462738491961e-08, + "loss": 0.7012, + "step": 32129 + }, + { + "epoch": 1.18, + "learning_rate": 8.252096532024755e-08, + "loss": 0.6968, + "step": 32130 + }, + { + "epoch": 1.18, + "learning_rate": 8.251730325557549e-08, + "loss": 0.7014, + "step": 32131 + }, + { + "epoch": 1.18, + "learning_rate": 8.251364119090343e-08, + "loss": 0.6974, + "step": 32132 + }, + { + "epoch": 1.18, + "learning_rate": 8.250997912623137e-08, + "loss": 0.6893, + "step": 32133 + }, + { + "epoch": 1.18, + "learning_rate": 8.250631706155931e-08, + "loss": 0.707, + "step": 32134 + }, + { + "epoch": 1.18, + "learning_rate": 8.250265499688725e-08, + "loss": 0.6847, + "step": 32135 + }, + { + "epoch": 1.18, + "learning_rate": 8.249899293221519e-08, + "loss": 0.6939, + "step": 32136 + }, + { + "epoch": 1.18, + "learning_rate": 8.249533086754313e-08, + "loss": 0.6966, + "step": 32137 + }, + { + "epoch": 1.18, + "learning_rate": 8.249166880287105e-08, + "loss": 0.7016, + "step": 32138 + }, + { + "epoch": 1.18, + "learning_rate": 8.248800673819899e-08, + "loss": 0.7077, + "step": 32139 + }, + { + "epoch": 1.18, + "learning_rate": 8.248434467352693e-08, + "loss": 0.6921, + "step": 32140 + }, + { + "epoch": 1.18, + "learning_rate": 8.248068260885487e-08, + "loss": 0.6859, + "step": 32141 + }, + { + "epoch": 1.18, + "learning_rate": 8.247702054418281e-08, + "loss": 0.7047, + "step": 32142 + }, + { + "epoch": 1.18, + "learning_rate": 8.247335847951074e-08, + "loss": 0.6874, + "step": 32143 + }, + { + "epoch": 1.18, + "learning_rate": 8.246969641483868e-08, + "loss": 0.7007, + "step": 32144 + }, + { + "epoch": 1.18, + "learning_rate": 8.246603435016662e-08, + "loss": 0.6919, + "step": 32145 + }, + { + "epoch": 1.18, + "learning_rate": 8.246237228549456e-08, + "loss": 0.701, + "step": 32146 + }, + { + "epoch": 1.18, + "learning_rate": 8.245871022082249e-08, + "loss": 0.6847, + "step": 32147 + }, + { + "epoch": 1.18, + "learning_rate": 8.245504815615043e-08, + "loss": 0.6962, + "step": 32148 + }, + { + "epoch": 1.18, + "learning_rate": 8.245138609147837e-08, + "loss": 0.6897, + "step": 32149 + }, + { + "epoch": 1.18, + "learning_rate": 8.24477240268063e-08, + "loss": 0.694, + "step": 32150 + }, + { + "epoch": 1.18, + "learning_rate": 8.244406196213425e-08, + "loss": 0.6988, + "step": 32151 + }, + { + "epoch": 1.18, + "learning_rate": 8.244039989746219e-08, + "loss": 0.6918, + "step": 32152 + }, + { + "epoch": 1.18, + "learning_rate": 8.243673783279013e-08, + "loss": 0.6939, + "step": 32153 + }, + { + "epoch": 1.18, + "learning_rate": 8.243307576811807e-08, + "loss": 0.6788, + "step": 32154 + }, + { + "epoch": 1.18, + "learning_rate": 8.2429413703446e-08, + "loss": 0.6904, + "step": 32155 + }, + { + "epoch": 1.18, + "learning_rate": 8.242575163877393e-08, + "loss": 0.6972, + "step": 32156 + }, + { + "epoch": 1.18, + "learning_rate": 8.242208957410187e-08, + "loss": 0.6954, + "step": 32157 + }, + { + "epoch": 1.18, + "learning_rate": 8.241842750942981e-08, + "loss": 0.6844, + "step": 32158 + }, + { + "epoch": 1.18, + "learning_rate": 8.241476544475775e-08, + "loss": 0.6919, + "step": 32159 + }, + { + "epoch": 1.18, + "learning_rate": 8.241110338008569e-08, + "loss": 0.6945, + "step": 32160 + }, + { + "epoch": 1.18, + "learning_rate": 8.240744131541363e-08, + "loss": 0.7011, + "step": 32161 + }, + { + "epoch": 1.18, + "learning_rate": 8.240377925074157e-08, + "loss": 0.686, + "step": 32162 + }, + { + "epoch": 1.18, + "learning_rate": 8.240011718606951e-08, + "loss": 0.6805, + "step": 32163 + }, + { + "epoch": 1.18, + "learning_rate": 8.239645512139745e-08, + "loss": 0.6854, + "step": 32164 + }, + { + "epoch": 1.18, + "learning_rate": 8.239279305672538e-08, + "loss": 0.6889, + "step": 32165 + }, + { + "epoch": 1.18, + "learning_rate": 8.238913099205332e-08, + "loss": 0.6908, + "step": 32166 + }, + { + "epoch": 1.18, + "learning_rate": 8.238546892738126e-08, + "loss": 0.6982, + "step": 32167 + }, + { + "epoch": 1.18, + "learning_rate": 8.238180686270919e-08, + "loss": 0.6947, + "step": 32168 + }, + { + "epoch": 1.18, + "learning_rate": 8.237814479803712e-08, + "loss": 0.6986, + "step": 32169 + }, + { + "epoch": 1.18, + "learning_rate": 8.237448273336506e-08, + "loss": 0.6863, + "step": 32170 + }, + { + "epoch": 1.18, + "learning_rate": 8.2370820668693e-08, + "loss": 0.6805, + "step": 32171 + }, + { + "epoch": 1.18, + "learning_rate": 8.236715860402094e-08, + "loss": 0.6933, + "step": 32172 + }, + { + "epoch": 1.18, + "learning_rate": 8.236349653934887e-08, + "loss": 0.7033, + "step": 32173 + }, + { + "epoch": 1.18, + "learning_rate": 8.235983447467681e-08, + "loss": 0.6979, + "step": 32174 + }, + { + "epoch": 1.18, + "learning_rate": 8.235617241000475e-08, + "loss": 0.6839, + "step": 32175 + }, + { + "epoch": 1.18, + "learning_rate": 8.235251034533269e-08, + "loss": 0.687, + "step": 32176 + }, + { + "epoch": 1.18, + "learning_rate": 8.234884828066063e-08, + "loss": 0.6889, + "step": 32177 + }, + { + "epoch": 1.18, + "learning_rate": 8.234518621598857e-08, + "loss": 0.7002, + "step": 32178 + }, + { + "epoch": 1.18, + "learning_rate": 8.234152415131651e-08, + "loss": 0.6902, + "step": 32179 + }, + { + "epoch": 1.18, + "learning_rate": 8.233786208664445e-08, + "loss": 0.6917, + "step": 32180 + }, + { + "epoch": 1.18, + "learning_rate": 8.233420002197239e-08, + "loss": 0.7149, + "step": 32181 + }, + { + "epoch": 1.18, + "learning_rate": 8.233053795730032e-08, + "loss": 0.7024, + "step": 32182 + }, + { + "epoch": 1.18, + "learning_rate": 8.232687589262826e-08, + "loss": 0.6991, + "step": 32183 + }, + { + "epoch": 1.18, + "learning_rate": 8.23232138279562e-08, + "loss": 0.6955, + "step": 32184 + }, + { + "epoch": 1.18, + "learning_rate": 8.231955176328414e-08, + "loss": 0.6879, + "step": 32185 + }, + { + "epoch": 1.18, + "learning_rate": 8.231588969861208e-08, + "loss": 0.6847, + "step": 32186 + }, + { + "epoch": 1.18, + "learning_rate": 8.231222763394002e-08, + "loss": 0.7072, + "step": 32187 + }, + { + "epoch": 1.18, + "learning_rate": 8.230856556926796e-08, + "loss": 0.6943, + "step": 32188 + }, + { + "epoch": 1.18, + "learning_rate": 8.23049035045959e-08, + "loss": 0.6726, + "step": 32189 + }, + { + "epoch": 1.18, + "learning_rate": 8.230124143992382e-08, + "loss": 0.7035, + "step": 32190 + }, + { + "epoch": 1.18, + "learning_rate": 8.229757937525176e-08, + "loss": 0.6925, + "step": 32191 + }, + { + "epoch": 1.18, + "learning_rate": 8.229391731057969e-08, + "loss": 0.6863, + "step": 32192 + }, + { + "epoch": 1.18, + "learning_rate": 8.229025524590763e-08, + "loss": 0.6992, + "step": 32193 + }, + { + "epoch": 1.18, + "learning_rate": 8.228659318123557e-08, + "loss": 0.6932, + "step": 32194 + }, + { + "epoch": 1.18, + "learning_rate": 8.228293111656351e-08, + "loss": 0.6835, + "step": 32195 + }, + { + "epoch": 1.18, + "learning_rate": 8.227926905189145e-08, + "loss": 0.6985, + "step": 32196 + }, + { + "epoch": 1.18, + "learning_rate": 8.227560698721939e-08, + "loss": 0.6823, + "step": 32197 + }, + { + "epoch": 1.18, + "learning_rate": 8.227194492254733e-08, + "loss": 0.7024, + "step": 32198 + }, + { + "epoch": 1.18, + "learning_rate": 8.226828285787527e-08, + "loss": 0.6824, + "step": 32199 + }, + { + "epoch": 1.18, + "learning_rate": 8.22646207932032e-08, + "loss": 0.7001, + "step": 32200 + }, + { + "epoch": 1.18, + "learning_rate": 8.226095872853114e-08, + "loss": 0.6938, + "step": 32201 + }, + { + "epoch": 1.18, + "learning_rate": 8.225729666385908e-08, + "loss": 0.6835, + "step": 32202 + }, + { + "epoch": 1.18, + "learning_rate": 8.225363459918702e-08, + "loss": 0.6948, + "step": 32203 + }, + { + "epoch": 1.18, + "learning_rate": 8.224997253451496e-08, + "loss": 0.6941, + "step": 32204 + }, + { + "epoch": 1.18, + "learning_rate": 8.22463104698429e-08, + "loss": 0.6881, + "step": 32205 + }, + { + "epoch": 1.18, + "learning_rate": 8.224264840517084e-08, + "loss": 0.7013, + "step": 32206 + }, + { + "epoch": 1.18, + "learning_rate": 8.223898634049878e-08, + "loss": 0.6857, + "step": 32207 + }, + { + "epoch": 1.18, + "learning_rate": 8.223532427582672e-08, + "loss": 0.6934, + "step": 32208 + }, + { + "epoch": 1.18, + "learning_rate": 8.223166221115464e-08, + "loss": 0.6908, + "step": 32209 + }, + { + "epoch": 1.18, + "learning_rate": 8.222800014648258e-08, + "loss": 0.6894, + "step": 32210 + }, + { + "epoch": 1.18, + "learning_rate": 8.222433808181052e-08, + "loss": 0.6848, + "step": 32211 + }, + { + "epoch": 1.18, + "learning_rate": 8.222067601713846e-08, + "loss": 0.6997, + "step": 32212 + }, + { + "epoch": 1.18, + "learning_rate": 8.22170139524664e-08, + "loss": 0.6894, + "step": 32213 + }, + { + "epoch": 1.18, + "learning_rate": 8.221335188779433e-08, + "loss": 0.6948, + "step": 32214 + }, + { + "epoch": 1.18, + "learning_rate": 8.220968982312227e-08, + "loss": 0.6933, + "step": 32215 + }, + { + "epoch": 1.18, + "learning_rate": 8.220602775845021e-08, + "loss": 0.6972, + "step": 32216 + }, + { + "epoch": 1.18, + "learning_rate": 8.220236569377815e-08, + "loss": 0.6936, + "step": 32217 + }, + { + "epoch": 1.18, + "learning_rate": 8.219870362910608e-08, + "loss": 0.6997, + "step": 32218 + }, + { + "epoch": 1.18, + "learning_rate": 8.219504156443402e-08, + "loss": 0.675, + "step": 32219 + }, + { + "epoch": 1.18, + "learning_rate": 8.219137949976196e-08, + "loss": 0.6971, + "step": 32220 + }, + { + "epoch": 1.18, + "learning_rate": 8.21877174350899e-08, + "loss": 0.6833, + "step": 32221 + }, + { + "epoch": 1.18, + "learning_rate": 8.218405537041784e-08, + "loss": 0.6933, + "step": 32222 + }, + { + "epoch": 1.18, + "learning_rate": 8.218039330574578e-08, + "loss": 0.6941, + "step": 32223 + }, + { + "epoch": 1.18, + "learning_rate": 8.217673124107372e-08, + "loss": 0.6794, + "step": 32224 + }, + { + "epoch": 1.18, + "learning_rate": 8.217306917640166e-08, + "loss": 0.6911, + "step": 32225 + }, + { + "epoch": 1.18, + "learning_rate": 8.21694071117296e-08, + "loss": 0.6929, + "step": 32226 + }, + { + "epoch": 1.18, + "learning_rate": 8.216574504705752e-08, + "loss": 0.6925, + "step": 32227 + }, + { + "epoch": 1.18, + "learning_rate": 8.216208298238546e-08, + "loss": 0.6869, + "step": 32228 + }, + { + "epoch": 1.18, + "learning_rate": 8.21584209177134e-08, + "loss": 0.6794, + "step": 32229 + }, + { + "epoch": 1.18, + "learning_rate": 8.215475885304134e-08, + "loss": 0.6867, + "step": 32230 + }, + { + "epoch": 1.18, + "learning_rate": 8.215109678836928e-08, + "loss": 0.6884, + "step": 32231 + }, + { + "epoch": 1.18, + "learning_rate": 8.214743472369722e-08, + "loss": 0.6868, + "step": 32232 + }, + { + "epoch": 1.18, + "learning_rate": 8.214377265902516e-08, + "loss": 0.689, + "step": 32233 + }, + { + "epoch": 1.18, + "learning_rate": 8.21401105943531e-08, + "loss": 0.6883, + "step": 32234 + }, + { + "epoch": 1.18, + "learning_rate": 8.213644852968104e-08, + "loss": 0.6832, + "step": 32235 + }, + { + "epoch": 1.18, + "learning_rate": 8.213278646500897e-08, + "loss": 0.7033, + "step": 32236 + }, + { + "epoch": 1.18, + "learning_rate": 8.212912440033691e-08, + "loss": 0.6912, + "step": 32237 + }, + { + "epoch": 1.18, + "learning_rate": 8.212546233566485e-08, + "loss": 0.6838, + "step": 32238 + }, + { + "epoch": 1.18, + "learning_rate": 8.212180027099277e-08, + "loss": 0.6912, + "step": 32239 + }, + { + "epoch": 1.18, + "learning_rate": 8.211813820632071e-08, + "loss": 0.6807, + "step": 32240 + }, + { + "epoch": 1.18, + "learning_rate": 8.211447614164865e-08, + "loss": 0.7142, + "step": 32241 + }, + { + "epoch": 1.18, + "learning_rate": 8.21108140769766e-08, + "loss": 0.6963, + "step": 32242 + }, + { + "epoch": 1.18, + "learning_rate": 8.210715201230453e-08, + "loss": 0.6982, + "step": 32243 + }, + { + "epoch": 1.18, + "learning_rate": 8.210348994763247e-08, + "loss": 0.6806, + "step": 32244 + }, + { + "epoch": 1.18, + "learning_rate": 8.20998278829604e-08, + "loss": 0.6962, + "step": 32245 + }, + { + "epoch": 1.18, + "learning_rate": 8.209616581828834e-08, + "loss": 0.6918, + "step": 32246 + }, + { + "epoch": 1.18, + "learning_rate": 8.209250375361628e-08, + "loss": 0.6951, + "step": 32247 + }, + { + "epoch": 1.18, + "learning_rate": 8.208884168894422e-08, + "loss": 0.6979, + "step": 32248 + }, + { + "epoch": 1.18, + "learning_rate": 8.208517962427216e-08, + "loss": 0.694, + "step": 32249 + }, + { + "epoch": 1.18, + "learning_rate": 8.20815175596001e-08, + "loss": 0.6819, + "step": 32250 + }, + { + "epoch": 1.18, + "learning_rate": 8.207785549492804e-08, + "loss": 0.7002, + "step": 32251 + }, + { + "epoch": 1.18, + "learning_rate": 8.207419343025598e-08, + "loss": 0.7017, + "step": 32252 + }, + { + "epoch": 1.18, + "learning_rate": 8.207053136558392e-08, + "loss": 0.6976, + "step": 32253 + }, + { + "epoch": 1.18, + "learning_rate": 8.206686930091185e-08, + "loss": 0.6796, + "step": 32254 + }, + { + "epoch": 1.18, + "learning_rate": 8.206320723623979e-08, + "loss": 0.703, + "step": 32255 + }, + { + "epoch": 1.18, + "learning_rate": 8.205954517156773e-08, + "loss": 0.6867, + "step": 32256 + }, + { + "epoch": 1.18, + "learning_rate": 8.205588310689567e-08, + "loss": 0.6863, + "step": 32257 + }, + { + "epoch": 1.18, + "learning_rate": 8.20522210422236e-08, + "loss": 0.6942, + "step": 32258 + }, + { + "epoch": 1.18, + "learning_rate": 8.204855897755155e-08, + "loss": 0.6956, + "step": 32259 + }, + { + "epoch": 1.18, + "learning_rate": 8.204489691287949e-08, + "loss": 0.7003, + "step": 32260 + }, + { + "epoch": 1.18, + "learning_rate": 8.204123484820741e-08, + "loss": 0.6917, + "step": 32261 + }, + { + "epoch": 1.18, + "learning_rate": 8.203757278353535e-08, + "loss": 0.6977, + "step": 32262 + }, + { + "epoch": 1.18, + "learning_rate": 8.203391071886328e-08, + "loss": 0.6929, + "step": 32263 + }, + { + "epoch": 1.18, + "learning_rate": 8.203024865419122e-08, + "loss": 0.6917, + "step": 32264 + }, + { + "epoch": 1.18, + "learning_rate": 8.202658658951916e-08, + "loss": 0.6874, + "step": 32265 + }, + { + "epoch": 1.18, + "learning_rate": 8.20229245248471e-08, + "loss": 0.6953, + "step": 32266 + }, + { + "epoch": 1.18, + "learning_rate": 8.201926246017504e-08, + "loss": 0.6956, + "step": 32267 + }, + { + "epoch": 1.18, + "learning_rate": 8.201560039550298e-08, + "loss": 0.6889, + "step": 32268 + }, + { + "epoch": 1.18, + "learning_rate": 8.201193833083092e-08, + "loss": 0.6876, + "step": 32269 + }, + { + "epoch": 1.18, + "learning_rate": 8.200827626615886e-08, + "loss": 0.6908, + "step": 32270 + }, + { + "epoch": 1.18, + "learning_rate": 8.20046142014868e-08, + "loss": 0.678, + "step": 32271 + }, + { + "epoch": 1.18, + "learning_rate": 8.200095213681473e-08, + "loss": 0.6979, + "step": 32272 + }, + { + "epoch": 1.18, + "learning_rate": 8.199729007214267e-08, + "loss": 0.6765, + "step": 32273 + }, + { + "epoch": 1.18, + "learning_rate": 8.19936280074706e-08, + "loss": 0.6868, + "step": 32274 + }, + { + "epoch": 1.18, + "learning_rate": 8.198996594279855e-08, + "loss": 0.6951, + "step": 32275 + }, + { + "epoch": 1.18, + "learning_rate": 8.198630387812649e-08, + "loss": 0.7008, + "step": 32276 + }, + { + "epoch": 1.18, + "learning_rate": 8.198264181345443e-08, + "loss": 0.6827, + "step": 32277 + }, + { + "epoch": 1.18, + "learning_rate": 8.197897974878237e-08, + "loss": 0.6804, + "step": 32278 + }, + { + "epoch": 1.18, + "learning_rate": 8.19753176841103e-08, + "loss": 0.6944, + "step": 32279 + }, + { + "epoch": 1.18, + "learning_rate": 8.197165561943825e-08, + "loss": 0.6918, + "step": 32280 + }, + { + "epoch": 1.18, + "learning_rate": 8.196799355476617e-08, + "loss": 0.6957, + "step": 32281 + }, + { + "epoch": 1.18, + "learning_rate": 8.196433149009411e-08, + "loss": 0.6902, + "step": 32282 + }, + { + "epoch": 1.18, + "learning_rate": 8.196066942542205e-08, + "loss": 0.7076, + "step": 32283 + }, + { + "epoch": 1.18, + "learning_rate": 8.195700736074999e-08, + "loss": 0.7007, + "step": 32284 + }, + { + "epoch": 1.18, + "learning_rate": 8.195334529607792e-08, + "loss": 0.6919, + "step": 32285 + }, + { + "epoch": 1.18, + "learning_rate": 8.194968323140586e-08, + "loss": 0.6837, + "step": 32286 + }, + { + "epoch": 1.18, + "learning_rate": 8.19460211667338e-08, + "loss": 0.7046, + "step": 32287 + }, + { + "epoch": 1.18, + "learning_rate": 8.194235910206174e-08, + "loss": 0.6967, + "step": 32288 + }, + { + "epoch": 1.18, + "learning_rate": 8.193869703738968e-08, + "loss": 0.6967, + "step": 32289 + }, + { + "epoch": 1.18, + "learning_rate": 8.19350349727176e-08, + "loss": 0.7026, + "step": 32290 + }, + { + "epoch": 1.18, + "learning_rate": 8.193137290804554e-08, + "loss": 0.6869, + "step": 32291 + }, + { + "epoch": 1.18, + "learning_rate": 8.192771084337348e-08, + "loss": 0.6968, + "step": 32292 + }, + { + "epoch": 1.18, + "learning_rate": 8.192404877870142e-08, + "loss": 0.6836, + "step": 32293 + }, + { + "epoch": 1.18, + "learning_rate": 8.192038671402936e-08, + "loss": 0.6778, + "step": 32294 + }, + { + "epoch": 1.18, + "learning_rate": 8.19167246493573e-08, + "loss": 0.6893, + "step": 32295 + }, + { + "epoch": 1.18, + "learning_rate": 8.191306258468524e-08, + "loss": 0.685, + "step": 32296 + }, + { + "epoch": 1.18, + "learning_rate": 8.190940052001318e-08, + "loss": 0.7027, + "step": 32297 + }, + { + "epoch": 1.18, + "learning_rate": 8.190573845534112e-08, + "loss": 0.701, + "step": 32298 + }, + { + "epoch": 1.18, + "learning_rate": 8.190207639066905e-08, + "loss": 0.6975, + "step": 32299 + }, + { + "epoch": 1.18, + "learning_rate": 8.189841432599699e-08, + "loss": 0.6886, + "step": 32300 + }, + { + "epoch": 1.18, + "learning_rate": 8.189475226132493e-08, + "loss": 0.6941, + "step": 32301 + }, + { + "epoch": 1.18, + "learning_rate": 8.189109019665287e-08, + "loss": 0.6862, + "step": 32302 + }, + { + "epoch": 1.18, + "learning_rate": 8.188742813198081e-08, + "loss": 0.6976, + "step": 32303 + }, + { + "epoch": 1.18, + "learning_rate": 8.188376606730875e-08, + "loss": 0.7041, + "step": 32304 + }, + { + "epoch": 1.18, + "learning_rate": 8.188010400263669e-08, + "loss": 0.6893, + "step": 32305 + }, + { + "epoch": 1.18, + "learning_rate": 8.187644193796463e-08, + "loss": 0.6837, + "step": 32306 + }, + { + "epoch": 1.18, + "learning_rate": 8.187277987329256e-08, + "loss": 0.6904, + "step": 32307 + }, + { + "epoch": 1.18, + "learning_rate": 8.18691178086205e-08, + "loss": 0.7195, + "step": 32308 + }, + { + "epoch": 1.18, + "learning_rate": 8.186545574394844e-08, + "loss": 0.6876, + "step": 32309 + }, + { + "epoch": 1.18, + "learning_rate": 8.186179367927636e-08, + "loss": 0.6947, + "step": 32310 + }, + { + "epoch": 1.18, + "learning_rate": 8.18581316146043e-08, + "loss": 0.7077, + "step": 32311 + }, + { + "epoch": 1.18, + "learning_rate": 8.185446954993224e-08, + "loss": 0.7036, + "step": 32312 + }, + { + "epoch": 1.18, + "learning_rate": 8.185080748526018e-08, + "loss": 0.6835, + "step": 32313 + }, + { + "epoch": 1.18, + "learning_rate": 8.184714542058812e-08, + "loss": 0.6961, + "step": 32314 + }, + { + "epoch": 1.18, + "learning_rate": 8.184348335591606e-08, + "loss": 0.7025, + "step": 32315 + }, + { + "epoch": 1.18, + "learning_rate": 8.1839821291244e-08, + "loss": 0.696, + "step": 32316 + }, + { + "epoch": 1.18, + "learning_rate": 8.183615922657193e-08, + "loss": 0.6979, + "step": 32317 + }, + { + "epoch": 1.18, + "learning_rate": 8.183249716189987e-08, + "loss": 0.7049, + "step": 32318 + }, + { + "epoch": 1.18, + "learning_rate": 8.182883509722781e-08, + "loss": 0.6853, + "step": 32319 + }, + { + "epoch": 1.18, + "learning_rate": 8.182517303255575e-08, + "loss": 0.6892, + "step": 32320 + }, + { + "epoch": 1.18, + "learning_rate": 8.182151096788369e-08, + "loss": 0.6918, + "step": 32321 + }, + { + "epoch": 1.18, + "learning_rate": 8.181784890321163e-08, + "loss": 0.6891, + "step": 32322 + }, + { + "epoch": 1.18, + "learning_rate": 8.181418683853957e-08, + "loss": 0.7079, + "step": 32323 + }, + { + "epoch": 1.18, + "learning_rate": 8.181052477386751e-08, + "loss": 0.6927, + "step": 32324 + }, + { + "epoch": 1.18, + "learning_rate": 8.180686270919545e-08, + "loss": 0.6819, + "step": 32325 + }, + { + "epoch": 1.18, + "learning_rate": 8.180320064452338e-08, + "loss": 0.6956, + "step": 32326 + }, + { + "epoch": 1.18, + "learning_rate": 8.179953857985132e-08, + "loss": 0.6951, + "step": 32327 + }, + { + "epoch": 1.18, + "learning_rate": 8.179587651517926e-08, + "loss": 0.696, + "step": 32328 + }, + { + "epoch": 1.18, + "learning_rate": 8.17922144505072e-08, + "loss": 0.6999, + "step": 32329 + }, + { + "epoch": 1.18, + "learning_rate": 8.178855238583514e-08, + "loss": 0.6971, + "step": 32330 + }, + { + "epoch": 1.18, + "learning_rate": 8.178489032116308e-08, + "loss": 0.6942, + "step": 32331 + }, + { + "epoch": 1.18, + "learning_rate": 8.1781228256491e-08, + "loss": 0.6993, + "step": 32332 + }, + { + "epoch": 1.18, + "learning_rate": 8.177756619181894e-08, + "loss": 0.6945, + "step": 32333 + }, + { + "epoch": 1.18, + "learning_rate": 8.177390412714688e-08, + "loss": 0.6807, + "step": 32334 + }, + { + "epoch": 1.18, + "learning_rate": 8.177024206247481e-08, + "loss": 0.699, + "step": 32335 + }, + { + "epoch": 1.18, + "learning_rate": 8.176657999780275e-08, + "loss": 0.6923, + "step": 32336 + }, + { + "epoch": 1.18, + "learning_rate": 8.176291793313069e-08, + "loss": 0.6986, + "step": 32337 + }, + { + "epoch": 1.18, + "learning_rate": 8.175925586845863e-08, + "loss": 0.6909, + "step": 32338 + }, + { + "epoch": 1.18, + "learning_rate": 8.175559380378657e-08, + "loss": 0.6902, + "step": 32339 + }, + { + "epoch": 1.18, + "learning_rate": 8.175193173911451e-08, + "loss": 0.7039, + "step": 32340 + }, + { + "epoch": 1.18, + "learning_rate": 8.174826967444245e-08, + "loss": 0.6903, + "step": 32341 + }, + { + "epoch": 1.18, + "learning_rate": 8.174460760977039e-08, + "loss": 0.6901, + "step": 32342 + }, + { + "epoch": 1.18, + "learning_rate": 8.174094554509833e-08, + "loss": 0.6897, + "step": 32343 + }, + { + "epoch": 1.18, + "learning_rate": 8.173728348042626e-08, + "loss": 0.701, + "step": 32344 + }, + { + "epoch": 1.18, + "learning_rate": 8.17336214157542e-08, + "loss": 0.6737, + "step": 32345 + }, + { + "epoch": 1.18, + "learning_rate": 8.172995935108213e-08, + "loss": 0.6812, + "step": 32346 + }, + { + "epoch": 1.18, + "learning_rate": 8.172629728641007e-08, + "loss": 0.688, + "step": 32347 + }, + { + "epoch": 1.18, + "learning_rate": 8.172263522173801e-08, + "loss": 0.6957, + "step": 32348 + }, + { + "epoch": 1.18, + "learning_rate": 8.171897315706595e-08, + "loss": 0.692, + "step": 32349 + }, + { + "epoch": 1.18, + "learning_rate": 8.17153110923939e-08, + "loss": 0.6777, + "step": 32350 + }, + { + "epoch": 1.18, + "learning_rate": 8.171164902772183e-08, + "loss": 0.6975, + "step": 32351 + }, + { + "epoch": 1.18, + "learning_rate": 8.170798696304977e-08, + "loss": 0.6887, + "step": 32352 + }, + { + "epoch": 1.18, + "learning_rate": 8.17043248983777e-08, + "loss": 0.6985, + "step": 32353 + }, + { + "epoch": 1.18, + "learning_rate": 8.170066283370564e-08, + "loss": 0.7097, + "step": 32354 + }, + { + "epoch": 1.18, + "learning_rate": 8.169700076903358e-08, + "loss": 0.6939, + "step": 32355 + }, + { + "epoch": 1.18, + "learning_rate": 8.169333870436151e-08, + "loss": 0.7004, + "step": 32356 + }, + { + "epoch": 1.18, + "learning_rate": 8.168967663968945e-08, + "loss": 0.6875, + "step": 32357 + }, + { + "epoch": 1.18, + "learning_rate": 8.168601457501739e-08, + "loss": 0.7011, + "step": 32358 + }, + { + "epoch": 1.18, + "learning_rate": 8.168235251034533e-08, + "loss": 0.6986, + "step": 32359 + }, + { + "epoch": 1.18, + "learning_rate": 8.167869044567327e-08, + "loss": 0.7023, + "step": 32360 + }, + { + "epoch": 1.18, + "learning_rate": 8.167502838100121e-08, + "loss": 0.6909, + "step": 32361 + }, + { + "epoch": 1.18, + "learning_rate": 8.167136631632913e-08, + "loss": 0.6968, + "step": 32362 + }, + { + "epoch": 1.18, + "learning_rate": 8.166770425165707e-08, + "loss": 0.6974, + "step": 32363 + }, + { + "epoch": 1.18, + "learning_rate": 8.166404218698501e-08, + "loss": 0.6964, + "step": 32364 + }, + { + "epoch": 1.18, + "learning_rate": 8.166038012231295e-08, + "loss": 0.6885, + "step": 32365 + }, + { + "epoch": 1.18, + "learning_rate": 8.16567180576409e-08, + "loss": 0.7064, + "step": 32366 + }, + { + "epoch": 1.18, + "learning_rate": 8.165305599296883e-08, + "loss": 0.6988, + "step": 32367 + }, + { + "epoch": 1.18, + "learning_rate": 8.164939392829677e-08, + "loss": 0.6954, + "step": 32368 + }, + { + "epoch": 1.18, + "learning_rate": 8.164573186362471e-08, + "loss": 0.6775, + "step": 32369 + }, + { + "epoch": 1.18, + "learning_rate": 8.164206979895265e-08, + "loss": 0.7042, + "step": 32370 + }, + { + "epoch": 1.18, + "learning_rate": 8.163840773428058e-08, + "loss": 0.6921, + "step": 32371 + }, + { + "epoch": 1.18, + "learning_rate": 8.163474566960852e-08, + "loss": 0.6914, + "step": 32372 + }, + { + "epoch": 1.18, + "learning_rate": 8.163108360493646e-08, + "loss": 0.6974, + "step": 32373 + }, + { + "epoch": 1.18, + "learning_rate": 8.16274215402644e-08, + "loss": 0.691, + "step": 32374 + }, + { + "epoch": 1.18, + "learning_rate": 8.162375947559234e-08, + "loss": 0.6991, + "step": 32375 + }, + { + "epoch": 1.18, + "learning_rate": 8.162009741092028e-08, + "loss": 0.6854, + "step": 32376 + }, + { + "epoch": 1.18, + "learning_rate": 8.161643534624822e-08, + "loss": 0.708, + "step": 32377 + }, + { + "epoch": 1.18, + "learning_rate": 8.161277328157615e-08, + "loss": 0.6887, + "step": 32378 + }, + { + "epoch": 1.18, + "learning_rate": 8.160911121690409e-08, + "loss": 0.6898, + "step": 32379 + }, + { + "epoch": 1.18, + "learning_rate": 8.160544915223203e-08, + "loss": 0.6895, + "step": 32380 + }, + { + "epoch": 1.18, + "learning_rate": 8.160178708755995e-08, + "loss": 0.6903, + "step": 32381 + }, + { + "epoch": 1.18, + "learning_rate": 8.159812502288789e-08, + "loss": 0.699, + "step": 32382 + }, + { + "epoch": 1.18, + "learning_rate": 8.159446295821583e-08, + "loss": 0.6966, + "step": 32383 + }, + { + "epoch": 1.18, + "learning_rate": 8.159080089354377e-08, + "loss": 0.6862, + "step": 32384 + }, + { + "epoch": 1.18, + "learning_rate": 8.158713882887171e-08, + "loss": 0.6941, + "step": 32385 + }, + { + "epoch": 1.18, + "learning_rate": 8.158347676419965e-08, + "loss": 0.6866, + "step": 32386 + }, + { + "epoch": 1.18, + "learning_rate": 8.157981469952759e-08, + "loss": 0.6803, + "step": 32387 + }, + { + "epoch": 1.18, + "learning_rate": 8.157615263485553e-08, + "loss": 0.6947, + "step": 32388 + }, + { + "epoch": 1.19, + "learning_rate": 8.157249057018346e-08, + "loss": 0.6853, + "step": 32389 + }, + { + "epoch": 1.19, + "learning_rate": 8.15688285055114e-08, + "loss": 0.6939, + "step": 32390 + }, + { + "epoch": 1.19, + "learning_rate": 8.156516644083934e-08, + "loss": 0.693, + "step": 32391 + }, + { + "epoch": 1.19, + "learning_rate": 8.156150437616728e-08, + "loss": 0.6989, + "step": 32392 + }, + { + "epoch": 1.19, + "learning_rate": 8.155784231149522e-08, + "loss": 0.6849, + "step": 32393 + }, + { + "epoch": 1.19, + "learning_rate": 8.155418024682316e-08, + "loss": 0.7007, + "step": 32394 + }, + { + "epoch": 1.19, + "learning_rate": 8.15505181821511e-08, + "loss": 0.681, + "step": 32395 + }, + { + "epoch": 1.19, + "learning_rate": 8.154685611747904e-08, + "loss": 0.6985, + "step": 32396 + }, + { + "epoch": 1.19, + "learning_rate": 8.154319405280698e-08, + "loss": 0.6956, + "step": 32397 + }, + { + "epoch": 1.19, + "learning_rate": 8.15395319881349e-08, + "loss": 0.7031, + "step": 32398 + }, + { + "epoch": 1.19, + "learning_rate": 8.153586992346285e-08, + "loss": 0.6901, + "step": 32399 + }, + { + "epoch": 1.19, + "learning_rate": 8.153220785879079e-08, + "loss": 0.6971, + "step": 32400 + }, + { + "epoch": 1.19, + "learning_rate": 8.152854579411873e-08, + "loss": 0.6921, + "step": 32401 + }, + { + "epoch": 1.19, + "learning_rate": 8.152488372944667e-08, + "loss": 0.6931, + "step": 32402 + }, + { + "epoch": 1.19, + "learning_rate": 8.152122166477459e-08, + "loss": 0.7036, + "step": 32403 + }, + { + "epoch": 1.19, + "learning_rate": 8.151755960010253e-08, + "loss": 0.6946, + "step": 32404 + }, + { + "epoch": 1.19, + "learning_rate": 8.151389753543047e-08, + "loss": 0.6878, + "step": 32405 + }, + { + "epoch": 1.19, + "learning_rate": 8.151023547075841e-08, + "loss": 0.6858, + "step": 32406 + }, + { + "epoch": 1.19, + "learning_rate": 8.150657340608634e-08, + "loss": 0.6835, + "step": 32407 + }, + { + "epoch": 1.19, + "learning_rate": 8.150291134141428e-08, + "loss": 0.6965, + "step": 32408 + }, + { + "epoch": 1.19, + "learning_rate": 8.149924927674222e-08, + "loss": 0.6849, + "step": 32409 + }, + { + "epoch": 1.19, + "learning_rate": 8.149558721207016e-08, + "loss": 0.698, + "step": 32410 + }, + { + "epoch": 1.19, + "learning_rate": 8.14919251473981e-08, + "loss": 0.6774, + "step": 32411 + }, + { + "epoch": 1.19, + "learning_rate": 8.148826308272604e-08, + "loss": 0.6842, + "step": 32412 + }, + { + "epoch": 1.19, + "learning_rate": 8.148460101805398e-08, + "loss": 0.6822, + "step": 32413 + }, + { + "epoch": 1.19, + "learning_rate": 8.148093895338192e-08, + "loss": 0.6842, + "step": 32414 + }, + { + "epoch": 1.19, + "learning_rate": 8.147727688870986e-08, + "loss": 0.711, + "step": 32415 + }, + { + "epoch": 1.19, + "learning_rate": 8.147361482403778e-08, + "loss": 0.6853, + "step": 32416 + }, + { + "epoch": 1.19, + "learning_rate": 8.146995275936572e-08, + "loss": 0.6872, + "step": 32417 + }, + { + "epoch": 1.19, + "learning_rate": 8.146629069469366e-08, + "loss": 0.702, + "step": 32418 + }, + { + "epoch": 1.19, + "learning_rate": 8.14626286300216e-08, + "loss": 0.6906, + "step": 32419 + }, + { + "epoch": 1.19, + "learning_rate": 8.145896656534954e-08, + "loss": 0.6908, + "step": 32420 + }, + { + "epoch": 1.19, + "learning_rate": 8.145530450067748e-08, + "loss": 0.6982, + "step": 32421 + }, + { + "epoch": 1.19, + "learning_rate": 8.145164243600542e-08, + "loss": 0.688, + "step": 32422 + }, + { + "epoch": 1.19, + "learning_rate": 8.144798037133336e-08, + "loss": 0.6964, + "step": 32423 + }, + { + "epoch": 1.19, + "learning_rate": 8.144431830666129e-08, + "loss": 0.6882, + "step": 32424 + }, + { + "epoch": 1.19, + "learning_rate": 8.144065624198923e-08, + "loss": 0.6885, + "step": 32425 + }, + { + "epoch": 1.19, + "learning_rate": 8.143699417731717e-08, + "loss": 0.6984, + "step": 32426 + }, + { + "epoch": 1.19, + "learning_rate": 8.14333321126451e-08, + "loss": 0.6916, + "step": 32427 + }, + { + "epoch": 1.19, + "learning_rate": 8.142967004797304e-08, + "loss": 0.6997, + "step": 32428 + }, + { + "epoch": 1.19, + "learning_rate": 8.142600798330098e-08, + "loss": 0.6922, + "step": 32429 + }, + { + "epoch": 1.19, + "learning_rate": 8.142234591862892e-08, + "loss": 0.6971, + "step": 32430 + }, + { + "epoch": 1.19, + "learning_rate": 8.141868385395686e-08, + "loss": 0.6919, + "step": 32431 + }, + { + "epoch": 1.19, + "learning_rate": 8.14150217892848e-08, + "loss": 0.6903, + "step": 32432 + }, + { + "epoch": 1.19, + "learning_rate": 8.141135972461274e-08, + "loss": 0.6885, + "step": 32433 + }, + { + "epoch": 1.19, + "learning_rate": 8.140769765994066e-08, + "loss": 0.6911, + "step": 32434 + }, + { + "epoch": 1.19, + "learning_rate": 8.14040355952686e-08, + "loss": 0.681, + "step": 32435 + }, + { + "epoch": 1.19, + "learning_rate": 8.140037353059654e-08, + "loss": 0.7041, + "step": 32436 + }, + { + "epoch": 1.19, + "learning_rate": 8.139671146592448e-08, + "loss": 0.6926, + "step": 32437 + }, + { + "epoch": 1.19, + "learning_rate": 8.139304940125242e-08, + "loss": 0.7061, + "step": 32438 + }, + { + "epoch": 1.19, + "learning_rate": 8.138938733658036e-08, + "loss": 0.6911, + "step": 32439 + }, + { + "epoch": 1.19, + "learning_rate": 8.13857252719083e-08, + "loss": 0.7018, + "step": 32440 + }, + { + "epoch": 1.19, + "learning_rate": 8.138206320723624e-08, + "loss": 0.6919, + "step": 32441 + }, + { + "epoch": 1.19, + "learning_rate": 8.137840114256418e-08, + "loss": 0.6921, + "step": 32442 + }, + { + "epoch": 1.19, + "learning_rate": 8.137473907789211e-08, + "loss": 0.6953, + "step": 32443 + }, + { + "epoch": 1.19, + "learning_rate": 8.137107701322005e-08, + "loss": 0.6994, + "step": 32444 + }, + { + "epoch": 1.19, + "learning_rate": 8.136741494854799e-08, + "loss": 0.6938, + "step": 32445 + }, + { + "epoch": 1.19, + "learning_rate": 8.136375288387593e-08, + "loss": 0.694, + "step": 32446 + }, + { + "epoch": 1.19, + "learning_rate": 8.136009081920387e-08, + "loss": 0.6905, + "step": 32447 + }, + { + "epoch": 1.19, + "learning_rate": 8.135642875453181e-08, + "loss": 0.7061, + "step": 32448 + }, + { + "epoch": 1.19, + "learning_rate": 8.135276668985974e-08, + "loss": 0.7037, + "step": 32449 + }, + { + "epoch": 1.19, + "learning_rate": 8.134910462518768e-08, + "loss": 0.6884, + "step": 32450 + }, + { + "epoch": 1.19, + "learning_rate": 8.134544256051562e-08, + "loss": 0.6835, + "step": 32451 + }, + { + "epoch": 1.19, + "learning_rate": 8.134178049584354e-08, + "loss": 0.6898, + "step": 32452 + }, + { + "epoch": 1.19, + "learning_rate": 8.133811843117148e-08, + "loss": 0.6892, + "step": 32453 + }, + { + "epoch": 1.19, + "learning_rate": 8.133445636649942e-08, + "loss": 0.7051, + "step": 32454 + }, + { + "epoch": 1.19, + "learning_rate": 8.133079430182736e-08, + "loss": 0.6876, + "step": 32455 + }, + { + "epoch": 1.19, + "learning_rate": 8.13271322371553e-08, + "loss": 0.6834, + "step": 32456 + }, + { + "epoch": 1.19, + "learning_rate": 8.132347017248324e-08, + "loss": 0.6949, + "step": 32457 + }, + { + "epoch": 1.19, + "learning_rate": 8.131980810781118e-08, + "loss": 0.6965, + "step": 32458 + }, + { + "epoch": 1.19, + "learning_rate": 8.131614604313912e-08, + "loss": 0.6939, + "step": 32459 + }, + { + "epoch": 1.19, + "learning_rate": 8.131248397846706e-08, + "loss": 0.713, + "step": 32460 + }, + { + "epoch": 1.19, + "learning_rate": 8.130882191379499e-08, + "loss": 0.6897, + "step": 32461 + }, + { + "epoch": 1.19, + "learning_rate": 8.130515984912293e-08, + "loss": 0.6883, + "step": 32462 + }, + { + "epoch": 1.19, + "learning_rate": 8.130149778445087e-08, + "loss": 0.6955, + "step": 32463 + }, + { + "epoch": 1.19, + "learning_rate": 8.129783571977881e-08, + "loss": 0.6772, + "step": 32464 + }, + { + "epoch": 1.19, + "learning_rate": 8.129417365510675e-08, + "loss": 0.6954, + "step": 32465 + }, + { + "epoch": 1.19, + "learning_rate": 8.129051159043469e-08, + "loss": 0.683, + "step": 32466 + }, + { + "epoch": 1.19, + "learning_rate": 8.128684952576263e-08, + "loss": 0.7082, + "step": 32467 + }, + { + "epoch": 1.19, + "learning_rate": 8.128318746109057e-08, + "loss": 0.6845, + "step": 32468 + }, + { + "epoch": 1.19, + "learning_rate": 8.127952539641851e-08, + "loss": 0.6858, + "step": 32469 + }, + { + "epoch": 1.19, + "learning_rate": 8.127586333174643e-08, + "loss": 0.6956, + "step": 32470 + }, + { + "epoch": 1.19, + "learning_rate": 8.127220126707437e-08, + "loss": 0.6967, + "step": 32471 + }, + { + "epoch": 1.19, + "learning_rate": 8.126853920240231e-08, + "loss": 0.6897, + "step": 32472 + }, + { + "epoch": 1.19, + "learning_rate": 8.126487713773024e-08, + "loss": 0.69, + "step": 32473 + }, + { + "epoch": 1.19, + "learning_rate": 8.126121507305818e-08, + "loss": 0.6952, + "step": 32474 + }, + { + "epoch": 1.19, + "learning_rate": 8.125755300838612e-08, + "loss": 0.7057, + "step": 32475 + }, + { + "epoch": 1.19, + "learning_rate": 8.125389094371406e-08, + "loss": 0.6937, + "step": 32476 + }, + { + "epoch": 1.19, + "learning_rate": 8.1250228879042e-08, + "loss": 0.6942, + "step": 32477 + }, + { + "epoch": 1.19, + "learning_rate": 8.124656681436994e-08, + "loss": 0.6771, + "step": 32478 + }, + { + "epoch": 1.19, + "learning_rate": 8.124290474969787e-08, + "loss": 0.691, + "step": 32479 + }, + { + "epoch": 1.19, + "learning_rate": 8.123924268502581e-08, + "loss": 0.6947, + "step": 32480 + }, + { + "epoch": 1.19, + "learning_rate": 8.123558062035375e-08, + "loss": 0.6768, + "step": 32481 + }, + { + "epoch": 1.19, + "learning_rate": 8.123191855568169e-08, + "loss": 0.7093, + "step": 32482 + }, + { + "epoch": 1.19, + "learning_rate": 8.122825649100963e-08, + "loss": 0.6866, + "step": 32483 + }, + { + "epoch": 1.19, + "learning_rate": 8.122459442633757e-08, + "loss": 0.6864, + "step": 32484 + }, + { + "epoch": 1.19, + "learning_rate": 8.122093236166551e-08, + "loss": 0.6931, + "step": 32485 + }, + { + "epoch": 1.19, + "learning_rate": 8.121727029699345e-08, + "loss": 0.6971, + "step": 32486 + }, + { + "epoch": 1.19, + "learning_rate": 8.121360823232139e-08, + "loss": 0.6893, + "step": 32487 + }, + { + "epoch": 1.19, + "learning_rate": 8.120994616764931e-08, + "loss": 0.6861, + "step": 32488 + }, + { + "epoch": 1.19, + "learning_rate": 8.120628410297725e-08, + "loss": 0.6941, + "step": 32489 + }, + { + "epoch": 1.19, + "learning_rate": 8.12026220383052e-08, + "loss": 0.6975, + "step": 32490 + }, + { + "epoch": 1.19, + "learning_rate": 8.119895997363313e-08, + "loss": 0.7092, + "step": 32491 + }, + { + "epoch": 1.19, + "learning_rate": 8.119529790896107e-08, + "loss": 0.6827, + "step": 32492 + }, + { + "epoch": 1.19, + "learning_rate": 8.119163584428901e-08, + "loss": 0.6937, + "step": 32493 + }, + { + "epoch": 1.19, + "learning_rate": 8.118797377961695e-08, + "loss": 0.6943, + "step": 32494 + }, + { + "epoch": 1.19, + "learning_rate": 8.118431171494488e-08, + "loss": 0.6801, + "step": 32495 + }, + { + "epoch": 1.19, + "learning_rate": 8.118064965027282e-08, + "loss": 0.6754, + "step": 32496 + }, + { + "epoch": 1.19, + "learning_rate": 8.117698758560076e-08, + "loss": 0.6832, + "step": 32497 + }, + { + "epoch": 1.19, + "learning_rate": 8.117332552092869e-08, + "loss": 0.6989, + "step": 32498 + }, + { + "epoch": 1.19, + "learning_rate": 8.116966345625663e-08, + "loss": 0.6933, + "step": 32499 + }, + { + "epoch": 1.19, + "learning_rate": 8.116600139158457e-08, + "loss": 0.7042, + "step": 32500 + }, + { + "epoch": 1.19, + "learning_rate": 8.11623393269125e-08, + "loss": 0.6694, + "step": 32501 + }, + { + "epoch": 1.19, + "learning_rate": 8.115867726224045e-08, + "loss": 0.6885, + "step": 32502 + }, + { + "epoch": 1.19, + "learning_rate": 8.115501519756839e-08, + "loss": 0.7014, + "step": 32503 + }, + { + "epoch": 1.19, + "learning_rate": 8.115135313289633e-08, + "loss": 0.7039, + "step": 32504 + }, + { + "epoch": 1.19, + "learning_rate": 8.114769106822427e-08, + "loss": 0.6982, + "step": 32505 + }, + { + "epoch": 1.19, + "learning_rate": 8.114402900355219e-08, + "loss": 0.7021, + "step": 32506 + }, + { + "epoch": 1.19, + "learning_rate": 8.114036693888013e-08, + "loss": 0.6787, + "step": 32507 + }, + { + "epoch": 1.19, + "learning_rate": 8.113670487420807e-08, + "loss": 0.7158, + "step": 32508 + }, + { + "epoch": 1.19, + "learning_rate": 8.113304280953601e-08, + "loss": 0.6917, + "step": 32509 + }, + { + "epoch": 1.19, + "learning_rate": 8.112938074486395e-08, + "loss": 0.6898, + "step": 32510 + }, + { + "epoch": 1.19, + "learning_rate": 8.112571868019189e-08, + "loss": 0.6896, + "step": 32511 + }, + { + "epoch": 1.19, + "learning_rate": 8.112205661551983e-08, + "loss": 0.683, + "step": 32512 + }, + { + "epoch": 1.19, + "learning_rate": 8.111839455084777e-08, + "loss": 0.6927, + "step": 32513 + }, + { + "epoch": 1.19, + "learning_rate": 8.111473248617571e-08, + "loss": 0.6931, + "step": 32514 + }, + { + "epoch": 1.19, + "learning_rate": 8.111107042150364e-08, + "loss": 0.6843, + "step": 32515 + }, + { + "epoch": 1.19, + "learning_rate": 8.110740835683158e-08, + "loss": 0.6625, + "step": 32516 + }, + { + "epoch": 1.19, + "learning_rate": 8.110374629215952e-08, + "loss": 0.6996, + "step": 32517 + }, + { + "epoch": 1.19, + "learning_rate": 8.110008422748746e-08, + "loss": 0.6905, + "step": 32518 + }, + { + "epoch": 1.19, + "learning_rate": 8.10964221628154e-08, + "loss": 0.6887, + "step": 32519 + }, + { + "epoch": 1.19, + "learning_rate": 8.109276009814333e-08, + "loss": 0.6838, + "step": 32520 + }, + { + "epoch": 1.19, + "learning_rate": 8.108909803347127e-08, + "loss": 0.6902, + "step": 32521 + }, + { + "epoch": 1.19, + "learning_rate": 8.10854359687992e-08, + "loss": 0.7012, + "step": 32522 + }, + { + "epoch": 1.19, + "learning_rate": 8.108177390412714e-08, + "loss": 0.6828, + "step": 32523 + }, + { + "epoch": 1.19, + "learning_rate": 8.107811183945507e-08, + "loss": 0.6765, + "step": 32524 + }, + { + "epoch": 1.19, + "learning_rate": 8.107444977478301e-08, + "loss": 0.6865, + "step": 32525 + }, + { + "epoch": 1.19, + "learning_rate": 8.107078771011095e-08, + "loss": 0.6896, + "step": 32526 + }, + { + "epoch": 1.19, + "learning_rate": 8.106712564543889e-08, + "loss": 0.6736, + "step": 32527 + }, + { + "epoch": 1.19, + "learning_rate": 8.106346358076683e-08, + "loss": 0.6782, + "step": 32528 + }, + { + "epoch": 1.19, + "learning_rate": 8.105980151609477e-08, + "loss": 0.6839, + "step": 32529 + }, + { + "epoch": 1.19, + "learning_rate": 8.105613945142271e-08, + "loss": 0.6896, + "step": 32530 + }, + { + "epoch": 1.19, + "learning_rate": 8.105247738675065e-08, + "loss": 0.6824, + "step": 32531 + }, + { + "epoch": 1.19, + "learning_rate": 8.104881532207859e-08, + "loss": 0.6858, + "step": 32532 + }, + { + "epoch": 1.19, + "learning_rate": 8.104515325740652e-08, + "loss": 0.6962, + "step": 32533 + }, + { + "epoch": 1.19, + "learning_rate": 8.104149119273446e-08, + "loss": 0.7029, + "step": 32534 + }, + { + "epoch": 1.19, + "learning_rate": 8.10378291280624e-08, + "loss": 0.6811, + "step": 32535 + }, + { + "epoch": 1.19, + "learning_rate": 8.103416706339034e-08, + "loss": 0.7005, + "step": 32536 + }, + { + "epoch": 1.19, + "learning_rate": 8.103050499871828e-08, + "loss": 0.6988, + "step": 32537 + }, + { + "epoch": 1.19, + "learning_rate": 8.102684293404622e-08, + "loss": 0.7028, + "step": 32538 + }, + { + "epoch": 1.19, + "learning_rate": 8.102318086937416e-08, + "loss": 0.7009, + "step": 32539 + }, + { + "epoch": 1.19, + "learning_rate": 8.10195188047021e-08, + "loss": 0.6766, + "step": 32540 + }, + { + "epoch": 1.19, + "learning_rate": 8.101585674003004e-08, + "loss": 0.7, + "step": 32541 + }, + { + "epoch": 1.19, + "learning_rate": 8.101219467535796e-08, + "loss": 0.6869, + "step": 32542 + }, + { + "epoch": 1.19, + "learning_rate": 8.10085326106859e-08, + "loss": 0.7036, + "step": 32543 + }, + { + "epoch": 1.19, + "learning_rate": 8.100487054601383e-08, + "loss": 0.684, + "step": 32544 + }, + { + "epoch": 1.19, + "learning_rate": 8.100120848134177e-08, + "loss": 0.6996, + "step": 32545 + }, + { + "epoch": 1.19, + "learning_rate": 8.099754641666971e-08, + "loss": 0.6955, + "step": 32546 + }, + { + "epoch": 1.19, + "learning_rate": 8.099388435199765e-08, + "loss": 0.708, + "step": 32547 + }, + { + "epoch": 1.19, + "learning_rate": 8.099022228732559e-08, + "loss": 0.6941, + "step": 32548 + }, + { + "epoch": 1.19, + "learning_rate": 8.098656022265353e-08, + "loss": 0.6963, + "step": 32549 + }, + { + "epoch": 1.19, + "learning_rate": 8.098289815798146e-08, + "loss": 0.6914, + "step": 32550 + }, + { + "epoch": 1.19, + "learning_rate": 8.09792360933094e-08, + "loss": 0.6906, + "step": 32551 + }, + { + "epoch": 1.19, + "learning_rate": 8.097557402863734e-08, + "loss": 0.6836, + "step": 32552 + }, + { + "epoch": 1.19, + "learning_rate": 8.097191196396528e-08, + "loss": 0.7021, + "step": 32553 + }, + { + "epoch": 1.19, + "learning_rate": 8.096824989929322e-08, + "loss": 0.6875, + "step": 32554 + }, + { + "epoch": 1.19, + "learning_rate": 8.096458783462116e-08, + "loss": 0.71, + "step": 32555 + }, + { + "epoch": 1.19, + "learning_rate": 8.09609257699491e-08, + "loss": 0.6814, + "step": 32556 + }, + { + "epoch": 1.19, + "learning_rate": 8.095726370527704e-08, + "loss": 0.7004, + "step": 32557 + }, + { + "epoch": 1.19, + "learning_rate": 8.095360164060498e-08, + "loss": 0.6793, + "step": 32558 + }, + { + "epoch": 1.19, + "learning_rate": 8.09499395759329e-08, + "loss": 0.7211, + "step": 32559 + }, + { + "epoch": 1.19, + "learning_rate": 8.094627751126084e-08, + "loss": 0.7045, + "step": 32560 + }, + { + "epoch": 1.19, + "learning_rate": 8.094261544658878e-08, + "loss": 0.6937, + "step": 32561 + }, + { + "epoch": 1.19, + "learning_rate": 8.093895338191672e-08, + "loss": 0.6927, + "step": 32562 + }, + { + "epoch": 1.19, + "learning_rate": 8.093529131724466e-08, + "loss": 0.6786, + "step": 32563 + }, + { + "epoch": 1.19, + "learning_rate": 8.09316292525726e-08, + "loss": 0.6825, + "step": 32564 + }, + { + "epoch": 1.19, + "learning_rate": 8.092796718790054e-08, + "loss": 0.695, + "step": 32565 + }, + { + "epoch": 1.19, + "learning_rate": 8.092430512322847e-08, + "loss": 0.6839, + "step": 32566 + }, + { + "epoch": 1.19, + "learning_rate": 8.092064305855641e-08, + "loss": 0.6992, + "step": 32567 + }, + { + "epoch": 1.19, + "learning_rate": 8.091698099388435e-08, + "loss": 0.6946, + "step": 32568 + }, + { + "epoch": 1.19, + "learning_rate": 8.091331892921228e-08, + "loss": 0.6907, + "step": 32569 + }, + { + "epoch": 1.19, + "learning_rate": 8.090965686454022e-08, + "loss": 0.6936, + "step": 32570 + }, + { + "epoch": 1.19, + "learning_rate": 8.090599479986816e-08, + "loss": 0.6915, + "step": 32571 + }, + { + "epoch": 1.19, + "learning_rate": 8.09023327351961e-08, + "loss": 0.7048, + "step": 32572 + }, + { + "epoch": 1.19, + "learning_rate": 8.089867067052404e-08, + "loss": 0.7083, + "step": 32573 + }, + { + "epoch": 1.19, + "learning_rate": 8.089500860585198e-08, + "loss": 0.691, + "step": 32574 + }, + { + "epoch": 1.19, + "learning_rate": 8.089134654117992e-08, + "loss": 0.7177, + "step": 32575 + }, + { + "epoch": 1.19, + "learning_rate": 8.088768447650786e-08, + "loss": 0.6808, + "step": 32576 + }, + { + "epoch": 1.19, + "learning_rate": 8.088402241183578e-08, + "loss": 0.6964, + "step": 32577 + }, + { + "epoch": 1.19, + "learning_rate": 8.088036034716372e-08, + "loss": 0.6993, + "step": 32578 + }, + { + "epoch": 1.19, + "learning_rate": 8.087669828249166e-08, + "loss": 0.707, + "step": 32579 + }, + { + "epoch": 1.19, + "learning_rate": 8.08730362178196e-08, + "loss": 0.6874, + "step": 32580 + }, + { + "epoch": 1.19, + "learning_rate": 8.086937415314754e-08, + "loss": 0.6964, + "step": 32581 + }, + { + "epoch": 1.19, + "learning_rate": 8.086571208847548e-08, + "loss": 0.7034, + "step": 32582 + }, + { + "epoch": 1.19, + "learning_rate": 8.086205002380342e-08, + "loss": 0.7031, + "step": 32583 + }, + { + "epoch": 1.19, + "learning_rate": 8.085838795913136e-08, + "loss": 0.6901, + "step": 32584 + }, + { + "epoch": 1.19, + "learning_rate": 8.08547258944593e-08, + "loss": 0.6962, + "step": 32585 + }, + { + "epoch": 1.19, + "learning_rate": 8.085106382978723e-08, + "loss": 0.6794, + "step": 32586 + }, + { + "epoch": 1.19, + "learning_rate": 8.084740176511517e-08, + "loss": 0.699, + "step": 32587 + }, + { + "epoch": 1.19, + "learning_rate": 8.084373970044311e-08, + "loss": 0.6872, + "step": 32588 + }, + { + "epoch": 1.19, + "learning_rate": 8.084007763577105e-08, + "loss": 0.7041, + "step": 32589 + }, + { + "epoch": 1.19, + "learning_rate": 8.083641557109899e-08, + "loss": 0.7066, + "step": 32590 + }, + { + "epoch": 1.19, + "learning_rate": 8.083275350642691e-08, + "loss": 0.6971, + "step": 32591 + }, + { + "epoch": 1.19, + "learning_rate": 8.082909144175485e-08, + "loss": 0.6823, + "step": 32592 + }, + { + "epoch": 1.19, + "learning_rate": 8.08254293770828e-08, + "loss": 0.6881, + "step": 32593 + }, + { + "epoch": 1.19, + "learning_rate": 8.082176731241073e-08, + "loss": 0.679, + "step": 32594 + }, + { + "epoch": 1.19, + "learning_rate": 8.081810524773866e-08, + "loss": 0.6847, + "step": 32595 + }, + { + "epoch": 1.19, + "learning_rate": 8.08144431830666e-08, + "loss": 0.6796, + "step": 32596 + }, + { + "epoch": 1.19, + "learning_rate": 8.081078111839454e-08, + "loss": 0.6981, + "step": 32597 + }, + { + "epoch": 1.19, + "learning_rate": 8.080711905372248e-08, + "loss": 0.6741, + "step": 32598 + }, + { + "epoch": 1.19, + "learning_rate": 8.080345698905042e-08, + "loss": 0.7036, + "step": 32599 + }, + { + "epoch": 1.19, + "learning_rate": 8.079979492437836e-08, + "loss": 0.6922, + "step": 32600 + }, + { + "epoch": 1.19, + "learning_rate": 8.07961328597063e-08, + "loss": 0.688, + "step": 32601 + }, + { + "epoch": 1.19, + "learning_rate": 8.079247079503424e-08, + "loss": 0.6889, + "step": 32602 + }, + { + "epoch": 1.19, + "learning_rate": 8.078880873036218e-08, + "loss": 0.683, + "step": 32603 + }, + { + "epoch": 1.19, + "learning_rate": 8.078514666569011e-08, + "loss": 0.6882, + "step": 32604 + }, + { + "epoch": 1.19, + "learning_rate": 8.078148460101805e-08, + "loss": 0.6977, + "step": 32605 + }, + { + "epoch": 1.19, + "learning_rate": 8.077782253634599e-08, + "loss": 0.6764, + "step": 32606 + }, + { + "epoch": 1.19, + "learning_rate": 8.077416047167393e-08, + "loss": 0.693, + "step": 32607 + }, + { + "epoch": 1.19, + "learning_rate": 8.077049840700187e-08, + "loss": 0.6783, + "step": 32608 + }, + { + "epoch": 1.19, + "learning_rate": 8.076683634232981e-08, + "loss": 0.6929, + "step": 32609 + }, + { + "epoch": 1.19, + "learning_rate": 8.076317427765775e-08, + "loss": 0.6984, + "step": 32610 + }, + { + "epoch": 1.19, + "learning_rate": 8.075951221298569e-08, + "loss": 0.7097, + "step": 32611 + }, + { + "epoch": 1.19, + "learning_rate": 8.075585014831363e-08, + "loss": 0.6969, + "step": 32612 + }, + { + "epoch": 1.19, + "learning_rate": 8.075218808364155e-08, + "loss": 0.6805, + "step": 32613 + }, + { + "epoch": 1.19, + "learning_rate": 8.074852601896949e-08, + "loss": 0.7076, + "step": 32614 + }, + { + "epoch": 1.19, + "learning_rate": 8.074486395429742e-08, + "loss": 0.6853, + "step": 32615 + }, + { + "epoch": 1.19, + "learning_rate": 8.074120188962536e-08, + "loss": 0.7058, + "step": 32616 + }, + { + "epoch": 1.19, + "learning_rate": 8.07375398249533e-08, + "loss": 0.6956, + "step": 32617 + }, + { + "epoch": 1.19, + "learning_rate": 8.073387776028124e-08, + "loss": 0.7061, + "step": 32618 + }, + { + "epoch": 1.19, + "learning_rate": 8.073021569560918e-08, + "loss": 0.7028, + "step": 32619 + }, + { + "epoch": 1.19, + "learning_rate": 8.072655363093712e-08, + "loss": 0.6955, + "step": 32620 + }, + { + "epoch": 1.19, + "learning_rate": 8.072289156626506e-08, + "loss": 0.6966, + "step": 32621 + }, + { + "epoch": 1.19, + "learning_rate": 8.071922950159299e-08, + "loss": 0.7113, + "step": 32622 + }, + { + "epoch": 1.19, + "learning_rate": 8.071556743692093e-08, + "loss": 0.7015, + "step": 32623 + }, + { + "epoch": 1.19, + "learning_rate": 8.071190537224887e-08, + "loss": 0.6834, + "step": 32624 + }, + { + "epoch": 1.19, + "learning_rate": 8.07082433075768e-08, + "loss": 0.6845, + "step": 32625 + }, + { + "epoch": 1.19, + "learning_rate": 8.070458124290475e-08, + "loss": 0.694, + "step": 32626 + }, + { + "epoch": 1.19, + "learning_rate": 8.070091917823269e-08, + "loss": 0.6889, + "step": 32627 + }, + { + "epoch": 1.19, + "learning_rate": 8.069725711356063e-08, + "loss": 0.6934, + "step": 32628 + }, + { + "epoch": 1.19, + "learning_rate": 8.069359504888857e-08, + "loss": 0.6755, + "step": 32629 + }, + { + "epoch": 1.19, + "learning_rate": 8.06899329842165e-08, + "loss": 0.7166, + "step": 32630 + }, + { + "epoch": 1.19, + "learning_rate": 8.068627091954443e-08, + "loss": 0.6974, + "step": 32631 + }, + { + "epoch": 1.19, + "learning_rate": 8.068260885487237e-08, + "loss": 0.7109, + "step": 32632 + }, + { + "epoch": 1.19, + "learning_rate": 8.067894679020031e-08, + "loss": 0.6992, + "step": 32633 + }, + { + "epoch": 1.19, + "learning_rate": 8.067528472552825e-08, + "loss": 0.6751, + "step": 32634 + }, + { + "epoch": 1.19, + "learning_rate": 8.067162266085619e-08, + "loss": 0.6928, + "step": 32635 + }, + { + "epoch": 1.19, + "learning_rate": 8.066796059618413e-08, + "loss": 0.7007, + "step": 32636 + }, + { + "epoch": 1.19, + "learning_rate": 8.066429853151206e-08, + "loss": 0.7122, + "step": 32637 + }, + { + "epoch": 1.19, + "learning_rate": 8.066063646684e-08, + "loss": 0.6936, + "step": 32638 + }, + { + "epoch": 1.19, + "learning_rate": 8.065697440216794e-08, + "loss": 0.683, + "step": 32639 + }, + { + "epoch": 1.19, + "learning_rate": 8.065331233749587e-08, + "loss": 0.6847, + "step": 32640 + }, + { + "epoch": 1.19, + "learning_rate": 8.06496502728238e-08, + "loss": 0.6978, + "step": 32641 + }, + { + "epoch": 1.19, + "learning_rate": 8.064598820815175e-08, + "loss": 0.6913, + "step": 32642 + }, + { + "epoch": 1.19, + "learning_rate": 8.064232614347968e-08, + "loss": 0.6971, + "step": 32643 + }, + { + "epoch": 1.19, + "learning_rate": 8.063866407880762e-08, + "loss": 0.6862, + "step": 32644 + }, + { + "epoch": 1.19, + "learning_rate": 8.063500201413556e-08, + "loss": 0.6914, + "step": 32645 + }, + { + "epoch": 1.19, + "learning_rate": 8.06313399494635e-08, + "loss": 0.7008, + "step": 32646 + }, + { + "epoch": 1.19, + "learning_rate": 8.062767788479144e-08, + "loss": 0.6933, + "step": 32647 + }, + { + "epoch": 1.19, + "learning_rate": 8.062401582011938e-08, + "loss": 0.6955, + "step": 32648 + }, + { + "epoch": 1.19, + "learning_rate": 8.062035375544731e-08, + "loss": 0.6921, + "step": 32649 + }, + { + "epoch": 1.19, + "learning_rate": 8.061669169077525e-08, + "loss": 0.6828, + "step": 32650 + }, + { + "epoch": 1.19, + "learning_rate": 8.061302962610319e-08, + "loss": 0.6927, + "step": 32651 + }, + { + "epoch": 1.19, + "learning_rate": 8.060936756143113e-08, + "loss": 0.6619, + "step": 32652 + }, + { + "epoch": 1.19, + "learning_rate": 8.060570549675907e-08, + "loss": 0.6956, + "step": 32653 + }, + { + "epoch": 1.19, + "learning_rate": 8.060204343208701e-08, + "loss": 0.7011, + "step": 32654 + }, + { + "epoch": 1.19, + "learning_rate": 8.059838136741495e-08, + "loss": 0.6907, + "step": 32655 + }, + { + "epoch": 1.19, + "learning_rate": 8.059471930274289e-08, + "loss": 0.7044, + "step": 32656 + }, + { + "epoch": 1.19, + "learning_rate": 8.059105723807083e-08, + "loss": 0.6929, + "step": 32657 + }, + { + "epoch": 1.19, + "learning_rate": 8.058739517339876e-08, + "loss": 0.6955, + "step": 32658 + }, + { + "epoch": 1.19, + "learning_rate": 8.05837331087267e-08, + "loss": 0.6872, + "step": 32659 + }, + { + "epoch": 1.19, + "learning_rate": 8.058007104405464e-08, + "loss": 0.6954, + "step": 32660 + }, + { + "epoch": 1.19, + "learning_rate": 8.057640897938258e-08, + "loss": 0.6871, + "step": 32661 + }, + { + "epoch": 1.2, + "learning_rate": 8.05727469147105e-08, + "loss": 0.6999, + "step": 32662 + }, + { + "epoch": 1.2, + "learning_rate": 8.056908485003844e-08, + "loss": 0.7005, + "step": 32663 + }, + { + "epoch": 1.2, + "learning_rate": 8.056542278536638e-08, + "loss": 0.6857, + "step": 32664 + }, + { + "epoch": 1.2, + "learning_rate": 8.056176072069432e-08, + "loss": 0.7095, + "step": 32665 + }, + { + "epoch": 1.2, + "learning_rate": 8.055809865602226e-08, + "loss": 0.6897, + "step": 32666 + }, + { + "epoch": 1.2, + "learning_rate": 8.055443659135019e-08, + "loss": 0.7027, + "step": 32667 + }, + { + "epoch": 1.2, + "learning_rate": 8.055077452667813e-08, + "loss": 0.6929, + "step": 32668 + }, + { + "epoch": 1.2, + "learning_rate": 8.054711246200607e-08, + "loss": 0.715, + "step": 32669 + }, + { + "epoch": 1.2, + "learning_rate": 8.054345039733401e-08, + "loss": 0.6862, + "step": 32670 + }, + { + "epoch": 1.2, + "learning_rate": 8.053978833266195e-08, + "loss": 0.6879, + "step": 32671 + }, + { + "epoch": 1.2, + "learning_rate": 8.053612626798989e-08, + "loss": 0.6924, + "step": 32672 + }, + { + "epoch": 1.2, + "learning_rate": 8.053246420331783e-08, + "loss": 0.6907, + "step": 32673 + }, + { + "epoch": 1.2, + "learning_rate": 8.052880213864577e-08, + "loss": 0.6923, + "step": 32674 + }, + { + "epoch": 1.2, + "learning_rate": 8.052514007397371e-08, + "loss": 0.683, + "step": 32675 + }, + { + "epoch": 1.2, + "learning_rate": 8.052147800930164e-08, + "loss": 0.6735, + "step": 32676 + }, + { + "epoch": 1.2, + "learning_rate": 8.051781594462958e-08, + "loss": 0.6948, + "step": 32677 + }, + { + "epoch": 1.2, + "learning_rate": 8.051415387995752e-08, + "loss": 0.6938, + "step": 32678 + }, + { + "epoch": 1.2, + "learning_rate": 8.051049181528546e-08, + "loss": 0.6961, + "step": 32679 + }, + { + "epoch": 1.2, + "learning_rate": 8.05068297506134e-08, + "loss": 0.695, + "step": 32680 + }, + { + "epoch": 1.2, + "learning_rate": 8.050316768594134e-08, + "loss": 0.7063, + "step": 32681 + }, + { + "epoch": 1.2, + "learning_rate": 8.049950562126928e-08, + "loss": 0.6824, + "step": 32682 + }, + { + "epoch": 1.2, + "learning_rate": 8.049584355659722e-08, + "loss": 0.6928, + "step": 32683 + }, + { + "epoch": 1.2, + "learning_rate": 8.049218149192514e-08, + "loss": 0.6993, + "step": 32684 + }, + { + "epoch": 1.2, + "learning_rate": 8.048851942725308e-08, + "loss": 0.694, + "step": 32685 + }, + { + "epoch": 1.2, + "learning_rate": 8.048485736258101e-08, + "loss": 0.682, + "step": 32686 + }, + { + "epoch": 1.2, + "learning_rate": 8.048119529790895e-08, + "loss": 0.7053, + "step": 32687 + }, + { + "epoch": 1.2, + "learning_rate": 8.047753323323689e-08, + "loss": 0.6938, + "step": 32688 + }, + { + "epoch": 1.2, + "learning_rate": 8.047387116856483e-08, + "loss": 0.6996, + "step": 32689 + }, + { + "epoch": 1.2, + "learning_rate": 8.047020910389277e-08, + "loss": 0.6879, + "step": 32690 + }, + { + "epoch": 1.2, + "learning_rate": 8.046654703922071e-08, + "loss": 0.6901, + "step": 32691 + }, + { + "epoch": 1.2, + "learning_rate": 8.046288497454865e-08, + "loss": 0.6912, + "step": 32692 + }, + { + "epoch": 1.2, + "learning_rate": 8.045922290987659e-08, + "loss": 0.6903, + "step": 32693 + }, + { + "epoch": 1.2, + "learning_rate": 8.045556084520452e-08, + "loss": 0.6876, + "step": 32694 + }, + { + "epoch": 1.2, + "learning_rate": 8.045189878053246e-08, + "loss": 0.6839, + "step": 32695 + }, + { + "epoch": 1.2, + "learning_rate": 8.04482367158604e-08, + "loss": 0.6971, + "step": 32696 + }, + { + "epoch": 1.2, + "learning_rate": 8.044457465118834e-08, + "loss": 0.6901, + "step": 32697 + }, + { + "epoch": 1.2, + "learning_rate": 8.044091258651628e-08, + "loss": 0.6946, + "step": 32698 + }, + { + "epoch": 1.2, + "learning_rate": 8.043725052184422e-08, + "loss": 0.6929, + "step": 32699 + }, + { + "epoch": 1.2, + "learning_rate": 8.043358845717216e-08, + "loss": 0.6922, + "step": 32700 + }, + { + "epoch": 1.2, + "learning_rate": 8.04299263925001e-08, + "loss": 0.6833, + "step": 32701 + }, + { + "epoch": 1.2, + "learning_rate": 8.042626432782803e-08, + "loss": 0.6955, + "step": 32702 + }, + { + "epoch": 1.2, + "learning_rate": 8.042260226315596e-08, + "loss": 0.682, + "step": 32703 + }, + { + "epoch": 1.2, + "learning_rate": 8.04189401984839e-08, + "loss": 0.6844, + "step": 32704 + }, + { + "epoch": 1.2, + "learning_rate": 8.041527813381184e-08, + "loss": 0.7024, + "step": 32705 + }, + { + "epoch": 1.2, + "learning_rate": 8.041161606913978e-08, + "loss": 0.6888, + "step": 32706 + }, + { + "epoch": 1.2, + "learning_rate": 8.040795400446772e-08, + "loss": 0.6783, + "step": 32707 + }, + { + "epoch": 1.2, + "learning_rate": 8.040429193979565e-08, + "loss": 0.6744, + "step": 32708 + }, + { + "epoch": 1.2, + "learning_rate": 8.040062987512359e-08, + "loss": 0.6937, + "step": 32709 + }, + { + "epoch": 1.2, + "learning_rate": 8.039696781045153e-08, + "loss": 0.6964, + "step": 32710 + }, + { + "epoch": 1.2, + "learning_rate": 8.039330574577947e-08, + "loss": 0.6982, + "step": 32711 + }, + { + "epoch": 1.2, + "learning_rate": 8.03896436811074e-08, + "loss": 0.6892, + "step": 32712 + }, + { + "epoch": 1.2, + "learning_rate": 8.038598161643533e-08, + "loss": 0.6896, + "step": 32713 + }, + { + "epoch": 1.2, + "learning_rate": 8.038231955176327e-08, + "loss": 0.6962, + "step": 32714 + }, + { + "epoch": 1.2, + "learning_rate": 8.037865748709121e-08, + "loss": 0.6837, + "step": 32715 + }, + { + "epoch": 1.2, + "learning_rate": 8.037499542241915e-08, + "loss": 0.6889, + "step": 32716 + }, + { + "epoch": 1.2, + "learning_rate": 8.03713333577471e-08, + "loss": 0.7002, + "step": 32717 + }, + { + "epoch": 1.2, + "learning_rate": 8.036767129307503e-08, + "loss": 0.6883, + "step": 32718 + }, + { + "epoch": 1.2, + "learning_rate": 8.036400922840297e-08, + "loss": 0.6805, + "step": 32719 + }, + { + "epoch": 1.2, + "learning_rate": 8.036034716373091e-08, + "loss": 0.6922, + "step": 32720 + }, + { + "epoch": 1.2, + "learning_rate": 8.035668509905884e-08, + "loss": 0.6864, + "step": 32721 + }, + { + "epoch": 1.2, + "learning_rate": 8.035302303438678e-08, + "loss": 0.6983, + "step": 32722 + }, + { + "epoch": 1.2, + "learning_rate": 8.034936096971472e-08, + "loss": 0.6941, + "step": 32723 + }, + { + "epoch": 1.2, + "learning_rate": 8.034569890504266e-08, + "loss": 0.7, + "step": 32724 + }, + { + "epoch": 1.2, + "learning_rate": 8.03420368403706e-08, + "loss": 0.6911, + "step": 32725 + }, + { + "epoch": 1.2, + "learning_rate": 8.033837477569854e-08, + "loss": 0.6892, + "step": 32726 + }, + { + "epoch": 1.2, + "learning_rate": 8.033471271102648e-08, + "loss": 0.6871, + "step": 32727 + }, + { + "epoch": 1.2, + "learning_rate": 8.033105064635442e-08, + "loss": 0.6992, + "step": 32728 + }, + { + "epoch": 1.2, + "learning_rate": 8.032738858168236e-08, + "loss": 0.7067, + "step": 32729 + }, + { + "epoch": 1.2, + "learning_rate": 8.032372651701029e-08, + "loss": 0.699, + "step": 32730 + }, + { + "epoch": 1.2, + "learning_rate": 8.032006445233823e-08, + "loss": 0.6887, + "step": 32731 + }, + { + "epoch": 1.2, + "learning_rate": 8.031640238766617e-08, + "loss": 0.6962, + "step": 32732 + }, + { + "epoch": 1.2, + "learning_rate": 8.031274032299409e-08, + "loss": 0.6907, + "step": 32733 + }, + { + "epoch": 1.2, + "learning_rate": 8.030907825832203e-08, + "loss": 0.6943, + "step": 32734 + }, + { + "epoch": 1.2, + "learning_rate": 8.030541619364997e-08, + "loss": 0.6852, + "step": 32735 + }, + { + "epoch": 1.2, + "learning_rate": 8.030175412897791e-08, + "loss": 0.685, + "step": 32736 + }, + { + "epoch": 1.2, + "learning_rate": 8.029809206430585e-08, + "loss": 0.6975, + "step": 32737 + }, + { + "epoch": 1.2, + "learning_rate": 8.029442999963379e-08, + "loss": 0.6924, + "step": 32738 + }, + { + "epoch": 1.2, + "learning_rate": 8.029076793496172e-08, + "loss": 0.6994, + "step": 32739 + }, + { + "epoch": 1.2, + "learning_rate": 8.028710587028966e-08, + "loss": 0.6815, + "step": 32740 + }, + { + "epoch": 1.2, + "learning_rate": 8.02834438056176e-08, + "loss": 0.6868, + "step": 32741 + }, + { + "epoch": 1.2, + "learning_rate": 8.027978174094554e-08, + "loss": 0.6978, + "step": 32742 + }, + { + "epoch": 1.2, + "learning_rate": 8.027611967627348e-08, + "loss": 0.6803, + "step": 32743 + }, + { + "epoch": 1.2, + "learning_rate": 8.027245761160142e-08, + "loss": 0.6998, + "step": 32744 + }, + { + "epoch": 1.2, + "learning_rate": 8.026879554692936e-08, + "loss": 0.7037, + "step": 32745 + }, + { + "epoch": 1.2, + "learning_rate": 8.02651334822573e-08, + "loss": 0.6917, + "step": 32746 + }, + { + "epoch": 1.2, + "learning_rate": 8.026147141758524e-08, + "loss": 0.6853, + "step": 32747 + }, + { + "epoch": 1.2, + "learning_rate": 8.025780935291317e-08, + "loss": 0.689, + "step": 32748 + }, + { + "epoch": 1.2, + "learning_rate": 8.02541472882411e-08, + "loss": 0.6909, + "step": 32749 + }, + { + "epoch": 1.2, + "learning_rate": 8.025048522356905e-08, + "loss": 0.6905, + "step": 32750 + }, + { + "epoch": 1.2, + "learning_rate": 8.024682315889699e-08, + "loss": 0.7001, + "step": 32751 + }, + { + "epoch": 1.2, + "learning_rate": 8.024316109422493e-08, + "loss": 0.682, + "step": 32752 + }, + { + "epoch": 1.2, + "learning_rate": 8.023949902955287e-08, + "loss": 0.7047, + "step": 32753 + }, + { + "epoch": 1.2, + "learning_rate": 8.02358369648808e-08, + "loss": 0.69, + "step": 32754 + }, + { + "epoch": 1.2, + "learning_rate": 8.023217490020873e-08, + "loss": 0.7032, + "step": 32755 + }, + { + "epoch": 1.2, + "learning_rate": 8.022851283553667e-08, + "loss": 0.6785, + "step": 32756 + }, + { + "epoch": 1.2, + "learning_rate": 8.02248507708646e-08, + "loss": 0.6858, + "step": 32757 + }, + { + "epoch": 1.2, + "learning_rate": 8.022118870619254e-08, + "loss": 0.6856, + "step": 32758 + }, + { + "epoch": 1.2, + "learning_rate": 8.021752664152048e-08, + "loss": 0.7029, + "step": 32759 + }, + { + "epoch": 1.2, + "learning_rate": 8.021386457684842e-08, + "loss": 0.6944, + "step": 32760 + }, + { + "epoch": 1.2, + "learning_rate": 8.021020251217636e-08, + "loss": 0.683, + "step": 32761 + }, + { + "epoch": 1.2, + "learning_rate": 8.02065404475043e-08, + "loss": 0.7045, + "step": 32762 + }, + { + "epoch": 1.2, + "learning_rate": 8.020287838283224e-08, + "loss": 0.6958, + "step": 32763 + }, + { + "epoch": 1.2, + "learning_rate": 8.019921631816018e-08, + "loss": 0.6857, + "step": 32764 + }, + { + "epoch": 1.2, + "learning_rate": 8.019555425348812e-08, + "loss": 0.7003, + "step": 32765 + }, + { + "epoch": 1.2, + "learning_rate": 8.019189218881604e-08, + "loss": 0.7086, + "step": 32766 + }, + { + "epoch": 1.2, + "learning_rate": 8.018823012414398e-08, + "loss": 0.6922, + "step": 32767 + }, + { + "epoch": 1.2, + "learning_rate": 8.018456805947192e-08, + "loss": 0.6893, + "step": 32768 + }, + { + "epoch": 1.2, + "learning_rate": 8.018090599479986e-08, + "loss": 0.6737, + "step": 32769 + }, + { + "epoch": 1.2, + "learning_rate": 8.01772439301278e-08, + "loss": 0.6923, + "step": 32770 + }, + { + "epoch": 1.2, + "learning_rate": 8.017358186545574e-08, + "loss": 0.6983, + "step": 32771 + }, + { + "epoch": 1.2, + "learning_rate": 8.016991980078368e-08, + "loss": 0.6883, + "step": 32772 + }, + { + "epoch": 1.2, + "learning_rate": 8.016625773611162e-08, + "loss": 0.6927, + "step": 32773 + }, + { + "epoch": 1.2, + "learning_rate": 8.016259567143956e-08, + "loss": 0.6807, + "step": 32774 + }, + { + "epoch": 1.2, + "learning_rate": 8.015893360676749e-08, + "loss": 0.7006, + "step": 32775 + }, + { + "epoch": 1.2, + "learning_rate": 8.015527154209543e-08, + "loss": 0.6913, + "step": 32776 + }, + { + "epoch": 1.2, + "learning_rate": 8.015160947742337e-08, + "loss": 0.6822, + "step": 32777 + }, + { + "epoch": 1.2, + "learning_rate": 8.014794741275131e-08, + "loss": 0.684, + "step": 32778 + }, + { + "epoch": 1.2, + "learning_rate": 8.014428534807924e-08, + "loss": 0.685, + "step": 32779 + }, + { + "epoch": 1.2, + "learning_rate": 8.014062328340718e-08, + "loss": 0.7047, + "step": 32780 + }, + { + "epoch": 1.2, + "learning_rate": 8.013696121873512e-08, + "loss": 0.6993, + "step": 32781 + }, + { + "epoch": 1.2, + "learning_rate": 8.013329915406306e-08, + "loss": 0.6873, + "step": 32782 + }, + { + "epoch": 1.2, + "learning_rate": 8.0129637089391e-08, + "loss": 0.6912, + "step": 32783 + }, + { + "epoch": 1.2, + "learning_rate": 8.012597502471892e-08, + "loss": 0.6957, + "step": 32784 + }, + { + "epoch": 1.2, + "learning_rate": 8.012231296004686e-08, + "loss": 0.6898, + "step": 32785 + }, + { + "epoch": 1.2, + "learning_rate": 8.01186508953748e-08, + "loss": 0.6924, + "step": 32786 + }, + { + "epoch": 1.2, + "learning_rate": 8.011498883070274e-08, + "loss": 0.6932, + "step": 32787 + }, + { + "epoch": 1.2, + "learning_rate": 8.011132676603068e-08, + "loss": 0.7018, + "step": 32788 + }, + { + "epoch": 1.2, + "learning_rate": 8.010766470135862e-08, + "loss": 0.6974, + "step": 32789 + }, + { + "epoch": 1.2, + "learning_rate": 8.010400263668656e-08, + "loss": 0.6858, + "step": 32790 + }, + { + "epoch": 1.2, + "learning_rate": 8.01003405720145e-08, + "loss": 0.6982, + "step": 32791 + }, + { + "epoch": 1.2, + "learning_rate": 8.009667850734244e-08, + "loss": 0.7036, + "step": 32792 + }, + { + "epoch": 1.2, + "learning_rate": 8.009301644267037e-08, + "loss": 0.6877, + "step": 32793 + }, + { + "epoch": 1.2, + "learning_rate": 8.008935437799831e-08, + "loss": 0.6913, + "step": 32794 + }, + { + "epoch": 1.2, + "learning_rate": 8.008569231332625e-08, + "loss": 0.6882, + "step": 32795 + }, + { + "epoch": 1.2, + "learning_rate": 8.008203024865419e-08, + "loss": 0.6956, + "step": 32796 + }, + { + "epoch": 1.2, + "learning_rate": 8.007836818398213e-08, + "loss": 0.7005, + "step": 32797 + }, + { + "epoch": 1.2, + "learning_rate": 8.007470611931007e-08, + "loss": 0.7088, + "step": 32798 + }, + { + "epoch": 1.2, + "learning_rate": 8.007104405463801e-08, + "loss": 0.6978, + "step": 32799 + }, + { + "epoch": 1.2, + "learning_rate": 8.006738198996595e-08, + "loss": 0.6998, + "step": 32800 + }, + { + "epoch": 1.2, + "learning_rate": 8.006371992529388e-08, + "loss": 0.6871, + "step": 32801 + }, + { + "epoch": 1.2, + "learning_rate": 8.006005786062182e-08, + "loss": 0.6834, + "step": 32802 + }, + { + "epoch": 1.2, + "learning_rate": 8.005639579594976e-08, + "loss": 0.6921, + "step": 32803 + }, + { + "epoch": 1.2, + "learning_rate": 8.005273373127768e-08, + "loss": 0.6812, + "step": 32804 + }, + { + "epoch": 1.2, + "learning_rate": 8.004907166660562e-08, + "loss": 0.7016, + "step": 32805 + }, + { + "epoch": 1.2, + "learning_rate": 8.004540960193356e-08, + "loss": 0.7008, + "step": 32806 + }, + { + "epoch": 1.2, + "learning_rate": 8.00417475372615e-08, + "loss": 0.7031, + "step": 32807 + }, + { + "epoch": 1.2, + "learning_rate": 8.003808547258944e-08, + "loss": 0.7073, + "step": 32808 + }, + { + "epoch": 1.2, + "learning_rate": 8.003442340791738e-08, + "loss": 0.7024, + "step": 32809 + }, + { + "epoch": 1.2, + "learning_rate": 8.003076134324532e-08, + "loss": 0.7013, + "step": 32810 + }, + { + "epoch": 1.2, + "learning_rate": 8.002709927857325e-08, + "loss": 0.7016, + "step": 32811 + }, + { + "epoch": 1.2, + "learning_rate": 8.002343721390119e-08, + "loss": 0.6904, + "step": 32812 + }, + { + "epoch": 1.2, + "learning_rate": 8.001977514922913e-08, + "loss": 0.6924, + "step": 32813 + }, + { + "epoch": 1.2, + "learning_rate": 8.001611308455707e-08, + "loss": 0.702, + "step": 32814 + }, + { + "epoch": 1.2, + "learning_rate": 8.001245101988501e-08, + "loss": 0.6843, + "step": 32815 + }, + { + "epoch": 1.2, + "learning_rate": 8.000878895521295e-08, + "loss": 0.6856, + "step": 32816 + }, + { + "epoch": 1.2, + "learning_rate": 8.000512689054089e-08, + "loss": 0.6922, + "step": 32817 + }, + { + "epoch": 1.2, + "learning_rate": 8.000146482586883e-08, + "loss": 0.7045, + "step": 32818 + }, + { + "epoch": 1.2, + "learning_rate": 7.999780276119677e-08, + "loss": 0.6831, + "step": 32819 + }, + { + "epoch": 1.2, + "learning_rate": 7.99941406965247e-08, + "loss": 0.6958, + "step": 32820 + }, + { + "epoch": 1.2, + "learning_rate": 7.999047863185263e-08, + "loss": 0.7041, + "step": 32821 + }, + { + "epoch": 1.2, + "learning_rate": 7.998681656718057e-08, + "loss": 0.6897, + "step": 32822 + }, + { + "epoch": 1.2, + "learning_rate": 7.998315450250851e-08, + "loss": 0.6965, + "step": 32823 + }, + { + "epoch": 1.2, + "learning_rate": 7.997949243783645e-08, + "loss": 0.6942, + "step": 32824 + }, + { + "epoch": 1.2, + "learning_rate": 7.997583037316438e-08, + "loss": 0.6977, + "step": 32825 + }, + { + "epoch": 1.2, + "learning_rate": 7.997216830849232e-08, + "loss": 0.6848, + "step": 32826 + }, + { + "epoch": 1.2, + "learning_rate": 7.996850624382026e-08, + "loss": 0.6838, + "step": 32827 + }, + { + "epoch": 1.2, + "learning_rate": 7.99648441791482e-08, + "loss": 0.7015, + "step": 32828 + }, + { + "epoch": 1.2, + "learning_rate": 7.996118211447613e-08, + "loss": 0.7071, + "step": 32829 + }, + { + "epoch": 1.2, + "learning_rate": 7.995752004980407e-08, + "loss": 0.6963, + "step": 32830 + }, + { + "epoch": 1.2, + "learning_rate": 7.995385798513201e-08, + "loss": 0.6903, + "step": 32831 + }, + { + "epoch": 1.2, + "learning_rate": 7.995019592045995e-08, + "loss": 0.6973, + "step": 32832 + }, + { + "epoch": 1.2, + "learning_rate": 7.994653385578789e-08, + "loss": 0.6813, + "step": 32833 + }, + { + "epoch": 1.2, + "learning_rate": 7.994287179111583e-08, + "loss": 0.7045, + "step": 32834 + }, + { + "epoch": 1.2, + "learning_rate": 7.993920972644377e-08, + "loss": 0.6953, + "step": 32835 + }, + { + "epoch": 1.2, + "learning_rate": 7.993554766177171e-08, + "loss": 0.6963, + "step": 32836 + }, + { + "epoch": 1.2, + "learning_rate": 7.993188559709965e-08, + "loss": 0.6886, + "step": 32837 + }, + { + "epoch": 1.2, + "learning_rate": 7.992822353242757e-08, + "loss": 0.701, + "step": 32838 + }, + { + "epoch": 1.2, + "learning_rate": 7.992456146775551e-08, + "loss": 0.6924, + "step": 32839 + }, + { + "epoch": 1.2, + "learning_rate": 7.992089940308345e-08, + "loss": 0.6959, + "step": 32840 + }, + { + "epoch": 1.2, + "learning_rate": 7.99172373384114e-08, + "loss": 0.6946, + "step": 32841 + }, + { + "epoch": 1.2, + "learning_rate": 7.991357527373933e-08, + "loss": 0.6974, + "step": 32842 + }, + { + "epoch": 1.2, + "learning_rate": 7.990991320906727e-08, + "loss": 0.7159, + "step": 32843 + }, + { + "epoch": 1.2, + "learning_rate": 7.990625114439521e-08, + "loss": 0.7028, + "step": 32844 + }, + { + "epoch": 1.2, + "learning_rate": 7.990258907972315e-08, + "loss": 0.697, + "step": 32845 + }, + { + "epoch": 1.2, + "learning_rate": 7.98989270150511e-08, + "loss": 0.6796, + "step": 32846 + }, + { + "epoch": 1.2, + "learning_rate": 7.989526495037902e-08, + "loss": 0.6893, + "step": 32847 + }, + { + "epoch": 1.2, + "learning_rate": 7.989160288570696e-08, + "loss": 0.6906, + "step": 32848 + }, + { + "epoch": 1.2, + "learning_rate": 7.98879408210349e-08, + "loss": 0.6835, + "step": 32849 + }, + { + "epoch": 1.2, + "learning_rate": 7.988427875636283e-08, + "loss": 0.696, + "step": 32850 + }, + { + "epoch": 1.2, + "learning_rate": 7.988061669169077e-08, + "loss": 0.6889, + "step": 32851 + }, + { + "epoch": 1.2, + "learning_rate": 7.98769546270187e-08, + "loss": 0.7007, + "step": 32852 + }, + { + "epoch": 1.2, + "learning_rate": 7.987329256234665e-08, + "loss": 0.6955, + "step": 32853 + }, + { + "epoch": 1.2, + "learning_rate": 7.986963049767459e-08, + "loss": 0.7055, + "step": 32854 + }, + { + "epoch": 1.2, + "learning_rate": 7.986596843300253e-08, + "loss": 0.6995, + "step": 32855 + }, + { + "epoch": 1.2, + "learning_rate": 7.986230636833045e-08, + "loss": 0.6949, + "step": 32856 + }, + { + "epoch": 1.2, + "learning_rate": 7.985864430365839e-08, + "loss": 0.7085, + "step": 32857 + }, + { + "epoch": 1.2, + "learning_rate": 7.985498223898633e-08, + "loss": 0.6863, + "step": 32858 + }, + { + "epoch": 1.2, + "learning_rate": 7.985132017431427e-08, + "loss": 0.6898, + "step": 32859 + }, + { + "epoch": 1.2, + "learning_rate": 7.984765810964221e-08, + "loss": 0.6878, + "step": 32860 + }, + { + "epoch": 1.2, + "learning_rate": 7.984399604497015e-08, + "loss": 0.6977, + "step": 32861 + }, + { + "epoch": 1.2, + "learning_rate": 7.984033398029809e-08, + "loss": 0.6979, + "step": 32862 + }, + { + "epoch": 1.2, + "learning_rate": 7.983667191562603e-08, + "loss": 0.6877, + "step": 32863 + }, + { + "epoch": 1.2, + "learning_rate": 7.983300985095397e-08, + "loss": 0.6915, + "step": 32864 + }, + { + "epoch": 1.2, + "learning_rate": 7.98293477862819e-08, + "loss": 0.6928, + "step": 32865 + }, + { + "epoch": 1.2, + "learning_rate": 7.982568572160984e-08, + "loss": 0.6844, + "step": 32866 + }, + { + "epoch": 1.2, + "learning_rate": 7.982202365693778e-08, + "loss": 0.6985, + "step": 32867 + }, + { + "epoch": 1.2, + "learning_rate": 7.981836159226572e-08, + "loss": 0.6957, + "step": 32868 + }, + { + "epoch": 1.2, + "learning_rate": 7.981469952759366e-08, + "loss": 0.6882, + "step": 32869 + }, + { + "epoch": 1.2, + "learning_rate": 7.98110374629216e-08, + "loss": 0.7025, + "step": 32870 + }, + { + "epoch": 1.2, + "learning_rate": 7.980737539824954e-08, + "loss": 0.6836, + "step": 32871 + }, + { + "epoch": 1.2, + "learning_rate": 7.980371333357747e-08, + "loss": 0.6923, + "step": 32872 + }, + { + "epoch": 1.2, + "learning_rate": 7.98000512689054e-08, + "loss": 0.6943, + "step": 32873 + }, + { + "epoch": 1.2, + "learning_rate": 7.979638920423335e-08, + "loss": 0.6912, + "step": 32874 + }, + { + "epoch": 1.2, + "learning_rate": 7.979272713956127e-08, + "loss": 0.689, + "step": 32875 + }, + { + "epoch": 1.2, + "learning_rate": 7.978906507488921e-08, + "loss": 0.6937, + "step": 32876 + }, + { + "epoch": 1.2, + "learning_rate": 7.978540301021715e-08, + "loss": 0.6841, + "step": 32877 + }, + { + "epoch": 1.2, + "learning_rate": 7.978174094554509e-08, + "loss": 0.6974, + "step": 32878 + }, + { + "epoch": 1.2, + "learning_rate": 7.977807888087303e-08, + "loss": 0.6926, + "step": 32879 + }, + { + "epoch": 1.2, + "learning_rate": 7.977441681620097e-08, + "loss": 0.6797, + "step": 32880 + }, + { + "epoch": 1.2, + "learning_rate": 7.977075475152891e-08, + "loss": 0.6989, + "step": 32881 + }, + { + "epoch": 1.2, + "learning_rate": 7.976709268685685e-08, + "loss": 0.6829, + "step": 32882 + }, + { + "epoch": 1.2, + "learning_rate": 7.976343062218478e-08, + "loss": 0.703, + "step": 32883 + }, + { + "epoch": 1.2, + "learning_rate": 7.975976855751272e-08, + "loss": 0.6957, + "step": 32884 + }, + { + "epoch": 1.2, + "learning_rate": 7.975610649284066e-08, + "loss": 0.6957, + "step": 32885 + }, + { + "epoch": 1.2, + "learning_rate": 7.97524444281686e-08, + "loss": 0.6906, + "step": 32886 + }, + { + "epoch": 1.2, + "learning_rate": 7.974878236349654e-08, + "loss": 0.7046, + "step": 32887 + }, + { + "epoch": 1.2, + "learning_rate": 7.974512029882448e-08, + "loss": 0.6942, + "step": 32888 + }, + { + "epoch": 1.2, + "learning_rate": 7.974145823415242e-08, + "loss": 0.6845, + "step": 32889 + }, + { + "epoch": 1.2, + "learning_rate": 7.973779616948036e-08, + "loss": 0.6838, + "step": 32890 + }, + { + "epoch": 1.2, + "learning_rate": 7.97341341048083e-08, + "loss": 0.6829, + "step": 32891 + }, + { + "epoch": 1.2, + "learning_rate": 7.973047204013622e-08, + "loss": 0.6883, + "step": 32892 + }, + { + "epoch": 1.2, + "learning_rate": 7.972680997546416e-08, + "loss": 0.689, + "step": 32893 + }, + { + "epoch": 1.2, + "learning_rate": 7.97231479107921e-08, + "loss": 0.7027, + "step": 32894 + }, + { + "epoch": 1.2, + "learning_rate": 7.971948584612004e-08, + "loss": 0.6919, + "step": 32895 + }, + { + "epoch": 1.2, + "learning_rate": 7.971582378144797e-08, + "loss": 0.6825, + "step": 32896 + }, + { + "epoch": 1.2, + "learning_rate": 7.971216171677591e-08, + "loss": 0.7004, + "step": 32897 + }, + { + "epoch": 1.2, + "learning_rate": 7.970849965210385e-08, + "loss": 0.7005, + "step": 32898 + }, + { + "epoch": 1.2, + "learning_rate": 7.970483758743179e-08, + "loss": 0.6921, + "step": 32899 + }, + { + "epoch": 1.2, + "learning_rate": 7.970117552275973e-08, + "loss": 0.6998, + "step": 32900 + }, + { + "epoch": 1.2, + "learning_rate": 7.969751345808766e-08, + "loss": 0.687, + "step": 32901 + }, + { + "epoch": 1.2, + "learning_rate": 7.96938513934156e-08, + "loss": 0.6785, + "step": 32902 + }, + { + "epoch": 1.2, + "learning_rate": 7.969018932874354e-08, + "loss": 0.6939, + "step": 32903 + }, + { + "epoch": 1.2, + "learning_rate": 7.968652726407148e-08, + "loss": 0.693, + "step": 32904 + }, + { + "epoch": 1.2, + "learning_rate": 7.968286519939942e-08, + "loss": 0.6874, + "step": 32905 + }, + { + "epoch": 1.2, + "learning_rate": 7.967920313472736e-08, + "loss": 0.6858, + "step": 32906 + }, + { + "epoch": 1.2, + "learning_rate": 7.96755410700553e-08, + "loss": 0.7004, + "step": 32907 + }, + { + "epoch": 1.2, + "learning_rate": 7.967187900538324e-08, + "loss": 0.6916, + "step": 32908 + }, + { + "epoch": 1.2, + "learning_rate": 7.966821694071118e-08, + "loss": 0.7045, + "step": 32909 + }, + { + "epoch": 1.2, + "learning_rate": 7.96645548760391e-08, + "loss": 0.6967, + "step": 32910 + }, + { + "epoch": 1.2, + "learning_rate": 7.966089281136704e-08, + "loss": 0.6996, + "step": 32911 + }, + { + "epoch": 1.2, + "learning_rate": 7.965723074669498e-08, + "loss": 0.6926, + "step": 32912 + }, + { + "epoch": 1.2, + "learning_rate": 7.965356868202292e-08, + "loss": 0.6932, + "step": 32913 + }, + { + "epoch": 1.2, + "learning_rate": 7.964990661735086e-08, + "loss": 0.6867, + "step": 32914 + }, + { + "epoch": 1.2, + "learning_rate": 7.96462445526788e-08, + "loss": 0.7103, + "step": 32915 + }, + { + "epoch": 1.2, + "learning_rate": 7.964258248800674e-08, + "loss": 0.6851, + "step": 32916 + }, + { + "epoch": 1.2, + "learning_rate": 7.963892042333468e-08, + "loss": 0.6976, + "step": 32917 + }, + { + "epoch": 1.2, + "learning_rate": 7.963525835866261e-08, + "loss": 0.6979, + "step": 32918 + }, + { + "epoch": 1.2, + "learning_rate": 7.963159629399055e-08, + "loss": 0.7014, + "step": 32919 + }, + { + "epoch": 1.2, + "learning_rate": 7.962793422931849e-08, + "loss": 0.6979, + "step": 32920 + }, + { + "epoch": 1.2, + "learning_rate": 7.962427216464642e-08, + "loss": 0.6849, + "step": 32921 + }, + { + "epoch": 1.2, + "learning_rate": 7.962061009997436e-08, + "loss": 0.6852, + "step": 32922 + }, + { + "epoch": 1.2, + "learning_rate": 7.96169480353023e-08, + "loss": 0.6902, + "step": 32923 + }, + { + "epoch": 1.2, + "learning_rate": 7.961328597063024e-08, + "loss": 0.6976, + "step": 32924 + }, + { + "epoch": 1.2, + "learning_rate": 7.960962390595818e-08, + "loss": 0.6918, + "step": 32925 + }, + { + "epoch": 1.2, + "learning_rate": 7.960596184128612e-08, + "loss": 0.6936, + "step": 32926 + }, + { + "epoch": 1.2, + "learning_rate": 7.960229977661406e-08, + "loss": 0.7044, + "step": 32927 + }, + { + "epoch": 1.2, + "learning_rate": 7.959863771194198e-08, + "loss": 0.6936, + "step": 32928 + }, + { + "epoch": 1.2, + "learning_rate": 7.959497564726992e-08, + "loss": 0.7053, + "step": 32929 + }, + { + "epoch": 1.2, + "learning_rate": 7.959131358259786e-08, + "loss": 0.6698, + "step": 32930 + }, + { + "epoch": 1.2, + "learning_rate": 7.95876515179258e-08, + "loss": 0.7027, + "step": 32931 + }, + { + "epoch": 1.2, + "learning_rate": 7.958398945325374e-08, + "loss": 0.6968, + "step": 32932 + }, + { + "epoch": 1.2, + "learning_rate": 7.958032738858168e-08, + "loss": 0.7076, + "step": 32933 + }, + { + "epoch": 1.2, + "learning_rate": 7.957666532390962e-08, + "loss": 0.6999, + "step": 32934 + }, + { + "epoch": 1.2, + "learning_rate": 7.957300325923756e-08, + "loss": 0.6866, + "step": 32935 + }, + { + "epoch": 1.21, + "learning_rate": 7.956934119456549e-08, + "loss": 0.6935, + "step": 32936 + }, + { + "epoch": 1.21, + "learning_rate": 7.956567912989343e-08, + "loss": 0.6908, + "step": 32937 + }, + { + "epoch": 1.21, + "learning_rate": 7.956201706522137e-08, + "loss": 0.6922, + "step": 32938 + }, + { + "epoch": 1.21, + "learning_rate": 7.955835500054931e-08, + "loss": 0.7048, + "step": 32939 + }, + { + "epoch": 1.21, + "learning_rate": 7.955469293587725e-08, + "loss": 0.6852, + "step": 32940 + }, + { + "epoch": 1.21, + "learning_rate": 7.955103087120519e-08, + "loss": 0.6938, + "step": 32941 + }, + { + "epoch": 1.21, + "learning_rate": 7.954736880653313e-08, + "loss": 0.7051, + "step": 32942 + }, + { + "epoch": 1.21, + "learning_rate": 7.954370674186105e-08, + "loss": 0.6863, + "step": 32943 + }, + { + "epoch": 1.21, + "learning_rate": 7.9540044677189e-08, + "loss": 0.6756, + "step": 32944 + }, + { + "epoch": 1.21, + "learning_rate": 7.953638261251692e-08, + "loss": 0.6986, + "step": 32945 + }, + { + "epoch": 1.21, + "learning_rate": 7.953272054784486e-08, + "loss": 0.7077, + "step": 32946 + }, + { + "epoch": 1.21, + "learning_rate": 7.95290584831728e-08, + "loss": 0.7014, + "step": 32947 + }, + { + "epoch": 1.21, + "learning_rate": 7.952539641850074e-08, + "loss": 0.6908, + "step": 32948 + }, + { + "epoch": 1.21, + "learning_rate": 7.952173435382868e-08, + "loss": 0.6965, + "step": 32949 + }, + { + "epoch": 1.21, + "learning_rate": 7.951807228915662e-08, + "loss": 0.6971, + "step": 32950 + }, + { + "epoch": 1.21, + "learning_rate": 7.951441022448456e-08, + "loss": 0.6863, + "step": 32951 + }, + { + "epoch": 1.21, + "learning_rate": 7.95107481598125e-08, + "loss": 0.6947, + "step": 32952 + }, + { + "epoch": 1.21, + "learning_rate": 7.950708609514044e-08, + "loss": 0.6909, + "step": 32953 + }, + { + "epoch": 1.21, + "learning_rate": 7.950342403046837e-08, + "loss": 0.6961, + "step": 32954 + }, + { + "epoch": 1.21, + "learning_rate": 7.949976196579631e-08, + "loss": 0.696, + "step": 32955 + }, + { + "epoch": 1.21, + "learning_rate": 7.949609990112425e-08, + "loss": 0.6827, + "step": 32956 + }, + { + "epoch": 1.21, + "learning_rate": 7.949243783645219e-08, + "loss": 0.688, + "step": 32957 + }, + { + "epoch": 1.21, + "learning_rate": 7.948877577178013e-08, + "loss": 0.6751, + "step": 32958 + }, + { + "epoch": 1.21, + "learning_rate": 7.948511370710807e-08, + "loss": 0.6909, + "step": 32959 + }, + { + "epoch": 1.21, + "learning_rate": 7.948145164243601e-08, + "loss": 0.6976, + "step": 32960 + }, + { + "epoch": 1.21, + "learning_rate": 7.947778957776395e-08, + "loss": 0.6931, + "step": 32961 + }, + { + "epoch": 1.21, + "learning_rate": 7.947412751309189e-08, + "loss": 0.6903, + "step": 32962 + }, + { + "epoch": 1.21, + "learning_rate": 7.947046544841981e-08, + "loss": 0.6975, + "step": 32963 + }, + { + "epoch": 1.21, + "learning_rate": 7.946680338374775e-08, + "loss": 0.71, + "step": 32964 + }, + { + "epoch": 1.21, + "learning_rate": 7.94631413190757e-08, + "loss": 0.6918, + "step": 32965 + }, + { + "epoch": 1.21, + "learning_rate": 7.945947925440363e-08, + "loss": 0.69, + "step": 32966 + }, + { + "epoch": 1.21, + "learning_rate": 7.945581718973156e-08, + "loss": 0.7018, + "step": 32967 + }, + { + "epoch": 1.21, + "learning_rate": 7.94521551250595e-08, + "loss": 0.6847, + "step": 32968 + }, + { + "epoch": 1.21, + "learning_rate": 7.944849306038744e-08, + "loss": 0.6894, + "step": 32969 + }, + { + "epoch": 1.21, + "learning_rate": 7.944483099571538e-08, + "loss": 0.6989, + "step": 32970 + }, + { + "epoch": 1.21, + "learning_rate": 7.944116893104332e-08, + "loss": 0.6995, + "step": 32971 + }, + { + "epoch": 1.21, + "learning_rate": 7.943750686637125e-08, + "loss": 0.6852, + "step": 32972 + }, + { + "epoch": 1.21, + "learning_rate": 7.943384480169919e-08, + "loss": 0.6786, + "step": 32973 + }, + { + "epoch": 1.21, + "learning_rate": 7.943018273702713e-08, + "loss": 0.6899, + "step": 32974 + }, + { + "epoch": 1.21, + "learning_rate": 7.942652067235507e-08, + "loss": 0.7051, + "step": 32975 + }, + { + "epoch": 1.21, + "learning_rate": 7.9422858607683e-08, + "loss": 0.6841, + "step": 32976 + }, + { + "epoch": 1.21, + "learning_rate": 7.941919654301095e-08, + "loss": 0.7046, + "step": 32977 + }, + { + "epoch": 1.21, + "learning_rate": 7.941553447833889e-08, + "loss": 0.6921, + "step": 32978 + }, + { + "epoch": 1.21, + "learning_rate": 7.941187241366683e-08, + "loss": 0.6967, + "step": 32979 + }, + { + "epoch": 1.21, + "learning_rate": 7.940821034899477e-08, + "loss": 0.6909, + "step": 32980 + }, + { + "epoch": 1.21, + "learning_rate": 7.940454828432269e-08, + "loss": 0.6842, + "step": 32981 + }, + { + "epoch": 1.21, + "learning_rate": 7.940088621965063e-08, + "loss": 0.6976, + "step": 32982 + }, + { + "epoch": 1.21, + "learning_rate": 7.939722415497857e-08, + "loss": 0.67, + "step": 32983 + }, + { + "epoch": 1.21, + "learning_rate": 7.939356209030651e-08, + "loss": 0.696, + "step": 32984 + }, + { + "epoch": 1.21, + "learning_rate": 7.938990002563445e-08, + "loss": 0.6961, + "step": 32985 + }, + { + "epoch": 1.21, + "learning_rate": 7.938623796096239e-08, + "loss": 0.6885, + "step": 32986 + }, + { + "epoch": 1.21, + "learning_rate": 7.938257589629033e-08, + "loss": 0.6945, + "step": 32987 + }, + { + "epoch": 1.21, + "learning_rate": 7.937891383161827e-08, + "loss": 0.6842, + "step": 32988 + }, + { + "epoch": 1.21, + "learning_rate": 7.93752517669462e-08, + "loss": 0.6847, + "step": 32989 + }, + { + "epoch": 1.21, + "learning_rate": 7.937158970227414e-08, + "loss": 0.7069, + "step": 32990 + }, + { + "epoch": 1.21, + "learning_rate": 7.936792763760208e-08, + "loss": 0.6953, + "step": 32991 + }, + { + "epoch": 1.21, + "learning_rate": 7.936426557293e-08, + "loss": 0.6869, + "step": 32992 + }, + { + "epoch": 1.21, + "learning_rate": 7.936060350825795e-08, + "loss": 0.6898, + "step": 32993 + }, + { + "epoch": 1.21, + "learning_rate": 7.935694144358589e-08, + "loss": 0.6945, + "step": 32994 + }, + { + "epoch": 1.21, + "learning_rate": 7.935327937891383e-08, + "loss": 0.6984, + "step": 32995 + }, + { + "epoch": 1.21, + "learning_rate": 7.934961731424177e-08, + "loss": 0.6991, + "step": 32996 + }, + { + "epoch": 1.21, + "learning_rate": 7.93459552495697e-08, + "loss": 0.6917, + "step": 32997 + }, + { + "epoch": 1.21, + "learning_rate": 7.934229318489764e-08, + "loss": 0.6923, + "step": 32998 + }, + { + "epoch": 1.21, + "learning_rate": 7.933863112022557e-08, + "loss": 0.6835, + "step": 32999 + }, + { + "epoch": 1.21, + "learning_rate": 7.933496905555351e-08, + "loss": 0.6988, + "step": 33000 + }, + { + "epoch": 1.21, + "learning_rate": 7.933130699088145e-08, + "loss": 0.6962, + "step": 33001 + }, + { + "epoch": 1.21, + "learning_rate": 7.932764492620939e-08, + "loss": 0.6886, + "step": 33002 + }, + { + "epoch": 1.21, + "learning_rate": 7.932398286153733e-08, + "loss": 0.6947, + "step": 33003 + }, + { + "epoch": 1.21, + "learning_rate": 7.932032079686527e-08, + "loss": 0.6746, + "step": 33004 + }, + { + "epoch": 1.21, + "learning_rate": 7.931665873219321e-08, + "loss": 0.6813, + "step": 33005 + }, + { + "epoch": 1.21, + "learning_rate": 7.931299666752115e-08, + "loss": 0.69, + "step": 33006 + }, + { + "epoch": 1.21, + "learning_rate": 7.930933460284909e-08, + "loss": 0.6863, + "step": 33007 + }, + { + "epoch": 1.21, + "learning_rate": 7.930567253817702e-08, + "loss": 0.6939, + "step": 33008 + }, + { + "epoch": 1.21, + "learning_rate": 7.930201047350496e-08, + "loss": 0.6939, + "step": 33009 + }, + { + "epoch": 1.21, + "learning_rate": 7.92983484088329e-08, + "loss": 0.6882, + "step": 33010 + }, + { + "epoch": 1.21, + "learning_rate": 7.929468634416084e-08, + "loss": 0.6931, + "step": 33011 + }, + { + "epoch": 1.21, + "learning_rate": 7.929102427948878e-08, + "loss": 0.682, + "step": 33012 + }, + { + "epoch": 1.21, + "learning_rate": 7.928736221481672e-08, + "loss": 0.7, + "step": 33013 + }, + { + "epoch": 1.21, + "learning_rate": 7.928370015014464e-08, + "loss": 0.6756, + "step": 33014 + }, + { + "epoch": 1.21, + "learning_rate": 7.928003808547258e-08, + "loss": 0.7027, + "step": 33015 + }, + { + "epoch": 1.21, + "learning_rate": 7.927637602080052e-08, + "loss": 0.7038, + "step": 33016 + }, + { + "epoch": 1.21, + "learning_rate": 7.927271395612845e-08, + "loss": 0.701, + "step": 33017 + }, + { + "epoch": 1.21, + "learning_rate": 7.926905189145639e-08, + "loss": 0.6917, + "step": 33018 + }, + { + "epoch": 1.21, + "learning_rate": 7.926538982678433e-08, + "loss": 0.6893, + "step": 33019 + }, + { + "epoch": 1.21, + "learning_rate": 7.926172776211227e-08, + "loss": 0.691, + "step": 33020 + }, + { + "epoch": 1.21, + "learning_rate": 7.925806569744021e-08, + "loss": 0.6885, + "step": 33021 + }, + { + "epoch": 1.21, + "learning_rate": 7.925440363276815e-08, + "loss": 0.7013, + "step": 33022 + }, + { + "epoch": 1.21, + "learning_rate": 7.925074156809609e-08, + "loss": 0.6744, + "step": 33023 + }, + { + "epoch": 1.21, + "learning_rate": 7.924707950342403e-08, + "loss": 0.6877, + "step": 33024 + }, + { + "epoch": 1.21, + "learning_rate": 7.924341743875197e-08, + "loss": 0.7056, + "step": 33025 + }, + { + "epoch": 1.21, + "learning_rate": 7.92397553740799e-08, + "loss": 0.6911, + "step": 33026 + }, + { + "epoch": 1.21, + "learning_rate": 7.923609330940784e-08, + "loss": 0.6753, + "step": 33027 + }, + { + "epoch": 1.21, + "learning_rate": 7.923243124473578e-08, + "loss": 0.6671, + "step": 33028 + }, + { + "epoch": 1.21, + "learning_rate": 7.922876918006372e-08, + "loss": 0.7044, + "step": 33029 + }, + { + "epoch": 1.21, + "learning_rate": 7.922510711539166e-08, + "loss": 0.6953, + "step": 33030 + }, + { + "epoch": 1.21, + "learning_rate": 7.92214450507196e-08, + "loss": 0.7033, + "step": 33031 + }, + { + "epoch": 1.21, + "learning_rate": 7.921778298604754e-08, + "loss": 0.6919, + "step": 33032 + }, + { + "epoch": 1.21, + "learning_rate": 7.921412092137548e-08, + "loss": 0.6994, + "step": 33033 + }, + { + "epoch": 1.21, + "learning_rate": 7.921045885670342e-08, + "loss": 0.682, + "step": 33034 + }, + { + "epoch": 1.21, + "learning_rate": 7.920679679203134e-08, + "loss": 0.6841, + "step": 33035 + }, + { + "epoch": 1.21, + "learning_rate": 7.920313472735928e-08, + "loss": 0.6939, + "step": 33036 + }, + { + "epoch": 1.21, + "learning_rate": 7.919947266268722e-08, + "loss": 0.6923, + "step": 33037 + }, + { + "epoch": 1.21, + "learning_rate": 7.919581059801515e-08, + "loss": 0.6943, + "step": 33038 + }, + { + "epoch": 1.21, + "learning_rate": 7.919214853334309e-08, + "loss": 0.6966, + "step": 33039 + }, + { + "epoch": 1.21, + "learning_rate": 7.918848646867103e-08, + "loss": 0.6941, + "step": 33040 + }, + { + "epoch": 1.21, + "learning_rate": 7.918482440399897e-08, + "loss": 0.7055, + "step": 33041 + }, + { + "epoch": 1.21, + "learning_rate": 7.918116233932691e-08, + "loss": 0.7022, + "step": 33042 + }, + { + "epoch": 1.21, + "learning_rate": 7.917750027465485e-08, + "loss": 0.6919, + "step": 33043 + }, + { + "epoch": 1.21, + "learning_rate": 7.917383820998278e-08, + "loss": 0.7002, + "step": 33044 + }, + { + "epoch": 1.21, + "learning_rate": 7.917017614531072e-08, + "loss": 0.7098, + "step": 33045 + }, + { + "epoch": 1.21, + "learning_rate": 7.916651408063866e-08, + "loss": 0.6885, + "step": 33046 + }, + { + "epoch": 1.21, + "learning_rate": 7.91628520159666e-08, + "loss": 0.6719, + "step": 33047 + }, + { + "epoch": 1.21, + "learning_rate": 7.915918995129454e-08, + "loss": 0.6985, + "step": 33048 + }, + { + "epoch": 1.21, + "learning_rate": 7.915552788662248e-08, + "loss": 0.6946, + "step": 33049 + }, + { + "epoch": 1.21, + "learning_rate": 7.915186582195042e-08, + "loss": 0.7022, + "step": 33050 + }, + { + "epoch": 1.21, + "learning_rate": 7.914820375727836e-08, + "loss": 0.6884, + "step": 33051 + }, + { + "epoch": 1.21, + "learning_rate": 7.91445416926063e-08, + "loss": 0.693, + "step": 33052 + }, + { + "epoch": 1.21, + "learning_rate": 7.914087962793422e-08, + "loss": 0.7108, + "step": 33053 + }, + { + "epoch": 1.21, + "learning_rate": 7.913721756326216e-08, + "loss": 0.6907, + "step": 33054 + }, + { + "epoch": 1.21, + "learning_rate": 7.91335554985901e-08, + "loss": 0.7089, + "step": 33055 + }, + { + "epoch": 1.21, + "learning_rate": 7.912989343391804e-08, + "loss": 0.6927, + "step": 33056 + }, + { + "epoch": 1.21, + "learning_rate": 7.912623136924598e-08, + "loss": 0.6958, + "step": 33057 + }, + { + "epoch": 1.21, + "learning_rate": 7.912256930457392e-08, + "loss": 0.6775, + "step": 33058 + }, + { + "epoch": 1.21, + "learning_rate": 7.911890723990186e-08, + "loss": 0.7006, + "step": 33059 + }, + { + "epoch": 1.21, + "learning_rate": 7.911524517522979e-08, + "loss": 0.7011, + "step": 33060 + }, + { + "epoch": 1.21, + "learning_rate": 7.911158311055773e-08, + "loss": 0.6913, + "step": 33061 + }, + { + "epoch": 1.21, + "learning_rate": 7.910792104588567e-08, + "loss": 0.6935, + "step": 33062 + }, + { + "epoch": 1.21, + "learning_rate": 7.91042589812136e-08, + "loss": 0.6771, + "step": 33063 + }, + { + "epoch": 1.21, + "learning_rate": 7.910059691654153e-08, + "loss": 0.6794, + "step": 33064 + }, + { + "epoch": 1.21, + "learning_rate": 7.909693485186947e-08, + "loss": 0.6906, + "step": 33065 + }, + { + "epoch": 1.21, + "learning_rate": 7.909327278719741e-08, + "loss": 0.7028, + "step": 33066 + }, + { + "epoch": 1.21, + "learning_rate": 7.908961072252535e-08, + "loss": 0.684, + "step": 33067 + }, + { + "epoch": 1.21, + "learning_rate": 7.90859486578533e-08, + "loss": 0.6808, + "step": 33068 + }, + { + "epoch": 1.21, + "learning_rate": 7.908228659318123e-08, + "loss": 0.6954, + "step": 33069 + }, + { + "epoch": 1.21, + "learning_rate": 7.907862452850917e-08, + "loss": 0.6879, + "step": 33070 + }, + { + "epoch": 1.21, + "learning_rate": 7.90749624638371e-08, + "loss": 0.7038, + "step": 33071 + }, + { + "epoch": 1.21, + "learning_rate": 7.907130039916504e-08, + "loss": 0.7017, + "step": 33072 + }, + { + "epoch": 1.21, + "learning_rate": 7.906763833449298e-08, + "loss": 0.6896, + "step": 33073 + }, + { + "epoch": 1.21, + "learning_rate": 7.906397626982092e-08, + "loss": 0.6891, + "step": 33074 + }, + { + "epoch": 1.21, + "learning_rate": 7.906031420514886e-08, + "loss": 0.6803, + "step": 33075 + }, + { + "epoch": 1.21, + "learning_rate": 7.90566521404768e-08, + "loss": 0.6941, + "step": 33076 + }, + { + "epoch": 1.21, + "learning_rate": 7.905299007580474e-08, + "loss": 0.6897, + "step": 33077 + }, + { + "epoch": 1.21, + "learning_rate": 7.904932801113268e-08, + "loss": 0.706, + "step": 33078 + }, + { + "epoch": 1.21, + "learning_rate": 7.904566594646062e-08, + "loss": 0.6978, + "step": 33079 + }, + { + "epoch": 1.21, + "learning_rate": 7.904200388178855e-08, + "loss": 0.6791, + "step": 33080 + }, + { + "epoch": 1.21, + "learning_rate": 7.903834181711649e-08, + "loss": 0.6787, + "step": 33081 + }, + { + "epoch": 1.21, + "learning_rate": 7.903467975244443e-08, + "loss": 0.7174, + "step": 33082 + }, + { + "epoch": 1.21, + "learning_rate": 7.903101768777237e-08, + "loss": 0.6916, + "step": 33083 + }, + { + "epoch": 1.21, + "learning_rate": 7.902735562310031e-08, + "loss": 0.6933, + "step": 33084 + }, + { + "epoch": 1.21, + "learning_rate": 7.902369355842823e-08, + "loss": 0.7125, + "step": 33085 + }, + { + "epoch": 1.21, + "learning_rate": 7.902003149375617e-08, + "loss": 0.6872, + "step": 33086 + }, + { + "epoch": 1.21, + "learning_rate": 7.901636942908411e-08, + "loss": 0.6862, + "step": 33087 + }, + { + "epoch": 1.21, + "learning_rate": 7.901270736441205e-08, + "loss": 0.6912, + "step": 33088 + }, + { + "epoch": 1.21, + "learning_rate": 7.900904529973998e-08, + "loss": 0.6846, + "step": 33089 + }, + { + "epoch": 1.21, + "learning_rate": 7.900538323506792e-08, + "loss": 0.7007, + "step": 33090 + }, + { + "epoch": 1.21, + "learning_rate": 7.900172117039586e-08, + "loss": 0.7031, + "step": 33091 + }, + { + "epoch": 1.21, + "learning_rate": 7.89980591057238e-08, + "loss": 0.6922, + "step": 33092 + }, + { + "epoch": 1.21, + "learning_rate": 7.899439704105174e-08, + "loss": 0.7114, + "step": 33093 + }, + { + "epoch": 1.21, + "learning_rate": 7.899073497637968e-08, + "loss": 0.6867, + "step": 33094 + }, + { + "epoch": 1.21, + "learning_rate": 7.898707291170762e-08, + "loss": 0.7013, + "step": 33095 + }, + { + "epoch": 1.21, + "learning_rate": 7.898341084703556e-08, + "loss": 0.6912, + "step": 33096 + }, + { + "epoch": 1.21, + "learning_rate": 7.89797487823635e-08, + "loss": 0.689, + "step": 33097 + }, + { + "epoch": 1.21, + "learning_rate": 7.897608671769143e-08, + "loss": 0.6898, + "step": 33098 + }, + { + "epoch": 1.21, + "learning_rate": 7.897242465301937e-08, + "loss": 0.6902, + "step": 33099 + }, + { + "epoch": 1.21, + "learning_rate": 7.89687625883473e-08, + "loss": 0.6807, + "step": 33100 + }, + { + "epoch": 1.21, + "learning_rate": 7.896510052367525e-08, + "loss": 0.6897, + "step": 33101 + }, + { + "epoch": 1.21, + "learning_rate": 7.896143845900319e-08, + "loss": 0.6883, + "step": 33102 + }, + { + "epoch": 1.21, + "learning_rate": 7.895777639433113e-08, + "loss": 0.7003, + "step": 33103 + }, + { + "epoch": 1.21, + "learning_rate": 7.895411432965907e-08, + "loss": 0.6861, + "step": 33104 + }, + { + "epoch": 1.21, + "learning_rate": 7.8950452264987e-08, + "loss": 0.6701, + "step": 33105 + }, + { + "epoch": 1.21, + "learning_rate": 7.894679020031495e-08, + "loss": 0.6959, + "step": 33106 + }, + { + "epoch": 1.21, + "learning_rate": 7.894312813564287e-08, + "loss": 0.697, + "step": 33107 + }, + { + "epoch": 1.21, + "learning_rate": 7.893946607097081e-08, + "loss": 0.6786, + "step": 33108 + }, + { + "epoch": 1.21, + "learning_rate": 7.893580400629874e-08, + "loss": 0.6831, + "step": 33109 + }, + { + "epoch": 1.21, + "learning_rate": 7.893214194162668e-08, + "loss": 0.6864, + "step": 33110 + }, + { + "epoch": 1.21, + "learning_rate": 7.892847987695462e-08, + "loss": 0.6849, + "step": 33111 + }, + { + "epoch": 1.21, + "learning_rate": 7.892481781228256e-08, + "loss": 0.7094, + "step": 33112 + }, + { + "epoch": 1.21, + "learning_rate": 7.89211557476105e-08, + "loss": 0.6987, + "step": 33113 + }, + { + "epoch": 1.21, + "learning_rate": 7.891749368293844e-08, + "loss": 0.6965, + "step": 33114 + }, + { + "epoch": 1.21, + "learning_rate": 7.891383161826638e-08, + "loss": 0.7052, + "step": 33115 + }, + { + "epoch": 1.21, + "learning_rate": 7.89101695535943e-08, + "loss": 0.6851, + "step": 33116 + }, + { + "epoch": 1.21, + "learning_rate": 7.890650748892224e-08, + "loss": 0.6816, + "step": 33117 + }, + { + "epoch": 1.21, + "learning_rate": 7.890284542425018e-08, + "loss": 0.6774, + "step": 33118 + }, + { + "epoch": 1.21, + "learning_rate": 7.889918335957812e-08, + "loss": 0.6766, + "step": 33119 + }, + { + "epoch": 1.21, + "learning_rate": 7.889552129490606e-08, + "loss": 0.6842, + "step": 33120 + }, + { + "epoch": 1.21, + "learning_rate": 7.8891859230234e-08, + "loss": 0.6971, + "step": 33121 + }, + { + "epoch": 1.21, + "learning_rate": 7.888819716556194e-08, + "loss": 0.6866, + "step": 33122 + }, + { + "epoch": 1.21, + "learning_rate": 7.888453510088988e-08, + "loss": 0.6979, + "step": 33123 + }, + { + "epoch": 1.21, + "learning_rate": 7.888087303621782e-08, + "loss": 0.6853, + "step": 33124 + }, + { + "epoch": 1.21, + "learning_rate": 7.887721097154575e-08, + "loss": 0.7097, + "step": 33125 + }, + { + "epoch": 1.21, + "learning_rate": 7.887354890687369e-08, + "loss": 0.6968, + "step": 33126 + }, + { + "epoch": 1.21, + "learning_rate": 7.886988684220163e-08, + "loss": 0.6704, + "step": 33127 + }, + { + "epoch": 1.21, + "learning_rate": 7.886622477752957e-08, + "loss": 0.6996, + "step": 33128 + }, + { + "epoch": 1.21, + "learning_rate": 7.886256271285751e-08, + "loss": 0.6771, + "step": 33129 + }, + { + "epoch": 1.21, + "learning_rate": 7.885890064818545e-08, + "loss": 0.6774, + "step": 33130 + }, + { + "epoch": 1.21, + "learning_rate": 7.885523858351338e-08, + "loss": 0.6998, + "step": 33131 + }, + { + "epoch": 1.21, + "learning_rate": 7.885157651884132e-08, + "loss": 0.6937, + "step": 33132 + }, + { + "epoch": 1.21, + "learning_rate": 7.884791445416926e-08, + "loss": 0.6964, + "step": 33133 + }, + { + "epoch": 1.21, + "learning_rate": 7.884425238949718e-08, + "loss": 0.6916, + "step": 33134 + }, + { + "epoch": 1.21, + "learning_rate": 7.884059032482512e-08, + "loss": 0.6745, + "step": 33135 + }, + { + "epoch": 1.21, + "learning_rate": 7.883692826015306e-08, + "loss": 0.6957, + "step": 33136 + }, + { + "epoch": 1.21, + "learning_rate": 7.8833266195481e-08, + "loss": 0.6961, + "step": 33137 + }, + { + "epoch": 1.21, + "learning_rate": 7.882960413080894e-08, + "loss": 0.6865, + "step": 33138 + }, + { + "epoch": 1.21, + "learning_rate": 7.882594206613688e-08, + "loss": 0.6851, + "step": 33139 + }, + { + "epoch": 1.21, + "learning_rate": 7.882228000146482e-08, + "loss": 0.6862, + "step": 33140 + }, + { + "epoch": 1.21, + "learning_rate": 7.881861793679276e-08, + "loss": 0.7053, + "step": 33141 + }, + { + "epoch": 1.21, + "learning_rate": 7.88149558721207e-08, + "loss": 0.6973, + "step": 33142 + }, + { + "epoch": 1.21, + "learning_rate": 7.881129380744863e-08, + "loss": 0.6799, + "step": 33143 + }, + { + "epoch": 1.21, + "learning_rate": 7.880763174277657e-08, + "loss": 0.6906, + "step": 33144 + }, + { + "epoch": 1.21, + "learning_rate": 7.880396967810451e-08, + "loss": 0.6902, + "step": 33145 + }, + { + "epoch": 1.21, + "learning_rate": 7.880030761343245e-08, + "loss": 0.6979, + "step": 33146 + }, + { + "epoch": 1.21, + "learning_rate": 7.879664554876039e-08, + "loss": 0.7118, + "step": 33147 + }, + { + "epoch": 1.21, + "learning_rate": 7.879298348408833e-08, + "loss": 0.6918, + "step": 33148 + }, + { + "epoch": 1.21, + "learning_rate": 7.878932141941627e-08, + "loss": 0.6934, + "step": 33149 + }, + { + "epoch": 1.21, + "learning_rate": 7.878565935474421e-08, + "loss": 0.6911, + "step": 33150 + }, + { + "epoch": 1.21, + "learning_rate": 7.878199729007215e-08, + "loss": 0.6872, + "step": 33151 + }, + { + "epoch": 1.21, + "learning_rate": 7.877833522540008e-08, + "loss": 0.707, + "step": 33152 + }, + { + "epoch": 1.21, + "learning_rate": 7.877467316072802e-08, + "loss": 0.6893, + "step": 33153 + }, + { + "epoch": 1.21, + "learning_rate": 7.877101109605596e-08, + "loss": 0.6883, + "step": 33154 + }, + { + "epoch": 1.21, + "learning_rate": 7.87673490313839e-08, + "loss": 0.6688, + "step": 33155 + }, + { + "epoch": 1.21, + "learning_rate": 7.876368696671182e-08, + "loss": 0.6965, + "step": 33156 + }, + { + "epoch": 1.21, + "learning_rate": 7.876002490203976e-08, + "loss": 0.6881, + "step": 33157 + }, + { + "epoch": 1.21, + "learning_rate": 7.87563628373677e-08, + "loss": 0.6887, + "step": 33158 + }, + { + "epoch": 1.21, + "learning_rate": 7.875270077269564e-08, + "loss": 0.7018, + "step": 33159 + }, + { + "epoch": 1.21, + "learning_rate": 7.874903870802358e-08, + "loss": 0.6997, + "step": 33160 + }, + { + "epoch": 1.21, + "learning_rate": 7.874537664335151e-08, + "loss": 0.6853, + "step": 33161 + }, + { + "epoch": 1.21, + "learning_rate": 7.874171457867945e-08, + "loss": 0.682, + "step": 33162 + }, + { + "epoch": 1.21, + "learning_rate": 7.873805251400739e-08, + "loss": 0.6873, + "step": 33163 + }, + { + "epoch": 1.21, + "learning_rate": 7.873439044933533e-08, + "loss": 0.6899, + "step": 33164 + }, + { + "epoch": 1.21, + "learning_rate": 7.873072838466327e-08, + "loss": 0.6785, + "step": 33165 + }, + { + "epoch": 1.21, + "learning_rate": 7.872706631999121e-08, + "loss": 0.7004, + "step": 33166 + }, + { + "epoch": 1.21, + "learning_rate": 7.872340425531915e-08, + "loss": 0.6878, + "step": 33167 + }, + { + "epoch": 1.21, + "learning_rate": 7.871974219064709e-08, + "loss": 0.6867, + "step": 33168 + }, + { + "epoch": 1.21, + "learning_rate": 7.871608012597503e-08, + "loss": 0.6978, + "step": 33169 + }, + { + "epoch": 1.21, + "learning_rate": 7.871241806130296e-08, + "loss": 0.7044, + "step": 33170 + }, + { + "epoch": 1.21, + "learning_rate": 7.87087559966309e-08, + "loss": 0.6902, + "step": 33171 + }, + { + "epoch": 1.21, + "learning_rate": 7.870509393195884e-08, + "loss": 0.6975, + "step": 33172 + }, + { + "epoch": 1.21, + "learning_rate": 7.870143186728678e-08, + "loss": 0.7004, + "step": 33173 + }, + { + "epoch": 1.21, + "learning_rate": 7.869776980261472e-08, + "loss": 0.6924, + "step": 33174 + }, + { + "epoch": 1.21, + "learning_rate": 7.869410773794265e-08, + "loss": 0.6774, + "step": 33175 + }, + { + "epoch": 1.21, + "learning_rate": 7.86904456732706e-08, + "loss": 0.707, + "step": 33176 + }, + { + "epoch": 1.21, + "learning_rate": 7.868678360859852e-08, + "loss": 0.681, + "step": 33177 + }, + { + "epoch": 1.21, + "learning_rate": 7.868312154392646e-08, + "loss": 0.7014, + "step": 33178 + }, + { + "epoch": 1.21, + "learning_rate": 7.86794594792544e-08, + "loss": 0.6998, + "step": 33179 + }, + { + "epoch": 1.21, + "learning_rate": 7.867579741458233e-08, + "loss": 0.7062, + "step": 33180 + }, + { + "epoch": 1.21, + "learning_rate": 7.867213534991027e-08, + "loss": 0.6969, + "step": 33181 + }, + { + "epoch": 1.21, + "learning_rate": 7.866847328523821e-08, + "loss": 0.7012, + "step": 33182 + }, + { + "epoch": 1.21, + "learning_rate": 7.866481122056615e-08, + "loss": 0.6892, + "step": 33183 + }, + { + "epoch": 1.21, + "learning_rate": 7.866114915589409e-08, + "loss": 0.7003, + "step": 33184 + }, + { + "epoch": 1.21, + "learning_rate": 7.865748709122203e-08, + "loss": 0.6773, + "step": 33185 + }, + { + "epoch": 1.21, + "learning_rate": 7.865382502654997e-08, + "loss": 0.6941, + "step": 33186 + }, + { + "epoch": 1.21, + "learning_rate": 7.865016296187791e-08, + "loss": 0.6965, + "step": 33187 + }, + { + "epoch": 1.21, + "learning_rate": 7.864650089720583e-08, + "loss": 0.6742, + "step": 33188 + }, + { + "epoch": 1.21, + "learning_rate": 7.864283883253377e-08, + "loss": 0.6826, + "step": 33189 + }, + { + "epoch": 1.21, + "learning_rate": 7.863917676786171e-08, + "loss": 0.7004, + "step": 33190 + }, + { + "epoch": 1.21, + "learning_rate": 7.863551470318965e-08, + "loss": 0.6827, + "step": 33191 + }, + { + "epoch": 1.21, + "learning_rate": 7.86318526385176e-08, + "loss": 0.6914, + "step": 33192 + }, + { + "epoch": 1.21, + "learning_rate": 7.862819057384553e-08, + "loss": 0.6765, + "step": 33193 + }, + { + "epoch": 1.21, + "learning_rate": 7.862452850917347e-08, + "loss": 0.7085, + "step": 33194 + }, + { + "epoch": 1.21, + "learning_rate": 7.862086644450141e-08, + "loss": 0.685, + "step": 33195 + }, + { + "epoch": 1.21, + "learning_rate": 7.861720437982935e-08, + "loss": 0.7016, + "step": 33196 + }, + { + "epoch": 1.21, + "learning_rate": 7.861354231515728e-08, + "loss": 0.6844, + "step": 33197 + }, + { + "epoch": 1.21, + "learning_rate": 7.860988025048522e-08, + "loss": 0.6895, + "step": 33198 + }, + { + "epoch": 1.21, + "learning_rate": 7.860621818581316e-08, + "loss": 0.6878, + "step": 33199 + }, + { + "epoch": 1.21, + "learning_rate": 7.86025561211411e-08, + "loss": 0.6803, + "step": 33200 + }, + { + "epoch": 1.21, + "learning_rate": 7.859889405646904e-08, + "loss": 0.7052, + "step": 33201 + }, + { + "epoch": 1.21, + "learning_rate": 7.859523199179697e-08, + "loss": 0.6887, + "step": 33202 + }, + { + "epoch": 1.21, + "learning_rate": 7.859156992712491e-08, + "loss": 0.7152, + "step": 33203 + }, + { + "epoch": 1.21, + "learning_rate": 7.858790786245285e-08, + "loss": 0.7032, + "step": 33204 + }, + { + "epoch": 1.21, + "learning_rate": 7.858424579778079e-08, + "loss": 0.6941, + "step": 33205 + }, + { + "epoch": 1.21, + "learning_rate": 7.858058373310871e-08, + "loss": 0.7083, + "step": 33206 + }, + { + "epoch": 1.21, + "learning_rate": 7.857692166843665e-08, + "loss": 0.6989, + "step": 33207 + }, + { + "epoch": 1.21, + "learning_rate": 7.857325960376459e-08, + "loss": 0.6912, + "step": 33208 + }, + { + "epoch": 1.22, + "learning_rate": 7.856959753909253e-08, + "loss": 0.6943, + "step": 33209 + }, + { + "epoch": 1.22, + "learning_rate": 7.856593547442047e-08, + "loss": 0.696, + "step": 33210 + }, + { + "epoch": 1.22, + "learning_rate": 7.856227340974841e-08, + "loss": 0.6734, + "step": 33211 + }, + { + "epoch": 1.22, + "learning_rate": 7.855861134507635e-08, + "loss": 0.6938, + "step": 33212 + }, + { + "epoch": 1.22, + "learning_rate": 7.855494928040429e-08, + "loss": 0.6913, + "step": 33213 + }, + { + "epoch": 1.22, + "learning_rate": 7.855128721573223e-08, + "loss": 0.6927, + "step": 33214 + }, + { + "epoch": 1.22, + "learning_rate": 7.854762515106016e-08, + "loss": 0.6901, + "step": 33215 + }, + { + "epoch": 1.22, + "learning_rate": 7.85439630863881e-08, + "loss": 0.6958, + "step": 33216 + }, + { + "epoch": 1.22, + "learning_rate": 7.854030102171604e-08, + "loss": 0.6929, + "step": 33217 + }, + { + "epoch": 1.22, + "learning_rate": 7.853663895704398e-08, + "loss": 0.6934, + "step": 33218 + }, + { + "epoch": 1.22, + "learning_rate": 7.853297689237192e-08, + "loss": 0.7057, + "step": 33219 + }, + { + "epoch": 1.22, + "learning_rate": 7.852931482769986e-08, + "loss": 0.6951, + "step": 33220 + }, + { + "epoch": 1.22, + "learning_rate": 7.85256527630278e-08, + "loss": 0.684, + "step": 33221 + }, + { + "epoch": 1.22, + "learning_rate": 7.852199069835574e-08, + "loss": 0.6926, + "step": 33222 + }, + { + "epoch": 1.22, + "learning_rate": 7.851832863368368e-08, + "loss": 0.6952, + "step": 33223 + }, + { + "epoch": 1.22, + "learning_rate": 7.85146665690116e-08, + "loss": 0.7094, + "step": 33224 + }, + { + "epoch": 1.22, + "learning_rate": 7.851100450433955e-08, + "loss": 0.6859, + "step": 33225 + }, + { + "epoch": 1.22, + "learning_rate": 7.850734243966749e-08, + "loss": 0.6814, + "step": 33226 + }, + { + "epoch": 1.22, + "learning_rate": 7.850368037499541e-08, + "loss": 0.7, + "step": 33227 + }, + { + "epoch": 1.22, + "learning_rate": 7.850001831032335e-08, + "loss": 0.7063, + "step": 33228 + }, + { + "epoch": 1.22, + "learning_rate": 7.849635624565129e-08, + "loss": 0.6973, + "step": 33229 + }, + { + "epoch": 1.22, + "learning_rate": 7.849269418097923e-08, + "loss": 0.7059, + "step": 33230 + }, + { + "epoch": 1.22, + "learning_rate": 7.848903211630717e-08, + "loss": 0.6906, + "step": 33231 + }, + { + "epoch": 1.22, + "learning_rate": 7.848537005163511e-08, + "loss": 0.6956, + "step": 33232 + }, + { + "epoch": 1.22, + "learning_rate": 7.848170798696304e-08, + "loss": 0.6988, + "step": 33233 + }, + { + "epoch": 1.22, + "learning_rate": 7.847804592229098e-08, + "loss": 0.6897, + "step": 33234 + }, + { + "epoch": 1.22, + "learning_rate": 7.847438385761892e-08, + "loss": 0.6958, + "step": 33235 + }, + { + "epoch": 1.22, + "learning_rate": 7.847072179294686e-08, + "loss": 0.6927, + "step": 33236 + }, + { + "epoch": 1.22, + "learning_rate": 7.84670597282748e-08, + "loss": 0.6792, + "step": 33237 + }, + { + "epoch": 1.22, + "learning_rate": 7.846339766360274e-08, + "loss": 0.6899, + "step": 33238 + }, + { + "epoch": 1.22, + "learning_rate": 7.845973559893068e-08, + "loss": 0.6868, + "step": 33239 + }, + { + "epoch": 1.22, + "learning_rate": 7.845607353425862e-08, + "loss": 0.7029, + "step": 33240 + }, + { + "epoch": 1.22, + "learning_rate": 7.845241146958656e-08, + "loss": 0.6886, + "step": 33241 + }, + { + "epoch": 1.22, + "learning_rate": 7.844874940491448e-08, + "loss": 0.6873, + "step": 33242 + }, + { + "epoch": 1.22, + "learning_rate": 7.844508734024242e-08, + "loss": 0.6939, + "step": 33243 + }, + { + "epoch": 1.22, + "learning_rate": 7.844142527557036e-08, + "loss": 0.691, + "step": 33244 + }, + { + "epoch": 1.22, + "learning_rate": 7.84377632108983e-08, + "loss": 0.6967, + "step": 33245 + }, + { + "epoch": 1.22, + "learning_rate": 7.843410114622624e-08, + "loss": 0.6824, + "step": 33246 + }, + { + "epoch": 1.22, + "learning_rate": 7.843043908155418e-08, + "loss": 0.6964, + "step": 33247 + }, + { + "epoch": 1.22, + "learning_rate": 7.842677701688211e-08, + "loss": 0.6979, + "step": 33248 + }, + { + "epoch": 1.22, + "learning_rate": 7.842311495221005e-08, + "loss": 0.7032, + "step": 33249 + }, + { + "epoch": 1.22, + "learning_rate": 7.841945288753799e-08, + "loss": 0.7071, + "step": 33250 + }, + { + "epoch": 1.22, + "learning_rate": 7.841579082286592e-08, + "loss": 0.6907, + "step": 33251 + }, + { + "epoch": 1.22, + "learning_rate": 7.841212875819386e-08, + "loss": 0.7071, + "step": 33252 + }, + { + "epoch": 1.22, + "learning_rate": 7.84084666935218e-08, + "loss": 0.699, + "step": 33253 + }, + { + "epoch": 1.22, + "learning_rate": 7.840480462884974e-08, + "loss": 0.6888, + "step": 33254 + }, + { + "epoch": 1.22, + "learning_rate": 7.840114256417768e-08, + "loss": 0.6796, + "step": 33255 + }, + { + "epoch": 1.22, + "learning_rate": 7.839748049950562e-08, + "loss": 0.6866, + "step": 33256 + }, + { + "epoch": 1.22, + "learning_rate": 7.839381843483356e-08, + "loss": 0.695, + "step": 33257 + }, + { + "epoch": 1.22, + "learning_rate": 7.83901563701615e-08, + "loss": 0.7151, + "step": 33258 + }, + { + "epoch": 1.22, + "learning_rate": 7.838649430548944e-08, + "loss": 0.7001, + "step": 33259 + }, + { + "epoch": 1.22, + "learning_rate": 7.838283224081736e-08, + "loss": 0.6888, + "step": 33260 + }, + { + "epoch": 1.22, + "learning_rate": 7.83791701761453e-08, + "loss": 0.6926, + "step": 33261 + }, + { + "epoch": 1.22, + "learning_rate": 7.837550811147324e-08, + "loss": 0.6961, + "step": 33262 + }, + { + "epoch": 1.22, + "learning_rate": 7.837184604680118e-08, + "loss": 0.6854, + "step": 33263 + }, + { + "epoch": 1.22, + "learning_rate": 7.836818398212912e-08, + "loss": 0.6941, + "step": 33264 + }, + { + "epoch": 1.22, + "learning_rate": 7.836452191745706e-08, + "loss": 0.7002, + "step": 33265 + }, + { + "epoch": 1.22, + "learning_rate": 7.8360859852785e-08, + "loss": 0.7041, + "step": 33266 + }, + { + "epoch": 1.22, + "learning_rate": 7.835719778811294e-08, + "loss": 0.6911, + "step": 33267 + }, + { + "epoch": 1.22, + "learning_rate": 7.835353572344088e-08, + "loss": 0.6806, + "step": 33268 + }, + { + "epoch": 1.22, + "learning_rate": 7.834987365876881e-08, + "loss": 0.6921, + "step": 33269 + }, + { + "epoch": 1.22, + "learning_rate": 7.834621159409675e-08, + "loss": 0.6881, + "step": 33270 + }, + { + "epoch": 1.22, + "learning_rate": 7.834254952942469e-08, + "loss": 0.7034, + "step": 33271 + }, + { + "epoch": 1.22, + "learning_rate": 7.833888746475263e-08, + "loss": 0.6922, + "step": 33272 + }, + { + "epoch": 1.22, + "learning_rate": 7.833522540008056e-08, + "loss": 0.7022, + "step": 33273 + }, + { + "epoch": 1.22, + "learning_rate": 7.83315633354085e-08, + "loss": 0.7007, + "step": 33274 + }, + { + "epoch": 1.22, + "learning_rate": 7.832790127073644e-08, + "loss": 0.6742, + "step": 33275 + }, + { + "epoch": 1.22, + "learning_rate": 7.832423920606438e-08, + "loss": 0.7105, + "step": 33276 + }, + { + "epoch": 1.22, + "learning_rate": 7.832057714139232e-08, + "loss": 0.686, + "step": 33277 + }, + { + "epoch": 1.22, + "learning_rate": 7.831691507672024e-08, + "loss": 0.6874, + "step": 33278 + }, + { + "epoch": 1.22, + "learning_rate": 7.831325301204818e-08, + "loss": 0.7099, + "step": 33279 + }, + { + "epoch": 1.22, + "learning_rate": 7.830959094737612e-08, + "loss": 0.7015, + "step": 33280 + }, + { + "epoch": 1.22, + "learning_rate": 7.830592888270406e-08, + "loss": 0.6872, + "step": 33281 + }, + { + "epoch": 1.22, + "learning_rate": 7.8302266818032e-08, + "loss": 0.6943, + "step": 33282 + }, + { + "epoch": 1.22, + "learning_rate": 7.829860475335994e-08, + "loss": 0.6926, + "step": 33283 + }, + { + "epoch": 1.22, + "learning_rate": 7.829494268868788e-08, + "loss": 0.6895, + "step": 33284 + }, + { + "epoch": 1.22, + "learning_rate": 7.829128062401582e-08, + "loss": 0.6937, + "step": 33285 + }, + { + "epoch": 1.22, + "learning_rate": 7.828761855934376e-08, + "loss": 0.6911, + "step": 33286 + }, + { + "epoch": 1.22, + "learning_rate": 7.828395649467169e-08, + "loss": 0.6898, + "step": 33287 + }, + { + "epoch": 1.22, + "learning_rate": 7.828029442999963e-08, + "loss": 0.692, + "step": 33288 + }, + { + "epoch": 1.22, + "learning_rate": 7.827663236532757e-08, + "loss": 0.6997, + "step": 33289 + }, + { + "epoch": 1.22, + "learning_rate": 7.827297030065551e-08, + "loss": 0.7002, + "step": 33290 + }, + { + "epoch": 1.22, + "learning_rate": 7.826930823598345e-08, + "loss": 0.6798, + "step": 33291 + }, + { + "epoch": 1.22, + "learning_rate": 7.826564617131139e-08, + "loss": 0.6946, + "step": 33292 + }, + { + "epoch": 1.22, + "learning_rate": 7.826198410663933e-08, + "loss": 0.6896, + "step": 33293 + }, + { + "epoch": 1.22, + "learning_rate": 7.825832204196727e-08, + "loss": 0.6867, + "step": 33294 + }, + { + "epoch": 1.22, + "learning_rate": 7.82546599772952e-08, + "loss": 0.6943, + "step": 33295 + }, + { + "epoch": 1.22, + "learning_rate": 7.825099791262313e-08, + "loss": 0.6904, + "step": 33296 + }, + { + "epoch": 1.22, + "learning_rate": 7.824733584795106e-08, + "loss": 0.6899, + "step": 33297 + }, + { + "epoch": 1.22, + "learning_rate": 7.8243673783279e-08, + "loss": 0.6918, + "step": 33298 + }, + { + "epoch": 1.22, + "learning_rate": 7.824001171860694e-08, + "loss": 0.7125, + "step": 33299 + }, + { + "epoch": 1.22, + "learning_rate": 7.823634965393488e-08, + "loss": 0.6781, + "step": 33300 + }, + { + "epoch": 1.22, + "learning_rate": 7.823268758926282e-08, + "loss": 0.6949, + "step": 33301 + }, + { + "epoch": 1.22, + "learning_rate": 7.822902552459076e-08, + "loss": 0.6891, + "step": 33302 + }, + { + "epoch": 1.22, + "learning_rate": 7.82253634599187e-08, + "loss": 0.6985, + "step": 33303 + }, + { + "epoch": 1.22, + "learning_rate": 7.822170139524664e-08, + "loss": 0.6944, + "step": 33304 + }, + { + "epoch": 1.22, + "learning_rate": 7.821803933057457e-08, + "loss": 0.6901, + "step": 33305 + }, + { + "epoch": 1.22, + "learning_rate": 7.821437726590251e-08, + "loss": 0.6935, + "step": 33306 + }, + { + "epoch": 1.22, + "learning_rate": 7.821071520123045e-08, + "loss": 0.6878, + "step": 33307 + }, + { + "epoch": 1.22, + "learning_rate": 7.820705313655839e-08, + "loss": 0.6938, + "step": 33308 + }, + { + "epoch": 1.22, + "learning_rate": 7.820339107188633e-08, + "loss": 0.6837, + "step": 33309 + }, + { + "epoch": 1.22, + "learning_rate": 7.819972900721427e-08, + "loss": 0.6842, + "step": 33310 + }, + { + "epoch": 1.22, + "learning_rate": 7.819606694254221e-08, + "loss": 0.6915, + "step": 33311 + }, + { + "epoch": 1.22, + "learning_rate": 7.819240487787015e-08, + "loss": 0.6934, + "step": 33312 + }, + { + "epoch": 1.22, + "learning_rate": 7.818874281319809e-08, + "loss": 0.6961, + "step": 33313 + }, + { + "epoch": 1.22, + "learning_rate": 7.818508074852601e-08, + "loss": 0.6935, + "step": 33314 + }, + { + "epoch": 1.22, + "learning_rate": 7.818141868385395e-08, + "loss": 0.6876, + "step": 33315 + }, + { + "epoch": 1.22, + "learning_rate": 7.81777566191819e-08, + "loss": 0.6987, + "step": 33316 + }, + { + "epoch": 1.22, + "learning_rate": 7.817409455450983e-08, + "loss": 0.6833, + "step": 33317 + }, + { + "epoch": 1.22, + "learning_rate": 7.817043248983777e-08, + "loss": 0.6921, + "step": 33318 + }, + { + "epoch": 1.22, + "learning_rate": 7.81667704251657e-08, + "loss": 0.6849, + "step": 33319 + }, + { + "epoch": 1.22, + "learning_rate": 7.816310836049364e-08, + "loss": 0.6935, + "step": 33320 + }, + { + "epoch": 1.22, + "learning_rate": 7.815944629582158e-08, + "loss": 0.6746, + "step": 33321 + }, + { + "epoch": 1.22, + "learning_rate": 7.815578423114951e-08, + "loss": 0.7092, + "step": 33322 + }, + { + "epoch": 1.22, + "learning_rate": 7.815212216647745e-08, + "loss": 0.6948, + "step": 33323 + }, + { + "epoch": 1.22, + "learning_rate": 7.814846010180539e-08, + "loss": 0.6839, + "step": 33324 + }, + { + "epoch": 1.22, + "learning_rate": 7.814479803713333e-08, + "loss": 0.6991, + "step": 33325 + }, + { + "epoch": 1.22, + "learning_rate": 7.814113597246127e-08, + "loss": 0.702, + "step": 33326 + }, + { + "epoch": 1.22, + "learning_rate": 7.81374739077892e-08, + "loss": 0.6897, + "step": 33327 + }, + { + "epoch": 1.22, + "learning_rate": 7.813381184311715e-08, + "loss": 0.6971, + "step": 33328 + }, + { + "epoch": 1.22, + "learning_rate": 7.813014977844509e-08, + "loss": 0.6903, + "step": 33329 + }, + { + "epoch": 1.22, + "learning_rate": 7.812648771377303e-08, + "loss": 0.6821, + "step": 33330 + }, + { + "epoch": 1.22, + "learning_rate": 7.812282564910095e-08, + "loss": 0.6968, + "step": 33331 + }, + { + "epoch": 1.22, + "learning_rate": 7.811916358442889e-08, + "loss": 0.6941, + "step": 33332 + }, + { + "epoch": 1.22, + "learning_rate": 7.811550151975683e-08, + "loss": 0.7041, + "step": 33333 + }, + { + "epoch": 1.22, + "learning_rate": 7.811183945508477e-08, + "loss": 0.6987, + "step": 33334 + }, + { + "epoch": 1.22, + "learning_rate": 7.810817739041271e-08, + "loss": 0.6872, + "step": 33335 + }, + { + "epoch": 1.22, + "learning_rate": 7.810451532574065e-08, + "loss": 0.708, + "step": 33336 + }, + { + "epoch": 1.22, + "learning_rate": 7.810085326106859e-08, + "loss": 0.6936, + "step": 33337 + }, + { + "epoch": 1.22, + "learning_rate": 7.809719119639653e-08, + "loss": 0.71, + "step": 33338 + }, + { + "epoch": 1.22, + "learning_rate": 7.809352913172447e-08, + "loss": 0.6995, + "step": 33339 + }, + { + "epoch": 1.22, + "learning_rate": 7.80898670670524e-08, + "loss": 0.6857, + "step": 33340 + }, + { + "epoch": 1.22, + "learning_rate": 7.808620500238034e-08, + "loss": 0.6915, + "step": 33341 + }, + { + "epoch": 1.22, + "learning_rate": 7.808254293770828e-08, + "loss": 0.7009, + "step": 33342 + }, + { + "epoch": 1.22, + "learning_rate": 7.807888087303622e-08, + "loss": 0.6938, + "step": 33343 + }, + { + "epoch": 1.22, + "learning_rate": 7.807521880836415e-08, + "loss": 0.674, + "step": 33344 + }, + { + "epoch": 1.22, + "learning_rate": 7.807155674369209e-08, + "loss": 0.7026, + "step": 33345 + }, + { + "epoch": 1.22, + "learning_rate": 7.806789467902003e-08, + "loss": 0.7037, + "step": 33346 + }, + { + "epoch": 1.22, + "learning_rate": 7.806423261434797e-08, + "loss": 0.6863, + "step": 33347 + }, + { + "epoch": 1.22, + "learning_rate": 7.80605705496759e-08, + "loss": 0.7046, + "step": 33348 + }, + { + "epoch": 1.22, + "learning_rate": 7.805690848500383e-08, + "loss": 0.6829, + "step": 33349 + }, + { + "epoch": 1.22, + "learning_rate": 7.805324642033177e-08, + "loss": 0.7021, + "step": 33350 + }, + { + "epoch": 1.22, + "learning_rate": 7.804958435565971e-08, + "loss": 0.6849, + "step": 33351 + }, + { + "epoch": 1.22, + "learning_rate": 7.804592229098765e-08, + "loss": 0.6901, + "step": 33352 + }, + { + "epoch": 1.22, + "learning_rate": 7.804226022631559e-08, + "loss": 0.689, + "step": 33353 + }, + { + "epoch": 1.22, + "learning_rate": 7.803859816164353e-08, + "loss": 0.6928, + "step": 33354 + }, + { + "epoch": 1.22, + "learning_rate": 7.803493609697147e-08, + "loss": 0.7198, + "step": 33355 + }, + { + "epoch": 1.22, + "learning_rate": 7.803127403229941e-08, + "loss": 0.6926, + "step": 33356 + }, + { + "epoch": 1.22, + "learning_rate": 7.802761196762735e-08, + "loss": 0.6746, + "step": 33357 + }, + { + "epoch": 1.22, + "learning_rate": 7.802394990295528e-08, + "loss": 0.6971, + "step": 33358 + }, + { + "epoch": 1.22, + "learning_rate": 7.802028783828322e-08, + "loss": 0.7043, + "step": 33359 + }, + { + "epoch": 1.22, + "learning_rate": 7.801662577361116e-08, + "loss": 0.678, + "step": 33360 + }, + { + "epoch": 1.22, + "learning_rate": 7.80129637089391e-08, + "loss": 0.6991, + "step": 33361 + }, + { + "epoch": 1.22, + "learning_rate": 7.800930164426704e-08, + "loss": 0.6977, + "step": 33362 + }, + { + "epoch": 1.22, + "learning_rate": 7.800563957959498e-08, + "loss": 0.677, + "step": 33363 + }, + { + "epoch": 1.22, + "learning_rate": 7.800197751492292e-08, + "loss": 0.6771, + "step": 33364 + }, + { + "epoch": 1.22, + "learning_rate": 7.799831545025086e-08, + "loss": 0.6831, + "step": 33365 + }, + { + "epoch": 1.22, + "learning_rate": 7.799465338557878e-08, + "loss": 0.6908, + "step": 33366 + }, + { + "epoch": 1.22, + "learning_rate": 7.799099132090672e-08, + "loss": 0.6926, + "step": 33367 + }, + { + "epoch": 1.22, + "learning_rate": 7.798732925623465e-08, + "loss": 0.6984, + "step": 33368 + }, + { + "epoch": 1.22, + "learning_rate": 7.798366719156259e-08, + "loss": 0.6908, + "step": 33369 + }, + { + "epoch": 1.22, + "learning_rate": 7.798000512689053e-08, + "loss": 0.687, + "step": 33370 + }, + { + "epoch": 1.22, + "learning_rate": 7.797634306221847e-08, + "loss": 0.705, + "step": 33371 + }, + { + "epoch": 1.22, + "learning_rate": 7.797268099754641e-08, + "loss": 0.6832, + "step": 33372 + }, + { + "epoch": 1.22, + "learning_rate": 7.796901893287435e-08, + "loss": 0.7006, + "step": 33373 + }, + { + "epoch": 1.22, + "learning_rate": 7.796535686820229e-08, + "loss": 0.6837, + "step": 33374 + }, + { + "epoch": 1.22, + "learning_rate": 7.796169480353023e-08, + "loss": 0.7015, + "step": 33375 + }, + { + "epoch": 1.22, + "learning_rate": 7.795803273885816e-08, + "loss": 0.6889, + "step": 33376 + }, + { + "epoch": 1.22, + "learning_rate": 7.79543706741861e-08, + "loss": 0.6884, + "step": 33377 + }, + { + "epoch": 1.22, + "learning_rate": 7.795070860951404e-08, + "loss": 0.6953, + "step": 33378 + }, + { + "epoch": 1.22, + "learning_rate": 7.794704654484198e-08, + "loss": 0.6998, + "step": 33379 + }, + { + "epoch": 1.22, + "learning_rate": 7.794338448016992e-08, + "loss": 0.6843, + "step": 33380 + }, + { + "epoch": 1.22, + "learning_rate": 7.793972241549786e-08, + "loss": 0.699, + "step": 33381 + }, + { + "epoch": 1.22, + "learning_rate": 7.79360603508258e-08, + "loss": 0.6949, + "step": 33382 + }, + { + "epoch": 1.22, + "learning_rate": 7.793239828615374e-08, + "loss": 0.699, + "step": 33383 + }, + { + "epoch": 1.22, + "learning_rate": 7.792873622148168e-08, + "loss": 0.6939, + "step": 33384 + }, + { + "epoch": 1.22, + "learning_rate": 7.79250741568096e-08, + "loss": 0.6942, + "step": 33385 + }, + { + "epoch": 1.22, + "learning_rate": 7.792141209213754e-08, + "loss": 0.7031, + "step": 33386 + }, + { + "epoch": 1.22, + "learning_rate": 7.791775002746548e-08, + "loss": 0.6789, + "step": 33387 + }, + { + "epoch": 1.22, + "learning_rate": 7.791408796279342e-08, + "loss": 0.6988, + "step": 33388 + }, + { + "epoch": 1.22, + "learning_rate": 7.791042589812136e-08, + "loss": 0.6993, + "step": 33389 + }, + { + "epoch": 1.22, + "learning_rate": 7.790676383344929e-08, + "loss": 0.7042, + "step": 33390 + }, + { + "epoch": 1.22, + "learning_rate": 7.790310176877723e-08, + "loss": 0.6883, + "step": 33391 + }, + { + "epoch": 1.22, + "learning_rate": 7.789943970410517e-08, + "loss": 0.6876, + "step": 33392 + }, + { + "epoch": 1.22, + "learning_rate": 7.789577763943311e-08, + "loss": 0.6825, + "step": 33393 + }, + { + "epoch": 1.22, + "learning_rate": 7.789211557476104e-08, + "loss": 0.6969, + "step": 33394 + }, + { + "epoch": 1.22, + "learning_rate": 7.788845351008898e-08, + "loss": 0.6895, + "step": 33395 + }, + { + "epoch": 1.22, + "learning_rate": 7.788479144541692e-08, + "loss": 0.6975, + "step": 33396 + }, + { + "epoch": 1.22, + "learning_rate": 7.788112938074486e-08, + "loss": 0.7003, + "step": 33397 + }, + { + "epoch": 1.22, + "learning_rate": 7.78774673160728e-08, + "loss": 0.6993, + "step": 33398 + }, + { + "epoch": 1.22, + "learning_rate": 7.787380525140074e-08, + "loss": 0.6694, + "step": 33399 + }, + { + "epoch": 1.22, + "learning_rate": 7.787014318672868e-08, + "loss": 0.7043, + "step": 33400 + }, + { + "epoch": 1.22, + "learning_rate": 7.786648112205662e-08, + "loss": 0.686, + "step": 33401 + }, + { + "epoch": 1.22, + "learning_rate": 7.786281905738456e-08, + "loss": 0.6977, + "step": 33402 + }, + { + "epoch": 1.22, + "learning_rate": 7.785915699271248e-08, + "loss": 0.7005, + "step": 33403 + }, + { + "epoch": 1.22, + "learning_rate": 7.785549492804042e-08, + "loss": 0.7, + "step": 33404 + }, + { + "epoch": 1.22, + "learning_rate": 7.785183286336836e-08, + "loss": 0.6996, + "step": 33405 + }, + { + "epoch": 1.22, + "learning_rate": 7.78481707986963e-08, + "loss": 0.6846, + "step": 33406 + }, + { + "epoch": 1.22, + "learning_rate": 7.784450873402424e-08, + "loss": 0.6994, + "step": 33407 + }, + { + "epoch": 1.22, + "learning_rate": 7.784084666935218e-08, + "loss": 0.6838, + "step": 33408 + }, + { + "epoch": 1.22, + "learning_rate": 7.783718460468012e-08, + "loss": 0.6889, + "step": 33409 + }, + { + "epoch": 1.22, + "learning_rate": 7.783352254000806e-08, + "loss": 0.6971, + "step": 33410 + }, + { + "epoch": 1.22, + "learning_rate": 7.7829860475336e-08, + "loss": 0.6962, + "step": 33411 + }, + { + "epoch": 1.22, + "learning_rate": 7.782619841066393e-08, + "loss": 0.6925, + "step": 33412 + }, + { + "epoch": 1.22, + "learning_rate": 7.782253634599187e-08, + "loss": 0.6917, + "step": 33413 + }, + { + "epoch": 1.22, + "learning_rate": 7.781887428131981e-08, + "loss": 0.7033, + "step": 33414 + }, + { + "epoch": 1.22, + "learning_rate": 7.781521221664773e-08, + "loss": 0.6901, + "step": 33415 + }, + { + "epoch": 1.22, + "learning_rate": 7.781155015197567e-08, + "loss": 0.6826, + "step": 33416 + }, + { + "epoch": 1.22, + "learning_rate": 7.780788808730361e-08, + "loss": 0.6912, + "step": 33417 + }, + { + "epoch": 1.22, + "learning_rate": 7.780422602263155e-08, + "loss": 0.6825, + "step": 33418 + }, + { + "epoch": 1.22, + "learning_rate": 7.78005639579595e-08, + "loss": 0.688, + "step": 33419 + }, + { + "epoch": 1.22, + "learning_rate": 7.779690189328743e-08, + "loss": 0.6881, + "step": 33420 + }, + { + "epoch": 1.22, + "learning_rate": 7.779323982861536e-08, + "loss": 0.6933, + "step": 33421 + }, + { + "epoch": 1.22, + "learning_rate": 7.77895777639433e-08, + "loss": 0.6992, + "step": 33422 + }, + { + "epoch": 1.22, + "learning_rate": 7.778591569927124e-08, + "loss": 0.6848, + "step": 33423 + }, + { + "epoch": 1.22, + "learning_rate": 7.778225363459918e-08, + "loss": 0.6944, + "step": 33424 + }, + { + "epoch": 1.22, + "learning_rate": 7.777859156992712e-08, + "loss": 0.7012, + "step": 33425 + }, + { + "epoch": 1.22, + "learning_rate": 7.777492950525506e-08, + "loss": 0.6901, + "step": 33426 + }, + { + "epoch": 1.22, + "learning_rate": 7.7771267440583e-08, + "loss": 0.6888, + "step": 33427 + }, + { + "epoch": 1.22, + "learning_rate": 7.776760537591094e-08, + "loss": 0.6975, + "step": 33428 + }, + { + "epoch": 1.22, + "learning_rate": 7.776394331123888e-08, + "loss": 0.6899, + "step": 33429 + }, + { + "epoch": 1.22, + "learning_rate": 7.776028124656681e-08, + "loss": 0.706, + "step": 33430 + }, + { + "epoch": 1.22, + "learning_rate": 7.775661918189475e-08, + "loss": 0.6934, + "step": 33431 + }, + { + "epoch": 1.22, + "learning_rate": 7.775295711722269e-08, + "loss": 0.6886, + "step": 33432 + }, + { + "epoch": 1.22, + "learning_rate": 7.774929505255063e-08, + "loss": 0.686, + "step": 33433 + }, + { + "epoch": 1.22, + "learning_rate": 7.774563298787857e-08, + "loss": 0.705, + "step": 33434 + }, + { + "epoch": 1.22, + "learning_rate": 7.774197092320651e-08, + "loss": 0.6811, + "step": 33435 + }, + { + "epoch": 1.22, + "learning_rate": 7.773830885853445e-08, + "loss": 0.6996, + "step": 33436 + }, + { + "epoch": 1.22, + "learning_rate": 7.773464679386237e-08, + "loss": 0.6892, + "step": 33437 + }, + { + "epoch": 1.22, + "learning_rate": 7.773098472919031e-08, + "loss": 0.6882, + "step": 33438 + }, + { + "epoch": 1.22, + "learning_rate": 7.772732266451824e-08, + "loss": 0.7159, + "step": 33439 + }, + { + "epoch": 1.22, + "learning_rate": 7.772366059984618e-08, + "loss": 0.689, + "step": 33440 + }, + { + "epoch": 1.22, + "learning_rate": 7.771999853517412e-08, + "loss": 0.6963, + "step": 33441 + }, + { + "epoch": 1.22, + "learning_rate": 7.771633647050206e-08, + "loss": 0.7138, + "step": 33442 + }, + { + "epoch": 1.22, + "learning_rate": 7.771267440583e-08, + "loss": 0.6693, + "step": 33443 + }, + { + "epoch": 1.22, + "learning_rate": 7.770901234115794e-08, + "loss": 0.6884, + "step": 33444 + }, + { + "epoch": 1.22, + "learning_rate": 7.770535027648588e-08, + "loss": 0.683, + "step": 33445 + }, + { + "epoch": 1.22, + "learning_rate": 7.770168821181382e-08, + "loss": 0.6903, + "step": 33446 + }, + { + "epoch": 1.22, + "learning_rate": 7.769802614714176e-08, + "loss": 0.6954, + "step": 33447 + }, + { + "epoch": 1.22, + "learning_rate": 7.769436408246969e-08, + "loss": 0.6978, + "step": 33448 + }, + { + "epoch": 1.22, + "learning_rate": 7.769070201779763e-08, + "loss": 0.695, + "step": 33449 + }, + { + "epoch": 1.22, + "learning_rate": 7.768703995312557e-08, + "loss": 0.6924, + "step": 33450 + }, + { + "epoch": 1.22, + "learning_rate": 7.76833778884535e-08, + "loss": 0.6911, + "step": 33451 + }, + { + "epoch": 1.22, + "learning_rate": 7.767971582378145e-08, + "loss": 0.6991, + "step": 33452 + }, + { + "epoch": 1.22, + "learning_rate": 7.767605375910939e-08, + "loss": 0.6889, + "step": 33453 + }, + { + "epoch": 1.22, + "learning_rate": 7.767239169443733e-08, + "loss": 0.6862, + "step": 33454 + }, + { + "epoch": 1.22, + "learning_rate": 7.766872962976527e-08, + "loss": 0.7027, + "step": 33455 + }, + { + "epoch": 1.22, + "learning_rate": 7.76650675650932e-08, + "loss": 0.6938, + "step": 33456 + }, + { + "epoch": 1.22, + "learning_rate": 7.766140550042113e-08, + "loss": 0.7015, + "step": 33457 + }, + { + "epoch": 1.22, + "learning_rate": 7.765774343574907e-08, + "loss": 0.6933, + "step": 33458 + }, + { + "epoch": 1.22, + "learning_rate": 7.765408137107701e-08, + "loss": 0.6875, + "step": 33459 + }, + { + "epoch": 1.22, + "learning_rate": 7.765041930640495e-08, + "loss": 0.6889, + "step": 33460 + }, + { + "epoch": 1.22, + "learning_rate": 7.764675724173288e-08, + "loss": 0.6948, + "step": 33461 + }, + { + "epoch": 1.22, + "learning_rate": 7.764309517706082e-08, + "loss": 0.6949, + "step": 33462 + }, + { + "epoch": 1.22, + "learning_rate": 7.763943311238876e-08, + "loss": 0.6908, + "step": 33463 + }, + { + "epoch": 1.22, + "learning_rate": 7.76357710477167e-08, + "loss": 0.6848, + "step": 33464 + }, + { + "epoch": 1.22, + "learning_rate": 7.763210898304464e-08, + "loss": 0.6966, + "step": 33465 + }, + { + "epoch": 1.22, + "learning_rate": 7.762844691837257e-08, + "loss": 0.7066, + "step": 33466 + }, + { + "epoch": 1.22, + "learning_rate": 7.76247848537005e-08, + "loss": 0.6906, + "step": 33467 + }, + { + "epoch": 1.22, + "learning_rate": 7.762112278902845e-08, + "loss": 0.6779, + "step": 33468 + }, + { + "epoch": 1.22, + "learning_rate": 7.761746072435639e-08, + "loss": 0.699, + "step": 33469 + }, + { + "epoch": 1.22, + "learning_rate": 7.761379865968433e-08, + "loss": 0.7003, + "step": 33470 + }, + { + "epoch": 1.22, + "learning_rate": 7.761013659501227e-08, + "loss": 0.683, + "step": 33471 + }, + { + "epoch": 1.22, + "learning_rate": 7.76064745303402e-08, + "loss": 0.6993, + "step": 33472 + }, + { + "epoch": 1.22, + "learning_rate": 7.760281246566814e-08, + "loss": 0.696, + "step": 33473 + }, + { + "epoch": 1.22, + "learning_rate": 7.759915040099608e-08, + "loss": 0.7018, + "step": 33474 + }, + { + "epoch": 1.22, + "learning_rate": 7.759548833632401e-08, + "loss": 0.7019, + "step": 33475 + }, + { + "epoch": 1.22, + "learning_rate": 7.759182627165195e-08, + "loss": 0.6895, + "step": 33476 + }, + { + "epoch": 1.22, + "learning_rate": 7.758816420697989e-08, + "loss": 0.7029, + "step": 33477 + }, + { + "epoch": 1.22, + "learning_rate": 7.758450214230783e-08, + "loss": 0.6898, + "step": 33478 + }, + { + "epoch": 1.22, + "learning_rate": 7.758084007763577e-08, + "loss": 0.6986, + "step": 33479 + }, + { + "epoch": 1.22, + "learning_rate": 7.757717801296371e-08, + "loss": 0.6882, + "step": 33480 + }, + { + "epoch": 1.22, + "learning_rate": 7.757351594829165e-08, + "loss": 0.6892, + "step": 33481 + }, + { + "epoch": 1.23, + "learning_rate": 7.756985388361959e-08, + "loss": 0.6944, + "step": 33482 + }, + { + "epoch": 1.23, + "learning_rate": 7.756619181894752e-08, + "loss": 0.6918, + "step": 33483 + }, + { + "epoch": 1.23, + "learning_rate": 7.756252975427546e-08, + "loss": 0.6926, + "step": 33484 + }, + { + "epoch": 1.23, + "learning_rate": 7.75588676896034e-08, + "loss": 0.6953, + "step": 33485 + }, + { + "epoch": 1.23, + "learning_rate": 7.755520562493132e-08, + "loss": 0.7055, + "step": 33486 + }, + { + "epoch": 1.23, + "learning_rate": 7.755154356025926e-08, + "loss": 0.6962, + "step": 33487 + }, + { + "epoch": 1.23, + "learning_rate": 7.75478814955872e-08, + "loss": 0.6838, + "step": 33488 + }, + { + "epoch": 1.23, + "learning_rate": 7.754421943091514e-08, + "loss": 0.6924, + "step": 33489 + }, + { + "epoch": 1.23, + "learning_rate": 7.754055736624308e-08, + "loss": 0.7034, + "step": 33490 + }, + { + "epoch": 1.23, + "learning_rate": 7.753689530157102e-08, + "loss": 0.6917, + "step": 33491 + }, + { + "epoch": 1.23, + "learning_rate": 7.753323323689896e-08, + "loss": 0.6949, + "step": 33492 + }, + { + "epoch": 1.23, + "learning_rate": 7.752957117222689e-08, + "loss": 0.6961, + "step": 33493 + }, + { + "epoch": 1.23, + "learning_rate": 7.752590910755483e-08, + "loss": 0.7037, + "step": 33494 + }, + { + "epoch": 1.23, + "learning_rate": 7.752224704288277e-08, + "loss": 0.6899, + "step": 33495 + }, + { + "epoch": 1.23, + "learning_rate": 7.751858497821071e-08, + "loss": 0.69, + "step": 33496 + }, + { + "epoch": 1.23, + "learning_rate": 7.751492291353865e-08, + "loss": 0.6894, + "step": 33497 + }, + { + "epoch": 1.23, + "learning_rate": 7.751126084886659e-08, + "loss": 0.6896, + "step": 33498 + }, + { + "epoch": 1.23, + "learning_rate": 7.750759878419453e-08, + "loss": 0.7003, + "step": 33499 + }, + { + "epoch": 1.23, + "learning_rate": 7.750393671952247e-08, + "loss": 0.6997, + "step": 33500 + }, + { + "epoch": 1.23, + "learning_rate": 7.750027465485041e-08, + "loss": 0.6877, + "step": 33501 + }, + { + "epoch": 1.23, + "learning_rate": 7.749661259017834e-08, + "loss": 0.701, + "step": 33502 + }, + { + "epoch": 1.23, + "learning_rate": 7.749295052550628e-08, + "loss": 0.6907, + "step": 33503 + }, + { + "epoch": 1.23, + "learning_rate": 7.748928846083422e-08, + "loss": 0.6941, + "step": 33504 + }, + { + "epoch": 1.23, + "learning_rate": 7.748562639616216e-08, + "loss": 0.7037, + "step": 33505 + }, + { + "epoch": 1.23, + "learning_rate": 7.74819643314901e-08, + "loss": 0.6972, + "step": 33506 + }, + { + "epoch": 1.23, + "learning_rate": 7.747830226681804e-08, + "loss": 0.6945, + "step": 33507 + }, + { + "epoch": 1.23, + "learning_rate": 7.747464020214596e-08, + "loss": 0.6857, + "step": 33508 + }, + { + "epoch": 1.23, + "learning_rate": 7.74709781374739e-08, + "loss": 0.7147, + "step": 33509 + }, + { + "epoch": 1.23, + "learning_rate": 7.746731607280184e-08, + "loss": 0.6852, + "step": 33510 + }, + { + "epoch": 1.23, + "learning_rate": 7.746365400812977e-08, + "loss": 0.6904, + "step": 33511 + }, + { + "epoch": 1.23, + "learning_rate": 7.745999194345771e-08, + "loss": 0.6942, + "step": 33512 + }, + { + "epoch": 1.23, + "learning_rate": 7.745632987878565e-08, + "loss": 0.6899, + "step": 33513 + }, + { + "epoch": 1.23, + "learning_rate": 7.745266781411359e-08, + "loss": 0.6867, + "step": 33514 + }, + { + "epoch": 1.23, + "learning_rate": 7.744900574944153e-08, + "loss": 0.7, + "step": 33515 + }, + { + "epoch": 1.23, + "learning_rate": 7.744534368476947e-08, + "loss": 0.7188, + "step": 33516 + }, + { + "epoch": 1.23, + "learning_rate": 7.744168162009741e-08, + "loss": 0.6909, + "step": 33517 + }, + { + "epoch": 1.23, + "learning_rate": 7.743801955542535e-08, + "loss": 0.6991, + "step": 33518 + }, + { + "epoch": 1.23, + "learning_rate": 7.743435749075329e-08, + "loss": 0.6789, + "step": 33519 + }, + { + "epoch": 1.23, + "learning_rate": 7.743069542608122e-08, + "loss": 0.6928, + "step": 33520 + }, + { + "epoch": 1.23, + "learning_rate": 7.742703336140916e-08, + "loss": 0.6958, + "step": 33521 + }, + { + "epoch": 1.23, + "learning_rate": 7.74233712967371e-08, + "loss": 0.6918, + "step": 33522 + }, + { + "epoch": 1.23, + "learning_rate": 7.741970923206504e-08, + "loss": 0.7077, + "step": 33523 + }, + { + "epoch": 1.23, + "learning_rate": 7.741604716739298e-08, + "loss": 0.6885, + "step": 33524 + }, + { + "epoch": 1.23, + "learning_rate": 7.741238510272092e-08, + "loss": 0.7003, + "step": 33525 + }, + { + "epoch": 1.23, + "learning_rate": 7.740872303804886e-08, + "loss": 0.6835, + "step": 33526 + }, + { + "epoch": 1.23, + "learning_rate": 7.74050609733768e-08, + "loss": 0.7013, + "step": 33527 + }, + { + "epoch": 1.23, + "learning_rate": 7.740139890870474e-08, + "loss": 0.6941, + "step": 33528 + }, + { + "epoch": 1.23, + "learning_rate": 7.739773684403266e-08, + "loss": 0.6873, + "step": 33529 + }, + { + "epoch": 1.23, + "learning_rate": 7.73940747793606e-08, + "loss": 0.6995, + "step": 33530 + }, + { + "epoch": 1.23, + "learning_rate": 7.739041271468854e-08, + "loss": 0.6873, + "step": 33531 + }, + { + "epoch": 1.23, + "learning_rate": 7.738675065001647e-08, + "loss": 0.6947, + "step": 33532 + }, + { + "epoch": 1.23, + "learning_rate": 7.738308858534441e-08, + "loss": 0.6903, + "step": 33533 + }, + { + "epoch": 1.23, + "learning_rate": 7.737942652067235e-08, + "loss": 0.6935, + "step": 33534 + }, + { + "epoch": 1.23, + "learning_rate": 7.737576445600029e-08, + "loss": 0.6954, + "step": 33535 + }, + { + "epoch": 1.23, + "learning_rate": 7.737210239132823e-08, + "loss": 0.7044, + "step": 33536 + }, + { + "epoch": 1.23, + "learning_rate": 7.736844032665617e-08, + "loss": 0.6989, + "step": 33537 + }, + { + "epoch": 1.23, + "learning_rate": 7.73647782619841e-08, + "loss": 0.701, + "step": 33538 + }, + { + "epoch": 1.23, + "learning_rate": 7.736111619731203e-08, + "loss": 0.692, + "step": 33539 + }, + { + "epoch": 1.23, + "learning_rate": 7.735745413263997e-08, + "loss": 0.686, + "step": 33540 + }, + { + "epoch": 1.23, + "learning_rate": 7.735379206796791e-08, + "loss": 0.6883, + "step": 33541 + }, + { + "epoch": 1.23, + "learning_rate": 7.735013000329585e-08, + "loss": 0.6771, + "step": 33542 + }, + { + "epoch": 1.23, + "learning_rate": 7.73464679386238e-08, + "loss": 0.7089, + "step": 33543 + }, + { + "epoch": 1.23, + "learning_rate": 7.734280587395173e-08, + "loss": 0.6806, + "step": 33544 + }, + { + "epoch": 1.23, + "learning_rate": 7.733914380927967e-08, + "loss": 0.6999, + "step": 33545 + }, + { + "epoch": 1.23, + "learning_rate": 7.733548174460761e-08, + "loss": 0.7099, + "step": 33546 + }, + { + "epoch": 1.23, + "learning_rate": 7.733181967993554e-08, + "loss": 0.6905, + "step": 33547 + }, + { + "epoch": 1.23, + "learning_rate": 7.732815761526348e-08, + "loss": 0.6887, + "step": 33548 + }, + { + "epoch": 1.23, + "learning_rate": 7.732449555059142e-08, + "loss": 0.701, + "step": 33549 + }, + { + "epoch": 1.23, + "learning_rate": 7.732083348591936e-08, + "loss": 0.6921, + "step": 33550 + }, + { + "epoch": 1.23, + "learning_rate": 7.73171714212473e-08, + "loss": 0.6862, + "step": 33551 + }, + { + "epoch": 1.23, + "learning_rate": 7.731350935657524e-08, + "loss": 0.7107, + "step": 33552 + }, + { + "epoch": 1.23, + "learning_rate": 7.730984729190318e-08, + "loss": 0.6962, + "step": 33553 + }, + { + "epoch": 1.23, + "learning_rate": 7.730618522723111e-08, + "loss": 0.6902, + "step": 33554 + }, + { + "epoch": 1.23, + "learning_rate": 7.730252316255905e-08, + "loss": 0.6936, + "step": 33555 + }, + { + "epoch": 1.23, + "learning_rate": 7.729886109788699e-08, + "loss": 0.7011, + "step": 33556 + }, + { + "epoch": 1.23, + "learning_rate": 7.729519903321491e-08, + "loss": 0.6885, + "step": 33557 + }, + { + "epoch": 1.23, + "learning_rate": 7.729153696854285e-08, + "loss": 0.6982, + "step": 33558 + }, + { + "epoch": 1.23, + "learning_rate": 7.72878749038708e-08, + "loss": 0.7016, + "step": 33559 + }, + { + "epoch": 1.23, + "learning_rate": 7.728421283919873e-08, + "loss": 0.6939, + "step": 33560 + }, + { + "epoch": 1.23, + "learning_rate": 7.728055077452667e-08, + "loss": 0.6971, + "step": 33561 + }, + { + "epoch": 1.23, + "learning_rate": 7.727688870985461e-08, + "loss": 0.6861, + "step": 33562 + }, + { + "epoch": 1.23, + "learning_rate": 7.727322664518255e-08, + "loss": 0.7113, + "step": 33563 + }, + { + "epoch": 1.23, + "learning_rate": 7.726956458051049e-08, + "loss": 0.6802, + "step": 33564 + }, + { + "epoch": 1.23, + "learning_rate": 7.726590251583842e-08, + "loss": 0.6868, + "step": 33565 + }, + { + "epoch": 1.23, + "learning_rate": 7.726224045116636e-08, + "loss": 0.6913, + "step": 33566 + }, + { + "epoch": 1.23, + "learning_rate": 7.72585783864943e-08, + "loss": 0.6894, + "step": 33567 + }, + { + "epoch": 1.23, + "learning_rate": 7.725491632182224e-08, + "loss": 0.6936, + "step": 33568 + }, + { + "epoch": 1.23, + "learning_rate": 7.725125425715018e-08, + "loss": 0.6935, + "step": 33569 + }, + { + "epoch": 1.23, + "learning_rate": 7.724759219247812e-08, + "loss": 0.6854, + "step": 33570 + }, + { + "epoch": 1.23, + "learning_rate": 7.724393012780606e-08, + "loss": 0.6922, + "step": 33571 + }, + { + "epoch": 1.23, + "learning_rate": 7.7240268063134e-08, + "loss": 0.6964, + "step": 33572 + }, + { + "epoch": 1.23, + "learning_rate": 7.723660599846194e-08, + "loss": 0.6945, + "step": 33573 + }, + { + "epoch": 1.23, + "learning_rate": 7.723294393378987e-08, + "loss": 0.6879, + "step": 33574 + }, + { + "epoch": 1.23, + "learning_rate": 7.72292818691178e-08, + "loss": 0.6909, + "step": 33575 + }, + { + "epoch": 1.23, + "learning_rate": 7.722561980444575e-08, + "loss": 0.6953, + "step": 33576 + }, + { + "epoch": 1.23, + "learning_rate": 7.722195773977369e-08, + "loss": 0.687, + "step": 33577 + }, + { + "epoch": 1.23, + "learning_rate": 7.721829567510163e-08, + "loss": 0.7083, + "step": 33578 + }, + { + "epoch": 1.23, + "learning_rate": 7.721463361042955e-08, + "loss": 0.6965, + "step": 33579 + }, + { + "epoch": 1.23, + "learning_rate": 7.721097154575749e-08, + "loss": 0.6953, + "step": 33580 + }, + { + "epoch": 1.23, + "learning_rate": 7.720730948108543e-08, + "loss": 0.6869, + "step": 33581 + }, + { + "epoch": 1.23, + "learning_rate": 7.720364741641337e-08, + "loss": 0.6977, + "step": 33582 + }, + { + "epoch": 1.23, + "learning_rate": 7.71999853517413e-08, + "loss": 0.6967, + "step": 33583 + }, + { + "epoch": 1.23, + "learning_rate": 7.719632328706924e-08, + "loss": 0.6947, + "step": 33584 + }, + { + "epoch": 1.23, + "learning_rate": 7.719266122239718e-08, + "loss": 0.6949, + "step": 33585 + }, + { + "epoch": 1.23, + "learning_rate": 7.718899915772512e-08, + "loss": 0.6905, + "step": 33586 + }, + { + "epoch": 1.23, + "learning_rate": 7.718533709305306e-08, + "loss": 0.7021, + "step": 33587 + }, + { + "epoch": 1.23, + "learning_rate": 7.7181675028381e-08, + "loss": 0.7119, + "step": 33588 + }, + { + "epoch": 1.23, + "learning_rate": 7.717801296370894e-08, + "loss": 0.6867, + "step": 33589 + }, + { + "epoch": 1.23, + "learning_rate": 7.717435089903688e-08, + "loss": 0.7028, + "step": 33590 + }, + { + "epoch": 1.23, + "learning_rate": 7.717068883436482e-08, + "loss": 0.6948, + "step": 33591 + }, + { + "epoch": 1.23, + "learning_rate": 7.716702676969274e-08, + "loss": 0.6915, + "step": 33592 + }, + { + "epoch": 1.23, + "learning_rate": 7.716336470502068e-08, + "loss": 0.7121, + "step": 33593 + }, + { + "epoch": 1.23, + "learning_rate": 7.715970264034862e-08, + "loss": 0.6992, + "step": 33594 + }, + { + "epoch": 1.23, + "learning_rate": 7.715604057567656e-08, + "loss": 0.703, + "step": 33595 + }, + { + "epoch": 1.23, + "learning_rate": 7.71523785110045e-08, + "loss": 0.6897, + "step": 33596 + }, + { + "epoch": 1.23, + "learning_rate": 7.714871644633244e-08, + "loss": 0.6944, + "step": 33597 + }, + { + "epoch": 1.23, + "learning_rate": 7.714505438166038e-08, + "loss": 0.6864, + "step": 33598 + }, + { + "epoch": 1.23, + "learning_rate": 7.714139231698832e-08, + "loss": 0.6854, + "step": 33599 + }, + { + "epoch": 1.23, + "learning_rate": 7.713773025231625e-08, + "loss": 0.6971, + "step": 33600 + }, + { + "epoch": 1.23, + "learning_rate": 7.713406818764419e-08, + "loss": 0.7027, + "step": 33601 + }, + { + "epoch": 1.23, + "learning_rate": 7.713040612297213e-08, + "loss": 0.6868, + "step": 33602 + }, + { + "epoch": 1.23, + "learning_rate": 7.712674405830006e-08, + "loss": 0.6919, + "step": 33603 + }, + { + "epoch": 1.23, + "learning_rate": 7.7123081993628e-08, + "loss": 0.6928, + "step": 33604 + }, + { + "epoch": 1.23, + "learning_rate": 7.711941992895594e-08, + "loss": 0.6969, + "step": 33605 + }, + { + "epoch": 1.23, + "learning_rate": 7.711575786428388e-08, + "loss": 0.6948, + "step": 33606 + }, + { + "epoch": 1.23, + "learning_rate": 7.711209579961182e-08, + "loss": 0.6822, + "step": 33607 + }, + { + "epoch": 1.23, + "learning_rate": 7.710843373493976e-08, + "loss": 0.6929, + "step": 33608 + }, + { + "epoch": 1.23, + "learning_rate": 7.71047716702677e-08, + "loss": 0.6983, + "step": 33609 + }, + { + "epoch": 1.23, + "learning_rate": 7.710110960559562e-08, + "loss": 0.6888, + "step": 33610 + }, + { + "epoch": 1.23, + "learning_rate": 7.709744754092356e-08, + "loss": 0.6861, + "step": 33611 + }, + { + "epoch": 1.23, + "learning_rate": 7.70937854762515e-08, + "loss": 0.6915, + "step": 33612 + }, + { + "epoch": 1.23, + "learning_rate": 7.709012341157944e-08, + "loss": 0.6966, + "step": 33613 + }, + { + "epoch": 1.23, + "learning_rate": 7.708646134690738e-08, + "loss": 0.6932, + "step": 33614 + }, + { + "epoch": 1.23, + "learning_rate": 7.708279928223532e-08, + "loss": 0.6972, + "step": 33615 + }, + { + "epoch": 1.23, + "learning_rate": 7.707913721756326e-08, + "loss": 0.6929, + "step": 33616 + }, + { + "epoch": 1.23, + "learning_rate": 7.70754751528912e-08, + "loss": 0.6854, + "step": 33617 + }, + { + "epoch": 1.23, + "learning_rate": 7.707181308821914e-08, + "loss": 0.6925, + "step": 33618 + }, + { + "epoch": 1.23, + "learning_rate": 7.706815102354707e-08, + "loss": 0.7012, + "step": 33619 + }, + { + "epoch": 1.23, + "learning_rate": 7.706448895887501e-08, + "loss": 0.6886, + "step": 33620 + }, + { + "epoch": 1.23, + "learning_rate": 7.706082689420295e-08, + "loss": 0.6888, + "step": 33621 + }, + { + "epoch": 1.23, + "learning_rate": 7.705716482953089e-08, + "loss": 0.6926, + "step": 33622 + }, + { + "epoch": 1.23, + "learning_rate": 7.705350276485883e-08, + "loss": 0.6944, + "step": 33623 + }, + { + "epoch": 1.23, + "learning_rate": 7.704984070018677e-08, + "loss": 0.6773, + "step": 33624 + }, + { + "epoch": 1.23, + "learning_rate": 7.70461786355147e-08, + "loss": 0.7016, + "step": 33625 + }, + { + "epoch": 1.23, + "learning_rate": 7.704251657084264e-08, + "loss": 0.7039, + "step": 33626 + }, + { + "epoch": 1.23, + "learning_rate": 7.703885450617058e-08, + "loss": 0.6902, + "step": 33627 + }, + { + "epoch": 1.23, + "learning_rate": 7.70351924414985e-08, + "loss": 0.7015, + "step": 33628 + }, + { + "epoch": 1.23, + "learning_rate": 7.703153037682644e-08, + "loss": 0.7028, + "step": 33629 + }, + { + "epoch": 1.23, + "learning_rate": 7.702786831215438e-08, + "loss": 0.7004, + "step": 33630 + }, + { + "epoch": 1.23, + "learning_rate": 7.702420624748232e-08, + "loss": 0.6958, + "step": 33631 + }, + { + "epoch": 1.23, + "learning_rate": 7.702054418281026e-08, + "loss": 0.7041, + "step": 33632 + }, + { + "epoch": 1.23, + "learning_rate": 7.70168821181382e-08, + "loss": 0.687, + "step": 33633 + }, + { + "epoch": 1.23, + "learning_rate": 7.701322005346614e-08, + "loss": 0.6892, + "step": 33634 + }, + { + "epoch": 1.23, + "learning_rate": 7.700955798879408e-08, + "loss": 0.7044, + "step": 33635 + }, + { + "epoch": 1.23, + "learning_rate": 7.700589592412202e-08, + "loss": 0.6956, + "step": 33636 + }, + { + "epoch": 1.23, + "learning_rate": 7.700223385944995e-08, + "loss": 0.6805, + "step": 33637 + }, + { + "epoch": 1.23, + "learning_rate": 7.699857179477789e-08, + "loss": 0.6848, + "step": 33638 + }, + { + "epoch": 1.23, + "learning_rate": 7.699490973010583e-08, + "loss": 0.6811, + "step": 33639 + }, + { + "epoch": 1.23, + "learning_rate": 7.699124766543377e-08, + "loss": 0.6951, + "step": 33640 + }, + { + "epoch": 1.23, + "learning_rate": 7.698758560076171e-08, + "loss": 0.6898, + "step": 33641 + }, + { + "epoch": 1.23, + "learning_rate": 7.698392353608965e-08, + "loss": 0.6793, + "step": 33642 + }, + { + "epoch": 1.23, + "learning_rate": 7.698026147141759e-08, + "loss": 0.6876, + "step": 33643 + }, + { + "epoch": 1.23, + "learning_rate": 7.697659940674553e-08, + "loss": 0.6878, + "step": 33644 + }, + { + "epoch": 1.23, + "learning_rate": 7.697293734207347e-08, + "loss": 0.6832, + "step": 33645 + }, + { + "epoch": 1.23, + "learning_rate": 7.69692752774014e-08, + "loss": 0.6906, + "step": 33646 + }, + { + "epoch": 1.23, + "learning_rate": 7.696561321272934e-08, + "loss": 0.6907, + "step": 33647 + }, + { + "epoch": 1.23, + "learning_rate": 7.696195114805728e-08, + "loss": 0.6902, + "step": 33648 + }, + { + "epoch": 1.23, + "learning_rate": 7.69582890833852e-08, + "loss": 0.6944, + "step": 33649 + }, + { + "epoch": 1.23, + "learning_rate": 7.695462701871314e-08, + "loss": 0.7013, + "step": 33650 + }, + { + "epoch": 1.23, + "learning_rate": 7.695096495404108e-08, + "loss": 0.7086, + "step": 33651 + }, + { + "epoch": 1.23, + "learning_rate": 7.694730288936902e-08, + "loss": 0.6859, + "step": 33652 + }, + { + "epoch": 1.23, + "learning_rate": 7.694364082469696e-08, + "loss": 0.6869, + "step": 33653 + }, + { + "epoch": 1.23, + "learning_rate": 7.69399787600249e-08, + "loss": 0.6948, + "step": 33654 + }, + { + "epoch": 1.23, + "learning_rate": 7.693631669535283e-08, + "loss": 0.6864, + "step": 33655 + }, + { + "epoch": 1.23, + "learning_rate": 7.693265463068077e-08, + "loss": 0.6981, + "step": 33656 + }, + { + "epoch": 1.23, + "learning_rate": 7.692899256600871e-08, + "loss": 0.6969, + "step": 33657 + }, + { + "epoch": 1.23, + "learning_rate": 7.692533050133665e-08, + "loss": 0.7002, + "step": 33658 + }, + { + "epoch": 1.23, + "learning_rate": 7.692166843666459e-08, + "loss": 0.7035, + "step": 33659 + }, + { + "epoch": 1.23, + "learning_rate": 7.691800637199253e-08, + "loss": 0.6937, + "step": 33660 + }, + { + "epoch": 1.23, + "learning_rate": 7.691434430732047e-08, + "loss": 0.6756, + "step": 33661 + }, + { + "epoch": 1.23, + "learning_rate": 7.691068224264841e-08, + "loss": 0.6724, + "step": 33662 + }, + { + "epoch": 1.23, + "learning_rate": 7.690702017797635e-08, + "loss": 0.6969, + "step": 33663 + }, + { + "epoch": 1.23, + "learning_rate": 7.690335811330427e-08, + "loss": 0.68, + "step": 33664 + }, + { + "epoch": 1.23, + "learning_rate": 7.689969604863221e-08, + "loss": 0.6903, + "step": 33665 + }, + { + "epoch": 1.23, + "learning_rate": 7.689603398396015e-08, + "loss": 0.6797, + "step": 33666 + }, + { + "epoch": 1.23, + "learning_rate": 7.68923719192881e-08, + "loss": 0.6978, + "step": 33667 + }, + { + "epoch": 1.23, + "learning_rate": 7.688870985461603e-08, + "loss": 0.688, + "step": 33668 + }, + { + "epoch": 1.23, + "learning_rate": 7.688504778994397e-08, + "loss": 0.6904, + "step": 33669 + }, + { + "epoch": 1.23, + "learning_rate": 7.688138572527191e-08, + "loss": 0.7046, + "step": 33670 + }, + { + "epoch": 1.23, + "learning_rate": 7.687772366059984e-08, + "loss": 0.6931, + "step": 33671 + }, + { + "epoch": 1.23, + "learning_rate": 7.687406159592778e-08, + "loss": 0.6896, + "step": 33672 + }, + { + "epoch": 1.23, + "learning_rate": 7.687039953125572e-08, + "loss": 0.6883, + "step": 33673 + }, + { + "epoch": 1.23, + "learning_rate": 7.686673746658365e-08, + "loss": 0.6865, + "step": 33674 + }, + { + "epoch": 1.23, + "learning_rate": 7.686307540191159e-08, + "loss": 0.6949, + "step": 33675 + }, + { + "epoch": 1.23, + "learning_rate": 7.685941333723953e-08, + "loss": 0.6665, + "step": 33676 + }, + { + "epoch": 1.23, + "learning_rate": 7.685575127256747e-08, + "loss": 0.6836, + "step": 33677 + }, + { + "epoch": 1.23, + "learning_rate": 7.685208920789541e-08, + "loss": 0.6879, + "step": 33678 + }, + { + "epoch": 1.23, + "learning_rate": 7.684842714322335e-08, + "loss": 0.6927, + "step": 33679 + }, + { + "epoch": 1.23, + "learning_rate": 7.684476507855129e-08, + "loss": 0.6956, + "step": 33680 + }, + { + "epoch": 1.23, + "learning_rate": 7.684110301387923e-08, + "loss": 0.691, + "step": 33681 + }, + { + "epoch": 1.23, + "learning_rate": 7.683744094920715e-08, + "loss": 0.6866, + "step": 33682 + }, + { + "epoch": 1.23, + "learning_rate": 7.683377888453509e-08, + "loss": 0.6916, + "step": 33683 + }, + { + "epoch": 1.23, + "learning_rate": 7.683011681986303e-08, + "loss": 0.7043, + "step": 33684 + }, + { + "epoch": 1.23, + "learning_rate": 7.682645475519097e-08, + "loss": 0.6884, + "step": 33685 + }, + { + "epoch": 1.23, + "learning_rate": 7.682279269051891e-08, + "loss": 0.6912, + "step": 33686 + }, + { + "epoch": 1.23, + "learning_rate": 7.681913062584685e-08, + "loss": 0.7005, + "step": 33687 + }, + { + "epoch": 1.23, + "learning_rate": 7.681546856117479e-08, + "loss": 0.6853, + "step": 33688 + }, + { + "epoch": 1.23, + "learning_rate": 7.681180649650273e-08, + "loss": 0.6979, + "step": 33689 + }, + { + "epoch": 1.23, + "learning_rate": 7.680814443183067e-08, + "loss": 0.6869, + "step": 33690 + }, + { + "epoch": 1.23, + "learning_rate": 7.68044823671586e-08, + "loss": 0.6978, + "step": 33691 + }, + { + "epoch": 1.23, + "learning_rate": 7.680082030248654e-08, + "loss": 0.6868, + "step": 33692 + }, + { + "epoch": 1.23, + "learning_rate": 7.679715823781448e-08, + "loss": 0.6901, + "step": 33693 + }, + { + "epoch": 1.23, + "learning_rate": 7.679349617314242e-08, + "loss": 0.6992, + "step": 33694 + }, + { + "epoch": 1.23, + "learning_rate": 7.678983410847036e-08, + "loss": 0.6773, + "step": 33695 + }, + { + "epoch": 1.23, + "learning_rate": 7.678617204379829e-08, + "loss": 0.688, + "step": 33696 + }, + { + "epoch": 1.23, + "learning_rate": 7.678250997912623e-08, + "loss": 0.6834, + "step": 33697 + }, + { + "epoch": 1.23, + "learning_rate": 7.677884791445417e-08, + "loss": 0.699, + "step": 33698 + }, + { + "epoch": 1.23, + "learning_rate": 7.67751858497821e-08, + "loss": 0.6783, + "step": 33699 + }, + { + "epoch": 1.23, + "learning_rate": 7.677152378511003e-08, + "loss": 0.6991, + "step": 33700 + }, + { + "epoch": 1.23, + "learning_rate": 7.676786172043797e-08, + "loss": 0.6948, + "step": 33701 + }, + { + "epoch": 1.23, + "learning_rate": 7.676419965576591e-08, + "loss": 0.6967, + "step": 33702 + }, + { + "epoch": 1.23, + "learning_rate": 7.676053759109385e-08, + "loss": 0.6817, + "step": 33703 + }, + { + "epoch": 1.23, + "learning_rate": 7.675687552642179e-08, + "loss": 0.6916, + "step": 33704 + }, + { + "epoch": 1.23, + "learning_rate": 7.675321346174973e-08, + "loss": 0.688, + "step": 33705 + }, + { + "epoch": 1.23, + "learning_rate": 7.674955139707767e-08, + "loss": 0.6806, + "step": 33706 + }, + { + "epoch": 1.23, + "learning_rate": 7.674588933240561e-08, + "loss": 0.7079, + "step": 33707 + }, + { + "epoch": 1.23, + "learning_rate": 7.674222726773354e-08, + "loss": 0.6911, + "step": 33708 + }, + { + "epoch": 1.23, + "learning_rate": 7.673856520306148e-08, + "loss": 0.6913, + "step": 33709 + }, + { + "epoch": 1.23, + "learning_rate": 7.673490313838942e-08, + "loss": 0.699, + "step": 33710 + }, + { + "epoch": 1.23, + "learning_rate": 7.673124107371736e-08, + "loss": 0.6927, + "step": 33711 + }, + { + "epoch": 1.23, + "learning_rate": 7.67275790090453e-08, + "loss": 0.6879, + "step": 33712 + }, + { + "epoch": 1.23, + "learning_rate": 7.672391694437324e-08, + "loss": 0.6837, + "step": 33713 + }, + { + "epoch": 1.23, + "learning_rate": 7.672025487970118e-08, + "loss": 0.6906, + "step": 33714 + }, + { + "epoch": 1.23, + "learning_rate": 7.671659281502912e-08, + "loss": 0.6947, + "step": 33715 + }, + { + "epoch": 1.23, + "learning_rate": 7.671293075035706e-08, + "loss": 0.7048, + "step": 33716 + }, + { + "epoch": 1.23, + "learning_rate": 7.670926868568498e-08, + "loss": 0.6834, + "step": 33717 + }, + { + "epoch": 1.23, + "learning_rate": 7.670560662101292e-08, + "loss": 0.6803, + "step": 33718 + }, + { + "epoch": 1.23, + "learning_rate": 7.670194455634086e-08, + "loss": 0.6917, + "step": 33719 + }, + { + "epoch": 1.23, + "learning_rate": 7.669828249166879e-08, + "loss": 0.6908, + "step": 33720 + }, + { + "epoch": 1.23, + "learning_rate": 7.669462042699673e-08, + "loss": 0.6943, + "step": 33721 + }, + { + "epoch": 1.23, + "learning_rate": 7.669095836232467e-08, + "loss": 0.6844, + "step": 33722 + }, + { + "epoch": 1.23, + "learning_rate": 7.668729629765261e-08, + "loss": 0.6966, + "step": 33723 + }, + { + "epoch": 1.23, + "learning_rate": 7.668363423298055e-08, + "loss": 0.6972, + "step": 33724 + }, + { + "epoch": 1.23, + "learning_rate": 7.667997216830849e-08, + "loss": 0.6876, + "step": 33725 + }, + { + "epoch": 1.23, + "learning_rate": 7.667631010363642e-08, + "loss": 0.6894, + "step": 33726 + }, + { + "epoch": 1.23, + "learning_rate": 7.667264803896436e-08, + "loss": 0.6926, + "step": 33727 + }, + { + "epoch": 1.23, + "learning_rate": 7.66689859742923e-08, + "loss": 0.696, + "step": 33728 + }, + { + "epoch": 1.23, + "learning_rate": 7.666532390962024e-08, + "loss": 0.6902, + "step": 33729 + }, + { + "epoch": 1.23, + "learning_rate": 7.666166184494818e-08, + "loss": 0.6849, + "step": 33730 + }, + { + "epoch": 1.23, + "learning_rate": 7.665799978027612e-08, + "loss": 0.7012, + "step": 33731 + }, + { + "epoch": 1.23, + "learning_rate": 7.665433771560406e-08, + "loss": 0.6958, + "step": 33732 + }, + { + "epoch": 1.23, + "learning_rate": 7.6650675650932e-08, + "loss": 0.6844, + "step": 33733 + }, + { + "epoch": 1.23, + "learning_rate": 7.664701358625994e-08, + "loss": 0.6761, + "step": 33734 + }, + { + "epoch": 1.23, + "learning_rate": 7.664335152158786e-08, + "loss": 0.6758, + "step": 33735 + }, + { + "epoch": 1.23, + "learning_rate": 7.66396894569158e-08, + "loss": 0.683, + "step": 33736 + }, + { + "epoch": 1.23, + "learning_rate": 7.663602739224374e-08, + "loss": 0.6889, + "step": 33737 + }, + { + "epoch": 1.23, + "learning_rate": 7.663236532757168e-08, + "loss": 0.6938, + "step": 33738 + }, + { + "epoch": 1.23, + "learning_rate": 7.662870326289962e-08, + "loss": 0.696, + "step": 33739 + }, + { + "epoch": 1.23, + "learning_rate": 7.662504119822756e-08, + "loss": 0.6948, + "step": 33740 + }, + { + "epoch": 1.23, + "learning_rate": 7.66213791335555e-08, + "loss": 0.6991, + "step": 33741 + }, + { + "epoch": 1.23, + "learning_rate": 7.661771706888343e-08, + "loss": 0.6967, + "step": 33742 + }, + { + "epoch": 1.23, + "learning_rate": 7.661405500421137e-08, + "loss": 0.6991, + "step": 33743 + }, + { + "epoch": 1.23, + "learning_rate": 7.661039293953931e-08, + "loss": 0.6963, + "step": 33744 + }, + { + "epoch": 1.23, + "learning_rate": 7.660673087486724e-08, + "loss": 0.6964, + "step": 33745 + }, + { + "epoch": 1.23, + "learning_rate": 7.660306881019518e-08, + "loss": 0.6915, + "step": 33746 + }, + { + "epoch": 1.23, + "learning_rate": 7.659940674552312e-08, + "loss": 0.6926, + "step": 33747 + }, + { + "epoch": 1.23, + "learning_rate": 7.659574468085106e-08, + "loss": 0.6959, + "step": 33748 + }, + { + "epoch": 1.23, + "learning_rate": 7.6592082616179e-08, + "loss": 0.7139, + "step": 33749 + }, + { + "epoch": 1.23, + "learning_rate": 7.658842055150694e-08, + "loss": 0.6863, + "step": 33750 + }, + { + "epoch": 1.23, + "learning_rate": 7.658475848683488e-08, + "loss": 0.6974, + "step": 33751 + }, + { + "epoch": 1.23, + "learning_rate": 7.658109642216282e-08, + "loss": 0.6998, + "step": 33752 + }, + { + "epoch": 1.23, + "learning_rate": 7.657743435749074e-08, + "loss": 0.6816, + "step": 33753 + }, + { + "epoch": 1.23, + "learning_rate": 7.657377229281868e-08, + "loss": 0.696, + "step": 33754 + }, + { + "epoch": 1.23, + "learning_rate": 7.657011022814662e-08, + "loss": 0.6943, + "step": 33755 + }, + { + "epoch": 1.24, + "learning_rate": 7.656644816347456e-08, + "loss": 0.6904, + "step": 33756 + }, + { + "epoch": 1.24, + "learning_rate": 7.65627860988025e-08, + "loss": 0.6843, + "step": 33757 + }, + { + "epoch": 1.24, + "learning_rate": 7.655912403413044e-08, + "loss": 0.6795, + "step": 33758 + }, + { + "epoch": 1.24, + "learning_rate": 7.655546196945838e-08, + "loss": 0.6787, + "step": 33759 + }, + { + "epoch": 1.24, + "learning_rate": 7.655179990478632e-08, + "loss": 0.6921, + "step": 33760 + }, + { + "epoch": 1.24, + "learning_rate": 7.654813784011426e-08, + "loss": 0.6931, + "step": 33761 + }, + { + "epoch": 1.24, + "learning_rate": 7.654447577544219e-08, + "loss": 0.6792, + "step": 33762 + }, + { + "epoch": 1.24, + "learning_rate": 7.654081371077013e-08, + "loss": 0.6783, + "step": 33763 + }, + { + "epoch": 1.24, + "learning_rate": 7.653715164609807e-08, + "loss": 0.7051, + "step": 33764 + }, + { + "epoch": 1.24, + "learning_rate": 7.653348958142601e-08, + "loss": 0.6863, + "step": 33765 + }, + { + "epoch": 1.24, + "learning_rate": 7.652982751675395e-08, + "loss": 0.6955, + "step": 33766 + }, + { + "epoch": 1.24, + "learning_rate": 7.652616545208188e-08, + "loss": 0.6893, + "step": 33767 + }, + { + "epoch": 1.24, + "learning_rate": 7.652250338740982e-08, + "loss": 0.6873, + "step": 33768 + }, + { + "epoch": 1.24, + "learning_rate": 7.651884132273775e-08, + "loss": 0.6954, + "step": 33769 + }, + { + "epoch": 1.24, + "learning_rate": 7.65151792580657e-08, + "loss": 0.6866, + "step": 33770 + }, + { + "epoch": 1.24, + "learning_rate": 7.651151719339362e-08, + "loss": 0.6973, + "step": 33771 + }, + { + "epoch": 1.24, + "learning_rate": 7.650785512872156e-08, + "loss": 0.683, + "step": 33772 + }, + { + "epoch": 1.24, + "learning_rate": 7.65041930640495e-08, + "loss": 0.7037, + "step": 33773 + }, + { + "epoch": 1.24, + "learning_rate": 7.650053099937744e-08, + "loss": 0.6789, + "step": 33774 + }, + { + "epoch": 1.24, + "learning_rate": 7.649686893470538e-08, + "loss": 0.7004, + "step": 33775 + }, + { + "epoch": 1.24, + "learning_rate": 7.649320687003332e-08, + "loss": 0.7073, + "step": 33776 + }, + { + "epoch": 1.24, + "learning_rate": 7.648954480536126e-08, + "loss": 0.6869, + "step": 33777 + }, + { + "epoch": 1.24, + "learning_rate": 7.64858827406892e-08, + "loss": 0.6904, + "step": 33778 + }, + { + "epoch": 1.24, + "learning_rate": 7.648222067601714e-08, + "loss": 0.6934, + "step": 33779 + }, + { + "epoch": 1.24, + "learning_rate": 7.647855861134507e-08, + "loss": 0.6907, + "step": 33780 + }, + { + "epoch": 1.24, + "learning_rate": 7.647489654667301e-08, + "loss": 0.6958, + "step": 33781 + }, + { + "epoch": 1.24, + "learning_rate": 7.647123448200095e-08, + "loss": 0.6997, + "step": 33782 + }, + { + "epoch": 1.24, + "learning_rate": 7.646757241732889e-08, + "loss": 0.6948, + "step": 33783 + }, + { + "epoch": 1.24, + "learning_rate": 7.646391035265683e-08, + "loss": 0.6915, + "step": 33784 + }, + { + "epoch": 1.24, + "learning_rate": 7.646024828798477e-08, + "loss": 0.7015, + "step": 33785 + }, + { + "epoch": 1.24, + "learning_rate": 7.645658622331271e-08, + "loss": 0.6876, + "step": 33786 + }, + { + "epoch": 1.24, + "learning_rate": 7.645292415864065e-08, + "loss": 0.6979, + "step": 33787 + }, + { + "epoch": 1.24, + "learning_rate": 7.644926209396859e-08, + "loss": 0.6962, + "step": 33788 + }, + { + "epoch": 1.24, + "learning_rate": 7.644560002929651e-08, + "loss": 0.6811, + "step": 33789 + }, + { + "epoch": 1.24, + "learning_rate": 7.644193796462445e-08, + "loss": 0.6883, + "step": 33790 + }, + { + "epoch": 1.24, + "learning_rate": 7.643827589995238e-08, + "loss": 0.7082, + "step": 33791 + }, + { + "epoch": 1.24, + "learning_rate": 7.643461383528032e-08, + "loss": 0.7067, + "step": 33792 + }, + { + "epoch": 1.24, + "learning_rate": 7.643095177060826e-08, + "loss": 0.6894, + "step": 33793 + }, + { + "epoch": 1.24, + "learning_rate": 7.64272897059362e-08, + "loss": 0.6919, + "step": 33794 + }, + { + "epoch": 1.24, + "learning_rate": 7.642362764126414e-08, + "loss": 0.6934, + "step": 33795 + }, + { + "epoch": 1.24, + "learning_rate": 7.641996557659208e-08, + "loss": 0.6899, + "step": 33796 + }, + { + "epoch": 1.24, + "learning_rate": 7.641630351192002e-08, + "loss": 0.6832, + "step": 33797 + }, + { + "epoch": 1.24, + "learning_rate": 7.641264144724795e-08, + "loss": 0.6996, + "step": 33798 + }, + { + "epoch": 1.24, + "learning_rate": 7.640897938257589e-08, + "loss": 0.6999, + "step": 33799 + }, + { + "epoch": 1.24, + "learning_rate": 7.640531731790383e-08, + "loss": 0.6937, + "step": 33800 + }, + { + "epoch": 1.24, + "learning_rate": 7.640165525323177e-08, + "loss": 0.6827, + "step": 33801 + }, + { + "epoch": 1.24, + "learning_rate": 7.63979931885597e-08, + "loss": 0.7001, + "step": 33802 + }, + { + "epoch": 1.24, + "learning_rate": 7.639433112388765e-08, + "loss": 0.6997, + "step": 33803 + }, + { + "epoch": 1.24, + "learning_rate": 7.639066905921559e-08, + "loss": 0.6979, + "step": 33804 + }, + { + "epoch": 1.24, + "learning_rate": 7.638700699454353e-08, + "loss": 0.6914, + "step": 33805 + }, + { + "epoch": 1.24, + "learning_rate": 7.638334492987147e-08, + "loss": 0.6874, + "step": 33806 + }, + { + "epoch": 1.24, + "learning_rate": 7.637968286519939e-08, + "loss": 0.7052, + "step": 33807 + }, + { + "epoch": 1.24, + "learning_rate": 7.637602080052733e-08, + "loss": 0.6988, + "step": 33808 + }, + { + "epoch": 1.24, + "learning_rate": 7.637235873585527e-08, + "loss": 0.6948, + "step": 33809 + }, + { + "epoch": 1.24, + "learning_rate": 7.636869667118321e-08, + "loss": 0.6948, + "step": 33810 + }, + { + "epoch": 1.24, + "learning_rate": 7.636503460651115e-08, + "loss": 0.7001, + "step": 33811 + }, + { + "epoch": 1.24, + "learning_rate": 7.636137254183909e-08, + "loss": 0.7022, + "step": 33812 + }, + { + "epoch": 1.24, + "learning_rate": 7.635771047716702e-08, + "loss": 0.7053, + "step": 33813 + }, + { + "epoch": 1.24, + "learning_rate": 7.635404841249496e-08, + "loss": 0.6911, + "step": 33814 + }, + { + "epoch": 1.24, + "learning_rate": 7.63503863478229e-08, + "loss": 0.7064, + "step": 33815 + }, + { + "epoch": 1.24, + "learning_rate": 7.634672428315083e-08, + "loss": 0.6857, + "step": 33816 + }, + { + "epoch": 1.24, + "learning_rate": 7.634306221847877e-08, + "loss": 0.701, + "step": 33817 + }, + { + "epoch": 1.24, + "learning_rate": 7.63394001538067e-08, + "loss": 0.6933, + "step": 33818 + }, + { + "epoch": 1.24, + "learning_rate": 7.633573808913465e-08, + "loss": 0.6813, + "step": 33819 + }, + { + "epoch": 1.24, + "learning_rate": 7.633207602446259e-08, + "loss": 0.6838, + "step": 33820 + }, + { + "epoch": 1.24, + "learning_rate": 7.632841395979053e-08, + "loss": 0.6818, + "step": 33821 + }, + { + "epoch": 1.24, + "learning_rate": 7.632475189511847e-08, + "loss": 0.6924, + "step": 33822 + }, + { + "epoch": 1.24, + "learning_rate": 7.63210898304464e-08, + "loss": 0.6738, + "step": 33823 + }, + { + "epoch": 1.24, + "learning_rate": 7.631742776577435e-08, + "loss": 0.7039, + "step": 33824 + }, + { + "epoch": 1.24, + "learning_rate": 7.631376570110227e-08, + "loss": 0.6838, + "step": 33825 + }, + { + "epoch": 1.24, + "learning_rate": 7.631010363643021e-08, + "loss": 0.6939, + "step": 33826 + }, + { + "epoch": 1.24, + "learning_rate": 7.630644157175815e-08, + "loss": 0.6986, + "step": 33827 + }, + { + "epoch": 1.24, + "learning_rate": 7.630277950708609e-08, + "loss": 0.6873, + "step": 33828 + }, + { + "epoch": 1.24, + "learning_rate": 7.629911744241403e-08, + "loss": 0.6946, + "step": 33829 + }, + { + "epoch": 1.24, + "learning_rate": 7.629545537774197e-08, + "loss": 0.6933, + "step": 33830 + }, + { + "epoch": 1.24, + "learning_rate": 7.629179331306991e-08, + "loss": 0.6949, + "step": 33831 + }, + { + "epoch": 1.24, + "learning_rate": 7.628813124839785e-08, + "loss": 0.7028, + "step": 33832 + }, + { + "epoch": 1.24, + "learning_rate": 7.628446918372579e-08, + "loss": 0.6942, + "step": 33833 + }, + { + "epoch": 1.24, + "learning_rate": 7.628080711905372e-08, + "loss": 0.6922, + "step": 33834 + }, + { + "epoch": 1.24, + "learning_rate": 7.627714505438166e-08, + "loss": 0.6972, + "step": 33835 + }, + { + "epoch": 1.24, + "learning_rate": 7.62734829897096e-08, + "loss": 0.6926, + "step": 33836 + }, + { + "epoch": 1.24, + "learning_rate": 7.626982092503754e-08, + "loss": 0.6903, + "step": 33837 + }, + { + "epoch": 1.24, + "learning_rate": 7.626615886036546e-08, + "loss": 0.7006, + "step": 33838 + }, + { + "epoch": 1.24, + "learning_rate": 7.62624967956934e-08, + "loss": 0.6647, + "step": 33839 + }, + { + "epoch": 1.24, + "learning_rate": 7.625883473102134e-08, + "loss": 0.6986, + "step": 33840 + }, + { + "epoch": 1.24, + "learning_rate": 7.625517266634928e-08, + "loss": 0.6957, + "step": 33841 + }, + { + "epoch": 1.24, + "learning_rate": 7.625151060167722e-08, + "loss": 0.6946, + "step": 33842 + }, + { + "epoch": 1.24, + "learning_rate": 7.624784853700515e-08, + "loss": 0.6737, + "step": 33843 + }, + { + "epoch": 1.24, + "learning_rate": 7.624418647233309e-08, + "loss": 0.6963, + "step": 33844 + }, + { + "epoch": 1.24, + "learning_rate": 7.624052440766103e-08, + "loss": 0.6935, + "step": 33845 + }, + { + "epoch": 1.24, + "learning_rate": 7.623686234298897e-08, + "loss": 0.6957, + "step": 33846 + }, + { + "epoch": 1.24, + "learning_rate": 7.623320027831691e-08, + "loss": 0.6963, + "step": 33847 + }, + { + "epoch": 1.24, + "learning_rate": 7.622953821364485e-08, + "loss": 0.6974, + "step": 33848 + }, + { + "epoch": 1.24, + "learning_rate": 7.622587614897279e-08, + "loss": 0.697, + "step": 33849 + }, + { + "epoch": 1.24, + "learning_rate": 7.622221408430073e-08, + "loss": 0.7027, + "step": 33850 + }, + { + "epoch": 1.24, + "learning_rate": 7.621855201962867e-08, + "loss": 0.6891, + "step": 33851 + }, + { + "epoch": 1.24, + "learning_rate": 7.62148899549566e-08, + "loss": 0.7012, + "step": 33852 + }, + { + "epoch": 1.24, + "learning_rate": 7.621122789028454e-08, + "loss": 0.6787, + "step": 33853 + }, + { + "epoch": 1.24, + "learning_rate": 7.620756582561248e-08, + "loss": 0.6932, + "step": 33854 + }, + { + "epoch": 1.24, + "learning_rate": 7.620390376094042e-08, + "loss": 0.6987, + "step": 33855 + }, + { + "epoch": 1.24, + "learning_rate": 7.620024169626836e-08, + "loss": 0.7012, + "step": 33856 + }, + { + "epoch": 1.24, + "learning_rate": 7.61965796315963e-08, + "loss": 0.7047, + "step": 33857 + }, + { + "epoch": 1.24, + "learning_rate": 7.619291756692424e-08, + "loss": 0.6801, + "step": 33858 + }, + { + "epoch": 1.24, + "learning_rate": 7.618925550225218e-08, + "loss": 0.6948, + "step": 33859 + }, + { + "epoch": 1.24, + "learning_rate": 7.61855934375801e-08, + "loss": 0.6929, + "step": 33860 + }, + { + "epoch": 1.24, + "learning_rate": 7.618193137290804e-08, + "loss": 0.6914, + "step": 33861 + }, + { + "epoch": 1.24, + "learning_rate": 7.617826930823597e-08, + "loss": 0.6878, + "step": 33862 + }, + { + "epoch": 1.24, + "learning_rate": 7.617460724356391e-08, + "loss": 0.6912, + "step": 33863 + }, + { + "epoch": 1.24, + "learning_rate": 7.617094517889185e-08, + "loss": 0.6806, + "step": 33864 + }, + { + "epoch": 1.24, + "learning_rate": 7.616728311421979e-08, + "loss": 0.683, + "step": 33865 + }, + { + "epoch": 1.24, + "learning_rate": 7.616362104954773e-08, + "loss": 0.6843, + "step": 33866 + }, + { + "epoch": 1.24, + "learning_rate": 7.615995898487567e-08, + "loss": 0.6997, + "step": 33867 + }, + { + "epoch": 1.24, + "learning_rate": 7.615629692020361e-08, + "loss": 0.682, + "step": 33868 + }, + { + "epoch": 1.24, + "learning_rate": 7.615263485553155e-08, + "loss": 0.6971, + "step": 33869 + }, + { + "epoch": 1.24, + "learning_rate": 7.614897279085948e-08, + "loss": 0.701, + "step": 33870 + }, + { + "epoch": 1.24, + "learning_rate": 7.614531072618742e-08, + "loss": 0.6998, + "step": 33871 + }, + { + "epoch": 1.24, + "learning_rate": 7.614164866151536e-08, + "loss": 0.6974, + "step": 33872 + }, + { + "epoch": 1.24, + "learning_rate": 7.61379865968433e-08, + "loss": 0.7069, + "step": 33873 + }, + { + "epoch": 1.24, + "learning_rate": 7.613432453217124e-08, + "loss": 0.6971, + "step": 33874 + }, + { + "epoch": 1.24, + "learning_rate": 7.613066246749918e-08, + "loss": 0.6906, + "step": 33875 + }, + { + "epoch": 1.24, + "learning_rate": 7.612700040282712e-08, + "loss": 0.6937, + "step": 33876 + }, + { + "epoch": 1.24, + "learning_rate": 7.612333833815506e-08, + "loss": 0.6988, + "step": 33877 + }, + { + "epoch": 1.24, + "learning_rate": 7.6119676273483e-08, + "loss": 0.6926, + "step": 33878 + }, + { + "epoch": 1.24, + "learning_rate": 7.611601420881092e-08, + "loss": 0.707, + "step": 33879 + }, + { + "epoch": 1.24, + "learning_rate": 7.611235214413886e-08, + "loss": 0.6905, + "step": 33880 + }, + { + "epoch": 1.24, + "learning_rate": 7.61086900794668e-08, + "loss": 0.6936, + "step": 33881 + }, + { + "epoch": 1.24, + "learning_rate": 7.610502801479474e-08, + "loss": 0.6765, + "step": 33882 + }, + { + "epoch": 1.24, + "learning_rate": 7.610136595012268e-08, + "loss": 0.6943, + "step": 33883 + }, + { + "epoch": 1.24, + "learning_rate": 7.609770388545061e-08, + "loss": 0.6945, + "step": 33884 + }, + { + "epoch": 1.24, + "learning_rate": 7.609404182077855e-08, + "loss": 0.7139, + "step": 33885 + }, + { + "epoch": 1.24, + "learning_rate": 7.609037975610649e-08, + "loss": 0.6962, + "step": 33886 + }, + { + "epoch": 1.24, + "learning_rate": 7.608671769143443e-08, + "loss": 0.7061, + "step": 33887 + }, + { + "epoch": 1.24, + "learning_rate": 7.608305562676235e-08, + "loss": 0.6764, + "step": 33888 + }, + { + "epoch": 1.24, + "learning_rate": 7.60793935620903e-08, + "loss": 0.687, + "step": 33889 + }, + { + "epoch": 1.24, + "learning_rate": 7.607573149741823e-08, + "loss": 0.6946, + "step": 33890 + }, + { + "epoch": 1.24, + "learning_rate": 7.607206943274617e-08, + "loss": 0.6866, + "step": 33891 + }, + { + "epoch": 1.24, + "learning_rate": 7.606840736807411e-08, + "loss": 0.6857, + "step": 33892 + }, + { + "epoch": 1.24, + "learning_rate": 7.606474530340205e-08, + "loss": 0.693, + "step": 33893 + }, + { + "epoch": 1.24, + "learning_rate": 7.606108323873e-08, + "loss": 0.6858, + "step": 33894 + }, + { + "epoch": 1.24, + "learning_rate": 7.605742117405793e-08, + "loss": 0.7006, + "step": 33895 + }, + { + "epoch": 1.24, + "learning_rate": 7.605375910938587e-08, + "loss": 0.6926, + "step": 33896 + }, + { + "epoch": 1.24, + "learning_rate": 7.60500970447138e-08, + "loss": 0.6986, + "step": 33897 + }, + { + "epoch": 1.24, + "learning_rate": 7.604643498004174e-08, + "loss": 0.6986, + "step": 33898 + }, + { + "epoch": 1.24, + "learning_rate": 7.604277291536968e-08, + "loss": 0.6913, + "step": 33899 + }, + { + "epoch": 1.24, + "learning_rate": 7.603911085069762e-08, + "loss": 0.6937, + "step": 33900 + }, + { + "epoch": 1.24, + "learning_rate": 7.603544878602556e-08, + "loss": 0.6795, + "step": 33901 + }, + { + "epoch": 1.24, + "learning_rate": 7.60317867213535e-08, + "loss": 0.7008, + "step": 33902 + }, + { + "epoch": 1.24, + "learning_rate": 7.602812465668144e-08, + "loss": 0.6819, + "step": 33903 + }, + { + "epoch": 1.24, + "learning_rate": 7.602446259200938e-08, + "loss": 0.7002, + "step": 33904 + }, + { + "epoch": 1.24, + "learning_rate": 7.602080052733732e-08, + "loss": 0.6914, + "step": 33905 + }, + { + "epoch": 1.24, + "learning_rate": 7.601713846266525e-08, + "loss": 0.7002, + "step": 33906 + }, + { + "epoch": 1.24, + "learning_rate": 7.601347639799319e-08, + "loss": 0.693, + "step": 33907 + }, + { + "epoch": 1.24, + "learning_rate": 7.600981433332113e-08, + "loss": 0.6901, + "step": 33908 + }, + { + "epoch": 1.24, + "learning_rate": 7.600615226864905e-08, + "loss": 0.7027, + "step": 33909 + }, + { + "epoch": 1.24, + "learning_rate": 7.6002490203977e-08, + "loss": 0.696, + "step": 33910 + }, + { + "epoch": 1.24, + "learning_rate": 7.599882813930493e-08, + "loss": 0.6969, + "step": 33911 + }, + { + "epoch": 1.24, + "learning_rate": 7.599516607463287e-08, + "loss": 0.7034, + "step": 33912 + }, + { + "epoch": 1.24, + "learning_rate": 7.599150400996081e-08, + "loss": 0.7008, + "step": 33913 + }, + { + "epoch": 1.24, + "learning_rate": 7.598784194528875e-08, + "loss": 0.6942, + "step": 33914 + }, + { + "epoch": 1.24, + "learning_rate": 7.598417988061668e-08, + "loss": 0.6955, + "step": 33915 + }, + { + "epoch": 1.24, + "learning_rate": 7.598051781594462e-08, + "loss": 0.6889, + "step": 33916 + }, + { + "epoch": 1.24, + "learning_rate": 7.597685575127256e-08, + "loss": 0.6736, + "step": 33917 + }, + { + "epoch": 1.24, + "learning_rate": 7.59731936866005e-08, + "loss": 0.6947, + "step": 33918 + }, + { + "epoch": 1.24, + "learning_rate": 7.596953162192844e-08, + "loss": 0.7023, + "step": 33919 + }, + { + "epoch": 1.24, + "learning_rate": 7.596586955725638e-08, + "loss": 0.6995, + "step": 33920 + }, + { + "epoch": 1.24, + "learning_rate": 7.596220749258432e-08, + "loss": 0.7038, + "step": 33921 + }, + { + "epoch": 1.24, + "learning_rate": 7.595854542791226e-08, + "loss": 0.6998, + "step": 33922 + }, + { + "epoch": 1.24, + "learning_rate": 7.59548833632402e-08, + "loss": 0.6897, + "step": 33923 + }, + { + "epoch": 1.24, + "learning_rate": 7.595122129856813e-08, + "loss": 0.6908, + "step": 33924 + }, + { + "epoch": 1.24, + "learning_rate": 7.594755923389607e-08, + "loss": 0.6923, + "step": 33925 + }, + { + "epoch": 1.24, + "learning_rate": 7.5943897169224e-08, + "loss": 0.6951, + "step": 33926 + }, + { + "epoch": 1.24, + "learning_rate": 7.594023510455195e-08, + "loss": 0.6808, + "step": 33927 + }, + { + "epoch": 1.24, + "learning_rate": 7.593657303987989e-08, + "loss": 0.6987, + "step": 33928 + }, + { + "epoch": 1.24, + "learning_rate": 7.593291097520783e-08, + "loss": 0.6994, + "step": 33929 + }, + { + "epoch": 1.24, + "learning_rate": 7.592924891053577e-08, + "loss": 0.69, + "step": 33930 + }, + { + "epoch": 1.24, + "learning_rate": 7.592558684586369e-08, + "loss": 0.7017, + "step": 33931 + }, + { + "epoch": 1.24, + "learning_rate": 7.592192478119163e-08, + "loss": 0.7014, + "step": 33932 + }, + { + "epoch": 1.24, + "learning_rate": 7.591826271651956e-08, + "loss": 0.6956, + "step": 33933 + }, + { + "epoch": 1.24, + "learning_rate": 7.59146006518475e-08, + "loss": 0.6885, + "step": 33934 + }, + { + "epoch": 1.24, + "learning_rate": 7.591093858717544e-08, + "loss": 0.6829, + "step": 33935 + }, + { + "epoch": 1.24, + "learning_rate": 7.590727652250338e-08, + "loss": 0.6955, + "step": 33936 + }, + { + "epoch": 1.24, + "learning_rate": 7.590361445783132e-08, + "loss": 0.6956, + "step": 33937 + }, + { + "epoch": 1.24, + "learning_rate": 7.589995239315926e-08, + "loss": 0.6856, + "step": 33938 + }, + { + "epoch": 1.24, + "learning_rate": 7.58962903284872e-08, + "loss": 0.7023, + "step": 33939 + }, + { + "epoch": 1.24, + "learning_rate": 7.589262826381514e-08, + "loss": 0.6942, + "step": 33940 + }, + { + "epoch": 1.24, + "learning_rate": 7.588896619914308e-08, + "loss": 0.6989, + "step": 33941 + }, + { + "epoch": 1.24, + "learning_rate": 7.5885304134471e-08, + "loss": 0.6998, + "step": 33942 + }, + { + "epoch": 1.24, + "learning_rate": 7.588164206979895e-08, + "loss": 0.6872, + "step": 33943 + }, + { + "epoch": 1.24, + "learning_rate": 7.587798000512689e-08, + "loss": 0.6904, + "step": 33944 + }, + { + "epoch": 1.24, + "learning_rate": 7.587431794045483e-08, + "loss": 0.6989, + "step": 33945 + }, + { + "epoch": 1.24, + "learning_rate": 7.587065587578276e-08, + "loss": 0.6863, + "step": 33946 + }, + { + "epoch": 1.24, + "learning_rate": 7.58669938111107e-08, + "loss": 0.692, + "step": 33947 + }, + { + "epoch": 1.24, + "learning_rate": 7.586333174643864e-08, + "loss": 0.7037, + "step": 33948 + }, + { + "epoch": 1.24, + "learning_rate": 7.585966968176658e-08, + "loss": 0.6814, + "step": 33949 + }, + { + "epoch": 1.24, + "learning_rate": 7.585600761709452e-08, + "loss": 0.6978, + "step": 33950 + }, + { + "epoch": 1.24, + "learning_rate": 7.585234555242245e-08, + "loss": 0.6902, + "step": 33951 + }, + { + "epoch": 1.24, + "learning_rate": 7.584868348775039e-08, + "loss": 0.6867, + "step": 33952 + }, + { + "epoch": 1.24, + "learning_rate": 7.584502142307833e-08, + "loss": 0.6992, + "step": 33953 + }, + { + "epoch": 1.24, + "learning_rate": 7.584135935840627e-08, + "loss": 0.6898, + "step": 33954 + }, + { + "epoch": 1.24, + "learning_rate": 7.58376972937342e-08, + "loss": 0.6978, + "step": 33955 + }, + { + "epoch": 1.24, + "learning_rate": 7.583403522906214e-08, + "loss": 0.687, + "step": 33956 + }, + { + "epoch": 1.24, + "learning_rate": 7.583037316439008e-08, + "loss": 0.6844, + "step": 33957 + }, + { + "epoch": 1.24, + "learning_rate": 7.582671109971802e-08, + "loss": 0.6934, + "step": 33958 + }, + { + "epoch": 1.24, + "learning_rate": 7.582304903504596e-08, + "loss": 0.6838, + "step": 33959 + }, + { + "epoch": 1.24, + "learning_rate": 7.581938697037388e-08, + "loss": 0.7047, + "step": 33960 + }, + { + "epoch": 1.24, + "learning_rate": 7.581572490570182e-08, + "loss": 0.6746, + "step": 33961 + }, + { + "epoch": 1.24, + "learning_rate": 7.581206284102976e-08, + "loss": 0.6836, + "step": 33962 + }, + { + "epoch": 1.24, + "learning_rate": 7.58084007763577e-08, + "loss": 0.7095, + "step": 33963 + }, + { + "epoch": 1.24, + "learning_rate": 7.580473871168564e-08, + "loss": 0.6944, + "step": 33964 + }, + { + "epoch": 1.24, + "learning_rate": 7.580107664701358e-08, + "loss": 0.6881, + "step": 33965 + }, + { + "epoch": 1.24, + "learning_rate": 7.579741458234152e-08, + "loss": 0.6953, + "step": 33966 + }, + { + "epoch": 1.24, + "learning_rate": 7.579375251766946e-08, + "loss": 0.6853, + "step": 33967 + }, + { + "epoch": 1.24, + "learning_rate": 7.57900904529974e-08, + "loss": 0.6858, + "step": 33968 + }, + { + "epoch": 1.24, + "learning_rate": 7.578642838832533e-08, + "loss": 0.6883, + "step": 33969 + }, + { + "epoch": 1.24, + "learning_rate": 7.578276632365327e-08, + "loss": 0.6927, + "step": 33970 + }, + { + "epoch": 1.24, + "learning_rate": 7.577910425898121e-08, + "loss": 0.7026, + "step": 33971 + }, + { + "epoch": 1.24, + "learning_rate": 7.577544219430915e-08, + "loss": 0.7014, + "step": 33972 + }, + { + "epoch": 1.24, + "learning_rate": 7.577178012963709e-08, + "loss": 0.691, + "step": 33973 + }, + { + "epoch": 1.24, + "learning_rate": 7.576811806496503e-08, + "loss": 0.6902, + "step": 33974 + }, + { + "epoch": 1.24, + "learning_rate": 7.576445600029297e-08, + "loss": 0.6862, + "step": 33975 + }, + { + "epoch": 1.24, + "learning_rate": 7.576079393562091e-08, + "loss": 0.6885, + "step": 33976 + }, + { + "epoch": 1.24, + "learning_rate": 7.575713187094884e-08, + "loss": 0.6975, + "step": 33977 + }, + { + "epoch": 1.24, + "learning_rate": 7.575346980627678e-08, + "loss": 0.6982, + "step": 33978 + }, + { + "epoch": 1.24, + "learning_rate": 7.574980774160472e-08, + "loss": 0.6968, + "step": 33979 + }, + { + "epoch": 1.24, + "learning_rate": 7.574614567693264e-08, + "loss": 0.6875, + "step": 33980 + }, + { + "epoch": 1.24, + "learning_rate": 7.574248361226058e-08, + "loss": 0.6859, + "step": 33981 + }, + { + "epoch": 1.24, + "learning_rate": 7.573882154758852e-08, + "loss": 0.7067, + "step": 33982 + }, + { + "epoch": 1.24, + "learning_rate": 7.573515948291646e-08, + "loss": 0.6836, + "step": 33983 + }, + { + "epoch": 1.24, + "learning_rate": 7.57314974182444e-08, + "loss": 0.6998, + "step": 33984 + }, + { + "epoch": 1.24, + "learning_rate": 7.572783535357234e-08, + "loss": 0.6942, + "step": 33985 + }, + { + "epoch": 1.24, + "learning_rate": 7.572417328890028e-08, + "loss": 0.6989, + "step": 33986 + }, + { + "epoch": 1.24, + "learning_rate": 7.572051122422821e-08, + "loss": 0.6936, + "step": 33987 + }, + { + "epoch": 1.24, + "learning_rate": 7.571684915955615e-08, + "loss": 0.6829, + "step": 33988 + }, + { + "epoch": 1.24, + "learning_rate": 7.571318709488409e-08, + "loss": 0.6945, + "step": 33989 + }, + { + "epoch": 1.24, + "learning_rate": 7.570952503021203e-08, + "loss": 0.6941, + "step": 33990 + }, + { + "epoch": 1.24, + "learning_rate": 7.570586296553997e-08, + "loss": 0.6829, + "step": 33991 + }, + { + "epoch": 1.24, + "learning_rate": 7.570220090086791e-08, + "loss": 0.7252, + "step": 33992 + }, + { + "epoch": 1.24, + "learning_rate": 7.569853883619585e-08, + "loss": 0.6973, + "step": 33993 + }, + { + "epoch": 1.24, + "learning_rate": 7.569487677152379e-08, + "loss": 0.6868, + "step": 33994 + }, + { + "epoch": 1.24, + "learning_rate": 7.569121470685173e-08, + "loss": 0.6798, + "step": 33995 + }, + { + "epoch": 1.24, + "learning_rate": 7.568755264217966e-08, + "loss": 0.6873, + "step": 33996 + }, + { + "epoch": 1.24, + "learning_rate": 7.56838905775076e-08, + "loss": 0.6874, + "step": 33997 + }, + { + "epoch": 1.24, + "learning_rate": 7.568022851283554e-08, + "loss": 0.6863, + "step": 33998 + }, + { + "epoch": 1.24, + "learning_rate": 7.567656644816348e-08, + "loss": 0.7085, + "step": 33999 + }, + { + "epoch": 1.24, + "learning_rate": 7.567290438349142e-08, + "loss": 0.7007, + "step": 34000 + }, + { + "epoch": 1.24, + "learning_rate": 7.566924231881934e-08, + "loss": 0.7007, + "step": 34001 + }, + { + "epoch": 1.24, + "learning_rate": 7.566558025414728e-08, + "loss": 0.7005, + "step": 34002 + }, + { + "epoch": 1.24, + "learning_rate": 7.566191818947522e-08, + "loss": 0.6982, + "step": 34003 + }, + { + "epoch": 1.24, + "learning_rate": 7.565825612480316e-08, + "loss": 0.7069, + "step": 34004 + }, + { + "epoch": 1.24, + "learning_rate": 7.565459406013109e-08, + "loss": 0.6931, + "step": 34005 + }, + { + "epoch": 1.24, + "learning_rate": 7.565093199545903e-08, + "loss": 0.6829, + "step": 34006 + }, + { + "epoch": 1.24, + "learning_rate": 7.564726993078697e-08, + "loss": 0.6927, + "step": 34007 + }, + { + "epoch": 1.24, + "learning_rate": 7.564360786611491e-08, + "loss": 0.6996, + "step": 34008 + }, + { + "epoch": 1.24, + "learning_rate": 7.563994580144285e-08, + "loss": 0.7012, + "step": 34009 + }, + { + "epoch": 1.24, + "learning_rate": 7.563628373677079e-08, + "loss": 0.6894, + "step": 34010 + }, + { + "epoch": 1.24, + "learning_rate": 7.563262167209873e-08, + "loss": 0.685, + "step": 34011 + }, + { + "epoch": 1.24, + "learning_rate": 7.562895960742667e-08, + "loss": 0.6806, + "step": 34012 + }, + { + "epoch": 1.24, + "learning_rate": 7.562529754275461e-08, + "loss": 0.673, + "step": 34013 + }, + { + "epoch": 1.24, + "learning_rate": 7.562163547808253e-08, + "loss": 0.6816, + "step": 34014 + }, + { + "epoch": 1.24, + "learning_rate": 7.561797341341047e-08, + "loss": 0.714, + "step": 34015 + }, + { + "epoch": 1.24, + "learning_rate": 7.561431134873841e-08, + "loss": 0.6977, + "step": 34016 + }, + { + "epoch": 1.24, + "learning_rate": 7.561064928406635e-08, + "loss": 0.684, + "step": 34017 + }, + { + "epoch": 1.24, + "learning_rate": 7.56069872193943e-08, + "loss": 0.6831, + "step": 34018 + }, + { + "epoch": 1.24, + "learning_rate": 7.560332515472223e-08, + "loss": 0.6897, + "step": 34019 + }, + { + "epoch": 1.24, + "learning_rate": 7.559966309005017e-08, + "loss": 0.6961, + "step": 34020 + }, + { + "epoch": 1.24, + "learning_rate": 7.559600102537811e-08, + "loss": 0.6826, + "step": 34021 + }, + { + "epoch": 1.24, + "learning_rate": 7.559233896070605e-08, + "loss": 0.6954, + "step": 34022 + }, + { + "epoch": 1.24, + "learning_rate": 7.558867689603398e-08, + "loss": 0.6934, + "step": 34023 + }, + { + "epoch": 1.24, + "learning_rate": 7.558501483136192e-08, + "loss": 0.6959, + "step": 34024 + }, + { + "epoch": 1.24, + "learning_rate": 7.558135276668986e-08, + "loss": 0.699, + "step": 34025 + }, + { + "epoch": 1.24, + "learning_rate": 7.557769070201779e-08, + "loss": 0.6897, + "step": 34026 + }, + { + "epoch": 1.24, + "learning_rate": 7.557402863734573e-08, + "loss": 0.7033, + "step": 34027 + }, + { + "epoch": 1.24, + "learning_rate": 7.557036657267367e-08, + "loss": 0.6858, + "step": 34028 + }, + { + "epoch": 1.25, + "learning_rate": 7.556670450800161e-08, + "loss": 0.7031, + "step": 34029 + }, + { + "epoch": 1.25, + "learning_rate": 7.556304244332955e-08, + "loss": 0.7065, + "step": 34030 + }, + { + "epoch": 1.25, + "learning_rate": 7.555938037865749e-08, + "loss": 0.6973, + "step": 34031 + }, + { + "epoch": 1.25, + "learning_rate": 7.555571831398541e-08, + "loss": 0.7019, + "step": 34032 + }, + { + "epoch": 1.25, + "learning_rate": 7.555205624931335e-08, + "loss": 0.6883, + "step": 34033 + }, + { + "epoch": 1.25, + "learning_rate": 7.55483941846413e-08, + "loss": 0.6853, + "step": 34034 + }, + { + "epoch": 1.25, + "learning_rate": 7.554473211996923e-08, + "loss": 0.6893, + "step": 34035 + }, + { + "epoch": 1.25, + "learning_rate": 7.554107005529717e-08, + "loss": 0.7008, + "step": 34036 + }, + { + "epoch": 1.25, + "learning_rate": 7.553740799062511e-08, + "loss": 0.7062, + "step": 34037 + }, + { + "epoch": 1.25, + "learning_rate": 7.553374592595305e-08, + "loss": 0.6935, + "step": 34038 + }, + { + "epoch": 1.25, + "learning_rate": 7.553008386128099e-08, + "loss": 0.7088, + "step": 34039 + }, + { + "epoch": 1.25, + "learning_rate": 7.552642179660893e-08, + "loss": 0.7143, + "step": 34040 + }, + { + "epoch": 1.25, + "learning_rate": 7.552275973193686e-08, + "loss": 0.6872, + "step": 34041 + }, + { + "epoch": 1.25, + "learning_rate": 7.55190976672648e-08, + "loss": 0.7031, + "step": 34042 + }, + { + "epoch": 1.25, + "learning_rate": 7.551543560259274e-08, + "loss": 0.6936, + "step": 34043 + }, + { + "epoch": 1.25, + "learning_rate": 7.551177353792068e-08, + "loss": 0.7002, + "step": 34044 + }, + { + "epoch": 1.25, + "learning_rate": 7.550811147324862e-08, + "loss": 0.6914, + "step": 34045 + }, + { + "epoch": 1.25, + "learning_rate": 7.550444940857656e-08, + "loss": 0.6872, + "step": 34046 + }, + { + "epoch": 1.25, + "learning_rate": 7.55007873439045e-08, + "loss": 0.6925, + "step": 34047 + }, + { + "epoch": 1.25, + "learning_rate": 7.549712527923243e-08, + "loss": 0.6861, + "step": 34048 + }, + { + "epoch": 1.25, + "learning_rate": 7.549346321456037e-08, + "loss": 0.691, + "step": 34049 + }, + { + "epoch": 1.25, + "learning_rate": 7.54898011498883e-08, + "loss": 0.7056, + "step": 34050 + }, + { + "epoch": 1.25, + "learning_rate": 7.548613908521623e-08, + "loss": 0.6981, + "step": 34051 + }, + { + "epoch": 1.25, + "learning_rate": 7.548247702054417e-08, + "loss": 0.7016, + "step": 34052 + }, + { + "epoch": 1.25, + "learning_rate": 7.547881495587211e-08, + "loss": 0.698, + "step": 34053 + }, + { + "epoch": 1.25, + "learning_rate": 7.547515289120005e-08, + "loss": 0.7, + "step": 34054 + }, + { + "epoch": 1.25, + "learning_rate": 7.547149082652799e-08, + "loss": 0.6965, + "step": 34055 + }, + { + "epoch": 1.25, + "learning_rate": 7.546782876185593e-08, + "loss": 0.6923, + "step": 34056 + }, + { + "epoch": 1.25, + "learning_rate": 7.546416669718387e-08, + "loss": 0.6893, + "step": 34057 + }, + { + "epoch": 1.25, + "learning_rate": 7.546050463251181e-08, + "loss": 0.6947, + "step": 34058 + }, + { + "epoch": 1.25, + "learning_rate": 7.545684256783974e-08, + "loss": 0.6869, + "step": 34059 + }, + { + "epoch": 1.25, + "learning_rate": 7.545318050316768e-08, + "loss": 0.6872, + "step": 34060 + }, + { + "epoch": 1.25, + "learning_rate": 7.544951843849562e-08, + "loss": 0.6816, + "step": 34061 + }, + { + "epoch": 1.25, + "learning_rate": 7.544585637382356e-08, + "loss": 0.6926, + "step": 34062 + }, + { + "epoch": 1.25, + "learning_rate": 7.54421943091515e-08, + "loss": 0.6878, + "step": 34063 + }, + { + "epoch": 1.25, + "learning_rate": 7.543853224447944e-08, + "loss": 0.6936, + "step": 34064 + }, + { + "epoch": 1.25, + "learning_rate": 7.543487017980738e-08, + "loss": 0.688, + "step": 34065 + }, + { + "epoch": 1.25, + "learning_rate": 7.543120811513532e-08, + "loss": 0.678, + "step": 34066 + }, + { + "epoch": 1.25, + "learning_rate": 7.542754605046326e-08, + "loss": 0.6962, + "step": 34067 + }, + { + "epoch": 1.25, + "learning_rate": 7.542388398579118e-08, + "loss": 0.6952, + "step": 34068 + }, + { + "epoch": 1.25, + "learning_rate": 7.542022192111912e-08, + "loss": 0.6751, + "step": 34069 + }, + { + "epoch": 1.25, + "learning_rate": 7.541655985644706e-08, + "loss": 0.7054, + "step": 34070 + }, + { + "epoch": 1.25, + "learning_rate": 7.5412897791775e-08, + "loss": 0.6878, + "step": 34071 + }, + { + "epoch": 1.25, + "learning_rate": 7.540923572710293e-08, + "loss": 0.6919, + "step": 34072 + }, + { + "epoch": 1.25, + "learning_rate": 7.540557366243087e-08, + "loss": 0.685, + "step": 34073 + }, + { + "epoch": 1.25, + "learning_rate": 7.540191159775881e-08, + "loss": 0.7014, + "step": 34074 + }, + { + "epoch": 1.25, + "learning_rate": 7.539824953308675e-08, + "loss": 0.685, + "step": 34075 + }, + { + "epoch": 1.25, + "learning_rate": 7.539458746841469e-08, + "loss": 0.6829, + "step": 34076 + }, + { + "epoch": 1.25, + "learning_rate": 7.539092540374262e-08, + "loss": 0.6961, + "step": 34077 + }, + { + "epoch": 1.25, + "learning_rate": 7.538726333907056e-08, + "loss": 0.6886, + "step": 34078 + }, + { + "epoch": 1.25, + "learning_rate": 7.53836012743985e-08, + "loss": 0.6959, + "step": 34079 + }, + { + "epoch": 1.25, + "learning_rate": 7.537993920972644e-08, + "loss": 0.6807, + "step": 34080 + }, + { + "epoch": 1.25, + "learning_rate": 7.537627714505438e-08, + "loss": 0.6974, + "step": 34081 + }, + { + "epoch": 1.25, + "learning_rate": 7.537261508038232e-08, + "loss": 0.6864, + "step": 34082 + }, + { + "epoch": 1.25, + "learning_rate": 7.536895301571026e-08, + "loss": 0.6807, + "step": 34083 + }, + { + "epoch": 1.25, + "learning_rate": 7.53652909510382e-08, + "loss": 0.6985, + "step": 34084 + }, + { + "epoch": 1.25, + "learning_rate": 7.536162888636614e-08, + "loss": 0.6861, + "step": 34085 + }, + { + "epoch": 1.25, + "learning_rate": 7.535796682169406e-08, + "loss": 0.7035, + "step": 34086 + }, + { + "epoch": 1.25, + "learning_rate": 7.5354304757022e-08, + "loss": 0.6805, + "step": 34087 + }, + { + "epoch": 1.25, + "learning_rate": 7.535064269234994e-08, + "loss": 0.7018, + "step": 34088 + }, + { + "epoch": 1.25, + "learning_rate": 7.534698062767788e-08, + "loss": 0.6901, + "step": 34089 + }, + { + "epoch": 1.25, + "learning_rate": 7.534331856300582e-08, + "loss": 0.6808, + "step": 34090 + }, + { + "epoch": 1.25, + "learning_rate": 7.533965649833376e-08, + "loss": 0.6864, + "step": 34091 + }, + { + "epoch": 1.25, + "learning_rate": 7.53359944336617e-08, + "loss": 0.6887, + "step": 34092 + }, + { + "epoch": 1.25, + "learning_rate": 7.533233236898964e-08, + "loss": 0.693, + "step": 34093 + }, + { + "epoch": 1.25, + "learning_rate": 7.532867030431757e-08, + "loss": 0.672, + "step": 34094 + }, + { + "epoch": 1.25, + "learning_rate": 7.532500823964551e-08, + "loss": 0.687, + "step": 34095 + }, + { + "epoch": 1.25, + "learning_rate": 7.532134617497345e-08, + "loss": 0.6991, + "step": 34096 + }, + { + "epoch": 1.25, + "learning_rate": 7.531768411030138e-08, + "loss": 0.6955, + "step": 34097 + }, + { + "epoch": 1.25, + "learning_rate": 7.531402204562932e-08, + "loss": 0.698, + "step": 34098 + }, + { + "epoch": 1.25, + "learning_rate": 7.531035998095726e-08, + "loss": 0.7008, + "step": 34099 + }, + { + "epoch": 1.25, + "learning_rate": 7.53066979162852e-08, + "loss": 0.6834, + "step": 34100 + }, + { + "epoch": 1.25, + "learning_rate": 7.530303585161314e-08, + "loss": 0.7005, + "step": 34101 + }, + { + "epoch": 1.25, + "learning_rate": 7.529937378694108e-08, + "loss": 0.6777, + "step": 34102 + }, + { + "epoch": 1.25, + "learning_rate": 7.5295711722269e-08, + "loss": 0.6954, + "step": 34103 + }, + { + "epoch": 1.25, + "learning_rate": 7.529204965759694e-08, + "loss": 0.7141, + "step": 34104 + }, + { + "epoch": 1.25, + "learning_rate": 7.528838759292488e-08, + "loss": 0.6904, + "step": 34105 + }, + { + "epoch": 1.25, + "learning_rate": 7.528472552825282e-08, + "loss": 0.6844, + "step": 34106 + }, + { + "epoch": 1.25, + "learning_rate": 7.528106346358076e-08, + "loss": 0.6844, + "step": 34107 + }, + { + "epoch": 1.25, + "learning_rate": 7.52774013989087e-08, + "loss": 0.6859, + "step": 34108 + }, + { + "epoch": 1.25, + "learning_rate": 7.527373933423664e-08, + "loss": 0.7085, + "step": 34109 + }, + { + "epoch": 1.25, + "learning_rate": 7.527007726956458e-08, + "loss": 0.6892, + "step": 34110 + }, + { + "epoch": 1.25, + "learning_rate": 7.526641520489252e-08, + "loss": 0.6975, + "step": 34111 + }, + { + "epoch": 1.25, + "learning_rate": 7.526275314022045e-08, + "loss": 0.7009, + "step": 34112 + }, + { + "epoch": 1.25, + "learning_rate": 7.525909107554839e-08, + "loss": 0.7031, + "step": 34113 + }, + { + "epoch": 1.25, + "learning_rate": 7.525542901087633e-08, + "loss": 0.6999, + "step": 34114 + }, + { + "epoch": 1.25, + "learning_rate": 7.525176694620427e-08, + "loss": 0.6954, + "step": 34115 + }, + { + "epoch": 1.25, + "learning_rate": 7.524810488153221e-08, + "loss": 0.6941, + "step": 34116 + }, + { + "epoch": 1.25, + "learning_rate": 7.524444281686015e-08, + "loss": 0.7045, + "step": 34117 + }, + { + "epoch": 1.25, + "learning_rate": 7.524078075218809e-08, + "loss": 0.6923, + "step": 34118 + }, + { + "epoch": 1.25, + "learning_rate": 7.523711868751602e-08, + "loss": 0.7012, + "step": 34119 + }, + { + "epoch": 1.25, + "learning_rate": 7.523345662284396e-08, + "loss": 0.7038, + "step": 34120 + }, + { + "epoch": 1.25, + "learning_rate": 7.522979455817188e-08, + "loss": 0.6869, + "step": 34121 + }, + { + "epoch": 1.25, + "learning_rate": 7.522613249349982e-08, + "loss": 0.688, + "step": 34122 + }, + { + "epoch": 1.25, + "learning_rate": 7.522247042882776e-08, + "loss": 0.6867, + "step": 34123 + }, + { + "epoch": 1.25, + "learning_rate": 7.52188083641557e-08, + "loss": 0.6961, + "step": 34124 + }, + { + "epoch": 1.25, + "learning_rate": 7.521514629948364e-08, + "loss": 0.6878, + "step": 34125 + }, + { + "epoch": 1.25, + "learning_rate": 7.521148423481158e-08, + "loss": 0.6945, + "step": 34126 + }, + { + "epoch": 1.25, + "learning_rate": 7.520782217013952e-08, + "loss": 0.6724, + "step": 34127 + }, + { + "epoch": 1.25, + "learning_rate": 7.520416010546746e-08, + "loss": 0.6903, + "step": 34128 + }, + { + "epoch": 1.25, + "learning_rate": 7.52004980407954e-08, + "loss": 0.6909, + "step": 34129 + }, + { + "epoch": 1.25, + "learning_rate": 7.519683597612333e-08, + "loss": 0.6808, + "step": 34130 + }, + { + "epoch": 1.25, + "learning_rate": 7.519317391145127e-08, + "loss": 0.6925, + "step": 34131 + }, + { + "epoch": 1.25, + "learning_rate": 7.518951184677921e-08, + "loss": 0.699, + "step": 34132 + }, + { + "epoch": 1.25, + "learning_rate": 7.518584978210715e-08, + "loss": 0.7021, + "step": 34133 + }, + { + "epoch": 1.25, + "learning_rate": 7.518218771743509e-08, + "loss": 0.6873, + "step": 34134 + }, + { + "epoch": 1.25, + "learning_rate": 7.517852565276303e-08, + "loss": 0.6741, + "step": 34135 + }, + { + "epoch": 1.25, + "learning_rate": 7.517486358809097e-08, + "loss": 0.6817, + "step": 34136 + }, + { + "epoch": 1.25, + "learning_rate": 7.517120152341891e-08, + "loss": 0.6965, + "step": 34137 + }, + { + "epoch": 1.25, + "learning_rate": 7.516753945874685e-08, + "loss": 0.6975, + "step": 34138 + }, + { + "epoch": 1.25, + "learning_rate": 7.516387739407477e-08, + "loss": 0.7011, + "step": 34139 + }, + { + "epoch": 1.25, + "learning_rate": 7.516021532940271e-08, + "loss": 0.6914, + "step": 34140 + }, + { + "epoch": 1.25, + "learning_rate": 7.515655326473065e-08, + "loss": 0.6806, + "step": 34141 + }, + { + "epoch": 1.25, + "learning_rate": 7.51528912000586e-08, + "loss": 0.6912, + "step": 34142 + }, + { + "epoch": 1.25, + "learning_rate": 7.514922913538652e-08, + "loss": 0.6956, + "step": 34143 + }, + { + "epoch": 1.25, + "learning_rate": 7.514556707071446e-08, + "loss": 0.6863, + "step": 34144 + }, + { + "epoch": 1.25, + "learning_rate": 7.51419050060424e-08, + "loss": 0.6986, + "step": 34145 + }, + { + "epoch": 1.25, + "learning_rate": 7.513824294137034e-08, + "loss": 0.6993, + "step": 34146 + }, + { + "epoch": 1.25, + "learning_rate": 7.513458087669828e-08, + "loss": 0.6932, + "step": 34147 + }, + { + "epoch": 1.25, + "learning_rate": 7.513091881202621e-08, + "loss": 0.6848, + "step": 34148 + }, + { + "epoch": 1.25, + "learning_rate": 7.512725674735415e-08, + "loss": 0.6942, + "step": 34149 + }, + { + "epoch": 1.25, + "learning_rate": 7.512359468268209e-08, + "loss": 0.7036, + "step": 34150 + }, + { + "epoch": 1.25, + "learning_rate": 7.511993261801003e-08, + "loss": 0.7002, + "step": 34151 + }, + { + "epoch": 1.25, + "learning_rate": 7.511627055333797e-08, + "loss": 0.685, + "step": 34152 + }, + { + "epoch": 1.25, + "learning_rate": 7.511260848866591e-08, + "loss": 0.7064, + "step": 34153 + }, + { + "epoch": 1.25, + "learning_rate": 7.510894642399385e-08, + "loss": 0.7105, + "step": 34154 + }, + { + "epoch": 1.25, + "learning_rate": 7.510528435932179e-08, + "loss": 0.6945, + "step": 34155 + }, + { + "epoch": 1.25, + "learning_rate": 7.510162229464973e-08, + "loss": 0.6976, + "step": 34156 + }, + { + "epoch": 1.25, + "learning_rate": 7.509796022997765e-08, + "loss": 0.7011, + "step": 34157 + }, + { + "epoch": 1.25, + "learning_rate": 7.509429816530559e-08, + "loss": 0.6889, + "step": 34158 + }, + { + "epoch": 1.25, + "learning_rate": 7.509063610063353e-08, + "loss": 0.6898, + "step": 34159 + }, + { + "epoch": 1.25, + "learning_rate": 7.508697403596147e-08, + "loss": 0.6967, + "step": 34160 + }, + { + "epoch": 1.25, + "learning_rate": 7.508331197128941e-08, + "loss": 0.6958, + "step": 34161 + }, + { + "epoch": 1.25, + "learning_rate": 7.507964990661735e-08, + "loss": 0.6894, + "step": 34162 + }, + { + "epoch": 1.25, + "learning_rate": 7.507598784194529e-08, + "loss": 0.6881, + "step": 34163 + }, + { + "epoch": 1.25, + "learning_rate": 7.507232577727323e-08, + "loss": 0.6959, + "step": 34164 + }, + { + "epoch": 1.25, + "learning_rate": 7.506866371260116e-08, + "loss": 0.6799, + "step": 34165 + }, + { + "epoch": 1.25, + "learning_rate": 7.50650016479291e-08, + "loss": 0.7061, + "step": 34166 + }, + { + "epoch": 1.25, + "learning_rate": 7.506133958325704e-08, + "loss": 0.6811, + "step": 34167 + }, + { + "epoch": 1.25, + "learning_rate": 7.505767751858497e-08, + "loss": 0.698, + "step": 34168 + }, + { + "epoch": 1.25, + "learning_rate": 7.50540154539129e-08, + "loss": 0.6985, + "step": 34169 + }, + { + "epoch": 1.25, + "learning_rate": 7.505035338924085e-08, + "loss": 0.6879, + "step": 34170 + }, + { + "epoch": 1.25, + "learning_rate": 7.504669132456879e-08, + "loss": 0.6958, + "step": 34171 + }, + { + "epoch": 1.25, + "learning_rate": 7.504302925989673e-08, + "loss": 0.6908, + "step": 34172 + }, + { + "epoch": 1.25, + "learning_rate": 7.503936719522467e-08, + "loss": 0.6934, + "step": 34173 + }, + { + "epoch": 1.25, + "learning_rate": 7.50357051305526e-08, + "loss": 0.6954, + "step": 34174 + }, + { + "epoch": 1.25, + "learning_rate": 7.503204306588053e-08, + "loss": 0.6963, + "step": 34175 + }, + { + "epoch": 1.25, + "learning_rate": 7.502838100120847e-08, + "loss": 0.699, + "step": 34176 + }, + { + "epoch": 1.25, + "learning_rate": 7.502471893653641e-08, + "loss": 0.6862, + "step": 34177 + }, + { + "epoch": 1.25, + "learning_rate": 7.502105687186435e-08, + "loss": 0.6797, + "step": 34178 + }, + { + "epoch": 1.25, + "learning_rate": 7.501739480719229e-08, + "loss": 0.6841, + "step": 34179 + }, + { + "epoch": 1.25, + "learning_rate": 7.501373274252023e-08, + "loss": 0.6938, + "step": 34180 + }, + { + "epoch": 1.25, + "learning_rate": 7.501007067784817e-08, + "loss": 0.6956, + "step": 34181 + }, + { + "epoch": 1.25, + "learning_rate": 7.500640861317611e-08, + "loss": 0.6917, + "step": 34182 + }, + { + "epoch": 1.25, + "learning_rate": 7.500274654850405e-08, + "loss": 0.7004, + "step": 34183 + }, + { + "epoch": 1.25, + "learning_rate": 7.499908448383198e-08, + "loss": 0.7009, + "step": 34184 + }, + { + "epoch": 1.25, + "learning_rate": 7.499542241915992e-08, + "loss": 0.694, + "step": 34185 + }, + { + "epoch": 1.25, + "learning_rate": 7.499176035448786e-08, + "loss": 0.6807, + "step": 34186 + }, + { + "epoch": 1.25, + "learning_rate": 7.49880982898158e-08, + "loss": 0.687, + "step": 34187 + }, + { + "epoch": 1.25, + "learning_rate": 7.498443622514374e-08, + "loss": 0.6905, + "step": 34188 + }, + { + "epoch": 1.25, + "learning_rate": 7.498077416047168e-08, + "loss": 0.6843, + "step": 34189 + }, + { + "epoch": 1.25, + "learning_rate": 7.49771120957996e-08, + "loss": 0.6979, + "step": 34190 + }, + { + "epoch": 1.25, + "learning_rate": 7.497345003112754e-08, + "loss": 0.7029, + "step": 34191 + }, + { + "epoch": 1.25, + "learning_rate": 7.496978796645548e-08, + "loss": 0.6954, + "step": 34192 + }, + { + "epoch": 1.25, + "learning_rate": 7.496612590178341e-08, + "loss": 0.6949, + "step": 34193 + }, + { + "epoch": 1.25, + "learning_rate": 7.496246383711135e-08, + "loss": 0.6963, + "step": 34194 + }, + { + "epoch": 1.25, + "learning_rate": 7.495880177243929e-08, + "loss": 0.6993, + "step": 34195 + }, + { + "epoch": 1.25, + "learning_rate": 7.495513970776723e-08, + "loss": 0.6973, + "step": 34196 + }, + { + "epoch": 1.25, + "learning_rate": 7.495147764309517e-08, + "loss": 0.691, + "step": 34197 + }, + { + "epoch": 1.25, + "learning_rate": 7.494781557842311e-08, + "loss": 0.7051, + "step": 34198 + }, + { + "epoch": 1.25, + "learning_rate": 7.494415351375105e-08, + "loss": 0.7045, + "step": 34199 + }, + { + "epoch": 1.25, + "learning_rate": 7.494049144907899e-08, + "loss": 0.6986, + "step": 34200 + }, + { + "epoch": 1.25, + "learning_rate": 7.493682938440693e-08, + "loss": 0.6893, + "step": 34201 + }, + { + "epoch": 1.25, + "learning_rate": 7.493316731973486e-08, + "loss": 0.6971, + "step": 34202 + }, + { + "epoch": 1.25, + "learning_rate": 7.49295052550628e-08, + "loss": 0.6798, + "step": 34203 + }, + { + "epoch": 1.25, + "learning_rate": 7.492584319039074e-08, + "loss": 0.6995, + "step": 34204 + }, + { + "epoch": 1.25, + "learning_rate": 7.492218112571868e-08, + "loss": 0.6932, + "step": 34205 + }, + { + "epoch": 1.25, + "learning_rate": 7.491851906104662e-08, + "loss": 0.6958, + "step": 34206 + }, + { + "epoch": 1.25, + "learning_rate": 7.491485699637456e-08, + "loss": 0.7014, + "step": 34207 + }, + { + "epoch": 1.25, + "learning_rate": 7.49111949317025e-08, + "loss": 0.6893, + "step": 34208 + }, + { + "epoch": 1.25, + "learning_rate": 7.490753286703044e-08, + "loss": 0.7143, + "step": 34209 + }, + { + "epoch": 1.25, + "learning_rate": 7.490387080235838e-08, + "loss": 0.7026, + "step": 34210 + }, + { + "epoch": 1.25, + "learning_rate": 7.49002087376863e-08, + "loss": 0.6936, + "step": 34211 + }, + { + "epoch": 1.25, + "learning_rate": 7.489654667301424e-08, + "loss": 0.709, + "step": 34212 + }, + { + "epoch": 1.25, + "learning_rate": 7.489288460834218e-08, + "loss": 0.7021, + "step": 34213 + }, + { + "epoch": 1.25, + "learning_rate": 7.488922254367011e-08, + "loss": 0.7051, + "step": 34214 + }, + { + "epoch": 1.25, + "learning_rate": 7.488556047899805e-08, + "loss": 0.6908, + "step": 34215 + }, + { + "epoch": 1.25, + "learning_rate": 7.488189841432599e-08, + "loss": 0.6973, + "step": 34216 + }, + { + "epoch": 1.25, + "learning_rate": 7.487823634965393e-08, + "loss": 0.6894, + "step": 34217 + }, + { + "epoch": 1.25, + "learning_rate": 7.487457428498187e-08, + "loss": 0.686, + "step": 34218 + }, + { + "epoch": 1.25, + "learning_rate": 7.487091222030981e-08, + "loss": 0.6811, + "step": 34219 + }, + { + "epoch": 1.25, + "learning_rate": 7.486725015563774e-08, + "loss": 0.6842, + "step": 34220 + }, + { + "epoch": 1.25, + "learning_rate": 7.486358809096568e-08, + "loss": 0.6832, + "step": 34221 + }, + { + "epoch": 1.25, + "learning_rate": 7.485992602629362e-08, + "loss": 0.6979, + "step": 34222 + }, + { + "epoch": 1.25, + "learning_rate": 7.485626396162156e-08, + "loss": 0.6936, + "step": 34223 + }, + { + "epoch": 1.25, + "learning_rate": 7.48526018969495e-08, + "loss": 0.7033, + "step": 34224 + }, + { + "epoch": 1.25, + "learning_rate": 7.484893983227744e-08, + "loss": 0.6937, + "step": 34225 + }, + { + "epoch": 1.25, + "learning_rate": 7.484527776760538e-08, + "loss": 0.6958, + "step": 34226 + }, + { + "epoch": 1.25, + "learning_rate": 7.484161570293332e-08, + "loss": 0.6952, + "step": 34227 + }, + { + "epoch": 1.25, + "learning_rate": 7.483795363826126e-08, + "loss": 0.6825, + "step": 34228 + }, + { + "epoch": 1.25, + "learning_rate": 7.483429157358918e-08, + "loss": 0.6819, + "step": 34229 + }, + { + "epoch": 1.25, + "learning_rate": 7.483062950891712e-08, + "loss": 0.6987, + "step": 34230 + }, + { + "epoch": 1.25, + "learning_rate": 7.482696744424506e-08, + "loss": 0.6943, + "step": 34231 + }, + { + "epoch": 1.25, + "learning_rate": 7.4823305379573e-08, + "loss": 0.6899, + "step": 34232 + }, + { + "epoch": 1.25, + "learning_rate": 7.481964331490094e-08, + "loss": 0.703, + "step": 34233 + }, + { + "epoch": 1.25, + "learning_rate": 7.481598125022888e-08, + "loss": 0.6947, + "step": 34234 + }, + { + "epoch": 1.25, + "learning_rate": 7.481231918555682e-08, + "loss": 0.687, + "step": 34235 + }, + { + "epoch": 1.25, + "learning_rate": 7.480865712088475e-08, + "loss": 0.6891, + "step": 34236 + }, + { + "epoch": 1.25, + "learning_rate": 7.480499505621269e-08, + "loss": 0.701, + "step": 34237 + }, + { + "epoch": 1.25, + "learning_rate": 7.480133299154063e-08, + "loss": 0.6946, + "step": 34238 + }, + { + "epoch": 1.25, + "learning_rate": 7.479767092686856e-08, + "loss": 0.7021, + "step": 34239 + }, + { + "epoch": 1.25, + "learning_rate": 7.47940088621965e-08, + "loss": 0.7, + "step": 34240 + }, + { + "epoch": 1.25, + "learning_rate": 7.479034679752444e-08, + "loss": 0.708, + "step": 34241 + }, + { + "epoch": 1.25, + "learning_rate": 7.478668473285238e-08, + "loss": 0.688, + "step": 34242 + }, + { + "epoch": 1.25, + "learning_rate": 7.478302266818031e-08, + "loss": 0.6804, + "step": 34243 + }, + { + "epoch": 1.25, + "learning_rate": 7.477936060350825e-08, + "loss": 0.6769, + "step": 34244 + }, + { + "epoch": 1.25, + "learning_rate": 7.47756985388362e-08, + "loss": 0.6856, + "step": 34245 + }, + { + "epoch": 1.25, + "learning_rate": 7.477203647416413e-08, + "loss": 0.6992, + "step": 34246 + }, + { + "epoch": 1.25, + "learning_rate": 7.476837440949206e-08, + "loss": 0.681, + "step": 34247 + }, + { + "epoch": 1.25, + "learning_rate": 7.476471234482e-08, + "loss": 0.6991, + "step": 34248 + }, + { + "epoch": 1.25, + "learning_rate": 7.476105028014794e-08, + "loss": 0.6938, + "step": 34249 + }, + { + "epoch": 1.25, + "learning_rate": 7.475738821547588e-08, + "loss": 0.6921, + "step": 34250 + }, + { + "epoch": 1.25, + "learning_rate": 7.475372615080382e-08, + "loss": 0.7076, + "step": 34251 + }, + { + "epoch": 1.25, + "learning_rate": 7.475006408613176e-08, + "loss": 0.6965, + "step": 34252 + }, + { + "epoch": 1.25, + "learning_rate": 7.47464020214597e-08, + "loss": 0.6834, + "step": 34253 + }, + { + "epoch": 1.25, + "learning_rate": 7.474273995678764e-08, + "loss": 0.712, + "step": 34254 + }, + { + "epoch": 1.25, + "learning_rate": 7.473907789211558e-08, + "loss": 0.6961, + "step": 34255 + }, + { + "epoch": 1.25, + "learning_rate": 7.473541582744351e-08, + "loss": 0.7041, + "step": 34256 + }, + { + "epoch": 1.25, + "learning_rate": 7.473175376277145e-08, + "loss": 0.6869, + "step": 34257 + }, + { + "epoch": 1.25, + "learning_rate": 7.472809169809939e-08, + "loss": 0.6962, + "step": 34258 + }, + { + "epoch": 1.25, + "learning_rate": 7.472442963342733e-08, + "loss": 0.7002, + "step": 34259 + }, + { + "epoch": 1.25, + "learning_rate": 7.472076756875527e-08, + "loss": 0.6912, + "step": 34260 + }, + { + "epoch": 1.25, + "learning_rate": 7.47171055040832e-08, + "loss": 0.6829, + "step": 34261 + }, + { + "epoch": 1.25, + "learning_rate": 7.471344343941113e-08, + "loss": 0.6845, + "step": 34262 + }, + { + "epoch": 1.25, + "learning_rate": 7.470978137473907e-08, + "loss": 0.6928, + "step": 34263 + }, + { + "epoch": 1.25, + "learning_rate": 7.470611931006701e-08, + "loss": 0.6957, + "step": 34264 + }, + { + "epoch": 1.25, + "learning_rate": 7.470245724539494e-08, + "loss": 0.6905, + "step": 34265 + }, + { + "epoch": 1.25, + "learning_rate": 7.469879518072288e-08, + "loss": 0.6894, + "step": 34266 + }, + { + "epoch": 1.25, + "learning_rate": 7.469513311605082e-08, + "loss": 0.6877, + "step": 34267 + }, + { + "epoch": 1.25, + "learning_rate": 7.469147105137876e-08, + "loss": 0.701, + "step": 34268 + }, + { + "epoch": 1.25, + "learning_rate": 7.46878089867067e-08, + "loss": 0.6888, + "step": 34269 + }, + { + "epoch": 1.25, + "learning_rate": 7.468414692203464e-08, + "loss": 0.6831, + "step": 34270 + }, + { + "epoch": 1.25, + "learning_rate": 7.468048485736258e-08, + "loss": 0.6729, + "step": 34271 + }, + { + "epoch": 1.25, + "learning_rate": 7.467682279269052e-08, + "loss": 0.6882, + "step": 34272 + }, + { + "epoch": 1.25, + "learning_rate": 7.467316072801846e-08, + "loss": 0.712, + "step": 34273 + }, + { + "epoch": 1.25, + "learning_rate": 7.466949866334639e-08, + "loss": 0.7003, + "step": 34274 + }, + { + "epoch": 1.25, + "learning_rate": 7.466583659867433e-08, + "loss": 0.6888, + "step": 34275 + }, + { + "epoch": 1.25, + "learning_rate": 7.466217453400227e-08, + "loss": 0.6842, + "step": 34276 + }, + { + "epoch": 1.25, + "learning_rate": 7.46585124693302e-08, + "loss": 0.6867, + "step": 34277 + }, + { + "epoch": 1.25, + "learning_rate": 7.465485040465815e-08, + "loss": 0.6995, + "step": 34278 + }, + { + "epoch": 1.25, + "learning_rate": 7.465118833998609e-08, + "loss": 0.694, + "step": 34279 + }, + { + "epoch": 1.25, + "learning_rate": 7.464752627531403e-08, + "loss": 0.6991, + "step": 34280 + }, + { + "epoch": 1.25, + "learning_rate": 7.464386421064197e-08, + "loss": 0.6909, + "step": 34281 + }, + { + "epoch": 1.25, + "learning_rate": 7.46402021459699e-08, + "loss": 0.7035, + "step": 34282 + }, + { + "epoch": 1.25, + "learning_rate": 7.463654008129783e-08, + "loss": 0.6951, + "step": 34283 + }, + { + "epoch": 1.25, + "learning_rate": 7.463287801662577e-08, + "loss": 0.6972, + "step": 34284 + }, + { + "epoch": 1.25, + "learning_rate": 7.46292159519537e-08, + "loss": 0.7014, + "step": 34285 + }, + { + "epoch": 1.25, + "learning_rate": 7.462555388728164e-08, + "loss": 0.6998, + "step": 34286 + }, + { + "epoch": 1.25, + "learning_rate": 7.462189182260958e-08, + "loss": 0.6988, + "step": 34287 + }, + { + "epoch": 1.25, + "learning_rate": 7.461822975793752e-08, + "loss": 0.6875, + "step": 34288 + }, + { + "epoch": 1.25, + "learning_rate": 7.461456769326546e-08, + "loss": 0.6799, + "step": 34289 + }, + { + "epoch": 1.25, + "learning_rate": 7.46109056285934e-08, + "loss": 0.6984, + "step": 34290 + }, + { + "epoch": 1.25, + "learning_rate": 7.460724356392134e-08, + "loss": 0.6801, + "step": 34291 + }, + { + "epoch": 1.25, + "learning_rate": 7.460358149924927e-08, + "loss": 0.6889, + "step": 34292 + }, + { + "epoch": 1.25, + "learning_rate": 7.45999194345772e-08, + "loss": 0.6969, + "step": 34293 + }, + { + "epoch": 1.25, + "learning_rate": 7.459625736990515e-08, + "loss": 0.6991, + "step": 34294 + }, + { + "epoch": 1.25, + "learning_rate": 7.459259530523309e-08, + "loss": 0.6973, + "step": 34295 + }, + { + "epoch": 1.25, + "learning_rate": 7.458893324056103e-08, + "loss": 0.6779, + "step": 34296 + }, + { + "epoch": 1.25, + "learning_rate": 7.458527117588897e-08, + "loss": 0.6847, + "step": 34297 + }, + { + "epoch": 1.25, + "learning_rate": 7.45816091112169e-08, + "loss": 0.6894, + "step": 34298 + }, + { + "epoch": 1.25, + "learning_rate": 7.457794704654485e-08, + "loss": 0.6874, + "step": 34299 + }, + { + "epoch": 1.25, + "learning_rate": 7.457428498187279e-08, + "loss": 0.6916, + "step": 34300 + }, + { + "epoch": 1.25, + "learning_rate": 7.457062291720071e-08, + "loss": 0.6919, + "step": 34301 + }, + { + "epoch": 1.26, + "learning_rate": 7.456696085252865e-08, + "loss": 0.6902, + "step": 34302 + }, + { + "epoch": 1.26, + "learning_rate": 7.456329878785659e-08, + "loss": 0.6832, + "step": 34303 + }, + { + "epoch": 1.26, + "learning_rate": 7.455963672318453e-08, + "loss": 0.6829, + "step": 34304 + }, + { + "epoch": 1.26, + "learning_rate": 7.455597465851247e-08, + "loss": 0.6894, + "step": 34305 + }, + { + "epoch": 1.26, + "learning_rate": 7.455231259384041e-08, + "loss": 0.6954, + "step": 34306 + }, + { + "epoch": 1.26, + "learning_rate": 7.454865052916834e-08, + "loss": 0.7042, + "step": 34307 + }, + { + "epoch": 1.26, + "learning_rate": 7.454498846449628e-08, + "loss": 0.6825, + "step": 34308 + }, + { + "epoch": 1.26, + "learning_rate": 7.454132639982422e-08, + "loss": 0.6886, + "step": 34309 + }, + { + "epoch": 1.26, + "learning_rate": 7.453766433515214e-08, + "loss": 0.6964, + "step": 34310 + }, + { + "epoch": 1.26, + "learning_rate": 7.453400227048008e-08, + "loss": 0.6952, + "step": 34311 + }, + { + "epoch": 1.26, + "learning_rate": 7.453034020580802e-08, + "loss": 0.6904, + "step": 34312 + }, + { + "epoch": 1.26, + "learning_rate": 7.452667814113596e-08, + "loss": 0.6821, + "step": 34313 + }, + { + "epoch": 1.26, + "learning_rate": 7.45230160764639e-08, + "loss": 0.6905, + "step": 34314 + }, + { + "epoch": 1.26, + "learning_rate": 7.451935401179184e-08, + "loss": 0.7009, + "step": 34315 + }, + { + "epoch": 1.26, + "learning_rate": 7.451569194711978e-08, + "loss": 0.6982, + "step": 34316 + }, + { + "epoch": 1.26, + "learning_rate": 7.451202988244772e-08, + "loss": 0.6847, + "step": 34317 + }, + { + "epoch": 1.26, + "learning_rate": 7.450836781777566e-08, + "loss": 0.6912, + "step": 34318 + }, + { + "epoch": 1.26, + "learning_rate": 7.450470575310359e-08, + "loss": 0.7099, + "step": 34319 + }, + { + "epoch": 1.26, + "learning_rate": 7.450104368843153e-08, + "loss": 0.6928, + "step": 34320 + }, + { + "epoch": 1.26, + "learning_rate": 7.449738162375947e-08, + "loss": 0.6887, + "step": 34321 + }, + { + "epoch": 1.26, + "learning_rate": 7.449371955908741e-08, + "loss": 0.6915, + "step": 34322 + }, + { + "epoch": 1.26, + "learning_rate": 7.449005749441535e-08, + "loss": 0.6845, + "step": 34323 + }, + { + "epoch": 1.26, + "learning_rate": 7.448639542974329e-08, + "loss": 0.6783, + "step": 34324 + }, + { + "epoch": 1.26, + "learning_rate": 7.448273336507123e-08, + "loss": 0.6986, + "step": 34325 + }, + { + "epoch": 1.26, + "learning_rate": 7.447907130039917e-08, + "loss": 0.706, + "step": 34326 + }, + { + "epoch": 1.26, + "learning_rate": 7.447540923572711e-08, + "loss": 0.6983, + "step": 34327 + }, + { + "epoch": 1.26, + "learning_rate": 7.447174717105504e-08, + "loss": 0.6943, + "step": 34328 + }, + { + "epoch": 1.26, + "learning_rate": 7.446808510638298e-08, + "loss": 0.7038, + "step": 34329 + }, + { + "epoch": 1.26, + "learning_rate": 7.446442304171092e-08, + "loss": 0.6833, + "step": 34330 + }, + { + "epoch": 1.26, + "learning_rate": 7.446076097703886e-08, + "loss": 0.6962, + "step": 34331 + }, + { + "epoch": 1.26, + "learning_rate": 7.445709891236678e-08, + "loss": 0.6867, + "step": 34332 + }, + { + "epoch": 1.26, + "learning_rate": 7.445343684769472e-08, + "loss": 0.6891, + "step": 34333 + }, + { + "epoch": 1.26, + "learning_rate": 7.444977478302266e-08, + "loss": 0.6914, + "step": 34334 + }, + { + "epoch": 1.26, + "learning_rate": 7.44461127183506e-08, + "loss": 0.6749, + "step": 34335 + }, + { + "epoch": 1.26, + "learning_rate": 7.444245065367854e-08, + "loss": 0.6876, + "step": 34336 + }, + { + "epoch": 1.26, + "learning_rate": 7.443878858900647e-08, + "loss": 0.7002, + "step": 34337 + }, + { + "epoch": 1.26, + "learning_rate": 7.443512652433441e-08, + "loss": 0.6788, + "step": 34338 + }, + { + "epoch": 1.26, + "learning_rate": 7.443146445966235e-08, + "loss": 0.6937, + "step": 34339 + }, + { + "epoch": 1.26, + "learning_rate": 7.442780239499029e-08, + "loss": 0.7065, + "step": 34340 + }, + { + "epoch": 1.26, + "learning_rate": 7.442414033031823e-08, + "loss": 0.7031, + "step": 34341 + }, + { + "epoch": 1.26, + "learning_rate": 7.442047826564617e-08, + "loss": 0.695, + "step": 34342 + }, + { + "epoch": 1.26, + "learning_rate": 7.441681620097411e-08, + "loss": 0.704, + "step": 34343 + }, + { + "epoch": 1.26, + "learning_rate": 7.441315413630205e-08, + "loss": 0.6992, + "step": 34344 + }, + { + "epoch": 1.26, + "learning_rate": 7.440949207162999e-08, + "loss": 0.6875, + "step": 34345 + }, + { + "epoch": 1.26, + "learning_rate": 7.440583000695792e-08, + "loss": 0.6812, + "step": 34346 + }, + { + "epoch": 1.26, + "learning_rate": 7.440216794228586e-08, + "loss": 0.6972, + "step": 34347 + }, + { + "epoch": 1.26, + "learning_rate": 7.43985058776138e-08, + "loss": 0.7037, + "step": 34348 + }, + { + "epoch": 1.26, + "learning_rate": 7.439484381294174e-08, + "loss": 0.696, + "step": 34349 + }, + { + "epoch": 1.26, + "learning_rate": 7.439118174826968e-08, + "loss": 0.6962, + "step": 34350 + }, + { + "epoch": 1.26, + "learning_rate": 7.438751968359762e-08, + "loss": 0.68, + "step": 34351 + }, + { + "epoch": 1.26, + "learning_rate": 7.438385761892556e-08, + "loss": 0.6897, + "step": 34352 + }, + { + "epoch": 1.26, + "learning_rate": 7.438019555425348e-08, + "loss": 0.6872, + "step": 34353 + }, + { + "epoch": 1.26, + "learning_rate": 7.437653348958142e-08, + "loss": 0.6977, + "step": 34354 + }, + { + "epoch": 1.26, + "learning_rate": 7.437287142490936e-08, + "loss": 0.6987, + "step": 34355 + }, + { + "epoch": 1.26, + "learning_rate": 7.436920936023729e-08, + "loss": 0.7085, + "step": 34356 + }, + { + "epoch": 1.26, + "learning_rate": 7.436554729556523e-08, + "loss": 0.691, + "step": 34357 + }, + { + "epoch": 1.26, + "learning_rate": 7.436188523089317e-08, + "loss": 0.7037, + "step": 34358 + }, + { + "epoch": 1.26, + "learning_rate": 7.435822316622111e-08, + "loss": 0.6988, + "step": 34359 + }, + { + "epoch": 1.26, + "learning_rate": 7.435456110154905e-08, + "loss": 0.6858, + "step": 34360 + }, + { + "epoch": 1.26, + "learning_rate": 7.435089903687699e-08, + "loss": 0.6956, + "step": 34361 + }, + { + "epoch": 1.26, + "learning_rate": 7.434723697220493e-08, + "loss": 0.7022, + "step": 34362 + }, + { + "epoch": 1.26, + "learning_rate": 7.434357490753287e-08, + "loss": 0.6989, + "step": 34363 + }, + { + "epoch": 1.26, + "learning_rate": 7.43399128428608e-08, + "loss": 0.6924, + "step": 34364 + }, + { + "epoch": 1.26, + "learning_rate": 7.433625077818873e-08, + "loss": 0.6957, + "step": 34365 + }, + { + "epoch": 1.26, + "learning_rate": 7.433258871351667e-08, + "loss": 0.6812, + "step": 34366 + }, + { + "epoch": 1.26, + "learning_rate": 7.432892664884461e-08, + "loss": 0.7168, + "step": 34367 + }, + { + "epoch": 1.26, + "learning_rate": 7.432526458417255e-08, + "loss": 0.6884, + "step": 34368 + }, + { + "epoch": 1.26, + "learning_rate": 7.43216025195005e-08, + "loss": 0.6929, + "step": 34369 + }, + { + "epoch": 1.26, + "learning_rate": 7.431794045482843e-08, + "loss": 0.7013, + "step": 34370 + }, + { + "epoch": 1.26, + "learning_rate": 7.431427839015637e-08, + "loss": 0.7057, + "step": 34371 + }, + { + "epoch": 1.26, + "learning_rate": 7.431061632548431e-08, + "loss": 0.6906, + "step": 34372 + }, + { + "epoch": 1.26, + "learning_rate": 7.430695426081224e-08, + "loss": 0.7018, + "step": 34373 + }, + { + "epoch": 1.26, + "learning_rate": 7.430329219614018e-08, + "loss": 0.697, + "step": 34374 + }, + { + "epoch": 1.26, + "learning_rate": 7.429963013146812e-08, + "loss": 0.6848, + "step": 34375 + }, + { + "epoch": 1.26, + "learning_rate": 7.429596806679606e-08, + "loss": 0.6964, + "step": 34376 + }, + { + "epoch": 1.26, + "learning_rate": 7.4292306002124e-08, + "loss": 0.6921, + "step": 34377 + }, + { + "epoch": 1.26, + "learning_rate": 7.428864393745193e-08, + "loss": 0.6902, + "step": 34378 + }, + { + "epoch": 1.26, + "learning_rate": 7.428498187277987e-08, + "loss": 0.6874, + "step": 34379 + }, + { + "epoch": 1.26, + "learning_rate": 7.428131980810781e-08, + "loss": 0.7022, + "step": 34380 + }, + { + "epoch": 1.26, + "learning_rate": 7.427765774343575e-08, + "loss": 0.6828, + "step": 34381 + }, + { + "epoch": 1.26, + "learning_rate": 7.427399567876367e-08, + "loss": 0.6907, + "step": 34382 + }, + { + "epoch": 1.26, + "learning_rate": 7.427033361409161e-08, + "loss": 0.6898, + "step": 34383 + }, + { + "epoch": 1.26, + "learning_rate": 7.426667154941955e-08, + "loss": 0.6801, + "step": 34384 + }, + { + "epoch": 1.26, + "learning_rate": 7.42630094847475e-08, + "loss": 0.6752, + "step": 34385 + }, + { + "epoch": 1.26, + "learning_rate": 7.425934742007543e-08, + "loss": 0.6779, + "step": 34386 + }, + { + "epoch": 1.26, + "learning_rate": 7.425568535540337e-08, + "loss": 0.6906, + "step": 34387 + }, + { + "epoch": 1.26, + "learning_rate": 7.425202329073131e-08, + "loss": 0.6802, + "step": 34388 + }, + { + "epoch": 1.26, + "learning_rate": 7.424836122605925e-08, + "loss": 0.6852, + "step": 34389 + }, + { + "epoch": 1.26, + "learning_rate": 7.424469916138719e-08, + "loss": 0.6975, + "step": 34390 + }, + { + "epoch": 1.26, + "learning_rate": 7.424103709671512e-08, + "loss": 0.6864, + "step": 34391 + }, + { + "epoch": 1.26, + "learning_rate": 7.423737503204306e-08, + "loss": 0.696, + "step": 34392 + }, + { + "epoch": 1.26, + "learning_rate": 7.4233712967371e-08, + "loss": 0.6837, + "step": 34393 + }, + { + "epoch": 1.26, + "learning_rate": 7.423005090269894e-08, + "loss": 0.6893, + "step": 34394 + }, + { + "epoch": 1.26, + "learning_rate": 7.422638883802688e-08, + "loss": 0.7079, + "step": 34395 + }, + { + "epoch": 1.26, + "learning_rate": 7.422272677335482e-08, + "loss": 0.7013, + "step": 34396 + }, + { + "epoch": 1.26, + "learning_rate": 7.421906470868276e-08, + "loss": 0.7117, + "step": 34397 + }, + { + "epoch": 1.26, + "learning_rate": 7.42154026440107e-08, + "loss": 0.7141, + "step": 34398 + }, + { + "epoch": 1.26, + "learning_rate": 7.421174057933864e-08, + "loss": 0.6999, + "step": 34399 + }, + { + "epoch": 1.26, + "learning_rate": 7.420807851466657e-08, + "loss": 0.6766, + "step": 34400 + }, + { + "epoch": 1.26, + "learning_rate": 7.42044164499945e-08, + "loss": 0.6965, + "step": 34401 + }, + { + "epoch": 1.26, + "learning_rate": 7.420075438532245e-08, + "loss": 0.6899, + "step": 34402 + }, + { + "epoch": 1.26, + "learning_rate": 7.419709232065037e-08, + "loss": 0.7098, + "step": 34403 + }, + { + "epoch": 1.26, + "learning_rate": 7.419343025597831e-08, + "loss": 0.6968, + "step": 34404 + }, + { + "epoch": 1.26, + "learning_rate": 7.418976819130625e-08, + "loss": 0.6864, + "step": 34405 + }, + { + "epoch": 1.26, + "learning_rate": 7.418610612663419e-08, + "loss": 0.7045, + "step": 34406 + }, + { + "epoch": 1.26, + "learning_rate": 7.418244406196213e-08, + "loss": 0.6873, + "step": 34407 + }, + { + "epoch": 1.26, + "learning_rate": 7.417878199729007e-08, + "loss": 0.7105, + "step": 34408 + }, + { + "epoch": 1.26, + "learning_rate": 7.4175119932618e-08, + "loss": 0.6954, + "step": 34409 + }, + { + "epoch": 1.26, + "learning_rate": 7.417145786794594e-08, + "loss": 0.7016, + "step": 34410 + }, + { + "epoch": 1.26, + "learning_rate": 7.416779580327388e-08, + "loss": 0.6866, + "step": 34411 + }, + { + "epoch": 1.26, + "learning_rate": 7.416413373860182e-08, + "loss": 0.6922, + "step": 34412 + }, + { + "epoch": 1.26, + "learning_rate": 7.416047167392976e-08, + "loss": 0.7021, + "step": 34413 + }, + { + "epoch": 1.26, + "learning_rate": 7.41568096092577e-08, + "loss": 0.6777, + "step": 34414 + }, + { + "epoch": 1.26, + "learning_rate": 7.415314754458564e-08, + "loss": 0.7123, + "step": 34415 + }, + { + "epoch": 1.26, + "learning_rate": 7.414948547991358e-08, + "loss": 0.6972, + "step": 34416 + }, + { + "epoch": 1.26, + "learning_rate": 7.414582341524152e-08, + "loss": 0.7026, + "step": 34417 + }, + { + "epoch": 1.26, + "learning_rate": 7.414216135056945e-08, + "loss": 0.6916, + "step": 34418 + }, + { + "epoch": 1.26, + "learning_rate": 7.413849928589739e-08, + "loss": 0.6852, + "step": 34419 + }, + { + "epoch": 1.26, + "learning_rate": 7.413483722122532e-08, + "loss": 0.6975, + "step": 34420 + }, + { + "epoch": 1.26, + "learning_rate": 7.413117515655326e-08, + "loss": 0.6982, + "step": 34421 + }, + { + "epoch": 1.26, + "learning_rate": 7.41275130918812e-08, + "loss": 0.7021, + "step": 34422 + }, + { + "epoch": 1.26, + "learning_rate": 7.412385102720914e-08, + "loss": 0.696, + "step": 34423 + }, + { + "epoch": 1.26, + "learning_rate": 7.412018896253707e-08, + "loss": 0.7127, + "step": 34424 + }, + { + "epoch": 1.26, + "learning_rate": 7.411652689786501e-08, + "loss": 0.6868, + "step": 34425 + }, + { + "epoch": 1.26, + "learning_rate": 7.411286483319295e-08, + "loss": 0.6916, + "step": 34426 + }, + { + "epoch": 1.26, + "learning_rate": 7.410920276852088e-08, + "loss": 0.6993, + "step": 34427 + }, + { + "epoch": 1.26, + "learning_rate": 7.410554070384882e-08, + "loss": 0.6828, + "step": 34428 + }, + { + "epoch": 1.26, + "learning_rate": 7.410187863917676e-08, + "loss": 0.6972, + "step": 34429 + }, + { + "epoch": 1.26, + "learning_rate": 7.40982165745047e-08, + "loss": 0.6927, + "step": 34430 + }, + { + "epoch": 1.26, + "learning_rate": 7.409455450983264e-08, + "loss": 0.691, + "step": 34431 + }, + { + "epoch": 1.26, + "learning_rate": 7.409089244516058e-08, + "loss": 0.7009, + "step": 34432 + }, + { + "epoch": 1.26, + "learning_rate": 7.408723038048852e-08, + "loss": 0.6907, + "step": 34433 + }, + { + "epoch": 1.26, + "learning_rate": 7.408356831581646e-08, + "loss": 0.6897, + "step": 34434 + }, + { + "epoch": 1.26, + "learning_rate": 7.40799062511444e-08, + "loss": 0.6713, + "step": 34435 + }, + { + "epoch": 1.26, + "learning_rate": 7.407624418647232e-08, + "loss": 0.7137, + "step": 34436 + }, + { + "epoch": 1.26, + "learning_rate": 7.407258212180026e-08, + "loss": 0.694, + "step": 34437 + }, + { + "epoch": 1.26, + "learning_rate": 7.40689200571282e-08, + "loss": 0.7013, + "step": 34438 + }, + { + "epoch": 1.26, + "learning_rate": 7.406525799245614e-08, + "loss": 0.697, + "step": 34439 + }, + { + "epoch": 1.26, + "learning_rate": 7.406159592778408e-08, + "loss": 0.6965, + "step": 34440 + }, + { + "epoch": 1.26, + "learning_rate": 7.405793386311202e-08, + "loss": 0.7013, + "step": 34441 + }, + { + "epoch": 1.26, + "learning_rate": 7.405427179843996e-08, + "loss": 0.6884, + "step": 34442 + }, + { + "epoch": 1.26, + "learning_rate": 7.40506097337679e-08, + "loss": 0.6913, + "step": 34443 + }, + { + "epoch": 1.26, + "learning_rate": 7.404694766909584e-08, + "loss": 0.6937, + "step": 34444 + }, + { + "epoch": 1.26, + "learning_rate": 7.404328560442377e-08, + "loss": 0.6907, + "step": 34445 + }, + { + "epoch": 1.26, + "learning_rate": 7.403962353975171e-08, + "loss": 0.6958, + "step": 34446 + }, + { + "epoch": 1.26, + "learning_rate": 7.403596147507965e-08, + "loss": 0.7075, + "step": 34447 + }, + { + "epoch": 1.26, + "learning_rate": 7.403229941040759e-08, + "loss": 0.6864, + "step": 34448 + }, + { + "epoch": 1.26, + "learning_rate": 7.402863734573552e-08, + "loss": 0.6948, + "step": 34449 + }, + { + "epoch": 1.26, + "learning_rate": 7.402497528106346e-08, + "loss": 0.7035, + "step": 34450 + }, + { + "epoch": 1.26, + "learning_rate": 7.40213132163914e-08, + "loss": 0.6986, + "step": 34451 + }, + { + "epoch": 1.26, + "learning_rate": 7.401765115171934e-08, + "loss": 0.7004, + "step": 34452 + }, + { + "epoch": 1.26, + "learning_rate": 7.401398908704728e-08, + "loss": 0.6816, + "step": 34453 + }, + { + "epoch": 1.26, + "learning_rate": 7.40103270223752e-08, + "loss": 0.6947, + "step": 34454 + }, + { + "epoch": 1.26, + "learning_rate": 7.400666495770314e-08, + "loss": 0.6936, + "step": 34455 + }, + { + "epoch": 1.26, + "learning_rate": 7.400300289303108e-08, + "loss": 0.6916, + "step": 34456 + }, + { + "epoch": 1.26, + "learning_rate": 7.399934082835902e-08, + "loss": 0.6976, + "step": 34457 + }, + { + "epoch": 1.26, + "learning_rate": 7.399567876368696e-08, + "loss": 0.6954, + "step": 34458 + }, + { + "epoch": 1.26, + "learning_rate": 7.39920166990149e-08, + "loss": 0.7052, + "step": 34459 + }, + { + "epoch": 1.26, + "learning_rate": 7.398835463434284e-08, + "loss": 0.7066, + "step": 34460 + }, + { + "epoch": 1.26, + "learning_rate": 7.398469256967078e-08, + "loss": 0.6856, + "step": 34461 + }, + { + "epoch": 1.26, + "learning_rate": 7.398103050499872e-08, + "loss": 0.6976, + "step": 34462 + }, + { + "epoch": 1.26, + "learning_rate": 7.397736844032665e-08, + "loss": 0.6945, + "step": 34463 + }, + { + "epoch": 1.26, + "learning_rate": 7.397370637565459e-08, + "loss": 0.7009, + "step": 34464 + }, + { + "epoch": 1.26, + "learning_rate": 7.397004431098253e-08, + "loss": 0.7115, + "step": 34465 + }, + { + "epoch": 1.26, + "learning_rate": 7.396638224631047e-08, + "loss": 0.6909, + "step": 34466 + }, + { + "epoch": 1.26, + "learning_rate": 7.396272018163841e-08, + "loss": 0.7028, + "step": 34467 + }, + { + "epoch": 1.26, + "learning_rate": 7.395905811696635e-08, + "loss": 0.6917, + "step": 34468 + }, + { + "epoch": 1.26, + "learning_rate": 7.395539605229429e-08, + "loss": 0.6983, + "step": 34469 + }, + { + "epoch": 1.26, + "learning_rate": 7.395173398762223e-08, + "loss": 0.6993, + "step": 34470 + }, + { + "epoch": 1.26, + "learning_rate": 7.394807192295016e-08, + "loss": 0.6946, + "step": 34471 + }, + { + "epoch": 1.26, + "learning_rate": 7.39444098582781e-08, + "loss": 0.694, + "step": 34472 + }, + { + "epoch": 1.26, + "learning_rate": 7.394074779360602e-08, + "loss": 0.6984, + "step": 34473 + }, + { + "epoch": 1.26, + "learning_rate": 7.393708572893396e-08, + "loss": 0.6879, + "step": 34474 + }, + { + "epoch": 1.26, + "learning_rate": 7.39334236642619e-08, + "loss": 0.6938, + "step": 34475 + }, + { + "epoch": 1.26, + "learning_rate": 7.392976159958984e-08, + "loss": 0.7001, + "step": 34476 + }, + { + "epoch": 1.26, + "learning_rate": 7.392609953491778e-08, + "loss": 0.6995, + "step": 34477 + }, + { + "epoch": 1.26, + "learning_rate": 7.392243747024572e-08, + "loss": 0.7022, + "step": 34478 + }, + { + "epoch": 1.26, + "learning_rate": 7.391877540557366e-08, + "loss": 0.6967, + "step": 34479 + }, + { + "epoch": 1.26, + "learning_rate": 7.391511334090159e-08, + "loss": 0.6934, + "step": 34480 + }, + { + "epoch": 1.26, + "learning_rate": 7.391145127622953e-08, + "loss": 0.698, + "step": 34481 + }, + { + "epoch": 1.26, + "learning_rate": 7.390778921155747e-08, + "loss": 0.6809, + "step": 34482 + }, + { + "epoch": 1.26, + "learning_rate": 7.390412714688541e-08, + "loss": 0.6842, + "step": 34483 + }, + { + "epoch": 1.26, + "learning_rate": 7.390046508221335e-08, + "loss": 0.6969, + "step": 34484 + }, + { + "epoch": 1.26, + "learning_rate": 7.389680301754129e-08, + "loss": 0.6889, + "step": 34485 + }, + { + "epoch": 1.26, + "learning_rate": 7.389314095286923e-08, + "loss": 0.7117, + "step": 34486 + }, + { + "epoch": 1.26, + "learning_rate": 7.388947888819717e-08, + "loss": 0.6822, + "step": 34487 + }, + { + "epoch": 1.26, + "learning_rate": 7.388581682352511e-08, + "loss": 0.6959, + "step": 34488 + }, + { + "epoch": 1.26, + "learning_rate": 7.388215475885303e-08, + "loss": 0.6923, + "step": 34489 + }, + { + "epoch": 1.26, + "learning_rate": 7.387849269418097e-08, + "loss": 0.6796, + "step": 34490 + }, + { + "epoch": 1.26, + "learning_rate": 7.387483062950891e-08, + "loss": 0.671, + "step": 34491 + }, + { + "epoch": 1.26, + "learning_rate": 7.387116856483685e-08, + "loss": 0.6965, + "step": 34492 + }, + { + "epoch": 1.26, + "learning_rate": 7.38675065001648e-08, + "loss": 0.6887, + "step": 34493 + }, + { + "epoch": 1.26, + "learning_rate": 7.386384443549273e-08, + "loss": 0.6848, + "step": 34494 + }, + { + "epoch": 1.26, + "learning_rate": 7.386018237082066e-08, + "loss": 0.6897, + "step": 34495 + }, + { + "epoch": 1.26, + "learning_rate": 7.38565203061486e-08, + "loss": 0.6819, + "step": 34496 + }, + { + "epoch": 1.26, + "learning_rate": 7.385285824147654e-08, + "loss": 0.6884, + "step": 34497 + }, + { + "epoch": 1.26, + "learning_rate": 7.384919617680447e-08, + "loss": 0.6904, + "step": 34498 + }, + { + "epoch": 1.26, + "learning_rate": 7.384553411213241e-08, + "loss": 0.682, + "step": 34499 + }, + { + "epoch": 1.26, + "learning_rate": 7.384187204746035e-08, + "loss": 0.6912, + "step": 34500 + }, + { + "epoch": 1.26, + "learning_rate": 7.383820998278829e-08, + "loss": 0.6855, + "step": 34501 + }, + { + "epoch": 1.26, + "learning_rate": 7.383454791811623e-08, + "loss": 0.6902, + "step": 34502 + }, + { + "epoch": 1.26, + "learning_rate": 7.383088585344417e-08, + "loss": 0.6833, + "step": 34503 + }, + { + "epoch": 1.26, + "learning_rate": 7.382722378877211e-08, + "loss": 0.6985, + "step": 34504 + }, + { + "epoch": 1.26, + "learning_rate": 7.382356172410005e-08, + "loss": 0.6965, + "step": 34505 + }, + { + "epoch": 1.26, + "learning_rate": 7.381989965942799e-08, + "loss": 0.6837, + "step": 34506 + }, + { + "epoch": 1.26, + "learning_rate": 7.381623759475591e-08, + "loss": 0.6935, + "step": 34507 + }, + { + "epoch": 1.26, + "learning_rate": 7.381257553008385e-08, + "loss": 0.7013, + "step": 34508 + }, + { + "epoch": 1.26, + "learning_rate": 7.380891346541179e-08, + "loss": 0.6927, + "step": 34509 + }, + { + "epoch": 1.26, + "learning_rate": 7.380525140073973e-08, + "loss": 0.6937, + "step": 34510 + }, + { + "epoch": 1.26, + "learning_rate": 7.380158933606767e-08, + "loss": 0.6991, + "step": 34511 + }, + { + "epoch": 1.26, + "learning_rate": 7.379792727139561e-08, + "loss": 0.6815, + "step": 34512 + }, + { + "epoch": 1.26, + "learning_rate": 7.379426520672355e-08, + "loss": 0.6837, + "step": 34513 + }, + { + "epoch": 1.26, + "learning_rate": 7.379060314205149e-08, + "loss": 0.6999, + "step": 34514 + }, + { + "epoch": 1.26, + "learning_rate": 7.378694107737943e-08, + "loss": 0.7083, + "step": 34515 + }, + { + "epoch": 1.26, + "learning_rate": 7.378327901270736e-08, + "loss": 0.6773, + "step": 34516 + }, + { + "epoch": 1.26, + "learning_rate": 7.37796169480353e-08, + "loss": 0.6904, + "step": 34517 + }, + { + "epoch": 1.26, + "learning_rate": 7.377595488336324e-08, + "loss": 0.6811, + "step": 34518 + }, + { + "epoch": 1.26, + "learning_rate": 7.377229281869118e-08, + "loss": 0.6884, + "step": 34519 + }, + { + "epoch": 1.26, + "learning_rate": 7.37686307540191e-08, + "loss": 0.6927, + "step": 34520 + }, + { + "epoch": 1.26, + "learning_rate": 7.376496868934705e-08, + "loss": 0.6988, + "step": 34521 + }, + { + "epoch": 1.26, + "learning_rate": 7.376130662467499e-08, + "loss": 0.7086, + "step": 34522 + }, + { + "epoch": 1.26, + "learning_rate": 7.375764456000293e-08, + "loss": 0.6916, + "step": 34523 + }, + { + "epoch": 1.26, + "learning_rate": 7.375398249533087e-08, + "loss": 0.6884, + "step": 34524 + }, + { + "epoch": 1.26, + "learning_rate": 7.375032043065879e-08, + "loss": 0.6943, + "step": 34525 + }, + { + "epoch": 1.26, + "learning_rate": 7.374665836598673e-08, + "loss": 0.6792, + "step": 34526 + }, + { + "epoch": 1.26, + "learning_rate": 7.374299630131467e-08, + "loss": 0.6983, + "step": 34527 + }, + { + "epoch": 1.26, + "learning_rate": 7.373933423664261e-08, + "loss": 0.6944, + "step": 34528 + }, + { + "epoch": 1.26, + "learning_rate": 7.373567217197055e-08, + "loss": 0.6819, + "step": 34529 + }, + { + "epoch": 1.26, + "learning_rate": 7.373201010729849e-08, + "loss": 0.682, + "step": 34530 + }, + { + "epoch": 1.26, + "learning_rate": 7.372834804262643e-08, + "loss": 0.6694, + "step": 34531 + }, + { + "epoch": 1.26, + "learning_rate": 7.372468597795437e-08, + "loss": 0.6756, + "step": 34532 + }, + { + "epoch": 1.26, + "learning_rate": 7.372102391328231e-08, + "loss": 0.7014, + "step": 34533 + }, + { + "epoch": 1.26, + "learning_rate": 7.371736184861024e-08, + "loss": 0.7018, + "step": 34534 + }, + { + "epoch": 1.26, + "learning_rate": 7.371369978393818e-08, + "loss": 0.7019, + "step": 34535 + }, + { + "epoch": 1.26, + "learning_rate": 7.371003771926612e-08, + "loss": 0.6996, + "step": 34536 + }, + { + "epoch": 1.26, + "learning_rate": 7.370637565459406e-08, + "loss": 0.6914, + "step": 34537 + }, + { + "epoch": 1.26, + "learning_rate": 7.3702713589922e-08, + "loss": 0.6984, + "step": 34538 + }, + { + "epoch": 1.26, + "learning_rate": 7.369905152524994e-08, + "loss": 0.6912, + "step": 34539 + }, + { + "epoch": 1.26, + "learning_rate": 7.369538946057788e-08, + "loss": 0.6881, + "step": 34540 + }, + { + "epoch": 1.26, + "learning_rate": 7.369172739590582e-08, + "loss": 0.7007, + "step": 34541 + }, + { + "epoch": 1.26, + "learning_rate": 7.368806533123374e-08, + "loss": 0.7107, + "step": 34542 + }, + { + "epoch": 1.26, + "learning_rate": 7.368440326656168e-08, + "loss": 0.7009, + "step": 34543 + }, + { + "epoch": 1.26, + "learning_rate": 7.368074120188961e-08, + "loss": 0.6899, + "step": 34544 + }, + { + "epoch": 1.26, + "learning_rate": 7.367707913721755e-08, + "loss": 0.6996, + "step": 34545 + }, + { + "epoch": 1.26, + "learning_rate": 7.367341707254549e-08, + "loss": 0.7118, + "step": 34546 + }, + { + "epoch": 1.26, + "learning_rate": 7.366975500787343e-08, + "loss": 0.694, + "step": 34547 + }, + { + "epoch": 1.26, + "learning_rate": 7.366609294320137e-08, + "loss": 0.6746, + "step": 34548 + }, + { + "epoch": 1.26, + "learning_rate": 7.366243087852931e-08, + "loss": 0.7087, + "step": 34549 + }, + { + "epoch": 1.26, + "learning_rate": 7.365876881385725e-08, + "loss": 0.6874, + "step": 34550 + }, + { + "epoch": 1.26, + "learning_rate": 7.365510674918519e-08, + "loss": 0.699, + "step": 34551 + }, + { + "epoch": 1.26, + "learning_rate": 7.365144468451312e-08, + "loss": 0.698, + "step": 34552 + }, + { + "epoch": 1.26, + "learning_rate": 7.364778261984106e-08, + "loss": 0.6982, + "step": 34553 + }, + { + "epoch": 1.26, + "learning_rate": 7.3644120555169e-08, + "loss": 0.6832, + "step": 34554 + }, + { + "epoch": 1.26, + "learning_rate": 7.364045849049694e-08, + "loss": 0.6889, + "step": 34555 + }, + { + "epoch": 1.26, + "learning_rate": 7.363679642582488e-08, + "loss": 0.6754, + "step": 34556 + }, + { + "epoch": 1.26, + "learning_rate": 7.363313436115282e-08, + "loss": 0.6802, + "step": 34557 + }, + { + "epoch": 1.26, + "learning_rate": 7.362947229648076e-08, + "loss": 0.6912, + "step": 34558 + }, + { + "epoch": 1.26, + "learning_rate": 7.36258102318087e-08, + "loss": 0.6831, + "step": 34559 + }, + { + "epoch": 1.26, + "learning_rate": 7.362214816713664e-08, + "loss": 0.6857, + "step": 34560 + }, + { + "epoch": 1.26, + "learning_rate": 7.361848610246456e-08, + "loss": 0.6896, + "step": 34561 + }, + { + "epoch": 1.26, + "learning_rate": 7.36148240377925e-08, + "loss": 0.6919, + "step": 34562 + }, + { + "epoch": 1.26, + "learning_rate": 7.361116197312044e-08, + "loss": 0.695, + "step": 34563 + }, + { + "epoch": 1.26, + "learning_rate": 7.360749990844838e-08, + "loss": 0.6773, + "step": 34564 + }, + { + "epoch": 1.26, + "learning_rate": 7.360383784377632e-08, + "loss": 0.6922, + "step": 34565 + }, + { + "epoch": 1.26, + "learning_rate": 7.360017577910425e-08, + "loss": 0.6922, + "step": 34566 + }, + { + "epoch": 1.26, + "learning_rate": 7.359651371443219e-08, + "loss": 0.6886, + "step": 34567 + }, + { + "epoch": 1.26, + "learning_rate": 7.359285164976013e-08, + "loss": 0.6927, + "step": 34568 + }, + { + "epoch": 1.26, + "learning_rate": 7.358918958508807e-08, + "loss": 0.6783, + "step": 34569 + }, + { + "epoch": 1.26, + "learning_rate": 7.3585527520416e-08, + "loss": 0.689, + "step": 34570 + }, + { + "epoch": 1.26, + "learning_rate": 7.358186545574394e-08, + "loss": 0.7054, + "step": 34571 + }, + { + "epoch": 1.26, + "learning_rate": 7.357820339107188e-08, + "loss": 0.6666, + "step": 34572 + }, + { + "epoch": 1.26, + "learning_rate": 7.357454132639982e-08, + "loss": 0.6862, + "step": 34573 + }, + { + "epoch": 1.26, + "learning_rate": 7.357087926172776e-08, + "loss": 0.6884, + "step": 34574 + }, + { + "epoch": 1.27, + "learning_rate": 7.35672171970557e-08, + "loss": 0.6995, + "step": 34575 + }, + { + "epoch": 1.27, + "learning_rate": 7.356355513238364e-08, + "loss": 0.6928, + "step": 34576 + }, + { + "epoch": 1.27, + "learning_rate": 7.355989306771158e-08, + "loss": 0.686, + "step": 34577 + }, + { + "epoch": 1.27, + "learning_rate": 7.355623100303952e-08, + "loss": 0.6999, + "step": 34578 + }, + { + "epoch": 1.27, + "learning_rate": 7.355256893836744e-08, + "loss": 0.6939, + "step": 34579 + }, + { + "epoch": 1.27, + "learning_rate": 7.354890687369538e-08, + "loss": 0.6731, + "step": 34580 + }, + { + "epoch": 1.27, + "learning_rate": 7.354524480902332e-08, + "loss": 0.7002, + "step": 34581 + }, + { + "epoch": 1.27, + "learning_rate": 7.354158274435126e-08, + "loss": 0.6898, + "step": 34582 + }, + { + "epoch": 1.27, + "learning_rate": 7.35379206796792e-08, + "loss": 0.7074, + "step": 34583 + }, + { + "epoch": 1.27, + "learning_rate": 7.353425861500714e-08, + "loss": 0.6982, + "step": 34584 + }, + { + "epoch": 1.27, + "learning_rate": 7.353059655033508e-08, + "loss": 0.6956, + "step": 34585 + }, + { + "epoch": 1.27, + "learning_rate": 7.352693448566302e-08, + "loss": 0.7005, + "step": 34586 + }, + { + "epoch": 1.27, + "learning_rate": 7.352327242099096e-08, + "loss": 0.679, + "step": 34587 + }, + { + "epoch": 1.27, + "learning_rate": 7.351961035631889e-08, + "loss": 0.6844, + "step": 34588 + }, + { + "epoch": 1.27, + "learning_rate": 7.351594829164683e-08, + "loss": 0.6932, + "step": 34589 + }, + { + "epoch": 1.27, + "learning_rate": 7.351228622697477e-08, + "loss": 0.6942, + "step": 34590 + }, + { + "epoch": 1.27, + "learning_rate": 7.35086241623027e-08, + "loss": 0.6973, + "step": 34591 + }, + { + "epoch": 1.27, + "learning_rate": 7.350496209763064e-08, + "loss": 0.6925, + "step": 34592 + }, + { + "epoch": 1.27, + "learning_rate": 7.350130003295858e-08, + "loss": 0.6958, + "step": 34593 + }, + { + "epoch": 1.27, + "learning_rate": 7.349763796828652e-08, + "loss": 0.6752, + "step": 34594 + }, + { + "epoch": 1.27, + "learning_rate": 7.349397590361446e-08, + "loss": 0.6898, + "step": 34595 + }, + { + "epoch": 1.27, + "learning_rate": 7.34903138389424e-08, + "loss": 0.7027, + "step": 34596 + }, + { + "epoch": 1.27, + "learning_rate": 7.348665177427032e-08, + "loss": 0.7025, + "step": 34597 + }, + { + "epoch": 1.27, + "learning_rate": 7.348298970959826e-08, + "loss": 0.7021, + "step": 34598 + }, + { + "epoch": 1.27, + "learning_rate": 7.34793276449262e-08, + "loss": 0.6885, + "step": 34599 + }, + { + "epoch": 1.27, + "learning_rate": 7.347566558025414e-08, + "loss": 0.7034, + "step": 34600 + }, + { + "epoch": 1.27, + "learning_rate": 7.347200351558208e-08, + "loss": 0.6979, + "step": 34601 + }, + { + "epoch": 1.27, + "learning_rate": 7.346834145091002e-08, + "loss": 0.6963, + "step": 34602 + }, + { + "epoch": 1.27, + "learning_rate": 7.346467938623796e-08, + "loss": 0.6949, + "step": 34603 + }, + { + "epoch": 1.27, + "learning_rate": 7.34610173215659e-08, + "loss": 0.7001, + "step": 34604 + }, + { + "epoch": 1.27, + "learning_rate": 7.345735525689384e-08, + "loss": 0.6805, + "step": 34605 + }, + { + "epoch": 1.27, + "learning_rate": 7.345369319222177e-08, + "loss": 0.6753, + "step": 34606 + }, + { + "epoch": 1.27, + "learning_rate": 7.345003112754971e-08, + "loss": 0.6752, + "step": 34607 + }, + { + "epoch": 1.27, + "learning_rate": 7.344636906287765e-08, + "loss": 0.6904, + "step": 34608 + }, + { + "epoch": 1.27, + "learning_rate": 7.344270699820559e-08, + "loss": 0.7055, + "step": 34609 + }, + { + "epoch": 1.27, + "learning_rate": 7.343904493353353e-08, + "loss": 0.6807, + "step": 34610 + }, + { + "epoch": 1.27, + "learning_rate": 7.343538286886147e-08, + "loss": 0.689, + "step": 34611 + }, + { + "epoch": 1.27, + "learning_rate": 7.343172080418941e-08, + "loss": 0.6999, + "step": 34612 + }, + { + "epoch": 1.27, + "learning_rate": 7.342805873951733e-08, + "loss": 0.6943, + "step": 34613 + }, + { + "epoch": 1.27, + "learning_rate": 7.342439667484527e-08, + "loss": 0.6924, + "step": 34614 + }, + { + "epoch": 1.27, + "learning_rate": 7.34207346101732e-08, + "loss": 0.6857, + "step": 34615 + }, + { + "epoch": 1.27, + "learning_rate": 7.341707254550114e-08, + "loss": 0.6942, + "step": 34616 + }, + { + "epoch": 1.27, + "learning_rate": 7.341341048082908e-08, + "loss": 0.6976, + "step": 34617 + }, + { + "epoch": 1.27, + "learning_rate": 7.340974841615702e-08, + "loss": 0.704, + "step": 34618 + }, + { + "epoch": 1.27, + "learning_rate": 7.340608635148496e-08, + "loss": 0.705, + "step": 34619 + }, + { + "epoch": 1.27, + "learning_rate": 7.34024242868129e-08, + "loss": 0.6914, + "step": 34620 + }, + { + "epoch": 1.27, + "learning_rate": 7.339876222214084e-08, + "loss": 0.6926, + "step": 34621 + }, + { + "epoch": 1.27, + "learning_rate": 7.339510015746878e-08, + "loss": 0.6916, + "step": 34622 + }, + { + "epoch": 1.27, + "learning_rate": 7.339143809279672e-08, + "loss": 0.6855, + "step": 34623 + }, + { + "epoch": 1.27, + "learning_rate": 7.338777602812465e-08, + "loss": 0.6941, + "step": 34624 + }, + { + "epoch": 1.27, + "learning_rate": 7.338411396345259e-08, + "loss": 0.6935, + "step": 34625 + }, + { + "epoch": 1.27, + "learning_rate": 7.338045189878053e-08, + "loss": 0.6934, + "step": 34626 + }, + { + "epoch": 1.27, + "learning_rate": 7.337678983410847e-08, + "loss": 0.6937, + "step": 34627 + }, + { + "epoch": 1.27, + "learning_rate": 7.33731277694364e-08, + "loss": 0.6941, + "step": 34628 + }, + { + "epoch": 1.27, + "learning_rate": 7.336946570476435e-08, + "loss": 0.6984, + "step": 34629 + }, + { + "epoch": 1.27, + "learning_rate": 7.336580364009229e-08, + "loss": 0.6783, + "step": 34630 + }, + { + "epoch": 1.27, + "learning_rate": 7.336214157542023e-08, + "loss": 0.6888, + "step": 34631 + }, + { + "epoch": 1.27, + "learning_rate": 7.335847951074817e-08, + "loss": 0.7131, + "step": 34632 + }, + { + "epoch": 1.27, + "learning_rate": 7.335481744607609e-08, + "loss": 0.6944, + "step": 34633 + }, + { + "epoch": 1.27, + "learning_rate": 7.335115538140403e-08, + "loss": 0.7002, + "step": 34634 + }, + { + "epoch": 1.27, + "learning_rate": 7.334749331673197e-08, + "loss": 0.6825, + "step": 34635 + }, + { + "epoch": 1.27, + "learning_rate": 7.334383125205991e-08, + "loss": 0.6878, + "step": 34636 + }, + { + "epoch": 1.27, + "learning_rate": 7.334016918738784e-08, + "loss": 0.7141, + "step": 34637 + }, + { + "epoch": 1.27, + "learning_rate": 7.333650712271578e-08, + "loss": 0.6915, + "step": 34638 + }, + { + "epoch": 1.27, + "learning_rate": 7.333284505804372e-08, + "loss": 0.6947, + "step": 34639 + }, + { + "epoch": 1.27, + "learning_rate": 7.332918299337166e-08, + "loss": 0.6948, + "step": 34640 + }, + { + "epoch": 1.27, + "learning_rate": 7.33255209286996e-08, + "loss": 0.703, + "step": 34641 + }, + { + "epoch": 1.27, + "learning_rate": 7.332185886402753e-08, + "loss": 0.6987, + "step": 34642 + }, + { + "epoch": 1.27, + "learning_rate": 7.331819679935547e-08, + "loss": 0.699, + "step": 34643 + }, + { + "epoch": 1.27, + "learning_rate": 7.33145347346834e-08, + "loss": 0.6847, + "step": 34644 + }, + { + "epoch": 1.27, + "learning_rate": 7.331087267001135e-08, + "loss": 0.6984, + "step": 34645 + }, + { + "epoch": 1.27, + "learning_rate": 7.330721060533929e-08, + "loss": 0.703, + "step": 34646 + }, + { + "epoch": 1.27, + "learning_rate": 7.330354854066723e-08, + "loss": 0.6856, + "step": 34647 + }, + { + "epoch": 1.27, + "learning_rate": 7.329988647599517e-08, + "loss": 0.6869, + "step": 34648 + }, + { + "epoch": 1.27, + "learning_rate": 7.32962244113231e-08, + "loss": 0.6997, + "step": 34649 + }, + { + "epoch": 1.27, + "learning_rate": 7.329256234665105e-08, + "loss": 0.7037, + "step": 34650 + }, + { + "epoch": 1.27, + "learning_rate": 7.328890028197897e-08, + "loss": 0.704, + "step": 34651 + }, + { + "epoch": 1.27, + "learning_rate": 7.328523821730691e-08, + "loss": 0.7086, + "step": 34652 + }, + { + "epoch": 1.27, + "learning_rate": 7.328157615263485e-08, + "loss": 0.6921, + "step": 34653 + }, + { + "epoch": 1.27, + "learning_rate": 7.327791408796279e-08, + "loss": 0.6763, + "step": 34654 + }, + { + "epoch": 1.27, + "learning_rate": 7.327425202329073e-08, + "loss": 0.6767, + "step": 34655 + }, + { + "epoch": 1.27, + "learning_rate": 7.327058995861867e-08, + "loss": 0.6864, + "step": 34656 + }, + { + "epoch": 1.27, + "learning_rate": 7.326692789394661e-08, + "loss": 0.6944, + "step": 34657 + }, + { + "epoch": 1.27, + "learning_rate": 7.326326582927455e-08, + "loss": 0.6983, + "step": 34658 + }, + { + "epoch": 1.27, + "learning_rate": 7.325960376460248e-08, + "loss": 0.6952, + "step": 34659 + }, + { + "epoch": 1.27, + "learning_rate": 7.325594169993042e-08, + "loss": 0.6971, + "step": 34660 + }, + { + "epoch": 1.27, + "learning_rate": 7.325227963525836e-08, + "loss": 0.6966, + "step": 34661 + }, + { + "epoch": 1.27, + "learning_rate": 7.324861757058628e-08, + "loss": 0.6947, + "step": 34662 + }, + { + "epoch": 1.27, + "learning_rate": 7.324495550591422e-08, + "loss": 0.6898, + "step": 34663 + }, + { + "epoch": 1.27, + "learning_rate": 7.324129344124216e-08, + "loss": 0.687, + "step": 34664 + }, + { + "epoch": 1.27, + "learning_rate": 7.32376313765701e-08, + "loss": 0.6932, + "step": 34665 + }, + { + "epoch": 1.27, + "learning_rate": 7.323396931189804e-08, + "loss": 0.6814, + "step": 34666 + }, + { + "epoch": 1.27, + "learning_rate": 7.323030724722598e-08, + "loss": 0.6962, + "step": 34667 + }, + { + "epoch": 1.27, + "learning_rate": 7.322664518255392e-08, + "loss": 0.6782, + "step": 34668 + }, + { + "epoch": 1.27, + "learning_rate": 7.322298311788185e-08, + "loss": 0.6916, + "step": 34669 + }, + { + "epoch": 1.27, + "learning_rate": 7.321932105320979e-08, + "loss": 0.7003, + "step": 34670 + }, + { + "epoch": 1.27, + "learning_rate": 7.321565898853773e-08, + "loss": 0.681, + "step": 34671 + }, + { + "epoch": 1.27, + "learning_rate": 7.321199692386567e-08, + "loss": 0.6933, + "step": 34672 + }, + { + "epoch": 1.27, + "learning_rate": 7.320833485919361e-08, + "loss": 0.6958, + "step": 34673 + }, + { + "epoch": 1.27, + "learning_rate": 7.320467279452155e-08, + "loss": 0.6942, + "step": 34674 + }, + { + "epoch": 1.27, + "learning_rate": 7.320101072984949e-08, + "loss": 0.6937, + "step": 34675 + }, + { + "epoch": 1.27, + "learning_rate": 7.319734866517743e-08, + "loss": 0.6855, + "step": 34676 + }, + { + "epoch": 1.27, + "learning_rate": 7.319368660050537e-08, + "loss": 0.6898, + "step": 34677 + }, + { + "epoch": 1.27, + "learning_rate": 7.31900245358333e-08, + "loss": 0.6907, + "step": 34678 + }, + { + "epoch": 1.27, + "learning_rate": 7.318636247116124e-08, + "loss": 0.6878, + "step": 34679 + }, + { + "epoch": 1.27, + "learning_rate": 7.318270040648918e-08, + "loss": 0.6864, + "step": 34680 + }, + { + "epoch": 1.27, + "learning_rate": 7.317903834181712e-08, + "loss": 0.6972, + "step": 34681 + }, + { + "epoch": 1.27, + "learning_rate": 7.317537627714506e-08, + "loss": 0.697, + "step": 34682 + }, + { + "epoch": 1.27, + "learning_rate": 7.3171714212473e-08, + "loss": 0.6941, + "step": 34683 + }, + { + "epoch": 1.27, + "learning_rate": 7.316805214780092e-08, + "loss": 0.7008, + "step": 34684 + }, + { + "epoch": 1.27, + "learning_rate": 7.316439008312886e-08, + "loss": 0.6948, + "step": 34685 + }, + { + "epoch": 1.27, + "learning_rate": 7.31607280184568e-08, + "loss": 0.7007, + "step": 34686 + }, + { + "epoch": 1.27, + "learning_rate": 7.315706595378473e-08, + "loss": 0.695, + "step": 34687 + }, + { + "epoch": 1.27, + "learning_rate": 7.315340388911267e-08, + "loss": 0.6965, + "step": 34688 + }, + { + "epoch": 1.27, + "learning_rate": 7.314974182444061e-08, + "loss": 0.6977, + "step": 34689 + }, + { + "epoch": 1.27, + "learning_rate": 7.314607975976855e-08, + "loss": 0.6806, + "step": 34690 + }, + { + "epoch": 1.27, + "learning_rate": 7.314241769509649e-08, + "loss": 0.6899, + "step": 34691 + }, + { + "epoch": 1.27, + "learning_rate": 7.313875563042443e-08, + "loss": 0.6947, + "step": 34692 + }, + { + "epoch": 1.27, + "learning_rate": 7.313509356575237e-08, + "loss": 0.6733, + "step": 34693 + }, + { + "epoch": 1.27, + "learning_rate": 7.313143150108031e-08, + "loss": 0.6775, + "step": 34694 + }, + { + "epoch": 1.27, + "learning_rate": 7.312776943640825e-08, + "loss": 0.7, + "step": 34695 + }, + { + "epoch": 1.27, + "learning_rate": 7.312410737173618e-08, + "loss": 0.6884, + "step": 34696 + }, + { + "epoch": 1.27, + "learning_rate": 7.312044530706412e-08, + "loss": 0.6984, + "step": 34697 + }, + { + "epoch": 1.27, + "learning_rate": 7.311678324239206e-08, + "loss": 0.7001, + "step": 34698 + }, + { + "epoch": 1.27, + "learning_rate": 7.311312117772e-08, + "loss": 0.6843, + "step": 34699 + }, + { + "epoch": 1.27, + "learning_rate": 7.310945911304794e-08, + "loss": 0.6811, + "step": 34700 + }, + { + "epoch": 1.27, + "learning_rate": 7.310579704837588e-08, + "loss": 0.7122, + "step": 34701 + }, + { + "epoch": 1.27, + "learning_rate": 7.310213498370382e-08, + "loss": 0.6885, + "step": 34702 + }, + { + "epoch": 1.27, + "learning_rate": 7.309847291903176e-08, + "loss": 0.701, + "step": 34703 + }, + { + "epoch": 1.27, + "learning_rate": 7.30948108543597e-08, + "loss": 0.7044, + "step": 34704 + }, + { + "epoch": 1.27, + "learning_rate": 7.309114878968762e-08, + "loss": 0.688, + "step": 34705 + }, + { + "epoch": 1.27, + "learning_rate": 7.308748672501556e-08, + "loss": 0.6902, + "step": 34706 + }, + { + "epoch": 1.27, + "learning_rate": 7.30838246603435e-08, + "loss": 0.6881, + "step": 34707 + }, + { + "epoch": 1.27, + "learning_rate": 7.308016259567143e-08, + "loss": 0.6826, + "step": 34708 + }, + { + "epoch": 1.27, + "learning_rate": 7.307650053099937e-08, + "loss": 0.6893, + "step": 34709 + }, + { + "epoch": 1.27, + "learning_rate": 7.307283846632731e-08, + "loss": 0.7031, + "step": 34710 + }, + { + "epoch": 1.27, + "learning_rate": 7.306917640165525e-08, + "loss": 0.6901, + "step": 34711 + }, + { + "epoch": 1.27, + "learning_rate": 7.306551433698319e-08, + "loss": 0.6966, + "step": 34712 + }, + { + "epoch": 1.27, + "learning_rate": 7.306185227231113e-08, + "loss": 0.6996, + "step": 34713 + }, + { + "epoch": 1.27, + "learning_rate": 7.305819020763906e-08, + "loss": 0.6963, + "step": 34714 + }, + { + "epoch": 1.27, + "learning_rate": 7.3054528142967e-08, + "loss": 0.6974, + "step": 34715 + }, + { + "epoch": 1.27, + "learning_rate": 7.305086607829494e-08, + "loss": 0.6984, + "step": 34716 + }, + { + "epoch": 1.27, + "learning_rate": 7.304720401362287e-08, + "loss": 0.6854, + "step": 34717 + }, + { + "epoch": 1.27, + "learning_rate": 7.304354194895081e-08, + "loss": 0.682, + "step": 34718 + }, + { + "epoch": 1.27, + "learning_rate": 7.303987988427875e-08, + "loss": 0.6872, + "step": 34719 + }, + { + "epoch": 1.27, + "learning_rate": 7.30362178196067e-08, + "loss": 0.6996, + "step": 34720 + }, + { + "epoch": 1.27, + "learning_rate": 7.303255575493463e-08, + "loss": 0.6917, + "step": 34721 + }, + { + "epoch": 1.27, + "learning_rate": 7.302889369026257e-08, + "loss": 0.6993, + "step": 34722 + }, + { + "epoch": 1.27, + "learning_rate": 7.30252316255905e-08, + "loss": 0.6979, + "step": 34723 + }, + { + "epoch": 1.27, + "learning_rate": 7.302156956091844e-08, + "loss": 0.6959, + "step": 34724 + }, + { + "epoch": 1.27, + "learning_rate": 7.301790749624638e-08, + "loss": 0.6974, + "step": 34725 + }, + { + "epoch": 1.27, + "learning_rate": 7.301424543157432e-08, + "loss": 0.6775, + "step": 34726 + }, + { + "epoch": 1.27, + "learning_rate": 7.301058336690226e-08, + "loss": 0.7075, + "step": 34727 + }, + { + "epoch": 1.27, + "learning_rate": 7.30069213022302e-08, + "loss": 0.6825, + "step": 34728 + }, + { + "epoch": 1.27, + "learning_rate": 7.300325923755814e-08, + "loss": 0.6862, + "step": 34729 + }, + { + "epoch": 1.27, + "learning_rate": 7.299959717288607e-08, + "loss": 0.6893, + "step": 34730 + }, + { + "epoch": 1.27, + "learning_rate": 7.299593510821401e-08, + "loss": 0.6897, + "step": 34731 + }, + { + "epoch": 1.27, + "learning_rate": 7.299227304354195e-08, + "loss": 0.6857, + "step": 34732 + }, + { + "epoch": 1.27, + "learning_rate": 7.298861097886987e-08, + "loss": 0.7037, + "step": 34733 + }, + { + "epoch": 1.27, + "learning_rate": 7.298494891419781e-08, + "loss": 0.7017, + "step": 34734 + }, + { + "epoch": 1.27, + "learning_rate": 7.298128684952575e-08, + "loss": 0.7015, + "step": 34735 + }, + { + "epoch": 1.27, + "learning_rate": 7.29776247848537e-08, + "loss": 0.6987, + "step": 34736 + }, + { + "epoch": 1.27, + "learning_rate": 7.297396272018163e-08, + "loss": 0.6888, + "step": 34737 + }, + { + "epoch": 1.27, + "learning_rate": 7.297030065550957e-08, + "loss": 0.6701, + "step": 34738 + }, + { + "epoch": 1.27, + "learning_rate": 7.296663859083751e-08, + "loss": 0.6749, + "step": 34739 + }, + { + "epoch": 1.27, + "learning_rate": 7.296297652616545e-08, + "loss": 0.6934, + "step": 34740 + }, + { + "epoch": 1.27, + "learning_rate": 7.295931446149338e-08, + "loss": 0.7076, + "step": 34741 + }, + { + "epoch": 1.27, + "learning_rate": 7.295565239682132e-08, + "loss": 0.6844, + "step": 34742 + }, + { + "epoch": 1.27, + "learning_rate": 7.295199033214926e-08, + "loss": 0.6964, + "step": 34743 + }, + { + "epoch": 1.27, + "learning_rate": 7.29483282674772e-08, + "loss": 0.6955, + "step": 34744 + }, + { + "epoch": 1.27, + "learning_rate": 7.294466620280514e-08, + "loss": 0.692, + "step": 34745 + }, + { + "epoch": 1.27, + "learning_rate": 7.294100413813308e-08, + "loss": 0.7001, + "step": 34746 + }, + { + "epoch": 1.27, + "learning_rate": 7.293734207346102e-08, + "loss": 0.7, + "step": 34747 + }, + { + "epoch": 1.27, + "learning_rate": 7.293368000878896e-08, + "loss": 0.7016, + "step": 34748 + }, + { + "epoch": 1.27, + "learning_rate": 7.29300179441169e-08, + "loss": 0.7109, + "step": 34749 + }, + { + "epoch": 1.27, + "learning_rate": 7.292635587944483e-08, + "loss": 0.6764, + "step": 34750 + }, + { + "epoch": 1.27, + "learning_rate": 7.292269381477277e-08, + "loss": 0.6981, + "step": 34751 + }, + { + "epoch": 1.27, + "learning_rate": 7.29190317501007e-08, + "loss": 0.6799, + "step": 34752 + }, + { + "epoch": 1.27, + "learning_rate": 7.291536968542865e-08, + "loss": 0.6899, + "step": 34753 + }, + { + "epoch": 1.27, + "learning_rate": 7.291170762075659e-08, + "loss": 0.7029, + "step": 34754 + }, + { + "epoch": 1.27, + "learning_rate": 7.290804555608451e-08, + "loss": 0.6923, + "step": 34755 + }, + { + "epoch": 1.27, + "learning_rate": 7.290438349141245e-08, + "loss": 0.6942, + "step": 34756 + }, + { + "epoch": 1.27, + "learning_rate": 7.290072142674039e-08, + "loss": 0.709, + "step": 34757 + }, + { + "epoch": 1.27, + "learning_rate": 7.289705936206833e-08, + "loss": 0.6883, + "step": 34758 + }, + { + "epoch": 1.27, + "learning_rate": 7.289339729739626e-08, + "loss": 0.6938, + "step": 34759 + }, + { + "epoch": 1.27, + "learning_rate": 7.28897352327242e-08, + "loss": 0.6865, + "step": 34760 + }, + { + "epoch": 1.27, + "learning_rate": 7.288607316805214e-08, + "loss": 0.6913, + "step": 34761 + }, + { + "epoch": 1.27, + "learning_rate": 7.288241110338008e-08, + "loss": 0.6982, + "step": 34762 + }, + { + "epoch": 1.27, + "learning_rate": 7.287874903870802e-08, + "loss": 0.6793, + "step": 34763 + }, + { + "epoch": 1.27, + "learning_rate": 7.287508697403596e-08, + "loss": 0.6962, + "step": 34764 + }, + { + "epoch": 1.27, + "learning_rate": 7.28714249093639e-08, + "loss": 0.6978, + "step": 34765 + }, + { + "epoch": 1.27, + "learning_rate": 7.286776284469184e-08, + "loss": 0.6923, + "step": 34766 + }, + { + "epoch": 1.27, + "learning_rate": 7.286410078001978e-08, + "loss": 0.7045, + "step": 34767 + }, + { + "epoch": 1.27, + "learning_rate": 7.28604387153477e-08, + "loss": 0.6777, + "step": 34768 + }, + { + "epoch": 1.27, + "learning_rate": 7.285677665067565e-08, + "loss": 0.6819, + "step": 34769 + }, + { + "epoch": 1.27, + "learning_rate": 7.285311458600359e-08, + "loss": 0.6975, + "step": 34770 + }, + { + "epoch": 1.27, + "learning_rate": 7.284945252133153e-08, + "loss": 0.6985, + "step": 34771 + }, + { + "epoch": 1.27, + "learning_rate": 7.284579045665947e-08, + "loss": 0.7016, + "step": 34772 + }, + { + "epoch": 1.27, + "learning_rate": 7.28421283919874e-08, + "loss": 0.693, + "step": 34773 + }, + { + "epoch": 1.27, + "learning_rate": 7.283846632731535e-08, + "loss": 0.6885, + "step": 34774 + }, + { + "epoch": 1.27, + "learning_rate": 7.283480426264328e-08, + "loss": 0.6874, + "step": 34775 + }, + { + "epoch": 1.27, + "learning_rate": 7.283114219797121e-08, + "loss": 0.7059, + "step": 34776 + }, + { + "epoch": 1.27, + "learning_rate": 7.282748013329915e-08, + "loss": 0.7046, + "step": 34777 + }, + { + "epoch": 1.27, + "learning_rate": 7.282381806862709e-08, + "loss": 0.7042, + "step": 34778 + }, + { + "epoch": 1.27, + "learning_rate": 7.282015600395502e-08, + "loss": 0.7101, + "step": 34779 + }, + { + "epoch": 1.27, + "learning_rate": 7.281649393928296e-08, + "loss": 0.7069, + "step": 34780 + }, + { + "epoch": 1.27, + "learning_rate": 7.28128318746109e-08, + "loss": 0.6965, + "step": 34781 + }, + { + "epoch": 1.27, + "learning_rate": 7.280916980993884e-08, + "loss": 0.6972, + "step": 34782 + }, + { + "epoch": 1.27, + "learning_rate": 7.280550774526678e-08, + "loss": 0.6875, + "step": 34783 + }, + { + "epoch": 1.27, + "learning_rate": 7.280184568059472e-08, + "loss": 0.7051, + "step": 34784 + }, + { + "epoch": 1.27, + "learning_rate": 7.279818361592266e-08, + "loss": 0.6957, + "step": 34785 + }, + { + "epoch": 1.27, + "learning_rate": 7.279452155125058e-08, + "loss": 0.6848, + "step": 34786 + }, + { + "epoch": 1.27, + "learning_rate": 7.279085948657852e-08, + "loss": 0.6895, + "step": 34787 + }, + { + "epoch": 1.27, + "learning_rate": 7.278719742190646e-08, + "loss": 0.7075, + "step": 34788 + }, + { + "epoch": 1.27, + "learning_rate": 7.27835353572344e-08, + "loss": 0.6893, + "step": 34789 + }, + { + "epoch": 1.27, + "learning_rate": 7.277987329256234e-08, + "loss": 0.6823, + "step": 34790 + }, + { + "epoch": 1.27, + "learning_rate": 7.277621122789028e-08, + "loss": 0.6875, + "step": 34791 + }, + { + "epoch": 1.27, + "learning_rate": 7.277254916321822e-08, + "loss": 0.689, + "step": 34792 + }, + { + "epoch": 1.27, + "learning_rate": 7.276888709854616e-08, + "loss": 0.7043, + "step": 34793 + }, + { + "epoch": 1.27, + "learning_rate": 7.27652250338741e-08, + "loss": 0.7043, + "step": 34794 + }, + { + "epoch": 1.27, + "learning_rate": 7.276156296920203e-08, + "loss": 0.6745, + "step": 34795 + }, + { + "epoch": 1.27, + "learning_rate": 7.275790090452997e-08, + "loss": 0.6936, + "step": 34796 + }, + { + "epoch": 1.27, + "learning_rate": 7.275423883985791e-08, + "loss": 0.6969, + "step": 34797 + }, + { + "epoch": 1.27, + "learning_rate": 7.275057677518585e-08, + "loss": 0.6893, + "step": 34798 + }, + { + "epoch": 1.27, + "learning_rate": 7.274691471051379e-08, + "loss": 0.697, + "step": 34799 + }, + { + "epoch": 1.27, + "learning_rate": 7.274325264584173e-08, + "loss": 0.7058, + "step": 34800 + }, + { + "epoch": 1.27, + "learning_rate": 7.273959058116966e-08, + "loss": 0.6986, + "step": 34801 + }, + { + "epoch": 1.27, + "learning_rate": 7.27359285164976e-08, + "loss": 0.6927, + "step": 34802 + }, + { + "epoch": 1.27, + "learning_rate": 7.273226645182554e-08, + "loss": 0.7067, + "step": 34803 + }, + { + "epoch": 1.27, + "learning_rate": 7.272860438715346e-08, + "loss": 0.6878, + "step": 34804 + }, + { + "epoch": 1.27, + "learning_rate": 7.27249423224814e-08, + "loss": 0.6983, + "step": 34805 + }, + { + "epoch": 1.27, + "learning_rate": 7.272128025780934e-08, + "loss": 0.7054, + "step": 34806 + }, + { + "epoch": 1.27, + "learning_rate": 7.271761819313728e-08, + "loss": 0.6912, + "step": 34807 + }, + { + "epoch": 1.27, + "learning_rate": 7.271395612846522e-08, + "loss": 0.6973, + "step": 34808 + }, + { + "epoch": 1.27, + "learning_rate": 7.271029406379316e-08, + "loss": 0.6985, + "step": 34809 + }, + { + "epoch": 1.27, + "learning_rate": 7.27066319991211e-08, + "loss": 0.6882, + "step": 34810 + }, + { + "epoch": 1.27, + "learning_rate": 7.270296993444904e-08, + "loss": 0.6922, + "step": 34811 + }, + { + "epoch": 1.27, + "learning_rate": 7.269930786977698e-08, + "loss": 0.6979, + "step": 34812 + }, + { + "epoch": 1.27, + "learning_rate": 7.269564580510491e-08, + "loss": 0.6904, + "step": 34813 + }, + { + "epoch": 1.27, + "learning_rate": 7.269198374043285e-08, + "loss": 0.7171, + "step": 34814 + }, + { + "epoch": 1.27, + "learning_rate": 7.268832167576079e-08, + "loss": 0.6873, + "step": 34815 + }, + { + "epoch": 1.27, + "learning_rate": 7.268465961108873e-08, + "loss": 0.7051, + "step": 34816 + }, + { + "epoch": 1.27, + "learning_rate": 7.268099754641667e-08, + "loss": 0.6975, + "step": 34817 + }, + { + "epoch": 1.27, + "learning_rate": 7.267733548174461e-08, + "loss": 0.6977, + "step": 34818 + }, + { + "epoch": 1.27, + "learning_rate": 7.267367341707255e-08, + "loss": 0.6863, + "step": 34819 + }, + { + "epoch": 1.27, + "learning_rate": 7.267001135240049e-08, + "loss": 0.6996, + "step": 34820 + }, + { + "epoch": 1.27, + "learning_rate": 7.266634928772843e-08, + "loss": 0.6966, + "step": 34821 + }, + { + "epoch": 1.27, + "learning_rate": 7.266268722305636e-08, + "loss": 0.7021, + "step": 34822 + }, + { + "epoch": 1.27, + "learning_rate": 7.26590251583843e-08, + "loss": 0.6818, + "step": 34823 + }, + { + "epoch": 1.27, + "learning_rate": 7.265536309371224e-08, + "loss": 0.6867, + "step": 34824 + }, + { + "epoch": 1.27, + "learning_rate": 7.265170102904016e-08, + "loss": 0.6945, + "step": 34825 + }, + { + "epoch": 1.27, + "learning_rate": 7.26480389643681e-08, + "loss": 0.6935, + "step": 34826 + }, + { + "epoch": 1.27, + "learning_rate": 7.264437689969604e-08, + "loss": 0.6922, + "step": 34827 + }, + { + "epoch": 1.27, + "learning_rate": 7.264071483502398e-08, + "loss": 0.7014, + "step": 34828 + }, + { + "epoch": 1.27, + "learning_rate": 7.263705277035192e-08, + "loss": 0.6912, + "step": 34829 + }, + { + "epoch": 1.27, + "learning_rate": 7.263339070567986e-08, + "loss": 0.6931, + "step": 34830 + }, + { + "epoch": 1.27, + "learning_rate": 7.262972864100779e-08, + "loss": 0.6897, + "step": 34831 + }, + { + "epoch": 1.27, + "learning_rate": 7.262606657633573e-08, + "loss": 0.7103, + "step": 34832 + }, + { + "epoch": 1.27, + "learning_rate": 7.262240451166367e-08, + "loss": 0.6857, + "step": 34833 + }, + { + "epoch": 1.27, + "learning_rate": 7.261874244699161e-08, + "loss": 0.6886, + "step": 34834 + }, + { + "epoch": 1.27, + "learning_rate": 7.261508038231955e-08, + "loss": 0.6912, + "step": 34835 + }, + { + "epoch": 1.27, + "learning_rate": 7.261141831764749e-08, + "loss": 0.6886, + "step": 34836 + }, + { + "epoch": 1.27, + "learning_rate": 7.260775625297543e-08, + "loss": 0.6894, + "step": 34837 + }, + { + "epoch": 1.27, + "learning_rate": 7.260409418830337e-08, + "loss": 0.6857, + "step": 34838 + }, + { + "epoch": 1.27, + "learning_rate": 7.260043212363131e-08, + "loss": 0.689, + "step": 34839 + }, + { + "epoch": 1.27, + "learning_rate": 7.259677005895923e-08, + "loss": 0.6947, + "step": 34840 + }, + { + "epoch": 1.27, + "learning_rate": 7.259310799428717e-08, + "loss": 0.6991, + "step": 34841 + }, + { + "epoch": 1.27, + "learning_rate": 7.258944592961511e-08, + "loss": 0.693, + "step": 34842 + }, + { + "epoch": 1.27, + "learning_rate": 7.258578386494305e-08, + "loss": 0.6877, + "step": 34843 + }, + { + "epoch": 1.27, + "learning_rate": 7.2582121800271e-08, + "loss": 0.697, + "step": 34844 + }, + { + "epoch": 1.27, + "learning_rate": 7.257845973559893e-08, + "loss": 0.6905, + "step": 34845 + }, + { + "epoch": 1.27, + "learning_rate": 7.257479767092687e-08, + "loss": 0.684, + "step": 34846 + }, + { + "epoch": 1.27, + "learning_rate": 7.25711356062548e-08, + "loss": 0.7014, + "step": 34847 + }, + { + "epoch": 1.27, + "learning_rate": 7.256747354158274e-08, + "loss": 0.7081, + "step": 34848 + }, + { + "epoch": 1.28, + "learning_rate": 7.256381147691068e-08, + "loss": 0.6895, + "step": 34849 + }, + { + "epoch": 1.28, + "learning_rate": 7.256014941223861e-08, + "loss": 0.6854, + "step": 34850 + }, + { + "epoch": 1.28, + "learning_rate": 7.255648734756655e-08, + "loss": 0.6834, + "step": 34851 + }, + { + "epoch": 1.28, + "learning_rate": 7.255282528289449e-08, + "loss": 0.6955, + "step": 34852 + }, + { + "epoch": 1.28, + "learning_rate": 7.254916321822243e-08, + "loss": 0.6891, + "step": 34853 + }, + { + "epoch": 1.28, + "learning_rate": 7.254550115355037e-08, + "loss": 0.713, + "step": 34854 + }, + { + "epoch": 1.28, + "learning_rate": 7.254183908887831e-08, + "loss": 0.6921, + "step": 34855 + }, + { + "epoch": 1.28, + "learning_rate": 7.253817702420625e-08, + "loss": 0.6994, + "step": 34856 + }, + { + "epoch": 1.28, + "learning_rate": 7.253451495953417e-08, + "loss": 0.6884, + "step": 34857 + }, + { + "epoch": 1.28, + "learning_rate": 7.253085289486211e-08, + "loss": 0.6882, + "step": 34858 + }, + { + "epoch": 1.28, + "learning_rate": 7.252719083019005e-08, + "loss": 0.6898, + "step": 34859 + }, + { + "epoch": 1.28, + "learning_rate": 7.2523528765518e-08, + "loss": 0.6942, + "step": 34860 + }, + { + "epoch": 1.28, + "learning_rate": 7.251986670084593e-08, + "loss": 0.7175, + "step": 34861 + }, + { + "epoch": 1.28, + "learning_rate": 7.251620463617387e-08, + "loss": 0.6734, + "step": 34862 + }, + { + "epoch": 1.28, + "learning_rate": 7.251254257150181e-08, + "loss": 0.6985, + "step": 34863 + }, + { + "epoch": 1.28, + "learning_rate": 7.250888050682975e-08, + "loss": 0.6936, + "step": 34864 + }, + { + "epoch": 1.28, + "learning_rate": 7.250521844215769e-08, + "loss": 0.6876, + "step": 34865 + }, + { + "epoch": 1.28, + "learning_rate": 7.250155637748562e-08, + "loss": 0.7004, + "step": 34866 + }, + { + "epoch": 1.28, + "learning_rate": 7.249789431281356e-08, + "loss": 0.7067, + "step": 34867 + }, + { + "epoch": 1.28, + "learning_rate": 7.24942322481415e-08, + "loss": 0.6893, + "step": 34868 + }, + { + "epoch": 1.28, + "learning_rate": 7.249057018346944e-08, + "loss": 0.7081, + "step": 34869 + }, + { + "epoch": 1.28, + "learning_rate": 7.248690811879738e-08, + "loss": 0.6949, + "step": 34870 + }, + { + "epoch": 1.28, + "learning_rate": 7.248324605412532e-08, + "loss": 0.6863, + "step": 34871 + }, + { + "epoch": 1.28, + "learning_rate": 7.247958398945325e-08, + "loss": 0.6941, + "step": 34872 + }, + { + "epoch": 1.28, + "learning_rate": 7.247592192478119e-08, + "loss": 0.6969, + "step": 34873 + }, + { + "epoch": 1.28, + "learning_rate": 7.247225986010913e-08, + "loss": 0.7043, + "step": 34874 + }, + { + "epoch": 1.28, + "learning_rate": 7.246859779543705e-08, + "loss": 0.687, + "step": 34875 + }, + { + "epoch": 1.28, + "learning_rate": 7.246493573076499e-08, + "loss": 0.6871, + "step": 34876 + }, + { + "epoch": 1.28, + "learning_rate": 7.246127366609293e-08, + "loss": 0.6832, + "step": 34877 + }, + { + "epoch": 1.28, + "learning_rate": 7.245761160142087e-08, + "loss": 0.6962, + "step": 34878 + }, + { + "epoch": 1.28, + "learning_rate": 7.245394953674881e-08, + "loss": 0.695, + "step": 34879 + }, + { + "epoch": 1.28, + "learning_rate": 7.245028747207675e-08, + "loss": 0.6993, + "step": 34880 + }, + { + "epoch": 1.28, + "learning_rate": 7.244662540740469e-08, + "loss": 0.6958, + "step": 34881 + }, + { + "epoch": 1.28, + "learning_rate": 7.244296334273263e-08, + "loss": 0.6796, + "step": 34882 + }, + { + "epoch": 1.28, + "learning_rate": 7.243930127806057e-08, + "loss": 0.7098, + "step": 34883 + }, + { + "epoch": 1.28, + "learning_rate": 7.24356392133885e-08, + "loss": 0.6986, + "step": 34884 + }, + { + "epoch": 1.28, + "learning_rate": 7.243197714871644e-08, + "loss": 0.7022, + "step": 34885 + }, + { + "epoch": 1.28, + "learning_rate": 7.242831508404438e-08, + "loss": 0.6895, + "step": 34886 + }, + { + "epoch": 1.28, + "learning_rate": 7.242465301937232e-08, + "loss": 0.6937, + "step": 34887 + }, + { + "epoch": 1.28, + "learning_rate": 7.242099095470026e-08, + "loss": 0.7042, + "step": 34888 + }, + { + "epoch": 1.28, + "learning_rate": 7.24173288900282e-08, + "loss": 0.691, + "step": 34889 + }, + { + "epoch": 1.28, + "learning_rate": 7.241366682535614e-08, + "loss": 0.683, + "step": 34890 + }, + { + "epoch": 1.28, + "learning_rate": 7.241000476068408e-08, + "loss": 0.6884, + "step": 34891 + }, + { + "epoch": 1.28, + "learning_rate": 7.240634269601202e-08, + "loss": 0.6864, + "step": 34892 + }, + { + "epoch": 1.28, + "learning_rate": 7.240268063133995e-08, + "loss": 0.6962, + "step": 34893 + }, + { + "epoch": 1.28, + "learning_rate": 7.239901856666788e-08, + "loss": 0.7004, + "step": 34894 + }, + { + "epoch": 1.28, + "learning_rate": 7.239535650199582e-08, + "loss": 0.6992, + "step": 34895 + }, + { + "epoch": 1.28, + "learning_rate": 7.239169443732375e-08, + "loss": 0.6907, + "step": 34896 + }, + { + "epoch": 1.28, + "learning_rate": 7.238803237265169e-08, + "loss": 0.6804, + "step": 34897 + }, + { + "epoch": 1.28, + "learning_rate": 7.238437030797963e-08, + "loss": 0.694, + "step": 34898 + }, + { + "epoch": 1.28, + "learning_rate": 7.238070824330757e-08, + "loss": 0.6859, + "step": 34899 + }, + { + "epoch": 1.28, + "learning_rate": 7.237704617863551e-08, + "loss": 0.6834, + "step": 34900 + }, + { + "epoch": 1.28, + "learning_rate": 7.237338411396345e-08, + "loss": 0.6905, + "step": 34901 + }, + { + "epoch": 1.28, + "learning_rate": 7.236972204929138e-08, + "loss": 0.6804, + "step": 34902 + }, + { + "epoch": 1.28, + "learning_rate": 7.236605998461932e-08, + "loss": 0.6962, + "step": 34903 + }, + { + "epoch": 1.28, + "learning_rate": 7.236239791994726e-08, + "loss": 0.6952, + "step": 34904 + }, + { + "epoch": 1.28, + "learning_rate": 7.23587358552752e-08, + "loss": 0.6798, + "step": 34905 + }, + { + "epoch": 1.28, + "learning_rate": 7.235507379060314e-08, + "loss": 0.6993, + "step": 34906 + }, + { + "epoch": 1.28, + "learning_rate": 7.235141172593108e-08, + "loss": 0.7004, + "step": 34907 + }, + { + "epoch": 1.28, + "learning_rate": 7.234774966125902e-08, + "loss": 0.6883, + "step": 34908 + }, + { + "epoch": 1.28, + "learning_rate": 7.234408759658696e-08, + "loss": 0.6887, + "step": 34909 + }, + { + "epoch": 1.28, + "learning_rate": 7.23404255319149e-08, + "loss": 0.6843, + "step": 34910 + }, + { + "epoch": 1.28, + "learning_rate": 7.233676346724282e-08, + "loss": 0.6921, + "step": 34911 + }, + { + "epoch": 1.28, + "learning_rate": 7.233310140257076e-08, + "loss": 0.6885, + "step": 34912 + }, + { + "epoch": 1.28, + "learning_rate": 7.23294393378987e-08, + "loss": 0.6832, + "step": 34913 + }, + { + "epoch": 1.28, + "learning_rate": 7.232577727322664e-08, + "loss": 0.693, + "step": 34914 + }, + { + "epoch": 1.28, + "learning_rate": 7.232211520855458e-08, + "loss": 0.6978, + "step": 34915 + }, + { + "epoch": 1.28, + "learning_rate": 7.231845314388252e-08, + "loss": 0.6887, + "step": 34916 + }, + { + "epoch": 1.28, + "learning_rate": 7.231479107921046e-08, + "loss": 0.696, + "step": 34917 + }, + { + "epoch": 1.28, + "learning_rate": 7.231112901453839e-08, + "loss": 0.7002, + "step": 34918 + }, + { + "epoch": 1.28, + "learning_rate": 7.230746694986633e-08, + "loss": 0.6779, + "step": 34919 + }, + { + "epoch": 1.28, + "learning_rate": 7.230380488519427e-08, + "loss": 0.6912, + "step": 34920 + }, + { + "epoch": 1.28, + "learning_rate": 7.23001428205222e-08, + "loss": 0.7001, + "step": 34921 + }, + { + "epoch": 1.28, + "learning_rate": 7.229648075585014e-08, + "loss": 0.7014, + "step": 34922 + }, + { + "epoch": 1.28, + "learning_rate": 7.229281869117808e-08, + "loss": 0.6837, + "step": 34923 + }, + { + "epoch": 1.28, + "learning_rate": 7.228915662650602e-08, + "loss": 0.6822, + "step": 34924 + }, + { + "epoch": 1.28, + "learning_rate": 7.228549456183396e-08, + "loss": 0.7103, + "step": 34925 + }, + { + "epoch": 1.28, + "learning_rate": 7.22818324971619e-08, + "loss": 0.6874, + "step": 34926 + }, + { + "epoch": 1.28, + "learning_rate": 7.227817043248984e-08, + "loss": 0.6919, + "step": 34927 + }, + { + "epoch": 1.28, + "learning_rate": 7.227450836781778e-08, + "loss": 0.714, + "step": 34928 + }, + { + "epoch": 1.28, + "learning_rate": 7.22708463031457e-08, + "loss": 0.6856, + "step": 34929 + }, + { + "epoch": 1.28, + "learning_rate": 7.226718423847364e-08, + "loss": 0.6912, + "step": 34930 + }, + { + "epoch": 1.28, + "learning_rate": 7.226352217380158e-08, + "loss": 0.6894, + "step": 34931 + }, + { + "epoch": 1.28, + "learning_rate": 7.225986010912952e-08, + "loss": 0.7041, + "step": 34932 + }, + { + "epoch": 1.28, + "learning_rate": 7.225619804445746e-08, + "loss": 0.6964, + "step": 34933 + }, + { + "epoch": 1.28, + "learning_rate": 7.22525359797854e-08, + "loss": 0.6865, + "step": 34934 + }, + { + "epoch": 1.28, + "learning_rate": 7.224887391511334e-08, + "loss": 0.687, + "step": 34935 + }, + { + "epoch": 1.28, + "learning_rate": 7.224521185044128e-08, + "loss": 0.699, + "step": 34936 + }, + { + "epoch": 1.28, + "learning_rate": 7.224154978576922e-08, + "loss": 0.6844, + "step": 34937 + }, + { + "epoch": 1.28, + "learning_rate": 7.223788772109715e-08, + "loss": 0.684, + "step": 34938 + }, + { + "epoch": 1.28, + "learning_rate": 7.223422565642509e-08, + "loss": 0.6829, + "step": 34939 + }, + { + "epoch": 1.28, + "learning_rate": 7.223056359175303e-08, + "loss": 0.7, + "step": 34940 + }, + { + "epoch": 1.28, + "learning_rate": 7.222690152708097e-08, + "loss": 0.6956, + "step": 34941 + }, + { + "epoch": 1.28, + "learning_rate": 7.222323946240891e-08, + "loss": 0.7044, + "step": 34942 + }, + { + "epoch": 1.28, + "learning_rate": 7.221957739773684e-08, + "loss": 0.6804, + "step": 34943 + }, + { + "epoch": 1.28, + "learning_rate": 7.221591533306478e-08, + "loss": 0.709, + "step": 34944 + }, + { + "epoch": 1.28, + "learning_rate": 7.221225326839272e-08, + "loss": 0.696, + "step": 34945 + }, + { + "epoch": 1.28, + "learning_rate": 7.220859120372066e-08, + "loss": 0.6994, + "step": 34946 + }, + { + "epoch": 1.28, + "learning_rate": 7.220492913904858e-08, + "loss": 0.6922, + "step": 34947 + }, + { + "epoch": 1.28, + "learning_rate": 7.220126707437652e-08, + "loss": 0.6995, + "step": 34948 + }, + { + "epoch": 1.28, + "learning_rate": 7.219760500970446e-08, + "loss": 0.6959, + "step": 34949 + }, + { + "epoch": 1.28, + "learning_rate": 7.21939429450324e-08, + "loss": 0.7028, + "step": 34950 + }, + { + "epoch": 1.28, + "learning_rate": 7.219028088036034e-08, + "loss": 0.6811, + "step": 34951 + }, + { + "epoch": 1.28, + "learning_rate": 7.218661881568828e-08, + "loss": 0.6646, + "step": 34952 + }, + { + "epoch": 1.28, + "learning_rate": 7.218295675101622e-08, + "loss": 0.707, + "step": 34953 + }, + { + "epoch": 1.28, + "learning_rate": 7.217929468634416e-08, + "loss": 0.6761, + "step": 34954 + }, + { + "epoch": 1.28, + "learning_rate": 7.21756326216721e-08, + "loss": 0.6876, + "step": 34955 + }, + { + "epoch": 1.28, + "learning_rate": 7.217197055700003e-08, + "loss": 0.6911, + "step": 34956 + }, + { + "epoch": 1.28, + "learning_rate": 7.216830849232797e-08, + "loss": 0.6918, + "step": 34957 + }, + { + "epoch": 1.28, + "learning_rate": 7.216464642765591e-08, + "loss": 0.7113, + "step": 34958 + }, + { + "epoch": 1.28, + "learning_rate": 7.216098436298385e-08, + "loss": 0.6763, + "step": 34959 + }, + { + "epoch": 1.28, + "learning_rate": 7.215732229831179e-08, + "loss": 0.7095, + "step": 34960 + }, + { + "epoch": 1.28, + "learning_rate": 7.215366023363973e-08, + "loss": 0.6891, + "step": 34961 + }, + { + "epoch": 1.28, + "learning_rate": 7.214999816896767e-08, + "loss": 0.7083, + "step": 34962 + }, + { + "epoch": 1.28, + "learning_rate": 7.214633610429561e-08, + "loss": 0.6942, + "step": 34963 + }, + { + "epoch": 1.28, + "learning_rate": 7.214267403962355e-08, + "loss": 0.7017, + "step": 34964 + }, + { + "epoch": 1.28, + "learning_rate": 7.213901197495147e-08, + "loss": 0.6925, + "step": 34965 + }, + { + "epoch": 1.28, + "learning_rate": 7.213534991027941e-08, + "loss": 0.6837, + "step": 34966 + }, + { + "epoch": 1.28, + "learning_rate": 7.213168784560734e-08, + "loss": 0.6989, + "step": 34967 + }, + { + "epoch": 1.28, + "learning_rate": 7.212802578093528e-08, + "loss": 0.6991, + "step": 34968 + }, + { + "epoch": 1.28, + "learning_rate": 7.212436371626322e-08, + "loss": 0.6955, + "step": 34969 + }, + { + "epoch": 1.28, + "learning_rate": 7.212070165159116e-08, + "loss": 0.7064, + "step": 34970 + }, + { + "epoch": 1.28, + "learning_rate": 7.21170395869191e-08, + "loss": 0.6976, + "step": 34971 + }, + { + "epoch": 1.28, + "learning_rate": 7.211337752224704e-08, + "loss": 0.7077, + "step": 34972 + }, + { + "epoch": 1.28, + "learning_rate": 7.210971545757498e-08, + "loss": 0.6971, + "step": 34973 + }, + { + "epoch": 1.28, + "learning_rate": 7.210605339290291e-08, + "loss": 0.7011, + "step": 34974 + }, + { + "epoch": 1.28, + "learning_rate": 7.210239132823085e-08, + "loss": 0.6844, + "step": 34975 + }, + { + "epoch": 1.28, + "learning_rate": 7.209872926355879e-08, + "loss": 0.6945, + "step": 34976 + }, + { + "epoch": 1.28, + "learning_rate": 7.209506719888673e-08, + "loss": 0.7005, + "step": 34977 + }, + { + "epoch": 1.28, + "learning_rate": 7.209140513421467e-08, + "loss": 0.6904, + "step": 34978 + }, + { + "epoch": 1.28, + "learning_rate": 7.208774306954261e-08, + "loss": 0.6824, + "step": 34979 + }, + { + "epoch": 1.28, + "learning_rate": 7.208408100487055e-08, + "loss": 0.6908, + "step": 34980 + }, + { + "epoch": 1.28, + "learning_rate": 7.208041894019849e-08, + "loss": 0.6937, + "step": 34981 + }, + { + "epoch": 1.28, + "learning_rate": 7.207675687552643e-08, + "loss": 0.7011, + "step": 34982 + }, + { + "epoch": 1.28, + "learning_rate": 7.207309481085435e-08, + "loss": 0.6916, + "step": 34983 + }, + { + "epoch": 1.28, + "learning_rate": 7.206943274618229e-08, + "loss": 0.6898, + "step": 34984 + }, + { + "epoch": 1.28, + "learning_rate": 7.206577068151023e-08, + "loss": 0.6966, + "step": 34985 + }, + { + "epoch": 1.28, + "learning_rate": 7.206210861683817e-08, + "loss": 0.696, + "step": 34986 + }, + { + "epoch": 1.28, + "learning_rate": 7.205844655216611e-08, + "loss": 0.6961, + "step": 34987 + }, + { + "epoch": 1.28, + "learning_rate": 7.205478448749405e-08, + "loss": 0.692, + "step": 34988 + }, + { + "epoch": 1.28, + "learning_rate": 7.205112242282198e-08, + "loss": 0.6949, + "step": 34989 + }, + { + "epoch": 1.28, + "learning_rate": 7.204746035814992e-08, + "loss": 0.6985, + "step": 34990 + }, + { + "epoch": 1.28, + "learning_rate": 7.204379829347786e-08, + "loss": 0.6879, + "step": 34991 + }, + { + "epoch": 1.28, + "learning_rate": 7.204013622880579e-08, + "loss": 0.7014, + "step": 34992 + }, + { + "epoch": 1.28, + "learning_rate": 7.203647416413373e-08, + "loss": 0.7039, + "step": 34993 + }, + { + "epoch": 1.28, + "learning_rate": 7.203281209946167e-08, + "loss": 0.682, + "step": 34994 + }, + { + "epoch": 1.28, + "learning_rate": 7.20291500347896e-08, + "loss": 0.703, + "step": 34995 + }, + { + "epoch": 1.28, + "learning_rate": 7.202548797011755e-08, + "loss": 0.7009, + "step": 34996 + }, + { + "epoch": 1.28, + "learning_rate": 7.202182590544549e-08, + "loss": 0.6735, + "step": 34997 + }, + { + "epoch": 1.28, + "learning_rate": 7.201816384077343e-08, + "loss": 0.695, + "step": 34998 + }, + { + "epoch": 1.28, + "learning_rate": 7.201450177610137e-08, + "loss": 0.7002, + "step": 34999 + }, + { + "epoch": 1.28, + "learning_rate": 7.20108397114293e-08, + "loss": 0.6919, + "step": 35000 + }, + { + "epoch": 1.28, + "learning_rate": 7.200717764675723e-08, + "loss": 0.6874, + "step": 35001 + }, + { + "epoch": 1.28, + "learning_rate": 7.200351558208517e-08, + "loss": 0.6993, + "step": 35002 + }, + { + "epoch": 1.28, + "learning_rate": 7.199985351741311e-08, + "loss": 0.693, + "step": 35003 + }, + { + "epoch": 1.28, + "learning_rate": 7.199619145274105e-08, + "loss": 0.6888, + "step": 35004 + }, + { + "epoch": 1.28, + "learning_rate": 7.199252938806899e-08, + "loss": 0.6892, + "step": 35005 + }, + { + "epoch": 1.28, + "learning_rate": 7.198886732339693e-08, + "loss": 0.6903, + "step": 35006 + }, + { + "epoch": 1.28, + "learning_rate": 7.198520525872487e-08, + "loss": 0.6983, + "step": 35007 + }, + { + "epoch": 1.28, + "learning_rate": 7.198154319405281e-08, + "loss": 0.6914, + "step": 35008 + }, + { + "epoch": 1.28, + "learning_rate": 7.197788112938075e-08, + "loss": 0.6936, + "step": 35009 + }, + { + "epoch": 1.28, + "learning_rate": 7.197421906470868e-08, + "loss": 0.7035, + "step": 35010 + }, + { + "epoch": 1.28, + "learning_rate": 7.197055700003662e-08, + "loss": 0.6986, + "step": 35011 + }, + { + "epoch": 1.28, + "learning_rate": 7.196689493536456e-08, + "loss": 0.7137, + "step": 35012 + }, + { + "epoch": 1.28, + "learning_rate": 7.19632328706925e-08, + "loss": 0.6894, + "step": 35013 + }, + { + "epoch": 1.28, + "learning_rate": 7.195957080602042e-08, + "loss": 0.697, + "step": 35014 + }, + { + "epoch": 1.28, + "learning_rate": 7.195590874134836e-08, + "loss": 0.6949, + "step": 35015 + }, + { + "epoch": 1.28, + "learning_rate": 7.19522466766763e-08, + "loss": 0.7025, + "step": 35016 + }, + { + "epoch": 1.28, + "learning_rate": 7.194858461200424e-08, + "loss": 0.7035, + "step": 35017 + }, + { + "epoch": 1.28, + "learning_rate": 7.194492254733218e-08, + "loss": 0.7055, + "step": 35018 + }, + { + "epoch": 1.28, + "learning_rate": 7.194126048266011e-08, + "loss": 0.6983, + "step": 35019 + }, + { + "epoch": 1.28, + "learning_rate": 7.193759841798805e-08, + "loss": 0.6983, + "step": 35020 + }, + { + "epoch": 1.28, + "learning_rate": 7.193393635331599e-08, + "loss": 0.6836, + "step": 35021 + }, + { + "epoch": 1.28, + "learning_rate": 7.193027428864393e-08, + "loss": 0.6787, + "step": 35022 + }, + { + "epoch": 1.28, + "learning_rate": 7.192661222397187e-08, + "loss": 0.6851, + "step": 35023 + }, + { + "epoch": 1.28, + "learning_rate": 7.192295015929981e-08, + "loss": 0.6794, + "step": 35024 + }, + { + "epoch": 1.28, + "learning_rate": 7.191928809462775e-08, + "loss": 0.701, + "step": 35025 + }, + { + "epoch": 1.28, + "learning_rate": 7.191562602995569e-08, + "loss": 0.6832, + "step": 35026 + }, + { + "epoch": 1.28, + "learning_rate": 7.191196396528363e-08, + "loss": 0.6962, + "step": 35027 + }, + { + "epoch": 1.28, + "learning_rate": 7.190830190061156e-08, + "loss": 0.706, + "step": 35028 + }, + { + "epoch": 1.28, + "learning_rate": 7.19046398359395e-08, + "loss": 0.7098, + "step": 35029 + }, + { + "epoch": 1.28, + "learning_rate": 7.190097777126744e-08, + "loss": 0.6914, + "step": 35030 + }, + { + "epoch": 1.28, + "learning_rate": 7.189731570659538e-08, + "loss": 0.6997, + "step": 35031 + }, + { + "epoch": 1.28, + "learning_rate": 7.189365364192332e-08, + "loss": 0.6925, + "step": 35032 + }, + { + "epoch": 1.28, + "learning_rate": 7.188999157725126e-08, + "loss": 0.7095, + "step": 35033 + }, + { + "epoch": 1.28, + "learning_rate": 7.18863295125792e-08, + "loss": 0.6894, + "step": 35034 + }, + { + "epoch": 1.28, + "learning_rate": 7.188266744790714e-08, + "loss": 0.6769, + "step": 35035 + }, + { + "epoch": 1.28, + "learning_rate": 7.187900538323506e-08, + "loss": 0.6963, + "step": 35036 + }, + { + "epoch": 1.28, + "learning_rate": 7.1875343318563e-08, + "loss": 0.6868, + "step": 35037 + }, + { + "epoch": 1.28, + "learning_rate": 7.187168125389093e-08, + "loss": 0.6829, + "step": 35038 + }, + { + "epoch": 1.28, + "learning_rate": 7.186801918921887e-08, + "loss": 0.6907, + "step": 35039 + }, + { + "epoch": 1.28, + "learning_rate": 7.186435712454681e-08, + "loss": 0.7038, + "step": 35040 + }, + { + "epoch": 1.28, + "learning_rate": 7.186069505987475e-08, + "loss": 0.6984, + "step": 35041 + }, + { + "epoch": 1.28, + "learning_rate": 7.185703299520269e-08, + "loss": 0.6907, + "step": 35042 + }, + { + "epoch": 1.28, + "learning_rate": 7.185337093053063e-08, + "loss": 0.699, + "step": 35043 + }, + { + "epoch": 1.28, + "learning_rate": 7.184970886585857e-08, + "loss": 0.6682, + "step": 35044 + }, + { + "epoch": 1.28, + "learning_rate": 7.184604680118651e-08, + "loss": 0.6825, + "step": 35045 + }, + { + "epoch": 1.28, + "learning_rate": 7.184238473651444e-08, + "loss": 0.6902, + "step": 35046 + }, + { + "epoch": 1.28, + "learning_rate": 7.183872267184238e-08, + "loss": 0.693, + "step": 35047 + }, + { + "epoch": 1.28, + "learning_rate": 7.183506060717032e-08, + "loss": 0.6762, + "step": 35048 + }, + { + "epoch": 1.28, + "learning_rate": 7.183139854249826e-08, + "loss": 0.6861, + "step": 35049 + }, + { + "epoch": 1.28, + "learning_rate": 7.18277364778262e-08, + "loss": 0.6963, + "step": 35050 + }, + { + "epoch": 1.28, + "learning_rate": 7.182407441315414e-08, + "loss": 0.691, + "step": 35051 + }, + { + "epoch": 1.28, + "learning_rate": 7.182041234848208e-08, + "loss": 0.6739, + "step": 35052 + }, + { + "epoch": 1.28, + "learning_rate": 7.181675028381002e-08, + "loss": 0.6979, + "step": 35053 + }, + { + "epoch": 1.28, + "learning_rate": 7.181308821913796e-08, + "loss": 0.6806, + "step": 35054 + }, + { + "epoch": 1.28, + "learning_rate": 7.180942615446588e-08, + "loss": 0.6935, + "step": 35055 + }, + { + "epoch": 1.28, + "learning_rate": 7.180576408979382e-08, + "loss": 0.6886, + "step": 35056 + }, + { + "epoch": 1.28, + "learning_rate": 7.180210202512176e-08, + "loss": 0.6834, + "step": 35057 + }, + { + "epoch": 1.28, + "learning_rate": 7.17984399604497e-08, + "loss": 0.6949, + "step": 35058 + }, + { + "epoch": 1.28, + "learning_rate": 7.179477789577764e-08, + "loss": 0.7, + "step": 35059 + }, + { + "epoch": 1.28, + "learning_rate": 7.179111583110557e-08, + "loss": 0.6839, + "step": 35060 + }, + { + "epoch": 1.28, + "learning_rate": 7.178745376643351e-08, + "loss": 0.7021, + "step": 35061 + }, + { + "epoch": 1.28, + "learning_rate": 7.178379170176145e-08, + "loss": 0.6851, + "step": 35062 + }, + { + "epoch": 1.28, + "learning_rate": 7.178012963708939e-08, + "loss": 0.6981, + "step": 35063 + }, + { + "epoch": 1.28, + "learning_rate": 7.177646757241732e-08, + "loss": 0.6905, + "step": 35064 + }, + { + "epoch": 1.28, + "learning_rate": 7.177280550774526e-08, + "loss": 0.7013, + "step": 35065 + }, + { + "epoch": 1.28, + "learning_rate": 7.17691434430732e-08, + "loss": 0.6868, + "step": 35066 + }, + { + "epoch": 1.28, + "learning_rate": 7.176548137840114e-08, + "loss": 0.7009, + "step": 35067 + }, + { + "epoch": 1.28, + "learning_rate": 7.176181931372908e-08, + "loss": 0.6832, + "step": 35068 + }, + { + "epoch": 1.28, + "learning_rate": 7.175815724905702e-08, + "loss": 0.6966, + "step": 35069 + }, + { + "epoch": 1.28, + "learning_rate": 7.175449518438496e-08, + "loss": 0.6905, + "step": 35070 + }, + { + "epoch": 1.28, + "learning_rate": 7.17508331197129e-08, + "loss": 0.6803, + "step": 35071 + }, + { + "epoch": 1.28, + "learning_rate": 7.174717105504083e-08, + "loss": 0.6878, + "step": 35072 + }, + { + "epoch": 1.28, + "learning_rate": 7.174350899036876e-08, + "loss": 0.6753, + "step": 35073 + }, + { + "epoch": 1.28, + "learning_rate": 7.17398469256967e-08, + "loss": 0.6835, + "step": 35074 + }, + { + "epoch": 1.28, + "learning_rate": 7.173618486102464e-08, + "loss": 0.6996, + "step": 35075 + }, + { + "epoch": 1.28, + "learning_rate": 7.173252279635258e-08, + "loss": 0.7078, + "step": 35076 + }, + { + "epoch": 1.28, + "learning_rate": 7.172886073168052e-08, + "loss": 0.6811, + "step": 35077 + }, + { + "epoch": 1.28, + "learning_rate": 7.172519866700846e-08, + "loss": 0.6802, + "step": 35078 + }, + { + "epoch": 1.28, + "learning_rate": 7.17215366023364e-08, + "loss": 0.6719, + "step": 35079 + }, + { + "epoch": 1.28, + "learning_rate": 7.171787453766434e-08, + "loss": 0.6963, + "step": 35080 + }, + { + "epoch": 1.28, + "learning_rate": 7.171421247299228e-08, + "loss": 0.6869, + "step": 35081 + }, + { + "epoch": 1.28, + "learning_rate": 7.171055040832021e-08, + "loss": 0.7017, + "step": 35082 + }, + { + "epoch": 1.28, + "learning_rate": 7.170688834364815e-08, + "loss": 0.6917, + "step": 35083 + }, + { + "epoch": 1.28, + "learning_rate": 7.170322627897609e-08, + "loss": 0.6931, + "step": 35084 + }, + { + "epoch": 1.28, + "learning_rate": 7.169956421430401e-08, + "loss": 0.6893, + "step": 35085 + }, + { + "epoch": 1.28, + "learning_rate": 7.169590214963195e-08, + "loss": 0.6957, + "step": 35086 + }, + { + "epoch": 1.28, + "learning_rate": 7.16922400849599e-08, + "loss": 0.6913, + "step": 35087 + }, + { + "epoch": 1.28, + "learning_rate": 7.168857802028783e-08, + "loss": 0.6734, + "step": 35088 + }, + { + "epoch": 1.28, + "learning_rate": 7.168491595561577e-08, + "loss": 0.6841, + "step": 35089 + }, + { + "epoch": 1.28, + "learning_rate": 7.168125389094371e-08, + "loss": 0.6876, + "step": 35090 + }, + { + "epoch": 1.28, + "learning_rate": 7.167759182627164e-08, + "loss": 0.6943, + "step": 35091 + }, + { + "epoch": 1.28, + "learning_rate": 7.167392976159958e-08, + "loss": 0.7017, + "step": 35092 + }, + { + "epoch": 1.28, + "learning_rate": 7.167026769692752e-08, + "loss": 0.6921, + "step": 35093 + }, + { + "epoch": 1.28, + "learning_rate": 7.166660563225546e-08, + "loss": 0.6945, + "step": 35094 + }, + { + "epoch": 1.28, + "learning_rate": 7.16629435675834e-08, + "loss": 0.6822, + "step": 35095 + }, + { + "epoch": 1.28, + "learning_rate": 7.165928150291134e-08, + "loss": 0.6924, + "step": 35096 + }, + { + "epoch": 1.28, + "learning_rate": 7.165561943823928e-08, + "loss": 0.6917, + "step": 35097 + }, + { + "epoch": 1.28, + "learning_rate": 7.165195737356722e-08, + "loss": 0.7025, + "step": 35098 + }, + { + "epoch": 1.28, + "learning_rate": 7.164829530889516e-08, + "loss": 0.697, + "step": 35099 + }, + { + "epoch": 1.28, + "learning_rate": 7.164463324422309e-08, + "loss": 0.7019, + "step": 35100 + }, + { + "epoch": 1.28, + "learning_rate": 7.164097117955103e-08, + "loss": 0.6937, + "step": 35101 + }, + { + "epoch": 1.28, + "learning_rate": 7.163730911487897e-08, + "loss": 0.6868, + "step": 35102 + }, + { + "epoch": 1.28, + "learning_rate": 7.16336470502069e-08, + "loss": 0.6815, + "step": 35103 + }, + { + "epoch": 1.28, + "learning_rate": 7.162998498553485e-08, + "loss": 0.6892, + "step": 35104 + }, + { + "epoch": 1.28, + "learning_rate": 7.162632292086279e-08, + "loss": 0.7042, + "step": 35105 + }, + { + "epoch": 1.28, + "learning_rate": 7.162266085619073e-08, + "loss": 0.6832, + "step": 35106 + }, + { + "epoch": 1.28, + "learning_rate": 7.161899879151865e-08, + "loss": 0.6979, + "step": 35107 + }, + { + "epoch": 1.28, + "learning_rate": 7.161533672684659e-08, + "loss": 0.6943, + "step": 35108 + }, + { + "epoch": 1.28, + "learning_rate": 7.161167466217452e-08, + "loss": 0.691, + "step": 35109 + }, + { + "epoch": 1.28, + "learning_rate": 7.160801259750246e-08, + "loss": 0.701, + "step": 35110 + }, + { + "epoch": 1.28, + "learning_rate": 7.16043505328304e-08, + "loss": 0.697, + "step": 35111 + }, + { + "epoch": 1.28, + "learning_rate": 7.160068846815834e-08, + "loss": 0.6934, + "step": 35112 + }, + { + "epoch": 1.28, + "learning_rate": 7.159702640348628e-08, + "loss": 0.6928, + "step": 35113 + }, + { + "epoch": 1.28, + "learning_rate": 7.159336433881422e-08, + "loss": 0.6994, + "step": 35114 + }, + { + "epoch": 1.28, + "learning_rate": 7.158970227414216e-08, + "loss": 0.6993, + "step": 35115 + }, + { + "epoch": 1.28, + "learning_rate": 7.15860402094701e-08, + "loss": 0.6929, + "step": 35116 + }, + { + "epoch": 1.28, + "learning_rate": 7.158237814479804e-08, + "loss": 0.7045, + "step": 35117 + }, + { + "epoch": 1.28, + "learning_rate": 7.157871608012597e-08, + "loss": 0.6864, + "step": 35118 + }, + { + "epoch": 1.28, + "learning_rate": 7.15750540154539e-08, + "loss": 0.7051, + "step": 35119 + }, + { + "epoch": 1.28, + "learning_rate": 7.157139195078185e-08, + "loss": 0.7018, + "step": 35120 + }, + { + "epoch": 1.28, + "learning_rate": 7.156772988610979e-08, + "loss": 0.6866, + "step": 35121 + }, + { + "epoch": 1.29, + "learning_rate": 7.156406782143773e-08, + "loss": 0.6987, + "step": 35122 + }, + { + "epoch": 1.29, + "learning_rate": 7.156040575676567e-08, + "loss": 0.7166, + "step": 35123 + }, + { + "epoch": 1.29, + "learning_rate": 7.15567436920936e-08, + "loss": 0.7022, + "step": 35124 + }, + { + "epoch": 1.29, + "learning_rate": 7.155308162742155e-08, + "loss": 0.6883, + "step": 35125 + }, + { + "epoch": 1.29, + "learning_rate": 7.154941956274949e-08, + "loss": 0.7046, + "step": 35126 + }, + { + "epoch": 1.29, + "learning_rate": 7.154575749807741e-08, + "loss": 0.6998, + "step": 35127 + }, + { + "epoch": 1.29, + "learning_rate": 7.154209543340535e-08, + "loss": 0.6976, + "step": 35128 + }, + { + "epoch": 1.29, + "learning_rate": 7.153843336873329e-08, + "loss": 0.7022, + "step": 35129 + }, + { + "epoch": 1.29, + "learning_rate": 7.153477130406123e-08, + "loss": 0.6804, + "step": 35130 + }, + { + "epoch": 1.29, + "learning_rate": 7.153110923938916e-08, + "loss": 0.6965, + "step": 35131 + }, + { + "epoch": 1.29, + "learning_rate": 7.15274471747171e-08, + "loss": 0.6799, + "step": 35132 + }, + { + "epoch": 1.29, + "learning_rate": 7.152378511004504e-08, + "loss": 0.689, + "step": 35133 + }, + { + "epoch": 1.29, + "learning_rate": 7.152012304537298e-08, + "loss": 0.6889, + "step": 35134 + }, + { + "epoch": 1.29, + "learning_rate": 7.151646098070092e-08, + "loss": 0.6976, + "step": 35135 + }, + { + "epoch": 1.29, + "learning_rate": 7.151279891602884e-08, + "loss": 0.6944, + "step": 35136 + }, + { + "epoch": 1.29, + "learning_rate": 7.150913685135678e-08, + "loss": 0.6852, + "step": 35137 + }, + { + "epoch": 1.29, + "learning_rate": 7.150547478668472e-08, + "loss": 0.6897, + "step": 35138 + }, + { + "epoch": 1.29, + "learning_rate": 7.150181272201266e-08, + "loss": 0.6932, + "step": 35139 + }, + { + "epoch": 1.29, + "learning_rate": 7.14981506573406e-08, + "loss": 0.6903, + "step": 35140 + }, + { + "epoch": 1.29, + "learning_rate": 7.149448859266854e-08, + "loss": 0.6857, + "step": 35141 + }, + { + "epoch": 1.29, + "learning_rate": 7.149082652799648e-08, + "loss": 0.6861, + "step": 35142 + }, + { + "epoch": 1.29, + "learning_rate": 7.148716446332442e-08, + "loss": 0.709, + "step": 35143 + }, + { + "epoch": 1.29, + "learning_rate": 7.148350239865236e-08, + "loss": 0.7116, + "step": 35144 + }, + { + "epoch": 1.29, + "learning_rate": 7.147984033398029e-08, + "loss": 0.6878, + "step": 35145 + }, + { + "epoch": 1.29, + "learning_rate": 7.147617826930823e-08, + "loss": 0.6863, + "step": 35146 + }, + { + "epoch": 1.29, + "learning_rate": 7.147251620463617e-08, + "loss": 0.7079, + "step": 35147 + }, + { + "epoch": 1.29, + "learning_rate": 7.146885413996411e-08, + "loss": 0.6948, + "step": 35148 + }, + { + "epoch": 1.29, + "learning_rate": 7.146519207529205e-08, + "loss": 0.7133, + "step": 35149 + }, + { + "epoch": 1.29, + "learning_rate": 7.146153001061999e-08, + "loss": 0.7052, + "step": 35150 + }, + { + "epoch": 1.29, + "learning_rate": 7.145786794594793e-08, + "loss": 0.6993, + "step": 35151 + }, + { + "epoch": 1.29, + "learning_rate": 7.145420588127587e-08, + "loss": 0.6982, + "step": 35152 + }, + { + "epoch": 1.29, + "learning_rate": 7.14505438166038e-08, + "loss": 0.6892, + "step": 35153 + }, + { + "epoch": 1.29, + "learning_rate": 7.144688175193174e-08, + "loss": 0.6948, + "step": 35154 + }, + { + "epoch": 1.29, + "learning_rate": 7.144321968725968e-08, + "loss": 0.6961, + "step": 35155 + }, + { + "epoch": 1.29, + "learning_rate": 7.14395576225876e-08, + "loss": 0.7007, + "step": 35156 + }, + { + "epoch": 1.29, + "learning_rate": 7.143589555791554e-08, + "loss": 0.6973, + "step": 35157 + }, + { + "epoch": 1.29, + "learning_rate": 7.143223349324348e-08, + "loss": 0.6872, + "step": 35158 + }, + { + "epoch": 1.29, + "learning_rate": 7.142857142857142e-08, + "loss": 0.6969, + "step": 35159 + }, + { + "epoch": 1.29, + "learning_rate": 7.142490936389936e-08, + "loss": 0.6834, + "step": 35160 + }, + { + "epoch": 1.29, + "learning_rate": 7.14212472992273e-08, + "loss": 0.6934, + "step": 35161 + }, + { + "epoch": 1.29, + "learning_rate": 7.141758523455524e-08, + "loss": 0.6814, + "step": 35162 + }, + { + "epoch": 1.29, + "learning_rate": 7.141392316988317e-08, + "loss": 0.7107, + "step": 35163 + }, + { + "epoch": 1.29, + "learning_rate": 7.141026110521111e-08, + "loss": 0.6918, + "step": 35164 + }, + { + "epoch": 1.29, + "learning_rate": 7.140659904053905e-08, + "loss": 0.7076, + "step": 35165 + }, + { + "epoch": 1.29, + "learning_rate": 7.140293697586699e-08, + "loss": 0.6924, + "step": 35166 + }, + { + "epoch": 1.29, + "learning_rate": 7.139927491119493e-08, + "loss": 0.6926, + "step": 35167 + }, + { + "epoch": 1.29, + "learning_rate": 7.139561284652287e-08, + "loss": 0.6839, + "step": 35168 + }, + { + "epoch": 1.29, + "learning_rate": 7.139195078185081e-08, + "loss": 0.6935, + "step": 35169 + }, + { + "epoch": 1.29, + "learning_rate": 7.138828871717875e-08, + "loss": 0.7164, + "step": 35170 + }, + { + "epoch": 1.29, + "learning_rate": 7.138462665250669e-08, + "loss": 0.6926, + "step": 35171 + }, + { + "epoch": 1.29, + "learning_rate": 7.138096458783462e-08, + "loss": 0.6863, + "step": 35172 + }, + { + "epoch": 1.29, + "learning_rate": 7.137730252316256e-08, + "loss": 0.7, + "step": 35173 + }, + { + "epoch": 1.29, + "learning_rate": 7.13736404584905e-08, + "loss": 0.6933, + "step": 35174 + }, + { + "epoch": 1.29, + "learning_rate": 7.136997839381844e-08, + "loss": 0.6836, + "step": 35175 + }, + { + "epoch": 1.29, + "learning_rate": 7.136631632914638e-08, + "loss": 0.6801, + "step": 35176 + }, + { + "epoch": 1.29, + "learning_rate": 7.13626542644743e-08, + "loss": 0.6954, + "step": 35177 + }, + { + "epoch": 1.29, + "learning_rate": 7.135899219980224e-08, + "loss": 0.7001, + "step": 35178 + }, + { + "epoch": 1.29, + "learning_rate": 7.135533013513018e-08, + "loss": 0.7062, + "step": 35179 + }, + { + "epoch": 1.29, + "learning_rate": 7.135166807045812e-08, + "loss": 0.6902, + "step": 35180 + }, + { + "epoch": 1.29, + "learning_rate": 7.134800600578605e-08, + "loss": 0.6973, + "step": 35181 + }, + { + "epoch": 1.29, + "learning_rate": 7.134434394111399e-08, + "loss": 0.6996, + "step": 35182 + }, + { + "epoch": 1.29, + "learning_rate": 7.134068187644193e-08, + "loss": 0.696, + "step": 35183 + }, + { + "epoch": 1.29, + "learning_rate": 7.133701981176987e-08, + "loss": 0.681, + "step": 35184 + }, + { + "epoch": 1.29, + "learning_rate": 7.133335774709781e-08, + "loss": 0.6784, + "step": 35185 + }, + { + "epoch": 1.29, + "learning_rate": 7.132969568242575e-08, + "loss": 0.6852, + "step": 35186 + }, + { + "epoch": 1.29, + "learning_rate": 7.132603361775369e-08, + "loss": 0.6898, + "step": 35187 + }, + { + "epoch": 1.29, + "learning_rate": 7.132237155308163e-08, + "loss": 0.6925, + "step": 35188 + }, + { + "epoch": 1.29, + "learning_rate": 7.131870948840957e-08, + "loss": 0.6968, + "step": 35189 + }, + { + "epoch": 1.29, + "learning_rate": 7.13150474237375e-08, + "loss": 0.6989, + "step": 35190 + }, + { + "epoch": 1.29, + "learning_rate": 7.131138535906543e-08, + "loss": 0.7007, + "step": 35191 + }, + { + "epoch": 1.29, + "learning_rate": 7.130772329439337e-08, + "loss": 0.6969, + "step": 35192 + }, + { + "epoch": 1.29, + "learning_rate": 7.130406122972131e-08, + "loss": 0.6938, + "step": 35193 + }, + { + "epoch": 1.29, + "learning_rate": 7.130039916504925e-08, + "loss": 0.6955, + "step": 35194 + }, + { + "epoch": 1.29, + "learning_rate": 7.12967371003772e-08, + "loss": 0.6959, + "step": 35195 + }, + { + "epoch": 1.29, + "learning_rate": 7.129307503570513e-08, + "loss": 0.6986, + "step": 35196 + }, + { + "epoch": 1.29, + "learning_rate": 7.128941297103307e-08, + "loss": 0.6843, + "step": 35197 + }, + { + "epoch": 1.29, + "learning_rate": 7.128575090636101e-08, + "loss": 0.6776, + "step": 35198 + }, + { + "epoch": 1.29, + "learning_rate": 7.128208884168894e-08, + "loss": 0.6945, + "step": 35199 + }, + { + "epoch": 1.29, + "learning_rate": 7.127842677701688e-08, + "loss": 0.695, + "step": 35200 + }, + { + "epoch": 1.29, + "learning_rate": 7.127476471234482e-08, + "loss": 0.693, + "step": 35201 + }, + { + "epoch": 1.29, + "learning_rate": 7.127110264767275e-08, + "loss": 0.6885, + "step": 35202 + }, + { + "epoch": 1.29, + "learning_rate": 7.126744058300069e-08, + "loss": 0.6919, + "step": 35203 + }, + { + "epoch": 1.29, + "learning_rate": 7.126377851832863e-08, + "loss": 0.6914, + "step": 35204 + }, + { + "epoch": 1.29, + "learning_rate": 7.126011645365657e-08, + "loss": 0.6947, + "step": 35205 + }, + { + "epoch": 1.29, + "learning_rate": 7.125645438898451e-08, + "loss": 0.6948, + "step": 35206 + }, + { + "epoch": 1.29, + "learning_rate": 7.125279232431245e-08, + "loss": 0.6955, + "step": 35207 + }, + { + "epoch": 1.29, + "learning_rate": 7.124913025964037e-08, + "loss": 0.6987, + "step": 35208 + }, + { + "epoch": 1.29, + "learning_rate": 7.124546819496831e-08, + "loss": 0.7088, + "step": 35209 + }, + { + "epoch": 1.29, + "learning_rate": 7.124180613029625e-08, + "loss": 0.6876, + "step": 35210 + }, + { + "epoch": 1.29, + "learning_rate": 7.12381440656242e-08, + "loss": 0.6863, + "step": 35211 + }, + { + "epoch": 1.29, + "learning_rate": 7.123448200095213e-08, + "loss": 0.6934, + "step": 35212 + }, + { + "epoch": 1.29, + "learning_rate": 7.123081993628007e-08, + "loss": 0.6833, + "step": 35213 + }, + { + "epoch": 1.29, + "learning_rate": 7.122715787160801e-08, + "loss": 0.6904, + "step": 35214 + }, + { + "epoch": 1.29, + "learning_rate": 7.122349580693595e-08, + "loss": 0.6899, + "step": 35215 + }, + { + "epoch": 1.29, + "learning_rate": 7.12198337422639e-08, + "loss": 0.696, + "step": 35216 + }, + { + "epoch": 1.29, + "learning_rate": 7.121617167759182e-08, + "loss": 0.6943, + "step": 35217 + }, + { + "epoch": 1.29, + "learning_rate": 7.121250961291976e-08, + "loss": 0.6883, + "step": 35218 + }, + { + "epoch": 1.29, + "learning_rate": 7.12088475482477e-08, + "loss": 0.7052, + "step": 35219 + }, + { + "epoch": 1.29, + "learning_rate": 7.120518548357564e-08, + "loss": 0.678, + "step": 35220 + }, + { + "epoch": 1.29, + "learning_rate": 7.120152341890358e-08, + "loss": 0.682, + "step": 35221 + }, + { + "epoch": 1.29, + "learning_rate": 7.119786135423152e-08, + "loss": 0.7073, + "step": 35222 + }, + { + "epoch": 1.29, + "learning_rate": 7.119419928955946e-08, + "loss": 0.6995, + "step": 35223 + }, + { + "epoch": 1.29, + "learning_rate": 7.119053722488739e-08, + "loss": 0.7058, + "step": 35224 + }, + { + "epoch": 1.29, + "learning_rate": 7.118687516021533e-08, + "loss": 0.7052, + "step": 35225 + }, + { + "epoch": 1.29, + "learning_rate": 7.118321309554327e-08, + "loss": 0.7098, + "step": 35226 + }, + { + "epoch": 1.29, + "learning_rate": 7.117955103087119e-08, + "loss": 0.6957, + "step": 35227 + }, + { + "epoch": 1.29, + "learning_rate": 7.117588896619913e-08, + "loss": 0.6804, + "step": 35228 + }, + { + "epoch": 1.29, + "learning_rate": 7.117222690152707e-08, + "loss": 0.6961, + "step": 35229 + }, + { + "epoch": 1.29, + "learning_rate": 7.116856483685501e-08, + "loss": 0.6928, + "step": 35230 + }, + { + "epoch": 1.29, + "learning_rate": 7.116490277218295e-08, + "loss": 0.6846, + "step": 35231 + }, + { + "epoch": 1.29, + "learning_rate": 7.116124070751089e-08, + "loss": 0.6898, + "step": 35232 + }, + { + "epoch": 1.29, + "learning_rate": 7.115757864283883e-08, + "loss": 0.689, + "step": 35233 + }, + { + "epoch": 1.29, + "learning_rate": 7.115391657816677e-08, + "loss": 0.6895, + "step": 35234 + }, + { + "epoch": 1.29, + "learning_rate": 7.11502545134947e-08, + "loss": 0.7073, + "step": 35235 + }, + { + "epoch": 1.29, + "learning_rate": 7.114659244882264e-08, + "loss": 0.704, + "step": 35236 + }, + { + "epoch": 1.29, + "learning_rate": 7.114293038415058e-08, + "loss": 0.6821, + "step": 35237 + }, + { + "epoch": 1.29, + "learning_rate": 7.113926831947852e-08, + "loss": 0.696, + "step": 35238 + }, + { + "epoch": 1.29, + "learning_rate": 7.113560625480646e-08, + "loss": 0.6838, + "step": 35239 + }, + { + "epoch": 1.29, + "learning_rate": 7.11319441901344e-08, + "loss": 0.6865, + "step": 35240 + }, + { + "epoch": 1.29, + "learning_rate": 7.112828212546234e-08, + "loss": 0.6981, + "step": 35241 + }, + { + "epoch": 1.29, + "learning_rate": 7.112462006079028e-08, + "loss": 0.6937, + "step": 35242 + }, + { + "epoch": 1.29, + "learning_rate": 7.11209579961182e-08, + "loss": 0.695, + "step": 35243 + }, + { + "epoch": 1.29, + "learning_rate": 7.111729593144615e-08, + "loss": 0.7088, + "step": 35244 + }, + { + "epoch": 1.29, + "learning_rate": 7.111363386677409e-08, + "loss": 0.6741, + "step": 35245 + }, + { + "epoch": 1.29, + "learning_rate": 7.110997180210203e-08, + "loss": 0.6878, + "step": 35246 + }, + { + "epoch": 1.29, + "learning_rate": 7.110630973742997e-08, + "loss": 0.6918, + "step": 35247 + }, + { + "epoch": 1.29, + "learning_rate": 7.110264767275789e-08, + "loss": 0.6847, + "step": 35248 + }, + { + "epoch": 1.29, + "learning_rate": 7.109898560808583e-08, + "loss": 0.6986, + "step": 35249 + }, + { + "epoch": 1.29, + "learning_rate": 7.109532354341377e-08, + "loss": 0.7047, + "step": 35250 + }, + { + "epoch": 1.29, + "learning_rate": 7.109166147874171e-08, + "loss": 0.7012, + "step": 35251 + }, + { + "epoch": 1.29, + "learning_rate": 7.108799941406964e-08, + "loss": 0.7041, + "step": 35252 + }, + { + "epoch": 1.29, + "learning_rate": 7.108433734939758e-08, + "loss": 0.6764, + "step": 35253 + }, + { + "epoch": 1.29, + "learning_rate": 7.108067528472552e-08, + "loss": 0.691, + "step": 35254 + }, + { + "epoch": 1.29, + "learning_rate": 7.107701322005346e-08, + "loss": 0.697, + "step": 35255 + }, + { + "epoch": 1.29, + "learning_rate": 7.10733511553814e-08, + "loss": 0.6964, + "step": 35256 + }, + { + "epoch": 1.29, + "learning_rate": 7.106968909070934e-08, + "loss": 0.6924, + "step": 35257 + }, + { + "epoch": 1.29, + "learning_rate": 7.106602702603728e-08, + "loss": 0.6923, + "step": 35258 + }, + { + "epoch": 1.29, + "learning_rate": 7.106236496136522e-08, + "loss": 0.697, + "step": 35259 + }, + { + "epoch": 1.29, + "learning_rate": 7.105870289669316e-08, + "loss": 0.6955, + "step": 35260 + }, + { + "epoch": 1.29, + "learning_rate": 7.105504083202108e-08, + "loss": 0.6828, + "step": 35261 + }, + { + "epoch": 1.29, + "learning_rate": 7.105137876734902e-08, + "loss": 0.6991, + "step": 35262 + }, + { + "epoch": 1.29, + "learning_rate": 7.104771670267696e-08, + "loss": 0.6974, + "step": 35263 + }, + { + "epoch": 1.29, + "learning_rate": 7.10440546380049e-08, + "loss": 0.6873, + "step": 35264 + }, + { + "epoch": 1.29, + "learning_rate": 7.104039257333284e-08, + "loss": 0.6839, + "step": 35265 + }, + { + "epoch": 1.29, + "learning_rate": 7.103673050866078e-08, + "loss": 0.6927, + "step": 35266 + }, + { + "epoch": 1.29, + "learning_rate": 7.103306844398872e-08, + "loss": 0.6963, + "step": 35267 + }, + { + "epoch": 1.29, + "learning_rate": 7.102940637931666e-08, + "loss": 0.7, + "step": 35268 + }, + { + "epoch": 1.29, + "learning_rate": 7.10257443146446e-08, + "loss": 0.6972, + "step": 35269 + }, + { + "epoch": 1.29, + "learning_rate": 7.102208224997253e-08, + "loss": 0.6799, + "step": 35270 + }, + { + "epoch": 1.29, + "learning_rate": 7.101842018530047e-08, + "loss": 0.6811, + "step": 35271 + }, + { + "epoch": 1.29, + "learning_rate": 7.101475812062841e-08, + "loss": 0.6926, + "step": 35272 + }, + { + "epoch": 1.29, + "learning_rate": 7.101109605595634e-08, + "loss": 0.6802, + "step": 35273 + }, + { + "epoch": 1.29, + "learning_rate": 7.100743399128428e-08, + "loss": 0.6888, + "step": 35274 + }, + { + "epoch": 1.29, + "learning_rate": 7.100377192661222e-08, + "loss": 0.6917, + "step": 35275 + }, + { + "epoch": 1.29, + "learning_rate": 7.100010986194016e-08, + "loss": 0.6866, + "step": 35276 + }, + { + "epoch": 1.29, + "learning_rate": 7.09964477972681e-08, + "loss": 0.6799, + "step": 35277 + }, + { + "epoch": 1.29, + "learning_rate": 7.099278573259604e-08, + "loss": 0.6811, + "step": 35278 + }, + { + "epoch": 1.29, + "learning_rate": 7.098912366792396e-08, + "loss": 0.691, + "step": 35279 + }, + { + "epoch": 1.29, + "learning_rate": 7.09854616032519e-08, + "loss": 0.7014, + "step": 35280 + }, + { + "epoch": 1.29, + "learning_rate": 7.098179953857984e-08, + "loss": 0.6855, + "step": 35281 + }, + { + "epoch": 1.29, + "learning_rate": 7.097813747390778e-08, + "loss": 0.6957, + "step": 35282 + }, + { + "epoch": 1.29, + "learning_rate": 7.097447540923572e-08, + "loss": 0.6907, + "step": 35283 + }, + { + "epoch": 1.29, + "learning_rate": 7.097081334456366e-08, + "loss": 0.6884, + "step": 35284 + }, + { + "epoch": 1.29, + "learning_rate": 7.09671512798916e-08, + "loss": 0.6961, + "step": 35285 + }, + { + "epoch": 1.29, + "learning_rate": 7.096348921521954e-08, + "loss": 0.7076, + "step": 35286 + }, + { + "epoch": 1.29, + "learning_rate": 7.095982715054748e-08, + "loss": 0.7046, + "step": 35287 + }, + { + "epoch": 1.29, + "learning_rate": 7.095616508587541e-08, + "loss": 0.6861, + "step": 35288 + }, + { + "epoch": 1.29, + "learning_rate": 7.095250302120335e-08, + "loss": 0.6885, + "step": 35289 + }, + { + "epoch": 1.29, + "learning_rate": 7.094884095653129e-08, + "loss": 0.6923, + "step": 35290 + }, + { + "epoch": 1.29, + "learning_rate": 7.094517889185923e-08, + "loss": 0.6922, + "step": 35291 + }, + { + "epoch": 1.29, + "learning_rate": 7.094151682718717e-08, + "loss": 0.7015, + "step": 35292 + }, + { + "epoch": 1.29, + "learning_rate": 7.093785476251511e-08, + "loss": 0.688, + "step": 35293 + }, + { + "epoch": 1.29, + "learning_rate": 7.093419269784305e-08, + "loss": 0.6865, + "step": 35294 + }, + { + "epoch": 1.29, + "learning_rate": 7.093053063317098e-08, + "loss": 0.6984, + "step": 35295 + }, + { + "epoch": 1.29, + "learning_rate": 7.092686856849892e-08, + "loss": 0.6984, + "step": 35296 + }, + { + "epoch": 1.29, + "learning_rate": 7.092320650382684e-08, + "loss": 0.6841, + "step": 35297 + }, + { + "epoch": 1.29, + "learning_rate": 7.091954443915478e-08, + "loss": 0.693, + "step": 35298 + }, + { + "epoch": 1.29, + "learning_rate": 7.091588237448272e-08, + "loss": 0.7054, + "step": 35299 + }, + { + "epoch": 1.29, + "learning_rate": 7.091222030981066e-08, + "loss": 0.6886, + "step": 35300 + }, + { + "epoch": 1.29, + "learning_rate": 7.09085582451386e-08, + "loss": 0.6954, + "step": 35301 + }, + { + "epoch": 1.29, + "learning_rate": 7.090489618046654e-08, + "loss": 0.7022, + "step": 35302 + }, + { + "epoch": 1.29, + "learning_rate": 7.090123411579448e-08, + "loss": 0.7036, + "step": 35303 + }, + { + "epoch": 1.29, + "learning_rate": 7.089757205112242e-08, + "loss": 0.6806, + "step": 35304 + }, + { + "epoch": 1.29, + "learning_rate": 7.089390998645036e-08, + "loss": 0.6929, + "step": 35305 + }, + { + "epoch": 1.29, + "learning_rate": 7.089024792177829e-08, + "loss": 0.7035, + "step": 35306 + }, + { + "epoch": 1.29, + "learning_rate": 7.088658585710623e-08, + "loss": 0.7032, + "step": 35307 + }, + { + "epoch": 1.29, + "learning_rate": 7.088292379243417e-08, + "loss": 0.6987, + "step": 35308 + }, + { + "epoch": 1.29, + "learning_rate": 7.087926172776211e-08, + "loss": 0.6963, + "step": 35309 + }, + { + "epoch": 1.29, + "learning_rate": 7.087559966309005e-08, + "loss": 0.7054, + "step": 35310 + }, + { + "epoch": 1.29, + "learning_rate": 7.087193759841799e-08, + "loss": 0.6862, + "step": 35311 + }, + { + "epoch": 1.29, + "learning_rate": 7.086827553374593e-08, + "loss": 0.6908, + "step": 35312 + }, + { + "epoch": 1.29, + "learning_rate": 7.086461346907387e-08, + "loss": 0.6917, + "step": 35313 + }, + { + "epoch": 1.29, + "learning_rate": 7.086095140440181e-08, + "loss": 0.6857, + "step": 35314 + }, + { + "epoch": 1.29, + "learning_rate": 7.085728933972973e-08, + "loss": 0.6938, + "step": 35315 + }, + { + "epoch": 1.29, + "learning_rate": 7.085362727505767e-08, + "loss": 0.6931, + "step": 35316 + }, + { + "epoch": 1.29, + "learning_rate": 7.084996521038561e-08, + "loss": 0.6983, + "step": 35317 + }, + { + "epoch": 1.29, + "learning_rate": 7.084630314571355e-08, + "loss": 0.6928, + "step": 35318 + }, + { + "epoch": 1.29, + "learning_rate": 7.084264108104148e-08, + "loss": 0.6928, + "step": 35319 + }, + { + "epoch": 1.29, + "learning_rate": 7.083897901636942e-08, + "loss": 0.6907, + "step": 35320 + }, + { + "epoch": 1.29, + "learning_rate": 7.083531695169736e-08, + "loss": 0.688, + "step": 35321 + }, + { + "epoch": 1.29, + "learning_rate": 7.08316548870253e-08, + "loss": 0.693, + "step": 35322 + }, + { + "epoch": 1.29, + "learning_rate": 7.082799282235324e-08, + "loss": 0.6973, + "step": 35323 + }, + { + "epoch": 1.29, + "learning_rate": 7.082433075768117e-08, + "loss": 0.7056, + "step": 35324 + }, + { + "epoch": 1.29, + "learning_rate": 7.082066869300911e-08, + "loss": 0.6822, + "step": 35325 + }, + { + "epoch": 1.29, + "learning_rate": 7.081700662833705e-08, + "loss": 0.6879, + "step": 35326 + }, + { + "epoch": 1.29, + "learning_rate": 7.081334456366499e-08, + "loss": 0.6726, + "step": 35327 + }, + { + "epoch": 1.29, + "learning_rate": 7.080968249899293e-08, + "loss": 0.6997, + "step": 35328 + }, + { + "epoch": 1.29, + "learning_rate": 7.080602043432087e-08, + "loss": 0.6777, + "step": 35329 + }, + { + "epoch": 1.29, + "learning_rate": 7.080235836964881e-08, + "loss": 0.7124, + "step": 35330 + }, + { + "epoch": 1.29, + "learning_rate": 7.079869630497675e-08, + "loss": 0.706, + "step": 35331 + }, + { + "epoch": 1.29, + "learning_rate": 7.079503424030469e-08, + "loss": 0.6884, + "step": 35332 + }, + { + "epoch": 1.29, + "learning_rate": 7.079137217563261e-08, + "loss": 0.6999, + "step": 35333 + }, + { + "epoch": 1.29, + "learning_rate": 7.078771011096055e-08, + "loss": 0.6836, + "step": 35334 + }, + { + "epoch": 1.29, + "learning_rate": 7.07840480462885e-08, + "loss": 0.6998, + "step": 35335 + }, + { + "epoch": 1.29, + "learning_rate": 7.078038598161643e-08, + "loss": 0.69, + "step": 35336 + }, + { + "epoch": 1.29, + "learning_rate": 7.077672391694437e-08, + "loss": 0.6898, + "step": 35337 + }, + { + "epoch": 1.29, + "learning_rate": 7.077306185227231e-08, + "loss": 0.6974, + "step": 35338 + }, + { + "epoch": 1.29, + "learning_rate": 7.076939978760025e-08, + "loss": 0.6916, + "step": 35339 + }, + { + "epoch": 1.29, + "learning_rate": 7.076573772292819e-08, + "loss": 0.682, + "step": 35340 + }, + { + "epoch": 1.29, + "learning_rate": 7.076207565825612e-08, + "loss": 0.6982, + "step": 35341 + }, + { + "epoch": 1.29, + "learning_rate": 7.075841359358406e-08, + "loss": 0.7139, + "step": 35342 + }, + { + "epoch": 1.29, + "learning_rate": 7.0754751528912e-08, + "loss": 0.6962, + "step": 35343 + }, + { + "epoch": 1.29, + "learning_rate": 7.075108946423993e-08, + "loss": 0.6961, + "step": 35344 + }, + { + "epoch": 1.29, + "learning_rate": 7.074742739956787e-08, + "loss": 0.6805, + "step": 35345 + }, + { + "epoch": 1.29, + "learning_rate": 7.07437653348958e-08, + "loss": 0.6911, + "step": 35346 + }, + { + "epoch": 1.29, + "learning_rate": 7.074010327022375e-08, + "loss": 0.6931, + "step": 35347 + }, + { + "epoch": 1.29, + "learning_rate": 7.073644120555169e-08, + "loss": 0.6888, + "step": 35348 + }, + { + "epoch": 1.29, + "learning_rate": 7.073277914087963e-08, + "loss": 0.6784, + "step": 35349 + }, + { + "epoch": 1.29, + "learning_rate": 7.072911707620757e-08, + "loss": 0.6894, + "step": 35350 + }, + { + "epoch": 1.29, + "learning_rate": 7.072545501153549e-08, + "loss": 0.6961, + "step": 35351 + }, + { + "epoch": 1.29, + "learning_rate": 7.072179294686343e-08, + "loss": 0.7032, + "step": 35352 + }, + { + "epoch": 1.29, + "learning_rate": 7.071813088219137e-08, + "loss": 0.7175, + "step": 35353 + }, + { + "epoch": 1.29, + "learning_rate": 7.071446881751931e-08, + "loss": 0.6817, + "step": 35354 + }, + { + "epoch": 1.29, + "learning_rate": 7.071080675284725e-08, + "loss": 0.6954, + "step": 35355 + }, + { + "epoch": 1.29, + "learning_rate": 7.070714468817519e-08, + "loss": 0.7038, + "step": 35356 + }, + { + "epoch": 1.29, + "learning_rate": 7.070348262350313e-08, + "loss": 0.6817, + "step": 35357 + }, + { + "epoch": 1.29, + "learning_rate": 7.069982055883107e-08, + "loss": 0.7067, + "step": 35358 + }, + { + "epoch": 1.29, + "learning_rate": 7.069615849415901e-08, + "loss": 0.6961, + "step": 35359 + }, + { + "epoch": 1.29, + "learning_rate": 7.069249642948694e-08, + "loss": 0.6967, + "step": 35360 + }, + { + "epoch": 1.29, + "learning_rate": 7.068883436481488e-08, + "loss": 0.6885, + "step": 35361 + }, + { + "epoch": 1.29, + "learning_rate": 7.068517230014282e-08, + "loss": 0.6886, + "step": 35362 + }, + { + "epoch": 1.29, + "learning_rate": 7.068151023547076e-08, + "loss": 0.6909, + "step": 35363 + }, + { + "epoch": 1.29, + "learning_rate": 7.06778481707987e-08, + "loss": 0.704, + "step": 35364 + }, + { + "epoch": 1.29, + "learning_rate": 7.067418610612664e-08, + "loss": 0.6873, + "step": 35365 + }, + { + "epoch": 1.29, + "learning_rate": 7.067052404145457e-08, + "loss": 0.6951, + "step": 35366 + }, + { + "epoch": 1.29, + "learning_rate": 7.06668619767825e-08, + "loss": 0.7031, + "step": 35367 + }, + { + "epoch": 1.29, + "learning_rate": 7.066319991211045e-08, + "loss": 0.6902, + "step": 35368 + }, + { + "epoch": 1.29, + "learning_rate": 7.065953784743837e-08, + "loss": 0.7062, + "step": 35369 + }, + { + "epoch": 1.29, + "learning_rate": 7.065587578276631e-08, + "loss": 0.677, + "step": 35370 + }, + { + "epoch": 1.29, + "learning_rate": 7.065221371809425e-08, + "loss": 0.6905, + "step": 35371 + }, + { + "epoch": 1.29, + "learning_rate": 7.064855165342219e-08, + "loss": 0.6974, + "step": 35372 + }, + { + "epoch": 1.29, + "learning_rate": 7.064488958875013e-08, + "loss": 0.6829, + "step": 35373 + }, + { + "epoch": 1.29, + "learning_rate": 7.064122752407807e-08, + "loss": 0.6923, + "step": 35374 + }, + { + "epoch": 1.29, + "learning_rate": 7.063756545940601e-08, + "loss": 0.6955, + "step": 35375 + }, + { + "epoch": 1.29, + "learning_rate": 7.063390339473395e-08, + "loss": 0.6921, + "step": 35376 + }, + { + "epoch": 1.29, + "learning_rate": 7.063024133006189e-08, + "loss": 0.6898, + "step": 35377 + }, + { + "epoch": 1.29, + "learning_rate": 7.062657926538982e-08, + "loss": 0.689, + "step": 35378 + }, + { + "epoch": 1.29, + "learning_rate": 7.062291720071776e-08, + "loss": 0.6889, + "step": 35379 + }, + { + "epoch": 1.29, + "learning_rate": 7.06192551360457e-08, + "loss": 0.6854, + "step": 35380 + }, + { + "epoch": 1.29, + "learning_rate": 7.061559307137364e-08, + "loss": 0.7006, + "step": 35381 + }, + { + "epoch": 1.29, + "learning_rate": 7.061193100670158e-08, + "loss": 0.6808, + "step": 35382 + }, + { + "epoch": 1.29, + "learning_rate": 7.060826894202952e-08, + "loss": 0.6964, + "step": 35383 + }, + { + "epoch": 1.29, + "learning_rate": 7.060460687735746e-08, + "loss": 0.6931, + "step": 35384 + }, + { + "epoch": 1.29, + "learning_rate": 7.06009448126854e-08, + "loss": 0.6796, + "step": 35385 + }, + { + "epoch": 1.29, + "learning_rate": 7.059728274801334e-08, + "loss": 0.6994, + "step": 35386 + }, + { + "epoch": 1.29, + "learning_rate": 7.059362068334126e-08, + "loss": 0.6957, + "step": 35387 + }, + { + "epoch": 1.29, + "learning_rate": 7.05899586186692e-08, + "loss": 0.6941, + "step": 35388 + }, + { + "epoch": 1.29, + "learning_rate": 7.058629655399714e-08, + "loss": 0.7039, + "step": 35389 + }, + { + "epoch": 1.29, + "learning_rate": 7.058263448932507e-08, + "loss": 0.6973, + "step": 35390 + }, + { + "epoch": 1.29, + "learning_rate": 7.057897242465301e-08, + "loss": 0.6979, + "step": 35391 + }, + { + "epoch": 1.29, + "learning_rate": 7.057531035998095e-08, + "loss": 0.6864, + "step": 35392 + }, + { + "epoch": 1.29, + "learning_rate": 7.057164829530889e-08, + "loss": 0.6738, + "step": 35393 + }, + { + "epoch": 1.29, + "learning_rate": 7.056798623063683e-08, + "loss": 0.6854, + "step": 35394 + }, + { + "epoch": 1.3, + "learning_rate": 7.056432416596477e-08, + "loss": 0.6941, + "step": 35395 + }, + { + "epoch": 1.3, + "learning_rate": 7.05606621012927e-08, + "loss": 0.7024, + "step": 35396 + }, + { + "epoch": 1.3, + "learning_rate": 7.055700003662064e-08, + "loss": 0.6897, + "step": 35397 + }, + { + "epoch": 1.3, + "learning_rate": 7.055333797194858e-08, + "loss": 0.6826, + "step": 35398 + }, + { + "epoch": 1.3, + "learning_rate": 7.054967590727652e-08, + "loss": 0.701, + "step": 35399 + }, + { + "epoch": 1.3, + "learning_rate": 7.054601384260446e-08, + "loss": 0.6948, + "step": 35400 + }, + { + "epoch": 1.3, + "learning_rate": 7.05423517779324e-08, + "loss": 0.6766, + "step": 35401 + }, + { + "epoch": 1.3, + "learning_rate": 7.053868971326034e-08, + "loss": 0.6853, + "step": 35402 + }, + { + "epoch": 1.3, + "learning_rate": 7.053502764858828e-08, + "loss": 0.7051, + "step": 35403 + }, + { + "epoch": 1.3, + "learning_rate": 7.053136558391622e-08, + "loss": 0.6983, + "step": 35404 + }, + { + "epoch": 1.3, + "learning_rate": 7.052770351924414e-08, + "loss": 0.6927, + "step": 35405 + }, + { + "epoch": 1.3, + "learning_rate": 7.052404145457208e-08, + "loss": 0.6983, + "step": 35406 + }, + { + "epoch": 1.3, + "learning_rate": 7.052037938990002e-08, + "loss": 0.6904, + "step": 35407 + }, + { + "epoch": 1.3, + "learning_rate": 7.051671732522796e-08, + "loss": 0.6811, + "step": 35408 + }, + { + "epoch": 1.3, + "learning_rate": 7.05130552605559e-08, + "loss": 0.7001, + "step": 35409 + }, + { + "epoch": 1.3, + "learning_rate": 7.050939319588384e-08, + "loss": 0.6734, + "step": 35410 + }, + { + "epoch": 1.3, + "learning_rate": 7.050573113121178e-08, + "loss": 0.6976, + "step": 35411 + }, + { + "epoch": 1.3, + "learning_rate": 7.050206906653971e-08, + "loss": 0.6973, + "step": 35412 + }, + { + "epoch": 1.3, + "learning_rate": 7.049840700186765e-08, + "loss": 0.677, + "step": 35413 + }, + { + "epoch": 1.3, + "learning_rate": 7.049474493719559e-08, + "loss": 0.6969, + "step": 35414 + }, + { + "epoch": 1.3, + "learning_rate": 7.049108287252352e-08, + "loss": 0.6948, + "step": 35415 + }, + { + "epoch": 1.3, + "learning_rate": 7.048742080785146e-08, + "loss": 0.6929, + "step": 35416 + }, + { + "epoch": 1.3, + "learning_rate": 7.04837587431794e-08, + "loss": 0.6886, + "step": 35417 + }, + { + "epoch": 1.3, + "learning_rate": 7.048009667850734e-08, + "loss": 0.697, + "step": 35418 + }, + { + "epoch": 1.3, + "learning_rate": 7.047643461383528e-08, + "loss": 0.7067, + "step": 35419 + }, + { + "epoch": 1.3, + "learning_rate": 7.047277254916322e-08, + "loss": 0.6997, + "step": 35420 + }, + { + "epoch": 1.3, + "learning_rate": 7.046911048449116e-08, + "loss": 0.6934, + "step": 35421 + }, + { + "epoch": 1.3, + "learning_rate": 7.04654484198191e-08, + "loss": 0.6919, + "step": 35422 + }, + { + "epoch": 1.3, + "learning_rate": 7.046178635514702e-08, + "loss": 0.697, + "step": 35423 + }, + { + "epoch": 1.3, + "learning_rate": 7.045812429047496e-08, + "loss": 0.6887, + "step": 35424 + }, + { + "epoch": 1.3, + "learning_rate": 7.04544622258029e-08, + "loss": 0.6913, + "step": 35425 + }, + { + "epoch": 1.3, + "learning_rate": 7.045080016113084e-08, + "loss": 0.6929, + "step": 35426 + }, + { + "epoch": 1.3, + "learning_rate": 7.044713809645878e-08, + "loss": 0.6939, + "step": 35427 + }, + { + "epoch": 1.3, + "learning_rate": 7.044347603178672e-08, + "loss": 0.6987, + "step": 35428 + }, + { + "epoch": 1.3, + "learning_rate": 7.043981396711466e-08, + "loss": 0.6885, + "step": 35429 + }, + { + "epoch": 1.3, + "learning_rate": 7.04361519024426e-08, + "loss": 0.7061, + "step": 35430 + }, + { + "epoch": 1.3, + "learning_rate": 7.043248983777054e-08, + "loss": 0.6876, + "step": 35431 + }, + { + "epoch": 1.3, + "learning_rate": 7.042882777309847e-08, + "loss": 0.6891, + "step": 35432 + }, + { + "epoch": 1.3, + "learning_rate": 7.042516570842641e-08, + "loss": 0.6805, + "step": 35433 + }, + { + "epoch": 1.3, + "learning_rate": 7.042150364375435e-08, + "loss": 0.6963, + "step": 35434 + }, + { + "epoch": 1.3, + "learning_rate": 7.041784157908229e-08, + "loss": 0.681, + "step": 35435 + }, + { + "epoch": 1.3, + "learning_rate": 7.041417951441023e-08, + "loss": 0.6928, + "step": 35436 + }, + { + "epoch": 1.3, + "learning_rate": 7.041051744973815e-08, + "loss": 0.6768, + "step": 35437 + }, + { + "epoch": 1.3, + "learning_rate": 7.04068553850661e-08, + "loss": 0.6943, + "step": 35438 + }, + { + "epoch": 1.3, + "learning_rate": 7.040319332039403e-08, + "loss": 0.6836, + "step": 35439 + }, + { + "epoch": 1.3, + "learning_rate": 7.039953125572197e-08, + "loss": 0.6882, + "step": 35440 + }, + { + "epoch": 1.3, + "learning_rate": 7.03958691910499e-08, + "loss": 0.6827, + "step": 35441 + }, + { + "epoch": 1.3, + "learning_rate": 7.039220712637784e-08, + "loss": 0.6837, + "step": 35442 + }, + { + "epoch": 1.3, + "learning_rate": 7.038854506170578e-08, + "loss": 0.6787, + "step": 35443 + }, + { + "epoch": 1.3, + "learning_rate": 7.038488299703372e-08, + "loss": 0.6991, + "step": 35444 + }, + { + "epoch": 1.3, + "learning_rate": 7.038122093236166e-08, + "loss": 0.6793, + "step": 35445 + }, + { + "epoch": 1.3, + "learning_rate": 7.03775588676896e-08, + "loss": 0.7088, + "step": 35446 + }, + { + "epoch": 1.3, + "learning_rate": 7.037389680301754e-08, + "loss": 0.705, + "step": 35447 + }, + { + "epoch": 1.3, + "learning_rate": 7.037023473834548e-08, + "loss": 0.6872, + "step": 35448 + }, + { + "epoch": 1.3, + "learning_rate": 7.036657267367342e-08, + "loss": 0.6881, + "step": 35449 + }, + { + "epoch": 1.3, + "learning_rate": 7.036291060900135e-08, + "loss": 0.6831, + "step": 35450 + }, + { + "epoch": 1.3, + "learning_rate": 7.035924854432929e-08, + "loss": 0.688, + "step": 35451 + }, + { + "epoch": 1.3, + "learning_rate": 7.035558647965723e-08, + "loss": 0.6762, + "step": 35452 + }, + { + "epoch": 1.3, + "learning_rate": 7.035192441498517e-08, + "loss": 0.6987, + "step": 35453 + }, + { + "epoch": 1.3, + "learning_rate": 7.034826235031311e-08, + "loss": 0.6895, + "step": 35454 + }, + { + "epoch": 1.3, + "learning_rate": 7.034460028564105e-08, + "loss": 0.6951, + "step": 35455 + }, + { + "epoch": 1.3, + "learning_rate": 7.034093822096899e-08, + "loss": 0.6825, + "step": 35456 + }, + { + "epoch": 1.3, + "learning_rate": 7.033727615629693e-08, + "loss": 0.6944, + "step": 35457 + }, + { + "epoch": 1.3, + "learning_rate": 7.033361409162487e-08, + "loss": 0.7042, + "step": 35458 + }, + { + "epoch": 1.3, + "learning_rate": 7.032995202695279e-08, + "loss": 0.6932, + "step": 35459 + }, + { + "epoch": 1.3, + "learning_rate": 7.032628996228073e-08, + "loss": 0.6944, + "step": 35460 + }, + { + "epoch": 1.3, + "learning_rate": 7.032262789760866e-08, + "loss": 0.6863, + "step": 35461 + }, + { + "epoch": 1.3, + "learning_rate": 7.03189658329366e-08, + "loss": 0.7105, + "step": 35462 + }, + { + "epoch": 1.3, + "learning_rate": 7.031530376826454e-08, + "loss": 0.6942, + "step": 35463 + }, + { + "epoch": 1.3, + "learning_rate": 7.031164170359248e-08, + "loss": 0.6895, + "step": 35464 + }, + { + "epoch": 1.3, + "learning_rate": 7.030797963892042e-08, + "loss": 0.6964, + "step": 35465 + }, + { + "epoch": 1.3, + "learning_rate": 7.030431757424836e-08, + "loss": 0.6946, + "step": 35466 + }, + { + "epoch": 1.3, + "learning_rate": 7.03006555095763e-08, + "loss": 0.6941, + "step": 35467 + }, + { + "epoch": 1.3, + "learning_rate": 7.029699344490423e-08, + "loss": 0.702, + "step": 35468 + }, + { + "epoch": 1.3, + "learning_rate": 7.029333138023217e-08, + "loss": 0.6961, + "step": 35469 + }, + { + "epoch": 1.3, + "learning_rate": 7.02896693155601e-08, + "loss": 0.7019, + "step": 35470 + }, + { + "epoch": 1.3, + "learning_rate": 7.028600725088805e-08, + "loss": 0.6909, + "step": 35471 + }, + { + "epoch": 1.3, + "learning_rate": 7.028234518621599e-08, + "loss": 0.6935, + "step": 35472 + }, + { + "epoch": 1.3, + "learning_rate": 7.027868312154393e-08, + "loss": 0.7015, + "step": 35473 + }, + { + "epoch": 1.3, + "learning_rate": 7.027502105687187e-08, + "loss": 0.6975, + "step": 35474 + }, + { + "epoch": 1.3, + "learning_rate": 7.02713589921998e-08, + "loss": 0.6897, + "step": 35475 + }, + { + "epoch": 1.3, + "learning_rate": 7.026769692752775e-08, + "loss": 0.6967, + "step": 35476 + }, + { + "epoch": 1.3, + "learning_rate": 7.026403486285567e-08, + "loss": 0.6869, + "step": 35477 + }, + { + "epoch": 1.3, + "learning_rate": 7.026037279818361e-08, + "loss": 0.7015, + "step": 35478 + }, + { + "epoch": 1.3, + "learning_rate": 7.025671073351155e-08, + "loss": 0.6989, + "step": 35479 + }, + { + "epoch": 1.3, + "learning_rate": 7.025304866883949e-08, + "loss": 0.694, + "step": 35480 + }, + { + "epoch": 1.3, + "learning_rate": 7.024938660416743e-08, + "loss": 0.6865, + "step": 35481 + }, + { + "epoch": 1.3, + "learning_rate": 7.024572453949537e-08, + "loss": 0.6929, + "step": 35482 + }, + { + "epoch": 1.3, + "learning_rate": 7.02420624748233e-08, + "loss": 0.6795, + "step": 35483 + }, + { + "epoch": 1.3, + "learning_rate": 7.023840041015124e-08, + "loss": 0.6821, + "step": 35484 + }, + { + "epoch": 1.3, + "learning_rate": 7.023473834547918e-08, + "loss": 0.704, + "step": 35485 + }, + { + "epoch": 1.3, + "learning_rate": 7.02310762808071e-08, + "loss": 0.7021, + "step": 35486 + }, + { + "epoch": 1.3, + "learning_rate": 7.022741421613505e-08, + "loss": 0.6904, + "step": 35487 + }, + { + "epoch": 1.3, + "learning_rate": 7.022375215146298e-08, + "loss": 0.6937, + "step": 35488 + }, + { + "epoch": 1.3, + "learning_rate": 7.022009008679092e-08, + "loss": 0.6982, + "step": 35489 + }, + { + "epoch": 1.3, + "learning_rate": 7.021642802211886e-08, + "loss": 0.6953, + "step": 35490 + }, + { + "epoch": 1.3, + "learning_rate": 7.02127659574468e-08, + "loss": 0.688, + "step": 35491 + }, + { + "epoch": 1.3, + "learning_rate": 7.020910389277474e-08, + "loss": 0.6966, + "step": 35492 + }, + { + "epoch": 1.3, + "learning_rate": 7.020544182810268e-08, + "loss": 0.6849, + "step": 35493 + }, + { + "epoch": 1.3, + "learning_rate": 7.020177976343062e-08, + "loss": 0.7033, + "step": 35494 + }, + { + "epoch": 1.3, + "learning_rate": 7.019811769875855e-08, + "loss": 0.696, + "step": 35495 + }, + { + "epoch": 1.3, + "learning_rate": 7.019445563408649e-08, + "loss": 0.6904, + "step": 35496 + }, + { + "epoch": 1.3, + "learning_rate": 7.019079356941443e-08, + "loss": 0.7009, + "step": 35497 + }, + { + "epoch": 1.3, + "learning_rate": 7.018713150474237e-08, + "loss": 0.6721, + "step": 35498 + }, + { + "epoch": 1.3, + "learning_rate": 7.018346944007031e-08, + "loss": 0.686, + "step": 35499 + }, + { + "epoch": 1.3, + "learning_rate": 7.017980737539825e-08, + "loss": 0.6864, + "step": 35500 + }, + { + "epoch": 1.3, + "learning_rate": 7.017614531072619e-08, + "loss": 0.6888, + "step": 35501 + }, + { + "epoch": 1.3, + "learning_rate": 7.017248324605413e-08, + "loss": 0.7017, + "step": 35502 + }, + { + "epoch": 1.3, + "learning_rate": 7.016882118138207e-08, + "loss": 0.6975, + "step": 35503 + }, + { + "epoch": 1.3, + "learning_rate": 7.016515911671e-08, + "loss": 0.6922, + "step": 35504 + }, + { + "epoch": 1.3, + "learning_rate": 7.016149705203794e-08, + "loss": 0.6753, + "step": 35505 + }, + { + "epoch": 1.3, + "learning_rate": 7.015783498736588e-08, + "loss": 0.688, + "step": 35506 + }, + { + "epoch": 1.3, + "learning_rate": 7.015417292269382e-08, + "loss": 0.7043, + "step": 35507 + }, + { + "epoch": 1.3, + "learning_rate": 7.015051085802174e-08, + "loss": 0.6889, + "step": 35508 + }, + { + "epoch": 1.3, + "learning_rate": 7.014684879334968e-08, + "loss": 0.6885, + "step": 35509 + }, + { + "epoch": 1.3, + "learning_rate": 7.014318672867762e-08, + "loss": 0.698, + "step": 35510 + }, + { + "epoch": 1.3, + "learning_rate": 7.013952466400556e-08, + "loss": 0.6911, + "step": 35511 + }, + { + "epoch": 1.3, + "learning_rate": 7.01358625993335e-08, + "loss": 0.6814, + "step": 35512 + }, + { + "epoch": 1.3, + "learning_rate": 7.013220053466143e-08, + "loss": 0.697, + "step": 35513 + }, + { + "epoch": 1.3, + "learning_rate": 7.012853846998937e-08, + "loss": 0.6948, + "step": 35514 + }, + { + "epoch": 1.3, + "learning_rate": 7.012487640531731e-08, + "loss": 0.7116, + "step": 35515 + }, + { + "epoch": 1.3, + "learning_rate": 7.012121434064525e-08, + "loss": 0.6979, + "step": 35516 + }, + { + "epoch": 1.3, + "learning_rate": 7.011755227597319e-08, + "loss": 0.7145, + "step": 35517 + }, + { + "epoch": 1.3, + "learning_rate": 7.011389021130113e-08, + "loss": 0.6986, + "step": 35518 + }, + { + "epoch": 1.3, + "learning_rate": 7.011022814662907e-08, + "loss": 0.6897, + "step": 35519 + }, + { + "epoch": 1.3, + "learning_rate": 7.010656608195701e-08, + "loss": 0.7076, + "step": 35520 + }, + { + "epoch": 1.3, + "learning_rate": 7.010290401728495e-08, + "loss": 0.6912, + "step": 35521 + }, + { + "epoch": 1.3, + "learning_rate": 7.009924195261288e-08, + "loss": 0.6877, + "step": 35522 + }, + { + "epoch": 1.3, + "learning_rate": 7.009557988794082e-08, + "loss": 0.6994, + "step": 35523 + }, + { + "epoch": 1.3, + "learning_rate": 7.009191782326876e-08, + "loss": 0.6879, + "step": 35524 + }, + { + "epoch": 1.3, + "learning_rate": 7.00882557585967e-08, + "loss": 0.6879, + "step": 35525 + }, + { + "epoch": 1.3, + "learning_rate": 7.008459369392464e-08, + "loss": 0.6982, + "step": 35526 + }, + { + "epoch": 1.3, + "learning_rate": 7.008093162925258e-08, + "loss": 0.7053, + "step": 35527 + }, + { + "epoch": 1.3, + "learning_rate": 7.007726956458052e-08, + "loss": 0.6933, + "step": 35528 + }, + { + "epoch": 1.3, + "learning_rate": 7.007360749990844e-08, + "loss": 0.6958, + "step": 35529 + }, + { + "epoch": 1.3, + "learning_rate": 7.006994543523638e-08, + "loss": 0.6989, + "step": 35530 + }, + { + "epoch": 1.3, + "learning_rate": 7.006628337056432e-08, + "loss": 0.6918, + "step": 35531 + }, + { + "epoch": 1.3, + "learning_rate": 7.006262130589225e-08, + "loss": 0.6774, + "step": 35532 + }, + { + "epoch": 1.3, + "learning_rate": 7.005895924122019e-08, + "loss": 0.6952, + "step": 35533 + }, + { + "epoch": 1.3, + "learning_rate": 7.005529717654813e-08, + "loss": 0.7035, + "step": 35534 + }, + { + "epoch": 1.3, + "learning_rate": 7.005163511187607e-08, + "loss": 0.6986, + "step": 35535 + }, + { + "epoch": 1.3, + "learning_rate": 7.004797304720401e-08, + "loss": 0.6974, + "step": 35536 + }, + { + "epoch": 1.3, + "learning_rate": 7.004431098253195e-08, + "loss": 0.683, + "step": 35537 + }, + { + "epoch": 1.3, + "learning_rate": 7.004064891785989e-08, + "loss": 0.6993, + "step": 35538 + }, + { + "epoch": 1.3, + "learning_rate": 7.003698685318783e-08, + "loss": 0.7003, + "step": 35539 + }, + { + "epoch": 1.3, + "learning_rate": 7.003332478851576e-08, + "loss": 0.6807, + "step": 35540 + }, + { + "epoch": 1.3, + "learning_rate": 7.00296627238437e-08, + "loss": 0.709, + "step": 35541 + }, + { + "epoch": 1.3, + "learning_rate": 7.002600065917164e-08, + "loss": 0.6998, + "step": 35542 + }, + { + "epoch": 1.3, + "learning_rate": 7.002233859449958e-08, + "loss": 0.6932, + "step": 35543 + }, + { + "epoch": 1.3, + "learning_rate": 7.001867652982752e-08, + "loss": 0.6984, + "step": 35544 + }, + { + "epoch": 1.3, + "learning_rate": 7.001501446515546e-08, + "loss": 0.6641, + "step": 35545 + }, + { + "epoch": 1.3, + "learning_rate": 7.00113524004834e-08, + "loss": 0.6747, + "step": 35546 + }, + { + "epoch": 1.3, + "learning_rate": 7.000769033581133e-08, + "loss": 0.7059, + "step": 35547 + }, + { + "epoch": 1.3, + "learning_rate": 7.000402827113927e-08, + "loss": 0.7066, + "step": 35548 + }, + { + "epoch": 1.3, + "learning_rate": 7.00003662064672e-08, + "loss": 0.697, + "step": 35549 + }, + { + "epoch": 1.3, + "learning_rate": 6.999670414179514e-08, + "loss": 0.701, + "step": 35550 + }, + { + "epoch": 1.3, + "learning_rate": 6.999304207712308e-08, + "loss": 0.6899, + "step": 35551 + }, + { + "epoch": 1.3, + "learning_rate": 6.998938001245102e-08, + "loss": 0.6868, + "step": 35552 + }, + { + "epoch": 1.3, + "learning_rate": 6.998571794777896e-08, + "loss": 0.6798, + "step": 35553 + }, + { + "epoch": 1.3, + "learning_rate": 6.998205588310689e-08, + "loss": 0.6874, + "step": 35554 + }, + { + "epoch": 1.3, + "learning_rate": 6.997839381843483e-08, + "loss": 0.684, + "step": 35555 + }, + { + "epoch": 1.3, + "learning_rate": 6.997473175376277e-08, + "loss": 0.6775, + "step": 35556 + }, + { + "epoch": 1.3, + "learning_rate": 6.997106968909071e-08, + "loss": 0.6862, + "step": 35557 + }, + { + "epoch": 1.3, + "learning_rate": 6.996740762441863e-08, + "loss": 0.6984, + "step": 35558 + }, + { + "epoch": 1.3, + "learning_rate": 6.996374555974657e-08, + "loss": 0.6855, + "step": 35559 + }, + { + "epoch": 1.3, + "learning_rate": 6.996008349507451e-08, + "loss": 0.7011, + "step": 35560 + }, + { + "epoch": 1.3, + "learning_rate": 6.995642143040245e-08, + "loss": 0.6906, + "step": 35561 + }, + { + "epoch": 1.3, + "learning_rate": 6.99527593657304e-08, + "loss": 0.6857, + "step": 35562 + }, + { + "epoch": 1.3, + "learning_rate": 6.994909730105833e-08, + "loss": 0.6846, + "step": 35563 + }, + { + "epoch": 1.3, + "learning_rate": 6.994543523638627e-08, + "loss": 0.6904, + "step": 35564 + }, + { + "epoch": 1.3, + "learning_rate": 6.994177317171421e-08, + "loss": 0.6917, + "step": 35565 + }, + { + "epoch": 1.3, + "learning_rate": 6.993811110704215e-08, + "loss": 0.6964, + "step": 35566 + }, + { + "epoch": 1.3, + "learning_rate": 6.993444904237008e-08, + "loss": 0.6958, + "step": 35567 + }, + { + "epoch": 1.3, + "learning_rate": 6.993078697769802e-08, + "loss": 0.702, + "step": 35568 + }, + { + "epoch": 1.3, + "learning_rate": 6.992712491302596e-08, + "loss": 0.6946, + "step": 35569 + }, + { + "epoch": 1.3, + "learning_rate": 6.99234628483539e-08, + "loss": 0.6981, + "step": 35570 + }, + { + "epoch": 1.3, + "learning_rate": 6.991980078368184e-08, + "loss": 0.6882, + "step": 35571 + }, + { + "epoch": 1.3, + "learning_rate": 6.991613871900978e-08, + "loss": 0.6859, + "step": 35572 + }, + { + "epoch": 1.3, + "learning_rate": 6.991247665433772e-08, + "loss": 0.7014, + "step": 35573 + }, + { + "epoch": 1.3, + "learning_rate": 6.990881458966566e-08, + "loss": 0.7018, + "step": 35574 + }, + { + "epoch": 1.3, + "learning_rate": 6.99051525249936e-08, + "loss": 0.6993, + "step": 35575 + }, + { + "epoch": 1.3, + "learning_rate": 6.990149046032153e-08, + "loss": 0.6858, + "step": 35576 + }, + { + "epoch": 1.3, + "learning_rate": 6.989782839564947e-08, + "loss": 0.7026, + "step": 35577 + }, + { + "epoch": 1.3, + "learning_rate": 6.98941663309774e-08, + "loss": 0.6918, + "step": 35578 + }, + { + "epoch": 1.3, + "learning_rate": 6.989050426630533e-08, + "loss": 0.6937, + "step": 35579 + }, + { + "epoch": 1.3, + "learning_rate": 6.988684220163327e-08, + "loss": 0.6942, + "step": 35580 + }, + { + "epoch": 1.3, + "learning_rate": 6.988318013696121e-08, + "loss": 0.6842, + "step": 35581 + }, + { + "epoch": 1.3, + "learning_rate": 6.987951807228915e-08, + "loss": 0.7071, + "step": 35582 + }, + { + "epoch": 1.3, + "learning_rate": 6.987585600761709e-08, + "loss": 0.6908, + "step": 35583 + }, + { + "epoch": 1.3, + "learning_rate": 6.987219394294503e-08, + "loss": 0.6892, + "step": 35584 + }, + { + "epoch": 1.3, + "learning_rate": 6.986853187827296e-08, + "loss": 0.7025, + "step": 35585 + }, + { + "epoch": 1.3, + "learning_rate": 6.98648698136009e-08, + "loss": 0.7008, + "step": 35586 + }, + { + "epoch": 1.3, + "learning_rate": 6.986120774892884e-08, + "loss": 0.6898, + "step": 35587 + }, + { + "epoch": 1.3, + "learning_rate": 6.985754568425678e-08, + "loss": 0.7113, + "step": 35588 + }, + { + "epoch": 1.3, + "learning_rate": 6.985388361958472e-08, + "loss": 0.6957, + "step": 35589 + }, + { + "epoch": 1.3, + "learning_rate": 6.985022155491266e-08, + "loss": 0.6867, + "step": 35590 + }, + { + "epoch": 1.3, + "learning_rate": 6.98465594902406e-08, + "loss": 0.6985, + "step": 35591 + }, + { + "epoch": 1.3, + "learning_rate": 6.984289742556854e-08, + "loss": 0.6861, + "step": 35592 + }, + { + "epoch": 1.3, + "learning_rate": 6.983923536089648e-08, + "loss": 0.6948, + "step": 35593 + }, + { + "epoch": 1.3, + "learning_rate": 6.98355732962244e-08, + "loss": 0.6965, + "step": 35594 + }, + { + "epoch": 1.3, + "learning_rate": 6.983191123155235e-08, + "loss": 0.6925, + "step": 35595 + }, + { + "epoch": 1.3, + "learning_rate": 6.982824916688029e-08, + "loss": 0.6934, + "step": 35596 + }, + { + "epoch": 1.3, + "learning_rate": 6.982458710220823e-08, + "loss": 0.6962, + "step": 35597 + }, + { + "epoch": 1.3, + "learning_rate": 6.982092503753617e-08, + "loss": 0.6807, + "step": 35598 + }, + { + "epoch": 1.3, + "learning_rate": 6.98172629728641e-08, + "loss": 0.6798, + "step": 35599 + }, + { + "epoch": 1.3, + "learning_rate": 6.981360090819203e-08, + "loss": 0.6901, + "step": 35600 + }, + { + "epoch": 1.3, + "learning_rate": 6.980993884351997e-08, + "loss": 0.6778, + "step": 35601 + }, + { + "epoch": 1.3, + "learning_rate": 6.980627677884791e-08, + "loss": 0.6972, + "step": 35602 + }, + { + "epoch": 1.3, + "learning_rate": 6.980261471417584e-08, + "loss": 0.6874, + "step": 35603 + }, + { + "epoch": 1.3, + "learning_rate": 6.979895264950378e-08, + "loss": 0.7082, + "step": 35604 + }, + { + "epoch": 1.3, + "learning_rate": 6.979529058483172e-08, + "loss": 0.6953, + "step": 35605 + }, + { + "epoch": 1.3, + "learning_rate": 6.979162852015966e-08, + "loss": 0.6899, + "step": 35606 + }, + { + "epoch": 1.3, + "learning_rate": 6.97879664554876e-08, + "loss": 0.6893, + "step": 35607 + }, + { + "epoch": 1.3, + "learning_rate": 6.978430439081554e-08, + "loss": 0.6992, + "step": 35608 + }, + { + "epoch": 1.3, + "learning_rate": 6.978064232614348e-08, + "loss": 0.6825, + "step": 35609 + }, + { + "epoch": 1.3, + "learning_rate": 6.977698026147142e-08, + "loss": 0.7032, + "step": 35610 + }, + { + "epoch": 1.3, + "learning_rate": 6.977331819679936e-08, + "loss": 0.6822, + "step": 35611 + }, + { + "epoch": 1.3, + "learning_rate": 6.976965613212728e-08, + "loss": 0.6857, + "step": 35612 + }, + { + "epoch": 1.3, + "learning_rate": 6.976599406745522e-08, + "loss": 0.6845, + "step": 35613 + }, + { + "epoch": 1.3, + "learning_rate": 6.976233200278316e-08, + "loss": 0.7004, + "step": 35614 + }, + { + "epoch": 1.3, + "learning_rate": 6.97586699381111e-08, + "loss": 0.6902, + "step": 35615 + }, + { + "epoch": 1.3, + "learning_rate": 6.975500787343904e-08, + "loss": 0.6845, + "step": 35616 + }, + { + "epoch": 1.3, + "learning_rate": 6.975134580876698e-08, + "loss": 0.7112, + "step": 35617 + }, + { + "epoch": 1.3, + "learning_rate": 6.974768374409492e-08, + "loss": 0.6907, + "step": 35618 + }, + { + "epoch": 1.3, + "learning_rate": 6.974402167942286e-08, + "loss": 0.6943, + "step": 35619 + }, + { + "epoch": 1.3, + "learning_rate": 6.97403596147508e-08, + "loss": 0.697, + "step": 35620 + }, + { + "epoch": 1.3, + "learning_rate": 6.973669755007873e-08, + "loss": 0.6956, + "step": 35621 + }, + { + "epoch": 1.3, + "learning_rate": 6.973303548540667e-08, + "loss": 0.6987, + "step": 35622 + }, + { + "epoch": 1.3, + "learning_rate": 6.972937342073461e-08, + "loss": 0.6887, + "step": 35623 + }, + { + "epoch": 1.3, + "learning_rate": 6.972571135606255e-08, + "loss": 0.6959, + "step": 35624 + }, + { + "epoch": 1.3, + "learning_rate": 6.972204929139048e-08, + "loss": 0.6945, + "step": 35625 + }, + { + "epoch": 1.3, + "learning_rate": 6.971838722671842e-08, + "loss": 0.692, + "step": 35626 + }, + { + "epoch": 1.3, + "learning_rate": 6.971472516204636e-08, + "loss": 0.6994, + "step": 35627 + }, + { + "epoch": 1.3, + "learning_rate": 6.97110630973743e-08, + "loss": 0.6999, + "step": 35628 + }, + { + "epoch": 1.3, + "learning_rate": 6.970740103270222e-08, + "loss": 0.7087, + "step": 35629 + }, + { + "epoch": 1.3, + "learning_rate": 6.970373896803016e-08, + "loss": 0.6905, + "step": 35630 + }, + { + "epoch": 1.3, + "learning_rate": 6.97000769033581e-08, + "loss": 0.6888, + "step": 35631 + }, + { + "epoch": 1.3, + "learning_rate": 6.969641483868604e-08, + "loss": 0.6926, + "step": 35632 + }, + { + "epoch": 1.3, + "learning_rate": 6.969275277401398e-08, + "loss": 0.6908, + "step": 35633 + }, + { + "epoch": 1.3, + "learning_rate": 6.968909070934192e-08, + "loss": 0.6951, + "step": 35634 + }, + { + "epoch": 1.3, + "learning_rate": 6.968542864466986e-08, + "loss": 0.6946, + "step": 35635 + }, + { + "epoch": 1.3, + "learning_rate": 6.96817665799978e-08, + "loss": 0.6929, + "step": 35636 + }, + { + "epoch": 1.3, + "learning_rate": 6.967810451532574e-08, + "loss": 0.6794, + "step": 35637 + }, + { + "epoch": 1.3, + "learning_rate": 6.967444245065367e-08, + "loss": 0.6981, + "step": 35638 + }, + { + "epoch": 1.3, + "learning_rate": 6.967078038598161e-08, + "loss": 0.6964, + "step": 35639 + }, + { + "epoch": 1.3, + "learning_rate": 6.966711832130955e-08, + "loss": 0.6901, + "step": 35640 + }, + { + "epoch": 1.3, + "learning_rate": 6.966345625663749e-08, + "loss": 0.7007, + "step": 35641 + }, + { + "epoch": 1.3, + "learning_rate": 6.965979419196543e-08, + "loss": 0.6993, + "step": 35642 + }, + { + "epoch": 1.3, + "learning_rate": 6.965613212729337e-08, + "loss": 0.6862, + "step": 35643 + }, + { + "epoch": 1.3, + "learning_rate": 6.965247006262131e-08, + "loss": 0.7095, + "step": 35644 + }, + { + "epoch": 1.3, + "learning_rate": 6.964880799794925e-08, + "loss": 0.683, + "step": 35645 + }, + { + "epoch": 1.3, + "learning_rate": 6.964514593327719e-08, + "loss": 0.6957, + "step": 35646 + }, + { + "epoch": 1.3, + "learning_rate": 6.964148386860512e-08, + "loss": 0.6902, + "step": 35647 + }, + { + "epoch": 1.3, + "learning_rate": 6.963782180393306e-08, + "loss": 0.6755, + "step": 35648 + }, + { + "epoch": 1.3, + "learning_rate": 6.963415973926098e-08, + "loss": 0.7177, + "step": 35649 + }, + { + "epoch": 1.3, + "learning_rate": 6.963049767458892e-08, + "loss": 0.6925, + "step": 35650 + }, + { + "epoch": 1.3, + "learning_rate": 6.962683560991686e-08, + "loss": 0.6924, + "step": 35651 + }, + { + "epoch": 1.3, + "learning_rate": 6.96231735452448e-08, + "loss": 0.6975, + "step": 35652 + }, + { + "epoch": 1.3, + "learning_rate": 6.961951148057274e-08, + "loss": 0.6913, + "step": 35653 + }, + { + "epoch": 1.3, + "learning_rate": 6.961584941590068e-08, + "loss": 0.6958, + "step": 35654 + }, + { + "epoch": 1.3, + "learning_rate": 6.961218735122862e-08, + "loss": 0.6939, + "step": 35655 + }, + { + "epoch": 1.3, + "learning_rate": 6.960852528655655e-08, + "loss": 0.6942, + "step": 35656 + }, + { + "epoch": 1.3, + "learning_rate": 6.960486322188449e-08, + "loss": 0.6878, + "step": 35657 + }, + { + "epoch": 1.3, + "learning_rate": 6.960120115721243e-08, + "loss": 0.6991, + "step": 35658 + }, + { + "epoch": 1.3, + "learning_rate": 6.959753909254037e-08, + "loss": 0.6912, + "step": 35659 + }, + { + "epoch": 1.3, + "learning_rate": 6.959387702786831e-08, + "loss": 0.6939, + "step": 35660 + }, + { + "epoch": 1.3, + "learning_rate": 6.959021496319625e-08, + "loss": 0.6918, + "step": 35661 + }, + { + "epoch": 1.3, + "learning_rate": 6.958655289852419e-08, + "loss": 0.6941, + "step": 35662 + }, + { + "epoch": 1.3, + "learning_rate": 6.958289083385213e-08, + "loss": 0.6894, + "step": 35663 + }, + { + "epoch": 1.3, + "learning_rate": 6.957922876918007e-08, + "loss": 0.6857, + "step": 35664 + }, + { + "epoch": 1.3, + "learning_rate": 6.9575566704508e-08, + "loss": 0.6902, + "step": 35665 + }, + { + "epoch": 1.3, + "learning_rate": 6.957190463983593e-08, + "loss": 0.7073, + "step": 35666 + }, + { + "epoch": 1.3, + "learning_rate": 6.956824257516387e-08, + "loss": 0.7102, + "step": 35667 + }, + { + "epoch": 1.3, + "learning_rate": 6.956458051049181e-08, + "loss": 0.6809, + "step": 35668 + }, + { + "epoch": 1.31, + "learning_rate": 6.956091844581975e-08, + "loss": 0.6788, + "step": 35669 + }, + { + "epoch": 1.31, + "learning_rate": 6.95572563811477e-08, + "loss": 0.7001, + "step": 35670 + }, + { + "epoch": 1.31, + "learning_rate": 6.955359431647562e-08, + "loss": 0.6835, + "step": 35671 + }, + { + "epoch": 1.31, + "learning_rate": 6.954993225180356e-08, + "loss": 0.6713, + "step": 35672 + }, + { + "epoch": 1.31, + "learning_rate": 6.95462701871315e-08, + "loss": 0.6926, + "step": 35673 + }, + { + "epoch": 1.31, + "learning_rate": 6.954260812245943e-08, + "loss": 0.6899, + "step": 35674 + }, + { + "epoch": 1.31, + "learning_rate": 6.953894605778737e-08, + "loss": 0.6874, + "step": 35675 + }, + { + "epoch": 1.31, + "learning_rate": 6.953528399311531e-08, + "loss": 0.6946, + "step": 35676 + }, + { + "epoch": 1.31, + "learning_rate": 6.953162192844325e-08, + "loss": 0.6926, + "step": 35677 + }, + { + "epoch": 1.31, + "learning_rate": 6.952795986377119e-08, + "loss": 0.6778, + "step": 35678 + }, + { + "epoch": 1.31, + "learning_rate": 6.952429779909913e-08, + "loss": 0.6977, + "step": 35679 + }, + { + "epoch": 1.31, + "learning_rate": 6.952063573442707e-08, + "loss": 0.7102, + "step": 35680 + }, + { + "epoch": 1.31, + "learning_rate": 6.951697366975501e-08, + "loss": 0.7039, + "step": 35681 + }, + { + "epoch": 1.31, + "learning_rate": 6.951331160508295e-08, + "loss": 0.6905, + "step": 35682 + }, + { + "epoch": 1.31, + "learning_rate": 6.950964954041087e-08, + "loss": 0.7061, + "step": 35683 + }, + { + "epoch": 1.31, + "learning_rate": 6.950598747573881e-08, + "loss": 0.6969, + "step": 35684 + }, + { + "epoch": 1.31, + "learning_rate": 6.950232541106675e-08, + "loss": 0.6911, + "step": 35685 + }, + { + "epoch": 1.31, + "learning_rate": 6.94986633463947e-08, + "loss": 0.6945, + "step": 35686 + }, + { + "epoch": 1.31, + "learning_rate": 6.949500128172263e-08, + "loss": 0.6803, + "step": 35687 + }, + { + "epoch": 1.31, + "learning_rate": 6.949133921705057e-08, + "loss": 0.7006, + "step": 35688 + }, + { + "epoch": 1.31, + "learning_rate": 6.948767715237851e-08, + "loss": 0.6771, + "step": 35689 + }, + { + "epoch": 1.31, + "learning_rate": 6.948401508770645e-08, + "loss": 0.7007, + "step": 35690 + }, + { + "epoch": 1.31, + "learning_rate": 6.94803530230344e-08, + "loss": 0.695, + "step": 35691 + }, + { + "epoch": 1.31, + "learning_rate": 6.947669095836232e-08, + "loss": 0.687, + "step": 35692 + }, + { + "epoch": 1.31, + "learning_rate": 6.947302889369026e-08, + "loss": 0.6934, + "step": 35693 + }, + { + "epoch": 1.31, + "learning_rate": 6.94693668290182e-08, + "loss": 0.6881, + "step": 35694 + }, + { + "epoch": 1.31, + "learning_rate": 6.946570476434614e-08, + "loss": 0.6937, + "step": 35695 + }, + { + "epoch": 1.31, + "learning_rate": 6.946204269967407e-08, + "loss": 0.6928, + "step": 35696 + }, + { + "epoch": 1.31, + "learning_rate": 6.9458380635002e-08, + "loss": 0.7024, + "step": 35697 + }, + { + "epoch": 1.31, + "learning_rate": 6.945471857032995e-08, + "loss": 0.6935, + "step": 35698 + }, + { + "epoch": 1.31, + "learning_rate": 6.945105650565789e-08, + "loss": 0.6835, + "step": 35699 + }, + { + "epoch": 1.31, + "learning_rate": 6.944739444098583e-08, + "loss": 0.6903, + "step": 35700 + }, + { + "epoch": 1.31, + "learning_rate": 6.944373237631375e-08, + "loss": 0.6911, + "step": 35701 + }, + { + "epoch": 1.31, + "learning_rate": 6.944007031164169e-08, + "loss": 0.7024, + "step": 35702 + }, + { + "epoch": 1.31, + "learning_rate": 6.943640824696963e-08, + "loss": 0.6878, + "step": 35703 + }, + { + "epoch": 1.31, + "learning_rate": 6.943274618229757e-08, + "loss": 0.6695, + "step": 35704 + }, + { + "epoch": 1.31, + "learning_rate": 6.942908411762551e-08, + "loss": 0.6889, + "step": 35705 + }, + { + "epoch": 1.31, + "learning_rate": 6.942542205295345e-08, + "loss": 0.706, + "step": 35706 + }, + { + "epoch": 1.31, + "learning_rate": 6.942175998828139e-08, + "loss": 0.6977, + "step": 35707 + }, + { + "epoch": 1.31, + "learning_rate": 6.941809792360933e-08, + "loss": 0.7051, + "step": 35708 + }, + { + "epoch": 1.31, + "learning_rate": 6.941443585893727e-08, + "loss": 0.6875, + "step": 35709 + }, + { + "epoch": 1.31, + "learning_rate": 6.94107737942652e-08, + "loss": 0.6966, + "step": 35710 + }, + { + "epoch": 1.31, + "learning_rate": 6.940711172959314e-08, + "loss": 0.6822, + "step": 35711 + }, + { + "epoch": 1.31, + "learning_rate": 6.940344966492108e-08, + "loss": 0.7058, + "step": 35712 + }, + { + "epoch": 1.31, + "learning_rate": 6.939978760024902e-08, + "loss": 0.6908, + "step": 35713 + }, + { + "epoch": 1.31, + "learning_rate": 6.939612553557696e-08, + "loss": 0.6905, + "step": 35714 + }, + { + "epoch": 1.31, + "learning_rate": 6.93924634709049e-08, + "loss": 0.6969, + "step": 35715 + }, + { + "epoch": 1.31, + "learning_rate": 6.938880140623284e-08, + "loss": 0.6929, + "step": 35716 + }, + { + "epoch": 1.31, + "learning_rate": 6.938513934156078e-08, + "loss": 0.688, + "step": 35717 + }, + { + "epoch": 1.31, + "learning_rate": 6.93814772768887e-08, + "loss": 0.6757, + "step": 35718 + }, + { + "epoch": 1.31, + "learning_rate": 6.937781521221665e-08, + "loss": 0.6885, + "step": 35719 + }, + { + "epoch": 1.31, + "learning_rate": 6.937415314754457e-08, + "loss": 0.7088, + "step": 35720 + }, + { + "epoch": 1.31, + "learning_rate": 6.937049108287251e-08, + "loss": 0.6882, + "step": 35721 + }, + { + "epoch": 1.31, + "learning_rate": 6.936682901820045e-08, + "loss": 0.6988, + "step": 35722 + }, + { + "epoch": 1.31, + "learning_rate": 6.936316695352839e-08, + "loss": 0.7082, + "step": 35723 + }, + { + "epoch": 1.31, + "learning_rate": 6.935950488885633e-08, + "loss": 0.686, + "step": 35724 + }, + { + "epoch": 1.31, + "learning_rate": 6.935584282418427e-08, + "loss": 0.7121, + "step": 35725 + }, + { + "epoch": 1.31, + "learning_rate": 6.935218075951221e-08, + "loss": 0.6855, + "step": 35726 + }, + { + "epoch": 1.31, + "learning_rate": 6.934851869484015e-08, + "loss": 0.6862, + "step": 35727 + }, + { + "epoch": 1.31, + "learning_rate": 6.934485663016808e-08, + "loss": 0.6963, + "step": 35728 + }, + { + "epoch": 1.31, + "learning_rate": 6.934119456549602e-08, + "loss": 0.6925, + "step": 35729 + }, + { + "epoch": 1.31, + "learning_rate": 6.933753250082396e-08, + "loss": 0.6917, + "step": 35730 + }, + { + "epoch": 1.31, + "learning_rate": 6.93338704361519e-08, + "loss": 0.7008, + "step": 35731 + }, + { + "epoch": 1.31, + "learning_rate": 6.933020837147984e-08, + "loss": 0.6926, + "step": 35732 + }, + { + "epoch": 1.31, + "learning_rate": 6.932654630680778e-08, + "loss": 0.7002, + "step": 35733 + }, + { + "epoch": 1.31, + "learning_rate": 6.932288424213572e-08, + "loss": 0.7073, + "step": 35734 + }, + { + "epoch": 1.31, + "learning_rate": 6.931922217746366e-08, + "loss": 0.7013, + "step": 35735 + }, + { + "epoch": 1.31, + "learning_rate": 6.93155601127916e-08, + "loss": 0.6947, + "step": 35736 + }, + { + "epoch": 1.31, + "learning_rate": 6.931189804811952e-08, + "loss": 0.6939, + "step": 35737 + }, + { + "epoch": 1.31, + "learning_rate": 6.930823598344746e-08, + "loss": 0.7223, + "step": 35738 + }, + { + "epoch": 1.31, + "learning_rate": 6.93045739187754e-08, + "loss": 0.6974, + "step": 35739 + }, + { + "epoch": 1.31, + "learning_rate": 6.930091185410334e-08, + "loss": 0.6766, + "step": 35740 + }, + { + "epoch": 1.31, + "learning_rate": 6.929724978943128e-08, + "loss": 0.6791, + "step": 35741 + }, + { + "epoch": 1.31, + "learning_rate": 6.929358772475921e-08, + "loss": 0.7, + "step": 35742 + }, + { + "epoch": 1.31, + "learning_rate": 6.928992566008715e-08, + "loss": 0.7029, + "step": 35743 + }, + { + "epoch": 1.31, + "learning_rate": 6.928626359541509e-08, + "loss": 0.6998, + "step": 35744 + }, + { + "epoch": 1.31, + "learning_rate": 6.928260153074303e-08, + "loss": 0.7104, + "step": 35745 + }, + { + "epoch": 1.31, + "learning_rate": 6.927893946607096e-08, + "loss": 0.6963, + "step": 35746 + }, + { + "epoch": 1.31, + "learning_rate": 6.92752774013989e-08, + "loss": 0.691, + "step": 35747 + }, + { + "epoch": 1.31, + "learning_rate": 6.927161533672684e-08, + "loss": 0.6972, + "step": 35748 + }, + { + "epoch": 1.31, + "learning_rate": 6.926795327205478e-08, + "loss": 0.6929, + "step": 35749 + }, + { + "epoch": 1.31, + "learning_rate": 6.926429120738272e-08, + "loss": 0.691, + "step": 35750 + }, + { + "epoch": 1.31, + "learning_rate": 6.926062914271066e-08, + "loss": 0.703, + "step": 35751 + }, + { + "epoch": 1.31, + "learning_rate": 6.92569670780386e-08, + "loss": 0.6913, + "step": 35752 + }, + { + "epoch": 1.31, + "learning_rate": 6.925330501336654e-08, + "loss": 0.7022, + "step": 35753 + }, + { + "epoch": 1.31, + "learning_rate": 6.924964294869448e-08, + "loss": 0.7074, + "step": 35754 + }, + { + "epoch": 1.31, + "learning_rate": 6.92459808840224e-08, + "loss": 0.701, + "step": 35755 + }, + { + "epoch": 1.31, + "learning_rate": 6.924231881935034e-08, + "loss": 0.6931, + "step": 35756 + }, + { + "epoch": 1.31, + "learning_rate": 6.923865675467828e-08, + "loss": 0.6966, + "step": 35757 + }, + { + "epoch": 1.31, + "learning_rate": 6.923499469000622e-08, + "loss": 0.6885, + "step": 35758 + }, + { + "epoch": 1.31, + "learning_rate": 6.923133262533416e-08, + "loss": 0.6814, + "step": 35759 + }, + { + "epoch": 1.31, + "learning_rate": 6.92276705606621e-08, + "loss": 0.6933, + "step": 35760 + }, + { + "epoch": 1.31, + "learning_rate": 6.922400849599004e-08, + "loss": 0.6975, + "step": 35761 + }, + { + "epoch": 1.31, + "learning_rate": 6.922034643131798e-08, + "loss": 0.6993, + "step": 35762 + }, + { + "epoch": 1.31, + "learning_rate": 6.921668436664592e-08, + "loss": 0.6895, + "step": 35763 + }, + { + "epoch": 1.31, + "learning_rate": 6.921302230197385e-08, + "loss": 0.6945, + "step": 35764 + }, + { + "epoch": 1.31, + "learning_rate": 6.920936023730179e-08, + "loss": 0.6856, + "step": 35765 + }, + { + "epoch": 1.31, + "learning_rate": 6.920569817262973e-08, + "loss": 0.6781, + "step": 35766 + }, + { + "epoch": 1.31, + "learning_rate": 6.920203610795766e-08, + "loss": 0.6991, + "step": 35767 + }, + { + "epoch": 1.31, + "learning_rate": 6.91983740432856e-08, + "loss": 0.698, + "step": 35768 + }, + { + "epoch": 1.31, + "learning_rate": 6.919471197861354e-08, + "loss": 0.6944, + "step": 35769 + }, + { + "epoch": 1.31, + "learning_rate": 6.919104991394148e-08, + "loss": 0.6915, + "step": 35770 + }, + { + "epoch": 1.31, + "learning_rate": 6.918738784926942e-08, + "loss": 0.7021, + "step": 35771 + }, + { + "epoch": 1.31, + "learning_rate": 6.918372578459736e-08, + "loss": 0.7083, + "step": 35772 + }, + { + "epoch": 1.31, + "learning_rate": 6.918006371992528e-08, + "loss": 0.7013, + "step": 35773 + }, + { + "epoch": 1.31, + "learning_rate": 6.917640165525322e-08, + "loss": 0.6984, + "step": 35774 + }, + { + "epoch": 1.31, + "learning_rate": 6.917273959058116e-08, + "loss": 0.6992, + "step": 35775 + }, + { + "epoch": 1.31, + "learning_rate": 6.91690775259091e-08, + "loss": 0.703, + "step": 35776 + }, + { + "epoch": 1.31, + "learning_rate": 6.916541546123704e-08, + "loss": 0.6963, + "step": 35777 + }, + { + "epoch": 1.31, + "learning_rate": 6.916175339656498e-08, + "loss": 0.6944, + "step": 35778 + }, + { + "epoch": 1.31, + "learning_rate": 6.915809133189292e-08, + "loss": 0.7027, + "step": 35779 + }, + { + "epoch": 1.31, + "learning_rate": 6.915442926722086e-08, + "loss": 0.6924, + "step": 35780 + }, + { + "epoch": 1.31, + "learning_rate": 6.91507672025488e-08, + "loss": 0.6956, + "step": 35781 + }, + { + "epoch": 1.31, + "learning_rate": 6.914710513787673e-08, + "loss": 0.7007, + "step": 35782 + }, + { + "epoch": 1.31, + "learning_rate": 6.914344307320467e-08, + "loss": 0.67, + "step": 35783 + }, + { + "epoch": 1.31, + "learning_rate": 6.913978100853261e-08, + "loss": 0.7034, + "step": 35784 + }, + { + "epoch": 1.31, + "learning_rate": 6.913611894386055e-08, + "loss": 0.6854, + "step": 35785 + }, + { + "epoch": 1.31, + "learning_rate": 6.913245687918849e-08, + "loss": 0.6899, + "step": 35786 + }, + { + "epoch": 1.31, + "learning_rate": 6.912879481451643e-08, + "loss": 0.7016, + "step": 35787 + }, + { + "epoch": 1.31, + "learning_rate": 6.912513274984437e-08, + "loss": 0.6959, + "step": 35788 + }, + { + "epoch": 1.31, + "learning_rate": 6.91214706851723e-08, + "loss": 0.6823, + "step": 35789 + }, + { + "epoch": 1.31, + "learning_rate": 6.911780862050023e-08, + "loss": 0.696, + "step": 35790 + }, + { + "epoch": 1.31, + "learning_rate": 6.911414655582816e-08, + "loss": 0.69, + "step": 35791 + }, + { + "epoch": 1.31, + "learning_rate": 6.91104844911561e-08, + "loss": 0.6932, + "step": 35792 + }, + { + "epoch": 1.31, + "learning_rate": 6.910682242648404e-08, + "loss": 0.6963, + "step": 35793 + }, + { + "epoch": 1.31, + "learning_rate": 6.910316036181198e-08, + "loss": 0.6949, + "step": 35794 + }, + { + "epoch": 1.31, + "learning_rate": 6.909949829713992e-08, + "loss": 0.6997, + "step": 35795 + }, + { + "epoch": 1.31, + "learning_rate": 6.909583623246786e-08, + "loss": 0.6766, + "step": 35796 + }, + { + "epoch": 1.31, + "learning_rate": 6.90921741677958e-08, + "loss": 0.6818, + "step": 35797 + }, + { + "epoch": 1.31, + "learning_rate": 6.908851210312374e-08, + "loss": 0.696, + "step": 35798 + }, + { + "epoch": 1.31, + "learning_rate": 6.908485003845168e-08, + "loss": 0.6943, + "step": 35799 + }, + { + "epoch": 1.31, + "learning_rate": 6.908118797377961e-08, + "loss": 0.7087, + "step": 35800 + }, + { + "epoch": 1.31, + "learning_rate": 6.907752590910755e-08, + "loss": 0.7114, + "step": 35801 + }, + { + "epoch": 1.31, + "learning_rate": 6.907386384443549e-08, + "loss": 0.6978, + "step": 35802 + }, + { + "epoch": 1.31, + "learning_rate": 6.907020177976343e-08, + "loss": 0.6906, + "step": 35803 + }, + { + "epoch": 1.31, + "learning_rate": 6.906653971509137e-08, + "loss": 0.6924, + "step": 35804 + }, + { + "epoch": 1.31, + "learning_rate": 6.906287765041931e-08, + "loss": 0.6947, + "step": 35805 + }, + { + "epoch": 1.31, + "learning_rate": 6.905921558574725e-08, + "loss": 0.7083, + "step": 35806 + }, + { + "epoch": 1.31, + "learning_rate": 6.905555352107519e-08, + "loss": 0.679, + "step": 35807 + }, + { + "epoch": 1.31, + "learning_rate": 6.905189145640313e-08, + "loss": 0.712, + "step": 35808 + }, + { + "epoch": 1.31, + "learning_rate": 6.904822939173105e-08, + "loss": 0.6884, + "step": 35809 + }, + { + "epoch": 1.31, + "learning_rate": 6.9044567327059e-08, + "loss": 0.698, + "step": 35810 + }, + { + "epoch": 1.31, + "learning_rate": 6.904090526238693e-08, + "loss": 0.7094, + "step": 35811 + }, + { + "epoch": 1.31, + "learning_rate": 6.903724319771487e-08, + "loss": 0.7078, + "step": 35812 + }, + { + "epoch": 1.31, + "learning_rate": 6.90335811330428e-08, + "loss": 0.7018, + "step": 35813 + }, + { + "epoch": 1.31, + "learning_rate": 6.902991906837074e-08, + "loss": 0.6948, + "step": 35814 + }, + { + "epoch": 1.31, + "learning_rate": 6.902625700369868e-08, + "loss": 0.727, + "step": 35815 + }, + { + "epoch": 1.31, + "learning_rate": 6.902259493902662e-08, + "loss": 0.6904, + "step": 35816 + }, + { + "epoch": 1.31, + "learning_rate": 6.901893287435456e-08, + "loss": 0.6936, + "step": 35817 + }, + { + "epoch": 1.31, + "learning_rate": 6.901527080968249e-08, + "loss": 0.6981, + "step": 35818 + }, + { + "epoch": 1.31, + "learning_rate": 6.901160874501043e-08, + "loss": 0.6892, + "step": 35819 + }, + { + "epoch": 1.31, + "learning_rate": 6.900794668033837e-08, + "loss": 0.6833, + "step": 35820 + }, + { + "epoch": 1.31, + "learning_rate": 6.90042846156663e-08, + "loss": 0.6978, + "step": 35821 + }, + { + "epoch": 1.31, + "learning_rate": 6.900062255099425e-08, + "loss": 0.6918, + "step": 35822 + }, + { + "epoch": 1.31, + "learning_rate": 6.899696048632219e-08, + "loss": 0.6952, + "step": 35823 + }, + { + "epoch": 1.31, + "learning_rate": 6.899329842165013e-08, + "loss": 0.6865, + "step": 35824 + }, + { + "epoch": 1.31, + "learning_rate": 6.898963635697807e-08, + "loss": 0.6962, + "step": 35825 + }, + { + "epoch": 1.31, + "learning_rate": 6.8985974292306e-08, + "loss": 0.7001, + "step": 35826 + }, + { + "epoch": 1.31, + "learning_rate": 6.898231222763393e-08, + "loss": 0.6837, + "step": 35827 + }, + { + "epoch": 1.31, + "learning_rate": 6.897865016296187e-08, + "loss": 0.6983, + "step": 35828 + }, + { + "epoch": 1.31, + "learning_rate": 6.897498809828981e-08, + "loss": 0.6829, + "step": 35829 + }, + { + "epoch": 1.31, + "learning_rate": 6.897132603361775e-08, + "loss": 0.7113, + "step": 35830 + }, + { + "epoch": 1.31, + "learning_rate": 6.896766396894569e-08, + "loss": 0.6915, + "step": 35831 + }, + { + "epoch": 1.31, + "learning_rate": 6.896400190427363e-08, + "loss": 0.6874, + "step": 35832 + }, + { + "epoch": 1.31, + "learning_rate": 6.896033983960157e-08, + "loss": 0.6928, + "step": 35833 + }, + { + "epoch": 1.31, + "learning_rate": 6.895667777492951e-08, + "loss": 0.6919, + "step": 35834 + }, + { + "epoch": 1.31, + "learning_rate": 6.895301571025744e-08, + "loss": 0.7035, + "step": 35835 + }, + { + "epoch": 1.31, + "learning_rate": 6.894935364558538e-08, + "loss": 0.7, + "step": 35836 + }, + { + "epoch": 1.31, + "learning_rate": 6.894569158091332e-08, + "loss": 0.6912, + "step": 35837 + }, + { + "epoch": 1.31, + "learning_rate": 6.894202951624125e-08, + "loss": 0.681, + "step": 35838 + }, + { + "epoch": 1.31, + "learning_rate": 6.893836745156919e-08, + "loss": 0.6932, + "step": 35839 + }, + { + "epoch": 1.31, + "learning_rate": 6.893470538689713e-08, + "loss": 0.692, + "step": 35840 + }, + { + "epoch": 1.31, + "learning_rate": 6.893104332222507e-08, + "loss": 0.6885, + "step": 35841 + }, + { + "epoch": 1.31, + "learning_rate": 6.8927381257553e-08, + "loss": 0.6833, + "step": 35842 + }, + { + "epoch": 1.31, + "learning_rate": 6.892371919288094e-08, + "loss": 0.6846, + "step": 35843 + }, + { + "epoch": 1.31, + "learning_rate": 6.892005712820888e-08, + "loss": 0.6868, + "step": 35844 + }, + { + "epoch": 1.31, + "learning_rate": 6.891639506353681e-08, + "loss": 0.7007, + "step": 35845 + }, + { + "epoch": 1.31, + "learning_rate": 6.891273299886475e-08, + "loss": 0.6829, + "step": 35846 + }, + { + "epoch": 1.31, + "learning_rate": 6.890907093419269e-08, + "loss": 0.6947, + "step": 35847 + }, + { + "epoch": 1.31, + "learning_rate": 6.890540886952063e-08, + "loss": 0.6875, + "step": 35848 + }, + { + "epoch": 1.31, + "learning_rate": 6.890174680484857e-08, + "loss": 0.6785, + "step": 35849 + }, + { + "epoch": 1.31, + "learning_rate": 6.889808474017651e-08, + "loss": 0.6891, + "step": 35850 + }, + { + "epoch": 1.31, + "learning_rate": 6.889442267550445e-08, + "loss": 0.7106, + "step": 35851 + }, + { + "epoch": 1.31, + "learning_rate": 6.889076061083239e-08, + "loss": 0.6935, + "step": 35852 + }, + { + "epoch": 1.31, + "learning_rate": 6.888709854616033e-08, + "loss": 0.6828, + "step": 35853 + }, + { + "epoch": 1.31, + "learning_rate": 6.888343648148826e-08, + "loss": 0.6893, + "step": 35854 + }, + { + "epoch": 1.31, + "learning_rate": 6.88797744168162e-08, + "loss": 0.6916, + "step": 35855 + }, + { + "epoch": 1.31, + "learning_rate": 6.887611235214414e-08, + "loss": 0.6945, + "step": 35856 + }, + { + "epoch": 1.31, + "learning_rate": 6.887245028747208e-08, + "loss": 0.7083, + "step": 35857 + }, + { + "epoch": 1.31, + "learning_rate": 6.886878822280002e-08, + "loss": 0.6961, + "step": 35858 + }, + { + "epoch": 1.31, + "learning_rate": 6.886512615812796e-08, + "loss": 0.6961, + "step": 35859 + }, + { + "epoch": 1.31, + "learning_rate": 6.886146409345588e-08, + "loss": 0.6797, + "step": 35860 + }, + { + "epoch": 1.31, + "learning_rate": 6.885780202878382e-08, + "loss": 0.7059, + "step": 35861 + }, + { + "epoch": 1.31, + "learning_rate": 6.885413996411176e-08, + "loss": 0.6922, + "step": 35862 + }, + { + "epoch": 1.31, + "learning_rate": 6.885047789943969e-08, + "loss": 0.7062, + "step": 35863 + }, + { + "epoch": 1.31, + "learning_rate": 6.884681583476763e-08, + "loss": 0.6834, + "step": 35864 + }, + { + "epoch": 1.31, + "learning_rate": 6.884315377009557e-08, + "loss": 0.6922, + "step": 35865 + }, + { + "epoch": 1.31, + "learning_rate": 6.883949170542351e-08, + "loss": 0.6968, + "step": 35866 + }, + { + "epoch": 1.31, + "learning_rate": 6.883582964075145e-08, + "loss": 0.6939, + "step": 35867 + }, + { + "epoch": 1.31, + "learning_rate": 6.883216757607939e-08, + "loss": 0.6855, + "step": 35868 + }, + { + "epoch": 1.31, + "learning_rate": 6.882850551140733e-08, + "loss": 0.6851, + "step": 35869 + }, + { + "epoch": 1.31, + "learning_rate": 6.882484344673527e-08, + "loss": 0.7108, + "step": 35870 + }, + { + "epoch": 1.31, + "learning_rate": 6.882118138206321e-08, + "loss": 0.7016, + "step": 35871 + }, + { + "epoch": 1.31, + "learning_rate": 6.881751931739114e-08, + "loss": 0.6935, + "step": 35872 + }, + { + "epoch": 1.31, + "learning_rate": 6.881385725271908e-08, + "loss": 0.6957, + "step": 35873 + }, + { + "epoch": 1.31, + "learning_rate": 6.881019518804702e-08, + "loss": 0.6929, + "step": 35874 + }, + { + "epoch": 1.31, + "learning_rate": 6.880653312337496e-08, + "loss": 0.6966, + "step": 35875 + }, + { + "epoch": 1.31, + "learning_rate": 6.88028710587029e-08, + "loss": 0.6901, + "step": 35876 + }, + { + "epoch": 1.31, + "learning_rate": 6.879920899403084e-08, + "loss": 0.7031, + "step": 35877 + }, + { + "epoch": 1.31, + "learning_rate": 6.879554692935878e-08, + "loss": 0.6982, + "step": 35878 + }, + { + "epoch": 1.31, + "learning_rate": 6.879188486468672e-08, + "loss": 0.6929, + "step": 35879 + }, + { + "epoch": 1.31, + "learning_rate": 6.878822280001466e-08, + "loss": 0.7037, + "step": 35880 + }, + { + "epoch": 1.31, + "learning_rate": 6.878456073534258e-08, + "loss": 0.6873, + "step": 35881 + }, + { + "epoch": 1.31, + "learning_rate": 6.878089867067052e-08, + "loss": 0.6966, + "step": 35882 + }, + { + "epoch": 1.31, + "learning_rate": 6.877723660599846e-08, + "loss": 0.6947, + "step": 35883 + }, + { + "epoch": 1.31, + "learning_rate": 6.877357454132639e-08, + "loss": 0.6886, + "step": 35884 + }, + { + "epoch": 1.31, + "learning_rate": 6.876991247665433e-08, + "loss": 0.7057, + "step": 35885 + }, + { + "epoch": 1.31, + "learning_rate": 6.876625041198227e-08, + "loss": 0.6914, + "step": 35886 + }, + { + "epoch": 1.31, + "learning_rate": 6.876258834731021e-08, + "loss": 0.6913, + "step": 35887 + }, + { + "epoch": 1.31, + "learning_rate": 6.875892628263815e-08, + "loss": 0.6866, + "step": 35888 + }, + { + "epoch": 1.31, + "learning_rate": 6.875526421796609e-08, + "loss": 0.6806, + "step": 35889 + }, + { + "epoch": 1.31, + "learning_rate": 6.875160215329402e-08, + "loss": 0.6828, + "step": 35890 + }, + { + "epoch": 1.31, + "learning_rate": 6.874794008862196e-08, + "loss": 0.6875, + "step": 35891 + }, + { + "epoch": 1.31, + "learning_rate": 6.87442780239499e-08, + "loss": 0.6917, + "step": 35892 + }, + { + "epoch": 1.31, + "learning_rate": 6.874061595927784e-08, + "loss": 0.6983, + "step": 35893 + }, + { + "epoch": 1.31, + "learning_rate": 6.873695389460578e-08, + "loss": 0.6892, + "step": 35894 + }, + { + "epoch": 1.31, + "learning_rate": 6.873329182993372e-08, + "loss": 0.681, + "step": 35895 + }, + { + "epoch": 1.31, + "learning_rate": 6.872962976526166e-08, + "loss": 0.6872, + "step": 35896 + }, + { + "epoch": 1.31, + "learning_rate": 6.87259677005896e-08, + "loss": 0.6926, + "step": 35897 + }, + { + "epoch": 1.31, + "learning_rate": 6.872230563591754e-08, + "loss": 0.6963, + "step": 35898 + }, + { + "epoch": 1.31, + "learning_rate": 6.871864357124546e-08, + "loss": 0.6896, + "step": 35899 + }, + { + "epoch": 1.31, + "learning_rate": 6.87149815065734e-08, + "loss": 0.6991, + "step": 35900 + }, + { + "epoch": 1.31, + "learning_rate": 6.871131944190134e-08, + "loss": 0.7047, + "step": 35901 + }, + { + "epoch": 1.31, + "learning_rate": 6.870765737722928e-08, + "loss": 0.6918, + "step": 35902 + }, + { + "epoch": 1.31, + "learning_rate": 6.870399531255722e-08, + "loss": 0.7039, + "step": 35903 + }, + { + "epoch": 1.31, + "learning_rate": 6.870033324788516e-08, + "loss": 0.6993, + "step": 35904 + }, + { + "epoch": 1.31, + "learning_rate": 6.86966711832131e-08, + "loss": 0.688, + "step": 35905 + }, + { + "epoch": 1.31, + "learning_rate": 6.869300911854103e-08, + "loss": 0.6965, + "step": 35906 + }, + { + "epoch": 1.31, + "learning_rate": 6.868934705386897e-08, + "loss": 0.693, + "step": 35907 + }, + { + "epoch": 1.31, + "learning_rate": 6.868568498919691e-08, + "loss": 0.6969, + "step": 35908 + }, + { + "epoch": 1.31, + "learning_rate": 6.868202292452483e-08, + "loss": 0.6881, + "step": 35909 + }, + { + "epoch": 1.31, + "learning_rate": 6.867836085985277e-08, + "loss": 0.6927, + "step": 35910 + }, + { + "epoch": 1.31, + "learning_rate": 6.867469879518071e-08, + "loss": 0.6793, + "step": 35911 + }, + { + "epoch": 1.31, + "learning_rate": 6.867103673050865e-08, + "loss": 0.697, + "step": 35912 + }, + { + "epoch": 1.31, + "learning_rate": 6.86673746658366e-08, + "loss": 0.6968, + "step": 35913 + }, + { + "epoch": 1.31, + "learning_rate": 6.866371260116453e-08, + "loss": 0.6963, + "step": 35914 + }, + { + "epoch": 1.31, + "learning_rate": 6.866005053649247e-08, + "loss": 0.694, + "step": 35915 + }, + { + "epoch": 1.31, + "learning_rate": 6.865638847182041e-08, + "loss": 0.6875, + "step": 35916 + }, + { + "epoch": 1.31, + "learning_rate": 6.865272640714834e-08, + "loss": 0.6785, + "step": 35917 + }, + { + "epoch": 1.31, + "learning_rate": 6.864906434247628e-08, + "loss": 0.6962, + "step": 35918 + }, + { + "epoch": 1.31, + "learning_rate": 6.864540227780422e-08, + "loss": 0.6914, + "step": 35919 + }, + { + "epoch": 1.31, + "learning_rate": 6.864174021313216e-08, + "loss": 0.7091, + "step": 35920 + }, + { + "epoch": 1.31, + "learning_rate": 6.86380781484601e-08, + "loss": 0.7197, + "step": 35921 + }, + { + "epoch": 1.31, + "learning_rate": 6.863441608378804e-08, + "loss": 0.6951, + "step": 35922 + }, + { + "epoch": 1.31, + "learning_rate": 6.863075401911598e-08, + "loss": 0.6977, + "step": 35923 + }, + { + "epoch": 1.31, + "learning_rate": 6.862709195444392e-08, + "loss": 0.702, + "step": 35924 + }, + { + "epoch": 1.31, + "learning_rate": 6.862342988977186e-08, + "loss": 0.701, + "step": 35925 + }, + { + "epoch": 1.31, + "learning_rate": 6.861976782509979e-08, + "loss": 0.6884, + "step": 35926 + }, + { + "epoch": 1.31, + "learning_rate": 6.861610576042773e-08, + "loss": 0.6941, + "step": 35927 + }, + { + "epoch": 1.31, + "learning_rate": 6.861244369575567e-08, + "loss": 0.6872, + "step": 35928 + }, + { + "epoch": 1.31, + "learning_rate": 6.860878163108361e-08, + "loss": 0.686, + "step": 35929 + }, + { + "epoch": 1.31, + "learning_rate": 6.860511956641155e-08, + "loss": 0.6862, + "step": 35930 + }, + { + "epoch": 1.31, + "learning_rate": 6.860145750173947e-08, + "loss": 0.6804, + "step": 35931 + }, + { + "epoch": 1.31, + "learning_rate": 6.859779543706741e-08, + "loss": 0.7083, + "step": 35932 + }, + { + "epoch": 1.31, + "learning_rate": 6.859413337239535e-08, + "loss": 0.6899, + "step": 35933 + }, + { + "epoch": 1.31, + "learning_rate": 6.859047130772329e-08, + "loss": 0.7106, + "step": 35934 + }, + { + "epoch": 1.31, + "learning_rate": 6.858680924305122e-08, + "loss": 0.6892, + "step": 35935 + }, + { + "epoch": 1.31, + "learning_rate": 6.858314717837916e-08, + "loss": 0.6914, + "step": 35936 + }, + { + "epoch": 1.31, + "learning_rate": 6.85794851137071e-08, + "loss": 0.6907, + "step": 35937 + }, + { + "epoch": 1.31, + "learning_rate": 6.857582304903504e-08, + "loss": 0.6954, + "step": 35938 + }, + { + "epoch": 1.31, + "learning_rate": 6.857216098436298e-08, + "loss": 0.69, + "step": 35939 + }, + { + "epoch": 1.31, + "learning_rate": 6.856849891969092e-08, + "loss": 0.6988, + "step": 35940 + }, + { + "epoch": 1.31, + "learning_rate": 6.856483685501886e-08, + "loss": 0.704, + "step": 35941 + }, + { + "epoch": 1.32, + "learning_rate": 6.85611747903468e-08, + "loss": 0.6949, + "step": 35942 + }, + { + "epoch": 1.32, + "learning_rate": 6.855751272567474e-08, + "loss": 0.6851, + "step": 35943 + }, + { + "epoch": 1.32, + "learning_rate": 6.855385066100267e-08, + "loss": 0.7053, + "step": 35944 + }, + { + "epoch": 1.32, + "learning_rate": 6.85501885963306e-08, + "loss": 0.6806, + "step": 35945 + }, + { + "epoch": 1.32, + "learning_rate": 6.854652653165855e-08, + "loss": 0.6749, + "step": 35946 + }, + { + "epoch": 1.32, + "learning_rate": 6.854286446698649e-08, + "loss": 0.6829, + "step": 35947 + }, + { + "epoch": 1.32, + "learning_rate": 6.853920240231443e-08, + "loss": 0.7094, + "step": 35948 + }, + { + "epoch": 1.32, + "learning_rate": 6.853554033764237e-08, + "loss": 0.6855, + "step": 35949 + }, + { + "epoch": 1.32, + "learning_rate": 6.85318782729703e-08, + "loss": 0.707, + "step": 35950 + }, + { + "epoch": 1.32, + "learning_rate": 6.852821620829825e-08, + "loss": 0.7013, + "step": 35951 + }, + { + "epoch": 1.32, + "learning_rate": 6.852455414362617e-08, + "loss": 0.6857, + "step": 35952 + }, + { + "epoch": 1.32, + "learning_rate": 6.852089207895411e-08, + "loss": 0.6902, + "step": 35953 + }, + { + "epoch": 1.32, + "learning_rate": 6.851723001428205e-08, + "loss": 0.6882, + "step": 35954 + }, + { + "epoch": 1.32, + "learning_rate": 6.851356794960998e-08, + "loss": 0.6779, + "step": 35955 + }, + { + "epoch": 1.32, + "learning_rate": 6.850990588493792e-08, + "loss": 0.692, + "step": 35956 + }, + { + "epoch": 1.32, + "learning_rate": 6.850624382026586e-08, + "loss": 0.6838, + "step": 35957 + }, + { + "epoch": 1.32, + "learning_rate": 6.85025817555938e-08, + "loss": 0.6934, + "step": 35958 + }, + { + "epoch": 1.32, + "learning_rate": 6.849891969092174e-08, + "loss": 0.6988, + "step": 35959 + }, + { + "epoch": 1.32, + "learning_rate": 6.849525762624968e-08, + "loss": 0.7001, + "step": 35960 + }, + { + "epoch": 1.32, + "learning_rate": 6.849159556157762e-08, + "loss": 0.6788, + "step": 35961 + }, + { + "epoch": 1.32, + "learning_rate": 6.848793349690555e-08, + "loss": 0.7006, + "step": 35962 + }, + { + "epoch": 1.32, + "learning_rate": 6.848427143223348e-08, + "loss": 0.712, + "step": 35963 + }, + { + "epoch": 1.32, + "learning_rate": 6.848060936756142e-08, + "loss": 0.7008, + "step": 35964 + }, + { + "epoch": 1.32, + "learning_rate": 6.847694730288936e-08, + "loss": 0.7062, + "step": 35965 + }, + { + "epoch": 1.32, + "learning_rate": 6.84732852382173e-08, + "loss": 0.7006, + "step": 35966 + }, + { + "epoch": 1.32, + "learning_rate": 6.846962317354524e-08, + "loss": 0.7015, + "step": 35967 + }, + { + "epoch": 1.32, + "learning_rate": 6.846596110887318e-08, + "loss": 0.6999, + "step": 35968 + }, + { + "epoch": 1.32, + "learning_rate": 6.846229904420112e-08, + "loss": 0.6945, + "step": 35969 + }, + { + "epoch": 1.32, + "learning_rate": 6.845863697952906e-08, + "loss": 0.6993, + "step": 35970 + }, + { + "epoch": 1.32, + "learning_rate": 6.845497491485699e-08, + "loss": 0.6912, + "step": 35971 + }, + { + "epoch": 1.32, + "learning_rate": 6.845131285018493e-08, + "loss": 0.6955, + "step": 35972 + }, + { + "epoch": 1.32, + "learning_rate": 6.844765078551287e-08, + "loss": 0.7067, + "step": 35973 + }, + { + "epoch": 1.32, + "learning_rate": 6.844398872084081e-08, + "loss": 0.6931, + "step": 35974 + }, + { + "epoch": 1.32, + "learning_rate": 6.844032665616875e-08, + "loss": 0.6891, + "step": 35975 + }, + { + "epoch": 1.32, + "learning_rate": 6.843666459149669e-08, + "loss": 0.6879, + "step": 35976 + }, + { + "epoch": 1.32, + "learning_rate": 6.843300252682462e-08, + "loss": 0.6972, + "step": 35977 + }, + { + "epoch": 1.32, + "learning_rate": 6.842934046215256e-08, + "loss": 0.6989, + "step": 35978 + }, + { + "epoch": 1.32, + "learning_rate": 6.84256783974805e-08, + "loss": 0.7079, + "step": 35979 + }, + { + "epoch": 1.32, + "learning_rate": 6.842201633280842e-08, + "loss": 0.6909, + "step": 35980 + }, + { + "epoch": 1.32, + "learning_rate": 6.841835426813636e-08, + "loss": 0.6931, + "step": 35981 + }, + { + "epoch": 1.32, + "learning_rate": 6.84146922034643e-08, + "loss": 0.6955, + "step": 35982 + }, + { + "epoch": 1.32, + "learning_rate": 6.841103013879224e-08, + "loss": 0.6884, + "step": 35983 + }, + { + "epoch": 1.32, + "learning_rate": 6.840736807412018e-08, + "loss": 0.7012, + "step": 35984 + }, + { + "epoch": 1.32, + "learning_rate": 6.840370600944812e-08, + "loss": 0.69, + "step": 35985 + }, + { + "epoch": 1.32, + "learning_rate": 6.840004394477606e-08, + "loss": 0.6779, + "step": 35986 + }, + { + "epoch": 1.32, + "learning_rate": 6.8396381880104e-08, + "loss": 0.6876, + "step": 35987 + }, + { + "epoch": 1.32, + "learning_rate": 6.839271981543194e-08, + "loss": 0.692, + "step": 35988 + }, + { + "epoch": 1.32, + "learning_rate": 6.838905775075987e-08, + "loss": 0.6882, + "step": 35989 + }, + { + "epoch": 1.32, + "learning_rate": 6.838539568608781e-08, + "loss": 0.6881, + "step": 35990 + }, + { + "epoch": 1.32, + "learning_rate": 6.838173362141575e-08, + "loss": 0.6835, + "step": 35991 + }, + { + "epoch": 1.32, + "learning_rate": 6.837807155674369e-08, + "loss": 0.6997, + "step": 35992 + }, + { + "epoch": 1.32, + "learning_rate": 6.837440949207163e-08, + "loss": 0.709, + "step": 35993 + }, + { + "epoch": 1.32, + "learning_rate": 6.837074742739957e-08, + "loss": 0.6908, + "step": 35994 + }, + { + "epoch": 1.32, + "learning_rate": 6.836708536272751e-08, + "loss": 0.6927, + "step": 35995 + }, + { + "epoch": 1.32, + "learning_rate": 6.836342329805545e-08, + "loss": 0.6897, + "step": 35996 + }, + { + "epoch": 1.32, + "learning_rate": 6.835976123338339e-08, + "loss": 0.6864, + "step": 35997 + }, + { + "epoch": 1.32, + "learning_rate": 6.835609916871132e-08, + "loss": 0.6975, + "step": 35998 + }, + { + "epoch": 1.32, + "learning_rate": 6.835243710403926e-08, + "loss": 0.6949, + "step": 35999 + }, + { + "epoch": 1.32, + "learning_rate": 6.83487750393672e-08, + "loss": 0.7001, + "step": 36000 + }, + { + "epoch": 1.32, + "learning_rate": 6.834511297469512e-08, + "loss": 0.6845, + "step": 36001 + }, + { + "epoch": 1.32, + "learning_rate": 6.834145091002306e-08, + "loss": 0.6812, + "step": 36002 + }, + { + "epoch": 1.32, + "learning_rate": 6.8337788845351e-08, + "loss": 0.6971, + "step": 36003 + }, + { + "epoch": 1.32, + "learning_rate": 6.833412678067894e-08, + "loss": 0.684, + "step": 36004 + }, + { + "epoch": 1.32, + "learning_rate": 6.833046471600688e-08, + "loss": 0.6882, + "step": 36005 + }, + { + "epoch": 1.32, + "learning_rate": 6.832680265133482e-08, + "loss": 0.6915, + "step": 36006 + }, + { + "epoch": 1.32, + "learning_rate": 6.832314058666275e-08, + "loss": 0.7059, + "step": 36007 + }, + { + "epoch": 1.32, + "learning_rate": 6.831947852199069e-08, + "loss": 0.684, + "step": 36008 + }, + { + "epoch": 1.32, + "learning_rate": 6.831581645731863e-08, + "loss": 0.6883, + "step": 36009 + }, + { + "epoch": 1.32, + "learning_rate": 6.831215439264657e-08, + "loss": 0.6802, + "step": 36010 + }, + { + "epoch": 1.32, + "learning_rate": 6.830849232797451e-08, + "loss": 0.7078, + "step": 36011 + }, + { + "epoch": 1.32, + "learning_rate": 6.830483026330245e-08, + "loss": 0.6877, + "step": 36012 + }, + { + "epoch": 1.32, + "learning_rate": 6.830116819863039e-08, + "loss": 0.6956, + "step": 36013 + }, + { + "epoch": 1.32, + "learning_rate": 6.829750613395833e-08, + "loss": 0.6968, + "step": 36014 + }, + { + "epoch": 1.32, + "learning_rate": 6.829384406928626e-08, + "loss": 0.6882, + "step": 36015 + }, + { + "epoch": 1.32, + "learning_rate": 6.82901820046142e-08, + "loss": 0.6974, + "step": 36016 + }, + { + "epoch": 1.32, + "learning_rate": 6.828651993994214e-08, + "loss": 0.7033, + "step": 36017 + }, + { + "epoch": 1.32, + "learning_rate": 6.828285787527008e-08, + "loss": 0.6837, + "step": 36018 + }, + { + "epoch": 1.32, + "learning_rate": 6.827919581059802e-08, + "loss": 0.6914, + "step": 36019 + }, + { + "epoch": 1.32, + "learning_rate": 6.827553374592595e-08, + "loss": 0.6827, + "step": 36020 + }, + { + "epoch": 1.32, + "learning_rate": 6.82718716812539e-08, + "loss": 0.6893, + "step": 36021 + }, + { + "epoch": 1.32, + "learning_rate": 6.826820961658183e-08, + "loss": 0.7095, + "step": 36022 + }, + { + "epoch": 1.32, + "learning_rate": 6.826454755190976e-08, + "loss": 0.7116, + "step": 36023 + }, + { + "epoch": 1.32, + "learning_rate": 6.82608854872377e-08, + "loss": 0.6814, + "step": 36024 + }, + { + "epoch": 1.32, + "learning_rate": 6.825722342256564e-08, + "loss": 0.6956, + "step": 36025 + }, + { + "epoch": 1.32, + "learning_rate": 6.825356135789357e-08, + "loss": 0.6959, + "step": 36026 + }, + { + "epoch": 1.32, + "learning_rate": 6.824989929322151e-08, + "loss": 0.6866, + "step": 36027 + }, + { + "epoch": 1.32, + "learning_rate": 6.824623722854945e-08, + "loss": 0.6772, + "step": 36028 + }, + { + "epoch": 1.32, + "learning_rate": 6.824257516387739e-08, + "loss": 0.6948, + "step": 36029 + }, + { + "epoch": 1.32, + "learning_rate": 6.823891309920533e-08, + "loss": 0.6905, + "step": 36030 + }, + { + "epoch": 1.32, + "learning_rate": 6.823525103453327e-08, + "loss": 0.6841, + "step": 36031 + }, + { + "epoch": 1.32, + "learning_rate": 6.823158896986121e-08, + "loss": 0.6964, + "step": 36032 + }, + { + "epoch": 1.32, + "learning_rate": 6.822792690518913e-08, + "loss": 0.7036, + "step": 36033 + }, + { + "epoch": 1.32, + "learning_rate": 6.822426484051707e-08, + "loss": 0.689, + "step": 36034 + }, + { + "epoch": 1.32, + "learning_rate": 6.822060277584501e-08, + "loss": 0.691, + "step": 36035 + }, + { + "epoch": 1.32, + "learning_rate": 6.821694071117295e-08, + "loss": 0.6886, + "step": 36036 + }, + { + "epoch": 1.32, + "learning_rate": 6.82132786465009e-08, + "loss": 0.7072, + "step": 36037 + }, + { + "epoch": 1.32, + "learning_rate": 6.820961658182883e-08, + "loss": 0.7072, + "step": 36038 + }, + { + "epoch": 1.32, + "learning_rate": 6.820595451715677e-08, + "loss": 0.6775, + "step": 36039 + }, + { + "epoch": 1.32, + "learning_rate": 6.820229245248471e-08, + "loss": 0.7022, + "step": 36040 + }, + { + "epoch": 1.32, + "learning_rate": 6.819863038781265e-08, + "loss": 0.6811, + "step": 36041 + }, + { + "epoch": 1.32, + "learning_rate": 6.819496832314058e-08, + "loss": 0.6975, + "step": 36042 + }, + { + "epoch": 1.32, + "learning_rate": 6.819130625846852e-08, + "loss": 0.7029, + "step": 36043 + }, + { + "epoch": 1.32, + "learning_rate": 6.818764419379646e-08, + "loss": 0.6946, + "step": 36044 + }, + { + "epoch": 1.32, + "learning_rate": 6.81839821291244e-08, + "loss": 0.695, + "step": 36045 + }, + { + "epoch": 1.32, + "learning_rate": 6.818032006445234e-08, + "loss": 0.6842, + "step": 36046 + }, + { + "epoch": 1.32, + "learning_rate": 6.817665799978028e-08, + "loss": 0.7019, + "step": 36047 + }, + { + "epoch": 1.32, + "learning_rate": 6.817299593510821e-08, + "loss": 0.6892, + "step": 36048 + }, + { + "epoch": 1.32, + "learning_rate": 6.816933387043615e-08, + "loss": 0.6951, + "step": 36049 + }, + { + "epoch": 1.32, + "learning_rate": 6.816567180576409e-08, + "loss": 0.6988, + "step": 36050 + }, + { + "epoch": 1.32, + "learning_rate": 6.816200974109201e-08, + "loss": 0.7117, + "step": 36051 + }, + { + "epoch": 1.32, + "learning_rate": 6.815834767641995e-08, + "loss": 0.7005, + "step": 36052 + }, + { + "epoch": 1.32, + "learning_rate": 6.815468561174789e-08, + "loss": 0.6953, + "step": 36053 + }, + { + "epoch": 1.32, + "learning_rate": 6.815102354707583e-08, + "loss": 0.6902, + "step": 36054 + }, + { + "epoch": 1.32, + "learning_rate": 6.814736148240377e-08, + "loss": 0.7078, + "step": 36055 + }, + { + "epoch": 1.32, + "learning_rate": 6.814369941773171e-08, + "loss": 0.6825, + "step": 36056 + }, + { + "epoch": 1.32, + "learning_rate": 6.814003735305965e-08, + "loss": 0.6985, + "step": 36057 + }, + { + "epoch": 1.32, + "learning_rate": 6.813637528838759e-08, + "loss": 0.7, + "step": 36058 + }, + { + "epoch": 1.32, + "learning_rate": 6.813271322371553e-08, + "loss": 0.7098, + "step": 36059 + }, + { + "epoch": 1.32, + "learning_rate": 6.812905115904346e-08, + "loss": 0.6907, + "step": 36060 + }, + { + "epoch": 1.32, + "learning_rate": 6.81253890943714e-08, + "loss": 0.687, + "step": 36061 + }, + { + "epoch": 1.32, + "learning_rate": 6.812172702969934e-08, + "loss": 0.6891, + "step": 36062 + }, + { + "epoch": 1.32, + "learning_rate": 6.811806496502728e-08, + "loss": 0.6905, + "step": 36063 + }, + { + "epoch": 1.32, + "learning_rate": 6.811440290035522e-08, + "loss": 0.6995, + "step": 36064 + }, + { + "epoch": 1.32, + "learning_rate": 6.811074083568316e-08, + "loss": 0.6961, + "step": 36065 + }, + { + "epoch": 1.32, + "learning_rate": 6.81070787710111e-08, + "loss": 0.6822, + "step": 36066 + }, + { + "epoch": 1.32, + "learning_rate": 6.810341670633904e-08, + "loss": 0.7009, + "step": 36067 + }, + { + "epoch": 1.32, + "learning_rate": 6.809975464166698e-08, + "loss": 0.6935, + "step": 36068 + }, + { + "epoch": 1.32, + "learning_rate": 6.80960925769949e-08, + "loss": 0.7021, + "step": 36069 + }, + { + "epoch": 1.32, + "learning_rate": 6.809243051232285e-08, + "loss": 0.692, + "step": 36070 + }, + { + "epoch": 1.32, + "learning_rate": 6.808876844765079e-08, + "loss": 0.6871, + "step": 36071 + }, + { + "epoch": 1.32, + "learning_rate": 6.808510638297871e-08, + "loss": 0.7089, + "step": 36072 + }, + { + "epoch": 1.32, + "learning_rate": 6.808144431830665e-08, + "loss": 0.7066, + "step": 36073 + }, + { + "epoch": 1.32, + "learning_rate": 6.807778225363459e-08, + "loss": 0.6815, + "step": 36074 + }, + { + "epoch": 1.32, + "learning_rate": 6.807412018896253e-08, + "loss": 0.6993, + "step": 36075 + }, + { + "epoch": 1.32, + "learning_rate": 6.807045812429047e-08, + "loss": 0.6898, + "step": 36076 + }, + { + "epoch": 1.32, + "learning_rate": 6.806679605961841e-08, + "loss": 0.692, + "step": 36077 + }, + { + "epoch": 1.32, + "learning_rate": 6.806313399494634e-08, + "loss": 0.6745, + "step": 36078 + }, + { + "epoch": 1.32, + "learning_rate": 6.805947193027428e-08, + "loss": 0.708, + "step": 36079 + }, + { + "epoch": 1.32, + "learning_rate": 6.805580986560222e-08, + "loss": 0.6824, + "step": 36080 + }, + { + "epoch": 1.32, + "learning_rate": 6.805214780093016e-08, + "loss": 0.6916, + "step": 36081 + }, + { + "epoch": 1.32, + "learning_rate": 6.80484857362581e-08, + "loss": 0.689, + "step": 36082 + }, + { + "epoch": 1.32, + "learning_rate": 6.804482367158604e-08, + "loss": 0.6807, + "step": 36083 + }, + { + "epoch": 1.32, + "learning_rate": 6.804116160691398e-08, + "loss": 0.7, + "step": 36084 + }, + { + "epoch": 1.32, + "learning_rate": 6.803749954224192e-08, + "loss": 0.6895, + "step": 36085 + }, + { + "epoch": 1.32, + "learning_rate": 6.803383747756986e-08, + "loss": 0.6879, + "step": 36086 + }, + { + "epoch": 1.32, + "learning_rate": 6.803017541289778e-08, + "loss": 0.6996, + "step": 36087 + }, + { + "epoch": 1.32, + "learning_rate": 6.802651334822572e-08, + "loss": 0.7016, + "step": 36088 + }, + { + "epoch": 1.32, + "learning_rate": 6.802285128355366e-08, + "loss": 0.6882, + "step": 36089 + }, + { + "epoch": 1.32, + "learning_rate": 6.80191892188816e-08, + "loss": 0.7018, + "step": 36090 + }, + { + "epoch": 1.32, + "learning_rate": 6.801552715420954e-08, + "loss": 0.6955, + "step": 36091 + }, + { + "epoch": 1.32, + "learning_rate": 6.801186508953748e-08, + "loss": 0.6862, + "step": 36092 + }, + { + "epoch": 1.32, + "learning_rate": 6.800820302486542e-08, + "loss": 0.6941, + "step": 36093 + }, + { + "epoch": 1.32, + "learning_rate": 6.800454096019335e-08, + "loss": 0.6956, + "step": 36094 + }, + { + "epoch": 1.32, + "learning_rate": 6.800087889552129e-08, + "loss": 0.7016, + "step": 36095 + }, + { + "epoch": 1.32, + "learning_rate": 6.799721683084923e-08, + "loss": 0.7005, + "step": 36096 + }, + { + "epoch": 1.32, + "learning_rate": 6.799355476617716e-08, + "loss": 0.6857, + "step": 36097 + }, + { + "epoch": 1.32, + "learning_rate": 6.79898927015051e-08, + "loss": 0.6925, + "step": 36098 + }, + { + "epoch": 1.32, + "learning_rate": 6.798623063683304e-08, + "loss": 0.7052, + "step": 36099 + }, + { + "epoch": 1.32, + "learning_rate": 6.798256857216098e-08, + "loss": 0.6855, + "step": 36100 + }, + { + "epoch": 1.32, + "learning_rate": 6.797890650748892e-08, + "loss": 0.6873, + "step": 36101 + }, + { + "epoch": 1.32, + "learning_rate": 6.797524444281686e-08, + "loss": 0.6868, + "step": 36102 + }, + { + "epoch": 1.32, + "learning_rate": 6.79715823781448e-08, + "loss": 0.6953, + "step": 36103 + }, + { + "epoch": 1.32, + "learning_rate": 6.796792031347274e-08, + "loss": 0.699, + "step": 36104 + }, + { + "epoch": 1.32, + "learning_rate": 6.796425824880066e-08, + "loss": 0.689, + "step": 36105 + }, + { + "epoch": 1.32, + "learning_rate": 6.79605961841286e-08, + "loss": 0.6989, + "step": 36106 + }, + { + "epoch": 1.32, + "learning_rate": 6.795693411945654e-08, + "loss": 0.6934, + "step": 36107 + }, + { + "epoch": 1.32, + "learning_rate": 6.795327205478448e-08, + "loss": 0.6907, + "step": 36108 + }, + { + "epoch": 1.32, + "learning_rate": 6.794960999011242e-08, + "loss": 0.6891, + "step": 36109 + }, + { + "epoch": 1.32, + "learning_rate": 6.794594792544036e-08, + "loss": 0.7017, + "step": 36110 + }, + { + "epoch": 1.32, + "learning_rate": 6.79422858607683e-08, + "loss": 0.6853, + "step": 36111 + }, + { + "epoch": 1.32, + "learning_rate": 6.793862379609624e-08, + "loss": 0.6894, + "step": 36112 + }, + { + "epoch": 1.32, + "learning_rate": 6.793496173142418e-08, + "loss": 0.6905, + "step": 36113 + }, + { + "epoch": 1.32, + "learning_rate": 6.793129966675211e-08, + "loss": 0.6972, + "step": 36114 + }, + { + "epoch": 1.32, + "learning_rate": 6.792763760208005e-08, + "loss": 0.6745, + "step": 36115 + }, + { + "epoch": 1.32, + "learning_rate": 6.792397553740799e-08, + "loss": 0.6809, + "step": 36116 + }, + { + "epoch": 1.32, + "learning_rate": 6.792031347273593e-08, + "loss": 0.6866, + "step": 36117 + }, + { + "epoch": 1.32, + "learning_rate": 6.791665140806387e-08, + "loss": 0.6904, + "step": 36118 + }, + { + "epoch": 1.32, + "learning_rate": 6.79129893433918e-08, + "loss": 0.7117, + "step": 36119 + }, + { + "epoch": 1.32, + "learning_rate": 6.790932727871974e-08, + "loss": 0.6968, + "step": 36120 + }, + { + "epoch": 1.32, + "learning_rate": 6.790566521404768e-08, + "loss": 0.6828, + "step": 36121 + }, + { + "epoch": 1.32, + "learning_rate": 6.790200314937562e-08, + "loss": 0.7065, + "step": 36122 + }, + { + "epoch": 1.32, + "learning_rate": 6.789834108470354e-08, + "loss": 0.6901, + "step": 36123 + }, + { + "epoch": 1.32, + "learning_rate": 6.789467902003148e-08, + "loss": 0.6894, + "step": 36124 + }, + { + "epoch": 1.32, + "learning_rate": 6.789101695535942e-08, + "loss": 0.6949, + "step": 36125 + }, + { + "epoch": 1.32, + "learning_rate": 6.788735489068736e-08, + "loss": 0.6914, + "step": 36126 + }, + { + "epoch": 1.32, + "learning_rate": 6.78836928260153e-08, + "loss": 0.6988, + "step": 36127 + }, + { + "epoch": 1.32, + "learning_rate": 6.788003076134324e-08, + "loss": 0.6966, + "step": 36128 + }, + { + "epoch": 1.32, + "learning_rate": 6.787636869667118e-08, + "loss": 0.696, + "step": 36129 + }, + { + "epoch": 1.32, + "learning_rate": 6.787270663199912e-08, + "loss": 0.6809, + "step": 36130 + }, + { + "epoch": 1.32, + "learning_rate": 6.786904456732706e-08, + "loss": 0.7013, + "step": 36131 + }, + { + "epoch": 1.32, + "learning_rate": 6.786538250265499e-08, + "loss": 0.6927, + "step": 36132 + }, + { + "epoch": 1.32, + "learning_rate": 6.786172043798293e-08, + "loss": 0.7027, + "step": 36133 + }, + { + "epoch": 1.32, + "learning_rate": 6.785805837331087e-08, + "loss": 0.6844, + "step": 36134 + }, + { + "epoch": 1.32, + "learning_rate": 6.785439630863881e-08, + "loss": 0.697, + "step": 36135 + }, + { + "epoch": 1.32, + "learning_rate": 6.785073424396675e-08, + "loss": 0.6957, + "step": 36136 + }, + { + "epoch": 1.32, + "learning_rate": 6.784707217929469e-08, + "loss": 0.6966, + "step": 36137 + }, + { + "epoch": 1.32, + "learning_rate": 6.784341011462263e-08, + "loss": 0.6842, + "step": 36138 + }, + { + "epoch": 1.32, + "learning_rate": 6.783974804995057e-08, + "loss": 0.691, + "step": 36139 + }, + { + "epoch": 1.32, + "learning_rate": 6.783608598527851e-08, + "loss": 0.6854, + "step": 36140 + }, + { + "epoch": 1.32, + "learning_rate": 6.783242392060643e-08, + "loss": 0.6998, + "step": 36141 + }, + { + "epoch": 1.32, + "learning_rate": 6.782876185593437e-08, + "loss": 0.7004, + "step": 36142 + }, + { + "epoch": 1.32, + "learning_rate": 6.78250997912623e-08, + "loss": 0.6991, + "step": 36143 + }, + { + "epoch": 1.32, + "learning_rate": 6.782143772659024e-08, + "loss": 0.6953, + "step": 36144 + }, + { + "epoch": 1.32, + "learning_rate": 6.781777566191818e-08, + "loss": 0.6882, + "step": 36145 + }, + { + "epoch": 1.32, + "learning_rate": 6.781411359724612e-08, + "loss": 0.707, + "step": 36146 + }, + { + "epoch": 1.32, + "learning_rate": 6.781045153257406e-08, + "loss": 0.701, + "step": 36147 + }, + { + "epoch": 1.32, + "learning_rate": 6.7806789467902e-08, + "loss": 0.6903, + "step": 36148 + }, + { + "epoch": 1.32, + "learning_rate": 6.780312740322994e-08, + "loss": 0.7006, + "step": 36149 + }, + { + "epoch": 1.32, + "learning_rate": 6.779946533855787e-08, + "loss": 0.6894, + "step": 36150 + }, + { + "epoch": 1.32, + "learning_rate": 6.779580327388581e-08, + "loss": 0.6887, + "step": 36151 + }, + { + "epoch": 1.32, + "learning_rate": 6.779214120921375e-08, + "loss": 0.6885, + "step": 36152 + }, + { + "epoch": 1.32, + "learning_rate": 6.778847914454169e-08, + "loss": 0.7005, + "step": 36153 + }, + { + "epoch": 1.32, + "learning_rate": 6.778481707986963e-08, + "loss": 0.6954, + "step": 36154 + }, + { + "epoch": 1.32, + "learning_rate": 6.778115501519757e-08, + "loss": 0.6933, + "step": 36155 + }, + { + "epoch": 1.32, + "learning_rate": 6.777749295052551e-08, + "loss": 0.6872, + "step": 36156 + }, + { + "epoch": 1.32, + "learning_rate": 6.777383088585345e-08, + "loss": 0.6941, + "step": 36157 + }, + { + "epoch": 1.32, + "learning_rate": 6.777016882118139e-08, + "loss": 0.6936, + "step": 36158 + }, + { + "epoch": 1.32, + "learning_rate": 6.776650675650931e-08, + "loss": 0.6877, + "step": 36159 + }, + { + "epoch": 1.32, + "learning_rate": 6.776284469183725e-08, + "loss": 0.694, + "step": 36160 + }, + { + "epoch": 1.32, + "learning_rate": 6.77591826271652e-08, + "loss": 0.6964, + "step": 36161 + }, + { + "epoch": 1.32, + "learning_rate": 6.775552056249313e-08, + "loss": 0.6976, + "step": 36162 + }, + { + "epoch": 1.32, + "learning_rate": 6.775185849782107e-08, + "loss": 0.6986, + "step": 36163 + }, + { + "epoch": 1.32, + "learning_rate": 6.774819643314901e-08, + "loss": 0.7015, + "step": 36164 + }, + { + "epoch": 1.32, + "learning_rate": 6.774453436847694e-08, + "loss": 0.703, + "step": 36165 + }, + { + "epoch": 1.32, + "learning_rate": 6.774087230380488e-08, + "loss": 0.6884, + "step": 36166 + }, + { + "epoch": 1.32, + "learning_rate": 6.773721023913282e-08, + "loss": 0.708, + "step": 36167 + }, + { + "epoch": 1.32, + "learning_rate": 6.773354817446075e-08, + "loss": 0.6911, + "step": 36168 + }, + { + "epoch": 1.32, + "learning_rate": 6.772988610978869e-08, + "loss": 0.6874, + "step": 36169 + }, + { + "epoch": 1.32, + "learning_rate": 6.772622404511663e-08, + "loss": 0.7102, + "step": 36170 + }, + { + "epoch": 1.32, + "learning_rate": 6.772256198044457e-08, + "loss": 0.7014, + "step": 36171 + }, + { + "epoch": 1.32, + "learning_rate": 6.77188999157725e-08, + "loss": 0.698, + "step": 36172 + }, + { + "epoch": 1.32, + "learning_rate": 6.771523785110045e-08, + "loss": 0.6888, + "step": 36173 + }, + { + "epoch": 1.32, + "learning_rate": 6.771157578642839e-08, + "loss": 0.689, + "step": 36174 + }, + { + "epoch": 1.32, + "learning_rate": 6.770791372175633e-08, + "loss": 0.6966, + "step": 36175 + }, + { + "epoch": 1.32, + "learning_rate": 6.770425165708427e-08, + "loss": 0.684, + "step": 36176 + }, + { + "epoch": 1.32, + "learning_rate": 6.770058959241219e-08, + "loss": 0.6954, + "step": 36177 + }, + { + "epoch": 1.32, + "learning_rate": 6.769692752774013e-08, + "loss": 0.6928, + "step": 36178 + }, + { + "epoch": 1.32, + "learning_rate": 6.769326546306807e-08, + "loss": 0.6872, + "step": 36179 + }, + { + "epoch": 1.32, + "learning_rate": 6.768960339839601e-08, + "loss": 0.6939, + "step": 36180 + }, + { + "epoch": 1.32, + "learning_rate": 6.768594133372395e-08, + "loss": 0.6936, + "step": 36181 + }, + { + "epoch": 1.32, + "learning_rate": 6.768227926905189e-08, + "loss": 0.6969, + "step": 36182 + }, + { + "epoch": 1.32, + "learning_rate": 6.767861720437983e-08, + "loss": 0.6908, + "step": 36183 + }, + { + "epoch": 1.32, + "learning_rate": 6.767495513970777e-08, + "loss": 0.6916, + "step": 36184 + }, + { + "epoch": 1.32, + "learning_rate": 6.767129307503571e-08, + "loss": 0.7079, + "step": 36185 + }, + { + "epoch": 1.32, + "learning_rate": 6.766763101036364e-08, + "loss": 0.6997, + "step": 36186 + }, + { + "epoch": 1.32, + "learning_rate": 6.766396894569158e-08, + "loss": 0.6897, + "step": 36187 + }, + { + "epoch": 1.32, + "learning_rate": 6.766030688101952e-08, + "loss": 0.7019, + "step": 36188 + }, + { + "epoch": 1.32, + "learning_rate": 6.765664481634746e-08, + "loss": 0.7085, + "step": 36189 + }, + { + "epoch": 1.32, + "learning_rate": 6.765298275167539e-08, + "loss": 0.6958, + "step": 36190 + }, + { + "epoch": 1.32, + "learning_rate": 6.764932068700333e-08, + "loss": 0.7078, + "step": 36191 + }, + { + "epoch": 1.32, + "learning_rate": 6.764565862233127e-08, + "loss": 0.7008, + "step": 36192 + }, + { + "epoch": 1.32, + "learning_rate": 6.76419965576592e-08, + "loss": 0.6944, + "step": 36193 + }, + { + "epoch": 1.32, + "learning_rate": 6.763833449298715e-08, + "loss": 0.6846, + "step": 36194 + }, + { + "epoch": 1.32, + "learning_rate": 6.763467242831507e-08, + "loss": 0.6879, + "step": 36195 + }, + { + "epoch": 1.32, + "learning_rate": 6.763101036364301e-08, + "loss": 0.6869, + "step": 36196 + }, + { + "epoch": 1.32, + "learning_rate": 6.762734829897095e-08, + "loss": 0.7002, + "step": 36197 + }, + { + "epoch": 1.32, + "learning_rate": 6.762368623429889e-08, + "loss": 0.7069, + "step": 36198 + }, + { + "epoch": 1.32, + "learning_rate": 6.762002416962683e-08, + "loss": 0.6924, + "step": 36199 + }, + { + "epoch": 1.32, + "learning_rate": 6.761636210495477e-08, + "loss": 0.7044, + "step": 36200 + }, + { + "epoch": 1.32, + "learning_rate": 6.761270004028271e-08, + "loss": 0.6819, + "step": 36201 + }, + { + "epoch": 1.32, + "learning_rate": 6.760903797561065e-08, + "loss": 0.6915, + "step": 36202 + }, + { + "epoch": 1.32, + "learning_rate": 6.760537591093859e-08, + "loss": 0.681, + "step": 36203 + }, + { + "epoch": 1.32, + "learning_rate": 6.760171384626652e-08, + "loss": 0.6929, + "step": 36204 + }, + { + "epoch": 1.32, + "learning_rate": 6.759805178159446e-08, + "loss": 0.6878, + "step": 36205 + }, + { + "epoch": 1.32, + "learning_rate": 6.75943897169224e-08, + "loss": 0.6839, + "step": 36206 + }, + { + "epoch": 1.32, + "learning_rate": 6.759072765225034e-08, + "loss": 0.6822, + "step": 36207 + }, + { + "epoch": 1.32, + "learning_rate": 6.758706558757828e-08, + "loss": 0.677, + "step": 36208 + }, + { + "epoch": 1.32, + "learning_rate": 6.758340352290622e-08, + "loss": 0.7014, + "step": 36209 + }, + { + "epoch": 1.32, + "learning_rate": 6.757974145823416e-08, + "loss": 0.6808, + "step": 36210 + }, + { + "epoch": 1.32, + "learning_rate": 6.75760793935621e-08, + "loss": 0.6945, + "step": 36211 + }, + { + "epoch": 1.32, + "learning_rate": 6.757241732889002e-08, + "loss": 0.7007, + "step": 36212 + }, + { + "epoch": 1.32, + "learning_rate": 6.756875526421796e-08, + "loss": 0.6958, + "step": 36213 + }, + { + "epoch": 1.32, + "learning_rate": 6.756509319954589e-08, + "loss": 0.6941, + "step": 36214 + }, + { + "epoch": 1.33, + "learning_rate": 6.756143113487383e-08, + "loss": 0.6878, + "step": 36215 + }, + { + "epoch": 1.33, + "learning_rate": 6.755776907020177e-08, + "loss": 0.6934, + "step": 36216 + }, + { + "epoch": 1.33, + "learning_rate": 6.755410700552971e-08, + "loss": 0.7105, + "step": 36217 + }, + { + "epoch": 1.33, + "learning_rate": 6.755044494085765e-08, + "loss": 0.6862, + "step": 36218 + }, + { + "epoch": 1.33, + "learning_rate": 6.754678287618559e-08, + "loss": 0.7054, + "step": 36219 + }, + { + "epoch": 1.33, + "learning_rate": 6.754312081151353e-08, + "loss": 0.7097, + "step": 36220 + }, + { + "epoch": 1.33, + "learning_rate": 6.753945874684147e-08, + "loss": 0.7111, + "step": 36221 + }, + { + "epoch": 1.33, + "learning_rate": 6.75357966821694e-08, + "loss": 0.7021, + "step": 36222 + }, + { + "epoch": 1.33, + "learning_rate": 6.753213461749734e-08, + "loss": 0.7042, + "step": 36223 + }, + { + "epoch": 1.33, + "learning_rate": 6.752847255282528e-08, + "loss": 0.6957, + "step": 36224 + }, + { + "epoch": 1.33, + "learning_rate": 6.752481048815322e-08, + "loss": 0.6985, + "step": 36225 + }, + { + "epoch": 1.33, + "learning_rate": 6.752114842348116e-08, + "loss": 0.6993, + "step": 36226 + }, + { + "epoch": 1.33, + "learning_rate": 6.75174863588091e-08, + "loss": 0.6924, + "step": 36227 + }, + { + "epoch": 1.33, + "learning_rate": 6.751382429413704e-08, + "loss": 0.7027, + "step": 36228 + }, + { + "epoch": 1.33, + "learning_rate": 6.751016222946498e-08, + "loss": 0.6982, + "step": 36229 + }, + { + "epoch": 1.33, + "learning_rate": 6.750650016479292e-08, + "loss": 0.685, + "step": 36230 + }, + { + "epoch": 1.33, + "learning_rate": 6.750283810012084e-08, + "loss": 0.6949, + "step": 36231 + }, + { + "epoch": 1.33, + "learning_rate": 6.749917603544878e-08, + "loss": 0.7054, + "step": 36232 + }, + { + "epoch": 1.33, + "learning_rate": 6.749551397077672e-08, + "loss": 0.6826, + "step": 36233 + }, + { + "epoch": 1.33, + "learning_rate": 6.749185190610466e-08, + "loss": 0.6924, + "step": 36234 + }, + { + "epoch": 1.33, + "learning_rate": 6.74881898414326e-08, + "loss": 0.6923, + "step": 36235 + }, + { + "epoch": 1.33, + "learning_rate": 6.748452777676053e-08, + "loss": 0.6897, + "step": 36236 + }, + { + "epoch": 1.33, + "learning_rate": 6.748086571208847e-08, + "loss": 0.7011, + "step": 36237 + }, + { + "epoch": 1.33, + "learning_rate": 6.747720364741641e-08, + "loss": 0.6872, + "step": 36238 + }, + { + "epoch": 1.33, + "learning_rate": 6.747354158274435e-08, + "loss": 0.6947, + "step": 36239 + }, + { + "epoch": 1.33, + "learning_rate": 6.746987951807228e-08, + "loss": 0.6938, + "step": 36240 + }, + { + "epoch": 1.33, + "learning_rate": 6.746621745340022e-08, + "loss": 0.6923, + "step": 36241 + }, + { + "epoch": 1.33, + "learning_rate": 6.746255538872816e-08, + "loss": 0.6927, + "step": 36242 + }, + { + "epoch": 1.33, + "learning_rate": 6.74588933240561e-08, + "loss": 0.688, + "step": 36243 + }, + { + "epoch": 1.33, + "learning_rate": 6.745523125938404e-08, + "loss": 0.6862, + "step": 36244 + }, + { + "epoch": 1.33, + "learning_rate": 6.745156919471198e-08, + "loss": 0.6991, + "step": 36245 + }, + { + "epoch": 1.33, + "learning_rate": 6.744790713003992e-08, + "loss": 0.7046, + "step": 36246 + }, + { + "epoch": 1.33, + "learning_rate": 6.744424506536786e-08, + "loss": 0.6868, + "step": 36247 + }, + { + "epoch": 1.33, + "learning_rate": 6.74405830006958e-08, + "loss": 0.6914, + "step": 36248 + }, + { + "epoch": 1.33, + "learning_rate": 6.743692093602372e-08, + "loss": 0.6981, + "step": 36249 + }, + { + "epoch": 1.33, + "learning_rate": 6.743325887135166e-08, + "loss": 0.6867, + "step": 36250 + }, + { + "epoch": 1.33, + "learning_rate": 6.74295968066796e-08, + "loss": 0.686, + "step": 36251 + }, + { + "epoch": 1.33, + "learning_rate": 6.742593474200754e-08, + "loss": 0.6888, + "step": 36252 + }, + { + "epoch": 1.33, + "learning_rate": 6.742227267733548e-08, + "loss": 0.7134, + "step": 36253 + }, + { + "epoch": 1.33, + "learning_rate": 6.741861061266342e-08, + "loss": 0.6918, + "step": 36254 + }, + { + "epoch": 1.33, + "learning_rate": 6.741494854799136e-08, + "loss": 0.6979, + "step": 36255 + }, + { + "epoch": 1.33, + "learning_rate": 6.74112864833193e-08, + "loss": 0.6878, + "step": 36256 + }, + { + "epoch": 1.33, + "learning_rate": 6.740762441864724e-08, + "loss": 0.6877, + "step": 36257 + }, + { + "epoch": 1.33, + "learning_rate": 6.740396235397517e-08, + "loss": 0.6956, + "step": 36258 + }, + { + "epoch": 1.33, + "learning_rate": 6.740030028930311e-08, + "loss": 0.7037, + "step": 36259 + }, + { + "epoch": 1.33, + "learning_rate": 6.739663822463105e-08, + "loss": 0.6885, + "step": 36260 + }, + { + "epoch": 1.33, + "learning_rate": 6.739297615995897e-08, + "loss": 0.6998, + "step": 36261 + }, + { + "epoch": 1.33, + "learning_rate": 6.738931409528691e-08, + "loss": 0.6928, + "step": 36262 + }, + { + "epoch": 1.33, + "learning_rate": 6.738565203061485e-08, + "loss": 0.6952, + "step": 36263 + }, + { + "epoch": 1.33, + "learning_rate": 6.73819899659428e-08, + "loss": 0.6981, + "step": 36264 + }, + { + "epoch": 1.33, + "learning_rate": 6.737832790127073e-08, + "loss": 0.6969, + "step": 36265 + }, + { + "epoch": 1.33, + "learning_rate": 6.737466583659867e-08, + "loss": 0.6843, + "step": 36266 + }, + { + "epoch": 1.33, + "learning_rate": 6.73710037719266e-08, + "loss": 0.6979, + "step": 36267 + }, + { + "epoch": 1.33, + "learning_rate": 6.736734170725454e-08, + "loss": 0.6834, + "step": 36268 + }, + { + "epoch": 1.33, + "learning_rate": 6.736367964258248e-08, + "loss": 0.6858, + "step": 36269 + }, + { + "epoch": 1.33, + "learning_rate": 6.736001757791042e-08, + "loss": 0.7039, + "step": 36270 + }, + { + "epoch": 1.33, + "learning_rate": 6.735635551323836e-08, + "loss": 0.6925, + "step": 36271 + }, + { + "epoch": 1.33, + "learning_rate": 6.73526934485663e-08, + "loss": 0.6885, + "step": 36272 + }, + { + "epoch": 1.33, + "learning_rate": 6.734903138389424e-08, + "loss": 0.6967, + "step": 36273 + }, + { + "epoch": 1.33, + "learning_rate": 6.734536931922218e-08, + "loss": 0.6811, + "step": 36274 + }, + { + "epoch": 1.33, + "learning_rate": 6.734170725455012e-08, + "loss": 0.7009, + "step": 36275 + }, + { + "epoch": 1.33, + "learning_rate": 6.733804518987805e-08, + "loss": 0.6989, + "step": 36276 + }, + { + "epoch": 1.33, + "learning_rate": 6.733438312520599e-08, + "loss": 0.7109, + "step": 36277 + }, + { + "epoch": 1.33, + "learning_rate": 6.733072106053393e-08, + "loss": 0.6912, + "step": 36278 + }, + { + "epoch": 1.33, + "learning_rate": 6.732705899586187e-08, + "loss": 0.6902, + "step": 36279 + }, + { + "epoch": 1.33, + "learning_rate": 6.732339693118981e-08, + "loss": 0.6994, + "step": 36280 + }, + { + "epoch": 1.33, + "learning_rate": 6.731973486651775e-08, + "loss": 0.7127, + "step": 36281 + }, + { + "epoch": 1.33, + "learning_rate": 6.731607280184569e-08, + "loss": 0.6854, + "step": 36282 + }, + { + "epoch": 1.33, + "learning_rate": 6.731241073717361e-08, + "loss": 0.6841, + "step": 36283 + }, + { + "epoch": 1.33, + "learning_rate": 6.730874867250155e-08, + "loss": 0.69, + "step": 36284 + }, + { + "epoch": 1.33, + "learning_rate": 6.730508660782948e-08, + "loss": 0.7059, + "step": 36285 + }, + { + "epoch": 1.33, + "learning_rate": 6.730142454315742e-08, + "loss": 0.6877, + "step": 36286 + }, + { + "epoch": 1.33, + "learning_rate": 6.729776247848536e-08, + "loss": 0.6953, + "step": 36287 + }, + { + "epoch": 1.33, + "learning_rate": 6.72941004138133e-08, + "loss": 0.698, + "step": 36288 + }, + { + "epoch": 1.33, + "learning_rate": 6.729043834914124e-08, + "loss": 0.6884, + "step": 36289 + }, + { + "epoch": 1.33, + "learning_rate": 6.728677628446918e-08, + "loss": 0.686, + "step": 36290 + }, + { + "epoch": 1.33, + "learning_rate": 6.728311421979712e-08, + "loss": 0.6867, + "step": 36291 + }, + { + "epoch": 1.33, + "learning_rate": 6.727945215512506e-08, + "loss": 0.6782, + "step": 36292 + }, + { + "epoch": 1.33, + "learning_rate": 6.7275790090453e-08, + "loss": 0.6915, + "step": 36293 + }, + { + "epoch": 1.33, + "learning_rate": 6.727212802578093e-08, + "loss": 0.7005, + "step": 36294 + }, + { + "epoch": 1.33, + "learning_rate": 6.726846596110887e-08, + "loss": 0.6955, + "step": 36295 + }, + { + "epoch": 1.33, + "learning_rate": 6.72648038964368e-08, + "loss": 0.7066, + "step": 36296 + }, + { + "epoch": 1.33, + "learning_rate": 6.726114183176475e-08, + "loss": 0.7053, + "step": 36297 + }, + { + "epoch": 1.33, + "learning_rate": 6.725747976709269e-08, + "loss": 0.6919, + "step": 36298 + }, + { + "epoch": 1.33, + "learning_rate": 6.725381770242063e-08, + "loss": 0.6935, + "step": 36299 + }, + { + "epoch": 1.33, + "learning_rate": 6.725015563774857e-08, + "loss": 0.6947, + "step": 36300 + }, + { + "epoch": 1.33, + "learning_rate": 6.72464935730765e-08, + "loss": 0.6929, + "step": 36301 + }, + { + "epoch": 1.33, + "learning_rate": 6.724283150840445e-08, + "loss": 0.6899, + "step": 36302 + }, + { + "epoch": 1.33, + "learning_rate": 6.723916944373237e-08, + "loss": 0.6959, + "step": 36303 + }, + { + "epoch": 1.33, + "learning_rate": 6.723550737906031e-08, + "loss": 0.6973, + "step": 36304 + }, + { + "epoch": 1.33, + "learning_rate": 6.723184531438825e-08, + "loss": 0.7072, + "step": 36305 + }, + { + "epoch": 1.33, + "learning_rate": 6.722818324971619e-08, + "loss": 0.6937, + "step": 36306 + }, + { + "epoch": 1.33, + "learning_rate": 6.722452118504412e-08, + "loss": 0.7074, + "step": 36307 + }, + { + "epoch": 1.33, + "learning_rate": 6.722085912037206e-08, + "loss": 0.6986, + "step": 36308 + }, + { + "epoch": 1.33, + "learning_rate": 6.72171970557e-08, + "loss": 0.6866, + "step": 36309 + }, + { + "epoch": 1.33, + "learning_rate": 6.721353499102794e-08, + "loss": 0.6952, + "step": 36310 + }, + { + "epoch": 1.33, + "learning_rate": 6.720987292635588e-08, + "loss": 0.7011, + "step": 36311 + }, + { + "epoch": 1.33, + "learning_rate": 6.72062108616838e-08, + "loss": 0.6853, + "step": 36312 + }, + { + "epoch": 1.33, + "learning_rate": 6.720254879701175e-08, + "loss": 0.6966, + "step": 36313 + }, + { + "epoch": 1.33, + "learning_rate": 6.719888673233969e-08, + "loss": 0.7003, + "step": 36314 + }, + { + "epoch": 1.33, + "learning_rate": 6.719522466766763e-08, + "loss": 0.6982, + "step": 36315 + }, + { + "epoch": 1.33, + "learning_rate": 6.719156260299557e-08, + "loss": 0.694, + "step": 36316 + }, + { + "epoch": 1.33, + "learning_rate": 6.71879005383235e-08, + "loss": 0.7002, + "step": 36317 + }, + { + "epoch": 1.33, + "learning_rate": 6.718423847365144e-08, + "loss": 0.6941, + "step": 36318 + }, + { + "epoch": 1.33, + "learning_rate": 6.718057640897938e-08, + "loss": 0.6861, + "step": 36319 + }, + { + "epoch": 1.33, + "learning_rate": 6.717691434430732e-08, + "loss": 0.693, + "step": 36320 + }, + { + "epoch": 1.33, + "learning_rate": 6.717325227963525e-08, + "loss": 0.6942, + "step": 36321 + }, + { + "epoch": 1.33, + "learning_rate": 6.716959021496319e-08, + "loss": 0.7027, + "step": 36322 + }, + { + "epoch": 1.33, + "learning_rate": 6.716592815029113e-08, + "loss": 0.6987, + "step": 36323 + }, + { + "epoch": 1.33, + "learning_rate": 6.716226608561907e-08, + "loss": 0.6878, + "step": 36324 + }, + { + "epoch": 1.33, + "learning_rate": 6.715860402094701e-08, + "loss": 0.6923, + "step": 36325 + }, + { + "epoch": 1.33, + "learning_rate": 6.715494195627495e-08, + "loss": 0.704, + "step": 36326 + }, + { + "epoch": 1.33, + "learning_rate": 6.715127989160289e-08, + "loss": 0.6843, + "step": 36327 + }, + { + "epoch": 1.33, + "learning_rate": 6.714761782693083e-08, + "loss": 0.6931, + "step": 36328 + }, + { + "epoch": 1.33, + "learning_rate": 6.714395576225876e-08, + "loss": 0.7026, + "step": 36329 + }, + { + "epoch": 1.33, + "learning_rate": 6.71402936975867e-08, + "loss": 0.6902, + "step": 36330 + }, + { + "epoch": 1.33, + "learning_rate": 6.713663163291464e-08, + "loss": 0.7006, + "step": 36331 + }, + { + "epoch": 1.33, + "learning_rate": 6.713296956824256e-08, + "loss": 0.6978, + "step": 36332 + }, + { + "epoch": 1.33, + "learning_rate": 6.71293075035705e-08, + "loss": 0.6918, + "step": 36333 + }, + { + "epoch": 1.33, + "learning_rate": 6.712564543889844e-08, + "loss": 0.7033, + "step": 36334 + }, + { + "epoch": 1.33, + "learning_rate": 6.712198337422638e-08, + "loss": 0.6954, + "step": 36335 + }, + { + "epoch": 1.33, + "learning_rate": 6.711832130955432e-08, + "loss": 0.6911, + "step": 36336 + }, + { + "epoch": 1.33, + "learning_rate": 6.711465924488226e-08, + "loss": 0.6987, + "step": 36337 + }, + { + "epoch": 1.33, + "learning_rate": 6.71109971802102e-08, + "loss": 0.6868, + "step": 36338 + }, + { + "epoch": 1.33, + "learning_rate": 6.710733511553813e-08, + "loss": 0.6985, + "step": 36339 + }, + { + "epoch": 1.33, + "learning_rate": 6.710367305086607e-08, + "loss": 0.6963, + "step": 36340 + }, + { + "epoch": 1.33, + "learning_rate": 6.710001098619401e-08, + "loss": 0.693, + "step": 36341 + }, + { + "epoch": 1.33, + "learning_rate": 6.709634892152195e-08, + "loss": 0.6941, + "step": 36342 + }, + { + "epoch": 1.33, + "learning_rate": 6.709268685684989e-08, + "loss": 0.6976, + "step": 36343 + }, + { + "epoch": 1.33, + "learning_rate": 6.708902479217783e-08, + "loss": 0.6825, + "step": 36344 + }, + { + "epoch": 1.33, + "learning_rate": 6.708536272750577e-08, + "loss": 0.6805, + "step": 36345 + }, + { + "epoch": 1.33, + "learning_rate": 6.708170066283371e-08, + "loss": 0.6858, + "step": 36346 + }, + { + "epoch": 1.33, + "learning_rate": 6.707803859816165e-08, + "loss": 0.6846, + "step": 36347 + }, + { + "epoch": 1.33, + "learning_rate": 6.707437653348958e-08, + "loss": 0.6948, + "step": 36348 + }, + { + "epoch": 1.33, + "learning_rate": 6.707071446881752e-08, + "loss": 0.7022, + "step": 36349 + }, + { + "epoch": 1.33, + "learning_rate": 6.706705240414546e-08, + "loss": 0.7032, + "step": 36350 + }, + { + "epoch": 1.33, + "learning_rate": 6.70633903394734e-08, + "loss": 0.6943, + "step": 36351 + }, + { + "epoch": 1.33, + "learning_rate": 6.705972827480134e-08, + "loss": 0.6967, + "step": 36352 + }, + { + "epoch": 1.33, + "learning_rate": 6.705606621012926e-08, + "loss": 0.7007, + "step": 36353 + }, + { + "epoch": 1.33, + "learning_rate": 6.70524041454572e-08, + "loss": 0.691, + "step": 36354 + }, + { + "epoch": 1.33, + "learning_rate": 6.704874208078514e-08, + "loss": 0.7031, + "step": 36355 + }, + { + "epoch": 1.33, + "learning_rate": 6.704508001611308e-08, + "loss": 0.6942, + "step": 36356 + }, + { + "epoch": 1.33, + "learning_rate": 6.704141795144101e-08, + "loss": 0.6796, + "step": 36357 + }, + { + "epoch": 1.33, + "learning_rate": 6.703775588676895e-08, + "loss": 0.6949, + "step": 36358 + }, + { + "epoch": 1.33, + "learning_rate": 6.703409382209689e-08, + "loss": 0.6938, + "step": 36359 + }, + { + "epoch": 1.33, + "learning_rate": 6.703043175742483e-08, + "loss": 0.7008, + "step": 36360 + }, + { + "epoch": 1.33, + "learning_rate": 6.702676969275277e-08, + "loss": 0.6783, + "step": 36361 + }, + { + "epoch": 1.33, + "learning_rate": 6.702310762808071e-08, + "loss": 0.6972, + "step": 36362 + }, + { + "epoch": 1.33, + "learning_rate": 6.701944556340865e-08, + "loss": 0.6957, + "step": 36363 + }, + { + "epoch": 1.33, + "learning_rate": 6.701578349873659e-08, + "loss": 0.6825, + "step": 36364 + }, + { + "epoch": 1.33, + "learning_rate": 6.701212143406453e-08, + "loss": 0.6941, + "step": 36365 + }, + { + "epoch": 1.33, + "learning_rate": 6.700845936939246e-08, + "loss": 0.6956, + "step": 36366 + }, + { + "epoch": 1.33, + "learning_rate": 6.70047973047204e-08, + "loss": 0.7043, + "step": 36367 + }, + { + "epoch": 1.33, + "learning_rate": 6.700113524004834e-08, + "loss": 0.6955, + "step": 36368 + }, + { + "epoch": 1.33, + "learning_rate": 6.699747317537628e-08, + "loss": 0.6928, + "step": 36369 + }, + { + "epoch": 1.33, + "learning_rate": 6.699381111070422e-08, + "loss": 0.6858, + "step": 36370 + }, + { + "epoch": 1.33, + "learning_rate": 6.699014904603216e-08, + "loss": 0.6872, + "step": 36371 + }, + { + "epoch": 1.33, + "learning_rate": 6.69864869813601e-08, + "loss": 0.7096, + "step": 36372 + }, + { + "epoch": 1.33, + "learning_rate": 6.698282491668804e-08, + "loss": 0.6935, + "step": 36373 + }, + { + "epoch": 1.33, + "learning_rate": 6.697916285201598e-08, + "loss": 0.6913, + "step": 36374 + }, + { + "epoch": 1.33, + "learning_rate": 6.69755007873439e-08, + "loss": 0.6971, + "step": 36375 + }, + { + "epoch": 1.33, + "learning_rate": 6.697183872267184e-08, + "loss": 0.6863, + "step": 36376 + }, + { + "epoch": 1.33, + "learning_rate": 6.696817665799978e-08, + "loss": 0.6927, + "step": 36377 + }, + { + "epoch": 1.33, + "learning_rate": 6.696451459332771e-08, + "loss": 0.6918, + "step": 36378 + }, + { + "epoch": 1.33, + "learning_rate": 6.696085252865565e-08, + "loss": 0.6932, + "step": 36379 + }, + { + "epoch": 1.33, + "learning_rate": 6.695719046398359e-08, + "loss": 0.6942, + "step": 36380 + }, + { + "epoch": 1.33, + "learning_rate": 6.695352839931153e-08, + "loss": 0.6982, + "step": 36381 + }, + { + "epoch": 1.33, + "learning_rate": 6.694986633463947e-08, + "loss": 0.7021, + "step": 36382 + }, + { + "epoch": 1.33, + "learning_rate": 6.694620426996741e-08, + "loss": 0.6872, + "step": 36383 + }, + { + "epoch": 1.33, + "learning_rate": 6.694254220529533e-08, + "loss": 0.6955, + "step": 36384 + }, + { + "epoch": 1.33, + "learning_rate": 6.693888014062327e-08, + "loss": 0.6892, + "step": 36385 + }, + { + "epoch": 1.33, + "learning_rate": 6.693521807595121e-08, + "loss": 0.7067, + "step": 36386 + }, + { + "epoch": 1.33, + "learning_rate": 6.693155601127915e-08, + "loss": 0.6881, + "step": 36387 + }, + { + "epoch": 1.33, + "learning_rate": 6.69278939466071e-08, + "loss": 0.6911, + "step": 36388 + }, + { + "epoch": 1.33, + "learning_rate": 6.692423188193503e-08, + "loss": 0.6944, + "step": 36389 + }, + { + "epoch": 1.33, + "learning_rate": 6.692056981726297e-08, + "loss": 0.6945, + "step": 36390 + }, + { + "epoch": 1.33, + "learning_rate": 6.691690775259091e-08, + "loss": 0.6913, + "step": 36391 + }, + { + "epoch": 1.33, + "learning_rate": 6.691324568791885e-08, + "loss": 0.6911, + "step": 36392 + }, + { + "epoch": 1.33, + "learning_rate": 6.690958362324678e-08, + "loss": 0.6993, + "step": 36393 + }, + { + "epoch": 1.33, + "learning_rate": 6.690592155857472e-08, + "loss": 0.6859, + "step": 36394 + }, + { + "epoch": 1.33, + "learning_rate": 6.690225949390266e-08, + "loss": 0.7006, + "step": 36395 + }, + { + "epoch": 1.33, + "learning_rate": 6.68985974292306e-08, + "loss": 0.6961, + "step": 36396 + }, + { + "epoch": 1.33, + "learning_rate": 6.689493536455854e-08, + "loss": 0.6991, + "step": 36397 + }, + { + "epoch": 1.33, + "learning_rate": 6.689127329988648e-08, + "loss": 0.6868, + "step": 36398 + }, + { + "epoch": 1.33, + "learning_rate": 6.688761123521442e-08, + "loss": 0.6976, + "step": 36399 + }, + { + "epoch": 1.33, + "learning_rate": 6.688394917054235e-08, + "loss": 0.681, + "step": 36400 + }, + { + "epoch": 1.33, + "learning_rate": 6.688028710587029e-08, + "loss": 0.6934, + "step": 36401 + }, + { + "epoch": 1.33, + "learning_rate": 6.687662504119821e-08, + "loss": 0.6919, + "step": 36402 + }, + { + "epoch": 1.33, + "learning_rate": 6.687296297652615e-08, + "loss": 0.6913, + "step": 36403 + }, + { + "epoch": 1.33, + "learning_rate": 6.68693009118541e-08, + "loss": 0.6966, + "step": 36404 + }, + { + "epoch": 1.33, + "learning_rate": 6.686563884718203e-08, + "loss": 0.6823, + "step": 36405 + }, + { + "epoch": 1.33, + "learning_rate": 6.686197678250997e-08, + "loss": 0.6776, + "step": 36406 + }, + { + "epoch": 1.33, + "learning_rate": 6.685831471783791e-08, + "loss": 0.6853, + "step": 36407 + }, + { + "epoch": 1.33, + "learning_rate": 6.685465265316585e-08, + "loss": 0.6829, + "step": 36408 + }, + { + "epoch": 1.33, + "learning_rate": 6.685099058849379e-08, + "loss": 0.6938, + "step": 36409 + }, + { + "epoch": 1.33, + "learning_rate": 6.684732852382172e-08, + "loss": 0.7059, + "step": 36410 + }, + { + "epoch": 1.33, + "learning_rate": 6.684366645914966e-08, + "loss": 0.6816, + "step": 36411 + }, + { + "epoch": 1.33, + "learning_rate": 6.68400043944776e-08, + "loss": 0.6913, + "step": 36412 + }, + { + "epoch": 1.33, + "learning_rate": 6.683634232980554e-08, + "loss": 0.6946, + "step": 36413 + }, + { + "epoch": 1.33, + "learning_rate": 6.683268026513348e-08, + "loss": 0.6985, + "step": 36414 + }, + { + "epoch": 1.33, + "learning_rate": 6.682901820046142e-08, + "loss": 0.6947, + "step": 36415 + }, + { + "epoch": 1.33, + "learning_rate": 6.682535613578936e-08, + "loss": 0.6928, + "step": 36416 + }, + { + "epoch": 1.33, + "learning_rate": 6.68216940711173e-08, + "loss": 0.7113, + "step": 36417 + }, + { + "epoch": 1.33, + "learning_rate": 6.681803200644524e-08, + "loss": 0.6965, + "step": 36418 + }, + { + "epoch": 1.33, + "learning_rate": 6.681436994177317e-08, + "loss": 0.6845, + "step": 36419 + }, + { + "epoch": 1.33, + "learning_rate": 6.68107078771011e-08, + "loss": 0.6861, + "step": 36420 + }, + { + "epoch": 1.33, + "learning_rate": 6.680704581242905e-08, + "loss": 0.6904, + "step": 36421 + }, + { + "epoch": 1.33, + "learning_rate": 6.680338374775699e-08, + "loss": 0.6966, + "step": 36422 + }, + { + "epoch": 1.33, + "learning_rate": 6.679972168308493e-08, + "loss": 0.6884, + "step": 36423 + }, + { + "epoch": 1.33, + "learning_rate": 6.679605961841285e-08, + "loss": 0.7021, + "step": 36424 + }, + { + "epoch": 1.33, + "learning_rate": 6.679239755374079e-08, + "loss": 0.6862, + "step": 36425 + }, + { + "epoch": 1.33, + "learning_rate": 6.678873548906873e-08, + "loss": 0.6897, + "step": 36426 + }, + { + "epoch": 1.33, + "learning_rate": 6.678507342439667e-08, + "loss": 0.693, + "step": 36427 + }, + { + "epoch": 1.33, + "learning_rate": 6.67814113597246e-08, + "loss": 0.6894, + "step": 36428 + }, + { + "epoch": 1.33, + "learning_rate": 6.677774929505254e-08, + "loss": 0.7097, + "step": 36429 + }, + { + "epoch": 1.33, + "learning_rate": 6.677408723038048e-08, + "loss": 0.6971, + "step": 36430 + }, + { + "epoch": 1.33, + "learning_rate": 6.677042516570842e-08, + "loss": 0.6966, + "step": 36431 + }, + { + "epoch": 1.33, + "learning_rate": 6.676676310103636e-08, + "loss": 0.716, + "step": 36432 + }, + { + "epoch": 1.33, + "learning_rate": 6.67631010363643e-08, + "loss": 0.6953, + "step": 36433 + }, + { + "epoch": 1.33, + "learning_rate": 6.675943897169224e-08, + "loss": 0.6989, + "step": 36434 + }, + { + "epoch": 1.33, + "learning_rate": 6.675577690702018e-08, + "loss": 0.6928, + "step": 36435 + }, + { + "epoch": 1.33, + "learning_rate": 6.675211484234812e-08, + "loss": 0.6859, + "step": 36436 + }, + { + "epoch": 1.33, + "learning_rate": 6.674845277767604e-08, + "loss": 0.6913, + "step": 36437 + }, + { + "epoch": 1.33, + "learning_rate": 6.674479071300398e-08, + "loss": 0.6914, + "step": 36438 + }, + { + "epoch": 1.33, + "learning_rate": 6.674112864833192e-08, + "loss": 0.6928, + "step": 36439 + }, + { + "epoch": 1.33, + "learning_rate": 6.673746658365986e-08, + "loss": 0.6977, + "step": 36440 + }, + { + "epoch": 1.33, + "learning_rate": 6.67338045189878e-08, + "loss": 0.7072, + "step": 36441 + }, + { + "epoch": 1.33, + "learning_rate": 6.673014245431574e-08, + "loss": 0.6877, + "step": 36442 + }, + { + "epoch": 1.33, + "learning_rate": 6.672648038964368e-08, + "loss": 0.6843, + "step": 36443 + }, + { + "epoch": 1.33, + "learning_rate": 6.672281832497162e-08, + "loss": 0.6885, + "step": 36444 + }, + { + "epoch": 1.33, + "learning_rate": 6.671915626029956e-08, + "loss": 0.6911, + "step": 36445 + }, + { + "epoch": 1.33, + "learning_rate": 6.671549419562749e-08, + "loss": 0.6858, + "step": 36446 + }, + { + "epoch": 1.33, + "learning_rate": 6.671183213095543e-08, + "loss": 0.7022, + "step": 36447 + }, + { + "epoch": 1.33, + "learning_rate": 6.670817006628337e-08, + "loss": 0.6855, + "step": 36448 + }, + { + "epoch": 1.33, + "learning_rate": 6.67045080016113e-08, + "loss": 0.6987, + "step": 36449 + }, + { + "epoch": 1.33, + "learning_rate": 6.670084593693924e-08, + "loss": 0.697, + "step": 36450 + }, + { + "epoch": 1.33, + "learning_rate": 6.669718387226718e-08, + "loss": 0.6844, + "step": 36451 + }, + { + "epoch": 1.33, + "learning_rate": 6.669352180759512e-08, + "loss": 0.6937, + "step": 36452 + }, + { + "epoch": 1.33, + "learning_rate": 6.668985974292306e-08, + "loss": 0.6976, + "step": 36453 + }, + { + "epoch": 1.33, + "learning_rate": 6.6686197678251e-08, + "loss": 0.6975, + "step": 36454 + }, + { + "epoch": 1.33, + "learning_rate": 6.668253561357892e-08, + "loss": 0.6936, + "step": 36455 + }, + { + "epoch": 1.33, + "learning_rate": 6.667887354890686e-08, + "loss": 0.6987, + "step": 36456 + }, + { + "epoch": 1.33, + "learning_rate": 6.66752114842348e-08, + "loss": 0.6882, + "step": 36457 + }, + { + "epoch": 1.33, + "learning_rate": 6.667154941956274e-08, + "loss": 0.6895, + "step": 36458 + }, + { + "epoch": 1.33, + "learning_rate": 6.666788735489068e-08, + "loss": 0.6901, + "step": 36459 + }, + { + "epoch": 1.33, + "learning_rate": 6.666422529021862e-08, + "loss": 0.6986, + "step": 36460 + }, + { + "epoch": 1.33, + "learning_rate": 6.666056322554656e-08, + "loss": 0.7028, + "step": 36461 + }, + { + "epoch": 1.33, + "learning_rate": 6.66569011608745e-08, + "loss": 0.708, + "step": 36462 + }, + { + "epoch": 1.33, + "learning_rate": 6.665323909620244e-08, + "loss": 0.6888, + "step": 36463 + }, + { + "epoch": 1.33, + "learning_rate": 6.664957703153037e-08, + "loss": 0.7025, + "step": 36464 + }, + { + "epoch": 1.33, + "learning_rate": 6.664591496685831e-08, + "loss": 0.6794, + "step": 36465 + }, + { + "epoch": 1.33, + "learning_rate": 6.664225290218625e-08, + "loss": 0.6934, + "step": 36466 + }, + { + "epoch": 1.33, + "learning_rate": 6.663859083751419e-08, + "loss": 0.684, + "step": 36467 + }, + { + "epoch": 1.33, + "learning_rate": 6.663492877284213e-08, + "loss": 0.6921, + "step": 36468 + }, + { + "epoch": 1.33, + "learning_rate": 6.663126670817007e-08, + "loss": 0.6937, + "step": 36469 + }, + { + "epoch": 1.33, + "learning_rate": 6.662760464349801e-08, + "loss": 0.6881, + "step": 36470 + }, + { + "epoch": 1.33, + "learning_rate": 6.662394257882594e-08, + "loss": 0.6958, + "step": 36471 + }, + { + "epoch": 1.33, + "learning_rate": 6.662028051415388e-08, + "loss": 0.6872, + "step": 36472 + }, + { + "epoch": 1.33, + "learning_rate": 6.66166184494818e-08, + "loss": 0.7093, + "step": 36473 + }, + { + "epoch": 1.33, + "learning_rate": 6.661295638480974e-08, + "loss": 0.6896, + "step": 36474 + }, + { + "epoch": 1.33, + "learning_rate": 6.660929432013768e-08, + "loss": 0.6771, + "step": 36475 + }, + { + "epoch": 1.33, + "learning_rate": 6.660563225546562e-08, + "loss": 0.6822, + "step": 36476 + }, + { + "epoch": 1.33, + "learning_rate": 6.660197019079356e-08, + "loss": 0.7004, + "step": 36477 + }, + { + "epoch": 1.33, + "learning_rate": 6.65983081261215e-08, + "loss": 0.6803, + "step": 36478 + }, + { + "epoch": 1.33, + "learning_rate": 6.659464606144944e-08, + "loss": 0.6961, + "step": 36479 + }, + { + "epoch": 1.33, + "learning_rate": 6.659098399677738e-08, + "loss": 0.6887, + "step": 36480 + }, + { + "epoch": 1.33, + "learning_rate": 6.658732193210532e-08, + "loss": 0.6892, + "step": 36481 + }, + { + "epoch": 1.33, + "learning_rate": 6.658365986743325e-08, + "loss": 0.6864, + "step": 36482 + }, + { + "epoch": 1.33, + "learning_rate": 6.657999780276119e-08, + "loss": 0.6931, + "step": 36483 + }, + { + "epoch": 1.33, + "learning_rate": 6.657633573808913e-08, + "loss": 0.6995, + "step": 36484 + }, + { + "epoch": 1.33, + "learning_rate": 6.657267367341707e-08, + "loss": 0.695, + "step": 36485 + }, + { + "epoch": 1.33, + "learning_rate": 6.656901160874501e-08, + "loss": 0.7018, + "step": 36486 + }, + { + "epoch": 1.33, + "learning_rate": 6.656534954407295e-08, + "loss": 0.7096, + "step": 36487 + }, + { + "epoch": 1.33, + "learning_rate": 6.656168747940089e-08, + "loss": 0.6947, + "step": 36488 + }, + { + "epoch": 1.34, + "learning_rate": 6.655802541472883e-08, + "loss": 0.6952, + "step": 36489 + }, + { + "epoch": 1.34, + "learning_rate": 6.655436335005677e-08, + "loss": 0.7153, + "step": 36490 + }, + { + "epoch": 1.34, + "learning_rate": 6.65507012853847e-08, + "loss": 0.6945, + "step": 36491 + }, + { + "epoch": 1.34, + "learning_rate": 6.654703922071264e-08, + "loss": 0.7058, + "step": 36492 + }, + { + "epoch": 1.34, + "learning_rate": 6.654337715604058e-08, + "loss": 0.6909, + "step": 36493 + }, + { + "epoch": 1.34, + "learning_rate": 6.653971509136852e-08, + "loss": 0.7054, + "step": 36494 + }, + { + "epoch": 1.34, + "learning_rate": 6.653605302669644e-08, + "loss": 0.7043, + "step": 36495 + }, + { + "epoch": 1.34, + "learning_rate": 6.653239096202438e-08, + "loss": 0.6928, + "step": 36496 + }, + { + "epoch": 1.34, + "learning_rate": 6.652872889735232e-08, + "loss": 0.6904, + "step": 36497 + }, + { + "epoch": 1.34, + "learning_rate": 6.652506683268026e-08, + "loss": 0.6962, + "step": 36498 + }, + { + "epoch": 1.34, + "learning_rate": 6.65214047680082e-08, + "loss": 0.6945, + "step": 36499 + }, + { + "epoch": 1.34, + "learning_rate": 6.651774270333613e-08, + "loss": 0.701, + "step": 36500 + }, + { + "epoch": 1.34, + "learning_rate": 6.651408063866407e-08, + "loss": 0.7123, + "step": 36501 + }, + { + "epoch": 1.34, + "learning_rate": 6.651041857399201e-08, + "loss": 0.6896, + "step": 36502 + }, + { + "epoch": 1.34, + "learning_rate": 6.650675650931995e-08, + "loss": 0.6909, + "step": 36503 + }, + { + "epoch": 1.34, + "learning_rate": 6.650309444464789e-08, + "loss": 0.696, + "step": 36504 + }, + { + "epoch": 1.34, + "learning_rate": 6.649943237997583e-08, + "loss": 0.697, + "step": 36505 + }, + { + "epoch": 1.34, + "learning_rate": 6.649577031530377e-08, + "loss": 0.6852, + "step": 36506 + }, + { + "epoch": 1.34, + "learning_rate": 6.649210825063171e-08, + "loss": 0.7095, + "step": 36507 + }, + { + "epoch": 1.34, + "learning_rate": 6.648844618595965e-08, + "loss": 0.681, + "step": 36508 + }, + { + "epoch": 1.34, + "learning_rate": 6.648478412128757e-08, + "loss": 0.6878, + "step": 36509 + }, + { + "epoch": 1.34, + "learning_rate": 6.648112205661551e-08, + "loss": 0.7047, + "step": 36510 + }, + { + "epoch": 1.34, + "learning_rate": 6.647745999194345e-08, + "loss": 0.7013, + "step": 36511 + }, + { + "epoch": 1.34, + "learning_rate": 6.64737979272714e-08, + "loss": 0.6856, + "step": 36512 + }, + { + "epoch": 1.34, + "learning_rate": 6.647013586259933e-08, + "loss": 0.6891, + "step": 36513 + }, + { + "epoch": 1.34, + "learning_rate": 6.646647379792727e-08, + "loss": 0.7097, + "step": 36514 + }, + { + "epoch": 1.34, + "learning_rate": 6.646281173325521e-08, + "loss": 0.6851, + "step": 36515 + }, + { + "epoch": 1.34, + "learning_rate": 6.645914966858315e-08, + "loss": 0.691, + "step": 36516 + }, + { + "epoch": 1.34, + "learning_rate": 6.645548760391108e-08, + "loss": 0.6874, + "step": 36517 + }, + { + "epoch": 1.34, + "learning_rate": 6.645182553923902e-08, + "loss": 0.6831, + "step": 36518 + }, + { + "epoch": 1.34, + "learning_rate": 6.644816347456696e-08, + "loss": 0.6803, + "step": 36519 + }, + { + "epoch": 1.34, + "learning_rate": 6.644450140989489e-08, + "loss": 0.6941, + "step": 36520 + }, + { + "epoch": 1.34, + "learning_rate": 6.644083934522283e-08, + "loss": 0.68, + "step": 36521 + }, + { + "epoch": 1.34, + "learning_rate": 6.643717728055077e-08, + "loss": 0.6969, + "step": 36522 + }, + { + "epoch": 1.34, + "learning_rate": 6.643351521587871e-08, + "loss": 0.6849, + "step": 36523 + }, + { + "epoch": 1.34, + "learning_rate": 6.642985315120665e-08, + "loss": 0.6848, + "step": 36524 + }, + { + "epoch": 1.34, + "learning_rate": 6.642619108653459e-08, + "loss": 0.6911, + "step": 36525 + }, + { + "epoch": 1.34, + "learning_rate": 6.642252902186253e-08, + "loss": 0.6999, + "step": 36526 + }, + { + "epoch": 1.34, + "learning_rate": 6.641886695719045e-08, + "loss": 0.6897, + "step": 36527 + }, + { + "epoch": 1.34, + "learning_rate": 6.641520489251839e-08, + "loss": 0.6732, + "step": 36528 + }, + { + "epoch": 1.34, + "learning_rate": 6.641154282784633e-08, + "loss": 0.7105, + "step": 36529 + }, + { + "epoch": 1.34, + "learning_rate": 6.640788076317427e-08, + "loss": 0.7035, + "step": 36530 + }, + { + "epoch": 1.34, + "learning_rate": 6.640421869850221e-08, + "loss": 0.6948, + "step": 36531 + }, + { + "epoch": 1.34, + "learning_rate": 6.640055663383015e-08, + "loss": 0.7129, + "step": 36532 + }, + { + "epoch": 1.34, + "learning_rate": 6.639689456915809e-08, + "loss": 0.6657, + "step": 36533 + }, + { + "epoch": 1.34, + "learning_rate": 6.639323250448603e-08, + "loss": 0.6953, + "step": 36534 + }, + { + "epoch": 1.34, + "learning_rate": 6.638957043981397e-08, + "loss": 0.6914, + "step": 36535 + }, + { + "epoch": 1.34, + "learning_rate": 6.63859083751419e-08, + "loss": 0.6983, + "step": 36536 + }, + { + "epoch": 1.34, + "learning_rate": 6.638224631046984e-08, + "loss": 0.7019, + "step": 36537 + }, + { + "epoch": 1.34, + "learning_rate": 6.637858424579778e-08, + "loss": 0.6879, + "step": 36538 + }, + { + "epoch": 1.34, + "learning_rate": 6.637492218112572e-08, + "loss": 0.6924, + "step": 36539 + }, + { + "epoch": 1.34, + "learning_rate": 6.637126011645366e-08, + "loss": 0.6932, + "step": 36540 + }, + { + "epoch": 1.34, + "learning_rate": 6.63675980517816e-08, + "loss": 0.6828, + "step": 36541 + }, + { + "epoch": 1.34, + "learning_rate": 6.636393598710953e-08, + "loss": 0.6854, + "step": 36542 + }, + { + "epoch": 1.34, + "learning_rate": 6.636027392243747e-08, + "loss": 0.699, + "step": 36543 + }, + { + "epoch": 1.34, + "learning_rate": 6.63566118577654e-08, + "loss": 0.6943, + "step": 36544 + }, + { + "epoch": 1.34, + "learning_rate": 6.635294979309333e-08, + "loss": 0.6963, + "step": 36545 + }, + { + "epoch": 1.34, + "learning_rate": 6.634928772842127e-08, + "loss": 0.6955, + "step": 36546 + }, + { + "epoch": 1.34, + "learning_rate": 6.634562566374921e-08, + "loss": 0.6899, + "step": 36547 + }, + { + "epoch": 1.34, + "learning_rate": 6.634196359907715e-08, + "loss": 0.6829, + "step": 36548 + }, + { + "epoch": 1.34, + "learning_rate": 6.633830153440509e-08, + "loss": 0.6934, + "step": 36549 + }, + { + "epoch": 1.34, + "learning_rate": 6.633463946973303e-08, + "loss": 0.6941, + "step": 36550 + }, + { + "epoch": 1.34, + "learning_rate": 6.633097740506097e-08, + "loss": 0.6953, + "step": 36551 + }, + { + "epoch": 1.34, + "learning_rate": 6.632731534038891e-08, + "loss": 0.6894, + "step": 36552 + }, + { + "epoch": 1.34, + "learning_rate": 6.632365327571685e-08, + "loss": 0.6909, + "step": 36553 + }, + { + "epoch": 1.34, + "learning_rate": 6.631999121104478e-08, + "loss": 0.7003, + "step": 36554 + }, + { + "epoch": 1.34, + "learning_rate": 6.631632914637272e-08, + "loss": 0.6951, + "step": 36555 + }, + { + "epoch": 1.34, + "learning_rate": 6.631266708170066e-08, + "loss": 0.7077, + "step": 36556 + }, + { + "epoch": 1.34, + "learning_rate": 6.63090050170286e-08, + "loss": 0.6978, + "step": 36557 + }, + { + "epoch": 1.34, + "learning_rate": 6.630534295235654e-08, + "loss": 0.676, + "step": 36558 + }, + { + "epoch": 1.34, + "learning_rate": 6.630168088768448e-08, + "loss": 0.6894, + "step": 36559 + }, + { + "epoch": 1.34, + "learning_rate": 6.629801882301242e-08, + "loss": 0.7037, + "step": 36560 + }, + { + "epoch": 1.34, + "learning_rate": 6.629435675834036e-08, + "loss": 0.6857, + "step": 36561 + }, + { + "epoch": 1.34, + "learning_rate": 6.62906946936683e-08, + "loss": 0.7011, + "step": 36562 + }, + { + "epoch": 1.34, + "learning_rate": 6.628703262899622e-08, + "loss": 0.6901, + "step": 36563 + }, + { + "epoch": 1.34, + "learning_rate": 6.628337056432416e-08, + "loss": 0.6956, + "step": 36564 + }, + { + "epoch": 1.34, + "learning_rate": 6.62797084996521e-08, + "loss": 0.6935, + "step": 36565 + }, + { + "epoch": 1.34, + "learning_rate": 6.627604643498003e-08, + "loss": 0.7061, + "step": 36566 + }, + { + "epoch": 1.34, + "learning_rate": 6.627238437030797e-08, + "loss": 0.6841, + "step": 36567 + }, + { + "epoch": 1.34, + "learning_rate": 6.626872230563591e-08, + "loss": 0.7086, + "step": 36568 + }, + { + "epoch": 1.34, + "learning_rate": 6.626506024096385e-08, + "loss": 0.6891, + "step": 36569 + }, + { + "epoch": 1.34, + "learning_rate": 6.626139817629179e-08, + "loss": 0.6979, + "step": 36570 + }, + { + "epoch": 1.34, + "learning_rate": 6.625773611161973e-08, + "loss": 0.6968, + "step": 36571 + }, + { + "epoch": 1.34, + "learning_rate": 6.625407404694766e-08, + "loss": 0.6907, + "step": 36572 + }, + { + "epoch": 1.34, + "learning_rate": 6.62504119822756e-08, + "loss": 0.6824, + "step": 36573 + }, + { + "epoch": 1.34, + "learning_rate": 6.624674991760354e-08, + "loss": 0.6955, + "step": 36574 + }, + { + "epoch": 1.34, + "learning_rate": 6.624308785293148e-08, + "loss": 0.6918, + "step": 36575 + }, + { + "epoch": 1.34, + "learning_rate": 6.623942578825942e-08, + "loss": 0.7006, + "step": 36576 + }, + { + "epoch": 1.34, + "learning_rate": 6.623576372358736e-08, + "loss": 0.6914, + "step": 36577 + }, + { + "epoch": 1.34, + "learning_rate": 6.62321016589153e-08, + "loss": 0.6837, + "step": 36578 + }, + { + "epoch": 1.34, + "learning_rate": 6.622843959424324e-08, + "loss": 0.7119, + "step": 36579 + }, + { + "epoch": 1.34, + "learning_rate": 6.622477752957118e-08, + "loss": 0.6957, + "step": 36580 + }, + { + "epoch": 1.34, + "learning_rate": 6.62211154648991e-08, + "loss": 0.6944, + "step": 36581 + }, + { + "epoch": 1.34, + "learning_rate": 6.621745340022704e-08, + "loss": 0.695, + "step": 36582 + }, + { + "epoch": 1.34, + "learning_rate": 6.621379133555498e-08, + "loss": 0.7021, + "step": 36583 + }, + { + "epoch": 1.34, + "learning_rate": 6.621012927088292e-08, + "loss": 0.6799, + "step": 36584 + }, + { + "epoch": 1.34, + "learning_rate": 6.620646720621086e-08, + "loss": 0.6902, + "step": 36585 + }, + { + "epoch": 1.34, + "learning_rate": 6.62028051415388e-08, + "loss": 0.6977, + "step": 36586 + }, + { + "epoch": 1.34, + "learning_rate": 6.619914307686674e-08, + "loss": 0.6909, + "step": 36587 + }, + { + "epoch": 1.34, + "learning_rate": 6.619548101219467e-08, + "loss": 0.6951, + "step": 36588 + }, + { + "epoch": 1.34, + "learning_rate": 6.619181894752261e-08, + "loss": 0.6842, + "step": 36589 + }, + { + "epoch": 1.34, + "learning_rate": 6.618815688285055e-08, + "loss": 0.7063, + "step": 36590 + }, + { + "epoch": 1.34, + "learning_rate": 6.618449481817848e-08, + "loss": 0.6964, + "step": 36591 + }, + { + "epoch": 1.34, + "learning_rate": 6.618083275350642e-08, + "loss": 0.7043, + "step": 36592 + }, + { + "epoch": 1.34, + "learning_rate": 6.617717068883436e-08, + "loss": 0.703, + "step": 36593 + }, + { + "epoch": 1.34, + "learning_rate": 6.61735086241623e-08, + "loss": 0.6795, + "step": 36594 + }, + { + "epoch": 1.34, + "learning_rate": 6.616984655949024e-08, + "loss": 0.6857, + "step": 36595 + }, + { + "epoch": 1.34, + "learning_rate": 6.616618449481818e-08, + "loss": 0.6884, + "step": 36596 + }, + { + "epoch": 1.34, + "learning_rate": 6.616252243014612e-08, + "loss": 0.6998, + "step": 36597 + }, + { + "epoch": 1.34, + "learning_rate": 6.615886036547406e-08, + "loss": 0.698, + "step": 36598 + }, + { + "epoch": 1.34, + "learning_rate": 6.615519830080198e-08, + "loss": 0.6937, + "step": 36599 + }, + { + "epoch": 1.34, + "learning_rate": 6.615153623612992e-08, + "loss": 0.7089, + "step": 36600 + }, + { + "epoch": 1.34, + "learning_rate": 6.614787417145786e-08, + "loss": 0.702, + "step": 36601 + }, + { + "epoch": 1.34, + "learning_rate": 6.61442121067858e-08, + "loss": 0.6945, + "step": 36602 + }, + { + "epoch": 1.34, + "learning_rate": 6.614055004211374e-08, + "loss": 0.6901, + "step": 36603 + }, + { + "epoch": 1.34, + "learning_rate": 6.613688797744168e-08, + "loss": 0.6907, + "step": 36604 + }, + { + "epoch": 1.34, + "learning_rate": 6.613322591276962e-08, + "loss": 0.6938, + "step": 36605 + }, + { + "epoch": 1.34, + "learning_rate": 6.612956384809756e-08, + "loss": 0.6894, + "step": 36606 + }, + { + "epoch": 1.34, + "learning_rate": 6.61259017834255e-08, + "loss": 0.6878, + "step": 36607 + }, + { + "epoch": 1.34, + "learning_rate": 6.612223971875343e-08, + "loss": 0.6987, + "step": 36608 + }, + { + "epoch": 1.34, + "learning_rate": 6.611857765408137e-08, + "loss": 0.7062, + "step": 36609 + }, + { + "epoch": 1.34, + "learning_rate": 6.611491558940931e-08, + "loss": 0.6857, + "step": 36610 + }, + { + "epoch": 1.34, + "learning_rate": 6.611125352473725e-08, + "loss": 0.6864, + "step": 36611 + }, + { + "epoch": 1.34, + "learning_rate": 6.610759146006519e-08, + "loss": 0.6989, + "step": 36612 + }, + { + "epoch": 1.34, + "learning_rate": 6.610392939539312e-08, + "loss": 0.7048, + "step": 36613 + }, + { + "epoch": 1.34, + "learning_rate": 6.610026733072105e-08, + "loss": 0.6957, + "step": 36614 + }, + { + "epoch": 1.34, + "learning_rate": 6.6096605266049e-08, + "loss": 0.6964, + "step": 36615 + }, + { + "epoch": 1.34, + "learning_rate": 6.609294320137693e-08, + "loss": 0.6943, + "step": 36616 + }, + { + "epoch": 1.34, + "learning_rate": 6.608928113670486e-08, + "loss": 0.6988, + "step": 36617 + }, + { + "epoch": 1.34, + "learning_rate": 6.60856190720328e-08, + "loss": 0.6936, + "step": 36618 + }, + { + "epoch": 1.34, + "learning_rate": 6.608195700736074e-08, + "loss": 0.6887, + "step": 36619 + }, + { + "epoch": 1.34, + "learning_rate": 6.607829494268868e-08, + "loss": 0.6982, + "step": 36620 + }, + { + "epoch": 1.34, + "learning_rate": 6.607463287801662e-08, + "loss": 0.7014, + "step": 36621 + }, + { + "epoch": 1.34, + "learning_rate": 6.607097081334456e-08, + "loss": 0.6979, + "step": 36622 + }, + { + "epoch": 1.34, + "learning_rate": 6.60673087486725e-08, + "loss": 0.6997, + "step": 36623 + }, + { + "epoch": 1.34, + "learning_rate": 6.606364668400044e-08, + "loss": 0.6945, + "step": 36624 + }, + { + "epoch": 1.34, + "learning_rate": 6.605998461932838e-08, + "loss": 0.7035, + "step": 36625 + }, + { + "epoch": 1.34, + "learning_rate": 6.605632255465631e-08, + "loss": 0.6928, + "step": 36626 + }, + { + "epoch": 1.34, + "learning_rate": 6.605266048998425e-08, + "loss": 0.6907, + "step": 36627 + }, + { + "epoch": 1.34, + "learning_rate": 6.604899842531219e-08, + "loss": 0.6795, + "step": 36628 + }, + { + "epoch": 1.34, + "learning_rate": 6.604533636064013e-08, + "loss": 0.6996, + "step": 36629 + }, + { + "epoch": 1.34, + "learning_rate": 6.604167429596807e-08, + "loss": 0.6855, + "step": 36630 + }, + { + "epoch": 1.34, + "learning_rate": 6.603801223129601e-08, + "loss": 0.6966, + "step": 36631 + }, + { + "epoch": 1.34, + "learning_rate": 6.603435016662395e-08, + "loss": 0.7088, + "step": 36632 + }, + { + "epoch": 1.34, + "learning_rate": 6.603068810195189e-08, + "loss": 0.6882, + "step": 36633 + }, + { + "epoch": 1.34, + "learning_rate": 6.602702603727981e-08, + "loss": 0.698, + "step": 36634 + }, + { + "epoch": 1.34, + "learning_rate": 6.602336397260775e-08, + "loss": 0.6963, + "step": 36635 + }, + { + "epoch": 1.34, + "learning_rate": 6.60197019079357e-08, + "loss": 0.6978, + "step": 36636 + }, + { + "epoch": 1.34, + "learning_rate": 6.601603984326362e-08, + "loss": 0.6806, + "step": 36637 + }, + { + "epoch": 1.34, + "learning_rate": 6.601237777859156e-08, + "loss": 0.6887, + "step": 36638 + }, + { + "epoch": 1.34, + "learning_rate": 6.60087157139195e-08, + "loss": 0.6907, + "step": 36639 + }, + { + "epoch": 1.34, + "learning_rate": 6.600505364924744e-08, + "loss": 0.6778, + "step": 36640 + }, + { + "epoch": 1.34, + "learning_rate": 6.600139158457538e-08, + "loss": 0.6829, + "step": 36641 + }, + { + "epoch": 1.34, + "learning_rate": 6.599772951990332e-08, + "loss": 0.6888, + "step": 36642 + }, + { + "epoch": 1.34, + "learning_rate": 6.599406745523126e-08, + "loss": 0.6994, + "step": 36643 + }, + { + "epoch": 1.34, + "learning_rate": 6.599040539055919e-08, + "loss": 0.703, + "step": 36644 + }, + { + "epoch": 1.34, + "learning_rate": 6.598674332588713e-08, + "loss": 0.6962, + "step": 36645 + }, + { + "epoch": 1.34, + "learning_rate": 6.598308126121507e-08, + "loss": 0.6886, + "step": 36646 + }, + { + "epoch": 1.34, + "learning_rate": 6.5979419196543e-08, + "loss": 0.6809, + "step": 36647 + }, + { + "epoch": 1.34, + "learning_rate": 6.597575713187095e-08, + "loss": 0.7079, + "step": 36648 + }, + { + "epoch": 1.34, + "learning_rate": 6.597209506719889e-08, + "loss": 0.6845, + "step": 36649 + }, + { + "epoch": 1.34, + "learning_rate": 6.596843300252683e-08, + "loss": 0.697, + "step": 36650 + }, + { + "epoch": 1.34, + "learning_rate": 6.596477093785477e-08, + "loss": 0.6986, + "step": 36651 + }, + { + "epoch": 1.34, + "learning_rate": 6.59611088731827e-08, + "loss": 0.692, + "step": 36652 + }, + { + "epoch": 1.34, + "learning_rate": 6.595744680851063e-08, + "loss": 0.6899, + "step": 36653 + }, + { + "epoch": 1.34, + "learning_rate": 6.595378474383857e-08, + "loss": 0.6864, + "step": 36654 + }, + { + "epoch": 1.34, + "learning_rate": 6.595012267916651e-08, + "loss": 0.6906, + "step": 36655 + }, + { + "epoch": 1.34, + "learning_rate": 6.594646061449445e-08, + "loss": 0.6917, + "step": 36656 + }, + { + "epoch": 1.34, + "learning_rate": 6.594279854982239e-08, + "loss": 0.6763, + "step": 36657 + }, + { + "epoch": 1.34, + "learning_rate": 6.593913648515033e-08, + "loss": 0.6811, + "step": 36658 + }, + { + "epoch": 1.34, + "learning_rate": 6.593547442047826e-08, + "loss": 0.6791, + "step": 36659 + }, + { + "epoch": 1.34, + "learning_rate": 6.59318123558062e-08, + "loss": 0.7026, + "step": 36660 + }, + { + "epoch": 1.34, + "learning_rate": 6.592815029113414e-08, + "loss": 0.7034, + "step": 36661 + }, + { + "epoch": 1.34, + "learning_rate": 6.592448822646207e-08, + "loss": 0.7028, + "step": 36662 + }, + { + "epoch": 1.34, + "learning_rate": 6.592082616179e-08, + "loss": 0.7113, + "step": 36663 + }, + { + "epoch": 1.34, + "learning_rate": 6.591716409711795e-08, + "loss": 0.6705, + "step": 36664 + }, + { + "epoch": 1.34, + "learning_rate": 6.591350203244589e-08, + "loss": 0.7022, + "step": 36665 + }, + { + "epoch": 1.34, + "learning_rate": 6.590983996777383e-08, + "loss": 0.6976, + "step": 36666 + }, + { + "epoch": 1.34, + "learning_rate": 6.590617790310177e-08, + "loss": 0.7039, + "step": 36667 + }, + { + "epoch": 1.34, + "learning_rate": 6.59025158384297e-08, + "loss": 0.691, + "step": 36668 + }, + { + "epoch": 1.34, + "learning_rate": 6.589885377375765e-08, + "loss": 0.6953, + "step": 36669 + }, + { + "epoch": 1.34, + "learning_rate": 6.589519170908559e-08, + "loss": 0.6957, + "step": 36670 + }, + { + "epoch": 1.34, + "learning_rate": 6.589152964441351e-08, + "loss": 0.6841, + "step": 36671 + }, + { + "epoch": 1.34, + "learning_rate": 6.588786757974145e-08, + "loss": 0.6873, + "step": 36672 + }, + { + "epoch": 1.34, + "learning_rate": 6.588420551506939e-08, + "loss": 0.6982, + "step": 36673 + }, + { + "epoch": 1.34, + "learning_rate": 6.588054345039733e-08, + "loss": 0.7006, + "step": 36674 + }, + { + "epoch": 1.34, + "learning_rate": 6.587688138572527e-08, + "loss": 0.696, + "step": 36675 + }, + { + "epoch": 1.34, + "learning_rate": 6.587321932105321e-08, + "loss": 0.6987, + "step": 36676 + }, + { + "epoch": 1.34, + "learning_rate": 6.586955725638115e-08, + "loss": 0.6931, + "step": 36677 + }, + { + "epoch": 1.34, + "learning_rate": 6.586589519170909e-08, + "loss": 0.7011, + "step": 36678 + }, + { + "epoch": 1.34, + "learning_rate": 6.586223312703703e-08, + "loss": 0.687, + "step": 36679 + }, + { + "epoch": 1.34, + "learning_rate": 6.585857106236496e-08, + "loss": 0.6969, + "step": 36680 + }, + { + "epoch": 1.34, + "learning_rate": 6.58549089976929e-08, + "loss": 0.6962, + "step": 36681 + }, + { + "epoch": 1.34, + "learning_rate": 6.585124693302084e-08, + "loss": 0.6848, + "step": 36682 + }, + { + "epoch": 1.34, + "learning_rate": 6.584758486834878e-08, + "loss": 0.69, + "step": 36683 + }, + { + "epoch": 1.34, + "learning_rate": 6.58439228036767e-08, + "loss": 0.6949, + "step": 36684 + }, + { + "epoch": 1.34, + "learning_rate": 6.584026073900464e-08, + "loss": 0.684, + "step": 36685 + }, + { + "epoch": 1.34, + "learning_rate": 6.583659867433258e-08, + "loss": 0.6945, + "step": 36686 + }, + { + "epoch": 1.34, + "learning_rate": 6.583293660966052e-08, + "loss": 0.6909, + "step": 36687 + }, + { + "epoch": 1.34, + "learning_rate": 6.582927454498846e-08, + "loss": 0.6988, + "step": 36688 + }, + { + "epoch": 1.34, + "learning_rate": 6.582561248031639e-08, + "loss": 0.6898, + "step": 36689 + }, + { + "epoch": 1.34, + "learning_rate": 6.582195041564433e-08, + "loss": 0.692, + "step": 36690 + }, + { + "epoch": 1.34, + "learning_rate": 6.581828835097227e-08, + "loss": 0.6919, + "step": 36691 + }, + { + "epoch": 1.34, + "learning_rate": 6.581462628630021e-08, + "loss": 0.6856, + "step": 36692 + }, + { + "epoch": 1.34, + "learning_rate": 6.581096422162815e-08, + "loss": 0.7013, + "step": 36693 + }, + { + "epoch": 1.34, + "learning_rate": 6.580730215695609e-08, + "loss": 0.689, + "step": 36694 + }, + { + "epoch": 1.34, + "learning_rate": 6.580364009228403e-08, + "loss": 0.6892, + "step": 36695 + }, + { + "epoch": 1.34, + "learning_rate": 6.579997802761197e-08, + "loss": 0.6889, + "step": 36696 + }, + { + "epoch": 1.34, + "learning_rate": 6.579631596293991e-08, + "loss": 0.6886, + "step": 36697 + }, + { + "epoch": 1.34, + "learning_rate": 6.579265389826784e-08, + "loss": 0.6968, + "step": 36698 + }, + { + "epoch": 1.34, + "learning_rate": 6.578899183359578e-08, + "loss": 0.6815, + "step": 36699 + }, + { + "epoch": 1.34, + "learning_rate": 6.578532976892372e-08, + "loss": 0.696, + "step": 36700 + }, + { + "epoch": 1.34, + "learning_rate": 6.578166770425166e-08, + "loss": 0.6892, + "step": 36701 + }, + { + "epoch": 1.34, + "learning_rate": 6.57780056395796e-08, + "loss": 0.7, + "step": 36702 + }, + { + "epoch": 1.34, + "learning_rate": 6.577434357490754e-08, + "loss": 0.6942, + "step": 36703 + }, + { + "epoch": 1.34, + "learning_rate": 6.577068151023548e-08, + "loss": 0.6899, + "step": 36704 + }, + { + "epoch": 1.34, + "learning_rate": 6.57670194455634e-08, + "loss": 0.702, + "step": 36705 + }, + { + "epoch": 1.34, + "learning_rate": 6.576335738089134e-08, + "loss": 0.698, + "step": 36706 + }, + { + "epoch": 1.34, + "learning_rate": 6.575969531621928e-08, + "loss": 0.6934, + "step": 36707 + }, + { + "epoch": 1.34, + "learning_rate": 6.575603325154721e-08, + "loss": 0.6821, + "step": 36708 + }, + { + "epoch": 1.34, + "learning_rate": 6.575237118687515e-08, + "loss": 0.6925, + "step": 36709 + }, + { + "epoch": 1.34, + "learning_rate": 6.574870912220309e-08, + "loss": 0.6967, + "step": 36710 + }, + { + "epoch": 1.34, + "learning_rate": 6.574504705753103e-08, + "loss": 0.6744, + "step": 36711 + }, + { + "epoch": 1.34, + "learning_rate": 6.574138499285897e-08, + "loss": 0.6916, + "step": 36712 + }, + { + "epoch": 1.34, + "learning_rate": 6.573772292818691e-08, + "loss": 0.6876, + "step": 36713 + }, + { + "epoch": 1.34, + "learning_rate": 6.573406086351485e-08, + "loss": 0.6893, + "step": 36714 + }, + { + "epoch": 1.34, + "learning_rate": 6.573039879884279e-08, + "loss": 0.6951, + "step": 36715 + }, + { + "epoch": 1.34, + "learning_rate": 6.572673673417072e-08, + "loss": 0.6848, + "step": 36716 + }, + { + "epoch": 1.34, + "learning_rate": 6.572307466949866e-08, + "loss": 0.6988, + "step": 36717 + }, + { + "epoch": 1.34, + "learning_rate": 6.57194126048266e-08, + "loss": 0.7031, + "step": 36718 + }, + { + "epoch": 1.34, + "learning_rate": 6.571575054015454e-08, + "loss": 0.6953, + "step": 36719 + }, + { + "epoch": 1.34, + "learning_rate": 6.571208847548248e-08, + "loss": 0.6886, + "step": 36720 + }, + { + "epoch": 1.34, + "learning_rate": 6.570842641081042e-08, + "loss": 0.7058, + "step": 36721 + }, + { + "epoch": 1.34, + "learning_rate": 6.570476434613836e-08, + "loss": 0.6993, + "step": 36722 + }, + { + "epoch": 1.34, + "learning_rate": 6.57011022814663e-08, + "loss": 0.6953, + "step": 36723 + }, + { + "epoch": 1.34, + "learning_rate": 6.569744021679424e-08, + "loss": 0.6849, + "step": 36724 + }, + { + "epoch": 1.34, + "learning_rate": 6.569377815212216e-08, + "loss": 0.6733, + "step": 36725 + }, + { + "epoch": 1.34, + "learning_rate": 6.56901160874501e-08, + "loss": 0.6786, + "step": 36726 + }, + { + "epoch": 1.34, + "learning_rate": 6.568645402277804e-08, + "loss": 0.6981, + "step": 36727 + }, + { + "epoch": 1.34, + "learning_rate": 6.568279195810598e-08, + "loss": 0.6799, + "step": 36728 + }, + { + "epoch": 1.34, + "learning_rate": 6.567912989343392e-08, + "loss": 0.6843, + "step": 36729 + }, + { + "epoch": 1.34, + "learning_rate": 6.567546782876185e-08, + "loss": 0.707, + "step": 36730 + }, + { + "epoch": 1.34, + "learning_rate": 6.567180576408979e-08, + "loss": 0.6974, + "step": 36731 + }, + { + "epoch": 1.34, + "learning_rate": 6.566814369941773e-08, + "loss": 0.6885, + "step": 36732 + }, + { + "epoch": 1.34, + "learning_rate": 6.566448163474567e-08, + "loss": 0.7066, + "step": 36733 + }, + { + "epoch": 1.34, + "learning_rate": 6.56608195700736e-08, + "loss": 0.6825, + "step": 36734 + }, + { + "epoch": 1.34, + "learning_rate": 6.565715750540153e-08, + "loss": 0.6996, + "step": 36735 + }, + { + "epoch": 1.34, + "learning_rate": 6.565349544072947e-08, + "loss": 0.6924, + "step": 36736 + }, + { + "epoch": 1.34, + "learning_rate": 6.564983337605741e-08, + "loss": 0.6982, + "step": 36737 + }, + { + "epoch": 1.34, + "learning_rate": 6.564617131138535e-08, + "loss": 0.6875, + "step": 36738 + }, + { + "epoch": 1.34, + "learning_rate": 6.56425092467133e-08, + "loss": 0.6897, + "step": 36739 + }, + { + "epoch": 1.34, + "learning_rate": 6.563884718204123e-08, + "loss": 0.7139, + "step": 36740 + }, + { + "epoch": 1.34, + "learning_rate": 6.563518511736917e-08, + "loss": 0.7033, + "step": 36741 + }, + { + "epoch": 1.34, + "learning_rate": 6.563152305269711e-08, + "loss": 0.6899, + "step": 36742 + }, + { + "epoch": 1.34, + "learning_rate": 6.562786098802504e-08, + "loss": 0.7002, + "step": 36743 + }, + { + "epoch": 1.34, + "learning_rate": 6.562419892335298e-08, + "loss": 0.6945, + "step": 36744 + }, + { + "epoch": 1.34, + "learning_rate": 6.562053685868092e-08, + "loss": 0.6928, + "step": 36745 + }, + { + "epoch": 1.34, + "learning_rate": 6.561687479400886e-08, + "loss": 0.6898, + "step": 36746 + }, + { + "epoch": 1.34, + "learning_rate": 6.56132127293368e-08, + "loss": 0.7001, + "step": 36747 + }, + { + "epoch": 1.34, + "learning_rate": 6.560955066466474e-08, + "loss": 0.6818, + "step": 36748 + }, + { + "epoch": 1.34, + "learning_rate": 6.560588859999268e-08, + "loss": 0.6845, + "step": 36749 + }, + { + "epoch": 1.34, + "learning_rate": 6.560222653532062e-08, + "loss": 0.6968, + "step": 36750 + }, + { + "epoch": 1.34, + "learning_rate": 6.559856447064856e-08, + "loss": 0.6906, + "step": 36751 + }, + { + "epoch": 1.34, + "learning_rate": 6.559490240597649e-08, + "loss": 0.6834, + "step": 36752 + }, + { + "epoch": 1.34, + "learning_rate": 6.559124034130443e-08, + "loss": 0.698, + "step": 36753 + }, + { + "epoch": 1.34, + "learning_rate": 6.558757827663235e-08, + "loss": 0.6856, + "step": 36754 + }, + { + "epoch": 1.34, + "learning_rate": 6.55839162119603e-08, + "loss": 0.6938, + "step": 36755 + }, + { + "epoch": 1.34, + "learning_rate": 6.558025414728823e-08, + "loss": 0.6927, + "step": 36756 + }, + { + "epoch": 1.34, + "learning_rate": 6.557659208261617e-08, + "loss": 0.6915, + "step": 36757 + }, + { + "epoch": 1.34, + "learning_rate": 6.557293001794411e-08, + "loss": 0.6919, + "step": 36758 + }, + { + "epoch": 1.34, + "learning_rate": 6.556926795327205e-08, + "loss": 0.7025, + "step": 36759 + }, + { + "epoch": 1.34, + "learning_rate": 6.55656058886e-08, + "loss": 0.6858, + "step": 36760 + }, + { + "epoch": 1.34, + "learning_rate": 6.556194382392792e-08, + "loss": 0.6894, + "step": 36761 + }, + { + "epoch": 1.35, + "learning_rate": 6.555828175925586e-08, + "loss": 0.6981, + "step": 36762 + }, + { + "epoch": 1.35, + "learning_rate": 6.55546196945838e-08, + "loss": 0.7035, + "step": 36763 + }, + { + "epoch": 1.35, + "learning_rate": 6.555095762991174e-08, + "loss": 0.6927, + "step": 36764 + }, + { + "epoch": 1.35, + "learning_rate": 6.554729556523968e-08, + "loss": 0.6934, + "step": 36765 + }, + { + "epoch": 1.35, + "learning_rate": 6.554363350056762e-08, + "loss": 0.6973, + "step": 36766 + }, + { + "epoch": 1.35, + "learning_rate": 6.553997143589556e-08, + "loss": 0.695, + "step": 36767 + }, + { + "epoch": 1.35, + "learning_rate": 6.55363093712235e-08, + "loss": 0.6974, + "step": 36768 + }, + { + "epoch": 1.35, + "learning_rate": 6.553264730655144e-08, + "loss": 0.6908, + "step": 36769 + }, + { + "epoch": 1.35, + "learning_rate": 6.552898524187937e-08, + "loss": 0.691, + "step": 36770 + }, + { + "epoch": 1.35, + "learning_rate": 6.55253231772073e-08, + "loss": 0.686, + "step": 36771 + }, + { + "epoch": 1.35, + "learning_rate": 6.552166111253525e-08, + "loss": 0.6939, + "step": 36772 + }, + { + "epoch": 1.35, + "learning_rate": 6.551799904786319e-08, + "loss": 0.6945, + "step": 36773 + }, + { + "epoch": 1.35, + "learning_rate": 6.551433698319113e-08, + "loss": 0.6945, + "step": 36774 + }, + { + "epoch": 1.35, + "learning_rate": 6.551067491851907e-08, + "loss": 0.6847, + "step": 36775 + }, + { + "epoch": 1.35, + "learning_rate": 6.550701285384699e-08, + "loss": 0.6995, + "step": 36776 + }, + { + "epoch": 1.35, + "learning_rate": 6.550335078917493e-08, + "loss": 0.6929, + "step": 36777 + }, + { + "epoch": 1.35, + "learning_rate": 6.549968872450287e-08, + "loss": 0.6833, + "step": 36778 + }, + { + "epoch": 1.35, + "learning_rate": 6.54960266598308e-08, + "loss": 0.6886, + "step": 36779 + }, + { + "epoch": 1.35, + "learning_rate": 6.549236459515874e-08, + "loss": 0.6799, + "step": 36780 + }, + { + "epoch": 1.35, + "learning_rate": 6.548870253048668e-08, + "loss": 0.6974, + "step": 36781 + }, + { + "epoch": 1.35, + "learning_rate": 6.548504046581462e-08, + "loss": 0.6936, + "step": 36782 + }, + { + "epoch": 1.35, + "learning_rate": 6.548137840114256e-08, + "loss": 0.698, + "step": 36783 + }, + { + "epoch": 1.35, + "learning_rate": 6.54777163364705e-08, + "loss": 0.7036, + "step": 36784 + }, + { + "epoch": 1.35, + "learning_rate": 6.547405427179844e-08, + "loss": 0.6922, + "step": 36785 + }, + { + "epoch": 1.35, + "learning_rate": 6.547039220712638e-08, + "loss": 0.7114, + "step": 36786 + }, + { + "epoch": 1.35, + "learning_rate": 6.54667301424543e-08, + "loss": 0.6944, + "step": 36787 + }, + { + "epoch": 1.35, + "learning_rate": 6.546306807778225e-08, + "loss": 0.7044, + "step": 36788 + }, + { + "epoch": 1.35, + "learning_rate": 6.545940601311019e-08, + "loss": 0.6889, + "step": 36789 + }, + { + "epoch": 1.35, + "learning_rate": 6.545574394843813e-08, + "loss": 0.666, + "step": 36790 + }, + { + "epoch": 1.35, + "learning_rate": 6.545208188376606e-08, + "loss": 0.6979, + "step": 36791 + }, + { + "epoch": 1.35, + "learning_rate": 6.5448419819094e-08, + "loss": 0.6886, + "step": 36792 + }, + { + "epoch": 1.35, + "learning_rate": 6.544475775442194e-08, + "loss": 0.7065, + "step": 36793 + }, + { + "epoch": 1.35, + "learning_rate": 6.544109568974988e-08, + "loss": 0.7057, + "step": 36794 + }, + { + "epoch": 1.35, + "learning_rate": 6.543743362507782e-08, + "loss": 0.7045, + "step": 36795 + }, + { + "epoch": 1.35, + "learning_rate": 6.543377156040575e-08, + "loss": 0.6827, + "step": 36796 + }, + { + "epoch": 1.35, + "learning_rate": 6.543010949573369e-08, + "loss": 0.6876, + "step": 36797 + }, + { + "epoch": 1.35, + "learning_rate": 6.542644743106163e-08, + "loss": 0.6843, + "step": 36798 + }, + { + "epoch": 1.35, + "learning_rate": 6.542278536638957e-08, + "loss": 0.7062, + "step": 36799 + }, + { + "epoch": 1.35, + "learning_rate": 6.541912330171751e-08, + "loss": 0.6955, + "step": 36800 + }, + { + "epoch": 1.35, + "learning_rate": 6.541546123704544e-08, + "loss": 0.6829, + "step": 36801 + }, + { + "epoch": 1.35, + "learning_rate": 6.541179917237338e-08, + "loss": 0.6979, + "step": 36802 + }, + { + "epoch": 1.35, + "learning_rate": 6.540813710770132e-08, + "loss": 0.7049, + "step": 36803 + }, + { + "epoch": 1.35, + "learning_rate": 6.540447504302926e-08, + "loss": 0.697, + "step": 36804 + }, + { + "epoch": 1.35, + "learning_rate": 6.540081297835718e-08, + "loss": 0.6958, + "step": 36805 + }, + { + "epoch": 1.35, + "learning_rate": 6.539715091368512e-08, + "loss": 0.6859, + "step": 36806 + }, + { + "epoch": 1.35, + "learning_rate": 6.539348884901306e-08, + "loss": 0.711, + "step": 36807 + }, + { + "epoch": 1.35, + "learning_rate": 6.5389826784341e-08, + "loss": 0.6904, + "step": 36808 + }, + { + "epoch": 1.35, + "learning_rate": 6.538616471966894e-08, + "loss": 0.6973, + "step": 36809 + }, + { + "epoch": 1.35, + "learning_rate": 6.538250265499688e-08, + "loss": 0.6871, + "step": 36810 + }, + { + "epoch": 1.35, + "learning_rate": 6.537884059032482e-08, + "loss": 0.6992, + "step": 36811 + }, + { + "epoch": 1.35, + "learning_rate": 6.537517852565276e-08, + "loss": 0.701, + "step": 36812 + }, + { + "epoch": 1.35, + "learning_rate": 6.53715164609807e-08, + "loss": 0.6862, + "step": 36813 + }, + { + "epoch": 1.35, + "learning_rate": 6.536785439630863e-08, + "loss": 0.6901, + "step": 36814 + }, + { + "epoch": 1.35, + "learning_rate": 6.536419233163657e-08, + "loss": 0.7138, + "step": 36815 + }, + { + "epoch": 1.35, + "learning_rate": 6.536053026696451e-08, + "loss": 0.6968, + "step": 36816 + }, + { + "epoch": 1.35, + "learning_rate": 6.535686820229245e-08, + "loss": 0.7003, + "step": 36817 + }, + { + "epoch": 1.35, + "learning_rate": 6.535320613762039e-08, + "loss": 0.6855, + "step": 36818 + }, + { + "epoch": 1.35, + "learning_rate": 6.534954407294833e-08, + "loss": 0.6885, + "step": 36819 + }, + { + "epoch": 1.35, + "learning_rate": 6.534588200827627e-08, + "loss": 0.6936, + "step": 36820 + }, + { + "epoch": 1.35, + "learning_rate": 6.534221994360421e-08, + "loss": 0.7025, + "step": 36821 + }, + { + "epoch": 1.35, + "learning_rate": 6.533855787893215e-08, + "loss": 0.6831, + "step": 36822 + }, + { + "epoch": 1.35, + "learning_rate": 6.533489581426008e-08, + "loss": 0.7028, + "step": 36823 + }, + { + "epoch": 1.35, + "learning_rate": 6.533123374958802e-08, + "loss": 0.6943, + "step": 36824 + }, + { + "epoch": 1.35, + "learning_rate": 6.532757168491594e-08, + "loss": 0.6949, + "step": 36825 + }, + { + "epoch": 1.35, + "learning_rate": 6.532390962024388e-08, + "loss": 0.6967, + "step": 36826 + }, + { + "epoch": 1.35, + "learning_rate": 6.532024755557182e-08, + "loss": 0.7072, + "step": 36827 + }, + { + "epoch": 1.35, + "learning_rate": 6.531658549089976e-08, + "loss": 0.704, + "step": 36828 + }, + { + "epoch": 1.35, + "learning_rate": 6.53129234262277e-08, + "loss": 0.697, + "step": 36829 + }, + { + "epoch": 1.35, + "learning_rate": 6.530926136155564e-08, + "loss": 0.7002, + "step": 36830 + }, + { + "epoch": 1.35, + "learning_rate": 6.530559929688358e-08, + "loss": 0.6923, + "step": 36831 + }, + { + "epoch": 1.35, + "learning_rate": 6.530193723221151e-08, + "loss": 0.6893, + "step": 36832 + }, + { + "epoch": 1.35, + "learning_rate": 6.529827516753945e-08, + "loss": 0.6907, + "step": 36833 + }, + { + "epoch": 1.35, + "learning_rate": 6.529461310286739e-08, + "loss": 0.6979, + "step": 36834 + }, + { + "epoch": 1.35, + "learning_rate": 6.529095103819533e-08, + "loss": 0.6836, + "step": 36835 + }, + { + "epoch": 1.35, + "learning_rate": 6.528728897352327e-08, + "loss": 0.6896, + "step": 36836 + }, + { + "epoch": 1.35, + "learning_rate": 6.528362690885121e-08, + "loss": 0.6693, + "step": 36837 + }, + { + "epoch": 1.35, + "learning_rate": 6.527996484417915e-08, + "loss": 0.6941, + "step": 36838 + }, + { + "epoch": 1.35, + "learning_rate": 6.527630277950709e-08, + "loss": 0.6885, + "step": 36839 + }, + { + "epoch": 1.35, + "learning_rate": 6.527264071483503e-08, + "loss": 0.6971, + "step": 36840 + }, + { + "epoch": 1.35, + "learning_rate": 6.526897865016296e-08, + "loss": 0.717, + "step": 36841 + }, + { + "epoch": 1.35, + "learning_rate": 6.52653165854909e-08, + "loss": 0.7054, + "step": 36842 + }, + { + "epoch": 1.35, + "learning_rate": 6.526165452081884e-08, + "loss": 0.6977, + "step": 36843 + }, + { + "epoch": 1.35, + "learning_rate": 6.525799245614678e-08, + "loss": 0.6912, + "step": 36844 + }, + { + "epoch": 1.35, + "learning_rate": 6.525433039147472e-08, + "loss": 0.6863, + "step": 36845 + }, + { + "epoch": 1.35, + "learning_rate": 6.525066832680266e-08, + "loss": 0.6915, + "step": 36846 + }, + { + "epoch": 1.35, + "learning_rate": 6.524700626213058e-08, + "loss": 0.6908, + "step": 36847 + }, + { + "epoch": 1.35, + "learning_rate": 6.524334419745852e-08, + "loss": 0.6856, + "step": 36848 + }, + { + "epoch": 1.35, + "learning_rate": 6.523968213278646e-08, + "loss": 0.6743, + "step": 36849 + }, + { + "epoch": 1.35, + "learning_rate": 6.523602006811439e-08, + "loss": 0.678, + "step": 36850 + }, + { + "epoch": 1.35, + "learning_rate": 6.523235800344233e-08, + "loss": 0.6925, + "step": 36851 + }, + { + "epoch": 1.35, + "learning_rate": 6.522869593877027e-08, + "loss": 0.7048, + "step": 36852 + }, + { + "epoch": 1.35, + "learning_rate": 6.522503387409821e-08, + "loss": 0.6876, + "step": 36853 + }, + { + "epoch": 1.35, + "learning_rate": 6.522137180942615e-08, + "loss": 0.6895, + "step": 36854 + }, + { + "epoch": 1.35, + "learning_rate": 6.521770974475409e-08, + "loss": 0.6846, + "step": 36855 + }, + { + "epoch": 1.35, + "learning_rate": 6.521404768008203e-08, + "loss": 0.6869, + "step": 36856 + }, + { + "epoch": 1.35, + "learning_rate": 6.521038561540997e-08, + "loss": 0.6893, + "step": 36857 + }, + { + "epoch": 1.35, + "learning_rate": 6.520672355073791e-08, + "loss": 0.6834, + "step": 36858 + }, + { + "epoch": 1.35, + "learning_rate": 6.520306148606583e-08, + "loss": 0.696, + "step": 36859 + }, + { + "epoch": 1.35, + "learning_rate": 6.519939942139377e-08, + "loss": 0.6998, + "step": 36860 + }, + { + "epoch": 1.35, + "learning_rate": 6.519573735672171e-08, + "loss": 0.6891, + "step": 36861 + }, + { + "epoch": 1.35, + "learning_rate": 6.519207529204965e-08, + "loss": 0.6781, + "step": 36862 + }, + { + "epoch": 1.35, + "learning_rate": 6.51884132273776e-08, + "loss": 0.7142, + "step": 36863 + }, + { + "epoch": 1.35, + "learning_rate": 6.518475116270553e-08, + "loss": 0.6926, + "step": 36864 + }, + { + "epoch": 1.35, + "learning_rate": 6.518108909803347e-08, + "loss": 0.6905, + "step": 36865 + }, + { + "epoch": 1.35, + "learning_rate": 6.517742703336141e-08, + "loss": 0.6865, + "step": 36866 + }, + { + "epoch": 1.35, + "learning_rate": 6.517376496868935e-08, + "loss": 0.6907, + "step": 36867 + }, + { + "epoch": 1.35, + "learning_rate": 6.517010290401728e-08, + "loss": 0.7008, + "step": 36868 + }, + { + "epoch": 1.35, + "learning_rate": 6.516644083934522e-08, + "loss": 0.683, + "step": 36869 + }, + { + "epoch": 1.35, + "learning_rate": 6.516277877467316e-08, + "loss": 0.6871, + "step": 36870 + }, + { + "epoch": 1.35, + "learning_rate": 6.51591167100011e-08, + "loss": 0.694, + "step": 36871 + }, + { + "epoch": 1.35, + "learning_rate": 6.515545464532903e-08, + "loss": 0.6873, + "step": 36872 + }, + { + "epoch": 1.35, + "learning_rate": 6.515179258065697e-08, + "loss": 0.69, + "step": 36873 + }, + { + "epoch": 1.35, + "learning_rate": 6.514813051598491e-08, + "loss": 0.6952, + "step": 36874 + }, + { + "epoch": 1.35, + "learning_rate": 6.514446845131285e-08, + "loss": 0.6895, + "step": 36875 + }, + { + "epoch": 1.35, + "learning_rate": 6.514080638664079e-08, + "loss": 0.6942, + "step": 36876 + }, + { + "epoch": 1.35, + "learning_rate": 6.513714432196871e-08, + "loss": 0.7002, + "step": 36877 + }, + { + "epoch": 1.35, + "learning_rate": 6.513348225729665e-08, + "loss": 0.6856, + "step": 36878 + }, + { + "epoch": 1.35, + "learning_rate": 6.51298201926246e-08, + "loss": 0.7012, + "step": 36879 + }, + { + "epoch": 1.35, + "learning_rate": 6.512615812795253e-08, + "loss": 0.6992, + "step": 36880 + }, + { + "epoch": 1.35, + "learning_rate": 6.512249606328047e-08, + "loss": 0.703, + "step": 36881 + }, + { + "epoch": 1.35, + "learning_rate": 6.511883399860841e-08, + "loss": 0.6921, + "step": 36882 + }, + { + "epoch": 1.35, + "learning_rate": 6.511517193393635e-08, + "loss": 0.6871, + "step": 36883 + }, + { + "epoch": 1.35, + "learning_rate": 6.511150986926429e-08, + "loss": 0.6893, + "step": 36884 + }, + { + "epoch": 1.35, + "learning_rate": 6.510784780459223e-08, + "loss": 0.7023, + "step": 36885 + }, + { + "epoch": 1.35, + "learning_rate": 6.510418573992016e-08, + "loss": 0.678, + "step": 36886 + }, + { + "epoch": 1.35, + "learning_rate": 6.51005236752481e-08, + "loss": 0.6995, + "step": 36887 + }, + { + "epoch": 1.35, + "learning_rate": 6.509686161057604e-08, + "loss": 0.6929, + "step": 36888 + }, + { + "epoch": 1.35, + "learning_rate": 6.509319954590398e-08, + "loss": 0.7112, + "step": 36889 + }, + { + "epoch": 1.35, + "learning_rate": 6.508953748123192e-08, + "loss": 0.6878, + "step": 36890 + }, + { + "epoch": 1.35, + "learning_rate": 6.508587541655986e-08, + "loss": 0.6934, + "step": 36891 + }, + { + "epoch": 1.35, + "learning_rate": 6.50822133518878e-08, + "loss": 0.6938, + "step": 36892 + }, + { + "epoch": 1.35, + "learning_rate": 6.507855128721574e-08, + "loss": 0.6998, + "step": 36893 + }, + { + "epoch": 1.35, + "learning_rate": 6.507488922254367e-08, + "loss": 0.6791, + "step": 36894 + }, + { + "epoch": 1.35, + "learning_rate": 6.50712271578716e-08, + "loss": 0.693, + "step": 36895 + }, + { + "epoch": 1.35, + "learning_rate": 6.506756509319953e-08, + "loss": 0.696, + "step": 36896 + }, + { + "epoch": 1.35, + "learning_rate": 6.506390302852747e-08, + "loss": 0.7025, + "step": 36897 + }, + { + "epoch": 1.35, + "learning_rate": 6.506024096385541e-08, + "loss": 0.6954, + "step": 36898 + }, + { + "epoch": 1.35, + "learning_rate": 6.505657889918335e-08, + "loss": 0.7009, + "step": 36899 + }, + { + "epoch": 1.35, + "learning_rate": 6.505291683451129e-08, + "loss": 0.6977, + "step": 36900 + }, + { + "epoch": 1.35, + "learning_rate": 6.504925476983923e-08, + "loss": 0.6886, + "step": 36901 + }, + { + "epoch": 1.35, + "learning_rate": 6.504559270516717e-08, + "loss": 0.7018, + "step": 36902 + }, + { + "epoch": 1.35, + "learning_rate": 6.504193064049511e-08, + "loss": 0.7046, + "step": 36903 + }, + { + "epoch": 1.35, + "learning_rate": 6.503826857582304e-08, + "loss": 0.6879, + "step": 36904 + }, + { + "epoch": 1.35, + "learning_rate": 6.503460651115098e-08, + "loss": 0.6941, + "step": 36905 + }, + { + "epoch": 1.35, + "learning_rate": 6.503094444647892e-08, + "loss": 0.6958, + "step": 36906 + }, + { + "epoch": 1.35, + "learning_rate": 6.502728238180686e-08, + "loss": 0.7008, + "step": 36907 + }, + { + "epoch": 1.35, + "learning_rate": 6.50236203171348e-08, + "loss": 0.6914, + "step": 36908 + }, + { + "epoch": 1.35, + "learning_rate": 6.501995825246274e-08, + "loss": 0.7057, + "step": 36909 + }, + { + "epoch": 1.35, + "learning_rate": 6.501629618779068e-08, + "loss": 0.6966, + "step": 36910 + }, + { + "epoch": 1.35, + "learning_rate": 6.501263412311862e-08, + "loss": 0.6861, + "step": 36911 + }, + { + "epoch": 1.35, + "learning_rate": 6.500897205844656e-08, + "loss": 0.6901, + "step": 36912 + }, + { + "epoch": 1.35, + "learning_rate": 6.500530999377448e-08, + "loss": 0.6875, + "step": 36913 + }, + { + "epoch": 1.35, + "learning_rate": 6.500164792910242e-08, + "loss": 0.6887, + "step": 36914 + }, + { + "epoch": 1.35, + "learning_rate": 6.499798586443036e-08, + "loss": 0.6905, + "step": 36915 + }, + { + "epoch": 1.35, + "learning_rate": 6.49943237997583e-08, + "loss": 0.686, + "step": 36916 + }, + { + "epoch": 1.35, + "learning_rate": 6.499066173508624e-08, + "loss": 0.6928, + "step": 36917 + }, + { + "epoch": 1.35, + "learning_rate": 6.498699967041417e-08, + "loss": 0.6889, + "step": 36918 + }, + { + "epoch": 1.35, + "learning_rate": 6.498333760574211e-08, + "loss": 0.6882, + "step": 36919 + }, + { + "epoch": 1.35, + "learning_rate": 6.497967554107005e-08, + "loss": 0.707, + "step": 36920 + }, + { + "epoch": 1.35, + "learning_rate": 6.497601347639799e-08, + "loss": 0.7021, + "step": 36921 + }, + { + "epoch": 1.35, + "learning_rate": 6.497235141172592e-08, + "loss": 0.6863, + "step": 36922 + }, + { + "epoch": 1.35, + "learning_rate": 6.496868934705386e-08, + "loss": 0.7026, + "step": 36923 + }, + { + "epoch": 1.35, + "learning_rate": 6.49650272823818e-08, + "loss": 0.6936, + "step": 36924 + }, + { + "epoch": 1.35, + "learning_rate": 6.496136521770974e-08, + "loss": 0.6906, + "step": 36925 + }, + { + "epoch": 1.35, + "learning_rate": 6.495770315303768e-08, + "loss": 0.6947, + "step": 36926 + }, + { + "epoch": 1.35, + "learning_rate": 6.495404108836562e-08, + "loss": 0.6843, + "step": 36927 + }, + { + "epoch": 1.35, + "learning_rate": 6.495037902369356e-08, + "loss": 0.6946, + "step": 36928 + }, + { + "epoch": 1.35, + "learning_rate": 6.49467169590215e-08, + "loss": 0.6905, + "step": 36929 + }, + { + "epoch": 1.35, + "learning_rate": 6.494305489434944e-08, + "loss": 0.6874, + "step": 36930 + }, + { + "epoch": 1.35, + "learning_rate": 6.493939282967736e-08, + "loss": 0.7081, + "step": 36931 + }, + { + "epoch": 1.35, + "learning_rate": 6.49357307650053e-08, + "loss": 0.6912, + "step": 36932 + }, + { + "epoch": 1.35, + "learning_rate": 6.493206870033324e-08, + "loss": 0.6975, + "step": 36933 + }, + { + "epoch": 1.35, + "learning_rate": 6.492840663566118e-08, + "loss": 0.6787, + "step": 36934 + }, + { + "epoch": 1.35, + "learning_rate": 6.492474457098912e-08, + "loss": 0.6913, + "step": 36935 + }, + { + "epoch": 1.35, + "learning_rate": 6.492108250631706e-08, + "loss": 0.7035, + "step": 36936 + }, + { + "epoch": 1.35, + "learning_rate": 6.4917420441645e-08, + "loss": 0.6846, + "step": 36937 + }, + { + "epoch": 1.35, + "learning_rate": 6.491375837697294e-08, + "loss": 0.6934, + "step": 36938 + }, + { + "epoch": 1.35, + "learning_rate": 6.491009631230088e-08, + "loss": 0.6943, + "step": 36939 + }, + { + "epoch": 1.35, + "learning_rate": 6.490643424762881e-08, + "loss": 0.7068, + "step": 36940 + }, + { + "epoch": 1.35, + "learning_rate": 6.490277218295675e-08, + "loss": 0.6978, + "step": 36941 + }, + { + "epoch": 1.35, + "learning_rate": 6.489911011828469e-08, + "loss": 0.7038, + "step": 36942 + }, + { + "epoch": 1.35, + "learning_rate": 6.489544805361262e-08, + "loss": 0.6957, + "step": 36943 + }, + { + "epoch": 1.35, + "learning_rate": 6.489178598894056e-08, + "loss": 0.6721, + "step": 36944 + }, + { + "epoch": 1.35, + "learning_rate": 6.48881239242685e-08, + "loss": 0.697, + "step": 36945 + }, + { + "epoch": 1.35, + "learning_rate": 6.488446185959644e-08, + "loss": 0.6844, + "step": 36946 + }, + { + "epoch": 1.35, + "learning_rate": 6.488079979492438e-08, + "loss": 0.7013, + "step": 36947 + }, + { + "epoch": 1.35, + "learning_rate": 6.487713773025232e-08, + "loss": 0.7025, + "step": 36948 + }, + { + "epoch": 1.35, + "learning_rate": 6.487347566558024e-08, + "loss": 0.6868, + "step": 36949 + }, + { + "epoch": 1.35, + "learning_rate": 6.486981360090818e-08, + "loss": 0.7048, + "step": 36950 + }, + { + "epoch": 1.35, + "learning_rate": 6.486615153623612e-08, + "loss": 0.698, + "step": 36951 + }, + { + "epoch": 1.35, + "learning_rate": 6.486248947156406e-08, + "loss": 0.6957, + "step": 36952 + }, + { + "epoch": 1.35, + "learning_rate": 6.4858827406892e-08, + "loss": 0.6869, + "step": 36953 + }, + { + "epoch": 1.35, + "learning_rate": 6.485516534221994e-08, + "loss": 0.6737, + "step": 36954 + }, + { + "epoch": 1.35, + "learning_rate": 6.485150327754788e-08, + "loss": 0.6863, + "step": 36955 + }, + { + "epoch": 1.35, + "learning_rate": 6.484784121287582e-08, + "loss": 0.7046, + "step": 36956 + }, + { + "epoch": 1.35, + "learning_rate": 6.484417914820376e-08, + "loss": 0.6891, + "step": 36957 + }, + { + "epoch": 1.35, + "learning_rate": 6.484051708353169e-08, + "loss": 0.7012, + "step": 36958 + }, + { + "epoch": 1.35, + "learning_rate": 6.483685501885963e-08, + "loss": 0.6992, + "step": 36959 + }, + { + "epoch": 1.35, + "learning_rate": 6.483319295418757e-08, + "loss": 0.6938, + "step": 36960 + }, + { + "epoch": 1.35, + "learning_rate": 6.482953088951551e-08, + "loss": 0.6922, + "step": 36961 + }, + { + "epoch": 1.35, + "learning_rate": 6.482586882484345e-08, + "loss": 0.6903, + "step": 36962 + }, + { + "epoch": 1.35, + "learning_rate": 6.482220676017139e-08, + "loss": 0.6929, + "step": 36963 + }, + { + "epoch": 1.35, + "learning_rate": 6.481854469549933e-08, + "loss": 0.7005, + "step": 36964 + }, + { + "epoch": 1.35, + "learning_rate": 6.481488263082726e-08, + "loss": 0.6924, + "step": 36965 + }, + { + "epoch": 1.35, + "learning_rate": 6.48112205661552e-08, + "loss": 0.6888, + "step": 36966 + }, + { + "epoch": 1.35, + "learning_rate": 6.480755850148312e-08, + "loss": 0.6957, + "step": 36967 + }, + { + "epoch": 1.35, + "learning_rate": 6.480389643681106e-08, + "loss": 0.6756, + "step": 36968 + }, + { + "epoch": 1.35, + "learning_rate": 6.4800234372139e-08, + "loss": 0.6805, + "step": 36969 + }, + { + "epoch": 1.35, + "learning_rate": 6.479657230746694e-08, + "loss": 0.7012, + "step": 36970 + }, + { + "epoch": 1.35, + "learning_rate": 6.479291024279488e-08, + "loss": 0.685, + "step": 36971 + }, + { + "epoch": 1.35, + "learning_rate": 6.478924817812282e-08, + "loss": 0.6933, + "step": 36972 + }, + { + "epoch": 1.35, + "learning_rate": 6.478558611345076e-08, + "loss": 0.6973, + "step": 36973 + }, + { + "epoch": 1.35, + "learning_rate": 6.47819240487787e-08, + "loss": 0.6889, + "step": 36974 + }, + { + "epoch": 1.35, + "learning_rate": 6.477826198410664e-08, + "loss": 0.6856, + "step": 36975 + }, + { + "epoch": 1.35, + "learning_rate": 6.477459991943457e-08, + "loss": 0.7009, + "step": 36976 + }, + { + "epoch": 1.35, + "learning_rate": 6.477093785476251e-08, + "loss": 0.7033, + "step": 36977 + }, + { + "epoch": 1.35, + "learning_rate": 6.476727579009045e-08, + "loss": 0.684, + "step": 36978 + }, + { + "epoch": 1.35, + "learning_rate": 6.476361372541839e-08, + "loss": 0.7108, + "step": 36979 + }, + { + "epoch": 1.35, + "learning_rate": 6.475995166074633e-08, + "loss": 0.7022, + "step": 36980 + }, + { + "epoch": 1.35, + "learning_rate": 6.475628959607427e-08, + "loss": 0.6856, + "step": 36981 + }, + { + "epoch": 1.35, + "learning_rate": 6.475262753140221e-08, + "loss": 0.6873, + "step": 36982 + }, + { + "epoch": 1.35, + "learning_rate": 6.474896546673015e-08, + "loss": 0.6966, + "step": 36983 + }, + { + "epoch": 1.35, + "learning_rate": 6.474530340205809e-08, + "loss": 0.6872, + "step": 36984 + }, + { + "epoch": 1.35, + "learning_rate": 6.474164133738601e-08, + "loss": 0.689, + "step": 36985 + }, + { + "epoch": 1.35, + "learning_rate": 6.473797927271395e-08, + "loss": 0.6865, + "step": 36986 + }, + { + "epoch": 1.35, + "learning_rate": 6.47343172080419e-08, + "loss": 0.6899, + "step": 36987 + }, + { + "epoch": 1.35, + "learning_rate": 6.473065514336983e-08, + "loss": 0.6934, + "step": 36988 + }, + { + "epoch": 1.35, + "learning_rate": 6.472699307869776e-08, + "loss": 0.6982, + "step": 36989 + }, + { + "epoch": 1.35, + "learning_rate": 6.47233310140257e-08, + "loss": 0.6958, + "step": 36990 + }, + { + "epoch": 1.35, + "learning_rate": 6.471966894935364e-08, + "loss": 0.6899, + "step": 36991 + }, + { + "epoch": 1.35, + "learning_rate": 6.471600688468158e-08, + "loss": 0.701, + "step": 36992 + }, + { + "epoch": 1.35, + "learning_rate": 6.471234482000952e-08, + "loss": 0.6903, + "step": 36993 + }, + { + "epoch": 1.35, + "learning_rate": 6.470868275533745e-08, + "loss": 0.6927, + "step": 36994 + }, + { + "epoch": 1.35, + "learning_rate": 6.470502069066539e-08, + "loss": 0.7194, + "step": 36995 + }, + { + "epoch": 1.35, + "learning_rate": 6.470135862599333e-08, + "loss": 0.6864, + "step": 36996 + }, + { + "epoch": 1.35, + "learning_rate": 6.469769656132127e-08, + "loss": 0.7079, + "step": 36997 + }, + { + "epoch": 1.35, + "learning_rate": 6.469403449664921e-08, + "loss": 0.6972, + "step": 36998 + }, + { + "epoch": 1.35, + "learning_rate": 6.469037243197715e-08, + "loss": 0.6836, + "step": 36999 + }, + { + "epoch": 1.35, + "learning_rate": 6.468671036730509e-08, + "loss": 0.6946, + "step": 37000 + }, + { + "epoch": 1.35, + "learning_rate": 6.468304830263303e-08, + "loss": 0.6989, + "step": 37001 + }, + { + "epoch": 1.35, + "learning_rate": 6.467938623796097e-08, + "loss": 0.7081, + "step": 37002 + }, + { + "epoch": 1.35, + "learning_rate": 6.467572417328889e-08, + "loss": 0.6874, + "step": 37003 + }, + { + "epoch": 1.35, + "learning_rate": 6.467206210861683e-08, + "loss": 0.7007, + "step": 37004 + }, + { + "epoch": 1.35, + "learning_rate": 6.466840004394477e-08, + "loss": 0.6891, + "step": 37005 + }, + { + "epoch": 1.35, + "learning_rate": 6.466473797927271e-08, + "loss": 0.6888, + "step": 37006 + }, + { + "epoch": 1.35, + "learning_rate": 6.466107591460065e-08, + "loss": 0.6913, + "step": 37007 + }, + { + "epoch": 1.35, + "learning_rate": 6.465741384992859e-08, + "loss": 0.682, + "step": 37008 + }, + { + "epoch": 1.35, + "learning_rate": 6.465375178525653e-08, + "loss": 0.6906, + "step": 37009 + }, + { + "epoch": 1.35, + "learning_rate": 6.465008972058447e-08, + "loss": 0.7068, + "step": 37010 + }, + { + "epoch": 1.35, + "learning_rate": 6.46464276559124e-08, + "loss": 0.6973, + "step": 37011 + }, + { + "epoch": 1.35, + "learning_rate": 6.464276559124034e-08, + "loss": 0.6914, + "step": 37012 + }, + { + "epoch": 1.35, + "learning_rate": 6.463910352656828e-08, + "loss": 0.7006, + "step": 37013 + }, + { + "epoch": 1.35, + "learning_rate": 6.46354414618962e-08, + "loss": 0.6865, + "step": 37014 + }, + { + "epoch": 1.35, + "learning_rate": 6.463177939722415e-08, + "loss": 0.7023, + "step": 37015 + }, + { + "epoch": 1.35, + "learning_rate": 6.462811733255209e-08, + "loss": 0.6885, + "step": 37016 + }, + { + "epoch": 1.35, + "learning_rate": 6.462445526788003e-08, + "loss": 0.6942, + "step": 37017 + }, + { + "epoch": 1.35, + "learning_rate": 6.462079320320797e-08, + "loss": 0.6894, + "step": 37018 + }, + { + "epoch": 1.35, + "learning_rate": 6.46171311385359e-08, + "loss": 0.688, + "step": 37019 + }, + { + "epoch": 1.35, + "learning_rate": 6.461346907386385e-08, + "loss": 0.6916, + "step": 37020 + }, + { + "epoch": 1.35, + "learning_rate": 6.460980700919177e-08, + "loss": 0.6926, + "step": 37021 + }, + { + "epoch": 1.35, + "learning_rate": 6.460614494451971e-08, + "loss": 0.7078, + "step": 37022 + }, + { + "epoch": 1.35, + "learning_rate": 6.460248287984765e-08, + "loss": 0.6976, + "step": 37023 + }, + { + "epoch": 1.35, + "learning_rate": 6.459882081517559e-08, + "loss": 0.6978, + "step": 37024 + }, + { + "epoch": 1.35, + "learning_rate": 6.459515875050353e-08, + "loss": 0.691, + "step": 37025 + }, + { + "epoch": 1.35, + "learning_rate": 6.459149668583147e-08, + "loss": 0.6963, + "step": 37026 + }, + { + "epoch": 1.35, + "learning_rate": 6.458783462115941e-08, + "loss": 0.6821, + "step": 37027 + }, + { + "epoch": 1.35, + "learning_rate": 6.458417255648735e-08, + "loss": 0.6981, + "step": 37028 + }, + { + "epoch": 1.35, + "learning_rate": 6.458051049181529e-08, + "loss": 0.696, + "step": 37029 + }, + { + "epoch": 1.35, + "learning_rate": 6.457684842714322e-08, + "loss": 0.6817, + "step": 37030 + }, + { + "epoch": 1.35, + "learning_rate": 6.457318636247116e-08, + "loss": 0.6942, + "step": 37031 + }, + { + "epoch": 1.35, + "learning_rate": 6.45695242977991e-08, + "loss": 0.6946, + "step": 37032 + }, + { + "epoch": 1.35, + "learning_rate": 6.456586223312704e-08, + "loss": 0.6845, + "step": 37033 + }, + { + "epoch": 1.35, + "learning_rate": 6.456220016845498e-08, + "loss": 0.679, + "step": 37034 + }, + { + "epoch": 1.36, + "learning_rate": 6.455853810378292e-08, + "loss": 0.6962, + "step": 37035 + }, + { + "epoch": 1.36, + "learning_rate": 6.455487603911084e-08, + "loss": 0.6967, + "step": 37036 + }, + { + "epoch": 1.36, + "learning_rate": 6.455121397443878e-08, + "loss": 0.6878, + "step": 37037 + }, + { + "epoch": 1.36, + "learning_rate": 6.454755190976672e-08, + "loss": 0.7045, + "step": 37038 + }, + { + "epoch": 1.36, + "learning_rate": 6.454388984509465e-08, + "loss": 0.6747, + "step": 37039 + }, + { + "epoch": 1.36, + "learning_rate": 6.454022778042259e-08, + "loss": 0.6825, + "step": 37040 + }, + { + "epoch": 1.36, + "learning_rate": 6.453656571575053e-08, + "loss": 0.6816, + "step": 37041 + }, + { + "epoch": 1.36, + "learning_rate": 6.453290365107847e-08, + "loss": 0.6848, + "step": 37042 + }, + { + "epoch": 1.36, + "learning_rate": 6.452924158640641e-08, + "loss": 0.6921, + "step": 37043 + }, + { + "epoch": 1.36, + "learning_rate": 6.452557952173435e-08, + "loss": 0.6972, + "step": 37044 + }, + { + "epoch": 1.36, + "learning_rate": 6.452191745706229e-08, + "loss": 0.6953, + "step": 37045 + }, + { + "epoch": 1.36, + "learning_rate": 6.451825539239023e-08, + "loss": 0.7059, + "step": 37046 + }, + { + "epoch": 1.36, + "learning_rate": 6.451459332771817e-08, + "loss": 0.6946, + "step": 37047 + }, + { + "epoch": 1.36, + "learning_rate": 6.45109312630461e-08, + "loss": 0.6936, + "step": 37048 + }, + { + "epoch": 1.36, + "learning_rate": 6.450726919837404e-08, + "loss": 0.6952, + "step": 37049 + }, + { + "epoch": 1.36, + "learning_rate": 6.450360713370198e-08, + "loss": 0.6835, + "step": 37050 + }, + { + "epoch": 1.36, + "learning_rate": 6.449994506902992e-08, + "loss": 0.6849, + "step": 37051 + }, + { + "epoch": 1.36, + "learning_rate": 6.449628300435786e-08, + "loss": 0.6962, + "step": 37052 + }, + { + "epoch": 1.36, + "learning_rate": 6.44926209396858e-08, + "loss": 0.6867, + "step": 37053 + }, + { + "epoch": 1.36, + "learning_rate": 6.448895887501374e-08, + "loss": 0.6823, + "step": 37054 + }, + { + "epoch": 1.36, + "learning_rate": 6.448529681034168e-08, + "loss": 0.6912, + "step": 37055 + }, + { + "epoch": 1.36, + "learning_rate": 6.448163474566962e-08, + "loss": 0.7051, + "step": 37056 + }, + { + "epoch": 1.36, + "learning_rate": 6.447797268099754e-08, + "loss": 0.6929, + "step": 37057 + }, + { + "epoch": 1.36, + "learning_rate": 6.447431061632548e-08, + "loss": 0.6936, + "step": 37058 + }, + { + "epoch": 1.36, + "learning_rate": 6.447064855165342e-08, + "loss": 0.6865, + "step": 37059 + }, + { + "epoch": 1.36, + "learning_rate": 6.446698648698135e-08, + "loss": 0.6936, + "step": 37060 + }, + { + "epoch": 1.36, + "learning_rate": 6.446332442230929e-08, + "loss": 0.6967, + "step": 37061 + }, + { + "epoch": 1.36, + "learning_rate": 6.445966235763723e-08, + "loss": 0.693, + "step": 37062 + }, + { + "epoch": 1.36, + "learning_rate": 6.445600029296517e-08, + "loss": 0.6853, + "step": 37063 + }, + { + "epoch": 1.36, + "learning_rate": 6.445233822829311e-08, + "loss": 0.7015, + "step": 37064 + }, + { + "epoch": 1.36, + "learning_rate": 6.444867616362105e-08, + "loss": 0.6865, + "step": 37065 + }, + { + "epoch": 1.36, + "learning_rate": 6.444501409894898e-08, + "loss": 0.6796, + "step": 37066 + }, + { + "epoch": 1.36, + "learning_rate": 6.444135203427692e-08, + "loss": 0.6966, + "step": 37067 + }, + { + "epoch": 1.36, + "learning_rate": 6.443768996960486e-08, + "loss": 0.697, + "step": 37068 + }, + { + "epoch": 1.36, + "learning_rate": 6.44340279049328e-08, + "loss": 0.6989, + "step": 37069 + }, + { + "epoch": 1.36, + "learning_rate": 6.443036584026074e-08, + "loss": 0.6952, + "step": 37070 + }, + { + "epoch": 1.36, + "learning_rate": 6.442670377558868e-08, + "loss": 0.6836, + "step": 37071 + }, + { + "epoch": 1.36, + "learning_rate": 6.442304171091662e-08, + "loss": 0.6748, + "step": 37072 + }, + { + "epoch": 1.36, + "learning_rate": 6.441937964624456e-08, + "loss": 0.6888, + "step": 37073 + }, + { + "epoch": 1.36, + "learning_rate": 6.44157175815725e-08, + "loss": 0.6944, + "step": 37074 + }, + { + "epoch": 1.36, + "learning_rate": 6.441205551690042e-08, + "loss": 0.6907, + "step": 37075 + }, + { + "epoch": 1.36, + "learning_rate": 6.440839345222836e-08, + "loss": 0.7026, + "step": 37076 + }, + { + "epoch": 1.36, + "learning_rate": 6.44047313875563e-08, + "loss": 0.7001, + "step": 37077 + }, + { + "epoch": 1.36, + "learning_rate": 6.440106932288424e-08, + "loss": 0.7042, + "step": 37078 + }, + { + "epoch": 1.36, + "learning_rate": 6.439740725821218e-08, + "loss": 0.6876, + "step": 37079 + }, + { + "epoch": 1.36, + "learning_rate": 6.439374519354012e-08, + "loss": 0.6906, + "step": 37080 + }, + { + "epoch": 1.36, + "learning_rate": 6.439008312886806e-08, + "loss": 0.6914, + "step": 37081 + }, + { + "epoch": 1.36, + "learning_rate": 6.438642106419599e-08, + "loss": 0.6839, + "step": 37082 + }, + { + "epoch": 1.36, + "learning_rate": 6.438275899952393e-08, + "loss": 0.6886, + "step": 37083 + }, + { + "epoch": 1.36, + "learning_rate": 6.437909693485187e-08, + "loss": 0.6907, + "step": 37084 + }, + { + "epoch": 1.36, + "learning_rate": 6.43754348701798e-08, + "loss": 0.6872, + "step": 37085 + }, + { + "epoch": 1.36, + "learning_rate": 6.437177280550774e-08, + "loss": 0.694, + "step": 37086 + }, + { + "epoch": 1.36, + "learning_rate": 6.436811074083568e-08, + "loss": 0.7096, + "step": 37087 + }, + { + "epoch": 1.36, + "learning_rate": 6.436444867616361e-08, + "loss": 0.6833, + "step": 37088 + }, + { + "epoch": 1.36, + "learning_rate": 6.436078661149155e-08, + "loss": 0.6936, + "step": 37089 + }, + { + "epoch": 1.36, + "learning_rate": 6.43571245468195e-08, + "loss": 0.7004, + "step": 37090 + }, + { + "epoch": 1.36, + "learning_rate": 6.435346248214743e-08, + "loss": 0.6916, + "step": 37091 + }, + { + "epoch": 1.36, + "learning_rate": 6.434980041747537e-08, + "loss": 0.6889, + "step": 37092 + }, + { + "epoch": 1.36, + "learning_rate": 6.43461383528033e-08, + "loss": 0.6955, + "step": 37093 + }, + { + "epoch": 1.36, + "learning_rate": 6.434247628813124e-08, + "loss": 0.6879, + "step": 37094 + }, + { + "epoch": 1.36, + "learning_rate": 6.433881422345918e-08, + "loss": 0.6743, + "step": 37095 + }, + { + "epoch": 1.36, + "learning_rate": 6.433515215878712e-08, + "loss": 0.6826, + "step": 37096 + }, + { + "epoch": 1.36, + "learning_rate": 6.433149009411506e-08, + "loss": 0.6997, + "step": 37097 + }, + { + "epoch": 1.36, + "learning_rate": 6.4327828029443e-08, + "loss": 0.6998, + "step": 37098 + }, + { + "epoch": 1.36, + "learning_rate": 6.432416596477094e-08, + "loss": 0.6911, + "step": 37099 + }, + { + "epoch": 1.36, + "learning_rate": 6.432050390009888e-08, + "loss": 0.6791, + "step": 37100 + }, + { + "epoch": 1.36, + "learning_rate": 6.431684183542682e-08, + "loss": 0.6824, + "step": 37101 + }, + { + "epoch": 1.36, + "learning_rate": 6.431317977075475e-08, + "loss": 0.6941, + "step": 37102 + }, + { + "epoch": 1.36, + "learning_rate": 6.430951770608269e-08, + "loss": 0.6977, + "step": 37103 + }, + { + "epoch": 1.36, + "learning_rate": 6.430585564141063e-08, + "loss": 0.6935, + "step": 37104 + }, + { + "epoch": 1.36, + "learning_rate": 6.430219357673857e-08, + "loss": 0.6954, + "step": 37105 + }, + { + "epoch": 1.36, + "learning_rate": 6.42985315120665e-08, + "loss": 0.6811, + "step": 37106 + }, + { + "epoch": 1.36, + "learning_rate": 6.429486944739443e-08, + "loss": 0.6929, + "step": 37107 + }, + { + "epoch": 1.36, + "learning_rate": 6.429120738272237e-08, + "loss": 0.7058, + "step": 37108 + }, + { + "epoch": 1.36, + "learning_rate": 6.428754531805031e-08, + "loss": 0.7124, + "step": 37109 + }, + { + "epoch": 1.36, + "learning_rate": 6.428388325337825e-08, + "loss": 0.7039, + "step": 37110 + }, + { + "epoch": 1.36, + "learning_rate": 6.428022118870618e-08, + "loss": 0.6928, + "step": 37111 + }, + { + "epoch": 1.36, + "learning_rate": 6.427655912403412e-08, + "loss": 0.6739, + "step": 37112 + }, + { + "epoch": 1.36, + "learning_rate": 6.427289705936206e-08, + "loss": 0.6914, + "step": 37113 + }, + { + "epoch": 1.36, + "learning_rate": 6.426923499469e-08, + "loss": 0.6993, + "step": 37114 + }, + { + "epoch": 1.36, + "learning_rate": 6.426557293001794e-08, + "loss": 0.6898, + "step": 37115 + }, + { + "epoch": 1.36, + "learning_rate": 6.426191086534588e-08, + "loss": 0.6856, + "step": 37116 + }, + { + "epoch": 1.36, + "learning_rate": 6.425824880067382e-08, + "loss": 0.6931, + "step": 37117 + }, + { + "epoch": 1.36, + "learning_rate": 6.425458673600176e-08, + "loss": 0.6808, + "step": 37118 + }, + { + "epoch": 1.36, + "learning_rate": 6.42509246713297e-08, + "loss": 0.6962, + "step": 37119 + }, + { + "epoch": 1.36, + "learning_rate": 6.424726260665763e-08, + "loss": 0.6947, + "step": 37120 + }, + { + "epoch": 1.36, + "learning_rate": 6.424360054198557e-08, + "loss": 0.7076, + "step": 37121 + }, + { + "epoch": 1.36, + "learning_rate": 6.42399384773135e-08, + "loss": 0.6956, + "step": 37122 + }, + { + "epoch": 1.36, + "learning_rate": 6.423627641264145e-08, + "loss": 0.6783, + "step": 37123 + }, + { + "epoch": 1.36, + "learning_rate": 6.423261434796939e-08, + "loss": 0.6949, + "step": 37124 + }, + { + "epoch": 1.36, + "learning_rate": 6.422895228329733e-08, + "loss": 0.6819, + "step": 37125 + }, + { + "epoch": 1.36, + "learning_rate": 6.422529021862527e-08, + "loss": 0.6981, + "step": 37126 + }, + { + "epoch": 1.36, + "learning_rate": 6.42216281539532e-08, + "loss": 0.6891, + "step": 37127 + }, + { + "epoch": 1.36, + "learning_rate": 6.421796608928113e-08, + "loss": 0.6921, + "step": 37128 + }, + { + "epoch": 1.36, + "learning_rate": 6.421430402460907e-08, + "loss": 0.7101, + "step": 37129 + }, + { + "epoch": 1.36, + "learning_rate": 6.421064195993701e-08, + "loss": 0.6893, + "step": 37130 + }, + { + "epoch": 1.36, + "learning_rate": 6.420697989526494e-08, + "loss": 0.6937, + "step": 37131 + }, + { + "epoch": 1.36, + "learning_rate": 6.420331783059288e-08, + "loss": 0.6921, + "step": 37132 + }, + { + "epoch": 1.36, + "learning_rate": 6.419965576592082e-08, + "loss": 0.6897, + "step": 37133 + }, + { + "epoch": 1.36, + "learning_rate": 6.419599370124876e-08, + "loss": 0.6975, + "step": 37134 + }, + { + "epoch": 1.36, + "learning_rate": 6.41923316365767e-08, + "loss": 0.6852, + "step": 37135 + }, + { + "epoch": 1.36, + "learning_rate": 6.418866957190464e-08, + "loss": 0.6834, + "step": 37136 + }, + { + "epoch": 1.36, + "learning_rate": 6.418500750723258e-08, + "loss": 0.6982, + "step": 37137 + }, + { + "epoch": 1.36, + "learning_rate": 6.41813454425605e-08, + "loss": 0.6874, + "step": 37138 + }, + { + "epoch": 1.36, + "learning_rate": 6.417768337788845e-08, + "loss": 0.6965, + "step": 37139 + }, + { + "epoch": 1.36, + "learning_rate": 6.417402131321639e-08, + "loss": 0.693, + "step": 37140 + }, + { + "epoch": 1.36, + "learning_rate": 6.417035924854433e-08, + "loss": 0.6924, + "step": 37141 + }, + { + "epoch": 1.36, + "learning_rate": 6.416669718387227e-08, + "loss": 0.6879, + "step": 37142 + }, + { + "epoch": 1.36, + "learning_rate": 6.41630351192002e-08, + "loss": 0.6949, + "step": 37143 + }, + { + "epoch": 1.36, + "learning_rate": 6.415937305452815e-08, + "loss": 0.6771, + "step": 37144 + }, + { + "epoch": 1.36, + "learning_rate": 6.415571098985609e-08, + "loss": 0.6933, + "step": 37145 + }, + { + "epoch": 1.36, + "learning_rate": 6.415204892518402e-08, + "loss": 0.68, + "step": 37146 + }, + { + "epoch": 1.36, + "learning_rate": 6.414838686051195e-08, + "loss": 0.7106, + "step": 37147 + }, + { + "epoch": 1.36, + "learning_rate": 6.414472479583989e-08, + "loss": 0.6848, + "step": 37148 + }, + { + "epoch": 1.36, + "learning_rate": 6.414106273116783e-08, + "loss": 0.7055, + "step": 37149 + }, + { + "epoch": 1.36, + "learning_rate": 6.413740066649577e-08, + "loss": 0.7034, + "step": 37150 + }, + { + "epoch": 1.36, + "learning_rate": 6.413373860182371e-08, + "loss": 0.6748, + "step": 37151 + }, + { + "epoch": 1.36, + "learning_rate": 6.413007653715165e-08, + "loss": 0.683, + "step": 37152 + }, + { + "epoch": 1.36, + "learning_rate": 6.412641447247958e-08, + "loss": 0.6851, + "step": 37153 + }, + { + "epoch": 1.36, + "learning_rate": 6.412275240780752e-08, + "loss": 0.6827, + "step": 37154 + }, + { + "epoch": 1.36, + "learning_rate": 6.411909034313546e-08, + "loss": 0.6953, + "step": 37155 + }, + { + "epoch": 1.36, + "learning_rate": 6.411542827846338e-08, + "loss": 0.7003, + "step": 37156 + }, + { + "epoch": 1.36, + "learning_rate": 6.411176621379132e-08, + "loss": 0.6881, + "step": 37157 + }, + { + "epoch": 1.36, + "learning_rate": 6.410810414911926e-08, + "loss": 0.6969, + "step": 37158 + }, + { + "epoch": 1.36, + "learning_rate": 6.41044420844472e-08, + "loss": 0.6886, + "step": 37159 + }, + { + "epoch": 1.36, + "learning_rate": 6.410078001977514e-08, + "loss": 0.6988, + "step": 37160 + }, + { + "epoch": 1.36, + "learning_rate": 6.409711795510308e-08, + "loss": 0.6857, + "step": 37161 + }, + { + "epoch": 1.36, + "learning_rate": 6.409345589043102e-08, + "loss": 0.6901, + "step": 37162 + }, + { + "epoch": 1.36, + "learning_rate": 6.408979382575896e-08, + "loss": 0.6773, + "step": 37163 + }, + { + "epoch": 1.36, + "learning_rate": 6.40861317610869e-08, + "loss": 0.6964, + "step": 37164 + }, + { + "epoch": 1.36, + "learning_rate": 6.408246969641483e-08, + "loss": 0.6949, + "step": 37165 + }, + { + "epoch": 1.36, + "learning_rate": 6.407880763174277e-08, + "loss": 0.6811, + "step": 37166 + }, + { + "epoch": 1.36, + "learning_rate": 6.407514556707071e-08, + "loss": 0.697, + "step": 37167 + }, + { + "epoch": 1.36, + "learning_rate": 6.407148350239865e-08, + "loss": 0.6933, + "step": 37168 + }, + { + "epoch": 1.36, + "learning_rate": 6.406782143772659e-08, + "loss": 0.6679, + "step": 37169 + }, + { + "epoch": 1.36, + "learning_rate": 6.406415937305453e-08, + "loss": 0.6993, + "step": 37170 + }, + { + "epoch": 1.36, + "learning_rate": 6.406049730838247e-08, + "loss": 0.6899, + "step": 37171 + }, + { + "epoch": 1.36, + "learning_rate": 6.405683524371041e-08, + "loss": 0.6912, + "step": 37172 + }, + { + "epoch": 1.36, + "learning_rate": 6.405317317903834e-08, + "loss": 0.6927, + "step": 37173 + }, + { + "epoch": 1.36, + "learning_rate": 6.404951111436628e-08, + "loss": 0.6955, + "step": 37174 + }, + { + "epoch": 1.36, + "learning_rate": 6.404584904969422e-08, + "loss": 0.6938, + "step": 37175 + }, + { + "epoch": 1.36, + "learning_rate": 6.404218698502216e-08, + "loss": 0.6937, + "step": 37176 + }, + { + "epoch": 1.36, + "learning_rate": 6.403852492035008e-08, + "loss": 0.6962, + "step": 37177 + }, + { + "epoch": 1.36, + "learning_rate": 6.403486285567802e-08, + "loss": 0.6729, + "step": 37178 + }, + { + "epoch": 1.36, + "learning_rate": 6.403120079100596e-08, + "loss": 0.6971, + "step": 37179 + }, + { + "epoch": 1.36, + "learning_rate": 6.40275387263339e-08, + "loss": 0.6942, + "step": 37180 + }, + { + "epoch": 1.36, + "learning_rate": 6.402387666166184e-08, + "loss": 0.6872, + "step": 37181 + }, + { + "epoch": 1.36, + "learning_rate": 6.402021459698977e-08, + "loss": 0.6902, + "step": 37182 + }, + { + "epoch": 1.36, + "learning_rate": 6.401655253231771e-08, + "loss": 0.6908, + "step": 37183 + }, + { + "epoch": 1.36, + "learning_rate": 6.401289046764565e-08, + "loss": 0.6987, + "step": 37184 + }, + { + "epoch": 1.36, + "learning_rate": 6.400922840297359e-08, + "loss": 0.6878, + "step": 37185 + }, + { + "epoch": 1.36, + "learning_rate": 6.400556633830153e-08, + "loss": 0.6934, + "step": 37186 + }, + { + "epoch": 1.36, + "learning_rate": 6.400190427362947e-08, + "loss": 0.676, + "step": 37187 + }, + { + "epoch": 1.36, + "learning_rate": 6.399824220895741e-08, + "loss": 0.7009, + "step": 37188 + }, + { + "epoch": 1.36, + "learning_rate": 6.399458014428535e-08, + "loss": 0.6867, + "step": 37189 + }, + { + "epoch": 1.36, + "learning_rate": 6.399091807961329e-08, + "loss": 0.6844, + "step": 37190 + }, + { + "epoch": 1.36, + "learning_rate": 6.398725601494122e-08, + "loss": 0.6945, + "step": 37191 + }, + { + "epoch": 1.36, + "learning_rate": 6.398359395026916e-08, + "loss": 0.6912, + "step": 37192 + }, + { + "epoch": 1.36, + "learning_rate": 6.39799318855971e-08, + "loss": 0.6892, + "step": 37193 + }, + { + "epoch": 1.36, + "learning_rate": 6.397626982092504e-08, + "loss": 0.6962, + "step": 37194 + }, + { + "epoch": 1.36, + "learning_rate": 6.397260775625298e-08, + "loss": 0.7023, + "step": 37195 + }, + { + "epoch": 1.36, + "learning_rate": 6.396894569158092e-08, + "loss": 0.6922, + "step": 37196 + }, + { + "epoch": 1.36, + "learning_rate": 6.396528362690886e-08, + "loss": 0.6842, + "step": 37197 + }, + { + "epoch": 1.36, + "learning_rate": 6.39616215622368e-08, + "loss": 0.6808, + "step": 37198 + }, + { + "epoch": 1.36, + "learning_rate": 6.395795949756472e-08, + "loss": 0.6954, + "step": 37199 + }, + { + "epoch": 1.36, + "learning_rate": 6.395429743289266e-08, + "loss": 0.6883, + "step": 37200 + }, + { + "epoch": 1.36, + "learning_rate": 6.39506353682206e-08, + "loss": 0.6777, + "step": 37201 + }, + { + "epoch": 1.36, + "learning_rate": 6.394697330354853e-08, + "loss": 0.6865, + "step": 37202 + }, + { + "epoch": 1.36, + "learning_rate": 6.394331123887647e-08, + "loss": 0.6961, + "step": 37203 + }, + { + "epoch": 1.36, + "learning_rate": 6.393964917420441e-08, + "loss": 0.6973, + "step": 37204 + }, + { + "epoch": 1.36, + "learning_rate": 6.393598710953235e-08, + "loss": 0.7076, + "step": 37205 + }, + { + "epoch": 1.36, + "learning_rate": 6.393232504486029e-08, + "loss": 0.6871, + "step": 37206 + }, + { + "epoch": 1.36, + "learning_rate": 6.392866298018823e-08, + "loss": 0.6993, + "step": 37207 + }, + { + "epoch": 1.36, + "learning_rate": 6.392500091551617e-08, + "loss": 0.6926, + "step": 37208 + }, + { + "epoch": 1.36, + "learning_rate": 6.39213388508441e-08, + "loss": 0.6821, + "step": 37209 + }, + { + "epoch": 1.36, + "learning_rate": 6.391767678617203e-08, + "loss": 0.6989, + "step": 37210 + }, + { + "epoch": 1.36, + "learning_rate": 6.391401472149997e-08, + "loss": 0.7026, + "step": 37211 + }, + { + "epoch": 1.36, + "learning_rate": 6.391035265682791e-08, + "loss": 0.7043, + "step": 37212 + }, + { + "epoch": 1.36, + "learning_rate": 6.390669059215585e-08, + "loss": 0.6828, + "step": 37213 + }, + { + "epoch": 1.36, + "learning_rate": 6.39030285274838e-08, + "loss": 0.7086, + "step": 37214 + }, + { + "epoch": 1.36, + "learning_rate": 6.389936646281173e-08, + "loss": 0.6936, + "step": 37215 + }, + { + "epoch": 1.36, + "learning_rate": 6.389570439813967e-08, + "loss": 0.7018, + "step": 37216 + }, + { + "epoch": 1.36, + "learning_rate": 6.389204233346761e-08, + "loss": 0.7025, + "step": 37217 + }, + { + "epoch": 1.36, + "learning_rate": 6.388838026879554e-08, + "loss": 0.687, + "step": 37218 + }, + { + "epoch": 1.36, + "learning_rate": 6.388471820412348e-08, + "loss": 0.6993, + "step": 37219 + }, + { + "epoch": 1.36, + "learning_rate": 6.388105613945142e-08, + "loss": 0.6871, + "step": 37220 + }, + { + "epoch": 1.36, + "learning_rate": 6.387739407477936e-08, + "loss": 0.6954, + "step": 37221 + }, + { + "epoch": 1.36, + "learning_rate": 6.38737320101073e-08, + "loss": 0.6886, + "step": 37222 + }, + { + "epoch": 1.36, + "learning_rate": 6.387006994543524e-08, + "loss": 0.71, + "step": 37223 + }, + { + "epoch": 1.36, + "learning_rate": 6.386640788076317e-08, + "loss": 0.6879, + "step": 37224 + }, + { + "epoch": 1.36, + "learning_rate": 6.386274581609111e-08, + "loss": 0.7082, + "step": 37225 + }, + { + "epoch": 1.36, + "learning_rate": 6.385908375141905e-08, + "loss": 0.689, + "step": 37226 + }, + { + "epoch": 1.36, + "learning_rate": 6.385542168674697e-08, + "loss": 0.6995, + "step": 37227 + }, + { + "epoch": 1.36, + "learning_rate": 6.385175962207491e-08, + "loss": 0.7, + "step": 37228 + }, + { + "epoch": 1.36, + "learning_rate": 6.384809755740285e-08, + "loss": 0.7091, + "step": 37229 + }, + { + "epoch": 1.36, + "learning_rate": 6.38444354927308e-08, + "loss": 0.7101, + "step": 37230 + }, + { + "epoch": 1.36, + "learning_rate": 6.384077342805873e-08, + "loss": 0.6944, + "step": 37231 + }, + { + "epoch": 1.36, + "learning_rate": 6.383711136338667e-08, + "loss": 0.6914, + "step": 37232 + }, + { + "epoch": 1.36, + "learning_rate": 6.383344929871461e-08, + "loss": 0.675, + "step": 37233 + }, + { + "epoch": 1.36, + "learning_rate": 6.382978723404255e-08, + "loss": 0.6903, + "step": 37234 + }, + { + "epoch": 1.36, + "learning_rate": 6.382612516937049e-08, + "loss": 0.6954, + "step": 37235 + }, + { + "epoch": 1.36, + "learning_rate": 6.382246310469842e-08, + "loss": 0.6885, + "step": 37236 + }, + { + "epoch": 1.36, + "learning_rate": 6.381880104002636e-08, + "loss": 0.7014, + "step": 37237 + }, + { + "epoch": 1.36, + "learning_rate": 6.38151389753543e-08, + "loss": 0.6963, + "step": 37238 + }, + { + "epoch": 1.36, + "learning_rate": 6.381147691068224e-08, + "loss": 0.6967, + "step": 37239 + }, + { + "epoch": 1.36, + "learning_rate": 6.380781484601018e-08, + "loss": 0.6813, + "step": 37240 + }, + { + "epoch": 1.36, + "learning_rate": 6.380415278133812e-08, + "loss": 0.6889, + "step": 37241 + }, + { + "epoch": 1.36, + "learning_rate": 6.380049071666606e-08, + "loss": 0.6966, + "step": 37242 + }, + { + "epoch": 1.36, + "learning_rate": 6.3796828651994e-08, + "loss": 0.6868, + "step": 37243 + }, + { + "epoch": 1.36, + "learning_rate": 6.379316658732194e-08, + "loss": 0.6888, + "step": 37244 + }, + { + "epoch": 1.36, + "learning_rate": 6.378950452264987e-08, + "loss": 0.6949, + "step": 37245 + }, + { + "epoch": 1.36, + "learning_rate": 6.37858424579778e-08, + "loss": 0.6958, + "step": 37246 + }, + { + "epoch": 1.36, + "learning_rate": 6.378218039330575e-08, + "loss": 0.689, + "step": 37247 + }, + { + "epoch": 1.36, + "learning_rate": 6.377851832863367e-08, + "loss": 0.7007, + "step": 37248 + }, + { + "epoch": 1.36, + "learning_rate": 6.377485626396161e-08, + "loss": 0.7007, + "step": 37249 + }, + { + "epoch": 1.36, + "learning_rate": 6.377119419928955e-08, + "loss": 0.6875, + "step": 37250 + }, + { + "epoch": 1.36, + "learning_rate": 6.376753213461749e-08, + "loss": 0.7007, + "step": 37251 + }, + { + "epoch": 1.36, + "learning_rate": 6.376387006994543e-08, + "loss": 0.712, + "step": 37252 + }, + { + "epoch": 1.36, + "learning_rate": 6.376020800527337e-08, + "loss": 0.6814, + "step": 37253 + }, + { + "epoch": 1.36, + "learning_rate": 6.37565459406013e-08, + "loss": 0.7026, + "step": 37254 + }, + { + "epoch": 1.36, + "learning_rate": 6.375288387592924e-08, + "loss": 0.6972, + "step": 37255 + }, + { + "epoch": 1.36, + "learning_rate": 6.374922181125718e-08, + "loss": 0.6992, + "step": 37256 + }, + { + "epoch": 1.36, + "learning_rate": 6.374555974658512e-08, + "loss": 0.6952, + "step": 37257 + }, + { + "epoch": 1.36, + "learning_rate": 6.374189768191306e-08, + "loss": 0.6968, + "step": 37258 + }, + { + "epoch": 1.36, + "learning_rate": 6.3738235617241e-08, + "loss": 0.6909, + "step": 37259 + }, + { + "epoch": 1.36, + "learning_rate": 6.373457355256894e-08, + "loss": 0.6862, + "step": 37260 + }, + { + "epoch": 1.36, + "learning_rate": 6.373091148789688e-08, + "loss": 0.6911, + "step": 37261 + }, + { + "epoch": 1.36, + "learning_rate": 6.372724942322482e-08, + "loss": 0.6823, + "step": 37262 + }, + { + "epoch": 1.36, + "learning_rate": 6.372358735855275e-08, + "loss": 0.6947, + "step": 37263 + }, + { + "epoch": 1.36, + "learning_rate": 6.371992529388069e-08, + "loss": 0.6948, + "step": 37264 + }, + { + "epoch": 1.36, + "learning_rate": 6.371626322920863e-08, + "loss": 0.6819, + "step": 37265 + }, + { + "epoch": 1.36, + "learning_rate": 6.371260116453656e-08, + "loss": 0.7034, + "step": 37266 + }, + { + "epoch": 1.36, + "learning_rate": 6.37089390998645e-08, + "loss": 0.671, + "step": 37267 + }, + { + "epoch": 1.36, + "learning_rate": 6.370527703519244e-08, + "loss": 0.6977, + "step": 37268 + }, + { + "epoch": 1.36, + "learning_rate": 6.370161497052038e-08, + "loss": 0.7034, + "step": 37269 + }, + { + "epoch": 1.36, + "learning_rate": 6.369795290584831e-08, + "loss": 0.6976, + "step": 37270 + }, + { + "epoch": 1.36, + "learning_rate": 6.369429084117625e-08, + "loss": 0.7057, + "step": 37271 + }, + { + "epoch": 1.36, + "learning_rate": 6.369062877650419e-08, + "loss": 0.7013, + "step": 37272 + }, + { + "epoch": 1.36, + "learning_rate": 6.368696671183212e-08, + "loss": 0.6855, + "step": 37273 + }, + { + "epoch": 1.36, + "learning_rate": 6.368330464716006e-08, + "loss": 0.6882, + "step": 37274 + }, + { + "epoch": 1.36, + "learning_rate": 6.3679642582488e-08, + "loss": 0.6818, + "step": 37275 + }, + { + "epoch": 1.36, + "learning_rate": 6.367598051781594e-08, + "loss": 0.7083, + "step": 37276 + }, + { + "epoch": 1.36, + "learning_rate": 6.367231845314388e-08, + "loss": 0.6741, + "step": 37277 + }, + { + "epoch": 1.36, + "learning_rate": 6.366865638847182e-08, + "loss": 0.6869, + "step": 37278 + }, + { + "epoch": 1.36, + "learning_rate": 6.366499432379976e-08, + "loss": 0.7003, + "step": 37279 + }, + { + "epoch": 1.36, + "learning_rate": 6.36613322591277e-08, + "loss": 0.6784, + "step": 37280 + }, + { + "epoch": 1.36, + "learning_rate": 6.365767019445562e-08, + "loss": 0.6806, + "step": 37281 + }, + { + "epoch": 1.36, + "learning_rate": 6.365400812978356e-08, + "loss": 0.6899, + "step": 37282 + }, + { + "epoch": 1.36, + "learning_rate": 6.36503460651115e-08, + "loss": 0.6954, + "step": 37283 + }, + { + "epoch": 1.36, + "learning_rate": 6.364668400043944e-08, + "loss": 0.6965, + "step": 37284 + }, + { + "epoch": 1.36, + "learning_rate": 6.364302193576738e-08, + "loss": 0.6928, + "step": 37285 + }, + { + "epoch": 1.36, + "learning_rate": 6.363935987109532e-08, + "loss": 0.6997, + "step": 37286 + }, + { + "epoch": 1.36, + "learning_rate": 6.363569780642326e-08, + "loss": 0.6754, + "step": 37287 + }, + { + "epoch": 1.36, + "learning_rate": 6.36320357417512e-08, + "loss": 0.696, + "step": 37288 + }, + { + "epoch": 1.36, + "learning_rate": 6.362837367707914e-08, + "loss": 0.6785, + "step": 37289 + }, + { + "epoch": 1.36, + "learning_rate": 6.362471161240707e-08, + "loss": 0.6959, + "step": 37290 + }, + { + "epoch": 1.36, + "learning_rate": 6.362104954773501e-08, + "loss": 0.6954, + "step": 37291 + }, + { + "epoch": 1.36, + "learning_rate": 6.361738748306295e-08, + "loss": 0.6971, + "step": 37292 + }, + { + "epoch": 1.36, + "learning_rate": 6.361372541839089e-08, + "loss": 0.6976, + "step": 37293 + }, + { + "epoch": 1.36, + "learning_rate": 6.361006335371883e-08, + "loss": 0.6961, + "step": 37294 + }, + { + "epoch": 1.36, + "learning_rate": 6.360640128904676e-08, + "loss": 0.7025, + "step": 37295 + }, + { + "epoch": 1.36, + "learning_rate": 6.36027392243747e-08, + "loss": 0.6872, + "step": 37296 + }, + { + "epoch": 1.36, + "learning_rate": 6.359907715970264e-08, + "loss": 0.6932, + "step": 37297 + }, + { + "epoch": 1.36, + "learning_rate": 6.359541509503058e-08, + "loss": 0.6945, + "step": 37298 + }, + { + "epoch": 1.36, + "learning_rate": 6.35917530303585e-08, + "loss": 0.6866, + "step": 37299 + }, + { + "epoch": 1.36, + "learning_rate": 6.358809096568644e-08, + "loss": 0.6892, + "step": 37300 + }, + { + "epoch": 1.36, + "learning_rate": 6.358442890101438e-08, + "loss": 0.6939, + "step": 37301 + }, + { + "epoch": 1.36, + "learning_rate": 6.358076683634232e-08, + "loss": 0.6851, + "step": 37302 + }, + { + "epoch": 1.36, + "learning_rate": 6.357710477167026e-08, + "loss": 0.6892, + "step": 37303 + }, + { + "epoch": 1.36, + "learning_rate": 6.35734427069982e-08, + "loss": 0.7128, + "step": 37304 + }, + { + "epoch": 1.36, + "learning_rate": 6.356978064232614e-08, + "loss": 0.6914, + "step": 37305 + }, + { + "epoch": 1.36, + "learning_rate": 6.356611857765408e-08, + "loss": 0.6805, + "step": 37306 + }, + { + "epoch": 1.36, + "learning_rate": 6.356245651298202e-08, + "loss": 0.6778, + "step": 37307 + }, + { + "epoch": 1.36, + "learning_rate": 6.355879444830995e-08, + "loss": 0.6815, + "step": 37308 + }, + { + "epoch": 1.37, + "learning_rate": 6.355513238363789e-08, + "loss": 0.6864, + "step": 37309 + }, + { + "epoch": 1.37, + "learning_rate": 6.355147031896583e-08, + "loss": 0.6751, + "step": 37310 + }, + { + "epoch": 1.37, + "learning_rate": 6.354780825429377e-08, + "loss": 0.7052, + "step": 37311 + }, + { + "epoch": 1.37, + "learning_rate": 6.354414618962171e-08, + "loss": 0.6897, + "step": 37312 + }, + { + "epoch": 1.37, + "learning_rate": 6.354048412494965e-08, + "loss": 0.7023, + "step": 37313 + }, + { + "epoch": 1.37, + "learning_rate": 6.353682206027759e-08, + "loss": 0.6913, + "step": 37314 + }, + { + "epoch": 1.37, + "learning_rate": 6.353315999560553e-08, + "loss": 0.6943, + "step": 37315 + }, + { + "epoch": 1.37, + "learning_rate": 6.352949793093347e-08, + "loss": 0.7039, + "step": 37316 + }, + { + "epoch": 1.37, + "learning_rate": 6.35258358662614e-08, + "loss": 0.6863, + "step": 37317 + }, + { + "epoch": 1.37, + "learning_rate": 6.352217380158934e-08, + "loss": 0.6865, + "step": 37318 + }, + { + "epoch": 1.37, + "learning_rate": 6.351851173691726e-08, + "loss": 0.7144, + "step": 37319 + }, + { + "epoch": 1.37, + "learning_rate": 6.35148496722452e-08, + "loss": 0.7003, + "step": 37320 + }, + { + "epoch": 1.37, + "learning_rate": 6.351118760757314e-08, + "loss": 0.695, + "step": 37321 + }, + { + "epoch": 1.37, + "learning_rate": 6.350752554290108e-08, + "loss": 0.6851, + "step": 37322 + }, + { + "epoch": 1.37, + "learning_rate": 6.350386347822902e-08, + "loss": 0.6979, + "step": 37323 + }, + { + "epoch": 1.37, + "learning_rate": 6.350020141355696e-08, + "loss": 0.6973, + "step": 37324 + }, + { + "epoch": 1.37, + "learning_rate": 6.34965393488849e-08, + "loss": 0.6956, + "step": 37325 + }, + { + "epoch": 1.37, + "learning_rate": 6.349287728421283e-08, + "loss": 0.6946, + "step": 37326 + }, + { + "epoch": 1.37, + "learning_rate": 6.348921521954077e-08, + "loss": 0.7025, + "step": 37327 + }, + { + "epoch": 1.37, + "learning_rate": 6.348555315486871e-08, + "loss": 0.7, + "step": 37328 + }, + { + "epoch": 1.37, + "learning_rate": 6.348189109019665e-08, + "loss": 0.6796, + "step": 37329 + }, + { + "epoch": 1.37, + "learning_rate": 6.347822902552459e-08, + "loss": 0.6973, + "step": 37330 + }, + { + "epoch": 1.37, + "learning_rate": 6.347456696085253e-08, + "loss": 0.7025, + "step": 37331 + }, + { + "epoch": 1.37, + "learning_rate": 6.347090489618047e-08, + "loss": 0.6947, + "step": 37332 + }, + { + "epoch": 1.37, + "learning_rate": 6.346724283150841e-08, + "loss": 0.6956, + "step": 37333 + }, + { + "epoch": 1.37, + "learning_rate": 6.346358076683635e-08, + "loss": 0.6991, + "step": 37334 + }, + { + "epoch": 1.37, + "learning_rate": 6.345991870216427e-08, + "loss": 0.6938, + "step": 37335 + }, + { + "epoch": 1.37, + "learning_rate": 6.345625663749221e-08, + "loss": 0.6888, + "step": 37336 + }, + { + "epoch": 1.37, + "learning_rate": 6.345259457282015e-08, + "loss": 0.6996, + "step": 37337 + }, + { + "epoch": 1.37, + "learning_rate": 6.34489325081481e-08, + "loss": 0.6974, + "step": 37338 + }, + { + "epoch": 1.37, + "learning_rate": 6.344527044347603e-08, + "loss": 0.6999, + "step": 37339 + }, + { + "epoch": 1.37, + "learning_rate": 6.344160837880397e-08, + "loss": 0.6891, + "step": 37340 + }, + { + "epoch": 1.37, + "learning_rate": 6.34379463141319e-08, + "loss": 0.6997, + "step": 37341 + }, + { + "epoch": 1.37, + "learning_rate": 6.343428424945984e-08, + "loss": 0.7011, + "step": 37342 + }, + { + "epoch": 1.37, + "learning_rate": 6.343062218478778e-08, + "loss": 0.7027, + "step": 37343 + }, + { + "epoch": 1.37, + "learning_rate": 6.342696012011571e-08, + "loss": 0.684, + "step": 37344 + }, + { + "epoch": 1.37, + "learning_rate": 6.342329805544365e-08, + "loss": 0.6954, + "step": 37345 + }, + { + "epoch": 1.37, + "learning_rate": 6.341963599077159e-08, + "loss": 0.7019, + "step": 37346 + }, + { + "epoch": 1.37, + "learning_rate": 6.341597392609953e-08, + "loss": 0.6953, + "step": 37347 + }, + { + "epoch": 1.37, + "learning_rate": 6.341231186142747e-08, + "loss": 0.6872, + "step": 37348 + }, + { + "epoch": 1.37, + "learning_rate": 6.340864979675541e-08, + "loss": 0.6987, + "step": 37349 + }, + { + "epoch": 1.37, + "learning_rate": 6.340498773208335e-08, + "loss": 0.6837, + "step": 37350 + }, + { + "epoch": 1.37, + "learning_rate": 6.340132566741129e-08, + "loss": 0.6829, + "step": 37351 + }, + { + "epoch": 1.37, + "learning_rate": 6.339766360273923e-08, + "loss": 0.6861, + "step": 37352 + }, + { + "epoch": 1.37, + "learning_rate": 6.339400153806715e-08, + "loss": 0.6954, + "step": 37353 + }, + { + "epoch": 1.37, + "learning_rate": 6.339033947339509e-08, + "loss": 0.6764, + "step": 37354 + }, + { + "epoch": 1.37, + "learning_rate": 6.338667740872303e-08, + "loss": 0.7034, + "step": 37355 + }, + { + "epoch": 1.37, + "learning_rate": 6.338301534405097e-08, + "loss": 0.6758, + "step": 37356 + }, + { + "epoch": 1.37, + "learning_rate": 6.337935327937891e-08, + "loss": 0.711, + "step": 37357 + }, + { + "epoch": 1.37, + "learning_rate": 6.337569121470685e-08, + "loss": 0.6926, + "step": 37358 + }, + { + "epoch": 1.37, + "learning_rate": 6.337202915003479e-08, + "loss": 0.7073, + "step": 37359 + }, + { + "epoch": 1.37, + "learning_rate": 6.336836708536273e-08, + "loss": 0.696, + "step": 37360 + }, + { + "epoch": 1.37, + "learning_rate": 6.336470502069067e-08, + "loss": 0.6759, + "step": 37361 + }, + { + "epoch": 1.37, + "learning_rate": 6.33610429560186e-08, + "loss": 0.6721, + "step": 37362 + }, + { + "epoch": 1.37, + "learning_rate": 6.335738089134654e-08, + "loss": 0.6949, + "step": 37363 + }, + { + "epoch": 1.37, + "learning_rate": 6.335371882667448e-08, + "loss": 0.6956, + "step": 37364 + }, + { + "epoch": 1.37, + "learning_rate": 6.335005676200242e-08, + "loss": 0.6997, + "step": 37365 + }, + { + "epoch": 1.37, + "learning_rate": 6.334639469733035e-08, + "loss": 0.6834, + "step": 37366 + }, + { + "epoch": 1.37, + "learning_rate": 6.334273263265829e-08, + "loss": 0.6977, + "step": 37367 + }, + { + "epoch": 1.37, + "learning_rate": 6.333907056798623e-08, + "loss": 0.6927, + "step": 37368 + }, + { + "epoch": 1.37, + "learning_rate": 6.333540850331417e-08, + "loss": 0.6918, + "step": 37369 + }, + { + "epoch": 1.37, + "learning_rate": 6.33317464386421e-08, + "loss": 0.7107, + "step": 37370 + }, + { + "epoch": 1.37, + "learning_rate": 6.332808437397003e-08, + "loss": 0.6909, + "step": 37371 + }, + { + "epoch": 1.37, + "learning_rate": 6.332442230929797e-08, + "loss": 0.7051, + "step": 37372 + }, + { + "epoch": 1.37, + "learning_rate": 6.332076024462591e-08, + "loss": 0.6948, + "step": 37373 + }, + { + "epoch": 1.37, + "learning_rate": 6.331709817995385e-08, + "loss": 0.6918, + "step": 37374 + }, + { + "epoch": 1.37, + "learning_rate": 6.331343611528179e-08, + "loss": 0.7024, + "step": 37375 + }, + { + "epoch": 1.37, + "learning_rate": 6.330977405060973e-08, + "loss": 0.7117, + "step": 37376 + }, + { + "epoch": 1.37, + "learning_rate": 6.330611198593767e-08, + "loss": 0.686, + "step": 37377 + }, + { + "epoch": 1.37, + "learning_rate": 6.330244992126561e-08, + "loss": 0.7073, + "step": 37378 + }, + { + "epoch": 1.37, + "learning_rate": 6.329878785659355e-08, + "loss": 0.7001, + "step": 37379 + }, + { + "epoch": 1.37, + "learning_rate": 6.329512579192148e-08, + "loss": 0.686, + "step": 37380 + }, + { + "epoch": 1.37, + "learning_rate": 6.329146372724942e-08, + "loss": 0.6888, + "step": 37381 + }, + { + "epoch": 1.37, + "learning_rate": 6.328780166257736e-08, + "loss": 0.6884, + "step": 37382 + }, + { + "epoch": 1.37, + "learning_rate": 6.32841395979053e-08, + "loss": 0.6846, + "step": 37383 + }, + { + "epoch": 1.37, + "learning_rate": 6.328047753323324e-08, + "loss": 0.6825, + "step": 37384 + }, + { + "epoch": 1.37, + "learning_rate": 6.327681546856118e-08, + "loss": 0.7071, + "step": 37385 + }, + { + "epoch": 1.37, + "learning_rate": 6.327315340388912e-08, + "loss": 0.6983, + "step": 37386 + }, + { + "epoch": 1.37, + "learning_rate": 6.326949133921706e-08, + "loss": 0.6878, + "step": 37387 + }, + { + "epoch": 1.37, + "learning_rate": 6.326582927454498e-08, + "loss": 0.6946, + "step": 37388 + }, + { + "epoch": 1.37, + "learning_rate": 6.326216720987292e-08, + "loss": 0.6649, + "step": 37389 + }, + { + "epoch": 1.37, + "learning_rate": 6.325850514520085e-08, + "loss": 0.7008, + "step": 37390 + }, + { + "epoch": 1.37, + "learning_rate": 6.325484308052879e-08, + "loss": 0.6968, + "step": 37391 + }, + { + "epoch": 1.37, + "learning_rate": 6.325118101585673e-08, + "loss": 0.6966, + "step": 37392 + }, + { + "epoch": 1.37, + "learning_rate": 6.324751895118467e-08, + "loss": 0.7028, + "step": 37393 + }, + { + "epoch": 1.37, + "learning_rate": 6.324385688651261e-08, + "loss": 0.6995, + "step": 37394 + }, + { + "epoch": 1.37, + "learning_rate": 6.324019482184055e-08, + "loss": 0.7006, + "step": 37395 + }, + { + "epoch": 1.37, + "learning_rate": 6.323653275716849e-08, + "loss": 0.7013, + "step": 37396 + }, + { + "epoch": 1.37, + "learning_rate": 6.323287069249643e-08, + "loss": 0.7041, + "step": 37397 + }, + { + "epoch": 1.37, + "learning_rate": 6.322920862782436e-08, + "loss": 0.6987, + "step": 37398 + }, + { + "epoch": 1.37, + "learning_rate": 6.32255465631523e-08, + "loss": 0.7106, + "step": 37399 + }, + { + "epoch": 1.37, + "learning_rate": 6.322188449848024e-08, + "loss": 0.7023, + "step": 37400 + }, + { + "epoch": 1.37, + "learning_rate": 6.321822243380818e-08, + "loss": 0.7001, + "step": 37401 + }, + { + "epoch": 1.37, + "learning_rate": 6.321456036913612e-08, + "loss": 0.6808, + "step": 37402 + }, + { + "epoch": 1.37, + "learning_rate": 6.321089830446406e-08, + "loss": 0.6951, + "step": 37403 + }, + { + "epoch": 1.37, + "learning_rate": 6.3207236239792e-08, + "loss": 0.6907, + "step": 37404 + }, + { + "epoch": 1.37, + "learning_rate": 6.320357417511994e-08, + "loss": 0.6909, + "step": 37405 + }, + { + "epoch": 1.37, + "learning_rate": 6.319991211044788e-08, + "loss": 0.6857, + "step": 37406 + }, + { + "epoch": 1.37, + "learning_rate": 6.31962500457758e-08, + "loss": 0.691, + "step": 37407 + }, + { + "epoch": 1.37, + "learning_rate": 6.319258798110374e-08, + "loss": 0.6842, + "step": 37408 + }, + { + "epoch": 1.37, + "learning_rate": 6.318892591643168e-08, + "loss": 0.7025, + "step": 37409 + }, + { + "epoch": 1.37, + "learning_rate": 6.318526385175962e-08, + "loss": 0.6936, + "step": 37410 + }, + { + "epoch": 1.37, + "learning_rate": 6.318160178708756e-08, + "loss": 0.6804, + "step": 37411 + }, + { + "epoch": 1.37, + "learning_rate": 6.317793972241549e-08, + "loss": 0.6885, + "step": 37412 + }, + { + "epoch": 1.37, + "learning_rate": 6.317427765774343e-08, + "loss": 0.6939, + "step": 37413 + }, + { + "epoch": 1.37, + "learning_rate": 6.317061559307137e-08, + "loss": 0.6956, + "step": 37414 + }, + { + "epoch": 1.37, + "learning_rate": 6.316695352839931e-08, + "loss": 0.6896, + "step": 37415 + }, + { + "epoch": 1.37, + "learning_rate": 6.316329146372724e-08, + "loss": 0.7001, + "step": 37416 + }, + { + "epoch": 1.37, + "learning_rate": 6.315962939905518e-08, + "loss": 0.6972, + "step": 37417 + }, + { + "epoch": 1.37, + "learning_rate": 6.315596733438312e-08, + "loss": 0.6867, + "step": 37418 + }, + { + "epoch": 1.37, + "learning_rate": 6.315230526971106e-08, + "loss": 0.6896, + "step": 37419 + }, + { + "epoch": 1.37, + "learning_rate": 6.3148643205039e-08, + "loss": 0.7027, + "step": 37420 + }, + { + "epoch": 1.37, + "learning_rate": 6.314498114036694e-08, + "loss": 0.6935, + "step": 37421 + }, + { + "epoch": 1.37, + "learning_rate": 6.314131907569488e-08, + "loss": 0.7124, + "step": 37422 + }, + { + "epoch": 1.37, + "learning_rate": 6.313765701102282e-08, + "loss": 0.6884, + "step": 37423 + }, + { + "epoch": 1.37, + "learning_rate": 6.313399494635076e-08, + "loss": 0.6981, + "step": 37424 + }, + { + "epoch": 1.37, + "learning_rate": 6.313033288167868e-08, + "loss": 0.6832, + "step": 37425 + }, + { + "epoch": 1.37, + "learning_rate": 6.312667081700662e-08, + "loss": 0.7001, + "step": 37426 + }, + { + "epoch": 1.37, + "learning_rate": 6.312300875233456e-08, + "loss": 0.695, + "step": 37427 + }, + { + "epoch": 1.37, + "learning_rate": 6.31193466876625e-08, + "loss": 0.667, + "step": 37428 + }, + { + "epoch": 1.37, + "learning_rate": 6.311568462299044e-08, + "loss": 0.6945, + "step": 37429 + }, + { + "epoch": 1.37, + "learning_rate": 6.311202255831838e-08, + "loss": 0.6977, + "step": 37430 + }, + { + "epoch": 1.37, + "learning_rate": 6.310836049364632e-08, + "loss": 0.6963, + "step": 37431 + }, + { + "epoch": 1.37, + "learning_rate": 6.310469842897426e-08, + "loss": 0.6788, + "step": 37432 + }, + { + "epoch": 1.37, + "learning_rate": 6.31010363643022e-08, + "loss": 0.6923, + "step": 37433 + }, + { + "epoch": 1.37, + "learning_rate": 6.309737429963013e-08, + "loss": 0.6883, + "step": 37434 + }, + { + "epoch": 1.37, + "learning_rate": 6.309371223495807e-08, + "loss": 0.6958, + "step": 37435 + }, + { + "epoch": 1.37, + "learning_rate": 6.309005017028601e-08, + "loss": 0.6919, + "step": 37436 + }, + { + "epoch": 1.37, + "learning_rate": 6.308638810561394e-08, + "loss": 0.6862, + "step": 37437 + }, + { + "epoch": 1.37, + "learning_rate": 6.308272604094188e-08, + "loss": 0.7002, + "step": 37438 + }, + { + "epoch": 1.37, + "learning_rate": 6.307906397626982e-08, + "loss": 0.6909, + "step": 37439 + }, + { + "epoch": 1.37, + "learning_rate": 6.307540191159776e-08, + "loss": 0.6827, + "step": 37440 + }, + { + "epoch": 1.37, + "learning_rate": 6.30717398469257e-08, + "loss": 0.6889, + "step": 37441 + }, + { + "epoch": 1.37, + "learning_rate": 6.306807778225364e-08, + "loss": 0.7011, + "step": 37442 + }, + { + "epoch": 1.37, + "learning_rate": 6.306441571758156e-08, + "loss": 0.6888, + "step": 37443 + }, + { + "epoch": 1.37, + "learning_rate": 6.30607536529095e-08, + "loss": 0.6931, + "step": 37444 + }, + { + "epoch": 1.37, + "learning_rate": 6.305709158823744e-08, + "loss": 0.6887, + "step": 37445 + }, + { + "epoch": 1.37, + "learning_rate": 6.305342952356538e-08, + "loss": 0.7005, + "step": 37446 + }, + { + "epoch": 1.37, + "learning_rate": 6.304976745889332e-08, + "loss": 0.7105, + "step": 37447 + }, + { + "epoch": 1.37, + "learning_rate": 6.304610539422126e-08, + "loss": 0.7035, + "step": 37448 + }, + { + "epoch": 1.37, + "learning_rate": 6.30424433295492e-08, + "loss": 0.7078, + "step": 37449 + }, + { + "epoch": 1.37, + "learning_rate": 6.303878126487714e-08, + "loss": 0.696, + "step": 37450 + }, + { + "epoch": 1.37, + "learning_rate": 6.303511920020508e-08, + "loss": 0.7067, + "step": 37451 + }, + { + "epoch": 1.37, + "learning_rate": 6.303145713553301e-08, + "loss": 0.7081, + "step": 37452 + }, + { + "epoch": 1.37, + "learning_rate": 6.302779507086095e-08, + "loss": 0.6859, + "step": 37453 + }, + { + "epoch": 1.37, + "learning_rate": 6.302413300618889e-08, + "loss": 0.6946, + "step": 37454 + }, + { + "epoch": 1.37, + "learning_rate": 6.302047094151683e-08, + "loss": 0.6929, + "step": 37455 + }, + { + "epoch": 1.37, + "learning_rate": 6.301680887684477e-08, + "loss": 0.6894, + "step": 37456 + }, + { + "epoch": 1.37, + "learning_rate": 6.301314681217271e-08, + "loss": 0.6858, + "step": 37457 + }, + { + "epoch": 1.37, + "learning_rate": 6.300948474750063e-08, + "loss": 0.7051, + "step": 37458 + }, + { + "epoch": 1.37, + "learning_rate": 6.300582268282857e-08, + "loss": 0.703, + "step": 37459 + }, + { + "epoch": 1.37, + "learning_rate": 6.300216061815651e-08, + "loss": 0.6911, + "step": 37460 + }, + { + "epoch": 1.37, + "learning_rate": 6.299849855348444e-08, + "loss": 0.6814, + "step": 37461 + }, + { + "epoch": 1.37, + "learning_rate": 6.299483648881238e-08, + "loss": 0.7097, + "step": 37462 + }, + { + "epoch": 1.37, + "learning_rate": 6.299117442414032e-08, + "loss": 0.6912, + "step": 37463 + }, + { + "epoch": 1.37, + "learning_rate": 6.298751235946826e-08, + "loss": 0.6786, + "step": 37464 + }, + { + "epoch": 1.37, + "learning_rate": 6.29838502947962e-08, + "loss": 0.6948, + "step": 37465 + }, + { + "epoch": 1.37, + "learning_rate": 6.298018823012414e-08, + "loss": 0.7068, + "step": 37466 + }, + { + "epoch": 1.37, + "learning_rate": 6.297652616545208e-08, + "loss": 0.6797, + "step": 37467 + }, + { + "epoch": 1.37, + "learning_rate": 6.297286410078002e-08, + "loss": 0.6838, + "step": 37468 + }, + { + "epoch": 1.37, + "learning_rate": 6.296920203610796e-08, + "loss": 0.6904, + "step": 37469 + }, + { + "epoch": 1.37, + "learning_rate": 6.296553997143589e-08, + "loss": 0.6976, + "step": 37470 + }, + { + "epoch": 1.37, + "learning_rate": 6.296187790676383e-08, + "loss": 0.6904, + "step": 37471 + }, + { + "epoch": 1.37, + "learning_rate": 6.295821584209177e-08, + "loss": 0.6966, + "step": 37472 + }, + { + "epoch": 1.37, + "learning_rate": 6.295455377741971e-08, + "loss": 0.6905, + "step": 37473 + }, + { + "epoch": 1.37, + "learning_rate": 6.295089171274765e-08, + "loss": 0.6955, + "step": 37474 + }, + { + "epoch": 1.37, + "learning_rate": 6.294722964807559e-08, + "loss": 0.6868, + "step": 37475 + }, + { + "epoch": 1.37, + "learning_rate": 6.294356758340353e-08, + "loss": 0.6973, + "step": 37476 + }, + { + "epoch": 1.37, + "learning_rate": 6.293990551873147e-08, + "loss": 0.6862, + "step": 37477 + }, + { + "epoch": 1.37, + "learning_rate": 6.29362434540594e-08, + "loss": 0.6982, + "step": 37478 + }, + { + "epoch": 1.37, + "learning_rate": 6.293258138938733e-08, + "loss": 0.6898, + "step": 37479 + }, + { + "epoch": 1.37, + "learning_rate": 6.292891932471527e-08, + "loss": 0.6841, + "step": 37480 + }, + { + "epoch": 1.37, + "learning_rate": 6.292525726004321e-08, + "loss": 0.6956, + "step": 37481 + }, + { + "epoch": 1.37, + "learning_rate": 6.292159519537115e-08, + "loss": 0.6958, + "step": 37482 + }, + { + "epoch": 1.37, + "learning_rate": 6.291793313069908e-08, + "loss": 0.6966, + "step": 37483 + }, + { + "epoch": 1.37, + "learning_rate": 6.291427106602702e-08, + "loss": 0.6856, + "step": 37484 + }, + { + "epoch": 1.37, + "learning_rate": 6.291060900135496e-08, + "loss": 0.6894, + "step": 37485 + }, + { + "epoch": 1.37, + "learning_rate": 6.29069469366829e-08, + "loss": 0.694, + "step": 37486 + }, + { + "epoch": 1.37, + "learning_rate": 6.290328487201084e-08, + "loss": 0.6847, + "step": 37487 + }, + { + "epoch": 1.37, + "learning_rate": 6.289962280733877e-08, + "loss": 0.6933, + "step": 37488 + }, + { + "epoch": 1.37, + "learning_rate": 6.28959607426667e-08, + "loss": 0.676, + "step": 37489 + }, + { + "epoch": 1.37, + "learning_rate": 6.289229867799465e-08, + "loss": 0.6932, + "step": 37490 + }, + { + "epoch": 1.37, + "learning_rate": 6.288863661332259e-08, + "loss": 0.6922, + "step": 37491 + }, + { + "epoch": 1.37, + "learning_rate": 6.288497454865053e-08, + "loss": 0.6907, + "step": 37492 + }, + { + "epoch": 1.37, + "learning_rate": 6.288131248397847e-08, + "loss": 0.6871, + "step": 37493 + }, + { + "epoch": 1.37, + "learning_rate": 6.28776504193064e-08, + "loss": 0.7132, + "step": 37494 + }, + { + "epoch": 1.37, + "learning_rate": 6.287398835463435e-08, + "loss": 0.7103, + "step": 37495 + }, + { + "epoch": 1.37, + "learning_rate": 6.287032628996229e-08, + "loss": 0.688, + "step": 37496 + }, + { + "epoch": 1.37, + "learning_rate": 6.286666422529021e-08, + "loss": 0.7091, + "step": 37497 + }, + { + "epoch": 1.37, + "learning_rate": 6.286300216061815e-08, + "loss": 0.6935, + "step": 37498 + }, + { + "epoch": 1.37, + "learning_rate": 6.285934009594609e-08, + "loss": 0.6934, + "step": 37499 + }, + { + "epoch": 1.37, + "learning_rate": 6.285567803127403e-08, + "loss": 0.7021, + "step": 37500 + }, + { + "epoch": 1.37, + "learning_rate": 6.285201596660197e-08, + "loss": 0.6953, + "step": 37501 + }, + { + "epoch": 1.37, + "learning_rate": 6.284835390192991e-08, + "loss": 0.6917, + "step": 37502 + }, + { + "epoch": 1.37, + "learning_rate": 6.284469183725785e-08, + "loss": 0.6896, + "step": 37503 + }, + { + "epoch": 1.37, + "learning_rate": 6.284102977258579e-08, + "loss": 0.6896, + "step": 37504 + }, + { + "epoch": 1.37, + "learning_rate": 6.283736770791372e-08, + "loss": 0.7002, + "step": 37505 + }, + { + "epoch": 1.37, + "learning_rate": 6.283370564324166e-08, + "loss": 0.7045, + "step": 37506 + }, + { + "epoch": 1.37, + "learning_rate": 6.28300435785696e-08, + "loss": 0.6926, + "step": 37507 + }, + { + "epoch": 1.37, + "learning_rate": 6.282638151389752e-08, + "loss": 0.6883, + "step": 37508 + }, + { + "epoch": 1.37, + "learning_rate": 6.282271944922546e-08, + "loss": 0.7019, + "step": 37509 + }, + { + "epoch": 1.37, + "learning_rate": 6.28190573845534e-08, + "loss": 0.6946, + "step": 37510 + }, + { + "epoch": 1.37, + "learning_rate": 6.281539531988134e-08, + "loss": 0.6955, + "step": 37511 + }, + { + "epoch": 1.37, + "learning_rate": 6.281173325520928e-08, + "loss": 0.6967, + "step": 37512 + }, + { + "epoch": 1.37, + "learning_rate": 6.280807119053722e-08, + "loss": 0.6949, + "step": 37513 + }, + { + "epoch": 1.37, + "learning_rate": 6.280440912586516e-08, + "loss": 0.6908, + "step": 37514 + }, + { + "epoch": 1.37, + "learning_rate": 6.280074706119309e-08, + "loss": 0.6897, + "step": 37515 + }, + { + "epoch": 1.37, + "learning_rate": 6.279708499652103e-08, + "loss": 0.7071, + "step": 37516 + }, + { + "epoch": 1.37, + "learning_rate": 6.279342293184897e-08, + "loss": 0.6926, + "step": 37517 + }, + { + "epoch": 1.37, + "learning_rate": 6.278976086717691e-08, + "loss": 0.6974, + "step": 37518 + }, + { + "epoch": 1.37, + "learning_rate": 6.278609880250485e-08, + "loss": 0.6881, + "step": 37519 + }, + { + "epoch": 1.37, + "learning_rate": 6.278243673783279e-08, + "loss": 0.6837, + "step": 37520 + }, + { + "epoch": 1.37, + "learning_rate": 6.277877467316073e-08, + "loss": 0.6794, + "step": 37521 + }, + { + "epoch": 1.37, + "learning_rate": 6.277511260848867e-08, + "loss": 0.6982, + "step": 37522 + }, + { + "epoch": 1.37, + "learning_rate": 6.277145054381661e-08, + "loss": 0.7084, + "step": 37523 + }, + { + "epoch": 1.37, + "learning_rate": 6.276778847914454e-08, + "loss": 0.6861, + "step": 37524 + }, + { + "epoch": 1.37, + "learning_rate": 6.276412641447248e-08, + "loss": 0.6973, + "step": 37525 + }, + { + "epoch": 1.37, + "learning_rate": 6.276046434980042e-08, + "loss": 0.6867, + "step": 37526 + }, + { + "epoch": 1.37, + "learning_rate": 6.275680228512836e-08, + "loss": 0.6959, + "step": 37527 + }, + { + "epoch": 1.37, + "learning_rate": 6.27531402204563e-08, + "loss": 0.6795, + "step": 37528 + }, + { + "epoch": 1.37, + "learning_rate": 6.274947815578422e-08, + "loss": 0.6914, + "step": 37529 + }, + { + "epoch": 1.37, + "learning_rate": 6.274581609111216e-08, + "loss": 0.6966, + "step": 37530 + }, + { + "epoch": 1.37, + "learning_rate": 6.27421540264401e-08, + "loss": 0.6687, + "step": 37531 + }, + { + "epoch": 1.37, + "learning_rate": 6.273849196176804e-08, + "loss": 0.7044, + "step": 37532 + }, + { + "epoch": 1.37, + "learning_rate": 6.273482989709597e-08, + "loss": 0.6793, + "step": 37533 + }, + { + "epoch": 1.37, + "learning_rate": 6.273116783242391e-08, + "loss": 0.6889, + "step": 37534 + }, + { + "epoch": 1.37, + "learning_rate": 6.272750576775185e-08, + "loss": 0.7045, + "step": 37535 + }, + { + "epoch": 1.37, + "learning_rate": 6.272384370307979e-08, + "loss": 0.6933, + "step": 37536 + }, + { + "epoch": 1.37, + "learning_rate": 6.272018163840773e-08, + "loss": 0.6873, + "step": 37537 + }, + { + "epoch": 1.37, + "learning_rate": 6.271651957373567e-08, + "loss": 0.686, + "step": 37538 + }, + { + "epoch": 1.37, + "learning_rate": 6.271285750906361e-08, + "loss": 0.6981, + "step": 37539 + }, + { + "epoch": 1.37, + "learning_rate": 6.270919544439155e-08, + "loss": 0.6922, + "step": 37540 + }, + { + "epoch": 1.37, + "learning_rate": 6.270553337971949e-08, + "loss": 0.7065, + "step": 37541 + }, + { + "epoch": 1.37, + "learning_rate": 6.270187131504742e-08, + "loss": 0.7085, + "step": 37542 + }, + { + "epoch": 1.37, + "learning_rate": 6.269820925037536e-08, + "loss": 0.7031, + "step": 37543 + }, + { + "epoch": 1.37, + "learning_rate": 6.26945471857033e-08, + "loss": 0.6885, + "step": 37544 + }, + { + "epoch": 1.37, + "learning_rate": 6.269088512103124e-08, + "loss": 0.6965, + "step": 37545 + }, + { + "epoch": 1.37, + "learning_rate": 6.268722305635918e-08, + "loss": 0.6893, + "step": 37546 + }, + { + "epoch": 1.37, + "learning_rate": 6.268356099168712e-08, + "loss": 0.6884, + "step": 37547 + }, + { + "epoch": 1.37, + "learning_rate": 6.267989892701506e-08, + "loss": 0.6896, + "step": 37548 + }, + { + "epoch": 1.37, + "learning_rate": 6.2676236862343e-08, + "loss": 0.6748, + "step": 37549 + }, + { + "epoch": 1.37, + "learning_rate": 6.267257479767092e-08, + "loss": 0.691, + "step": 37550 + }, + { + "epoch": 1.37, + "learning_rate": 6.266891273299886e-08, + "loss": 0.6909, + "step": 37551 + }, + { + "epoch": 1.37, + "learning_rate": 6.26652506683268e-08, + "loss": 0.7022, + "step": 37552 + }, + { + "epoch": 1.37, + "learning_rate": 6.266158860365474e-08, + "loss": 0.678, + "step": 37553 + }, + { + "epoch": 1.37, + "learning_rate": 6.265792653898267e-08, + "loss": 0.7007, + "step": 37554 + }, + { + "epoch": 1.37, + "learning_rate": 6.265426447431061e-08, + "loss": 0.6821, + "step": 37555 + }, + { + "epoch": 1.37, + "learning_rate": 6.265060240963855e-08, + "loss": 0.6906, + "step": 37556 + }, + { + "epoch": 1.37, + "learning_rate": 6.264694034496649e-08, + "loss": 0.695, + "step": 37557 + }, + { + "epoch": 1.37, + "learning_rate": 6.264327828029443e-08, + "loss": 0.685, + "step": 37558 + }, + { + "epoch": 1.37, + "learning_rate": 6.263961621562236e-08, + "loss": 0.6835, + "step": 37559 + }, + { + "epoch": 1.37, + "learning_rate": 6.26359541509503e-08, + "loss": 0.7053, + "step": 37560 + }, + { + "epoch": 1.37, + "learning_rate": 6.263229208627824e-08, + "loss": 0.691, + "step": 37561 + }, + { + "epoch": 1.37, + "learning_rate": 6.262863002160618e-08, + "loss": 0.709, + "step": 37562 + }, + { + "epoch": 1.37, + "learning_rate": 6.262496795693411e-08, + "loss": 0.6952, + "step": 37563 + }, + { + "epoch": 1.37, + "learning_rate": 6.262130589226205e-08, + "loss": 0.696, + "step": 37564 + }, + { + "epoch": 1.37, + "learning_rate": 6.261764382759e-08, + "loss": 0.6963, + "step": 37565 + }, + { + "epoch": 1.37, + "learning_rate": 6.261398176291793e-08, + "loss": 0.7078, + "step": 37566 + }, + { + "epoch": 1.37, + "learning_rate": 6.261031969824587e-08, + "loss": 0.6959, + "step": 37567 + }, + { + "epoch": 1.37, + "learning_rate": 6.26066576335738e-08, + "loss": 0.6949, + "step": 37568 + }, + { + "epoch": 1.37, + "learning_rate": 6.260299556890174e-08, + "loss": 0.6889, + "step": 37569 + }, + { + "epoch": 1.37, + "learning_rate": 6.259933350422968e-08, + "loss": 0.6964, + "step": 37570 + }, + { + "epoch": 1.37, + "learning_rate": 6.259567143955762e-08, + "loss": 0.6986, + "step": 37571 + }, + { + "epoch": 1.37, + "learning_rate": 6.259200937488556e-08, + "loss": 0.6815, + "step": 37572 + }, + { + "epoch": 1.37, + "learning_rate": 6.25883473102135e-08, + "loss": 0.6887, + "step": 37573 + }, + { + "epoch": 1.37, + "learning_rate": 6.258468524554144e-08, + "loss": 0.6865, + "step": 37574 + }, + { + "epoch": 1.37, + "learning_rate": 6.258102318086938e-08, + "loss": 0.6825, + "step": 37575 + }, + { + "epoch": 1.37, + "learning_rate": 6.257736111619731e-08, + "loss": 0.6808, + "step": 37576 + }, + { + "epoch": 1.37, + "learning_rate": 6.257369905152525e-08, + "loss": 0.6873, + "step": 37577 + }, + { + "epoch": 1.37, + "learning_rate": 6.257003698685317e-08, + "loss": 0.6784, + "step": 37578 + }, + { + "epoch": 1.37, + "learning_rate": 6.256637492218111e-08, + "loss": 0.6777, + "step": 37579 + }, + { + "epoch": 1.37, + "learning_rate": 6.256271285750905e-08, + "loss": 0.695, + "step": 37580 + }, + { + "epoch": 1.37, + "learning_rate": 6.2559050792837e-08, + "loss": 0.6929, + "step": 37581 + }, + { + "epoch": 1.38, + "learning_rate": 6.255538872816493e-08, + "loss": 0.6811, + "step": 37582 + }, + { + "epoch": 1.38, + "learning_rate": 6.255172666349287e-08, + "loss": 0.6867, + "step": 37583 + }, + { + "epoch": 1.38, + "learning_rate": 6.254806459882081e-08, + "loss": 0.6851, + "step": 37584 + }, + { + "epoch": 1.38, + "learning_rate": 6.254440253414875e-08, + "loss": 0.7096, + "step": 37585 + }, + { + "epoch": 1.38, + "learning_rate": 6.254074046947668e-08, + "loss": 0.6934, + "step": 37586 + }, + { + "epoch": 1.38, + "learning_rate": 6.253707840480462e-08, + "loss": 0.7027, + "step": 37587 + }, + { + "epoch": 1.38, + "learning_rate": 6.253341634013256e-08, + "loss": 0.6865, + "step": 37588 + }, + { + "epoch": 1.38, + "learning_rate": 6.25297542754605e-08, + "loss": 0.6909, + "step": 37589 + }, + { + "epoch": 1.38, + "learning_rate": 6.252609221078844e-08, + "loss": 0.6985, + "step": 37590 + }, + { + "epoch": 1.38, + "learning_rate": 6.252243014611638e-08, + "loss": 0.6931, + "step": 37591 + }, + { + "epoch": 1.38, + "learning_rate": 6.251876808144432e-08, + "loss": 0.6912, + "step": 37592 + }, + { + "epoch": 1.38, + "learning_rate": 6.251510601677226e-08, + "loss": 0.7006, + "step": 37593 + }, + { + "epoch": 1.38, + "learning_rate": 6.25114439521002e-08, + "loss": 0.7016, + "step": 37594 + }, + { + "epoch": 1.38, + "learning_rate": 6.250778188742813e-08, + "loss": 0.6925, + "step": 37595 + }, + { + "epoch": 1.38, + "learning_rate": 6.250411982275607e-08, + "loss": 0.7248, + "step": 37596 + }, + { + "epoch": 1.38, + "learning_rate": 6.2500457758084e-08, + "loss": 0.6892, + "step": 37597 + }, + { + "epoch": 1.38, + "learning_rate": 6.249679569341195e-08, + "loss": 0.6987, + "step": 37598 + }, + { + "epoch": 1.38, + "learning_rate": 6.249313362873989e-08, + "loss": 0.6979, + "step": 37599 + }, + { + "epoch": 1.38, + "learning_rate": 6.248947156406781e-08, + "loss": 0.7003, + "step": 37600 + }, + { + "epoch": 1.38, + "learning_rate": 6.248580949939575e-08, + "loss": 0.6946, + "step": 37601 + }, + { + "epoch": 1.38, + "learning_rate": 6.248214743472369e-08, + "loss": 0.6989, + "step": 37602 + }, + { + "epoch": 1.38, + "learning_rate": 6.247848537005163e-08, + "loss": 0.6983, + "step": 37603 + }, + { + "epoch": 1.38, + "learning_rate": 6.247482330537956e-08, + "loss": 0.6857, + "step": 37604 + }, + { + "epoch": 1.38, + "learning_rate": 6.24711612407075e-08, + "loss": 0.6915, + "step": 37605 + }, + { + "epoch": 1.38, + "learning_rate": 6.246749917603544e-08, + "loss": 0.6904, + "step": 37606 + }, + { + "epoch": 1.38, + "learning_rate": 6.246383711136338e-08, + "loss": 0.7003, + "step": 37607 + }, + { + "epoch": 1.38, + "learning_rate": 6.246017504669132e-08, + "loss": 0.6966, + "step": 37608 + }, + { + "epoch": 1.38, + "learning_rate": 6.245651298201926e-08, + "loss": 0.6904, + "step": 37609 + }, + { + "epoch": 1.38, + "learning_rate": 6.24528509173472e-08, + "loss": 0.7024, + "step": 37610 + }, + { + "epoch": 1.38, + "learning_rate": 6.244918885267514e-08, + "loss": 0.6915, + "step": 37611 + }, + { + "epoch": 1.38, + "learning_rate": 6.244552678800308e-08, + "loss": 0.6841, + "step": 37612 + }, + { + "epoch": 1.38, + "learning_rate": 6.2441864723331e-08, + "loss": 0.7017, + "step": 37613 + }, + { + "epoch": 1.38, + "learning_rate": 6.243820265865895e-08, + "loss": 0.6884, + "step": 37614 + }, + { + "epoch": 1.38, + "learning_rate": 6.243454059398689e-08, + "loss": 0.6996, + "step": 37615 + }, + { + "epoch": 1.38, + "learning_rate": 6.243087852931483e-08, + "loss": 0.677, + "step": 37616 + }, + { + "epoch": 1.38, + "learning_rate": 6.242721646464277e-08, + "loss": 0.6907, + "step": 37617 + }, + { + "epoch": 1.38, + "learning_rate": 6.24235543999707e-08, + "loss": 0.6862, + "step": 37618 + }, + { + "epoch": 1.38, + "learning_rate": 6.241989233529865e-08, + "loss": 0.6938, + "step": 37619 + }, + { + "epoch": 1.38, + "learning_rate": 6.241623027062658e-08, + "loss": 0.6816, + "step": 37620 + }, + { + "epoch": 1.38, + "learning_rate": 6.241256820595452e-08, + "loss": 0.6843, + "step": 37621 + }, + { + "epoch": 1.38, + "learning_rate": 6.240890614128245e-08, + "loss": 0.6946, + "step": 37622 + }, + { + "epoch": 1.38, + "learning_rate": 6.240524407661039e-08, + "loss": 0.6945, + "step": 37623 + }, + { + "epoch": 1.38, + "learning_rate": 6.240158201193833e-08, + "loss": 0.6775, + "step": 37624 + }, + { + "epoch": 1.38, + "learning_rate": 6.239791994726626e-08, + "loss": 0.7066, + "step": 37625 + }, + { + "epoch": 1.38, + "learning_rate": 6.23942578825942e-08, + "loss": 0.693, + "step": 37626 + }, + { + "epoch": 1.38, + "learning_rate": 6.239059581792214e-08, + "loss": 0.6818, + "step": 37627 + }, + { + "epoch": 1.38, + "learning_rate": 6.238693375325008e-08, + "loss": 0.7057, + "step": 37628 + }, + { + "epoch": 1.38, + "learning_rate": 6.238327168857802e-08, + "loss": 0.6965, + "step": 37629 + }, + { + "epoch": 1.38, + "learning_rate": 6.237960962390596e-08, + "loss": 0.6692, + "step": 37630 + }, + { + "epoch": 1.38, + "learning_rate": 6.237594755923388e-08, + "loss": 0.688, + "step": 37631 + }, + { + "epoch": 1.38, + "learning_rate": 6.237228549456182e-08, + "loss": 0.691, + "step": 37632 + }, + { + "epoch": 1.38, + "learning_rate": 6.236862342988976e-08, + "loss": 0.7006, + "step": 37633 + }, + { + "epoch": 1.38, + "learning_rate": 6.23649613652177e-08, + "loss": 0.6863, + "step": 37634 + }, + { + "epoch": 1.38, + "learning_rate": 6.236129930054564e-08, + "loss": 0.688, + "step": 37635 + }, + { + "epoch": 1.38, + "learning_rate": 6.235763723587358e-08, + "loss": 0.6842, + "step": 37636 + }, + { + "epoch": 1.38, + "learning_rate": 6.235397517120152e-08, + "loss": 0.6916, + "step": 37637 + }, + { + "epoch": 1.38, + "learning_rate": 6.235031310652946e-08, + "loss": 0.6907, + "step": 37638 + }, + { + "epoch": 1.38, + "learning_rate": 6.23466510418574e-08, + "loss": 0.6998, + "step": 37639 + }, + { + "epoch": 1.38, + "learning_rate": 6.234298897718533e-08, + "loss": 0.6903, + "step": 37640 + }, + { + "epoch": 1.38, + "learning_rate": 6.233932691251327e-08, + "loss": 0.6887, + "step": 37641 + }, + { + "epoch": 1.38, + "learning_rate": 6.233566484784121e-08, + "loss": 0.6938, + "step": 37642 + }, + { + "epoch": 1.38, + "learning_rate": 6.233200278316915e-08, + "loss": 0.6921, + "step": 37643 + }, + { + "epoch": 1.38, + "learning_rate": 6.232834071849709e-08, + "loss": 0.6732, + "step": 37644 + }, + { + "epoch": 1.38, + "learning_rate": 6.232467865382503e-08, + "loss": 0.6851, + "step": 37645 + }, + { + "epoch": 1.38, + "learning_rate": 6.232101658915297e-08, + "loss": 0.6905, + "step": 37646 + }, + { + "epoch": 1.38, + "learning_rate": 6.23173545244809e-08, + "loss": 0.6927, + "step": 37647 + }, + { + "epoch": 1.38, + "learning_rate": 6.231369245980884e-08, + "loss": 0.6863, + "step": 37648 + }, + { + "epoch": 1.38, + "learning_rate": 6.231003039513676e-08, + "loss": 0.6895, + "step": 37649 + }, + { + "epoch": 1.38, + "learning_rate": 6.23063683304647e-08, + "loss": 0.6988, + "step": 37650 + }, + { + "epoch": 1.38, + "learning_rate": 6.230270626579264e-08, + "loss": 0.6974, + "step": 37651 + }, + { + "epoch": 1.38, + "learning_rate": 6.229904420112058e-08, + "loss": 0.683, + "step": 37652 + }, + { + "epoch": 1.38, + "learning_rate": 6.229538213644852e-08, + "loss": 0.695, + "step": 37653 + }, + { + "epoch": 1.38, + "learning_rate": 6.229172007177646e-08, + "loss": 0.6896, + "step": 37654 + }, + { + "epoch": 1.38, + "learning_rate": 6.22880580071044e-08, + "loss": 0.6933, + "step": 37655 + }, + { + "epoch": 1.38, + "learning_rate": 6.228439594243234e-08, + "loss": 0.6934, + "step": 37656 + }, + { + "epoch": 1.38, + "learning_rate": 6.228073387776028e-08, + "loss": 0.6931, + "step": 37657 + }, + { + "epoch": 1.38, + "learning_rate": 6.227707181308821e-08, + "loss": 0.6996, + "step": 37658 + }, + { + "epoch": 1.38, + "learning_rate": 6.227340974841615e-08, + "loss": 0.6886, + "step": 37659 + }, + { + "epoch": 1.38, + "learning_rate": 6.226974768374409e-08, + "loss": 0.6881, + "step": 37660 + }, + { + "epoch": 1.38, + "learning_rate": 6.226608561907203e-08, + "loss": 0.6966, + "step": 37661 + }, + { + "epoch": 1.38, + "learning_rate": 6.226242355439997e-08, + "loss": 0.688, + "step": 37662 + }, + { + "epoch": 1.38, + "learning_rate": 6.225876148972791e-08, + "loss": 0.7028, + "step": 37663 + }, + { + "epoch": 1.38, + "learning_rate": 6.225509942505585e-08, + "loss": 0.6876, + "step": 37664 + }, + { + "epoch": 1.38, + "learning_rate": 6.225143736038379e-08, + "loss": 0.6938, + "step": 37665 + }, + { + "epoch": 1.38, + "learning_rate": 6.224777529571173e-08, + "loss": 0.6969, + "step": 37666 + }, + { + "epoch": 1.38, + "learning_rate": 6.224411323103966e-08, + "loss": 0.6895, + "step": 37667 + }, + { + "epoch": 1.38, + "learning_rate": 6.22404511663676e-08, + "loss": 0.6871, + "step": 37668 + }, + { + "epoch": 1.38, + "learning_rate": 6.223678910169554e-08, + "loss": 0.6858, + "step": 37669 + }, + { + "epoch": 1.38, + "learning_rate": 6.223312703702348e-08, + "loss": 0.6896, + "step": 37670 + }, + { + "epoch": 1.38, + "learning_rate": 6.22294649723514e-08, + "loss": 0.6942, + "step": 37671 + }, + { + "epoch": 1.38, + "learning_rate": 6.222580290767934e-08, + "loss": 0.6957, + "step": 37672 + }, + { + "epoch": 1.38, + "learning_rate": 6.222214084300728e-08, + "loss": 0.7014, + "step": 37673 + }, + { + "epoch": 1.38, + "learning_rate": 6.221847877833522e-08, + "loss": 0.6933, + "step": 37674 + }, + { + "epoch": 1.38, + "learning_rate": 6.221481671366316e-08, + "loss": 0.702, + "step": 37675 + }, + { + "epoch": 1.38, + "learning_rate": 6.221115464899109e-08, + "loss": 0.6725, + "step": 37676 + }, + { + "epoch": 1.38, + "learning_rate": 6.220749258431903e-08, + "loss": 0.6932, + "step": 37677 + }, + { + "epoch": 1.38, + "learning_rate": 6.220383051964697e-08, + "loss": 0.7104, + "step": 37678 + }, + { + "epoch": 1.38, + "learning_rate": 6.220016845497491e-08, + "loss": 0.6876, + "step": 37679 + }, + { + "epoch": 1.38, + "learning_rate": 6.219650639030285e-08, + "loss": 0.6882, + "step": 37680 + }, + { + "epoch": 1.38, + "learning_rate": 6.219284432563079e-08, + "loss": 0.7077, + "step": 37681 + }, + { + "epoch": 1.38, + "learning_rate": 6.218918226095873e-08, + "loss": 0.697, + "step": 37682 + }, + { + "epoch": 1.38, + "learning_rate": 6.218552019628667e-08, + "loss": 0.7044, + "step": 37683 + }, + { + "epoch": 1.38, + "learning_rate": 6.218185813161461e-08, + "loss": 0.6803, + "step": 37684 + }, + { + "epoch": 1.38, + "learning_rate": 6.217819606694253e-08, + "loss": 0.6955, + "step": 37685 + }, + { + "epoch": 1.38, + "learning_rate": 6.217453400227047e-08, + "loss": 0.7114, + "step": 37686 + }, + { + "epoch": 1.38, + "learning_rate": 6.217087193759841e-08, + "loss": 0.6803, + "step": 37687 + }, + { + "epoch": 1.38, + "learning_rate": 6.216720987292635e-08, + "loss": 0.691, + "step": 37688 + }, + { + "epoch": 1.38, + "learning_rate": 6.21635478082543e-08, + "loss": 0.6883, + "step": 37689 + }, + { + "epoch": 1.38, + "learning_rate": 6.215988574358223e-08, + "loss": 0.7044, + "step": 37690 + }, + { + "epoch": 1.38, + "learning_rate": 6.215622367891017e-08, + "loss": 0.6969, + "step": 37691 + }, + { + "epoch": 1.38, + "learning_rate": 6.215256161423811e-08, + "loss": 0.6991, + "step": 37692 + }, + { + "epoch": 1.38, + "learning_rate": 6.214889954956604e-08, + "loss": 0.6915, + "step": 37693 + }, + { + "epoch": 1.38, + "learning_rate": 6.214523748489398e-08, + "loss": 0.7039, + "step": 37694 + }, + { + "epoch": 1.38, + "learning_rate": 6.214157542022192e-08, + "loss": 0.682, + "step": 37695 + }, + { + "epoch": 1.38, + "learning_rate": 6.213791335554985e-08, + "loss": 0.6845, + "step": 37696 + }, + { + "epoch": 1.38, + "learning_rate": 6.213425129087779e-08, + "loss": 0.6962, + "step": 37697 + }, + { + "epoch": 1.38, + "learning_rate": 6.213058922620573e-08, + "loss": 0.6962, + "step": 37698 + }, + { + "epoch": 1.38, + "learning_rate": 6.212692716153367e-08, + "loss": 0.6879, + "step": 37699 + }, + { + "epoch": 1.38, + "learning_rate": 6.212326509686161e-08, + "loss": 0.7084, + "step": 37700 + }, + { + "epoch": 1.38, + "learning_rate": 6.211960303218955e-08, + "loss": 0.675, + "step": 37701 + }, + { + "epoch": 1.38, + "learning_rate": 6.211594096751749e-08, + "loss": 0.6745, + "step": 37702 + }, + { + "epoch": 1.38, + "learning_rate": 6.211227890284541e-08, + "loss": 0.6835, + "step": 37703 + }, + { + "epoch": 1.38, + "learning_rate": 6.210861683817335e-08, + "loss": 0.7009, + "step": 37704 + }, + { + "epoch": 1.38, + "learning_rate": 6.21049547735013e-08, + "loss": 0.6892, + "step": 37705 + }, + { + "epoch": 1.38, + "learning_rate": 6.210129270882923e-08, + "loss": 0.6955, + "step": 37706 + }, + { + "epoch": 1.38, + "learning_rate": 6.209763064415717e-08, + "loss": 0.6946, + "step": 37707 + }, + { + "epoch": 1.38, + "learning_rate": 6.209396857948511e-08, + "loss": 0.6929, + "step": 37708 + }, + { + "epoch": 1.38, + "learning_rate": 6.209030651481305e-08, + "loss": 0.6894, + "step": 37709 + }, + { + "epoch": 1.38, + "learning_rate": 6.208664445014099e-08, + "loss": 0.6954, + "step": 37710 + }, + { + "epoch": 1.38, + "learning_rate": 6.208298238546893e-08, + "loss": 0.7065, + "step": 37711 + }, + { + "epoch": 1.38, + "learning_rate": 6.207932032079686e-08, + "loss": 0.7011, + "step": 37712 + }, + { + "epoch": 1.38, + "learning_rate": 6.20756582561248e-08, + "loss": 0.6893, + "step": 37713 + }, + { + "epoch": 1.38, + "learning_rate": 6.207199619145274e-08, + "loss": 0.6964, + "step": 37714 + }, + { + "epoch": 1.38, + "learning_rate": 6.206833412678068e-08, + "loss": 0.6885, + "step": 37715 + }, + { + "epoch": 1.38, + "learning_rate": 6.206467206210862e-08, + "loss": 0.6992, + "step": 37716 + }, + { + "epoch": 1.38, + "learning_rate": 6.206100999743656e-08, + "loss": 0.6845, + "step": 37717 + }, + { + "epoch": 1.38, + "learning_rate": 6.205734793276449e-08, + "loss": 0.706, + "step": 37718 + }, + { + "epoch": 1.38, + "learning_rate": 6.205368586809243e-08, + "loss": 0.6876, + "step": 37719 + }, + { + "epoch": 1.38, + "learning_rate": 6.205002380342037e-08, + "loss": 0.6916, + "step": 37720 + }, + { + "epoch": 1.38, + "learning_rate": 6.204636173874829e-08, + "loss": 0.6862, + "step": 37721 + }, + { + "epoch": 1.38, + "learning_rate": 6.204269967407623e-08, + "loss": 0.6981, + "step": 37722 + }, + { + "epoch": 1.38, + "learning_rate": 6.203903760940417e-08, + "loss": 0.6972, + "step": 37723 + }, + { + "epoch": 1.38, + "learning_rate": 6.203537554473211e-08, + "loss": 0.6886, + "step": 37724 + }, + { + "epoch": 1.38, + "learning_rate": 6.203171348006005e-08, + "loss": 0.6946, + "step": 37725 + }, + { + "epoch": 1.38, + "learning_rate": 6.202805141538799e-08, + "loss": 0.7066, + "step": 37726 + }, + { + "epoch": 1.38, + "learning_rate": 6.202438935071593e-08, + "loss": 0.6944, + "step": 37727 + }, + { + "epoch": 1.38, + "learning_rate": 6.202072728604387e-08, + "loss": 0.7009, + "step": 37728 + }, + { + "epoch": 1.38, + "learning_rate": 6.201706522137181e-08, + "loss": 0.6979, + "step": 37729 + }, + { + "epoch": 1.38, + "learning_rate": 6.201340315669974e-08, + "loss": 0.6954, + "step": 37730 + }, + { + "epoch": 1.38, + "learning_rate": 6.200974109202768e-08, + "loss": 0.7097, + "step": 37731 + }, + { + "epoch": 1.38, + "learning_rate": 6.200607902735562e-08, + "loss": 0.6847, + "step": 37732 + }, + { + "epoch": 1.38, + "learning_rate": 6.200241696268356e-08, + "loss": 0.706, + "step": 37733 + }, + { + "epoch": 1.38, + "learning_rate": 6.19987548980115e-08, + "loss": 0.6911, + "step": 37734 + }, + { + "epoch": 1.38, + "learning_rate": 6.199509283333944e-08, + "loss": 0.6991, + "step": 37735 + }, + { + "epoch": 1.38, + "learning_rate": 6.199143076866738e-08, + "loss": 0.7101, + "step": 37736 + }, + { + "epoch": 1.38, + "learning_rate": 6.198776870399532e-08, + "loss": 0.6854, + "step": 37737 + }, + { + "epoch": 1.38, + "learning_rate": 6.198410663932326e-08, + "loss": 0.6844, + "step": 37738 + }, + { + "epoch": 1.38, + "learning_rate": 6.198044457465119e-08, + "loss": 0.6894, + "step": 37739 + }, + { + "epoch": 1.38, + "learning_rate": 6.197678250997912e-08, + "loss": 0.6904, + "step": 37740 + }, + { + "epoch": 1.38, + "learning_rate": 6.197312044530706e-08, + "loss": 0.6962, + "step": 37741 + }, + { + "epoch": 1.38, + "learning_rate": 6.196945838063499e-08, + "loss": 0.7087, + "step": 37742 + }, + { + "epoch": 1.38, + "learning_rate": 6.196579631596293e-08, + "loss": 0.6734, + "step": 37743 + }, + { + "epoch": 1.38, + "learning_rate": 6.196213425129087e-08, + "loss": 0.69, + "step": 37744 + }, + { + "epoch": 1.38, + "learning_rate": 6.195847218661881e-08, + "loss": 0.6894, + "step": 37745 + }, + { + "epoch": 1.38, + "learning_rate": 6.195481012194675e-08, + "loss": 0.6922, + "step": 37746 + }, + { + "epoch": 1.38, + "learning_rate": 6.195114805727469e-08, + "loss": 0.6987, + "step": 37747 + }, + { + "epoch": 1.38, + "learning_rate": 6.194748599260262e-08, + "loss": 0.6905, + "step": 37748 + }, + { + "epoch": 1.38, + "learning_rate": 6.194382392793056e-08, + "loss": 0.7001, + "step": 37749 + }, + { + "epoch": 1.38, + "learning_rate": 6.19401618632585e-08, + "loss": 0.6884, + "step": 37750 + }, + { + "epoch": 1.38, + "learning_rate": 6.193649979858644e-08, + "loss": 0.6922, + "step": 37751 + }, + { + "epoch": 1.38, + "learning_rate": 6.193283773391438e-08, + "loss": 0.6843, + "step": 37752 + }, + { + "epoch": 1.38, + "learning_rate": 6.192917566924232e-08, + "loss": 0.6954, + "step": 37753 + }, + { + "epoch": 1.38, + "learning_rate": 6.192551360457026e-08, + "loss": 0.6994, + "step": 37754 + }, + { + "epoch": 1.38, + "learning_rate": 6.19218515398982e-08, + "loss": 0.69, + "step": 37755 + }, + { + "epoch": 1.38, + "learning_rate": 6.191818947522614e-08, + "loss": 0.6872, + "step": 37756 + }, + { + "epoch": 1.38, + "learning_rate": 6.191452741055406e-08, + "loss": 0.709, + "step": 37757 + }, + { + "epoch": 1.38, + "learning_rate": 6.1910865345882e-08, + "loss": 0.6827, + "step": 37758 + }, + { + "epoch": 1.38, + "learning_rate": 6.190720328120994e-08, + "loss": 0.6923, + "step": 37759 + }, + { + "epoch": 1.38, + "learning_rate": 6.190354121653788e-08, + "loss": 0.7014, + "step": 37760 + }, + { + "epoch": 1.38, + "learning_rate": 6.189987915186582e-08, + "loss": 0.7038, + "step": 37761 + }, + { + "epoch": 1.38, + "learning_rate": 6.189621708719376e-08, + "loss": 0.6983, + "step": 37762 + }, + { + "epoch": 1.38, + "learning_rate": 6.18925550225217e-08, + "loss": 0.7079, + "step": 37763 + }, + { + "epoch": 1.38, + "learning_rate": 6.188889295784963e-08, + "loss": 0.7039, + "step": 37764 + }, + { + "epoch": 1.38, + "learning_rate": 6.188523089317757e-08, + "loss": 0.7003, + "step": 37765 + }, + { + "epoch": 1.38, + "learning_rate": 6.188156882850551e-08, + "loss": 0.6959, + "step": 37766 + }, + { + "epoch": 1.38, + "learning_rate": 6.187790676383344e-08, + "loss": 0.701, + "step": 37767 + }, + { + "epoch": 1.38, + "learning_rate": 6.187424469916138e-08, + "loss": 0.6927, + "step": 37768 + }, + { + "epoch": 1.38, + "learning_rate": 6.187058263448932e-08, + "loss": 0.696, + "step": 37769 + }, + { + "epoch": 1.38, + "learning_rate": 6.186692056981726e-08, + "loss": 0.6859, + "step": 37770 + }, + { + "epoch": 1.38, + "learning_rate": 6.18632585051452e-08, + "loss": 0.6994, + "step": 37771 + }, + { + "epoch": 1.38, + "learning_rate": 6.185959644047314e-08, + "loss": 0.6975, + "step": 37772 + }, + { + "epoch": 1.38, + "learning_rate": 6.185593437580108e-08, + "loss": 0.6888, + "step": 37773 + }, + { + "epoch": 1.38, + "learning_rate": 6.185227231112902e-08, + "loss": 0.6937, + "step": 37774 + }, + { + "epoch": 1.38, + "learning_rate": 6.184861024645694e-08, + "loss": 0.6848, + "step": 37775 + }, + { + "epoch": 1.38, + "learning_rate": 6.184494818178488e-08, + "loss": 0.6927, + "step": 37776 + }, + { + "epoch": 1.38, + "learning_rate": 6.184128611711282e-08, + "loss": 0.6979, + "step": 37777 + }, + { + "epoch": 1.38, + "learning_rate": 6.183762405244076e-08, + "loss": 0.6946, + "step": 37778 + }, + { + "epoch": 1.38, + "learning_rate": 6.18339619877687e-08, + "loss": 0.6924, + "step": 37779 + }, + { + "epoch": 1.38, + "learning_rate": 6.183029992309664e-08, + "loss": 0.6927, + "step": 37780 + }, + { + "epoch": 1.38, + "learning_rate": 6.182663785842458e-08, + "loss": 0.7173, + "step": 37781 + }, + { + "epoch": 1.38, + "learning_rate": 6.182297579375252e-08, + "loss": 0.7006, + "step": 37782 + }, + { + "epoch": 1.38, + "learning_rate": 6.181931372908046e-08, + "loss": 0.6905, + "step": 37783 + }, + { + "epoch": 1.38, + "learning_rate": 6.181565166440839e-08, + "loss": 0.687, + "step": 37784 + }, + { + "epoch": 1.38, + "learning_rate": 6.181198959973633e-08, + "loss": 0.701, + "step": 37785 + }, + { + "epoch": 1.38, + "learning_rate": 6.180832753506427e-08, + "loss": 0.6828, + "step": 37786 + }, + { + "epoch": 1.38, + "learning_rate": 6.180466547039221e-08, + "loss": 0.6918, + "step": 37787 + }, + { + "epoch": 1.38, + "learning_rate": 6.180100340572015e-08, + "loss": 0.6974, + "step": 37788 + }, + { + "epoch": 1.38, + "learning_rate": 6.179734134104808e-08, + "loss": 0.6949, + "step": 37789 + }, + { + "epoch": 1.38, + "learning_rate": 6.179367927637602e-08, + "loss": 0.687, + "step": 37790 + }, + { + "epoch": 1.38, + "learning_rate": 6.179001721170396e-08, + "loss": 0.683, + "step": 37791 + }, + { + "epoch": 1.38, + "learning_rate": 6.17863551470319e-08, + "loss": 0.7107, + "step": 37792 + }, + { + "epoch": 1.38, + "learning_rate": 6.178269308235982e-08, + "loss": 0.706, + "step": 37793 + }, + { + "epoch": 1.38, + "learning_rate": 6.177903101768776e-08, + "loss": 0.6986, + "step": 37794 + }, + { + "epoch": 1.38, + "learning_rate": 6.17753689530157e-08, + "loss": 0.6898, + "step": 37795 + }, + { + "epoch": 1.38, + "learning_rate": 6.177170688834364e-08, + "loss": 0.6946, + "step": 37796 + }, + { + "epoch": 1.38, + "learning_rate": 6.176804482367158e-08, + "loss": 0.6787, + "step": 37797 + }, + { + "epoch": 1.38, + "learning_rate": 6.176438275899952e-08, + "loss": 0.6848, + "step": 37798 + }, + { + "epoch": 1.38, + "learning_rate": 6.176072069432746e-08, + "loss": 0.6874, + "step": 37799 + }, + { + "epoch": 1.38, + "learning_rate": 6.17570586296554e-08, + "loss": 0.6958, + "step": 37800 + }, + { + "epoch": 1.38, + "learning_rate": 6.175339656498334e-08, + "loss": 0.693, + "step": 37801 + }, + { + "epoch": 1.38, + "learning_rate": 6.174973450031127e-08, + "loss": 0.7035, + "step": 37802 + }, + { + "epoch": 1.38, + "learning_rate": 6.174607243563921e-08, + "loss": 0.6855, + "step": 37803 + }, + { + "epoch": 1.38, + "learning_rate": 6.174241037096715e-08, + "loss": 0.693, + "step": 37804 + }, + { + "epoch": 1.38, + "learning_rate": 6.173874830629509e-08, + "loss": 0.6829, + "step": 37805 + }, + { + "epoch": 1.38, + "learning_rate": 6.173508624162303e-08, + "loss": 0.6922, + "step": 37806 + }, + { + "epoch": 1.38, + "learning_rate": 6.173142417695097e-08, + "loss": 0.689, + "step": 37807 + }, + { + "epoch": 1.38, + "learning_rate": 6.172776211227891e-08, + "loss": 0.6969, + "step": 37808 + }, + { + "epoch": 1.38, + "learning_rate": 6.172410004760685e-08, + "loss": 0.6862, + "step": 37809 + }, + { + "epoch": 1.38, + "learning_rate": 6.172043798293477e-08, + "loss": 0.7012, + "step": 37810 + }, + { + "epoch": 1.38, + "learning_rate": 6.171677591826271e-08, + "loss": 0.6975, + "step": 37811 + }, + { + "epoch": 1.38, + "learning_rate": 6.171311385359065e-08, + "loss": 0.71, + "step": 37812 + }, + { + "epoch": 1.38, + "learning_rate": 6.170945178891858e-08, + "loss": 0.6898, + "step": 37813 + }, + { + "epoch": 1.38, + "learning_rate": 6.170578972424652e-08, + "loss": 0.7004, + "step": 37814 + }, + { + "epoch": 1.38, + "learning_rate": 6.170212765957446e-08, + "loss": 0.6966, + "step": 37815 + }, + { + "epoch": 1.38, + "learning_rate": 6.16984655949024e-08, + "loss": 0.6925, + "step": 37816 + }, + { + "epoch": 1.38, + "learning_rate": 6.169480353023034e-08, + "loss": 0.6754, + "step": 37817 + }, + { + "epoch": 1.38, + "learning_rate": 6.169114146555828e-08, + "loss": 0.6848, + "step": 37818 + }, + { + "epoch": 1.38, + "learning_rate": 6.168747940088622e-08, + "loss": 0.6865, + "step": 37819 + }, + { + "epoch": 1.38, + "learning_rate": 6.168381733621415e-08, + "loss": 0.6986, + "step": 37820 + }, + { + "epoch": 1.38, + "learning_rate": 6.168015527154209e-08, + "loss": 0.6924, + "step": 37821 + }, + { + "epoch": 1.38, + "learning_rate": 6.167649320687003e-08, + "loss": 0.6908, + "step": 37822 + }, + { + "epoch": 1.38, + "learning_rate": 6.167283114219797e-08, + "loss": 0.7001, + "step": 37823 + }, + { + "epoch": 1.38, + "learning_rate": 6.166916907752591e-08, + "loss": 0.6994, + "step": 37824 + }, + { + "epoch": 1.38, + "learning_rate": 6.166550701285385e-08, + "loss": 0.6914, + "step": 37825 + }, + { + "epoch": 1.38, + "learning_rate": 6.166184494818179e-08, + "loss": 0.6957, + "step": 37826 + }, + { + "epoch": 1.38, + "learning_rate": 6.165818288350973e-08, + "loss": 0.6949, + "step": 37827 + }, + { + "epoch": 1.38, + "learning_rate": 6.165452081883767e-08, + "loss": 0.6951, + "step": 37828 + }, + { + "epoch": 1.38, + "learning_rate": 6.165085875416559e-08, + "loss": 0.699, + "step": 37829 + }, + { + "epoch": 1.38, + "learning_rate": 6.164719668949353e-08, + "loss": 0.6841, + "step": 37830 + }, + { + "epoch": 1.38, + "learning_rate": 6.164353462482147e-08, + "loss": 0.6947, + "step": 37831 + }, + { + "epoch": 1.38, + "learning_rate": 6.163987256014941e-08, + "loss": 0.6931, + "step": 37832 + }, + { + "epoch": 1.38, + "learning_rate": 6.163621049547735e-08, + "loss": 0.6941, + "step": 37833 + }, + { + "epoch": 1.38, + "learning_rate": 6.163254843080529e-08, + "loss": 0.6978, + "step": 37834 + }, + { + "epoch": 1.38, + "learning_rate": 6.162888636613322e-08, + "loss": 0.6972, + "step": 37835 + }, + { + "epoch": 1.38, + "learning_rate": 6.162522430146116e-08, + "loss": 0.7012, + "step": 37836 + }, + { + "epoch": 1.38, + "learning_rate": 6.16215622367891e-08, + "loss": 0.6941, + "step": 37837 + }, + { + "epoch": 1.38, + "learning_rate": 6.161790017211703e-08, + "loss": 0.6907, + "step": 37838 + }, + { + "epoch": 1.38, + "learning_rate": 6.161423810744497e-08, + "loss": 0.6956, + "step": 37839 + }, + { + "epoch": 1.38, + "learning_rate": 6.16105760427729e-08, + "loss": 0.685, + "step": 37840 + }, + { + "epoch": 1.38, + "learning_rate": 6.160691397810085e-08, + "loss": 0.6955, + "step": 37841 + }, + { + "epoch": 1.38, + "learning_rate": 6.160325191342879e-08, + "loss": 0.6982, + "step": 37842 + }, + { + "epoch": 1.38, + "learning_rate": 6.159958984875673e-08, + "loss": 0.6962, + "step": 37843 + }, + { + "epoch": 1.38, + "learning_rate": 6.159592778408467e-08, + "loss": 0.6896, + "step": 37844 + }, + { + "epoch": 1.38, + "learning_rate": 6.15922657194126e-08, + "loss": 0.685, + "step": 37845 + }, + { + "epoch": 1.38, + "learning_rate": 6.158860365474055e-08, + "loss": 0.695, + "step": 37846 + }, + { + "epoch": 1.38, + "learning_rate": 6.158494159006847e-08, + "loss": 0.6963, + "step": 37847 + }, + { + "epoch": 1.38, + "learning_rate": 6.158127952539641e-08, + "loss": 0.6806, + "step": 37848 + }, + { + "epoch": 1.38, + "learning_rate": 6.157761746072435e-08, + "loss": 0.6877, + "step": 37849 + }, + { + "epoch": 1.38, + "learning_rate": 6.157395539605229e-08, + "loss": 0.7008, + "step": 37850 + }, + { + "epoch": 1.38, + "learning_rate": 6.157029333138023e-08, + "loss": 0.6891, + "step": 37851 + }, + { + "epoch": 1.38, + "learning_rate": 6.156663126670817e-08, + "loss": 0.713, + "step": 37852 + }, + { + "epoch": 1.38, + "learning_rate": 6.156296920203611e-08, + "loss": 0.6935, + "step": 37853 + }, + { + "epoch": 1.38, + "learning_rate": 6.155930713736405e-08, + "loss": 0.7011, + "step": 37854 + }, + { + "epoch": 1.39, + "learning_rate": 6.155564507269199e-08, + "loss": 0.6819, + "step": 37855 + }, + { + "epoch": 1.39, + "learning_rate": 6.155198300801992e-08, + "loss": 0.6972, + "step": 37856 + }, + { + "epoch": 1.39, + "learning_rate": 6.154832094334786e-08, + "loss": 0.6826, + "step": 37857 + }, + { + "epoch": 1.39, + "learning_rate": 6.15446588786758e-08, + "loss": 0.7031, + "step": 37858 + }, + { + "epoch": 1.39, + "learning_rate": 6.154099681400374e-08, + "loss": 0.7036, + "step": 37859 + }, + { + "epoch": 1.39, + "learning_rate": 6.153733474933166e-08, + "loss": 0.7018, + "step": 37860 + }, + { + "epoch": 1.39, + "learning_rate": 6.15336726846596e-08, + "loss": 0.6825, + "step": 37861 + }, + { + "epoch": 1.39, + "learning_rate": 6.153001061998754e-08, + "loss": 0.7006, + "step": 37862 + }, + { + "epoch": 1.39, + "learning_rate": 6.152634855531548e-08, + "loss": 0.6928, + "step": 37863 + }, + { + "epoch": 1.39, + "learning_rate": 6.152268649064342e-08, + "loss": 0.6884, + "step": 37864 + }, + { + "epoch": 1.39, + "learning_rate": 6.151902442597135e-08, + "loss": 0.6796, + "step": 37865 + }, + { + "epoch": 1.39, + "learning_rate": 6.151536236129929e-08, + "loss": 0.6944, + "step": 37866 + }, + { + "epoch": 1.39, + "learning_rate": 6.151170029662723e-08, + "loss": 0.7026, + "step": 37867 + }, + { + "epoch": 1.39, + "learning_rate": 6.150803823195517e-08, + "loss": 0.6852, + "step": 37868 + }, + { + "epoch": 1.39, + "learning_rate": 6.150437616728311e-08, + "loss": 0.6893, + "step": 37869 + }, + { + "epoch": 1.39, + "learning_rate": 6.150071410261105e-08, + "loss": 0.6753, + "step": 37870 + }, + { + "epoch": 1.39, + "learning_rate": 6.149705203793899e-08, + "loss": 0.6809, + "step": 37871 + }, + { + "epoch": 1.39, + "learning_rate": 6.149338997326693e-08, + "loss": 0.6968, + "step": 37872 + }, + { + "epoch": 1.39, + "learning_rate": 6.148972790859487e-08, + "loss": 0.6949, + "step": 37873 + }, + { + "epoch": 1.39, + "learning_rate": 6.14860658439228e-08, + "loss": 0.7, + "step": 37874 + }, + { + "epoch": 1.39, + "learning_rate": 6.148240377925074e-08, + "loss": 0.6818, + "step": 37875 + }, + { + "epoch": 1.39, + "learning_rate": 6.147874171457868e-08, + "loss": 0.6735, + "step": 37876 + }, + { + "epoch": 1.39, + "learning_rate": 6.147507964990662e-08, + "loss": 0.7081, + "step": 37877 + }, + { + "epoch": 1.39, + "learning_rate": 6.147141758523456e-08, + "loss": 0.6936, + "step": 37878 + }, + { + "epoch": 1.39, + "learning_rate": 6.14677555205625e-08, + "loss": 0.6938, + "step": 37879 + }, + { + "epoch": 1.39, + "learning_rate": 6.146409345589044e-08, + "loss": 0.6893, + "step": 37880 + }, + { + "epoch": 1.39, + "learning_rate": 6.146043139121836e-08, + "loss": 0.6947, + "step": 37881 + }, + { + "epoch": 1.39, + "learning_rate": 6.14567693265463e-08, + "loss": 0.692, + "step": 37882 + }, + { + "epoch": 1.39, + "learning_rate": 6.145310726187424e-08, + "loss": 0.6914, + "step": 37883 + }, + { + "epoch": 1.39, + "learning_rate": 6.144944519720217e-08, + "loss": 0.6904, + "step": 37884 + }, + { + "epoch": 1.39, + "learning_rate": 6.144578313253011e-08, + "loss": 0.6701, + "step": 37885 + }, + { + "epoch": 1.39, + "learning_rate": 6.144212106785805e-08, + "loss": 0.7071, + "step": 37886 + }, + { + "epoch": 1.39, + "learning_rate": 6.143845900318599e-08, + "loss": 0.6849, + "step": 37887 + }, + { + "epoch": 1.39, + "learning_rate": 6.143479693851393e-08, + "loss": 0.6836, + "step": 37888 + }, + { + "epoch": 1.39, + "learning_rate": 6.143113487384187e-08, + "loss": 0.7016, + "step": 37889 + }, + { + "epoch": 1.39, + "learning_rate": 6.142747280916981e-08, + "loss": 0.6833, + "step": 37890 + }, + { + "epoch": 1.39, + "learning_rate": 6.142381074449775e-08, + "loss": 0.6863, + "step": 37891 + }, + { + "epoch": 1.39, + "learning_rate": 6.142014867982568e-08, + "loss": 0.697, + "step": 37892 + }, + { + "epoch": 1.39, + "learning_rate": 6.141648661515362e-08, + "loss": 0.6852, + "step": 37893 + }, + { + "epoch": 1.39, + "learning_rate": 6.141282455048156e-08, + "loss": 0.6974, + "step": 37894 + }, + { + "epoch": 1.39, + "learning_rate": 6.14091624858095e-08, + "loss": 0.6799, + "step": 37895 + }, + { + "epoch": 1.39, + "learning_rate": 6.140550042113744e-08, + "loss": 0.7061, + "step": 37896 + }, + { + "epoch": 1.39, + "learning_rate": 6.140183835646538e-08, + "loss": 0.6869, + "step": 37897 + }, + { + "epoch": 1.39, + "learning_rate": 6.139817629179332e-08, + "loss": 0.6935, + "step": 37898 + }, + { + "epoch": 1.39, + "learning_rate": 6.139451422712126e-08, + "loss": 0.7031, + "step": 37899 + }, + { + "epoch": 1.39, + "learning_rate": 6.13908521624492e-08, + "loss": 0.6873, + "step": 37900 + }, + { + "epoch": 1.39, + "learning_rate": 6.138719009777712e-08, + "loss": 0.7065, + "step": 37901 + }, + { + "epoch": 1.39, + "learning_rate": 6.138352803310506e-08, + "loss": 0.6936, + "step": 37902 + }, + { + "epoch": 1.39, + "learning_rate": 6.1379865968433e-08, + "loss": 0.6907, + "step": 37903 + }, + { + "epoch": 1.39, + "learning_rate": 6.137620390376094e-08, + "loss": 0.6914, + "step": 37904 + }, + { + "epoch": 1.39, + "learning_rate": 6.137254183908888e-08, + "loss": 0.6906, + "step": 37905 + }, + { + "epoch": 1.39, + "learning_rate": 6.136887977441681e-08, + "loss": 0.6997, + "step": 37906 + }, + { + "epoch": 1.39, + "learning_rate": 6.136521770974475e-08, + "loss": 0.6938, + "step": 37907 + }, + { + "epoch": 1.39, + "learning_rate": 6.136155564507269e-08, + "loss": 0.6809, + "step": 37908 + }, + { + "epoch": 1.39, + "learning_rate": 6.135789358040063e-08, + "loss": 0.6915, + "step": 37909 + }, + { + "epoch": 1.39, + "learning_rate": 6.135423151572856e-08, + "loss": 0.6861, + "step": 37910 + }, + { + "epoch": 1.39, + "learning_rate": 6.13505694510565e-08, + "loss": 0.6853, + "step": 37911 + }, + { + "epoch": 1.39, + "learning_rate": 6.134690738638444e-08, + "loss": 0.6832, + "step": 37912 + }, + { + "epoch": 1.39, + "learning_rate": 6.134324532171238e-08, + "loss": 0.7106, + "step": 37913 + }, + { + "epoch": 1.39, + "learning_rate": 6.133958325704032e-08, + "loss": 0.6938, + "step": 37914 + }, + { + "epoch": 1.39, + "learning_rate": 6.133592119236826e-08, + "loss": 0.7122, + "step": 37915 + }, + { + "epoch": 1.39, + "learning_rate": 6.13322591276962e-08, + "loss": 0.6863, + "step": 37916 + }, + { + "epoch": 1.39, + "learning_rate": 6.132859706302413e-08, + "loss": 0.6948, + "step": 37917 + }, + { + "epoch": 1.39, + "learning_rate": 6.132493499835207e-08, + "loss": 0.697, + "step": 37918 + }, + { + "epoch": 1.39, + "learning_rate": 6.132127293368e-08, + "loss": 0.6936, + "step": 37919 + }, + { + "epoch": 1.39, + "learning_rate": 6.131761086900794e-08, + "loss": 0.7011, + "step": 37920 + }, + { + "epoch": 1.39, + "learning_rate": 6.131394880433588e-08, + "loss": 0.6994, + "step": 37921 + }, + { + "epoch": 1.39, + "learning_rate": 6.131028673966382e-08, + "loss": 0.6892, + "step": 37922 + }, + { + "epoch": 1.39, + "learning_rate": 6.130662467499176e-08, + "loss": 0.7, + "step": 37923 + }, + { + "epoch": 1.39, + "learning_rate": 6.13029626103197e-08, + "loss": 0.6989, + "step": 37924 + }, + { + "epoch": 1.39, + "learning_rate": 6.129930054564764e-08, + "loss": 0.6872, + "step": 37925 + }, + { + "epoch": 1.39, + "learning_rate": 6.129563848097558e-08, + "loss": 0.6898, + "step": 37926 + }, + { + "epoch": 1.39, + "learning_rate": 6.129197641630352e-08, + "loss": 0.7048, + "step": 37927 + }, + { + "epoch": 1.39, + "learning_rate": 6.128831435163145e-08, + "loss": 0.6949, + "step": 37928 + }, + { + "epoch": 1.39, + "learning_rate": 6.128465228695939e-08, + "loss": 0.6979, + "step": 37929 + }, + { + "epoch": 1.39, + "learning_rate": 6.128099022228731e-08, + "loss": 0.7065, + "step": 37930 + }, + { + "epoch": 1.39, + "learning_rate": 6.127732815761525e-08, + "loss": 0.6934, + "step": 37931 + }, + { + "epoch": 1.39, + "learning_rate": 6.12736660929432e-08, + "loss": 0.6812, + "step": 37932 + }, + { + "epoch": 1.39, + "learning_rate": 6.127000402827113e-08, + "loss": 0.6929, + "step": 37933 + }, + { + "epoch": 1.39, + "learning_rate": 6.126634196359907e-08, + "loss": 0.7013, + "step": 37934 + }, + { + "epoch": 1.39, + "learning_rate": 6.126267989892701e-08, + "loss": 0.6889, + "step": 37935 + }, + { + "epoch": 1.39, + "learning_rate": 6.125901783425494e-08, + "loss": 0.7031, + "step": 37936 + }, + { + "epoch": 1.39, + "learning_rate": 6.125535576958288e-08, + "loss": 0.7007, + "step": 37937 + }, + { + "epoch": 1.39, + "learning_rate": 6.125169370491082e-08, + "loss": 0.6919, + "step": 37938 + }, + { + "epoch": 1.39, + "learning_rate": 6.124803164023876e-08, + "loss": 0.6961, + "step": 37939 + }, + { + "epoch": 1.39, + "learning_rate": 6.12443695755667e-08, + "loss": 0.6902, + "step": 37940 + }, + { + "epoch": 1.39, + "learning_rate": 6.124070751089464e-08, + "loss": 0.6918, + "step": 37941 + }, + { + "epoch": 1.39, + "learning_rate": 6.123704544622258e-08, + "loss": 0.687, + "step": 37942 + }, + { + "epoch": 1.39, + "learning_rate": 6.123338338155052e-08, + "loss": 0.6982, + "step": 37943 + }, + { + "epoch": 1.39, + "learning_rate": 6.122972131687846e-08, + "loss": 0.6772, + "step": 37944 + }, + { + "epoch": 1.39, + "learning_rate": 6.122605925220639e-08, + "loss": 0.6942, + "step": 37945 + }, + { + "epoch": 1.39, + "learning_rate": 6.122239718753433e-08, + "loss": 0.6899, + "step": 37946 + }, + { + "epoch": 1.39, + "learning_rate": 6.121873512286227e-08, + "loss": 0.6887, + "step": 37947 + }, + { + "epoch": 1.39, + "learning_rate": 6.12150730581902e-08, + "loss": 0.7059, + "step": 37948 + }, + { + "epoch": 1.39, + "learning_rate": 6.121141099351815e-08, + "loss": 0.6946, + "step": 37949 + }, + { + "epoch": 1.39, + "learning_rate": 6.120774892884609e-08, + "loss": 0.6979, + "step": 37950 + }, + { + "epoch": 1.39, + "learning_rate": 6.120408686417403e-08, + "loss": 0.675, + "step": 37951 + }, + { + "epoch": 1.39, + "learning_rate": 6.120042479950195e-08, + "loss": 0.6907, + "step": 37952 + }, + { + "epoch": 1.39, + "learning_rate": 6.119676273482989e-08, + "loss": 0.6878, + "step": 37953 + }, + { + "epoch": 1.39, + "learning_rate": 6.119310067015783e-08, + "loss": 0.6778, + "step": 37954 + }, + { + "epoch": 1.39, + "learning_rate": 6.118943860548576e-08, + "loss": 0.6968, + "step": 37955 + }, + { + "epoch": 1.39, + "learning_rate": 6.11857765408137e-08, + "loss": 0.699, + "step": 37956 + }, + { + "epoch": 1.39, + "learning_rate": 6.118211447614164e-08, + "loss": 0.6932, + "step": 37957 + }, + { + "epoch": 1.39, + "learning_rate": 6.117845241146958e-08, + "loss": 0.7, + "step": 37958 + }, + { + "epoch": 1.39, + "learning_rate": 6.117479034679752e-08, + "loss": 0.6899, + "step": 37959 + }, + { + "epoch": 1.39, + "learning_rate": 6.117112828212546e-08, + "loss": 0.6854, + "step": 37960 + }, + { + "epoch": 1.39, + "learning_rate": 6.11674662174534e-08, + "loss": 0.7011, + "step": 37961 + }, + { + "epoch": 1.39, + "learning_rate": 6.116380415278134e-08, + "loss": 0.684, + "step": 37962 + }, + { + "epoch": 1.39, + "learning_rate": 6.116014208810927e-08, + "loss": 0.7013, + "step": 37963 + }, + { + "epoch": 1.39, + "learning_rate": 6.11564800234372e-08, + "loss": 0.6946, + "step": 37964 + }, + { + "epoch": 1.39, + "learning_rate": 6.115281795876515e-08, + "loss": 0.6899, + "step": 37965 + }, + { + "epoch": 1.39, + "learning_rate": 6.114915589409309e-08, + "loss": 0.6881, + "step": 37966 + }, + { + "epoch": 1.39, + "learning_rate": 6.114549382942103e-08, + "loss": 0.6995, + "step": 37967 + }, + { + "epoch": 1.39, + "learning_rate": 6.114183176474897e-08, + "loss": 0.6816, + "step": 37968 + }, + { + "epoch": 1.39, + "learning_rate": 6.11381697000769e-08, + "loss": 0.7208, + "step": 37969 + }, + { + "epoch": 1.39, + "learning_rate": 6.113450763540485e-08, + "loss": 0.6866, + "step": 37970 + }, + { + "epoch": 1.39, + "learning_rate": 6.113084557073279e-08, + "loss": 0.6924, + "step": 37971 + }, + { + "epoch": 1.39, + "learning_rate": 6.112718350606071e-08, + "loss": 0.7057, + "step": 37972 + }, + { + "epoch": 1.39, + "learning_rate": 6.112352144138865e-08, + "loss": 0.7065, + "step": 37973 + }, + { + "epoch": 1.39, + "learning_rate": 6.111985937671659e-08, + "loss": 0.6929, + "step": 37974 + }, + { + "epoch": 1.39, + "learning_rate": 6.111619731204453e-08, + "loss": 0.6965, + "step": 37975 + }, + { + "epoch": 1.39, + "learning_rate": 6.111253524737247e-08, + "loss": 0.7017, + "step": 37976 + }, + { + "epoch": 1.39, + "learning_rate": 6.11088731827004e-08, + "loss": 0.6994, + "step": 37977 + }, + { + "epoch": 1.39, + "learning_rate": 6.110521111802834e-08, + "loss": 0.6844, + "step": 37978 + }, + { + "epoch": 1.39, + "learning_rate": 6.110154905335628e-08, + "loss": 0.71, + "step": 37979 + }, + { + "epoch": 1.39, + "learning_rate": 6.109788698868422e-08, + "loss": 0.6894, + "step": 37980 + }, + { + "epoch": 1.39, + "learning_rate": 6.109422492401214e-08, + "loss": 0.6905, + "step": 37981 + }, + { + "epoch": 1.39, + "learning_rate": 6.109056285934008e-08, + "loss": 0.7006, + "step": 37982 + }, + { + "epoch": 1.39, + "learning_rate": 6.108690079466802e-08, + "loss": 0.6831, + "step": 37983 + }, + { + "epoch": 1.39, + "learning_rate": 6.108323872999596e-08, + "loss": 0.6997, + "step": 37984 + }, + { + "epoch": 1.39, + "learning_rate": 6.10795766653239e-08, + "loss": 0.6937, + "step": 37985 + }, + { + "epoch": 1.39, + "learning_rate": 6.107591460065184e-08, + "loss": 0.6984, + "step": 37986 + }, + { + "epoch": 1.39, + "learning_rate": 6.107225253597978e-08, + "loss": 0.704, + "step": 37987 + }, + { + "epoch": 1.39, + "learning_rate": 6.106859047130772e-08, + "loss": 0.6904, + "step": 37988 + }, + { + "epoch": 1.39, + "learning_rate": 6.106492840663566e-08, + "loss": 0.7034, + "step": 37989 + }, + { + "epoch": 1.39, + "learning_rate": 6.106126634196359e-08, + "loss": 0.7007, + "step": 37990 + }, + { + "epoch": 1.39, + "learning_rate": 6.105760427729153e-08, + "loss": 0.688, + "step": 37991 + }, + { + "epoch": 1.39, + "learning_rate": 6.105394221261947e-08, + "loss": 0.6854, + "step": 37992 + }, + { + "epoch": 1.39, + "learning_rate": 6.105028014794741e-08, + "loss": 0.6901, + "step": 37993 + }, + { + "epoch": 1.39, + "learning_rate": 6.104661808327535e-08, + "loss": 0.6855, + "step": 37994 + }, + { + "epoch": 1.39, + "learning_rate": 6.104295601860329e-08, + "loss": 0.6829, + "step": 37995 + }, + { + "epoch": 1.39, + "learning_rate": 6.103929395393123e-08, + "loss": 0.702, + "step": 37996 + }, + { + "epoch": 1.39, + "learning_rate": 6.103563188925917e-08, + "loss": 0.7017, + "step": 37997 + }, + { + "epoch": 1.39, + "learning_rate": 6.103196982458711e-08, + "loss": 0.6745, + "step": 37998 + }, + { + "epoch": 1.39, + "learning_rate": 6.102830775991504e-08, + "loss": 0.6835, + "step": 37999 + }, + { + "epoch": 1.39, + "learning_rate": 6.102464569524298e-08, + "loss": 0.6807, + "step": 38000 + }, + { + "epoch": 1.39, + "learning_rate": 6.10209836305709e-08, + "loss": 0.6855, + "step": 38001 + }, + { + "epoch": 1.39, + "learning_rate": 6.101732156589884e-08, + "loss": 0.7011, + "step": 38002 + }, + { + "epoch": 1.39, + "learning_rate": 6.101365950122678e-08, + "loss": 0.6821, + "step": 38003 + }, + { + "epoch": 1.39, + "learning_rate": 6.100999743655472e-08, + "loss": 0.7109, + "step": 38004 + }, + { + "epoch": 1.39, + "learning_rate": 6.100633537188266e-08, + "loss": 0.692, + "step": 38005 + }, + { + "epoch": 1.39, + "learning_rate": 6.10026733072106e-08, + "loss": 0.7019, + "step": 38006 + }, + { + "epoch": 1.39, + "learning_rate": 6.099901124253854e-08, + "loss": 0.6855, + "step": 38007 + }, + { + "epoch": 1.39, + "learning_rate": 6.099534917786647e-08, + "loss": 0.6855, + "step": 38008 + }, + { + "epoch": 1.39, + "learning_rate": 6.099168711319441e-08, + "loss": 0.6958, + "step": 38009 + }, + { + "epoch": 1.39, + "learning_rate": 6.098802504852235e-08, + "loss": 0.699, + "step": 38010 + }, + { + "epoch": 1.39, + "learning_rate": 6.098436298385029e-08, + "loss": 0.6877, + "step": 38011 + }, + { + "epoch": 1.39, + "learning_rate": 6.098070091917823e-08, + "loss": 0.6801, + "step": 38012 + }, + { + "epoch": 1.39, + "learning_rate": 6.097703885450617e-08, + "loss": 0.7029, + "step": 38013 + }, + { + "epoch": 1.39, + "learning_rate": 6.097337678983411e-08, + "loss": 0.7038, + "step": 38014 + }, + { + "epoch": 1.39, + "learning_rate": 6.096971472516205e-08, + "loss": 0.6995, + "step": 38015 + }, + { + "epoch": 1.39, + "learning_rate": 6.096605266048999e-08, + "loss": 0.6727, + "step": 38016 + }, + { + "epoch": 1.39, + "learning_rate": 6.096239059581792e-08, + "loss": 0.6811, + "step": 38017 + }, + { + "epoch": 1.39, + "learning_rate": 6.095872853114586e-08, + "loss": 0.6879, + "step": 38018 + }, + { + "epoch": 1.39, + "learning_rate": 6.09550664664738e-08, + "loss": 0.6884, + "step": 38019 + }, + { + "epoch": 1.39, + "learning_rate": 6.095140440180174e-08, + "loss": 0.6804, + "step": 38020 + }, + { + "epoch": 1.39, + "learning_rate": 6.094774233712968e-08, + "loss": 0.6934, + "step": 38021 + }, + { + "epoch": 1.39, + "learning_rate": 6.094408027245762e-08, + "loss": 0.7098, + "step": 38022 + }, + { + "epoch": 1.39, + "learning_rate": 6.094041820778554e-08, + "loss": 0.6905, + "step": 38023 + }, + { + "epoch": 1.39, + "learning_rate": 6.093675614311348e-08, + "loss": 0.7041, + "step": 38024 + }, + { + "epoch": 1.39, + "learning_rate": 6.093309407844142e-08, + "loss": 0.6894, + "step": 38025 + }, + { + "epoch": 1.39, + "learning_rate": 6.092943201376935e-08, + "loss": 0.6848, + "step": 38026 + }, + { + "epoch": 1.39, + "learning_rate": 6.092576994909729e-08, + "loss": 0.6988, + "step": 38027 + }, + { + "epoch": 1.39, + "learning_rate": 6.092210788442523e-08, + "loss": 0.6953, + "step": 38028 + }, + { + "epoch": 1.39, + "learning_rate": 6.091844581975317e-08, + "loss": 0.705, + "step": 38029 + }, + { + "epoch": 1.39, + "learning_rate": 6.091478375508111e-08, + "loss": 0.6841, + "step": 38030 + }, + { + "epoch": 1.39, + "learning_rate": 6.091112169040905e-08, + "loss": 0.6832, + "step": 38031 + }, + { + "epoch": 1.39, + "learning_rate": 6.090745962573699e-08, + "loss": 0.7011, + "step": 38032 + }, + { + "epoch": 1.39, + "learning_rate": 6.090379756106493e-08, + "loss": 0.6853, + "step": 38033 + }, + { + "epoch": 1.39, + "learning_rate": 6.090013549639287e-08, + "loss": 0.6917, + "step": 38034 + }, + { + "epoch": 1.39, + "learning_rate": 6.08964734317208e-08, + "loss": 0.6968, + "step": 38035 + }, + { + "epoch": 1.39, + "learning_rate": 6.089281136704874e-08, + "loss": 0.7073, + "step": 38036 + }, + { + "epoch": 1.39, + "learning_rate": 6.088914930237667e-08, + "loss": 0.6884, + "step": 38037 + }, + { + "epoch": 1.39, + "learning_rate": 6.088548723770461e-08, + "loss": 0.7046, + "step": 38038 + }, + { + "epoch": 1.39, + "learning_rate": 6.088182517303255e-08, + "loss": 0.6799, + "step": 38039 + }, + { + "epoch": 1.39, + "learning_rate": 6.08781631083605e-08, + "loss": 0.698, + "step": 38040 + }, + { + "epoch": 1.39, + "learning_rate": 6.087450104368843e-08, + "loss": 0.6904, + "step": 38041 + }, + { + "epoch": 1.39, + "learning_rate": 6.087083897901637e-08, + "loss": 0.6878, + "step": 38042 + }, + { + "epoch": 1.39, + "learning_rate": 6.086717691434431e-08, + "loss": 0.7, + "step": 38043 + }, + { + "epoch": 1.39, + "learning_rate": 6.086351484967224e-08, + "loss": 0.701, + "step": 38044 + }, + { + "epoch": 1.39, + "learning_rate": 6.085985278500018e-08, + "loss": 0.7018, + "step": 38045 + }, + { + "epoch": 1.39, + "learning_rate": 6.085619072032812e-08, + "loss": 0.6964, + "step": 38046 + }, + { + "epoch": 1.39, + "learning_rate": 6.085252865565606e-08, + "loss": 0.6884, + "step": 38047 + }, + { + "epoch": 1.39, + "learning_rate": 6.084886659098399e-08, + "loss": 0.7067, + "step": 38048 + }, + { + "epoch": 1.39, + "learning_rate": 6.084520452631193e-08, + "loss": 0.7019, + "step": 38049 + }, + { + "epoch": 1.39, + "learning_rate": 6.084154246163987e-08, + "loss": 0.6887, + "step": 38050 + }, + { + "epoch": 1.39, + "learning_rate": 6.083788039696781e-08, + "loss": 0.6894, + "step": 38051 + }, + { + "epoch": 1.39, + "learning_rate": 6.083421833229575e-08, + "loss": 0.7152, + "step": 38052 + }, + { + "epoch": 1.39, + "learning_rate": 6.083055626762367e-08, + "loss": 0.6892, + "step": 38053 + }, + { + "epoch": 1.39, + "learning_rate": 6.082689420295161e-08, + "loss": 0.6871, + "step": 38054 + }, + { + "epoch": 1.39, + "learning_rate": 6.082323213827955e-08, + "loss": 0.6771, + "step": 38055 + }, + { + "epoch": 1.39, + "learning_rate": 6.08195700736075e-08, + "loss": 0.6895, + "step": 38056 + }, + { + "epoch": 1.39, + "learning_rate": 6.081590800893543e-08, + "loss": 0.6906, + "step": 38057 + }, + { + "epoch": 1.39, + "learning_rate": 6.081224594426337e-08, + "loss": 0.695, + "step": 38058 + }, + { + "epoch": 1.39, + "learning_rate": 6.080858387959131e-08, + "loss": 0.693, + "step": 38059 + }, + { + "epoch": 1.39, + "learning_rate": 6.080492181491925e-08, + "loss": 0.6876, + "step": 38060 + }, + { + "epoch": 1.39, + "learning_rate": 6.08012597502472e-08, + "loss": 0.7111, + "step": 38061 + }, + { + "epoch": 1.39, + "learning_rate": 6.079759768557512e-08, + "loss": 0.696, + "step": 38062 + }, + { + "epoch": 1.39, + "learning_rate": 6.079393562090306e-08, + "loss": 0.6749, + "step": 38063 + }, + { + "epoch": 1.39, + "learning_rate": 6.0790273556231e-08, + "loss": 0.6844, + "step": 38064 + }, + { + "epoch": 1.39, + "learning_rate": 6.078661149155894e-08, + "loss": 0.6837, + "step": 38065 + }, + { + "epoch": 1.39, + "learning_rate": 6.078294942688688e-08, + "loss": 0.696, + "step": 38066 + }, + { + "epoch": 1.39, + "learning_rate": 6.077928736221482e-08, + "loss": 0.6817, + "step": 38067 + }, + { + "epoch": 1.39, + "learning_rate": 6.077562529754276e-08, + "loss": 0.6925, + "step": 38068 + }, + { + "epoch": 1.39, + "learning_rate": 6.07719632328707e-08, + "loss": 0.6992, + "step": 38069 + }, + { + "epoch": 1.39, + "learning_rate": 6.076830116819863e-08, + "loss": 0.6898, + "step": 38070 + }, + { + "epoch": 1.39, + "learning_rate": 6.076463910352657e-08, + "loss": 0.6976, + "step": 38071 + }, + { + "epoch": 1.39, + "learning_rate": 6.076097703885449e-08, + "loss": 0.6911, + "step": 38072 + }, + { + "epoch": 1.39, + "learning_rate": 6.075731497418243e-08, + "loss": 0.7092, + "step": 38073 + }, + { + "epoch": 1.39, + "learning_rate": 6.075365290951037e-08, + "loss": 0.6896, + "step": 38074 + }, + { + "epoch": 1.39, + "learning_rate": 6.074999084483831e-08, + "loss": 0.6907, + "step": 38075 + }, + { + "epoch": 1.39, + "learning_rate": 6.074632878016625e-08, + "loss": 0.6941, + "step": 38076 + }, + { + "epoch": 1.39, + "learning_rate": 6.074266671549419e-08, + "loss": 0.7127, + "step": 38077 + }, + { + "epoch": 1.39, + "learning_rate": 6.073900465082213e-08, + "loss": 0.6962, + "step": 38078 + }, + { + "epoch": 1.39, + "learning_rate": 6.073534258615007e-08, + "loss": 0.6867, + "step": 38079 + }, + { + "epoch": 1.39, + "learning_rate": 6.0731680521478e-08, + "loss": 0.6976, + "step": 38080 + }, + { + "epoch": 1.39, + "learning_rate": 6.072801845680594e-08, + "loss": 0.6983, + "step": 38081 + }, + { + "epoch": 1.39, + "learning_rate": 6.072435639213388e-08, + "loss": 0.7027, + "step": 38082 + }, + { + "epoch": 1.39, + "learning_rate": 6.072069432746182e-08, + "loss": 0.6831, + "step": 38083 + }, + { + "epoch": 1.39, + "learning_rate": 6.071703226278976e-08, + "loss": 0.6773, + "step": 38084 + }, + { + "epoch": 1.39, + "learning_rate": 6.07133701981177e-08, + "loss": 0.6771, + "step": 38085 + }, + { + "epoch": 1.39, + "learning_rate": 6.070970813344564e-08, + "loss": 0.703, + "step": 38086 + }, + { + "epoch": 1.39, + "learning_rate": 6.070604606877358e-08, + "loss": 0.6844, + "step": 38087 + }, + { + "epoch": 1.39, + "learning_rate": 6.070238400410152e-08, + "loss": 0.6905, + "step": 38088 + }, + { + "epoch": 1.39, + "learning_rate": 6.069872193942945e-08, + "loss": 0.6888, + "step": 38089 + }, + { + "epoch": 1.39, + "learning_rate": 6.069505987475739e-08, + "loss": 0.6944, + "step": 38090 + }, + { + "epoch": 1.39, + "learning_rate": 6.069139781008533e-08, + "loss": 0.6944, + "step": 38091 + }, + { + "epoch": 1.39, + "learning_rate": 6.068773574541327e-08, + "loss": 0.6962, + "step": 38092 + }, + { + "epoch": 1.39, + "learning_rate": 6.06840736807412e-08, + "loss": 0.6824, + "step": 38093 + }, + { + "epoch": 1.39, + "learning_rate": 6.068041161606913e-08, + "loss": 0.6951, + "step": 38094 + }, + { + "epoch": 1.39, + "learning_rate": 6.067674955139707e-08, + "loss": 0.6861, + "step": 38095 + }, + { + "epoch": 1.39, + "learning_rate": 6.067308748672501e-08, + "loss": 0.6867, + "step": 38096 + }, + { + "epoch": 1.39, + "learning_rate": 6.066942542205295e-08, + "loss": 0.6887, + "step": 38097 + }, + { + "epoch": 1.39, + "learning_rate": 6.066576335738088e-08, + "loss": 0.6934, + "step": 38098 + }, + { + "epoch": 1.39, + "learning_rate": 6.066210129270882e-08, + "loss": 0.6795, + "step": 38099 + }, + { + "epoch": 1.39, + "learning_rate": 6.065843922803676e-08, + "loss": 0.6939, + "step": 38100 + }, + { + "epoch": 1.39, + "learning_rate": 6.06547771633647e-08, + "loss": 0.7039, + "step": 38101 + }, + { + "epoch": 1.39, + "learning_rate": 6.065111509869264e-08, + "loss": 0.6992, + "step": 38102 + }, + { + "epoch": 1.39, + "learning_rate": 6.064745303402058e-08, + "loss": 0.6874, + "step": 38103 + }, + { + "epoch": 1.39, + "learning_rate": 6.064379096934852e-08, + "loss": 0.6908, + "step": 38104 + }, + { + "epoch": 1.39, + "learning_rate": 6.064012890467646e-08, + "loss": 0.6803, + "step": 38105 + }, + { + "epoch": 1.39, + "learning_rate": 6.06364668400044e-08, + "loss": 0.6919, + "step": 38106 + }, + { + "epoch": 1.39, + "learning_rate": 6.063280477533232e-08, + "loss": 0.695, + "step": 38107 + }, + { + "epoch": 1.39, + "learning_rate": 6.062914271066026e-08, + "loss": 0.6837, + "step": 38108 + }, + { + "epoch": 1.39, + "learning_rate": 6.06254806459882e-08, + "loss": 0.6835, + "step": 38109 + }, + { + "epoch": 1.39, + "learning_rate": 6.062181858131614e-08, + "loss": 0.7064, + "step": 38110 + }, + { + "epoch": 1.39, + "learning_rate": 6.061815651664408e-08, + "loss": 0.6654, + "step": 38111 + }, + { + "epoch": 1.39, + "learning_rate": 6.061449445197202e-08, + "loss": 0.7028, + "step": 38112 + }, + { + "epoch": 1.39, + "learning_rate": 6.061083238729996e-08, + "loss": 0.6888, + "step": 38113 + }, + { + "epoch": 1.39, + "learning_rate": 6.06071703226279e-08, + "loss": 0.6774, + "step": 38114 + }, + { + "epoch": 1.39, + "learning_rate": 6.060350825795584e-08, + "loss": 0.6956, + "step": 38115 + }, + { + "epoch": 1.39, + "learning_rate": 6.059984619328377e-08, + "loss": 0.6725, + "step": 38116 + }, + { + "epoch": 1.39, + "learning_rate": 6.059618412861171e-08, + "loss": 0.7029, + "step": 38117 + }, + { + "epoch": 1.39, + "learning_rate": 6.059252206393965e-08, + "loss": 0.6891, + "step": 38118 + }, + { + "epoch": 1.39, + "learning_rate": 6.058885999926758e-08, + "loss": 0.6926, + "step": 38119 + }, + { + "epoch": 1.39, + "learning_rate": 6.058519793459552e-08, + "loss": 0.7038, + "step": 38120 + }, + { + "epoch": 1.39, + "learning_rate": 6.058153586992346e-08, + "loss": 0.6981, + "step": 38121 + }, + { + "epoch": 1.39, + "learning_rate": 6.05778738052514e-08, + "loss": 0.6835, + "step": 38122 + }, + { + "epoch": 1.39, + "learning_rate": 6.057421174057934e-08, + "loss": 0.6994, + "step": 38123 + }, + { + "epoch": 1.39, + "learning_rate": 6.057054967590728e-08, + "loss": 0.6982, + "step": 38124 + }, + { + "epoch": 1.39, + "learning_rate": 6.05668876112352e-08, + "loss": 0.6779, + "step": 38125 + }, + { + "epoch": 1.39, + "learning_rate": 6.056322554656314e-08, + "loss": 0.6881, + "step": 38126 + }, + { + "epoch": 1.39, + "learning_rate": 6.055956348189108e-08, + "loss": 0.69, + "step": 38127 + }, + { + "epoch": 1.39, + "learning_rate": 6.055590141721902e-08, + "loss": 0.6971, + "step": 38128 + }, + { + "epoch": 1.4, + "learning_rate": 6.055223935254696e-08, + "loss": 0.6894, + "step": 38129 + }, + { + "epoch": 1.4, + "learning_rate": 6.05485772878749e-08, + "loss": 0.6843, + "step": 38130 + }, + { + "epoch": 1.4, + "learning_rate": 6.054491522320284e-08, + "loss": 0.6928, + "step": 38131 + }, + { + "epoch": 1.4, + "learning_rate": 6.054125315853078e-08, + "loss": 0.6879, + "step": 38132 + }, + { + "epoch": 1.4, + "learning_rate": 6.053759109385872e-08, + "loss": 0.6929, + "step": 38133 + }, + { + "epoch": 1.4, + "learning_rate": 6.053392902918665e-08, + "loss": 0.7104, + "step": 38134 + }, + { + "epoch": 1.4, + "learning_rate": 6.053026696451459e-08, + "loss": 0.6913, + "step": 38135 + }, + { + "epoch": 1.4, + "learning_rate": 6.052660489984253e-08, + "loss": 0.699, + "step": 38136 + }, + { + "epoch": 1.4, + "learning_rate": 6.052294283517047e-08, + "loss": 0.6805, + "step": 38137 + }, + { + "epoch": 1.4, + "learning_rate": 6.051928077049841e-08, + "loss": 0.6893, + "step": 38138 + }, + { + "epoch": 1.4, + "learning_rate": 6.051561870582635e-08, + "loss": 0.6887, + "step": 38139 + }, + { + "epoch": 1.4, + "learning_rate": 6.051195664115429e-08, + "loss": 0.6986, + "step": 38140 + }, + { + "epoch": 1.4, + "learning_rate": 6.050829457648222e-08, + "loss": 0.7025, + "step": 38141 + }, + { + "epoch": 1.4, + "learning_rate": 6.050463251181016e-08, + "loss": 0.6983, + "step": 38142 + }, + { + "epoch": 1.4, + "learning_rate": 6.050097044713808e-08, + "loss": 0.707, + "step": 38143 + }, + { + "epoch": 1.4, + "learning_rate": 6.049730838246602e-08, + "loss": 0.6854, + "step": 38144 + }, + { + "epoch": 1.4, + "learning_rate": 6.049364631779396e-08, + "loss": 0.7004, + "step": 38145 + }, + { + "epoch": 1.4, + "learning_rate": 6.04899842531219e-08, + "loss": 0.6925, + "step": 38146 + }, + { + "epoch": 1.4, + "learning_rate": 6.048632218844984e-08, + "loss": 0.6745, + "step": 38147 + }, + { + "epoch": 1.4, + "learning_rate": 6.048266012377778e-08, + "loss": 0.6868, + "step": 38148 + }, + { + "epoch": 1.4, + "learning_rate": 6.047899805910572e-08, + "loss": 0.6785, + "step": 38149 + }, + { + "epoch": 1.4, + "learning_rate": 6.047533599443366e-08, + "loss": 0.6857, + "step": 38150 + }, + { + "epoch": 1.4, + "learning_rate": 6.04716739297616e-08, + "loss": 0.7081, + "step": 38151 + }, + { + "epoch": 1.4, + "learning_rate": 6.046801186508953e-08, + "loss": 0.6911, + "step": 38152 + }, + { + "epoch": 1.4, + "learning_rate": 6.046434980041747e-08, + "loss": 0.6891, + "step": 38153 + }, + { + "epoch": 1.4, + "learning_rate": 6.046068773574541e-08, + "loss": 0.7057, + "step": 38154 + }, + { + "epoch": 1.4, + "learning_rate": 6.045702567107335e-08, + "loss": 0.6839, + "step": 38155 + }, + { + "epoch": 1.4, + "learning_rate": 6.045336360640129e-08, + "loss": 0.6994, + "step": 38156 + }, + { + "epoch": 1.4, + "learning_rate": 6.044970154172923e-08, + "loss": 0.6923, + "step": 38157 + }, + { + "epoch": 1.4, + "learning_rate": 6.044603947705717e-08, + "loss": 0.6953, + "step": 38158 + }, + { + "epoch": 1.4, + "learning_rate": 6.044237741238511e-08, + "loss": 0.6903, + "step": 38159 + }, + { + "epoch": 1.4, + "learning_rate": 6.043871534771305e-08, + "loss": 0.6828, + "step": 38160 + }, + { + "epoch": 1.4, + "learning_rate": 6.043505328304097e-08, + "loss": 0.6878, + "step": 38161 + }, + { + "epoch": 1.4, + "learning_rate": 6.043139121836891e-08, + "loss": 0.6988, + "step": 38162 + }, + { + "epoch": 1.4, + "learning_rate": 6.042772915369685e-08, + "loss": 0.7024, + "step": 38163 + }, + { + "epoch": 1.4, + "learning_rate": 6.04240670890248e-08, + "loss": 0.7028, + "step": 38164 + }, + { + "epoch": 1.4, + "learning_rate": 6.042040502435272e-08, + "loss": 0.7032, + "step": 38165 + }, + { + "epoch": 1.4, + "learning_rate": 6.041674295968066e-08, + "loss": 0.6899, + "step": 38166 + }, + { + "epoch": 1.4, + "learning_rate": 6.04130808950086e-08, + "loss": 0.6833, + "step": 38167 + }, + { + "epoch": 1.4, + "learning_rate": 6.040941883033654e-08, + "loss": 0.7001, + "step": 38168 + }, + { + "epoch": 1.4, + "learning_rate": 6.040575676566448e-08, + "loss": 0.6879, + "step": 38169 + }, + { + "epoch": 1.4, + "learning_rate": 6.040209470099241e-08, + "loss": 0.696, + "step": 38170 + }, + { + "epoch": 1.4, + "learning_rate": 6.039843263632035e-08, + "loss": 0.683, + "step": 38171 + }, + { + "epoch": 1.4, + "learning_rate": 6.039477057164829e-08, + "loss": 0.6912, + "step": 38172 + }, + { + "epoch": 1.4, + "learning_rate": 6.039110850697623e-08, + "loss": 0.688, + "step": 38173 + }, + { + "epoch": 1.4, + "learning_rate": 6.038744644230417e-08, + "loss": 0.6893, + "step": 38174 + }, + { + "epoch": 1.4, + "learning_rate": 6.038378437763211e-08, + "loss": 0.6995, + "step": 38175 + }, + { + "epoch": 1.4, + "learning_rate": 6.038012231296005e-08, + "loss": 0.6983, + "step": 38176 + }, + { + "epoch": 1.4, + "learning_rate": 6.037646024828799e-08, + "loss": 0.6782, + "step": 38177 + }, + { + "epoch": 1.4, + "learning_rate": 6.037279818361593e-08, + "loss": 0.6917, + "step": 38178 + }, + { + "epoch": 1.4, + "learning_rate": 6.036913611894385e-08, + "loss": 0.6932, + "step": 38179 + }, + { + "epoch": 1.4, + "learning_rate": 6.03654740542718e-08, + "loss": 0.6846, + "step": 38180 + }, + { + "epoch": 1.4, + "learning_rate": 6.036181198959973e-08, + "loss": 0.6895, + "step": 38181 + }, + { + "epoch": 1.4, + "learning_rate": 6.035814992492767e-08, + "loss": 0.6926, + "step": 38182 + }, + { + "epoch": 1.4, + "learning_rate": 6.035448786025561e-08, + "loss": 0.6816, + "step": 38183 + }, + { + "epoch": 1.4, + "learning_rate": 6.035082579558355e-08, + "loss": 0.6958, + "step": 38184 + }, + { + "epoch": 1.4, + "learning_rate": 6.034716373091149e-08, + "loss": 0.6961, + "step": 38185 + }, + { + "epoch": 1.4, + "learning_rate": 6.034350166623943e-08, + "loss": 0.6797, + "step": 38186 + }, + { + "epoch": 1.4, + "learning_rate": 6.033983960156736e-08, + "loss": 0.6858, + "step": 38187 + }, + { + "epoch": 1.4, + "learning_rate": 6.03361775368953e-08, + "loss": 0.7136, + "step": 38188 + }, + { + "epoch": 1.4, + "learning_rate": 6.033251547222324e-08, + "loss": 0.7046, + "step": 38189 + }, + { + "epoch": 1.4, + "learning_rate": 6.032885340755117e-08, + "loss": 0.6973, + "step": 38190 + }, + { + "epoch": 1.4, + "learning_rate": 6.03251913428791e-08, + "loss": 0.7052, + "step": 38191 + }, + { + "epoch": 1.4, + "learning_rate": 6.032152927820705e-08, + "loss": 0.6996, + "step": 38192 + }, + { + "epoch": 1.4, + "learning_rate": 6.031786721353499e-08, + "loss": 0.6996, + "step": 38193 + }, + { + "epoch": 1.4, + "learning_rate": 6.031420514886293e-08, + "loss": 0.6998, + "step": 38194 + }, + { + "epoch": 1.4, + "learning_rate": 6.031054308419087e-08, + "loss": 0.699, + "step": 38195 + }, + { + "epoch": 1.4, + "learning_rate": 6.03068810195188e-08, + "loss": 0.6956, + "step": 38196 + }, + { + "epoch": 1.4, + "learning_rate": 6.030321895484673e-08, + "loss": 0.6965, + "step": 38197 + }, + { + "epoch": 1.4, + "learning_rate": 6.029955689017467e-08, + "loss": 0.6831, + "step": 38198 + }, + { + "epoch": 1.4, + "learning_rate": 6.029589482550261e-08, + "loss": 0.7048, + "step": 38199 + }, + { + "epoch": 1.4, + "learning_rate": 6.029223276083055e-08, + "loss": 0.6972, + "step": 38200 + }, + { + "epoch": 1.4, + "learning_rate": 6.028857069615849e-08, + "loss": 0.6956, + "step": 38201 + }, + { + "epoch": 1.4, + "learning_rate": 6.028490863148643e-08, + "loss": 0.6798, + "step": 38202 + }, + { + "epoch": 1.4, + "learning_rate": 6.028124656681437e-08, + "loss": 0.6932, + "step": 38203 + }, + { + "epoch": 1.4, + "learning_rate": 6.027758450214231e-08, + "loss": 0.6877, + "step": 38204 + }, + { + "epoch": 1.4, + "learning_rate": 6.027392243747025e-08, + "loss": 0.6921, + "step": 38205 + }, + { + "epoch": 1.4, + "learning_rate": 6.027026037279818e-08, + "loss": 0.6967, + "step": 38206 + }, + { + "epoch": 1.4, + "learning_rate": 6.026659830812612e-08, + "loss": 0.6819, + "step": 38207 + }, + { + "epoch": 1.4, + "learning_rate": 6.026293624345406e-08, + "loss": 0.6789, + "step": 38208 + }, + { + "epoch": 1.4, + "learning_rate": 6.0259274178782e-08, + "loss": 0.718, + "step": 38209 + }, + { + "epoch": 1.4, + "learning_rate": 6.025561211410994e-08, + "loss": 0.6918, + "step": 38210 + }, + { + "epoch": 1.4, + "learning_rate": 6.025195004943788e-08, + "loss": 0.6955, + "step": 38211 + }, + { + "epoch": 1.4, + "learning_rate": 6.02482879847658e-08, + "loss": 0.6907, + "step": 38212 + }, + { + "epoch": 1.4, + "learning_rate": 6.024462592009375e-08, + "loss": 0.6988, + "step": 38213 + }, + { + "epoch": 1.4, + "learning_rate": 6.024096385542168e-08, + "loss": 0.704, + "step": 38214 + }, + { + "epoch": 1.4, + "learning_rate": 6.023730179074961e-08, + "loss": 0.6864, + "step": 38215 + }, + { + "epoch": 1.4, + "learning_rate": 6.023363972607755e-08, + "loss": 0.6897, + "step": 38216 + }, + { + "epoch": 1.4, + "learning_rate": 6.022997766140549e-08, + "loss": 0.6901, + "step": 38217 + }, + { + "epoch": 1.4, + "learning_rate": 6.022631559673343e-08, + "loss": 0.6934, + "step": 38218 + }, + { + "epoch": 1.4, + "learning_rate": 6.022265353206137e-08, + "loss": 0.6892, + "step": 38219 + }, + { + "epoch": 1.4, + "learning_rate": 6.021899146738931e-08, + "loss": 0.6962, + "step": 38220 + }, + { + "epoch": 1.4, + "learning_rate": 6.021532940271725e-08, + "loss": 0.6688, + "step": 38221 + }, + { + "epoch": 1.4, + "learning_rate": 6.021166733804519e-08, + "loss": 0.6897, + "step": 38222 + }, + { + "epoch": 1.4, + "learning_rate": 6.020800527337313e-08, + "loss": 0.7014, + "step": 38223 + }, + { + "epoch": 1.4, + "learning_rate": 6.020434320870106e-08, + "loss": 0.6835, + "step": 38224 + }, + { + "epoch": 1.4, + "learning_rate": 6.0200681144029e-08, + "loss": 0.7019, + "step": 38225 + }, + { + "epoch": 1.4, + "learning_rate": 6.019701907935694e-08, + "loss": 0.6941, + "step": 38226 + }, + { + "epoch": 1.4, + "learning_rate": 6.019335701468488e-08, + "loss": 0.6953, + "step": 38227 + }, + { + "epoch": 1.4, + "learning_rate": 6.018969495001282e-08, + "loss": 0.6942, + "step": 38228 + }, + { + "epoch": 1.4, + "learning_rate": 6.018603288534076e-08, + "loss": 0.6978, + "step": 38229 + }, + { + "epoch": 1.4, + "learning_rate": 6.01823708206687e-08, + "loss": 0.6943, + "step": 38230 + }, + { + "epoch": 1.4, + "learning_rate": 6.017870875599664e-08, + "loss": 0.6934, + "step": 38231 + }, + { + "epoch": 1.4, + "learning_rate": 6.017504669132458e-08, + "loss": 0.6865, + "step": 38232 + }, + { + "epoch": 1.4, + "learning_rate": 6.01713846266525e-08, + "loss": 0.6879, + "step": 38233 + }, + { + "epoch": 1.4, + "learning_rate": 6.016772256198044e-08, + "loss": 0.6958, + "step": 38234 + }, + { + "epoch": 1.4, + "learning_rate": 6.016406049730838e-08, + "loss": 0.693, + "step": 38235 + }, + { + "epoch": 1.4, + "learning_rate": 6.016039843263631e-08, + "loss": 0.7013, + "step": 38236 + }, + { + "epoch": 1.4, + "learning_rate": 6.015673636796425e-08, + "loss": 0.6956, + "step": 38237 + }, + { + "epoch": 1.4, + "learning_rate": 6.015307430329219e-08, + "loss": 0.6938, + "step": 38238 + }, + { + "epoch": 1.4, + "learning_rate": 6.014941223862013e-08, + "loss": 0.6926, + "step": 38239 + }, + { + "epoch": 1.4, + "learning_rate": 6.014575017394807e-08, + "loss": 0.7081, + "step": 38240 + }, + { + "epoch": 1.4, + "learning_rate": 6.014208810927601e-08, + "loss": 0.697, + "step": 38241 + }, + { + "epoch": 1.4, + "learning_rate": 6.013842604460394e-08, + "loss": 0.6962, + "step": 38242 + }, + { + "epoch": 1.4, + "learning_rate": 6.013476397993188e-08, + "loss": 0.6871, + "step": 38243 + }, + { + "epoch": 1.4, + "learning_rate": 6.013110191525982e-08, + "loss": 0.6923, + "step": 38244 + }, + { + "epoch": 1.4, + "learning_rate": 6.012743985058776e-08, + "loss": 0.6978, + "step": 38245 + }, + { + "epoch": 1.4, + "learning_rate": 6.01237777859157e-08, + "loss": 0.6945, + "step": 38246 + }, + { + "epoch": 1.4, + "learning_rate": 6.012011572124364e-08, + "loss": 0.6938, + "step": 38247 + }, + { + "epoch": 1.4, + "learning_rate": 6.011645365657158e-08, + "loss": 0.6924, + "step": 38248 + }, + { + "epoch": 1.4, + "learning_rate": 6.011279159189952e-08, + "loss": 0.694, + "step": 38249 + }, + { + "epoch": 1.4, + "learning_rate": 6.010912952722746e-08, + "loss": 0.6873, + "step": 38250 + }, + { + "epoch": 1.4, + "learning_rate": 6.010546746255538e-08, + "loss": 0.695, + "step": 38251 + }, + { + "epoch": 1.4, + "learning_rate": 6.010180539788332e-08, + "loss": 0.6816, + "step": 38252 + }, + { + "epoch": 1.4, + "learning_rate": 6.009814333321126e-08, + "loss": 0.6968, + "step": 38253 + }, + { + "epoch": 1.4, + "learning_rate": 6.00944812685392e-08, + "loss": 0.6805, + "step": 38254 + }, + { + "epoch": 1.4, + "learning_rate": 6.009081920386714e-08, + "loss": 0.6884, + "step": 38255 + }, + { + "epoch": 1.4, + "learning_rate": 6.008715713919508e-08, + "loss": 0.6972, + "step": 38256 + }, + { + "epoch": 1.4, + "learning_rate": 6.008349507452302e-08, + "loss": 0.6946, + "step": 38257 + }, + { + "epoch": 1.4, + "learning_rate": 6.007983300985095e-08, + "loss": 0.6962, + "step": 38258 + }, + { + "epoch": 1.4, + "learning_rate": 6.007617094517889e-08, + "loss": 0.6848, + "step": 38259 + }, + { + "epoch": 1.4, + "learning_rate": 6.007250888050683e-08, + "loss": 0.6875, + "step": 38260 + }, + { + "epoch": 1.4, + "learning_rate": 6.006884681583476e-08, + "loss": 0.6843, + "step": 38261 + }, + { + "epoch": 1.4, + "learning_rate": 6.00651847511627e-08, + "loss": 0.6862, + "step": 38262 + }, + { + "epoch": 1.4, + "learning_rate": 6.006152268649064e-08, + "loss": 0.6974, + "step": 38263 + }, + { + "epoch": 1.4, + "learning_rate": 6.005786062181858e-08, + "loss": 0.6823, + "step": 38264 + }, + { + "epoch": 1.4, + "learning_rate": 6.005419855714652e-08, + "loss": 0.6963, + "step": 38265 + }, + { + "epoch": 1.4, + "learning_rate": 6.005053649247446e-08, + "loss": 0.6965, + "step": 38266 + }, + { + "epoch": 1.4, + "learning_rate": 6.00468744278024e-08, + "loss": 0.6895, + "step": 38267 + }, + { + "epoch": 1.4, + "learning_rate": 6.004321236313034e-08, + "loss": 0.7128, + "step": 38268 + }, + { + "epoch": 1.4, + "learning_rate": 6.003955029845826e-08, + "loss": 0.7053, + "step": 38269 + }, + { + "epoch": 1.4, + "learning_rate": 6.00358882337862e-08, + "loss": 0.6848, + "step": 38270 + }, + { + "epoch": 1.4, + "learning_rate": 6.003222616911414e-08, + "loss": 0.6988, + "step": 38271 + }, + { + "epoch": 1.4, + "learning_rate": 6.002856410444208e-08, + "loss": 0.6987, + "step": 38272 + }, + { + "epoch": 1.4, + "learning_rate": 6.002490203977002e-08, + "loss": 0.6941, + "step": 38273 + }, + { + "epoch": 1.4, + "learning_rate": 6.002123997509796e-08, + "loss": 0.6988, + "step": 38274 + }, + { + "epoch": 1.4, + "learning_rate": 6.00175779104259e-08, + "loss": 0.6944, + "step": 38275 + }, + { + "epoch": 1.4, + "learning_rate": 6.001391584575384e-08, + "loss": 0.6696, + "step": 38276 + }, + { + "epoch": 1.4, + "learning_rate": 6.001025378108178e-08, + "loss": 0.708, + "step": 38277 + }, + { + "epoch": 1.4, + "learning_rate": 6.000659171640971e-08, + "loss": 0.6977, + "step": 38278 + }, + { + "epoch": 1.4, + "learning_rate": 6.000292965173765e-08, + "loss": 0.6905, + "step": 38279 + }, + { + "epoch": 1.4, + "learning_rate": 5.999926758706559e-08, + "loss": 0.6814, + "step": 38280 + }, + { + "epoch": 1.4, + "learning_rate": 5.999560552239353e-08, + "loss": 0.6877, + "step": 38281 + }, + { + "epoch": 1.4, + "learning_rate": 5.999194345772145e-08, + "loss": 0.691, + "step": 38282 + }, + { + "epoch": 1.4, + "learning_rate": 5.99882813930494e-08, + "loss": 0.6885, + "step": 38283 + }, + { + "epoch": 1.4, + "learning_rate": 5.998461932837733e-08, + "loss": 0.6991, + "step": 38284 + }, + { + "epoch": 1.4, + "learning_rate": 5.998095726370527e-08, + "loss": 0.6979, + "step": 38285 + }, + { + "epoch": 1.4, + "learning_rate": 5.997729519903321e-08, + "loss": 0.6894, + "step": 38286 + }, + { + "epoch": 1.4, + "learning_rate": 5.997363313436114e-08, + "loss": 0.704, + "step": 38287 + }, + { + "epoch": 1.4, + "learning_rate": 5.996997106968908e-08, + "loss": 0.6762, + "step": 38288 + }, + { + "epoch": 1.4, + "learning_rate": 5.996630900501702e-08, + "loss": 0.7007, + "step": 38289 + }, + { + "epoch": 1.4, + "learning_rate": 5.996264694034496e-08, + "loss": 0.703, + "step": 38290 + }, + { + "epoch": 1.4, + "learning_rate": 5.99589848756729e-08, + "loss": 0.6766, + "step": 38291 + }, + { + "epoch": 1.4, + "learning_rate": 5.995532281100084e-08, + "loss": 0.6916, + "step": 38292 + }, + { + "epoch": 1.4, + "learning_rate": 5.995166074632878e-08, + "loss": 0.6901, + "step": 38293 + }, + { + "epoch": 1.4, + "learning_rate": 5.994799868165672e-08, + "loss": 0.6778, + "step": 38294 + }, + { + "epoch": 1.4, + "learning_rate": 5.994433661698466e-08, + "loss": 0.6792, + "step": 38295 + }, + { + "epoch": 1.4, + "learning_rate": 5.994067455231259e-08, + "loss": 0.6842, + "step": 38296 + }, + { + "epoch": 1.4, + "learning_rate": 5.993701248764053e-08, + "loss": 0.6931, + "step": 38297 + }, + { + "epoch": 1.4, + "learning_rate": 5.993335042296847e-08, + "loss": 0.7143, + "step": 38298 + }, + { + "epoch": 1.4, + "learning_rate": 5.992968835829641e-08, + "loss": 0.6995, + "step": 38299 + }, + { + "epoch": 1.4, + "learning_rate": 5.992602629362435e-08, + "loss": 0.7154, + "step": 38300 + }, + { + "epoch": 1.4, + "learning_rate": 5.992236422895229e-08, + "loss": 0.695, + "step": 38301 + }, + { + "epoch": 1.4, + "learning_rate": 5.991870216428023e-08, + "loss": 0.6975, + "step": 38302 + }, + { + "epoch": 1.4, + "learning_rate": 5.991504009960817e-08, + "loss": 0.6892, + "step": 38303 + }, + { + "epoch": 1.4, + "learning_rate": 5.991137803493609e-08, + "loss": 0.7108, + "step": 38304 + }, + { + "epoch": 1.4, + "learning_rate": 5.990771597026403e-08, + "loss": 0.698, + "step": 38305 + }, + { + "epoch": 1.4, + "learning_rate": 5.990405390559197e-08, + "loss": 0.6923, + "step": 38306 + }, + { + "epoch": 1.4, + "learning_rate": 5.99003918409199e-08, + "loss": 0.692, + "step": 38307 + }, + { + "epoch": 1.4, + "learning_rate": 5.989672977624784e-08, + "loss": 0.6999, + "step": 38308 + }, + { + "epoch": 1.4, + "learning_rate": 5.989306771157578e-08, + "loss": 0.7012, + "step": 38309 + }, + { + "epoch": 1.4, + "learning_rate": 5.988940564690372e-08, + "loss": 0.6858, + "step": 38310 + }, + { + "epoch": 1.4, + "learning_rate": 5.988574358223166e-08, + "loss": 0.6999, + "step": 38311 + }, + { + "epoch": 1.4, + "learning_rate": 5.98820815175596e-08, + "loss": 0.699, + "step": 38312 + }, + { + "epoch": 1.4, + "learning_rate": 5.987841945288754e-08, + "loss": 0.6925, + "step": 38313 + }, + { + "epoch": 1.4, + "learning_rate": 5.987475738821547e-08, + "loss": 0.6966, + "step": 38314 + }, + { + "epoch": 1.4, + "learning_rate": 5.98710953235434e-08, + "loss": 0.6894, + "step": 38315 + }, + { + "epoch": 1.4, + "learning_rate": 5.986743325887135e-08, + "loss": 0.6996, + "step": 38316 + }, + { + "epoch": 1.4, + "learning_rate": 5.986377119419929e-08, + "loss": 0.6905, + "step": 38317 + }, + { + "epoch": 1.4, + "learning_rate": 5.986010912952723e-08, + "loss": 0.6921, + "step": 38318 + }, + { + "epoch": 1.4, + "learning_rate": 5.985644706485517e-08, + "loss": 0.6981, + "step": 38319 + }, + { + "epoch": 1.4, + "learning_rate": 5.98527850001831e-08, + "loss": 0.7017, + "step": 38320 + }, + { + "epoch": 1.4, + "learning_rate": 5.984912293551105e-08, + "loss": 0.6893, + "step": 38321 + }, + { + "epoch": 1.4, + "learning_rate": 5.984546087083897e-08, + "loss": 0.6799, + "step": 38322 + }, + { + "epoch": 1.4, + "learning_rate": 5.984179880616691e-08, + "loss": 0.6961, + "step": 38323 + }, + { + "epoch": 1.4, + "learning_rate": 5.983813674149485e-08, + "loss": 0.6913, + "step": 38324 + }, + { + "epoch": 1.4, + "learning_rate": 5.983447467682279e-08, + "loss": 0.702, + "step": 38325 + }, + { + "epoch": 1.4, + "learning_rate": 5.983081261215073e-08, + "loss": 0.7091, + "step": 38326 + }, + { + "epoch": 1.4, + "learning_rate": 5.982715054747867e-08, + "loss": 0.69, + "step": 38327 + }, + { + "epoch": 1.4, + "learning_rate": 5.982348848280661e-08, + "loss": 0.6976, + "step": 38328 + }, + { + "epoch": 1.4, + "learning_rate": 5.981982641813454e-08, + "loss": 0.6995, + "step": 38329 + }, + { + "epoch": 1.4, + "learning_rate": 5.981616435346248e-08, + "loss": 0.6745, + "step": 38330 + }, + { + "epoch": 1.4, + "learning_rate": 5.981250228879042e-08, + "loss": 0.6997, + "step": 38331 + }, + { + "epoch": 1.4, + "learning_rate": 5.980884022411835e-08, + "loss": 0.7075, + "step": 38332 + }, + { + "epoch": 1.4, + "learning_rate": 5.980517815944629e-08, + "loss": 0.6862, + "step": 38333 + }, + { + "epoch": 1.4, + "learning_rate": 5.980151609477422e-08, + "loss": 0.6866, + "step": 38334 + }, + { + "epoch": 1.4, + "learning_rate": 5.979785403010216e-08, + "loss": 0.6923, + "step": 38335 + }, + { + "epoch": 1.4, + "learning_rate": 5.97941919654301e-08, + "loss": 0.6927, + "step": 38336 + }, + { + "epoch": 1.4, + "learning_rate": 5.979052990075804e-08, + "loss": 0.6974, + "step": 38337 + }, + { + "epoch": 1.4, + "learning_rate": 5.978686783608598e-08, + "loss": 0.6957, + "step": 38338 + }, + { + "epoch": 1.4, + "learning_rate": 5.978320577141392e-08, + "loss": 0.682, + "step": 38339 + }, + { + "epoch": 1.4, + "learning_rate": 5.977954370674185e-08, + "loss": 0.6822, + "step": 38340 + }, + { + "epoch": 1.4, + "learning_rate": 5.977588164206979e-08, + "loss": 0.6921, + "step": 38341 + }, + { + "epoch": 1.4, + "learning_rate": 5.977221957739773e-08, + "loss": 0.6899, + "step": 38342 + }, + { + "epoch": 1.4, + "learning_rate": 5.976855751272567e-08, + "loss": 0.6993, + "step": 38343 + }, + { + "epoch": 1.4, + "learning_rate": 5.976489544805361e-08, + "loss": 0.7024, + "step": 38344 + }, + { + "epoch": 1.4, + "learning_rate": 5.976123338338155e-08, + "loss": 0.6809, + "step": 38345 + }, + { + "epoch": 1.4, + "learning_rate": 5.975757131870949e-08, + "loss": 0.6871, + "step": 38346 + }, + { + "epoch": 1.4, + "learning_rate": 5.975390925403743e-08, + "loss": 0.6837, + "step": 38347 + }, + { + "epoch": 1.4, + "learning_rate": 5.975024718936537e-08, + "loss": 0.6896, + "step": 38348 + }, + { + "epoch": 1.4, + "learning_rate": 5.97465851246933e-08, + "loss": 0.6805, + "step": 38349 + }, + { + "epoch": 1.4, + "learning_rate": 5.974292306002124e-08, + "loss": 0.7032, + "step": 38350 + }, + { + "epoch": 1.4, + "learning_rate": 5.973926099534918e-08, + "loss": 0.7027, + "step": 38351 + }, + { + "epoch": 1.4, + "learning_rate": 5.973559893067712e-08, + "loss": 0.6979, + "step": 38352 + }, + { + "epoch": 1.4, + "learning_rate": 5.973193686600504e-08, + "loss": 0.6897, + "step": 38353 + }, + { + "epoch": 1.4, + "learning_rate": 5.972827480133298e-08, + "loss": 0.7003, + "step": 38354 + }, + { + "epoch": 1.4, + "learning_rate": 5.972461273666092e-08, + "loss": 0.6744, + "step": 38355 + }, + { + "epoch": 1.4, + "learning_rate": 5.972095067198886e-08, + "loss": 0.7158, + "step": 38356 + }, + { + "epoch": 1.4, + "learning_rate": 5.97172886073168e-08, + "loss": 0.6859, + "step": 38357 + }, + { + "epoch": 1.4, + "learning_rate": 5.971362654264473e-08, + "loss": 0.682, + "step": 38358 + }, + { + "epoch": 1.4, + "learning_rate": 5.970996447797267e-08, + "loss": 0.6996, + "step": 38359 + }, + { + "epoch": 1.4, + "learning_rate": 5.970630241330061e-08, + "loss": 0.6911, + "step": 38360 + }, + { + "epoch": 1.4, + "learning_rate": 5.970264034862855e-08, + "loss": 0.685, + "step": 38361 + }, + { + "epoch": 1.4, + "learning_rate": 5.969897828395649e-08, + "loss": 0.6954, + "step": 38362 + }, + { + "epoch": 1.4, + "learning_rate": 5.969531621928443e-08, + "loss": 0.6927, + "step": 38363 + }, + { + "epoch": 1.4, + "learning_rate": 5.969165415461237e-08, + "loss": 0.6833, + "step": 38364 + }, + { + "epoch": 1.4, + "learning_rate": 5.968799208994031e-08, + "loss": 0.6821, + "step": 38365 + }, + { + "epoch": 1.4, + "learning_rate": 5.968433002526825e-08, + "loss": 0.7012, + "step": 38366 + }, + { + "epoch": 1.4, + "learning_rate": 5.968066796059618e-08, + "loss": 0.6831, + "step": 38367 + }, + { + "epoch": 1.4, + "learning_rate": 5.967700589592412e-08, + "loss": 0.6929, + "step": 38368 + }, + { + "epoch": 1.4, + "learning_rate": 5.967334383125206e-08, + "loss": 0.6983, + "step": 38369 + }, + { + "epoch": 1.4, + "learning_rate": 5.966968176658e-08, + "loss": 0.6878, + "step": 38370 + }, + { + "epoch": 1.4, + "learning_rate": 5.966601970190794e-08, + "loss": 0.7021, + "step": 38371 + }, + { + "epoch": 1.4, + "learning_rate": 5.966235763723588e-08, + "loss": 0.7073, + "step": 38372 + }, + { + "epoch": 1.4, + "learning_rate": 5.965869557256382e-08, + "loss": 0.6839, + "step": 38373 + }, + { + "epoch": 1.4, + "learning_rate": 5.965503350789176e-08, + "loss": 0.6891, + "step": 38374 + }, + { + "epoch": 1.4, + "learning_rate": 5.965137144321968e-08, + "loss": 0.689, + "step": 38375 + }, + { + "epoch": 1.4, + "learning_rate": 5.964770937854762e-08, + "loss": 0.6994, + "step": 38376 + }, + { + "epoch": 1.4, + "learning_rate": 5.964404731387556e-08, + "loss": 0.6784, + "step": 38377 + }, + { + "epoch": 1.4, + "learning_rate": 5.964038524920349e-08, + "loss": 0.6946, + "step": 38378 + }, + { + "epoch": 1.4, + "learning_rate": 5.963672318453143e-08, + "loss": 0.702, + "step": 38379 + }, + { + "epoch": 1.4, + "learning_rate": 5.963306111985937e-08, + "loss": 0.6897, + "step": 38380 + }, + { + "epoch": 1.4, + "learning_rate": 5.962939905518731e-08, + "loss": 0.6899, + "step": 38381 + }, + { + "epoch": 1.4, + "learning_rate": 5.962573699051525e-08, + "loss": 0.7014, + "step": 38382 + }, + { + "epoch": 1.4, + "learning_rate": 5.962207492584319e-08, + "loss": 0.7062, + "step": 38383 + }, + { + "epoch": 1.4, + "learning_rate": 5.961841286117113e-08, + "loss": 0.6889, + "step": 38384 + }, + { + "epoch": 1.4, + "learning_rate": 5.961475079649906e-08, + "loss": 0.6929, + "step": 38385 + }, + { + "epoch": 1.4, + "learning_rate": 5.9611088731827e-08, + "loss": 0.7067, + "step": 38386 + }, + { + "epoch": 1.4, + "learning_rate": 5.960742666715494e-08, + "loss": 0.6916, + "step": 38387 + }, + { + "epoch": 1.4, + "learning_rate": 5.9603764602482875e-08, + "loss": 0.6976, + "step": 38388 + }, + { + "epoch": 1.4, + "learning_rate": 5.9600102537810815e-08, + "loss": 0.6866, + "step": 38389 + }, + { + "epoch": 1.4, + "learning_rate": 5.9596440473138755e-08, + "loss": 0.7074, + "step": 38390 + }, + { + "epoch": 1.4, + "learning_rate": 5.9592778408466695e-08, + "loss": 0.7099, + "step": 38391 + }, + { + "epoch": 1.4, + "learning_rate": 5.9589116343794635e-08, + "loss": 0.6933, + "step": 38392 + }, + { + "epoch": 1.4, + "learning_rate": 5.958545427912257e-08, + "loss": 0.6844, + "step": 38393 + }, + { + "epoch": 1.4, + "learning_rate": 5.95817922144505e-08, + "loss": 0.7005, + "step": 38394 + }, + { + "epoch": 1.4, + "learning_rate": 5.957813014977844e-08, + "loss": 0.6981, + "step": 38395 + }, + { + "epoch": 1.4, + "learning_rate": 5.9574468085106375e-08, + "loss": 0.6927, + "step": 38396 + }, + { + "epoch": 1.4, + "learning_rate": 5.9570806020434315e-08, + "loss": 0.6943, + "step": 38397 + }, + { + "epoch": 1.4, + "learning_rate": 5.9567143955762255e-08, + "loss": 0.6945, + "step": 38398 + }, + { + "epoch": 1.4, + "learning_rate": 5.9563481891090195e-08, + "loss": 0.6897, + "step": 38399 + }, + { + "epoch": 1.4, + "learning_rate": 5.9559819826418135e-08, + "loss": 0.6906, + "step": 38400 + }, + { + "epoch": 1.4, + "learning_rate": 5.9556157761746074e-08, + "loss": 0.6981, + "step": 38401 + }, + { + "epoch": 1.41, + "learning_rate": 5.9552495697074014e-08, + "loss": 0.7044, + "step": 38402 + }, + { + "epoch": 1.41, + "learning_rate": 5.954883363240194e-08, + "loss": 0.7025, + "step": 38403 + }, + { + "epoch": 1.41, + "learning_rate": 5.954517156772988e-08, + "loss": 0.7021, + "step": 38404 + }, + { + "epoch": 1.41, + "learning_rate": 5.954150950305782e-08, + "loss": 0.7001, + "step": 38405 + }, + { + "epoch": 1.41, + "learning_rate": 5.953784743838576e-08, + "loss": 0.6775, + "step": 38406 + }, + { + "epoch": 1.41, + "learning_rate": 5.9534185373713694e-08, + "loss": 0.6703, + "step": 38407 + }, + { + "epoch": 1.41, + "learning_rate": 5.9530523309041634e-08, + "loss": 0.6927, + "step": 38408 + }, + { + "epoch": 1.41, + "learning_rate": 5.9526861244369574e-08, + "loss": 0.7134, + "step": 38409 + }, + { + "epoch": 1.41, + "learning_rate": 5.9523199179697514e-08, + "loss": 0.6885, + "step": 38410 + }, + { + "epoch": 1.41, + "learning_rate": 5.9519537115025454e-08, + "loss": 0.6828, + "step": 38411 + }, + { + "epoch": 1.41, + "learning_rate": 5.951587505035338e-08, + "loss": 0.7001, + "step": 38412 + }, + { + "epoch": 1.41, + "learning_rate": 5.951221298568132e-08, + "loss": 0.6834, + "step": 38413 + }, + { + "epoch": 1.41, + "learning_rate": 5.950855092100926e-08, + "loss": 0.6955, + "step": 38414 + }, + { + "epoch": 1.41, + "learning_rate": 5.95048888563372e-08, + "loss": 0.6871, + "step": 38415 + }, + { + "epoch": 1.41, + "learning_rate": 5.950122679166514e-08, + "loss": 0.6897, + "step": 38416 + }, + { + "epoch": 1.41, + "learning_rate": 5.949756472699308e-08, + "loss": 0.691, + "step": 38417 + }, + { + "epoch": 1.41, + "learning_rate": 5.9493902662321014e-08, + "loss": 0.695, + "step": 38418 + }, + { + "epoch": 1.41, + "learning_rate": 5.9490240597648953e-08, + "loss": 0.6953, + "step": 38419 + }, + { + "epoch": 1.41, + "learning_rate": 5.9486578532976893e-08, + "loss": 0.7071, + "step": 38420 + }, + { + "epoch": 1.41, + "learning_rate": 5.948291646830482e-08, + "loss": 0.6884, + "step": 38421 + }, + { + "epoch": 1.41, + "learning_rate": 5.947925440363276e-08, + "loss": 0.6962, + "step": 38422 + }, + { + "epoch": 1.41, + "learning_rate": 5.94755923389607e-08, + "loss": 0.7018, + "step": 38423 + }, + { + "epoch": 1.41, + "learning_rate": 5.947193027428864e-08, + "loss": 0.6974, + "step": 38424 + }, + { + "epoch": 1.41, + "learning_rate": 5.946826820961658e-08, + "loss": 0.6791, + "step": 38425 + }, + { + "epoch": 1.41, + "learning_rate": 5.946460614494452e-08, + "loss": 0.7049, + "step": 38426 + }, + { + "epoch": 1.41, + "learning_rate": 5.946094408027246e-08, + "loss": 0.6889, + "step": 38427 + }, + { + "epoch": 1.41, + "learning_rate": 5.945728201560039e-08, + "loss": 0.6978, + "step": 38428 + }, + { + "epoch": 1.41, + "learning_rate": 5.945361995092833e-08, + "loss": 0.7085, + "step": 38429 + }, + { + "epoch": 1.41, + "learning_rate": 5.9449957886256266e-08, + "loss": 0.6908, + "step": 38430 + }, + { + "epoch": 1.41, + "learning_rate": 5.94462958215842e-08, + "loss": 0.7026, + "step": 38431 + }, + { + "epoch": 1.41, + "learning_rate": 5.944263375691214e-08, + "loss": 0.6952, + "step": 38432 + }, + { + "epoch": 1.41, + "learning_rate": 5.943897169224008e-08, + "loss": 0.6938, + "step": 38433 + }, + { + "epoch": 1.41, + "learning_rate": 5.943530962756802e-08, + "loss": 0.7067, + "step": 38434 + }, + { + "epoch": 1.41, + "learning_rate": 5.943164756289596e-08, + "loss": 0.6998, + "step": 38435 + }, + { + "epoch": 1.41, + "learning_rate": 5.94279854982239e-08, + "loss": 0.6842, + "step": 38436 + }, + { + "epoch": 1.41, + "learning_rate": 5.942432343355184e-08, + "loss": 0.6932, + "step": 38437 + }, + { + "epoch": 1.41, + "learning_rate": 5.942066136887978e-08, + "loss": 0.6982, + "step": 38438 + }, + { + "epoch": 1.41, + "learning_rate": 5.9416999304207706e-08, + "loss": 0.6669, + "step": 38439 + }, + { + "epoch": 1.41, + "learning_rate": 5.9413337239535646e-08, + "loss": 0.7007, + "step": 38440 + }, + { + "epoch": 1.41, + "learning_rate": 5.9409675174863586e-08, + "loss": 0.6805, + "step": 38441 + }, + { + "epoch": 1.41, + "learning_rate": 5.940601311019152e-08, + "loss": 0.7073, + "step": 38442 + }, + { + "epoch": 1.41, + "learning_rate": 5.940235104551946e-08, + "loss": 0.6853, + "step": 38443 + }, + { + "epoch": 1.41, + "learning_rate": 5.93986889808474e-08, + "loss": 0.6938, + "step": 38444 + }, + { + "epoch": 1.41, + "learning_rate": 5.939502691617534e-08, + "loss": 0.7012, + "step": 38445 + }, + { + "epoch": 1.41, + "learning_rate": 5.939136485150328e-08, + "loss": 0.6706, + "step": 38446 + }, + { + "epoch": 1.41, + "learning_rate": 5.938770278683122e-08, + "loss": 0.702, + "step": 38447 + }, + { + "epoch": 1.41, + "learning_rate": 5.9384040722159145e-08, + "loss": 0.6936, + "step": 38448 + }, + { + "epoch": 1.41, + "learning_rate": 5.9380378657487085e-08, + "loss": 0.708, + "step": 38449 + }, + { + "epoch": 1.41, + "learning_rate": 5.9376716592815025e-08, + "loss": 0.683, + "step": 38450 + }, + { + "epoch": 1.41, + "learning_rate": 5.9373054528142965e-08, + "loss": 0.7058, + "step": 38451 + }, + { + "epoch": 1.41, + "learning_rate": 5.9369392463470905e-08, + "loss": 0.6857, + "step": 38452 + }, + { + "epoch": 1.41, + "learning_rate": 5.936573039879884e-08, + "loss": 0.6769, + "step": 38453 + }, + { + "epoch": 1.41, + "learning_rate": 5.936206833412678e-08, + "loss": 0.6916, + "step": 38454 + }, + { + "epoch": 1.41, + "learning_rate": 5.935840626945472e-08, + "loss": 0.6951, + "step": 38455 + }, + { + "epoch": 1.41, + "learning_rate": 5.935474420478266e-08, + "loss": 0.6967, + "step": 38456 + }, + { + "epoch": 1.41, + "learning_rate": 5.9351082140110585e-08, + "loss": 0.6885, + "step": 38457 + }, + { + "epoch": 1.41, + "learning_rate": 5.9347420075438525e-08, + "loss": 0.7058, + "step": 38458 + }, + { + "epoch": 1.41, + "learning_rate": 5.9343758010766465e-08, + "loss": 0.6998, + "step": 38459 + }, + { + "epoch": 1.41, + "learning_rate": 5.9340095946094404e-08, + "loss": 0.6916, + "step": 38460 + }, + { + "epoch": 1.41, + "learning_rate": 5.9336433881422344e-08, + "loss": 0.6956, + "step": 38461 + }, + { + "epoch": 1.41, + "learning_rate": 5.9332771816750284e-08, + "loss": 0.6849, + "step": 38462 + }, + { + "epoch": 1.41, + "learning_rate": 5.9329109752078224e-08, + "loss": 0.6963, + "step": 38463 + }, + { + "epoch": 1.41, + "learning_rate": 5.932544768740616e-08, + "loss": 0.6883, + "step": 38464 + }, + { + "epoch": 1.41, + "learning_rate": 5.93217856227341e-08, + "loss": 0.6854, + "step": 38465 + }, + { + "epoch": 1.41, + "learning_rate": 5.931812355806203e-08, + "loss": 0.6844, + "step": 38466 + }, + { + "epoch": 1.41, + "learning_rate": 5.9314461493389964e-08, + "loss": 0.6928, + "step": 38467 + }, + { + "epoch": 1.41, + "learning_rate": 5.9310799428717904e-08, + "loss": 0.6837, + "step": 38468 + }, + { + "epoch": 1.41, + "learning_rate": 5.9307137364045844e-08, + "loss": 0.6877, + "step": 38469 + }, + { + "epoch": 1.41, + "learning_rate": 5.9303475299373784e-08, + "loss": 0.6845, + "step": 38470 + }, + { + "epoch": 1.41, + "learning_rate": 5.9299813234701724e-08, + "loss": 0.6981, + "step": 38471 + }, + { + "epoch": 1.41, + "learning_rate": 5.9296151170029664e-08, + "loss": 0.692, + "step": 38472 + }, + { + "epoch": 1.41, + "learning_rate": 5.9292489105357604e-08, + "loss": 0.6898, + "step": 38473 + }, + { + "epoch": 1.41, + "learning_rate": 5.9288827040685544e-08, + "loss": 0.6924, + "step": 38474 + }, + { + "epoch": 1.41, + "learning_rate": 5.928516497601347e-08, + "loss": 0.6841, + "step": 38475 + }, + { + "epoch": 1.41, + "learning_rate": 5.928150291134141e-08, + "loss": 0.6914, + "step": 38476 + }, + { + "epoch": 1.41, + "learning_rate": 5.927784084666935e-08, + "loss": 0.6911, + "step": 38477 + }, + { + "epoch": 1.41, + "learning_rate": 5.9274178781997284e-08, + "loss": 0.7005, + "step": 38478 + }, + { + "epoch": 1.41, + "learning_rate": 5.9270516717325223e-08, + "loss": 0.6932, + "step": 38479 + }, + { + "epoch": 1.41, + "learning_rate": 5.9266854652653163e-08, + "loss": 0.6686, + "step": 38480 + }, + { + "epoch": 1.41, + "learning_rate": 5.92631925879811e-08, + "loss": 0.7029, + "step": 38481 + }, + { + "epoch": 1.41, + "learning_rate": 5.925953052330904e-08, + "loss": 0.6958, + "step": 38482 + }, + { + "epoch": 1.41, + "learning_rate": 5.925586845863698e-08, + "loss": 0.689, + "step": 38483 + }, + { + "epoch": 1.41, + "learning_rate": 5.925220639396491e-08, + "loss": 0.698, + "step": 38484 + }, + { + "epoch": 1.41, + "learning_rate": 5.924854432929285e-08, + "loss": 0.6714, + "step": 38485 + }, + { + "epoch": 1.41, + "learning_rate": 5.924488226462079e-08, + "loss": 0.6849, + "step": 38486 + }, + { + "epoch": 1.41, + "learning_rate": 5.924122019994873e-08, + "loss": 0.6967, + "step": 38487 + }, + { + "epoch": 1.41, + "learning_rate": 5.923755813527666e-08, + "loss": 0.7015, + "step": 38488 + }, + { + "epoch": 1.41, + "learning_rate": 5.92338960706046e-08, + "loss": 0.6893, + "step": 38489 + }, + { + "epoch": 1.41, + "learning_rate": 5.923023400593254e-08, + "loss": 0.6939, + "step": 38490 + }, + { + "epoch": 1.41, + "learning_rate": 5.922657194126048e-08, + "loss": 0.6901, + "step": 38491 + }, + { + "epoch": 1.41, + "learning_rate": 5.922290987658842e-08, + "loss": 0.7046, + "step": 38492 + }, + { + "epoch": 1.41, + "learning_rate": 5.921924781191635e-08, + "loss": 0.6783, + "step": 38493 + }, + { + "epoch": 1.41, + "learning_rate": 5.921558574724429e-08, + "loss": 0.6895, + "step": 38494 + }, + { + "epoch": 1.41, + "learning_rate": 5.921192368257223e-08, + "loss": 0.6929, + "step": 38495 + }, + { + "epoch": 1.41, + "learning_rate": 5.920826161790017e-08, + "loss": 0.6919, + "step": 38496 + }, + { + "epoch": 1.41, + "learning_rate": 5.920459955322811e-08, + "loss": 0.706, + "step": 38497 + }, + { + "epoch": 1.41, + "learning_rate": 5.920093748855605e-08, + "loss": 0.6923, + "step": 38498 + }, + { + "epoch": 1.41, + "learning_rate": 5.919727542388398e-08, + "loss": 0.6894, + "step": 38499 + }, + { + "epoch": 1.41, + "learning_rate": 5.919361335921192e-08, + "loss": 0.6969, + "step": 38500 + }, + { + "epoch": 1.41, + "learning_rate": 5.918995129453986e-08, + "loss": 0.7032, + "step": 38501 + }, + { + "epoch": 1.41, + "learning_rate": 5.918628922986779e-08, + "loss": 0.6909, + "step": 38502 + }, + { + "epoch": 1.41, + "learning_rate": 5.918262716519573e-08, + "loss": 0.7019, + "step": 38503 + }, + { + "epoch": 1.41, + "learning_rate": 5.917896510052367e-08, + "loss": 0.6917, + "step": 38504 + }, + { + "epoch": 1.41, + "learning_rate": 5.917530303585161e-08, + "loss": 0.6919, + "step": 38505 + }, + { + "epoch": 1.41, + "learning_rate": 5.917164097117955e-08, + "loss": 0.7053, + "step": 38506 + }, + { + "epoch": 1.41, + "learning_rate": 5.916797890650749e-08, + "loss": 0.688, + "step": 38507 + }, + { + "epoch": 1.41, + "learning_rate": 5.916431684183543e-08, + "loss": 0.6962, + "step": 38508 + }, + { + "epoch": 1.41, + "learning_rate": 5.916065477716337e-08, + "loss": 0.6762, + "step": 38509 + }, + { + "epoch": 1.41, + "learning_rate": 5.91569927124913e-08, + "loss": 0.708, + "step": 38510 + }, + { + "epoch": 1.41, + "learning_rate": 5.9153330647819235e-08, + "loss": 0.6837, + "step": 38511 + }, + { + "epoch": 1.41, + "learning_rate": 5.9149668583147175e-08, + "loss": 0.7025, + "step": 38512 + }, + { + "epoch": 1.41, + "learning_rate": 5.914600651847511e-08, + "loss": 0.6899, + "step": 38513 + }, + { + "epoch": 1.41, + "learning_rate": 5.914234445380305e-08, + "loss": 0.6925, + "step": 38514 + }, + { + "epoch": 1.41, + "learning_rate": 5.913868238913099e-08, + "loss": 0.7008, + "step": 38515 + }, + { + "epoch": 1.41, + "learning_rate": 5.913502032445893e-08, + "loss": 0.694, + "step": 38516 + }, + { + "epoch": 1.41, + "learning_rate": 5.913135825978687e-08, + "loss": 0.6985, + "step": 38517 + }, + { + "epoch": 1.41, + "learning_rate": 5.912769619511481e-08, + "loss": 0.7075, + "step": 38518 + }, + { + "epoch": 1.41, + "learning_rate": 5.912403413044275e-08, + "loss": 0.6942, + "step": 38519 + }, + { + "epoch": 1.41, + "learning_rate": 5.9120372065770674e-08, + "loss": 0.6735, + "step": 38520 + }, + { + "epoch": 1.41, + "learning_rate": 5.9116710001098614e-08, + "loss": 0.6881, + "step": 38521 + }, + { + "epoch": 1.41, + "learning_rate": 5.9113047936426554e-08, + "loss": 0.6795, + "step": 38522 + }, + { + "epoch": 1.41, + "learning_rate": 5.9109385871754494e-08, + "loss": 0.711, + "step": 38523 + }, + { + "epoch": 1.41, + "learning_rate": 5.910572380708243e-08, + "loss": 0.685, + "step": 38524 + }, + { + "epoch": 1.41, + "learning_rate": 5.910206174241037e-08, + "loss": 0.6883, + "step": 38525 + }, + { + "epoch": 1.41, + "learning_rate": 5.909839967773831e-08, + "loss": 0.6926, + "step": 38526 + }, + { + "epoch": 1.41, + "learning_rate": 5.909473761306625e-08, + "loss": 0.6844, + "step": 38527 + }, + { + "epoch": 1.41, + "learning_rate": 5.909107554839419e-08, + "loss": 0.6948, + "step": 38528 + }, + { + "epoch": 1.41, + "learning_rate": 5.9087413483722114e-08, + "loss": 0.6956, + "step": 38529 + }, + { + "epoch": 1.41, + "learning_rate": 5.9083751419050054e-08, + "loss": 0.6935, + "step": 38530 + }, + { + "epoch": 1.41, + "learning_rate": 5.9080089354377994e-08, + "loss": 0.6978, + "step": 38531 + }, + { + "epoch": 1.41, + "learning_rate": 5.9076427289705934e-08, + "loss": 0.6939, + "step": 38532 + }, + { + "epoch": 1.41, + "learning_rate": 5.9072765225033874e-08, + "loss": 0.6974, + "step": 38533 + }, + { + "epoch": 1.41, + "learning_rate": 5.9069103160361814e-08, + "loss": 0.6934, + "step": 38534 + }, + { + "epoch": 1.41, + "learning_rate": 5.906544109568975e-08, + "loss": 0.6949, + "step": 38535 + }, + { + "epoch": 1.41, + "learning_rate": 5.906177903101769e-08, + "loss": 0.7027, + "step": 38536 + }, + { + "epoch": 1.41, + "learning_rate": 5.905811696634563e-08, + "loss": 0.666, + "step": 38537 + }, + { + "epoch": 1.41, + "learning_rate": 5.9054454901673553e-08, + "loss": 0.7023, + "step": 38538 + }, + { + "epoch": 1.41, + "learning_rate": 5.9050792837001493e-08, + "loss": 0.7038, + "step": 38539 + }, + { + "epoch": 1.41, + "learning_rate": 5.9047130772329433e-08, + "loss": 0.6974, + "step": 38540 + }, + { + "epoch": 1.41, + "learning_rate": 5.904346870765737e-08, + "loss": 0.6943, + "step": 38541 + }, + { + "epoch": 1.41, + "learning_rate": 5.903980664298531e-08, + "loss": 0.7029, + "step": 38542 + }, + { + "epoch": 1.41, + "learning_rate": 5.903614457831325e-08, + "loss": 0.6914, + "step": 38543 + }, + { + "epoch": 1.41, + "learning_rate": 5.903248251364119e-08, + "loss": 0.6829, + "step": 38544 + }, + { + "epoch": 1.41, + "learning_rate": 5.902882044896913e-08, + "loss": 0.7016, + "step": 38545 + }, + { + "epoch": 1.41, + "learning_rate": 5.9025158384297066e-08, + "loss": 0.6739, + "step": 38546 + }, + { + "epoch": 1.41, + "learning_rate": 5.9021496319625e-08, + "loss": 0.6914, + "step": 38547 + }, + { + "epoch": 1.41, + "learning_rate": 5.901783425495293e-08, + "loss": 0.6943, + "step": 38548 + }, + { + "epoch": 1.41, + "learning_rate": 5.901417219028087e-08, + "loss": 0.6986, + "step": 38549 + }, + { + "epoch": 1.41, + "learning_rate": 5.901051012560881e-08, + "loss": 0.6876, + "step": 38550 + }, + { + "epoch": 1.41, + "learning_rate": 5.900684806093675e-08, + "loss": 0.7013, + "step": 38551 + }, + { + "epoch": 1.41, + "learning_rate": 5.900318599626469e-08, + "loss": 0.6869, + "step": 38552 + }, + { + "epoch": 1.41, + "learning_rate": 5.899952393159263e-08, + "loss": 0.6923, + "step": 38553 + }, + { + "epoch": 1.41, + "learning_rate": 5.899586186692057e-08, + "loss": 0.6975, + "step": 38554 + }, + { + "epoch": 1.41, + "learning_rate": 5.899219980224851e-08, + "loss": 0.6944, + "step": 38555 + }, + { + "epoch": 1.41, + "learning_rate": 5.898853773757644e-08, + "loss": 0.6964, + "step": 38556 + }, + { + "epoch": 1.41, + "learning_rate": 5.898487567290438e-08, + "loss": 0.6856, + "step": 38557 + }, + { + "epoch": 1.41, + "learning_rate": 5.898121360823232e-08, + "loss": 0.7068, + "step": 38558 + }, + { + "epoch": 1.41, + "learning_rate": 5.897755154356025e-08, + "loss": 0.6907, + "step": 38559 + }, + { + "epoch": 1.41, + "learning_rate": 5.897388947888819e-08, + "loss": 0.696, + "step": 38560 + }, + { + "epoch": 1.41, + "learning_rate": 5.897022741421613e-08, + "loss": 0.6969, + "step": 38561 + }, + { + "epoch": 1.41, + "learning_rate": 5.896656534954407e-08, + "loss": 0.7023, + "step": 38562 + }, + { + "epoch": 1.41, + "learning_rate": 5.896290328487201e-08, + "loss": 0.6868, + "step": 38563 + }, + { + "epoch": 1.41, + "learning_rate": 5.895924122019995e-08, + "loss": 0.6825, + "step": 38564 + }, + { + "epoch": 1.41, + "learning_rate": 5.895557915552788e-08, + "loss": 0.6934, + "step": 38565 + }, + { + "epoch": 1.41, + "learning_rate": 5.895191709085582e-08, + "loss": 0.6943, + "step": 38566 + }, + { + "epoch": 1.41, + "learning_rate": 5.894825502618376e-08, + "loss": 0.6995, + "step": 38567 + }, + { + "epoch": 1.41, + "learning_rate": 5.89445929615117e-08, + "loss": 0.7028, + "step": 38568 + }, + { + "epoch": 1.41, + "learning_rate": 5.894093089683964e-08, + "loss": 0.6972, + "step": 38569 + }, + { + "epoch": 1.41, + "learning_rate": 5.893726883216757e-08, + "loss": 0.6929, + "step": 38570 + }, + { + "epoch": 1.41, + "learning_rate": 5.893360676749551e-08, + "loss": 0.7017, + "step": 38571 + }, + { + "epoch": 1.41, + "learning_rate": 5.892994470282345e-08, + "loss": 0.6974, + "step": 38572 + }, + { + "epoch": 1.41, + "learning_rate": 5.892628263815139e-08, + "loss": 0.6861, + "step": 38573 + }, + { + "epoch": 1.41, + "learning_rate": 5.892262057347932e-08, + "loss": 0.6818, + "step": 38574 + }, + { + "epoch": 1.41, + "learning_rate": 5.891895850880726e-08, + "loss": 0.6937, + "step": 38575 + }, + { + "epoch": 1.41, + "learning_rate": 5.89152964441352e-08, + "loss": 0.7017, + "step": 38576 + }, + { + "epoch": 1.41, + "learning_rate": 5.891163437946314e-08, + "loss": 0.7061, + "step": 38577 + }, + { + "epoch": 1.41, + "learning_rate": 5.890797231479108e-08, + "loss": 0.6935, + "step": 38578 + }, + { + "epoch": 1.41, + "learning_rate": 5.890431025011902e-08, + "loss": 0.6934, + "step": 38579 + }, + { + "epoch": 1.41, + "learning_rate": 5.890064818544696e-08, + "loss": 0.7033, + "step": 38580 + }, + { + "epoch": 1.41, + "learning_rate": 5.889698612077489e-08, + "loss": 0.7039, + "step": 38581 + }, + { + "epoch": 1.41, + "learning_rate": 5.889332405610283e-08, + "loss": 0.6969, + "step": 38582 + }, + { + "epoch": 1.41, + "learning_rate": 5.8889661991430764e-08, + "loss": 0.6897, + "step": 38583 + }, + { + "epoch": 1.41, + "learning_rate": 5.88859999267587e-08, + "loss": 0.6908, + "step": 38584 + }, + { + "epoch": 1.41, + "learning_rate": 5.888233786208664e-08, + "loss": 0.6956, + "step": 38585 + }, + { + "epoch": 1.41, + "learning_rate": 5.887867579741458e-08, + "loss": 0.7069, + "step": 38586 + }, + { + "epoch": 1.41, + "learning_rate": 5.887501373274252e-08, + "loss": 0.682, + "step": 38587 + }, + { + "epoch": 1.41, + "learning_rate": 5.887135166807046e-08, + "loss": 0.6843, + "step": 38588 + }, + { + "epoch": 1.41, + "learning_rate": 5.88676896033984e-08, + "loss": 0.6989, + "step": 38589 + }, + { + "epoch": 1.41, + "learning_rate": 5.886402753872634e-08, + "loss": 0.6831, + "step": 38590 + }, + { + "epoch": 1.41, + "learning_rate": 5.886036547405428e-08, + "loss": 0.6903, + "step": 38591 + }, + { + "epoch": 1.41, + "learning_rate": 5.8856703409382204e-08, + "loss": 0.6961, + "step": 38592 + }, + { + "epoch": 1.41, + "learning_rate": 5.8853041344710144e-08, + "loss": 0.6824, + "step": 38593 + }, + { + "epoch": 1.41, + "learning_rate": 5.8849379280038084e-08, + "loss": 0.6942, + "step": 38594 + }, + { + "epoch": 1.41, + "learning_rate": 5.884571721536602e-08, + "loss": 0.6906, + "step": 38595 + }, + { + "epoch": 1.41, + "learning_rate": 5.884205515069396e-08, + "loss": 0.7023, + "step": 38596 + }, + { + "epoch": 1.41, + "learning_rate": 5.88383930860219e-08, + "loss": 0.686, + "step": 38597 + }, + { + "epoch": 1.41, + "learning_rate": 5.883473102134984e-08, + "loss": 0.6871, + "step": 38598 + }, + { + "epoch": 1.41, + "learning_rate": 5.8831068956677777e-08, + "loss": 0.689, + "step": 38599 + }, + { + "epoch": 1.41, + "learning_rate": 5.8827406892005717e-08, + "loss": 0.6969, + "step": 38600 + }, + { + "epoch": 1.41, + "learning_rate": 5.882374482733364e-08, + "loss": 0.6903, + "step": 38601 + }, + { + "epoch": 1.41, + "learning_rate": 5.882008276266158e-08, + "loss": 0.6881, + "step": 38602 + }, + { + "epoch": 1.41, + "learning_rate": 5.881642069798952e-08, + "loss": 0.7123, + "step": 38603 + }, + { + "epoch": 1.41, + "learning_rate": 5.881275863331746e-08, + "loss": 0.6903, + "step": 38604 + }, + { + "epoch": 1.41, + "learning_rate": 5.88090965686454e-08, + "loss": 0.6901, + "step": 38605 + }, + { + "epoch": 1.41, + "learning_rate": 5.8805434503973336e-08, + "loss": 0.6691, + "step": 38606 + }, + { + "epoch": 1.41, + "learning_rate": 5.8801772439301276e-08, + "loss": 0.6974, + "step": 38607 + }, + { + "epoch": 1.41, + "learning_rate": 5.8798110374629216e-08, + "loss": 0.6975, + "step": 38608 + }, + { + "epoch": 1.41, + "learning_rate": 5.8794448309957156e-08, + "loss": 0.6916, + "step": 38609 + }, + { + "epoch": 1.41, + "learning_rate": 5.879078624528508e-08, + "loss": 0.6784, + "step": 38610 + }, + { + "epoch": 1.41, + "learning_rate": 5.878712418061302e-08, + "loss": 0.6912, + "step": 38611 + }, + { + "epoch": 1.41, + "learning_rate": 5.878346211594096e-08, + "loss": 0.6981, + "step": 38612 + }, + { + "epoch": 1.41, + "learning_rate": 5.87798000512689e-08, + "loss": 0.6947, + "step": 38613 + }, + { + "epoch": 1.41, + "learning_rate": 5.877613798659684e-08, + "loss": 0.6988, + "step": 38614 + }, + { + "epoch": 1.41, + "learning_rate": 5.877247592192478e-08, + "loss": 0.7028, + "step": 38615 + }, + { + "epoch": 1.41, + "learning_rate": 5.876881385725272e-08, + "loss": 0.6967, + "step": 38616 + }, + { + "epoch": 1.41, + "learning_rate": 5.8765151792580656e-08, + "loss": 0.6804, + "step": 38617 + }, + { + "epoch": 1.41, + "learning_rate": 5.8761489727908596e-08, + "loss": 0.6991, + "step": 38618 + }, + { + "epoch": 1.41, + "learning_rate": 5.875782766323652e-08, + "loss": 0.6821, + "step": 38619 + }, + { + "epoch": 1.41, + "learning_rate": 5.875416559856446e-08, + "loss": 0.7029, + "step": 38620 + }, + { + "epoch": 1.41, + "learning_rate": 5.87505035338924e-08, + "loss": 0.6757, + "step": 38621 + }, + { + "epoch": 1.41, + "learning_rate": 5.874684146922034e-08, + "loss": 0.6915, + "step": 38622 + }, + { + "epoch": 1.41, + "learning_rate": 5.874317940454828e-08, + "loss": 0.7051, + "step": 38623 + }, + { + "epoch": 1.41, + "learning_rate": 5.873951733987622e-08, + "loss": 0.7111, + "step": 38624 + }, + { + "epoch": 1.41, + "learning_rate": 5.873585527520416e-08, + "loss": 0.6917, + "step": 38625 + }, + { + "epoch": 1.41, + "learning_rate": 5.87321932105321e-08, + "loss": 0.6768, + "step": 38626 + }, + { + "epoch": 1.41, + "learning_rate": 5.872853114586004e-08, + "loss": 0.6958, + "step": 38627 + }, + { + "epoch": 1.41, + "learning_rate": 5.872486908118797e-08, + "loss": 0.6848, + "step": 38628 + }, + { + "epoch": 1.41, + "learning_rate": 5.872120701651591e-08, + "loss": 0.6993, + "step": 38629 + }, + { + "epoch": 1.41, + "learning_rate": 5.871754495184384e-08, + "loss": 0.687, + "step": 38630 + }, + { + "epoch": 1.41, + "learning_rate": 5.871388288717178e-08, + "loss": 0.6922, + "step": 38631 + }, + { + "epoch": 1.41, + "learning_rate": 5.871022082249972e-08, + "loss": 0.6951, + "step": 38632 + }, + { + "epoch": 1.41, + "learning_rate": 5.870655875782766e-08, + "loss": 0.6814, + "step": 38633 + }, + { + "epoch": 1.41, + "learning_rate": 5.87028966931556e-08, + "loss": 0.6976, + "step": 38634 + }, + { + "epoch": 1.41, + "learning_rate": 5.869923462848354e-08, + "loss": 0.6835, + "step": 38635 + }, + { + "epoch": 1.41, + "learning_rate": 5.869557256381148e-08, + "loss": 0.6845, + "step": 38636 + }, + { + "epoch": 1.41, + "learning_rate": 5.869191049913941e-08, + "loss": 0.695, + "step": 38637 + }, + { + "epoch": 1.41, + "learning_rate": 5.868824843446735e-08, + "loss": 0.6788, + "step": 38638 + }, + { + "epoch": 1.41, + "learning_rate": 5.868458636979529e-08, + "loss": 0.7059, + "step": 38639 + }, + { + "epoch": 1.41, + "learning_rate": 5.868092430512323e-08, + "loss": 0.6903, + "step": 38640 + }, + { + "epoch": 1.41, + "learning_rate": 5.867726224045116e-08, + "loss": 0.6956, + "step": 38641 + }, + { + "epoch": 1.41, + "learning_rate": 5.86736001757791e-08, + "loss": 0.6869, + "step": 38642 + }, + { + "epoch": 1.41, + "learning_rate": 5.866993811110704e-08, + "loss": 0.7002, + "step": 38643 + }, + { + "epoch": 1.41, + "learning_rate": 5.866627604643498e-08, + "loss": 0.6976, + "step": 38644 + }, + { + "epoch": 1.41, + "learning_rate": 5.866261398176292e-08, + "loss": 0.6927, + "step": 38645 + }, + { + "epoch": 1.41, + "learning_rate": 5.865895191709085e-08, + "loss": 0.6894, + "step": 38646 + }, + { + "epoch": 1.41, + "learning_rate": 5.865528985241879e-08, + "loss": 0.6829, + "step": 38647 + }, + { + "epoch": 1.41, + "learning_rate": 5.865162778774673e-08, + "loss": 0.7145, + "step": 38648 + }, + { + "epoch": 1.41, + "learning_rate": 5.864796572307467e-08, + "loss": 0.6938, + "step": 38649 + }, + { + "epoch": 1.41, + "learning_rate": 5.864430365840261e-08, + "loss": 0.6872, + "step": 38650 + }, + { + "epoch": 1.41, + "learning_rate": 5.864064159373055e-08, + "loss": 0.6899, + "step": 38651 + }, + { + "epoch": 1.41, + "learning_rate": 5.863697952905848e-08, + "loss": 0.6847, + "step": 38652 + }, + { + "epoch": 1.41, + "learning_rate": 5.863331746438642e-08, + "loss": 0.6924, + "step": 38653 + }, + { + "epoch": 1.41, + "learning_rate": 5.862965539971436e-08, + "loss": 0.686, + "step": 38654 + }, + { + "epoch": 1.41, + "learning_rate": 5.862599333504229e-08, + "loss": 0.7007, + "step": 38655 + }, + { + "epoch": 1.41, + "learning_rate": 5.862233127037023e-08, + "loss": 0.6861, + "step": 38656 + }, + { + "epoch": 1.41, + "learning_rate": 5.861866920569817e-08, + "loss": 0.6953, + "step": 38657 + }, + { + "epoch": 1.41, + "learning_rate": 5.861500714102611e-08, + "loss": 0.7023, + "step": 38658 + }, + { + "epoch": 1.41, + "learning_rate": 5.8611345076354047e-08, + "loss": 0.7074, + "step": 38659 + }, + { + "epoch": 1.41, + "learning_rate": 5.8607683011681987e-08, + "loss": 0.6808, + "step": 38660 + }, + { + "epoch": 1.41, + "learning_rate": 5.8604020947009926e-08, + "loss": 0.693, + "step": 38661 + }, + { + "epoch": 1.41, + "learning_rate": 5.8600358882337866e-08, + "loss": 0.6799, + "step": 38662 + }, + { + "epoch": 1.41, + "learning_rate": 5.85966968176658e-08, + "loss": 0.6959, + "step": 38663 + }, + { + "epoch": 1.41, + "learning_rate": 5.859303475299373e-08, + "loss": 0.6988, + "step": 38664 + }, + { + "epoch": 1.41, + "learning_rate": 5.858937268832167e-08, + "loss": 0.6945, + "step": 38665 + }, + { + "epoch": 1.41, + "learning_rate": 5.8585710623649606e-08, + "loss": 0.6958, + "step": 38666 + }, + { + "epoch": 1.41, + "learning_rate": 5.8582048558977546e-08, + "loss": 0.6974, + "step": 38667 + }, + { + "epoch": 1.41, + "learning_rate": 5.8578386494305486e-08, + "loss": 0.6857, + "step": 38668 + }, + { + "epoch": 1.41, + "learning_rate": 5.8574724429633426e-08, + "loss": 0.6773, + "step": 38669 + }, + { + "epoch": 1.41, + "learning_rate": 5.8571062364961366e-08, + "loss": 0.7035, + "step": 38670 + }, + { + "epoch": 1.41, + "learning_rate": 5.8567400300289306e-08, + "loss": 0.6803, + "step": 38671 + }, + { + "epoch": 1.41, + "learning_rate": 5.8563738235617246e-08, + "loss": 0.6993, + "step": 38672 + }, + { + "epoch": 1.41, + "learning_rate": 5.856007617094517e-08, + "loss": 0.6905, + "step": 38673 + }, + { + "epoch": 1.41, + "learning_rate": 5.855641410627311e-08, + "loss": 0.7002, + "step": 38674 + }, + { + "epoch": 1.42, + "learning_rate": 5.855275204160105e-08, + "loss": 0.6942, + "step": 38675 + }, + { + "epoch": 1.42, + "learning_rate": 5.854908997692899e-08, + "loss": 0.6947, + "step": 38676 + }, + { + "epoch": 1.42, + "learning_rate": 5.8545427912256926e-08, + "loss": 0.6995, + "step": 38677 + }, + { + "epoch": 1.42, + "learning_rate": 5.8541765847584866e-08, + "loss": 0.6826, + "step": 38678 + }, + { + "epoch": 1.42, + "learning_rate": 5.8538103782912805e-08, + "loss": 0.6918, + "step": 38679 + }, + { + "epoch": 1.42, + "learning_rate": 5.8534441718240745e-08, + "loss": 0.6933, + "step": 38680 + }, + { + "epoch": 1.42, + "learning_rate": 5.8530779653568685e-08, + "loss": 0.6846, + "step": 38681 + }, + { + "epoch": 1.42, + "learning_rate": 5.852711758889661e-08, + "loss": 0.6916, + "step": 38682 + }, + { + "epoch": 1.42, + "learning_rate": 5.852345552422455e-08, + "loss": 0.7095, + "step": 38683 + }, + { + "epoch": 1.42, + "learning_rate": 5.851979345955249e-08, + "loss": 0.7102, + "step": 38684 + }, + { + "epoch": 1.42, + "learning_rate": 5.851613139488043e-08, + "loss": 0.6953, + "step": 38685 + }, + { + "epoch": 1.42, + "learning_rate": 5.851246933020837e-08, + "loss": 0.6965, + "step": 38686 + }, + { + "epoch": 1.42, + "learning_rate": 5.850880726553631e-08, + "loss": 0.6967, + "step": 38687 + }, + { + "epoch": 1.42, + "learning_rate": 5.8505145200864245e-08, + "loss": 0.6835, + "step": 38688 + }, + { + "epoch": 1.42, + "learning_rate": 5.8501483136192185e-08, + "loss": 0.6946, + "step": 38689 + }, + { + "epoch": 1.42, + "learning_rate": 5.8497821071520125e-08, + "loss": 0.6793, + "step": 38690 + }, + { + "epoch": 1.42, + "learning_rate": 5.849415900684805e-08, + "loss": 0.6855, + "step": 38691 + }, + { + "epoch": 1.42, + "learning_rate": 5.849049694217599e-08, + "loss": 0.6978, + "step": 38692 + }, + { + "epoch": 1.42, + "learning_rate": 5.848683487750393e-08, + "loss": 0.6933, + "step": 38693 + }, + { + "epoch": 1.42, + "learning_rate": 5.848317281283187e-08, + "loss": 0.6873, + "step": 38694 + }, + { + "epoch": 1.42, + "learning_rate": 5.847951074815981e-08, + "loss": 0.6787, + "step": 38695 + }, + { + "epoch": 1.42, + "learning_rate": 5.847584868348775e-08, + "loss": 0.6925, + "step": 38696 + }, + { + "epoch": 1.42, + "learning_rate": 5.847218661881569e-08, + "loss": 0.7016, + "step": 38697 + }, + { + "epoch": 1.42, + "learning_rate": 5.846852455414363e-08, + "loss": 0.6802, + "step": 38698 + }, + { + "epoch": 1.42, + "learning_rate": 5.8464862489471564e-08, + "loss": 0.7014, + "step": 38699 + }, + { + "epoch": 1.42, + "learning_rate": 5.84612004247995e-08, + "loss": 0.6898, + "step": 38700 + }, + { + "epoch": 1.42, + "learning_rate": 5.845753836012743e-08, + "loss": 0.6648, + "step": 38701 + }, + { + "epoch": 1.42, + "learning_rate": 5.845387629545537e-08, + "loss": 0.6801, + "step": 38702 + }, + { + "epoch": 1.42, + "learning_rate": 5.845021423078331e-08, + "loss": 0.6808, + "step": 38703 + }, + { + "epoch": 1.42, + "learning_rate": 5.844655216611125e-08, + "loss": 0.6979, + "step": 38704 + }, + { + "epoch": 1.42, + "learning_rate": 5.844289010143919e-08, + "loss": 0.6923, + "step": 38705 + }, + { + "epoch": 1.42, + "learning_rate": 5.843922803676713e-08, + "loss": 0.6902, + "step": 38706 + }, + { + "epoch": 1.42, + "learning_rate": 5.843556597209507e-08, + "loss": 0.677, + "step": 38707 + }, + { + "epoch": 1.42, + "learning_rate": 5.8431903907423e-08, + "loss": 0.7034, + "step": 38708 + }, + { + "epoch": 1.42, + "learning_rate": 5.842824184275094e-08, + "loss": 0.7115, + "step": 38709 + }, + { + "epoch": 1.42, + "learning_rate": 5.842457977807888e-08, + "loss": 0.7102, + "step": 38710 + }, + { + "epoch": 1.42, + "learning_rate": 5.842091771340682e-08, + "loss": 0.697, + "step": 38711 + }, + { + "epoch": 1.42, + "learning_rate": 5.841725564873475e-08, + "loss": 0.7042, + "step": 38712 + }, + { + "epoch": 1.42, + "learning_rate": 5.841359358406269e-08, + "loss": 0.6785, + "step": 38713 + }, + { + "epoch": 1.42, + "learning_rate": 5.840993151939063e-08, + "loss": 0.6982, + "step": 38714 + }, + { + "epoch": 1.42, + "learning_rate": 5.840626945471857e-08, + "loss": 0.7123, + "step": 38715 + }, + { + "epoch": 1.42, + "learning_rate": 5.840260739004651e-08, + "loss": 0.706, + "step": 38716 + }, + { + "epoch": 1.42, + "learning_rate": 5.839894532537444e-08, + "loss": 0.6905, + "step": 38717 + }, + { + "epoch": 1.42, + "learning_rate": 5.8395283260702377e-08, + "loss": 0.7031, + "step": 38718 + }, + { + "epoch": 1.42, + "learning_rate": 5.8391621196030317e-08, + "loss": 0.7075, + "step": 38719 + }, + { + "epoch": 1.42, + "learning_rate": 5.8387959131358257e-08, + "loss": 0.6914, + "step": 38720 + }, + { + "epoch": 1.42, + "learning_rate": 5.8384297066686196e-08, + "loss": 0.7038, + "step": 38721 + }, + { + "epoch": 1.42, + "learning_rate": 5.8380635002014136e-08, + "loss": 0.7013, + "step": 38722 + }, + { + "epoch": 1.42, + "learning_rate": 5.837697293734207e-08, + "loss": 0.683, + "step": 38723 + }, + { + "epoch": 1.42, + "learning_rate": 5.837331087267001e-08, + "loss": 0.6853, + "step": 38724 + }, + { + "epoch": 1.42, + "learning_rate": 5.836964880799795e-08, + "loss": 0.6868, + "step": 38725 + }, + { + "epoch": 1.42, + "learning_rate": 5.8365986743325876e-08, + "loss": 0.6954, + "step": 38726 + }, + { + "epoch": 1.42, + "learning_rate": 5.8362324678653816e-08, + "loss": 0.6779, + "step": 38727 + }, + { + "epoch": 1.42, + "learning_rate": 5.8358662613981756e-08, + "loss": 0.6887, + "step": 38728 + }, + { + "epoch": 1.42, + "learning_rate": 5.8355000549309696e-08, + "loss": 0.7037, + "step": 38729 + }, + { + "epoch": 1.42, + "learning_rate": 5.8351338484637636e-08, + "loss": 0.6979, + "step": 38730 + }, + { + "epoch": 1.42, + "learning_rate": 5.8347676419965576e-08, + "loss": 0.7025, + "step": 38731 + }, + { + "epoch": 1.42, + "learning_rate": 5.8344014355293516e-08, + "loss": 0.6904, + "step": 38732 + }, + { + "epoch": 1.42, + "learning_rate": 5.8340352290621456e-08, + "loss": 0.6913, + "step": 38733 + }, + { + "epoch": 1.42, + "learning_rate": 5.833669022594939e-08, + "loss": 0.706, + "step": 38734 + }, + { + "epoch": 1.42, + "learning_rate": 5.833302816127732e-08, + "loss": 0.6894, + "step": 38735 + }, + { + "epoch": 1.42, + "learning_rate": 5.832936609660526e-08, + "loss": 0.686, + "step": 38736 + }, + { + "epoch": 1.42, + "learning_rate": 5.8325704031933196e-08, + "loss": 0.694, + "step": 38737 + }, + { + "epoch": 1.42, + "learning_rate": 5.8322041967261136e-08, + "loss": 0.7025, + "step": 38738 + }, + { + "epoch": 1.42, + "learning_rate": 5.8318379902589075e-08, + "loss": 0.6985, + "step": 38739 + }, + { + "epoch": 1.42, + "learning_rate": 5.8314717837917015e-08, + "loss": 0.7024, + "step": 38740 + }, + { + "epoch": 1.42, + "learning_rate": 5.8311055773244955e-08, + "loss": 0.7042, + "step": 38741 + }, + { + "epoch": 1.42, + "learning_rate": 5.8307393708572895e-08, + "loss": 0.6947, + "step": 38742 + }, + { + "epoch": 1.42, + "learning_rate": 5.8303731643900835e-08, + "loss": 0.7014, + "step": 38743 + }, + { + "epoch": 1.42, + "learning_rate": 5.830006957922876e-08, + "loss": 0.6885, + "step": 38744 + }, + { + "epoch": 1.42, + "learning_rate": 5.82964075145567e-08, + "loss": 0.7103, + "step": 38745 + }, + { + "epoch": 1.42, + "learning_rate": 5.829274544988464e-08, + "loss": 0.6915, + "step": 38746 + }, + { + "epoch": 1.42, + "learning_rate": 5.828908338521258e-08, + "loss": 0.6882, + "step": 38747 + }, + { + "epoch": 1.42, + "learning_rate": 5.8285421320540515e-08, + "loss": 0.6944, + "step": 38748 + }, + { + "epoch": 1.42, + "learning_rate": 5.8281759255868455e-08, + "loss": 0.6813, + "step": 38749 + }, + { + "epoch": 1.42, + "learning_rate": 5.8278097191196395e-08, + "loss": 0.6874, + "step": 38750 + }, + { + "epoch": 1.42, + "learning_rate": 5.8274435126524335e-08, + "loss": 0.6766, + "step": 38751 + }, + { + "epoch": 1.42, + "learning_rate": 5.8270773061852275e-08, + "loss": 0.6909, + "step": 38752 + }, + { + "epoch": 1.42, + "learning_rate": 5.82671109971802e-08, + "loss": 0.6918, + "step": 38753 + }, + { + "epoch": 1.42, + "learning_rate": 5.826344893250814e-08, + "loss": 0.7098, + "step": 38754 + }, + { + "epoch": 1.42, + "learning_rate": 5.825978686783608e-08, + "loss": 0.6841, + "step": 38755 + }, + { + "epoch": 1.42, + "learning_rate": 5.825612480316402e-08, + "loss": 0.698, + "step": 38756 + }, + { + "epoch": 1.42, + "learning_rate": 5.825246273849196e-08, + "loss": 0.6878, + "step": 38757 + }, + { + "epoch": 1.42, + "learning_rate": 5.82488006738199e-08, + "loss": 0.6907, + "step": 38758 + }, + { + "epoch": 1.42, + "learning_rate": 5.8245138609147834e-08, + "loss": 0.699, + "step": 38759 + }, + { + "epoch": 1.42, + "learning_rate": 5.8241476544475774e-08, + "loss": 0.6899, + "step": 38760 + }, + { + "epoch": 1.42, + "learning_rate": 5.8237814479803714e-08, + "loss": 0.6913, + "step": 38761 + }, + { + "epoch": 1.42, + "learning_rate": 5.823415241513164e-08, + "loss": 0.6927, + "step": 38762 + }, + { + "epoch": 1.42, + "learning_rate": 5.823049035045958e-08, + "loss": 0.6892, + "step": 38763 + }, + { + "epoch": 1.42, + "learning_rate": 5.822682828578752e-08, + "loss": 0.703, + "step": 38764 + }, + { + "epoch": 1.42, + "learning_rate": 5.822316622111546e-08, + "loss": 0.6873, + "step": 38765 + }, + { + "epoch": 1.42, + "learning_rate": 5.82195041564434e-08, + "loss": 0.6926, + "step": 38766 + }, + { + "epoch": 1.42, + "learning_rate": 5.821584209177134e-08, + "loss": 0.7065, + "step": 38767 + }, + { + "epoch": 1.42, + "learning_rate": 5.821218002709928e-08, + "loss": 0.6859, + "step": 38768 + }, + { + "epoch": 1.42, + "learning_rate": 5.820851796242722e-08, + "loss": 0.6754, + "step": 38769 + }, + { + "epoch": 1.42, + "learning_rate": 5.8204855897755154e-08, + "loss": 0.7157, + "step": 38770 + }, + { + "epoch": 1.42, + "learning_rate": 5.820119383308309e-08, + "loss": 0.6849, + "step": 38771 + }, + { + "epoch": 1.42, + "learning_rate": 5.819753176841102e-08, + "loss": 0.7107, + "step": 38772 + }, + { + "epoch": 1.42, + "learning_rate": 5.819386970373896e-08, + "loss": 0.6903, + "step": 38773 + }, + { + "epoch": 1.42, + "learning_rate": 5.81902076390669e-08, + "loss": 0.6622, + "step": 38774 + }, + { + "epoch": 1.42, + "learning_rate": 5.818654557439484e-08, + "loss": 0.6852, + "step": 38775 + }, + { + "epoch": 1.42, + "learning_rate": 5.818288350972278e-08, + "loss": 0.6996, + "step": 38776 + }, + { + "epoch": 1.42, + "learning_rate": 5.817922144505072e-08, + "loss": 0.6948, + "step": 38777 + }, + { + "epoch": 1.42, + "learning_rate": 5.817555938037866e-08, + "loss": 0.688, + "step": 38778 + }, + { + "epoch": 1.42, + "learning_rate": 5.81718973157066e-08, + "loss": 0.6946, + "step": 38779 + }, + { + "epoch": 1.42, + "learning_rate": 5.8168235251034527e-08, + "loss": 0.6967, + "step": 38780 + }, + { + "epoch": 1.42, + "learning_rate": 5.8164573186362466e-08, + "loss": 0.7039, + "step": 38781 + }, + { + "epoch": 1.42, + "learning_rate": 5.8160911121690406e-08, + "loss": 0.6893, + "step": 38782 + }, + { + "epoch": 1.42, + "learning_rate": 5.815724905701834e-08, + "loss": 0.701, + "step": 38783 + }, + { + "epoch": 1.42, + "learning_rate": 5.815358699234628e-08, + "loss": 0.704, + "step": 38784 + }, + { + "epoch": 1.42, + "learning_rate": 5.814992492767422e-08, + "loss": 0.6943, + "step": 38785 + }, + { + "epoch": 1.42, + "learning_rate": 5.814626286300216e-08, + "loss": 0.696, + "step": 38786 + }, + { + "epoch": 1.42, + "learning_rate": 5.81426007983301e-08, + "loss": 0.6714, + "step": 38787 + }, + { + "epoch": 1.42, + "learning_rate": 5.813893873365804e-08, + "loss": 0.6915, + "step": 38788 + }, + { + "epoch": 1.42, + "learning_rate": 5.8135276668985966e-08, + "loss": 0.6868, + "step": 38789 + }, + { + "epoch": 1.42, + "learning_rate": 5.8131614604313906e-08, + "loss": 0.6963, + "step": 38790 + }, + { + "epoch": 1.42, + "learning_rate": 5.8127952539641846e-08, + "loss": 0.6895, + "step": 38791 + }, + { + "epoch": 1.42, + "learning_rate": 5.8124290474969786e-08, + "loss": 0.6942, + "step": 38792 + }, + { + "epoch": 1.42, + "learning_rate": 5.8120628410297726e-08, + "loss": 0.6953, + "step": 38793 + }, + { + "epoch": 1.42, + "learning_rate": 5.811696634562566e-08, + "loss": 0.6858, + "step": 38794 + }, + { + "epoch": 1.42, + "learning_rate": 5.81133042809536e-08, + "loss": 0.6861, + "step": 38795 + }, + { + "epoch": 1.42, + "learning_rate": 5.810964221628154e-08, + "loss": 0.686, + "step": 38796 + }, + { + "epoch": 1.42, + "learning_rate": 5.810598015160948e-08, + "loss": 0.6862, + "step": 38797 + }, + { + "epoch": 1.42, + "learning_rate": 5.8102318086937406e-08, + "loss": 0.6972, + "step": 38798 + }, + { + "epoch": 1.42, + "learning_rate": 5.8098656022265345e-08, + "loss": 0.6968, + "step": 38799 + }, + { + "epoch": 1.42, + "learning_rate": 5.8094993957593285e-08, + "loss": 0.6959, + "step": 38800 + }, + { + "epoch": 1.42, + "learning_rate": 5.8091331892921225e-08, + "loss": 0.6881, + "step": 38801 + }, + { + "epoch": 1.42, + "learning_rate": 5.8087669828249165e-08, + "loss": 0.7002, + "step": 38802 + }, + { + "epoch": 1.42, + "learning_rate": 5.8084007763577105e-08, + "loss": 0.688, + "step": 38803 + }, + { + "epoch": 1.42, + "learning_rate": 5.8080345698905045e-08, + "loss": 0.6958, + "step": 38804 + }, + { + "epoch": 1.42, + "learning_rate": 5.807668363423298e-08, + "loss": 0.6731, + "step": 38805 + }, + { + "epoch": 1.42, + "learning_rate": 5.807302156956092e-08, + "loss": 0.6975, + "step": 38806 + }, + { + "epoch": 1.42, + "learning_rate": 5.806935950488885e-08, + "loss": 0.6972, + "step": 38807 + }, + { + "epoch": 1.42, + "learning_rate": 5.8065697440216785e-08, + "loss": 0.6885, + "step": 38808 + }, + { + "epoch": 1.42, + "learning_rate": 5.8062035375544725e-08, + "loss": 0.695, + "step": 38809 + }, + { + "epoch": 1.42, + "learning_rate": 5.8058373310872665e-08, + "loss": 0.7001, + "step": 38810 + }, + { + "epoch": 1.42, + "learning_rate": 5.8054711246200605e-08, + "loss": 0.6916, + "step": 38811 + }, + { + "epoch": 1.42, + "learning_rate": 5.8051049181528545e-08, + "loss": 0.7093, + "step": 38812 + }, + { + "epoch": 1.42, + "learning_rate": 5.8047387116856485e-08, + "loss": 0.6942, + "step": 38813 + }, + { + "epoch": 1.42, + "learning_rate": 5.8043725052184425e-08, + "loss": 0.6822, + "step": 38814 + }, + { + "epoch": 1.42, + "learning_rate": 5.8040062987512364e-08, + "loss": 0.6835, + "step": 38815 + }, + { + "epoch": 1.42, + "learning_rate": 5.803640092284029e-08, + "loss": 0.6975, + "step": 38816 + }, + { + "epoch": 1.42, + "learning_rate": 5.803273885816823e-08, + "loss": 0.6944, + "step": 38817 + }, + { + "epoch": 1.42, + "learning_rate": 5.802907679349617e-08, + "loss": 0.6866, + "step": 38818 + }, + { + "epoch": 1.42, + "learning_rate": 5.8025414728824104e-08, + "loss": 0.6948, + "step": 38819 + }, + { + "epoch": 1.42, + "learning_rate": 5.8021752664152044e-08, + "loss": 0.6898, + "step": 38820 + }, + { + "epoch": 1.42, + "learning_rate": 5.8018090599479984e-08, + "loss": 0.6817, + "step": 38821 + }, + { + "epoch": 1.42, + "learning_rate": 5.8014428534807924e-08, + "loss": 0.6911, + "step": 38822 + }, + { + "epoch": 1.42, + "learning_rate": 5.8010766470135864e-08, + "loss": 0.6997, + "step": 38823 + }, + { + "epoch": 1.42, + "learning_rate": 5.8007104405463804e-08, + "loss": 0.7093, + "step": 38824 + }, + { + "epoch": 1.42, + "learning_rate": 5.800344234079173e-08, + "loss": 0.6858, + "step": 38825 + }, + { + "epoch": 1.42, + "learning_rate": 5.799978027611967e-08, + "loss": 0.7108, + "step": 38826 + }, + { + "epoch": 1.42, + "learning_rate": 5.799611821144761e-08, + "loss": 0.6952, + "step": 38827 + }, + { + "epoch": 1.42, + "learning_rate": 5.799245614677555e-08, + "loss": 0.703, + "step": 38828 + }, + { + "epoch": 1.42, + "learning_rate": 5.798879408210349e-08, + "loss": 0.69, + "step": 38829 + }, + { + "epoch": 1.42, + "learning_rate": 5.7985132017431424e-08, + "loss": 0.6959, + "step": 38830 + }, + { + "epoch": 1.42, + "learning_rate": 5.7981469952759364e-08, + "loss": 0.7184, + "step": 38831 + }, + { + "epoch": 1.42, + "learning_rate": 5.7977807888087304e-08, + "loss": 0.7034, + "step": 38832 + }, + { + "epoch": 1.42, + "learning_rate": 5.7974145823415243e-08, + "loss": 0.6954, + "step": 38833 + }, + { + "epoch": 1.42, + "learning_rate": 5.797048375874317e-08, + "loss": 0.7062, + "step": 38834 + }, + { + "epoch": 1.42, + "learning_rate": 5.796682169407111e-08, + "loss": 0.6896, + "step": 38835 + }, + { + "epoch": 1.42, + "learning_rate": 5.796315962939905e-08, + "loss": 0.6953, + "step": 38836 + }, + { + "epoch": 1.42, + "learning_rate": 5.795949756472699e-08, + "loss": 0.7054, + "step": 38837 + }, + { + "epoch": 1.42, + "learning_rate": 5.795583550005493e-08, + "loss": 0.6872, + "step": 38838 + }, + { + "epoch": 1.42, + "learning_rate": 5.795217343538287e-08, + "loss": 0.6823, + "step": 38839 + }, + { + "epoch": 1.42, + "learning_rate": 5.79485113707108e-08, + "loss": 0.6868, + "step": 38840 + }, + { + "epoch": 1.42, + "learning_rate": 5.794484930603874e-08, + "loss": 0.6858, + "step": 38841 + }, + { + "epoch": 1.42, + "learning_rate": 5.794118724136668e-08, + "loss": 0.6781, + "step": 38842 + }, + { + "epoch": 1.42, + "learning_rate": 5.793752517669461e-08, + "loss": 0.6821, + "step": 38843 + }, + { + "epoch": 1.42, + "learning_rate": 5.793386311202255e-08, + "loss": 0.6974, + "step": 38844 + }, + { + "epoch": 1.42, + "learning_rate": 5.793020104735049e-08, + "loss": 0.6815, + "step": 38845 + }, + { + "epoch": 1.42, + "learning_rate": 5.792653898267843e-08, + "loss": 0.699, + "step": 38846 + }, + { + "epoch": 1.42, + "learning_rate": 5.792287691800637e-08, + "loss": 0.7092, + "step": 38847 + }, + { + "epoch": 1.42, + "learning_rate": 5.791921485333431e-08, + "loss": 0.6901, + "step": 38848 + }, + { + "epoch": 1.42, + "learning_rate": 5.791555278866225e-08, + "loss": 0.7036, + "step": 38849 + }, + { + "epoch": 1.42, + "learning_rate": 5.791189072399019e-08, + "loss": 0.6871, + "step": 38850 + }, + { + "epoch": 1.42, + "learning_rate": 5.790822865931812e-08, + "loss": 0.6957, + "step": 38851 + }, + { + "epoch": 1.42, + "learning_rate": 5.7904566594646056e-08, + "loss": 0.6991, + "step": 38852 + }, + { + "epoch": 1.42, + "learning_rate": 5.7900904529973996e-08, + "loss": 0.7052, + "step": 38853 + }, + { + "epoch": 1.42, + "learning_rate": 5.789724246530193e-08, + "loss": 0.6821, + "step": 38854 + }, + { + "epoch": 1.42, + "learning_rate": 5.789358040062987e-08, + "loss": 0.6977, + "step": 38855 + }, + { + "epoch": 1.42, + "learning_rate": 5.788991833595781e-08, + "loss": 0.6786, + "step": 38856 + }, + { + "epoch": 1.42, + "learning_rate": 5.788625627128575e-08, + "loss": 0.6914, + "step": 38857 + }, + { + "epoch": 1.42, + "learning_rate": 5.788259420661369e-08, + "loss": 0.6901, + "step": 38858 + }, + { + "epoch": 1.42, + "learning_rate": 5.787893214194163e-08, + "loss": 0.6896, + "step": 38859 + }, + { + "epoch": 1.42, + "learning_rate": 5.787527007726957e-08, + "loss": 0.681, + "step": 38860 + }, + { + "epoch": 1.42, + "learning_rate": 5.7871608012597495e-08, + "loss": 0.6915, + "step": 38861 + }, + { + "epoch": 1.42, + "learning_rate": 5.7867945947925435e-08, + "loss": 0.6894, + "step": 38862 + }, + { + "epoch": 1.42, + "learning_rate": 5.7864283883253375e-08, + "loss": 0.6718, + "step": 38863 + }, + { + "epoch": 1.42, + "learning_rate": 5.7860621818581315e-08, + "loss": 0.6862, + "step": 38864 + }, + { + "epoch": 1.42, + "learning_rate": 5.785695975390925e-08, + "loss": 0.6831, + "step": 38865 + }, + { + "epoch": 1.42, + "learning_rate": 5.785329768923719e-08, + "loss": 0.69, + "step": 38866 + }, + { + "epoch": 1.42, + "learning_rate": 5.784963562456513e-08, + "loss": 0.6951, + "step": 38867 + }, + { + "epoch": 1.42, + "learning_rate": 5.784597355989307e-08, + "loss": 0.6888, + "step": 38868 + }, + { + "epoch": 1.42, + "learning_rate": 5.784231149522101e-08, + "loss": 0.7044, + "step": 38869 + }, + { + "epoch": 1.42, + "learning_rate": 5.7838649430548935e-08, + "loss": 0.6973, + "step": 38870 + }, + { + "epoch": 1.42, + "learning_rate": 5.7834987365876875e-08, + "loss": 0.7154, + "step": 38871 + }, + { + "epoch": 1.42, + "learning_rate": 5.7831325301204815e-08, + "loss": 0.6847, + "step": 38872 + }, + { + "epoch": 1.42, + "learning_rate": 5.7827663236532755e-08, + "loss": 0.6883, + "step": 38873 + }, + { + "epoch": 1.42, + "learning_rate": 5.7824001171860695e-08, + "loss": 0.6929, + "step": 38874 + }, + { + "epoch": 1.42, + "learning_rate": 5.7820339107188634e-08, + "loss": 0.6927, + "step": 38875 + }, + { + "epoch": 1.42, + "learning_rate": 5.781667704251657e-08, + "loss": 0.6906, + "step": 38876 + }, + { + "epoch": 1.42, + "learning_rate": 5.781301497784451e-08, + "loss": 0.6968, + "step": 38877 + }, + { + "epoch": 1.42, + "learning_rate": 5.780935291317245e-08, + "loss": 0.6978, + "step": 38878 + }, + { + "epoch": 1.42, + "learning_rate": 5.7805690848500374e-08, + "loss": 0.6828, + "step": 38879 + }, + { + "epoch": 1.42, + "learning_rate": 5.7802028783828314e-08, + "loss": 0.7014, + "step": 38880 + }, + { + "epoch": 1.42, + "learning_rate": 5.7798366719156254e-08, + "loss": 0.6907, + "step": 38881 + }, + { + "epoch": 1.42, + "learning_rate": 5.7794704654484194e-08, + "loss": 0.6964, + "step": 38882 + }, + { + "epoch": 1.42, + "learning_rate": 5.7791042589812134e-08, + "loss": 0.6961, + "step": 38883 + }, + { + "epoch": 1.42, + "learning_rate": 5.7787380525140074e-08, + "loss": 0.6801, + "step": 38884 + }, + { + "epoch": 1.42, + "learning_rate": 5.7783718460468014e-08, + "loss": 0.6901, + "step": 38885 + }, + { + "epoch": 1.42, + "learning_rate": 5.7780056395795954e-08, + "loss": 0.6922, + "step": 38886 + }, + { + "epoch": 1.42, + "learning_rate": 5.777639433112389e-08, + "loss": 0.6765, + "step": 38887 + }, + { + "epoch": 1.42, + "learning_rate": 5.777273226645182e-08, + "loss": 0.6877, + "step": 38888 + }, + { + "epoch": 1.42, + "learning_rate": 5.776907020177976e-08, + "loss": 0.7051, + "step": 38889 + }, + { + "epoch": 1.42, + "learning_rate": 5.7765408137107694e-08, + "loss": 0.6983, + "step": 38890 + }, + { + "epoch": 1.42, + "learning_rate": 5.7761746072435634e-08, + "loss": 0.6895, + "step": 38891 + }, + { + "epoch": 1.42, + "learning_rate": 5.7758084007763574e-08, + "loss": 0.6835, + "step": 38892 + }, + { + "epoch": 1.42, + "learning_rate": 5.7754421943091513e-08, + "loss": 0.6895, + "step": 38893 + }, + { + "epoch": 1.42, + "learning_rate": 5.7750759878419453e-08, + "loss": 0.6712, + "step": 38894 + }, + { + "epoch": 1.42, + "learning_rate": 5.7747097813747393e-08, + "loss": 0.7075, + "step": 38895 + }, + { + "epoch": 1.42, + "learning_rate": 5.774343574907533e-08, + "loss": 0.6863, + "step": 38896 + }, + { + "epoch": 1.42, + "learning_rate": 5.773977368440326e-08, + "loss": 0.7126, + "step": 38897 + }, + { + "epoch": 1.42, + "learning_rate": 5.77361116197312e-08, + "loss": 0.6945, + "step": 38898 + }, + { + "epoch": 1.42, + "learning_rate": 5.773244955505914e-08, + "loss": 0.689, + "step": 38899 + }, + { + "epoch": 1.42, + "learning_rate": 5.772878749038707e-08, + "loss": 0.7016, + "step": 38900 + }, + { + "epoch": 1.42, + "learning_rate": 5.772512542571501e-08, + "loss": 0.69, + "step": 38901 + }, + { + "epoch": 1.42, + "learning_rate": 5.772146336104295e-08, + "loss": 0.6961, + "step": 38902 + }, + { + "epoch": 1.42, + "learning_rate": 5.771780129637089e-08, + "loss": 0.6888, + "step": 38903 + }, + { + "epoch": 1.42, + "learning_rate": 5.771413923169883e-08, + "loss": 0.7006, + "step": 38904 + }, + { + "epoch": 1.42, + "learning_rate": 5.771047716702677e-08, + "loss": 0.6748, + "step": 38905 + }, + { + "epoch": 1.42, + "learning_rate": 5.77068151023547e-08, + "loss": 0.6962, + "step": 38906 + }, + { + "epoch": 1.42, + "learning_rate": 5.770315303768264e-08, + "loss": 0.7031, + "step": 38907 + }, + { + "epoch": 1.42, + "learning_rate": 5.769949097301058e-08, + "loss": 0.6757, + "step": 38908 + }, + { + "epoch": 1.42, + "learning_rate": 5.769582890833852e-08, + "loss": 0.6934, + "step": 38909 + }, + { + "epoch": 1.42, + "learning_rate": 5.769216684366646e-08, + "loss": 0.6958, + "step": 38910 + }, + { + "epoch": 1.42, + "learning_rate": 5.768850477899439e-08, + "loss": 0.6867, + "step": 38911 + }, + { + "epoch": 1.42, + "learning_rate": 5.768484271432233e-08, + "loss": 0.7036, + "step": 38912 + }, + { + "epoch": 1.42, + "learning_rate": 5.768118064965027e-08, + "loss": 0.6841, + "step": 38913 + }, + { + "epoch": 1.42, + "learning_rate": 5.767751858497821e-08, + "loss": 0.7074, + "step": 38914 + }, + { + "epoch": 1.42, + "learning_rate": 5.767385652030614e-08, + "loss": 0.7035, + "step": 38915 + }, + { + "epoch": 1.42, + "learning_rate": 5.767019445563408e-08, + "loss": 0.6873, + "step": 38916 + }, + { + "epoch": 1.42, + "learning_rate": 5.766653239096202e-08, + "loss": 0.7048, + "step": 38917 + }, + { + "epoch": 1.42, + "learning_rate": 5.766287032628996e-08, + "loss": 0.6809, + "step": 38918 + }, + { + "epoch": 1.42, + "learning_rate": 5.76592082616179e-08, + "loss": 0.6787, + "step": 38919 + }, + { + "epoch": 1.42, + "learning_rate": 5.765554619694584e-08, + "loss": 0.7124, + "step": 38920 + }, + { + "epoch": 1.42, + "learning_rate": 5.765188413227378e-08, + "loss": 0.7001, + "step": 38921 + }, + { + "epoch": 1.42, + "learning_rate": 5.764822206760171e-08, + "loss": 0.6859, + "step": 38922 + }, + { + "epoch": 1.42, + "learning_rate": 5.764456000292965e-08, + "loss": 0.6908, + "step": 38923 + }, + { + "epoch": 1.42, + "learning_rate": 5.7640897938257585e-08, + "loss": 0.7056, + "step": 38924 + }, + { + "epoch": 1.42, + "learning_rate": 5.763723587358552e-08, + "loss": 0.696, + "step": 38925 + }, + { + "epoch": 1.42, + "learning_rate": 5.763357380891346e-08, + "loss": 0.7033, + "step": 38926 + }, + { + "epoch": 1.42, + "learning_rate": 5.76299117442414e-08, + "loss": 0.6843, + "step": 38927 + }, + { + "epoch": 1.42, + "learning_rate": 5.762624967956934e-08, + "loss": 0.6961, + "step": 38928 + }, + { + "epoch": 1.42, + "learning_rate": 5.762258761489728e-08, + "loss": 0.7124, + "step": 38929 + }, + { + "epoch": 1.42, + "learning_rate": 5.761892555022522e-08, + "loss": 0.6999, + "step": 38930 + }, + { + "epoch": 1.42, + "learning_rate": 5.761526348555316e-08, + "loss": 0.6904, + "step": 38931 + }, + { + "epoch": 1.42, + "learning_rate": 5.76116014208811e-08, + "loss": 0.7088, + "step": 38932 + }, + { + "epoch": 1.42, + "learning_rate": 5.7607939356209025e-08, + "loss": 0.6975, + "step": 38933 + }, + { + "epoch": 1.42, + "learning_rate": 5.7604277291536965e-08, + "loss": 0.6826, + "step": 38934 + }, + { + "epoch": 1.42, + "learning_rate": 5.7600615226864904e-08, + "loss": 0.6994, + "step": 38935 + }, + { + "epoch": 1.42, + "learning_rate": 5.759695316219284e-08, + "loss": 0.688, + "step": 38936 + }, + { + "epoch": 1.42, + "learning_rate": 5.759329109752078e-08, + "loss": 0.6757, + "step": 38937 + }, + { + "epoch": 1.42, + "learning_rate": 5.758962903284872e-08, + "loss": 0.6831, + "step": 38938 + }, + { + "epoch": 1.42, + "learning_rate": 5.758596696817666e-08, + "loss": 0.6846, + "step": 38939 + }, + { + "epoch": 1.42, + "learning_rate": 5.75823049035046e-08, + "loss": 0.7121, + "step": 38940 + }, + { + "epoch": 1.42, + "learning_rate": 5.757864283883254e-08, + "loss": 0.7005, + "step": 38941 + }, + { + "epoch": 1.42, + "learning_rate": 5.7574980774160464e-08, + "loss": 0.6938, + "step": 38942 + }, + { + "epoch": 1.42, + "learning_rate": 5.7571318709488404e-08, + "loss": 0.6978, + "step": 38943 + }, + { + "epoch": 1.42, + "learning_rate": 5.7567656644816344e-08, + "loss": 0.6932, + "step": 38944 + }, + { + "epoch": 1.42, + "learning_rate": 5.7563994580144284e-08, + "loss": 0.692, + "step": 38945 + }, + { + "epoch": 1.42, + "learning_rate": 5.7560332515472224e-08, + "loss": 0.6914, + "step": 38946 + }, + { + "epoch": 1.42, + "learning_rate": 5.755667045080016e-08, + "loss": 0.6863, + "step": 38947 + }, + { + "epoch": 1.42, + "learning_rate": 5.75530083861281e-08, + "loss": 0.6992, + "step": 38948 + }, + { + "epoch": 1.43, + "learning_rate": 5.754934632145604e-08, + "loss": 0.6921, + "step": 38949 + }, + { + "epoch": 1.43, + "learning_rate": 5.754568425678398e-08, + "loss": 0.6984, + "step": 38950 + }, + { + "epoch": 1.43, + "learning_rate": 5.7542022192111904e-08, + "loss": 0.6767, + "step": 38951 + }, + { + "epoch": 1.43, + "learning_rate": 5.7538360127439844e-08, + "loss": 0.6997, + "step": 38952 + }, + { + "epoch": 1.43, + "learning_rate": 5.7534698062767783e-08, + "loss": 0.6798, + "step": 38953 + }, + { + "epoch": 1.43, + "learning_rate": 5.7531035998095723e-08, + "loss": 0.7043, + "step": 38954 + }, + { + "epoch": 1.43, + "learning_rate": 5.752737393342366e-08, + "loss": 0.6818, + "step": 38955 + }, + { + "epoch": 1.43, + "learning_rate": 5.75237118687516e-08, + "loss": 0.6926, + "step": 38956 + }, + { + "epoch": 1.43, + "learning_rate": 5.752004980407954e-08, + "loss": 0.6951, + "step": 38957 + }, + { + "epoch": 1.43, + "learning_rate": 5.7516387739407476e-08, + "loss": 0.6977, + "step": 38958 + }, + { + "epoch": 1.43, + "learning_rate": 5.7512725674735416e-08, + "loss": 0.6982, + "step": 38959 + }, + { + "epoch": 1.43, + "learning_rate": 5.750906361006334e-08, + "loss": 0.7012, + "step": 38960 + }, + { + "epoch": 1.43, + "learning_rate": 5.750540154539128e-08, + "loss": 0.6864, + "step": 38961 + }, + { + "epoch": 1.43, + "learning_rate": 5.750173948071922e-08, + "loss": 0.686, + "step": 38962 + }, + { + "epoch": 1.43, + "learning_rate": 5.749807741604716e-08, + "loss": 0.6972, + "step": 38963 + }, + { + "epoch": 1.43, + "learning_rate": 5.74944153513751e-08, + "loss": 0.6846, + "step": 38964 + }, + { + "epoch": 1.43, + "learning_rate": 5.749075328670304e-08, + "loss": 0.6951, + "step": 38965 + }, + { + "epoch": 1.43, + "learning_rate": 5.748709122203098e-08, + "loss": 0.6737, + "step": 38966 + }, + { + "epoch": 1.43, + "learning_rate": 5.748342915735892e-08, + "loss": 0.7101, + "step": 38967 + }, + { + "epoch": 1.43, + "learning_rate": 5.747976709268686e-08, + "loss": 0.6871, + "step": 38968 + }, + { + "epoch": 1.43, + "learning_rate": 5.747610502801479e-08, + "loss": 0.704, + "step": 38969 + }, + { + "epoch": 1.43, + "learning_rate": 5.747244296334273e-08, + "loss": 0.6861, + "step": 38970 + }, + { + "epoch": 1.43, + "learning_rate": 5.746878089867066e-08, + "loss": 0.6754, + "step": 38971 + }, + { + "epoch": 1.43, + "learning_rate": 5.74651188339986e-08, + "loss": 0.7082, + "step": 38972 + }, + { + "epoch": 1.43, + "learning_rate": 5.746145676932654e-08, + "loss": 0.7231, + "step": 38973 + }, + { + "epoch": 1.43, + "learning_rate": 5.745779470465448e-08, + "loss": 0.6922, + "step": 38974 + }, + { + "epoch": 1.43, + "learning_rate": 5.745413263998242e-08, + "loss": 0.6863, + "step": 38975 + }, + { + "epoch": 1.43, + "learning_rate": 5.745047057531036e-08, + "loss": 0.6922, + "step": 38976 + }, + { + "epoch": 1.43, + "learning_rate": 5.74468085106383e-08, + "loss": 0.6862, + "step": 38977 + }, + { + "epoch": 1.43, + "learning_rate": 5.744314644596623e-08, + "loss": 0.6885, + "step": 38978 + }, + { + "epoch": 1.43, + "learning_rate": 5.743948438129417e-08, + "loss": 0.6929, + "step": 38979 + }, + { + "epoch": 1.43, + "learning_rate": 5.743582231662211e-08, + "loss": 0.6864, + "step": 38980 + }, + { + "epoch": 1.43, + "learning_rate": 5.743216025195005e-08, + "loss": 0.6954, + "step": 38981 + }, + { + "epoch": 1.43, + "learning_rate": 5.742849818727798e-08, + "loss": 0.7127, + "step": 38982 + }, + { + "epoch": 1.43, + "learning_rate": 5.742483612260592e-08, + "loss": 0.7008, + "step": 38983 + }, + { + "epoch": 1.43, + "learning_rate": 5.742117405793386e-08, + "loss": 0.7036, + "step": 38984 + }, + { + "epoch": 1.43, + "learning_rate": 5.74175119932618e-08, + "loss": 0.6963, + "step": 38985 + }, + { + "epoch": 1.43, + "learning_rate": 5.741384992858974e-08, + "loss": 0.6888, + "step": 38986 + }, + { + "epoch": 1.43, + "learning_rate": 5.741018786391767e-08, + "loss": 0.6956, + "step": 38987 + }, + { + "epoch": 1.43, + "learning_rate": 5.740652579924561e-08, + "loss": 0.714, + "step": 38988 + }, + { + "epoch": 1.43, + "learning_rate": 5.740286373457355e-08, + "loss": 0.6841, + "step": 38989 + }, + { + "epoch": 1.43, + "learning_rate": 5.739920166990149e-08, + "loss": 0.7092, + "step": 38990 + }, + { + "epoch": 1.43, + "learning_rate": 5.739553960522943e-08, + "loss": 0.684, + "step": 38991 + }, + { + "epoch": 1.43, + "learning_rate": 5.739187754055737e-08, + "loss": 0.6955, + "step": 38992 + }, + { + "epoch": 1.43, + "learning_rate": 5.73882154758853e-08, + "loss": 0.6945, + "step": 38993 + }, + { + "epoch": 1.43, + "learning_rate": 5.738455341121324e-08, + "loss": 0.6897, + "step": 38994 + }, + { + "epoch": 1.43, + "learning_rate": 5.738089134654118e-08, + "loss": 0.7011, + "step": 38995 + }, + { + "epoch": 1.43, + "learning_rate": 5.737722928186911e-08, + "loss": 0.7021, + "step": 38996 + }, + { + "epoch": 1.43, + "learning_rate": 5.737356721719705e-08, + "loss": 0.6955, + "step": 38997 + }, + { + "epoch": 1.43, + "learning_rate": 5.736990515252499e-08, + "loss": 0.7041, + "step": 38998 + }, + { + "epoch": 1.43, + "learning_rate": 5.736624308785293e-08, + "loss": 0.6896, + "step": 38999 + }, + { + "epoch": 1.43, + "learning_rate": 5.736258102318087e-08, + "loss": 0.6915, + "step": 39000 + }, + { + "epoch": 1.43, + "learning_rate": 5.735891895850881e-08, + "loss": 0.696, + "step": 39001 + }, + { + "epoch": 1.43, + "learning_rate": 5.735525689383675e-08, + "loss": 0.685, + "step": 39002 + }, + { + "epoch": 1.43, + "learning_rate": 5.735159482916469e-08, + "loss": 0.6893, + "step": 39003 + }, + { + "epoch": 1.43, + "learning_rate": 5.734793276449262e-08, + "loss": 0.6867, + "step": 39004 + }, + { + "epoch": 1.43, + "learning_rate": 5.7344270699820554e-08, + "loss": 0.7058, + "step": 39005 + }, + { + "epoch": 1.43, + "learning_rate": 5.7340608635148494e-08, + "loss": 0.6999, + "step": 39006 + }, + { + "epoch": 1.43, + "learning_rate": 5.733694657047643e-08, + "loss": 0.6977, + "step": 39007 + }, + { + "epoch": 1.43, + "learning_rate": 5.733328450580437e-08, + "loss": 0.6983, + "step": 39008 + }, + { + "epoch": 1.43, + "learning_rate": 5.732962244113231e-08, + "loss": 0.7011, + "step": 39009 + }, + { + "epoch": 1.43, + "learning_rate": 5.732596037646025e-08, + "loss": 0.6907, + "step": 39010 + }, + { + "epoch": 1.43, + "learning_rate": 5.732229831178819e-08, + "loss": 0.6965, + "step": 39011 + }, + { + "epoch": 1.43, + "learning_rate": 5.731863624711613e-08, + "loss": 0.6972, + "step": 39012 + }, + { + "epoch": 1.43, + "learning_rate": 5.7314974182444067e-08, + "loss": 0.683, + "step": 39013 + }, + { + "epoch": 1.43, + "learning_rate": 5.7311312117771993e-08, + "loss": 0.6907, + "step": 39014 + }, + { + "epoch": 1.43, + "learning_rate": 5.730765005309993e-08, + "loss": 0.7018, + "step": 39015 + }, + { + "epoch": 1.43, + "learning_rate": 5.730398798842787e-08, + "loss": 0.7015, + "step": 39016 + }, + { + "epoch": 1.43, + "learning_rate": 5.730032592375581e-08, + "loss": 0.682, + "step": 39017 + }, + { + "epoch": 1.43, + "learning_rate": 5.7296663859083746e-08, + "loss": 0.6918, + "step": 39018 + }, + { + "epoch": 1.43, + "learning_rate": 5.7293001794411686e-08, + "loss": 0.6937, + "step": 39019 + }, + { + "epoch": 1.43, + "learning_rate": 5.7289339729739626e-08, + "loss": 0.6872, + "step": 39020 + }, + { + "epoch": 1.43, + "learning_rate": 5.7285677665067566e-08, + "loss": 0.6958, + "step": 39021 + }, + { + "epoch": 1.43, + "learning_rate": 5.7282015600395506e-08, + "loss": 0.7062, + "step": 39022 + }, + { + "epoch": 1.43, + "learning_rate": 5.727835353572343e-08, + "loss": 0.6872, + "step": 39023 + }, + { + "epoch": 1.43, + "learning_rate": 5.727469147105137e-08, + "loss": 0.6997, + "step": 39024 + }, + { + "epoch": 1.43, + "learning_rate": 5.727102940637931e-08, + "loss": 0.6917, + "step": 39025 + }, + { + "epoch": 1.43, + "learning_rate": 5.726736734170725e-08, + "loss": 0.685, + "step": 39026 + }, + { + "epoch": 1.43, + "learning_rate": 5.726370527703519e-08, + "loss": 0.6997, + "step": 39027 + }, + { + "epoch": 1.43, + "learning_rate": 5.726004321236313e-08, + "loss": 0.6776, + "step": 39028 + }, + { + "epoch": 1.43, + "learning_rate": 5.7256381147691066e-08, + "loss": 0.694, + "step": 39029 + }, + { + "epoch": 1.43, + "learning_rate": 5.7252719083019006e-08, + "loss": 0.7032, + "step": 39030 + }, + { + "epoch": 1.43, + "learning_rate": 5.7249057018346946e-08, + "loss": 0.6954, + "step": 39031 + }, + { + "epoch": 1.43, + "learning_rate": 5.724539495367487e-08, + "loss": 0.6994, + "step": 39032 + }, + { + "epoch": 1.43, + "learning_rate": 5.724173288900281e-08, + "loss": 0.6842, + "step": 39033 + }, + { + "epoch": 1.43, + "learning_rate": 5.723807082433075e-08, + "loss": 0.6995, + "step": 39034 + }, + { + "epoch": 1.43, + "learning_rate": 5.723440875965869e-08, + "loss": 0.7054, + "step": 39035 + }, + { + "epoch": 1.43, + "learning_rate": 5.723074669498663e-08, + "loss": 0.6906, + "step": 39036 + }, + { + "epoch": 1.43, + "learning_rate": 5.722708463031457e-08, + "loss": 0.6956, + "step": 39037 + }, + { + "epoch": 1.43, + "learning_rate": 5.722342256564251e-08, + "loss": 0.7002, + "step": 39038 + }, + { + "epoch": 1.43, + "learning_rate": 5.721976050097045e-08, + "loss": 0.6928, + "step": 39039 + }, + { + "epoch": 1.43, + "learning_rate": 5.7216098436298385e-08, + "loss": 0.7113, + "step": 39040 + }, + { + "epoch": 1.43, + "learning_rate": 5.721243637162632e-08, + "loss": 0.6891, + "step": 39041 + }, + { + "epoch": 1.43, + "learning_rate": 5.720877430695425e-08, + "loss": 0.6953, + "step": 39042 + }, + { + "epoch": 1.43, + "learning_rate": 5.720511224228219e-08, + "loss": 0.6897, + "step": 39043 + }, + { + "epoch": 1.43, + "learning_rate": 5.720145017761013e-08, + "loss": 0.6934, + "step": 39044 + }, + { + "epoch": 1.43, + "learning_rate": 5.719778811293807e-08, + "loss": 0.6922, + "step": 39045 + }, + { + "epoch": 1.43, + "learning_rate": 5.719412604826601e-08, + "loss": 0.6839, + "step": 39046 + }, + { + "epoch": 1.43, + "learning_rate": 5.719046398359395e-08, + "loss": 0.6741, + "step": 39047 + }, + { + "epoch": 1.43, + "learning_rate": 5.718680191892189e-08, + "loss": 0.6866, + "step": 39048 + }, + { + "epoch": 1.43, + "learning_rate": 5.718313985424983e-08, + "loss": 0.696, + "step": 39049 + }, + { + "epoch": 1.43, + "learning_rate": 5.717947778957776e-08, + "loss": 0.6913, + "step": 39050 + }, + { + "epoch": 1.43, + "learning_rate": 5.71758157249057e-08, + "loss": 0.6954, + "step": 39051 + }, + { + "epoch": 1.43, + "learning_rate": 5.717215366023364e-08, + "loss": 0.6928, + "step": 39052 + }, + { + "epoch": 1.43, + "learning_rate": 5.716849159556157e-08, + "loss": 0.686, + "step": 39053 + }, + { + "epoch": 1.43, + "learning_rate": 5.716482953088951e-08, + "loss": 0.6922, + "step": 39054 + }, + { + "epoch": 1.43, + "learning_rate": 5.716116746621745e-08, + "loss": 0.6935, + "step": 39055 + }, + { + "epoch": 1.43, + "learning_rate": 5.715750540154539e-08, + "loss": 0.6855, + "step": 39056 + }, + { + "epoch": 1.43, + "learning_rate": 5.715384333687333e-08, + "loss": 0.6933, + "step": 39057 + }, + { + "epoch": 1.43, + "learning_rate": 5.715018127220127e-08, + "loss": 0.6842, + "step": 39058 + }, + { + "epoch": 1.43, + "learning_rate": 5.71465192075292e-08, + "loss": 0.6739, + "step": 39059 + }, + { + "epoch": 1.43, + "learning_rate": 5.714285714285714e-08, + "loss": 0.6813, + "step": 39060 + }, + { + "epoch": 1.43, + "learning_rate": 5.713919507818508e-08, + "loss": 0.7004, + "step": 39061 + }, + { + "epoch": 1.43, + "learning_rate": 5.713553301351302e-08, + "loss": 0.6945, + "step": 39062 + }, + { + "epoch": 1.43, + "learning_rate": 5.713187094884096e-08, + "loss": 0.7131, + "step": 39063 + }, + { + "epoch": 1.43, + "learning_rate": 5.712820888416889e-08, + "loss": 0.6936, + "step": 39064 + }, + { + "epoch": 1.43, + "learning_rate": 5.712454681949683e-08, + "loss": 0.7229, + "step": 39065 + }, + { + "epoch": 1.43, + "learning_rate": 5.712088475482477e-08, + "loss": 0.6687, + "step": 39066 + }, + { + "epoch": 1.43, + "learning_rate": 5.711722269015271e-08, + "loss": 0.687, + "step": 39067 + }, + { + "epoch": 1.43, + "learning_rate": 5.711356062548064e-08, + "loss": 0.7098, + "step": 39068 + }, + { + "epoch": 1.43, + "learning_rate": 5.710989856080858e-08, + "loss": 0.6992, + "step": 39069 + }, + { + "epoch": 1.43, + "learning_rate": 5.710623649613652e-08, + "loss": 0.7013, + "step": 39070 + }, + { + "epoch": 1.43, + "learning_rate": 5.710257443146446e-08, + "loss": 0.6852, + "step": 39071 + }, + { + "epoch": 1.43, + "learning_rate": 5.70989123667924e-08, + "loss": 0.7026, + "step": 39072 + }, + { + "epoch": 1.43, + "learning_rate": 5.7095250302120337e-08, + "loss": 0.6827, + "step": 39073 + }, + { + "epoch": 1.43, + "learning_rate": 5.7091588237448277e-08, + "loss": 0.6988, + "step": 39074 + }, + { + "epoch": 1.43, + "learning_rate": 5.708792617277621e-08, + "loss": 0.6825, + "step": 39075 + }, + { + "epoch": 1.43, + "learning_rate": 5.708426410810415e-08, + "loss": 0.6976, + "step": 39076 + }, + { + "epoch": 1.43, + "learning_rate": 5.708060204343208e-08, + "loss": 0.6975, + "step": 39077 + }, + { + "epoch": 1.43, + "learning_rate": 5.7076939978760016e-08, + "loss": 0.6801, + "step": 39078 + }, + { + "epoch": 1.43, + "learning_rate": 5.7073277914087956e-08, + "loss": 0.6941, + "step": 39079 + }, + { + "epoch": 1.43, + "learning_rate": 5.7069615849415896e-08, + "loss": 0.6939, + "step": 39080 + }, + { + "epoch": 1.43, + "learning_rate": 5.7065953784743836e-08, + "loss": 0.6801, + "step": 39081 + }, + { + "epoch": 1.43, + "learning_rate": 5.7062291720071776e-08, + "loss": 0.6981, + "step": 39082 + }, + { + "epoch": 1.43, + "learning_rate": 5.7058629655399716e-08, + "loss": 0.6922, + "step": 39083 + }, + { + "epoch": 1.43, + "learning_rate": 5.7054967590727656e-08, + "loss": 0.6968, + "step": 39084 + }, + { + "epoch": 1.43, + "learning_rate": 5.7051305526055596e-08, + "loss": 0.6793, + "step": 39085 + }, + { + "epoch": 1.43, + "learning_rate": 5.704764346138352e-08, + "loss": 0.7086, + "step": 39086 + }, + { + "epoch": 1.43, + "learning_rate": 5.704398139671146e-08, + "loss": 0.6819, + "step": 39087 + }, + { + "epoch": 1.43, + "learning_rate": 5.70403193320394e-08, + "loss": 0.7135, + "step": 39088 + }, + { + "epoch": 1.43, + "learning_rate": 5.7036657267367336e-08, + "loss": 0.6821, + "step": 39089 + }, + { + "epoch": 1.43, + "learning_rate": 5.7032995202695276e-08, + "loss": 0.7029, + "step": 39090 + }, + { + "epoch": 1.43, + "learning_rate": 5.7029333138023216e-08, + "loss": 0.6925, + "step": 39091 + }, + { + "epoch": 1.43, + "learning_rate": 5.7025671073351156e-08, + "loss": 0.6868, + "step": 39092 + }, + { + "epoch": 1.43, + "learning_rate": 5.7022009008679096e-08, + "loss": 0.6993, + "step": 39093 + }, + { + "epoch": 1.43, + "learning_rate": 5.701834694400702e-08, + "loss": 0.7117, + "step": 39094 + }, + { + "epoch": 1.43, + "learning_rate": 5.701468487933496e-08, + "loss": 0.6971, + "step": 39095 + }, + { + "epoch": 1.43, + "learning_rate": 5.70110228146629e-08, + "loss": 0.7007, + "step": 39096 + }, + { + "epoch": 1.43, + "learning_rate": 5.700736074999084e-08, + "loss": 0.6914, + "step": 39097 + }, + { + "epoch": 1.43, + "learning_rate": 5.700369868531878e-08, + "loss": 0.7071, + "step": 39098 + }, + { + "epoch": 1.43, + "learning_rate": 5.700003662064672e-08, + "loss": 0.6902, + "step": 39099 + }, + { + "epoch": 1.43, + "learning_rate": 5.6996374555974655e-08, + "loss": 0.6857, + "step": 39100 + }, + { + "epoch": 1.43, + "learning_rate": 5.6992712491302595e-08, + "loss": 0.7038, + "step": 39101 + }, + { + "epoch": 1.43, + "learning_rate": 5.6989050426630535e-08, + "loss": 0.6953, + "step": 39102 + }, + { + "epoch": 1.43, + "learning_rate": 5.698538836195846e-08, + "loss": 0.6944, + "step": 39103 + }, + { + "epoch": 1.43, + "learning_rate": 5.69817262972864e-08, + "loss": 0.6994, + "step": 39104 + }, + { + "epoch": 1.43, + "learning_rate": 5.697806423261434e-08, + "loss": 0.6849, + "step": 39105 + }, + { + "epoch": 1.43, + "learning_rate": 5.697440216794228e-08, + "loss": 0.6892, + "step": 39106 + }, + { + "epoch": 1.43, + "learning_rate": 5.697074010327022e-08, + "loss": 0.6879, + "step": 39107 + }, + { + "epoch": 1.43, + "learning_rate": 5.696707803859816e-08, + "loss": 0.6996, + "step": 39108 + }, + { + "epoch": 1.43, + "learning_rate": 5.69634159739261e-08, + "loss": 0.6735, + "step": 39109 + }, + { + "epoch": 1.43, + "learning_rate": 5.695975390925404e-08, + "loss": 0.6927, + "step": 39110 + }, + { + "epoch": 1.43, + "learning_rate": 5.6956091844581975e-08, + "loss": 0.6992, + "step": 39111 + }, + { + "epoch": 1.43, + "learning_rate": 5.695242977990991e-08, + "loss": 0.6901, + "step": 39112 + }, + { + "epoch": 1.43, + "learning_rate": 5.694876771523784e-08, + "loss": 0.6967, + "step": 39113 + }, + { + "epoch": 1.43, + "learning_rate": 5.694510565056578e-08, + "loss": 0.6808, + "step": 39114 + }, + { + "epoch": 1.43, + "learning_rate": 5.694144358589372e-08, + "loss": 0.6998, + "step": 39115 + }, + { + "epoch": 1.43, + "learning_rate": 5.693778152122166e-08, + "loss": 0.6949, + "step": 39116 + }, + { + "epoch": 1.43, + "learning_rate": 5.69341194565496e-08, + "loss": 0.6973, + "step": 39117 + }, + { + "epoch": 1.43, + "learning_rate": 5.693045739187754e-08, + "loss": 0.6834, + "step": 39118 + }, + { + "epoch": 1.43, + "learning_rate": 5.692679532720548e-08, + "loss": 0.6903, + "step": 39119 + }, + { + "epoch": 1.43, + "learning_rate": 5.692313326253342e-08, + "loss": 0.6918, + "step": 39120 + }, + { + "epoch": 1.43, + "learning_rate": 5.691947119786135e-08, + "loss": 0.6961, + "step": 39121 + }, + { + "epoch": 1.43, + "learning_rate": 5.691580913318929e-08, + "loss": 0.6952, + "step": 39122 + }, + { + "epoch": 1.43, + "learning_rate": 5.691214706851723e-08, + "loss": 0.6832, + "step": 39123 + }, + { + "epoch": 1.43, + "learning_rate": 5.690848500384516e-08, + "loss": 0.6958, + "step": 39124 + }, + { + "epoch": 1.43, + "learning_rate": 5.69048229391731e-08, + "loss": 0.6957, + "step": 39125 + }, + { + "epoch": 1.43, + "learning_rate": 5.690116087450104e-08, + "loss": 0.6969, + "step": 39126 + }, + { + "epoch": 1.43, + "learning_rate": 5.689749880982898e-08, + "loss": 0.7043, + "step": 39127 + }, + { + "epoch": 1.43, + "learning_rate": 5.689383674515692e-08, + "loss": 0.6869, + "step": 39128 + }, + { + "epoch": 1.43, + "learning_rate": 5.689017468048486e-08, + "loss": 0.6894, + "step": 39129 + }, + { + "epoch": 1.43, + "learning_rate": 5.688651261581279e-08, + "loss": 0.6816, + "step": 39130 + }, + { + "epoch": 1.43, + "learning_rate": 5.688285055114073e-08, + "loss": 0.6847, + "step": 39131 + }, + { + "epoch": 1.43, + "learning_rate": 5.687918848646867e-08, + "loss": 0.6957, + "step": 39132 + }, + { + "epoch": 1.43, + "learning_rate": 5.6875526421796607e-08, + "loss": 0.6952, + "step": 39133 + }, + { + "epoch": 1.43, + "learning_rate": 5.6871864357124547e-08, + "loss": 0.6847, + "step": 39134 + }, + { + "epoch": 1.43, + "learning_rate": 5.686820229245248e-08, + "loss": 0.7026, + "step": 39135 + }, + { + "epoch": 1.43, + "learning_rate": 5.686454022778042e-08, + "loss": 0.6913, + "step": 39136 + }, + { + "epoch": 1.43, + "learning_rate": 5.686087816310836e-08, + "loss": 0.6949, + "step": 39137 + }, + { + "epoch": 1.43, + "learning_rate": 5.68572160984363e-08, + "loss": 0.6918, + "step": 39138 + }, + { + "epoch": 1.43, + "learning_rate": 5.6853554033764226e-08, + "loss": 0.6796, + "step": 39139 + }, + { + "epoch": 1.43, + "learning_rate": 5.6849891969092166e-08, + "loss": 0.7033, + "step": 39140 + }, + { + "epoch": 1.43, + "learning_rate": 5.6846229904420106e-08, + "loss": 0.6853, + "step": 39141 + }, + { + "epoch": 1.43, + "learning_rate": 5.6842567839748046e-08, + "loss": 0.6975, + "step": 39142 + }, + { + "epoch": 1.43, + "learning_rate": 5.6838905775075986e-08, + "loss": 0.6972, + "step": 39143 + }, + { + "epoch": 1.43, + "learning_rate": 5.6835243710403926e-08, + "loss": 0.6896, + "step": 39144 + }, + { + "epoch": 1.43, + "learning_rate": 5.6831581645731866e-08, + "loss": 0.702, + "step": 39145 + }, + { + "epoch": 1.43, + "learning_rate": 5.68279195810598e-08, + "loss": 0.6872, + "step": 39146 + }, + { + "epoch": 1.43, + "learning_rate": 5.682425751638774e-08, + "loss": 0.7024, + "step": 39147 + }, + { + "epoch": 1.43, + "learning_rate": 5.682059545171567e-08, + "loss": 0.6923, + "step": 39148 + }, + { + "epoch": 1.43, + "learning_rate": 5.6816933387043606e-08, + "loss": 0.6922, + "step": 39149 + }, + { + "epoch": 1.43, + "learning_rate": 5.6813271322371546e-08, + "loss": 0.6983, + "step": 39150 + }, + { + "epoch": 1.43, + "learning_rate": 5.6809609257699486e-08, + "loss": 0.6978, + "step": 39151 + }, + { + "epoch": 1.43, + "learning_rate": 5.6805947193027426e-08, + "loss": 0.7005, + "step": 39152 + }, + { + "epoch": 1.43, + "learning_rate": 5.6802285128355366e-08, + "loss": 0.6998, + "step": 39153 + }, + { + "epoch": 1.43, + "learning_rate": 5.6798623063683305e-08, + "loss": 0.7162, + "step": 39154 + }, + { + "epoch": 1.43, + "learning_rate": 5.6794960999011245e-08, + "loss": 0.6984, + "step": 39155 + }, + { + "epoch": 1.43, + "learning_rate": 5.6791298934339185e-08, + "loss": 0.6963, + "step": 39156 + }, + { + "epoch": 1.43, + "learning_rate": 5.678763686966711e-08, + "loss": 0.7008, + "step": 39157 + }, + { + "epoch": 1.43, + "learning_rate": 5.678397480499505e-08, + "loss": 0.6899, + "step": 39158 + }, + { + "epoch": 1.43, + "learning_rate": 5.678031274032299e-08, + "loss": 0.695, + "step": 39159 + }, + { + "epoch": 1.43, + "learning_rate": 5.6776650675650925e-08, + "loss": 0.6951, + "step": 39160 + }, + { + "epoch": 1.43, + "learning_rate": 5.6772988610978865e-08, + "loss": 0.6988, + "step": 39161 + }, + { + "epoch": 1.43, + "learning_rate": 5.6769326546306805e-08, + "loss": 0.6853, + "step": 39162 + }, + { + "epoch": 1.43, + "learning_rate": 5.6765664481634745e-08, + "loss": 0.6929, + "step": 39163 + }, + { + "epoch": 1.43, + "learning_rate": 5.6762002416962685e-08, + "loss": 0.6937, + "step": 39164 + }, + { + "epoch": 1.43, + "learning_rate": 5.6758340352290625e-08, + "loss": 0.7053, + "step": 39165 + }, + { + "epoch": 1.43, + "learning_rate": 5.675467828761855e-08, + "loss": 0.6902, + "step": 39166 + }, + { + "epoch": 1.43, + "learning_rate": 5.675101622294649e-08, + "loss": 0.6964, + "step": 39167 + }, + { + "epoch": 1.43, + "learning_rate": 5.674735415827443e-08, + "loss": 0.6972, + "step": 39168 + }, + { + "epoch": 1.43, + "learning_rate": 5.674369209360237e-08, + "loss": 0.706, + "step": 39169 + }, + { + "epoch": 1.43, + "learning_rate": 5.674003002893031e-08, + "loss": 0.7025, + "step": 39170 + }, + { + "epoch": 1.43, + "learning_rate": 5.6736367964258245e-08, + "loss": 0.6908, + "step": 39171 + }, + { + "epoch": 1.43, + "learning_rate": 5.6732705899586184e-08, + "loss": 0.6982, + "step": 39172 + }, + { + "epoch": 1.43, + "learning_rate": 5.6729043834914124e-08, + "loss": 0.6851, + "step": 39173 + }, + { + "epoch": 1.43, + "learning_rate": 5.6725381770242064e-08, + "loss": 0.6939, + "step": 39174 + }, + { + "epoch": 1.43, + "learning_rate": 5.672171970556999e-08, + "loss": 0.7016, + "step": 39175 + }, + { + "epoch": 1.43, + "learning_rate": 5.671805764089793e-08, + "loss": 0.6917, + "step": 39176 + }, + { + "epoch": 1.43, + "learning_rate": 5.671439557622587e-08, + "loss": 0.6893, + "step": 39177 + }, + { + "epoch": 1.43, + "learning_rate": 5.671073351155381e-08, + "loss": 0.6848, + "step": 39178 + }, + { + "epoch": 1.43, + "learning_rate": 5.670707144688175e-08, + "loss": 0.6898, + "step": 39179 + }, + { + "epoch": 1.43, + "learning_rate": 5.670340938220969e-08, + "loss": 0.6823, + "step": 39180 + }, + { + "epoch": 1.43, + "learning_rate": 5.669974731753763e-08, + "loss": 0.693, + "step": 39181 + }, + { + "epoch": 1.43, + "learning_rate": 5.6696085252865564e-08, + "loss": 0.6827, + "step": 39182 + }, + { + "epoch": 1.43, + "learning_rate": 5.6692423188193504e-08, + "loss": 0.699, + "step": 39183 + }, + { + "epoch": 1.43, + "learning_rate": 5.668876112352143e-08, + "loss": 0.6961, + "step": 39184 + }, + { + "epoch": 1.43, + "learning_rate": 5.668509905884937e-08, + "loss": 0.6963, + "step": 39185 + }, + { + "epoch": 1.43, + "learning_rate": 5.668143699417731e-08, + "loss": 0.6925, + "step": 39186 + }, + { + "epoch": 1.43, + "learning_rate": 5.667777492950525e-08, + "loss": 0.6923, + "step": 39187 + }, + { + "epoch": 1.43, + "learning_rate": 5.667411286483319e-08, + "loss": 0.676, + "step": 39188 + }, + { + "epoch": 1.43, + "learning_rate": 5.667045080016113e-08, + "loss": 0.6939, + "step": 39189 + }, + { + "epoch": 1.43, + "learning_rate": 5.666678873548907e-08, + "loss": 0.6953, + "step": 39190 + }, + { + "epoch": 1.43, + "learning_rate": 5.666312667081701e-08, + "loss": 0.6911, + "step": 39191 + }, + { + "epoch": 1.43, + "learning_rate": 5.6659464606144943e-08, + "loss": 0.6768, + "step": 39192 + }, + { + "epoch": 1.43, + "learning_rate": 5.6655802541472877e-08, + "loss": 0.6867, + "step": 39193 + }, + { + "epoch": 1.43, + "learning_rate": 5.6652140476800817e-08, + "loss": 0.6944, + "step": 39194 + }, + { + "epoch": 1.43, + "learning_rate": 5.664847841212875e-08, + "loss": 0.6922, + "step": 39195 + }, + { + "epoch": 1.43, + "learning_rate": 5.664481634745669e-08, + "loss": 0.6913, + "step": 39196 + }, + { + "epoch": 1.43, + "learning_rate": 5.664115428278463e-08, + "loss": 0.7068, + "step": 39197 + }, + { + "epoch": 1.43, + "learning_rate": 5.663749221811257e-08, + "loss": 0.6952, + "step": 39198 + }, + { + "epoch": 1.43, + "learning_rate": 5.663383015344051e-08, + "loss": 0.6852, + "step": 39199 + }, + { + "epoch": 1.43, + "learning_rate": 5.663016808876845e-08, + "loss": 0.7015, + "step": 39200 + }, + { + "epoch": 1.43, + "learning_rate": 5.662650602409639e-08, + "loss": 0.6955, + "step": 39201 + }, + { + "epoch": 1.43, + "learning_rate": 5.6622843959424316e-08, + "loss": 0.6965, + "step": 39202 + }, + { + "epoch": 1.43, + "learning_rate": 5.6619181894752256e-08, + "loss": 0.7076, + "step": 39203 + }, + { + "epoch": 1.43, + "learning_rate": 5.6615519830080196e-08, + "loss": 0.6913, + "step": 39204 + }, + { + "epoch": 1.43, + "learning_rate": 5.6611857765408136e-08, + "loss": 0.697, + "step": 39205 + }, + { + "epoch": 1.43, + "learning_rate": 5.660819570073607e-08, + "loss": 0.7059, + "step": 39206 + }, + { + "epoch": 1.43, + "learning_rate": 5.660453363606401e-08, + "loss": 0.6711, + "step": 39207 + }, + { + "epoch": 1.43, + "learning_rate": 5.660087157139195e-08, + "loss": 0.7004, + "step": 39208 + }, + { + "epoch": 1.43, + "learning_rate": 5.659720950671989e-08, + "loss": 0.6878, + "step": 39209 + }, + { + "epoch": 1.43, + "learning_rate": 5.659354744204783e-08, + "loss": 0.6885, + "step": 39210 + }, + { + "epoch": 1.43, + "learning_rate": 5.6589885377375756e-08, + "loss": 0.6891, + "step": 39211 + }, + { + "epoch": 1.43, + "learning_rate": 5.6586223312703696e-08, + "loss": 0.696, + "step": 39212 + }, + { + "epoch": 1.43, + "learning_rate": 5.6582561248031635e-08, + "loss": 0.7056, + "step": 39213 + }, + { + "epoch": 1.43, + "learning_rate": 5.6578899183359575e-08, + "loss": 0.693, + "step": 39214 + }, + { + "epoch": 1.43, + "learning_rate": 5.6575237118687515e-08, + "loss": 0.6744, + "step": 39215 + }, + { + "epoch": 1.43, + "learning_rate": 5.6571575054015455e-08, + "loss": 0.6956, + "step": 39216 + }, + { + "epoch": 1.43, + "learning_rate": 5.656791298934339e-08, + "loss": 0.693, + "step": 39217 + }, + { + "epoch": 1.43, + "learning_rate": 5.656425092467133e-08, + "loss": 0.6954, + "step": 39218 + }, + { + "epoch": 1.43, + "learning_rate": 5.656058885999927e-08, + "loss": 0.6917, + "step": 39219 + }, + { + "epoch": 1.43, + "learning_rate": 5.6556926795327195e-08, + "loss": 0.6976, + "step": 39220 + }, + { + "epoch": 1.43, + "learning_rate": 5.6553264730655135e-08, + "loss": 0.6872, + "step": 39221 + }, + { + "epoch": 1.44, + "learning_rate": 5.6549602665983075e-08, + "loss": 0.6898, + "step": 39222 + }, + { + "epoch": 1.44, + "learning_rate": 5.6545940601311015e-08, + "loss": 0.6915, + "step": 39223 + }, + { + "epoch": 1.44, + "learning_rate": 5.6542278536638955e-08, + "loss": 0.6818, + "step": 39224 + }, + { + "epoch": 1.44, + "learning_rate": 5.6538616471966895e-08, + "loss": 0.6878, + "step": 39225 + }, + { + "epoch": 1.44, + "learning_rate": 5.6534954407294835e-08, + "loss": 0.6899, + "step": 39226 + }, + { + "epoch": 1.44, + "learning_rate": 5.6531292342622775e-08, + "loss": 0.6933, + "step": 39227 + }, + { + "epoch": 1.44, + "learning_rate": 5.652763027795071e-08, + "loss": 0.68, + "step": 39228 + }, + { + "epoch": 1.44, + "learning_rate": 5.652396821327864e-08, + "loss": 0.7001, + "step": 39229 + }, + { + "epoch": 1.44, + "learning_rate": 5.652030614860658e-08, + "loss": 0.6813, + "step": 39230 + }, + { + "epoch": 1.44, + "learning_rate": 5.6516644083934515e-08, + "loss": 0.6865, + "step": 39231 + }, + { + "epoch": 1.44, + "learning_rate": 5.6512982019262454e-08, + "loss": 0.7008, + "step": 39232 + }, + { + "epoch": 1.44, + "learning_rate": 5.6509319954590394e-08, + "loss": 0.6871, + "step": 39233 + }, + { + "epoch": 1.44, + "learning_rate": 5.6505657889918334e-08, + "loss": 0.6902, + "step": 39234 + }, + { + "epoch": 1.44, + "learning_rate": 5.6501995825246274e-08, + "loss": 0.6786, + "step": 39235 + }, + { + "epoch": 1.44, + "learning_rate": 5.6498333760574214e-08, + "loss": 0.6952, + "step": 39236 + }, + { + "epoch": 1.44, + "learning_rate": 5.6494671695902154e-08, + "loss": 0.6968, + "step": 39237 + }, + { + "epoch": 1.44, + "learning_rate": 5.649100963123008e-08, + "loss": 0.7004, + "step": 39238 + }, + { + "epoch": 1.44, + "learning_rate": 5.648734756655802e-08, + "loss": 0.6991, + "step": 39239 + }, + { + "epoch": 1.44, + "learning_rate": 5.648368550188596e-08, + "loss": 0.6951, + "step": 39240 + }, + { + "epoch": 1.44, + "learning_rate": 5.64800234372139e-08, + "loss": 0.6815, + "step": 39241 + }, + { + "epoch": 1.44, + "learning_rate": 5.6476361372541834e-08, + "loss": 0.6985, + "step": 39242 + }, + { + "epoch": 1.44, + "learning_rate": 5.6472699307869774e-08, + "loss": 0.6823, + "step": 39243 + }, + { + "epoch": 1.44, + "learning_rate": 5.6469037243197714e-08, + "loss": 0.6909, + "step": 39244 + }, + { + "epoch": 1.44, + "learning_rate": 5.6465375178525654e-08, + "loss": 0.6895, + "step": 39245 + }, + { + "epoch": 1.44, + "learning_rate": 5.6461713113853594e-08, + "loss": 0.6889, + "step": 39246 + }, + { + "epoch": 1.44, + "learning_rate": 5.645805104918152e-08, + "loss": 0.6973, + "step": 39247 + }, + { + "epoch": 1.44, + "learning_rate": 5.645438898450946e-08, + "loss": 0.6879, + "step": 39248 + }, + { + "epoch": 1.44, + "learning_rate": 5.64507269198374e-08, + "loss": 0.6937, + "step": 39249 + }, + { + "epoch": 1.44, + "learning_rate": 5.644706485516534e-08, + "loss": 0.6855, + "step": 39250 + }, + { + "epoch": 1.44, + "learning_rate": 5.644340279049328e-08, + "loss": 0.7025, + "step": 39251 + }, + { + "epoch": 1.44, + "learning_rate": 5.643974072582121e-08, + "loss": 0.6887, + "step": 39252 + }, + { + "epoch": 1.44, + "learning_rate": 5.643607866114915e-08, + "loss": 0.6996, + "step": 39253 + }, + { + "epoch": 1.44, + "learning_rate": 5.643241659647709e-08, + "loss": 0.6929, + "step": 39254 + }, + { + "epoch": 1.44, + "learning_rate": 5.642875453180503e-08, + "loss": 0.6983, + "step": 39255 + }, + { + "epoch": 1.44, + "learning_rate": 5.642509246713296e-08, + "loss": 0.7068, + "step": 39256 + }, + { + "epoch": 1.44, + "learning_rate": 5.64214304024609e-08, + "loss": 0.6994, + "step": 39257 + }, + { + "epoch": 1.44, + "learning_rate": 5.641776833778884e-08, + "loss": 0.6833, + "step": 39258 + }, + { + "epoch": 1.44, + "learning_rate": 5.641410627311678e-08, + "loss": 0.6862, + "step": 39259 + }, + { + "epoch": 1.44, + "learning_rate": 5.641044420844472e-08, + "loss": 0.6872, + "step": 39260 + }, + { + "epoch": 1.44, + "learning_rate": 5.640678214377266e-08, + "loss": 0.6853, + "step": 39261 + }, + { + "epoch": 1.44, + "learning_rate": 5.64031200791006e-08, + "loss": 0.6956, + "step": 39262 + }, + { + "epoch": 1.44, + "learning_rate": 5.639945801442853e-08, + "loss": 0.7019, + "step": 39263 + }, + { + "epoch": 1.44, + "learning_rate": 5.639579594975647e-08, + "loss": 0.6976, + "step": 39264 + }, + { + "epoch": 1.44, + "learning_rate": 5.6392133885084406e-08, + "loss": 0.6976, + "step": 39265 + }, + { + "epoch": 1.44, + "learning_rate": 5.638847182041234e-08, + "loss": 0.7023, + "step": 39266 + }, + { + "epoch": 1.44, + "learning_rate": 5.638480975574028e-08, + "loss": 0.6974, + "step": 39267 + }, + { + "epoch": 1.44, + "learning_rate": 5.638114769106822e-08, + "loss": 0.6906, + "step": 39268 + }, + { + "epoch": 1.44, + "learning_rate": 5.637748562639616e-08, + "loss": 0.6859, + "step": 39269 + }, + { + "epoch": 1.44, + "learning_rate": 5.63738235617241e-08, + "loss": 0.6955, + "step": 39270 + }, + { + "epoch": 1.44, + "learning_rate": 5.637016149705204e-08, + "loss": 0.6929, + "step": 39271 + }, + { + "epoch": 1.44, + "learning_rate": 5.636649943237998e-08, + "loss": 0.6907, + "step": 39272 + }, + { + "epoch": 1.44, + "learning_rate": 5.636283736770792e-08, + "loss": 0.69, + "step": 39273 + }, + { + "epoch": 1.44, + "learning_rate": 5.6359175303035845e-08, + "loss": 0.6927, + "step": 39274 + }, + { + "epoch": 1.44, + "learning_rate": 5.6355513238363785e-08, + "loss": 0.6829, + "step": 39275 + }, + { + "epoch": 1.44, + "learning_rate": 5.6351851173691725e-08, + "loss": 0.7018, + "step": 39276 + }, + { + "epoch": 1.44, + "learning_rate": 5.634818910901966e-08, + "loss": 0.6835, + "step": 39277 + }, + { + "epoch": 1.44, + "learning_rate": 5.63445270443476e-08, + "loss": 0.7048, + "step": 39278 + }, + { + "epoch": 1.44, + "learning_rate": 5.634086497967554e-08, + "loss": 0.6891, + "step": 39279 + }, + { + "epoch": 1.44, + "learning_rate": 5.633720291500348e-08, + "loss": 0.692, + "step": 39280 + }, + { + "epoch": 1.44, + "learning_rate": 5.633354085033142e-08, + "loss": 0.6918, + "step": 39281 + }, + { + "epoch": 1.44, + "learning_rate": 5.632987878565936e-08, + "loss": 0.708, + "step": 39282 + }, + { + "epoch": 1.44, + "learning_rate": 5.6326216720987285e-08, + "loss": 0.6986, + "step": 39283 + }, + { + "epoch": 1.44, + "learning_rate": 5.6322554656315225e-08, + "loss": 0.7081, + "step": 39284 + }, + { + "epoch": 1.44, + "learning_rate": 5.6318892591643165e-08, + "loss": 0.6878, + "step": 39285 + }, + { + "epoch": 1.44, + "learning_rate": 5.6315230526971105e-08, + "loss": 0.6789, + "step": 39286 + }, + { + "epoch": 1.44, + "learning_rate": 5.6311568462299045e-08, + "loss": 0.6801, + "step": 39287 + }, + { + "epoch": 1.44, + "learning_rate": 5.630790639762698e-08, + "loss": 0.6959, + "step": 39288 + }, + { + "epoch": 1.44, + "learning_rate": 5.630424433295492e-08, + "loss": 0.6944, + "step": 39289 + }, + { + "epoch": 1.44, + "learning_rate": 5.630058226828286e-08, + "loss": 0.7053, + "step": 39290 + }, + { + "epoch": 1.44, + "learning_rate": 5.62969202036108e-08, + "loss": 0.6765, + "step": 39291 + }, + { + "epoch": 1.44, + "learning_rate": 5.6293258138938724e-08, + "loss": 0.691, + "step": 39292 + }, + { + "epoch": 1.44, + "learning_rate": 5.6289596074266664e-08, + "loss": 0.6869, + "step": 39293 + }, + { + "epoch": 1.44, + "learning_rate": 5.6285934009594604e-08, + "loss": 0.6841, + "step": 39294 + }, + { + "epoch": 1.44, + "learning_rate": 5.6282271944922544e-08, + "loss": 0.6922, + "step": 39295 + }, + { + "epoch": 1.44, + "learning_rate": 5.6278609880250484e-08, + "loss": 0.7038, + "step": 39296 + }, + { + "epoch": 1.44, + "learning_rate": 5.6274947815578424e-08, + "loss": 0.6936, + "step": 39297 + }, + { + "epoch": 1.44, + "learning_rate": 5.6271285750906364e-08, + "loss": 0.6983, + "step": 39298 + }, + { + "epoch": 1.44, + "learning_rate": 5.62676236862343e-08, + "loss": 0.6919, + "step": 39299 + }, + { + "epoch": 1.44, + "learning_rate": 5.626396162156224e-08, + "loss": 0.6928, + "step": 39300 + }, + { + "epoch": 1.44, + "learning_rate": 5.626029955689017e-08, + "loss": 0.7029, + "step": 39301 + }, + { + "epoch": 1.44, + "learning_rate": 5.6256637492218104e-08, + "loss": 0.6901, + "step": 39302 + }, + { + "epoch": 1.44, + "learning_rate": 5.6252975427546044e-08, + "loss": 0.695, + "step": 39303 + }, + { + "epoch": 1.44, + "learning_rate": 5.6249313362873984e-08, + "loss": 0.6876, + "step": 39304 + }, + { + "epoch": 1.44, + "learning_rate": 5.6245651298201924e-08, + "loss": 0.6969, + "step": 39305 + }, + { + "epoch": 1.44, + "learning_rate": 5.6241989233529864e-08, + "loss": 0.6917, + "step": 39306 + }, + { + "epoch": 1.44, + "learning_rate": 5.6238327168857803e-08, + "loss": 0.6916, + "step": 39307 + }, + { + "epoch": 1.44, + "learning_rate": 5.6234665104185743e-08, + "loss": 0.6884, + "step": 39308 + }, + { + "epoch": 1.44, + "learning_rate": 5.6231003039513683e-08, + "loss": 0.7092, + "step": 39309 + }, + { + "epoch": 1.44, + "learning_rate": 5.622734097484161e-08, + "loss": 0.6962, + "step": 39310 + }, + { + "epoch": 1.44, + "learning_rate": 5.622367891016955e-08, + "loss": 0.6903, + "step": 39311 + }, + { + "epoch": 1.44, + "learning_rate": 5.622001684549748e-08, + "loss": 0.7076, + "step": 39312 + }, + { + "epoch": 1.44, + "learning_rate": 5.621635478082542e-08, + "loss": 0.7062, + "step": 39313 + }, + { + "epoch": 1.44, + "learning_rate": 5.621269271615336e-08, + "loss": 0.6806, + "step": 39314 + }, + { + "epoch": 1.44, + "learning_rate": 5.62090306514813e-08, + "loss": 0.6845, + "step": 39315 + }, + { + "epoch": 1.44, + "learning_rate": 5.620536858680924e-08, + "loss": 0.6793, + "step": 39316 + }, + { + "epoch": 1.44, + "learning_rate": 5.620170652213718e-08, + "loss": 0.6949, + "step": 39317 + }, + { + "epoch": 1.44, + "learning_rate": 5.619804445746512e-08, + "loss": 0.6896, + "step": 39318 + }, + { + "epoch": 1.44, + "learning_rate": 5.619438239279305e-08, + "loss": 0.6874, + "step": 39319 + }, + { + "epoch": 1.44, + "learning_rate": 5.619072032812099e-08, + "loss": 0.6928, + "step": 39320 + }, + { + "epoch": 1.44, + "learning_rate": 5.618705826344893e-08, + "loss": 0.7035, + "step": 39321 + }, + { + "epoch": 1.44, + "learning_rate": 5.618339619877687e-08, + "loss": 0.6901, + "step": 39322 + }, + { + "epoch": 1.44, + "learning_rate": 5.61797341341048e-08, + "loss": 0.6877, + "step": 39323 + }, + { + "epoch": 1.44, + "learning_rate": 5.617607206943274e-08, + "loss": 0.6998, + "step": 39324 + }, + { + "epoch": 1.44, + "learning_rate": 5.617241000476068e-08, + "loss": 0.7171, + "step": 39325 + }, + { + "epoch": 1.44, + "learning_rate": 5.616874794008862e-08, + "loss": 0.7063, + "step": 39326 + }, + { + "epoch": 1.44, + "learning_rate": 5.616508587541656e-08, + "loss": 0.6966, + "step": 39327 + }, + { + "epoch": 1.44, + "learning_rate": 5.616142381074449e-08, + "loss": 0.6977, + "step": 39328 + }, + { + "epoch": 1.44, + "learning_rate": 5.615776174607243e-08, + "loss": 0.7062, + "step": 39329 + }, + { + "epoch": 1.44, + "learning_rate": 5.615409968140037e-08, + "loss": 0.69, + "step": 39330 + }, + { + "epoch": 1.44, + "learning_rate": 5.615043761672831e-08, + "loss": 0.6952, + "step": 39331 + }, + { + "epoch": 1.44, + "learning_rate": 5.614677555205625e-08, + "loss": 0.6885, + "step": 39332 + }, + { + "epoch": 1.44, + "learning_rate": 5.614311348738419e-08, + "loss": 0.691, + "step": 39333 + }, + { + "epoch": 1.44, + "learning_rate": 5.613945142271212e-08, + "loss": 0.6916, + "step": 39334 + }, + { + "epoch": 1.44, + "learning_rate": 5.613578935804006e-08, + "loss": 0.6919, + "step": 39335 + }, + { + "epoch": 1.44, + "learning_rate": 5.6132127293368e-08, + "loss": 0.6927, + "step": 39336 + }, + { + "epoch": 1.44, + "learning_rate": 5.612846522869593e-08, + "loss": 0.6965, + "step": 39337 + }, + { + "epoch": 1.44, + "learning_rate": 5.612480316402387e-08, + "loss": 0.6919, + "step": 39338 + }, + { + "epoch": 1.44, + "learning_rate": 5.612114109935181e-08, + "loss": 0.6891, + "step": 39339 + }, + { + "epoch": 1.44, + "learning_rate": 5.611747903467975e-08, + "loss": 0.6839, + "step": 39340 + }, + { + "epoch": 1.44, + "learning_rate": 5.611381697000769e-08, + "loss": 0.6974, + "step": 39341 + }, + { + "epoch": 1.44, + "learning_rate": 5.611015490533563e-08, + "loss": 0.6971, + "step": 39342 + }, + { + "epoch": 1.44, + "learning_rate": 5.610649284066357e-08, + "loss": 0.6812, + "step": 39343 + }, + { + "epoch": 1.44, + "learning_rate": 5.610283077599151e-08, + "loss": 0.6626, + "step": 39344 + }, + { + "epoch": 1.44, + "learning_rate": 5.609916871131944e-08, + "loss": 0.6804, + "step": 39345 + }, + { + "epoch": 1.44, + "learning_rate": 5.6095506646647375e-08, + "loss": 0.6924, + "step": 39346 + }, + { + "epoch": 1.44, + "learning_rate": 5.6091844581975315e-08, + "loss": 0.6879, + "step": 39347 + }, + { + "epoch": 1.44, + "learning_rate": 5.608818251730325e-08, + "loss": 0.6807, + "step": 39348 + }, + { + "epoch": 1.44, + "learning_rate": 5.608452045263119e-08, + "loss": 0.6977, + "step": 39349 + }, + { + "epoch": 1.44, + "learning_rate": 5.608085838795913e-08, + "loss": 0.6823, + "step": 39350 + }, + { + "epoch": 1.44, + "learning_rate": 5.607719632328707e-08, + "loss": 0.6857, + "step": 39351 + }, + { + "epoch": 1.44, + "learning_rate": 5.607353425861501e-08, + "loss": 0.6898, + "step": 39352 + }, + { + "epoch": 1.44, + "learning_rate": 5.606987219394295e-08, + "loss": 0.6946, + "step": 39353 + }, + { + "epoch": 1.44, + "learning_rate": 5.606621012927089e-08, + "loss": 0.6866, + "step": 39354 + }, + { + "epoch": 1.44, + "learning_rate": 5.6062548064598814e-08, + "loss": 0.6911, + "step": 39355 + }, + { + "epoch": 1.44, + "learning_rate": 5.6058885999926754e-08, + "loss": 0.6935, + "step": 39356 + }, + { + "epoch": 1.44, + "learning_rate": 5.6055223935254694e-08, + "loss": 0.6959, + "step": 39357 + }, + { + "epoch": 1.44, + "learning_rate": 5.6051561870582634e-08, + "loss": 0.6995, + "step": 39358 + }, + { + "epoch": 1.44, + "learning_rate": 5.604789980591057e-08, + "loss": 0.6839, + "step": 39359 + }, + { + "epoch": 1.44, + "learning_rate": 5.604423774123851e-08, + "loss": 0.6769, + "step": 39360 + }, + { + "epoch": 1.44, + "learning_rate": 5.604057567656645e-08, + "loss": 0.6911, + "step": 39361 + }, + { + "epoch": 1.44, + "learning_rate": 5.603691361189439e-08, + "loss": 0.6691, + "step": 39362 + }, + { + "epoch": 1.44, + "learning_rate": 5.603325154722233e-08, + "loss": 0.6963, + "step": 39363 + }, + { + "epoch": 1.44, + "learning_rate": 5.6029589482550254e-08, + "loss": 0.6931, + "step": 39364 + }, + { + "epoch": 1.44, + "learning_rate": 5.6025927417878194e-08, + "loss": 0.6991, + "step": 39365 + }, + { + "epoch": 1.44, + "learning_rate": 5.6022265353206134e-08, + "loss": 0.6946, + "step": 39366 + }, + { + "epoch": 1.44, + "learning_rate": 5.6018603288534073e-08, + "loss": 0.6947, + "step": 39367 + }, + { + "epoch": 1.44, + "learning_rate": 5.6014941223862013e-08, + "loss": 0.6839, + "step": 39368 + }, + { + "epoch": 1.44, + "learning_rate": 5.6011279159189953e-08, + "loss": 0.6909, + "step": 39369 + }, + { + "epoch": 1.44, + "learning_rate": 5.6007617094517887e-08, + "loss": 0.6929, + "step": 39370 + }, + { + "epoch": 1.44, + "learning_rate": 5.6003955029845827e-08, + "loss": 0.6948, + "step": 39371 + }, + { + "epoch": 1.44, + "learning_rate": 5.6000292965173767e-08, + "loss": 0.6952, + "step": 39372 + }, + { + "epoch": 1.44, + "learning_rate": 5.599663090050169e-08, + "loss": 0.679, + "step": 39373 + }, + { + "epoch": 1.44, + "learning_rate": 5.599296883582963e-08, + "loss": 0.699, + "step": 39374 + }, + { + "epoch": 1.44, + "learning_rate": 5.598930677115757e-08, + "loss": 0.6916, + "step": 39375 + }, + { + "epoch": 1.44, + "learning_rate": 5.598564470648551e-08, + "loss": 0.6967, + "step": 39376 + }, + { + "epoch": 1.44, + "learning_rate": 5.598198264181345e-08, + "loss": 0.7024, + "step": 39377 + }, + { + "epoch": 1.44, + "learning_rate": 5.597832057714139e-08, + "loss": 0.7019, + "step": 39378 + }, + { + "epoch": 1.44, + "learning_rate": 5.597465851246933e-08, + "loss": 0.6923, + "step": 39379 + }, + { + "epoch": 1.44, + "learning_rate": 5.597099644779727e-08, + "loss": 0.7003, + "step": 39380 + }, + { + "epoch": 1.44, + "learning_rate": 5.5967334383125206e-08, + "loss": 0.6961, + "step": 39381 + }, + { + "epoch": 1.44, + "learning_rate": 5.596367231845314e-08, + "loss": 0.6942, + "step": 39382 + }, + { + "epoch": 1.44, + "learning_rate": 5.596001025378107e-08, + "loss": 0.6973, + "step": 39383 + }, + { + "epoch": 1.44, + "learning_rate": 5.595634818910901e-08, + "loss": 0.6816, + "step": 39384 + }, + { + "epoch": 1.44, + "learning_rate": 5.595268612443695e-08, + "loss": 0.7024, + "step": 39385 + }, + { + "epoch": 1.44, + "learning_rate": 5.594902405976489e-08, + "loss": 0.703, + "step": 39386 + }, + { + "epoch": 1.44, + "learning_rate": 5.594536199509283e-08, + "loss": 0.6945, + "step": 39387 + }, + { + "epoch": 1.44, + "learning_rate": 5.594169993042077e-08, + "loss": 0.6976, + "step": 39388 + }, + { + "epoch": 1.44, + "learning_rate": 5.593803786574871e-08, + "loss": 0.7009, + "step": 39389 + }, + { + "epoch": 1.44, + "learning_rate": 5.593437580107665e-08, + "loss": 0.705, + "step": 39390 + }, + { + "epoch": 1.44, + "learning_rate": 5.593071373640458e-08, + "loss": 0.6936, + "step": 39391 + }, + { + "epoch": 1.44, + "learning_rate": 5.592705167173252e-08, + "loss": 0.6893, + "step": 39392 + }, + { + "epoch": 1.44, + "learning_rate": 5.592338960706046e-08, + "loss": 0.6762, + "step": 39393 + }, + { + "epoch": 1.44, + "learning_rate": 5.591972754238839e-08, + "loss": 0.6998, + "step": 39394 + }, + { + "epoch": 1.44, + "learning_rate": 5.591606547771633e-08, + "loss": 0.6969, + "step": 39395 + }, + { + "epoch": 1.44, + "learning_rate": 5.591240341304427e-08, + "loss": 0.6902, + "step": 39396 + }, + { + "epoch": 1.44, + "learning_rate": 5.590874134837221e-08, + "loss": 0.6934, + "step": 39397 + }, + { + "epoch": 1.44, + "learning_rate": 5.590507928370015e-08, + "loss": 0.6957, + "step": 39398 + }, + { + "epoch": 1.44, + "learning_rate": 5.590141721902809e-08, + "loss": 0.6978, + "step": 39399 + }, + { + "epoch": 1.44, + "learning_rate": 5.589775515435602e-08, + "loss": 0.6948, + "step": 39400 + }, + { + "epoch": 1.44, + "learning_rate": 5.589409308968396e-08, + "loss": 0.6919, + "step": 39401 + }, + { + "epoch": 1.44, + "learning_rate": 5.58904310250119e-08, + "loss": 0.7036, + "step": 39402 + }, + { + "epoch": 1.44, + "learning_rate": 5.588676896033984e-08, + "loss": 0.7014, + "step": 39403 + }, + { + "epoch": 1.44, + "learning_rate": 5.588310689566778e-08, + "loss": 0.675, + "step": 39404 + }, + { + "epoch": 1.44, + "learning_rate": 5.587944483099571e-08, + "loss": 0.6978, + "step": 39405 + }, + { + "epoch": 1.44, + "learning_rate": 5.587578276632365e-08, + "loss": 0.689, + "step": 39406 + }, + { + "epoch": 1.44, + "learning_rate": 5.587212070165159e-08, + "loss": 0.6973, + "step": 39407 + }, + { + "epoch": 1.44, + "learning_rate": 5.586845863697953e-08, + "loss": 0.693, + "step": 39408 + }, + { + "epoch": 1.44, + "learning_rate": 5.586479657230746e-08, + "loss": 0.7053, + "step": 39409 + }, + { + "epoch": 1.44, + "learning_rate": 5.58611345076354e-08, + "loss": 0.6806, + "step": 39410 + }, + { + "epoch": 1.44, + "learning_rate": 5.585747244296334e-08, + "loss": 0.6899, + "step": 39411 + }, + { + "epoch": 1.44, + "learning_rate": 5.585381037829128e-08, + "loss": 0.6805, + "step": 39412 + }, + { + "epoch": 1.44, + "learning_rate": 5.585014831361922e-08, + "loss": 0.6961, + "step": 39413 + }, + { + "epoch": 1.44, + "learning_rate": 5.584648624894716e-08, + "loss": 0.6808, + "step": 39414 + }, + { + "epoch": 1.44, + "learning_rate": 5.58428241842751e-08, + "loss": 0.6926, + "step": 39415 + }, + { + "epoch": 1.44, + "learning_rate": 5.583916211960303e-08, + "loss": 0.6965, + "step": 39416 + }, + { + "epoch": 1.44, + "learning_rate": 5.583550005493097e-08, + "loss": 0.6979, + "step": 39417 + }, + { + "epoch": 1.44, + "learning_rate": 5.5831837990258904e-08, + "loss": 0.6907, + "step": 39418 + }, + { + "epoch": 1.44, + "learning_rate": 5.582817592558684e-08, + "loss": 0.6938, + "step": 39419 + }, + { + "epoch": 1.44, + "learning_rate": 5.582451386091478e-08, + "loss": 0.7011, + "step": 39420 + }, + { + "epoch": 1.44, + "learning_rate": 5.582085179624272e-08, + "loss": 0.6954, + "step": 39421 + }, + { + "epoch": 1.44, + "learning_rate": 5.581718973157066e-08, + "loss": 0.6883, + "step": 39422 + }, + { + "epoch": 1.44, + "learning_rate": 5.58135276668986e-08, + "loss": 0.6985, + "step": 39423 + }, + { + "epoch": 1.44, + "learning_rate": 5.580986560222654e-08, + "loss": 0.6914, + "step": 39424 + }, + { + "epoch": 1.44, + "learning_rate": 5.580620353755448e-08, + "loss": 0.6965, + "step": 39425 + }, + { + "epoch": 1.44, + "learning_rate": 5.580254147288242e-08, + "loss": 0.6916, + "step": 39426 + }, + { + "epoch": 1.44, + "learning_rate": 5.5798879408210343e-08, + "loss": 0.685, + "step": 39427 + }, + { + "epoch": 1.44, + "learning_rate": 5.5795217343538283e-08, + "loss": 0.6942, + "step": 39428 + }, + { + "epoch": 1.44, + "learning_rate": 5.5791555278866223e-08, + "loss": 0.7026, + "step": 39429 + }, + { + "epoch": 1.44, + "learning_rate": 5.5787893214194157e-08, + "loss": 0.6923, + "step": 39430 + }, + { + "epoch": 1.44, + "learning_rate": 5.5784231149522097e-08, + "loss": 0.7053, + "step": 39431 + }, + { + "epoch": 1.44, + "learning_rate": 5.5780569084850036e-08, + "loss": 0.6877, + "step": 39432 + }, + { + "epoch": 1.44, + "learning_rate": 5.5776907020177976e-08, + "loss": 0.7049, + "step": 39433 + }, + { + "epoch": 1.44, + "learning_rate": 5.5773244955505916e-08, + "loss": 0.6971, + "step": 39434 + }, + { + "epoch": 1.44, + "learning_rate": 5.5769582890833856e-08, + "loss": 0.6921, + "step": 39435 + }, + { + "epoch": 1.44, + "learning_rate": 5.576592082616178e-08, + "loss": 0.7049, + "step": 39436 + }, + { + "epoch": 1.44, + "learning_rate": 5.576225876148972e-08, + "loss": 0.6915, + "step": 39437 + }, + { + "epoch": 1.44, + "learning_rate": 5.575859669681766e-08, + "loss": 0.6838, + "step": 39438 + }, + { + "epoch": 1.44, + "learning_rate": 5.57549346321456e-08, + "loss": 0.6831, + "step": 39439 + }, + { + "epoch": 1.44, + "learning_rate": 5.575127256747354e-08, + "loss": 0.7037, + "step": 39440 + }, + { + "epoch": 1.44, + "learning_rate": 5.5747610502801476e-08, + "loss": 0.7013, + "step": 39441 + }, + { + "epoch": 1.44, + "learning_rate": 5.5743948438129416e-08, + "loss": 0.693, + "step": 39442 + }, + { + "epoch": 1.44, + "learning_rate": 5.5740286373457356e-08, + "loss": 0.6741, + "step": 39443 + }, + { + "epoch": 1.44, + "learning_rate": 5.5736624308785296e-08, + "loss": 0.7203, + "step": 39444 + }, + { + "epoch": 1.44, + "learning_rate": 5.573296224411322e-08, + "loss": 0.7121, + "step": 39445 + }, + { + "epoch": 1.44, + "learning_rate": 5.572930017944116e-08, + "loss": 0.7161, + "step": 39446 + }, + { + "epoch": 1.44, + "learning_rate": 5.57256381147691e-08, + "loss": 0.6949, + "step": 39447 + }, + { + "epoch": 1.44, + "learning_rate": 5.572197605009704e-08, + "loss": 0.6791, + "step": 39448 + }, + { + "epoch": 1.44, + "learning_rate": 5.571831398542498e-08, + "loss": 0.7023, + "step": 39449 + }, + { + "epoch": 1.44, + "learning_rate": 5.571465192075292e-08, + "loss": 0.6899, + "step": 39450 + }, + { + "epoch": 1.44, + "learning_rate": 5.571098985608086e-08, + "loss": 0.7019, + "step": 39451 + }, + { + "epoch": 1.44, + "learning_rate": 5.5707327791408795e-08, + "loss": 0.694, + "step": 39452 + }, + { + "epoch": 1.44, + "learning_rate": 5.5703665726736735e-08, + "loss": 0.6885, + "step": 39453 + }, + { + "epoch": 1.44, + "learning_rate": 5.570000366206466e-08, + "loss": 0.7016, + "step": 39454 + }, + { + "epoch": 1.44, + "learning_rate": 5.56963415973926e-08, + "loss": 0.6925, + "step": 39455 + }, + { + "epoch": 1.44, + "learning_rate": 5.569267953272054e-08, + "loss": 0.7082, + "step": 39456 + }, + { + "epoch": 1.44, + "learning_rate": 5.568901746804848e-08, + "loss": 0.6963, + "step": 39457 + }, + { + "epoch": 1.44, + "learning_rate": 5.568535540337642e-08, + "loss": 0.6957, + "step": 39458 + }, + { + "epoch": 1.44, + "learning_rate": 5.568169333870436e-08, + "loss": 0.6808, + "step": 39459 + }, + { + "epoch": 1.44, + "learning_rate": 5.56780312740323e-08, + "loss": 0.6981, + "step": 39460 + }, + { + "epoch": 1.44, + "learning_rate": 5.567436920936024e-08, + "loss": 0.6854, + "step": 39461 + }, + { + "epoch": 1.44, + "learning_rate": 5.567070714468818e-08, + "loss": 0.689, + "step": 39462 + }, + { + "epoch": 1.44, + "learning_rate": 5.566704508001611e-08, + "loss": 0.6958, + "step": 39463 + }, + { + "epoch": 1.44, + "learning_rate": 5.566338301534405e-08, + "loss": 0.699, + "step": 39464 + }, + { + "epoch": 1.44, + "learning_rate": 5.565972095067198e-08, + "loss": 0.7037, + "step": 39465 + }, + { + "epoch": 1.44, + "learning_rate": 5.565605888599992e-08, + "loss": 0.684, + "step": 39466 + }, + { + "epoch": 1.44, + "learning_rate": 5.565239682132786e-08, + "loss": 0.6842, + "step": 39467 + }, + { + "epoch": 1.44, + "learning_rate": 5.56487347566558e-08, + "loss": 0.7021, + "step": 39468 + }, + { + "epoch": 1.44, + "learning_rate": 5.564507269198374e-08, + "loss": 0.7039, + "step": 39469 + }, + { + "epoch": 1.44, + "learning_rate": 5.564141062731168e-08, + "loss": 0.6891, + "step": 39470 + }, + { + "epoch": 1.44, + "learning_rate": 5.563774856263962e-08, + "loss": 0.6966, + "step": 39471 + }, + { + "epoch": 1.44, + "learning_rate": 5.563408649796755e-08, + "loss": 0.6889, + "step": 39472 + }, + { + "epoch": 1.44, + "learning_rate": 5.563042443329549e-08, + "loss": 0.6895, + "step": 39473 + }, + { + "epoch": 1.44, + "learning_rate": 5.562676236862343e-08, + "loss": 0.684, + "step": 39474 + }, + { + "epoch": 1.44, + "learning_rate": 5.562310030395137e-08, + "loss": 0.6921, + "step": 39475 + }, + { + "epoch": 1.44, + "learning_rate": 5.56194382392793e-08, + "loss": 0.7092, + "step": 39476 + }, + { + "epoch": 1.44, + "learning_rate": 5.561577617460724e-08, + "loss": 0.6827, + "step": 39477 + }, + { + "epoch": 1.44, + "learning_rate": 5.561211410993518e-08, + "loss": 0.7008, + "step": 39478 + }, + { + "epoch": 1.44, + "learning_rate": 5.560845204526312e-08, + "loss": 0.7007, + "step": 39479 + }, + { + "epoch": 1.44, + "learning_rate": 5.560478998059105e-08, + "loss": 0.7104, + "step": 39480 + }, + { + "epoch": 1.44, + "learning_rate": 5.560112791591899e-08, + "loss": 0.7069, + "step": 39481 + }, + { + "epoch": 1.44, + "learning_rate": 5.559746585124693e-08, + "loss": 0.6879, + "step": 39482 + }, + { + "epoch": 1.44, + "learning_rate": 5.559380378657487e-08, + "loss": 0.6943, + "step": 39483 + }, + { + "epoch": 1.44, + "learning_rate": 5.559014172190281e-08, + "loss": 0.6937, + "step": 39484 + }, + { + "epoch": 1.44, + "learning_rate": 5.558647965723075e-08, + "loss": 0.698, + "step": 39485 + }, + { + "epoch": 1.44, + "learning_rate": 5.558281759255869e-08, + "loss": 0.6875, + "step": 39486 + }, + { + "epoch": 1.44, + "learning_rate": 5.557915552788662e-08, + "loss": 0.6989, + "step": 39487 + }, + { + "epoch": 1.44, + "learning_rate": 5.557549346321456e-08, + "loss": 0.6943, + "step": 39488 + }, + { + "epoch": 1.44, + "learning_rate": 5.5571831398542493e-08, + "loss": 0.6911, + "step": 39489 + }, + { + "epoch": 1.44, + "learning_rate": 5.5568169333870427e-08, + "loss": 0.6994, + "step": 39490 + }, + { + "epoch": 1.44, + "learning_rate": 5.5564507269198367e-08, + "loss": 0.6908, + "step": 39491 + }, + { + "epoch": 1.44, + "learning_rate": 5.5560845204526306e-08, + "loss": 0.6781, + "step": 39492 + }, + { + "epoch": 1.44, + "learning_rate": 5.5557183139854246e-08, + "loss": 0.6953, + "step": 39493 + }, + { + "epoch": 1.44, + "learning_rate": 5.5553521075182186e-08, + "loss": 0.6898, + "step": 39494 + }, + { + "epoch": 1.45, + "learning_rate": 5.5549859010510126e-08, + "loss": 0.6756, + "step": 39495 + }, + { + "epoch": 1.45, + "learning_rate": 5.5546196945838066e-08, + "loss": 0.6934, + "step": 39496 + }, + { + "epoch": 1.45, + "learning_rate": 5.5542534881166006e-08, + "loss": 0.699, + "step": 39497 + }, + { + "epoch": 1.45, + "learning_rate": 5.553887281649393e-08, + "loss": 0.6906, + "step": 39498 + }, + { + "epoch": 1.45, + "learning_rate": 5.553521075182187e-08, + "loss": 0.7023, + "step": 39499 + }, + { + "epoch": 1.45, + "learning_rate": 5.553154868714981e-08, + "loss": 0.6809, + "step": 39500 + }, + { + "epoch": 1.45, + "learning_rate": 5.5527886622477746e-08, + "loss": 0.6882, + "step": 39501 + }, + { + "epoch": 1.45, + "learning_rate": 5.5524224557805686e-08, + "loss": 0.7042, + "step": 39502 + }, + { + "epoch": 1.45, + "learning_rate": 5.5520562493133626e-08, + "loss": 0.6959, + "step": 39503 + }, + { + "epoch": 1.45, + "learning_rate": 5.5516900428461566e-08, + "loss": 0.7031, + "step": 39504 + }, + { + "epoch": 1.45, + "learning_rate": 5.5513238363789506e-08, + "loss": 0.705, + "step": 39505 + }, + { + "epoch": 1.45, + "learning_rate": 5.5509576299117446e-08, + "loss": 0.7006, + "step": 39506 + }, + { + "epoch": 1.45, + "learning_rate": 5.550591423444537e-08, + "loss": 0.696, + "step": 39507 + }, + { + "epoch": 1.45, + "learning_rate": 5.550225216977331e-08, + "loss": 0.6898, + "step": 39508 + }, + { + "epoch": 1.45, + "learning_rate": 5.549859010510125e-08, + "loss": 0.6811, + "step": 39509 + }, + { + "epoch": 1.45, + "learning_rate": 5.549492804042919e-08, + "loss": 0.6871, + "step": 39510 + }, + { + "epoch": 1.45, + "learning_rate": 5.549126597575713e-08, + "loss": 0.6912, + "step": 39511 + }, + { + "epoch": 1.45, + "learning_rate": 5.5487603911085065e-08, + "loss": 0.6774, + "step": 39512 + }, + { + "epoch": 1.45, + "learning_rate": 5.5483941846413005e-08, + "loss": 0.6796, + "step": 39513 + }, + { + "epoch": 1.45, + "learning_rate": 5.5480279781740945e-08, + "loss": 0.6884, + "step": 39514 + }, + { + "epoch": 1.45, + "learning_rate": 5.5476617717068885e-08, + "loss": 0.6925, + "step": 39515 + }, + { + "epoch": 1.45, + "learning_rate": 5.547295565239681e-08, + "loss": 0.7064, + "step": 39516 + }, + { + "epoch": 1.45, + "learning_rate": 5.546929358772475e-08, + "loss": 0.7017, + "step": 39517 + }, + { + "epoch": 1.45, + "learning_rate": 5.546563152305269e-08, + "loss": 0.6969, + "step": 39518 + }, + { + "epoch": 1.45, + "learning_rate": 5.546196945838063e-08, + "loss": 0.6823, + "step": 39519 + }, + { + "epoch": 1.45, + "learning_rate": 5.545830739370857e-08, + "loss": 0.6968, + "step": 39520 + }, + { + "epoch": 1.45, + "learning_rate": 5.545464532903651e-08, + "loss": 0.6759, + "step": 39521 + }, + { + "epoch": 1.45, + "learning_rate": 5.545098326436445e-08, + "loss": 0.6975, + "step": 39522 + }, + { + "epoch": 1.45, + "learning_rate": 5.5447321199692385e-08, + "loss": 0.6845, + "step": 39523 + }, + { + "epoch": 1.45, + "learning_rate": 5.5443659135020325e-08, + "loss": 0.6912, + "step": 39524 + }, + { + "epoch": 1.45, + "learning_rate": 5.543999707034825e-08, + "loss": 0.6922, + "step": 39525 + }, + { + "epoch": 1.45, + "learning_rate": 5.543633500567619e-08, + "loss": 0.6904, + "step": 39526 + }, + { + "epoch": 1.45, + "learning_rate": 5.543267294100413e-08, + "loss": 0.7054, + "step": 39527 + }, + { + "epoch": 1.45, + "learning_rate": 5.542901087633207e-08, + "loss": 0.698, + "step": 39528 + }, + { + "epoch": 1.45, + "learning_rate": 5.542534881166001e-08, + "loss": 0.7009, + "step": 39529 + }, + { + "epoch": 1.45, + "learning_rate": 5.542168674698795e-08, + "loss": 0.6927, + "step": 39530 + }, + { + "epoch": 1.45, + "learning_rate": 5.541802468231589e-08, + "loss": 0.6941, + "step": 39531 + }, + { + "epoch": 1.45, + "learning_rate": 5.541436261764383e-08, + "loss": 0.6898, + "step": 39532 + }, + { + "epoch": 1.45, + "learning_rate": 5.541070055297177e-08, + "loss": 0.6981, + "step": 39533 + }, + { + "epoch": 1.45, + "learning_rate": 5.54070384882997e-08, + "loss": 0.687, + "step": 39534 + }, + { + "epoch": 1.45, + "learning_rate": 5.540337642362764e-08, + "loss": 0.6844, + "step": 39535 + }, + { + "epoch": 1.45, + "learning_rate": 5.539971435895557e-08, + "loss": 0.7072, + "step": 39536 + }, + { + "epoch": 1.45, + "learning_rate": 5.539605229428351e-08, + "loss": 0.6936, + "step": 39537 + }, + { + "epoch": 1.45, + "learning_rate": 5.539239022961145e-08, + "loss": 0.6877, + "step": 39538 + }, + { + "epoch": 1.45, + "learning_rate": 5.538872816493939e-08, + "loss": 0.6846, + "step": 39539 + }, + { + "epoch": 1.45, + "learning_rate": 5.538506610026733e-08, + "loss": 0.6933, + "step": 39540 + }, + { + "epoch": 1.45, + "learning_rate": 5.538140403559527e-08, + "loss": 0.6829, + "step": 39541 + }, + { + "epoch": 1.45, + "learning_rate": 5.537774197092321e-08, + "loss": 0.695, + "step": 39542 + }, + { + "epoch": 1.45, + "learning_rate": 5.537407990625114e-08, + "loss": 0.6842, + "step": 39543 + }, + { + "epoch": 1.45, + "learning_rate": 5.537041784157908e-08, + "loss": 0.6914, + "step": 39544 + }, + { + "epoch": 1.45, + "learning_rate": 5.536675577690702e-08, + "loss": 0.6914, + "step": 39545 + }, + { + "epoch": 1.45, + "learning_rate": 5.536309371223496e-08, + "loss": 0.6759, + "step": 39546 + }, + { + "epoch": 1.45, + "learning_rate": 5.535943164756289e-08, + "loss": 0.6907, + "step": 39547 + }, + { + "epoch": 1.45, + "learning_rate": 5.535576958289083e-08, + "loss": 0.7033, + "step": 39548 + }, + { + "epoch": 1.45, + "learning_rate": 5.535210751821877e-08, + "loss": 0.6965, + "step": 39549 + }, + { + "epoch": 1.45, + "learning_rate": 5.534844545354671e-08, + "loss": 0.6892, + "step": 39550 + }, + { + "epoch": 1.45, + "learning_rate": 5.534478338887465e-08, + "loss": 0.7041, + "step": 39551 + }, + { + "epoch": 1.45, + "learning_rate": 5.5341121324202576e-08, + "loss": 0.6939, + "step": 39552 + }, + { + "epoch": 1.45, + "learning_rate": 5.5337459259530516e-08, + "loss": 0.6863, + "step": 39553 + }, + { + "epoch": 1.45, + "learning_rate": 5.5333797194858456e-08, + "loss": 0.6998, + "step": 39554 + }, + { + "epoch": 1.45, + "learning_rate": 5.5330135130186396e-08, + "loss": 0.6878, + "step": 39555 + }, + { + "epoch": 1.45, + "learning_rate": 5.5326473065514336e-08, + "loss": 0.6729, + "step": 39556 + }, + { + "epoch": 1.45, + "learning_rate": 5.5322811000842276e-08, + "loss": 0.6941, + "step": 39557 + }, + { + "epoch": 1.45, + "learning_rate": 5.531914893617021e-08, + "loss": 0.7041, + "step": 39558 + }, + { + "epoch": 1.45, + "learning_rate": 5.531548687149815e-08, + "loss": 0.6748, + "step": 39559 + }, + { + "epoch": 1.45, + "learning_rate": 5.531182480682609e-08, + "loss": 0.6924, + "step": 39560 + }, + { + "epoch": 1.45, + "learning_rate": 5.5308162742154016e-08, + "loss": 0.6878, + "step": 39561 + }, + { + "epoch": 1.45, + "learning_rate": 5.5304500677481956e-08, + "loss": 0.6876, + "step": 39562 + }, + { + "epoch": 1.45, + "learning_rate": 5.5300838612809896e-08, + "loss": 0.705, + "step": 39563 + }, + { + "epoch": 1.45, + "learning_rate": 5.5297176548137836e-08, + "loss": 0.7175, + "step": 39564 + }, + { + "epoch": 1.45, + "learning_rate": 5.5293514483465776e-08, + "loss": 0.6996, + "step": 39565 + }, + { + "epoch": 1.45, + "learning_rate": 5.5289852418793716e-08, + "loss": 0.7096, + "step": 39566 + }, + { + "epoch": 1.45, + "learning_rate": 5.5286190354121656e-08, + "loss": 0.6999, + "step": 39567 + }, + { + "epoch": 1.45, + "learning_rate": 5.5282528289449595e-08, + "loss": 0.7008, + "step": 39568 + }, + { + "epoch": 1.45, + "learning_rate": 5.527886622477753e-08, + "loss": 0.6943, + "step": 39569 + }, + { + "epoch": 1.45, + "learning_rate": 5.527520416010546e-08, + "loss": 0.7094, + "step": 39570 + }, + { + "epoch": 1.45, + "learning_rate": 5.52715420954334e-08, + "loss": 0.6915, + "step": 39571 + }, + { + "epoch": 1.45, + "learning_rate": 5.5267880030761335e-08, + "loss": 0.6896, + "step": 39572 + }, + { + "epoch": 1.45, + "learning_rate": 5.5264217966089275e-08, + "loss": 0.6932, + "step": 39573 + }, + { + "epoch": 1.45, + "learning_rate": 5.5260555901417215e-08, + "loss": 0.6886, + "step": 39574 + }, + { + "epoch": 1.45, + "learning_rate": 5.5256893836745155e-08, + "loss": 0.6965, + "step": 39575 + }, + { + "epoch": 1.45, + "learning_rate": 5.5253231772073095e-08, + "loss": 0.6905, + "step": 39576 + }, + { + "epoch": 1.45, + "learning_rate": 5.5249569707401035e-08, + "loss": 0.7025, + "step": 39577 + }, + { + "epoch": 1.45, + "learning_rate": 5.5245907642728975e-08, + "loss": 0.6937, + "step": 39578 + }, + { + "epoch": 1.45, + "learning_rate": 5.52422455780569e-08, + "loss": 0.7075, + "step": 39579 + }, + { + "epoch": 1.45, + "learning_rate": 5.523858351338484e-08, + "loss": 0.6894, + "step": 39580 + }, + { + "epoch": 1.45, + "learning_rate": 5.523492144871278e-08, + "loss": 0.6975, + "step": 39581 + }, + { + "epoch": 1.45, + "learning_rate": 5.523125938404072e-08, + "loss": 0.6839, + "step": 39582 + }, + { + "epoch": 1.45, + "learning_rate": 5.5227597319368655e-08, + "loss": 0.6918, + "step": 39583 + }, + { + "epoch": 1.45, + "learning_rate": 5.5223935254696595e-08, + "loss": 0.6958, + "step": 39584 + }, + { + "epoch": 1.45, + "learning_rate": 5.5220273190024535e-08, + "loss": 0.6996, + "step": 39585 + }, + { + "epoch": 1.45, + "learning_rate": 5.5216611125352474e-08, + "loss": 0.6848, + "step": 39586 + }, + { + "epoch": 1.45, + "learning_rate": 5.5212949060680414e-08, + "loss": 0.6983, + "step": 39587 + }, + { + "epoch": 1.45, + "learning_rate": 5.520928699600834e-08, + "loss": 0.7037, + "step": 39588 + }, + { + "epoch": 1.45, + "learning_rate": 5.520562493133628e-08, + "loss": 0.6874, + "step": 39589 + }, + { + "epoch": 1.45, + "learning_rate": 5.520196286666422e-08, + "loss": 0.6801, + "step": 39590 + }, + { + "epoch": 1.45, + "learning_rate": 5.519830080199216e-08, + "loss": 0.6911, + "step": 39591 + }, + { + "epoch": 1.45, + "learning_rate": 5.51946387373201e-08, + "loss": 0.6893, + "step": 39592 + }, + { + "epoch": 1.45, + "learning_rate": 5.519097667264804e-08, + "loss": 0.6908, + "step": 39593 + }, + { + "epoch": 1.45, + "learning_rate": 5.5187314607975974e-08, + "loss": 0.6824, + "step": 39594 + }, + { + "epoch": 1.45, + "learning_rate": 5.5183652543303914e-08, + "loss": 0.6938, + "step": 39595 + }, + { + "epoch": 1.45, + "learning_rate": 5.5179990478631854e-08, + "loss": 0.6805, + "step": 39596 + }, + { + "epoch": 1.45, + "learning_rate": 5.517632841395978e-08, + "loss": 0.6976, + "step": 39597 + }, + { + "epoch": 1.45, + "learning_rate": 5.517266634928772e-08, + "loss": 0.6941, + "step": 39598 + }, + { + "epoch": 1.45, + "learning_rate": 5.516900428461566e-08, + "loss": 0.6895, + "step": 39599 + }, + { + "epoch": 1.45, + "learning_rate": 5.51653422199436e-08, + "loss": 0.6944, + "step": 39600 + }, + { + "epoch": 1.45, + "learning_rate": 5.516168015527154e-08, + "loss": 0.6832, + "step": 39601 + }, + { + "epoch": 1.45, + "learning_rate": 5.515801809059948e-08, + "loss": 0.7041, + "step": 39602 + }, + { + "epoch": 1.45, + "learning_rate": 5.515435602592742e-08, + "loss": 0.6922, + "step": 39603 + }, + { + "epoch": 1.45, + "learning_rate": 5.5150693961255353e-08, + "loss": 0.6754, + "step": 39604 + }, + { + "epoch": 1.45, + "learning_rate": 5.5147031896583293e-08, + "loss": 0.6958, + "step": 39605 + }, + { + "epoch": 1.45, + "learning_rate": 5.514336983191123e-08, + "loss": 0.6954, + "step": 39606 + }, + { + "epoch": 1.45, + "learning_rate": 5.513970776723916e-08, + "loss": 0.6979, + "step": 39607 + }, + { + "epoch": 1.45, + "learning_rate": 5.51360457025671e-08, + "loss": 0.6962, + "step": 39608 + }, + { + "epoch": 1.45, + "learning_rate": 5.513238363789504e-08, + "loss": 0.6779, + "step": 39609 + }, + { + "epoch": 1.45, + "learning_rate": 5.512872157322298e-08, + "loss": 0.6866, + "step": 39610 + }, + { + "epoch": 1.45, + "learning_rate": 5.512505950855092e-08, + "loss": 0.7116, + "step": 39611 + }, + { + "epoch": 1.45, + "learning_rate": 5.512139744387886e-08, + "loss": 0.6867, + "step": 39612 + }, + { + "epoch": 1.45, + "learning_rate": 5.51177353792068e-08, + "loss": 0.7036, + "step": 39613 + }, + { + "epoch": 1.45, + "learning_rate": 5.511407331453474e-08, + "loss": 0.6929, + "step": 39614 + }, + { + "epoch": 1.45, + "learning_rate": 5.5110411249862666e-08, + "loss": 0.6942, + "step": 39615 + }, + { + "epoch": 1.45, + "learning_rate": 5.5106749185190606e-08, + "loss": 0.6876, + "step": 39616 + }, + { + "epoch": 1.45, + "learning_rate": 5.5103087120518546e-08, + "loss": 0.7173, + "step": 39617 + }, + { + "epoch": 1.45, + "learning_rate": 5.509942505584648e-08, + "loss": 0.7018, + "step": 39618 + }, + { + "epoch": 1.45, + "learning_rate": 5.509576299117442e-08, + "loss": 0.7006, + "step": 39619 + }, + { + "epoch": 1.45, + "learning_rate": 5.509210092650236e-08, + "loss": 0.6859, + "step": 39620 + }, + { + "epoch": 1.45, + "learning_rate": 5.50884388618303e-08, + "loss": 0.6994, + "step": 39621 + }, + { + "epoch": 1.45, + "learning_rate": 5.508477679715824e-08, + "loss": 0.6897, + "step": 39622 + }, + { + "epoch": 1.45, + "learning_rate": 5.508111473248618e-08, + "loss": 0.6944, + "step": 39623 + }, + { + "epoch": 1.45, + "learning_rate": 5.5077452667814106e-08, + "loss": 0.6994, + "step": 39624 + }, + { + "epoch": 1.45, + "learning_rate": 5.5073790603142046e-08, + "loss": 0.6872, + "step": 39625 + }, + { + "epoch": 1.45, + "learning_rate": 5.5070128538469986e-08, + "loss": 0.6734, + "step": 39626 + }, + { + "epoch": 1.45, + "learning_rate": 5.5066466473797926e-08, + "loss": 0.694, + "step": 39627 + }, + { + "epoch": 1.45, + "learning_rate": 5.5062804409125865e-08, + "loss": 0.6981, + "step": 39628 + }, + { + "epoch": 1.45, + "learning_rate": 5.50591423444538e-08, + "loss": 0.6913, + "step": 39629 + }, + { + "epoch": 1.45, + "learning_rate": 5.505548027978174e-08, + "loss": 0.6951, + "step": 39630 + }, + { + "epoch": 1.45, + "learning_rate": 5.505181821510968e-08, + "loss": 0.6833, + "step": 39631 + }, + { + "epoch": 1.45, + "learning_rate": 5.504815615043762e-08, + "loss": 0.701, + "step": 39632 + }, + { + "epoch": 1.45, + "learning_rate": 5.5044494085765545e-08, + "loss": 0.69, + "step": 39633 + }, + { + "epoch": 1.45, + "learning_rate": 5.5040832021093485e-08, + "loss": 0.6756, + "step": 39634 + }, + { + "epoch": 1.45, + "learning_rate": 5.5037169956421425e-08, + "loss": 0.7122, + "step": 39635 + }, + { + "epoch": 1.45, + "learning_rate": 5.5033507891749365e-08, + "loss": 0.691, + "step": 39636 + }, + { + "epoch": 1.45, + "learning_rate": 5.5029845827077305e-08, + "loss": 0.6918, + "step": 39637 + }, + { + "epoch": 1.45, + "learning_rate": 5.5026183762405245e-08, + "loss": 0.6981, + "step": 39638 + }, + { + "epoch": 1.45, + "learning_rate": 5.5022521697733185e-08, + "loss": 0.6887, + "step": 39639 + }, + { + "epoch": 1.45, + "learning_rate": 5.501885963306112e-08, + "loss": 0.7075, + "step": 39640 + }, + { + "epoch": 1.45, + "learning_rate": 5.501519756838906e-08, + "loss": 0.6902, + "step": 39641 + }, + { + "epoch": 1.45, + "learning_rate": 5.501153550371699e-08, + "loss": 0.6985, + "step": 39642 + }, + { + "epoch": 1.45, + "learning_rate": 5.5007873439044925e-08, + "loss": 0.6858, + "step": 39643 + }, + { + "epoch": 1.45, + "learning_rate": 5.5004211374372865e-08, + "loss": 0.6875, + "step": 39644 + }, + { + "epoch": 1.45, + "learning_rate": 5.5000549309700805e-08, + "loss": 0.7005, + "step": 39645 + }, + { + "epoch": 1.45, + "learning_rate": 5.4996887245028744e-08, + "loss": 0.6888, + "step": 39646 + }, + { + "epoch": 1.45, + "learning_rate": 5.4993225180356684e-08, + "loss": 0.6947, + "step": 39647 + }, + { + "epoch": 1.45, + "learning_rate": 5.4989563115684624e-08, + "loss": 0.6918, + "step": 39648 + }, + { + "epoch": 1.45, + "learning_rate": 5.4985901051012564e-08, + "loss": 0.6935, + "step": 39649 + }, + { + "epoch": 1.45, + "learning_rate": 5.4982238986340504e-08, + "loss": 0.6828, + "step": 39650 + }, + { + "epoch": 1.45, + "learning_rate": 5.497857692166843e-08, + "loss": 0.6871, + "step": 39651 + }, + { + "epoch": 1.45, + "learning_rate": 5.497491485699637e-08, + "loss": 0.7035, + "step": 39652 + }, + { + "epoch": 1.45, + "learning_rate": 5.497125279232431e-08, + "loss": 0.6869, + "step": 39653 + }, + { + "epoch": 1.45, + "learning_rate": 5.4967590727652244e-08, + "loss": 0.6905, + "step": 39654 + }, + { + "epoch": 1.45, + "learning_rate": 5.4963928662980184e-08, + "loss": 0.6936, + "step": 39655 + }, + { + "epoch": 1.45, + "learning_rate": 5.4960266598308124e-08, + "loss": 0.688, + "step": 39656 + }, + { + "epoch": 1.45, + "learning_rate": 5.4956604533636064e-08, + "loss": 0.7063, + "step": 39657 + }, + { + "epoch": 1.45, + "learning_rate": 5.4952942468964004e-08, + "loss": 0.7006, + "step": 39658 + }, + { + "epoch": 1.45, + "learning_rate": 5.4949280404291944e-08, + "loss": 0.702, + "step": 39659 + }, + { + "epoch": 1.45, + "learning_rate": 5.494561833961987e-08, + "loss": 0.6929, + "step": 39660 + }, + { + "epoch": 1.45, + "learning_rate": 5.494195627494781e-08, + "loss": 0.708, + "step": 39661 + }, + { + "epoch": 1.45, + "learning_rate": 5.493829421027575e-08, + "loss": 0.6904, + "step": 39662 + }, + { + "epoch": 1.45, + "learning_rate": 5.493463214560369e-08, + "loss": 0.689, + "step": 39663 + }, + { + "epoch": 1.45, + "learning_rate": 5.4930970080931623e-08, + "loss": 0.686, + "step": 39664 + }, + { + "epoch": 1.45, + "learning_rate": 5.4927308016259563e-08, + "loss": 0.6942, + "step": 39665 + }, + { + "epoch": 1.45, + "learning_rate": 5.4923645951587503e-08, + "loss": 0.7036, + "step": 39666 + }, + { + "epoch": 1.45, + "learning_rate": 5.491998388691544e-08, + "loss": 0.703, + "step": 39667 + }, + { + "epoch": 1.45, + "learning_rate": 5.491632182224338e-08, + "loss": 0.6899, + "step": 39668 + }, + { + "epoch": 1.45, + "learning_rate": 5.491265975757131e-08, + "loss": 0.6894, + "step": 39669 + }, + { + "epoch": 1.45, + "learning_rate": 5.490899769289925e-08, + "loss": 0.6966, + "step": 39670 + }, + { + "epoch": 1.45, + "learning_rate": 5.490533562822719e-08, + "loss": 0.6978, + "step": 39671 + }, + { + "epoch": 1.45, + "learning_rate": 5.490167356355513e-08, + "loss": 0.6955, + "step": 39672 + }, + { + "epoch": 1.45, + "learning_rate": 5.489801149888307e-08, + "loss": 0.6798, + "step": 39673 + }, + { + "epoch": 1.45, + "learning_rate": 5.489434943421101e-08, + "loss": 0.6978, + "step": 39674 + }, + { + "epoch": 1.45, + "learning_rate": 5.489068736953894e-08, + "loss": 0.705, + "step": 39675 + }, + { + "epoch": 1.45, + "learning_rate": 5.488702530486688e-08, + "loss": 0.6883, + "step": 39676 + }, + { + "epoch": 1.45, + "learning_rate": 5.488336324019482e-08, + "loss": 0.6951, + "step": 39677 + }, + { + "epoch": 1.45, + "learning_rate": 5.487970117552275e-08, + "loss": 0.6886, + "step": 39678 + }, + { + "epoch": 1.45, + "learning_rate": 5.487603911085069e-08, + "loss": 0.6947, + "step": 39679 + }, + { + "epoch": 1.45, + "learning_rate": 5.487237704617863e-08, + "loss": 0.684, + "step": 39680 + }, + { + "epoch": 1.45, + "learning_rate": 5.486871498150657e-08, + "loss": 0.6936, + "step": 39681 + }, + { + "epoch": 1.45, + "learning_rate": 5.486505291683451e-08, + "loss": 0.6825, + "step": 39682 + }, + { + "epoch": 1.45, + "learning_rate": 5.486139085216245e-08, + "loss": 0.6934, + "step": 39683 + }, + { + "epoch": 1.45, + "learning_rate": 5.485772878749039e-08, + "loss": 0.6905, + "step": 39684 + }, + { + "epoch": 1.45, + "learning_rate": 5.485406672281833e-08, + "loss": 0.7, + "step": 39685 + }, + { + "epoch": 1.45, + "learning_rate": 5.485040465814626e-08, + "loss": 0.705, + "step": 39686 + }, + { + "epoch": 1.45, + "learning_rate": 5.4846742593474196e-08, + "loss": 0.6853, + "step": 39687 + }, + { + "epoch": 1.45, + "learning_rate": 5.4843080528802135e-08, + "loss": 0.6918, + "step": 39688 + }, + { + "epoch": 1.45, + "learning_rate": 5.483941846413007e-08, + "loss": 0.7018, + "step": 39689 + }, + { + "epoch": 1.45, + "learning_rate": 5.483575639945801e-08, + "loss": 0.6807, + "step": 39690 + }, + { + "epoch": 1.45, + "learning_rate": 5.483209433478595e-08, + "loss": 0.6773, + "step": 39691 + }, + { + "epoch": 1.45, + "learning_rate": 5.482843227011389e-08, + "loss": 0.7006, + "step": 39692 + }, + { + "epoch": 1.45, + "learning_rate": 5.482477020544183e-08, + "loss": 0.6851, + "step": 39693 + }, + { + "epoch": 1.45, + "learning_rate": 5.482110814076977e-08, + "loss": 0.6804, + "step": 39694 + }, + { + "epoch": 1.45, + "learning_rate": 5.481744607609771e-08, + "loss": 0.6996, + "step": 39695 + }, + { + "epoch": 1.45, + "learning_rate": 5.4813784011425635e-08, + "loss": 0.6867, + "step": 39696 + }, + { + "epoch": 1.45, + "learning_rate": 5.4810121946753575e-08, + "loss": 0.6979, + "step": 39697 + }, + { + "epoch": 1.45, + "learning_rate": 5.4806459882081515e-08, + "loss": 0.6907, + "step": 39698 + }, + { + "epoch": 1.45, + "learning_rate": 5.4802797817409455e-08, + "loss": 0.6919, + "step": 39699 + }, + { + "epoch": 1.45, + "learning_rate": 5.479913575273739e-08, + "loss": 0.6998, + "step": 39700 + }, + { + "epoch": 1.45, + "learning_rate": 5.479547368806533e-08, + "loss": 0.6946, + "step": 39701 + }, + { + "epoch": 1.45, + "learning_rate": 5.479181162339327e-08, + "loss": 0.6916, + "step": 39702 + }, + { + "epoch": 1.45, + "learning_rate": 5.478814955872121e-08, + "loss": 0.6696, + "step": 39703 + }, + { + "epoch": 1.45, + "learning_rate": 5.478448749404915e-08, + "loss": 0.685, + "step": 39704 + }, + { + "epoch": 1.45, + "learning_rate": 5.4780825429377075e-08, + "loss": 0.6836, + "step": 39705 + }, + { + "epoch": 1.45, + "learning_rate": 5.4777163364705014e-08, + "loss": 0.6938, + "step": 39706 + }, + { + "epoch": 1.45, + "learning_rate": 5.4773501300032954e-08, + "loss": 0.7033, + "step": 39707 + }, + { + "epoch": 1.45, + "learning_rate": 5.4769839235360894e-08, + "loss": 0.6872, + "step": 39708 + }, + { + "epoch": 1.45, + "learning_rate": 5.4766177170688834e-08, + "loss": 0.6982, + "step": 39709 + }, + { + "epoch": 1.45, + "learning_rate": 5.4762515106016774e-08, + "loss": 0.6961, + "step": 39710 + }, + { + "epoch": 1.45, + "learning_rate": 5.475885304134471e-08, + "loss": 0.6891, + "step": 39711 + }, + { + "epoch": 1.45, + "learning_rate": 5.475519097667265e-08, + "loss": 0.6883, + "step": 39712 + }, + { + "epoch": 1.45, + "learning_rate": 5.475152891200059e-08, + "loss": 0.692, + "step": 39713 + }, + { + "epoch": 1.45, + "learning_rate": 5.4747866847328514e-08, + "loss": 0.6989, + "step": 39714 + }, + { + "epoch": 1.45, + "learning_rate": 5.4744204782656454e-08, + "loss": 0.6854, + "step": 39715 + }, + { + "epoch": 1.45, + "learning_rate": 5.4740542717984394e-08, + "loss": 0.7021, + "step": 39716 + }, + { + "epoch": 1.45, + "learning_rate": 5.4736880653312334e-08, + "loss": 0.6946, + "step": 39717 + }, + { + "epoch": 1.45, + "learning_rate": 5.4733218588640274e-08, + "loss": 0.6976, + "step": 39718 + }, + { + "epoch": 1.45, + "learning_rate": 5.4729556523968214e-08, + "loss": 0.6932, + "step": 39719 + }, + { + "epoch": 1.45, + "learning_rate": 5.4725894459296154e-08, + "loss": 0.7032, + "step": 39720 + }, + { + "epoch": 1.45, + "learning_rate": 5.4722232394624094e-08, + "loss": 0.6901, + "step": 39721 + }, + { + "epoch": 1.45, + "learning_rate": 5.471857032995203e-08, + "loss": 0.6843, + "step": 39722 + }, + { + "epoch": 1.45, + "learning_rate": 5.471490826527996e-08, + "loss": 0.6966, + "step": 39723 + }, + { + "epoch": 1.45, + "learning_rate": 5.4711246200607893e-08, + "loss": 0.6984, + "step": 39724 + }, + { + "epoch": 1.45, + "learning_rate": 5.4707584135935833e-08, + "loss": 0.692, + "step": 39725 + }, + { + "epoch": 1.45, + "learning_rate": 5.4703922071263773e-08, + "loss": 0.6897, + "step": 39726 + }, + { + "epoch": 1.45, + "learning_rate": 5.470026000659171e-08, + "loss": 0.6986, + "step": 39727 + }, + { + "epoch": 1.45, + "learning_rate": 5.469659794191965e-08, + "loss": 0.6918, + "step": 39728 + }, + { + "epoch": 1.45, + "learning_rate": 5.469293587724759e-08, + "loss": 0.6888, + "step": 39729 + }, + { + "epoch": 1.45, + "learning_rate": 5.468927381257553e-08, + "loss": 0.6857, + "step": 39730 + }, + { + "epoch": 1.45, + "learning_rate": 5.468561174790347e-08, + "loss": 0.6991, + "step": 39731 + }, + { + "epoch": 1.45, + "learning_rate": 5.46819496832314e-08, + "loss": 0.6887, + "step": 39732 + }, + { + "epoch": 1.45, + "learning_rate": 5.467828761855934e-08, + "loss": 0.6951, + "step": 39733 + }, + { + "epoch": 1.45, + "learning_rate": 5.467462555388728e-08, + "loss": 0.6801, + "step": 39734 + }, + { + "epoch": 1.45, + "learning_rate": 5.467096348921521e-08, + "loss": 0.6993, + "step": 39735 + }, + { + "epoch": 1.45, + "learning_rate": 5.466730142454315e-08, + "loss": 0.7042, + "step": 39736 + }, + { + "epoch": 1.45, + "learning_rate": 5.466363935987109e-08, + "loss": 0.6885, + "step": 39737 + }, + { + "epoch": 1.45, + "learning_rate": 5.465997729519903e-08, + "loss": 0.6952, + "step": 39738 + }, + { + "epoch": 1.45, + "learning_rate": 5.465631523052697e-08, + "loss": 0.6809, + "step": 39739 + }, + { + "epoch": 1.45, + "learning_rate": 5.465265316585491e-08, + "loss": 0.6789, + "step": 39740 + }, + { + "epoch": 1.45, + "learning_rate": 5.464899110118284e-08, + "loss": 0.6823, + "step": 39741 + }, + { + "epoch": 1.45, + "learning_rate": 5.464532903651078e-08, + "loss": 0.6872, + "step": 39742 + }, + { + "epoch": 1.45, + "learning_rate": 5.464166697183872e-08, + "loss": 0.694, + "step": 39743 + }, + { + "epoch": 1.45, + "learning_rate": 5.463800490716666e-08, + "loss": 0.7012, + "step": 39744 + }, + { + "epoch": 1.45, + "learning_rate": 5.46343428424946e-08, + "loss": 0.6987, + "step": 39745 + }, + { + "epoch": 1.45, + "learning_rate": 5.463068077782253e-08, + "loss": 0.7007, + "step": 39746 + }, + { + "epoch": 1.45, + "learning_rate": 5.462701871315047e-08, + "loss": 0.6967, + "step": 39747 + }, + { + "epoch": 1.45, + "learning_rate": 5.462335664847841e-08, + "loss": 0.6867, + "step": 39748 + }, + { + "epoch": 1.45, + "learning_rate": 5.461969458380635e-08, + "loss": 0.6964, + "step": 39749 + }, + { + "epoch": 1.45, + "learning_rate": 5.461603251913428e-08, + "loss": 0.6933, + "step": 39750 + }, + { + "epoch": 1.45, + "learning_rate": 5.461237045446222e-08, + "loss": 0.6766, + "step": 39751 + }, + { + "epoch": 1.45, + "learning_rate": 5.460870838979016e-08, + "loss": 0.6884, + "step": 39752 + }, + { + "epoch": 1.45, + "learning_rate": 5.46050463251181e-08, + "loss": 0.6824, + "step": 39753 + }, + { + "epoch": 1.45, + "learning_rate": 5.460138426044604e-08, + "loss": 0.6942, + "step": 39754 + }, + { + "epoch": 1.45, + "learning_rate": 5.459772219577398e-08, + "loss": 0.684, + "step": 39755 + }, + { + "epoch": 1.45, + "learning_rate": 5.459406013110192e-08, + "loss": 0.6917, + "step": 39756 + }, + { + "epoch": 1.45, + "learning_rate": 5.459039806642985e-08, + "loss": 0.6894, + "step": 39757 + }, + { + "epoch": 1.45, + "learning_rate": 5.458673600175779e-08, + "loss": 0.6936, + "step": 39758 + }, + { + "epoch": 1.45, + "learning_rate": 5.4583073937085725e-08, + "loss": 0.6866, + "step": 39759 + }, + { + "epoch": 1.45, + "learning_rate": 5.457941187241366e-08, + "loss": 0.7005, + "step": 39760 + }, + { + "epoch": 1.45, + "learning_rate": 5.45757498077416e-08, + "loss": 0.6928, + "step": 39761 + }, + { + "epoch": 1.45, + "learning_rate": 5.457208774306954e-08, + "loss": 0.695, + "step": 39762 + }, + { + "epoch": 1.45, + "learning_rate": 5.456842567839748e-08, + "loss": 0.6953, + "step": 39763 + }, + { + "epoch": 1.45, + "learning_rate": 5.456476361372542e-08, + "loss": 0.6934, + "step": 39764 + }, + { + "epoch": 1.45, + "learning_rate": 5.456110154905336e-08, + "loss": 0.7023, + "step": 39765 + }, + { + "epoch": 1.45, + "learning_rate": 5.45574394843813e-08, + "loss": 0.6923, + "step": 39766 + }, + { + "epoch": 1.45, + "learning_rate": 5.455377741970924e-08, + "loss": 0.6894, + "step": 39767 + }, + { + "epoch": 1.45, + "learning_rate": 5.4550115355037164e-08, + "loss": 0.7088, + "step": 39768 + }, + { + "epoch": 1.46, + "learning_rate": 5.4546453290365104e-08, + "loss": 0.6992, + "step": 39769 + }, + { + "epoch": 1.46, + "learning_rate": 5.4542791225693044e-08, + "loss": 0.7048, + "step": 39770 + }, + { + "epoch": 1.46, + "learning_rate": 5.453912916102098e-08, + "loss": 0.6923, + "step": 39771 + }, + { + "epoch": 1.46, + "learning_rate": 5.453546709634892e-08, + "loss": 0.6762, + "step": 39772 + }, + { + "epoch": 1.46, + "learning_rate": 5.453180503167686e-08, + "loss": 0.6971, + "step": 39773 + }, + { + "epoch": 1.46, + "learning_rate": 5.45281429670048e-08, + "loss": 0.699, + "step": 39774 + }, + { + "epoch": 1.46, + "learning_rate": 5.452448090233274e-08, + "loss": 0.694, + "step": 39775 + }, + { + "epoch": 1.46, + "learning_rate": 5.452081883766068e-08, + "loss": 0.6981, + "step": 39776 + }, + { + "epoch": 1.46, + "learning_rate": 5.4517156772988604e-08, + "loss": 0.7036, + "step": 39777 + }, + { + "epoch": 1.46, + "learning_rate": 5.4513494708316544e-08, + "loss": 0.6924, + "step": 39778 + }, + { + "epoch": 1.46, + "learning_rate": 5.4509832643644484e-08, + "loss": 0.6987, + "step": 39779 + }, + { + "epoch": 1.46, + "learning_rate": 5.4506170578972424e-08, + "loss": 0.6949, + "step": 39780 + }, + { + "epoch": 1.46, + "learning_rate": 5.4502508514300364e-08, + "loss": 0.6959, + "step": 39781 + }, + { + "epoch": 1.46, + "learning_rate": 5.44988464496283e-08, + "loss": 0.698, + "step": 39782 + }, + { + "epoch": 1.46, + "learning_rate": 5.449518438495624e-08, + "loss": 0.6991, + "step": 39783 + }, + { + "epoch": 1.46, + "learning_rate": 5.4491522320284177e-08, + "loss": 0.6916, + "step": 39784 + }, + { + "epoch": 1.46, + "learning_rate": 5.4487860255612117e-08, + "loss": 0.6822, + "step": 39785 + }, + { + "epoch": 1.46, + "learning_rate": 5.448419819094004e-08, + "loss": 0.6894, + "step": 39786 + }, + { + "epoch": 1.46, + "learning_rate": 5.448053612626798e-08, + "loss": 0.7104, + "step": 39787 + }, + { + "epoch": 1.46, + "learning_rate": 5.447687406159592e-08, + "loss": 0.6982, + "step": 39788 + }, + { + "epoch": 1.46, + "learning_rate": 5.447321199692386e-08, + "loss": 0.6974, + "step": 39789 + }, + { + "epoch": 1.46, + "learning_rate": 5.44695499322518e-08, + "loss": 0.7022, + "step": 39790 + }, + { + "epoch": 1.46, + "learning_rate": 5.446588786757974e-08, + "loss": 0.6795, + "step": 39791 + }, + { + "epoch": 1.46, + "learning_rate": 5.446222580290768e-08, + "loss": 0.698, + "step": 39792 + }, + { + "epoch": 1.46, + "learning_rate": 5.4458563738235616e-08, + "loss": 0.694, + "step": 39793 + }, + { + "epoch": 1.46, + "learning_rate": 5.4454901673563556e-08, + "loss": 0.6887, + "step": 39794 + }, + { + "epoch": 1.46, + "learning_rate": 5.445123960889148e-08, + "loss": 0.6894, + "step": 39795 + }, + { + "epoch": 1.46, + "learning_rate": 5.444757754421942e-08, + "loss": 0.6755, + "step": 39796 + }, + { + "epoch": 1.46, + "learning_rate": 5.444391547954736e-08, + "loss": 0.7132, + "step": 39797 + }, + { + "epoch": 1.46, + "learning_rate": 5.44402534148753e-08, + "loss": 0.6913, + "step": 39798 + }, + { + "epoch": 1.46, + "learning_rate": 5.443659135020324e-08, + "loss": 0.6775, + "step": 39799 + }, + { + "epoch": 1.46, + "learning_rate": 5.443292928553118e-08, + "loss": 0.6763, + "step": 39800 + }, + { + "epoch": 1.46, + "learning_rate": 5.442926722085912e-08, + "loss": 0.7061, + "step": 39801 + }, + { + "epoch": 1.46, + "learning_rate": 5.442560515618706e-08, + "loss": 0.6967, + "step": 39802 + }, + { + "epoch": 1.46, + "learning_rate": 5.4421943091515e-08, + "loss": 0.7068, + "step": 39803 + }, + { + "epoch": 1.46, + "learning_rate": 5.441828102684293e-08, + "loss": 0.6906, + "step": 39804 + }, + { + "epoch": 1.46, + "learning_rate": 5.441461896217087e-08, + "loss": 0.6948, + "step": 39805 + }, + { + "epoch": 1.46, + "learning_rate": 5.44109568974988e-08, + "loss": 0.688, + "step": 39806 + }, + { + "epoch": 1.46, + "learning_rate": 5.440729483282674e-08, + "loss": 0.6844, + "step": 39807 + }, + { + "epoch": 1.46, + "learning_rate": 5.440363276815468e-08, + "loss": 0.6912, + "step": 39808 + }, + { + "epoch": 1.46, + "learning_rate": 5.439997070348262e-08, + "loss": 0.6687, + "step": 39809 + }, + { + "epoch": 1.46, + "learning_rate": 5.439630863881056e-08, + "loss": 0.6812, + "step": 39810 + }, + { + "epoch": 1.46, + "learning_rate": 5.43926465741385e-08, + "loss": 0.6939, + "step": 39811 + }, + { + "epoch": 1.46, + "learning_rate": 5.438898450946644e-08, + "loss": 0.6889, + "step": 39812 + }, + { + "epoch": 1.46, + "learning_rate": 5.438532244479437e-08, + "loss": 0.6921, + "step": 39813 + }, + { + "epoch": 1.46, + "learning_rate": 5.438166038012231e-08, + "loss": 0.6824, + "step": 39814 + }, + { + "epoch": 1.46, + "learning_rate": 5.437799831545025e-08, + "loss": 0.7001, + "step": 39815 + }, + { + "epoch": 1.46, + "learning_rate": 5.437433625077819e-08, + "loss": 0.6851, + "step": 39816 + }, + { + "epoch": 1.46, + "learning_rate": 5.437067418610612e-08, + "loss": 0.6939, + "step": 39817 + }, + { + "epoch": 1.46, + "learning_rate": 5.436701212143406e-08, + "loss": 0.685, + "step": 39818 + }, + { + "epoch": 1.46, + "learning_rate": 5.4363350056762e-08, + "loss": 0.6815, + "step": 39819 + }, + { + "epoch": 1.46, + "learning_rate": 5.435968799208994e-08, + "loss": 0.6879, + "step": 39820 + }, + { + "epoch": 1.46, + "learning_rate": 5.435602592741788e-08, + "loss": 0.6967, + "step": 39821 + }, + { + "epoch": 1.46, + "learning_rate": 5.435236386274581e-08, + "loss": 0.6822, + "step": 39822 + }, + { + "epoch": 1.46, + "learning_rate": 5.434870179807375e-08, + "loss": 0.6952, + "step": 39823 + }, + { + "epoch": 1.46, + "learning_rate": 5.434503973340169e-08, + "loss": 0.6823, + "step": 39824 + }, + { + "epoch": 1.46, + "learning_rate": 5.434137766872963e-08, + "loss": 0.6933, + "step": 39825 + }, + { + "epoch": 1.46, + "learning_rate": 5.433771560405757e-08, + "loss": 0.6953, + "step": 39826 + }, + { + "epoch": 1.46, + "learning_rate": 5.433405353938551e-08, + "loss": 0.6936, + "step": 39827 + }, + { + "epoch": 1.46, + "learning_rate": 5.433039147471344e-08, + "loss": 0.6923, + "step": 39828 + }, + { + "epoch": 1.46, + "learning_rate": 5.432672941004138e-08, + "loss": 0.684, + "step": 39829 + }, + { + "epoch": 1.46, + "learning_rate": 5.432306734536932e-08, + "loss": 0.6975, + "step": 39830 + }, + { + "epoch": 1.46, + "learning_rate": 5.431940528069725e-08, + "loss": 0.6872, + "step": 39831 + }, + { + "epoch": 1.46, + "learning_rate": 5.431574321602519e-08, + "loss": 0.6875, + "step": 39832 + }, + { + "epoch": 1.46, + "learning_rate": 5.431208115135313e-08, + "loss": 0.6912, + "step": 39833 + }, + { + "epoch": 1.46, + "learning_rate": 5.430841908668107e-08, + "loss": 0.7011, + "step": 39834 + }, + { + "epoch": 1.46, + "learning_rate": 5.430475702200901e-08, + "loss": 0.6922, + "step": 39835 + }, + { + "epoch": 1.46, + "learning_rate": 5.430109495733695e-08, + "loss": 0.7023, + "step": 39836 + }, + { + "epoch": 1.46, + "learning_rate": 5.429743289266489e-08, + "loss": 0.6884, + "step": 39837 + }, + { + "epoch": 1.46, + "learning_rate": 5.429377082799283e-08, + "loss": 0.6913, + "step": 39838 + }, + { + "epoch": 1.46, + "learning_rate": 5.429010876332076e-08, + "loss": 0.7029, + "step": 39839 + }, + { + "epoch": 1.46, + "learning_rate": 5.4286446698648694e-08, + "loss": 0.6892, + "step": 39840 + }, + { + "epoch": 1.46, + "learning_rate": 5.4282784633976633e-08, + "loss": 0.7053, + "step": 39841 + }, + { + "epoch": 1.46, + "learning_rate": 5.427912256930457e-08, + "loss": 0.6934, + "step": 39842 + }, + { + "epoch": 1.46, + "learning_rate": 5.427546050463251e-08, + "loss": 0.7042, + "step": 39843 + }, + { + "epoch": 1.46, + "learning_rate": 5.4271798439960447e-08, + "loss": 0.6963, + "step": 39844 + }, + { + "epoch": 1.46, + "learning_rate": 5.4268136375288387e-08, + "loss": 0.6838, + "step": 39845 + }, + { + "epoch": 1.46, + "learning_rate": 5.4264474310616327e-08, + "loss": 0.7165, + "step": 39846 + }, + { + "epoch": 1.46, + "learning_rate": 5.4260812245944266e-08, + "loss": 0.7047, + "step": 39847 + }, + { + "epoch": 1.46, + "learning_rate": 5.4257150181272206e-08, + "loss": 0.7055, + "step": 39848 + }, + { + "epoch": 1.46, + "learning_rate": 5.425348811660013e-08, + "loss": 0.6981, + "step": 39849 + }, + { + "epoch": 1.46, + "learning_rate": 5.424982605192807e-08, + "loss": 0.6914, + "step": 39850 + }, + { + "epoch": 1.46, + "learning_rate": 5.424616398725601e-08, + "loss": 0.6903, + "step": 39851 + }, + { + "epoch": 1.46, + "learning_rate": 5.424250192258395e-08, + "loss": 0.6928, + "step": 39852 + }, + { + "epoch": 1.46, + "learning_rate": 5.4238839857911886e-08, + "loss": 0.6981, + "step": 39853 + }, + { + "epoch": 1.46, + "learning_rate": 5.4235177793239826e-08, + "loss": 0.6753, + "step": 39854 + }, + { + "epoch": 1.46, + "learning_rate": 5.4231515728567766e-08, + "loss": 0.7026, + "step": 39855 + }, + { + "epoch": 1.46, + "learning_rate": 5.4227853663895706e-08, + "loss": 0.7001, + "step": 39856 + }, + { + "epoch": 1.46, + "learning_rate": 5.422419159922363e-08, + "loss": 0.6959, + "step": 39857 + }, + { + "epoch": 1.46, + "learning_rate": 5.422052953455157e-08, + "loss": 0.6935, + "step": 39858 + }, + { + "epoch": 1.46, + "learning_rate": 5.421686746987951e-08, + "loss": 0.7019, + "step": 39859 + }, + { + "epoch": 1.46, + "learning_rate": 5.421320540520745e-08, + "loss": 0.6895, + "step": 39860 + }, + { + "epoch": 1.46, + "learning_rate": 5.420954334053539e-08, + "loss": 0.7058, + "step": 39861 + }, + { + "epoch": 1.46, + "learning_rate": 5.420588127586333e-08, + "loss": 0.7142, + "step": 39862 + }, + { + "epoch": 1.46, + "learning_rate": 5.420221921119127e-08, + "loss": 0.6967, + "step": 39863 + }, + { + "epoch": 1.46, + "learning_rate": 5.4198557146519206e-08, + "loss": 0.6988, + "step": 39864 + }, + { + "epoch": 1.46, + "learning_rate": 5.4194895081847145e-08, + "loss": 0.6939, + "step": 39865 + }, + { + "epoch": 1.46, + "learning_rate": 5.419123301717507e-08, + "loss": 0.6838, + "step": 39866 + }, + { + "epoch": 1.46, + "learning_rate": 5.418757095250301e-08, + "loss": 0.6917, + "step": 39867 + }, + { + "epoch": 1.46, + "learning_rate": 5.418390888783095e-08, + "loss": 0.7071, + "step": 39868 + }, + { + "epoch": 1.46, + "learning_rate": 5.418024682315889e-08, + "loss": 0.6939, + "step": 39869 + }, + { + "epoch": 1.46, + "learning_rate": 5.417658475848683e-08, + "loss": 0.6928, + "step": 39870 + }, + { + "epoch": 1.46, + "learning_rate": 5.417292269381477e-08, + "loss": 0.6793, + "step": 39871 + }, + { + "epoch": 1.46, + "learning_rate": 5.416926062914271e-08, + "loss": 0.6901, + "step": 39872 + }, + { + "epoch": 1.46, + "learning_rate": 5.416559856447065e-08, + "loss": 0.6866, + "step": 39873 + }, + { + "epoch": 1.46, + "learning_rate": 5.416193649979859e-08, + "loss": 0.6873, + "step": 39874 + }, + { + "epoch": 1.46, + "learning_rate": 5.415827443512652e-08, + "loss": 0.6927, + "step": 39875 + }, + { + "epoch": 1.46, + "learning_rate": 5.415461237045446e-08, + "loss": 0.6896, + "step": 39876 + }, + { + "epoch": 1.46, + "learning_rate": 5.415095030578239e-08, + "loss": 0.6772, + "step": 39877 + }, + { + "epoch": 1.46, + "learning_rate": 5.414728824111033e-08, + "loss": 0.6906, + "step": 39878 + }, + { + "epoch": 1.46, + "learning_rate": 5.414362617643827e-08, + "loss": 0.7085, + "step": 39879 + }, + { + "epoch": 1.46, + "learning_rate": 5.413996411176621e-08, + "loss": 0.6967, + "step": 39880 + }, + { + "epoch": 1.46, + "learning_rate": 5.413630204709415e-08, + "loss": 0.6976, + "step": 39881 + }, + { + "epoch": 1.46, + "learning_rate": 5.413263998242209e-08, + "loss": 0.6919, + "step": 39882 + }, + { + "epoch": 1.46, + "learning_rate": 5.412897791775003e-08, + "loss": 0.6896, + "step": 39883 + }, + { + "epoch": 1.46, + "learning_rate": 5.412531585307796e-08, + "loss": 0.7013, + "step": 39884 + }, + { + "epoch": 1.46, + "learning_rate": 5.41216537884059e-08, + "loss": 0.6822, + "step": 39885 + }, + { + "epoch": 1.46, + "learning_rate": 5.411799172373384e-08, + "loss": 0.6932, + "step": 39886 + }, + { + "epoch": 1.46, + "learning_rate": 5.411432965906178e-08, + "loss": 0.6861, + "step": 39887 + }, + { + "epoch": 1.46, + "learning_rate": 5.411066759438971e-08, + "loss": 0.6864, + "step": 39888 + }, + { + "epoch": 1.46, + "learning_rate": 5.410700552971765e-08, + "loss": 0.6959, + "step": 39889 + }, + { + "epoch": 1.46, + "learning_rate": 5.410334346504559e-08, + "loss": 0.7026, + "step": 39890 + }, + { + "epoch": 1.46, + "learning_rate": 5.409968140037353e-08, + "loss": 0.692, + "step": 39891 + }, + { + "epoch": 1.46, + "learning_rate": 5.409601933570147e-08, + "loss": 0.708, + "step": 39892 + }, + { + "epoch": 1.46, + "learning_rate": 5.40923572710294e-08, + "loss": 0.6922, + "step": 39893 + }, + { + "epoch": 1.46, + "learning_rate": 5.408869520635734e-08, + "loss": 0.6819, + "step": 39894 + }, + { + "epoch": 1.46, + "learning_rate": 5.408503314168528e-08, + "loss": 0.6896, + "step": 39895 + }, + { + "epoch": 1.46, + "learning_rate": 5.408137107701322e-08, + "loss": 0.7004, + "step": 39896 + }, + { + "epoch": 1.46, + "learning_rate": 5.407770901234116e-08, + "loss": 0.6845, + "step": 39897 + }, + { + "epoch": 1.46, + "learning_rate": 5.40740469476691e-08, + "loss": 0.6697, + "step": 39898 + }, + { + "epoch": 1.46, + "learning_rate": 5.407038488299703e-08, + "loss": 0.6711, + "step": 39899 + }, + { + "epoch": 1.46, + "learning_rate": 5.406672281832497e-08, + "loss": 0.696, + "step": 39900 + }, + { + "epoch": 1.46, + "learning_rate": 5.406306075365291e-08, + "loss": 0.6908, + "step": 39901 + }, + { + "epoch": 1.46, + "learning_rate": 5.405939868898084e-08, + "loss": 0.6991, + "step": 39902 + }, + { + "epoch": 1.46, + "learning_rate": 5.405573662430878e-08, + "loss": 0.7011, + "step": 39903 + }, + { + "epoch": 1.46, + "learning_rate": 5.4052074559636717e-08, + "loss": 0.698, + "step": 39904 + }, + { + "epoch": 1.46, + "learning_rate": 5.4048412494964657e-08, + "loss": 0.6875, + "step": 39905 + }, + { + "epoch": 1.46, + "learning_rate": 5.4044750430292597e-08, + "loss": 0.698, + "step": 39906 + }, + { + "epoch": 1.46, + "learning_rate": 5.4041088365620536e-08, + "loss": 0.6885, + "step": 39907 + }, + { + "epoch": 1.46, + "learning_rate": 5.4037426300948476e-08, + "loss": 0.6919, + "step": 39908 + }, + { + "epoch": 1.46, + "learning_rate": 5.4033764236276416e-08, + "loss": 0.6981, + "step": 39909 + }, + { + "epoch": 1.46, + "learning_rate": 5.403010217160435e-08, + "loss": 0.7156, + "step": 39910 + }, + { + "epoch": 1.46, + "learning_rate": 5.402644010693228e-08, + "loss": 0.6887, + "step": 39911 + }, + { + "epoch": 1.46, + "learning_rate": 5.402277804226022e-08, + "loss": 0.686, + "step": 39912 + }, + { + "epoch": 1.46, + "learning_rate": 5.4019115977588156e-08, + "loss": 0.6989, + "step": 39913 + }, + { + "epoch": 1.46, + "learning_rate": 5.4015453912916096e-08, + "loss": 0.6971, + "step": 39914 + }, + { + "epoch": 1.46, + "learning_rate": 5.4011791848244036e-08, + "loss": 0.6907, + "step": 39915 + }, + { + "epoch": 1.46, + "learning_rate": 5.4008129783571976e-08, + "loss": 0.6902, + "step": 39916 + }, + { + "epoch": 1.46, + "learning_rate": 5.4004467718899916e-08, + "loss": 0.6947, + "step": 39917 + }, + { + "epoch": 1.46, + "learning_rate": 5.4000805654227856e-08, + "loss": 0.6971, + "step": 39918 + }, + { + "epoch": 1.46, + "learning_rate": 5.3997143589555796e-08, + "loss": 0.694, + "step": 39919 + }, + { + "epoch": 1.46, + "learning_rate": 5.399348152488372e-08, + "loss": 0.6905, + "step": 39920 + }, + { + "epoch": 1.46, + "learning_rate": 5.398981946021166e-08, + "loss": 0.6901, + "step": 39921 + }, + { + "epoch": 1.46, + "learning_rate": 5.39861573955396e-08, + "loss": 0.6909, + "step": 39922 + }, + { + "epoch": 1.46, + "learning_rate": 5.398249533086754e-08, + "loss": 0.6962, + "step": 39923 + }, + { + "epoch": 1.46, + "learning_rate": 5.3978833266195476e-08, + "loss": 0.7018, + "step": 39924 + }, + { + "epoch": 1.46, + "learning_rate": 5.3975171201523415e-08, + "loss": 0.6921, + "step": 39925 + }, + { + "epoch": 1.46, + "learning_rate": 5.3971509136851355e-08, + "loss": 0.6688, + "step": 39926 + }, + { + "epoch": 1.46, + "learning_rate": 5.3967847072179295e-08, + "loss": 0.6921, + "step": 39927 + }, + { + "epoch": 1.46, + "learning_rate": 5.3964185007507235e-08, + "loss": 0.6964, + "step": 39928 + }, + { + "epoch": 1.46, + "learning_rate": 5.396052294283516e-08, + "loss": 0.6996, + "step": 39929 + }, + { + "epoch": 1.46, + "learning_rate": 5.39568608781631e-08, + "loss": 0.6871, + "step": 39930 + }, + { + "epoch": 1.46, + "learning_rate": 5.395319881349104e-08, + "loss": 0.685, + "step": 39931 + }, + { + "epoch": 1.46, + "learning_rate": 5.394953674881898e-08, + "loss": 0.6799, + "step": 39932 + }, + { + "epoch": 1.46, + "learning_rate": 5.394587468414692e-08, + "loss": 0.6943, + "step": 39933 + }, + { + "epoch": 1.46, + "learning_rate": 5.394221261947486e-08, + "loss": 0.6993, + "step": 39934 + }, + { + "epoch": 1.46, + "learning_rate": 5.3938550554802795e-08, + "loss": 0.6915, + "step": 39935 + }, + { + "epoch": 1.46, + "learning_rate": 5.3934888490130735e-08, + "loss": 0.6904, + "step": 39936 + }, + { + "epoch": 1.46, + "learning_rate": 5.3931226425458675e-08, + "loss": 0.6912, + "step": 39937 + }, + { + "epoch": 1.46, + "learning_rate": 5.39275643607866e-08, + "loss": 0.6827, + "step": 39938 + }, + { + "epoch": 1.46, + "learning_rate": 5.392390229611454e-08, + "loss": 0.6922, + "step": 39939 + }, + { + "epoch": 1.46, + "learning_rate": 5.392024023144248e-08, + "loss": 0.7094, + "step": 39940 + }, + { + "epoch": 1.46, + "learning_rate": 5.391657816677042e-08, + "loss": 0.678, + "step": 39941 + }, + { + "epoch": 1.46, + "learning_rate": 5.391291610209836e-08, + "loss": 0.6885, + "step": 39942 + }, + { + "epoch": 1.46, + "learning_rate": 5.39092540374263e-08, + "loss": 0.6782, + "step": 39943 + }, + { + "epoch": 1.46, + "learning_rate": 5.390559197275424e-08, + "loss": 0.6902, + "step": 39944 + }, + { + "epoch": 1.46, + "learning_rate": 5.390192990808218e-08, + "loss": 0.6927, + "step": 39945 + }, + { + "epoch": 1.46, + "learning_rate": 5.3898267843410114e-08, + "loss": 0.6915, + "step": 39946 + }, + { + "epoch": 1.46, + "learning_rate": 5.389460577873805e-08, + "loss": 0.6906, + "step": 39947 + }, + { + "epoch": 1.46, + "learning_rate": 5.389094371406598e-08, + "loss": 0.7045, + "step": 39948 + }, + { + "epoch": 1.46, + "learning_rate": 5.388728164939392e-08, + "loss": 0.7031, + "step": 39949 + }, + { + "epoch": 1.46, + "learning_rate": 5.388361958472186e-08, + "loss": 0.6924, + "step": 39950 + }, + { + "epoch": 1.46, + "learning_rate": 5.38799575200498e-08, + "loss": 0.693, + "step": 39951 + }, + { + "epoch": 1.46, + "learning_rate": 5.387629545537774e-08, + "loss": 0.6892, + "step": 39952 + }, + { + "epoch": 1.46, + "learning_rate": 5.387263339070568e-08, + "loss": 0.6968, + "step": 39953 + }, + { + "epoch": 1.46, + "learning_rate": 5.386897132603362e-08, + "loss": 0.6958, + "step": 39954 + }, + { + "epoch": 1.46, + "learning_rate": 5.386530926136156e-08, + "loss": 0.6902, + "step": 39955 + }, + { + "epoch": 1.46, + "learning_rate": 5.386164719668949e-08, + "loss": 0.6967, + "step": 39956 + }, + { + "epoch": 1.46, + "learning_rate": 5.385798513201743e-08, + "loss": 0.6842, + "step": 39957 + }, + { + "epoch": 1.46, + "learning_rate": 5.385432306734537e-08, + "loss": 0.695, + "step": 39958 + }, + { + "epoch": 1.46, + "learning_rate": 5.38506610026733e-08, + "loss": 0.6872, + "step": 39959 + }, + { + "epoch": 1.46, + "learning_rate": 5.384699893800124e-08, + "loss": 0.6912, + "step": 39960 + }, + { + "epoch": 1.46, + "learning_rate": 5.384333687332918e-08, + "loss": 0.7041, + "step": 39961 + }, + { + "epoch": 1.46, + "learning_rate": 5.383967480865712e-08, + "loss": 0.6849, + "step": 39962 + }, + { + "epoch": 1.46, + "learning_rate": 5.383601274398506e-08, + "loss": 0.7084, + "step": 39963 + }, + { + "epoch": 1.46, + "learning_rate": 5.3832350679313e-08, + "loss": 0.6967, + "step": 39964 + }, + { + "epoch": 1.46, + "learning_rate": 5.3828688614640927e-08, + "loss": 0.6991, + "step": 39965 + }, + { + "epoch": 1.46, + "learning_rate": 5.3825026549968866e-08, + "loss": 0.6913, + "step": 39966 + }, + { + "epoch": 1.46, + "learning_rate": 5.3821364485296806e-08, + "loss": 0.6843, + "step": 39967 + }, + { + "epoch": 1.46, + "learning_rate": 5.3817702420624746e-08, + "loss": 0.683, + "step": 39968 + }, + { + "epoch": 1.46, + "learning_rate": 5.3814040355952686e-08, + "loss": 0.6855, + "step": 39969 + }, + { + "epoch": 1.46, + "learning_rate": 5.381037829128062e-08, + "loss": 0.7052, + "step": 39970 + }, + { + "epoch": 1.46, + "learning_rate": 5.380671622660856e-08, + "loss": 0.679, + "step": 39971 + }, + { + "epoch": 1.46, + "learning_rate": 5.38030541619365e-08, + "loss": 0.6814, + "step": 39972 + }, + { + "epoch": 1.46, + "learning_rate": 5.379939209726444e-08, + "loss": 0.6846, + "step": 39973 + }, + { + "epoch": 1.46, + "learning_rate": 5.3795730032592366e-08, + "loss": 0.693, + "step": 39974 + }, + { + "epoch": 1.46, + "learning_rate": 5.3792067967920306e-08, + "loss": 0.6922, + "step": 39975 + }, + { + "epoch": 1.46, + "learning_rate": 5.3788405903248246e-08, + "loss": 0.6799, + "step": 39976 + }, + { + "epoch": 1.46, + "learning_rate": 5.3784743838576186e-08, + "loss": 0.674, + "step": 39977 + }, + { + "epoch": 1.46, + "learning_rate": 5.3781081773904126e-08, + "loss": 0.6844, + "step": 39978 + }, + { + "epoch": 1.46, + "learning_rate": 5.3777419709232066e-08, + "loss": 0.6903, + "step": 39979 + }, + { + "epoch": 1.46, + "learning_rate": 5.3773757644560006e-08, + "loss": 0.7079, + "step": 39980 + }, + { + "epoch": 1.46, + "learning_rate": 5.377009557988794e-08, + "loss": 0.6792, + "step": 39981 + }, + { + "epoch": 1.46, + "learning_rate": 5.376643351521588e-08, + "loss": 0.6897, + "step": 39982 + }, + { + "epoch": 1.46, + "learning_rate": 5.376277145054381e-08, + "loss": 0.6931, + "step": 39983 + }, + { + "epoch": 1.46, + "learning_rate": 5.3759109385871746e-08, + "loss": 0.6833, + "step": 39984 + }, + { + "epoch": 1.46, + "learning_rate": 5.3755447321199685e-08, + "loss": 0.701, + "step": 39985 + }, + { + "epoch": 1.46, + "learning_rate": 5.3751785256527625e-08, + "loss": 0.7017, + "step": 39986 + }, + { + "epoch": 1.46, + "learning_rate": 5.3748123191855565e-08, + "loss": 0.6845, + "step": 39987 + }, + { + "epoch": 1.46, + "learning_rate": 5.3744461127183505e-08, + "loss": 0.7041, + "step": 39988 + }, + { + "epoch": 1.46, + "learning_rate": 5.3740799062511445e-08, + "loss": 0.6898, + "step": 39989 + }, + { + "epoch": 1.46, + "learning_rate": 5.3737136997839385e-08, + "loss": 0.6902, + "step": 39990 + }, + { + "epoch": 1.46, + "learning_rate": 5.3733474933167325e-08, + "loss": 0.683, + "step": 39991 + }, + { + "epoch": 1.46, + "learning_rate": 5.372981286849525e-08, + "loss": 0.6896, + "step": 39992 + }, + { + "epoch": 1.46, + "learning_rate": 5.372615080382319e-08, + "loss": 0.6983, + "step": 39993 + }, + { + "epoch": 1.46, + "learning_rate": 5.372248873915113e-08, + "loss": 0.6978, + "step": 39994 + }, + { + "epoch": 1.46, + "learning_rate": 5.3718826674479065e-08, + "loss": 0.6865, + "step": 39995 + }, + { + "epoch": 1.46, + "learning_rate": 5.3715164609807005e-08, + "loss": 0.6858, + "step": 39996 + }, + { + "epoch": 1.46, + "learning_rate": 5.3711502545134945e-08, + "loss": 0.6836, + "step": 39997 + }, + { + "epoch": 1.46, + "learning_rate": 5.3707840480462885e-08, + "loss": 0.6991, + "step": 39998 + }, + { + "epoch": 1.46, + "learning_rate": 5.3704178415790825e-08, + "loss": 0.6805, + "step": 39999 + }, + { + "epoch": 1.46, + "learning_rate": 5.3700516351118765e-08, + "loss": 0.7048, + "step": 40000 + }, + { + "epoch": 1.46, + "learning_rate": 5.369685428644669e-08, + "loss": 0.7032, + "step": 40001 + }, + { + "epoch": 1.46, + "learning_rate": 5.369319222177463e-08, + "loss": 0.6864, + "step": 40002 + }, + { + "epoch": 1.46, + "learning_rate": 5.368953015710257e-08, + "loss": 0.6834, + "step": 40003 + }, + { + "epoch": 1.46, + "learning_rate": 5.368586809243051e-08, + "loss": 0.7027, + "step": 40004 + }, + { + "epoch": 1.46, + "learning_rate": 5.368220602775845e-08, + "loss": 0.6955, + "step": 40005 + }, + { + "epoch": 1.46, + "learning_rate": 5.3678543963086384e-08, + "loss": 0.6949, + "step": 40006 + }, + { + "epoch": 1.46, + "learning_rate": 5.3674881898414324e-08, + "loss": 0.7031, + "step": 40007 + }, + { + "epoch": 1.46, + "learning_rate": 5.3671219833742264e-08, + "loss": 0.69, + "step": 40008 + }, + { + "epoch": 1.46, + "learning_rate": 5.3667557769070204e-08, + "loss": 0.6985, + "step": 40009 + }, + { + "epoch": 1.46, + "learning_rate": 5.366389570439813e-08, + "loss": 0.6964, + "step": 40010 + }, + { + "epoch": 1.46, + "learning_rate": 5.366023363972607e-08, + "loss": 0.6918, + "step": 40011 + }, + { + "epoch": 1.46, + "learning_rate": 5.365657157505401e-08, + "loss": 0.6959, + "step": 40012 + }, + { + "epoch": 1.46, + "learning_rate": 5.365290951038195e-08, + "loss": 0.6915, + "step": 40013 + }, + { + "epoch": 1.46, + "learning_rate": 5.364924744570989e-08, + "loss": 0.6781, + "step": 40014 + }, + { + "epoch": 1.46, + "learning_rate": 5.364558538103783e-08, + "loss": 0.6948, + "step": 40015 + }, + { + "epoch": 1.46, + "learning_rate": 5.3641923316365764e-08, + "loss": 0.6788, + "step": 40016 + }, + { + "epoch": 1.46, + "learning_rate": 5.3638261251693704e-08, + "loss": 0.6945, + "step": 40017 + }, + { + "epoch": 1.46, + "learning_rate": 5.3634599187021644e-08, + "loss": 0.6802, + "step": 40018 + }, + { + "epoch": 1.46, + "learning_rate": 5.363093712234957e-08, + "loss": 0.6856, + "step": 40019 + }, + { + "epoch": 1.46, + "learning_rate": 5.362727505767751e-08, + "loss": 0.7025, + "step": 40020 + }, + { + "epoch": 1.46, + "learning_rate": 5.362361299300545e-08, + "loss": 0.6857, + "step": 40021 + }, + { + "epoch": 1.46, + "learning_rate": 5.361995092833339e-08, + "loss": 0.6934, + "step": 40022 + }, + { + "epoch": 1.46, + "learning_rate": 5.361628886366133e-08, + "loss": 0.6994, + "step": 40023 + }, + { + "epoch": 1.46, + "learning_rate": 5.361262679898927e-08, + "loss": 0.6906, + "step": 40024 + }, + { + "epoch": 1.46, + "learning_rate": 5.360896473431721e-08, + "loss": 0.6868, + "step": 40025 + }, + { + "epoch": 1.46, + "learning_rate": 5.360530266964515e-08, + "loss": 0.7028, + "step": 40026 + }, + { + "epoch": 1.46, + "learning_rate": 5.360164060497308e-08, + "loss": 0.6859, + "step": 40027 + }, + { + "epoch": 1.46, + "learning_rate": 5.3597978540301016e-08, + "loss": 0.7057, + "step": 40028 + }, + { + "epoch": 1.46, + "learning_rate": 5.3594316475628956e-08, + "loss": 0.691, + "step": 40029 + }, + { + "epoch": 1.46, + "learning_rate": 5.359065441095689e-08, + "loss": 0.6816, + "step": 40030 + }, + { + "epoch": 1.46, + "learning_rate": 5.358699234628483e-08, + "loss": 0.6912, + "step": 40031 + }, + { + "epoch": 1.46, + "learning_rate": 5.358333028161277e-08, + "loss": 0.6872, + "step": 40032 + }, + { + "epoch": 1.46, + "learning_rate": 5.357966821694071e-08, + "loss": 0.681, + "step": 40033 + }, + { + "epoch": 1.46, + "learning_rate": 5.357600615226865e-08, + "loss": 0.6904, + "step": 40034 + }, + { + "epoch": 1.46, + "learning_rate": 5.357234408759659e-08, + "loss": 0.6878, + "step": 40035 + }, + { + "epoch": 1.46, + "learning_rate": 5.356868202292453e-08, + "loss": 0.7023, + "step": 40036 + }, + { + "epoch": 1.46, + "learning_rate": 5.3565019958252456e-08, + "loss": 0.6909, + "step": 40037 + }, + { + "epoch": 1.46, + "learning_rate": 5.3561357893580396e-08, + "loss": 0.691, + "step": 40038 + }, + { + "epoch": 1.46, + "learning_rate": 5.3557695828908336e-08, + "loss": 0.6951, + "step": 40039 + }, + { + "epoch": 1.46, + "learning_rate": 5.3554033764236276e-08, + "loss": 0.7006, + "step": 40040 + }, + { + "epoch": 1.46, + "learning_rate": 5.355037169956421e-08, + "loss": 0.6716, + "step": 40041 + }, + { + "epoch": 1.47, + "learning_rate": 5.354670963489215e-08, + "loss": 0.7076, + "step": 40042 + }, + { + "epoch": 1.47, + "learning_rate": 5.354304757022009e-08, + "loss": 0.6884, + "step": 40043 + }, + { + "epoch": 1.47, + "learning_rate": 5.353938550554803e-08, + "loss": 0.7033, + "step": 40044 + }, + { + "epoch": 1.47, + "learning_rate": 5.353572344087597e-08, + "loss": 0.6841, + "step": 40045 + }, + { + "epoch": 1.47, + "learning_rate": 5.3532061376203895e-08, + "loss": 0.7035, + "step": 40046 + }, + { + "epoch": 1.47, + "learning_rate": 5.3528399311531835e-08, + "loss": 0.6922, + "step": 40047 + }, + { + "epoch": 1.47, + "learning_rate": 5.3524737246859775e-08, + "loss": 0.6893, + "step": 40048 + }, + { + "epoch": 1.47, + "learning_rate": 5.3521075182187715e-08, + "loss": 0.6943, + "step": 40049 + }, + { + "epoch": 1.47, + "learning_rate": 5.3517413117515655e-08, + "loss": 0.7048, + "step": 40050 + }, + { + "epoch": 1.47, + "learning_rate": 5.3513751052843595e-08, + "loss": 0.6854, + "step": 40051 + }, + { + "epoch": 1.47, + "learning_rate": 5.351008898817153e-08, + "loss": 0.713, + "step": 40052 + }, + { + "epoch": 1.47, + "learning_rate": 5.350642692349947e-08, + "loss": 0.6953, + "step": 40053 + }, + { + "epoch": 1.47, + "learning_rate": 5.350276485882741e-08, + "loss": 0.6838, + "step": 40054 + }, + { + "epoch": 1.47, + "learning_rate": 5.3499102794155335e-08, + "loss": 0.6929, + "step": 40055 + }, + { + "epoch": 1.47, + "learning_rate": 5.3495440729483275e-08, + "loss": 0.6967, + "step": 40056 + }, + { + "epoch": 1.47, + "learning_rate": 5.3491778664811215e-08, + "loss": 0.7008, + "step": 40057 + }, + { + "epoch": 1.47, + "learning_rate": 5.3488116600139155e-08, + "loss": 0.692, + "step": 40058 + }, + { + "epoch": 1.47, + "learning_rate": 5.3484454535467095e-08, + "loss": 0.6869, + "step": 40059 + }, + { + "epoch": 1.47, + "learning_rate": 5.3480792470795034e-08, + "loss": 0.7062, + "step": 40060 + }, + { + "epoch": 1.47, + "learning_rate": 5.3477130406122974e-08, + "loss": 0.7001, + "step": 40061 + }, + { + "epoch": 1.47, + "learning_rate": 5.3473468341450914e-08, + "loss": 0.7013, + "step": 40062 + }, + { + "epoch": 1.47, + "learning_rate": 5.346980627677885e-08, + "loss": 0.6943, + "step": 40063 + }, + { + "epoch": 1.47, + "learning_rate": 5.346614421210678e-08, + "loss": 0.703, + "step": 40064 + }, + { + "epoch": 1.47, + "learning_rate": 5.346248214743472e-08, + "loss": 0.7026, + "step": 40065 + }, + { + "epoch": 1.47, + "learning_rate": 5.3458820082762654e-08, + "loss": 0.6931, + "step": 40066 + }, + { + "epoch": 1.47, + "learning_rate": 5.3455158018090594e-08, + "loss": 0.7016, + "step": 40067 + }, + { + "epoch": 1.47, + "learning_rate": 5.3451495953418534e-08, + "loss": 0.698, + "step": 40068 + }, + { + "epoch": 1.47, + "learning_rate": 5.3447833888746474e-08, + "loss": 0.7034, + "step": 40069 + }, + { + "epoch": 1.47, + "learning_rate": 5.3444171824074414e-08, + "loss": 0.6876, + "step": 40070 + }, + { + "epoch": 1.47, + "learning_rate": 5.3440509759402354e-08, + "loss": 0.6868, + "step": 40071 + }, + { + "epoch": 1.47, + "learning_rate": 5.3436847694730294e-08, + "loss": 0.702, + "step": 40072 + }, + { + "epoch": 1.47, + "learning_rate": 5.343318563005822e-08, + "loss": 0.6894, + "step": 40073 + }, + { + "epoch": 1.47, + "learning_rate": 5.342952356538616e-08, + "loss": 0.6879, + "step": 40074 + }, + { + "epoch": 1.47, + "learning_rate": 5.34258615007141e-08, + "loss": 0.6892, + "step": 40075 + }, + { + "epoch": 1.47, + "learning_rate": 5.3422199436042034e-08, + "loss": 0.7027, + "step": 40076 + }, + { + "epoch": 1.47, + "learning_rate": 5.3418537371369974e-08, + "loss": 0.683, + "step": 40077 + }, + { + "epoch": 1.47, + "learning_rate": 5.3414875306697914e-08, + "loss": 0.6976, + "step": 40078 + }, + { + "epoch": 1.47, + "learning_rate": 5.3411213242025853e-08, + "loss": 0.7069, + "step": 40079 + }, + { + "epoch": 1.47, + "learning_rate": 5.3407551177353793e-08, + "loss": 0.6853, + "step": 40080 + }, + { + "epoch": 1.47, + "learning_rate": 5.340388911268173e-08, + "loss": 0.6899, + "step": 40081 + }, + { + "epoch": 1.47, + "learning_rate": 5.340022704800966e-08, + "loss": 0.6979, + "step": 40082 + }, + { + "epoch": 1.47, + "learning_rate": 5.33965649833376e-08, + "loss": 0.693, + "step": 40083 + }, + { + "epoch": 1.47, + "learning_rate": 5.339290291866554e-08, + "loss": 0.6937, + "step": 40084 + }, + { + "epoch": 1.47, + "learning_rate": 5.338924085399348e-08, + "loss": 0.7014, + "step": 40085 + }, + { + "epoch": 1.47, + "learning_rate": 5.338557878932142e-08, + "loss": 0.6906, + "step": 40086 + }, + { + "epoch": 1.47, + "learning_rate": 5.338191672464935e-08, + "loss": 0.6933, + "step": 40087 + }, + { + "epoch": 1.47, + "learning_rate": 5.337825465997729e-08, + "loss": 0.7002, + "step": 40088 + }, + { + "epoch": 1.47, + "learning_rate": 5.337459259530523e-08, + "loss": 0.6924, + "step": 40089 + }, + { + "epoch": 1.47, + "learning_rate": 5.337093053063317e-08, + "loss": 0.7116, + "step": 40090 + }, + { + "epoch": 1.47, + "learning_rate": 5.33672684659611e-08, + "loss": 0.6868, + "step": 40091 + }, + { + "epoch": 1.47, + "learning_rate": 5.336360640128904e-08, + "loss": 0.6909, + "step": 40092 + }, + { + "epoch": 1.47, + "learning_rate": 5.335994433661698e-08, + "loss": 0.6993, + "step": 40093 + }, + { + "epoch": 1.47, + "learning_rate": 5.335628227194492e-08, + "loss": 0.6952, + "step": 40094 + }, + { + "epoch": 1.47, + "learning_rate": 5.335262020727286e-08, + "loss": 0.6951, + "step": 40095 + }, + { + "epoch": 1.47, + "learning_rate": 5.33489581426008e-08, + "loss": 0.6985, + "step": 40096 + }, + { + "epoch": 1.47, + "learning_rate": 5.334529607792874e-08, + "loss": 0.7018, + "step": 40097 + }, + { + "epoch": 1.47, + "learning_rate": 5.334163401325667e-08, + "loss": 0.7117, + "step": 40098 + }, + { + "epoch": 1.47, + "learning_rate": 5.333797194858461e-08, + "loss": 0.6931, + "step": 40099 + }, + { + "epoch": 1.47, + "learning_rate": 5.3334309883912546e-08, + "loss": 0.6824, + "step": 40100 + }, + { + "epoch": 1.47, + "learning_rate": 5.333064781924048e-08, + "loss": 0.6945, + "step": 40101 + }, + { + "epoch": 1.47, + "learning_rate": 5.332698575456842e-08, + "loss": 0.6761, + "step": 40102 + }, + { + "epoch": 1.47, + "learning_rate": 5.332332368989636e-08, + "loss": 0.6812, + "step": 40103 + }, + { + "epoch": 1.47, + "learning_rate": 5.33196616252243e-08, + "loss": 0.6899, + "step": 40104 + }, + { + "epoch": 1.47, + "learning_rate": 5.331599956055224e-08, + "loss": 0.7042, + "step": 40105 + }, + { + "epoch": 1.47, + "learning_rate": 5.331233749588018e-08, + "loss": 0.6884, + "step": 40106 + }, + { + "epoch": 1.47, + "learning_rate": 5.330867543120812e-08, + "loss": 0.6973, + "step": 40107 + }, + { + "epoch": 1.47, + "learning_rate": 5.330501336653606e-08, + "loss": 0.6866, + "step": 40108 + }, + { + "epoch": 1.47, + "learning_rate": 5.3301351301863985e-08, + "loss": 0.6933, + "step": 40109 + }, + { + "epoch": 1.47, + "learning_rate": 5.3297689237191925e-08, + "loss": 0.6845, + "step": 40110 + }, + { + "epoch": 1.47, + "learning_rate": 5.3294027172519865e-08, + "loss": 0.6896, + "step": 40111 + }, + { + "epoch": 1.47, + "learning_rate": 5.32903651078478e-08, + "loss": 0.7015, + "step": 40112 + }, + { + "epoch": 1.47, + "learning_rate": 5.328670304317574e-08, + "loss": 0.7152, + "step": 40113 + }, + { + "epoch": 1.47, + "learning_rate": 5.328304097850368e-08, + "loss": 0.7007, + "step": 40114 + }, + { + "epoch": 1.47, + "learning_rate": 5.327937891383162e-08, + "loss": 0.6889, + "step": 40115 + }, + { + "epoch": 1.47, + "learning_rate": 5.327571684915956e-08, + "loss": 0.6827, + "step": 40116 + }, + { + "epoch": 1.47, + "learning_rate": 5.32720547844875e-08, + "loss": 0.6966, + "step": 40117 + }, + { + "epoch": 1.47, + "learning_rate": 5.3268392719815425e-08, + "loss": 0.7007, + "step": 40118 + }, + { + "epoch": 1.47, + "learning_rate": 5.3264730655143365e-08, + "loss": 0.7118, + "step": 40119 + }, + { + "epoch": 1.47, + "learning_rate": 5.3261068590471304e-08, + "loss": 0.6907, + "step": 40120 + }, + { + "epoch": 1.47, + "learning_rate": 5.3257406525799244e-08, + "loss": 0.6773, + "step": 40121 + }, + { + "epoch": 1.47, + "learning_rate": 5.3253744461127184e-08, + "loss": 0.6935, + "step": 40122 + }, + { + "epoch": 1.47, + "learning_rate": 5.325008239645512e-08, + "loss": 0.6873, + "step": 40123 + }, + { + "epoch": 1.47, + "learning_rate": 5.324642033178306e-08, + "loss": 0.675, + "step": 40124 + }, + { + "epoch": 1.47, + "learning_rate": 5.3242758267111e-08, + "loss": 0.6901, + "step": 40125 + }, + { + "epoch": 1.47, + "learning_rate": 5.323909620243894e-08, + "loss": 0.6961, + "step": 40126 + }, + { + "epoch": 1.47, + "learning_rate": 5.3235434137766864e-08, + "loss": 0.6973, + "step": 40127 + }, + { + "epoch": 1.47, + "learning_rate": 5.3231772073094804e-08, + "loss": 0.6818, + "step": 40128 + }, + { + "epoch": 1.47, + "learning_rate": 5.3228110008422744e-08, + "loss": 0.7048, + "step": 40129 + }, + { + "epoch": 1.47, + "learning_rate": 5.3224447943750684e-08, + "loss": 0.6969, + "step": 40130 + }, + { + "epoch": 1.47, + "learning_rate": 5.3220785879078624e-08, + "loss": 0.7184, + "step": 40131 + }, + { + "epoch": 1.47, + "learning_rate": 5.3217123814406564e-08, + "loss": 0.7213, + "step": 40132 + }, + { + "epoch": 1.47, + "learning_rate": 5.3213461749734504e-08, + "loss": 0.6883, + "step": 40133 + }, + { + "epoch": 1.47, + "learning_rate": 5.320979968506244e-08, + "loss": 0.6738, + "step": 40134 + }, + { + "epoch": 1.47, + "learning_rate": 5.320613762039038e-08, + "loss": 0.685, + "step": 40135 + }, + { + "epoch": 1.47, + "learning_rate": 5.3202475555718304e-08, + "loss": 0.6868, + "step": 40136 + }, + { + "epoch": 1.47, + "learning_rate": 5.3198813491046244e-08, + "loss": 0.6892, + "step": 40137 + }, + { + "epoch": 1.47, + "learning_rate": 5.3195151426374183e-08, + "loss": 0.7006, + "step": 40138 + }, + { + "epoch": 1.47, + "learning_rate": 5.3191489361702123e-08, + "loss": 0.6864, + "step": 40139 + }, + { + "epoch": 1.47, + "learning_rate": 5.3187827297030063e-08, + "loss": 0.7054, + "step": 40140 + }, + { + "epoch": 1.47, + "learning_rate": 5.3184165232358e-08, + "loss": 0.7013, + "step": 40141 + }, + { + "epoch": 1.47, + "learning_rate": 5.318050316768594e-08, + "loss": 0.6848, + "step": 40142 + }, + { + "epoch": 1.47, + "learning_rate": 5.317684110301388e-08, + "loss": 0.6883, + "step": 40143 + }, + { + "epoch": 1.47, + "learning_rate": 5.317317903834182e-08, + "loss": 0.6982, + "step": 40144 + }, + { + "epoch": 1.47, + "learning_rate": 5.316951697366975e-08, + "loss": 0.7099, + "step": 40145 + }, + { + "epoch": 1.47, + "learning_rate": 5.316585490899769e-08, + "loss": 0.7061, + "step": 40146 + }, + { + "epoch": 1.47, + "learning_rate": 5.316219284432562e-08, + "loss": 0.7109, + "step": 40147 + }, + { + "epoch": 1.47, + "learning_rate": 5.315853077965356e-08, + "loss": 0.6961, + "step": 40148 + }, + { + "epoch": 1.47, + "learning_rate": 5.31548687149815e-08, + "loss": 0.7099, + "step": 40149 + }, + { + "epoch": 1.47, + "learning_rate": 5.315120665030944e-08, + "loss": 0.6796, + "step": 40150 + }, + { + "epoch": 1.47, + "learning_rate": 5.314754458563738e-08, + "loss": 0.695, + "step": 40151 + }, + { + "epoch": 1.47, + "learning_rate": 5.314388252096532e-08, + "loss": 0.6902, + "step": 40152 + }, + { + "epoch": 1.47, + "learning_rate": 5.314022045629326e-08, + "loss": 0.6889, + "step": 40153 + }, + { + "epoch": 1.47, + "learning_rate": 5.313655839162119e-08, + "loss": 0.6748, + "step": 40154 + }, + { + "epoch": 1.47, + "learning_rate": 5.313289632694913e-08, + "loss": 0.6906, + "step": 40155 + }, + { + "epoch": 1.47, + "learning_rate": 5.312923426227707e-08, + "loss": 0.7055, + "step": 40156 + }, + { + "epoch": 1.47, + "learning_rate": 5.312557219760501e-08, + "loss": 0.6988, + "step": 40157 + }, + { + "epoch": 1.47, + "learning_rate": 5.312191013293294e-08, + "loss": 0.7068, + "step": 40158 + }, + { + "epoch": 1.47, + "learning_rate": 5.311824806826088e-08, + "loss": 0.6895, + "step": 40159 + }, + { + "epoch": 1.47, + "learning_rate": 5.311458600358882e-08, + "loss": 0.694, + "step": 40160 + }, + { + "epoch": 1.47, + "learning_rate": 5.311092393891676e-08, + "loss": 0.7041, + "step": 40161 + }, + { + "epoch": 1.47, + "learning_rate": 5.31072618742447e-08, + "loss": 0.6922, + "step": 40162 + }, + { + "epoch": 1.47, + "learning_rate": 5.310359980957263e-08, + "loss": 0.6964, + "step": 40163 + }, + { + "epoch": 1.47, + "learning_rate": 5.309993774490057e-08, + "loss": 0.6976, + "step": 40164 + }, + { + "epoch": 1.47, + "learning_rate": 5.309627568022851e-08, + "loss": 0.6896, + "step": 40165 + }, + { + "epoch": 1.47, + "learning_rate": 5.309261361555645e-08, + "loss": 0.689, + "step": 40166 + }, + { + "epoch": 1.47, + "learning_rate": 5.308895155088439e-08, + "loss": 0.7069, + "step": 40167 + }, + { + "epoch": 1.47, + "learning_rate": 5.308528948621233e-08, + "loss": 0.6993, + "step": 40168 + }, + { + "epoch": 1.47, + "learning_rate": 5.308162742154026e-08, + "loss": 0.701, + "step": 40169 + }, + { + "epoch": 1.47, + "learning_rate": 5.30779653568682e-08, + "loss": 0.6707, + "step": 40170 + }, + { + "epoch": 1.47, + "learning_rate": 5.307430329219614e-08, + "loss": 0.6918, + "step": 40171 + }, + { + "epoch": 1.47, + "learning_rate": 5.307064122752407e-08, + "loss": 0.6918, + "step": 40172 + }, + { + "epoch": 1.47, + "learning_rate": 5.306697916285201e-08, + "loss": 0.6943, + "step": 40173 + }, + { + "epoch": 1.47, + "learning_rate": 5.306331709817995e-08, + "loss": 0.6842, + "step": 40174 + }, + { + "epoch": 1.47, + "learning_rate": 5.305965503350789e-08, + "loss": 0.7025, + "step": 40175 + }, + { + "epoch": 1.47, + "learning_rate": 5.305599296883583e-08, + "loss": 0.6988, + "step": 40176 + }, + { + "epoch": 1.47, + "learning_rate": 5.305233090416377e-08, + "loss": 0.6986, + "step": 40177 + }, + { + "epoch": 1.47, + "learning_rate": 5.304866883949171e-08, + "loss": 0.7048, + "step": 40178 + }, + { + "epoch": 1.47, + "learning_rate": 5.304500677481965e-08, + "loss": 0.6916, + "step": 40179 + }, + { + "epoch": 1.47, + "learning_rate": 5.304134471014758e-08, + "loss": 0.7004, + "step": 40180 + }, + { + "epoch": 1.47, + "learning_rate": 5.3037682645475514e-08, + "loss": 0.6878, + "step": 40181 + }, + { + "epoch": 1.47, + "learning_rate": 5.3034020580803454e-08, + "loss": 0.6884, + "step": 40182 + }, + { + "epoch": 1.47, + "learning_rate": 5.303035851613139e-08, + "loss": 0.7058, + "step": 40183 + }, + { + "epoch": 1.47, + "learning_rate": 5.302669645145933e-08, + "loss": 0.6956, + "step": 40184 + }, + { + "epoch": 1.47, + "learning_rate": 5.302303438678727e-08, + "loss": 0.6971, + "step": 40185 + }, + { + "epoch": 1.47, + "learning_rate": 5.301937232211521e-08, + "loss": 0.7084, + "step": 40186 + }, + { + "epoch": 1.47, + "learning_rate": 5.301571025744315e-08, + "loss": 0.6901, + "step": 40187 + }, + { + "epoch": 1.47, + "learning_rate": 5.301204819277109e-08, + "loss": 0.7091, + "step": 40188 + }, + { + "epoch": 1.47, + "learning_rate": 5.300838612809903e-08, + "loss": 0.6902, + "step": 40189 + }, + { + "epoch": 1.47, + "learning_rate": 5.3004724063426954e-08, + "loss": 0.7055, + "step": 40190 + }, + { + "epoch": 1.47, + "learning_rate": 5.3001061998754894e-08, + "loss": 0.6811, + "step": 40191 + }, + { + "epoch": 1.47, + "learning_rate": 5.2997399934082834e-08, + "loss": 0.7021, + "step": 40192 + }, + { + "epoch": 1.47, + "learning_rate": 5.2993737869410774e-08, + "loss": 0.6867, + "step": 40193 + }, + { + "epoch": 1.47, + "learning_rate": 5.299007580473871e-08, + "loss": 0.6937, + "step": 40194 + }, + { + "epoch": 1.47, + "learning_rate": 5.298641374006665e-08, + "loss": 0.7008, + "step": 40195 + }, + { + "epoch": 1.47, + "learning_rate": 5.298275167539459e-08, + "loss": 0.706, + "step": 40196 + }, + { + "epoch": 1.47, + "learning_rate": 5.297908961072253e-08, + "loss": 0.6946, + "step": 40197 + }, + { + "epoch": 1.47, + "learning_rate": 5.297542754605047e-08, + "loss": 0.6857, + "step": 40198 + }, + { + "epoch": 1.47, + "learning_rate": 5.2971765481378393e-08, + "loss": 0.6882, + "step": 40199 + }, + { + "epoch": 1.47, + "learning_rate": 5.2968103416706333e-08, + "loss": 0.6827, + "step": 40200 + }, + { + "epoch": 1.47, + "learning_rate": 5.296444135203427e-08, + "loss": 0.7022, + "step": 40201 + }, + { + "epoch": 1.47, + "learning_rate": 5.296077928736221e-08, + "loss": 0.6982, + "step": 40202 + }, + { + "epoch": 1.47, + "learning_rate": 5.295711722269015e-08, + "loss": 0.686, + "step": 40203 + }, + { + "epoch": 1.47, + "learning_rate": 5.295345515801809e-08, + "loss": 0.6994, + "step": 40204 + }, + { + "epoch": 1.47, + "learning_rate": 5.2949793093346026e-08, + "loss": 0.6966, + "step": 40205 + }, + { + "epoch": 1.47, + "learning_rate": 5.2946131028673966e-08, + "loss": 0.6974, + "step": 40206 + }, + { + "epoch": 1.47, + "learning_rate": 5.2942468964001906e-08, + "loss": 0.6823, + "step": 40207 + }, + { + "epoch": 1.47, + "learning_rate": 5.293880689932983e-08, + "loss": 0.6936, + "step": 40208 + }, + { + "epoch": 1.47, + "learning_rate": 5.293514483465777e-08, + "loss": 0.6871, + "step": 40209 + }, + { + "epoch": 1.47, + "learning_rate": 5.293148276998571e-08, + "loss": 0.6902, + "step": 40210 + }, + { + "epoch": 1.47, + "learning_rate": 5.292782070531365e-08, + "loss": 0.6872, + "step": 40211 + }, + { + "epoch": 1.47, + "learning_rate": 5.292415864064159e-08, + "loss": 0.6922, + "step": 40212 + }, + { + "epoch": 1.47, + "learning_rate": 5.292049657596953e-08, + "loss": 0.6989, + "step": 40213 + }, + { + "epoch": 1.47, + "learning_rate": 5.291683451129747e-08, + "loss": 0.7, + "step": 40214 + }, + { + "epoch": 1.47, + "learning_rate": 5.291317244662541e-08, + "loss": 0.6934, + "step": 40215 + }, + { + "epoch": 1.47, + "learning_rate": 5.2909510381953346e-08, + "loss": 0.6883, + "step": 40216 + }, + { + "epoch": 1.47, + "learning_rate": 5.290584831728128e-08, + "loss": 0.6891, + "step": 40217 + }, + { + "epoch": 1.47, + "learning_rate": 5.290218625260921e-08, + "loss": 0.7113, + "step": 40218 + }, + { + "epoch": 1.47, + "learning_rate": 5.289852418793715e-08, + "loss": 0.6873, + "step": 40219 + }, + { + "epoch": 1.47, + "learning_rate": 5.289486212326509e-08, + "loss": 0.7, + "step": 40220 + }, + { + "epoch": 1.47, + "learning_rate": 5.289120005859303e-08, + "loss": 0.6934, + "step": 40221 + }, + { + "epoch": 1.47, + "learning_rate": 5.288753799392097e-08, + "loss": 0.6878, + "step": 40222 + }, + { + "epoch": 1.47, + "learning_rate": 5.288387592924891e-08, + "loss": 0.6884, + "step": 40223 + }, + { + "epoch": 1.47, + "learning_rate": 5.288021386457685e-08, + "loss": 0.6886, + "step": 40224 + }, + { + "epoch": 1.47, + "learning_rate": 5.287655179990479e-08, + "loss": 0.6795, + "step": 40225 + }, + { + "epoch": 1.47, + "learning_rate": 5.287288973523272e-08, + "loss": 0.6962, + "step": 40226 + }, + { + "epoch": 1.47, + "learning_rate": 5.286922767056066e-08, + "loss": 0.7048, + "step": 40227 + }, + { + "epoch": 1.47, + "learning_rate": 5.28655656058886e-08, + "loss": 0.6811, + "step": 40228 + }, + { + "epoch": 1.47, + "learning_rate": 5.286190354121653e-08, + "loss": 0.6917, + "step": 40229 + }, + { + "epoch": 1.47, + "learning_rate": 5.285824147654447e-08, + "loss": 0.7053, + "step": 40230 + }, + { + "epoch": 1.47, + "learning_rate": 5.285457941187241e-08, + "loss": 0.6953, + "step": 40231 + }, + { + "epoch": 1.47, + "learning_rate": 5.285091734720035e-08, + "loss": 0.6969, + "step": 40232 + }, + { + "epoch": 1.47, + "learning_rate": 5.284725528252829e-08, + "loss": 0.6759, + "step": 40233 + }, + { + "epoch": 1.47, + "learning_rate": 5.284359321785623e-08, + "loss": 0.6664, + "step": 40234 + }, + { + "epoch": 1.47, + "learning_rate": 5.283993115318416e-08, + "loss": 0.6924, + "step": 40235 + }, + { + "epoch": 1.47, + "learning_rate": 5.28362690885121e-08, + "loss": 0.706, + "step": 40236 + }, + { + "epoch": 1.47, + "learning_rate": 5.283260702384004e-08, + "loss": 0.694, + "step": 40237 + }, + { + "epoch": 1.47, + "learning_rate": 5.282894495916798e-08, + "loss": 0.6909, + "step": 40238 + }, + { + "epoch": 1.47, + "learning_rate": 5.282528289449592e-08, + "loss": 0.7046, + "step": 40239 + }, + { + "epoch": 1.47, + "learning_rate": 5.282162082982385e-08, + "loss": 0.6931, + "step": 40240 + }, + { + "epoch": 1.47, + "learning_rate": 5.281795876515179e-08, + "loss": 0.6717, + "step": 40241 + }, + { + "epoch": 1.47, + "learning_rate": 5.281429670047973e-08, + "loss": 0.6916, + "step": 40242 + }, + { + "epoch": 1.47, + "learning_rate": 5.281063463580766e-08, + "loss": 0.7042, + "step": 40243 + }, + { + "epoch": 1.47, + "learning_rate": 5.28069725711356e-08, + "loss": 0.6867, + "step": 40244 + }, + { + "epoch": 1.47, + "learning_rate": 5.280331050646354e-08, + "loss": 0.6947, + "step": 40245 + }, + { + "epoch": 1.47, + "learning_rate": 5.279964844179148e-08, + "loss": 0.6994, + "step": 40246 + }, + { + "epoch": 1.47, + "learning_rate": 5.279598637711942e-08, + "loss": 0.6824, + "step": 40247 + }, + { + "epoch": 1.47, + "learning_rate": 5.279232431244736e-08, + "loss": 0.6935, + "step": 40248 + }, + { + "epoch": 1.47, + "learning_rate": 5.27886622477753e-08, + "loss": 0.692, + "step": 40249 + }, + { + "epoch": 1.47, + "learning_rate": 5.278500018310324e-08, + "loss": 0.6881, + "step": 40250 + }, + { + "epoch": 1.47, + "learning_rate": 5.278133811843117e-08, + "loss": 0.7052, + "step": 40251 + }, + { + "epoch": 1.47, + "learning_rate": 5.2777676053759104e-08, + "loss": 0.693, + "step": 40252 + }, + { + "epoch": 1.47, + "learning_rate": 5.2774013989087044e-08, + "loss": 0.7023, + "step": 40253 + }, + { + "epoch": 1.47, + "learning_rate": 5.277035192441498e-08, + "loss": 0.6946, + "step": 40254 + }, + { + "epoch": 1.47, + "learning_rate": 5.276668985974292e-08, + "loss": 0.6914, + "step": 40255 + }, + { + "epoch": 1.47, + "learning_rate": 5.276302779507086e-08, + "loss": 0.6881, + "step": 40256 + }, + { + "epoch": 1.47, + "learning_rate": 5.27593657303988e-08, + "loss": 0.6815, + "step": 40257 + }, + { + "epoch": 1.47, + "learning_rate": 5.275570366572674e-08, + "loss": 0.6985, + "step": 40258 + }, + { + "epoch": 1.47, + "learning_rate": 5.2752041601054677e-08, + "loss": 0.691, + "step": 40259 + }, + { + "epoch": 1.47, + "learning_rate": 5.2748379536382617e-08, + "loss": 0.6921, + "step": 40260 + }, + { + "epoch": 1.47, + "learning_rate": 5.274471747171054e-08, + "loss": 0.6917, + "step": 40261 + }, + { + "epoch": 1.47, + "learning_rate": 5.274105540703848e-08, + "loss": 0.6895, + "step": 40262 + }, + { + "epoch": 1.47, + "learning_rate": 5.273739334236642e-08, + "loss": 0.6837, + "step": 40263 + }, + { + "epoch": 1.47, + "learning_rate": 5.273373127769436e-08, + "loss": 0.6931, + "step": 40264 + }, + { + "epoch": 1.47, + "learning_rate": 5.2730069213022296e-08, + "loss": 0.6916, + "step": 40265 + }, + { + "epoch": 1.47, + "learning_rate": 5.2726407148350236e-08, + "loss": 0.6841, + "step": 40266 + }, + { + "epoch": 1.47, + "learning_rate": 5.2722745083678176e-08, + "loss": 0.6983, + "step": 40267 + }, + { + "epoch": 1.47, + "learning_rate": 5.2719083019006116e-08, + "loss": 0.6929, + "step": 40268 + }, + { + "epoch": 1.47, + "learning_rate": 5.2715420954334056e-08, + "loss": 0.6936, + "step": 40269 + }, + { + "epoch": 1.47, + "learning_rate": 5.271175888966198e-08, + "loss": 0.6982, + "step": 40270 + }, + { + "epoch": 1.47, + "learning_rate": 5.270809682498992e-08, + "loss": 0.6851, + "step": 40271 + }, + { + "epoch": 1.47, + "learning_rate": 5.270443476031786e-08, + "loss": 0.6995, + "step": 40272 + }, + { + "epoch": 1.47, + "learning_rate": 5.27007726956458e-08, + "loss": 0.686, + "step": 40273 + }, + { + "epoch": 1.47, + "learning_rate": 5.269711063097374e-08, + "loss": 0.7057, + "step": 40274 + }, + { + "epoch": 1.47, + "learning_rate": 5.269344856630168e-08, + "loss": 0.6847, + "step": 40275 + }, + { + "epoch": 1.47, + "learning_rate": 5.2689786501629616e-08, + "loss": 0.6929, + "step": 40276 + }, + { + "epoch": 1.47, + "learning_rate": 5.2686124436957556e-08, + "loss": 0.6958, + "step": 40277 + }, + { + "epoch": 1.47, + "learning_rate": 5.2682462372285496e-08, + "loss": 0.6964, + "step": 40278 + }, + { + "epoch": 1.47, + "learning_rate": 5.267880030761342e-08, + "loss": 0.6924, + "step": 40279 + }, + { + "epoch": 1.47, + "learning_rate": 5.267513824294136e-08, + "loss": 0.7031, + "step": 40280 + }, + { + "epoch": 1.47, + "learning_rate": 5.26714761782693e-08, + "loss": 0.6814, + "step": 40281 + }, + { + "epoch": 1.47, + "learning_rate": 5.266781411359724e-08, + "loss": 0.6938, + "step": 40282 + }, + { + "epoch": 1.47, + "learning_rate": 5.266415204892518e-08, + "loss": 0.6977, + "step": 40283 + }, + { + "epoch": 1.47, + "learning_rate": 5.266048998425312e-08, + "loss": 0.6983, + "step": 40284 + }, + { + "epoch": 1.47, + "learning_rate": 5.265682791958106e-08, + "loss": 0.6875, + "step": 40285 + }, + { + "epoch": 1.47, + "learning_rate": 5.2653165854909e-08, + "loss": 0.6985, + "step": 40286 + }, + { + "epoch": 1.47, + "learning_rate": 5.2649503790236935e-08, + "loss": 0.6935, + "step": 40287 + }, + { + "epoch": 1.47, + "learning_rate": 5.264584172556487e-08, + "loss": 0.6998, + "step": 40288 + }, + { + "epoch": 1.47, + "learning_rate": 5.26421796608928e-08, + "loss": 0.7015, + "step": 40289 + }, + { + "epoch": 1.47, + "learning_rate": 5.263851759622074e-08, + "loss": 0.6922, + "step": 40290 + }, + { + "epoch": 1.47, + "learning_rate": 5.263485553154868e-08, + "loss": 0.7013, + "step": 40291 + }, + { + "epoch": 1.47, + "learning_rate": 5.263119346687662e-08, + "loss": 0.6915, + "step": 40292 + }, + { + "epoch": 1.47, + "learning_rate": 5.262753140220456e-08, + "loss": 0.6956, + "step": 40293 + }, + { + "epoch": 1.47, + "learning_rate": 5.26238693375325e-08, + "loss": 0.7032, + "step": 40294 + }, + { + "epoch": 1.47, + "learning_rate": 5.262020727286044e-08, + "loss": 0.6987, + "step": 40295 + }, + { + "epoch": 1.47, + "learning_rate": 5.261654520818838e-08, + "loss": 0.6975, + "step": 40296 + }, + { + "epoch": 1.47, + "learning_rate": 5.261288314351631e-08, + "loss": 0.6934, + "step": 40297 + }, + { + "epoch": 1.47, + "learning_rate": 5.260922107884425e-08, + "loss": 0.6851, + "step": 40298 + }, + { + "epoch": 1.47, + "learning_rate": 5.260555901417219e-08, + "loss": 0.6996, + "step": 40299 + }, + { + "epoch": 1.47, + "learning_rate": 5.260189694950012e-08, + "loss": 0.7006, + "step": 40300 + }, + { + "epoch": 1.47, + "learning_rate": 5.259823488482806e-08, + "loss": 0.7111, + "step": 40301 + }, + { + "epoch": 1.47, + "learning_rate": 5.2594572820156e-08, + "loss": 0.6993, + "step": 40302 + }, + { + "epoch": 1.47, + "learning_rate": 5.259091075548394e-08, + "loss": 0.7078, + "step": 40303 + }, + { + "epoch": 1.47, + "learning_rate": 5.258724869081188e-08, + "loss": 0.6961, + "step": 40304 + }, + { + "epoch": 1.47, + "learning_rate": 5.258358662613982e-08, + "loss": 0.6913, + "step": 40305 + }, + { + "epoch": 1.47, + "learning_rate": 5.257992456146775e-08, + "loss": 0.6888, + "step": 40306 + }, + { + "epoch": 1.47, + "learning_rate": 5.257626249679569e-08, + "loss": 0.6907, + "step": 40307 + }, + { + "epoch": 1.47, + "learning_rate": 5.257260043212363e-08, + "loss": 0.7077, + "step": 40308 + }, + { + "epoch": 1.47, + "learning_rate": 5.256893836745157e-08, + "loss": 0.6839, + "step": 40309 + }, + { + "epoch": 1.47, + "learning_rate": 5.256527630277951e-08, + "loss": 0.6877, + "step": 40310 + }, + { + "epoch": 1.47, + "learning_rate": 5.256161423810744e-08, + "loss": 0.7082, + "step": 40311 + }, + { + "epoch": 1.47, + "learning_rate": 5.255795217343538e-08, + "loss": 0.693, + "step": 40312 + }, + { + "epoch": 1.47, + "learning_rate": 5.255429010876332e-08, + "loss": 0.6947, + "step": 40313 + }, + { + "epoch": 1.47, + "learning_rate": 5.255062804409126e-08, + "loss": 0.6806, + "step": 40314 + }, + { + "epoch": 1.48, + "learning_rate": 5.254696597941919e-08, + "loss": 0.6796, + "step": 40315 + }, + { + "epoch": 1.48, + "learning_rate": 5.254330391474713e-08, + "loss": 0.6976, + "step": 40316 + }, + { + "epoch": 1.48, + "learning_rate": 5.253964185007507e-08, + "loss": 0.6923, + "step": 40317 + }, + { + "epoch": 1.48, + "learning_rate": 5.2535979785403007e-08, + "loss": 0.6946, + "step": 40318 + }, + { + "epoch": 1.48, + "learning_rate": 5.2532317720730947e-08, + "loss": 0.694, + "step": 40319 + }, + { + "epoch": 1.48, + "learning_rate": 5.2528655656058887e-08, + "loss": 0.6988, + "step": 40320 + }, + { + "epoch": 1.48, + "learning_rate": 5.2524993591386826e-08, + "loss": 0.6862, + "step": 40321 + }, + { + "epoch": 1.48, + "learning_rate": 5.252133152671476e-08, + "loss": 0.7082, + "step": 40322 + }, + { + "epoch": 1.48, + "learning_rate": 5.25176694620427e-08, + "loss": 0.6837, + "step": 40323 + }, + { + "epoch": 1.48, + "learning_rate": 5.251400739737063e-08, + "loss": 0.6894, + "step": 40324 + }, + { + "epoch": 1.48, + "learning_rate": 5.2510345332698566e-08, + "loss": 0.6844, + "step": 40325 + }, + { + "epoch": 1.48, + "learning_rate": 5.2506683268026506e-08, + "loss": 0.7078, + "step": 40326 + }, + { + "epoch": 1.48, + "learning_rate": 5.2503021203354446e-08, + "loss": 0.684, + "step": 40327 + }, + { + "epoch": 1.48, + "learning_rate": 5.2499359138682386e-08, + "loss": 0.6952, + "step": 40328 + }, + { + "epoch": 1.48, + "learning_rate": 5.2495697074010326e-08, + "loss": 0.7096, + "step": 40329 + }, + { + "epoch": 1.48, + "learning_rate": 5.2492035009338266e-08, + "loss": 0.7177, + "step": 40330 + }, + { + "epoch": 1.48, + "learning_rate": 5.2488372944666206e-08, + "loss": 0.6962, + "step": 40331 + }, + { + "epoch": 1.48, + "learning_rate": 5.2484710879994146e-08, + "loss": 0.6883, + "step": 40332 + }, + { + "epoch": 1.48, + "learning_rate": 5.248104881532207e-08, + "loss": 0.6918, + "step": 40333 + }, + { + "epoch": 1.48, + "learning_rate": 5.247738675065001e-08, + "loss": 0.7032, + "step": 40334 + }, + { + "epoch": 1.48, + "learning_rate": 5.247372468597795e-08, + "loss": 0.6844, + "step": 40335 + }, + { + "epoch": 1.48, + "learning_rate": 5.2470062621305886e-08, + "loss": 0.697, + "step": 40336 + }, + { + "epoch": 1.48, + "learning_rate": 5.2466400556633826e-08, + "loss": 0.6944, + "step": 40337 + }, + { + "epoch": 1.48, + "learning_rate": 5.2462738491961766e-08, + "loss": 0.6886, + "step": 40338 + }, + { + "epoch": 1.48, + "learning_rate": 5.2459076427289705e-08, + "loss": 0.676, + "step": 40339 + }, + { + "epoch": 1.48, + "learning_rate": 5.2455414362617645e-08, + "loss": 0.689, + "step": 40340 + }, + { + "epoch": 1.48, + "learning_rate": 5.2451752297945585e-08, + "loss": 0.6893, + "step": 40341 + }, + { + "epoch": 1.48, + "learning_rate": 5.244809023327351e-08, + "loss": 0.6916, + "step": 40342 + }, + { + "epoch": 1.48, + "learning_rate": 5.244442816860145e-08, + "loss": 0.7031, + "step": 40343 + }, + { + "epoch": 1.48, + "learning_rate": 5.244076610392939e-08, + "loss": 0.6951, + "step": 40344 + }, + { + "epoch": 1.48, + "learning_rate": 5.243710403925733e-08, + "loss": 0.7, + "step": 40345 + }, + { + "epoch": 1.48, + "learning_rate": 5.243344197458527e-08, + "loss": 0.6882, + "step": 40346 + }, + { + "epoch": 1.48, + "learning_rate": 5.2429779909913205e-08, + "loss": 0.7092, + "step": 40347 + }, + { + "epoch": 1.48, + "learning_rate": 5.2426117845241145e-08, + "loss": 0.6968, + "step": 40348 + }, + { + "epoch": 1.48, + "learning_rate": 5.2422455780569085e-08, + "loss": 0.6817, + "step": 40349 + }, + { + "epoch": 1.48, + "learning_rate": 5.2418793715897025e-08, + "loss": 0.6796, + "step": 40350 + }, + { + "epoch": 1.48, + "learning_rate": 5.241513165122495e-08, + "loss": 0.6897, + "step": 40351 + }, + { + "epoch": 1.48, + "learning_rate": 5.241146958655289e-08, + "loss": 0.6908, + "step": 40352 + }, + { + "epoch": 1.48, + "learning_rate": 5.240780752188083e-08, + "loss": 0.6817, + "step": 40353 + }, + { + "epoch": 1.48, + "learning_rate": 5.240414545720877e-08, + "loss": 0.6854, + "step": 40354 + }, + { + "epoch": 1.48, + "learning_rate": 5.240048339253671e-08, + "loss": 0.687, + "step": 40355 + }, + { + "epoch": 1.48, + "learning_rate": 5.239682132786465e-08, + "loss": 0.6873, + "step": 40356 + }, + { + "epoch": 1.48, + "learning_rate": 5.239315926319259e-08, + "loss": 0.6968, + "step": 40357 + }, + { + "epoch": 1.48, + "learning_rate": 5.2389497198520524e-08, + "loss": 0.694, + "step": 40358 + }, + { + "epoch": 1.48, + "learning_rate": 5.2385835133848464e-08, + "loss": 0.6975, + "step": 40359 + }, + { + "epoch": 1.48, + "learning_rate": 5.238217306917639e-08, + "loss": 0.7031, + "step": 40360 + }, + { + "epoch": 1.48, + "learning_rate": 5.237851100450433e-08, + "loss": 0.6896, + "step": 40361 + }, + { + "epoch": 1.48, + "learning_rate": 5.237484893983227e-08, + "loss": 0.6933, + "step": 40362 + }, + { + "epoch": 1.48, + "learning_rate": 5.237118687516021e-08, + "loss": 0.6815, + "step": 40363 + }, + { + "epoch": 1.48, + "learning_rate": 5.236752481048815e-08, + "loss": 0.7077, + "step": 40364 + }, + { + "epoch": 1.48, + "learning_rate": 5.236386274581609e-08, + "loss": 0.7027, + "step": 40365 + }, + { + "epoch": 1.48, + "learning_rate": 5.236020068114403e-08, + "loss": 0.6892, + "step": 40366 + }, + { + "epoch": 1.48, + "learning_rate": 5.235653861647197e-08, + "loss": 0.6958, + "step": 40367 + }, + { + "epoch": 1.48, + "learning_rate": 5.2352876551799904e-08, + "loss": 0.6898, + "step": 40368 + }, + { + "epoch": 1.48, + "learning_rate": 5.234921448712784e-08, + "loss": 0.6964, + "step": 40369 + }, + { + "epoch": 1.48, + "learning_rate": 5.234555242245578e-08, + "loss": 0.7031, + "step": 40370 + }, + { + "epoch": 1.48, + "learning_rate": 5.234189035778371e-08, + "loss": 0.7048, + "step": 40371 + }, + { + "epoch": 1.48, + "learning_rate": 5.233822829311165e-08, + "loss": 0.6935, + "step": 40372 + }, + { + "epoch": 1.48, + "learning_rate": 5.233456622843959e-08, + "loss": 0.709, + "step": 40373 + }, + { + "epoch": 1.48, + "learning_rate": 5.233090416376753e-08, + "loss": 0.7141, + "step": 40374 + }, + { + "epoch": 1.48, + "learning_rate": 5.232724209909547e-08, + "loss": 0.686, + "step": 40375 + }, + { + "epoch": 1.48, + "learning_rate": 5.232358003442341e-08, + "loss": 0.6789, + "step": 40376 + }, + { + "epoch": 1.48, + "learning_rate": 5.231991796975135e-08, + "loss": 0.6929, + "step": 40377 + }, + { + "epoch": 1.48, + "learning_rate": 5.2316255905079277e-08, + "loss": 0.6709, + "step": 40378 + }, + { + "epoch": 1.48, + "learning_rate": 5.2312593840407217e-08, + "loss": 0.7047, + "step": 40379 + }, + { + "epoch": 1.48, + "learning_rate": 5.2308931775735157e-08, + "loss": 0.6889, + "step": 40380 + }, + { + "epoch": 1.48, + "learning_rate": 5.2305269711063096e-08, + "loss": 0.687, + "step": 40381 + }, + { + "epoch": 1.48, + "learning_rate": 5.230160764639103e-08, + "loss": 0.6905, + "step": 40382 + }, + { + "epoch": 1.48, + "learning_rate": 5.229794558171897e-08, + "loss": 0.6892, + "step": 40383 + }, + { + "epoch": 1.48, + "learning_rate": 5.229428351704691e-08, + "loss": 0.6835, + "step": 40384 + }, + { + "epoch": 1.48, + "learning_rate": 5.229062145237485e-08, + "loss": 0.6975, + "step": 40385 + }, + { + "epoch": 1.48, + "learning_rate": 5.228695938770279e-08, + "loss": 0.6903, + "step": 40386 + }, + { + "epoch": 1.48, + "learning_rate": 5.2283297323030716e-08, + "loss": 0.683, + "step": 40387 + }, + { + "epoch": 1.48, + "learning_rate": 5.2279635258358656e-08, + "loss": 0.6991, + "step": 40388 + }, + { + "epoch": 1.48, + "learning_rate": 5.2275973193686596e-08, + "loss": 0.705, + "step": 40389 + }, + { + "epoch": 1.48, + "learning_rate": 5.2272311129014536e-08, + "loss": 0.6923, + "step": 40390 + }, + { + "epoch": 1.48, + "learning_rate": 5.2268649064342476e-08, + "loss": 0.6949, + "step": 40391 + }, + { + "epoch": 1.48, + "learning_rate": 5.2264986999670416e-08, + "loss": 0.7055, + "step": 40392 + }, + { + "epoch": 1.48, + "learning_rate": 5.226132493499835e-08, + "loss": 0.7046, + "step": 40393 + }, + { + "epoch": 1.48, + "learning_rate": 5.225766287032629e-08, + "loss": 0.6946, + "step": 40394 + }, + { + "epoch": 1.48, + "learning_rate": 5.225400080565423e-08, + "loss": 0.6789, + "step": 40395 + }, + { + "epoch": 1.48, + "learning_rate": 5.2250338740982156e-08, + "loss": 0.7088, + "step": 40396 + }, + { + "epoch": 1.48, + "learning_rate": 5.2246676676310096e-08, + "loss": 0.6839, + "step": 40397 + }, + { + "epoch": 1.48, + "learning_rate": 5.2243014611638036e-08, + "loss": 0.6921, + "step": 40398 + }, + { + "epoch": 1.48, + "learning_rate": 5.2239352546965975e-08, + "loss": 0.6864, + "step": 40399 + }, + { + "epoch": 1.48, + "learning_rate": 5.2235690482293915e-08, + "loss": 0.6913, + "step": 40400 + }, + { + "epoch": 1.48, + "learning_rate": 5.2232028417621855e-08, + "loss": 0.6964, + "step": 40401 + }, + { + "epoch": 1.48, + "learning_rate": 5.2228366352949795e-08, + "loss": 0.6948, + "step": 40402 + }, + { + "epoch": 1.48, + "learning_rate": 5.2224704288277735e-08, + "loss": 0.6974, + "step": 40403 + }, + { + "epoch": 1.48, + "learning_rate": 5.222104222360567e-08, + "loss": 0.6798, + "step": 40404 + }, + { + "epoch": 1.48, + "learning_rate": 5.22173801589336e-08, + "loss": 0.6961, + "step": 40405 + }, + { + "epoch": 1.48, + "learning_rate": 5.221371809426154e-08, + "loss": 0.6879, + "step": 40406 + }, + { + "epoch": 1.48, + "learning_rate": 5.2210056029589475e-08, + "loss": 0.6835, + "step": 40407 + }, + { + "epoch": 1.48, + "learning_rate": 5.2206393964917415e-08, + "loss": 0.6967, + "step": 40408 + }, + { + "epoch": 1.48, + "learning_rate": 5.2202731900245355e-08, + "loss": 0.6839, + "step": 40409 + }, + { + "epoch": 1.48, + "learning_rate": 5.2199069835573295e-08, + "loss": 0.691, + "step": 40410 + }, + { + "epoch": 1.48, + "learning_rate": 5.2195407770901235e-08, + "loss": 0.6859, + "step": 40411 + }, + { + "epoch": 1.48, + "learning_rate": 5.2191745706229175e-08, + "loss": 0.6942, + "step": 40412 + }, + { + "epoch": 1.48, + "learning_rate": 5.2188083641557115e-08, + "loss": 0.6904, + "step": 40413 + }, + { + "epoch": 1.48, + "learning_rate": 5.218442157688504e-08, + "loss": 0.6956, + "step": 40414 + }, + { + "epoch": 1.48, + "learning_rate": 5.218075951221298e-08, + "loss": 0.6966, + "step": 40415 + }, + { + "epoch": 1.48, + "learning_rate": 5.217709744754092e-08, + "loss": 0.7051, + "step": 40416 + }, + { + "epoch": 1.48, + "learning_rate": 5.217343538286886e-08, + "loss": 0.6953, + "step": 40417 + }, + { + "epoch": 1.48, + "learning_rate": 5.2169773318196794e-08, + "loss": 0.6834, + "step": 40418 + }, + { + "epoch": 1.48, + "learning_rate": 5.2166111253524734e-08, + "loss": 0.6907, + "step": 40419 + }, + { + "epoch": 1.48, + "learning_rate": 5.2162449188852674e-08, + "loss": 0.6914, + "step": 40420 + }, + { + "epoch": 1.48, + "learning_rate": 5.2158787124180614e-08, + "loss": 0.6927, + "step": 40421 + }, + { + "epoch": 1.48, + "learning_rate": 5.2155125059508554e-08, + "loss": 0.6925, + "step": 40422 + }, + { + "epoch": 1.48, + "learning_rate": 5.215146299483648e-08, + "loss": 0.6943, + "step": 40423 + }, + { + "epoch": 1.48, + "learning_rate": 5.214780093016442e-08, + "loss": 0.6996, + "step": 40424 + }, + { + "epoch": 1.48, + "learning_rate": 5.214413886549236e-08, + "loss": 0.6934, + "step": 40425 + }, + { + "epoch": 1.48, + "learning_rate": 5.21404768008203e-08, + "loss": 0.7058, + "step": 40426 + }, + { + "epoch": 1.48, + "learning_rate": 5.213681473614824e-08, + "loss": 0.6953, + "step": 40427 + }, + { + "epoch": 1.48, + "learning_rate": 5.2133152671476174e-08, + "loss": 0.6835, + "step": 40428 + }, + { + "epoch": 1.48, + "learning_rate": 5.2129490606804114e-08, + "loss": 0.696, + "step": 40429 + }, + { + "epoch": 1.48, + "learning_rate": 5.2125828542132054e-08, + "loss": 0.7062, + "step": 40430 + }, + { + "epoch": 1.48, + "learning_rate": 5.2122166477459994e-08, + "loss": 0.6845, + "step": 40431 + }, + { + "epoch": 1.48, + "learning_rate": 5.211850441278792e-08, + "loss": 0.7007, + "step": 40432 + }, + { + "epoch": 1.48, + "learning_rate": 5.211484234811586e-08, + "loss": 0.6869, + "step": 40433 + }, + { + "epoch": 1.48, + "learning_rate": 5.21111802834438e-08, + "loss": 0.7048, + "step": 40434 + }, + { + "epoch": 1.48, + "learning_rate": 5.210751821877174e-08, + "loss": 0.6914, + "step": 40435 + }, + { + "epoch": 1.48, + "learning_rate": 5.210385615409968e-08, + "loss": 0.6926, + "step": 40436 + }, + { + "epoch": 1.48, + "learning_rate": 5.210019408942762e-08, + "loss": 0.6927, + "step": 40437 + }, + { + "epoch": 1.48, + "learning_rate": 5.209653202475556e-08, + "loss": 0.6936, + "step": 40438 + }, + { + "epoch": 1.48, + "learning_rate": 5.209286996008349e-08, + "loss": 0.6952, + "step": 40439 + }, + { + "epoch": 1.48, + "learning_rate": 5.208920789541143e-08, + "loss": 0.6817, + "step": 40440 + }, + { + "epoch": 1.48, + "learning_rate": 5.2085545830739366e-08, + "loss": 0.6956, + "step": 40441 + }, + { + "epoch": 1.48, + "learning_rate": 5.20818837660673e-08, + "loss": 0.703, + "step": 40442 + }, + { + "epoch": 1.48, + "learning_rate": 5.207822170139524e-08, + "loss": 0.6802, + "step": 40443 + }, + { + "epoch": 1.48, + "learning_rate": 5.207455963672318e-08, + "loss": 0.6868, + "step": 40444 + }, + { + "epoch": 1.48, + "learning_rate": 5.207089757205112e-08, + "loss": 0.7034, + "step": 40445 + }, + { + "epoch": 1.48, + "learning_rate": 5.206723550737906e-08, + "loss": 0.699, + "step": 40446 + }, + { + "epoch": 1.48, + "learning_rate": 5.2063573442707e-08, + "loss": 0.7054, + "step": 40447 + }, + { + "epoch": 1.48, + "learning_rate": 5.205991137803494e-08, + "loss": 0.7031, + "step": 40448 + }, + { + "epoch": 1.48, + "learning_rate": 5.205624931336288e-08, + "loss": 0.6912, + "step": 40449 + }, + { + "epoch": 1.48, + "learning_rate": 5.2052587248690806e-08, + "loss": 0.6833, + "step": 40450 + }, + { + "epoch": 1.48, + "learning_rate": 5.2048925184018746e-08, + "loss": 0.693, + "step": 40451 + }, + { + "epoch": 1.48, + "learning_rate": 5.2045263119346686e-08, + "loss": 0.7074, + "step": 40452 + }, + { + "epoch": 1.48, + "learning_rate": 5.204160105467462e-08, + "loss": 0.6893, + "step": 40453 + }, + { + "epoch": 1.48, + "learning_rate": 5.203793899000256e-08, + "loss": 0.6927, + "step": 40454 + }, + { + "epoch": 1.48, + "learning_rate": 5.20342769253305e-08, + "loss": 0.6912, + "step": 40455 + }, + { + "epoch": 1.48, + "learning_rate": 5.203061486065844e-08, + "loss": 0.7122, + "step": 40456 + }, + { + "epoch": 1.48, + "learning_rate": 5.202695279598638e-08, + "loss": 0.6754, + "step": 40457 + }, + { + "epoch": 1.48, + "learning_rate": 5.202329073131432e-08, + "loss": 0.6925, + "step": 40458 + }, + { + "epoch": 1.48, + "learning_rate": 5.2019628666642245e-08, + "loss": 0.6956, + "step": 40459 + }, + { + "epoch": 1.48, + "learning_rate": 5.2015966601970185e-08, + "loss": 0.6847, + "step": 40460 + }, + { + "epoch": 1.48, + "learning_rate": 5.2012304537298125e-08, + "loss": 0.6966, + "step": 40461 + }, + { + "epoch": 1.48, + "learning_rate": 5.2008642472626065e-08, + "loss": 0.7009, + "step": 40462 + }, + { + "epoch": 1.48, + "learning_rate": 5.2004980407954005e-08, + "loss": 0.6961, + "step": 40463 + }, + { + "epoch": 1.48, + "learning_rate": 5.200131834328194e-08, + "loss": 0.7039, + "step": 40464 + }, + { + "epoch": 1.48, + "learning_rate": 5.199765627860988e-08, + "loss": 0.7059, + "step": 40465 + }, + { + "epoch": 1.48, + "learning_rate": 5.199399421393782e-08, + "loss": 0.688, + "step": 40466 + }, + { + "epoch": 1.48, + "learning_rate": 5.199033214926576e-08, + "loss": 0.6932, + "step": 40467 + }, + { + "epoch": 1.48, + "learning_rate": 5.1986670084593685e-08, + "loss": 0.6884, + "step": 40468 + }, + { + "epoch": 1.48, + "learning_rate": 5.1983008019921625e-08, + "loss": 0.706, + "step": 40469 + }, + { + "epoch": 1.48, + "learning_rate": 5.1979345955249565e-08, + "loss": 0.6931, + "step": 40470 + }, + { + "epoch": 1.48, + "learning_rate": 5.1975683890577505e-08, + "loss": 0.693, + "step": 40471 + }, + { + "epoch": 1.48, + "learning_rate": 5.1972021825905445e-08, + "loss": 0.6784, + "step": 40472 + }, + { + "epoch": 1.48, + "learning_rate": 5.1968359761233385e-08, + "loss": 0.6816, + "step": 40473 + }, + { + "epoch": 1.48, + "learning_rate": 5.1964697696561325e-08, + "loss": 0.6845, + "step": 40474 + }, + { + "epoch": 1.48, + "learning_rate": 5.196103563188926e-08, + "loss": 0.6931, + "step": 40475 + }, + { + "epoch": 1.48, + "learning_rate": 5.19573735672172e-08, + "loss": 0.6968, + "step": 40476 + }, + { + "epoch": 1.48, + "learning_rate": 5.195371150254513e-08, + "loss": 0.6852, + "step": 40477 + }, + { + "epoch": 1.48, + "learning_rate": 5.1950049437873064e-08, + "loss": 0.6845, + "step": 40478 + }, + { + "epoch": 1.48, + "learning_rate": 5.1946387373201004e-08, + "loss": 0.688, + "step": 40479 + }, + { + "epoch": 1.48, + "learning_rate": 5.1942725308528944e-08, + "loss": 0.6832, + "step": 40480 + }, + { + "epoch": 1.48, + "learning_rate": 5.1939063243856884e-08, + "loss": 0.6971, + "step": 40481 + }, + { + "epoch": 1.48, + "learning_rate": 5.1935401179184824e-08, + "loss": 0.6956, + "step": 40482 + }, + { + "epoch": 1.48, + "learning_rate": 5.1931739114512764e-08, + "loss": 0.6951, + "step": 40483 + }, + { + "epoch": 1.48, + "learning_rate": 5.1928077049840704e-08, + "loss": 0.6897, + "step": 40484 + }, + { + "epoch": 1.48, + "learning_rate": 5.1924414985168644e-08, + "loss": 0.6957, + "step": 40485 + }, + { + "epoch": 1.48, + "learning_rate": 5.192075292049657e-08, + "loss": 0.6791, + "step": 40486 + }, + { + "epoch": 1.48, + "learning_rate": 5.191709085582451e-08, + "loss": 0.6901, + "step": 40487 + }, + { + "epoch": 1.48, + "learning_rate": 5.1913428791152444e-08, + "loss": 0.7066, + "step": 40488 + }, + { + "epoch": 1.48, + "learning_rate": 5.1909766726480384e-08, + "loss": 0.6981, + "step": 40489 + }, + { + "epoch": 1.48, + "learning_rate": 5.1906104661808324e-08, + "loss": 0.6939, + "step": 40490 + }, + { + "epoch": 1.48, + "learning_rate": 5.1902442597136264e-08, + "loss": 0.6869, + "step": 40491 + }, + { + "epoch": 1.48, + "learning_rate": 5.1898780532464204e-08, + "loss": 0.6998, + "step": 40492 + }, + { + "epoch": 1.48, + "learning_rate": 5.1895118467792143e-08, + "loss": 0.6996, + "step": 40493 + }, + { + "epoch": 1.48, + "learning_rate": 5.1891456403120083e-08, + "loss": 0.6957, + "step": 40494 + }, + { + "epoch": 1.48, + "learning_rate": 5.188779433844801e-08, + "loss": 0.6798, + "step": 40495 + }, + { + "epoch": 1.48, + "learning_rate": 5.188413227377595e-08, + "loss": 0.6999, + "step": 40496 + }, + { + "epoch": 1.48, + "learning_rate": 5.188047020910389e-08, + "loss": 0.6781, + "step": 40497 + }, + { + "epoch": 1.48, + "learning_rate": 5.187680814443183e-08, + "loss": 0.7011, + "step": 40498 + }, + { + "epoch": 1.48, + "learning_rate": 5.187314607975976e-08, + "loss": 0.6963, + "step": 40499 + }, + { + "epoch": 1.48, + "learning_rate": 5.18694840150877e-08, + "loss": 0.7071, + "step": 40500 + }, + { + "epoch": 1.48, + "learning_rate": 5.186582195041564e-08, + "loss": 0.689, + "step": 40501 + }, + { + "epoch": 1.48, + "learning_rate": 5.186215988574358e-08, + "loss": 0.6988, + "step": 40502 + }, + { + "epoch": 1.48, + "learning_rate": 5.185849782107152e-08, + "loss": 0.6928, + "step": 40503 + }, + { + "epoch": 1.48, + "learning_rate": 5.185483575639945e-08, + "loss": 0.6947, + "step": 40504 + }, + { + "epoch": 1.48, + "learning_rate": 5.185117369172739e-08, + "loss": 0.687, + "step": 40505 + }, + { + "epoch": 1.48, + "learning_rate": 5.184751162705533e-08, + "loss": 0.6784, + "step": 40506 + }, + { + "epoch": 1.48, + "learning_rate": 5.184384956238327e-08, + "loss": 0.6886, + "step": 40507 + }, + { + "epoch": 1.48, + "learning_rate": 5.184018749771121e-08, + "loss": 0.6909, + "step": 40508 + }, + { + "epoch": 1.48, + "learning_rate": 5.183652543303915e-08, + "loss": 0.6763, + "step": 40509 + }, + { + "epoch": 1.48, + "learning_rate": 5.183286336836708e-08, + "loss": 0.6688, + "step": 40510 + }, + { + "epoch": 1.48, + "learning_rate": 5.182920130369502e-08, + "loss": 0.6905, + "step": 40511 + }, + { + "epoch": 1.48, + "learning_rate": 5.182553923902296e-08, + "loss": 0.6961, + "step": 40512 + }, + { + "epoch": 1.48, + "learning_rate": 5.182187717435089e-08, + "loss": 0.6792, + "step": 40513 + }, + { + "epoch": 1.48, + "learning_rate": 5.181821510967883e-08, + "loss": 0.7028, + "step": 40514 + }, + { + "epoch": 1.48, + "learning_rate": 5.181455304500677e-08, + "loss": 0.6866, + "step": 40515 + }, + { + "epoch": 1.48, + "learning_rate": 5.181089098033471e-08, + "loss": 0.6987, + "step": 40516 + }, + { + "epoch": 1.48, + "learning_rate": 5.180722891566265e-08, + "loss": 0.7063, + "step": 40517 + }, + { + "epoch": 1.48, + "learning_rate": 5.180356685099059e-08, + "loss": 0.6792, + "step": 40518 + }, + { + "epoch": 1.48, + "learning_rate": 5.179990478631853e-08, + "loss": 0.6883, + "step": 40519 + }, + { + "epoch": 1.48, + "learning_rate": 5.179624272164647e-08, + "loss": 0.6945, + "step": 40520 + }, + { + "epoch": 1.48, + "learning_rate": 5.17925806569744e-08, + "loss": 0.683, + "step": 40521 + }, + { + "epoch": 1.48, + "learning_rate": 5.1788918592302335e-08, + "loss": 0.7021, + "step": 40522 + }, + { + "epoch": 1.48, + "learning_rate": 5.1785256527630275e-08, + "loss": 0.6967, + "step": 40523 + }, + { + "epoch": 1.48, + "learning_rate": 5.178159446295821e-08, + "loss": 0.6875, + "step": 40524 + }, + { + "epoch": 1.48, + "learning_rate": 5.177793239828615e-08, + "loss": 0.6908, + "step": 40525 + }, + { + "epoch": 1.48, + "learning_rate": 5.177427033361409e-08, + "loss": 0.6903, + "step": 40526 + }, + { + "epoch": 1.48, + "learning_rate": 5.177060826894203e-08, + "loss": 0.6971, + "step": 40527 + }, + { + "epoch": 1.48, + "learning_rate": 5.176694620426997e-08, + "loss": 0.693, + "step": 40528 + }, + { + "epoch": 1.48, + "learning_rate": 5.176328413959791e-08, + "loss": 0.6929, + "step": 40529 + }, + { + "epoch": 1.48, + "learning_rate": 5.175962207492585e-08, + "loss": 0.6876, + "step": 40530 + }, + { + "epoch": 1.48, + "learning_rate": 5.1755960010253775e-08, + "loss": 0.6804, + "step": 40531 + }, + { + "epoch": 1.48, + "learning_rate": 5.1752297945581715e-08, + "loss": 0.6864, + "step": 40532 + }, + { + "epoch": 1.48, + "learning_rate": 5.1748635880909655e-08, + "loss": 0.6876, + "step": 40533 + }, + { + "epoch": 1.48, + "learning_rate": 5.1744973816237595e-08, + "loss": 0.7142, + "step": 40534 + }, + { + "epoch": 1.48, + "learning_rate": 5.174131175156553e-08, + "loss": 0.6854, + "step": 40535 + }, + { + "epoch": 1.48, + "learning_rate": 5.173764968689347e-08, + "loss": 0.7031, + "step": 40536 + }, + { + "epoch": 1.48, + "learning_rate": 5.173398762222141e-08, + "loss": 0.7036, + "step": 40537 + }, + { + "epoch": 1.48, + "learning_rate": 5.173032555754935e-08, + "loss": 0.686, + "step": 40538 + }, + { + "epoch": 1.48, + "learning_rate": 5.172666349287729e-08, + "loss": 0.7005, + "step": 40539 + }, + { + "epoch": 1.48, + "learning_rate": 5.1723001428205214e-08, + "loss": 0.7068, + "step": 40540 + }, + { + "epoch": 1.48, + "learning_rate": 5.1719339363533154e-08, + "loss": 0.7007, + "step": 40541 + }, + { + "epoch": 1.48, + "learning_rate": 5.1715677298861094e-08, + "loss": 0.6886, + "step": 40542 + }, + { + "epoch": 1.48, + "learning_rate": 5.1712015234189034e-08, + "loss": 0.699, + "step": 40543 + }, + { + "epoch": 1.48, + "learning_rate": 5.1708353169516974e-08, + "loss": 0.6972, + "step": 40544 + }, + { + "epoch": 1.48, + "learning_rate": 5.1704691104844914e-08, + "loss": 0.6793, + "step": 40545 + }, + { + "epoch": 1.48, + "learning_rate": 5.170102904017285e-08, + "loss": 0.6941, + "step": 40546 + }, + { + "epoch": 1.48, + "learning_rate": 5.169736697550079e-08, + "loss": 0.6962, + "step": 40547 + }, + { + "epoch": 1.48, + "learning_rate": 5.169370491082873e-08, + "loss": 0.6915, + "step": 40548 + }, + { + "epoch": 1.48, + "learning_rate": 5.1690042846156654e-08, + "loss": 0.6846, + "step": 40549 + }, + { + "epoch": 1.48, + "learning_rate": 5.1686380781484594e-08, + "loss": 0.7009, + "step": 40550 + }, + { + "epoch": 1.48, + "learning_rate": 5.1682718716812534e-08, + "loss": 0.6939, + "step": 40551 + }, + { + "epoch": 1.48, + "learning_rate": 5.1679056652140474e-08, + "loss": 0.7043, + "step": 40552 + }, + { + "epoch": 1.48, + "learning_rate": 5.1675394587468413e-08, + "loss": 0.693, + "step": 40553 + }, + { + "epoch": 1.48, + "learning_rate": 5.1671732522796353e-08, + "loss": 0.6858, + "step": 40554 + }, + { + "epoch": 1.48, + "learning_rate": 5.1668070458124293e-08, + "loss": 0.6944, + "step": 40555 + }, + { + "epoch": 1.48, + "learning_rate": 5.166440839345223e-08, + "loss": 0.6958, + "step": 40556 + }, + { + "epoch": 1.48, + "learning_rate": 5.1660746328780167e-08, + "loss": 0.6841, + "step": 40557 + }, + { + "epoch": 1.48, + "learning_rate": 5.16570842641081e-08, + "loss": 0.695, + "step": 40558 + }, + { + "epoch": 1.48, + "learning_rate": 5.165342219943603e-08, + "loss": 0.6946, + "step": 40559 + }, + { + "epoch": 1.48, + "learning_rate": 5.164976013476397e-08, + "loss": 0.698, + "step": 40560 + }, + { + "epoch": 1.48, + "learning_rate": 5.164609807009191e-08, + "loss": 0.7012, + "step": 40561 + }, + { + "epoch": 1.48, + "learning_rate": 5.164243600541985e-08, + "loss": 0.6874, + "step": 40562 + }, + { + "epoch": 1.48, + "learning_rate": 5.163877394074779e-08, + "loss": 0.6942, + "step": 40563 + }, + { + "epoch": 1.48, + "learning_rate": 5.163511187607573e-08, + "loss": 0.7029, + "step": 40564 + }, + { + "epoch": 1.48, + "learning_rate": 5.163144981140367e-08, + "loss": 0.7021, + "step": 40565 + }, + { + "epoch": 1.48, + "learning_rate": 5.162778774673161e-08, + "loss": 0.7067, + "step": 40566 + }, + { + "epoch": 1.48, + "learning_rate": 5.162412568205954e-08, + "loss": 0.6918, + "step": 40567 + }, + { + "epoch": 1.48, + "learning_rate": 5.162046361738748e-08, + "loss": 0.686, + "step": 40568 + }, + { + "epoch": 1.48, + "learning_rate": 5.161680155271542e-08, + "loss": 0.693, + "step": 40569 + }, + { + "epoch": 1.48, + "learning_rate": 5.161313948804335e-08, + "loss": 0.69, + "step": 40570 + }, + { + "epoch": 1.48, + "learning_rate": 5.160947742337129e-08, + "loss": 0.69, + "step": 40571 + }, + { + "epoch": 1.48, + "learning_rate": 5.160581535869923e-08, + "loss": 0.673, + "step": 40572 + }, + { + "epoch": 1.48, + "learning_rate": 5.160215329402717e-08, + "loss": 0.6978, + "step": 40573 + }, + { + "epoch": 1.48, + "learning_rate": 5.159849122935511e-08, + "loss": 0.6952, + "step": 40574 + }, + { + "epoch": 1.48, + "learning_rate": 5.159482916468305e-08, + "loss": 0.6974, + "step": 40575 + }, + { + "epoch": 1.48, + "learning_rate": 5.159116710001098e-08, + "loss": 0.675, + "step": 40576 + }, + { + "epoch": 1.48, + "learning_rate": 5.158750503533892e-08, + "loss": 0.6872, + "step": 40577 + }, + { + "epoch": 1.48, + "learning_rate": 5.158384297066686e-08, + "loss": 0.6951, + "step": 40578 + }, + { + "epoch": 1.48, + "learning_rate": 5.15801809059948e-08, + "loss": 0.7017, + "step": 40579 + }, + { + "epoch": 1.48, + "learning_rate": 5.157651884132274e-08, + "loss": 0.6807, + "step": 40580 + }, + { + "epoch": 1.48, + "learning_rate": 5.157285677665067e-08, + "loss": 0.6908, + "step": 40581 + }, + { + "epoch": 1.48, + "learning_rate": 5.156919471197861e-08, + "loss": 0.6861, + "step": 40582 + }, + { + "epoch": 1.48, + "learning_rate": 5.156553264730655e-08, + "loss": 0.6958, + "step": 40583 + }, + { + "epoch": 1.48, + "learning_rate": 5.156187058263449e-08, + "loss": 0.6957, + "step": 40584 + }, + { + "epoch": 1.48, + "learning_rate": 5.155820851796242e-08, + "loss": 0.7, + "step": 40585 + }, + { + "epoch": 1.48, + "learning_rate": 5.155454645329036e-08, + "loss": 0.6941, + "step": 40586 + }, + { + "epoch": 1.48, + "learning_rate": 5.15508843886183e-08, + "loss": 0.6996, + "step": 40587 + }, + { + "epoch": 1.48, + "learning_rate": 5.154722232394624e-08, + "loss": 0.6921, + "step": 40588 + }, + { + "epoch": 1.49, + "learning_rate": 5.154356025927418e-08, + "loss": 0.696, + "step": 40589 + }, + { + "epoch": 1.49, + "learning_rate": 5.153989819460212e-08, + "loss": 0.6909, + "step": 40590 + }, + { + "epoch": 1.49, + "learning_rate": 5.153623612993006e-08, + "loss": 0.6824, + "step": 40591 + }, + { + "epoch": 1.49, + "learning_rate": 5.153257406525799e-08, + "loss": 0.7025, + "step": 40592 + }, + { + "epoch": 1.49, + "learning_rate": 5.152891200058593e-08, + "loss": 0.6894, + "step": 40593 + }, + { + "epoch": 1.49, + "learning_rate": 5.1525249935913865e-08, + "loss": 0.7049, + "step": 40594 + }, + { + "epoch": 1.49, + "learning_rate": 5.15215878712418e-08, + "loss": 0.7005, + "step": 40595 + }, + { + "epoch": 1.49, + "learning_rate": 5.151792580656974e-08, + "loss": 0.6876, + "step": 40596 + }, + { + "epoch": 1.49, + "learning_rate": 5.151426374189768e-08, + "loss": 0.6978, + "step": 40597 + }, + { + "epoch": 1.49, + "learning_rate": 5.151060167722562e-08, + "loss": 0.6898, + "step": 40598 + }, + { + "epoch": 1.49, + "learning_rate": 5.150693961255356e-08, + "loss": 0.6876, + "step": 40599 + }, + { + "epoch": 1.49, + "learning_rate": 5.15032775478815e-08, + "loss": 0.6926, + "step": 40600 + }, + { + "epoch": 1.49, + "learning_rate": 5.149961548320944e-08, + "loss": 0.6875, + "step": 40601 + }, + { + "epoch": 1.49, + "learning_rate": 5.149595341853738e-08, + "loss": 0.7052, + "step": 40602 + }, + { + "epoch": 1.49, + "learning_rate": 5.1492291353865304e-08, + "loss": 0.6939, + "step": 40603 + }, + { + "epoch": 1.49, + "learning_rate": 5.1488629289193244e-08, + "loss": 0.7019, + "step": 40604 + }, + { + "epoch": 1.49, + "learning_rate": 5.1484967224521184e-08, + "loss": 0.6976, + "step": 40605 + }, + { + "epoch": 1.49, + "learning_rate": 5.148130515984912e-08, + "loss": 0.7121, + "step": 40606 + }, + { + "epoch": 1.49, + "learning_rate": 5.147764309517706e-08, + "loss": 0.6976, + "step": 40607 + }, + { + "epoch": 1.49, + "learning_rate": 5.1473981030505e-08, + "loss": 0.7088, + "step": 40608 + }, + { + "epoch": 1.49, + "learning_rate": 5.147031896583294e-08, + "loss": 0.6987, + "step": 40609 + }, + { + "epoch": 1.49, + "learning_rate": 5.146665690116088e-08, + "loss": 0.6855, + "step": 40610 + }, + { + "epoch": 1.49, + "learning_rate": 5.146299483648882e-08, + "loss": 0.688, + "step": 40611 + }, + { + "epoch": 1.49, + "learning_rate": 5.1459332771816744e-08, + "loss": 0.6953, + "step": 40612 + }, + { + "epoch": 1.49, + "learning_rate": 5.1455670707144683e-08, + "loss": 0.6978, + "step": 40613 + }, + { + "epoch": 1.49, + "learning_rate": 5.1452008642472623e-08, + "loss": 0.6752, + "step": 40614 + }, + { + "epoch": 1.49, + "learning_rate": 5.144834657780056e-08, + "loss": 0.707, + "step": 40615 + }, + { + "epoch": 1.49, + "learning_rate": 5.14446845131285e-08, + "loss": 0.6979, + "step": 40616 + }, + { + "epoch": 1.49, + "learning_rate": 5.1441022448456437e-08, + "loss": 0.6877, + "step": 40617 + }, + { + "epoch": 1.49, + "learning_rate": 5.1437360383784376e-08, + "loss": 0.6958, + "step": 40618 + }, + { + "epoch": 1.49, + "learning_rate": 5.1433698319112316e-08, + "loss": 0.6906, + "step": 40619 + }, + { + "epoch": 1.49, + "learning_rate": 5.1430036254440256e-08, + "loss": 0.6984, + "step": 40620 + }, + { + "epoch": 1.49, + "learning_rate": 5.142637418976818e-08, + "loss": 0.6892, + "step": 40621 + }, + { + "epoch": 1.49, + "learning_rate": 5.142271212509612e-08, + "loss": 0.6927, + "step": 40622 + }, + { + "epoch": 1.49, + "learning_rate": 5.141905006042406e-08, + "loss": 0.699, + "step": 40623 + }, + { + "epoch": 1.49, + "learning_rate": 5.1415387995752e-08, + "loss": 0.6907, + "step": 40624 + }, + { + "epoch": 1.49, + "learning_rate": 5.141172593107994e-08, + "loss": 0.6879, + "step": 40625 + }, + { + "epoch": 1.49, + "learning_rate": 5.140806386640788e-08, + "loss": 0.6871, + "step": 40626 + }, + { + "epoch": 1.49, + "learning_rate": 5.140440180173582e-08, + "loss": 0.7047, + "step": 40627 + }, + { + "epoch": 1.49, + "learning_rate": 5.1400739737063756e-08, + "loss": 0.6886, + "step": 40628 + }, + { + "epoch": 1.49, + "learning_rate": 5.139707767239169e-08, + "loss": 0.6931, + "step": 40629 + }, + { + "epoch": 1.49, + "learning_rate": 5.139341560771962e-08, + "loss": 0.6917, + "step": 40630 + }, + { + "epoch": 1.49, + "learning_rate": 5.138975354304756e-08, + "loss": 0.6876, + "step": 40631 + }, + { + "epoch": 1.49, + "learning_rate": 5.13860914783755e-08, + "loss": 0.6904, + "step": 40632 + }, + { + "epoch": 1.49, + "learning_rate": 5.138242941370344e-08, + "loss": 0.7001, + "step": 40633 + }, + { + "epoch": 1.49, + "learning_rate": 5.137876734903138e-08, + "loss": 0.7053, + "step": 40634 + }, + { + "epoch": 1.49, + "learning_rate": 5.137510528435932e-08, + "loss": 0.7065, + "step": 40635 + }, + { + "epoch": 1.49, + "learning_rate": 5.137144321968726e-08, + "loss": 0.6957, + "step": 40636 + }, + { + "epoch": 1.49, + "learning_rate": 5.13677811550152e-08, + "loss": 0.691, + "step": 40637 + }, + { + "epoch": 1.49, + "learning_rate": 5.136411909034313e-08, + "loss": 0.6948, + "step": 40638 + }, + { + "epoch": 1.49, + "learning_rate": 5.136045702567107e-08, + "loss": 0.6898, + "step": 40639 + }, + { + "epoch": 1.49, + "learning_rate": 5.135679496099901e-08, + "loss": 0.7017, + "step": 40640 + }, + { + "epoch": 1.49, + "learning_rate": 5.135313289632694e-08, + "loss": 0.6939, + "step": 40641 + }, + { + "epoch": 1.49, + "learning_rate": 5.134947083165488e-08, + "loss": 0.6895, + "step": 40642 + }, + { + "epoch": 1.49, + "learning_rate": 5.134580876698282e-08, + "loss": 0.6938, + "step": 40643 + }, + { + "epoch": 1.49, + "learning_rate": 5.134214670231076e-08, + "loss": 0.6768, + "step": 40644 + }, + { + "epoch": 1.49, + "learning_rate": 5.13384846376387e-08, + "loss": 0.6881, + "step": 40645 + }, + { + "epoch": 1.49, + "learning_rate": 5.133482257296664e-08, + "loss": 0.6981, + "step": 40646 + }, + { + "epoch": 1.49, + "learning_rate": 5.133116050829457e-08, + "loss": 0.6883, + "step": 40647 + }, + { + "epoch": 1.49, + "learning_rate": 5.132749844362251e-08, + "loss": 0.6943, + "step": 40648 + }, + { + "epoch": 1.49, + "learning_rate": 5.132383637895045e-08, + "loss": 0.6954, + "step": 40649 + }, + { + "epoch": 1.49, + "learning_rate": 5.132017431427839e-08, + "loss": 0.6911, + "step": 40650 + }, + { + "epoch": 1.49, + "learning_rate": 5.131651224960633e-08, + "loss": 0.6999, + "step": 40651 + }, + { + "epoch": 1.49, + "learning_rate": 5.131285018493426e-08, + "loss": 0.6965, + "step": 40652 + }, + { + "epoch": 1.49, + "learning_rate": 5.13091881202622e-08, + "loss": 0.7026, + "step": 40653 + }, + { + "epoch": 1.49, + "learning_rate": 5.130552605559014e-08, + "loss": 0.6923, + "step": 40654 + }, + { + "epoch": 1.49, + "learning_rate": 5.130186399091808e-08, + "loss": 0.7028, + "step": 40655 + }, + { + "epoch": 1.49, + "learning_rate": 5.129820192624601e-08, + "loss": 0.6855, + "step": 40656 + }, + { + "epoch": 1.49, + "learning_rate": 5.129453986157395e-08, + "loss": 0.7072, + "step": 40657 + }, + { + "epoch": 1.49, + "learning_rate": 5.129087779690189e-08, + "loss": 0.6914, + "step": 40658 + }, + { + "epoch": 1.49, + "learning_rate": 5.128721573222983e-08, + "loss": 0.7005, + "step": 40659 + }, + { + "epoch": 1.49, + "learning_rate": 5.128355366755777e-08, + "loss": 0.6957, + "step": 40660 + }, + { + "epoch": 1.49, + "learning_rate": 5.127989160288571e-08, + "loss": 0.688, + "step": 40661 + }, + { + "epoch": 1.49, + "learning_rate": 5.127622953821365e-08, + "loss": 0.6854, + "step": 40662 + }, + { + "epoch": 1.49, + "learning_rate": 5.127256747354158e-08, + "loss": 0.6963, + "step": 40663 + }, + { + "epoch": 1.49, + "learning_rate": 5.126890540886952e-08, + "loss": 0.7008, + "step": 40664 + }, + { + "epoch": 1.49, + "learning_rate": 5.1265243344197454e-08, + "loss": 0.6942, + "step": 40665 + }, + { + "epoch": 1.49, + "learning_rate": 5.126158127952539e-08, + "loss": 0.6944, + "step": 40666 + }, + { + "epoch": 1.49, + "learning_rate": 5.125791921485333e-08, + "loss": 0.6956, + "step": 40667 + }, + { + "epoch": 1.49, + "learning_rate": 5.125425715018127e-08, + "loss": 0.6898, + "step": 40668 + }, + { + "epoch": 1.49, + "learning_rate": 5.125059508550921e-08, + "loss": 0.6912, + "step": 40669 + }, + { + "epoch": 1.49, + "learning_rate": 5.124693302083715e-08, + "loss": 0.6925, + "step": 40670 + }, + { + "epoch": 1.49, + "learning_rate": 5.124327095616509e-08, + "loss": 0.6978, + "step": 40671 + }, + { + "epoch": 1.49, + "learning_rate": 5.123960889149303e-08, + "loss": 0.696, + "step": 40672 + }, + { + "epoch": 1.49, + "learning_rate": 5.1235946826820967e-08, + "loss": 0.6939, + "step": 40673 + }, + { + "epoch": 1.49, + "learning_rate": 5.1232284762148893e-08, + "loss": 0.6876, + "step": 40674 + }, + { + "epoch": 1.49, + "learning_rate": 5.122862269747683e-08, + "loss": 0.7065, + "step": 40675 + }, + { + "epoch": 1.49, + "learning_rate": 5.122496063280477e-08, + "loss": 0.6957, + "step": 40676 + }, + { + "epoch": 1.49, + "learning_rate": 5.1221298568132707e-08, + "loss": 0.6888, + "step": 40677 + }, + { + "epoch": 1.49, + "learning_rate": 5.1217636503460646e-08, + "loss": 0.6928, + "step": 40678 + }, + { + "epoch": 1.49, + "learning_rate": 5.1213974438788586e-08, + "loss": 0.6916, + "step": 40679 + }, + { + "epoch": 1.49, + "learning_rate": 5.1210312374116526e-08, + "loss": 0.6859, + "step": 40680 + }, + { + "epoch": 1.49, + "learning_rate": 5.1206650309444466e-08, + "loss": 0.6801, + "step": 40681 + }, + { + "epoch": 1.49, + "learning_rate": 5.1202988244772406e-08, + "loss": 0.6983, + "step": 40682 + }, + { + "epoch": 1.49, + "learning_rate": 5.119932618010033e-08, + "loss": 0.6872, + "step": 40683 + }, + { + "epoch": 1.49, + "learning_rate": 5.119566411542827e-08, + "loss": 0.6875, + "step": 40684 + }, + { + "epoch": 1.49, + "learning_rate": 5.119200205075621e-08, + "loss": 0.6909, + "step": 40685 + }, + { + "epoch": 1.49, + "learning_rate": 5.118833998608415e-08, + "loss": 0.6915, + "step": 40686 + }, + { + "epoch": 1.49, + "learning_rate": 5.118467792141209e-08, + "loss": 0.6967, + "step": 40687 + }, + { + "epoch": 1.49, + "learning_rate": 5.1181015856740026e-08, + "loss": 0.6919, + "step": 40688 + }, + { + "epoch": 1.49, + "learning_rate": 5.1177353792067966e-08, + "loss": 0.6951, + "step": 40689 + }, + { + "epoch": 1.49, + "learning_rate": 5.1173691727395906e-08, + "loss": 0.6831, + "step": 40690 + }, + { + "epoch": 1.49, + "learning_rate": 5.1170029662723846e-08, + "loss": 0.7074, + "step": 40691 + }, + { + "epoch": 1.49, + "learning_rate": 5.116636759805177e-08, + "loss": 0.7003, + "step": 40692 + }, + { + "epoch": 1.49, + "learning_rate": 5.116270553337971e-08, + "loss": 0.696, + "step": 40693 + }, + { + "epoch": 1.49, + "learning_rate": 5.115904346870765e-08, + "loss": 0.699, + "step": 40694 + }, + { + "epoch": 1.49, + "learning_rate": 5.115538140403559e-08, + "loss": 0.7007, + "step": 40695 + }, + { + "epoch": 1.49, + "learning_rate": 5.115171933936353e-08, + "loss": 0.6922, + "step": 40696 + }, + { + "epoch": 1.49, + "learning_rate": 5.114805727469147e-08, + "loss": 0.6965, + "step": 40697 + }, + { + "epoch": 1.49, + "learning_rate": 5.114439521001941e-08, + "loss": 0.7031, + "step": 40698 + }, + { + "epoch": 1.49, + "learning_rate": 5.1140733145347345e-08, + "loss": 0.6929, + "step": 40699 + }, + { + "epoch": 1.49, + "learning_rate": 5.1137071080675285e-08, + "loss": 0.7033, + "step": 40700 + }, + { + "epoch": 1.49, + "learning_rate": 5.113340901600321e-08, + "loss": 0.689, + "step": 40701 + }, + { + "epoch": 1.49, + "learning_rate": 5.112974695133115e-08, + "loss": 0.7035, + "step": 40702 + }, + { + "epoch": 1.49, + "learning_rate": 5.112608488665909e-08, + "loss": 0.6866, + "step": 40703 + }, + { + "epoch": 1.49, + "learning_rate": 5.112242282198703e-08, + "loss": 0.6905, + "step": 40704 + }, + { + "epoch": 1.49, + "learning_rate": 5.111876075731497e-08, + "loss": 0.6881, + "step": 40705 + }, + { + "epoch": 1.49, + "learning_rate": 5.111509869264291e-08, + "loss": 0.6837, + "step": 40706 + }, + { + "epoch": 1.49, + "learning_rate": 5.111143662797085e-08, + "loss": 0.6912, + "step": 40707 + }, + { + "epoch": 1.49, + "learning_rate": 5.110777456329879e-08, + "loss": 0.6975, + "step": 40708 + }, + { + "epoch": 1.49, + "learning_rate": 5.110411249862673e-08, + "loss": 0.6962, + "step": 40709 + }, + { + "epoch": 1.49, + "learning_rate": 5.110045043395466e-08, + "loss": 0.6864, + "step": 40710 + }, + { + "epoch": 1.49, + "learning_rate": 5.10967883692826e-08, + "loss": 0.6906, + "step": 40711 + }, + { + "epoch": 1.49, + "learning_rate": 5.109312630461053e-08, + "loss": 0.6816, + "step": 40712 + }, + { + "epoch": 1.49, + "learning_rate": 5.108946423993847e-08, + "loss": 0.6877, + "step": 40713 + }, + { + "epoch": 1.49, + "learning_rate": 5.108580217526641e-08, + "loss": 0.7013, + "step": 40714 + }, + { + "epoch": 1.49, + "learning_rate": 5.108214011059435e-08, + "loss": 0.6923, + "step": 40715 + }, + { + "epoch": 1.49, + "learning_rate": 5.107847804592229e-08, + "loss": 0.7066, + "step": 40716 + }, + { + "epoch": 1.49, + "learning_rate": 5.107481598125023e-08, + "loss": 0.6875, + "step": 40717 + }, + { + "epoch": 1.49, + "learning_rate": 5.107115391657817e-08, + "loss": 0.6905, + "step": 40718 + }, + { + "epoch": 1.49, + "learning_rate": 5.10674918519061e-08, + "loss": 0.7009, + "step": 40719 + }, + { + "epoch": 1.49, + "learning_rate": 5.106382978723404e-08, + "loss": 0.6879, + "step": 40720 + }, + { + "epoch": 1.49, + "learning_rate": 5.106016772256198e-08, + "loss": 0.6805, + "step": 40721 + }, + { + "epoch": 1.49, + "learning_rate": 5.105650565788992e-08, + "loss": 0.692, + "step": 40722 + }, + { + "epoch": 1.49, + "learning_rate": 5.105284359321785e-08, + "loss": 0.6801, + "step": 40723 + }, + { + "epoch": 1.49, + "learning_rate": 5.104918152854579e-08, + "loss": 0.7002, + "step": 40724 + }, + { + "epoch": 1.49, + "learning_rate": 5.104551946387373e-08, + "loss": 0.6928, + "step": 40725 + }, + { + "epoch": 1.49, + "learning_rate": 5.104185739920167e-08, + "loss": 0.7051, + "step": 40726 + }, + { + "epoch": 1.49, + "learning_rate": 5.103819533452961e-08, + "loss": 0.689, + "step": 40727 + }, + { + "epoch": 1.49, + "learning_rate": 5.103453326985754e-08, + "loss": 0.7145, + "step": 40728 + }, + { + "epoch": 1.49, + "learning_rate": 5.103087120518548e-08, + "loss": 0.6873, + "step": 40729 + }, + { + "epoch": 1.49, + "learning_rate": 5.102720914051342e-08, + "loss": 0.6964, + "step": 40730 + }, + { + "epoch": 1.49, + "learning_rate": 5.102354707584136e-08, + "loss": 0.6824, + "step": 40731 + }, + { + "epoch": 1.49, + "learning_rate": 5.10198850111693e-08, + "loss": 0.697, + "step": 40732 + }, + { + "epoch": 1.49, + "learning_rate": 5.1016222946497237e-08, + "loss": 0.6856, + "step": 40733 + }, + { + "epoch": 1.49, + "learning_rate": 5.101256088182517e-08, + "loss": 0.6961, + "step": 40734 + }, + { + "epoch": 1.49, + "learning_rate": 5.100889881715311e-08, + "loss": 0.697, + "step": 40735 + }, + { + "epoch": 1.49, + "learning_rate": 5.100523675248105e-08, + "loss": 0.6912, + "step": 40736 + }, + { + "epoch": 1.49, + "learning_rate": 5.1001574687808977e-08, + "loss": 0.6911, + "step": 40737 + }, + { + "epoch": 1.49, + "learning_rate": 5.0997912623136916e-08, + "loss": 0.6813, + "step": 40738 + }, + { + "epoch": 1.49, + "learning_rate": 5.0994250558464856e-08, + "loss": 0.7022, + "step": 40739 + }, + { + "epoch": 1.49, + "learning_rate": 5.0990588493792796e-08, + "loss": 0.6938, + "step": 40740 + }, + { + "epoch": 1.49, + "learning_rate": 5.0986926429120736e-08, + "loss": 0.6837, + "step": 40741 + }, + { + "epoch": 1.49, + "learning_rate": 5.0983264364448676e-08, + "loss": 0.6823, + "step": 40742 + }, + { + "epoch": 1.49, + "learning_rate": 5.0979602299776616e-08, + "loss": 0.6871, + "step": 40743 + }, + { + "epoch": 1.49, + "learning_rate": 5.0975940235104556e-08, + "loss": 0.6847, + "step": 40744 + }, + { + "epoch": 1.49, + "learning_rate": 5.097227817043249e-08, + "loss": 0.701, + "step": 40745 + }, + { + "epoch": 1.49, + "learning_rate": 5.096861610576042e-08, + "loss": 0.6858, + "step": 40746 + }, + { + "epoch": 1.49, + "learning_rate": 5.096495404108836e-08, + "loss": 0.6983, + "step": 40747 + }, + { + "epoch": 1.49, + "learning_rate": 5.0961291976416296e-08, + "loss": 0.6925, + "step": 40748 + }, + { + "epoch": 1.49, + "learning_rate": 5.0957629911744236e-08, + "loss": 0.6911, + "step": 40749 + }, + { + "epoch": 1.49, + "learning_rate": 5.0953967847072176e-08, + "loss": 0.6977, + "step": 40750 + }, + { + "epoch": 1.49, + "learning_rate": 5.0950305782400116e-08, + "loss": 0.6902, + "step": 40751 + }, + { + "epoch": 1.49, + "learning_rate": 5.0946643717728056e-08, + "loss": 0.6844, + "step": 40752 + }, + { + "epoch": 1.49, + "learning_rate": 5.0942981653055996e-08, + "loss": 0.6921, + "step": 40753 + }, + { + "epoch": 1.49, + "learning_rate": 5.0939319588383935e-08, + "loss": 0.6918, + "step": 40754 + }, + { + "epoch": 1.49, + "learning_rate": 5.093565752371186e-08, + "loss": 0.7008, + "step": 40755 + }, + { + "epoch": 1.49, + "learning_rate": 5.09319954590398e-08, + "loss": 0.6772, + "step": 40756 + }, + { + "epoch": 1.49, + "learning_rate": 5.092833339436774e-08, + "loss": 0.6891, + "step": 40757 + }, + { + "epoch": 1.49, + "learning_rate": 5.092467132969568e-08, + "loss": 0.6836, + "step": 40758 + }, + { + "epoch": 1.49, + "learning_rate": 5.0921009265023615e-08, + "loss": 0.6841, + "step": 40759 + }, + { + "epoch": 1.49, + "learning_rate": 5.0917347200351555e-08, + "loss": 0.6897, + "step": 40760 + }, + { + "epoch": 1.49, + "learning_rate": 5.0913685135679495e-08, + "loss": 0.6947, + "step": 40761 + }, + { + "epoch": 1.49, + "learning_rate": 5.0910023071007435e-08, + "loss": 0.7036, + "step": 40762 + }, + { + "epoch": 1.49, + "learning_rate": 5.0906361006335375e-08, + "loss": 0.6985, + "step": 40763 + }, + { + "epoch": 1.49, + "learning_rate": 5.09026989416633e-08, + "loss": 0.6933, + "step": 40764 + }, + { + "epoch": 1.49, + "learning_rate": 5.089903687699124e-08, + "loss": 0.6888, + "step": 40765 + }, + { + "epoch": 1.49, + "learning_rate": 5.089537481231918e-08, + "loss": 0.6974, + "step": 40766 + }, + { + "epoch": 1.49, + "learning_rate": 5.089171274764712e-08, + "loss": 0.6931, + "step": 40767 + }, + { + "epoch": 1.49, + "learning_rate": 5.088805068297506e-08, + "loss": 0.6823, + "step": 40768 + }, + { + "epoch": 1.49, + "learning_rate": 5.0884388618303e-08, + "loss": 0.688, + "step": 40769 + }, + { + "epoch": 1.49, + "learning_rate": 5.0880726553630935e-08, + "loss": 0.6881, + "step": 40770 + }, + { + "epoch": 1.49, + "learning_rate": 5.0877064488958875e-08, + "loss": 0.6924, + "step": 40771 + }, + { + "epoch": 1.49, + "learning_rate": 5.0873402424286814e-08, + "loss": 0.696, + "step": 40772 + }, + { + "epoch": 1.49, + "learning_rate": 5.086974035961474e-08, + "loss": 0.6962, + "step": 40773 + }, + { + "epoch": 1.49, + "learning_rate": 5.086607829494268e-08, + "loss": 0.6964, + "step": 40774 + }, + { + "epoch": 1.49, + "learning_rate": 5.086241623027062e-08, + "loss": 0.6974, + "step": 40775 + }, + { + "epoch": 1.49, + "learning_rate": 5.085875416559856e-08, + "loss": 0.6971, + "step": 40776 + }, + { + "epoch": 1.49, + "learning_rate": 5.08550921009265e-08, + "loss": 0.6774, + "step": 40777 + }, + { + "epoch": 1.49, + "learning_rate": 5.085143003625444e-08, + "loss": 0.6933, + "step": 40778 + }, + { + "epoch": 1.49, + "learning_rate": 5.084776797158238e-08, + "loss": 0.6906, + "step": 40779 + }, + { + "epoch": 1.49, + "learning_rate": 5.0844105906910314e-08, + "loss": 0.7024, + "step": 40780 + }, + { + "epoch": 1.49, + "learning_rate": 5.0840443842238254e-08, + "loss": 0.6826, + "step": 40781 + }, + { + "epoch": 1.49, + "learning_rate": 5.083678177756619e-08, + "loss": 0.7003, + "step": 40782 + }, + { + "epoch": 1.49, + "learning_rate": 5.083311971289412e-08, + "loss": 0.706, + "step": 40783 + }, + { + "epoch": 1.49, + "learning_rate": 5.082945764822206e-08, + "loss": 0.691, + "step": 40784 + }, + { + "epoch": 1.49, + "learning_rate": 5.082579558355e-08, + "loss": 0.7056, + "step": 40785 + }, + { + "epoch": 1.49, + "learning_rate": 5.082213351887794e-08, + "loss": 0.6966, + "step": 40786 + }, + { + "epoch": 1.49, + "learning_rate": 5.081847145420588e-08, + "loss": 0.6871, + "step": 40787 + }, + { + "epoch": 1.49, + "learning_rate": 5.081480938953382e-08, + "loss": 0.6843, + "step": 40788 + }, + { + "epoch": 1.49, + "learning_rate": 5.081114732486176e-08, + "loss": 0.6854, + "step": 40789 + }, + { + "epoch": 1.49, + "learning_rate": 5.08074852601897e-08, + "loss": 0.6882, + "step": 40790 + }, + { + "epoch": 1.49, + "learning_rate": 5.080382319551763e-08, + "loss": 0.6887, + "step": 40791 + }, + { + "epoch": 1.49, + "learning_rate": 5.080016113084557e-08, + "loss": 0.6915, + "step": 40792 + }, + { + "epoch": 1.49, + "learning_rate": 5.0796499066173507e-08, + "loss": 0.6878, + "step": 40793 + }, + { + "epoch": 1.49, + "learning_rate": 5.079283700150144e-08, + "loss": 0.6965, + "step": 40794 + }, + { + "epoch": 1.49, + "learning_rate": 5.078917493682938e-08, + "loss": 0.6992, + "step": 40795 + }, + { + "epoch": 1.49, + "learning_rate": 5.078551287215732e-08, + "loss": 0.6794, + "step": 40796 + }, + { + "epoch": 1.49, + "learning_rate": 5.078185080748526e-08, + "loss": 0.6996, + "step": 40797 + }, + { + "epoch": 1.49, + "learning_rate": 5.07781887428132e-08, + "loss": 0.692, + "step": 40798 + }, + { + "epoch": 1.49, + "learning_rate": 5.077452667814114e-08, + "loss": 0.7012, + "step": 40799 + }, + { + "epoch": 1.49, + "learning_rate": 5.0770864613469066e-08, + "loss": 0.6831, + "step": 40800 + }, + { + "epoch": 1.49, + "learning_rate": 5.0767202548797006e-08, + "loss": 0.6941, + "step": 40801 + }, + { + "epoch": 1.49, + "learning_rate": 5.0763540484124946e-08, + "loss": 0.6888, + "step": 40802 + }, + { + "epoch": 1.49, + "learning_rate": 5.0759878419452886e-08, + "loss": 0.6911, + "step": 40803 + }, + { + "epoch": 1.49, + "learning_rate": 5.0756216354780826e-08, + "loss": 0.6869, + "step": 40804 + }, + { + "epoch": 1.49, + "learning_rate": 5.075255429010876e-08, + "loss": 0.7075, + "step": 40805 + }, + { + "epoch": 1.49, + "learning_rate": 5.07488922254367e-08, + "loss": 0.6958, + "step": 40806 + }, + { + "epoch": 1.49, + "learning_rate": 5.074523016076464e-08, + "loss": 0.6963, + "step": 40807 + }, + { + "epoch": 1.49, + "learning_rate": 5.074156809609258e-08, + "loss": 0.6849, + "step": 40808 + }, + { + "epoch": 1.49, + "learning_rate": 5.0737906031420506e-08, + "loss": 0.6991, + "step": 40809 + }, + { + "epoch": 1.49, + "learning_rate": 5.0734243966748446e-08, + "loss": 0.6895, + "step": 40810 + }, + { + "epoch": 1.49, + "learning_rate": 5.0730581902076386e-08, + "loss": 0.689, + "step": 40811 + }, + { + "epoch": 1.49, + "learning_rate": 5.0726919837404326e-08, + "loss": 0.7045, + "step": 40812 + }, + { + "epoch": 1.49, + "learning_rate": 5.0723257772732265e-08, + "loss": 0.6727, + "step": 40813 + }, + { + "epoch": 1.49, + "learning_rate": 5.0719595708060205e-08, + "loss": 0.6927, + "step": 40814 + }, + { + "epoch": 1.49, + "learning_rate": 5.0715933643388145e-08, + "loss": 0.6968, + "step": 40815 + }, + { + "epoch": 1.49, + "learning_rate": 5.071227157871608e-08, + "loss": 0.689, + "step": 40816 + }, + { + "epoch": 1.49, + "learning_rate": 5.070860951404402e-08, + "loss": 0.6933, + "step": 40817 + }, + { + "epoch": 1.49, + "learning_rate": 5.070494744937195e-08, + "loss": 0.6917, + "step": 40818 + }, + { + "epoch": 1.49, + "learning_rate": 5.0701285384699885e-08, + "loss": 0.7023, + "step": 40819 + }, + { + "epoch": 1.49, + "learning_rate": 5.0697623320027825e-08, + "loss": 0.6848, + "step": 40820 + }, + { + "epoch": 1.49, + "learning_rate": 5.0693961255355765e-08, + "loss": 0.6874, + "step": 40821 + }, + { + "epoch": 1.49, + "learning_rate": 5.0690299190683705e-08, + "loss": 0.6816, + "step": 40822 + }, + { + "epoch": 1.49, + "learning_rate": 5.0686637126011645e-08, + "loss": 0.6991, + "step": 40823 + }, + { + "epoch": 1.49, + "learning_rate": 5.0682975061339585e-08, + "loss": 0.6989, + "step": 40824 + }, + { + "epoch": 1.49, + "learning_rate": 5.0679312996667525e-08, + "loss": 0.6945, + "step": 40825 + }, + { + "epoch": 1.49, + "learning_rate": 5.0675650931995465e-08, + "loss": 0.6982, + "step": 40826 + }, + { + "epoch": 1.49, + "learning_rate": 5.067198886732339e-08, + "loss": 0.7025, + "step": 40827 + }, + { + "epoch": 1.49, + "learning_rate": 5.066832680265133e-08, + "loss": 0.6947, + "step": 40828 + }, + { + "epoch": 1.49, + "learning_rate": 5.066466473797927e-08, + "loss": 0.6969, + "step": 40829 + }, + { + "epoch": 1.49, + "learning_rate": 5.0661002673307205e-08, + "loss": 0.6932, + "step": 40830 + }, + { + "epoch": 1.49, + "learning_rate": 5.0657340608635145e-08, + "loss": 0.688, + "step": 40831 + }, + { + "epoch": 1.49, + "learning_rate": 5.0653678543963084e-08, + "loss": 0.6968, + "step": 40832 + }, + { + "epoch": 1.49, + "learning_rate": 5.0650016479291024e-08, + "loss": 0.6878, + "step": 40833 + }, + { + "epoch": 1.49, + "learning_rate": 5.0646354414618964e-08, + "loss": 0.7009, + "step": 40834 + }, + { + "epoch": 1.49, + "learning_rate": 5.0642692349946904e-08, + "loss": 0.6833, + "step": 40835 + }, + { + "epoch": 1.49, + "learning_rate": 5.063903028527483e-08, + "loss": 0.6978, + "step": 40836 + }, + { + "epoch": 1.49, + "learning_rate": 5.063536822060277e-08, + "loss": 0.6938, + "step": 40837 + }, + { + "epoch": 1.49, + "learning_rate": 5.063170615593071e-08, + "loss": 0.6898, + "step": 40838 + }, + { + "epoch": 1.49, + "learning_rate": 5.062804409125865e-08, + "loss": 0.6937, + "step": 40839 + }, + { + "epoch": 1.49, + "learning_rate": 5.0624382026586584e-08, + "loss": 0.6925, + "step": 40840 + }, + { + "epoch": 1.49, + "learning_rate": 5.0620719961914524e-08, + "loss": 0.6993, + "step": 40841 + }, + { + "epoch": 1.49, + "learning_rate": 5.0617057897242464e-08, + "loss": 0.6996, + "step": 40842 + }, + { + "epoch": 1.49, + "learning_rate": 5.0613395832570404e-08, + "loss": 0.7026, + "step": 40843 + }, + { + "epoch": 1.49, + "learning_rate": 5.0609733767898344e-08, + "loss": 0.6989, + "step": 40844 + }, + { + "epoch": 1.49, + "learning_rate": 5.060607170322627e-08, + "loss": 0.6785, + "step": 40845 + }, + { + "epoch": 1.49, + "learning_rate": 5.060240963855421e-08, + "loss": 0.7038, + "step": 40846 + }, + { + "epoch": 1.49, + "learning_rate": 5.059874757388215e-08, + "loss": 0.6936, + "step": 40847 + }, + { + "epoch": 1.49, + "learning_rate": 5.059508550921009e-08, + "loss": 0.6992, + "step": 40848 + }, + { + "epoch": 1.49, + "learning_rate": 5.059142344453803e-08, + "loss": 0.6828, + "step": 40849 + }, + { + "epoch": 1.49, + "learning_rate": 5.058776137986597e-08, + "loss": 0.7065, + "step": 40850 + }, + { + "epoch": 1.49, + "learning_rate": 5.0584099315193903e-08, + "loss": 0.6952, + "step": 40851 + }, + { + "epoch": 1.49, + "learning_rate": 5.058043725052184e-08, + "loss": 0.6844, + "step": 40852 + }, + { + "epoch": 1.49, + "learning_rate": 5.057677518584978e-08, + "loss": 0.6745, + "step": 40853 + }, + { + "epoch": 1.49, + "learning_rate": 5.057311312117771e-08, + "loss": 0.7075, + "step": 40854 + }, + { + "epoch": 1.49, + "learning_rate": 5.056945105650565e-08, + "loss": 0.6944, + "step": 40855 + }, + { + "epoch": 1.49, + "learning_rate": 5.056578899183359e-08, + "loss": 0.686, + "step": 40856 + }, + { + "epoch": 1.49, + "learning_rate": 5.056212692716153e-08, + "loss": 0.6971, + "step": 40857 + }, + { + "epoch": 1.49, + "learning_rate": 5.055846486248947e-08, + "loss": 0.6792, + "step": 40858 + }, + { + "epoch": 1.49, + "learning_rate": 5.055480279781741e-08, + "loss": 0.7058, + "step": 40859 + }, + { + "epoch": 1.49, + "learning_rate": 5.055114073314535e-08, + "loss": 0.6966, + "step": 40860 + }, + { + "epoch": 1.49, + "learning_rate": 5.054747866847329e-08, + "loss": 0.6962, + "step": 40861 + }, + { + "epoch": 1.5, + "learning_rate": 5.054381660380122e-08, + "loss": 0.7033, + "step": 40862 + }, + { + "epoch": 1.5, + "learning_rate": 5.0540154539129156e-08, + "loss": 0.6909, + "step": 40863 + }, + { + "epoch": 1.5, + "learning_rate": 5.0536492474457096e-08, + "loss": 0.6961, + "step": 40864 + }, + { + "epoch": 1.5, + "learning_rate": 5.053283040978503e-08, + "loss": 0.6917, + "step": 40865 + }, + { + "epoch": 1.5, + "learning_rate": 5.052916834511297e-08, + "loss": 0.7073, + "step": 40866 + }, + { + "epoch": 1.5, + "learning_rate": 5.052550628044091e-08, + "loss": 0.705, + "step": 40867 + }, + { + "epoch": 1.5, + "learning_rate": 5.052184421576885e-08, + "loss": 0.7085, + "step": 40868 + }, + { + "epoch": 1.5, + "learning_rate": 5.051818215109679e-08, + "loss": 0.6891, + "step": 40869 + }, + { + "epoch": 1.5, + "learning_rate": 5.051452008642473e-08, + "loss": 0.6941, + "step": 40870 + }, + { + "epoch": 1.5, + "learning_rate": 5.051085802175267e-08, + "loss": 0.6805, + "step": 40871 + }, + { + "epoch": 1.5, + "learning_rate": 5.0507195957080596e-08, + "loss": 0.6878, + "step": 40872 + }, + { + "epoch": 1.5, + "learning_rate": 5.0503533892408535e-08, + "loss": 0.6811, + "step": 40873 + }, + { + "epoch": 1.5, + "learning_rate": 5.0499871827736475e-08, + "loss": 0.6851, + "step": 40874 + }, + { + "epoch": 1.5, + "learning_rate": 5.0496209763064415e-08, + "loss": 0.7052, + "step": 40875 + }, + { + "epoch": 1.5, + "learning_rate": 5.049254769839235e-08, + "loss": 0.6872, + "step": 40876 + }, + { + "epoch": 1.5, + "learning_rate": 5.048888563372029e-08, + "loss": 0.6911, + "step": 40877 + }, + { + "epoch": 1.5, + "learning_rate": 5.048522356904823e-08, + "loss": 0.6967, + "step": 40878 + }, + { + "epoch": 1.5, + "learning_rate": 5.048156150437617e-08, + "loss": 0.6735, + "step": 40879 + }, + { + "epoch": 1.5, + "learning_rate": 5.047789943970411e-08, + "loss": 0.6942, + "step": 40880 + }, + { + "epoch": 1.5, + "learning_rate": 5.0474237375032035e-08, + "loss": 0.6838, + "step": 40881 + }, + { + "epoch": 1.5, + "learning_rate": 5.0470575310359975e-08, + "loss": 0.6899, + "step": 40882 + }, + { + "epoch": 1.5, + "learning_rate": 5.0466913245687915e-08, + "loss": 0.6939, + "step": 40883 + }, + { + "epoch": 1.5, + "learning_rate": 5.0463251181015855e-08, + "loss": 0.6852, + "step": 40884 + }, + { + "epoch": 1.5, + "learning_rate": 5.0459589116343795e-08, + "loss": 0.6918, + "step": 40885 + }, + { + "epoch": 1.5, + "learning_rate": 5.0455927051671735e-08, + "loss": 0.6964, + "step": 40886 + }, + { + "epoch": 1.5, + "learning_rate": 5.045226498699967e-08, + "loss": 0.6973, + "step": 40887 + }, + { + "epoch": 1.5, + "learning_rate": 5.044860292232761e-08, + "loss": 0.688, + "step": 40888 + }, + { + "epoch": 1.5, + "learning_rate": 5.044494085765555e-08, + "loss": 0.6865, + "step": 40889 + }, + { + "epoch": 1.5, + "learning_rate": 5.0441278792983475e-08, + "loss": 0.7024, + "step": 40890 + }, + { + "epoch": 1.5, + "learning_rate": 5.0437616728311414e-08, + "loss": 0.6968, + "step": 40891 + }, + { + "epoch": 1.5, + "learning_rate": 5.0433954663639354e-08, + "loss": 0.692, + "step": 40892 + }, + { + "epoch": 1.5, + "learning_rate": 5.0430292598967294e-08, + "loss": 0.6949, + "step": 40893 + }, + { + "epoch": 1.5, + "learning_rate": 5.0426630534295234e-08, + "loss": 0.6884, + "step": 40894 + }, + { + "epoch": 1.5, + "learning_rate": 5.0422968469623174e-08, + "loss": 0.693, + "step": 40895 + }, + { + "epoch": 1.5, + "learning_rate": 5.0419306404951114e-08, + "loss": 0.7001, + "step": 40896 + }, + { + "epoch": 1.5, + "learning_rate": 5.0415644340279054e-08, + "loss": 0.6951, + "step": 40897 + }, + { + "epoch": 1.5, + "learning_rate": 5.041198227560699e-08, + "loss": 0.6895, + "step": 40898 + }, + { + "epoch": 1.5, + "learning_rate": 5.040832021093492e-08, + "loss": 0.696, + "step": 40899 + }, + { + "epoch": 1.5, + "learning_rate": 5.0404658146262854e-08, + "loss": 0.6848, + "step": 40900 + }, + { + "epoch": 1.5, + "learning_rate": 5.0400996081590794e-08, + "loss": 0.6977, + "step": 40901 + }, + { + "epoch": 1.5, + "learning_rate": 5.0397334016918734e-08, + "loss": 0.6998, + "step": 40902 + }, + { + "epoch": 1.5, + "learning_rate": 5.0393671952246674e-08, + "loss": 0.7182, + "step": 40903 + }, + { + "epoch": 1.5, + "learning_rate": 5.0390009887574614e-08, + "loss": 0.6917, + "step": 40904 + }, + { + "epoch": 1.5, + "learning_rate": 5.0386347822902554e-08, + "loss": 0.7018, + "step": 40905 + }, + { + "epoch": 1.5, + "learning_rate": 5.0382685758230494e-08, + "loss": 0.6974, + "step": 40906 + }, + { + "epoch": 1.5, + "learning_rate": 5.0379023693558434e-08, + "loss": 0.6916, + "step": 40907 + }, + { + "epoch": 1.5, + "learning_rate": 5.037536162888636e-08, + "loss": 0.7073, + "step": 40908 + }, + { + "epoch": 1.5, + "learning_rate": 5.03716995642143e-08, + "loss": 0.7116, + "step": 40909 + }, + { + "epoch": 1.5, + "learning_rate": 5.036803749954224e-08, + "loss": 0.6918, + "step": 40910 + }, + { + "epoch": 1.5, + "learning_rate": 5.0364375434870173e-08, + "loss": 0.6938, + "step": 40911 + }, + { + "epoch": 1.5, + "learning_rate": 5.036071337019811e-08, + "loss": 0.6988, + "step": 40912 + }, + { + "epoch": 1.5, + "learning_rate": 5.035705130552605e-08, + "loss": 0.6953, + "step": 40913 + }, + { + "epoch": 1.5, + "learning_rate": 5.035338924085399e-08, + "loss": 0.695, + "step": 40914 + }, + { + "epoch": 1.5, + "learning_rate": 5.034972717618193e-08, + "loss": 0.6858, + "step": 40915 + }, + { + "epoch": 1.5, + "learning_rate": 5.034606511150987e-08, + "loss": 0.6902, + "step": 40916 + }, + { + "epoch": 1.5, + "learning_rate": 5.03424030468378e-08, + "loss": 0.6984, + "step": 40917 + }, + { + "epoch": 1.5, + "learning_rate": 5.033874098216574e-08, + "loss": 0.6928, + "step": 40918 + }, + { + "epoch": 1.5, + "learning_rate": 5.033507891749368e-08, + "loss": 0.7055, + "step": 40919 + }, + { + "epoch": 1.5, + "learning_rate": 5.033141685282162e-08, + "loss": 0.6959, + "step": 40920 + }, + { + "epoch": 1.5, + "learning_rate": 5.032775478814956e-08, + "loss": 0.6865, + "step": 40921 + }, + { + "epoch": 1.5, + "learning_rate": 5.032409272347749e-08, + "loss": 0.6862, + "step": 40922 + }, + { + "epoch": 1.5, + "learning_rate": 5.032043065880543e-08, + "loss": 0.693, + "step": 40923 + }, + { + "epoch": 1.5, + "learning_rate": 5.031676859413337e-08, + "loss": 0.6892, + "step": 40924 + }, + { + "epoch": 1.5, + "learning_rate": 5.031310652946131e-08, + "loss": 0.7074, + "step": 40925 + }, + { + "epoch": 1.5, + "learning_rate": 5.030944446478924e-08, + "loss": 0.6891, + "step": 40926 + }, + { + "epoch": 1.5, + "learning_rate": 5.030578240011718e-08, + "loss": 0.7076, + "step": 40927 + }, + { + "epoch": 1.5, + "learning_rate": 5.030212033544512e-08, + "loss": 0.6843, + "step": 40928 + }, + { + "epoch": 1.5, + "learning_rate": 5.029845827077306e-08, + "loss": 0.7063, + "step": 40929 + }, + { + "epoch": 1.5, + "learning_rate": 5.0294796206101e-08, + "loss": 0.6874, + "step": 40930 + }, + { + "epoch": 1.5, + "learning_rate": 5.029113414142894e-08, + "loss": 0.6883, + "step": 40931 + }, + { + "epoch": 1.5, + "learning_rate": 5.028747207675688e-08, + "loss": 0.6997, + "step": 40932 + }, + { + "epoch": 1.5, + "learning_rate": 5.028381001208481e-08, + "loss": 0.6971, + "step": 40933 + }, + { + "epoch": 1.5, + "learning_rate": 5.028014794741275e-08, + "loss": 0.7053, + "step": 40934 + }, + { + "epoch": 1.5, + "learning_rate": 5.0276485882740685e-08, + "loss": 0.6939, + "step": 40935 + }, + { + "epoch": 1.5, + "learning_rate": 5.027282381806862e-08, + "loss": 0.6974, + "step": 40936 + }, + { + "epoch": 1.5, + "learning_rate": 5.026916175339656e-08, + "loss": 0.6961, + "step": 40937 + }, + { + "epoch": 1.5, + "learning_rate": 5.02654996887245e-08, + "loss": 0.7049, + "step": 40938 + }, + { + "epoch": 1.5, + "learning_rate": 5.026183762405244e-08, + "loss": 0.6932, + "step": 40939 + }, + { + "epoch": 1.5, + "learning_rate": 5.025817555938038e-08, + "loss": 0.6919, + "step": 40940 + }, + { + "epoch": 1.5, + "learning_rate": 5.025451349470832e-08, + "loss": 0.6891, + "step": 40941 + }, + { + "epoch": 1.5, + "learning_rate": 5.025085143003626e-08, + "loss": 0.6934, + "step": 40942 + }, + { + "epoch": 1.5, + "learning_rate": 5.02471893653642e-08, + "loss": 0.6963, + "step": 40943 + }, + { + "epoch": 1.5, + "learning_rate": 5.0243527300692125e-08, + "loss": 0.6947, + "step": 40944 + }, + { + "epoch": 1.5, + "learning_rate": 5.0239865236020065e-08, + "loss": 0.691, + "step": 40945 + }, + { + "epoch": 1.5, + "learning_rate": 5.0236203171348005e-08, + "loss": 0.7041, + "step": 40946 + }, + { + "epoch": 1.5, + "learning_rate": 5.023254110667594e-08, + "loss": 0.6989, + "step": 40947 + }, + { + "epoch": 1.5, + "learning_rate": 5.022887904200388e-08, + "loss": 0.6944, + "step": 40948 + }, + { + "epoch": 1.5, + "learning_rate": 5.022521697733182e-08, + "loss": 0.696, + "step": 40949 + }, + { + "epoch": 1.5, + "learning_rate": 5.022155491265976e-08, + "loss": 0.6983, + "step": 40950 + }, + { + "epoch": 1.5, + "learning_rate": 5.02178928479877e-08, + "loss": 0.6979, + "step": 40951 + }, + { + "epoch": 1.5, + "learning_rate": 5.021423078331564e-08, + "loss": 0.6925, + "step": 40952 + }, + { + "epoch": 1.5, + "learning_rate": 5.0210568718643564e-08, + "loss": 0.6862, + "step": 40953 + }, + { + "epoch": 1.5, + "learning_rate": 5.0206906653971504e-08, + "loss": 0.6783, + "step": 40954 + }, + { + "epoch": 1.5, + "learning_rate": 5.0203244589299444e-08, + "loss": 0.6952, + "step": 40955 + }, + { + "epoch": 1.5, + "learning_rate": 5.0199582524627384e-08, + "loss": 0.7072, + "step": 40956 + }, + { + "epoch": 1.5, + "learning_rate": 5.0195920459955324e-08, + "loss": 0.6926, + "step": 40957 + }, + { + "epoch": 1.5, + "learning_rate": 5.019225839528326e-08, + "loss": 0.7069, + "step": 40958 + }, + { + "epoch": 1.5, + "learning_rate": 5.01885963306112e-08, + "loss": 0.684, + "step": 40959 + }, + { + "epoch": 1.5, + "learning_rate": 5.018493426593914e-08, + "loss": 0.6866, + "step": 40960 + }, + { + "epoch": 1.5, + "learning_rate": 5.018127220126708e-08, + "loss": 0.6803, + "step": 40961 + }, + { + "epoch": 1.5, + "learning_rate": 5.0177610136595004e-08, + "loss": 0.6943, + "step": 40962 + }, + { + "epoch": 1.5, + "learning_rate": 5.0173948071922944e-08, + "loss": 0.6942, + "step": 40963 + }, + { + "epoch": 1.5, + "learning_rate": 5.0170286007250884e-08, + "loss": 0.6974, + "step": 40964 + }, + { + "epoch": 1.5, + "learning_rate": 5.0166623942578824e-08, + "loss": 0.6749, + "step": 40965 + }, + { + "epoch": 1.5, + "learning_rate": 5.0162961877906764e-08, + "loss": 0.6977, + "step": 40966 + }, + { + "epoch": 1.5, + "learning_rate": 5.0159299813234703e-08, + "loss": 0.6745, + "step": 40967 + }, + { + "epoch": 1.5, + "learning_rate": 5.0155637748562643e-08, + "loss": 0.6738, + "step": 40968 + }, + { + "epoch": 1.5, + "learning_rate": 5.015197568389058e-08, + "loss": 0.6911, + "step": 40969 + }, + { + "epoch": 1.5, + "learning_rate": 5.0148313619218517e-08, + "loss": 0.6941, + "step": 40970 + }, + { + "epoch": 1.5, + "learning_rate": 5.0144651554546443e-08, + "loss": 0.6964, + "step": 40971 + }, + { + "epoch": 1.5, + "learning_rate": 5.014098948987438e-08, + "loss": 0.6863, + "step": 40972 + }, + { + "epoch": 1.5, + "learning_rate": 5.013732742520232e-08, + "loss": 0.6942, + "step": 40973 + }, + { + "epoch": 1.5, + "learning_rate": 5.013366536053026e-08, + "loss": 0.6903, + "step": 40974 + }, + { + "epoch": 1.5, + "learning_rate": 5.01300032958582e-08, + "loss": 0.6912, + "step": 40975 + }, + { + "epoch": 1.5, + "learning_rate": 5.012634123118614e-08, + "loss": 0.6842, + "step": 40976 + }, + { + "epoch": 1.5, + "learning_rate": 5.012267916651408e-08, + "loss": 0.6795, + "step": 40977 + }, + { + "epoch": 1.5, + "learning_rate": 5.011901710184202e-08, + "loss": 0.6943, + "step": 40978 + }, + { + "epoch": 1.5, + "learning_rate": 5.011535503716996e-08, + "loss": 0.6827, + "step": 40979 + }, + { + "epoch": 1.5, + "learning_rate": 5.011169297249789e-08, + "loss": 0.683, + "step": 40980 + }, + { + "epoch": 1.5, + "learning_rate": 5.010803090782583e-08, + "loss": 0.7004, + "step": 40981 + }, + { + "epoch": 1.5, + "learning_rate": 5.010436884315376e-08, + "loss": 0.6973, + "step": 40982 + }, + { + "epoch": 1.5, + "learning_rate": 5.01007067784817e-08, + "loss": 0.692, + "step": 40983 + }, + { + "epoch": 1.5, + "learning_rate": 5.009704471380964e-08, + "loss": 0.6889, + "step": 40984 + }, + { + "epoch": 1.5, + "learning_rate": 5.009338264913758e-08, + "loss": 0.7005, + "step": 40985 + }, + { + "epoch": 1.5, + "learning_rate": 5.008972058446552e-08, + "loss": 0.6882, + "step": 40986 + }, + { + "epoch": 1.5, + "learning_rate": 5.008605851979346e-08, + "loss": 0.7015, + "step": 40987 + }, + { + "epoch": 1.5, + "learning_rate": 5.00823964551214e-08, + "loss": 0.6944, + "step": 40988 + }, + { + "epoch": 1.5, + "learning_rate": 5.007873439044933e-08, + "loss": 0.6738, + "step": 40989 + }, + { + "epoch": 1.5, + "learning_rate": 5.007507232577727e-08, + "loss": 0.7009, + "step": 40990 + }, + { + "epoch": 1.5, + "learning_rate": 5.007141026110521e-08, + "loss": 0.693, + "step": 40991 + }, + { + "epoch": 1.5, + "learning_rate": 5.006774819643315e-08, + "loss": 0.6961, + "step": 40992 + }, + { + "epoch": 1.5, + "learning_rate": 5.006408613176108e-08, + "loss": 0.6968, + "step": 40993 + }, + { + "epoch": 1.5, + "learning_rate": 5.006042406708902e-08, + "loss": 0.7006, + "step": 40994 + }, + { + "epoch": 1.5, + "learning_rate": 5.005676200241696e-08, + "loss": 0.6882, + "step": 40995 + }, + { + "epoch": 1.5, + "learning_rate": 5.00530999377449e-08, + "loss": 0.7034, + "step": 40996 + }, + { + "epoch": 1.5, + "learning_rate": 5.004943787307284e-08, + "loss": 0.697, + "step": 40997 + }, + { + "epoch": 1.5, + "learning_rate": 5.004577580840077e-08, + "loss": 0.6953, + "step": 40998 + }, + { + "epoch": 1.5, + "learning_rate": 5.004211374372871e-08, + "loss": 0.6947, + "step": 40999 + }, + { + "epoch": 1.5, + "learning_rate": 5.003845167905665e-08, + "loss": 0.6964, + "step": 41000 + }, + { + "epoch": 1.5, + "learning_rate": 5.003478961438459e-08, + "loss": 0.6937, + "step": 41001 + }, + { + "epoch": 1.5, + "learning_rate": 5.003112754971253e-08, + "loss": 0.6886, + "step": 41002 + }, + { + "epoch": 1.5, + "learning_rate": 5.002746548504047e-08, + "loss": 0.7073, + "step": 41003 + }, + { + "epoch": 1.5, + "learning_rate": 5.00238034203684e-08, + "loss": 0.6944, + "step": 41004 + }, + { + "epoch": 1.5, + "learning_rate": 5.002014135569634e-08, + "loss": 0.6901, + "step": 41005 + }, + { + "epoch": 1.5, + "learning_rate": 5.001647929102428e-08, + "loss": 0.6981, + "step": 41006 + }, + { + "epoch": 1.5, + "learning_rate": 5.001281722635221e-08, + "loss": 0.6936, + "step": 41007 + }, + { + "epoch": 1.5, + "learning_rate": 5.000915516168015e-08, + "loss": 0.6831, + "step": 41008 + }, + { + "epoch": 1.5, + "learning_rate": 5.000549309700809e-08, + "loss": 0.6965, + "step": 41009 + }, + { + "epoch": 1.5, + "learning_rate": 5.000183103233603e-08, + "loss": 0.6941, + "step": 41010 + }, + { + "epoch": 1.5, + "learning_rate": 4.999816896766397e-08, + "loss": 0.6961, + "step": 41011 + }, + { + "epoch": 1.5, + "learning_rate": 4.999450690299191e-08, + "loss": 0.7048, + "step": 41012 + }, + { + "epoch": 1.5, + "learning_rate": 4.999084483831984e-08, + "loss": 0.6927, + "step": 41013 + }, + { + "epoch": 1.5, + "learning_rate": 4.998718277364778e-08, + "loss": 0.6892, + "step": 41014 + }, + { + "epoch": 1.5, + "learning_rate": 4.998352070897572e-08, + "loss": 0.7053, + "step": 41015 + }, + { + "epoch": 1.5, + "learning_rate": 4.9979858644303654e-08, + "loss": 0.7034, + "step": 41016 + }, + { + "epoch": 1.5, + "learning_rate": 4.9976196579631594e-08, + "loss": 0.7051, + "step": 41017 + }, + { + "epoch": 1.5, + "learning_rate": 4.997253451495953e-08, + "loss": 0.6935, + "step": 41018 + }, + { + "epoch": 1.5, + "learning_rate": 4.996887245028747e-08, + "loss": 0.6992, + "step": 41019 + }, + { + "epoch": 1.5, + "learning_rate": 4.996521038561541e-08, + "loss": 0.6906, + "step": 41020 + }, + { + "epoch": 1.5, + "learning_rate": 4.996154832094335e-08, + "loss": 0.6966, + "step": 41021 + }, + { + "epoch": 1.5, + "learning_rate": 4.995788625627128e-08, + "loss": 0.7069, + "step": 41022 + }, + { + "epoch": 1.5, + "learning_rate": 4.995422419159922e-08, + "loss": 0.6783, + "step": 41023 + }, + { + "epoch": 1.5, + "learning_rate": 4.995056212692716e-08, + "loss": 0.7043, + "step": 41024 + }, + { + "epoch": 1.5, + "learning_rate": 4.99469000622551e-08, + "loss": 0.6879, + "step": 41025 + }, + { + "epoch": 1.5, + "learning_rate": 4.994323799758304e-08, + "loss": 0.7022, + "step": 41026 + }, + { + "epoch": 1.5, + "learning_rate": 4.9939575932910973e-08, + "loss": 0.6882, + "step": 41027 + }, + { + "epoch": 1.5, + "learning_rate": 4.9935913868238913e-08, + "loss": 0.6881, + "step": 41028 + }, + { + "epoch": 1.5, + "learning_rate": 4.993225180356685e-08, + "loss": 0.6866, + "step": 41029 + }, + { + "epoch": 1.5, + "learning_rate": 4.9928589738894787e-08, + "loss": 0.7059, + "step": 41030 + }, + { + "epoch": 1.5, + "learning_rate": 4.992492767422272e-08, + "loss": 0.68, + "step": 41031 + }, + { + "epoch": 1.5, + "learning_rate": 4.992126560955066e-08, + "loss": 0.6866, + "step": 41032 + }, + { + "epoch": 1.5, + "learning_rate": 4.99176035448786e-08, + "loss": 0.6936, + "step": 41033 + }, + { + "epoch": 1.5, + "learning_rate": 4.991394148020654e-08, + "loss": 0.6924, + "step": 41034 + }, + { + "epoch": 1.5, + "learning_rate": 4.991027941553448e-08, + "loss": 0.6869, + "step": 41035 + }, + { + "epoch": 1.5, + "learning_rate": 4.990661735086241e-08, + "loss": 0.6949, + "step": 41036 + }, + { + "epoch": 1.5, + "learning_rate": 4.990295528619035e-08, + "loss": 0.7022, + "step": 41037 + }, + { + "epoch": 1.5, + "learning_rate": 4.989929322151829e-08, + "loss": 0.6919, + "step": 41038 + }, + { + "epoch": 1.5, + "learning_rate": 4.989563115684623e-08, + "loss": 0.6979, + "step": 41039 + }, + { + "epoch": 1.5, + "learning_rate": 4.9891969092174166e-08, + "loss": 0.7062, + "step": 41040 + }, + { + "epoch": 1.5, + "learning_rate": 4.98883070275021e-08, + "loss": 0.6949, + "step": 41041 + }, + { + "epoch": 1.5, + "learning_rate": 4.988464496283004e-08, + "loss": 0.6929, + "step": 41042 + }, + { + "epoch": 1.5, + "learning_rate": 4.988098289815798e-08, + "loss": 0.7009, + "step": 41043 + }, + { + "epoch": 1.5, + "learning_rate": 4.987732083348592e-08, + "loss": 0.6895, + "step": 41044 + }, + { + "epoch": 1.5, + "learning_rate": 4.987365876881385e-08, + "loss": 0.6942, + "step": 41045 + }, + { + "epoch": 1.5, + "learning_rate": 4.986999670414179e-08, + "loss": 0.7048, + "step": 41046 + }, + { + "epoch": 1.5, + "learning_rate": 4.986633463946973e-08, + "loss": 0.7066, + "step": 41047 + }, + { + "epoch": 1.5, + "learning_rate": 4.986267257479767e-08, + "loss": 0.6866, + "step": 41048 + }, + { + "epoch": 1.5, + "learning_rate": 4.9859010510125606e-08, + "loss": 0.6889, + "step": 41049 + }, + { + "epoch": 1.5, + "learning_rate": 4.9855348445453546e-08, + "loss": 0.6895, + "step": 41050 + }, + { + "epoch": 1.5, + "learning_rate": 4.9851686380781485e-08, + "loss": 0.6954, + "step": 41051 + }, + { + "epoch": 1.5, + "learning_rate": 4.984802431610942e-08, + "loss": 0.6871, + "step": 41052 + }, + { + "epoch": 1.5, + "learning_rate": 4.984436225143736e-08, + "loss": 0.701, + "step": 41053 + }, + { + "epoch": 1.5, + "learning_rate": 4.984070018676529e-08, + "loss": 0.6854, + "step": 41054 + }, + { + "epoch": 1.5, + "learning_rate": 4.983703812209323e-08, + "loss": 0.7018, + "step": 41055 + }, + { + "epoch": 1.5, + "learning_rate": 4.983337605742117e-08, + "loss": 0.6913, + "step": 41056 + }, + { + "epoch": 1.5, + "learning_rate": 4.982971399274911e-08, + "loss": 0.6924, + "step": 41057 + }, + { + "epoch": 1.5, + "learning_rate": 4.9826051928077045e-08, + "loss": 0.6942, + "step": 41058 + }, + { + "epoch": 1.5, + "learning_rate": 4.9822389863404985e-08, + "loss": 0.6961, + "step": 41059 + }, + { + "epoch": 1.5, + "learning_rate": 4.9818727798732925e-08, + "loss": 0.6957, + "step": 41060 + }, + { + "epoch": 1.5, + "learning_rate": 4.9815065734060865e-08, + "loss": 0.6987, + "step": 41061 + }, + { + "epoch": 1.5, + "learning_rate": 4.9811403669388805e-08, + "loss": 0.6875, + "step": 41062 + }, + { + "epoch": 1.5, + "learning_rate": 4.980774160471674e-08, + "loss": 0.6952, + "step": 41063 + }, + { + "epoch": 1.5, + "learning_rate": 4.980407954004467e-08, + "loss": 0.6861, + "step": 41064 + }, + { + "epoch": 1.5, + "learning_rate": 4.980041747537261e-08, + "loss": 0.6875, + "step": 41065 + }, + { + "epoch": 1.5, + "learning_rate": 4.979675541070055e-08, + "loss": 0.7167, + "step": 41066 + }, + { + "epoch": 1.5, + "learning_rate": 4.9793093346028485e-08, + "loss": 0.6987, + "step": 41067 + }, + { + "epoch": 1.5, + "learning_rate": 4.9789431281356425e-08, + "loss": 0.698, + "step": 41068 + }, + { + "epoch": 1.5, + "learning_rate": 4.9785769216684364e-08, + "loss": 0.7016, + "step": 41069 + }, + { + "epoch": 1.5, + "learning_rate": 4.9782107152012304e-08, + "loss": 0.6936, + "step": 41070 + }, + { + "epoch": 1.5, + "learning_rate": 4.9778445087340244e-08, + "loss": 0.6922, + "step": 41071 + }, + { + "epoch": 1.5, + "learning_rate": 4.977478302266818e-08, + "loss": 0.7094, + "step": 41072 + }, + { + "epoch": 1.5, + "learning_rate": 4.977112095799612e-08, + "loss": 0.6968, + "step": 41073 + }, + { + "epoch": 1.5, + "learning_rate": 4.976745889332406e-08, + "loss": 0.6892, + "step": 41074 + }, + { + "epoch": 1.5, + "learning_rate": 4.976379682865199e-08, + "loss": 0.6949, + "step": 41075 + }, + { + "epoch": 1.5, + "learning_rate": 4.9760134763979924e-08, + "loss": 0.689, + "step": 41076 + }, + { + "epoch": 1.5, + "learning_rate": 4.9756472699307864e-08, + "loss": 0.7134, + "step": 41077 + }, + { + "epoch": 1.5, + "learning_rate": 4.9752810634635804e-08, + "loss": 0.6926, + "step": 41078 + }, + { + "epoch": 1.5, + "learning_rate": 4.9749148569963744e-08, + "loss": 0.6914, + "step": 41079 + }, + { + "epoch": 1.5, + "learning_rate": 4.9745486505291684e-08, + "loss": 0.7022, + "step": 41080 + }, + { + "epoch": 1.5, + "learning_rate": 4.974182444061962e-08, + "loss": 0.7067, + "step": 41081 + }, + { + "epoch": 1.5, + "learning_rate": 4.973816237594756e-08, + "loss": 0.6855, + "step": 41082 + }, + { + "epoch": 1.5, + "learning_rate": 4.97345003112755e-08, + "loss": 0.6951, + "step": 41083 + }, + { + "epoch": 1.5, + "learning_rate": 4.973083824660344e-08, + "loss": 0.6851, + "step": 41084 + }, + { + "epoch": 1.5, + "learning_rate": 4.972717618193137e-08, + "loss": 0.6883, + "step": 41085 + }, + { + "epoch": 1.5, + "learning_rate": 4.972351411725931e-08, + "loss": 0.6942, + "step": 41086 + }, + { + "epoch": 1.5, + "learning_rate": 4.9719852052587243e-08, + "loss": 0.6901, + "step": 41087 + }, + { + "epoch": 1.5, + "learning_rate": 4.9716189987915183e-08, + "loss": 0.6979, + "step": 41088 + }, + { + "epoch": 1.5, + "learning_rate": 4.9712527923243123e-08, + "loss": 0.6955, + "step": 41089 + }, + { + "epoch": 1.5, + "learning_rate": 4.9708865858571057e-08, + "loss": 0.6891, + "step": 41090 + }, + { + "epoch": 1.5, + "learning_rate": 4.9705203793898997e-08, + "loss": 0.691, + "step": 41091 + }, + { + "epoch": 1.5, + "learning_rate": 4.9701541729226936e-08, + "loss": 0.6818, + "step": 41092 + }, + { + "epoch": 1.5, + "learning_rate": 4.9697879664554876e-08, + "loss": 0.7006, + "step": 41093 + }, + { + "epoch": 1.5, + "learning_rate": 4.969421759988281e-08, + "loss": 0.6907, + "step": 41094 + }, + { + "epoch": 1.5, + "learning_rate": 4.969055553521075e-08, + "loss": 0.6912, + "step": 41095 + }, + { + "epoch": 1.5, + "learning_rate": 4.968689347053869e-08, + "loss": 0.6939, + "step": 41096 + }, + { + "epoch": 1.5, + "learning_rate": 4.968323140586663e-08, + "loss": 0.7028, + "step": 41097 + }, + { + "epoch": 1.5, + "learning_rate": 4.967956934119456e-08, + "loss": 0.6926, + "step": 41098 + }, + { + "epoch": 1.5, + "learning_rate": 4.96759072765225e-08, + "loss": 0.6878, + "step": 41099 + }, + { + "epoch": 1.5, + "learning_rate": 4.9672245211850436e-08, + "loss": 0.6899, + "step": 41100 + }, + { + "epoch": 1.5, + "learning_rate": 4.9668583147178376e-08, + "loss": 0.681, + "step": 41101 + }, + { + "epoch": 1.5, + "learning_rate": 4.9664921082506316e-08, + "loss": 0.6768, + "step": 41102 + }, + { + "epoch": 1.5, + "learning_rate": 4.966125901783425e-08, + "loss": 0.6896, + "step": 41103 + }, + { + "epoch": 1.5, + "learning_rate": 4.965759695316219e-08, + "loss": 0.6958, + "step": 41104 + }, + { + "epoch": 1.5, + "learning_rate": 4.965393488849013e-08, + "loss": 0.6827, + "step": 41105 + }, + { + "epoch": 1.5, + "learning_rate": 4.965027282381807e-08, + "loss": 0.6921, + "step": 41106 + }, + { + "epoch": 1.5, + "learning_rate": 4.9646610759146e-08, + "loss": 0.6866, + "step": 41107 + }, + { + "epoch": 1.5, + "learning_rate": 4.964294869447394e-08, + "loss": 0.6893, + "step": 41108 + }, + { + "epoch": 1.5, + "learning_rate": 4.963928662980188e-08, + "loss": 0.6908, + "step": 41109 + }, + { + "epoch": 1.5, + "learning_rate": 4.963562456512982e-08, + "loss": 0.6918, + "step": 41110 + }, + { + "epoch": 1.5, + "learning_rate": 4.9631962500457755e-08, + "loss": 0.6858, + "step": 41111 + }, + { + "epoch": 1.5, + "learning_rate": 4.962830043578569e-08, + "loss": 0.6858, + "step": 41112 + }, + { + "epoch": 1.5, + "learning_rate": 4.962463837111363e-08, + "loss": 0.6745, + "step": 41113 + }, + { + "epoch": 1.5, + "learning_rate": 4.962097630644157e-08, + "loss": 0.6954, + "step": 41114 + }, + { + "epoch": 1.5, + "learning_rate": 4.961731424176951e-08, + "loss": 0.696, + "step": 41115 + }, + { + "epoch": 1.5, + "learning_rate": 4.961365217709744e-08, + "loss": 0.6972, + "step": 41116 + }, + { + "epoch": 1.5, + "learning_rate": 4.960999011242538e-08, + "loss": 0.684, + "step": 41117 + }, + { + "epoch": 1.5, + "learning_rate": 4.960632804775332e-08, + "loss": 0.7033, + "step": 41118 + }, + { + "epoch": 1.5, + "learning_rate": 4.960266598308126e-08, + "loss": 0.6927, + "step": 41119 + }, + { + "epoch": 1.5, + "learning_rate": 4.95990039184092e-08, + "loss": 0.7052, + "step": 41120 + }, + { + "epoch": 1.5, + "learning_rate": 4.9595341853737135e-08, + "loss": 0.6954, + "step": 41121 + }, + { + "epoch": 1.5, + "learning_rate": 4.9591679789065075e-08, + "loss": 0.6972, + "step": 41122 + }, + { + "epoch": 1.5, + "learning_rate": 4.958801772439301e-08, + "loss": 0.6882, + "step": 41123 + }, + { + "epoch": 1.5, + "learning_rate": 4.958435565972095e-08, + "loss": 0.6841, + "step": 41124 + }, + { + "epoch": 1.5, + "learning_rate": 4.958069359504888e-08, + "loss": 0.6923, + "step": 41125 + }, + { + "epoch": 1.5, + "learning_rate": 4.957703153037682e-08, + "loss": 0.6863, + "step": 41126 + }, + { + "epoch": 1.5, + "learning_rate": 4.957336946570476e-08, + "loss": 0.6926, + "step": 41127 + }, + { + "epoch": 1.5, + "learning_rate": 4.95697074010327e-08, + "loss": 0.7019, + "step": 41128 + }, + { + "epoch": 1.5, + "learning_rate": 4.956604533636064e-08, + "loss": 0.6955, + "step": 41129 + }, + { + "epoch": 1.5, + "learning_rate": 4.9562383271688574e-08, + "loss": 0.6954, + "step": 41130 + }, + { + "epoch": 1.5, + "learning_rate": 4.9558721207016514e-08, + "loss": 0.7037, + "step": 41131 + }, + { + "epoch": 1.5, + "learning_rate": 4.9555059142344454e-08, + "loss": 0.6978, + "step": 41132 + }, + { + "epoch": 1.5, + "learning_rate": 4.9551397077672394e-08, + "loss": 0.6941, + "step": 41133 + }, + { + "epoch": 1.5, + "learning_rate": 4.954773501300033e-08, + "loss": 0.6843, + "step": 41134 + }, + { + "epoch": 1.51, + "learning_rate": 4.954407294832826e-08, + "loss": 0.6884, + "step": 41135 + }, + { + "epoch": 1.51, + "learning_rate": 4.95404108836562e-08, + "loss": 0.689, + "step": 41136 + }, + { + "epoch": 1.51, + "learning_rate": 4.953674881898414e-08, + "loss": 0.6997, + "step": 41137 + }, + { + "epoch": 1.51, + "learning_rate": 4.953308675431208e-08, + "loss": 0.6972, + "step": 41138 + }, + { + "epoch": 1.51, + "learning_rate": 4.9529424689640014e-08, + "loss": 0.6866, + "step": 41139 + }, + { + "epoch": 1.51, + "learning_rate": 4.9525762624967954e-08, + "loss": 0.6947, + "step": 41140 + }, + { + "epoch": 1.51, + "learning_rate": 4.9522100560295894e-08, + "loss": 0.6895, + "step": 41141 + }, + { + "epoch": 1.51, + "learning_rate": 4.9518438495623834e-08, + "loss": 0.6871, + "step": 41142 + }, + { + "epoch": 1.51, + "learning_rate": 4.951477643095177e-08, + "loss": 0.7022, + "step": 41143 + }, + { + "epoch": 1.51, + "learning_rate": 4.951111436627971e-08, + "loss": 0.6944, + "step": 41144 + }, + { + "epoch": 1.51, + "learning_rate": 4.950745230160765e-08, + "loss": 0.6905, + "step": 41145 + }, + { + "epoch": 1.51, + "learning_rate": 4.950379023693558e-08, + "loss": 0.6891, + "step": 41146 + }, + { + "epoch": 1.51, + "learning_rate": 4.950012817226352e-08, + "loss": 0.6811, + "step": 41147 + }, + { + "epoch": 1.51, + "learning_rate": 4.9496466107591453e-08, + "loss": 0.7057, + "step": 41148 + }, + { + "epoch": 1.51, + "learning_rate": 4.949280404291939e-08, + "loss": 0.6863, + "step": 41149 + }, + { + "epoch": 1.51, + "learning_rate": 4.948914197824733e-08, + "loss": 0.6932, + "step": 41150 + }, + { + "epoch": 1.51, + "learning_rate": 4.948547991357527e-08, + "loss": 0.7029, + "step": 41151 + }, + { + "epoch": 1.51, + "learning_rate": 4.9481817848903206e-08, + "loss": 0.692, + "step": 41152 + }, + { + "epoch": 1.51, + "learning_rate": 4.9478155784231146e-08, + "loss": 0.6878, + "step": 41153 + }, + { + "epoch": 1.51, + "learning_rate": 4.9474493719559086e-08, + "loss": 0.6756, + "step": 41154 + }, + { + "epoch": 1.51, + "learning_rate": 4.9470831654887026e-08, + "loss": 0.6976, + "step": 41155 + }, + { + "epoch": 1.51, + "learning_rate": 4.9467169590214966e-08, + "loss": 0.696, + "step": 41156 + }, + { + "epoch": 1.51, + "learning_rate": 4.94635075255429e-08, + "loss": 0.6981, + "step": 41157 + }, + { + "epoch": 1.51, + "learning_rate": 4.945984546087083e-08, + "loss": 0.6904, + "step": 41158 + }, + { + "epoch": 1.51, + "learning_rate": 4.945618339619877e-08, + "loss": 0.6978, + "step": 41159 + }, + { + "epoch": 1.51, + "learning_rate": 4.945252133152671e-08, + "loss": 0.6867, + "step": 41160 + }, + { + "epoch": 1.51, + "learning_rate": 4.9448859266854646e-08, + "loss": 0.6944, + "step": 41161 + }, + { + "epoch": 1.51, + "learning_rate": 4.9445197202182586e-08, + "loss": 0.6934, + "step": 41162 + }, + { + "epoch": 1.51, + "learning_rate": 4.9441535137510526e-08, + "loss": 0.692, + "step": 41163 + }, + { + "epoch": 1.51, + "learning_rate": 4.9437873072838466e-08, + "loss": 0.6919, + "step": 41164 + }, + { + "epoch": 1.51, + "learning_rate": 4.9434211008166406e-08, + "loss": 0.6994, + "step": 41165 + }, + { + "epoch": 1.51, + "learning_rate": 4.943054894349434e-08, + "loss": 0.6883, + "step": 41166 + }, + { + "epoch": 1.51, + "learning_rate": 4.942688687882228e-08, + "loss": 0.6823, + "step": 41167 + }, + { + "epoch": 1.51, + "learning_rate": 4.942322481415022e-08, + "loss": 0.6885, + "step": 41168 + }, + { + "epoch": 1.51, + "learning_rate": 4.941956274947815e-08, + "loss": 0.6965, + "step": 41169 + }, + { + "epoch": 1.51, + "learning_rate": 4.941590068480609e-08, + "loss": 0.6891, + "step": 41170 + }, + { + "epoch": 1.51, + "learning_rate": 4.9412238620134025e-08, + "loss": 0.6783, + "step": 41171 + }, + { + "epoch": 1.51, + "learning_rate": 4.9408576555461965e-08, + "loss": 0.695, + "step": 41172 + }, + { + "epoch": 1.51, + "learning_rate": 4.9404914490789905e-08, + "loss": 0.7071, + "step": 41173 + }, + { + "epoch": 1.51, + "learning_rate": 4.9401252426117845e-08, + "loss": 0.6963, + "step": 41174 + }, + { + "epoch": 1.51, + "learning_rate": 4.939759036144578e-08, + "loss": 0.701, + "step": 41175 + }, + { + "epoch": 1.51, + "learning_rate": 4.939392829677372e-08, + "loss": 0.6924, + "step": 41176 + }, + { + "epoch": 1.51, + "learning_rate": 4.939026623210166e-08, + "loss": 0.6929, + "step": 41177 + }, + { + "epoch": 1.51, + "learning_rate": 4.93866041674296e-08, + "loss": 0.7044, + "step": 41178 + }, + { + "epoch": 1.51, + "learning_rate": 4.938294210275753e-08, + "loss": 0.6886, + "step": 41179 + }, + { + "epoch": 1.51, + "learning_rate": 4.937928003808547e-08, + "loss": 0.6967, + "step": 41180 + }, + { + "epoch": 1.51, + "learning_rate": 4.937561797341341e-08, + "loss": 0.675, + "step": 41181 + }, + { + "epoch": 1.51, + "learning_rate": 4.9371955908741345e-08, + "loss": 0.687, + "step": 41182 + }, + { + "epoch": 1.51, + "learning_rate": 4.9368293844069285e-08, + "loss": 0.7003, + "step": 41183 + }, + { + "epoch": 1.51, + "learning_rate": 4.936463177939722e-08, + "loss": 0.6822, + "step": 41184 + }, + { + "epoch": 1.51, + "learning_rate": 4.936096971472516e-08, + "loss": 0.6936, + "step": 41185 + }, + { + "epoch": 1.51, + "learning_rate": 4.93573076500531e-08, + "loss": 0.6848, + "step": 41186 + }, + { + "epoch": 1.51, + "learning_rate": 4.935364558538104e-08, + "loss": 0.694, + "step": 41187 + }, + { + "epoch": 1.51, + "learning_rate": 4.934998352070897e-08, + "loss": 0.6949, + "step": 41188 + }, + { + "epoch": 1.51, + "learning_rate": 4.934632145603691e-08, + "loss": 0.6934, + "step": 41189 + }, + { + "epoch": 1.51, + "learning_rate": 4.934265939136485e-08, + "loss": 0.6998, + "step": 41190 + }, + { + "epoch": 1.51, + "learning_rate": 4.933899732669279e-08, + "loss": 0.6939, + "step": 41191 + }, + { + "epoch": 1.51, + "learning_rate": 4.9335335262020724e-08, + "loss": 0.703, + "step": 41192 + }, + { + "epoch": 1.51, + "learning_rate": 4.9331673197348664e-08, + "loss": 0.7022, + "step": 41193 + }, + { + "epoch": 1.51, + "learning_rate": 4.93280111326766e-08, + "loss": 0.6855, + "step": 41194 + }, + { + "epoch": 1.51, + "learning_rate": 4.932434906800454e-08, + "loss": 0.6807, + "step": 41195 + }, + { + "epoch": 1.51, + "learning_rate": 4.932068700333248e-08, + "loss": 0.6797, + "step": 41196 + }, + { + "epoch": 1.51, + "learning_rate": 4.931702493866041e-08, + "loss": 0.6848, + "step": 41197 + }, + { + "epoch": 1.51, + "learning_rate": 4.931336287398835e-08, + "loss": 0.6931, + "step": 41198 + }, + { + "epoch": 1.51, + "learning_rate": 4.930970080931629e-08, + "loss": 0.7018, + "step": 41199 + }, + { + "epoch": 1.51, + "learning_rate": 4.930603874464423e-08, + "loss": 0.6906, + "step": 41200 + }, + { + "epoch": 1.51, + "learning_rate": 4.930237667997217e-08, + "loss": 0.7014, + "step": 41201 + }, + { + "epoch": 1.51, + "learning_rate": 4.9298714615300104e-08, + "loss": 0.6916, + "step": 41202 + }, + { + "epoch": 1.51, + "learning_rate": 4.9295052550628044e-08, + "loss": 0.7053, + "step": 41203 + }, + { + "epoch": 1.51, + "learning_rate": 4.9291390485955983e-08, + "loss": 0.695, + "step": 41204 + }, + { + "epoch": 1.51, + "learning_rate": 4.928772842128392e-08, + "loss": 0.7084, + "step": 41205 + }, + { + "epoch": 1.51, + "learning_rate": 4.928406635661185e-08, + "loss": 0.6967, + "step": 41206 + }, + { + "epoch": 1.51, + "learning_rate": 4.928040429193979e-08, + "loss": 0.6859, + "step": 41207 + }, + { + "epoch": 1.51, + "learning_rate": 4.927674222726773e-08, + "loss": 0.6885, + "step": 41208 + }, + { + "epoch": 1.51, + "learning_rate": 4.927308016259567e-08, + "loss": 0.6895, + "step": 41209 + }, + { + "epoch": 1.51, + "learning_rate": 4.926941809792361e-08, + "loss": 0.6931, + "step": 41210 + }, + { + "epoch": 1.51, + "learning_rate": 4.926575603325154e-08, + "loss": 0.705, + "step": 41211 + }, + { + "epoch": 1.51, + "learning_rate": 4.926209396857948e-08, + "loss": 0.6969, + "step": 41212 + }, + { + "epoch": 1.51, + "learning_rate": 4.925843190390742e-08, + "loss": 0.7132, + "step": 41213 + }, + { + "epoch": 1.51, + "learning_rate": 4.925476983923536e-08, + "loss": 0.6929, + "step": 41214 + }, + { + "epoch": 1.51, + "learning_rate": 4.9251107774563296e-08, + "loss": 0.6995, + "step": 41215 + }, + { + "epoch": 1.51, + "learning_rate": 4.9247445709891236e-08, + "loss": 0.6967, + "step": 41216 + }, + { + "epoch": 1.51, + "learning_rate": 4.924378364521917e-08, + "loss": 0.7031, + "step": 41217 + }, + { + "epoch": 1.51, + "learning_rate": 4.924012158054711e-08, + "loss": 0.7073, + "step": 41218 + }, + { + "epoch": 1.51, + "learning_rate": 4.923645951587505e-08, + "loss": 0.6966, + "step": 41219 + }, + { + "epoch": 1.51, + "learning_rate": 4.923279745120298e-08, + "loss": 0.6977, + "step": 41220 + }, + { + "epoch": 1.51, + "learning_rate": 4.922913538653092e-08, + "loss": 0.6999, + "step": 41221 + }, + { + "epoch": 1.51, + "learning_rate": 4.922547332185886e-08, + "loss": 0.6973, + "step": 41222 + }, + { + "epoch": 1.51, + "learning_rate": 4.92218112571868e-08, + "loss": 0.6894, + "step": 41223 + }, + { + "epoch": 1.51, + "learning_rate": 4.9218149192514736e-08, + "loss": 0.6924, + "step": 41224 + }, + { + "epoch": 1.51, + "learning_rate": 4.9214487127842676e-08, + "loss": 0.6959, + "step": 41225 + }, + { + "epoch": 1.51, + "learning_rate": 4.9210825063170616e-08, + "loss": 0.6935, + "step": 41226 + }, + { + "epoch": 1.51, + "learning_rate": 4.9207162998498556e-08, + "loss": 0.6962, + "step": 41227 + }, + { + "epoch": 1.51, + "learning_rate": 4.920350093382649e-08, + "loss": 0.6905, + "step": 41228 + }, + { + "epoch": 1.51, + "learning_rate": 4.919983886915442e-08, + "loss": 0.6836, + "step": 41229 + }, + { + "epoch": 1.51, + "learning_rate": 4.919617680448236e-08, + "loss": 0.7009, + "step": 41230 + }, + { + "epoch": 1.51, + "learning_rate": 4.91925147398103e-08, + "loss": 0.6791, + "step": 41231 + }, + { + "epoch": 1.51, + "learning_rate": 4.918885267513824e-08, + "loss": 0.6861, + "step": 41232 + }, + { + "epoch": 1.51, + "learning_rate": 4.9185190610466175e-08, + "loss": 0.6947, + "step": 41233 + }, + { + "epoch": 1.51, + "learning_rate": 4.9181528545794115e-08, + "loss": 0.6999, + "step": 41234 + }, + { + "epoch": 1.51, + "learning_rate": 4.9177866481122055e-08, + "loss": 0.6953, + "step": 41235 + }, + { + "epoch": 1.51, + "learning_rate": 4.9174204416449995e-08, + "loss": 0.6825, + "step": 41236 + }, + { + "epoch": 1.51, + "learning_rate": 4.9170542351777935e-08, + "loss": 0.6971, + "step": 41237 + }, + { + "epoch": 1.51, + "learning_rate": 4.916688028710587e-08, + "loss": 0.6904, + "step": 41238 + }, + { + "epoch": 1.51, + "learning_rate": 4.916321822243381e-08, + "loss": 0.693, + "step": 41239 + }, + { + "epoch": 1.51, + "learning_rate": 4.915955615776174e-08, + "loss": 0.6976, + "step": 41240 + }, + { + "epoch": 1.51, + "learning_rate": 4.915589409308968e-08, + "loss": 0.7035, + "step": 41241 + }, + { + "epoch": 1.51, + "learning_rate": 4.9152232028417615e-08, + "loss": 0.704, + "step": 41242 + }, + { + "epoch": 1.51, + "learning_rate": 4.9148569963745555e-08, + "loss": 0.7129, + "step": 41243 + }, + { + "epoch": 1.51, + "learning_rate": 4.9144907899073495e-08, + "loss": 0.6849, + "step": 41244 + }, + { + "epoch": 1.51, + "learning_rate": 4.9141245834401435e-08, + "loss": 0.6932, + "step": 41245 + }, + { + "epoch": 1.51, + "learning_rate": 4.9137583769729374e-08, + "loss": 0.6921, + "step": 41246 + }, + { + "epoch": 1.51, + "learning_rate": 4.913392170505731e-08, + "loss": 0.6971, + "step": 41247 + }, + { + "epoch": 1.51, + "learning_rate": 4.913025964038525e-08, + "loss": 0.7027, + "step": 41248 + }, + { + "epoch": 1.51, + "learning_rate": 4.912659757571319e-08, + "loss": 0.7073, + "step": 41249 + }, + { + "epoch": 1.51, + "learning_rate": 4.912293551104113e-08, + "loss": 0.6857, + "step": 41250 + }, + { + "epoch": 1.51, + "learning_rate": 4.911927344636906e-08, + "loss": 0.6993, + "step": 41251 + }, + { + "epoch": 1.51, + "learning_rate": 4.9115611381696994e-08, + "loss": 0.7023, + "step": 41252 + }, + { + "epoch": 1.51, + "learning_rate": 4.9111949317024934e-08, + "loss": 0.6927, + "step": 41253 + }, + { + "epoch": 1.51, + "learning_rate": 4.9108287252352874e-08, + "loss": 0.6903, + "step": 41254 + }, + { + "epoch": 1.51, + "learning_rate": 4.9104625187680814e-08, + "loss": 0.6878, + "step": 41255 + }, + { + "epoch": 1.51, + "learning_rate": 4.910096312300875e-08, + "loss": 0.6901, + "step": 41256 + }, + { + "epoch": 1.51, + "learning_rate": 4.909730105833669e-08, + "loss": 0.695, + "step": 41257 + }, + { + "epoch": 1.51, + "learning_rate": 4.909363899366463e-08, + "loss": 0.6875, + "step": 41258 + }, + { + "epoch": 1.51, + "learning_rate": 4.908997692899257e-08, + "loss": 0.6907, + "step": 41259 + }, + { + "epoch": 1.51, + "learning_rate": 4.90863148643205e-08, + "loss": 0.6985, + "step": 41260 + }, + { + "epoch": 1.51, + "learning_rate": 4.908265279964844e-08, + "loss": 0.6988, + "step": 41261 + }, + { + "epoch": 1.51, + "learning_rate": 4.907899073497638e-08, + "loss": 0.6855, + "step": 41262 + }, + { + "epoch": 1.51, + "learning_rate": 4.9075328670304314e-08, + "loss": 0.7071, + "step": 41263 + }, + { + "epoch": 1.51, + "learning_rate": 4.9071666605632253e-08, + "loss": 0.6975, + "step": 41264 + }, + { + "epoch": 1.51, + "learning_rate": 4.906800454096019e-08, + "loss": 0.7058, + "step": 41265 + }, + { + "epoch": 1.51, + "learning_rate": 4.906434247628813e-08, + "loss": 0.6983, + "step": 41266 + }, + { + "epoch": 1.51, + "learning_rate": 4.9060680411616067e-08, + "loss": 0.6835, + "step": 41267 + }, + { + "epoch": 1.51, + "learning_rate": 4.9057018346944007e-08, + "loss": 0.6888, + "step": 41268 + }, + { + "epoch": 1.51, + "learning_rate": 4.905335628227194e-08, + "loss": 0.6836, + "step": 41269 + }, + { + "epoch": 1.51, + "learning_rate": 4.904969421759988e-08, + "loss": 0.6917, + "step": 41270 + }, + { + "epoch": 1.51, + "learning_rate": 4.904603215292782e-08, + "loss": 0.7032, + "step": 41271 + }, + { + "epoch": 1.51, + "learning_rate": 4.904237008825576e-08, + "loss": 0.6884, + "step": 41272 + }, + { + "epoch": 1.51, + "learning_rate": 4.90387080235837e-08, + "loss": 0.6951, + "step": 41273 + }, + { + "epoch": 1.51, + "learning_rate": 4.903504595891163e-08, + "loss": 0.69, + "step": 41274 + }, + { + "epoch": 1.51, + "learning_rate": 4.903138389423957e-08, + "loss": 0.6872, + "step": 41275 + }, + { + "epoch": 1.51, + "learning_rate": 4.9027721829567506e-08, + "loss": 0.6836, + "step": 41276 + }, + { + "epoch": 1.51, + "learning_rate": 4.9024059764895446e-08, + "loss": 0.698, + "step": 41277 + }, + { + "epoch": 1.51, + "learning_rate": 4.902039770022338e-08, + "loss": 0.7013, + "step": 41278 + }, + { + "epoch": 1.51, + "learning_rate": 4.901673563555132e-08, + "loss": 0.6877, + "step": 41279 + }, + { + "epoch": 1.51, + "learning_rate": 4.901307357087926e-08, + "loss": 0.6941, + "step": 41280 + }, + { + "epoch": 1.51, + "learning_rate": 4.90094115062072e-08, + "loss": 0.6843, + "step": 41281 + }, + { + "epoch": 1.51, + "learning_rate": 4.900574944153514e-08, + "loss": 0.6876, + "step": 41282 + }, + { + "epoch": 1.51, + "learning_rate": 4.900208737686307e-08, + "loss": 0.6946, + "step": 41283 + }, + { + "epoch": 1.51, + "learning_rate": 4.899842531219101e-08, + "loss": 0.6843, + "step": 41284 + }, + { + "epoch": 1.51, + "learning_rate": 4.899476324751895e-08, + "loss": 0.6986, + "step": 41285 + }, + { + "epoch": 1.51, + "learning_rate": 4.899110118284689e-08, + "loss": 0.6915, + "step": 41286 + }, + { + "epoch": 1.51, + "learning_rate": 4.8987439118174826e-08, + "loss": 0.6942, + "step": 41287 + }, + { + "epoch": 1.51, + "learning_rate": 4.898377705350276e-08, + "loss": 0.6937, + "step": 41288 + }, + { + "epoch": 1.51, + "learning_rate": 4.89801149888307e-08, + "loss": 0.6978, + "step": 41289 + }, + { + "epoch": 1.51, + "learning_rate": 4.897645292415864e-08, + "loss": 0.6879, + "step": 41290 + }, + { + "epoch": 1.51, + "learning_rate": 4.897279085948658e-08, + "loss": 0.6901, + "step": 41291 + }, + { + "epoch": 1.51, + "learning_rate": 4.896912879481451e-08, + "loss": 0.699, + "step": 41292 + }, + { + "epoch": 1.51, + "learning_rate": 4.896546673014245e-08, + "loss": 0.6886, + "step": 41293 + }, + { + "epoch": 1.51, + "learning_rate": 4.896180466547039e-08, + "loss": 0.7054, + "step": 41294 + }, + { + "epoch": 1.51, + "learning_rate": 4.895814260079833e-08, + "loss": 0.6884, + "step": 41295 + }, + { + "epoch": 1.51, + "learning_rate": 4.8954480536126265e-08, + "loss": 0.6828, + "step": 41296 + }, + { + "epoch": 1.51, + "learning_rate": 4.8950818471454205e-08, + "loss": 0.6882, + "step": 41297 + }, + { + "epoch": 1.51, + "learning_rate": 4.8947156406782145e-08, + "loss": 0.6922, + "step": 41298 + }, + { + "epoch": 1.51, + "learning_rate": 4.894349434211008e-08, + "loss": 0.7035, + "step": 41299 + }, + { + "epoch": 1.51, + "learning_rate": 4.893983227743801e-08, + "loss": 0.6905, + "step": 41300 + }, + { + "epoch": 1.51, + "learning_rate": 4.893617021276595e-08, + "loss": 0.6922, + "step": 41301 + }, + { + "epoch": 1.51, + "learning_rate": 4.893250814809389e-08, + "loss": 0.6919, + "step": 41302 + }, + { + "epoch": 1.51, + "learning_rate": 4.892884608342183e-08, + "loss": 0.6865, + "step": 41303 + }, + { + "epoch": 1.51, + "learning_rate": 4.892518401874977e-08, + "loss": 0.6903, + "step": 41304 + }, + { + "epoch": 1.51, + "learning_rate": 4.8921521954077705e-08, + "loss": 0.6891, + "step": 41305 + }, + { + "epoch": 1.51, + "learning_rate": 4.8917859889405644e-08, + "loss": 0.6957, + "step": 41306 + }, + { + "epoch": 1.51, + "learning_rate": 4.8914197824733584e-08, + "loss": 0.6832, + "step": 41307 + }, + { + "epoch": 1.51, + "learning_rate": 4.8910535760061524e-08, + "loss": 0.7045, + "step": 41308 + }, + { + "epoch": 1.51, + "learning_rate": 4.890687369538946e-08, + "loss": 0.6956, + "step": 41309 + }, + { + "epoch": 1.51, + "learning_rate": 4.89032116307174e-08, + "loss": 0.6955, + "step": 41310 + }, + { + "epoch": 1.51, + "learning_rate": 4.889954956604533e-08, + "loss": 0.693, + "step": 41311 + }, + { + "epoch": 1.51, + "learning_rate": 4.889588750137327e-08, + "loss": 0.6929, + "step": 41312 + }, + { + "epoch": 1.51, + "learning_rate": 4.889222543670121e-08, + "loss": 0.6903, + "step": 41313 + }, + { + "epoch": 1.51, + "learning_rate": 4.8888563372029144e-08, + "loss": 0.7046, + "step": 41314 + }, + { + "epoch": 1.51, + "learning_rate": 4.8884901307357084e-08, + "loss": 0.6799, + "step": 41315 + }, + { + "epoch": 1.51, + "learning_rate": 4.8881239242685024e-08, + "loss": 0.6954, + "step": 41316 + }, + { + "epoch": 1.51, + "learning_rate": 4.8877577178012964e-08, + "loss": 0.7064, + "step": 41317 + }, + { + "epoch": 1.51, + "learning_rate": 4.88739151133409e-08, + "loss": 0.6924, + "step": 41318 + }, + { + "epoch": 1.51, + "learning_rate": 4.887025304866884e-08, + "loss": 0.713, + "step": 41319 + }, + { + "epoch": 1.51, + "learning_rate": 4.886659098399678e-08, + "loss": 0.6804, + "step": 41320 + }, + { + "epoch": 1.51, + "learning_rate": 4.886292891932472e-08, + "loss": 0.7077, + "step": 41321 + }, + { + "epoch": 1.51, + "learning_rate": 4.885926685465265e-08, + "loss": 0.6895, + "step": 41322 + }, + { + "epoch": 1.51, + "learning_rate": 4.8855604789980584e-08, + "loss": 0.6868, + "step": 41323 + }, + { + "epoch": 1.51, + "learning_rate": 4.8851942725308523e-08, + "loss": 0.7083, + "step": 41324 + }, + { + "epoch": 1.51, + "learning_rate": 4.8848280660636463e-08, + "loss": 0.6948, + "step": 41325 + }, + { + "epoch": 1.51, + "learning_rate": 4.8844618595964403e-08, + "loss": 0.6967, + "step": 41326 + }, + { + "epoch": 1.51, + "learning_rate": 4.8840956531292337e-08, + "loss": 0.7059, + "step": 41327 + }, + { + "epoch": 1.51, + "learning_rate": 4.8837294466620277e-08, + "loss": 0.704, + "step": 41328 + }, + { + "epoch": 1.51, + "learning_rate": 4.8833632401948216e-08, + "loss": 0.704, + "step": 41329 + }, + { + "epoch": 1.51, + "learning_rate": 4.8829970337276156e-08, + "loss": 0.6908, + "step": 41330 + }, + { + "epoch": 1.51, + "learning_rate": 4.8826308272604096e-08, + "loss": 0.6816, + "step": 41331 + }, + { + "epoch": 1.51, + "learning_rate": 4.882264620793203e-08, + "loss": 0.6878, + "step": 41332 + }, + { + "epoch": 1.51, + "learning_rate": 4.881898414325997e-08, + "loss": 0.6985, + "step": 41333 + }, + { + "epoch": 1.51, + "learning_rate": 4.88153220785879e-08, + "loss": 0.6925, + "step": 41334 + }, + { + "epoch": 1.51, + "learning_rate": 4.881166001391584e-08, + "loss": 0.6967, + "step": 41335 + }, + { + "epoch": 1.51, + "learning_rate": 4.8807997949243776e-08, + "loss": 0.6923, + "step": 41336 + }, + { + "epoch": 1.51, + "learning_rate": 4.8804335884571716e-08, + "loss": 0.6975, + "step": 41337 + }, + { + "epoch": 1.51, + "learning_rate": 4.8800673819899656e-08, + "loss": 0.6995, + "step": 41338 + }, + { + "epoch": 1.51, + "learning_rate": 4.8797011755227596e-08, + "loss": 0.6975, + "step": 41339 + }, + { + "epoch": 1.51, + "learning_rate": 4.8793349690555536e-08, + "loss": 0.6987, + "step": 41340 + }, + { + "epoch": 1.51, + "learning_rate": 4.878968762588347e-08, + "loss": 0.6792, + "step": 41341 + }, + { + "epoch": 1.51, + "learning_rate": 4.878602556121141e-08, + "loss": 0.6986, + "step": 41342 + }, + { + "epoch": 1.51, + "learning_rate": 4.878236349653935e-08, + "loss": 0.6912, + "step": 41343 + }, + { + "epoch": 1.51, + "learning_rate": 4.877870143186729e-08, + "loss": 0.6877, + "step": 41344 + }, + { + "epoch": 1.51, + "learning_rate": 4.877503936719522e-08, + "loss": 0.7076, + "step": 41345 + }, + { + "epoch": 1.51, + "learning_rate": 4.877137730252316e-08, + "loss": 0.6906, + "step": 41346 + }, + { + "epoch": 1.51, + "learning_rate": 4.8767715237851096e-08, + "loss": 0.7001, + "step": 41347 + }, + { + "epoch": 1.51, + "learning_rate": 4.8764053173179035e-08, + "loss": 0.6941, + "step": 41348 + }, + { + "epoch": 1.51, + "learning_rate": 4.8760391108506975e-08, + "loss": 0.6925, + "step": 41349 + }, + { + "epoch": 1.51, + "learning_rate": 4.875672904383491e-08, + "loss": 0.7019, + "step": 41350 + }, + { + "epoch": 1.51, + "learning_rate": 4.875306697916285e-08, + "loss": 0.7, + "step": 41351 + }, + { + "epoch": 1.51, + "learning_rate": 4.874940491449079e-08, + "loss": 0.6977, + "step": 41352 + }, + { + "epoch": 1.51, + "learning_rate": 4.874574284981873e-08, + "loss": 0.6965, + "step": 41353 + }, + { + "epoch": 1.51, + "learning_rate": 4.874208078514666e-08, + "loss": 0.7017, + "step": 41354 + }, + { + "epoch": 1.51, + "learning_rate": 4.87384187204746e-08, + "loss": 0.6912, + "step": 41355 + }, + { + "epoch": 1.51, + "learning_rate": 4.873475665580254e-08, + "loss": 0.6987, + "step": 41356 + }, + { + "epoch": 1.51, + "learning_rate": 4.873109459113048e-08, + "loss": 0.6982, + "step": 41357 + }, + { + "epoch": 1.51, + "learning_rate": 4.8727432526458415e-08, + "loss": 0.6917, + "step": 41358 + }, + { + "epoch": 1.51, + "learning_rate": 4.872377046178635e-08, + "loss": 0.7029, + "step": 41359 + }, + { + "epoch": 1.51, + "learning_rate": 4.872010839711429e-08, + "loss": 0.6765, + "step": 41360 + }, + { + "epoch": 1.51, + "learning_rate": 4.871644633244223e-08, + "loss": 0.7075, + "step": 41361 + }, + { + "epoch": 1.51, + "learning_rate": 4.871278426777017e-08, + "loss": 0.6876, + "step": 41362 + }, + { + "epoch": 1.51, + "learning_rate": 4.87091222030981e-08, + "loss": 0.6962, + "step": 41363 + }, + { + "epoch": 1.51, + "learning_rate": 4.870546013842604e-08, + "loss": 0.7006, + "step": 41364 + }, + { + "epoch": 1.51, + "learning_rate": 4.870179807375398e-08, + "loss": 0.6983, + "step": 41365 + }, + { + "epoch": 1.51, + "learning_rate": 4.869813600908192e-08, + "loss": 0.6813, + "step": 41366 + }, + { + "epoch": 1.51, + "learning_rate": 4.869447394440986e-08, + "loss": 0.6985, + "step": 41367 + }, + { + "epoch": 1.51, + "learning_rate": 4.8690811879737794e-08, + "loss": 0.6891, + "step": 41368 + }, + { + "epoch": 1.51, + "learning_rate": 4.8687149815065734e-08, + "loss": 0.7078, + "step": 41369 + }, + { + "epoch": 1.51, + "learning_rate": 4.868348775039367e-08, + "loss": 0.705, + "step": 41370 + }, + { + "epoch": 1.51, + "learning_rate": 4.867982568572161e-08, + "loss": 0.6963, + "step": 41371 + }, + { + "epoch": 1.51, + "learning_rate": 4.867616362104954e-08, + "loss": 0.6931, + "step": 41372 + }, + { + "epoch": 1.51, + "learning_rate": 4.867250155637748e-08, + "loss": 0.6968, + "step": 41373 + }, + { + "epoch": 1.51, + "learning_rate": 4.866883949170542e-08, + "loss": 0.6913, + "step": 41374 + }, + { + "epoch": 1.51, + "learning_rate": 4.866517742703336e-08, + "loss": 0.6773, + "step": 41375 + }, + { + "epoch": 1.51, + "learning_rate": 4.86615153623613e-08, + "loss": 0.6746, + "step": 41376 + }, + { + "epoch": 1.51, + "learning_rate": 4.8657853297689234e-08, + "loss": 0.6891, + "step": 41377 + }, + { + "epoch": 1.51, + "learning_rate": 4.8654191233017174e-08, + "loss": 0.7002, + "step": 41378 + }, + { + "epoch": 1.51, + "learning_rate": 4.8650529168345114e-08, + "loss": 0.6959, + "step": 41379 + }, + { + "epoch": 1.51, + "learning_rate": 4.8646867103673054e-08, + "loss": 0.696, + "step": 41380 + }, + { + "epoch": 1.51, + "learning_rate": 4.864320503900099e-08, + "loss": 0.6961, + "step": 41381 + }, + { + "epoch": 1.51, + "learning_rate": 4.863954297432892e-08, + "loss": 0.6904, + "step": 41382 + }, + { + "epoch": 1.51, + "learning_rate": 4.863588090965686e-08, + "loss": 0.6874, + "step": 41383 + }, + { + "epoch": 1.51, + "learning_rate": 4.86322188449848e-08, + "loss": 0.6891, + "step": 41384 + }, + { + "epoch": 1.51, + "learning_rate": 4.862855678031274e-08, + "loss": 0.694, + "step": 41385 + }, + { + "epoch": 1.51, + "learning_rate": 4.8624894715640673e-08, + "loss": 0.697, + "step": 41386 + }, + { + "epoch": 1.51, + "learning_rate": 4.862123265096861e-08, + "loss": 0.6866, + "step": 41387 + }, + { + "epoch": 1.51, + "learning_rate": 4.861757058629655e-08, + "loss": 0.6917, + "step": 41388 + }, + { + "epoch": 1.51, + "learning_rate": 4.861390852162449e-08, + "loss": 0.7001, + "step": 41389 + }, + { + "epoch": 1.51, + "learning_rate": 4.8610246456952426e-08, + "loss": 0.6958, + "step": 41390 + }, + { + "epoch": 1.51, + "learning_rate": 4.8606584392280366e-08, + "loss": 0.6948, + "step": 41391 + }, + { + "epoch": 1.51, + "learning_rate": 4.8602922327608306e-08, + "loss": 0.6969, + "step": 41392 + }, + { + "epoch": 1.51, + "learning_rate": 4.859926026293624e-08, + "loss": 0.7046, + "step": 41393 + }, + { + "epoch": 1.51, + "learning_rate": 4.859559819826418e-08, + "loss": 0.6926, + "step": 41394 + }, + { + "epoch": 1.51, + "learning_rate": 4.859193613359211e-08, + "loss": 0.6876, + "step": 41395 + }, + { + "epoch": 1.51, + "learning_rate": 4.858827406892005e-08, + "loss": 0.7036, + "step": 41396 + }, + { + "epoch": 1.51, + "learning_rate": 4.858461200424799e-08, + "loss": 0.6995, + "step": 41397 + }, + { + "epoch": 1.51, + "learning_rate": 4.858094993957593e-08, + "loss": 0.691, + "step": 41398 + }, + { + "epoch": 1.51, + "learning_rate": 4.8577287874903866e-08, + "loss": 0.6952, + "step": 41399 + }, + { + "epoch": 1.51, + "learning_rate": 4.8573625810231806e-08, + "loss": 0.6939, + "step": 41400 + }, + { + "epoch": 1.51, + "learning_rate": 4.8569963745559746e-08, + "loss": 0.6975, + "step": 41401 + }, + { + "epoch": 1.51, + "learning_rate": 4.8566301680887686e-08, + "loss": 0.6894, + "step": 41402 + }, + { + "epoch": 1.51, + "learning_rate": 4.8562639616215626e-08, + "loss": 0.7008, + "step": 41403 + }, + { + "epoch": 1.51, + "learning_rate": 4.855897755154356e-08, + "loss": 0.6817, + "step": 41404 + }, + { + "epoch": 1.51, + "learning_rate": 4.855531548687149e-08, + "loss": 0.6994, + "step": 41405 + }, + { + "epoch": 1.51, + "learning_rate": 4.855165342219943e-08, + "loss": 0.6851, + "step": 41406 + }, + { + "epoch": 1.51, + "learning_rate": 4.854799135752737e-08, + "loss": 0.6936, + "step": 41407 + }, + { + "epoch": 1.52, + "learning_rate": 4.8544329292855305e-08, + "loss": 0.7023, + "step": 41408 + }, + { + "epoch": 1.52, + "learning_rate": 4.8540667228183245e-08, + "loss": 0.7058, + "step": 41409 + }, + { + "epoch": 1.52, + "learning_rate": 4.8537005163511185e-08, + "loss": 0.6837, + "step": 41410 + }, + { + "epoch": 1.52, + "learning_rate": 4.8533343098839125e-08, + "loss": 0.7067, + "step": 41411 + }, + { + "epoch": 1.52, + "learning_rate": 4.8529681034167065e-08, + "loss": 0.6965, + "step": 41412 + }, + { + "epoch": 1.52, + "learning_rate": 4.8526018969495e-08, + "loss": 0.6893, + "step": 41413 + }, + { + "epoch": 1.52, + "learning_rate": 4.852235690482294e-08, + "loss": 0.6938, + "step": 41414 + }, + { + "epoch": 1.52, + "learning_rate": 4.851869484015088e-08, + "loss": 0.6995, + "step": 41415 + }, + { + "epoch": 1.52, + "learning_rate": 4.851503277547881e-08, + "loss": 0.6745, + "step": 41416 + }, + { + "epoch": 1.52, + "learning_rate": 4.851137071080675e-08, + "loss": 0.6997, + "step": 41417 + }, + { + "epoch": 1.52, + "learning_rate": 4.8507708646134685e-08, + "loss": 0.6875, + "step": 41418 + }, + { + "epoch": 1.52, + "learning_rate": 4.8504046581462625e-08, + "loss": 0.6962, + "step": 41419 + }, + { + "epoch": 1.52, + "learning_rate": 4.8500384516790565e-08, + "loss": 0.6849, + "step": 41420 + }, + { + "epoch": 1.52, + "learning_rate": 4.8496722452118505e-08, + "loss": 0.6973, + "step": 41421 + }, + { + "epoch": 1.52, + "learning_rate": 4.849306038744644e-08, + "loss": 0.6974, + "step": 41422 + }, + { + "epoch": 1.52, + "learning_rate": 4.848939832277438e-08, + "loss": 0.6904, + "step": 41423 + }, + { + "epoch": 1.52, + "learning_rate": 4.848573625810232e-08, + "loss": 0.6915, + "step": 41424 + }, + { + "epoch": 1.52, + "learning_rate": 4.848207419343026e-08, + "loss": 0.6859, + "step": 41425 + }, + { + "epoch": 1.52, + "learning_rate": 4.847841212875819e-08, + "loss": 0.7007, + "step": 41426 + }, + { + "epoch": 1.52, + "learning_rate": 4.847475006408613e-08, + "loss": 0.6805, + "step": 41427 + }, + { + "epoch": 1.52, + "learning_rate": 4.8471087999414064e-08, + "loss": 0.695, + "step": 41428 + }, + { + "epoch": 1.52, + "learning_rate": 4.8467425934742004e-08, + "loss": 0.6855, + "step": 41429 + }, + { + "epoch": 1.52, + "learning_rate": 4.8463763870069944e-08, + "loss": 0.6862, + "step": 41430 + }, + { + "epoch": 1.52, + "learning_rate": 4.846010180539788e-08, + "loss": 0.6895, + "step": 41431 + }, + { + "epoch": 1.52, + "learning_rate": 4.845643974072582e-08, + "loss": 0.681, + "step": 41432 + }, + { + "epoch": 1.52, + "learning_rate": 4.845277767605376e-08, + "loss": 0.697, + "step": 41433 + }, + { + "epoch": 1.52, + "learning_rate": 4.84491156113817e-08, + "loss": 0.6911, + "step": 41434 + }, + { + "epoch": 1.52, + "learning_rate": 4.844545354670963e-08, + "loss": 0.6864, + "step": 41435 + }, + { + "epoch": 1.52, + "learning_rate": 4.844179148203757e-08, + "loss": 0.6897, + "step": 41436 + }, + { + "epoch": 1.52, + "learning_rate": 4.843812941736551e-08, + "loss": 0.6821, + "step": 41437 + }, + { + "epoch": 1.52, + "learning_rate": 4.843446735269345e-08, + "loss": 0.6934, + "step": 41438 + }, + { + "epoch": 1.52, + "learning_rate": 4.8430805288021384e-08, + "loss": 0.6857, + "step": 41439 + }, + { + "epoch": 1.52, + "learning_rate": 4.8427143223349324e-08, + "loss": 0.6877, + "step": 41440 + }, + { + "epoch": 1.52, + "learning_rate": 4.842348115867726e-08, + "loss": 0.6809, + "step": 41441 + }, + { + "epoch": 1.52, + "learning_rate": 4.84198190940052e-08, + "loss": 0.6936, + "step": 41442 + }, + { + "epoch": 1.52, + "learning_rate": 4.841615702933314e-08, + "loss": 0.6919, + "step": 41443 + }, + { + "epoch": 1.52, + "learning_rate": 4.841249496466107e-08, + "loss": 0.6843, + "step": 41444 + }, + { + "epoch": 1.52, + "learning_rate": 4.840883289998901e-08, + "loss": 0.6973, + "step": 41445 + }, + { + "epoch": 1.52, + "learning_rate": 4.840517083531695e-08, + "loss": 0.6925, + "step": 41446 + }, + { + "epoch": 1.52, + "learning_rate": 4.840150877064489e-08, + "loss": 0.702, + "step": 41447 + }, + { + "epoch": 1.52, + "learning_rate": 4.839784670597283e-08, + "loss": 0.6848, + "step": 41448 + }, + { + "epoch": 1.52, + "learning_rate": 4.839418464130076e-08, + "loss": 0.6947, + "step": 41449 + }, + { + "epoch": 1.52, + "learning_rate": 4.83905225766287e-08, + "loss": 0.6873, + "step": 41450 + }, + { + "epoch": 1.52, + "learning_rate": 4.838686051195664e-08, + "loss": 0.6714, + "step": 41451 + }, + { + "epoch": 1.52, + "learning_rate": 4.8383198447284576e-08, + "loss": 0.6843, + "step": 41452 + }, + { + "epoch": 1.52, + "learning_rate": 4.837953638261251e-08, + "loss": 0.6974, + "step": 41453 + }, + { + "epoch": 1.52, + "learning_rate": 4.837587431794045e-08, + "loss": 0.7014, + "step": 41454 + }, + { + "epoch": 1.52, + "learning_rate": 4.837221225326839e-08, + "loss": 0.6928, + "step": 41455 + }, + { + "epoch": 1.52, + "learning_rate": 4.836855018859633e-08, + "loss": 0.6915, + "step": 41456 + }, + { + "epoch": 1.52, + "learning_rate": 4.836488812392427e-08, + "loss": 0.6987, + "step": 41457 + }, + { + "epoch": 1.52, + "learning_rate": 4.83612260592522e-08, + "loss": 0.6859, + "step": 41458 + }, + { + "epoch": 1.52, + "learning_rate": 4.835756399458014e-08, + "loss": 0.6995, + "step": 41459 + }, + { + "epoch": 1.52, + "learning_rate": 4.835390192990808e-08, + "loss": 0.6889, + "step": 41460 + }, + { + "epoch": 1.52, + "learning_rate": 4.835023986523602e-08, + "loss": 0.6927, + "step": 41461 + }, + { + "epoch": 1.52, + "learning_rate": 4.8346577800563956e-08, + "loss": 0.6885, + "step": 41462 + }, + { + "epoch": 1.52, + "learning_rate": 4.8342915735891896e-08, + "loss": 0.6834, + "step": 41463 + }, + { + "epoch": 1.52, + "learning_rate": 4.833925367121983e-08, + "loss": 0.6859, + "step": 41464 + }, + { + "epoch": 1.52, + "learning_rate": 4.833559160654777e-08, + "loss": 0.6846, + "step": 41465 + }, + { + "epoch": 1.52, + "learning_rate": 4.833192954187571e-08, + "loss": 0.692, + "step": 41466 + }, + { + "epoch": 1.52, + "learning_rate": 4.832826747720364e-08, + "loss": 0.693, + "step": 41467 + }, + { + "epoch": 1.52, + "learning_rate": 4.832460541253158e-08, + "loss": 0.7008, + "step": 41468 + }, + { + "epoch": 1.52, + "learning_rate": 4.832094334785952e-08, + "loss": 0.691, + "step": 41469 + }, + { + "epoch": 1.52, + "learning_rate": 4.831728128318746e-08, + "loss": 0.6938, + "step": 41470 + }, + { + "epoch": 1.52, + "learning_rate": 4.8313619218515395e-08, + "loss": 0.6933, + "step": 41471 + }, + { + "epoch": 1.52, + "learning_rate": 4.8309957153843335e-08, + "loss": 0.6972, + "step": 41472 + }, + { + "epoch": 1.52, + "learning_rate": 4.8306295089171275e-08, + "loss": 0.6883, + "step": 41473 + }, + { + "epoch": 1.52, + "learning_rate": 4.8302633024499215e-08, + "loss": 0.6979, + "step": 41474 + }, + { + "epoch": 1.52, + "learning_rate": 4.829897095982715e-08, + "loss": 0.703, + "step": 41475 + }, + { + "epoch": 1.52, + "learning_rate": 4.829530889515508e-08, + "loss": 0.7001, + "step": 41476 + }, + { + "epoch": 1.52, + "learning_rate": 4.829164683048302e-08, + "loss": 0.6901, + "step": 41477 + }, + { + "epoch": 1.52, + "learning_rate": 4.828798476581096e-08, + "loss": 0.6914, + "step": 41478 + }, + { + "epoch": 1.52, + "learning_rate": 4.82843227011389e-08, + "loss": 0.6858, + "step": 41479 + }, + { + "epoch": 1.52, + "learning_rate": 4.8280660636466835e-08, + "loss": 0.6944, + "step": 41480 + }, + { + "epoch": 1.52, + "learning_rate": 4.8276998571794775e-08, + "loss": 0.7005, + "step": 41481 + }, + { + "epoch": 1.52, + "learning_rate": 4.8273336507122715e-08, + "loss": 0.698, + "step": 41482 + }, + { + "epoch": 1.52, + "learning_rate": 4.8269674442450654e-08, + "loss": 0.6986, + "step": 41483 + }, + { + "epoch": 1.52, + "learning_rate": 4.8266012377778594e-08, + "loss": 0.688, + "step": 41484 + }, + { + "epoch": 1.52, + "learning_rate": 4.826235031310653e-08, + "loss": 0.6979, + "step": 41485 + }, + { + "epoch": 1.52, + "learning_rate": 4.825868824843447e-08, + "loss": 0.6866, + "step": 41486 + }, + { + "epoch": 1.52, + "learning_rate": 4.82550261837624e-08, + "loss": 0.6881, + "step": 41487 + }, + { + "epoch": 1.52, + "learning_rate": 4.825136411909034e-08, + "loss": 0.706, + "step": 41488 + }, + { + "epoch": 1.52, + "learning_rate": 4.8247702054418274e-08, + "loss": 0.6884, + "step": 41489 + }, + { + "epoch": 1.52, + "learning_rate": 4.8244039989746214e-08, + "loss": 0.6935, + "step": 41490 + }, + { + "epoch": 1.52, + "learning_rate": 4.8240377925074154e-08, + "loss": 0.6949, + "step": 41491 + }, + { + "epoch": 1.52, + "learning_rate": 4.8236715860402094e-08, + "loss": 0.7045, + "step": 41492 + }, + { + "epoch": 1.52, + "learning_rate": 4.823305379573003e-08, + "loss": 0.6853, + "step": 41493 + }, + { + "epoch": 1.52, + "learning_rate": 4.822939173105797e-08, + "loss": 0.7024, + "step": 41494 + }, + { + "epoch": 1.52, + "learning_rate": 4.822572966638591e-08, + "loss": 0.6929, + "step": 41495 + }, + { + "epoch": 1.52, + "learning_rate": 4.822206760171385e-08, + "loss": 0.6786, + "step": 41496 + }, + { + "epoch": 1.52, + "learning_rate": 4.821840553704179e-08, + "loss": 0.6878, + "step": 41497 + }, + { + "epoch": 1.52, + "learning_rate": 4.821474347236972e-08, + "loss": 0.6983, + "step": 41498 + }, + { + "epoch": 1.52, + "learning_rate": 4.8211081407697654e-08, + "loss": 0.6944, + "step": 41499 + }, + { + "epoch": 1.52, + "learning_rate": 4.8207419343025594e-08, + "loss": 0.7055, + "step": 41500 + }, + { + "epoch": 1.52, + "learning_rate": 4.8203757278353533e-08, + "loss": 0.7094, + "step": 41501 + }, + { + "epoch": 1.52, + "learning_rate": 4.820009521368147e-08, + "loss": 0.6945, + "step": 41502 + }, + { + "epoch": 1.52, + "learning_rate": 4.819643314900941e-08, + "loss": 0.6774, + "step": 41503 + }, + { + "epoch": 1.52, + "learning_rate": 4.8192771084337347e-08, + "loss": 0.6979, + "step": 41504 + }, + { + "epoch": 1.52, + "learning_rate": 4.8189109019665287e-08, + "loss": 0.697, + "step": 41505 + }, + { + "epoch": 1.52, + "learning_rate": 4.8185446954993227e-08, + "loss": 0.6953, + "step": 41506 + }, + { + "epoch": 1.52, + "learning_rate": 4.818178489032116e-08, + "loss": 0.6948, + "step": 41507 + }, + { + "epoch": 1.52, + "learning_rate": 4.81781228256491e-08, + "loss": 0.6946, + "step": 41508 + }, + { + "epoch": 1.52, + "learning_rate": 4.817446076097704e-08, + "loss": 0.6997, + "step": 41509 + }, + { + "epoch": 1.52, + "learning_rate": 4.817079869630497e-08, + "loss": 0.6859, + "step": 41510 + }, + { + "epoch": 1.52, + "learning_rate": 4.816713663163291e-08, + "loss": 0.6874, + "step": 41511 + }, + { + "epoch": 1.52, + "learning_rate": 4.8163474566960846e-08, + "loss": 0.6778, + "step": 41512 + }, + { + "epoch": 1.52, + "learning_rate": 4.8159812502288786e-08, + "loss": 0.6993, + "step": 41513 + }, + { + "epoch": 1.52, + "learning_rate": 4.8156150437616726e-08, + "loss": 0.6911, + "step": 41514 + }, + { + "epoch": 1.52, + "learning_rate": 4.8152488372944666e-08, + "loss": 0.6988, + "step": 41515 + }, + { + "epoch": 1.52, + "learning_rate": 4.81488263082726e-08, + "loss": 0.7005, + "step": 41516 + }, + { + "epoch": 1.52, + "learning_rate": 4.814516424360054e-08, + "loss": 0.6833, + "step": 41517 + }, + { + "epoch": 1.52, + "learning_rate": 4.814150217892848e-08, + "loss": 0.6844, + "step": 41518 + }, + { + "epoch": 1.52, + "learning_rate": 4.813784011425642e-08, + "loss": 0.6964, + "step": 41519 + }, + { + "epoch": 1.52, + "learning_rate": 4.813417804958435e-08, + "loss": 0.6859, + "step": 41520 + }, + { + "epoch": 1.52, + "learning_rate": 4.813051598491229e-08, + "loss": 0.7, + "step": 41521 + }, + { + "epoch": 1.52, + "learning_rate": 4.812685392024023e-08, + "loss": 0.6924, + "step": 41522 + }, + { + "epoch": 1.52, + "learning_rate": 4.8123191855568166e-08, + "loss": 0.6981, + "step": 41523 + }, + { + "epoch": 1.52, + "learning_rate": 4.8119529790896106e-08, + "loss": 0.6954, + "step": 41524 + }, + { + "epoch": 1.52, + "learning_rate": 4.811586772622404e-08, + "loss": 0.6789, + "step": 41525 + }, + { + "epoch": 1.52, + "learning_rate": 4.811220566155198e-08, + "loss": 0.693, + "step": 41526 + }, + { + "epoch": 1.52, + "learning_rate": 4.810854359687992e-08, + "loss": 0.6862, + "step": 41527 + }, + { + "epoch": 1.52, + "learning_rate": 4.810488153220786e-08, + "loss": 0.6862, + "step": 41528 + }, + { + "epoch": 1.52, + "learning_rate": 4.810121946753579e-08, + "loss": 0.7007, + "step": 41529 + }, + { + "epoch": 1.52, + "learning_rate": 4.809755740286373e-08, + "loss": 0.6872, + "step": 41530 + }, + { + "epoch": 1.52, + "learning_rate": 4.809389533819167e-08, + "loss": 0.6974, + "step": 41531 + }, + { + "epoch": 1.52, + "learning_rate": 4.809023327351961e-08, + "loss": 0.6938, + "step": 41532 + }, + { + "epoch": 1.52, + "learning_rate": 4.808657120884755e-08, + "loss": 0.6914, + "step": 41533 + }, + { + "epoch": 1.52, + "learning_rate": 4.8082909144175485e-08, + "loss": 0.7041, + "step": 41534 + }, + { + "epoch": 1.52, + "learning_rate": 4.807924707950342e-08, + "loss": 0.6846, + "step": 41535 + }, + { + "epoch": 1.52, + "learning_rate": 4.807558501483136e-08, + "loss": 0.6889, + "step": 41536 + }, + { + "epoch": 1.52, + "learning_rate": 4.80719229501593e-08, + "loss": 0.6776, + "step": 41537 + }, + { + "epoch": 1.52, + "learning_rate": 4.806826088548723e-08, + "loss": 0.7046, + "step": 41538 + }, + { + "epoch": 1.52, + "learning_rate": 4.806459882081517e-08, + "loss": 0.7002, + "step": 41539 + }, + { + "epoch": 1.52, + "learning_rate": 4.806093675614311e-08, + "loss": 0.6902, + "step": 41540 + }, + { + "epoch": 1.52, + "learning_rate": 4.805727469147105e-08, + "loss": 0.6872, + "step": 41541 + }, + { + "epoch": 1.52, + "learning_rate": 4.805361262679899e-08, + "loss": 0.6975, + "step": 41542 + }, + { + "epoch": 1.52, + "learning_rate": 4.8049950562126924e-08, + "loss": 0.7004, + "step": 41543 + }, + { + "epoch": 1.52, + "learning_rate": 4.8046288497454864e-08, + "loss": 0.6909, + "step": 41544 + }, + { + "epoch": 1.52, + "learning_rate": 4.8042626432782804e-08, + "loss": 0.6999, + "step": 41545 + }, + { + "epoch": 1.52, + "learning_rate": 4.803896436811074e-08, + "loss": 0.692, + "step": 41546 + }, + { + "epoch": 1.52, + "learning_rate": 4.803530230343867e-08, + "loss": 0.6942, + "step": 41547 + }, + { + "epoch": 1.52, + "learning_rate": 4.803164023876661e-08, + "loss": 0.6947, + "step": 41548 + }, + { + "epoch": 1.52, + "learning_rate": 4.802797817409455e-08, + "loss": 0.6938, + "step": 41549 + }, + { + "epoch": 1.52, + "learning_rate": 4.802431610942249e-08, + "loss": 0.6972, + "step": 41550 + }, + { + "epoch": 1.52, + "learning_rate": 4.802065404475043e-08, + "loss": 0.6932, + "step": 41551 + }, + { + "epoch": 1.52, + "learning_rate": 4.8016991980078364e-08, + "loss": 0.695, + "step": 41552 + }, + { + "epoch": 1.52, + "learning_rate": 4.8013329915406304e-08, + "loss": 0.6903, + "step": 41553 + }, + { + "epoch": 1.52, + "learning_rate": 4.8009667850734244e-08, + "loss": 0.6928, + "step": 41554 + }, + { + "epoch": 1.52, + "learning_rate": 4.8006005786062184e-08, + "loss": 0.6906, + "step": 41555 + }, + { + "epoch": 1.52, + "learning_rate": 4.800234372139012e-08, + "loss": 0.6892, + "step": 41556 + }, + { + "epoch": 1.52, + "learning_rate": 4.799868165671806e-08, + "loss": 0.6989, + "step": 41557 + }, + { + "epoch": 1.52, + "learning_rate": 4.799501959204599e-08, + "loss": 0.7012, + "step": 41558 + }, + { + "epoch": 1.52, + "learning_rate": 4.799135752737393e-08, + "loss": 0.6883, + "step": 41559 + }, + { + "epoch": 1.52, + "learning_rate": 4.798769546270187e-08, + "loss": 0.7034, + "step": 41560 + }, + { + "epoch": 1.52, + "learning_rate": 4.7984033398029803e-08, + "loss": 0.7008, + "step": 41561 + }, + { + "epoch": 1.52, + "learning_rate": 4.7980371333357743e-08, + "loss": 0.6997, + "step": 41562 + }, + { + "epoch": 1.52, + "learning_rate": 4.7976709268685683e-08, + "loss": 0.6841, + "step": 41563 + }, + { + "epoch": 1.52, + "learning_rate": 4.797304720401362e-08, + "loss": 0.6843, + "step": 41564 + }, + { + "epoch": 1.52, + "learning_rate": 4.7969385139341557e-08, + "loss": 0.7027, + "step": 41565 + }, + { + "epoch": 1.52, + "learning_rate": 4.7965723074669497e-08, + "loss": 0.6818, + "step": 41566 + }, + { + "epoch": 1.52, + "learning_rate": 4.7962061009997436e-08, + "loss": 0.6881, + "step": 41567 + }, + { + "epoch": 1.52, + "learning_rate": 4.7958398945325376e-08, + "loss": 0.6902, + "step": 41568 + }, + { + "epoch": 1.52, + "learning_rate": 4.795473688065331e-08, + "loss": 0.6888, + "step": 41569 + }, + { + "epoch": 1.52, + "learning_rate": 4.795107481598124e-08, + "loss": 0.6934, + "step": 41570 + }, + { + "epoch": 1.52, + "learning_rate": 4.794741275130918e-08, + "loss": 0.6994, + "step": 41571 + }, + { + "epoch": 1.52, + "learning_rate": 4.794375068663712e-08, + "loss": 0.6934, + "step": 41572 + }, + { + "epoch": 1.52, + "learning_rate": 4.794008862196506e-08, + "loss": 0.6886, + "step": 41573 + }, + { + "epoch": 1.52, + "learning_rate": 4.7936426557292996e-08, + "loss": 0.6853, + "step": 41574 + }, + { + "epoch": 1.52, + "learning_rate": 4.7932764492620936e-08, + "loss": 0.6821, + "step": 41575 + }, + { + "epoch": 1.52, + "learning_rate": 4.7929102427948876e-08, + "loss": 0.7005, + "step": 41576 + }, + { + "epoch": 1.52, + "learning_rate": 4.7925440363276816e-08, + "loss": 0.7049, + "step": 41577 + }, + { + "epoch": 1.52, + "learning_rate": 4.7921778298604756e-08, + "loss": 0.681, + "step": 41578 + }, + { + "epoch": 1.52, + "learning_rate": 4.791811623393269e-08, + "loss": 0.6893, + "step": 41579 + }, + { + "epoch": 1.52, + "learning_rate": 4.791445416926063e-08, + "loss": 0.7019, + "step": 41580 + }, + { + "epoch": 1.52, + "learning_rate": 4.791079210458856e-08, + "loss": 0.7045, + "step": 41581 + }, + { + "epoch": 1.52, + "learning_rate": 4.79071300399165e-08, + "loss": 0.7025, + "step": 41582 + }, + { + "epoch": 1.52, + "learning_rate": 4.7903467975244436e-08, + "loss": 0.7061, + "step": 41583 + }, + { + "epoch": 1.52, + "learning_rate": 4.7899805910572376e-08, + "loss": 0.697, + "step": 41584 + }, + { + "epoch": 1.52, + "learning_rate": 4.7896143845900315e-08, + "loss": 0.6822, + "step": 41585 + }, + { + "epoch": 1.52, + "learning_rate": 4.7892481781228255e-08, + "loss": 0.7049, + "step": 41586 + }, + { + "epoch": 1.52, + "learning_rate": 4.7888819716556195e-08, + "loss": 0.7074, + "step": 41587 + }, + { + "epoch": 1.52, + "learning_rate": 4.788515765188413e-08, + "loss": 0.7023, + "step": 41588 + }, + { + "epoch": 1.52, + "learning_rate": 4.788149558721207e-08, + "loss": 0.6868, + "step": 41589 + }, + { + "epoch": 1.52, + "learning_rate": 4.787783352254001e-08, + "loss": 0.6858, + "step": 41590 + }, + { + "epoch": 1.52, + "learning_rate": 4.787417145786795e-08, + "loss": 0.7075, + "step": 41591 + }, + { + "epoch": 1.52, + "learning_rate": 4.787050939319588e-08, + "loss": 0.6822, + "step": 41592 + }, + { + "epoch": 1.52, + "learning_rate": 4.786684732852382e-08, + "loss": 0.6935, + "step": 41593 + }, + { + "epoch": 1.52, + "learning_rate": 4.7863185263851755e-08, + "loss": 0.6999, + "step": 41594 + }, + { + "epoch": 1.52, + "learning_rate": 4.7859523199179695e-08, + "loss": 0.6962, + "step": 41595 + }, + { + "epoch": 1.52, + "learning_rate": 4.7855861134507635e-08, + "loss": 0.7038, + "step": 41596 + }, + { + "epoch": 1.52, + "learning_rate": 4.785219906983557e-08, + "loss": 0.6878, + "step": 41597 + }, + { + "epoch": 1.52, + "learning_rate": 4.784853700516351e-08, + "loss": 0.6929, + "step": 41598 + }, + { + "epoch": 1.52, + "learning_rate": 4.784487494049145e-08, + "loss": 0.702, + "step": 41599 + }, + { + "epoch": 1.52, + "learning_rate": 4.784121287581939e-08, + "loss": 0.6968, + "step": 41600 + }, + { + "epoch": 1.52, + "learning_rate": 4.783755081114732e-08, + "loss": 0.6996, + "step": 41601 + }, + { + "epoch": 1.52, + "learning_rate": 4.783388874647526e-08, + "loss": 0.7076, + "step": 41602 + }, + { + "epoch": 1.52, + "learning_rate": 4.78302266818032e-08, + "loss": 0.6767, + "step": 41603 + }, + { + "epoch": 1.52, + "learning_rate": 4.7826564617131134e-08, + "loss": 0.6905, + "step": 41604 + }, + { + "epoch": 1.52, + "learning_rate": 4.7822902552459074e-08, + "loss": 0.6869, + "step": 41605 + }, + { + "epoch": 1.52, + "learning_rate": 4.781924048778701e-08, + "loss": 0.6992, + "step": 41606 + }, + { + "epoch": 1.52, + "learning_rate": 4.781557842311495e-08, + "loss": 0.6985, + "step": 41607 + }, + { + "epoch": 1.52, + "learning_rate": 4.781191635844289e-08, + "loss": 0.6951, + "step": 41608 + }, + { + "epoch": 1.52, + "learning_rate": 4.780825429377083e-08, + "loss": 0.6932, + "step": 41609 + }, + { + "epoch": 1.52, + "learning_rate": 4.780459222909876e-08, + "loss": 0.703, + "step": 41610 + }, + { + "epoch": 1.52, + "learning_rate": 4.78009301644267e-08, + "loss": 0.6995, + "step": 41611 + }, + { + "epoch": 1.52, + "learning_rate": 4.779726809975464e-08, + "loss": 0.6866, + "step": 41612 + }, + { + "epoch": 1.52, + "learning_rate": 4.779360603508258e-08, + "loss": 0.6944, + "step": 41613 + }, + { + "epoch": 1.52, + "learning_rate": 4.778994397041052e-08, + "loss": 0.7012, + "step": 41614 + }, + { + "epoch": 1.52, + "learning_rate": 4.7786281905738454e-08, + "loss": 0.6993, + "step": 41615 + }, + { + "epoch": 1.52, + "learning_rate": 4.7782619841066394e-08, + "loss": 0.6947, + "step": 41616 + }, + { + "epoch": 1.52, + "learning_rate": 4.777895777639433e-08, + "loss": 0.7013, + "step": 41617 + }, + { + "epoch": 1.52, + "learning_rate": 4.777529571172227e-08, + "loss": 0.7002, + "step": 41618 + }, + { + "epoch": 1.52, + "learning_rate": 4.77716336470502e-08, + "loss": 0.6958, + "step": 41619 + }, + { + "epoch": 1.52, + "learning_rate": 4.776797158237814e-08, + "loss": 0.6906, + "step": 41620 + }, + { + "epoch": 1.52, + "learning_rate": 4.776430951770608e-08, + "loss": 0.6967, + "step": 41621 + }, + { + "epoch": 1.52, + "learning_rate": 4.776064745303402e-08, + "loss": 0.6954, + "step": 41622 + }, + { + "epoch": 1.52, + "learning_rate": 4.775698538836196e-08, + "loss": 0.6904, + "step": 41623 + }, + { + "epoch": 1.52, + "learning_rate": 4.775332332368989e-08, + "loss": 0.6979, + "step": 41624 + }, + { + "epoch": 1.52, + "learning_rate": 4.774966125901783e-08, + "loss": 0.6932, + "step": 41625 + }, + { + "epoch": 1.52, + "learning_rate": 4.774599919434577e-08, + "loss": 0.7029, + "step": 41626 + }, + { + "epoch": 1.52, + "learning_rate": 4.774233712967371e-08, + "loss": 0.6838, + "step": 41627 + }, + { + "epoch": 1.52, + "learning_rate": 4.7738675065001646e-08, + "loss": 0.6888, + "step": 41628 + }, + { + "epoch": 1.52, + "learning_rate": 4.773501300032958e-08, + "loss": 0.6909, + "step": 41629 + }, + { + "epoch": 1.52, + "learning_rate": 4.773135093565752e-08, + "loss": 0.7181, + "step": 41630 + }, + { + "epoch": 1.52, + "learning_rate": 4.772768887098546e-08, + "loss": 0.6988, + "step": 41631 + }, + { + "epoch": 1.52, + "learning_rate": 4.77240268063134e-08, + "loss": 0.6923, + "step": 41632 + }, + { + "epoch": 1.52, + "learning_rate": 4.772036474164133e-08, + "loss": 0.6911, + "step": 41633 + }, + { + "epoch": 1.52, + "learning_rate": 4.771670267696927e-08, + "loss": 0.6898, + "step": 41634 + }, + { + "epoch": 1.52, + "learning_rate": 4.771304061229721e-08, + "loss": 0.705, + "step": 41635 + }, + { + "epoch": 1.52, + "learning_rate": 4.770937854762515e-08, + "loss": 0.6915, + "step": 41636 + }, + { + "epoch": 1.52, + "learning_rate": 4.7705716482953086e-08, + "loss": 0.6923, + "step": 41637 + }, + { + "epoch": 1.52, + "learning_rate": 4.7702054418281026e-08, + "loss": 0.6934, + "step": 41638 + }, + { + "epoch": 1.52, + "learning_rate": 4.7698392353608966e-08, + "loss": 0.6981, + "step": 41639 + }, + { + "epoch": 1.52, + "learning_rate": 4.76947302889369e-08, + "loss": 0.6895, + "step": 41640 + }, + { + "epoch": 1.52, + "learning_rate": 4.769106822426484e-08, + "loss": 0.6809, + "step": 41641 + }, + { + "epoch": 1.52, + "learning_rate": 4.768740615959277e-08, + "loss": 0.6859, + "step": 41642 + }, + { + "epoch": 1.52, + "learning_rate": 4.768374409492071e-08, + "loss": 0.6864, + "step": 41643 + }, + { + "epoch": 1.52, + "learning_rate": 4.768008203024865e-08, + "loss": 0.7002, + "step": 41644 + }, + { + "epoch": 1.52, + "learning_rate": 4.767641996557659e-08, + "loss": 0.6833, + "step": 41645 + }, + { + "epoch": 1.52, + "learning_rate": 4.7672757900904525e-08, + "loss": 0.7027, + "step": 41646 + }, + { + "epoch": 1.52, + "learning_rate": 4.7669095836232465e-08, + "loss": 0.6934, + "step": 41647 + }, + { + "epoch": 1.52, + "learning_rate": 4.7665433771560405e-08, + "loss": 0.6921, + "step": 41648 + }, + { + "epoch": 1.52, + "learning_rate": 4.7661771706888345e-08, + "loss": 0.702, + "step": 41649 + }, + { + "epoch": 1.52, + "learning_rate": 4.7658109642216285e-08, + "loss": 0.687, + "step": 41650 + }, + { + "epoch": 1.52, + "learning_rate": 4.765444757754422e-08, + "loss": 0.6845, + "step": 41651 + }, + { + "epoch": 1.52, + "learning_rate": 4.765078551287215e-08, + "loss": 0.6875, + "step": 41652 + }, + { + "epoch": 1.52, + "learning_rate": 4.764712344820009e-08, + "loss": 0.6924, + "step": 41653 + }, + { + "epoch": 1.52, + "learning_rate": 4.764346138352803e-08, + "loss": 0.6852, + "step": 41654 + }, + { + "epoch": 1.52, + "learning_rate": 4.7639799318855965e-08, + "loss": 0.6969, + "step": 41655 + }, + { + "epoch": 1.52, + "learning_rate": 4.7636137254183905e-08, + "loss": 0.7042, + "step": 41656 + }, + { + "epoch": 1.52, + "learning_rate": 4.7632475189511845e-08, + "loss": 0.6885, + "step": 41657 + }, + { + "epoch": 1.52, + "learning_rate": 4.7628813124839785e-08, + "loss": 0.7039, + "step": 41658 + }, + { + "epoch": 1.52, + "learning_rate": 4.7625151060167725e-08, + "loss": 0.6824, + "step": 41659 + }, + { + "epoch": 1.52, + "learning_rate": 4.762148899549566e-08, + "loss": 0.697, + "step": 41660 + }, + { + "epoch": 1.52, + "learning_rate": 4.76178269308236e-08, + "loss": 0.6791, + "step": 41661 + }, + { + "epoch": 1.52, + "learning_rate": 4.761416486615154e-08, + "loss": 0.7041, + "step": 41662 + }, + { + "epoch": 1.52, + "learning_rate": 4.761050280147947e-08, + "loss": 0.6959, + "step": 41663 + }, + { + "epoch": 1.52, + "learning_rate": 4.7606840736807404e-08, + "loss": 0.6899, + "step": 41664 + }, + { + "epoch": 1.52, + "learning_rate": 4.7603178672135344e-08, + "loss": 0.6822, + "step": 41665 + }, + { + "epoch": 1.52, + "learning_rate": 4.7599516607463284e-08, + "loss": 0.6978, + "step": 41666 + }, + { + "epoch": 1.52, + "learning_rate": 4.7595854542791224e-08, + "loss": 0.6979, + "step": 41667 + }, + { + "epoch": 1.52, + "learning_rate": 4.7592192478119164e-08, + "loss": 0.6892, + "step": 41668 + }, + { + "epoch": 1.52, + "learning_rate": 4.75885304134471e-08, + "loss": 0.6987, + "step": 41669 + }, + { + "epoch": 1.52, + "learning_rate": 4.758486834877504e-08, + "loss": 0.6891, + "step": 41670 + }, + { + "epoch": 1.52, + "learning_rate": 4.758120628410298e-08, + "loss": 0.6919, + "step": 41671 + }, + { + "epoch": 1.52, + "learning_rate": 4.757754421943092e-08, + "loss": 0.7002, + "step": 41672 + }, + { + "epoch": 1.52, + "learning_rate": 4.757388215475885e-08, + "loss": 0.6938, + "step": 41673 + }, + { + "epoch": 1.52, + "learning_rate": 4.757022009008679e-08, + "loss": 0.691, + "step": 41674 + }, + { + "epoch": 1.52, + "learning_rate": 4.7566558025414724e-08, + "loss": 0.681, + "step": 41675 + }, + { + "epoch": 1.52, + "learning_rate": 4.7562895960742664e-08, + "loss": 0.6954, + "step": 41676 + }, + { + "epoch": 1.52, + "learning_rate": 4.7559233896070604e-08, + "loss": 0.6955, + "step": 41677 + }, + { + "epoch": 1.52, + "learning_rate": 4.755557183139854e-08, + "loss": 0.7038, + "step": 41678 + }, + { + "epoch": 1.52, + "learning_rate": 4.755190976672648e-08, + "loss": 0.6942, + "step": 41679 + }, + { + "epoch": 1.52, + "learning_rate": 4.754824770205442e-08, + "loss": 0.6868, + "step": 41680 + }, + { + "epoch": 1.52, + "learning_rate": 4.7544585637382357e-08, + "loss": 0.6946, + "step": 41681 + }, + { + "epoch": 1.53, + "learning_rate": 4.754092357271029e-08, + "loss": 0.6961, + "step": 41682 + }, + { + "epoch": 1.53, + "learning_rate": 4.753726150803823e-08, + "loss": 0.6986, + "step": 41683 + }, + { + "epoch": 1.53, + "learning_rate": 4.753359944336617e-08, + "loss": 0.7039, + "step": 41684 + }, + { + "epoch": 1.53, + "learning_rate": 4.752993737869411e-08, + "loss": 0.7025, + "step": 41685 + }, + { + "epoch": 1.53, + "learning_rate": 4.752627531402204e-08, + "loss": 0.703, + "step": 41686 + }, + { + "epoch": 1.53, + "learning_rate": 4.752261324934998e-08, + "loss": 0.7015, + "step": 41687 + }, + { + "epoch": 1.53, + "learning_rate": 4.7518951184677916e-08, + "loss": 0.6859, + "step": 41688 + }, + { + "epoch": 1.53, + "learning_rate": 4.7515289120005856e-08, + "loss": 0.6914, + "step": 41689 + }, + { + "epoch": 1.53, + "learning_rate": 4.7511627055333796e-08, + "loss": 0.6888, + "step": 41690 + }, + { + "epoch": 1.53, + "learning_rate": 4.750796499066173e-08, + "loss": 0.7039, + "step": 41691 + }, + { + "epoch": 1.53, + "learning_rate": 4.750430292598967e-08, + "loss": 0.7021, + "step": 41692 + }, + { + "epoch": 1.53, + "learning_rate": 4.750064086131761e-08, + "loss": 0.6862, + "step": 41693 + }, + { + "epoch": 1.53, + "learning_rate": 4.749697879664555e-08, + "loss": 0.6976, + "step": 41694 + }, + { + "epoch": 1.53, + "learning_rate": 4.749331673197348e-08, + "loss": 0.6883, + "step": 41695 + }, + { + "epoch": 1.53, + "learning_rate": 4.748965466730142e-08, + "loss": 0.6996, + "step": 41696 + }, + { + "epoch": 1.53, + "learning_rate": 4.748599260262936e-08, + "loss": 0.6871, + "step": 41697 + }, + { + "epoch": 1.53, + "learning_rate": 4.74823305379573e-08, + "loss": 0.7077, + "step": 41698 + }, + { + "epoch": 1.53, + "learning_rate": 4.7478668473285236e-08, + "loss": 0.6826, + "step": 41699 + }, + { + "epoch": 1.53, + "learning_rate": 4.747500640861317e-08, + "loss": 0.6878, + "step": 41700 + }, + { + "epoch": 1.53, + "learning_rate": 4.747134434394111e-08, + "loss": 0.7071, + "step": 41701 + }, + { + "epoch": 1.53, + "learning_rate": 4.746768227926905e-08, + "loss": 0.7037, + "step": 41702 + }, + { + "epoch": 1.53, + "learning_rate": 4.746402021459699e-08, + "loss": 0.6984, + "step": 41703 + }, + { + "epoch": 1.53, + "learning_rate": 4.746035814992492e-08, + "loss": 0.7129, + "step": 41704 + }, + { + "epoch": 1.53, + "learning_rate": 4.745669608525286e-08, + "loss": 0.6915, + "step": 41705 + }, + { + "epoch": 1.53, + "learning_rate": 4.74530340205808e-08, + "loss": 0.6957, + "step": 41706 + }, + { + "epoch": 1.53, + "learning_rate": 4.744937195590874e-08, + "loss": 0.6913, + "step": 41707 + }, + { + "epoch": 1.53, + "learning_rate": 4.744570989123668e-08, + "loss": 0.6936, + "step": 41708 + }, + { + "epoch": 1.53, + "learning_rate": 4.7442047826564615e-08, + "loss": 0.6847, + "step": 41709 + }, + { + "epoch": 1.53, + "learning_rate": 4.7438385761892555e-08, + "loss": 0.6871, + "step": 41710 + }, + { + "epoch": 1.53, + "learning_rate": 4.743472369722049e-08, + "loss": 0.6885, + "step": 41711 + }, + { + "epoch": 1.53, + "learning_rate": 4.743106163254843e-08, + "loss": 0.6888, + "step": 41712 + }, + { + "epoch": 1.53, + "learning_rate": 4.742739956787636e-08, + "loss": 0.6889, + "step": 41713 + }, + { + "epoch": 1.53, + "learning_rate": 4.74237375032043e-08, + "loss": 0.6894, + "step": 41714 + }, + { + "epoch": 1.53, + "learning_rate": 4.742007543853224e-08, + "loss": 0.6973, + "step": 41715 + }, + { + "epoch": 1.53, + "learning_rate": 4.741641337386018e-08, + "loss": 0.6901, + "step": 41716 + }, + { + "epoch": 1.53, + "learning_rate": 4.741275130918812e-08, + "loss": 0.6987, + "step": 41717 + }, + { + "epoch": 1.53, + "learning_rate": 4.7409089244516055e-08, + "loss": 0.6996, + "step": 41718 + }, + { + "epoch": 1.53, + "learning_rate": 4.7405427179843995e-08, + "loss": 0.7013, + "step": 41719 + }, + { + "epoch": 1.53, + "learning_rate": 4.7401765115171934e-08, + "loss": 0.6911, + "step": 41720 + }, + { + "epoch": 1.53, + "learning_rate": 4.7398103050499874e-08, + "loss": 0.7128, + "step": 41721 + }, + { + "epoch": 1.53, + "learning_rate": 4.739444098582781e-08, + "loss": 0.6855, + "step": 41722 + }, + { + "epoch": 1.53, + "learning_rate": 4.739077892115574e-08, + "loss": 0.6957, + "step": 41723 + }, + { + "epoch": 1.53, + "learning_rate": 4.738711685648368e-08, + "loss": 0.6855, + "step": 41724 + }, + { + "epoch": 1.53, + "learning_rate": 4.738345479181162e-08, + "loss": 0.6887, + "step": 41725 + }, + { + "epoch": 1.53, + "learning_rate": 4.737979272713956e-08, + "loss": 0.6981, + "step": 41726 + }, + { + "epoch": 1.53, + "learning_rate": 4.7376130662467494e-08, + "loss": 0.6778, + "step": 41727 + }, + { + "epoch": 1.53, + "learning_rate": 4.7372468597795434e-08, + "loss": 0.6904, + "step": 41728 + }, + { + "epoch": 1.53, + "learning_rate": 4.7368806533123374e-08, + "loss": 0.6997, + "step": 41729 + }, + { + "epoch": 1.53, + "learning_rate": 4.7365144468451314e-08, + "loss": 0.6997, + "step": 41730 + }, + { + "epoch": 1.53, + "learning_rate": 4.736148240377925e-08, + "loss": 0.7061, + "step": 41731 + }, + { + "epoch": 1.53, + "learning_rate": 4.735782033910719e-08, + "loss": 0.6963, + "step": 41732 + }, + { + "epoch": 1.53, + "learning_rate": 4.735415827443513e-08, + "loss": 0.701, + "step": 41733 + }, + { + "epoch": 1.53, + "learning_rate": 4.735049620976306e-08, + "loss": 0.6891, + "step": 41734 + }, + { + "epoch": 1.53, + "learning_rate": 4.7346834145091e-08, + "loss": 0.691, + "step": 41735 + }, + { + "epoch": 1.53, + "learning_rate": 4.7343172080418934e-08, + "loss": 0.6908, + "step": 41736 + }, + { + "epoch": 1.53, + "learning_rate": 4.7339510015746874e-08, + "loss": 0.6977, + "step": 41737 + }, + { + "epoch": 1.53, + "learning_rate": 4.7335847951074813e-08, + "loss": 0.6898, + "step": 41738 + }, + { + "epoch": 1.53, + "learning_rate": 4.7332185886402753e-08, + "loss": 0.701, + "step": 41739 + }, + { + "epoch": 1.53, + "learning_rate": 4.732852382173069e-08, + "loss": 0.6908, + "step": 41740 + }, + { + "epoch": 1.53, + "learning_rate": 4.7324861757058627e-08, + "loss": 0.6931, + "step": 41741 + }, + { + "epoch": 1.53, + "learning_rate": 4.7321199692386567e-08, + "loss": 0.6916, + "step": 41742 + }, + { + "epoch": 1.53, + "learning_rate": 4.7317537627714507e-08, + "loss": 0.6922, + "step": 41743 + }, + { + "epoch": 1.53, + "learning_rate": 4.7313875563042446e-08, + "loss": 0.6929, + "step": 41744 + }, + { + "epoch": 1.53, + "learning_rate": 4.731021349837038e-08, + "loss": 0.6968, + "step": 41745 + }, + { + "epoch": 1.53, + "learning_rate": 4.730655143369831e-08, + "loss": 0.6834, + "step": 41746 + }, + { + "epoch": 1.53, + "learning_rate": 4.730288936902625e-08, + "loss": 0.6925, + "step": 41747 + }, + { + "epoch": 1.53, + "learning_rate": 4.729922730435419e-08, + "loss": 0.6924, + "step": 41748 + }, + { + "epoch": 1.53, + "learning_rate": 4.7295565239682126e-08, + "loss": 0.6906, + "step": 41749 + }, + { + "epoch": 1.53, + "learning_rate": 4.7291903175010066e-08, + "loss": 0.6888, + "step": 41750 + }, + { + "epoch": 1.53, + "learning_rate": 4.7288241110338006e-08, + "loss": 0.6978, + "step": 41751 + }, + { + "epoch": 1.53, + "learning_rate": 4.7284579045665946e-08, + "loss": 0.6797, + "step": 41752 + }, + { + "epoch": 1.53, + "learning_rate": 4.7280916980993886e-08, + "loss": 0.6907, + "step": 41753 + }, + { + "epoch": 1.53, + "learning_rate": 4.727725491632182e-08, + "loss": 0.6884, + "step": 41754 + }, + { + "epoch": 1.53, + "learning_rate": 4.727359285164976e-08, + "loss": 0.6898, + "step": 41755 + }, + { + "epoch": 1.53, + "learning_rate": 4.72699307869777e-08, + "loss": 0.6952, + "step": 41756 + }, + { + "epoch": 1.53, + "learning_rate": 4.726626872230563e-08, + "loss": 0.7073, + "step": 41757 + }, + { + "epoch": 1.53, + "learning_rate": 4.726260665763357e-08, + "loss": 0.6935, + "step": 41758 + }, + { + "epoch": 1.53, + "learning_rate": 4.7258944592961506e-08, + "loss": 0.6979, + "step": 41759 + }, + { + "epoch": 1.53, + "learning_rate": 4.7255282528289446e-08, + "loss": 0.7045, + "step": 41760 + }, + { + "epoch": 1.53, + "learning_rate": 4.7251620463617386e-08, + "loss": 0.6912, + "step": 41761 + }, + { + "epoch": 1.53, + "learning_rate": 4.7247958398945325e-08, + "loss": 0.7034, + "step": 41762 + }, + { + "epoch": 1.53, + "learning_rate": 4.724429633427326e-08, + "loss": 0.6912, + "step": 41763 + }, + { + "epoch": 1.53, + "learning_rate": 4.72406342696012e-08, + "loss": 0.6978, + "step": 41764 + }, + { + "epoch": 1.53, + "learning_rate": 4.723697220492914e-08, + "loss": 0.6971, + "step": 41765 + }, + { + "epoch": 1.53, + "learning_rate": 4.723331014025708e-08, + "loss": 0.6808, + "step": 41766 + }, + { + "epoch": 1.53, + "learning_rate": 4.722964807558501e-08, + "loss": 0.708, + "step": 41767 + }, + { + "epoch": 1.53, + "learning_rate": 4.722598601091295e-08, + "loss": 0.682, + "step": 41768 + }, + { + "epoch": 1.53, + "learning_rate": 4.722232394624089e-08, + "loss": 0.6927, + "step": 41769 + }, + { + "epoch": 1.53, + "learning_rate": 4.7218661881568825e-08, + "loss": 0.6833, + "step": 41770 + }, + { + "epoch": 1.53, + "learning_rate": 4.7214999816896765e-08, + "loss": 0.6934, + "step": 41771 + }, + { + "epoch": 1.53, + "learning_rate": 4.72113377522247e-08, + "loss": 0.6984, + "step": 41772 + }, + { + "epoch": 1.53, + "learning_rate": 4.720767568755264e-08, + "loss": 0.7023, + "step": 41773 + }, + { + "epoch": 1.53, + "learning_rate": 4.720401362288058e-08, + "loss": 0.6956, + "step": 41774 + }, + { + "epoch": 1.53, + "learning_rate": 4.720035155820852e-08, + "loss": 0.6928, + "step": 41775 + }, + { + "epoch": 1.53, + "learning_rate": 4.719668949353645e-08, + "loss": 0.6994, + "step": 41776 + }, + { + "epoch": 1.53, + "learning_rate": 4.719302742886439e-08, + "loss": 0.69, + "step": 41777 + }, + { + "epoch": 1.53, + "learning_rate": 4.718936536419233e-08, + "loss": 0.7078, + "step": 41778 + }, + { + "epoch": 1.53, + "learning_rate": 4.718570329952027e-08, + "loss": 0.6946, + "step": 41779 + }, + { + "epoch": 1.53, + "learning_rate": 4.7182041234848204e-08, + "loss": 0.7071, + "step": 41780 + }, + { + "epoch": 1.53, + "learning_rate": 4.7178379170176144e-08, + "loss": 0.6902, + "step": 41781 + }, + { + "epoch": 1.53, + "learning_rate": 4.717471710550408e-08, + "loss": 0.6903, + "step": 41782 + }, + { + "epoch": 1.53, + "learning_rate": 4.717105504083202e-08, + "loss": 0.6874, + "step": 41783 + }, + { + "epoch": 1.53, + "learning_rate": 4.716739297615996e-08, + "loss": 0.6894, + "step": 41784 + }, + { + "epoch": 1.53, + "learning_rate": 4.716373091148789e-08, + "loss": 0.6951, + "step": 41785 + }, + { + "epoch": 1.53, + "learning_rate": 4.716006884681583e-08, + "loss": 0.6918, + "step": 41786 + }, + { + "epoch": 1.53, + "learning_rate": 4.715640678214377e-08, + "loss": 0.6973, + "step": 41787 + }, + { + "epoch": 1.53, + "learning_rate": 4.715274471747171e-08, + "loss": 0.6953, + "step": 41788 + }, + { + "epoch": 1.53, + "learning_rate": 4.714908265279965e-08, + "loss": 0.6863, + "step": 41789 + }, + { + "epoch": 1.53, + "learning_rate": 4.7145420588127584e-08, + "loss": 0.6843, + "step": 41790 + }, + { + "epoch": 1.53, + "learning_rate": 4.7141758523455524e-08, + "loss": 0.6883, + "step": 41791 + }, + { + "epoch": 1.53, + "learning_rate": 4.7138096458783464e-08, + "loss": 0.6971, + "step": 41792 + }, + { + "epoch": 1.53, + "learning_rate": 4.71344343941114e-08, + "loss": 0.6989, + "step": 41793 + }, + { + "epoch": 1.53, + "learning_rate": 4.713077232943933e-08, + "loss": 0.6917, + "step": 41794 + }, + { + "epoch": 1.53, + "learning_rate": 4.712711026476727e-08, + "loss": 0.6889, + "step": 41795 + }, + { + "epoch": 1.53, + "learning_rate": 4.712344820009521e-08, + "loss": 0.7042, + "step": 41796 + }, + { + "epoch": 1.53, + "learning_rate": 4.711978613542315e-08, + "loss": 0.6878, + "step": 41797 + }, + { + "epoch": 1.53, + "learning_rate": 4.711612407075109e-08, + "loss": 0.6974, + "step": 41798 + }, + { + "epoch": 1.53, + "learning_rate": 4.7112462006079023e-08, + "loss": 0.69, + "step": 41799 + }, + { + "epoch": 1.53, + "learning_rate": 4.7108799941406963e-08, + "loss": 0.6887, + "step": 41800 + }, + { + "epoch": 1.53, + "learning_rate": 4.71051378767349e-08, + "loss": 0.6853, + "step": 41801 + }, + { + "epoch": 1.53, + "learning_rate": 4.710147581206284e-08, + "loss": 0.6866, + "step": 41802 + }, + { + "epoch": 1.53, + "learning_rate": 4.7097813747390777e-08, + "loss": 0.7047, + "step": 41803 + }, + { + "epoch": 1.53, + "learning_rate": 4.7094151682718716e-08, + "loss": 0.6967, + "step": 41804 + }, + { + "epoch": 1.53, + "learning_rate": 4.709048961804665e-08, + "loss": 0.6935, + "step": 41805 + }, + { + "epoch": 1.53, + "learning_rate": 4.708682755337459e-08, + "loss": 0.6943, + "step": 41806 + }, + { + "epoch": 1.53, + "learning_rate": 4.708316548870253e-08, + "loss": 0.6952, + "step": 41807 + }, + { + "epoch": 1.53, + "learning_rate": 4.707950342403046e-08, + "loss": 0.6787, + "step": 41808 + }, + { + "epoch": 1.53, + "learning_rate": 4.70758413593584e-08, + "loss": 0.6867, + "step": 41809 + }, + { + "epoch": 1.53, + "learning_rate": 4.707217929468634e-08, + "loss": 0.7077, + "step": 41810 + }, + { + "epoch": 1.53, + "learning_rate": 4.706851723001428e-08, + "loss": 0.7051, + "step": 41811 + }, + { + "epoch": 1.53, + "learning_rate": 4.7064855165342216e-08, + "loss": 0.7072, + "step": 41812 + }, + { + "epoch": 1.53, + "learning_rate": 4.7061193100670156e-08, + "loss": 0.6953, + "step": 41813 + }, + { + "epoch": 1.53, + "learning_rate": 4.7057531035998096e-08, + "loss": 0.6984, + "step": 41814 + }, + { + "epoch": 1.53, + "learning_rate": 4.7053868971326036e-08, + "loss": 0.6898, + "step": 41815 + }, + { + "epoch": 1.53, + "learning_rate": 4.705020690665397e-08, + "loss": 0.6966, + "step": 41816 + }, + { + "epoch": 1.53, + "learning_rate": 4.70465448419819e-08, + "loss": 0.6884, + "step": 41817 + }, + { + "epoch": 1.53, + "learning_rate": 4.704288277730984e-08, + "loss": 0.696, + "step": 41818 + }, + { + "epoch": 1.53, + "learning_rate": 4.703922071263778e-08, + "loss": 0.6988, + "step": 41819 + }, + { + "epoch": 1.53, + "learning_rate": 4.703555864796572e-08, + "loss": 0.6945, + "step": 41820 + }, + { + "epoch": 1.53, + "learning_rate": 4.7031896583293656e-08, + "loss": 0.7262, + "step": 41821 + }, + { + "epoch": 1.53, + "learning_rate": 4.7028234518621595e-08, + "loss": 0.6992, + "step": 41822 + }, + { + "epoch": 1.53, + "learning_rate": 4.7024572453949535e-08, + "loss": 0.681, + "step": 41823 + }, + { + "epoch": 1.53, + "learning_rate": 4.7020910389277475e-08, + "loss": 0.6999, + "step": 41824 + }, + { + "epoch": 1.53, + "learning_rate": 4.7017248324605415e-08, + "loss": 0.695, + "step": 41825 + }, + { + "epoch": 1.53, + "learning_rate": 4.701358625993335e-08, + "loss": 0.6937, + "step": 41826 + }, + { + "epoch": 1.53, + "learning_rate": 4.700992419526129e-08, + "loss": 0.7023, + "step": 41827 + }, + { + "epoch": 1.53, + "learning_rate": 4.700626213058922e-08, + "loss": 0.6915, + "step": 41828 + }, + { + "epoch": 1.53, + "learning_rate": 4.700260006591716e-08, + "loss": 0.6992, + "step": 41829 + }, + { + "epoch": 1.53, + "learning_rate": 4.6998938001245095e-08, + "loss": 0.6976, + "step": 41830 + }, + { + "epoch": 1.53, + "learning_rate": 4.6995275936573035e-08, + "loss": 0.6811, + "step": 41831 + }, + { + "epoch": 1.53, + "learning_rate": 4.6991613871900975e-08, + "loss": 0.7064, + "step": 41832 + }, + { + "epoch": 1.53, + "learning_rate": 4.6987951807228915e-08, + "loss": 0.7013, + "step": 41833 + }, + { + "epoch": 1.53, + "learning_rate": 4.6984289742556855e-08, + "loss": 0.6833, + "step": 41834 + }, + { + "epoch": 1.53, + "learning_rate": 4.698062767788479e-08, + "loss": 0.6983, + "step": 41835 + }, + { + "epoch": 1.53, + "learning_rate": 4.697696561321273e-08, + "loss": 0.6983, + "step": 41836 + }, + { + "epoch": 1.53, + "learning_rate": 4.697330354854067e-08, + "loss": 0.6898, + "step": 41837 + }, + { + "epoch": 1.53, + "learning_rate": 4.696964148386861e-08, + "loss": 0.6958, + "step": 41838 + }, + { + "epoch": 1.53, + "learning_rate": 4.696597941919654e-08, + "loss": 0.7015, + "step": 41839 + }, + { + "epoch": 1.53, + "learning_rate": 4.6962317354524474e-08, + "loss": 0.696, + "step": 41840 + }, + { + "epoch": 1.53, + "learning_rate": 4.6958655289852414e-08, + "loss": 0.6928, + "step": 41841 + }, + { + "epoch": 1.53, + "learning_rate": 4.6954993225180354e-08, + "loss": 0.695, + "step": 41842 + }, + { + "epoch": 1.53, + "learning_rate": 4.6951331160508294e-08, + "loss": 0.6954, + "step": 41843 + }, + { + "epoch": 1.53, + "learning_rate": 4.694766909583623e-08, + "loss": 0.7044, + "step": 41844 + }, + { + "epoch": 1.53, + "learning_rate": 4.694400703116417e-08, + "loss": 0.6859, + "step": 41845 + }, + { + "epoch": 1.53, + "learning_rate": 4.694034496649211e-08, + "loss": 0.7117, + "step": 41846 + }, + { + "epoch": 1.53, + "learning_rate": 4.693668290182005e-08, + "loss": 0.6907, + "step": 41847 + }, + { + "epoch": 1.53, + "learning_rate": 4.693302083714798e-08, + "loss": 0.6998, + "step": 41848 + }, + { + "epoch": 1.53, + "learning_rate": 4.692935877247592e-08, + "loss": 0.6952, + "step": 41849 + }, + { + "epoch": 1.53, + "learning_rate": 4.692569670780386e-08, + "loss": 0.6961, + "step": 41850 + }, + { + "epoch": 1.53, + "learning_rate": 4.6922034643131794e-08, + "loss": 0.6972, + "step": 41851 + }, + { + "epoch": 1.53, + "learning_rate": 4.6918372578459734e-08, + "loss": 0.6955, + "step": 41852 + }, + { + "epoch": 1.53, + "learning_rate": 4.691471051378767e-08, + "loss": 0.6986, + "step": 41853 + }, + { + "epoch": 1.53, + "learning_rate": 4.691104844911561e-08, + "loss": 0.6925, + "step": 41854 + }, + { + "epoch": 1.53, + "learning_rate": 4.690738638444355e-08, + "loss": 0.6996, + "step": 41855 + }, + { + "epoch": 1.53, + "learning_rate": 4.690372431977149e-08, + "loss": 0.6952, + "step": 41856 + }, + { + "epoch": 1.53, + "learning_rate": 4.690006225509942e-08, + "loss": 0.6942, + "step": 41857 + }, + { + "epoch": 1.53, + "learning_rate": 4.689640019042736e-08, + "loss": 0.7062, + "step": 41858 + }, + { + "epoch": 1.53, + "learning_rate": 4.68927381257553e-08, + "loss": 0.6978, + "step": 41859 + }, + { + "epoch": 1.53, + "learning_rate": 4.688907606108324e-08, + "loss": 0.6846, + "step": 41860 + }, + { + "epoch": 1.53, + "learning_rate": 4.688541399641118e-08, + "loss": 0.6982, + "step": 41861 + }, + { + "epoch": 1.53, + "learning_rate": 4.688175193173911e-08, + "loss": 0.6841, + "step": 41862 + }, + { + "epoch": 1.53, + "learning_rate": 4.687808986706705e-08, + "loss": 0.6872, + "step": 41863 + }, + { + "epoch": 1.53, + "learning_rate": 4.6874427802394986e-08, + "loss": 0.6829, + "step": 41864 + }, + { + "epoch": 1.53, + "learning_rate": 4.6870765737722926e-08, + "loss": 0.7032, + "step": 41865 + }, + { + "epoch": 1.53, + "learning_rate": 4.686710367305086e-08, + "loss": 0.699, + "step": 41866 + }, + { + "epoch": 1.53, + "learning_rate": 4.68634416083788e-08, + "loss": 0.6873, + "step": 41867 + }, + { + "epoch": 1.53, + "learning_rate": 4.685977954370674e-08, + "loss": 0.7048, + "step": 41868 + }, + { + "epoch": 1.53, + "learning_rate": 4.685611747903468e-08, + "loss": 0.7035, + "step": 41869 + }, + { + "epoch": 1.53, + "learning_rate": 4.685245541436262e-08, + "loss": 0.6874, + "step": 41870 + }, + { + "epoch": 1.53, + "learning_rate": 4.684879334969055e-08, + "loss": 0.6957, + "step": 41871 + }, + { + "epoch": 1.53, + "learning_rate": 4.684513128501849e-08, + "loss": 0.6897, + "step": 41872 + }, + { + "epoch": 1.53, + "learning_rate": 4.684146922034643e-08, + "loss": 0.6968, + "step": 41873 + }, + { + "epoch": 1.53, + "learning_rate": 4.683780715567437e-08, + "loss": 0.6887, + "step": 41874 + }, + { + "epoch": 1.53, + "learning_rate": 4.6834145091002306e-08, + "loss": 0.6758, + "step": 41875 + }, + { + "epoch": 1.53, + "learning_rate": 4.683048302633024e-08, + "loss": 0.7085, + "step": 41876 + }, + { + "epoch": 1.53, + "learning_rate": 4.682682096165818e-08, + "loss": 0.7059, + "step": 41877 + }, + { + "epoch": 1.53, + "learning_rate": 4.682315889698612e-08, + "loss": 0.6981, + "step": 41878 + }, + { + "epoch": 1.53, + "learning_rate": 4.681949683231405e-08, + "loss": 0.7041, + "step": 41879 + }, + { + "epoch": 1.53, + "learning_rate": 4.681583476764199e-08, + "loss": 0.6928, + "step": 41880 + }, + { + "epoch": 1.53, + "learning_rate": 4.681217270296993e-08, + "loss": 0.6968, + "step": 41881 + }, + { + "epoch": 1.53, + "learning_rate": 4.680851063829787e-08, + "loss": 0.6992, + "step": 41882 + }, + { + "epoch": 1.53, + "learning_rate": 4.680484857362581e-08, + "loss": 0.6955, + "step": 41883 + }, + { + "epoch": 1.53, + "learning_rate": 4.6801186508953745e-08, + "loss": 0.6828, + "step": 41884 + }, + { + "epoch": 1.53, + "learning_rate": 4.6797524444281685e-08, + "loss": 0.6857, + "step": 41885 + }, + { + "epoch": 1.53, + "learning_rate": 4.6793862379609625e-08, + "loss": 0.6885, + "step": 41886 + }, + { + "epoch": 1.53, + "learning_rate": 4.679020031493756e-08, + "loss": 0.6964, + "step": 41887 + }, + { + "epoch": 1.53, + "learning_rate": 4.678653825026549e-08, + "loss": 0.6925, + "step": 41888 + }, + { + "epoch": 1.53, + "learning_rate": 4.678287618559343e-08, + "loss": 0.6993, + "step": 41889 + }, + { + "epoch": 1.53, + "learning_rate": 4.677921412092137e-08, + "loss": 0.7001, + "step": 41890 + }, + { + "epoch": 1.53, + "learning_rate": 4.677555205624931e-08, + "loss": 0.6968, + "step": 41891 + }, + { + "epoch": 1.53, + "learning_rate": 4.677188999157725e-08, + "loss": 0.6936, + "step": 41892 + }, + { + "epoch": 1.53, + "learning_rate": 4.6768227926905185e-08, + "loss": 0.6974, + "step": 41893 + }, + { + "epoch": 1.53, + "learning_rate": 4.6764565862233125e-08, + "loss": 0.6919, + "step": 41894 + }, + { + "epoch": 1.53, + "learning_rate": 4.6760903797561065e-08, + "loss": 0.6942, + "step": 41895 + }, + { + "epoch": 1.53, + "learning_rate": 4.6757241732889005e-08, + "loss": 0.6879, + "step": 41896 + }, + { + "epoch": 1.53, + "learning_rate": 4.675357966821694e-08, + "loss": 0.6942, + "step": 41897 + }, + { + "epoch": 1.53, + "learning_rate": 4.674991760354488e-08, + "loss": 0.6908, + "step": 41898 + }, + { + "epoch": 1.53, + "learning_rate": 4.674625553887281e-08, + "loss": 0.6871, + "step": 41899 + }, + { + "epoch": 1.53, + "learning_rate": 4.674259347420075e-08, + "loss": 0.6936, + "step": 41900 + }, + { + "epoch": 1.53, + "learning_rate": 4.673893140952869e-08, + "loss": 0.6799, + "step": 41901 + }, + { + "epoch": 1.53, + "learning_rate": 4.6735269344856624e-08, + "loss": 0.6954, + "step": 41902 + }, + { + "epoch": 1.53, + "learning_rate": 4.6731607280184564e-08, + "loss": 0.6961, + "step": 41903 + }, + { + "epoch": 1.53, + "learning_rate": 4.6727945215512504e-08, + "loss": 0.6976, + "step": 41904 + }, + { + "epoch": 1.53, + "learning_rate": 4.6724283150840444e-08, + "loss": 0.6757, + "step": 41905 + }, + { + "epoch": 1.53, + "learning_rate": 4.672062108616838e-08, + "loss": 0.6896, + "step": 41906 + }, + { + "epoch": 1.53, + "learning_rate": 4.671695902149632e-08, + "loss": 0.6791, + "step": 41907 + }, + { + "epoch": 1.53, + "learning_rate": 4.671329695682426e-08, + "loss": 0.6968, + "step": 41908 + }, + { + "epoch": 1.53, + "learning_rate": 4.67096348921522e-08, + "loss": 0.6968, + "step": 41909 + }, + { + "epoch": 1.53, + "learning_rate": 4.670597282748013e-08, + "loss": 0.7028, + "step": 41910 + }, + { + "epoch": 1.53, + "learning_rate": 4.6702310762808064e-08, + "loss": 0.6847, + "step": 41911 + }, + { + "epoch": 1.53, + "learning_rate": 4.6698648698136004e-08, + "loss": 0.6916, + "step": 41912 + }, + { + "epoch": 1.53, + "learning_rate": 4.6694986633463944e-08, + "loss": 0.6989, + "step": 41913 + }, + { + "epoch": 1.53, + "learning_rate": 4.6691324568791884e-08, + "loss": 0.7032, + "step": 41914 + }, + { + "epoch": 1.53, + "learning_rate": 4.668766250411982e-08, + "loss": 0.6907, + "step": 41915 + }, + { + "epoch": 1.53, + "learning_rate": 4.668400043944776e-08, + "loss": 0.6966, + "step": 41916 + }, + { + "epoch": 1.53, + "learning_rate": 4.66803383747757e-08, + "loss": 0.6987, + "step": 41917 + }, + { + "epoch": 1.53, + "learning_rate": 4.6676676310103637e-08, + "loss": 0.678, + "step": 41918 + }, + { + "epoch": 1.53, + "learning_rate": 4.6673014245431577e-08, + "loss": 0.6996, + "step": 41919 + }, + { + "epoch": 1.53, + "learning_rate": 4.666935218075951e-08, + "loss": 0.6998, + "step": 41920 + }, + { + "epoch": 1.53, + "learning_rate": 4.666569011608745e-08, + "loss": 0.6931, + "step": 41921 + }, + { + "epoch": 1.53, + "learning_rate": 4.666202805141538e-08, + "loss": 0.6903, + "step": 41922 + }, + { + "epoch": 1.53, + "learning_rate": 4.665836598674332e-08, + "loss": 0.6856, + "step": 41923 + }, + { + "epoch": 1.53, + "learning_rate": 4.6654703922071256e-08, + "loss": 0.6999, + "step": 41924 + }, + { + "epoch": 1.53, + "learning_rate": 4.6651041857399196e-08, + "loss": 0.6867, + "step": 41925 + }, + { + "epoch": 1.53, + "learning_rate": 4.6647379792727136e-08, + "loss": 0.6934, + "step": 41926 + }, + { + "epoch": 1.53, + "learning_rate": 4.6643717728055076e-08, + "loss": 0.6905, + "step": 41927 + }, + { + "epoch": 1.53, + "learning_rate": 4.6640055663383016e-08, + "loss": 0.6933, + "step": 41928 + }, + { + "epoch": 1.53, + "learning_rate": 4.663639359871095e-08, + "loss": 0.707, + "step": 41929 + }, + { + "epoch": 1.53, + "learning_rate": 4.663273153403889e-08, + "loss": 0.6862, + "step": 41930 + }, + { + "epoch": 1.53, + "learning_rate": 4.662906946936683e-08, + "loss": 0.6963, + "step": 41931 + }, + { + "epoch": 1.53, + "learning_rate": 4.662540740469477e-08, + "loss": 0.6989, + "step": 41932 + }, + { + "epoch": 1.53, + "learning_rate": 4.66217453400227e-08, + "loss": 0.6935, + "step": 41933 + }, + { + "epoch": 1.53, + "learning_rate": 4.661808327535064e-08, + "loss": 0.6978, + "step": 41934 + }, + { + "epoch": 1.53, + "learning_rate": 4.6614421210678576e-08, + "loss": 0.6777, + "step": 41935 + }, + { + "epoch": 1.53, + "learning_rate": 4.6610759146006516e-08, + "loss": 0.678, + "step": 41936 + }, + { + "epoch": 1.53, + "learning_rate": 4.6607097081334456e-08, + "loss": 0.6706, + "step": 41937 + }, + { + "epoch": 1.53, + "learning_rate": 4.660343501666239e-08, + "loss": 0.6864, + "step": 41938 + }, + { + "epoch": 1.53, + "learning_rate": 4.659977295199033e-08, + "loss": 0.697, + "step": 41939 + }, + { + "epoch": 1.53, + "learning_rate": 4.659611088731827e-08, + "loss": 0.7031, + "step": 41940 + }, + { + "epoch": 1.53, + "learning_rate": 4.659244882264621e-08, + "loss": 0.6999, + "step": 41941 + }, + { + "epoch": 1.53, + "learning_rate": 4.658878675797414e-08, + "loss": 0.6917, + "step": 41942 + }, + { + "epoch": 1.53, + "learning_rate": 4.658512469330208e-08, + "loss": 0.6883, + "step": 41943 + }, + { + "epoch": 1.53, + "learning_rate": 4.658146262863002e-08, + "loss": 0.6889, + "step": 41944 + }, + { + "epoch": 1.53, + "learning_rate": 4.657780056395796e-08, + "loss": 0.6917, + "step": 41945 + }, + { + "epoch": 1.53, + "learning_rate": 4.6574138499285895e-08, + "loss": 0.7008, + "step": 41946 + }, + { + "epoch": 1.53, + "learning_rate": 4.657047643461383e-08, + "loss": 0.6887, + "step": 41947 + }, + { + "epoch": 1.53, + "learning_rate": 4.656681436994177e-08, + "loss": 0.6928, + "step": 41948 + }, + { + "epoch": 1.53, + "learning_rate": 4.656315230526971e-08, + "loss": 0.6835, + "step": 41949 + }, + { + "epoch": 1.53, + "learning_rate": 4.655949024059765e-08, + "loss": 0.6976, + "step": 41950 + }, + { + "epoch": 1.53, + "learning_rate": 4.655582817592558e-08, + "loss": 0.6955, + "step": 41951 + }, + { + "epoch": 1.53, + "learning_rate": 4.655216611125352e-08, + "loss": 0.6963, + "step": 41952 + }, + { + "epoch": 1.53, + "learning_rate": 4.654850404658146e-08, + "loss": 0.6897, + "step": 41953 + }, + { + "epoch": 1.53, + "learning_rate": 4.65448419819094e-08, + "loss": 0.6948, + "step": 41954 + }, + { + "epoch": 1.54, + "learning_rate": 4.654117991723734e-08, + "loss": 0.6843, + "step": 41955 + }, + { + "epoch": 1.54, + "learning_rate": 4.6537517852565275e-08, + "loss": 0.689, + "step": 41956 + }, + { + "epoch": 1.54, + "learning_rate": 4.6533855787893214e-08, + "loss": 0.6919, + "step": 41957 + }, + { + "epoch": 1.54, + "learning_rate": 4.653019372322115e-08, + "loss": 0.691, + "step": 41958 + }, + { + "epoch": 1.54, + "learning_rate": 4.652653165854909e-08, + "loss": 0.6829, + "step": 41959 + }, + { + "epoch": 1.54, + "learning_rate": 4.652286959387702e-08, + "loss": 0.6907, + "step": 41960 + }, + { + "epoch": 1.54, + "learning_rate": 4.651920752920496e-08, + "loss": 0.6963, + "step": 41961 + }, + { + "epoch": 1.54, + "learning_rate": 4.65155454645329e-08, + "loss": 0.6929, + "step": 41962 + }, + { + "epoch": 1.54, + "learning_rate": 4.651188339986084e-08, + "loss": 0.7068, + "step": 41963 + }, + { + "epoch": 1.54, + "learning_rate": 4.650822133518878e-08, + "loss": 0.6989, + "step": 41964 + }, + { + "epoch": 1.54, + "learning_rate": 4.6504559270516714e-08, + "loss": 0.6899, + "step": 41965 + }, + { + "epoch": 1.54, + "learning_rate": 4.6500897205844654e-08, + "loss": 0.6972, + "step": 41966 + }, + { + "epoch": 1.54, + "learning_rate": 4.6497235141172594e-08, + "loss": 0.6927, + "step": 41967 + }, + { + "epoch": 1.54, + "learning_rate": 4.6493573076500534e-08, + "loss": 0.6916, + "step": 41968 + }, + { + "epoch": 1.54, + "learning_rate": 4.648991101182847e-08, + "loss": 0.698, + "step": 41969 + }, + { + "epoch": 1.54, + "learning_rate": 4.64862489471564e-08, + "loss": 0.7026, + "step": 41970 + }, + { + "epoch": 1.54, + "learning_rate": 4.648258688248434e-08, + "loss": 0.6967, + "step": 41971 + }, + { + "epoch": 1.54, + "learning_rate": 4.647892481781228e-08, + "loss": 0.6989, + "step": 41972 + }, + { + "epoch": 1.54, + "learning_rate": 4.647526275314022e-08, + "loss": 0.6958, + "step": 41973 + }, + { + "epoch": 1.54, + "learning_rate": 4.6471600688468154e-08, + "loss": 0.7015, + "step": 41974 + }, + { + "epoch": 1.54, + "learning_rate": 4.6467938623796094e-08, + "loss": 0.6847, + "step": 41975 + }, + { + "epoch": 1.54, + "learning_rate": 4.6464276559124033e-08, + "loss": 0.6976, + "step": 41976 + }, + { + "epoch": 1.54, + "learning_rate": 4.6460614494451973e-08, + "loss": 0.6957, + "step": 41977 + }, + { + "epoch": 1.54, + "learning_rate": 4.6456952429779907e-08, + "loss": 0.6947, + "step": 41978 + }, + { + "epoch": 1.54, + "learning_rate": 4.6453290365107847e-08, + "loss": 0.6934, + "step": 41979 + }, + { + "epoch": 1.54, + "learning_rate": 4.6449628300435787e-08, + "loss": 0.6903, + "step": 41980 + }, + { + "epoch": 1.54, + "learning_rate": 4.644596623576372e-08, + "loss": 0.6949, + "step": 41981 + }, + { + "epoch": 1.54, + "learning_rate": 4.644230417109166e-08, + "loss": 0.687, + "step": 41982 + }, + { + "epoch": 1.54, + "learning_rate": 4.643864210641959e-08, + "loss": 0.6824, + "step": 41983 + }, + { + "epoch": 1.54, + "learning_rate": 4.643498004174753e-08, + "loss": 0.679, + "step": 41984 + }, + { + "epoch": 1.54, + "learning_rate": 4.643131797707547e-08, + "loss": 0.6932, + "step": 41985 + }, + { + "epoch": 1.54, + "learning_rate": 4.642765591240341e-08, + "loss": 0.6814, + "step": 41986 + }, + { + "epoch": 1.54, + "learning_rate": 4.6423993847731346e-08, + "loss": 0.6884, + "step": 41987 + }, + { + "epoch": 1.54, + "learning_rate": 4.6420331783059286e-08, + "loss": 0.6797, + "step": 41988 + }, + { + "epoch": 1.54, + "learning_rate": 4.6416669718387226e-08, + "loss": 0.6975, + "step": 41989 + }, + { + "epoch": 1.54, + "learning_rate": 4.6413007653715166e-08, + "loss": 0.694, + "step": 41990 + }, + { + "epoch": 1.54, + "learning_rate": 4.6409345589043106e-08, + "loss": 0.7067, + "step": 41991 + }, + { + "epoch": 1.54, + "learning_rate": 4.640568352437104e-08, + "loss": 0.6982, + "step": 41992 + }, + { + "epoch": 1.54, + "learning_rate": 4.640202145969897e-08, + "loss": 0.6975, + "step": 41993 + }, + { + "epoch": 1.54, + "learning_rate": 4.639835939502691e-08, + "loss": 0.6945, + "step": 41994 + }, + { + "epoch": 1.54, + "learning_rate": 4.639469733035485e-08, + "loss": 0.676, + "step": 41995 + }, + { + "epoch": 1.54, + "learning_rate": 4.6391035265682786e-08, + "loss": 0.6937, + "step": 41996 + }, + { + "epoch": 1.54, + "learning_rate": 4.6387373201010726e-08, + "loss": 0.6835, + "step": 41997 + }, + { + "epoch": 1.54, + "learning_rate": 4.6383711136338666e-08, + "loss": 0.7011, + "step": 41998 + }, + { + "epoch": 1.54, + "learning_rate": 4.6380049071666605e-08, + "loss": 0.691, + "step": 41999 + }, + { + "epoch": 1.54, + "learning_rate": 4.6376387006994545e-08, + "loss": 0.6812, + "step": 42000 + }, + { + "epoch": 1.54, + "learning_rate": 4.637272494232248e-08, + "loss": 0.6936, + "step": 42001 + }, + { + "epoch": 1.54, + "learning_rate": 4.636906287765042e-08, + "loss": 0.6818, + "step": 42002 + }, + { + "epoch": 1.54, + "learning_rate": 4.636540081297836e-08, + "loss": 0.6942, + "step": 42003 + }, + { + "epoch": 1.54, + "learning_rate": 4.636173874830629e-08, + "loss": 0.7182, + "step": 42004 + }, + { + "epoch": 1.54, + "learning_rate": 4.635807668363423e-08, + "loss": 0.6901, + "step": 42005 + }, + { + "epoch": 1.54, + "learning_rate": 4.6354414618962165e-08, + "loss": 0.6933, + "step": 42006 + }, + { + "epoch": 1.54, + "learning_rate": 4.6350752554290105e-08, + "loss": 0.695, + "step": 42007 + }, + { + "epoch": 1.54, + "learning_rate": 4.6347090489618045e-08, + "loss": 0.7098, + "step": 42008 + }, + { + "epoch": 1.54, + "learning_rate": 4.6343428424945985e-08, + "loss": 0.6923, + "step": 42009 + }, + { + "epoch": 1.54, + "learning_rate": 4.633976636027392e-08, + "loss": 0.6939, + "step": 42010 + }, + { + "epoch": 1.54, + "learning_rate": 4.633610429560186e-08, + "loss": 0.6968, + "step": 42011 + }, + { + "epoch": 1.54, + "learning_rate": 4.63324422309298e-08, + "loss": 0.6775, + "step": 42012 + }, + { + "epoch": 1.54, + "learning_rate": 4.632878016625774e-08, + "loss": 0.686, + "step": 42013 + }, + { + "epoch": 1.54, + "learning_rate": 4.632511810158567e-08, + "loss": 0.687, + "step": 42014 + }, + { + "epoch": 1.54, + "learning_rate": 4.632145603691361e-08, + "loss": 0.6945, + "step": 42015 + }, + { + "epoch": 1.54, + "learning_rate": 4.6317793972241545e-08, + "loss": 0.6907, + "step": 42016 + }, + { + "epoch": 1.54, + "learning_rate": 4.6314131907569484e-08, + "loss": 0.6888, + "step": 42017 + }, + { + "epoch": 1.54, + "learning_rate": 4.6310469842897424e-08, + "loss": 0.6923, + "step": 42018 + }, + { + "epoch": 1.54, + "learning_rate": 4.630680777822536e-08, + "loss": 0.6877, + "step": 42019 + }, + { + "epoch": 1.54, + "learning_rate": 4.63031457135533e-08, + "loss": 0.6943, + "step": 42020 + }, + { + "epoch": 1.54, + "learning_rate": 4.629948364888124e-08, + "loss": 0.7022, + "step": 42021 + }, + { + "epoch": 1.54, + "learning_rate": 4.629582158420918e-08, + "loss": 0.6905, + "step": 42022 + }, + { + "epoch": 1.54, + "learning_rate": 4.629215951953711e-08, + "loss": 0.6876, + "step": 42023 + }, + { + "epoch": 1.54, + "learning_rate": 4.628849745486505e-08, + "loss": 0.6928, + "step": 42024 + }, + { + "epoch": 1.54, + "learning_rate": 4.628483539019299e-08, + "loss": 0.6944, + "step": 42025 + }, + { + "epoch": 1.54, + "learning_rate": 4.628117332552093e-08, + "loss": 0.6857, + "step": 42026 + }, + { + "epoch": 1.54, + "learning_rate": 4.6277511260848864e-08, + "loss": 0.677, + "step": 42027 + }, + { + "epoch": 1.54, + "learning_rate": 4.6273849196176804e-08, + "loss": 0.6944, + "step": 42028 + }, + { + "epoch": 1.54, + "learning_rate": 4.627018713150474e-08, + "loss": 0.6866, + "step": 42029 + }, + { + "epoch": 1.54, + "learning_rate": 4.626652506683268e-08, + "loss": 0.695, + "step": 42030 + }, + { + "epoch": 1.54, + "learning_rate": 4.626286300216062e-08, + "loss": 0.6953, + "step": 42031 + }, + { + "epoch": 1.54, + "learning_rate": 4.625920093748855e-08, + "loss": 0.7002, + "step": 42032 + }, + { + "epoch": 1.54, + "learning_rate": 4.625553887281649e-08, + "loss": 0.6895, + "step": 42033 + }, + { + "epoch": 1.54, + "learning_rate": 4.625187680814443e-08, + "loss": 0.6904, + "step": 42034 + }, + { + "epoch": 1.54, + "learning_rate": 4.624821474347237e-08, + "loss": 0.7069, + "step": 42035 + }, + { + "epoch": 1.54, + "learning_rate": 4.624455267880031e-08, + "loss": 0.6731, + "step": 42036 + }, + { + "epoch": 1.54, + "learning_rate": 4.6240890614128243e-08, + "loss": 0.7063, + "step": 42037 + }, + { + "epoch": 1.54, + "learning_rate": 4.623722854945618e-08, + "loss": 0.6866, + "step": 42038 + }, + { + "epoch": 1.54, + "learning_rate": 4.623356648478412e-08, + "loss": 0.6929, + "step": 42039 + }, + { + "epoch": 1.54, + "learning_rate": 4.6229904420112057e-08, + "loss": 0.6944, + "step": 42040 + }, + { + "epoch": 1.54, + "learning_rate": 4.622624235543999e-08, + "loss": 0.6848, + "step": 42041 + }, + { + "epoch": 1.54, + "learning_rate": 4.622258029076793e-08, + "loss": 0.6884, + "step": 42042 + }, + { + "epoch": 1.54, + "learning_rate": 4.621891822609587e-08, + "loss": 0.6889, + "step": 42043 + }, + { + "epoch": 1.54, + "learning_rate": 4.621525616142381e-08, + "loss": 0.6921, + "step": 42044 + }, + { + "epoch": 1.54, + "learning_rate": 4.621159409675175e-08, + "loss": 0.7002, + "step": 42045 + }, + { + "epoch": 1.54, + "learning_rate": 4.620793203207968e-08, + "loss": 0.6885, + "step": 42046 + }, + { + "epoch": 1.54, + "learning_rate": 4.620426996740762e-08, + "loss": 0.6923, + "step": 42047 + }, + { + "epoch": 1.54, + "learning_rate": 4.620060790273556e-08, + "loss": 0.68, + "step": 42048 + }, + { + "epoch": 1.54, + "learning_rate": 4.61969458380635e-08, + "loss": 0.693, + "step": 42049 + }, + { + "epoch": 1.54, + "learning_rate": 4.6193283773391436e-08, + "loss": 0.6996, + "step": 42050 + }, + { + "epoch": 1.54, + "learning_rate": 4.6189621708719376e-08, + "loss": 0.703, + "step": 42051 + }, + { + "epoch": 1.54, + "learning_rate": 4.618595964404731e-08, + "loss": 0.7046, + "step": 42052 + }, + { + "epoch": 1.54, + "learning_rate": 4.618229757937525e-08, + "loss": 0.6942, + "step": 42053 + }, + { + "epoch": 1.54, + "learning_rate": 4.617863551470319e-08, + "loss": 0.6743, + "step": 42054 + }, + { + "epoch": 1.54, + "learning_rate": 4.617497345003112e-08, + "loss": 0.689, + "step": 42055 + }, + { + "epoch": 1.54, + "learning_rate": 4.617131138535906e-08, + "loss": 0.6933, + "step": 42056 + }, + { + "epoch": 1.54, + "learning_rate": 4.6167649320687e-08, + "loss": 0.6971, + "step": 42057 + }, + { + "epoch": 1.54, + "learning_rate": 4.616398725601494e-08, + "loss": 0.6936, + "step": 42058 + }, + { + "epoch": 1.54, + "learning_rate": 4.6160325191342875e-08, + "loss": 0.6868, + "step": 42059 + }, + { + "epoch": 1.54, + "learning_rate": 4.6156663126670815e-08, + "loss": 0.6775, + "step": 42060 + }, + { + "epoch": 1.54, + "learning_rate": 4.6153001061998755e-08, + "loss": 0.6887, + "step": 42061 + }, + { + "epoch": 1.54, + "learning_rate": 4.6149338997326695e-08, + "loss": 0.6889, + "step": 42062 + }, + { + "epoch": 1.54, + "learning_rate": 4.614567693265463e-08, + "loss": 0.6895, + "step": 42063 + }, + { + "epoch": 1.54, + "learning_rate": 4.614201486798256e-08, + "loss": 0.6931, + "step": 42064 + }, + { + "epoch": 1.54, + "learning_rate": 4.61383528033105e-08, + "loss": 0.6876, + "step": 42065 + }, + { + "epoch": 1.54, + "learning_rate": 4.613469073863844e-08, + "loss": 0.6921, + "step": 42066 + }, + { + "epoch": 1.54, + "learning_rate": 4.613102867396638e-08, + "loss": 0.6941, + "step": 42067 + }, + { + "epoch": 1.54, + "learning_rate": 4.6127366609294315e-08, + "loss": 0.7007, + "step": 42068 + }, + { + "epoch": 1.54, + "learning_rate": 4.6123704544622255e-08, + "loss": 0.6941, + "step": 42069 + }, + { + "epoch": 1.54, + "learning_rate": 4.6120042479950195e-08, + "loss": 0.7069, + "step": 42070 + }, + { + "epoch": 1.54, + "learning_rate": 4.6116380415278135e-08, + "loss": 0.7029, + "step": 42071 + }, + { + "epoch": 1.54, + "learning_rate": 4.611271835060607e-08, + "loss": 0.6986, + "step": 42072 + }, + { + "epoch": 1.54, + "learning_rate": 4.610905628593401e-08, + "loss": 0.683, + "step": 42073 + }, + { + "epoch": 1.54, + "learning_rate": 4.610539422126195e-08, + "loss": 0.7051, + "step": 42074 + }, + { + "epoch": 1.54, + "learning_rate": 4.610173215658988e-08, + "loss": 0.6847, + "step": 42075 + }, + { + "epoch": 1.54, + "learning_rate": 4.609807009191782e-08, + "loss": 0.6867, + "step": 42076 + }, + { + "epoch": 1.54, + "learning_rate": 4.6094408027245754e-08, + "loss": 0.6946, + "step": 42077 + }, + { + "epoch": 1.54, + "learning_rate": 4.6090745962573694e-08, + "loss": 0.7121, + "step": 42078 + }, + { + "epoch": 1.54, + "learning_rate": 4.6087083897901634e-08, + "loss": 0.7087, + "step": 42079 + }, + { + "epoch": 1.54, + "learning_rate": 4.6083421833229574e-08, + "loss": 0.6938, + "step": 42080 + }, + { + "epoch": 1.54, + "learning_rate": 4.607975976855751e-08, + "loss": 0.6953, + "step": 42081 + }, + { + "epoch": 1.54, + "learning_rate": 4.607609770388545e-08, + "loss": 0.6914, + "step": 42082 + }, + { + "epoch": 1.54, + "learning_rate": 4.607243563921339e-08, + "loss": 0.6847, + "step": 42083 + }, + { + "epoch": 1.54, + "learning_rate": 4.606877357454133e-08, + "loss": 0.6942, + "step": 42084 + }, + { + "epoch": 1.54, + "learning_rate": 4.606511150986927e-08, + "loss": 0.6891, + "step": 42085 + }, + { + "epoch": 1.54, + "learning_rate": 4.60614494451972e-08, + "loss": 0.6918, + "step": 42086 + }, + { + "epoch": 1.54, + "learning_rate": 4.6057787380525134e-08, + "loss": 0.6939, + "step": 42087 + }, + { + "epoch": 1.54, + "learning_rate": 4.6054125315853074e-08, + "loss": 0.692, + "step": 42088 + }, + { + "epoch": 1.54, + "learning_rate": 4.6050463251181014e-08, + "loss": 0.6913, + "step": 42089 + }, + { + "epoch": 1.54, + "learning_rate": 4.604680118650895e-08, + "loss": 0.6874, + "step": 42090 + }, + { + "epoch": 1.54, + "learning_rate": 4.604313912183689e-08, + "loss": 0.7016, + "step": 42091 + }, + { + "epoch": 1.54, + "learning_rate": 4.603947705716483e-08, + "loss": 0.695, + "step": 42092 + }, + { + "epoch": 1.54, + "learning_rate": 4.603581499249277e-08, + "loss": 0.6912, + "step": 42093 + }, + { + "epoch": 1.54, + "learning_rate": 4.603215292782071e-08, + "loss": 0.6862, + "step": 42094 + }, + { + "epoch": 1.54, + "learning_rate": 4.602849086314864e-08, + "loss": 0.6895, + "step": 42095 + }, + { + "epoch": 1.54, + "learning_rate": 4.602482879847658e-08, + "loss": 0.6849, + "step": 42096 + }, + { + "epoch": 1.54, + "learning_rate": 4.602116673380452e-08, + "loss": 0.6853, + "step": 42097 + }, + { + "epoch": 1.54, + "learning_rate": 4.601750466913245e-08, + "loss": 0.7005, + "step": 42098 + }, + { + "epoch": 1.54, + "learning_rate": 4.601384260446039e-08, + "loss": 0.6886, + "step": 42099 + }, + { + "epoch": 1.54, + "learning_rate": 4.6010180539788327e-08, + "loss": 0.6946, + "step": 42100 + }, + { + "epoch": 1.54, + "learning_rate": 4.6006518475116266e-08, + "loss": 0.6763, + "step": 42101 + }, + { + "epoch": 1.54, + "learning_rate": 4.6002856410444206e-08, + "loss": 0.6923, + "step": 42102 + }, + { + "epoch": 1.54, + "learning_rate": 4.5999194345772146e-08, + "loss": 0.6874, + "step": 42103 + }, + { + "epoch": 1.54, + "learning_rate": 4.599553228110008e-08, + "loss": 0.6796, + "step": 42104 + }, + { + "epoch": 1.54, + "learning_rate": 4.599187021642802e-08, + "loss": 0.6924, + "step": 42105 + }, + { + "epoch": 1.54, + "learning_rate": 4.598820815175596e-08, + "loss": 0.6832, + "step": 42106 + }, + { + "epoch": 1.54, + "learning_rate": 4.59845460870839e-08, + "loss": 0.6873, + "step": 42107 + }, + { + "epoch": 1.54, + "learning_rate": 4.598088402241183e-08, + "loss": 0.6908, + "step": 42108 + }, + { + "epoch": 1.54, + "learning_rate": 4.597722195773977e-08, + "loss": 0.6849, + "step": 42109 + }, + { + "epoch": 1.54, + "learning_rate": 4.597355989306771e-08, + "loss": 0.6924, + "step": 42110 + }, + { + "epoch": 1.54, + "learning_rate": 4.5969897828395646e-08, + "loss": 0.6871, + "step": 42111 + }, + { + "epoch": 1.54, + "learning_rate": 4.5966235763723586e-08, + "loss": 0.6774, + "step": 42112 + }, + { + "epoch": 1.54, + "learning_rate": 4.596257369905152e-08, + "loss": 0.6991, + "step": 42113 + }, + { + "epoch": 1.54, + "learning_rate": 4.595891163437946e-08, + "loss": 0.6859, + "step": 42114 + }, + { + "epoch": 1.54, + "learning_rate": 4.59552495697074e-08, + "loss": 0.6905, + "step": 42115 + }, + { + "epoch": 1.54, + "learning_rate": 4.595158750503534e-08, + "loss": 0.7127, + "step": 42116 + }, + { + "epoch": 1.54, + "learning_rate": 4.594792544036327e-08, + "loss": 0.6915, + "step": 42117 + }, + { + "epoch": 1.54, + "learning_rate": 4.594426337569121e-08, + "loss": 0.6993, + "step": 42118 + }, + { + "epoch": 1.54, + "learning_rate": 4.594060131101915e-08, + "loss": 0.6991, + "step": 42119 + }, + { + "epoch": 1.54, + "learning_rate": 4.593693924634709e-08, + "loss": 0.7041, + "step": 42120 + }, + { + "epoch": 1.54, + "learning_rate": 4.593327718167503e-08, + "loss": 0.6958, + "step": 42121 + }, + { + "epoch": 1.54, + "learning_rate": 4.5929615117002965e-08, + "loss": 0.7017, + "step": 42122 + }, + { + "epoch": 1.54, + "learning_rate": 4.59259530523309e-08, + "loss": 0.6832, + "step": 42123 + }, + { + "epoch": 1.54, + "learning_rate": 4.592229098765884e-08, + "loss": 0.6997, + "step": 42124 + }, + { + "epoch": 1.54, + "learning_rate": 4.591862892298678e-08, + "loss": 0.6865, + "step": 42125 + }, + { + "epoch": 1.54, + "learning_rate": 4.591496685831471e-08, + "loss": 0.6883, + "step": 42126 + }, + { + "epoch": 1.54, + "learning_rate": 4.591130479364265e-08, + "loss": 0.6861, + "step": 42127 + }, + { + "epoch": 1.54, + "learning_rate": 4.590764272897059e-08, + "loss": 0.7021, + "step": 42128 + }, + { + "epoch": 1.54, + "learning_rate": 4.590398066429853e-08, + "loss": 0.6952, + "step": 42129 + }, + { + "epoch": 1.54, + "learning_rate": 4.590031859962647e-08, + "loss": 0.6933, + "step": 42130 + }, + { + "epoch": 1.54, + "learning_rate": 4.5896656534954405e-08, + "loss": 0.7039, + "step": 42131 + }, + { + "epoch": 1.54, + "learning_rate": 4.5892994470282345e-08, + "loss": 0.6977, + "step": 42132 + }, + { + "epoch": 1.54, + "learning_rate": 4.5889332405610285e-08, + "loss": 0.6949, + "step": 42133 + }, + { + "epoch": 1.54, + "learning_rate": 4.588567034093822e-08, + "loss": 0.697, + "step": 42134 + }, + { + "epoch": 1.54, + "learning_rate": 4.588200827626615e-08, + "loss": 0.6911, + "step": 42135 + }, + { + "epoch": 1.54, + "learning_rate": 4.587834621159409e-08, + "loss": 0.6996, + "step": 42136 + }, + { + "epoch": 1.54, + "learning_rate": 4.587468414692203e-08, + "loss": 0.6948, + "step": 42137 + }, + { + "epoch": 1.54, + "learning_rate": 4.587102208224997e-08, + "loss": 0.6893, + "step": 42138 + }, + { + "epoch": 1.54, + "learning_rate": 4.586736001757791e-08, + "loss": 0.6673, + "step": 42139 + }, + { + "epoch": 1.54, + "learning_rate": 4.5863697952905844e-08, + "loss": 0.6852, + "step": 42140 + }, + { + "epoch": 1.54, + "learning_rate": 4.5860035888233784e-08, + "loss": 0.6911, + "step": 42141 + }, + { + "epoch": 1.54, + "learning_rate": 4.5856373823561724e-08, + "loss": 0.6913, + "step": 42142 + }, + { + "epoch": 1.54, + "learning_rate": 4.5852711758889664e-08, + "loss": 0.6933, + "step": 42143 + }, + { + "epoch": 1.54, + "learning_rate": 4.58490496942176e-08, + "loss": 0.7002, + "step": 42144 + }, + { + "epoch": 1.54, + "learning_rate": 4.584538762954554e-08, + "loss": 0.7061, + "step": 42145 + }, + { + "epoch": 1.54, + "learning_rate": 4.584172556487347e-08, + "loss": 0.6915, + "step": 42146 + }, + { + "epoch": 1.54, + "learning_rate": 4.583806350020141e-08, + "loss": 0.7032, + "step": 42147 + }, + { + "epoch": 1.54, + "learning_rate": 4.583440143552935e-08, + "loss": 0.6802, + "step": 42148 + }, + { + "epoch": 1.54, + "learning_rate": 4.5830739370857284e-08, + "loss": 0.6821, + "step": 42149 + }, + { + "epoch": 1.54, + "learning_rate": 4.5827077306185224e-08, + "loss": 0.6848, + "step": 42150 + }, + { + "epoch": 1.54, + "learning_rate": 4.5823415241513164e-08, + "loss": 0.6995, + "step": 42151 + }, + { + "epoch": 1.54, + "learning_rate": 4.5819753176841104e-08, + "loss": 0.6962, + "step": 42152 + }, + { + "epoch": 1.54, + "learning_rate": 4.581609111216904e-08, + "loss": 0.7029, + "step": 42153 + }, + { + "epoch": 1.54, + "learning_rate": 4.581242904749698e-08, + "loss": 0.6933, + "step": 42154 + }, + { + "epoch": 1.54, + "learning_rate": 4.580876698282492e-08, + "loss": 0.6928, + "step": 42155 + }, + { + "epoch": 1.54, + "learning_rate": 4.5805104918152857e-08, + "loss": 0.6967, + "step": 42156 + }, + { + "epoch": 1.54, + "learning_rate": 4.580144285348079e-08, + "loss": 0.6945, + "step": 42157 + }, + { + "epoch": 1.54, + "learning_rate": 4.579778078880872e-08, + "loss": 0.6988, + "step": 42158 + }, + { + "epoch": 1.54, + "learning_rate": 4.579411872413666e-08, + "loss": 0.6951, + "step": 42159 + }, + { + "epoch": 1.54, + "learning_rate": 4.57904566594646e-08, + "loss": 0.694, + "step": 42160 + }, + { + "epoch": 1.54, + "learning_rate": 4.578679459479254e-08, + "loss": 0.6944, + "step": 42161 + }, + { + "epoch": 1.54, + "learning_rate": 4.5783132530120476e-08, + "loss": 0.682, + "step": 42162 + }, + { + "epoch": 1.54, + "learning_rate": 4.5779470465448416e-08, + "loss": 0.6957, + "step": 42163 + }, + { + "epoch": 1.54, + "learning_rate": 4.5775808400776356e-08, + "loss": 0.6908, + "step": 42164 + }, + { + "epoch": 1.54, + "learning_rate": 4.5772146336104296e-08, + "loss": 0.6916, + "step": 42165 + }, + { + "epoch": 1.54, + "learning_rate": 4.5768484271432236e-08, + "loss": 0.7007, + "step": 42166 + }, + { + "epoch": 1.54, + "learning_rate": 4.576482220676017e-08, + "loss": 0.6879, + "step": 42167 + }, + { + "epoch": 1.54, + "learning_rate": 4.576116014208811e-08, + "loss": 0.694, + "step": 42168 + }, + { + "epoch": 1.54, + "learning_rate": 4.575749807741604e-08, + "loss": 0.6834, + "step": 42169 + }, + { + "epoch": 1.54, + "learning_rate": 4.575383601274398e-08, + "loss": 0.7153, + "step": 42170 + }, + { + "epoch": 1.54, + "learning_rate": 4.5750173948071916e-08, + "loss": 0.6857, + "step": 42171 + }, + { + "epoch": 1.54, + "learning_rate": 4.5746511883399856e-08, + "loss": 0.6949, + "step": 42172 + }, + { + "epoch": 1.54, + "learning_rate": 4.5742849818727796e-08, + "loss": 0.6903, + "step": 42173 + }, + { + "epoch": 1.54, + "learning_rate": 4.5739187754055736e-08, + "loss": 0.6812, + "step": 42174 + }, + { + "epoch": 1.54, + "learning_rate": 4.5735525689383676e-08, + "loss": 0.6984, + "step": 42175 + }, + { + "epoch": 1.54, + "learning_rate": 4.573186362471161e-08, + "loss": 0.7017, + "step": 42176 + }, + { + "epoch": 1.54, + "learning_rate": 4.572820156003955e-08, + "loss": 0.6919, + "step": 42177 + }, + { + "epoch": 1.54, + "learning_rate": 4.572453949536749e-08, + "loss": 0.7004, + "step": 42178 + }, + { + "epoch": 1.54, + "learning_rate": 4.572087743069543e-08, + "loss": 0.6977, + "step": 42179 + }, + { + "epoch": 1.54, + "learning_rate": 4.571721536602336e-08, + "loss": 0.6833, + "step": 42180 + }, + { + "epoch": 1.54, + "learning_rate": 4.57135533013513e-08, + "loss": 0.7002, + "step": 42181 + }, + { + "epoch": 1.54, + "learning_rate": 4.5709891236679235e-08, + "loss": 0.696, + "step": 42182 + }, + { + "epoch": 1.54, + "learning_rate": 4.5706229172007175e-08, + "loss": 0.6959, + "step": 42183 + }, + { + "epoch": 1.54, + "learning_rate": 4.5702567107335115e-08, + "loss": 0.6847, + "step": 42184 + }, + { + "epoch": 1.54, + "learning_rate": 4.569890504266305e-08, + "loss": 0.6737, + "step": 42185 + }, + { + "epoch": 1.54, + "learning_rate": 4.569524297799099e-08, + "loss": 0.6978, + "step": 42186 + }, + { + "epoch": 1.54, + "learning_rate": 4.569158091331893e-08, + "loss": 0.6937, + "step": 42187 + }, + { + "epoch": 1.54, + "learning_rate": 4.568791884864687e-08, + "loss": 0.7039, + "step": 42188 + }, + { + "epoch": 1.54, + "learning_rate": 4.56842567839748e-08, + "loss": 0.6942, + "step": 42189 + }, + { + "epoch": 1.54, + "learning_rate": 4.568059471930274e-08, + "loss": 0.6881, + "step": 42190 + }, + { + "epoch": 1.54, + "learning_rate": 4.567693265463068e-08, + "loss": 0.6948, + "step": 42191 + }, + { + "epoch": 1.54, + "learning_rate": 4.5673270589958615e-08, + "loss": 0.7078, + "step": 42192 + }, + { + "epoch": 1.54, + "learning_rate": 4.5669608525286555e-08, + "loss": 0.7096, + "step": 42193 + }, + { + "epoch": 1.54, + "learning_rate": 4.566594646061449e-08, + "loss": 0.6907, + "step": 42194 + }, + { + "epoch": 1.54, + "learning_rate": 4.566228439594243e-08, + "loss": 0.7031, + "step": 42195 + }, + { + "epoch": 1.54, + "learning_rate": 4.565862233127037e-08, + "loss": 0.6967, + "step": 42196 + }, + { + "epoch": 1.54, + "learning_rate": 4.565496026659831e-08, + "loss": 0.6908, + "step": 42197 + }, + { + "epoch": 1.54, + "learning_rate": 4.565129820192624e-08, + "loss": 0.6864, + "step": 42198 + }, + { + "epoch": 1.54, + "learning_rate": 4.564763613725418e-08, + "loss": 0.6916, + "step": 42199 + }, + { + "epoch": 1.54, + "learning_rate": 4.564397407258212e-08, + "loss": 0.687, + "step": 42200 + }, + { + "epoch": 1.54, + "learning_rate": 4.564031200791006e-08, + "loss": 0.686, + "step": 42201 + }, + { + "epoch": 1.54, + "learning_rate": 4.5636649943238e-08, + "loss": 0.7034, + "step": 42202 + }, + { + "epoch": 1.54, + "learning_rate": 4.5632987878565934e-08, + "loss": 0.6893, + "step": 42203 + }, + { + "epoch": 1.54, + "learning_rate": 4.5629325813893874e-08, + "loss": 0.6908, + "step": 42204 + }, + { + "epoch": 1.54, + "learning_rate": 4.562566374922181e-08, + "loss": 0.6993, + "step": 42205 + }, + { + "epoch": 1.54, + "learning_rate": 4.562200168454975e-08, + "loss": 0.6991, + "step": 42206 + }, + { + "epoch": 1.54, + "learning_rate": 4.561833961987768e-08, + "loss": 0.7048, + "step": 42207 + }, + { + "epoch": 1.54, + "learning_rate": 4.561467755520562e-08, + "loss": 0.6948, + "step": 42208 + }, + { + "epoch": 1.54, + "learning_rate": 4.561101549053356e-08, + "loss": 0.6971, + "step": 42209 + }, + { + "epoch": 1.54, + "learning_rate": 4.56073534258615e-08, + "loss": 0.7013, + "step": 42210 + }, + { + "epoch": 1.54, + "learning_rate": 4.560369136118944e-08, + "loss": 0.6871, + "step": 42211 + }, + { + "epoch": 1.54, + "learning_rate": 4.5600029296517374e-08, + "loss": 0.6909, + "step": 42212 + }, + { + "epoch": 1.54, + "learning_rate": 4.5596367231845313e-08, + "loss": 0.6927, + "step": 42213 + }, + { + "epoch": 1.54, + "learning_rate": 4.5592705167173253e-08, + "loss": 0.691, + "step": 42214 + }, + { + "epoch": 1.54, + "learning_rate": 4.558904310250119e-08, + "loss": 0.6955, + "step": 42215 + }, + { + "epoch": 1.54, + "learning_rate": 4.5585381037829127e-08, + "loss": 0.6877, + "step": 42216 + }, + { + "epoch": 1.54, + "learning_rate": 4.558171897315706e-08, + "loss": 0.6882, + "step": 42217 + }, + { + "epoch": 1.54, + "learning_rate": 4.5578056908485e-08, + "loss": 0.6937, + "step": 42218 + }, + { + "epoch": 1.54, + "learning_rate": 4.557439484381294e-08, + "loss": 0.7038, + "step": 42219 + }, + { + "epoch": 1.54, + "learning_rate": 4.557073277914088e-08, + "loss": 0.6884, + "step": 42220 + }, + { + "epoch": 1.54, + "learning_rate": 4.556707071446881e-08, + "loss": 0.7131, + "step": 42221 + }, + { + "epoch": 1.54, + "learning_rate": 4.556340864979675e-08, + "loss": 0.6884, + "step": 42222 + }, + { + "epoch": 1.54, + "learning_rate": 4.555974658512469e-08, + "loss": 0.7031, + "step": 42223 + }, + { + "epoch": 1.54, + "learning_rate": 4.555608452045263e-08, + "loss": 0.691, + "step": 42224 + }, + { + "epoch": 1.54, + "learning_rate": 4.5552422455780566e-08, + "loss": 0.6873, + "step": 42225 + }, + { + "epoch": 1.54, + "learning_rate": 4.5548760391108506e-08, + "loss": 0.6915, + "step": 42226 + }, + { + "epoch": 1.54, + "learning_rate": 4.5545098326436446e-08, + "loss": 0.7011, + "step": 42227 + }, + { + "epoch": 1.55, + "learning_rate": 4.554143626176438e-08, + "loss": 0.7015, + "step": 42228 + }, + { + "epoch": 1.55, + "learning_rate": 4.553777419709232e-08, + "loss": 0.6949, + "step": 42229 + }, + { + "epoch": 1.55, + "learning_rate": 4.553411213242025e-08, + "loss": 0.6999, + "step": 42230 + }, + { + "epoch": 1.55, + "learning_rate": 4.553045006774819e-08, + "loss": 0.6867, + "step": 42231 + }, + { + "epoch": 1.55, + "learning_rate": 4.552678800307613e-08, + "loss": 0.694, + "step": 42232 + }, + { + "epoch": 1.55, + "learning_rate": 4.552312593840407e-08, + "loss": 0.691, + "step": 42233 + }, + { + "epoch": 1.55, + "learning_rate": 4.5519463873732006e-08, + "loss": 0.6836, + "step": 42234 + }, + { + "epoch": 1.55, + "learning_rate": 4.5515801809059946e-08, + "loss": 0.7002, + "step": 42235 + }, + { + "epoch": 1.55, + "learning_rate": 4.5512139744387885e-08, + "loss": 0.6921, + "step": 42236 + }, + { + "epoch": 1.55, + "learning_rate": 4.5508477679715825e-08, + "loss": 0.69, + "step": 42237 + }, + { + "epoch": 1.55, + "learning_rate": 4.5504815615043765e-08, + "loss": 0.7148, + "step": 42238 + }, + { + "epoch": 1.55, + "learning_rate": 4.55011535503717e-08, + "loss": 0.7118, + "step": 42239 + }, + { + "epoch": 1.55, + "learning_rate": 4.549749148569963e-08, + "loss": 0.6891, + "step": 42240 + }, + { + "epoch": 1.55, + "learning_rate": 4.549382942102757e-08, + "loss": 0.6996, + "step": 42241 + }, + { + "epoch": 1.55, + "learning_rate": 4.549016735635551e-08, + "loss": 0.6795, + "step": 42242 + }, + { + "epoch": 1.55, + "learning_rate": 4.5486505291683445e-08, + "loss": 0.6867, + "step": 42243 + }, + { + "epoch": 1.55, + "learning_rate": 4.5482843227011385e-08, + "loss": 0.7169, + "step": 42244 + }, + { + "epoch": 1.55, + "learning_rate": 4.5479181162339325e-08, + "loss": 0.6909, + "step": 42245 + }, + { + "epoch": 1.55, + "learning_rate": 4.5475519097667265e-08, + "loss": 0.7006, + "step": 42246 + }, + { + "epoch": 1.55, + "learning_rate": 4.5471857032995205e-08, + "loss": 0.6993, + "step": 42247 + }, + { + "epoch": 1.55, + "learning_rate": 4.546819496832314e-08, + "loss": 0.6936, + "step": 42248 + }, + { + "epoch": 1.55, + "learning_rate": 4.546453290365108e-08, + "loss": 0.6942, + "step": 42249 + }, + { + "epoch": 1.55, + "learning_rate": 4.546087083897902e-08, + "loss": 0.6814, + "step": 42250 + }, + { + "epoch": 1.55, + "learning_rate": 4.545720877430695e-08, + "loss": 0.6805, + "step": 42251 + }, + { + "epoch": 1.55, + "learning_rate": 4.5453546709634885e-08, + "loss": 0.6902, + "step": 42252 + }, + { + "epoch": 1.55, + "learning_rate": 4.5449884644962825e-08, + "loss": 0.6985, + "step": 42253 + }, + { + "epoch": 1.55, + "learning_rate": 4.5446222580290764e-08, + "loss": 0.6928, + "step": 42254 + }, + { + "epoch": 1.55, + "learning_rate": 4.5442560515618704e-08, + "loss": 0.6958, + "step": 42255 + }, + { + "epoch": 1.55, + "learning_rate": 4.5438898450946644e-08, + "loss": 0.7065, + "step": 42256 + }, + { + "epoch": 1.55, + "learning_rate": 4.543523638627458e-08, + "loss": 0.6834, + "step": 42257 + }, + { + "epoch": 1.55, + "learning_rate": 4.543157432160252e-08, + "loss": 0.6986, + "step": 42258 + }, + { + "epoch": 1.55, + "learning_rate": 4.542791225693046e-08, + "loss": 0.691, + "step": 42259 + }, + { + "epoch": 1.55, + "learning_rate": 4.54242501922584e-08, + "loss": 0.691, + "step": 42260 + }, + { + "epoch": 1.55, + "learning_rate": 4.542058812758633e-08, + "loss": 0.6883, + "step": 42261 + }, + { + "epoch": 1.55, + "learning_rate": 4.541692606291427e-08, + "loss": 0.6815, + "step": 42262 + }, + { + "epoch": 1.55, + "learning_rate": 4.5413263998242204e-08, + "loss": 0.6914, + "step": 42263 + }, + { + "epoch": 1.55, + "learning_rate": 4.5409601933570144e-08, + "loss": 0.6874, + "step": 42264 + }, + { + "epoch": 1.55, + "learning_rate": 4.540593986889808e-08, + "loss": 0.6987, + "step": 42265 + }, + { + "epoch": 1.55, + "learning_rate": 4.540227780422602e-08, + "loss": 0.6873, + "step": 42266 + }, + { + "epoch": 1.55, + "learning_rate": 4.539861573955396e-08, + "loss": 0.6985, + "step": 42267 + }, + { + "epoch": 1.55, + "learning_rate": 4.53949536748819e-08, + "loss": 0.711, + "step": 42268 + }, + { + "epoch": 1.55, + "learning_rate": 4.539129161020984e-08, + "loss": 0.6918, + "step": 42269 + }, + { + "epoch": 1.55, + "learning_rate": 4.538762954553777e-08, + "loss": 0.6769, + "step": 42270 + }, + { + "epoch": 1.55, + "learning_rate": 4.538396748086571e-08, + "loss": 0.6815, + "step": 42271 + }, + { + "epoch": 1.55, + "learning_rate": 4.538030541619365e-08, + "loss": 0.6934, + "step": 42272 + }, + { + "epoch": 1.55, + "learning_rate": 4.537664335152159e-08, + "loss": 0.7036, + "step": 42273 + }, + { + "epoch": 1.55, + "learning_rate": 4.5372981286849523e-08, + "loss": 0.6842, + "step": 42274 + }, + { + "epoch": 1.55, + "learning_rate": 4.536931922217746e-08, + "loss": 0.6873, + "step": 42275 + }, + { + "epoch": 1.55, + "learning_rate": 4.5365657157505397e-08, + "loss": 0.6871, + "step": 42276 + }, + { + "epoch": 1.55, + "learning_rate": 4.5361995092833337e-08, + "loss": 0.6912, + "step": 42277 + }, + { + "epoch": 1.55, + "learning_rate": 4.5358333028161276e-08, + "loss": 0.6906, + "step": 42278 + }, + { + "epoch": 1.55, + "learning_rate": 4.535467096348921e-08, + "loss": 0.6958, + "step": 42279 + }, + { + "epoch": 1.55, + "learning_rate": 4.535100889881715e-08, + "loss": 0.701, + "step": 42280 + }, + { + "epoch": 1.55, + "learning_rate": 4.534734683414509e-08, + "loss": 0.6984, + "step": 42281 + }, + { + "epoch": 1.55, + "learning_rate": 4.534368476947303e-08, + "loss": 0.6912, + "step": 42282 + }, + { + "epoch": 1.55, + "learning_rate": 4.534002270480096e-08, + "loss": 0.6869, + "step": 42283 + }, + { + "epoch": 1.55, + "learning_rate": 4.53363606401289e-08, + "loss": 0.6841, + "step": 42284 + }, + { + "epoch": 1.55, + "learning_rate": 4.533269857545684e-08, + "loss": 0.6954, + "step": 42285 + }, + { + "epoch": 1.55, + "learning_rate": 4.532903651078478e-08, + "loss": 0.6788, + "step": 42286 + }, + { + "epoch": 1.55, + "learning_rate": 4.5325374446112716e-08, + "loss": 0.6977, + "step": 42287 + }, + { + "epoch": 1.55, + "learning_rate": 4.532171238144065e-08, + "loss": 0.6968, + "step": 42288 + }, + { + "epoch": 1.55, + "learning_rate": 4.531805031676859e-08, + "loss": 0.7052, + "step": 42289 + }, + { + "epoch": 1.55, + "learning_rate": 4.531438825209653e-08, + "loss": 0.6921, + "step": 42290 + }, + { + "epoch": 1.55, + "learning_rate": 4.531072618742447e-08, + "loss": 0.6843, + "step": 42291 + }, + { + "epoch": 1.55, + "learning_rate": 4.53070641227524e-08, + "loss": 0.7018, + "step": 42292 + }, + { + "epoch": 1.55, + "learning_rate": 4.530340205808034e-08, + "loss": 0.6893, + "step": 42293 + }, + { + "epoch": 1.55, + "learning_rate": 4.529973999340828e-08, + "loss": 0.6965, + "step": 42294 + }, + { + "epoch": 1.55, + "learning_rate": 4.529607792873622e-08, + "loss": 0.7056, + "step": 42295 + }, + { + "epoch": 1.55, + "learning_rate": 4.529241586406416e-08, + "loss": 0.6964, + "step": 42296 + }, + { + "epoch": 1.55, + "learning_rate": 4.5288753799392095e-08, + "loss": 0.672, + "step": 42297 + }, + { + "epoch": 1.55, + "learning_rate": 4.5285091734720035e-08, + "loss": 0.6968, + "step": 42298 + }, + { + "epoch": 1.55, + "learning_rate": 4.528142967004797e-08, + "loss": 0.6886, + "step": 42299 + }, + { + "epoch": 1.55, + "learning_rate": 4.527776760537591e-08, + "loss": 0.702, + "step": 42300 + }, + { + "epoch": 1.55, + "learning_rate": 4.527410554070384e-08, + "loss": 0.6913, + "step": 42301 + }, + { + "epoch": 1.55, + "learning_rate": 4.527044347603178e-08, + "loss": 0.7027, + "step": 42302 + }, + { + "epoch": 1.55, + "learning_rate": 4.526678141135972e-08, + "loss": 0.7076, + "step": 42303 + }, + { + "epoch": 1.55, + "learning_rate": 4.526311934668766e-08, + "loss": 0.6953, + "step": 42304 + }, + { + "epoch": 1.55, + "learning_rate": 4.52594572820156e-08, + "loss": 0.6906, + "step": 42305 + }, + { + "epoch": 1.55, + "learning_rate": 4.5255795217343535e-08, + "loss": 0.6937, + "step": 42306 + }, + { + "epoch": 1.55, + "learning_rate": 4.5252133152671475e-08, + "loss": 0.6846, + "step": 42307 + }, + { + "epoch": 1.55, + "learning_rate": 4.5248471087999415e-08, + "loss": 0.6844, + "step": 42308 + }, + { + "epoch": 1.55, + "learning_rate": 4.5244809023327355e-08, + "loss": 0.6932, + "step": 42309 + }, + { + "epoch": 1.55, + "learning_rate": 4.524114695865529e-08, + "loss": 0.6855, + "step": 42310 + }, + { + "epoch": 1.55, + "learning_rate": 4.523748489398322e-08, + "loss": 0.6893, + "step": 42311 + }, + { + "epoch": 1.55, + "learning_rate": 4.523382282931116e-08, + "loss": 0.6787, + "step": 42312 + }, + { + "epoch": 1.55, + "learning_rate": 4.52301607646391e-08, + "loss": 0.6874, + "step": 42313 + }, + { + "epoch": 1.55, + "learning_rate": 4.522649869996704e-08, + "loss": 0.7, + "step": 42314 + }, + { + "epoch": 1.55, + "learning_rate": 4.5222836635294974e-08, + "loss": 0.6885, + "step": 42315 + }, + { + "epoch": 1.55, + "learning_rate": 4.5219174570622914e-08, + "loss": 0.688, + "step": 42316 + }, + { + "epoch": 1.55, + "learning_rate": 4.5215512505950854e-08, + "loss": 0.6856, + "step": 42317 + }, + { + "epoch": 1.55, + "learning_rate": 4.5211850441278794e-08, + "loss": 0.6819, + "step": 42318 + }, + { + "epoch": 1.55, + "learning_rate": 4.520818837660673e-08, + "loss": 0.698, + "step": 42319 + }, + { + "epoch": 1.55, + "learning_rate": 4.520452631193467e-08, + "loss": 0.6869, + "step": 42320 + }, + { + "epoch": 1.55, + "learning_rate": 4.520086424726261e-08, + "loss": 0.6869, + "step": 42321 + }, + { + "epoch": 1.55, + "learning_rate": 4.519720218259054e-08, + "loss": 0.6893, + "step": 42322 + }, + { + "epoch": 1.55, + "learning_rate": 4.519354011791848e-08, + "loss": 0.6871, + "step": 42323 + }, + { + "epoch": 1.55, + "learning_rate": 4.5189878053246414e-08, + "loss": 0.7006, + "step": 42324 + }, + { + "epoch": 1.55, + "learning_rate": 4.5186215988574354e-08, + "loss": 0.7043, + "step": 42325 + }, + { + "epoch": 1.55, + "learning_rate": 4.5182553923902294e-08, + "loss": 0.6908, + "step": 42326 + }, + { + "epoch": 1.55, + "learning_rate": 4.5178891859230234e-08, + "loss": 0.7055, + "step": 42327 + }, + { + "epoch": 1.55, + "learning_rate": 4.517522979455817e-08, + "loss": 0.7029, + "step": 42328 + }, + { + "epoch": 1.55, + "learning_rate": 4.517156772988611e-08, + "loss": 0.6931, + "step": 42329 + }, + { + "epoch": 1.55, + "learning_rate": 4.516790566521405e-08, + "loss": 0.6936, + "step": 42330 + }, + { + "epoch": 1.55, + "learning_rate": 4.516424360054199e-08, + "loss": 0.7003, + "step": 42331 + }, + { + "epoch": 1.55, + "learning_rate": 4.516058153586993e-08, + "loss": 0.6951, + "step": 42332 + }, + { + "epoch": 1.55, + "learning_rate": 4.515691947119786e-08, + "loss": 0.701, + "step": 42333 + }, + { + "epoch": 1.55, + "learning_rate": 4.5153257406525793e-08, + "loss": 0.6887, + "step": 42334 + }, + { + "epoch": 1.55, + "learning_rate": 4.514959534185373e-08, + "loss": 0.6955, + "step": 42335 + }, + { + "epoch": 1.55, + "learning_rate": 4.514593327718167e-08, + "loss": 0.7045, + "step": 42336 + }, + { + "epoch": 1.55, + "learning_rate": 4.5142271212509607e-08, + "loss": 0.6804, + "step": 42337 + }, + { + "epoch": 1.55, + "learning_rate": 4.5138609147837546e-08, + "loss": 0.7029, + "step": 42338 + }, + { + "epoch": 1.55, + "learning_rate": 4.5134947083165486e-08, + "loss": 0.6946, + "step": 42339 + }, + { + "epoch": 1.55, + "learning_rate": 4.5131285018493426e-08, + "loss": 0.6837, + "step": 42340 + }, + { + "epoch": 1.55, + "learning_rate": 4.5127622953821366e-08, + "loss": 0.689, + "step": 42341 + }, + { + "epoch": 1.55, + "learning_rate": 4.51239608891493e-08, + "loss": 0.6806, + "step": 42342 + }, + { + "epoch": 1.55, + "learning_rate": 4.512029882447724e-08, + "loss": 0.6891, + "step": 42343 + }, + { + "epoch": 1.55, + "learning_rate": 4.511663675980518e-08, + "loss": 0.6959, + "step": 42344 + }, + { + "epoch": 1.55, + "learning_rate": 4.511297469513311e-08, + "loss": 0.695, + "step": 42345 + }, + { + "epoch": 1.55, + "learning_rate": 4.510931263046105e-08, + "loss": 0.6889, + "step": 42346 + }, + { + "epoch": 1.55, + "learning_rate": 4.5105650565788986e-08, + "loss": 0.6934, + "step": 42347 + }, + { + "epoch": 1.55, + "learning_rate": 4.5101988501116926e-08, + "loss": 0.686, + "step": 42348 + }, + { + "epoch": 1.55, + "learning_rate": 4.5098326436444866e-08, + "loss": 0.7005, + "step": 42349 + }, + { + "epoch": 1.55, + "learning_rate": 4.5094664371772806e-08, + "loss": 0.6891, + "step": 42350 + }, + { + "epoch": 1.55, + "learning_rate": 4.509100230710074e-08, + "loss": 0.6897, + "step": 42351 + }, + { + "epoch": 1.55, + "learning_rate": 4.508734024242868e-08, + "loss": 0.687, + "step": 42352 + }, + { + "epoch": 1.55, + "learning_rate": 4.508367817775662e-08, + "loss": 0.7076, + "step": 42353 + }, + { + "epoch": 1.55, + "learning_rate": 4.508001611308456e-08, + "loss": 0.7017, + "step": 42354 + }, + { + "epoch": 1.55, + "learning_rate": 4.507635404841249e-08, + "loss": 0.7066, + "step": 42355 + }, + { + "epoch": 1.55, + "learning_rate": 4.507269198374043e-08, + "loss": 0.6922, + "step": 42356 + }, + { + "epoch": 1.55, + "learning_rate": 4.506902991906837e-08, + "loss": 0.6886, + "step": 42357 + }, + { + "epoch": 1.55, + "learning_rate": 4.5065367854396305e-08, + "loss": 0.6758, + "step": 42358 + }, + { + "epoch": 1.55, + "learning_rate": 4.5061705789724245e-08, + "loss": 0.708, + "step": 42359 + }, + { + "epoch": 1.55, + "learning_rate": 4.505804372505218e-08, + "loss": 0.6929, + "step": 42360 + }, + { + "epoch": 1.55, + "learning_rate": 4.505438166038012e-08, + "loss": 0.7083, + "step": 42361 + }, + { + "epoch": 1.55, + "learning_rate": 4.505071959570806e-08, + "loss": 0.6932, + "step": 42362 + }, + { + "epoch": 1.55, + "learning_rate": 4.5047057531036e-08, + "loss": 0.6951, + "step": 42363 + }, + { + "epoch": 1.55, + "learning_rate": 4.504339546636393e-08, + "loss": 0.6795, + "step": 42364 + }, + { + "epoch": 1.55, + "learning_rate": 4.503973340169187e-08, + "loss": 0.6923, + "step": 42365 + }, + { + "epoch": 1.55, + "learning_rate": 4.503607133701981e-08, + "loss": 0.6933, + "step": 42366 + }, + { + "epoch": 1.55, + "learning_rate": 4.503240927234775e-08, + "loss": 0.6969, + "step": 42367 + }, + { + "epoch": 1.55, + "learning_rate": 4.5028747207675685e-08, + "loss": 0.7045, + "step": 42368 + }, + { + "epoch": 1.55, + "learning_rate": 4.5025085143003625e-08, + "loss": 0.7073, + "step": 42369 + }, + { + "epoch": 1.55, + "learning_rate": 4.502142307833156e-08, + "loss": 0.6885, + "step": 42370 + }, + { + "epoch": 1.55, + "learning_rate": 4.50177610136595e-08, + "loss": 0.702, + "step": 42371 + }, + { + "epoch": 1.55, + "learning_rate": 4.501409894898744e-08, + "loss": 0.6812, + "step": 42372 + }, + { + "epoch": 1.55, + "learning_rate": 4.501043688431537e-08, + "loss": 0.6859, + "step": 42373 + }, + { + "epoch": 1.55, + "learning_rate": 4.500677481964331e-08, + "loss": 0.6849, + "step": 42374 + }, + { + "epoch": 1.55, + "learning_rate": 4.500311275497125e-08, + "loss": 0.6906, + "step": 42375 + }, + { + "epoch": 1.55, + "learning_rate": 4.499945069029919e-08, + "loss": 0.6945, + "step": 42376 + }, + { + "epoch": 1.55, + "learning_rate": 4.499578862562713e-08, + "loss": 0.6893, + "step": 42377 + }, + { + "epoch": 1.55, + "learning_rate": 4.4992126560955064e-08, + "loss": 0.6863, + "step": 42378 + }, + { + "epoch": 1.55, + "learning_rate": 4.4988464496283004e-08, + "loss": 0.6922, + "step": 42379 + }, + { + "epoch": 1.55, + "learning_rate": 4.4984802431610944e-08, + "loss": 0.6992, + "step": 42380 + }, + { + "epoch": 1.55, + "learning_rate": 4.498114036693888e-08, + "loss": 0.6885, + "step": 42381 + }, + { + "epoch": 1.55, + "learning_rate": 4.497747830226681e-08, + "loss": 0.6904, + "step": 42382 + }, + { + "epoch": 1.55, + "learning_rate": 4.497381623759475e-08, + "loss": 0.6962, + "step": 42383 + }, + { + "epoch": 1.55, + "learning_rate": 4.497015417292269e-08, + "loss": 0.6904, + "step": 42384 + }, + { + "epoch": 1.55, + "learning_rate": 4.496649210825063e-08, + "loss": 0.7002, + "step": 42385 + }, + { + "epoch": 1.55, + "learning_rate": 4.496283004357857e-08, + "loss": 0.6857, + "step": 42386 + }, + { + "epoch": 1.55, + "learning_rate": 4.4959167978906504e-08, + "loss": 0.6956, + "step": 42387 + }, + { + "epoch": 1.55, + "learning_rate": 4.4955505914234444e-08, + "loss": 0.6861, + "step": 42388 + }, + { + "epoch": 1.55, + "learning_rate": 4.4951843849562384e-08, + "loss": 0.7014, + "step": 42389 + }, + { + "epoch": 1.55, + "learning_rate": 4.4948181784890323e-08, + "loss": 0.6926, + "step": 42390 + }, + { + "epoch": 1.55, + "learning_rate": 4.494451972021826e-08, + "loss": 0.6827, + "step": 42391 + }, + { + "epoch": 1.55, + "learning_rate": 4.49408576555462e-08, + "loss": 0.6881, + "step": 42392 + }, + { + "epoch": 1.55, + "learning_rate": 4.493719559087413e-08, + "loss": 0.697, + "step": 42393 + }, + { + "epoch": 1.55, + "learning_rate": 4.493353352620207e-08, + "loss": 0.6831, + "step": 42394 + }, + { + "epoch": 1.55, + "learning_rate": 4.492987146153001e-08, + "loss": 0.6969, + "step": 42395 + }, + { + "epoch": 1.55, + "learning_rate": 4.492620939685794e-08, + "loss": 0.6722, + "step": 42396 + }, + { + "epoch": 1.55, + "learning_rate": 4.492254733218588e-08, + "loss": 0.6987, + "step": 42397 + }, + { + "epoch": 1.55, + "learning_rate": 4.491888526751382e-08, + "loss": 0.6972, + "step": 42398 + }, + { + "epoch": 1.55, + "learning_rate": 4.491522320284176e-08, + "loss": 0.7045, + "step": 42399 + }, + { + "epoch": 1.55, + "learning_rate": 4.4911561138169696e-08, + "loss": 0.6941, + "step": 42400 + }, + { + "epoch": 1.55, + "learning_rate": 4.4907899073497636e-08, + "loss": 0.6988, + "step": 42401 + }, + { + "epoch": 1.55, + "learning_rate": 4.4904237008825576e-08, + "loss": 0.6878, + "step": 42402 + }, + { + "epoch": 1.55, + "learning_rate": 4.4900574944153516e-08, + "loss": 0.6949, + "step": 42403 + }, + { + "epoch": 1.55, + "learning_rate": 4.489691287948145e-08, + "loss": 0.6927, + "step": 42404 + }, + { + "epoch": 1.55, + "learning_rate": 4.489325081480938e-08, + "loss": 0.686, + "step": 42405 + }, + { + "epoch": 1.55, + "learning_rate": 4.488958875013732e-08, + "loss": 0.6848, + "step": 42406 + }, + { + "epoch": 1.55, + "learning_rate": 4.488592668546526e-08, + "loss": 0.6935, + "step": 42407 + }, + { + "epoch": 1.55, + "learning_rate": 4.48822646207932e-08, + "loss": 0.6936, + "step": 42408 + }, + { + "epoch": 1.55, + "learning_rate": 4.4878602556121136e-08, + "loss": 0.7026, + "step": 42409 + }, + { + "epoch": 1.55, + "learning_rate": 4.4874940491449076e-08, + "loss": 0.6965, + "step": 42410 + }, + { + "epoch": 1.55, + "learning_rate": 4.4871278426777016e-08, + "loss": 0.6938, + "step": 42411 + }, + { + "epoch": 1.55, + "learning_rate": 4.4867616362104956e-08, + "loss": 0.6844, + "step": 42412 + }, + { + "epoch": 1.55, + "learning_rate": 4.4863954297432896e-08, + "loss": 0.6963, + "step": 42413 + }, + { + "epoch": 1.55, + "learning_rate": 4.486029223276083e-08, + "loss": 0.6895, + "step": 42414 + }, + { + "epoch": 1.55, + "learning_rate": 4.485663016808877e-08, + "loss": 0.6929, + "step": 42415 + }, + { + "epoch": 1.55, + "learning_rate": 4.48529681034167e-08, + "loss": 0.6969, + "step": 42416 + }, + { + "epoch": 1.55, + "learning_rate": 4.484930603874464e-08, + "loss": 0.6923, + "step": 42417 + }, + { + "epoch": 1.55, + "learning_rate": 4.4845643974072575e-08, + "loss": 0.6926, + "step": 42418 + }, + { + "epoch": 1.55, + "learning_rate": 4.4841981909400515e-08, + "loss": 0.6862, + "step": 42419 + }, + { + "epoch": 1.55, + "learning_rate": 4.4838319844728455e-08, + "loss": 0.6913, + "step": 42420 + }, + { + "epoch": 1.55, + "learning_rate": 4.4834657780056395e-08, + "loss": 0.6783, + "step": 42421 + }, + { + "epoch": 1.55, + "learning_rate": 4.4830995715384335e-08, + "loss": 0.6918, + "step": 42422 + }, + { + "epoch": 1.55, + "learning_rate": 4.482733365071227e-08, + "loss": 0.6867, + "step": 42423 + }, + { + "epoch": 1.55, + "learning_rate": 4.482367158604021e-08, + "loss": 0.6952, + "step": 42424 + }, + { + "epoch": 1.55, + "learning_rate": 4.482000952136815e-08, + "loss": 0.6916, + "step": 42425 + }, + { + "epoch": 1.55, + "learning_rate": 4.481634745669609e-08, + "loss": 0.6866, + "step": 42426 + }, + { + "epoch": 1.55, + "learning_rate": 4.481268539202402e-08, + "loss": 0.6929, + "step": 42427 + }, + { + "epoch": 1.55, + "learning_rate": 4.4809023327351955e-08, + "loss": 0.6891, + "step": 42428 + }, + { + "epoch": 1.55, + "learning_rate": 4.4805361262679895e-08, + "loss": 0.7015, + "step": 42429 + }, + { + "epoch": 1.55, + "learning_rate": 4.4801699198007835e-08, + "loss": 0.6992, + "step": 42430 + }, + { + "epoch": 1.55, + "learning_rate": 4.4798037133335775e-08, + "loss": 0.7088, + "step": 42431 + }, + { + "epoch": 1.55, + "learning_rate": 4.479437506866371e-08, + "loss": 0.6832, + "step": 42432 + }, + { + "epoch": 1.55, + "learning_rate": 4.479071300399165e-08, + "loss": 0.6923, + "step": 42433 + }, + { + "epoch": 1.55, + "learning_rate": 4.478705093931959e-08, + "loss": 0.6903, + "step": 42434 + }, + { + "epoch": 1.55, + "learning_rate": 4.478338887464753e-08, + "loss": 0.7065, + "step": 42435 + }, + { + "epoch": 1.55, + "learning_rate": 4.477972680997546e-08, + "loss": 0.6919, + "step": 42436 + }, + { + "epoch": 1.55, + "learning_rate": 4.47760647453034e-08, + "loss": 0.698, + "step": 42437 + }, + { + "epoch": 1.55, + "learning_rate": 4.477240268063134e-08, + "loss": 0.7053, + "step": 42438 + }, + { + "epoch": 1.55, + "learning_rate": 4.4768740615959274e-08, + "loss": 0.6935, + "step": 42439 + }, + { + "epoch": 1.55, + "learning_rate": 4.4765078551287214e-08, + "loss": 0.6845, + "step": 42440 + }, + { + "epoch": 1.55, + "learning_rate": 4.476141648661515e-08, + "loss": 0.6951, + "step": 42441 + }, + { + "epoch": 1.55, + "learning_rate": 4.475775442194309e-08, + "loss": 0.7007, + "step": 42442 + }, + { + "epoch": 1.55, + "learning_rate": 4.475409235727103e-08, + "loss": 0.6973, + "step": 42443 + }, + { + "epoch": 1.55, + "learning_rate": 4.475043029259897e-08, + "loss": 0.6994, + "step": 42444 + }, + { + "epoch": 1.55, + "learning_rate": 4.47467682279269e-08, + "loss": 0.683, + "step": 42445 + }, + { + "epoch": 1.55, + "learning_rate": 4.474310616325484e-08, + "loss": 0.7011, + "step": 42446 + }, + { + "epoch": 1.55, + "learning_rate": 4.473944409858278e-08, + "loss": 0.697, + "step": 42447 + }, + { + "epoch": 1.55, + "learning_rate": 4.473578203391072e-08, + "loss": 0.6924, + "step": 42448 + }, + { + "epoch": 1.55, + "learning_rate": 4.473211996923866e-08, + "loss": 0.7089, + "step": 42449 + }, + { + "epoch": 1.55, + "learning_rate": 4.4728457904566593e-08, + "loss": 0.6907, + "step": 42450 + }, + { + "epoch": 1.55, + "learning_rate": 4.4724795839894533e-08, + "loss": 0.6957, + "step": 42451 + }, + { + "epoch": 1.55, + "learning_rate": 4.472113377522247e-08, + "loss": 0.6927, + "step": 42452 + }, + { + "epoch": 1.55, + "learning_rate": 4.4717471710550407e-08, + "loss": 0.6965, + "step": 42453 + }, + { + "epoch": 1.55, + "learning_rate": 4.471380964587834e-08, + "loss": 0.6929, + "step": 42454 + }, + { + "epoch": 1.55, + "learning_rate": 4.471014758120628e-08, + "loss": 0.6924, + "step": 42455 + }, + { + "epoch": 1.55, + "learning_rate": 4.470648551653422e-08, + "loss": 0.6817, + "step": 42456 + }, + { + "epoch": 1.55, + "learning_rate": 4.470282345186216e-08, + "loss": 0.6959, + "step": 42457 + }, + { + "epoch": 1.55, + "learning_rate": 4.469916138719009e-08, + "loss": 0.6961, + "step": 42458 + }, + { + "epoch": 1.55, + "learning_rate": 4.469549932251803e-08, + "loss": 0.6945, + "step": 42459 + }, + { + "epoch": 1.55, + "learning_rate": 4.469183725784597e-08, + "loss": 0.6962, + "step": 42460 + }, + { + "epoch": 1.55, + "learning_rate": 4.468817519317391e-08, + "loss": 0.7005, + "step": 42461 + }, + { + "epoch": 1.55, + "learning_rate": 4.468451312850185e-08, + "loss": 0.7015, + "step": 42462 + }, + { + "epoch": 1.55, + "learning_rate": 4.4680851063829786e-08, + "loss": 0.7057, + "step": 42463 + }, + { + "epoch": 1.55, + "learning_rate": 4.467718899915772e-08, + "loss": 0.6932, + "step": 42464 + }, + { + "epoch": 1.55, + "learning_rate": 4.467352693448566e-08, + "loss": 0.688, + "step": 42465 + }, + { + "epoch": 1.55, + "learning_rate": 4.46698648698136e-08, + "loss": 0.6901, + "step": 42466 + }, + { + "epoch": 1.55, + "learning_rate": 4.466620280514153e-08, + "loss": 0.6871, + "step": 42467 + }, + { + "epoch": 1.55, + "learning_rate": 4.466254074046947e-08, + "loss": 0.695, + "step": 42468 + }, + { + "epoch": 1.55, + "learning_rate": 4.465887867579741e-08, + "loss": 0.6959, + "step": 42469 + }, + { + "epoch": 1.55, + "learning_rate": 4.465521661112535e-08, + "loss": 0.6853, + "step": 42470 + }, + { + "epoch": 1.55, + "learning_rate": 4.465155454645329e-08, + "loss": 0.6941, + "step": 42471 + }, + { + "epoch": 1.55, + "learning_rate": 4.4647892481781226e-08, + "loss": 0.6885, + "step": 42472 + }, + { + "epoch": 1.55, + "learning_rate": 4.4644230417109165e-08, + "loss": 0.7023, + "step": 42473 + }, + { + "epoch": 1.55, + "learning_rate": 4.4640568352437105e-08, + "loss": 0.7137, + "step": 42474 + }, + { + "epoch": 1.55, + "learning_rate": 4.463690628776504e-08, + "loss": 0.6935, + "step": 42475 + }, + { + "epoch": 1.55, + "learning_rate": 4.463324422309297e-08, + "loss": 0.6944, + "step": 42476 + }, + { + "epoch": 1.55, + "learning_rate": 4.462958215842091e-08, + "loss": 0.6886, + "step": 42477 + }, + { + "epoch": 1.55, + "learning_rate": 4.462592009374885e-08, + "loss": 0.6889, + "step": 42478 + }, + { + "epoch": 1.55, + "learning_rate": 4.462225802907679e-08, + "loss": 0.7075, + "step": 42479 + }, + { + "epoch": 1.55, + "learning_rate": 4.461859596440473e-08, + "loss": 0.6853, + "step": 42480 + }, + { + "epoch": 1.55, + "learning_rate": 4.4614933899732665e-08, + "loss": 0.6896, + "step": 42481 + }, + { + "epoch": 1.55, + "learning_rate": 4.4611271835060605e-08, + "loss": 0.6972, + "step": 42482 + }, + { + "epoch": 1.55, + "learning_rate": 4.4607609770388545e-08, + "loss": 0.6925, + "step": 42483 + }, + { + "epoch": 1.55, + "learning_rate": 4.4603947705716485e-08, + "loss": 0.6862, + "step": 42484 + }, + { + "epoch": 1.55, + "learning_rate": 4.460028564104442e-08, + "loss": 0.6916, + "step": 42485 + }, + { + "epoch": 1.55, + "learning_rate": 4.459662357637236e-08, + "loss": 0.6985, + "step": 42486 + }, + { + "epoch": 1.55, + "learning_rate": 4.459296151170029e-08, + "loss": 0.6954, + "step": 42487 + }, + { + "epoch": 1.55, + "learning_rate": 4.458929944702823e-08, + "loss": 0.7056, + "step": 42488 + }, + { + "epoch": 1.55, + "learning_rate": 4.458563738235617e-08, + "loss": 0.6953, + "step": 42489 + }, + { + "epoch": 1.55, + "learning_rate": 4.4581975317684105e-08, + "loss": 0.6859, + "step": 42490 + }, + { + "epoch": 1.55, + "learning_rate": 4.4578313253012045e-08, + "loss": 0.6703, + "step": 42491 + }, + { + "epoch": 1.55, + "learning_rate": 4.4574651188339984e-08, + "loss": 0.6887, + "step": 42492 + }, + { + "epoch": 1.55, + "learning_rate": 4.4570989123667924e-08, + "loss": 0.6887, + "step": 42493 + }, + { + "epoch": 1.55, + "learning_rate": 4.456732705899586e-08, + "loss": 0.686, + "step": 42494 + }, + { + "epoch": 1.55, + "learning_rate": 4.45636649943238e-08, + "loss": 0.6959, + "step": 42495 + }, + { + "epoch": 1.55, + "learning_rate": 4.456000292965174e-08, + "loss": 0.7026, + "step": 42496 + }, + { + "epoch": 1.55, + "learning_rate": 4.455634086497968e-08, + "loss": 0.6858, + "step": 42497 + }, + { + "epoch": 1.55, + "learning_rate": 4.455267880030761e-08, + "loss": 0.6899, + "step": 42498 + }, + { + "epoch": 1.55, + "learning_rate": 4.4549016735635544e-08, + "loss": 0.7073, + "step": 42499 + }, + { + "epoch": 1.55, + "learning_rate": 4.4545354670963484e-08, + "loss": 0.6853, + "step": 42500 + }, + { + "epoch": 1.55, + "learning_rate": 4.4541692606291424e-08, + "loss": 0.6888, + "step": 42501 + }, + { + "epoch": 1.56, + "learning_rate": 4.4538030541619364e-08, + "loss": 0.6941, + "step": 42502 + }, + { + "epoch": 1.56, + "learning_rate": 4.45343684769473e-08, + "loss": 0.6953, + "step": 42503 + }, + { + "epoch": 1.56, + "learning_rate": 4.453070641227524e-08, + "loss": 0.6965, + "step": 42504 + }, + { + "epoch": 1.56, + "learning_rate": 4.452704434760318e-08, + "loss": 0.6993, + "step": 42505 + }, + { + "epoch": 1.56, + "learning_rate": 4.452338228293112e-08, + "loss": 0.6838, + "step": 42506 + }, + { + "epoch": 1.56, + "learning_rate": 4.451972021825906e-08, + "loss": 0.6848, + "step": 42507 + }, + { + "epoch": 1.56, + "learning_rate": 4.451605815358699e-08, + "loss": 0.7012, + "step": 42508 + }, + { + "epoch": 1.56, + "learning_rate": 4.451239608891493e-08, + "loss": 0.6976, + "step": 42509 + }, + { + "epoch": 1.56, + "learning_rate": 4.4508734024242863e-08, + "loss": 0.6832, + "step": 42510 + }, + { + "epoch": 1.56, + "learning_rate": 4.4505071959570803e-08, + "loss": 0.7033, + "step": 42511 + }, + { + "epoch": 1.56, + "learning_rate": 4.4501409894898737e-08, + "loss": 0.6974, + "step": 42512 + }, + { + "epoch": 1.56, + "learning_rate": 4.4497747830226677e-08, + "loss": 0.6956, + "step": 42513 + }, + { + "epoch": 1.56, + "learning_rate": 4.4494085765554617e-08, + "loss": 0.691, + "step": 42514 + }, + { + "epoch": 1.56, + "learning_rate": 4.4490423700882556e-08, + "loss": 0.6971, + "step": 42515 + }, + { + "epoch": 1.56, + "learning_rate": 4.4486761636210496e-08, + "loss": 0.6904, + "step": 42516 + }, + { + "epoch": 1.56, + "learning_rate": 4.448309957153843e-08, + "loss": 0.6994, + "step": 42517 + }, + { + "epoch": 1.56, + "learning_rate": 4.447943750686637e-08, + "loss": 0.7019, + "step": 42518 + }, + { + "epoch": 1.56, + "learning_rate": 4.447577544219431e-08, + "loss": 0.7017, + "step": 42519 + }, + { + "epoch": 1.56, + "learning_rate": 4.447211337752225e-08, + "loss": 0.6871, + "step": 42520 + }, + { + "epoch": 1.56, + "learning_rate": 4.446845131285018e-08, + "loss": 0.6971, + "step": 42521 + }, + { + "epoch": 1.56, + "learning_rate": 4.446478924817812e-08, + "loss": 0.6924, + "step": 42522 + }, + { + "epoch": 1.56, + "learning_rate": 4.4461127183506056e-08, + "loss": 0.6953, + "step": 42523 + }, + { + "epoch": 1.56, + "learning_rate": 4.4457465118833996e-08, + "loss": 0.6846, + "step": 42524 + }, + { + "epoch": 1.56, + "learning_rate": 4.4453803054161936e-08, + "loss": 0.707, + "step": 42525 + }, + { + "epoch": 1.56, + "learning_rate": 4.445014098948987e-08, + "loss": 0.7014, + "step": 42526 + }, + { + "epoch": 1.56, + "learning_rate": 4.444647892481781e-08, + "loss": 0.6857, + "step": 42527 + }, + { + "epoch": 1.56, + "learning_rate": 4.444281686014575e-08, + "loss": 0.6992, + "step": 42528 + }, + { + "epoch": 1.56, + "learning_rate": 4.443915479547369e-08, + "loss": 0.689, + "step": 42529 + }, + { + "epoch": 1.56, + "learning_rate": 4.443549273080162e-08, + "loss": 0.6819, + "step": 42530 + }, + { + "epoch": 1.56, + "learning_rate": 4.443183066612956e-08, + "loss": 0.6979, + "step": 42531 + }, + { + "epoch": 1.56, + "learning_rate": 4.44281686014575e-08, + "loss": 0.6878, + "step": 42532 + }, + { + "epoch": 1.56, + "learning_rate": 4.442450653678544e-08, + "loss": 0.6975, + "step": 42533 + }, + { + "epoch": 1.56, + "learning_rate": 4.4420844472113375e-08, + "loss": 0.6999, + "step": 42534 + }, + { + "epoch": 1.56, + "learning_rate": 4.441718240744131e-08, + "loss": 0.6796, + "step": 42535 + }, + { + "epoch": 1.56, + "learning_rate": 4.441352034276925e-08, + "loss": 0.6999, + "step": 42536 + }, + { + "epoch": 1.56, + "learning_rate": 4.440985827809719e-08, + "loss": 0.6903, + "step": 42537 + }, + { + "epoch": 1.56, + "learning_rate": 4.440619621342513e-08, + "loss": 0.6938, + "step": 42538 + }, + { + "epoch": 1.56, + "learning_rate": 4.440253414875306e-08, + "loss": 0.6903, + "step": 42539 + }, + { + "epoch": 1.56, + "learning_rate": 4.4398872084081e-08, + "loss": 0.6881, + "step": 42540 + }, + { + "epoch": 1.56, + "learning_rate": 4.439521001940894e-08, + "loss": 0.6967, + "step": 42541 + }, + { + "epoch": 1.56, + "learning_rate": 4.439154795473688e-08, + "loss": 0.7036, + "step": 42542 + }, + { + "epoch": 1.56, + "learning_rate": 4.438788589006482e-08, + "loss": 0.6931, + "step": 42543 + }, + { + "epoch": 1.56, + "learning_rate": 4.4384223825392755e-08, + "loss": 0.6972, + "step": 42544 + }, + { + "epoch": 1.56, + "learning_rate": 4.4380561760720695e-08, + "loss": 0.6973, + "step": 42545 + }, + { + "epoch": 1.56, + "learning_rate": 4.437689969604863e-08, + "loss": 0.678, + "step": 42546 + }, + { + "epoch": 1.56, + "learning_rate": 4.437323763137657e-08, + "loss": 0.6892, + "step": 42547 + }, + { + "epoch": 1.56, + "learning_rate": 4.43695755667045e-08, + "loss": 0.6743, + "step": 42548 + }, + { + "epoch": 1.56, + "learning_rate": 4.436591350203244e-08, + "loss": 0.7017, + "step": 42549 + }, + { + "epoch": 1.56, + "learning_rate": 4.436225143736038e-08, + "loss": 0.6801, + "step": 42550 + }, + { + "epoch": 1.56, + "learning_rate": 4.435858937268832e-08, + "loss": 0.7018, + "step": 42551 + }, + { + "epoch": 1.56, + "learning_rate": 4.435492730801626e-08, + "loss": 0.6923, + "step": 42552 + }, + { + "epoch": 1.56, + "learning_rate": 4.4351265243344194e-08, + "loss": 0.6991, + "step": 42553 + }, + { + "epoch": 1.56, + "learning_rate": 4.4347603178672134e-08, + "loss": 0.698, + "step": 42554 + }, + { + "epoch": 1.56, + "learning_rate": 4.4343941114000074e-08, + "loss": 0.6791, + "step": 42555 + }, + { + "epoch": 1.56, + "learning_rate": 4.4340279049328014e-08, + "loss": 0.687, + "step": 42556 + }, + { + "epoch": 1.56, + "learning_rate": 4.433661698465595e-08, + "loss": 0.7017, + "step": 42557 + }, + { + "epoch": 1.56, + "learning_rate": 4.433295491998388e-08, + "loss": 0.6948, + "step": 42558 + }, + { + "epoch": 1.56, + "learning_rate": 4.432929285531182e-08, + "loss": 0.689, + "step": 42559 + }, + { + "epoch": 1.56, + "learning_rate": 4.432563079063976e-08, + "loss": 0.6963, + "step": 42560 + }, + { + "epoch": 1.56, + "learning_rate": 4.43219687259677e-08, + "loss": 0.6924, + "step": 42561 + }, + { + "epoch": 1.56, + "learning_rate": 4.4318306661295634e-08, + "loss": 0.6968, + "step": 42562 + }, + { + "epoch": 1.56, + "learning_rate": 4.4314644596623574e-08, + "loss": 0.6945, + "step": 42563 + }, + { + "epoch": 1.56, + "learning_rate": 4.4310982531951514e-08, + "loss": 0.6838, + "step": 42564 + }, + { + "epoch": 1.56, + "learning_rate": 4.4307320467279454e-08, + "loss": 0.6843, + "step": 42565 + }, + { + "epoch": 1.56, + "learning_rate": 4.430365840260739e-08, + "loss": 0.7011, + "step": 42566 + }, + { + "epoch": 1.56, + "learning_rate": 4.429999633793533e-08, + "loss": 0.7033, + "step": 42567 + }, + { + "epoch": 1.56, + "learning_rate": 4.429633427326327e-08, + "loss": 0.6966, + "step": 42568 + }, + { + "epoch": 1.56, + "learning_rate": 4.42926722085912e-08, + "loss": 0.6832, + "step": 42569 + }, + { + "epoch": 1.56, + "learning_rate": 4.428901014391914e-08, + "loss": 0.6903, + "step": 42570 + }, + { + "epoch": 1.56, + "learning_rate": 4.4285348079247073e-08, + "loss": 0.707, + "step": 42571 + }, + { + "epoch": 1.56, + "learning_rate": 4.428168601457501e-08, + "loss": 0.704, + "step": 42572 + }, + { + "epoch": 1.56, + "learning_rate": 4.427802394990295e-08, + "loss": 0.6923, + "step": 42573 + }, + { + "epoch": 1.56, + "learning_rate": 4.427436188523089e-08, + "loss": 0.7003, + "step": 42574 + }, + { + "epoch": 1.56, + "learning_rate": 4.4270699820558826e-08, + "loss": 0.6876, + "step": 42575 + }, + { + "epoch": 1.56, + "learning_rate": 4.4267037755886766e-08, + "loss": 0.6908, + "step": 42576 + }, + { + "epoch": 1.56, + "learning_rate": 4.4263375691214706e-08, + "loss": 0.6775, + "step": 42577 + }, + { + "epoch": 1.56, + "learning_rate": 4.4259713626542646e-08, + "loss": 0.6903, + "step": 42578 + }, + { + "epoch": 1.56, + "learning_rate": 4.4256051561870586e-08, + "loss": 0.6924, + "step": 42579 + }, + { + "epoch": 1.56, + "learning_rate": 4.425238949719852e-08, + "loss": 0.6792, + "step": 42580 + }, + { + "epoch": 1.56, + "learning_rate": 4.424872743252645e-08, + "loss": 0.6868, + "step": 42581 + }, + { + "epoch": 1.56, + "learning_rate": 4.424506536785439e-08, + "loss": 0.6891, + "step": 42582 + }, + { + "epoch": 1.56, + "learning_rate": 4.424140330318233e-08, + "loss": 0.6879, + "step": 42583 + }, + { + "epoch": 1.56, + "learning_rate": 4.4237741238510266e-08, + "loss": 0.6919, + "step": 42584 + }, + { + "epoch": 1.56, + "learning_rate": 4.4234079173838206e-08, + "loss": 0.7084, + "step": 42585 + }, + { + "epoch": 1.56, + "learning_rate": 4.4230417109166146e-08, + "loss": 0.6854, + "step": 42586 + }, + { + "epoch": 1.56, + "learning_rate": 4.4226755044494086e-08, + "loss": 0.6932, + "step": 42587 + }, + { + "epoch": 1.56, + "learning_rate": 4.4223092979822026e-08, + "loss": 0.6973, + "step": 42588 + }, + { + "epoch": 1.56, + "learning_rate": 4.421943091514996e-08, + "loss": 0.6934, + "step": 42589 + }, + { + "epoch": 1.56, + "learning_rate": 4.42157688504779e-08, + "loss": 0.6937, + "step": 42590 + }, + { + "epoch": 1.56, + "learning_rate": 4.421210678580584e-08, + "loss": 0.6948, + "step": 42591 + }, + { + "epoch": 1.56, + "learning_rate": 4.420844472113377e-08, + "loss": 0.6924, + "step": 42592 + }, + { + "epoch": 1.56, + "learning_rate": 4.420478265646171e-08, + "loss": 0.6877, + "step": 42593 + }, + { + "epoch": 1.56, + "learning_rate": 4.4201120591789645e-08, + "loss": 0.6949, + "step": 42594 + }, + { + "epoch": 1.56, + "learning_rate": 4.4197458527117585e-08, + "loss": 0.701, + "step": 42595 + }, + { + "epoch": 1.56, + "learning_rate": 4.4193796462445525e-08, + "loss": 0.6937, + "step": 42596 + }, + { + "epoch": 1.56, + "learning_rate": 4.4190134397773465e-08, + "loss": 0.6972, + "step": 42597 + }, + { + "epoch": 1.56, + "learning_rate": 4.41864723331014e-08, + "loss": 0.6823, + "step": 42598 + }, + { + "epoch": 1.56, + "learning_rate": 4.418281026842934e-08, + "loss": 0.6978, + "step": 42599 + }, + { + "epoch": 1.56, + "learning_rate": 4.417914820375728e-08, + "loss": 0.6751, + "step": 42600 + }, + { + "epoch": 1.56, + "learning_rate": 4.417548613908522e-08, + "loss": 0.6987, + "step": 42601 + }, + { + "epoch": 1.56, + "learning_rate": 4.417182407441315e-08, + "loss": 0.6901, + "step": 42602 + }, + { + "epoch": 1.56, + "learning_rate": 4.416816200974109e-08, + "loss": 0.6776, + "step": 42603 + }, + { + "epoch": 1.56, + "learning_rate": 4.4164499945069025e-08, + "loss": 0.6984, + "step": 42604 + }, + { + "epoch": 1.56, + "learning_rate": 4.4160837880396965e-08, + "loss": 0.6855, + "step": 42605 + }, + { + "epoch": 1.56, + "learning_rate": 4.4157175815724905e-08, + "loss": 0.7129, + "step": 42606 + }, + { + "epoch": 1.56, + "learning_rate": 4.415351375105284e-08, + "loss": 0.6841, + "step": 42607 + }, + { + "epoch": 1.56, + "learning_rate": 4.414985168638078e-08, + "loss": 0.7068, + "step": 42608 + }, + { + "epoch": 1.56, + "learning_rate": 4.414618962170872e-08, + "loss": 0.6908, + "step": 42609 + }, + { + "epoch": 1.56, + "learning_rate": 4.414252755703666e-08, + "loss": 0.6973, + "step": 42610 + }, + { + "epoch": 1.56, + "learning_rate": 4.413886549236459e-08, + "loss": 0.687, + "step": 42611 + }, + { + "epoch": 1.56, + "learning_rate": 4.413520342769253e-08, + "loss": 0.6953, + "step": 42612 + }, + { + "epoch": 1.56, + "learning_rate": 4.413154136302047e-08, + "loss": 0.6925, + "step": 42613 + }, + { + "epoch": 1.56, + "learning_rate": 4.412787929834841e-08, + "loss": 0.682, + "step": 42614 + }, + { + "epoch": 1.56, + "learning_rate": 4.4124217233676344e-08, + "loss": 0.701, + "step": 42615 + }, + { + "epoch": 1.56, + "learning_rate": 4.4120555169004284e-08, + "loss": 0.6932, + "step": 42616 + }, + { + "epoch": 1.56, + "learning_rate": 4.411689310433222e-08, + "loss": 0.6837, + "step": 42617 + }, + { + "epoch": 1.56, + "learning_rate": 4.411323103966016e-08, + "loss": 0.6901, + "step": 42618 + }, + { + "epoch": 1.56, + "learning_rate": 4.41095689749881e-08, + "loss": 0.6943, + "step": 42619 + }, + { + "epoch": 1.56, + "learning_rate": 4.410590691031603e-08, + "loss": 0.697, + "step": 42620 + }, + { + "epoch": 1.56, + "learning_rate": 4.410224484564397e-08, + "loss": 0.6944, + "step": 42621 + }, + { + "epoch": 1.56, + "learning_rate": 4.409858278097191e-08, + "loss": 0.6959, + "step": 42622 + }, + { + "epoch": 1.56, + "learning_rate": 4.409492071629985e-08, + "loss": 0.6976, + "step": 42623 + }, + { + "epoch": 1.56, + "learning_rate": 4.409125865162779e-08, + "loss": 0.7032, + "step": 42624 + }, + { + "epoch": 1.56, + "learning_rate": 4.4087596586955724e-08, + "loss": 0.6894, + "step": 42625 + }, + { + "epoch": 1.56, + "learning_rate": 4.4083934522283664e-08, + "loss": 0.6885, + "step": 42626 + }, + { + "epoch": 1.56, + "learning_rate": 4.4080272457611603e-08, + "loss": 0.6984, + "step": 42627 + }, + { + "epoch": 1.56, + "learning_rate": 4.407661039293954e-08, + "loss": 0.7041, + "step": 42628 + }, + { + "epoch": 1.56, + "learning_rate": 4.407294832826747e-08, + "loss": 0.6952, + "step": 42629 + }, + { + "epoch": 1.56, + "learning_rate": 4.406928626359541e-08, + "loss": 0.7095, + "step": 42630 + }, + { + "epoch": 1.56, + "learning_rate": 4.406562419892335e-08, + "loss": 0.6898, + "step": 42631 + }, + { + "epoch": 1.56, + "learning_rate": 4.406196213425129e-08, + "loss": 0.6908, + "step": 42632 + }, + { + "epoch": 1.56, + "learning_rate": 4.405830006957923e-08, + "loss": 0.7056, + "step": 42633 + }, + { + "epoch": 1.56, + "learning_rate": 4.405463800490716e-08, + "loss": 0.6924, + "step": 42634 + }, + { + "epoch": 1.56, + "learning_rate": 4.40509759402351e-08, + "loss": 0.6903, + "step": 42635 + }, + { + "epoch": 1.56, + "learning_rate": 4.404731387556304e-08, + "loss": 0.6994, + "step": 42636 + }, + { + "epoch": 1.56, + "learning_rate": 4.404365181089098e-08, + "loss": 0.6954, + "step": 42637 + }, + { + "epoch": 1.56, + "learning_rate": 4.4039989746218916e-08, + "loss": 0.6997, + "step": 42638 + }, + { + "epoch": 1.56, + "learning_rate": 4.4036327681546856e-08, + "loss": 0.6924, + "step": 42639 + }, + { + "epoch": 1.56, + "learning_rate": 4.403266561687479e-08, + "loss": 0.6982, + "step": 42640 + }, + { + "epoch": 1.56, + "learning_rate": 4.402900355220273e-08, + "loss": 0.6958, + "step": 42641 + }, + { + "epoch": 1.56, + "learning_rate": 4.402534148753067e-08, + "loss": 0.6954, + "step": 42642 + }, + { + "epoch": 1.56, + "learning_rate": 4.40216794228586e-08, + "loss": 0.6924, + "step": 42643 + }, + { + "epoch": 1.56, + "learning_rate": 4.401801735818654e-08, + "loss": 0.6948, + "step": 42644 + }, + { + "epoch": 1.56, + "learning_rate": 4.401435529351448e-08, + "loss": 0.7045, + "step": 42645 + }, + { + "epoch": 1.56, + "learning_rate": 4.401069322884242e-08, + "loss": 0.6782, + "step": 42646 + }, + { + "epoch": 1.56, + "learning_rate": 4.4007031164170356e-08, + "loss": 0.6849, + "step": 42647 + }, + { + "epoch": 1.56, + "learning_rate": 4.4003369099498296e-08, + "loss": 0.709, + "step": 42648 + }, + { + "epoch": 1.56, + "learning_rate": 4.3999707034826236e-08, + "loss": 0.6964, + "step": 42649 + }, + { + "epoch": 1.56, + "learning_rate": 4.3996044970154176e-08, + "loss": 0.6943, + "step": 42650 + }, + { + "epoch": 1.56, + "learning_rate": 4.399238290548211e-08, + "loss": 0.6863, + "step": 42651 + }, + { + "epoch": 1.56, + "learning_rate": 4.398872084081004e-08, + "loss": 0.702, + "step": 42652 + }, + { + "epoch": 1.56, + "learning_rate": 4.398505877613798e-08, + "loss": 0.6787, + "step": 42653 + }, + { + "epoch": 1.56, + "learning_rate": 4.398139671146592e-08, + "loss": 0.6933, + "step": 42654 + }, + { + "epoch": 1.56, + "learning_rate": 4.397773464679386e-08, + "loss": 0.7008, + "step": 42655 + }, + { + "epoch": 1.56, + "learning_rate": 4.3974072582121795e-08, + "loss": 0.6883, + "step": 42656 + }, + { + "epoch": 1.56, + "learning_rate": 4.3970410517449735e-08, + "loss": 0.6866, + "step": 42657 + }, + { + "epoch": 1.56, + "learning_rate": 4.3966748452777675e-08, + "loss": 0.6848, + "step": 42658 + }, + { + "epoch": 1.56, + "learning_rate": 4.3963086388105615e-08, + "loss": 0.69, + "step": 42659 + }, + { + "epoch": 1.56, + "learning_rate": 4.395942432343355e-08, + "loss": 0.6877, + "step": 42660 + }, + { + "epoch": 1.56, + "learning_rate": 4.395576225876149e-08, + "loss": 0.6982, + "step": 42661 + }, + { + "epoch": 1.56, + "learning_rate": 4.395210019408943e-08, + "loss": 0.7057, + "step": 42662 + }, + { + "epoch": 1.56, + "learning_rate": 4.394843812941736e-08, + "loss": 0.6878, + "step": 42663 + }, + { + "epoch": 1.56, + "learning_rate": 4.39447760647453e-08, + "loss": 0.6991, + "step": 42664 + }, + { + "epoch": 1.56, + "learning_rate": 4.3941114000073235e-08, + "loss": 0.6847, + "step": 42665 + }, + { + "epoch": 1.56, + "learning_rate": 4.3937451935401175e-08, + "loss": 0.6955, + "step": 42666 + }, + { + "epoch": 1.56, + "learning_rate": 4.3933789870729115e-08, + "loss": 0.6874, + "step": 42667 + }, + { + "epoch": 1.56, + "learning_rate": 4.3930127806057055e-08, + "loss": 0.6996, + "step": 42668 + }, + { + "epoch": 1.56, + "learning_rate": 4.392646574138499e-08, + "loss": 0.705, + "step": 42669 + }, + { + "epoch": 1.56, + "learning_rate": 4.392280367671293e-08, + "loss": 0.6948, + "step": 42670 + }, + { + "epoch": 1.56, + "learning_rate": 4.391914161204087e-08, + "loss": 0.7073, + "step": 42671 + }, + { + "epoch": 1.56, + "learning_rate": 4.391547954736881e-08, + "loss": 0.6942, + "step": 42672 + }, + { + "epoch": 1.56, + "learning_rate": 4.391181748269675e-08, + "loss": 0.6959, + "step": 42673 + }, + { + "epoch": 1.56, + "learning_rate": 4.390815541802468e-08, + "loss": 0.6975, + "step": 42674 + }, + { + "epoch": 1.56, + "learning_rate": 4.3904493353352614e-08, + "loss": 0.6849, + "step": 42675 + }, + { + "epoch": 1.56, + "learning_rate": 4.3900831288680554e-08, + "loss": 0.6841, + "step": 42676 + }, + { + "epoch": 1.56, + "learning_rate": 4.3897169224008494e-08, + "loss": 0.6946, + "step": 42677 + }, + { + "epoch": 1.56, + "learning_rate": 4.389350715933643e-08, + "loss": 0.6882, + "step": 42678 + }, + { + "epoch": 1.56, + "learning_rate": 4.388984509466437e-08, + "loss": 0.6821, + "step": 42679 + }, + { + "epoch": 1.56, + "learning_rate": 4.388618302999231e-08, + "loss": 0.6921, + "step": 42680 + }, + { + "epoch": 1.56, + "learning_rate": 4.388252096532025e-08, + "loss": 0.6884, + "step": 42681 + }, + { + "epoch": 1.56, + "learning_rate": 4.387885890064819e-08, + "loss": 0.6859, + "step": 42682 + }, + { + "epoch": 1.56, + "learning_rate": 4.387519683597612e-08, + "loss": 0.6946, + "step": 42683 + }, + { + "epoch": 1.56, + "learning_rate": 4.387153477130406e-08, + "loss": 0.6927, + "step": 42684 + }, + { + "epoch": 1.56, + "learning_rate": 4.3867872706632e-08, + "loss": 0.687, + "step": 42685 + }, + { + "epoch": 1.56, + "learning_rate": 4.3864210641959934e-08, + "loss": 0.6887, + "step": 42686 + }, + { + "epoch": 1.56, + "learning_rate": 4.3860548577287873e-08, + "loss": 0.6876, + "step": 42687 + }, + { + "epoch": 1.56, + "learning_rate": 4.385688651261581e-08, + "loss": 0.6994, + "step": 42688 + }, + { + "epoch": 1.56, + "learning_rate": 4.385322444794375e-08, + "loss": 0.6888, + "step": 42689 + }, + { + "epoch": 1.56, + "learning_rate": 4.3849562383271687e-08, + "loss": 0.686, + "step": 42690 + }, + { + "epoch": 1.56, + "learning_rate": 4.3845900318599627e-08, + "loss": 0.7023, + "step": 42691 + }, + { + "epoch": 1.56, + "learning_rate": 4.384223825392756e-08, + "loss": 0.679, + "step": 42692 + }, + { + "epoch": 1.56, + "learning_rate": 4.38385761892555e-08, + "loss": 0.6872, + "step": 42693 + }, + { + "epoch": 1.56, + "learning_rate": 4.383491412458344e-08, + "loss": 0.6914, + "step": 42694 + }, + { + "epoch": 1.56, + "learning_rate": 4.383125205991138e-08, + "loss": 0.6927, + "step": 42695 + }, + { + "epoch": 1.56, + "learning_rate": 4.382758999523931e-08, + "loss": 0.6907, + "step": 42696 + }, + { + "epoch": 1.56, + "learning_rate": 4.382392793056725e-08, + "loss": 0.6997, + "step": 42697 + }, + { + "epoch": 1.56, + "learning_rate": 4.382026586589519e-08, + "loss": 0.7074, + "step": 42698 + }, + { + "epoch": 1.56, + "learning_rate": 4.3816603801223126e-08, + "loss": 0.6958, + "step": 42699 + }, + { + "epoch": 1.56, + "learning_rate": 4.3812941736551066e-08, + "loss": 0.6922, + "step": 42700 + }, + { + "epoch": 1.56, + "learning_rate": 4.3809279671879e-08, + "loss": 0.6987, + "step": 42701 + }, + { + "epoch": 1.56, + "learning_rate": 4.380561760720694e-08, + "loss": 0.6852, + "step": 42702 + }, + { + "epoch": 1.56, + "learning_rate": 4.380195554253488e-08, + "loss": 0.6859, + "step": 42703 + }, + { + "epoch": 1.56, + "learning_rate": 4.379829347786282e-08, + "loss": 0.6879, + "step": 42704 + }, + { + "epoch": 1.56, + "learning_rate": 4.379463141319075e-08, + "loss": 0.6952, + "step": 42705 + }, + { + "epoch": 1.56, + "learning_rate": 4.379096934851869e-08, + "loss": 0.6882, + "step": 42706 + }, + { + "epoch": 1.56, + "learning_rate": 4.378730728384663e-08, + "loss": 0.7025, + "step": 42707 + }, + { + "epoch": 1.56, + "learning_rate": 4.378364521917457e-08, + "loss": 0.6944, + "step": 42708 + }, + { + "epoch": 1.56, + "learning_rate": 4.377998315450251e-08, + "loss": 0.6832, + "step": 42709 + }, + { + "epoch": 1.56, + "learning_rate": 4.3776321089830445e-08, + "loss": 0.7169, + "step": 42710 + }, + { + "epoch": 1.56, + "learning_rate": 4.377265902515838e-08, + "loss": 0.6851, + "step": 42711 + }, + { + "epoch": 1.56, + "learning_rate": 4.376899696048632e-08, + "loss": 0.6962, + "step": 42712 + }, + { + "epoch": 1.56, + "learning_rate": 4.376533489581426e-08, + "loss": 0.7034, + "step": 42713 + }, + { + "epoch": 1.56, + "learning_rate": 4.376167283114219e-08, + "loss": 0.7015, + "step": 42714 + }, + { + "epoch": 1.56, + "learning_rate": 4.375801076647013e-08, + "loss": 0.6912, + "step": 42715 + }, + { + "epoch": 1.56, + "learning_rate": 4.375434870179807e-08, + "loss": 0.6892, + "step": 42716 + }, + { + "epoch": 1.56, + "learning_rate": 4.375068663712601e-08, + "loss": 0.6823, + "step": 42717 + }, + { + "epoch": 1.56, + "learning_rate": 4.374702457245395e-08, + "loss": 0.6772, + "step": 42718 + }, + { + "epoch": 1.56, + "learning_rate": 4.3743362507781885e-08, + "loss": 0.7042, + "step": 42719 + }, + { + "epoch": 1.56, + "learning_rate": 4.3739700443109825e-08, + "loss": 0.6927, + "step": 42720 + }, + { + "epoch": 1.56, + "learning_rate": 4.3736038378437765e-08, + "loss": 0.7121, + "step": 42721 + }, + { + "epoch": 1.56, + "learning_rate": 4.37323763137657e-08, + "loss": 0.6832, + "step": 42722 + }, + { + "epoch": 1.56, + "learning_rate": 4.372871424909363e-08, + "loss": 0.7008, + "step": 42723 + }, + { + "epoch": 1.56, + "learning_rate": 4.372505218442157e-08, + "loss": 0.6857, + "step": 42724 + }, + { + "epoch": 1.56, + "learning_rate": 4.372139011974951e-08, + "loss": 0.6814, + "step": 42725 + }, + { + "epoch": 1.56, + "learning_rate": 4.371772805507745e-08, + "loss": 0.7132, + "step": 42726 + }, + { + "epoch": 1.56, + "learning_rate": 4.371406599040539e-08, + "loss": 0.6887, + "step": 42727 + }, + { + "epoch": 1.56, + "learning_rate": 4.3710403925733325e-08, + "loss": 0.7047, + "step": 42728 + }, + { + "epoch": 1.56, + "learning_rate": 4.3706741861061264e-08, + "loss": 0.6902, + "step": 42729 + }, + { + "epoch": 1.56, + "learning_rate": 4.3703079796389204e-08, + "loss": 0.6962, + "step": 42730 + }, + { + "epoch": 1.56, + "learning_rate": 4.3699417731717144e-08, + "loss": 0.7027, + "step": 42731 + }, + { + "epoch": 1.56, + "learning_rate": 4.369575566704508e-08, + "loss": 0.6861, + "step": 42732 + }, + { + "epoch": 1.56, + "learning_rate": 4.369209360237302e-08, + "loss": 0.688, + "step": 42733 + }, + { + "epoch": 1.56, + "learning_rate": 4.368843153770095e-08, + "loss": 0.7049, + "step": 42734 + }, + { + "epoch": 1.56, + "learning_rate": 4.368476947302889e-08, + "loss": 0.6949, + "step": 42735 + }, + { + "epoch": 1.56, + "learning_rate": 4.368110740835683e-08, + "loss": 0.6889, + "step": 42736 + }, + { + "epoch": 1.56, + "learning_rate": 4.3677445343684764e-08, + "loss": 0.7001, + "step": 42737 + }, + { + "epoch": 1.56, + "learning_rate": 4.3673783279012704e-08, + "loss": 0.6916, + "step": 42738 + }, + { + "epoch": 1.56, + "learning_rate": 4.3670121214340644e-08, + "loss": 0.6829, + "step": 42739 + }, + { + "epoch": 1.56, + "learning_rate": 4.3666459149668584e-08, + "loss": 0.6962, + "step": 42740 + }, + { + "epoch": 1.56, + "learning_rate": 4.366279708499652e-08, + "loss": 0.6786, + "step": 42741 + }, + { + "epoch": 1.56, + "learning_rate": 4.365913502032446e-08, + "loss": 0.6906, + "step": 42742 + }, + { + "epoch": 1.56, + "learning_rate": 4.36554729556524e-08, + "loss": 0.6871, + "step": 42743 + }, + { + "epoch": 1.56, + "learning_rate": 4.365181089098034e-08, + "loss": 0.6881, + "step": 42744 + }, + { + "epoch": 1.56, + "learning_rate": 4.364814882630827e-08, + "loss": 0.7093, + "step": 42745 + }, + { + "epoch": 1.56, + "learning_rate": 4.3644486761636204e-08, + "loss": 0.6847, + "step": 42746 + }, + { + "epoch": 1.56, + "learning_rate": 4.3640824696964143e-08, + "loss": 0.6881, + "step": 42747 + }, + { + "epoch": 1.56, + "learning_rate": 4.3637162632292083e-08, + "loss": 0.6806, + "step": 42748 + }, + { + "epoch": 1.56, + "learning_rate": 4.363350056762002e-08, + "loss": 0.6916, + "step": 42749 + }, + { + "epoch": 1.56, + "learning_rate": 4.3629838502947957e-08, + "loss": 0.6909, + "step": 42750 + }, + { + "epoch": 1.56, + "learning_rate": 4.3626176438275897e-08, + "loss": 0.6951, + "step": 42751 + }, + { + "epoch": 1.56, + "learning_rate": 4.3622514373603836e-08, + "loss": 0.6914, + "step": 42752 + }, + { + "epoch": 1.56, + "learning_rate": 4.3618852308931776e-08, + "loss": 0.7049, + "step": 42753 + }, + { + "epoch": 1.56, + "learning_rate": 4.3615190244259716e-08, + "loss": 0.6779, + "step": 42754 + }, + { + "epoch": 1.56, + "learning_rate": 4.361152817958765e-08, + "loss": 0.6924, + "step": 42755 + }, + { + "epoch": 1.56, + "learning_rate": 4.360786611491559e-08, + "loss": 0.6961, + "step": 42756 + }, + { + "epoch": 1.56, + "learning_rate": 4.360420405024352e-08, + "loss": 0.6979, + "step": 42757 + }, + { + "epoch": 1.56, + "learning_rate": 4.360054198557146e-08, + "loss": 0.6965, + "step": 42758 + }, + { + "epoch": 1.56, + "learning_rate": 4.3596879920899396e-08, + "loss": 0.6976, + "step": 42759 + }, + { + "epoch": 1.56, + "learning_rate": 4.3593217856227336e-08, + "loss": 0.6915, + "step": 42760 + }, + { + "epoch": 1.56, + "learning_rate": 4.3589555791555276e-08, + "loss": 0.6828, + "step": 42761 + }, + { + "epoch": 1.56, + "learning_rate": 4.3585893726883216e-08, + "loss": 0.7047, + "step": 42762 + }, + { + "epoch": 1.56, + "learning_rate": 4.3582231662211156e-08, + "loss": 0.7041, + "step": 42763 + }, + { + "epoch": 1.56, + "learning_rate": 4.357856959753909e-08, + "loss": 0.685, + "step": 42764 + }, + { + "epoch": 1.56, + "learning_rate": 4.357490753286703e-08, + "loss": 0.6901, + "step": 42765 + }, + { + "epoch": 1.56, + "learning_rate": 4.357124546819497e-08, + "loss": 0.6915, + "step": 42766 + }, + { + "epoch": 1.56, + "learning_rate": 4.356758340352291e-08, + "loss": 0.6974, + "step": 42767 + }, + { + "epoch": 1.56, + "learning_rate": 4.356392133885084e-08, + "loss": 0.6979, + "step": 42768 + }, + { + "epoch": 1.56, + "learning_rate": 4.356025927417878e-08, + "loss": 0.6936, + "step": 42769 + }, + { + "epoch": 1.56, + "learning_rate": 4.3556597209506715e-08, + "loss": 0.6989, + "step": 42770 + }, + { + "epoch": 1.56, + "learning_rate": 4.3552935144834655e-08, + "loss": 0.6808, + "step": 42771 + }, + { + "epoch": 1.56, + "learning_rate": 4.3549273080162595e-08, + "loss": 0.7038, + "step": 42772 + }, + { + "epoch": 1.56, + "learning_rate": 4.354561101549053e-08, + "loss": 0.6686, + "step": 42773 + }, + { + "epoch": 1.56, + "learning_rate": 4.354194895081847e-08, + "loss": 0.6921, + "step": 42774 + }, + { + "epoch": 1.57, + "learning_rate": 4.353828688614641e-08, + "loss": 0.6831, + "step": 42775 + }, + { + "epoch": 1.57, + "learning_rate": 4.353462482147435e-08, + "loss": 0.703, + "step": 42776 + }, + { + "epoch": 1.57, + "learning_rate": 4.353096275680228e-08, + "loss": 0.6902, + "step": 42777 + }, + { + "epoch": 1.57, + "learning_rate": 4.352730069213022e-08, + "loss": 0.7013, + "step": 42778 + }, + { + "epoch": 1.57, + "learning_rate": 4.352363862745816e-08, + "loss": 0.6864, + "step": 42779 + }, + { + "epoch": 1.57, + "learning_rate": 4.3519976562786095e-08, + "loss": 0.6873, + "step": 42780 + }, + { + "epoch": 1.57, + "learning_rate": 4.3516314498114035e-08, + "loss": 0.6891, + "step": 42781 + }, + { + "epoch": 1.57, + "learning_rate": 4.351265243344197e-08, + "loss": 0.6891, + "step": 42782 + }, + { + "epoch": 1.57, + "learning_rate": 4.350899036876991e-08, + "loss": 0.6859, + "step": 42783 + }, + { + "epoch": 1.57, + "learning_rate": 4.350532830409785e-08, + "loss": 0.6891, + "step": 42784 + }, + { + "epoch": 1.57, + "learning_rate": 4.350166623942579e-08, + "loss": 0.6964, + "step": 42785 + }, + { + "epoch": 1.57, + "learning_rate": 4.349800417475372e-08, + "loss": 0.6932, + "step": 42786 + }, + { + "epoch": 1.57, + "learning_rate": 4.349434211008166e-08, + "loss": 0.7078, + "step": 42787 + }, + { + "epoch": 1.57, + "learning_rate": 4.34906800454096e-08, + "loss": 0.6924, + "step": 42788 + }, + { + "epoch": 1.57, + "learning_rate": 4.348701798073754e-08, + "loss": 0.6903, + "step": 42789 + }, + { + "epoch": 1.57, + "learning_rate": 4.348335591606548e-08, + "loss": 0.694, + "step": 42790 + }, + { + "epoch": 1.57, + "learning_rate": 4.3479693851393414e-08, + "loss": 0.6938, + "step": 42791 + }, + { + "epoch": 1.57, + "learning_rate": 4.3476031786721354e-08, + "loss": 0.6933, + "step": 42792 + }, + { + "epoch": 1.57, + "learning_rate": 4.347236972204929e-08, + "loss": 0.6742, + "step": 42793 + }, + { + "epoch": 1.57, + "learning_rate": 4.346870765737723e-08, + "loss": 0.6928, + "step": 42794 + }, + { + "epoch": 1.57, + "learning_rate": 4.346504559270516e-08, + "loss": 0.6976, + "step": 42795 + }, + { + "epoch": 1.57, + "learning_rate": 4.34613835280331e-08, + "loss": 0.6847, + "step": 42796 + }, + { + "epoch": 1.57, + "learning_rate": 4.345772146336104e-08, + "loss": 0.6905, + "step": 42797 + }, + { + "epoch": 1.57, + "learning_rate": 4.345405939868898e-08, + "loss": 0.6886, + "step": 42798 + }, + { + "epoch": 1.57, + "learning_rate": 4.345039733401692e-08, + "loss": 0.6857, + "step": 42799 + }, + { + "epoch": 1.57, + "learning_rate": 4.3446735269344854e-08, + "loss": 0.6906, + "step": 42800 + }, + { + "epoch": 1.57, + "learning_rate": 4.3443073204672794e-08, + "loss": 0.7023, + "step": 42801 + }, + { + "epoch": 1.57, + "learning_rate": 4.3439411140000734e-08, + "loss": 0.6834, + "step": 42802 + }, + { + "epoch": 1.57, + "learning_rate": 4.3435749075328674e-08, + "loss": 0.707, + "step": 42803 + }, + { + "epoch": 1.57, + "learning_rate": 4.343208701065661e-08, + "loss": 0.7131, + "step": 42804 + }, + { + "epoch": 1.57, + "learning_rate": 4.342842494598454e-08, + "loss": 0.6874, + "step": 42805 + }, + { + "epoch": 1.57, + "learning_rate": 4.342476288131248e-08, + "loss": 0.697, + "step": 42806 + }, + { + "epoch": 1.57, + "learning_rate": 4.342110081664042e-08, + "loss": 0.6937, + "step": 42807 + }, + { + "epoch": 1.57, + "learning_rate": 4.341743875196836e-08, + "loss": 0.6957, + "step": 42808 + }, + { + "epoch": 1.57, + "learning_rate": 4.341377668729629e-08, + "loss": 0.6847, + "step": 42809 + }, + { + "epoch": 1.57, + "learning_rate": 4.341011462262423e-08, + "loss": 0.6913, + "step": 42810 + }, + { + "epoch": 1.57, + "learning_rate": 4.340645255795217e-08, + "loss": 0.6916, + "step": 42811 + }, + { + "epoch": 1.57, + "learning_rate": 4.340279049328011e-08, + "loss": 0.6916, + "step": 42812 + }, + { + "epoch": 1.57, + "learning_rate": 4.3399128428608046e-08, + "loss": 0.6914, + "step": 42813 + }, + { + "epoch": 1.57, + "learning_rate": 4.3395466363935986e-08, + "loss": 0.6971, + "step": 42814 + }, + { + "epoch": 1.57, + "learning_rate": 4.3391804299263926e-08, + "loss": 0.6869, + "step": 42815 + }, + { + "epoch": 1.57, + "learning_rate": 4.338814223459186e-08, + "loss": 0.6819, + "step": 42816 + }, + { + "epoch": 1.57, + "learning_rate": 4.33844801699198e-08, + "loss": 0.707, + "step": 42817 + }, + { + "epoch": 1.57, + "learning_rate": 4.338081810524773e-08, + "loss": 0.7059, + "step": 42818 + }, + { + "epoch": 1.57, + "learning_rate": 4.337715604057567e-08, + "loss": 0.6914, + "step": 42819 + }, + { + "epoch": 1.57, + "learning_rate": 4.337349397590361e-08, + "loss": 0.7002, + "step": 42820 + }, + { + "epoch": 1.57, + "learning_rate": 4.336983191123155e-08, + "loss": 0.6882, + "step": 42821 + }, + { + "epoch": 1.57, + "learning_rate": 4.3366169846559486e-08, + "loss": 0.6963, + "step": 42822 + }, + { + "epoch": 1.57, + "learning_rate": 4.3362507781887426e-08, + "loss": 0.7079, + "step": 42823 + }, + { + "epoch": 1.57, + "learning_rate": 4.3358845717215366e-08, + "loss": 0.6916, + "step": 42824 + }, + { + "epoch": 1.57, + "learning_rate": 4.3355183652543306e-08, + "loss": 0.6831, + "step": 42825 + }, + { + "epoch": 1.57, + "learning_rate": 4.3351521587871246e-08, + "loss": 0.6952, + "step": 42826 + }, + { + "epoch": 1.57, + "learning_rate": 4.334785952319918e-08, + "loss": 0.6839, + "step": 42827 + }, + { + "epoch": 1.57, + "learning_rate": 4.334419745852711e-08, + "loss": 0.6968, + "step": 42828 + }, + { + "epoch": 1.57, + "learning_rate": 4.334053539385505e-08, + "loss": 0.6883, + "step": 42829 + }, + { + "epoch": 1.57, + "learning_rate": 4.333687332918299e-08, + "loss": 0.6875, + "step": 42830 + }, + { + "epoch": 1.57, + "learning_rate": 4.3333211264510925e-08, + "loss": 0.7024, + "step": 42831 + }, + { + "epoch": 1.57, + "learning_rate": 4.3329549199838865e-08, + "loss": 0.7025, + "step": 42832 + }, + { + "epoch": 1.57, + "learning_rate": 4.3325887135166805e-08, + "loss": 0.6961, + "step": 42833 + }, + { + "epoch": 1.57, + "learning_rate": 4.3322225070494745e-08, + "loss": 0.6968, + "step": 42834 + }, + { + "epoch": 1.57, + "learning_rate": 4.3318563005822685e-08, + "loss": 0.6901, + "step": 42835 + }, + { + "epoch": 1.57, + "learning_rate": 4.331490094115062e-08, + "loss": 0.6956, + "step": 42836 + }, + { + "epoch": 1.57, + "learning_rate": 4.331123887647856e-08, + "loss": 0.6879, + "step": 42837 + }, + { + "epoch": 1.57, + "learning_rate": 4.33075768118065e-08, + "loss": 0.6816, + "step": 42838 + }, + { + "epoch": 1.57, + "learning_rate": 4.330391474713443e-08, + "loss": 0.7033, + "step": 42839 + }, + { + "epoch": 1.57, + "learning_rate": 4.3300252682462365e-08, + "loss": 0.6922, + "step": 42840 + }, + { + "epoch": 1.57, + "learning_rate": 4.3296590617790305e-08, + "loss": 0.6974, + "step": 42841 + }, + { + "epoch": 1.57, + "learning_rate": 4.3292928553118245e-08, + "loss": 0.6994, + "step": 42842 + }, + { + "epoch": 1.57, + "learning_rate": 4.3289266488446185e-08, + "loss": 0.6976, + "step": 42843 + }, + { + "epoch": 1.57, + "learning_rate": 4.328560442377412e-08, + "loss": 0.7033, + "step": 42844 + }, + { + "epoch": 1.57, + "learning_rate": 4.328194235910206e-08, + "loss": 0.6978, + "step": 42845 + }, + { + "epoch": 1.57, + "learning_rate": 4.327828029443e-08, + "loss": 0.695, + "step": 42846 + }, + { + "epoch": 1.57, + "learning_rate": 4.327461822975794e-08, + "loss": 0.683, + "step": 42847 + }, + { + "epoch": 1.57, + "learning_rate": 4.327095616508588e-08, + "loss": 0.6856, + "step": 42848 + }, + { + "epoch": 1.57, + "learning_rate": 4.326729410041381e-08, + "loss": 0.6987, + "step": 42849 + }, + { + "epoch": 1.57, + "learning_rate": 4.326363203574175e-08, + "loss": 0.6891, + "step": 42850 + }, + { + "epoch": 1.57, + "learning_rate": 4.3259969971069684e-08, + "loss": 0.7009, + "step": 42851 + }, + { + "epoch": 1.57, + "learning_rate": 4.3256307906397624e-08, + "loss": 0.6988, + "step": 42852 + }, + { + "epoch": 1.57, + "learning_rate": 4.325264584172556e-08, + "loss": 0.6937, + "step": 42853 + }, + { + "epoch": 1.57, + "learning_rate": 4.32489837770535e-08, + "loss": 0.6878, + "step": 42854 + }, + { + "epoch": 1.57, + "learning_rate": 4.324532171238144e-08, + "loss": 0.6919, + "step": 42855 + }, + { + "epoch": 1.57, + "learning_rate": 4.324165964770938e-08, + "loss": 0.6989, + "step": 42856 + }, + { + "epoch": 1.57, + "learning_rate": 4.323799758303732e-08, + "loss": 0.6937, + "step": 42857 + }, + { + "epoch": 1.57, + "learning_rate": 4.323433551836525e-08, + "loss": 0.6889, + "step": 42858 + }, + { + "epoch": 1.57, + "learning_rate": 4.323067345369319e-08, + "loss": 0.6925, + "step": 42859 + }, + { + "epoch": 1.57, + "learning_rate": 4.322701138902113e-08, + "loss": 0.6881, + "step": 42860 + }, + { + "epoch": 1.57, + "learning_rate": 4.322334932434907e-08, + "loss": 0.692, + "step": 42861 + }, + { + "epoch": 1.57, + "learning_rate": 4.3219687259677004e-08, + "loss": 0.6764, + "step": 42862 + }, + { + "epoch": 1.57, + "learning_rate": 4.3216025195004944e-08, + "loss": 0.7016, + "step": 42863 + }, + { + "epoch": 1.57, + "learning_rate": 4.321236313033288e-08, + "loss": 0.6934, + "step": 42864 + }, + { + "epoch": 1.57, + "learning_rate": 4.320870106566082e-08, + "loss": 0.6926, + "step": 42865 + }, + { + "epoch": 1.57, + "learning_rate": 4.320503900098876e-08, + "loss": 0.6984, + "step": 42866 + }, + { + "epoch": 1.57, + "learning_rate": 4.320137693631669e-08, + "loss": 0.7034, + "step": 42867 + }, + { + "epoch": 1.57, + "learning_rate": 4.319771487164463e-08, + "loss": 0.6925, + "step": 42868 + }, + { + "epoch": 1.57, + "learning_rate": 4.319405280697257e-08, + "loss": 0.6961, + "step": 42869 + }, + { + "epoch": 1.57, + "learning_rate": 4.319039074230051e-08, + "loss": 0.703, + "step": 42870 + }, + { + "epoch": 1.57, + "learning_rate": 4.318672867762844e-08, + "loss": 0.6933, + "step": 42871 + }, + { + "epoch": 1.57, + "learning_rate": 4.318306661295638e-08, + "loss": 0.6925, + "step": 42872 + }, + { + "epoch": 1.57, + "learning_rate": 4.317940454828432e-08, + "loss": 0.6927, + "step": 42873 + }, + { + "epoch": 1.57, + "learning_rate": 4.317574248361226e-08, + "loss": 0.6833, + "step": 42874 + }, + { + "epoch": 1.57, + "learning_rate": 4.3172080418940196e-08, + "loss": 0.7116, + "step": 42875 + }, + { + "epoch": 1.57, + "learning_rate": 4.316841835426813e-08, + "loss": 0.6973, + "step": 42876 + }, + { + "epoch": 1.57, + "learning_rate": 4.316475628959607e-08, + "loss": 0.6969, + "step": 42877 + }, + { + "epoch": 1.57, + "learning_rate": 4.316109422492401e-08, + "loss": 0.6817, + "step": 42878 + }, + { + "epoch": 1.57, + "learning_rate": 4.315743216025195e-08, + "loss": 0.6899, + "step": 42879 + }, + { + "epoch": 1.57, + "learning_rate": 4.315377009557988e-08, + "loss": 0.6933, + "step": 42880 + }, + { + "epoch": 1.57, + "learning_rate": 4.315010803090782e-08, + "loss": 0.6869, + "step": 42881 + }, + { + "epoch": 1.57, + "learning_rate": 4.314644596623576e-08, + "loss": 0.7072, + "step": 42882 + }, + { + "epoch": 1.57, + "learning_rate": 4.31427839015637e-08, + "loss": 0.6958, + "step": 42883 + }, + { + "epoch": 1.57, + "learning_rate": 4.313912183689164e-08, + "loss": 0.688, + "step": 42884 + }, + { + "epoch": 1.57, + "learning_rate": 4.3135459772219576e-08, + "loss": 0.7075, + "step": 42885 + }, + { + "epoch": 1.57, + "learning_rate": 4.3131797707547516e-08, + "loss": 0.6971, + "step": 42886 + }, + { + "epoch": 1.57, + "learning_rate": 4.312813564287545e-08, + "loss": 0.6858, + "step": 42887 + }, + { + "epoch": 1.57, + "learning_rate": 4.312447357820339e-08, + "loss": 0.6886, + "step": 42888 + }, + { + "epoch": 1.57, + "learning_rate": 4.312081151353132e-08, + "loss": 0.6808, + "step": 42889 + }, + { + "epoch": 1.57, + "learning_rate": 4.311714944885926e-08, + "loss": 0.6976, + "step": 42890 + }, + { + "epoch": 1.57, + "learning_rate": 4.31134873841872e-08, + "loss": 0.6924, + "step": 42891 + }, + { + "epoch": 1.57, + "learning_rate": 4.310982531951514e-08, + "loss": 0.6964, + "step": 42892 + }, + { + "epoch": 1.57, + "learning_rate": 4.310616325484308e-08, + "loss": 0.6964, + "step": 42893 + }, + { + "epoch": 1.57, + "learning_rate": 4.3102501190171015e-08, + "loss": 0.7036, + "step": 42894 + }, + { + "epoch": 1.57, + "learning_rate": 4.3098839125498955e-08, + "loss": 0.6931, + "step": 42895 + }, + { + "epoch": 1.57, + "learning_rate": 4.3095177060826895e-08, + "loss": 0.7026, + "step": 42896 + }, + { + "epoch": 1.57, + "learning_rate": 4.3091514996154835e-08, + "loss": 0.6994, + "step": 42897 + }, + { + "epoch": 1.57, + "learning_rate": 4.308785293148277e-08, + "loss": 0.7022, + "step": 42898 + }, + { + "epoch": 1.57, + "learning_rate": 4.30841908668107e-08, + "loss": 0.6861, + "step": 42899 + }, + { + "epoch": 1.57, + "learning_rate": 4.308052880213864e-08, + "loss": 0.6951, + "step": 42900 + }, + { + "epoch": 1.57, + "learning_rate": 4.307686673746658e-08, + "loss": 0.6993, + "step": 42901 + }, + { + "epoch": 1.57, + "learning_rate": 4.307320467279452e-08, + "loss": 0.688, + "step": 42902 + }, + { + "epoch": 1.57, + "learning_rate": 4.3069542608122455e-08, + "loss": 0.7014, + "step": 42903 + }, + { + "epoch": 1.57, + "learning_rate": 4.3065880543450395e-08, + "loss": 0.6969, + "step": 42904 + }, + { + "epoch": 1.57, + "learning_rate": 4.3062218478778335e-08, + "loss": 0.6989, + "step": 42905 + }, + { + "epoch": 1.57, + "learning_rate": 4.3058556414106274e-08, + "loss": 0.6859, + "step": 42906 + }, + { + "epoch": 1.57, + "learning_rate": 4.305489434943421e-08, + "loss": 0.6992, + "step": 42907 + }, + { + "epoch": 1.57, + "learning_rate": 4.305123228476215e-08, + "loss": 0.6921, + "step": 42908 + }, + { + "epoch": 1.57, + "learning_rate": 4.304757022009009e-08, + "loss": 0.6992, + "step": 42909 + }, + { + "epoch": 1.57, + "learning_rate": 4.304390815541802e-08, + "loss": 0.6955, + "step": 42910 + }, + { + "epoch": 1.57, + "learning_rate": 4.304024609074596e-08, + "loss": 0.6922, + "step": 42911 + }, + { + "epoch": 1.57, + "learning_rate": 4.3036584026073894e-08, + "loss": 0.6926, + "step": 42912 + }, + { + "epoch": 1.57, + "learning_rate": 4.3032921961401834e-08, + "loss": 0.6905, + "step": 42913 + }, + { + "epoch": 1.57, + "learning_rate": 4.3029259896729774e-08, + "loss": 0.6952, + "step": 42914 + }, + { + "epoch": 1.57, + "learning_rate": 4.3025597832057714e-08, + "loss": 0.6831, + "step": 42915 + }, + { + "epoch": 1.57, + "learning_rate": 4.302193576738565e-08, + "loss": 0.7012, + "step": 42916 + }, + { + "epoch": 1.57, + "learning_rate": 4.301827370271359e-08, + "loss": 0.6916, + "step": 42917 + }, + { + "epoch": 1.57, + "learning_rate": 4.301461163804153e-08, + "loss": 0.6956, + "step": 42918 + }, + { + "epoch": 1.57, + "learning_rate": 4.301094957336947e-08, + "loss": 0.6936, + "step": 42919 + }, + { + "epoch": 1.57, + "learning_rate": 4.300728750869741e-08, + "loss": 0.6977, + "step": 42920 + }, + { + "epoch": 1.57, + "learning_rate": 4.300362544402534e-08, + "loss": 0.6928, + "step": 42921 + }, + { + "epoch": 1.57, + "learning_rate": 4.2999963379353274e-08, + "loss": 0.6999, + "step": 42922 + }, + { + "epoch": 1.57, + "learning_rate": 4.2996301314681214e-08, + "loss": 0.6875, + "step": 42923 + }, + { + "epoch": 1.57, + "learning_rate": 4.2992639250009153e-08, + "loss": 0.6947, + "step": 42924 + }, + { + "epoch": 1.57, + "learning_rate": 4.298897718533709e-08, + "loss": 0.6926, + "step": 42925 + }, + { + "epoch": 1.57, + "learning_rate": 4.298531512066503e-08, + "loss": 0.7015, + "step": 42926 + }, + { + "epoch": 1.57, + "learning_rate": 4.2981653055992967e-08, + "loss": 0.6817, + "step": 42927 + }, + { + "epoch": 1.57, + "learning_rate": 4.2977990991320907e-08, + "loss": 0.7019, + "step": 42928 + }, + { + "epoch": 1.57, + "learning_rate": 4.2974328926648846e-08, + "loss": 0.6942, + "step": 42929 + }, + { + "epoch": 1.57, + "learning_rate": 4.297066686197678e-08, + "loss": 0.7146, + "step": 42930 + }, + { + "epoch": 1.57, + "learning_rate": 4.296700479730472e-08, + "loss": 0.6986, + "step": 42931 + }, + { + "epoch": 1.57, + "learning_rate": 4.296334273263266e-08, + "loss": 0.6838, + "step": 42932 + }, + { + "epoch": 1.57, + "learning_rate": 4.295968066796059e-08, + "loss": 0.6954, + "step": 42933 + }, + { + "epoch": 1.57, + "learning_rate": 4.295601860328853e-08, + "loss": 0.6943, + "step": 42934 + }, + { + "epoch": 1.57, + "learning_rate": 4.2952356538616466e-08, + "loss": 0.7042, + "step": 42935 + }, + { + "epoch": 1.57, + "learning_rate": 4.2948694473944406e-08, + "loss": 0.682, + "step": 42936 + }, + { + "epoch": 1.57, + "learning_rate": 4.2945032409272346e-08, + "loss": 0.7026, + "step": 42937 + }, + { + "epoch": 1.57, + "learning_rate": 4.2941370344600286e-08, + "loss": 0.7027, + "step": 42938 + }, + { + "epoch": 1.57, + "learning_rate": 4.293770827992822e-08, + "loss": 0.7025, + "step": 42939 + }, + { + "epoch": 1.57, + "learning_rate": 4.293404621525616e-08, + "loss": 0.6838, + "step": 42940 + }, + { + "epoch": 1.57, + "learning_rate": 4.29303841505841e-08, + "loss": 0.697, + "step": 42941 + }, + { + "epoch": 1.57, + "learning_rate": 4.292672208591204e-08, + "loss": 0.7011, + "step": 42942 + }, + { + "epoch": 1.57, + "learning_rate": 4.292306002123997e-08, + "loss": 0.6865, + "step": 42943 + }, + { + "epoch": 1.57, + "learning_rate": 4.291939795656791e-08, + "loss": 0.6903, + "step": 42944 + }, + { + "epoch": 1.57, + "learning_rate": 4.291573589189585e-08, + "loss": 0.6982, + "step": 42945 + }, + { + "epoch": 1.57, + "learning_rate": 4.2912073827223786e-08, + "loss": 0.6916, + "step": 42946 + }, + { + "epoch": 1.57, + "learning_rate": 4.2908411762551726e-08, + "loss": 0.6996, + "step": 42947 + }, + { + "epoch": 1.57, + "learning_rate": 4.290474969787966e-08, + "loss": 0.6969, + "step": 42948 + }, + { + "epoch": 1.57, + "learning_rate": 4.29010876332076e-08, + "loss": 0.6979, + "step": 42949 + }, + { + "epoch": 1.57, + "learning_rate": 4.289742556853554e-08, + "loss": 0.6828, + "step": 42950 + }, + { + "epoch": 1.57, + "learning_rate": 4.289376350386348e-08, + "loss": 0.7058, + "step": 42951 + }, + { + "epoch": 1.57, + "learning_rate": 4.289010143919141e-08, + "loss": 0.7103, + "step": 42952 + }, + { + "epoch": 1.57, + "learning_rate": 4.288643937451935e-08, + "loss": 0.6913, + "step": 42953 + }, + { + "epoch": 1.57, + "learning_rate": 4.288277730984729e-08, + "loss": 0.6992, + "step": 42954 + }, + { + "epoch": 1.57, + "learning_rate": 4.287911524517523e-08, + "loss": 0.7061, + "step": 42955 + }, + { + "epoch": 1.57, + "learning_rate": 4.2875453180503165e-08, + "loss": 0.6929, + "step": 42956 + }, + { + "epoch": 1.57, + "learning_rate": 4.2871791115831105e-08, + "loss": 0.6854, + "step": 42957 + }, + { + "epoch": 1.57, + "learning_rate": 4.286812905115904e-08, + "loss": 0.6992, + "step": 42958 + }, + { + "epoch": 1.57, + "learning_rate": 4.286446698648698e-08, + "loss": 0.6881, + "step": 42959 + }, + { + "epoch": 1.57, + "learning_rate": 4.286080492181492e-08, + "loss": 0.696, + "step": 42960 + }, + { + "epoch": 1.57, + "learning_rate": 4.285714285714285e-08, + "loss": 0.7046, + "step": 42961 + }, + { + "epoch": 1.57, + "learning_rate": 4.285348079247079e-08, + "loss": 0.6932, + "step": 42962 + }, + { + "epoch": 1.57, + "learning_rate": 4.284981872779873e-08, + "loss": 0.6838, + "step": 42963 + }, + { + "epoch": 1.57, + "learning_rate": 4.284615666312667e-08, + "loss": 0.6993, + "step": 42964 + }, + { + "epoch": 1.57, + "learning_rate": 4.284249459845461e-08, + "loss": 0.6765, + "step": 42965 + }, + { + "epoch": 1.57, + "learning_rate": 4.2838832533782544e-08, + "loss": 0.698, + "step": 42966 + }, + { + "epoch": 1.57, + "learning_rate": 4.2835170469110484e-08, + "loss": 0.6839, + "step": 42967 + }, + { + "epoch": 1.57, + "learning_rate": 4.2831508404438424e-08, + "loss": 0.6729, + "step": 42968 + }, + { + "epoch": 1.57, + "learning_rate": 4.282784633976636e-08, + "loss": 0.6977, + "step": 42969 + }, + { + "epoch": 1.57, + "learning_rate": 4.282418427509429e-08, + "loss": 0.7027, + "step": 42970 + }, + { + "epoch": 1.57, + "learning_rate": 4.282052221042223e-08, + "loss": 0.6868, + "step": 42971 + }, + { + "epoch": 1.57, + "learning_rate": 4.281686014575017e-08, + "loss": 0.6953, + "step": 42972 + }, + { + "epoch": 1.57, + "learning_rate": 4.281319808107811e-08, + "loss": 0.6949, + "step": 42973 + }, + { + "epoch": 1.57, + "learning_rate": 4.280953601640605e-08, + "loss": 0.6974, + "step": 42974 + }, + { + "epoch": 1.57, + "learning_rate": 4.2805873951733984e-08, + "loss": 0.6976, + "step": 42975 + }, + { + "epoch": 1.57, + "learning_rate": 4.2802211887061924e-08, + "loss": 0.6826, + "step": 42976 + }, + { + "epoch": 1.57, + "learning_rate": 4.2798549822389864e-08, + "loss": 0.6894, + "step": 42977 + }, + { + "epoch": 1.57, + "learning_rate": 4.2794887757717804e-08, + "loss": 0.6987, + "step": 42978 + }, + { + "epoch": 1.57, + "learning_rate": 4.279122569304574e-08, + "loss": 0.6799, + "step": 42979 + }, + { + "epoch": 1.57, + "learning_rate": 4.278756362837368e-08, + "loss": 0.6921, + "step": 42980 + }, + { + "epoch": 1.57, + "learning_rate": 4.278390156370161e-08, + "loss": 0.6955, + "step": 42981 + }, + { + "epoch": 1.57, + "learning_rate": 4.278023949902955e-08, + "loss": 0.6903, + "step": 42982 + }, + { + "epoch": 1.57, + "learning_rate": 4.277657743435749e-08, + "loss": 0.7004, + "step": 42983 + }, + { + "epoch": 1.57, + "learning_rate": 4.2772915369685423e-08, + "loss": 0.6919, + "step": 42984 + }, + { + "epoch": 1.57, + "learning_rate": 4.2769253305013363e-08, + "loss": 0.7057, + "step": 42985 + }, + { + "epoch": 1.57, + "learning_rate": 4.2765591240341303e-08, + "loss": 0.685, + "step": 42986 + }, + { + "epoch": 1.57, + "learning_rate": 4.276192917566924e-08, + "loss": 0.6923, + "step": 42987 + }, + { + "epoch": 1.57, + "learning_rate": 4.2758267110997177e-08, + "loss": 0.6972, + "step": 42988 + }, + { + "epoch": 1.57, + "learning_rate": 4.2754605046325116e-08, + "loss": 0.6899, + "step": 42989 + }, + { + "epoch": 1.57, + "learning_rate": 4.2750942981653056e-08, + "loss": 0.6924, + "step": 42990 + }, + { + "epoch": 1.57, + "learning_rate": 4.2747280916980996e-08, + "loss": 0.7008, + "step": 42991 + }, + { + "epoch": 1.57, + "learning_rate": 4.274361885230893e-08, + "loss": 0.6914, + "step": 42992 + }, + { + "epoch": 1.57, + "learning_rate": 4.273995678763686e-08, + "loss": 0.6965, + "step": 42993 + }, + { + "epoch": 1.57, + "learning_rate": 4.27362947229648e-08, + "loss": 0.6889, + "step": 42994 + }, + { + "epoch": 1.57, + "learning_rate": 4.273263265829274e-08, + "loss": 0.6941, + "step": 42995 + }, + { + "epoch": 1.57, + "learning_rate": 4.272897059362068e-08, + "loss": 0.6861, + "step": 42996 + }, + { + "epoch": 1.57, + "learning_rate": 4.2725308528948616e-08, + "loss": 0.6876, + "step": 42997 + }, + { + "epoch": 1.57, + "learning_rate": 4.2721646464276556e-08, + "loss": 0.7002, + "step": 42998 + }, + { + "epoch": 1.57, + "learning_rate": 4.2717984399604496e-08, + "loss": 0.6821, + "step": 42999 + }, + { + "epoch": 1.57, + "learning_rate": 4.2714322334932436e-08, + "loss": 0.6899, + "step": 43000 + }, + { + "epoch": 1.57, + "learning_rate": 4.2710660270260376e-08, + "loss": 0.7042, + "step": 43001 + }, + { + "epoch": 1.57, + "learning_rate": 4.270699820558831e-08, + "loss": 0.6951, + "step": 43002 + }, + { + "epoch": 1.57, + "learning_rate": 4.270333614091625e-08, + "loss": 0.6979, + "step": 43003 + }, + { + "epoch": 1.57, + "learning_rate": 4.269967407624418e-08, + "loss": 0.6915, + "step": 43004 + }, + { + "epoch": 1.57, + "learning_rate": 4.269601201157212e-08, + "loss": 0.6903, + "step": 43005 + }, + { + "epoch": 1.57, + "learning_rate": 4.2692349946900056e-08, + "loss": 0.7094, + "step": 43006 + }, + { + "epoch": 1.57, + "learning_rate": 4.2688687882227995e-08, + "loss": 0.699, + "step": 43007 + }, + { + "epoch": 1.57, + "learning_rate": 4.2685025817555935e-08, + "loss": 0.687, + "step": 43008 + }, + { + "epoch": 1.57, + "learning_rate": 4.2681363752883875e-08, + "loss": 0.6916, + "step": 43009 + }, + { + "epoch": 1.57, + "learning_rate": 4.2677701688211815e-08, + "loss": 0.6934, + "step": 43010 + }, + { + "epoch": 1.57, + "learning_rate": 4.267403962353975e-08, + "loss": 0.6956, + "step": 43011 + }, + { + "epoch": 1.57, + "learning_rate": 4.267037755886769e-08, + "loss": 0.6957, + "step": 43012 + }, + { + "epoch": 1.57, + "learning_rate": 4.266671549419563e-08, + "loss": 0.6948, + "step": 43013 + }, + { + "epoch": 1.57, + "learning_rate": 4.266305342952357e-08, + "loss": 0.6829, + "step": 43014 + }, + { + "epoch": 1.57, + "learning_rate": 4.26593913648515e-08, + "loss": 0.6982, + "step": 43015 + }, + { + "epoch": 1.57, + "learning_rate": 4.2655729300179435e-08, + "loss": 0.697, + "step": 43016 + }, + { + "epoch": 1.57, + "learning_rate": 4.2652067235507375e-08, + "loss": 0.6878, + "step": 43017 + }, + { + "epoch": 1.57, + "learning_rate": 4.2648405170835315e-08, + "loss": 0.6883, + "step": 43018 + }, + { + "epoch": 1.57, + "learning_rate": 4.2644743106163255e-08, + "loss": 0.6901, + "step": 43019 + }, + { + "epoch": 1.57, + "learning_rate": 4.264108104149119e-08, + "loss": 0.6942, + "step": 43020 + }, + { + "epoch": 1.57, + "learning_rate": 4.263741897681913e-08, + "loss": 0.6974, + "step": 43021 + }, + { + "epoch": 1.57, + "learning_rate": 4.263375691214707e-08, + "loss": 0.6818, + "step": 43022 + }, + { + "epoch": 1.57, + "learning_rate": 4.263009484747501e-08, + "loss": 0.6793, + "step": 43023 + }, + { + "epoch": 1.57, + "learning_rate": 4.262643278280294e-08, + "loss": 0.6948, + "step": 43024 + }, + { + "epoch": 1.57, + "learning_rate": 4.262277071813088e-08, + "loss": 0.6845, + "step": 43025 + }, + { + "epoch": 1.57, + "learning_rate": 4.261910865345882e-08, + "loss": 0.6848, + "step": 43026 + }, + { + "epoch": 1.57, + "learning_rate": 4.2615446588786754e-08, + "loss": 0.6912, + "step": 43027 + }, + { + "epoch": 1.57, + "learning_rate": 4.2611784524114694e-08, + "loss": 0.6826, + "step": 43028 + }, + { + "epoch": 1.57, + "learning_rate": 4.260812245944263e-08, + "loss": 0.6912, + "step": 43029 + }, + { + "epoch": 1.57, + "learning_rate": 4.260446039477057e-08, + "loss": 0.698, + "step": 43030 + }, + { + "epoch": 1.57, + "learning_rate": 4.260079833009851e-08, + "loss": 0.6932, + "step": 43031 + }, + { + "epoch": 1.57, + "learning_rate": 4.259713626542645e-08, + "loss": 0.7063, + "step": 43032 + }, + { + "epoch": 1.57, + "learning_rate": 4.259347420075438e-08, + "loss": 0.6842, + "step": 43033 + }, + { + "epoch": 1.57, + "learning_rate": 4.258981213608232e-08, + "loss": 0.6973, + "step": 43034 + }, + { + "epoch": 1.57, + "learning_rate": 4.258615007141026e-08, + "loss": 0.6919, + "step": 43035 + }, + { + "epoch": 1.57, + "learning_rate": 4.25824880067382e-08, + "loss": 0.6797, + "step": 43036 + }, + { + "epoch": 1.57, + "learning_rate": 4.2578825942066134e-08, + "loss": 0.6894, + "step": 43037 + }, + { + "epoch": 1.57, + "learning_rate": 4.2575163877394074e-08, + "loss": 0.685, + "step": 43038 + }, + { + "epoch": 1.57, + "learning_rate": 4.2571501812722014e-08, + "loss": 0.6952, + "step": 43039 + }, + { + "epoch": 1.57, + "learning_rate": 4.256783974804995e-08, + "loss": 0.6866, + "step": 43040 + }, + { + "epoch": 1.57, + "learning_rate": 4.256417768337789e-08, + "loss": 0.6905, + "step": 43041 + }, + { + "epoch": 1.57, + "learning_rate": 4.256051561870582e-08, + "loss": 0.6871, + "step": 43042 + }, + { + "epoch": 1.57, + "learning_rate": 4.255685355403376e-08, + "loss": 0.6961, + "step": 43043 + }, + { + "epoch": 1.57, + "learning_rate": 4.25531914893617e-08, + "loss": 0.6973, + "step": 43044 + }, + { + "epoch": 1.57, + "learning_rate": 4.254952942468964e-08, + "loss": 0.7118, + "step": 43045 + }, + { + "epoch": 1.57, + "learning_rate": 4.254586736001757e-08, + "loss": 0.6844, + "step": 43046 + }, + { + "epoch": 1.57, + "learning_rate": 4.254220529534551e-08, + "loss": 0.6875, + "step": 43047 + }, + { + "epoch": 1.58, + "learning_rate": 4.253854323067345e-08, + "loss": 0.705, + "step": 43048 + }, + { + "epoch": 1.58, + "learning_rate": 4.253488116600139e-08, + "loss": 0.6818, + "step": 43049 + }, + { + "epoch": 1.58, + "learning_rate": 4.253121910132933e-08, + "loss": 0.6904, + "step": 43050 + }, + { + "epoch": 1.58, + "learning_rate": 4.2527557036657266e-08, + "loss": 0.7045, + "step": 43051 + }, + { + "epoch": 1.58, + "learning_rate": 4.25238949719852e-08, + "loss": 0.6938, + "step": 43052 + }, + { + "epoch": 1.58, + "learning_rate": 4.252023290731314e-08, + "loss": 0.6988, + "step": 43053 + }, + { + "epoch": 1.58, + "learning_rate": 4.251657084264108e-08, + "loss": 0.7047, + "step": 43054 + }, + { + "epoch": 1.58, + "learning_rate": 4.251290877796901e-08, + "loss": 0.6809, + "step": 43055 + }, + { + "epoch": 1.58, + "learning_rate": 4.250924671329695e-08, + "loss": 0.6927, + "step": 43056 + }, + { + "epoch": 1.58, + "learning_rate": 4.250558464862489e-08, + "loss": 0.7077, + "step": 43057 + }, + { + "epoch": 1.58, + "learning_rate": 4.250192258395283e-08, + "loss": 0.6973, + "step": 43058 + }, + { + "epoch": 1.58, + "learning_rate": 4.249826051928077e-08, + "loss": 0.7084, + "step": 43059 + }, + { + "epoch": 1.58, + "learning_rate": 4.2494598454608706e-08, + "loss": 0.6948, + "step": 43060 + }, + { + "epoch": 1.58, + "learning_rate": 4.2490936389936646e-08, + "loss": 0.6994, + "step": 43061 + }, + { + "epoch": 1.58, + "learning_rate": 4.2487274325264586e-08, + "loss": 0.6847, + "step": 43062 + }, + { + "epoch": 1.58, + "learning_rate": 4.248361226059252e-08, + "loss": 0.708, + "step": 43063 + }, + { + "epoch": 1.58, + "learning_rate": 4.247995019592045e-08, + "loss": 0.6933, + "step": 43064 + }, + { + "epoch": 1.58, + "learning_rate": 4.247628813124839e-08, + "loss": 0.6869, + "step": 43065 + }, + { + "epoch": 1.58, + "learning_rate": 4.247262606657633e-08, + "loss": 0.6923, + "step": 43066 + }, + { + "epoch": 1.58, + "learning_rate": 4.246896400190427e-08, + "loss": 0.6965, + "step": 43067 + }, + { + "epoch": 1.58, + "learning_rate": 4.246530193723221e-08, + "loss": 0.6874, + "step": 43068 + }, + { + "epoch": 1.58, + "learning_rate": 4.2461639872560145e-08, + "loss": 0.6838, + "step": 43069 + }, + { + "epoch": 1.58, + "learning_rate": 4.2457977807888085e-08, + "loss": 0.6777, + "step": 43070 + }, + { + "epoch": 1.58, + "learning_rate": 4.2454315743216025e-08, + "loss": 0.6929, + "step": 43071 + }, + { + "epoch": 1.58, + "learning_rate": 4.2450653678543965e-08, + "loss": 0.6894, + "step": 43072 + }, + { + "epoch": 1.58, + "learning_rate": 4.24469916138719e-08, + "loss": 0.6944, + "step": 43073 + }, + { + "epoch": 1.58, + "learning_rate": 4.244332954919984e-08, + "loss": 0.699, + "step": 43074 + }, + { + "epoch": 1.58, + "learning_rate": 4.243966748452777e-08, + "loss": 0.6915, + "step": 43075 + }, + { + "epoch": 1.58, + "learning_rate": 4.243600541985571e-08, + "loss": 0.699, + "step": 43076 + }, + { + "epoch": 1.58, + "learning_rate": 4.243234335518365e-08, + "loss": 0.6909, + "step": 43077 + }, + { + "epoch": 1.58, + "learning_rate": 4.2428681290511585e-08, + "loss": 0.7056, + "step": 43078 + }, + { + "epoch": 1.58, + "learning_rate": 4.2425019225839525e-08, + "loss": 0.6878, + "step": 43079 + }, + { + "epoch": 1.58, + "learning_rate": 4.2421357161167465e-08, + "loss": 0.6813, + "step": 43080 + }, + { + "epoch": 1.58, + "learning_rate": 4.2417695096495405e-08, + "loss": 0.6918, + "step": 43081 + }, + { + "epoch": 1.58, + "learning_rate": 4.241403303182334e-08, + "loss": 0.6951, + "step": 43082 + }, + { + "epoch": 1.58, + "learning_rate": 4.241037096715128e-08, + "loss": 0.7073, + "step": 43083 + }, + { + "epoch": 1.58, + "learning_rate": 4.240670890247922e-08, + "loss": 0.6858, + "step": 43084 + }, + { + "epoch": 1.58, + "learning_rate": 4.240304683780716e-08, + "loss": 0.692, + "step": 43085 + }, + { + "epoch": 1.58, + "learning_rate": 4.239938477313509e-08, + "loss": 0.6961, + "step": 43086 + }, + { + "epoch": 1.58, + "learning_rate": 4.2395722708463024e-08, + "loss": 0.7066, + "step": 43087 + }, + { + "epoch": 1.58, + "learning_rate": 4.2392060643790964e-08, + "loss": 0.7033, + "step": 43088 + }, + { + "epoch": 1.58, + "learning_rate": 4.2388398579118904e-08, + "loss": 0.7058, + "step": 43089 + }, + { + "epoch": 1.58, + "learning_rate": 4.2384736514446844e-08, + "loss": 0.6996, + "step": 43090 + }, + { + "epoch": 1.58, + "learning_rate": 4.238107444977478e-08, + "loss": 0.6975, + "step": 43091 + }, + { + "epoch": 1.58, + "learning_rate": 4.237741238510272e-08, + "loss": 0.6925, + "step": 43092 + }, + { + "epoch": 1.58, + "learning_rate": 4.237375032043066e-08, + "loss": 0.7058, + "step": 43093 + }, + { + "epoch": 1.58, + "learning_rate": 4.23700882557586e-08, + "loss": 0.6895, + "step": 43094 + }, + { + "epoch": 1.58, + "learning_rate": 4.236642619108654e-08, + "loss": 0.6856, + "step": 43095 + }, + { + "epoch": 1.58, + "learning_rate": 4.236276412641447e-08, + "loss": 0.696, + "step": 43096 + }, + { + "epoch": 1.58, + "learning_rate": 4.235910206174241e-08, + "loss": 0.702, + "step": 43097 + }, + { + "epoch": 1.58, + "learning_rate": 4.2355439997070344e-08, + "loss": 0.6831, + "step": 43098 + }, + { + "epoch": 1.58, + "learning_rate": 4.2351777932398284e-08, + "loss": 0.6906, + "step": 43099 + }, + { + "epoch": 1.58, + "learning_rate": 4.234811586772622e-08, + "loss": 0.6908, + "step": 43100 + }, + { + "epoch": 1.58, + "learning_rate": 4.234445380305416e-08, + "loss": 0.7029, + "step": 43101 + }, + { + "epoch": 1.58, + "learning_rate": 4.23407917383821e-08, + "loss": 0.7043, + "step": 43102 + }, + { + "epoch": 1.58, + "learning_rate": 4.233712967371004e-08, + "loss": 0.6912, + "step": 43103 + }, + { + "epoch": 1.58, + "learning_rate": 4.2333467609037977e-08, + "loss": 0.6862, + "step": 43104 + }, + { + "epoch": 1.58, + "learning_rate": 4.232980554436591e-08, + "loss": 0.6968, + "step": 43105 + }, + { + "epoch": 1.58, + "learning_rate": 4.232614347969385e-08, + "loss": 0.7014, + "step": 43106 + }, + { + "epoch": 1.58, + "learning_rate": 4.232248141502179e-08, + "loss": 0.6973, + "step": 43107 + }, + { + "epoch": 1.58, + "learning_rate": 4.231881935034973e-08, + "loss": 0.6722, + "step": 43108 + }, + { + "epoch": 1.58, + "learning_rate": 4.231515728567766e-08, + "loss": 0.6873, + "step": 43109 + }, + { + "epoch": 1.58, + "learning_rate": 4.23114952210056e-08, + "loss": 0.7035, + "step": 43110 + }, + { + "epoch": 1.58, + "learning_rate": 4.2307833156333536e-08, + "loss": 0.6992, + "step": 43111 + }, + { + "epoch": 1.58, + "learning_rate": 4.2304171091661476e-08, + "loss": 0.6934, + "step": 43112 + }, + { + "epoch": 1.58, + "learning_rate": 4.2300509026989416e-08, + "loss": 0.7016, + "step": 43113 + }, + { + "epoch": 1.58, + "learning_rate": 4.229684696231735e-08, + "loss": 0.7044, + "step": 43114 + }, + { + "epoch": 1.58, + "learning_rate": 4.229318489764529e-08, + "loss": 0.6846, + "step": 43115 + }, + { + "epoch": 1.58, + "learning_rate": 4.228952283297323e-08, + "loss": 0.6843, + "step": 43116 + }, + { + "epoch": 1.58, + "learning_rate": 4.228586076830117e-08, + "loss": 0.6878, + "step": 43117 + }, + { + "epoch": 1.58, + "learning_rate": 4.22821987036291e-08, + "loss": 0.6917, + "step": 43118 + }, + { + "epoch": 1.58, + "learning_rate": 4.227853663895704e-08, + "loss": 0.6923, + "step": 43119 + }, + { + "epoch": 1.58, + "learning_rate": 4.227487457428498e-08, + "loss": 0.6957, + "step": 43120 + }, + { + "epoch": 1.58, + "learning_rate": 4.227121250961292e-08, + "loss": 0.696, + "step": 43121 + }, + { + "epoch": 1.58, + "learning_rate": 4.2267550444940856e-08, + "loss": 0.686, + "step": 43122 + }, + { + "epoch": 1.58, + "learning_rate": 4.226388838026879e-08, + "loss": 0.6867, + "step": 43123 + }, + { + "epoch": 1.58, + "learning_rate": 4.226022631559673e-08, + "loss": 0.6965, + "step": 43124 + }, + { + "epoch": 1.58, + "learning_rate": 4.225656425092467e-08, + "loss": 0.6946, + "step": 43125 + }, + { + "epoch": 1.58, + "learning_rate": 4.225290218625261e-08, + "loss": 0.689, + "step": 43126 + }, + { + "epoch": 1.58, + "learning_rate": 4.224924012158054e-08, + "loss": 0.6919, + "step": 43127 + }, + { + "epoch": 1.58, + "learning_rate": 4.224557805690848e-08, + "loss": 0.6793, + "step": 43128 + }, + { + "epoch": 1.58, + "learning_rate": 4.224191599223642e-08, + "loss": 0.6968, + "step": 43129 + }, + { + "epoch": 1.58, + "learning_rate": 4.223825392756436e-08, + "loss": 0.6982, + "step": 43130 + }, + { + "epoch": 1.58, + "learning_rate": 4.22345918628923e-08, + "loss": 0.698, + "step": 43131 + }, + { + "epoch": 1.58, + "learning_rate": 4.2230929798220235e-08, + "loss": 0.6989, + "step": 43132 + }, + { + "epoch": 1.58, + "learning_rate": 4.2227267733548175e-08, + "loss": 0.6994, + "step": 43133 + }, + { + "epoch": 1.58, + "learning_rate": 4.222360566887611e-08, + "loss": 0.6925, + "step": 43134 + }, + { + "epoch": 1.58, + "learning_rate": 4.221994360420405e-08, + "loss": 0.6836, + "step": 43135 + }, + { + "epoch": 1.58, + "learning_rate": 4.221628153953198e-08, + "loss": 0.6928, + "step": 43136 + }, + { + "epoch": 1.58, + "learning_rate": 4.221261947485992e-08, + "loss": 0.7011, + "step": 43137 + }, + { + "epoch": 1.58, + "learning_rate": 4.220895741018786e-08, + "loss": 0.6986, + "step": 43138 + }, + { + "epoch": 1.58, + "learning_rate": 4.22052953455158e-08, + "loss": 0.6998, + "step": 43139 + }, + { + "epoch": 1.58, + "learning_rate": 4.220163328084374e-08, + "loss": 0.6987, + "step": 43140 + }, + { + "epoch": 1.58, + "learning_rate": 4.2197971216171675e-08, + "loss": 0.6773, + "step": 43141 + }, + { + "epoch": 1.58, + "learning_rate": 4.2194309151499615e-08, + "loss": 0.6845, + "step": 43142 + }, + { + "epoch": 1.58, + "learning_rate": 4.2190647086827554e-08, + "loss": 0.6941, + "step": 43143 + }, + { + "epoch": 1.58, + "learning_rate": 4.2186985022155494e-08, + "loss": 0.7075, + "step": 43144 + }, + { + "epoch": 1.58, + "learning_rate": 4.218332295748343e-08, + "loss": 0.6989, + "step": 43145 + }, + { + "epoch": 1.58, + "learning_rate": 4.217966089281136e-08, + "loss": 0.6812, + "step": 43146 + }, + { + "epoch": 1.58, + "learning_rate": 4.21759988281393e-08, + "loss": 0.6903, + "step": 43147 + }, + { + "epoch": 1.58, + "learning_rate": 4.217233676346724e-08, + "loss": 0.7097, + "step": 43148 + }, + { + "epoch": 1.58, + "learning_rate": 4.216867469879518e-08, + "loss": 0.7004, + "step": 43149 + }, + { + "epoch": 1.58, + "learning_rate": 4.2165012634123114e-08, + "loss": 0.6991, + "step": 43150 + }, + { + "epoch": 1.58, + "learning_rate": 4.2161350569451054e-08, + "loss": 0.6885, + "step": 43151 + }, + { + "epoch": 1.58, + "learning_rate": 4.2157688504778994e-08, + "loss": 0.6935, + "step": 43152 + }, + { + "epoch": 1.58, + "learning_rate": 4.2154026440106934e-08, + "loss": 0.7117, + "step": 43153 + }, + { + "epoch": 1.58, + "learning_rate": 4.215036437543487e-08, + "loss": 0.6932, + "step": 43154 + }, + { + "epoch": 1.58, + "learning_rate": 4.214670231076281e-08, + "loss": 0.6896, + "step": 43155 + }, + { + "epoch": 1.58, + "learning_rate": 4.214304024609075e-08, + "loss": 0.6925, + "step": 43156 + }, + { + "epoch": 1.58, + "learning_rate": 4.213937818141868e-08, + "loss": 0.6904, + "step": 43157 + }, + { + "epoch": 1.58, + "learning_rate": 4.213571611674662e-08, + "loss": 0.7065, + "step": 43158 + }, + { + "epoch": 1.58, + "learning_rate": 4.2132054052074554e-08, + "loss": 0.6886, + "step": 43159 + }, + { + "epoch": 1.58, + "learning_rate": 4.2128391987402494e-08, + "loss": 0.6888, + "step": 43160 + }, + { + "epoch": 1.58, + "learning_rate": 4.2124729922730433e-08, + "loss": 0.7035, + "step": 43161 + }, + { + "epoch": 1.58, + "learning_rate": 4.2121067858058373e-08, + "loss": 0.6897, + "step": 43162 + }, + { + "epoch": 1.58, + "learning_rate": 4.211740579338631e-08, + "loss": 0.6927, + "step": 43163 + }, + { + "epoch": 1.58, + "learning_rate": 4.2113743728714247e-08, + "loss": 0.693, + "step": 43164 + }, + { + "epoch": 1.58, + "learning_rate": 4.2110081664042187e-08, + "loss": 0.6948, + "step": 43165 + }, + { + "epoch": 1.58, + "learning_rate": 4.2106419599370127e-08, + "loss": 0.6958, + "step": 43166 + }, + { + "epoch": 1.58, + "learning_rate": 4.2102757534698066e-08, + "loss": 0.6821, + "step": 43167 + }, + { + "epoch": 1.58, + "learning_rate": 4.2099095470026e-08, + "loss": 0.6998, + "step": 43168 + }, + { + "epoch": 1.58, + "learning_rate": 4.209543340535393e-08, + "loss": 0.6827, + "step": 43169 + }, + { + "epoch": 1.58, + "learning_rate": 4.209177134068187e-08, + "loss": 0.6926, + "step": 43170 + }, + { + "epoch": 1.58, + "learning_rate": 4.208810927600981e-08, + "loss": 0.6905, + "step": 43171 + }, + { + "epoch": 1.58, + "learning_rate": 4.2084447211337746e-08, + "loss": 0.6929, + "step": 43172 + }, + { + "epoch": 1.58, + "learning_rate": 4.2080785146665686e-08, + "loss": 0.7093, + "step": 43173 + }, + { + "epoch": 1.58, + "learning_rate": 4.2077123081993626e-08, + "loss": 0.696, + "step": 43174 + }, + { + "epoch": 1.58, + "learning_rate": 4.2073461017321566e-08, + "loss": 0.6866, + "step": 43175 + }, + { + "epoch": 1.58, + "learning_rate": 4.2069798952649506e-08, + "loss": 0.6872, + "step": 43176 + }, + { + "epoch": 1.58, + "learning_rate": 4.206613688797744e-08, + "loss": 0.6927, + "step": 43177 + }, + { + "epoch": 1.58, + "learning_rate": 4.206247482330538e-08, + "loss": 0.6959, + "step": 43178 + }, + { + "epoch": 1.58, + "learning_rate": 4.205881275863332e-08, + "loss": 0.6806, + "step": 43179 + }, + { + "epoch": 1.58, + "learning_rate": 4.205515069396125e-08, + "loss": 0.6997, + "step": 43180 + }, + { + "epoch": 1.58, + "learning_rate": 4.205148862928919e-08, + "loss": 0.7027, + "step": 43181 + }, + { + "epoch": 1.58, + "learning_rate": 4.2047826564617126e-08, + "loss": 0.6993, + "step": 43182 + }, + { + "epoch": 1.58, + "learning_rate": 4.2044164499945066e-08, + "loss": 0.6704, + "step": 43183 + }, + { + "epoch": 1.58, + "learning_rate": 4.2040502435273006e-08, + "loss": 0.6984, + "step": 43184 + }, + { + "epoch": 1.58, + "learning_rate": 4.2036840370600945e-08, + "loss": 0.7188, + "step": 43185 + }, + { + "epoch": 1.58, + "learning_rate": 4.203317830592888e-08, + "loss": 0.6874, + "step": 43186 + }, + { + "epoch": 1.58, + "learning_rate": 4.202951624125682e-08, + "loss": 0.685, + "step": 43187 + }, + { + "epoch": 1.58, + "learning_rate": 4.202585417658476e-08, + "loss": 0.695, + "step": 43188 + }, + { + "epoch": 1.58, + "learning_rate": 4.20221921119127e-08, + "loss": 0.688, + "step": 43189 + }, + { + "epoch": 1.58, + "learning_rate": 4.201853004724063e-08, + "loss": 0.6849, + "step": 43190 + }, + { + "epoch": 1.58, + "learning_rate": 4.201486798256857e-08, + "loss": 0.6932, + "step": 43191 + }, + { + "epoch": 1.58, + "learning_rate": 4.2011205917896505e-08, + "loss": 0.6921, + "step": 43192 + }, + { + "epoch": 1.58, + "learning_rate": 4.2007543853224445e-08, + "loss": 0.6948, + "step": 43193 + }, + { + "epoch": 1.58, + "learning_rate": 4.2003881788552385e-08, + "loss": 0.6926, + "step": 43194 + }, + { + "epoch": 1.58, + "learning_rate": 4.200021972388032e-08, + "loss": 0.6958, + "step": 43195 + }, + { + "epoch": 1.58, + "learning_rate": 4.199655765920826e-08, + "loss": 0.6848, + "step": 43196 + }, + { + "epoch": 1.58, + "learning_rate": 4.19928955945362e-08, + "loss": 0.6989, + "step": 43197 + }, + { + "epoch": 1.58, + "learning_rate": 4.198923352986414e-08, + "loss": 0.6896, + "step": 43198 + }, + { + "epoch": 1.58, + "learning_rate": 4.198557146519207e-08, + "loss": 0.6928, + "step": 43199 + }, + { + "epoch": 1.58, + "learning_rate": 4.198190940052001e-08, + "loss": 0.6928, + "step": 43200 + }, + { + "epoch": 1.58, + "learning_rate": 4.197824733584795e-08, + "loss": 0.689, + "step": 43201 + }, + { + "epoch": 1.58, + "learning_rate": 4.197458527117589e-08, + "loss": 0.6965, + "step": 43202 + }, + { + "epoch": 1.58, + "learning_rate": 4.1970923206503824e-08, + "loss": 0.6919, + "step": 43203 + }, + { + "epoch": 1.58, + "learning_rate": 4.1967261141831764e-08, + "loss": 0.687, + "step": 43204 + }, + { + "epoch": 1.58, + "learning_rate": 4.19635990771597e-08, + "loss": 0.7011, + "step": 43205 + }, + { + "epoch": 1.58, + "learning_rate": 4.195993701248764e-08, + "loss": 0.6976, + "step": 43206 + }, + { + "epoch": 1.58, + "learning_rate": 4.195627494781558e-08, + "loss": 0.7002, + "step": 43207 + }, + { + "epoch": 1.58, + "learning_rate": 4.195261288314351e-08, + "loss": 0.6912, + "step": 43208 + }, + { + "epoch": 1.58, + "learning_rate": 4.194895081847145e-08, + "loss": 0.7089, + "step": 43209 + }, + { + "epoch": 1.58, + "learning_rate": 4.194528875379939e-08, + "loss": 0.7025, + "step": 43210 + }, + { + "epoch": 1.58, + "learning_rate": 4.194162668912733e-08, + "loss": 0.6837, + "step": 43211 + }, + { + "epoch": 1.58, + "learning_rate": 4.193796462445527e-08, + "loss": 0.6994, + "step": 43212 + }, + { + "epoch": 1.58, + "learning_rate": 4.1934302559783204e-08, + "loss": 0.6912, + "step": 43213 + }, + { + "epoch": 1.58, + "learning_rate": 4.1930640495111144e-08, + "loss": 0.6926, + "step": 43214 + }, + { + "epoch": 1.58, + "learning_rate": 4.1926978430439084e-08, + "loss": 0.6881, + "step": 43215 + }, + { + "epoch": 1.58, + "learning_rate": 4.192331636576702e-08, + "loss": 0.6975, + "step": 43216 + }, + { + "epoch": 1.58, + "learning_rate": 4.191965430109495e-08, + "loss": 0.6853, + "step": 43217 + }, + { + "epoch": 1.58, + "learning_rate": 4.191599223642289e-08, + "loss": 0.6978, + "step": 43218 + }, + { + "epoch": 1.58, + "learning_rate": 4.191233017175083e-08, + "loss": 0.6908, + "step": 43219 + }, + { + "epoch": 1.58, + "learning_rate": 4.190866810707877e-08, + "loss": 0.684, + "step": 43220 + }, + { + "epoch": 1.58, + "learning_rate": 4.190500604240671e-08, + "loss": 0.6909, + "step": 43221 + }, + { + "epoch": 1.58, + "learning_rate": 4.1901343977734643e-08, + "loss": 0.6898, + "step": 43222 + }, + { + "epoch": 1.58, + "learning_rate": 4.1897681913062583e-08, + "loss": 0.6979, + "step": 43223 + }, + { + "epoch": 1.58, + "learning_rate": 4.189401984839052e-08, + "loss": 0.6878, + "step": 43224 + }, + { + "epoch": 1.58, + "learning_rate": 4.189035778371846e-08, + "loss": 0.6988, + "step": 43225 + }, + { + "epoch": 1.58, + "learning_rate": 4.1886695719046396e-08, + "loss": 0.7006, + "step": 43226 + }, + { + "epoch": 1.58, + "learning_rate": 4.1883033654374336e-08, + "loss": 0.703, + "step": 43227 + }, + { + "epoch": 1.58, + "learning_rate": 4.187937158970227e-08, + "loss": 0.6814, + "step": 43228 + }, + { + "epoch": 1.58, + "learning_rate": 4.187570952503021e-08, + "loss": 0.6903, + "step": 43229 + }, + { + "epoch": 1.58, + "learning_rate": 4.187204746035814e-08, + "loss": 0.6902, + "step": 43230 + }, + { + "epoch": 1.58, + "learning_rate": 4.186838539568608e-08, + "loss": 0.699, + "step": 43231 + }, + { + "epoch": 1.58, + "learning_rate": 4.186472333101402e-08, + "loss": 0.6976, + "step": 43232 + }, + { + "epoch": 1.58, + "learning_rate": 4.186106126634196e-08, + "loss": 0.6947, + "step": 43233 + }, + { + "epoch": 1.58, + "learning_rate": 4.18573992016699e-08, + "loss": 0.7011, + "step": 43234 + }, + { + "epoch": 1.58, + "learning_rate": 4.1853737136997836e-08, + "loss": 0.6909, + "step": 43235 + }, + { + "epoch": 1.58, + "learning_rate": 4.1850075072325776e-08, + "loss": 0.6948, + "step": 43236 + }, + { + "epoch": 1.58, + "learning_rate": 4.1846413007653716e-08, + "loss": 0.6836, + "step": 43237 + }, + { + "epoch": 1.58, + "learning_rate": 4.1842750942981656e-08, + "loss": 0.7082, + "step": 43238 + }, + { + "epoch": 1.58, + "learning_rate": 4.183908887830959e-08, + "loss": 0.6993, + "step": 43239 + }, + { + "epoch": 1.58, + "learning_rate": 4.183542681363752e-08, + "loss": 0.7094, + "step": 43240 + }, + { + "epoch": 1.58, + "learning_rate": 4.183176474896546e-08, + "loss": 0.6946, + "step": 43241 + }, + { + "epoch": 1.58, + "learning_rate": 4.18281026842934e-08, + "loss": 0.682, + "step": 43242 + }, + { + "epoch": 1.58, + "learning_rate": 4.182444061962134e-08, + "loss": 0.6936, + "step": 43243 + }, + { + "epoch": 1.58, + "learning_rate": 4.1820778554949276e-08, + "loss": 0.7055, + "step": 43244 + }, + { + "epoch": 1.58, + "learning_rate": 4.1817116490277215e-08, + "loss": 0.6939, + "step": 43245 + }, + { + "epoch": 1.58, + "learning_rate": 4.1813454425605155e-08, + "loss": 0.7033, + "step": 43246 + }, + { + "epoch": 1.58, + "learning_rate": 4.1809792360933095e-08, + "loss": 0.6919, + "step": 43247 + }, + { + "epoch": 1.58, + "learning_rate": 4.180613029626103e-08, + "loss": 0.6934, + "step": 43248 + }, + { + "epoch": 1.58, + "learning_rate": 4.180246823158897e-08, + "loss": 0.6944, + "step": 43249 + }, + { + "epoch": 1.58, + "learning_rate": 4.179880616691691e-08, + "loss": 0.6841, + "step": 43250 + }, + { + "epoch": 1.58, + "learning_rate": 4.179514410224484e-08, + "loss": 0.6894, + "step": 43251 + }, + { + "epoch": 1.58, + "learning_rate": 4.179148203757278e-08, + "loss": 0.6976, + "step": 43252 + }, + { + "epoch": 1.58, + "learning_rate": 4.1787819972900715e-08, + "loss": 0.6948, + "step": 43253 + }, + { + "epoch": 1.58, + "learning_rate": 4.1784157908228655e-08, + "loss": 0.6925, + "step": 43254 + }, + { + "epoch": 1.58, + "learning_rate": 4.1780495843556595e-08, + "loss": 0.6829, + "step": 43255 + }, + { + "epoch": 1.58, + "learning_rate": 4.1776833778884535e-08, + "loss": 0.6959, + "step": 43256 + }, + { + "epoch": 1.58, + "learning_rate": 4.177317171421247e-08, + "loss": 0.6874, + "step": 43257 + }, + { + "epoch": 1.58, + "learning_rate": 4.176950964954041e-08, + "loss": 0.6996, + "step": 43258 + }, + { + "epoch": 1.58, + "learning_rate": 4.176584758486835e-08, + "loss": 0.6905, + "step": 43259 + }, + { + "epoch": 1.58, + "learning_rate": 4.176218552019629e-08, + "loss": 0.6934, + "step": 43260 + }, + { + "epoch": 1.58, + "learning_rate": 4.175852345552423e-08, + "loss": 0.7012, + "step": 43261 + }, + { + "epoch": 1.58, + "learning_rate": 4.175486139085216e-08, + "loss": 0.6934, + "step": 43262 + }, + { + "epoch": 1.58, + "learning_rate": 4.1751199326180094e-08, + "loss": 0.6919, + "step": 43263 + }, + { + "epoch": 1.58, + "learning_rate": 4.1747537261508034e-08, + "loss": 0.7115, + "step": 43264 + }, + { + "epoch": 1.58, + "learning_rate": 4.1743875196835974e-08, + "loss": 0.6986, + "step": 43265 + }, + { + "epoch": 1.58, + "learning_rate": 4.174021313216391e-08, + "loss": 0.6805, + "step": 43266 + }, + { + "epoch": 1.58, + "learning_rate": 4.173655106749185e-08, + "loss": 0.6862, + "step": 43267 + }, + { + "epoch": 1.58, + "learning_rate": 4.173288900281979e-08, + "loss": 0.7034, + "step": 43268 + }, + { + "epoch": 1.58, + "learning_rate": 4.172922693814773e-08, + "loss": 0.7042, + "step": 43269 + }, + { + "epoch": 1.58, + "learning_rate": 4.172556487347567e-08, + "loss": 0.7035, + "step": 43270 + }, + { + "epoch": 1.58, + "learning_rate": 4.17219028088036e-08, + "loss": 0.6884, + "step": 43271 + }, + { + "epoch": 1.58, + "learning_rate": 4.171824074413154e-08, + "loss": 0.6958, + "step": 43272 + }, + { + "epoch": 1.58, + "learning_rate": 4.171457867945948e-08, + "loss": 0.6911, + "step": 43273 + }, + { + "epoch": 1.58, + "learning_rate": 4.1710916614787414e-08, + "loss": 0.7049, + "step": 43274 + }, + { + "epoch": 1.58, + "learning_rate": 4.1707254550115354e-08, + "loss": 0.6914, + "step": 43275 + }, + { + "epoch": 1.58, + "learning_rate": 4.170359248544329e-08, + "loss": 0.691, + "step": 43276 + }, + { + "epoch": 1.58, + "learning_rate": 4.169993042077123e-08, + "loss": 0.6967, + "step": 43277 + }, + { + "epoch": 1.58, + "learning_rate": 4.169626835609917e-08, + "loss": 0.7006, + "step": 43278 + }, + { + "epoch": 1.58, + "learning_rate": 4.169260629142711e-08, + "loss": 0.6973, + "step": 43279 + }, + { + "epoch": 1.58, + "learning_rate": 4.168894422675504e-08, + "loss": 0.6912, + "step": 43280 + }, + { + "epoch": 1.58, + "learning_rate": 4.168528216208298e-08, + "loss": 0.6929, + "step": 43281 + }, + { + "epoch": 1.58, + "learning_rate": 4.168162009741092e-08, + "loss": 0.6883, + "step": 43282 + }, + { + "epoch": 1.58, + "learning_rate": 4.167795803273886e-08, + "loss": 0.7048, + "step": 43283 + }, + { + "epoch": 1.58, + "learning_rate": 4.167429596806679e-08, + "loss": 0.6835, + "step": 43284 + }, + { + "epoch": 1.58, + "learning_rate": 4.167063390339473e-08, + "loss": 0.6909, + "step": 43285 + }, + { + "epoch": 1.58, + "learning_rate": 4.166697183872267e-08, + "loss": 0.6747, + "step": 43286 + }, + { + "epoch": 1.58, + "learning_rate": 4.1663309774050606e-08, + "loss": 0.6866, + "step": 43287 + }, + { + "epoch": 1.58, + "learning_rate": 4.1659647709378546e-08, + "loss": 0.6916, + "step": 43288 + }, + { + "epoch": 1.58, + "learning_rate": 4.165598564470648e-08, + "loss": 0.6985, + "step": 43289 + }, + { + "epoch": 1.58, + "learning_rate": 4.165232358003442e-08, + "loss": 0.6897, + "step": 43290 + }, + { + "epoch": 1.58, + "learning_rate": 4.164866151536236e-08, + "loss": 0.6989, + "step": 43291 + }, + { + "epoch": 1.58, + "learning_rate": 4.16449994506903e-08, + "loss": 0.6956, + "step": 43292 + }, + { + "epoch": 1.58, + "learning_rate": 4.164133738601823e-08, + "loss": 0.6791, + "step": 43293 + }, + { + "epoch": 1.58, + "learning_rate": 4.163767532134617e-08, + "loss": 0.6984, + "step": 43294 + }, + { + "epoch": 1.58, + "learning_rate": 4.163401325667411e-08, + "loss": 0.6961, + "step": 43295 + }, + { + "epoch": 1.58, + "learning_rate": 4.163035119200205e-08, + "loss": 0.6955, + "step": 43296 + }, + { + "epoch": 1.58, + "learning_rate": 4.162668912732999e-08, + "loss": 0.675, + "step": 43297 + }, + { + "epoch": 1.58, + "learning_rate": 4.1623027062657926e-08, + "loss": 0.6857, + "step": 43298 + }, + { + "epoch": 1.58, + "learning_rate": 4.161936499798586e-08, + "loss": 0.6984, + "step": 43299 + }, + { + "epoch": 1.58, + "learning_rate": 4.16157029333138e-08, + "loss": 0.6946, + "step": 43300 + }, + { + "epoch": 1.58, + "learning_rate": 4.161204086864174e-08, + "loss": 0.6789, + "step": 43301 + }, + { + "epoch": 1.58, + "learning_rate": 4.160837880396967e-08, + "loss": 0.6991, + "step": 43302 + }, + { + "epoch": 1.58, + "learning_rate": 4.160471673929761e-08, + "loss": 0.6918, + "step": 43303 + }, + { + "epoch": 1.58, + "learning_rate": 4.160105467462555e-08, + "loss": 0.6979, + "step": 43304 + }, + { + "epoch": 1.58, + "learning_rate": 4.159739260995349e-08, + "loss": 0.702, + "step": 43305 + }, + { + "epoch": 1.58, + "learning_rate": 4.159373054528143e-08, + "loss": 0.6988, + "step": 43306 + }, + { + "epoch": 1.58, + "learning_rate": 4.1590068480609365e-08, + "loss": 0.6892, + "step": 43307 + }, + { + "epoch": 1.58, + "learning_rate": 4.1586406415937305e-08, + "loss": 0.7012, + "step": 43308 + }, + { + "epoch": 1.58, + "learning_rate": 4.1582744351265245e-08, + "loss": 0.6863, + "step": 43309 + }, + { + "epoch": 1.58, + "learning_rate": 4.157908228659318e-08, + "loss": 0.6939, + "step": 43310 + }, + { + "epoch": 1.58, + "learning_rate": 4.157542022192111e-08, + "loss": 0.6906, + "step": 43311 + }, + { + "epoch": 1.58, + "learning_rate": 4.157175815724905e-08, + "loss": 0.6944, + "step": 43312 + }, + { + "epoch": 1.58, + "learning_rate": 4.156809609257699e-08, + "loss": 0.7026, + "step": 43313 + }, + { + "epoch": 1.58, + "learning_rate": 4.156443402790493e-08, + "loss": 0.7021, + "step": 43314 + }, + { + "epoch": 1.58, + "learning_rate": 4.156077196323287e-08, + "loss": 0.6986, + "step": 43315 + }, + { + "epoch": 1.58, + "learning_rate": 4.1557109898560805e-08, + "loss": 0.6977, + "step": 43316 + }, + { + "epoch": 1.58, + "learning_rate": 4.1553447833888745e-08, + "loss": 0.6998, + "step": 43317 + }, + { + "epoch": 1.58, + "learning_rate": 4.1549785769216685e-08, + "loss": 0.6929, + "step": 43318 + }, + { + "epoch": 1.58, + "learning_rate": 4.1546123704544625e-08, + "loss": 0.6867, + "step": 43319 + }, + { + "epoch": 1.58, + "learning_rate": 4.154246163987256e-08, + "loss": 0.7013, + "step": 43320 + }, + { + "epoch": 1.58, + "learning_rate": 4.15387995752005e-08, + "loss": 0.696, + "step": 43321 + }, + { + "epoch": 1.59, + "learning_rate": 4.153513751052843e-08, + "loss": 0.6944, + "step": 43322 + }, + { + "epoch": 1.59, + "learning_rate": 4.153147544585637e-08, + "loss": 0.6902, + "step": 43323 + }, + { + "epoch": 1.59, + "learning_rate": 4.152781338118431e-08, + "loss": 0.6956, + "step": 43324 + }, + { + "epoch": 1.59, + "learning_rate": 4.1524151316512244e-08, + "loss": 0.6975, + "step": 43325 + }, + { + "epoch": 1.59, + "learning_rate": 4.1520489251840184e-08, + "loss": 0.6889, + "step": 43326 + }, + { + "epoch": 1.59, + "learning_rate": 4.1516827187168124e-08, + "loss": 0.6934, + "step": 43327 + }, + { + "epoch": 1.59, + "learning_rate": 4.1513165122496064e-08, + "loss": 0.6882, + "step": 43328 + }, + { + "epoch": 1.59, + "learning_rate": 4.1509503057824e-08, + "loss": 0.6828, + "step": 43329 + }, + { + "epoch": 1.59, + "learning_rate": 4.150584099315194e-08, + "loss": 0.6908, + "step": 43330 + }, + { + "epoch": 1.59, + "learning_rate": 4.150217892847988e-08, + "loss": 0.6802, + "step": 43331 + }, + { + "epoch": 1.59, + "learning_rate": 4.149851686380782e-08, + "loss": 0.6982, + "step": 43332 + }, + { + "epoch": 1.59, + "learning_rate": 4.149485479913575e-08, + "loss": 0.6965, + "step": 43333 + }, + { + "epoch": 1.59, + "learning_rate": 4.1491192734463684e-08, + "loss": 0.7008, + "step": 43334 + }, + { + "epoch": 1.59, + "learning_rate": 4.1487530669791624e-08, + "loss": 0.6873, + "step": 43335 + }, + { + "epoch": 1.59, + "learning_rate": 4.1483868605119564e-08, + "loss": 0.6949, + "step": 43336 + }, + { + "epoch": 1.59, + "learning_rate": 4.1480206540447504e-08, + "loss": 0.7018, + "step": 43337 + }, + { + "epoch": 1.59, + "learning_rate": 4.147654447577544e-08, + "loss": 0.6991, + "step": 43338 + }, + { + "epoch": 1.59, + "learning_rate": 4.147288241110338e-08, + "loss": 0.6878, + "step": 43339 + }, + { + "epoch": 1.59, + "learning_rate": 4.146922034643132e-08, + "loss": 0.6869, + "step": 43340 + }, + { + "epoch": 1.59, + "learning_rate": 4.1465558281759257e-08, + "loss": 0.6972, + "step": 43341 + }, + { + "epoch": 1.59, + "learning_rate": 4.1461896217087197e-08, + "loss": 0.6953, + "step": 43342 + }, + { + "epoch": 1.59, + "learning_rate": 4.145823415241513e-08, + "loss": 0.6995, + "step": 43343 + }, + { + "epoch": 1.59, + "learning_rate": 4.145457208774307e-08, + "loss": 0.6926, + "step": 43344 + }, + { + "epoch": 1.59, + "learning_rate": 4.1450910023071e-08, + "loss": 0.6975, + "step": 43345 + }, + { + "epoch": 1.59, + "learning_rate": 4.144724795839894e-08, + "loss": 0.6936, + "step": 43346 + }, + { + "epoch": 1.59, + "learning_rate": 4.1443585893726876e-08, + "loss": 0.6859, + "step": 43347 + }, + { + "epoch": 1.59, + "learning_rate": 4.1439923829054816e-08, + "loss": 0.6794, + "step": 43348 + }, + { + "epoch": 1.59, + "learning_rate": 4.1436261764382756e-08, + "loss": 0.6945, + "step": 43349 + }, + { + "epoch": 1.59, + "learning_rate": 4.1432599699710696e-08, + "loss": 0.6826, + "step": 43350 + }, + { + "epoch": 1.59, + "learning_rate": 4.1428937635038636e-08, + "loss": 0.6888, + "step": 43351 + }, + { + "epoch": 1.59, + "learning_rate": 4.142527557036657e-08, + "loss": 0.6901, + "step": 43352 + }, + { + "epoch": 1.59, + "learning_rate": 4.142161350569451e-08, + "loss": 0.6985, + "step": 43353 + }, + { + "epoch": 1.59, + "learning_rate": 4.141795144102245e-08, + "loss": 0.6961, + "step": 43354 + }, + { + "epoch": 1.59, + "learning_rate": 4.141428937635039e-08, + "loss": 0.696, + "step": 43355 + }, + { + "epoch": 1.59, + "learning_rate": 4.141062731167832e-08, + "loss": 0.6899, + "step": 43356 + }, + { + "epoch": 1.59, + "learning_rate": 4.140696524700626e-08, + "loss": 0.6966, + "step": 43357 + }, + { + "epoch": 1.59, + "learning_rate": 4.1403303182334196e-08, + "loss": 0.7106, + "step": 43358 + }, + { + "epoch": 1.59, + "learning_rate": 4.1399641117662136e-08, + "loss": 0.6894, + "step": 43359 + }, + { + "epoch": 1.59, + "learning_rate": 4.1395979052990076e-08, + "loss": 0.6876, + "step": 43360 + }, + { + "epoch": 1.59, + "learning_rate": 4.139231698831801e-08, + "loss": 0.6951, + "step": 43361 + }, + { + "epoch": 1.59, + "learning_rate": 4.138865492364595e-08, + "loss": 0.6911, + "step": 43362 + }, + { + "epoch": 1.59, + "learning_rate": 4.138499285897389e-08, + "loss": 0.6984, + "step": 43363 + }, + { + "epoch": 1.59, + "learning_rate": 4.138133079430183e-08, + "loss": 0.6852, + "step": 43364 + }, + { + "epoch": 1.59, + "learning_rate": 4.137766872962976e-08, + "loss": 0.6926, + "step": 43365 + }, + { + "epoch": 1.59, + "learning_rate": 4.13740066649577e-08, + "loss": 0.7065, + "step": 43366 + }, + { + "epoch": 1.59, + "learning_rate": 4.137034460028564e-08, + "loss": 0.6876, + "step": 43367 + }, + { + "epoch": 1.59, + "learning_rate": 4.1366682535613575e-08, + "loss": 0.7008, + "step": 43368 + }, + { + "epoch": 1.59, + "learning_rate": 4.1363020470941515e-08, + "loss": 0.6938, + "step": 43369 + }, + { + "epoch": 1.59, + "learning_rate": 4.135935840626945e-08, + "loss": 0.6948, + "step": 43370 + }, + { + "epoch": 1.59, + "learning_rate": 4.135569634159739e-08, + "loss": 0.6919, + "step": 43371 + }, + { + "epoch": 1.59, + "learning_rate": 4.135203427692533e-08, + "loss": 0.6981, + "step": 43372 + }, + { + "epoch": 1.59, + "learning_rate": 4.134837221225327e-08, + "loss": 0.6933, + "step": 43373 + }, + { + "epoch": 1.59, + "learning_rate": 4.13447101475812e-08, + "loss": 0.6921, + "step": 43374 + }, + { + "epoch": 1.59, + "learning_rate": 4.134104808290914e-08, + "loss": 0.6901, + "step": 43375 + }, + { + "epoch": 1.59, + "learning_rate": 4.133738601823708e-08, + "loss": 0.7051, + "step": 43376 + }, + { + "epoch": 1.59, + "learning_rate": 4.133372395356502e-08, + "loss": 0.6813, + "step": 43377 + }, + { + "epoch": 1.59, + "learning_rate": 4.133006188889296e-08, + "loss": 0.6933, + "step": 43378 + }, + { + "epoch": 1.59, + "learning_rate": 4.1326399824220895e-08, + "loss": 0.6976, + "step": 43379 + }, + { + "epoch": 1.59, + "learning_rate": 4.1322737759548834e-08, + "loss": 0.6965, + "step": 43380 + }, + { + "epoch": 1.59, + "learning_rate": 4.131907569487677e-08, + "loss": 0.6949, + "step": 43381 + }, + { + "epoch": 1.59, + "learning_rate": 4.131541363020471e-08, + "loss": 0.6795, + "step": 43382 + }, + { + "epoch": 1.59, + "learning_rate": 4.131175156553264e-08, + "loss": 0.6934, + "step": 43383 + }, + { + "epoch": 1.59, + "learning_rate": 4.130808950086058e-08, + "loss": 0.694, + "step": 43384 + }, + { + "epoch": 1.59, + "learning_rate": 4.130442743618852e-08, + "loss": 0.7076, + "step": 43385 + }, + { + "epoch": 1.59, + "learning_rate": 4.130076537151646e-08, + "loss": 0.6958, + "step": 43386 + }, + { + "epoch": 1.59, + "learning_rate": 4.12971033068444e-08, + "loss": 0.6966, + "step": 43387 + }, + { + "epoch": 1.59, + "learning_rate": 4.1293441242172334e-08, + "loss": 0.6999, + "step": 43388 + }, + { + "epoch": 1.59, + "learning_rate": 4.1289779177500274e-08, + "loss": 0.7027, + "step": 43389 + }, + { + "epoch": 1.59, + "learning_rate": 4.1286117112828214e-08, + "loss": 0.6926, + "step": 43390 + }, + { + "epoch": 1.59, + "learning_rate": 4.1282455048156154e-08, + "loss": 0.6788, + "step": 43391 + }, + { + "epoch": 1.59, + "learning_rate": 4.127879298348409e-08, + "loss": 0.6963, + "step": 43392 + }, + { + "epoch": 1.59, + "learning_rate": 4.127513091881202e-08, + "loss": 0.6864, + "step": 43393 + }, + { + "epoch": 1.59, + "learning_rate": 4.127146885413996e-08, + "loss": 0.6977, + "step": 43394 + }, + { + "epoch": 1.59, + "learning_rate": 4.12678067894679e-08, + "loss": 0.6861, + "step": 43395 + }, + { + "epoch": 1.59, + "learning_rate": 4.126414472479584e-08, + "loss": 0.6947, + "step": 43396 + }, + { + "epoch": 1.59, + "learning_rate": 4.1260482660123774e-08, + "loss": 0.6978, + "step": 43397 + }, + { + "epoch": 1.59, + "learning_rate": 4.1256820595451713e-08, + "loss": 0.6852, + "step": 43398 + }, + { + "epoch": 1.59, + "learning_rate": 4.1253158530779653e-08, + "loss": 0.7049, + "step": 43399 + }, + { + "epoch": 1.59, + "learning_rate": 4.1249496466107593e-08, + "loss": 0.697, + "step": 43400 + }, + { + "epoch": 1.59, + "learning_rate": 4.1245834401435527e-08, + "loss": 0.7112, + "step": 43401 + }, + { + "epoch": 1.59, + "learning_rate": 4.1242172336763467e-08, + "loss": 0.6896, + "step": 43402 + }, + { + "epoch": 1.59, + "learning_rate": 4.1238510272091407e-08, + "loss": 0.6951, + "step": 43403 + }, + { + "epoch": 1.59, + "learning_rate": 4.123484820741934e-08, + "loss": 0.6813, + "step": 43404 + }, + { + "epoch": 1.59, + "learning_rate": 4.123118614274728e-08, + "loss": 0.6906, + "step": 43405 + }, + { + "epoch": 1.59, + "learning_rate": 4.122752407807521e-08, + "loss": 0.6948, + "step": 43406 + }, + { + "epoch": 1.59, + "learning_rate": 4.122386201340315e-08, + "loss": 0.6991, + "step": 43407 + }, + { + "epoch": 1.59, + "learning_rate": 4.122019994873109e-08, + "loss": 0.6975, + "step": 43408 + }, + { + "epoch": 1.59, + "learning_rate": 4.121653788405903e-08, + "loss": 0.6905, + "step": 43409 + }, + { + "epoch": 1.59, + "learning_rate": 4.1212875819386966e-08, + "loss": 0.6956, + "step": 43410 + }, + { + "epoch": 1.59, + "learning_rate": 4.1209213754714906e-08, + "loss": 0.6964, + "step": 43411 + }, + { + "epoch": 1.59, + "learning_rate": 4.1205551690042846e-08, + "loss": 0.7008, + "step": 43412 + }, + { + "epoch": 1.59, + "learning_rate": 4.1201889625370786e-08, + "loss": 0.6965, + "step": 43413 + }, + { + "epoch": 1.59, + "learning_rate": 4.1198227560698726e-08, + "loss": 0.6908, + "step": 43414 + }, + { + "epoch": 1.59, + "learning_rate": 4.119456549602666e-08, + "loss": 0.691, + "step": 43415 + }, + { + "epoch": 1.59, + "learning_rate": 4.119090343135459e-08, + "loss": 0.7, + "step": 43416 + }, + { + "epoch": 1.59, + "learning_rate": 4.118724136668253e-08, + "loss": 0.6973, + "step": 43417 + }, + { + "epoch": 1.59, + "learning_rate": 4.118357930201047e-08, + "loss": 0.6958, + "step": 43418 + }, + { + "epoch": 1.59, + "learning_rate": 4.1179917237338406e-08, + "loss": 0.6895, + "step": 43419 + }, + { + "epoch": 1.59, + "learning_rate": 4.1176255172666346e-08, + "loss": 0.6956, + "step": 43420 + }, + { + "epoch": 1.59, + "learning_rate": 4.1172593107994286e-08, + "loss": 0.6992, + "step": 43421 + }, + { + "epoch": 1.59, + "learning_rate": 4.1168931043322225e-08, + "loss": 0.6895, + "step": 43422 + }, + { + "epoch": 1.59, + "learning_rate": 4.116526897865016e-08, + "loss": 0.6941, + "step": 43423 + }, + { + "epoch": 1.59, + "learning_rate": 4.11616069139781e-08, + "loss": 0.6993, + "step": 43424 + }, + { + "epoch": 1.59, + "learning_rate": 4.115794484930604e-08, + "loss": 0.6934, + "step": 43425 + }, + { + "epoch": 1.59, + "learning_rate": 4.115428278463398e-08, + "loss": 0.6977, + "step": 43426 + }, + { + "epoch": 1.59, + "learning_rate": 4.115062071996191e-08, + "loss": 0.6954, + "step": 43427 + }, + { + "epoch": 1.59, + "learning_rate": 4.1146958655289845e-08, + "loss": 0.6995, + "step": 43428 + }, + { + "epoch": 1.59, + "learning_rate": 4.1143296590617785e-08, + "loss": 0.6922, + "step": 43429 + }, + { + "epoch": 1.59, + "learning_rate": 4.1139634525945725e-08, + "loss": 0.693, + "step": 43430 + }, + { + "epoch": 1.59, + "learning_rate": 4.1135972461273665e-08, + "loss": 0.6834, + "step": 43431 + }, + { + "epoch": 1.59, + "learning_rate": 4.11323103966016e-08, + "loss": 0.6961, + "step": 43432 + }, + { + "epoch": 1.59, + "learning_rate": 4.112864833192954e-08, + "loss": 0.7001, + "step": 43433 + }, + { + "epoch": 1.59, + "learning_rate": 4.112498626725748e-08, + "loss": 0.6916, + "step": 43434 + }, + { + "epoch": 1.59, + "learning_rate": 4.112132420258542e-08, + "loss": 0.693, + "step": 43435 + }, + { + "epoch": 1.59, + "learning_rate": 4.111766213791336e-08, + "loss": 0.6942, + "step": 43436 + }, + { + "epoch": 1.59, + "learning_rate": 4.111400007324129e-08, + "loss": 0.6872, + "step": 43437 + }, + { + "epoch": 1.59, + "learning_rate": 4.111033800856923e-08, + "loss": 0.6902, + "step": 43438 + }, + { + "epoch": 1.59, + "learning_rate": 4.1106675943897165e-08, + "loss": 0.692, + "step": 43439 + }, + { + "epoch": 1.59, + "learning_rate": 4.1103013879225104e-08, + "loss": 0.6791, + "step": 43440 + }, + { + "epoch": 1.59, + "learning_rate": 4.109935181455304e-08, + "loss": 0.6975, + "step": 43441 + }, + { + "epoch": 1.59, + "learning_rate": 4.109568974988098e-08, + "loss": 0.6911, + "step": 43442 + }, + { + "epoch": 1.59, + "learning_rate": 4.109202768520892e-08, + "loss": 0.6771, + "step": 43443 + }, + { + "epoch": 1.59, + "learning_rate": 4.108836562053686e-08, + "loss": 0.6979, + "step": 43444 + }, + { + "epoch": 1.59, + "learning_rate": 4.10847035558648e-08, + "loss": 0.6828, + "step": 43445 + }, + { + "epoch": 1.59, + "learning_rate": 4.108104149119273e-08, + "loss": 0.6992, + "step": 43446 + }, + { + "epoch": 1.59, + "learning_rate": 4.107737942652067e-08, + "loss": 0.692, + "step": 43447 + }, + { + "epoch": 1.59, + "learning_rate": 4.107371736184861e-08, + "loss": 0.689, + "step": 43448 + }, + { + "epoch": 1.59, + "learning_rate": 4.107005529717655e-08, + "loss": 0.7034, + "step": 43449 + }, + { + "epoch": 1.59, + "learning_rate": 4.1066393232504484e-08, + "loss": 0.6841, + "step": 43450 + }, + { + "epoch": 1.59, + "learning_rate": 4.1062731167832424e-08, + "loss": 0.6957, + "step": 43451 + }, + { + "epoch": 1.59, + "learning_rate": 4.105906910316036e-08, + "loss": 0.7102, + "step": 43452 + }, + { + "epoch": 1.59, + "learning_rate": 4.10554070384883e-08, + "loss": 0.6922, + "step": 43453 + }, + { + "epoch": 1.59, + "learning_rate": 4.105174497381624e-08, + "loss": 0.6969, + "step": 43454 + }, + { + "epoch": 1.59, + "learning_rate": 4.104808290914417e-08, + "loss": 0.6961, + "step": 43455 + }, + { + "epoch": 1.59, + "learning_rate": 4.104442084447211e-08, + "loss": 0.6925, + "step": 43456 + }, + { + "epoch": 1.59, + "learning_rate": 4.104075877980005e-08, + "loss": 0.6983, + "step": 43457 + }, + { + "epoch": 1.59, + "learning_rate": 4.103709671512799e-08, + "loss": 0.6935, + "step": 43458 + }, + { + "epoch": 1.59, + "learning_rate": 4.1033434650455923e-08, + "loss": 0.6836, + "step": 43459 + }, + { + "epoch": 1.59, + "learning_rate": 4.1029772585783863e-08, + "loss": 0.6942, + "step": 43460 + }, + { + "epoch": 1.59, + "learning_rate": 4.10261105211118e-08, + "loss": 0.6847, + "step": 43461 + }, + { + "epoch": 1.59, + "learning_rate": 4.102244845643974e-08, + "loss": 0.6955, + "step": 43462 + }, + { + "epoch": 1.59, + "learning_rate": 4.1018786391767677e-08, + "loss": 0.701, + "step": 43463 + }, + { + "epoch": 1.59, + "learning_rate": 4.101512432709561e-08, + "loss": 0.6979, + "step": 43464 + }, + { + "epoch": 1.59, + "learning_rate": 4.101146226242355e-08, + "loss": 0.6964, + "step": 43465 + }, + { + "epoch": 1.59, + "learning_rate": 4.100780019775149e-08, + "loss": 0.6824, + "step": 43466 + }, + { + "epoch": 1.59, + "learning_rate": 4.100413813307943e-08, + "loss": 0.7029, + "step": 43467 + }, + { + "epoch": 1.59, + "learning_rate": 4.100047606840736e-08, + "loss": 0.693, + "step": 43468 + }, + { + "epoch": 1.59, + "learning_rate": 4.09968140037353e-08, + "loss": 0.6888, + "step": 43469 + }, + { + "epoch": 1.59, + "learning_rate": 4.099315193906324e-08, + "loss": 0.694, + "step": 43470 + }, + { + "epoch": 1.59, + "learning_rate": 4.098948987439118e-08, + "loss": 0.7, + "step": 43471 + }, + { + "epoch": 1.59, + "learning_rate": 4.098582780971912e-08, + "loss": 0.7019, + "step": 43472 + }, + { + "epoch": 1.59, + "learning_rate": 4.0982165745047056e-08, + "loss": 0.6833, + "step": 43473 + }, + { + "epoch": 1.59, + "learning_rate": 4.0978503680374996e-08, + "loss": 0.692, + "step": 43474 + }, + { + "epoch": 1.59, + "learning_rate": 4.097484161570293e-08, + "loss": 0.6869, + "step": 43475 + }, + { + "epoch": 1.59, + "learning_rate": 4.097117955103087e-08, + "loss": 0.692, + "step": 43476 + }, + { + "epoch": 1.59, + "learning_rate": 4.09675174863588e-08, + "loss": 0.6927, + "step": 43477 + }, + { + "epoch": 1.59, + "learning_rate": 4.096385542168674e-08, + "loss": 0.6885, + "step": 43478 + }, + { + "epoch": 1.59, + "learning_rate": 4.096019335701468e-08, + "loss": 0.6971, + "step": 43479 + }, + { + "epoch": 1.59, + "learning_rate": 4.095653129234262e-08, + "loss": 0.6898, + "step": 43480 + }, + { + "epoch": 1.59, + "learning_rate": 4.095286922767056e-08, + "loss": 0.6944, + "step": 43481 + }, + { + "epoch": 1.59, + "learning_rate": 4.0949207162998495e-08, + "loss": 0.6921, + "step": 43482 + }, + { + "epoch": 1.59, + "learning_rate": 4.0945545098326435e-08, + "loss": 0.6981, + "step": 43483 + }, + { + "epoch": 1.59, + "learning_rate": 4.0941883033654375e-08, + "loss": 0.6822, + "step": 43484 + }, + { + "epoch": 1.59, + "learning_rate": 4.0938220968982315e-08, + "loss": 0.6867, + "step": 43485 + }, + { + "epoch": 1.59, + "learning_rate": 4.093455890431025e-08, + "loss": 0.6983, + "step": 43486 + }, + { + "epoch": 1.59, + "learning_rate": 4.093089683963818e-08, + "loss": 0.6882, + "step": 43487 + }, + { + "epoch": 1.59, + "learning_rate": 4.092723477496612e-08, + "loss": 0.7006, + "step": 43488 + }, + { + "epoch": 1.59, + "learning_rate": 4.092357271029406e-08, + "loss": 0.6912, + "step": 43489 + }, + { + "epoch": 1.59, + "learning_rate": 4.0919910645622e-08, + "loss": 0.6929, + "step": 43490 + }, + { + "epoch": 1.59, + "learning_rate": 4.0916248580949935e-08, + "loss": 0.6908, + "step": 43491 + }, + { + "epoch": 1.59, + "learning_rate": 4.0912586516277875e-08, + "loss": 0.7005, + "step": 43492 + }, + { + "epoch": 1.59, + "learning_rate": 4.0908924451605815e-08, + "loss": 0.6953, + "step": 43493 + }, + { + "epoch": 1.59, + "learning_rate": 4.0905262386933755e-08, + "loss": 0.6985, + "step": 43494 + }, + { + "epoch": 1.59, + "learning_rate": 4.090160032226169e-08, + "loss": 0.7055, + "step": 43495 + }, + { + "epoch": 1.59, + "learning_rate": 4.089793825758963e-08, + "loss": 0.6977, + "step": 43496 + }, + { + "epoch": 1.59, + "learning_rate": 4.089427619291757e-08, + "loss": 0.6997, + "step": 43497 + }, + { + "epoch": 1.59, + "learning_rate": 4.08906141282455e-08, + "loss": 0.6929, + "step": 43498 + }, + { + "epoch": 1.59, + "learning_rate": 4.088695206357344e-08, + "loss": 0.7175, + "step": 43499 + }, + { + "epoch": 1.59, + "learning_rate": 4.0883289998901374e-08, + "loss": 0.6876, + "step": 43500 + }, + { + "epoch": 1.59, + "learning_rate": 4.0879627934229314e-08, + "loss": 0.6991, + "step": 43501 + }, + { + "epoch": 1.59, + "learning_rate": 4.0875965869557254e-08, + "loss": 0.6944, + "step": 43502 + }, + { + "epoch": 1.59, + "learning_rate": 4.0872303804885194e-08, + "loss": 0.695, + "step": 43503 + }, + { + "epoch": 1.59, + "learning_rate": 4.086864174021313e-08, + "loss": 0.6929, + "step": 43504 + }, + { + "epoch": 1.59, + "learning_rate": 4.086497967554107e-08, + "loss": 0.7012, + "step": 43505 + }, + { + "epoch": 1.59, + "learning_rate": 4.086131761086901e-08, + "loss": 0.6931, + "step": 43506 + }, + { + "epoch": 1.59, + "learning_rate": 4.085765554619695e-08, + "loss": 0.69, + "step": 43507 + }, + { + "epoch": 1.59, + "learning_rate": 4.085399348152489e-08, + "loss": 0.6948, + "step": 43508 + }, + { + "epoch": 1.59, + "learning_rate": 4.085033141685282e-08, + "loss": 0.6891, + "step": 43509 + }, + { + "epoch": 1.59, + "learning_rate": 4.0846669352180754e-08, + "loss": 0.692, + "step": 43510 + }, + { + "epoch": 1.59, + "learning_rate": 4.0843007287508694e-08, + "loss": 0.6935, + "step": 43511 + }, + { + "epoch": 1.59, + "learning_rate": 4.0839345222836634e-08, + "loss": 0.6892, + "step": 43512 + }, + { + "epoch": 1.59, + "learning_rate": 4.083568315816457e-08, + "loss": 0.7016, + "step": 43513 + }, + { + "epoch": 1.59, + "learning_rate": 4.083202109349251e-08, + "loss": 0.6921, + "step": 43514 + }, + { + "epoch": 1.59, + "learning_rate": 4.082835902882045e-08, + "loss": 0.7019, + "step": 43515 + }, + { + "epoch": 1.59, + "learning_rate": 4.082469696414839e-08, + "loss": 0.692, + "step": 43516 + }, + { + "epoch": 1.59, + "learning_rate": 4.082103489947633e-08, + "loss": 0.6862, + "step": 43517 + }, + { + "epoch": 1.59, + "learning_rate": 4.081737283480426e-08, + "loss": 0.7029, + "step": 43518 + }, + { + "epoch": 1.59, + "learning_rate": 4.08137107701322e-08, + "loss": 0.6804, + "step": 43519 + }, + { + "epoch": 1.59, + "learning_rate": 4.081004870546014e-08, + "loss": 0.7144, + "step": 43520 + }, + { + "epoch": 1.59, + "learning_rate": 4.080638664078807e-08, + "loss": 0.6933, + "step": 43521 + }, + { + "epoch": 1.59, + "learning_rate": 4.080272457611601e-08, + "loss": 0.6916, + "step": 43522 + }, + { + "epoch": 1.59, + "learning_rate": 4.0799062511443946e-08, + "loss": 0.686, + "step": 43523 + }, + { + "epoch": 1.59, + "learning_rate": 4.0795400446771886e-08, + "loss": 0.6833, + "step": 43524 + }, + { + "epoch": 1.59, + "learning_rate": 4.0791738382099826e-08, + "loss": 0.6902, + "step": 43525 + }, + { + "epoch": 1.59, + "learning_rate": 4.0788076317427766e-08, + "loss": 0.6966, + "step": 43526 + }, + { + "epoch": 1.59, + "learning_rate": 4.07844142527557e-08, + "loss": 0.6878, + "step": 43527 + }, + { + "epoch": 1.59, + "learning_rate": 4.078075218808364e-08, + "loss": 0.6901, + "step": 43528 + }, + { + "epoch": 1.59, + "learning_rate": 4.077709012341158e-08, + "loss": 0.7012, + "step": 43529 + }, + { + "epoch": 1.59, + "learning_rate": 4.077342805873952e-08, + "loss": 0.697, + "step": 43530 + }, + { + "epoch": 1.59, + "learning_rate": 4.076976599406745e-08, + "loss": 0.6953, + "step": 43531 + }, + { + "epoch": 1.59, + "learning_rate": 4.076610392939539e-08, + "loss": 0.6927, + "step": 43532 + }, + { + "epoch": 1.59, + "learning_rate": 4.076244186472333e-08, + "loss": 0.6974, + "step": 43533 + }, + { + "epoch": 1.59, + "learning_rate": 4.0758779800051266e-08, + "loss": 0.6782, + "step": 43534 + }, + { + "epoch": 1.59, + "learning_rate": 4.0755117735379206e-08, + "loss": 0.6892, + "step": 43535 + }, + { + "epoch": 1.59, + "learning_rate": 4.075145567070714e-08, + "loss": 0.6861, + "step": 43536 + }, + { + "epoch": 1.59, + "learning_rate": 4.074779360603508e-08, + "loss": 0.7051, + "step": 43537 + }, + { + "epoch": 1.59, + "learning_rate": 4.074413154136302e-08, + "loss": 0.6941, + "step": 43538 + }, + { + "epoch": 1.59, + "learning_rate": 4.074046947669096e-08, + "loss": 0.6836, + "step": 43539 + }, + { + "epoch": 1.59, + "learning_rate": 4.073680741201889e-08, + "loss": 0.6866, + "step": 43540 + }, + { + "epoch": 1.59, + "learning_rate": 4.073314534734683e-08, + "loss": 0.6982, + "step": 43541 + }, + { + "epoch": 1.59, + "learning_rate": 4.072948328267477e-08, + "loss": 0.688, + "step": 43542 + }, + { + "epoch": 1.59, + "learning_rate": 4.072582121800271e-08, + "loss": 0.6989, + "step": 43543 + }, + { + "epoch": 1.59, + "learning_rate": 4.0722159153330645e-08, + "loss": 0.692, + "step": 43544 + }, + { + "epoch": 1.59, + "learning_rate": 4.0718497088658585e-08, + "loss": 0.6933, + "step": 43545 + }, + { + "epoch": 1.59, + "learning_rate": 4.071483502398652e-08, + "loss": 0.6869, + "step": 43546 + }, + { + "epoch": 1.59, + "learning_rate": 4.071117295931446e-08, + "loss": 0.7097, + "step": 43547 + }, + { + "epoch": 1.59, + "learning_rate": 4.07075108946424e-08, + "loss": 0.6877, + "step": 43548 + }, + { + "epoch": 1.59, + "learning_rate": 4.070384882997033e-08, + "loss": 0.6947, + "step": 43549 + }, + { + "epoch": 1.59, + "learning_rate": 4.070018676529827e-08, + "loss": 0.6959, + "step": 43550 + }, + { + "epoch": 1.59, + "learning_rate": 4.069652470062621e-08, + "loss": 0.6988, + "step": 43551 + }, + { + "epoch": 1.59, + "learning_rate": 4.069286263595415e-08, + "loss": 0.69, + "step": 43552 + }, + { + "epoch": 1.59, + "learning_rate": 4.068920057128209e-08, + "loss": 0.7159, + "step": 43553 + }, + { + "epoch": 1.59, + "learning_rate": 4.0685538506610025e-08, + "loss": 0.6942, + "step": 43554 + }, + { + "epoch": 1.59, + "learning_rate": 4.0681876441937965e-08, + "loss": 0.7058, + "step": 43555 + }, + { + "epoch": 1.59, + "learning_rate": 4.0678214377265905e-08, + "loss": 0.6933, + "step": 43556 + }, + { + "epoch": 1.59, + "learning_rate": 4.067455231259384e-08, + "loss": 0.6963, + "step": 43557 + }, + { + "epoch": 1.59, + "learning_rate": 4.067089024792177e-08, + "loss": 0.7016, + "step": 43558 + }, + { + "epoch": 1.59, + "learning_rate": 4.066722818324971e-08, + "loss": 0.6824, + "step": 43559 + }, + { + "epoch": 1.59, + "learning_rate": 4.066356611857765e-08, + "loss": 0.6961, + "step": 43560 + }, + { + "epoch": 1.59, + "learning_rate": 4.065990405390559e-08, + "loss": 0.6945, + "step": 43561 + }, + { + "epoch": 1.59, + "learning_rate": 4.065624198923353e-08, + "loss": 0.6823, + "step": 43562 + }, + { + "epoch": 1.59, + "learning_rate": 4.0652579924561464e-08, + "loss": 0.6937, + "step": 43563 + }, + { + "epoch": 1.59, + "learning_rate": 4.0648917859889404e-08, + "loss": 0.7039, + "step": 43564 + }, + { + "epoch": 1.59, + "learning_rate": 4.0645255795217344e-08, + "loss": 0.6907, + "step": 43565 + }, + { + "epoch": 1.59, + "learning_rate": 4.0641593730545284e-08, + "loss": 0.6989, + "step": 43566 + }, + { + "epoch": 1.59, + "learning_rate": 4.063793166587322e-08, + "loss": 0.6964, + "step": 43567 + }, + { + "epoch": 1.59, + "learning_rate": 4.063426960120116e-08, + "loss": 0.6905, + "step": 43568 + }, + { + "epoch": 1.59, + "learning_rate": 4.063060753652909e-08, + "loss": 0.6981, + "step": 43569 + }, + { + "epoch": 1.59, + "learning_rate": 4.062694547185703e-08, + "loss": 0.6987, + "step": 43570 + }, + { + "epoch": 1.59, + "learning_rate": 4.062328340718497e-08, + "loss": 0.7015, + "step": 43571 + }, + { + "epoch": 1.59, + "learning_rate": 4.0619621342512904e-08, + "loss": 0.6839, + "step": 43572 + }, + { + "epoch": 1.59, + "learning_rate": 4.0615959277840844e-08, + "loss": 0.7043, + "step": 43573 + }, + { + "epoch": 1.59, + "learning_rate": 4.0612297213168784e-08, + "loss": 0.6903, + "step": 43574 + }, + { + "epoch": 1.59, + "learning_rate": 4.0608635148496724e-08, + "loss": 0.6867, + "step": 43575 + }, + { + "epoch": 1.59, + "learning_rate": 4.060497308382466e-08, + "loss": 0.6943, + "step": 43576 + }, + { + "epoch": 1.59, + "learning_rate": 4.06013110191526e-08, + "loss": 0.6835, + "step": 43577 + }, + { + "epoch": 1.59, + "learning_rate": 4.0597648954480537e-08, + "loss": 0.6908, + "step": 43578 + }, + { + "epoch": 1.59, + "learning_rate": 4.0593986889808477e-08, + "loss": 0.6971, + "step": 43579 + }, + { + "epoch": 1.59, + "learning_rate": 4.059032482513641e-08, + "loss": 0.6862, + "step": 43580 + }, + { + "epoch": 1.59, + "learning_rate": 4.058666276046434e-08, + "loss": 0.6804, + "step": 43581 + }, + { + "epoch": 1.59, + "learning_rate": 4.058300069579228e-08, + "loss": 0.6793, + "step": 43582 + }, + { + "epoch": 1.59, + "learning_rate": 4.057933863112022e-08, + "loss": 0.6954, + "step": 43583 + }, + { + "epoch": 1.59, + "learning_rate": 4.057567656644816e-08, + "loss": 0.7148, + "step": 43584 + }, + { + "epoch": 1.59, + "learning_rate": 4.0572014501776096e-08, + "loss": 0.7035, + "step": 43585 + }, + { + "epoch": 1.59, + "learning_rate": 4.0568352437104036e-08, + "loss": 0.6939, + "step": 43586 + }, + { + "epoch": 1.59, + "learning_rate": 4.0564690372431976e-08, + "loss": 0.6984, + "step": 43587 + }, + { + "epoch": 1.59, + "learning_rate": 4.0561028307759916e-08, + "loss": 0.6877, + "step": 43588 + }, + { + "epoch": 1.59, + "learning_rate": 4.0557366243087856e-08, + "loss": 0.7009, + "step": 43589 + }, + { + "epoch": 1.59, + "learning_rate": 4.055370417841579e-08, + "loss": 0.6951, + "step": 43590 + }, + { + "epoch": 1.59, + "learning_rate": 4.055004211374373e-08, + "loss": 0.6809, + "step": 43591 + }, + { + "epoch": 1.59, + "learning_rate": 4.054638004907166e-08, + "loss": 0.694, + "step": 43592 + }, + { + "epoch": 1.59, + "learning_rate": 4.05427179843996e-08, + "loss": 0.695, + "step": 43593 + }, + { + "epoch": 1.59, + "learning_rate": 4.0539055919727536e-08, + "loss": 0.6852, + "step": 43594 + }, + { + "epoch": 1.6, + "learning_rate": 4.0535393855055476e-08, + "loss": 0.7083, + "step": 43595 + }, + { + "epoch": 1.6, + "learning_rate": 4.0531731790383416e-08, + "loss": 0.6822, + "step": 43596 + }, + { + "epoch": 1.6, + "learning_rate": 4.0528069725711356e-08, + "loss": 0.6848, + "step": 43597 + }, + { + "epoch": 1.6, + "learning_rate": 4.0524407661039296e-08, + "loss": 0.6846, + "step": 43598 + }, + { + "epoch": 1.6, + "learning_rate": 4.052074559636723e-08, + "loss": 0.6891, + "step": 43599 + }, + { + "epoch": 1.6, + "learning_rate": 4.051708353169517e-08, + "loss": 0.7, + "step": 43600 + }, + { + "epoch": 1.6, + "learning_rate": 4.051342146702311e-08, + "loss": 0.6866, + "step": 43601 + }, + { + "epoch": 1.6, + "learning_rate": 4.050975940235105e-08, + "loss": 0.6905, + "step": 43602 + }, + { + "epoch": 1.6, + "learning_rate": 4.050609733767898e-08, + "loss": 0.6902, + "step": 43603 + }, + { + "epoch": 1.6, + "learning_rate": 4.0502435273006915e-08, + "loss": 0.692, + "step": 43604 + }, + { + "epoch": 1.6, + "learning_rate": 4.0498773208334855e-08, + "loss": 0.6848, + "step": 43605 + }, + { + "epoch": 1.6, + "learning_rate": 4.0495111143662795e-08, + "loss": 0.7039, + "step": 43606 + }, + { + "epoch": 1.6, + "learning_rate": 4.049144907899073e-08, + "loss": 0.6931, + "step": 43607 + }, + { + "epoch": 1.6, + "learning_rate": 4.048778701431867e-08, + "loss": 0.6854, + "step": 43608 + }, + { + "epoch": 1.6, + "learning_rate": 4.048412494964661e-08, + "loss": 0.7008, + "step": 43609 + }, + { + "epoch": 1.6, + "learning_rate": 4.048046288497455e-08, + "loss": 0.6974, + "step": 43610 + }, + { + "epoch": 1.6, + "learning_rate": 4.047680082030249e-08, + "loss": 0.6922, + "step": 43611 + }, + { + "epoch": 1.6, + "learning_rate": 4.047313875563042e-08, + "loss": 0.6816, + "step": 43612 + }, + { + "epoch": 1.6, + "learning_rate": 4.046947669095836e-08, + "loss": 0.6795, + "step": 43613 + }, + { + "epoch": 1.6, + "learning_rate": 4.04658146262863e-08, + "loss": 0.6935, + "step": 43614 + }, + { + "epoch": 1.6, + "learning_rate": 4.0462152561614235e-08, + "loss": 0.688, + "step": 43615 + }, + { + "epoch": 1.6, + "learning_rate": 4.0458490496942175e-08, + "loss": 0.6867, + "step": 43616 + }, + { + "epoch": 1.6, + "learning_rate": 4.045482843227011e-08, + "loss": 0.6929, + "step": 43617 + }, + { + "epoch": 1.6, + "learning_rate": 4.045116636759805e-08, + "loss": 0.6877, + "step": 43618 + }, + { + "epoch": 1.6, + "learning_rate": 4.044750430292599e-08, + "loss": 0.6791, + "step": 43619 + }, + { + "epoch": 1.6, + "learning_rate": 4.044384223825393e-08, + "loss": 0.6904, + "step": 43620 + }, + { + "epoch": 1.6, + "learning_rate": 4.044018017358186e-08, + "loss": 0.688, + "step": 43621 + }, + { + "epoch": 1.6, + "learning_rate": 4.04365181089098e-08, + "loss": 0.6966, + "step": 43622 + }, + { + "epoch": 1.6, + "learning_rate": 4.043285604423774e-08, + "loss": 0.6994, + "step": 43623 + }, + { + "epoch": 1.6, + "learning_rate": 4.042919397956568e-08, + "loss": 0.6945, + "step": 43624 + }, + { + "epoch": 1.6, + "learning_rate": 4.0425531914893614e-08, + "loss": 0.6993, + "step": 43625 + }, + { + "epoch": 1.6, + "learning_rate": 4.0421869850221554e-08, + "loss": 0.69, + "step": 43626 + }, + { + "epoch": 1.6, + "learning_rate": 4.0418207785549494e-08, + "loss": 0.6816, + "step": 43627 + }, + { + "epoch": 1.6, + "learning_rate": 4.041454572087743e-08, + "loss": 0.6931, + "step": 43628 + }, + { + "epoch": 1.6, + "learning_rate": 4.041088365620537e-08, + "loss": 0.6974, + "step": 43629 + }, + { + "epoch": 1.6, + "learning_rate": 4.04072215915333e-08, + "loss": 0.6887, + "step": 43630 + }, + { + "epoch": 1.6, + "learning_rate": 4.040355952686124e-08, + "loss": 0.6906, + "step": 43631 + }, + { + "epoch": 1.6, + "learning_rate": 4.039989746218918e-08, + "loss": 0.6966, + "step": 43632 + }, + { + "epoch": 1.6, + "learning_rate": 4.039623539751712e-08, + "loss": 0.7041, + "step": 43633 + }, + { + "epoch": 1.6, + "learning_rate": 4.0392573332845054e-08, + "loss": 0.6905, + "step": 43634 + }, + { + "epoch": 1.6, + "learning_rate": 4.0388911268172993e-08, + "loss": 0.6941, + "step": 43635 + }, + { + "epoch": 1.6, + "learning_rate": 4.0385249203500933e-08, + "loss": 0.6822, + "step": 43636 + }, + { + "epoch": 1.6, + "learning_rate": 4.0381587138828873e-08, + "loss": 0.6993, + "step": 43637 + }, + { + "epoch": 1.6, + "learning_rate": 4.037792507415681e-08, + "loss": 0.6864, + "step": 43638 + }, + { + "epoch": 1.6, + "learning_rate": 4.0374263009484747e-08, + "loss": 0.692, + "step": 43639 + }, + { + "epoch": 1.6, + "learning_rate": 4.037060094481268e-08, + "loss": 0.6928, + "step": 43640 + }, + { + "epoch": 1.6, + "learning_rate": 4.036693888014062e-08, + "loss": 0.6828, + "step": 43641 + }, + { + "epoch": 1.6, + "learning_rate": 4.036327681546856e-08, + "loss": 0.6794, + "step": 43642 + }, + { + "epoch": 1.6, + "learning_rate": 4.035961475079649e-08, + "loss": 0.6963, + "step": 43643 + }, + { + "epoch": 1.6, + "learning_rate": 4.035595268612443e-08, + "loss": 0.6806, + "step": 43644 + }, + { + "epoch": 1.6, + "learning_rate": 4.035229062145237e-08, + "loss": 0.6872, + "step": 43645 + }, + { + "epoch": 1.6, + "learning_rate": 4.034862855678031e-08, + "loss": 0.6908, + "step": 43646 + }, + { + "epoch": 1.6, + "learning_rate": 4.034496649210825e-08, + "loss": 0.6937, + "step": 43647 + }, + { + "epoch": 1.6, + "learning_rate": 4.0341304427436186e-08, + "loss": 0.7007, + "step": 43648 + }, + { + "epoch": 1.6, + "learning_rate": 4.0337642362764126e-08, + "loss": 0.6877, + "step": 43649 + }, + { + "epoch": 1.6, + "learning_rate": 4.0333980298092066e-08, + "loss": 0.6968, + "step": 43650 + }, + { + "epoch": 1.6, + "learning_rate": 4.033031823342e-08, + "loss": 0.6907, + "step": 43651 + }, + { + "epoch": 1.6, + "learning_rate": 4.032665616874793e-08, + "loss": 0.6916, + "step": 43652 + }, + { + "epoch": 1.6, + "learning_rate": 4.032299410407587e-08, + "loss": 0.6963, + "step": 43653 + }, + { + "epoch": 1.6, + "learning_rate": 4.031933203940381e-08, + "loss": 0.6867, + "step": 43654 + }, + { + "epoch": 1.6, + "learning_rate": 4.031566997473175e-08, + "loss": 0.6995, + "step": 43655 + }, + { + "epoch": 1.6, + "learning_rate": 4.031200791005969e-08, + "loss": 0.6954, + "step": 43656 + }, + { + "epoch": 1.6, + "learning_rate": 4.0308345845387626e-08, + "loss": 0.6916, + "step": 43657 + }, + { + "epoch": 1.6, + "learning_rate": 4.0304683780715566e-08, + "loss": 0.6907, + "step": 43658 + }, + { + "epoch": 1.6, + "learning_rate": 4.0301021716043505e-08, + "loss": 0.685, + "step": 43659 + }, + { + "epoch": 1.6, + "learning_rate": 4.0297359651371445e-08, + "loss": 0.6943, + "step": 43660 + }, + { + "epoch": 1.6, + "learning_rate": 4.029369758669938e-08, + "loss": 0.6951, + "step": 43661 + }, + { + "epoch": 1.6, + "learning_rate": 4.029003552202732e-08, + "loss": 0.6984, + "step": 43662 + }, + { + "epoch": 1.6, + "learning_rate": 4.028637345735525e-08, + "loss": 0.6953, + "step": 43663 + }, + { + "epoch": 1.6, + "learning_rate": 4.028271139268319e-08, + "loss": 0.7025, + "step": 43664 + }, + { + "epoch": 1.6, + "learning_rate": 4.027904932801113e-08, + "loss": 0.7044, + "step": 43665 + }, + { + "epoch": 1.6, + "learning_rate": 4.0275387263339065e-08, + "loss": 0.6896, + "step": 43666 + }, + { + "epoch": 1.6, + "learning_rate": 4.0271725198667005e-08, + "loss": 0.7048, + "step": 43667 + }, + { + "epoch": 1.6, + "learning_rate": 4.0268063133994945e-08, + "loss": 0.7006, + "step": 43668 + }, + { + "epoch": 1.6, + "learning_rate": 4.0264401069322885e-08, + "loss": 0.6895, + "step": 43669 + }, + { + "epoch": 1.6, + "learning_rate": 4.026073900465082e-08, + "loss": 0.6935, + "step": 43670 + }, + { + "epoch": 1.6, + "learning_rate": 4.025707693997876e-08, + "loss": 0.6961, + "step": 43671 + }, + { + "epoch": 1.6, + "learning_rate": 4.02534148753067e-08, + "loss": 0.6938, + "step": 43672 + }, + { + "epoch": 1.6, + "learning_rate": 4.024975281063464e-08, + "loss": 0.6966, + "step": 43673 + }, + { + "epoch": 1.6, + "learning_rate": 4.024609074596257e-08, + "loss": 0.6936, + "step": 43674 + }, + { + "epoch": 1.6, + "learning_rate": 4.0242428681290505e-08, + "loss": 0.6906, + "step": 43675 + }, + { + "epoch": 1.6, + "learning_rate": 4.0238766616618445e-08, + "loss": 0.6967, + "step": 43676 + }, + { + "epoch": 1.6, + "learning_rate": 4.0235104551946384e-08, + "loss": 0.6889, + "step": 43677 + }, + { + "epoch": 1.6, + "learning_rate": 4.0231442487274324e-08, + "loss": 0.6894, + "step": 43678 + }, + { + "epoch": 1.6, + "learning_rate": 4.022778042260226e-08, + "loss": 0.6812, + "step": 43679 + }, + { + "epoch": 1.6, + "learning_rate": 4.02241183579302e-08, + "loss": 0.6994, + "step": 43680 + }, + { + "epoch": 1.6, + "learning_rate": 4.022045629325814e-08, + "loss": 0.701, + "step": 43681 + }, + { + "epoch": 1.6, + "learning_rate": 4.021679422858608e-08, + "loss": 0.6932, + "step": 43682 + }, + { + "epoch": 1.6, + "learning_rate": 4.021313216391402e-08, + "loss": 0.6978, + "step": 43683 + }, + { + "epoch": 1.6, + "learning_rate": 4.020947009924195e-08, + "loss": 0.6977, + "step": 43684 + }, + { + "epoch": 1.6, + "learning_rate": 4.020580803456989e-08, + "loss": 0.6833, + "step": 43685 + }, + { + "epoch": 1.6, + "learning_rate": 4.0202145969897824e-08, + "loss": 0.7069, + "step": 43686 + }, + { + "epoch": 1.6, + "learning_rate": 4.0198483905225764e-08, + "loss": 0.6919, + "step": 43687 + }, + { + "epoch": 1.6, + "learning_rate": 4.01948218405537e-08, + "loss": 0.6894, + "step": 43688 + }, + { + "epoch": 1.6, + "learning_rate": 4.019115977588164e-08, + "loss": 0.697, + "step": 43689 + }, + { + "epoch": 1.6, + "learning_rate": 4.018749771120958e-08, + "loss": 0.6918, + "step": 43690 + }, + { + "epoch": 1.6, + "learning_rate": 4.018383564653752e-08, + "loss": 0.7039, + "step": 43691 + }, + { + "epoch": 1.6, + "learning_rate": 4.018017358186546e-08, + "loss": 0.6828, + "step": 43692 + }, + { + "epoch": 1.6, + "learning_rate": 4.017651151719339e-08, + "loss": 0.705, + "step": 43693 + }, + { + "epoch": 1.6, + "learning_rate": 4.017284945252133e-08, + "loss": 0.6887, + "step": 43694 + }, + { + "epoch": 1.6, + "learning_rate": 4.016918738784927e-08, + "loss": 0.6932, + "step": 43695 + }, + { + "epoch": 1.6, + "learning_rate": 4.016552532317721e-08, + "loss": 0.6929, + "step": 43696 + }, + { + "epoch": 1.6, + "learning_rate": 4.0161863258505143e-08, + "loss": 0.6999, + "step": 43697 + }, + { + "epoch": 1.6, + "learning_rate": 4.015820119383308e-08, + "loss": 0.6853, + "step": 43698 + }, + { + "epoch": 1.6, + "learning_rate": 4.0154539129161017e-08, + "loss": 0.6838, + "step": 43699 + }, + { + "epoch": 1.6, + "learning_rate": 4.0150877064488957e-08, + "loss": 0.6909, + "step": 43700 + }, + { + "epoch": 1.6, + "learning_rate": 4.0147214999816896e-08, + "loss": 0.6843, + "step": 43701 + }, + { + "epoch": 1.6, + "learning_rate": 4.014355293514483e-08, + "loss": 0.6946, + "step": 43702 + }, + { + "epoch": 1.6, + "learning_rate": 4.013989087047277e-08, + "loss": 0.6882, + "step": 43703 + }, + { + "epoch": 1.6, + "learning_rate": 4.013622880580071e-08, + "loss": 0.7061, + "step": 43704 + }, + { + "epoch": 1.6, + "learning_rate": 4.013256674112865e-08, + "loss": 0.6884, + "step": 43705 + }, + { + "epoch": 1.6, + "learning_rate": 4.012890467645658e-08, + "loss": 0.6861, + "step": 43706 + }, + { + "epoch": 1.6, + "learning_rate": 4.012524261178452e-08, + "loss": 0.6843, + "step": 43707 + }, + { + "epoch": 1.6, + "learning_rate": 4.012158054711246e-08, + "loss": 0.6896, + "step": 43708 + }, + { + "epoch": 1.6, + "learning_rate": 4.01179184824404e-08, + "loss": 0.7007, + "step": 43709 + }, + { + "epoch": 1.6, + "learning_rate": 4.0114256417768336e-08, + "loss": 0.6851, + "step": 43710 + }, + { + "epoch": 1.6, + "learning_rate": 4.011059435309627e-08, + "loss": 0.7043, + "step": 43711 + }, + { + "epoch": 1.6, + "learning_rate": 4.010693228842421e-08, + "loss": 0.7003, + "step": 43712 + }, + { + "epoch": 1.6, + "learning_rate": 4.010327022375215e-08, + "loss": 0.6954, + "step": 43713 + }, + { + "epoch": 1.6, + "learning_rate": 4.009960815908009e-08, + "loss": 0.6991, + "step": 43714 + }, + { + "epoch": 1.6, + "learning_rate": 4.009594609440802e-08, + "loss": 0.6939, + "step": 43715 + }, + { + "epoch": 1.6, + "learning_rate": 4.009228402973596e-08, + "loss": 0.6883, + "step": 43716 + }, + { + "epoch": 1.6, + "learning_rate": 4.00886219650639e-08, + "loss": 0.7046, + "step": 43717 + }, + { + "epoch": 1.6, + "learning_rate": 4.008495990039184e-08, + "loss": 0.6912, + "step": 43718 + }, + { + "epoch": 1.6, + "learning_rate": 4.008129783571978e-08, + "loss": 0.6897, + "step": 43719 + }, + { + "epoch": 1.6, + "learning_rate": 4.0077635771047715e-08, + "loss": 0.7004, + "step": 43720 + }, + { + "epoch": 1.6, + "learning_rate": 4.0073973706375655e-08, + "loss": 0.6986, + "step": 43721 + }, + { + "epoch": 1.6, + "learning_rate": 4.007031164170359e-08, + "loss": 0.6966, + "step": 43722 + }, + { + "epoch": 1.6, + "learning_rate": 4.006664957703153e-08, + "loss": 0.6989, + "step": 43723 + }, + { + "epoch": 1.6, + "learning_rate": 4.006298751235946e-08, + "loss": 0.6904, + "step": 43724 + }, + { + "epoch": 1.6, + "learning_rate": 4.00593254476874e-08, + "loss": 0.6903, + "step": 43725 + }, + { + "epoch": 1.6, + "learning_rate": 4.005566338301534e-08, + "loss": 0.6863, + "step": 43726 + }, + { + "epoch": 1.6, + "learning_rate": 4.005200131834328e-08, + "loss": 0.6861, + "step": 43727 + }, + { + "epoch": 1.6, + "learning_rate": 4.004833925367122e-08, + "loss": 0.6931, + "step": 43728 + }, + { + "epoch": 1.6, + "learning_rate": 4.0044677188999155e-08, + "loss": 0.6964, + "step": 43729 + }, + { + "epoch": 1.6, + "learning_rate": 4.0041015124327095e-08, + "loss": 0.6879, + "step": 43730 + }, + { + "epoch": 1.6, + "learning_rate": 4.0037353059655035e-08, + "loss": 0.6795, + "step": 43731 + }, + { + "epoch": 1.6, + "learning_rate": 4.0033690994982975e-08, + "loss": 0.6986, + "step": 43732 + }, + { + "epoch": 1.6, + "learning_rate": 4.003002893031091e-08, + "loss": 0.6976, + "step": 43733 + }, + { + "epoch": 1.6, + "learning_rate": 4.002636686563884e-08, + "loss": 0.7002, + "step": 43734 + }, + { + "epoch": 1.6, + "learning_rate": 4.002270480096678e-08, + "loss": 0.6829, + "step": 43735 + }, + { + "epoch": 1.6, + "learning_rate": 4.001904273629472e-08, + "loss": 0.6809, + "step": 43736 + }, + { + "epoch": 1.6, + "learning_rate": 4.001538067162266e-08, + "loss": 0.7021, + "step": 43737 + }, + { + "epoch": 1.6, + "learning_rate": 4.0011718606950594e-08, + "loss": 0.7021, + "step": 43738 + }, + { + "epoch": 1.6, + "learning_rate": 4.0008056542278534e-08, + "loss": 0.6881, + "step": 43739 + }, + { + "epoch": 1.6, + "learning_rate": 4.0004394477606474e-08, + "loss": 0.7027, + "step": 43740 + }, + { + "epoch": 1.6, + "learning_rate": 4.0000732412934414e-08, + "loss": 0.6845, + "step": 43741 + }, + { + "epoch": 1.6, + "learning_rate": 3.999707034826235e-08, + "loss": 0.703, + "step": 43742 + }, + { + "epoch": 1.6, + "learning_rate": 3.999340828359029e-08, + "loss": 0.6902, + "step": 43743 + }, + { + "epoch": 1.6, + "learning_rate": 3.998974621891823e-08, + "loss": 0.6936, + "step": 43744 + }, + { + "epoch": 1.6, + "learning_rate": 3.998608415424616e-08, + "loss": 0.6884, + "step": 43745 + }, + { + "epoch": 1.6, + "learning_rate": 3.99824220895741e-08, + "loss": 0.689, + "step": 43746 + }, + { + "epoch": 1.6, + "learning_rate": 3.9978760024902034e-08, + "loss": 0.6822, + "step": 43747 + }, + { + "epoch": 1.6, + "learning_rate": 3.9975097960229974e-08, + "loss": 0.7018, + "step": 43748 + }, + { + "epoch": 1.6, + "learning_rate": 3.9971435895557914e-08, + "loss": 0.6933, + "step": 43749 + }, + { + "epoch": 1.6, + "learning_rate": 3.9967773830885854e-08, + "loss": 0.6973, + "step": 43750 + }, + { + "epoch": 1.6, + "learning_rate": 3.996411176621379e-08, + "loss": 0.6888, + "step": 43751 + }, + { + "epoch": 1.6, + "learning_rate": 3.996044970154173e-08, + "loss": 0.7036, + "step": 43752 + }, + { + "epoch": 1.6, + "learning_rate": 3.995678763686967e-08, + "loss": 0.6961, + "step": 43753 + }, + { + "epoch": 1.6, + "learning_rate": 3.995312557219761e-08, + "loss": 0.6876, + "step": 43754 + }, + { + "epoch": 1.6, + "learning_rate": 3.994946350752555e-08, + "loss": 0.6943, + "step": 43755 + }, + { + "epoch": 1.6, + "learning_rate": 3.994580144285348e-08, + "loss": 0.6913, + "step": 43756 + }, + { + "epoch": 1.6, + "learning_rate": 3.9942139378181413e-08, + "loss": 0.6956, + "step": 43757 + }, + { + "epoch": 1.6, + "learning_rate": 3.993847731350935e-08, + "loss": 0.6954, + "step": 43758 + }, + { + "epoch": 1.6, + "learning_rate": 3.993481524883729e-08, + "loss": 0.6903, + "step": 43759 + }, + { + "epoch": 1.6, + "learning_rate": 3.9931153184165226e-08, + "loss": 0.6901, + "step": 43760 + }, + { + "epoch": 1.6, + "learning_rate": 3.9927491119493166e-08, + "loss": 0.6999, + "step": 43761 + }, + { + "epoch": 1.6, + "learning_rate": 3.9923829054821106e-08, + "loss": 0.6906, + "step": 43762 + }, + { + "epoch": 1.6, + "learning_rate": 3.9920166990149046e-08, + "loss": 0.6933, + "step": 43763 + }, + { + "epoch": 1.6, + "learning_rate": 3.9916504925476986e-08, + "loss": 0.6994, + "step": 43764 + }, + { + "epoch": 1.6, + "learning_rate": 3.991284286080492e-08, + "loss": 0.706, + "step": 43765 + }, + { + "epoch": 1.6, + "learning_rate": 3.990918079613286e-08, + "loss": 0.7018, + "step": 43766 + }, + { + "epoch": 1.6, + "learning_rate": 3.99055187314608e-08, + "loss": 0.6953, + "step": 43767 + }, + { + "epoch": 1.6, + "learning_rate": 3.990185666678873e-08, + "loss": 0.7026, + "step": 43768 + }, + { + "epoch": 1.6, + "learning_rate": 3.989819460211667e-08, + "loss": 0.693, + "step": 43769 + }, + { + "epoch": 1.6, + "learning_rate": 3.9894532537444606e-08, + "loss": 0.7006, + "step": 43770 + }, + { + "epoch": 1.6, + "learning_rate": 3.9890870472772546e-08, + "loss": 0.6996, + "step": 43771 + }, + { + "epoch": 1.6, + "learning_rate": 3.9887208408100486e-08, + "loss": 0.6935, + "step": 43772 + }, + { + "epoch": 1.6, + "learning_rate": 3.9883546343428426e-08, + "loss": 0.6931, + "step": 43773 + }, + { + "epoch": 1.6, + "learning_rate": 3.987988427875636e-08, + "loss": 0.6996, + "step": 43774 + }, + { + "epoch": 1.6, + "learning_rate": 3.98762222140843e-08, + "loss": 0.6962, + "step": 43775 + }, + { + "epoch": 1.6, + "learning_rate": 3.987256014941224e-08, + "loss": 0.6907, + "step": 43776 + }, + { + "epoch": 1.6, + "learning_rate": 3.986889808474018e-08, + "loss": 0.6831, + "step": 43777 + }, + { + "epoch": 1.6, + "learning_rate": 3.986523602006811e-08, + "loss": 0.6924, + "step": 43778 + }, + { + "epoch": 1.6, + "learning_rate": 3.986157395539605e-08, + "loss": 0.6926, + "step": 43779 + }, + { + "epoch": 1.6, + "learning_rate": 3.9857911890723985e-08, + "loss": 0.6982, + "step": 43780 + }, + { + "epoch": 1.6, + "learning_rate": 3.9854249826051925e-08, + "loss": 0.6944, + "step": 43781 + }, + { + "epoch": 1.6, + "learning_rate": 3.9850587761379865e-08, + "loss": 0.6878, + "step": 43782 + }, + { + "epoch": 1.6, + "learning_rate": 3.98469256967078e-08, + "loss": 0.694, + "step": 43783 + }, + { + "epoch": 1.6, + "learning_rate": 3.984326363203574e-08, + "loss": 0.6914, + "step": 43784 + }, + { + "epoch": 1.6, + "learning_rate": 3.983960156736368e-08, + "loss": 0.695, + "step": 43785 + }, + { + "epoch": 1.6, + "learning_rate": 3.983593950269162e-08, + "loss": 0.6945, + "step": 43786 + }, + { + "epoch": 1.6, + "learning_rate": 3.983227743801955e-08, + "loss": 0.7039, + "step": 43787 + }, + { + "epoch": 1.6, + "learning_rate": 3.982861537334749e-08, + "loss": 0.6812, + "step": 43788 + }, + { + "epoch": 1.6, + "learning_rate": 3.982495330867543e-08, + "loss": 0.6855, + "step": 43789 + }, + { + "epoch": 1.6, + "learning_rate": 3.982129124400337e-08, + "loss": 0.6953, + "step": 43790 + }, + { + "epoch": 1.6, + "learning_rate": 3.9817629179331305e-08, + "loss": 0.6886, + "step": 43791 + }, + { + "epoch": 1.6, + "learning_rate": 3.9813967114659245e-08, + "loss": 0.7016, + "step": 43792 + }, + { + "epoch": 1.6, + "learning_rate": 3.981030504998718e-08, + "loss": 0.6982, + "step": 43793 + }, + { + "epoch": 1.6, + "learning_rate": 3.980664298531512e-08, + "loss": 0.6984, + "step": 43794 + }, + { + "epoch": 1.6, + "learning_rate": 3.980298092064306e-08, + "loss": 0.7013, + "step": 43795 + }, + { + "epoch": 1.6, + "learning_rate": 3.979931885597099e-08, + "loss": 0.6877, + "step": 43796 + }, + { + "epoch": 1.6, + "learning_rate": 3.979565679129893e-08, + "loss": 0.7003, + "step": 43797 + }, + { + "epoch": 1.6, + "learning_rate": 3.979199472662687e-08, + "loss": 0.7042, + "step": 43798 + }, + { + "epoch": 1.6, + "learning_rate": 3.978833266195481e-08, + "loss": 0.6969, + "step": 43799 + }, + { + "epoch": 1.6, + "learning_rate": 3.9784670597282744e-08, + "loss": 0.7072, + "step": 43800 + }, + { + "epoch": 1.6, + "learning_rate": 3.9781008532610684e-08, + "loss": 0.6912, + "step": 43801 + }, + { + "epoch": 1.6, + "learning_rate": 3.9777346467938624e-08, + "loss": 0.6923, + "step": 43802 + }, + { + "epoch": 1.6, + "learning_rate": 3.9773684403266564e-08, + "loss": 0.6972, + "step": 43803 + }, + { + "epoch": 1.6, + "learning_rate": 3.97700223385945e-08, + "loss": 0.6909, + "step": 43804 + }, + { + "epoch": 1.6, + "learning_rate": 3.976636027392243e-08, + "loss": 0.6834, + "step": 43805 + }, + { + "epoch": 1.6, + "learning_rate": 3.976269820925037e-08, + "loss": 0.6889, + "step": 43806 + }, + { + "epoch": 1.6, + "learning_rate": 3.975903614457831e-08, + "loss": 0.7112, + "step": 43807 + }, + { + "epoch": 1.6, + "learning_rate": 3.975537407990625e-08, + "loss": 0.6966, + "step": 43808 + }, + { + "epoch": 1.6, + "learning_rate": 3.9751712015234184e-08, + "loss": 0.708, + "step": 43809 + }, + { + "epoch": 1.6, + "learning_rate": 3.9748049950562124e-08, + "loss": 0.694, + "step": 43810 + }, + { + "epoch": 1.6, + "learning_rate": 3.9744387885890064e-08, + "loss": 0.7, + "step": 43811 + }, + { + "epoch": 1.6, + "learning_rate": 3.9740725821218004e-08, + "loss": 0.6973, + "step": 43812 + }, + { + "epoch": 1.6, + "learning_rate": 3.9737063756545943e-08, + "loss": 0.6862, + "step": 43813 + }, + { + "epoch": 1.6, + "learning_rate": 3.973340169187388e-08, + "loss": 0.6938, + "step": 43814 + }, + { + "epoch": 1.6, + "learning_rate": 3.9729739627201817e-08, + "loss": 0.6933, + "step": 43815 + }, + { + "epoch": 1.6, + "learning_rate": 3.972607756252975e-08, + "loss": 0.6908, + "step": 43816 + }, + { + "epoch": 1.6, + "learning_rate": 3.972241549785769e-08, + "loss": 0.6973, + "step": 43817 + }, + { + "epoch": 1.6, + "learning_rate": 3.971875343318562e-08, + "loss": 0.6806, + "step": 43818 + }, + { + "epoch": 1.6, + "learning_rate": 3.971509136851356e-08, + "loss": 0.69, + "step": 43819 + }, + { + "epoch": 1.6, + "learning_rate": 3.97114293038415e-08, + "loss": 0.6878, + "step": 43820 + }, + { + "epoch": 1.6, + "learning_rate": 3.970776723916944e-08, + "loss": 0.6905, + "step": 43821 + }, + { + "epoch": 1.6, + "learning_rate": 3.970410517449738e-08, + "loss": 0.7012, + "step": 43822 + }, + { + "epoch": 1.6, + "learning_rate": 3.9700443109825316e-08, + "loss": 0.6837, + "step": 43823 + }, + { + "epoch": 1.6, + "learning_rate": 3.9696781045153256e-08, + "loss": 0.7003, + "step": 43824 + }, + { + "epoch": 1.6, + "learning_rate": 3.9693118980481196e-08, + "loss": 0.706, + "step": 43825 + }, + { + "epoch": 1.6, + "learning_rate": 3.9689456915809136e-08, + "loss": 0.6852, + "step": 43826 + }, + { + "epoch": 1.6, + "learning_rate": 3.968579485113707e-08, + "loss": 0.6914, + "step": 43827 + }, + { + "epoch": 1.6, + "learning_rate": 3.9682132786465e-08, + "loss": 0.6998, + "step": 43828 + }, + { + "epoch": 1.6, + "learning_rate": 3.967847072179294e-08, + "loss": 0.6708, + "step": 43829 + }, + { + "epoch": 1.6, + "learning_rate": 3.967480865712088e-08, + "loss": 0.6835, + "step": 43830 + }, + { + "epoch": 1.6, + "learning_rate": 3.967114659244882e-08, + "loss": 0.7019, + "step": 43831 + }, + { + "epoch": 1.6, + "learning_rate": 3.9667484527776756e-08, + "loss": 0.6923, + "step": 43832 + }, + { + "epoch": 1.6, + "learning_rate": 3.9663822463104696e-08, + "loss": 0.6988, + "step": 43833 + }, + { + "epoch": 1.6, + "learning_rate": 3.9660160398432636e-08, + "loss": 0.6997, + "step": 43834 + }, + { + "epoch": 1.6, + "learning_rate": 3.9656498333760576e-08, + "loss": 0.6844, + "step": 43835 + }, + { + "epoch": 1.6, + "learning_rate": 3.965283626908851e-08, + "loss": 0.6843, + "step": 43836 + }, + { + "epoch": 1.6, + "learning_rate": 3.964917420441645e-08, + "loss": 0.6919, + "step": 43837 + }, + { + "epoch": 1.6, + "learning_rate": 3.964551213974439e-08, + "loss": 0.6949, + "step": 43838 + }, + { + "epoch": 1.6, + "learning_rate": 3.964185007507232e-08, + "loss": 0.68, + "step": 43839 + }, + { + "epoch": 1.6, + "learning_rate": 3.963818801040026e-08, + "loss": 0.7001, + "step": 43840 + }, + { + "epoch": 1.6, + "learning_rate": 3.9634525945728195e-08, + "loss": 0.6849, + "step": 43841 + }, + { + "epoch": 1.6, + "learning_rate": 3.9630863881056135e-08, + "loss": 0.6727, + "step": 43842 + }, + { + "epoch": 1.6, + "learning_rate": 3.9627201816384075e-08, + "loss": 0.6811, + "step": 43843 + }, + { + "epoch": 1.6, + "learning_rate": 3.9623539751712015e-08, + "loss": 0.6988, + "step": 43844 + }, + { + "epoch": 1.6, + "learning_rate": 3.961987768703995e-08, + "loss": 0.695, + "step": 43845 + }, + { + "epoch": 1.6, + "learning_rate": 3.961621562236789e-08, + "loss": 0.7029, + "step": 43846 + }, + { + "epoch": 1.6, + "learning_rate": 3.961255355769583e-08, + "loss": 0.6866, + "step": 43847 + }, + { + "epoch": 1.6, + "learning_rate": 3.960889149302377e-08, + "loss": 0.6966, + "step": 43848 + }, + { + "epoch": 1.6, + "learning_rate": 3.960522942835171e-08, + "loss": 0.6894, + "step": 43849 + }, + { + "epoch": 1.6, + "learning_rate": 3.960156736367964e-08, + "loss": 0.6926, + "step": 43850 + }, + { + "epoch": 1.6, + "learning_rate": 3.9597905299007575e-08, + "loss": 0.6958, + "step": 43851 + }, + { + "epoch": 1.6, + "learning_rate": 3.9594243234335515e-08, + "loss": 0.6934, + "step": 43852 + }, + { + "epoch": 1.6, + "learning_rate": 3.9590581169663455e-08, + "loss": 0.6831, + "step": 43853 + }, + { + "epoch": 1.6, + "learning_rate": 3.958691910499139e-08, + "loss": 0.6898, + "step": 43854 + }, + { + "epoch": 1.6, + "learning_rate": 3.958325704031933e-08, + "loss": 0.7034, + "step": 43855 + }, + { + "epoch": 1.6, + "learning_rate": 3.957959497564727e-08, + "loss": 0.684, + "step": 43856 + }, + { + "epoch": 1.6, + "learning_rate": 3.957593291097521e-08, + "loss": 0.6843, + "step": 43857 + }, + { + "epoch": 1.6, + "learning_rate": 3.957227084630315e-08, + "loss": 0.6739, + "step": 43858 + }, + { + "epoch": 1.6, + "learning_rate": 3.956860878163108e-08, + "loss": 0.6955, + "step": 43859 + }, + { + "epoch": 1.6, + "learning_rate": 3.956494671695902e-08, + "loss": 0.6961, + "step": 43860 + }, + { + "epoch": 1.6, + "learning_rate": 3.956128465228696e-08, + "loss": 0.7019, + "step": 43861 + }, + { + "epoch": 1.6, + "learning_rate": 3.9557622587614894e-08, + "loss": 0.6902, + "step": 43862 + }, + { + "epoch": 1.6, + "learning_rate": 3.9553960522942834e-08, + "loss": 0.6819, + "step": 43863 + }, + { + "epoch": 1.6, + "learning_rate": 3.955029845827077e-08, + "loss": 0.6979, + "step": 43864 + }, + { + "epoch": 1.6, + "learning_rate": 3.954663639359871e-08, + "loss": 0.6921, + "step": 43865 + }, + { + "epoch": 1.6, + "learning_rate": 3.954297432892665e-08, + "loss": 0.6917, + "step": 43866 + }, + { + "epoch": 1.6, + "learning_rate": 3.953931226425459e-08, + "loss": 0.6985, + "step": 43867 + }, + { + "epoch": 1.61, + "learning_rate": 3.953565019958252e-08, + "loss": 0.7019, + "step": 43868 + }, + { + "epoch": 1.61, + "learning_rate": 3.953198813491046e-08, + "loss": 0.6812, + "step": 43869 + }, + { + "epoch": 1.61, + "learning_rate": 3.95283260702384e-08, + "loss": 0.7122, + "step": 43870 + }, + { + "epoch": 1.61, + "learning_rate": 3.952466400556634e-08, + "loss": 0.6894, + "step": 43871 + }, + { + "epoch": 1.61, + "learning_rate": 3.9521001940894274e-08, + "loss": 0.6901, + "step": 43872 + }, + { + "epoch": 1.61, + "learning_rate": 3.9517339876222213e-08, + "loss": 0.7019, + "step": 43873 + }, + { + "epoch": 1.61, + "learning_rate": 3.9513677811550153e-08, + "loss": 0.6854, + "step": 43874 + }, + { + "epoch": 1.61, + "learning_rate": 3.9510015746878087e-08, + "loss": 0.7063, + "step": 43875 + }, + { + "epoch": 1.61, + "learning_rate": 3.9506353682206027e-08, + "loss": 0.6939, + "step": 43876 + }, + { + "epoch": 1.61, + "learning_rate": 3.950269161753396e-08, + "loss": 0.6978, + "step": 43877 + }, + { + "epoch": 1.61, + "learning_rate": 3.94990295528619e-08, + "loss": 0.6934, + "step": 43878 + }, + { + "epoch": 1.61, + "learning_rate": 3.949536748818984e-08, + "loss": 0.6935, + "step": 43879 + }, + { + "epoch": 1.61, + "learning_rate": 3.949170542351778e-08, + "loss": 0.6963, + "step": 43880 + }, + { + "epoch": 1.61, + "learning_rate": 3.948804335884571e-08, + "loss": 0.6936, + "step": 43881 + }, + { + "epoch": 1.61, + "learning_rate": 3.948438129417365e-08, + "loss": 0.6914, + "step": 43882 + }, + { + "epoch": 1.61, + "learning_rate": 3.948071922950159e-08, + "loss": 0.6906, + "step": 43883 + }, + { + "epoch": 1.61, + "learning_rate": 3.947705716482953e-08, + "loss": 0.7063, + "step": 43884 + }, + { + "epoch": 1.61, + "learning_rate": 3.947339510015747e-08, + "loss": 0.697, + "step": 43885 + }, + { + "epoch": 1.61, + "learning_rate": 3.9469733035485406e-08, + "loss": 0.6989, + "step": 43886 + }, + { + "epoch": 1.61, + "learning_rate": 3.946607097081334e-08, + "loss": 0.6911, + "step": 43887 + }, + { + "epoch": 1.61, + "learning_rate": 3.946240890614128e-08, + "loss": 0.6991, + "step": 43888 + }, + { + "epoch": 1.61, + "learning_rate": 3.945874684146922e-08, + "loss": 0.6833, + "step": 43889 + }, + { + "epoch": 1.61, + "learning_rate": 3.945508477679715e-08, + "loss": 0.6908, + "step": 43890 + }, + { + "epoch": 1.61, + "learning_rate": 3.945142271212509e-08, + "loss": 0.7042, + "step": 43891 + }, + { + "epoch": 1.61, + "learning_rate": 3.944776064745303e-08, + "loss": 0.6842, + "step": 43892 + }, + { + "epoch": 1.61, + "learning_rate": 3.944409858278097e-08, + "loss": 0.6955, + "step": 43893 + }, + { + "epoch": 1.61, + "learning_rate": 3.944043651810891e-08, + "loss": 0.6912, + "step": 43894 + }, + { + "epoch": 1.61, + "learning_rate": 3.9436774453436846e-08, + "loss": 0.6975, + "step": 43895 + }, + { + "epoch": 1.61, + "learning_rate": 3.9433112388764785e-08, + "loss": 0.7, + "step": 43896 + }, + { + "epoch": 1.61, + "learning_rate": 3.9429450324092725e-08, + "loss": 0.6987, + "step": 43897 + }, + { + "epoch": 1.61, + "learning_rate": 3.942578825942066e-08, + "loss": 0.6937, + "step": 43898 + }, + { + "epoch": 1.61, + "learning_rate": 3.942212619474859e-08, + "loss": 0.6844, + "step": 43899 + }, + { + "epoch": 1.61, + "learning_rate": 3.941846413007653e-08, + "loss": 0.6892, + "step": 43900 + }, + { + "epoch": 1.61, + "learning_rate": 3.941480206540447e-08, + "loss": 0.7002, + "step": 43901 + }, + { + "epoch": 1.61, + "learning_rate": 3.941114000073241e-08, + "loss": 0.6907, + "step": 43902 + }, + { + "epoch": 1.61, + "learning_rate": 3.940747793606035e-08, + "loss": 0.7009, + "step": 43903 + }, + { + "epoch": 1.61, + "learning_rate": 3.9403815871388285e-08, + "loss": 0.6785, + "step": 43904 + }, + { + "epoch": 1.61, + "learning_rate": 3.9400153806716225e-08, + "loss": 0.6942, + "step": 43905 + }, + { + "epoch": 1.61, + "learning_rate": 3.9396491742044165e-08, + "loss": 0.6953, + "step": 43906 + }, + { + "epoch": 1.61, + "learning_rate": 3.9392829677372105e-08, + "loss": 0.6928, + "step": 43907 + }, + { + "epoch": 1.61, + "learning_rate": 3.938916761270004e-08, + "loss": 0.6821, + "step": 43908 + }, + { + "epoch": 1.61, + "learning_rate": 3.938550554802798e-08, + "loss": 0.6937, + "step": 43909 + }, + { + "epoch": 1.61, + "learning_rate": 3.938184348335591e-08, + "loss": 0.6926, + "step": 43910 + }, + { + "epoch": 1.61, + "learning_rate": 3.937818141868385e-08, + "loss": 0.6898, + "step": 43911 + }, + { + "epoch": 1.61, + "learning_rate": 3.937451935401179e-08, + "loss": 0.6913, + "step": 43912 + }, + { + "epoch": 1.61, + "learning_rate": 3.9370857289339725e-08, + "loss": 0.6939, + "step": 43913 + }, + { + "epoch": 1.61, + "learning_rate": 3.9367195224667664e-08, + "loss": 0.6953, + "step": 43914 + }, + { + "epoch": 1.61, + "learning_rate": 3.9363533159995604e-08, + "loss": 0.6812, + "step": 43915 + }, + { + "epoch": 1.61, + "learning_rate": 3.9359871095323544e-08, + "loss": 0.689, + "step": 43916 + }, + { + "epoch": 1.61, + "learning_rate": 3.935620903065148e-08, + "loss": 0.6935, + "step": 43917 + }, + { + "epoch": 1.61, + "learning_rate": 3.935254696597942e-08, + "loss": 0.6986, + "step": 43918 + }, + { + "epoch": 1.61, + "learning_rate": 3.934888490130736e-08, + "loss": 0.7012, + "step": 43919 + }, + { + "epoch": 1.61, + "learning_rate": 3.93452228366353e-08, + "loss": 0.6999, + "step": 43920 + }, + { + "epoch": 1.61, + "learning_rate": 3.934156077196323e-08, + "loss": 0.6914, + "step": 43921 + }, + { + "epoch": 1.61, + "learning_rate": 3.9337898707291164e-08, + "loss": 0.6926, + "step": 43922 + }, + { + "epoch": 1.61, + "learning_rate": 3.9334236642619104e-08, + "loss": 0.679, + "step": 43923 + }, + { + "epoch": 1.61, + "learning_rate": 3.9330574577947044e-08, + "loss": 0.6961, + "step": 43924 + }, + { + "epoch": 1.61, + "learning_rate": 3.9326912513274984e-08, + "loss": 0.6803, + "step": 43925 + }, + { + "epoch": 1.61, + "learning_rate": 3.932325044860292e-08, + "loss": 0.6905, + "step": 43926 + }, + { + "epoch": 1.61, + "learning_rate": 3.931958838393086e-08, + "loss": 0.6913, + "step": 43927 + }, + { + "epoch": 1.61, + "learning_rate": 3.93159263192588e-08, + "loss": 0.6902, + "step": 43928 + }, + { + "epoch": 1.61, + "learning_rate": 3.931226425458674e-08, + "loss": 0.6925, + "step": 43929 + }, + { + "epoch": 1.61, + "learning_rate": 3.930860218991468e-08, + "loss": 0.6888, + "step": 43930 + }, + { + "epoch": 1.61, + "learning_rate": 3.930494012524261e-08, + "loss": 0.6878, + "step": 43931 + }, + { + "epoch": 1.61, + "learning_rate": 3.930127806057055e-08, + "loss": 0.6883, + "step": 43932 + }, + { + "epoch": 1.61, + "learning_rate": 3.9297615995898483e-08, + "loss": 0.6922, + "step": 43933 + }, + { + "epoch": 1.61, + "learning_rate": 3.9293953931226423e-08, + "loss": 0.7001, + "step": 43934 + }, + { + "epoch": 1.61, + "learning_rate": 3.9290291866554357e-08, + "loss": 0.6898, + "step": 43935 + }, + { + "epoch": 1.61, + "learning_rate": 3.9286629801882297e-08, + "loss": 0.6955, + "step": 43936 + }, + { + "epoch": 1.61, + "learning_rate": 3.9282967737210237e-08, + "loss": 0.7016, + "step": 43937 + }, + { + "epoch": 1.61, + "learning_rate": 3.9279305672538176e-08, + "loss": 0.6941, + "step": 43938 + }, + { + "epoch": 1.61, + "learning_rate": 3.9275643607866116e-08, + "loss": 0.6946, + "step": 43939 + }, + { + "epoch": 1.61, + "learning_rate": 3.927198154319405e-08, + "loss": 0.6962, + "step": 43940 + }, + { + "epoch": 1.61, + "learning_rate": 3.926831947852199e-08, + "loss": 0.6962, + "step": 43941 + }, + { + "epoch": 1.61, + "learning_rate": 3.926465741384993e-08, + "loss": 0.6826, + "step": 43942 + }, + { + "epoch": 1.61, + "learning_rate": 3.926099534917787e-08, + "loss": 0.6909, + "step": 43943 + }, + { + "epoch": 1.61, + "learning_rate": 3.92573332845058e-08, + "loss": 0.6875, + "step": 43944 + }, + { + "epoch": 1.61, + "learning_rate": 3.925367121983374e-08, + "loss": 0.6975, + "step": 43945 + }, + { + "epoch": 1.61, + "learning_rate": 3.9250009155161676e-08, + "loss": 0.6939, + "step": 43946 + }, + { + "epoch": 1.61, + "learning_rate": 3.9246347090489616e-08, + "loss": 0.6957, + "step": 43947 + }, + { + "epoch": 1.61, + "learning_rate": 3.9242685025817556e-08, + "loss": 0.6971, + "step": 43948 + }, + { + "epoch": 1.61, + "learning_rate": 3.923902296114549e-08, + "loss": 0.689, + "step": 43949 + }, + { + "epoch": 1.61, + "learning_rate": 3.923536089647343e-08, + "loss": 0.6894, + "step": 43950 + }, + { + "epoch": 1.61, + "learning_rate": 3.923169883180137e-08, + "loss": 0.7003, + "step": 43951 + }, + { + "epoch": 1.61, + "learning_rate": 3.922803676712931e-08, + "loss": 0.6917, + "step": 43952 + }, + { + "epoch": 1.61, + "learning_rate": 3.922437470245724e-08, + "loss": 0.6923, + "step": 43953 + }, + { + "epoch": 1.61, + "learning_rate": 3.922071263778518e-08, + "loss": 0.6825, + "step": 43954 + }, + { + "epoch": 1.61, + "learning_rate": 3.921705057311312e-08, + "loss": 0.6932, + "step": 43955 + }, + { + "epoch": 1.61, + "learning_rate": 3.9213388508441055e-08, + "loss": 0.6925, + "step": 43956 + }, + { + "epoch": 1.61, + "learning_rate": 3.9209726443768995e-08, + "loss": 0.6886, + "step": 43957 + }, + { + "epoch": 1.61, + "learning_rate": 3.920606437909693e-08, + "loss": 0.6947, + "step": 43958 + }, + { + "epoch": 1.61, + "learning_rate": 3.920240231442487e-08, + "loss": 0.696, + "step": 43959 + }, + { + "epoch": 1.61, + "learning_rate": 3.919874024975281e-08, + "loss": 0.6914, + "step": 43960 + }, + { + "epoch": 1.61, + "learning_rate": 3.919507818508075e-08, + "loss": 0.7048, + "step": 43961 + }, + { + "epoch": 1.61, + "learning_rate": 3.919141612040868e-08, + "loss": 0.6969, + "step": 43962 + }, + { + "epoch": 1.61, + "learning_rate": 3.918775405573662e-08, + "loss": 0.6969, + "step": 43963 + }, + { + "epoch": 1.61, + "learning_rate": 3.918409199106456e-08, + "loss": 0.6763, + "step": 43964 + }, + { + "epoch": 1.61, + "learning_rate": 3.91804299263925e-08, + "loss": 0.6874, + "step": 43965 + }, + { + "epoch": 1.61, + "learning_rate": 3.917676786172044e-08, + "loss": 0.6962, + "step": 43966 + }, + { + "epoch": 1.61, + "learning_rate": 3.9173105797048375e-08, + "loss": 0.6955, + "step": 43967 + }, + { + "epoch": 1.61, + "learning_rate": 3.9169443732376315e-08, + "loss": 0.6933, + "step": 43968 + }, + { + "epoch": 1.61, + "learning_rate": 3.916578166770425e-08, + "loss": 0.6921, + "step": 43969 + }, + { + "epoch": 1.61, + "learning_rate": 3.916211960303219e-08, + "loss": 0.6995, + "step": 43970 + }, + { + "epoch": 1.61, + "learning_rate": 3.915845753836012e-08, + "loss": 0.6987, + "step": 43971 + }, + { + "epoch": 1.61, + "learning_rate": 3.915479547368806e-08, + "loss": 0.6838, + "step": 43972 + }, + { + "epoch": 1.61, + "learning_rate": 3.9151133409016e-08, + "loss": 0.7017, + "step": 43973 + }, + { + "epoch": 1.61, + "learning_rate": 3.914747134434394e-08, + "loss": 0.6973, + "step": 43974 + }, + { + "epoch": 1.61, + "learning_rate": 3.914380927967188e-08, + "loss": 0.6913, + "step": 43975 + }, + { + "epoch": 1.61, + "learning_rate": 3.9140147214999814e-08, + "loss": 0.7004, + "step": 43976 + }, + { + "epoch": 1.61, + "learning_rate": 3.9136485150327754e-08, + "loss": 0.6912, + "step": 43977 + }, + { + "epoch": 1.61, + "learning_rate": 3.9132823085655694e-08, + "loss": 0.6971, + "step": 43978 + }, + { + "epoch": 1.61, + "learning_rate": 3.9129161020983634e-08, + "loss": 0.6958, + "step": 43979 + }, + { + "epoch": 1.61, + "learning_rate": 3.912549895631157e-08, + "loss": 0.6924, + "step": 43980 + }, + { + "epoch": 1.61, + "learning_rate": 3.91218368916395e-08, + "loss": 0.694, + "step": 43981 + }, + { + "epoch": 1.61, + "learning_rate": 3.911817482696744e-08, + "loss": 0.7047, + "step": 43982 + }, + { + "epoch": 1.61, + "learning_rate": 3.911451276229538e-08, + "loss": 0.7026, + "step": 43983 + }, + { + "epoch": 1.61, + "learning_rate": 3.911085069762332e-08, + "loss": 0.6892, + "step": 43984 + }, + { + "epoch": 1.61, + "learning_rate": 3.9107188632951254e-08, + "loss": 0.6899, + "step": 43985 + }, + { + "epoch": 1.61, + "learning_rate": 3.9103526568279194e-08, + "loss": 0.69, + "step": 43986 + }, + { + "epoch": 1.61, + "learning_rate": 3.9099864503607134e-08, + "loss": 0.6933, + "step": 43987 + }, + { + "epoch": 1.61, + "learning_rate": 3.9096202438935074e-08, + "loss": 0.6929, + "step": 43988 + }, + { + "epoch": 1.61, + "learning_rate": 3.909254037426301e-08, + "loss": 0.6949, + "step": 43989 + }, + { + "epoch": 1.61, + "learning_rate": 3.908887830959095e-08, + "loss": 0.6908, + "step": 43990 + }, + { + "epoch": 1.61, + "learning_rate": 3.908521624491889e-08, + "loss": 0.6915, + "step": 43991 + }, + { + "epoch": 1.61, + "learning_rate": 3.908155418024682e-08, + "loss": 0.6983, + "step": 43992 + }, + { + "epoch": 1.61, + "learning_rate": 3.9077892115574753e-08, + "loss": 0.6908, + "step": 43993 + }, + { + "epoch": 1.61, + "learning_rate": 3.9074230050902693e-08, + "loss": 0.6988, + "step": 43994 + }, + { + "epoch": 1.61, + "learning_rate": 3.907056798623063e-08, + "loss": 0.684, + "step": 43995 + }, + { + "epoch": 1.61, + "learning_rate": 3.906690592155857e-08, + "loss": 0.6815, + "step": 43996 + }, + { + "epoch": 1.61, + "learning_rate": 3.906324385688651e-08, + "loss": 0.6968, + "step": 43997 + }, + { + "epoch": 1.61, + "learning_rate": 3.9059581792214446e-08, + "loss": 0.6953, + "step": 43998 + }, + { + "epoch": 1.61, + "learning_rate": 3.9055919727542386e-08, + "loss": 0.6906, + "step": 43999 + }, + { + "epoch": 1.61, + "learning_rate": 3.9052257662870326e-08, + "loss": 0.6845, + "step": 44000 + }, + { + "epoch": 1.61, + "learning_rate": 3.9048595598198266e-08, + "loss": 0.694, + "step": 44001 + }, + { + "epoch": 1.61, + "learning_rate": 3.90449335335262e-08, + "loss": 0.6944, + "step": 44002 + }, + { + "epoch": 1.61, + "learning_rate": 3.904127146885414e-08, + "loss": 0.6971, + "step": 44003 + }, + { + "epoch": 1.61, + "learning_rate": 3.903760940418207e-08, + "loss": 0.711, + "step": 44004 + }, + { + "epoch": 1.61, + "learning_rate": 3.903394733951001e-08, + "loss": 0.6913, + "step": 44005 + }, + { + "epoch": 1.61, + "learning_rate": 3.903028527483795e-08, + "loss": 0.6917, + "step": 44006 + }, + { + "epoch": 1.61, + "learning_rate": 3.9026623210165886e-08, + "loss": 0.6895, + "step": 44007 + }, + { + "epoch": 1.61, + "learning_rate": 3.9022961145493826e-08, + "loss": 0.694, + "step": 44008 + }, + { + "epoch": 1.61, + "learning_rate": 3.9019299080821766e-08, + "loss": 0.6909, + "step": 44009 + }, + { + "epoch": 1.61, + "learning_rate": 3.9015637016149706e-08, + "loss": 0.6885, + "step": 44010 + }, + { + "epoch": 1.61, + "learning_rate": 3.901197495147764e-08, + "loss": 0.711, + "step": 44011 + }, + { + "epoch": 1.61, + "learning_rate": 3.900831288680558e-08, + "loss": 0.6949, + "step": 44012 + }, + { + "epoch": 1.61, + "learning_rate": 3.900465082213352e-08, + "loss": 0.6849, + "step": 44013 + }, + { + "epoch": 1.61, + "learning_rate": 3.900098875746146e-08, + "loss": 0.7019, + "step": 44014 + }, + { + "epoch": 1.61, + "learning_rate": 3.899732669278939e-08, + "loss": 0.6957, + "step": 44015 + }, + { + "epoch": 1.61, + "learning_rate": 3.8993664628117325e-08, + "loss": 0.681, + "step": 44016 + }, + { + "epoch": 1.61, + "learning_rate": 3.8990002563445265e-08, + "loss": 0.683, + "step": 44017 + }, + { + "epoch": 1.61, + "learning_rate": 3.8986340498773205e-08, + "loss": 0.6885, + "step": 44018 + }, + { + "epoch": 1.61, + "learning_rate": 3.8982678434101145e-08, + "loss": 0.6753, + "step": 44019 + }, + { + "epoch": 1.61, + "learning_rate": 3.897901636942908e-08, + "loss": 0.6884, + "step": 44020 + }, + { + "epoch": 1.61, + "learning_rate": 3.897535430475702e-08, + "loss": 0.6867, + "step": 44021 + }, + { + "epoch": 1.61, + "learning_rate": 3.897169224008496e-08, + "loss": 0.688, + "step": 44022 + }, + { + "epoch": 1.61, + "learning_rate": 3.89680301754129e-08, + "loss": 0.6733, + "step": 44023 + }, + { + "epoch": 1.61, + "learning_rate": 3.896436811074084e-08, + "loss": 0.6936, + "step": 44024 + }, + { + "epoch": 1.61, + "learning_rate": 3.896070604606877e-08, + "loss": 0.6872, + "step": 44025 + }, + { + "epoch": 1.61, + "learning_rate": 3.895704398139671e-08, + "loss": 0.6984, + "step": 44026 + }, + { + "epoch": 1.61, + "learning_rate": 3.8953381916724645e-08, + "loss": 0.6881, + "step": 44027 + }, + { + "epoch": 1.61, + "learning_rate": 3.8949719852052585e-08, + "loss": 0.6919, + "step": 44028 + }, + { + "epoch": 1.61, + "learning_rate": 3.894605778738052e-08, + "loss": 0.6945, + "step": 44029 + }, + { + "epoch": 1.61, + "learning_rate": 3.894239572270846e-08, + "loss": 0.7012, + "step": 44030 + }, + { + "epoch": 1.61, + "learning_rate": 3.89387336580364e-08, + "loss": 0.7024, + "step": 44031 + }, + { + "epoch": 1.61, + "learning_rate": 3.893507159336434e-08, + "loss": 0.7058, + "step": 44032 + }, + { + "epoch": 1.61, + "learning_rate": 3.893140952869228e-08, + "loss": 0.7008, + "step": 44033 + }, + { + "epoch": 1.61, + "learning_rate": 3.892774746402021e-08, + "loss": 0.6961, + "step": 44034 + }, + { + "epoch": 1.61, + "learning_rate": 3.892408539934815e-08, + "loss": 0.6855, + "step": 44035 + }, + { + "epoch": 1.61, + "learning_rate": 3.892042333467609e-08, + "loss": 0.7024, + "step": 44036 + }, + { + "epoch": 1.61, + "learning_rate": 3.891676127000403e-08, + "loss": 0.6943, + "step": 44037 + }, + { + "epoch": 1.61, + "learning_rate": 3.8913099205331964e-08, + "loss": 0.6961, + "step": 44038 + }, + { + "epoch": 1.61, + "learning_rate": 3.8909437140659904e-08, + "loss": 0.6991, + "step": 44039 + }, + { + "epoch": 1.61, + "learning_rate": 3.890577507598784e-08, + "loss": 0.6972, + "step": 44040 + }, + { + "epoch": 1.61, + "learning_rate": 3.890211301131578e-08, + "loss": 0.6883, + "step": 44041 + }, + { + "epoch": 1.61, + "learning_rate": 3.889845094664372e-08, + "loss": 0.6992, + "step": 44042 + }, + { + "epoch": 1.61, + "learning_rate": 3.889478888197165e-08, + "loss": 0.6911, + "step": 44043 + }, + { + "epoch": 1.61, + "learning_rate": 3.889112681729959e-08, + "loss": 0.6987, + "step": 44044 + }, + { + "epoch": 1.61, + "learning_rate": 3.888746475262753e-08, + "loss": 0.6975, + "step": 44045 + }, + { + "epoch": 1.61, + "learning_rate": 3.888380268795547e-08, + "loss": 0.6945, + "step": 44046 + }, + { + "epoch": 1.61, + "learning_rate": 3.8880140623283404e-08, + "loss": 0.7008, + "step": 44047 + }, + { + "epoch": 1.61, + "learning_rate": 3.8876478558611344e-08, + "loss": 0.6994, + "step": 44048 + }, + { + "epoch": 1.61, + "learning_rate": 3.8872816493939284e-08, + "loss": 0.6899, + "step": 44049 + }, + { + "epoch": 1.61, + "learning_rate": 3.8869154429267223e-08, + "loss": 0.6968, + "step": 44050 + }, + { + "epoch": 1.61, + "learning_rate": 3.886549236459516e-08, + "loss": 0.69, + "step": 44051 + }, + { + "epoch": 1.61, + "learning_rate": 3.886183029992309e-08, + "loss": 0.6882, + "step": 44052 + }, + { + "epoch": 1.61, + "learning_rate": 3.885816823525103e-08, + "loss": 0.6871, + "step": 44053 + }, + { + "epoch": 1.61, + "learning_rate": 3.885450617057897e-08, + "loss": 0.6969, + "step": 44054 + }, + { + "epoch": 1.61, + "learning_rate": 3.885084410590691e-08, + "loss": 0.6908, + "step": 44055 + }, + { + "epoch": 1.61, + "learning_rate": 3.884718204123484e-08, + "loss": 0.6984, + "step": 44056 + }, + { + "epoch": 1.61, + "learning_rate": 3.884351997656278e-08, + "loss": 0.6938, + "step": 44057 + }, + { + "epoch": 1.61, + "learning_rate": 3.883985791189072e-08, + "loss": 0.6966, + "step": 44058 + }, + { + "epoch": 1.61, + "learning_rate": 3.883619584721866e-08, + "loss": 0.6977, + "step": 44059 + }, + { + "epoch": 1.61, + "learning_rate": 3.88325337825466e-08, + "loss": 0.6937, + "step": 44060 + }, + { + "epoch": 1.61, + "learning_rate": 3.8828871717874536e-08, + "loss": 0.6941, + "step": 44061 + }, + { + "epoch": 1.61, + "learning_rate": 3.8825209653202476e-08, + "loss": 0.6932, + "step": 44062 + }, + { + "epoch": 1.61, + "learning_rate": 3.882154758853041e-08, + "loss": 0.6951, + "step": 44063 + }, + { + "epoch": 1.61, + "learning_rate": 3.881788552385835e-08, + "loss": 0.6987, + "step": 44064 + }, + { + "epoch": 1.61, + "learning_rate": 3.881422345918628e-08, + "loss": 0.6968, + "step": 44065 + }, + { + "epoch": 1.61, + "learning_rate": 3.881056139451422e-08, + "loss": 0.6937, + "step": 44066 + }, + { + "epoch": 1.61, + "learning_rate": 3.880689932984216e-08, + "loss": 0.6952, + "step": 44067 + }, + { + "epoch": 1.61, + "learning_rate": 3.88032372651701e-08, + "loss": 0.6928, + "step": 44068 + }, + { + "epoch": 1.61, + "learning_rate": 3.879957520049804e-08, + "loss": 0.6971, + "step": 44069 + }, + { + "epoch": 1.61, + "learning_rate": 3.8795913135825976e-08, + "loss": 0.6834, + "step": 44070 + }, + { + "epoch": 1.61, + "learning_rate": 3.8792251071153916e-08, + "loss": 0.6966, + "step": 44071 + }, + { + "epoch": 1.61, + "learning_rate": 3.8788589006481856e-08, + "loss": 0.6949, + "step": 44072 + }, + { + "epoch": 1.61, + "learning_rate": 3.8784926941809795e-08, + "loss": 0.6903, + "step": 44073 + }, + { + "epoch": 1.61, + "learning_rate": 3.878126487713773e-08, + "loss": 0.6971, + "step": 44074 + }, + { + "epoch": 1.61, + "learning_rate": 3.877760281246566e-08, + "loss": 0.6971, + "step": 44075 + }, + { + "epoch": 1.61, + "learning_rate": 3.87739407477936e-08, + "loss": 0.696, + "step": 44076 + }, + { + "epoch": 1.61, + "learning_rate": 3.877027868312154e-08, + "loss": 0.7003, + "step": 44077 + }, + { + "epoch": 1.61, + "learning_rate": 3.876661661844948e-08, + "loss": 0.6937, + "step": 44078 + }, + { + "epoch": 1.61, + "learning_rate": 3.8762954553777415e-08, + "loss": 0.695, + "step": 44079 + }, + { + "epoch": 1.61, + "learning_rate": 3.8759292489105355e-08, + "loss": 0.6911, + "step": 44080 + }, + { + "epoch": 1.61, + "learning_rate": 3.8755630424433295e-08, + "loss": 0.6874, + "step": 44081 + }, + { + "epoch": 1.61, + "learning_rate": 3.8751968359761235e-08, + "loss": 0.6984, + "step": 44082 + }, + { + "epoch": 1.61, + "learning_rate": 3.874830629508917e-08, + "loss": 0.6947, + "step": 44083 + }, + { + "epoch": 1.61, + "learning_rate": 3.874464423041711e-08, + "loss": 0.6784, + "step": 44084 + }, + { + "epoch": 1.61, + "learning_rate": 3.874098216574505e-08, + "loss": 0.7053, + "step": 44085 + }, + { + "epoch": 1.61, + "learning_rate": 3.873732010107298e-08, + "loss": 0.6857, + "step": 44086 + }, + { + "epoch": 1.61, + "learning_rate": 3.873365803640092e-08, + "loss": 0.6873, + "step": 44087 + }, + { + "epoch": 1.61, + "learning_rate": 3.8729995971728855e-08, + "loss": 0.6966, + "step": 44088 + }, + { + "epoch": 1.61, + "learning_rate": 3.8726333907056795e-08, + "loss": 0.6854, + "step": 44089 + }, + { + "epoch": 1.61, + "learning_rate": 3.8722671842384735e-08, + "loss": 0.7047, + "step": 44090 + }, + { + "epoch": 1.61, + "learning_rate": 3.8719009777712675e-08, + "loss": 0.688, + "step": 44091 + }, + { + "epoch": 1.61, + "learning_rate": 3.871534771304061e-08, + "loss": 0.689, + "step": 44092 + }, + { + "epoch": 1.61, + "learning_rate": 3.871168564836855e-08, + "loss": 0.7117, + "step": 44093 + }, + { + "epoch": 1.61, + "learning_rate": 3.870802358369649e-08, + "loss": 0.6873, + "step": 44094 + }, + { + "epoch": 1.61, + "learning_rate": 3.870436151902443e-08, + "loss": 0.6977, + "step": 44095 + }, + { + "epoch": 1.61, + "learning_rate": 3.870069945435237e-08, + "loss": 0.7016, + "step": 44096 + }, + { + "epoch": 1.61, + "learning_rate": 3.86970373896803e-08, + "loss": 0.6865, + "step": 44097 + }, + { + "epoch": 1.61, + "learning_rate": 3.8693375325008234e-08, + "loss": 0.6887, + "step": 44098 + }, + { + "epoch": 1.61, + "learning_rate": 3.8689713260336174e-08, + "loss": 0.6943, + "step": 44099 + }, + { + "epoch": 1.61, + "learning_rate": 3.8686051195664114e-08, + "loss": 0.683, + "step": 44100 + }, + { + "epoch": 1.61, + "learning_rate": 3.868238913099205e-08, + "loss": 0.6864, + "step": 44101 + }, + { + "epoch": 1.61, + "learning_rate": 3.867872706631999e-08, + "loss": 0.697, + "step": 44102 + }, + { + "epoch": 1.61, + "learning_rate": 3.867506500164793e-08, + "loss": 0.7014, + "step": 44103 + }, + { + "epoch": 1.61, + "learning_rate": 3.867140293697587e-08, + "loss": 0.6891, + "step": 44104 + }, + { + "epoch": 1.61, + "learning_rate": 3.866774087230381e-08, + "loss": 0.6953, + "step": 44105 + }, + { + "epoch": 1.61, + "learning_rate": 3.866407880763174e-08, + "loss": 0.6829, + "step": 44106 + }, + { + "epoch": 1.61, + "learning_rate": 3.866041674295968e-08, + "loss": 0.6986, + "step": 44107 + }, + { + "epoch": 1.61, + "learning_rate": 3.865675467828762e-08, + "loss": 0.6934, + "step": 44108 + }, + { + "epoch": 1.61, + "learning_rate": 3.8653092613615554e-08, + "loss": 0.6927, + "step": 44109 + }, + { + "epoch": 1.61, + "learning_rate": 3.8649430548943493e-08, + "loss": 0.6877, + "step": 44110 + }, + { + "epoch": 1.61, + "learning_rate": 3.864576848427143e-08, + "loss": 0.6992, + "step": 44111 + }, + { + "epoch": 1.61, + "learning_rate": 3.8642106419599367e-08, + "loss": 0.6815, + "step": 44112 + }, + { + "epoch": 1.61, + "learning_rate": 3.8638444354927307e-08, + "loss": 0.6835, + "step": 44113 + }, + { + "epoch": 1.61, + "learning_rate": 3.8634782290255247e-08, + "loss": 0.6883, + "step": 44114 + }, + { + "epoch": 1.61, + "learning_rate": 3.863112022558318e-08, + "loss": 0.6936, + "step": 44115 + }, + { + "epoch": 1.61, + "learning_rate": 3.862745816091112e-08, + "loss": 0.6859, + "step": 44116 + }, + { + "epoch": 1.61, + "learning_rate": 3.862379609623906e-08, + "loss": 0.6946, + "step": 44117 + }, + { + "epoch": 1.61, + "learning_rate": 3.8620134031567e-08, + "loss": 0.6984, + "step": 44118 + }, + { + "epoch": 1.61, + "learning_rate": 3.861647196689493e-08, + "loss": 0.6968, + "step": 44119 + }, + { + "epoch": 1.61, + "learning_rate": 3.861280990222287e-08, + "loss": 0.6976, + "step": 44120 + }, + { + "epoch": 1.61, + "learning_rate": 3.860914783755081e-08, + "loss": 0.6905, + "step": 44121 + }, + { + "epoch": 1.61, + "learning_rate": 3.8605485772878746e-08, + "loss": 0.69, + "step": 44122 + }, + { + "epoch": 1.61, + "learning_rate": 3.8601823708206686e-08, + "loss": 0.6873, + "step": 44123 + }, + { + "epoch": 1.61, + "learning_rate": 3.859816164353462e-08, + "loss": 0.6823, + "step": 44124 + }, + { + "epoch": 1.61, + "learning_rate": 3.859449957886256e-08, + "loss": 0.702, + "step": 44125 + }, + { + "epoch": 1.61, + "learning_rate": 3.85908375141905e-08, + "loss": 0.685, + "step": 44126 + }, + { + "epoch": 1.61, + "learning_rate": 3.858717544951844e-08, + "loss": 0.6963, + "step": 44127 + }, + { + "epoch": 1.61, + "learning_rate": 3.858351338484637e-08, + "loss": 0.6949, + "step": 44128 + }, + { + "epoch": 1.61, + "learning_rate": 3.857985132017431e-08, + "loss": 0.7092, + "step": 44129 + }, + { + "epoch": 1.61, + "learning_rate": 3.857618925550225e-08, + "loss": 0.6975, + "step": 44130 + }, + { + "epoch": 1.61, + "learning_rate": 3.857252719083019e-08, + "loss": 0.6933, + "step": 44131 + }, + { + "epoch": 1.61, + "learning_rate": 3.8568865126158126e-08, + "loss": 0.6857, + "step": 44132 + }, + { + "epoch": 1.61, + "learning_rate": 3.8565203061486065e-08, + "loss": 0.6842, + "step": 44133 + }, + { + "epoch": 1.61, + "learning_rate": 3.8561540996814e-08, + "loss": 0.6961, + "step": 44134 + }, + { + "epoch": 1.61, + "learning_rate": 3.855787893214194e-08, + "loss": 0.6834, + "step": 44135 + }, + { + "epoch": 1.61, + "learning_rate": 3.855421686746988e-08, + "loss": 0.6856, + "step": 44136 + }, + { + "epoch": 1.61, + "learning_rate": 3.855055480279781e-08, + "loss": 0.6963, + "step": 44137 + }, + { + "epoch": 1.61, + "learning_rate": 3.854689273812575e-08, + "loss": 0.683, + "step": 44138 + }, + { + "epoch": 1.61, + "learning_rate": 3.854323067345369e-08, + "loss": 0.6986, + "step": 44139 + }, + { + "epoch": 1.61, + "learning_rate": 3.853956860878163e-08, + "loss": 0.6892, + "step": 44140 + }, + { + "epoch": 1.61, + "learning_rate": 3.853590654410957e-08, + "loss": 0.6925, + "step": 44141 + }, + { + "epoch": 1.62, + "learning_rate": 3.8532244479437505e-08, + "loss": 0.6962, + "step": 44142 + }, + { + "epoch": 1.62, + "learning_rate": 3.8528582414765445e-08, + "loss": 0.6903, + "step": 44143 + }, + { + "epoch": 1.62, + "learning_rate": 3.8524920350093385e-08, + "loss": 0.6856, + "step": 44144 + }, + { + "epoch": 1.62, + "learning_rate": 3.852125828542132e-08, + "loss": 0.6876, + "step": 44145 + }, + { + "epoch": 1.62, + "learning_rate": 3.851759622074925e-08, + "loss": 0.6943, + "step": 44146 + }, + { + "epoch": 1.62, + "learning_rate": 3.851393415607719e-08, + "loss": 0.7023, + "step": 44147 + }, + { + "epoch": 1.62, + "learning_rate": 3.851027209140513e-08, + "loss": 0.7072, + "step": 44148 + }, + { + "epoch": 1.62, + "learning_rate": 3.850661002673307e-08, + "loss": 0.6856, + "step": 44149 + }, + { + "epoch": 1.62, + "learning_rate": 3.850294796206101e-08, + "loss": 0.7046, + "step": 44150 + }, + { + "epoch": 1.62, + "learning_rate": 3.8499285897388944e-08, + "loss": 0.6902, + "step": 44151 + }, + { + "epoch": 1.62, + "learning_rate": 3.8495623832716884e-08, + "loss": 0.7174, + "step": 44152 + }, + { + "epoch": 1.62, + "learning_rate": 3.8491961768044824e-08, + "loss": 0.6941, + "step": 44153 + }, + { + "epoch": 1.62, + "learning_rate": 3.8488299703372764e-08, + "loss": 0.7063, + "step": 44154 + }, + { + "epoch": 1.62, + "learning_rate": 3.84846376387007e-08, + "loss": 0.6995, + "step": 44155 + }, + { + "epoch": 1.62, + "learning_rate": 3.848097557402864e-08, + "loss": 0.6963, + "step": 44156 + }, + { + "epoch": 1.62, + "learning_rate": 3.847731350935657e-08, + "loss": 0.6879, + "step": 44157 + }, + { + "epoch": 1.62, + "learning_rate": 3.847365144468451e-08, + "loss": 0.6938, + "step": 44158 + }, + { + "epoch": 1.62, + "learning_rate": 3.846998938001245e-08, + "loss": 0.693, + "step": 44159 + }, + { + "epoch": 1.62, + "learning_rate": 3.8466327315340384e-08, + "loss": 0.6944, + "step": 44160 + }, + { + "epoch": 1.62, + "learning_rate": 3.8462665250668324e-08, + "loss": 0.6901, + "step": 44161 + }, + { + "epoch": 1.62, + "learning_rate": 3.8459003185996264e-08, + "loss": 0.7008, + "step": 44162 + }, + { + "epoch": 1.62, + "learning_rate": 3.8455341121324204e-08, + "loss": 0.6932, + "step": 44163 + }, + { + "epoch": 1.62, + "learning_rate": 3.845167905665214e-08, + "loss": 0.6989, + "step": 44164 + }, + { + "epoch": 1.62, + "learning_rate": 3.844801699198008e-08, + "loss": 0.6996, + "step": 44165 + }, + { + "epoch": 1.62, + "learning_rate": 3.844435492730802e-08, + "loss": 0.6862, + "step": 44166 + }, + { + "epoch": 1.62, + "learning_rate": 3.844069286263596e-08, + "loss": 0.6989, + "step": 44167 + }, + { + "epoch": 1.62, + "learning_rate": 3.843703079796389e-08, + "loss": 0.6884, + "step": 44168 + }, + { + "epoch": 1.62, + "learning_rate": 3.8433368733291824e-08, + "loss": 0.6879, + "step": 44169 + }, + { + "epoch": 1.62, + "learning_rate": 3.8429706668619763e-08, + "loss": 0.7007, + "step": 44170 + }, + { + "epoch": 1.62, + "learning_rate": 3.8426044603947703e-08, + "loss": 0.6941, + "step": 44171 + }, + { + "epoch": 1.62, + "learning_rate": 3.842238253927564e-08, + "loss": 0.6911, + "step": 44172 + }, + { + "epoch": 1.62, + "learning_rate": 3.8418720474603577e-08, + "loss": 0.6908, + "step": 44173 + }, + { + "epoch": 1.62, + "learning_rate": 3.8415058409931517e-08, + "loss": 0.7008, + "step": 44174 + }, + { + "epoch": 1.62, + "learning_rate": 3.8411396345259456e-08, + "loss": 0.6914, + "step": 44175 + }, + { + "epoch": 1.62, + "learning_rate": 3.8407734280587396e-08, + "loss": 0.6842, + "step": 44176 + }, + { + "epoch": 1.62, + "learning_rate": 3.8404072215915336e-08, + "loss": 0.6933, + "step": 44177 + }, + { + "epoch": 1.62, + "learning_rate": 3.840041015124327e-08, + "loss": 0.6924, + "step": 44178 + }, + { + "epoch": 1.62, + "learning_rate": 3.839674808657121e-08, + "loss": 0.6843, + "step": 44179 + }, + { + "epoch": 1.62, + "learning_rate": 3.839308602189914e-08, + "loss": 0.6961, + "step": 44180 + }, + { + "epoch": 1.62, + "learning_rate": 3.838942395722708e-08, + "loss": 0.6922, + "step": 44181 + }, + { + "epoch": 1.62, + "learning_rate": 3.8385761892555016e-08, + "loss": 0.6832, + "step": 44182 + }, + { + "epoch": 1.62, + "learning_rate": 3.8382099827882956e-08, + "loss": 0.6995, + "step": 44183 + }, + { + "epoch": 1.62, + "learning_rate": 3.8378437763210896e-08, + "loss": 0.6916, + "step": 44184 + }, + { + "epoch": 1.62, + "learning_rate": 3.8374775698538836e-08, + "loss": 0.6868, + "step": 44185 + }, + { + "epoch": 1.62, + "learning_rate": 3.837111363386677e-08, + "loss": 0.6891, + "step": 44186 + }, + { + "epoch": 1.62, + "learning_rate": 3.836745156919471e-08, + "loss": 0.691, + "step": 44187 + }, + { + "epoch": 1.62, + "learning_rate": 3.836378950452265e-08, + "loss": 0.6942, + "step": 44188 + }, + { + "epoch": 1.62, + "learning_rate": 3.836012743985059e-08, + "loss": 0.6852, + "step": 44189 + }, + { + "epoch": 1.62, + "learning_rate": 3.835646537517853e-08, + "loss": 0.6877, + "step": 44190 + }, + { + "epoch": 1.62, + "learning_rate": 3.835280331050646e-08, + "loss": 0.6867, + "step": 44191 + }, + { + "epoch": 1.62, + "learning_rate": 3.8349141245834396e-08, + "loss": 0.7049, + "step": 44192 + }, + { + "epoch": 1.62, + "learning_rate": 3.8345479181162335e-08, + "loss": 0.6999, + "step": 44193 + }, + { + "epoch": 1.62, + "learning_rate": 3.8341817116490275e-08, + "loss": 0.6932, + "step": 44194 + }, + { + "epoch": 1.62, + "learning_rate": 3.833815505181821e-08, + "loss": 0.6868, + "step": 44195 + }, + { + "epoch": 1.62, + "learning_rate": 3.833449298714615e-08, + "loss": 0.6925, + "step": 44196 + }, + { + "epoch": 1.62, + "learning_rate": 3.833083092247409e-08, + "loss": 0.6855, + "step": 44197 + }, + { + "epoch": 1.62, + "learning_rate": 3.832716885780203e-08, + "loss": 0.682, + "step": 44198 + }, + { + "epoch": 1.62, + "learning_rate": 3.832350679312997e-08, + "loss": 0.6957, + "step": 44199 + }, + { + "epoch": 1.62, + "learning_rate": 3.83198447284579e-08, + "loss": 0.6912, + "step": 44200 + }, + { + "epoch": 1.62, + "learning_rate": 3.831618266378584e-08, + "loss": 0.7084, + "step": 44201 + }, + { + "epoch": 1.62, + "learning_rate": 3.831252059911378e-08, + "loss": 0.683, + "step": 44202 + }, + { + "epoch": 1.62, + "learning_rate": 3.8308858534441715e-08, + "loss": 0.6895, + "step": 44203 + }, + { + "epoch": 1.62, + "learning_rate": 3.8305196469769655e-08, + "loss": 0.703, + "step": 44204 + }, + { + "epoch": 1.62, + "learning_rate": 3.830153440509759e-08, + "loss": 0.6895, + "step": 44205 + }, + { + "epoch": 1.62, + "learning_rate": 3.829787234042553e-08, + "loss": 0.6946, + "step": 44206 + }, + { + "epoch": 1.62, + "learning_rate": 3.829421027575347e-08, + "loss": 0.6847, + "step": 44207 + }, + { + "epoch": 1.62, + "learning_rate": 3.829054821108141e-08, + "loss": 0.7024, + "step": 44208 + }, + { + "epoch": 1.62, + "learning_rate": 3.828688614640934e-08, + "loss": 0.7034, + "step": 44209 + }, + { + "epoch": 1.62, + "learning_rate": 3.828322408173728e-08, + "loss": 0.6965, + "step": 44210 + }, + { + "epoch": 1.62, + "learning_rate": 3.827956201706522e-08, + "loss": 0.6954, + "step": 44211 + }, + { + "epoch": 1.62, + "learning_rate": 3.827589995239316e-08, + "loss": 0.6984, + "step": 44212 + }, + { + "epoch": 1.62, + "learning_rate": 3.8272237887721094e-08, + "loss": 0.6975, + "step": 44213 + }, + { + "epoch": 1.62, + "learning_rate": 3.8268575823049034e-08, + "loss": 0.7101, + "step": 44214 + }, + { + "epoch": 1.62, + "learning_rate": 3.8264913758376974e-08, + "loss": 0.674, + "step": 44215 + }, + { + "epoch": 1.62, + "learning_rate": 3.826125169370491e-08, + "loss": 0.6861, + "step": 44216 + }, + { + "epoch": 1.62, + "learning_rate": 3.825758962903285e-08, + "loss": 0.7055, + "step": 44217 + }, + { + "epoch": 1.62, + "learning_rate": 3.825392756436078e-08, + "loss": 0.6841, + "step": 44218 + }, + { + "epoch": 1.62, + "learning_rate": 3.825026549968872e-08, + "loss": 0.6974, + "step": 44219 + }, + { + "epoch": 1.62, + "learning_rate": 3.824660343501666e-08, + "loss": 0.6925, + "step": 44220 + }, + { + "epoch": 1.62, + "learning_rate": 3.82429413703446e-08, + "loss": 0.6849, + "step": 44221 + }, + { + "epoch": 1.62, + "learning_rate": 3.8239279305672534e-08, + "loss": 0.687, + "step": 44222 + }, + { + "epoch": 1.62, + "learning_rate": 3.8235617241000474e-08, + "loss": 0.6983, + "step": 44223 + }, + { + "epoch": 1.62, + "learning_rate": 3.8231955176328414e-08, + "loss": 0.6919, + "step": 44224 + }, + { + "epoch": 1.62, + "learning_rate": 3.8228293111656354e-08, + "loss": 0.6876, + "step": 44225 + }, + { + "epoch": 1.62, + "learning_rate": 3.8224631046984294e-08, + "loss": 0.6874, + "step": 44226 + }, + { + "epoch": 1.62, + "learning_rate": 3.822096898231223e-08, + "loss": 0.6862, + "step": 44227 + }, + { + "epoch": 1.62, + "learning_rate": 3.821730691764016e-08, + "loss": 0.6856, + "step": 44228 + }, + { + "epoch": 1.62, + "learning_rate": 3.82136448529681e-08, + "loss": 0.6891, + "step": 44229 + }, + { + "epoch": 1.62, + "learning_rate": 3.820998278829604e-08, + "loss": 0.6904, + "step": 44230 + }, + { + "epoch": 1.62, + "learning_rate": 3.8206320723623973e-08, + "loss": 0.6907, + "step": 44231 + }, + { + "epoch": 1.62, + "learning_rate": 3.820265865895191e-08, + "loss": 0.6928, + "step": 44232 + }, + { + "epoch": 1.62, + "learning_rate": 3.819899659427985e-08, + "loss": 0.6795, + "step": 44233 + }, + { + "epoch": 1.62, + "learning_rate": 3.819533452960779e-08, + "loss": 0.6725, + "step": 44234 + }, + { + "epoch": 1.62, + "learning_rate": 3.819167246493573e-08, + "loss": 0.6898, + "step": 44235 + }, + { + "epoch": 1.62, + "learning_rate": 3.8188010400263666e-08, + "loss": 0.6919, + "step": 44236 + }, + { + "epoch": 1.62, + "learning_rate": 3.8184348335591606e-08, + "loss": 0.6942, + "step": 44237 + }, + { + "epoch": 1.62, + "learning_rate": 3.8180686270919546e-08, + "loss": 0.6875, + "step": 44238 + }, + { + "epoch": 1.62, + "learning_rate": 3.817702420624748e-08, + "loss": 0.6898, + "step": 44239 + }, + { + "epoch": 1.62, + "learning_rate": 3.817336214157541e-08, + "loss": 0.6915, + "step": 44240 + }, + { + "epoch": 1.62, + "learning_rate": 3.816970007690335e-08, + "loss": 0.6903, + "step": 44241 + }, + { + "epoch": 1.62, + "learning_rate": 3.816603801223129e-08, + "loss": 0.6867, + "step": 44242 + }, + { + "epoch": 1.62, + "learning_rate": 3.816237594755923e-08, + "loss": 0.6998, + "step": 44243 + }, + { + "epoch": 1.62, + "learning_rate": 3.815871388288717e-08, + "loss": 0.6737, + "step": 44244 + }, + { + "epoch": 1.62, + "learning_rate": 3.8155051818215106e-08, + "loss": 0.7084, + "step": 44245 + }, + { + "epoch": 1.62, + "learning_rate": 3.8151389753543046e-08, + "loss": 0.6844, + "step": 44246 + }, + { + "epoch": 1.62, + "learning_rate": 3.8147727688870986e-08, + "loss": 0.6991, + "step": 44247 + }, + { + "epoch": 1.62, + "learning_rate": 3.8144065624198926e-08, + "loss": 0.6998, + "step": 44248 + }, + { + "epoch": 1.62, + "learning_rate": 3.814040355952686e-08, + "loss": 0.693, + "step": 44249 + }, + { + "epoch": 1.62, + "learning_rate": 3.81367414948548e-08, + "loss": 0.689, + "step": 44250 + }, + { + "epoch": 1.62, + "learning_rate": 3.813307943018273e-08, + "loss": 0.7054, + "step": 44251 + }, + { + "epoch": 1.62, + "learning_rate": 3.812941736551067e-08, + "loss": 0.7025, + "step": 44252 + }, + { + "epoch": 1.62, + "learning_rate": 3.812575530083861e-08, + "loss": 0.689, + "step": 44253 + }, + { + "epoch": 1.62, + "learning_rate": 3.8122093236166545e-08, + "loss": 0.7001, + "step": 44254 + }, + { + "epoch": 1.62, + "learning_rate": 3.8118431171494485e-08, + "loss": 0.6989, + "step": 44255 + }, + { + "epoch": 1.62, + "learning_rate": 3.8114769106822425e-08, + "loss": 0.6951, + "step": 44256 + }, + { + "epoch": 1.62, + "learning_rate": 3.8111107042150365e-08, + "loss": 0.6866, + "step": 44257 + }, + { + "epoch": 1.62, + "learning_rate": 3.81074449774783e-08, + "loss": 0.6906, + "step": 44258 + }, + { + "epoch": 1.62, + "learning_rate": 3.810378291280624e-08, + "loss": 0.7077, + "step": 44259 + }, + { + "epoch": 1.62, + "learning_rate": 3.810012084813418e-08, + "loss": 0.691, + "step": 44260 + }, + { + "epoch": 1.62, + "learning_rate": 3.809645878346212e-08, + "loss": 0.6896, + "step": 44261 + }, + { + "epoch": 1.62, + "learning_rate": 3.809279671879005e-08, + "loss": 0.6862, + "step": 44262 + }, + { + "epoch": 1.62, + "learning_rate": 3.8089134654117985e-08, + "loss": 0.6895, + "step": 44263 + }, + { + "epoch": 1.62, + "learning_rate": 3.8085472589445925e-08, + "loss": 0.692, + "step": 44264 + }, + { + "epoch": 1.62, + "learning_rate": 3.8081810524773865e-08, + "loss": 0.6914, + "step": 44265 + }, + { + "epoch": 1.62, + "learning_rate": 3.8078148460101805e-08, + "loss": 0.6985, + "step": 44266 + }, + { + "epoch": 1.62, + "learning_rate": 3.807448639542974e-08, + "loss": 0.6965, + "step": 44267 + }, + { + "epoch": 1.62, + "learning_rate": 3.807082433075768e-08, + "loss": 0.6712, + "step": 44268 + }, + { + "epoch": 1.62, + "learning_rate": 3.806716226608562e-08, + "loss": 0.6933, + "step": 44269 + }, + { + "epoch": 1.62, + "learning_rate": 3.806350020141356e-08, + "loss": 0.6877, + "step": 44270 + }, + { + "epoch": 1.62, + "learning_rate": 3.80598381367415e-08, + "loss": 0.6924, + "step": 44271 + }, + { + "epoch": 1.62, + "learning_rate": 3.805617607206943e-08, + "loss": 0.684, + "step": 44272 + }, + { + "epoch": 1.62, + "learning_rate": 3.805251400739737e-08, + "loss": 0.6991, + "step": 44273 + }, + { + "epoch": 1.62, + "learning_rate": 3.8048851942725304e-08, + "loss": 0.6911, + "step": 44274 + }, + { + "epoch": 1.62, + "learning_rate": 3.8045189878053244e-08, + "loss": 0.6893, + "step": 44275 + }, + { + "epoch": 1.62, + "learning_rate": 3.804152781338118e-08, + "loss": 0.6987, + "step": 44276 + }, + { + "epoch": 1.62, + "learning_rate": 3.803786574870912e-08, + "loss": 0.6903, + "step": 44277 + }, + { + "epoch": 1.62, + "learning_rate": 3.803420368403706e-08, + "loss": 0.6825, + "step": 44278 + }, + { + "epoch": 1.62, + "learning_rate": 3.8030541619365e-08, + "loss": 0.69, + "step": 44279 + }, + { + "epoch": 1.62, + "learning_rate": 3.802687955469294e-08, + "loss": 0.6805, + "step": 44280 + }, + { + "epoch": 1.62, + "learning_rate": 3.802321749002087e-08, + "loss": 0.6939, + "step": 44281 + }, + { + "epoch": 1.62, + "learning_rate": 3.801955542534881e-08, + "loss": 0.6909, + "step": 44282 + }, + { + "epoch": 1.62, + "learning_rate": 3.801589336067675e-08, + "loss": 0.6997, + "step": 44283 + }, + { + "epoch": 1.62, + "learning_rate": 3.801223129600469e-08, + "loss": 0.6925, + "step": 44284 + }, + { + "epoch": 1.62, + "learning_rate": 3.8008569231332624e-08, + "loss": 0.7022, + "step": 44285 + }, + { + "epoch": 1.62, + "learning_rate": 3.8004907166660564e-08, + "loss": 0.6964, + "step": 44286 + }, + { + "epoch": 1.62, + "learning_rate": 3.80012451019885e-08, + "loss": 0.6969, + "step": 44287 + }, + { + "epoch": 1.62, + "learning_rate": 3.799758303731644e-08, + "loss": 0.6984, + "step": 44288 + }, + { + "epoch": 1.62, + "learning_rate": 3.799392097264438e-08, + "loss": 0.7049, + "step": 44289 + }, + { + "epoch": 1.62, + "learning_rate": 3.799025890797231e-08, + "loss": 0.696, + "step": 44290 + }, + { + "epoch": 1.62, + "learning_rate": 3.798659684330025e-08, + "loss": 0.6921, + "step": 44291 + }, + { + "epoch": 1.62, + "learning_rate": 3.798293477862819e-08, + "loss": 0.7011, + "step": 44292 + }, + { + "epoch": 1.62, + "learning_rate": 3.797927271395613e-08, + "loss": 0.6842, + "step": 44293 + }, + { + "epoch": 1.62, + "learning_rate": 3.797561064928406e-08, + "loss": 0.6852, + "step": 44294 + }, + { + "epoch": 1.62, + "learning_rate": 3.7971948584612e-08, + "loss": 0.687, + "step": 44295 + }, + { + "epoch": 1.62, + "learning_rate": 3.796828651993994e-08, + "loss": 0.6905, + "step": 44296 + }, + { + "epoch": 1.62, + "learning_rate": 3.796462445526788e-08, + "loss": 0.6796, + "step": 44297 + }, + { + "epoch": 1.62, + "learning_rate": 3.7960962390595816e-08, + "loss": 0.681, + "step": 44298 + }, + { + "epoch": 1.62, + "learning_rate": 3.795730032592375e-08, + "loss": 0.6896, + "step": 44299 + }, + { + "epoch": 1.62, + "learning_rate": 3.795363826125169e-08, + "loss": 0.6985, + "step": 44300 + }, + { + "epoch": 1.62, + "learning_rate": 3.794997619657963e-08, + "loss": 0.6917, + "step": 44301 + }, + { + "epoch": 1.62, + "learning_rate": 3.794631413190757e-08, + "loss": 0.6889, + "step": 44302 + }, + { + "epoch": 1.62, + "learning_rate": 3.79426520672355e-08, + "loss": 0.6852, + "step": 44303 + }, + { + "epoch": 1.62, + "learning_rate": 3.793899000256344e-08, + "loss": 0.6992, + "step": 44304 + }, + { + "epoch": 1.62, + "learning_rate": 3.793532793789138e-08, + "loss": 0.6919, + "step": 44305 + }, + { + "epoch": 1.62, + "learning_rate": 3.793166587321932e-08, + "loss": 0.6916, + "step": 44306 + }, + { + "epoch": 1.62, + "learning_rate": 3.792800380854726e-08, + "loss": 0.6997, + "step": 44307 + }, + { + "epoch": 1.62, + "learning_rate": 3.7924341743875196e-08, + "loss": 0.6859, + "step": 44308 + }, + { + "epoch": 1.62, + "learning_rate": 3.7920679679203136e-08, + "loss": 0.6877, + "step": 44309 + }, + { + "epoch": 1.62, + "learning_rate": 3.791701761453107e-08, + "loss": 0.7058, + "step": 44310 + }, + { + "epoch": 1.62, + "learning_rate": 3.791335554985901e-08, + "loss": 0.6836, + "step": 44311 + }, + { + "epoch": 1.62, + "learning_rate": 3.790969348518694e-08, + "loss": 0.6922, + "step": 44312 + }, + { + "epoch": 1.62, + "learning_rate": 3.790603142051488e-08, + "loss": 0.689, + "step": 44313 + }, + { + "epoch": 1.62, + "learning_rate": 3.790236935584282e-08, + "loss": 0.6976, + "step": 44314 + }, + { + "epoch": 1.62, + "learning_rate": 3.789870729117076e-08, + "loss": 0.6859, + "step": 44315 + }, + { + "epoch": 1.62, + "learning_rate": 3.78950452264987e-08, + "loss": 0.6927, + "step": 44316 + }, + { + "epoch": 1.62, + "learning_rate": 3.7891383161826635e-08, + "loss": 0.7028, + "step": 44317 + }, + { + "epoch": 1.62, + "learning_rate": 3.7887721097154575e-08, + "loss": 0.7009, + "step": 44318 + }, + { + "epoch": 1.62, + "learning_rate": 3.7884059032482515e-08, + "loss": 0.6885, + "step": 44319 + }, + { + "epoch": 1.62, + "learning_rate": 3.7880396967810455e-08, + "loss": 0.6967, + "step": 44320 + }, + { + "epoch": 1.62, + "learning_rate": 3.787673490313839e-08, + "loss": 0.6957, + "step": 44321 + }, + { + "epoch": 1.62, + "learning_rate": 3.787307283846632e-08, + "loss": 0.7007, + "step": 44322 + }, + { + "epoch": 1.62, + "learning_rate": 3.786941077379426e-08, + "loss": 0.6872, + "step": 44323 + }, + { + "epoch": 1.62, + "learning_rate": 3.78657487091222e-08, + "loss": 0.7027, + "step": 44324 + }, + { + "epoch": 1.62, + "learning_rate": 3.786208664445014e-08, + "loss": 0.6966, + "step": 44325 + }, + { + "epoch": 1.62, + "learning_rate": 3.7858424579778075e-08, + "loss": 0.6922, + "step": 44326 + }, + { + "epoch": 1.62, + "learning_rate": 3.7854762515106015e-08, + "loss": 0.6927, + "step": 44327 + }, + { + "epoch": 1.62, + "learning_rate": 3.7851100450433955e-08, + "loss": 0.694, + "step": 44328 + }, + { + "epoch": 1.62, + "learning_rate": 3.7847438385761894e-08, + "loss": 0.6897, + "step": 44329 + }, + { + "epoch": 1.62, + "learning_rate": 3.784377632108983e-08, + "loss": 0.6871, + "step": 44330 + }, + { + "epoch": 1.62, + "learning_rate": 3.784011425641777e-08, + "loss": 0.6928, + "step": 44331 + }, + { + "epoch": 1.62, + "learning_rate": 3.783645219174571e-08, + "loss": 0.6917, + "step": 44332 + }, + { + "epoch": 1.62, + "learning_rate": 3.783279012707364e-08, + "loss": 0.6862, + "step": 44333 + }, + { + "epoch": 1.62, + "learning_rate": 3.782912806240158e-08, + "loss": 0.6958, + "step": 44334 + }, + { + "epoch": 1.62, + "learning_rate": 3.7825465997729514e-08, + "loss": 0.6956, + "step": 44335 + }, + { + "epoch": 1.62, + "learning_rate": 3.7821803933057454e-08, + "loss": 0.6928, + "step": 44336 + }, + { + "epoch": 1.62, + "learning_rate": 3.7818141868385394e-08, + "loss": 0.6928, + "step": 44337 + }, + { + "epoch": 1.62, + "learning_rate": 3.7814479803713334e-08, + "loss": 0.6889, + "step": 44338 + }, + { + "epoch": 1.62, + "learning_rate": 3.781081773904127e-08, + "loss": 0.7024, + "step": 44339 + }, + { + "epoch": 1.62, + "learning_rate": 3.780715567436921e-08, + "loss": 0.7074, + "step": 44340 + }, + { + "epoch": 1.62, + "learning_rate": 3.780349360969715e-08, + "loss": 0.6957, + "step": 44341 + }, + { + "epoch": 1.62, + "learning_rate": 3.779983154502509e-08, + "loss": 0.6768, + "step": 44342 + }, + { + "epoch": 1.62, + "learning_rate": 3.779616948035303e-08, + "loss": 0.6966, + "step": 44343 + }, + { + "epoch": 1.62, + "learning_rate": 3.779250741568096e-08, + "loss": 0.6989, + "step": 44344 + }, + { + "epoch": 1.62, + "learning_rate": 3.7788845351008894e-08, + "loss": 0.6959, + "step": 44345 + }, + { + "epoch": 1.62, + "learning_rate": 3.7785183286336834e-08, + "loss": 0.6939, + "step": 44346 + }, + { + "epoch": 1.62, + "learning_rate": 3.7781521221664773e-08, + "loss": 0.6981, + "step": 44347 + }, + { + "epoch": 1.62, + "learning_rate": 3.777785915699271e-08, + "loss": 0.7017, + "step": 44348 + }, + { + "epoch": 1.62, + "learning_rate": 3.777419709232065e-08, + "loss": 0.6974, + "step": 44349 + }, + { + "epoch": 1.62, + "learning_rate": 3.7770535027648587e-08, + "loss": 0.6956, + "step": 44350 + }, + { + "epoch": 1.62, + "learning_rate": 3.7766872962976527e-08, + "loss": 0.6971, + "step": 44351 + }, + { + "epoch": 1.62, + "learning_rate": 3.7763210898304466e-08, + "loss": 0.6947, + "step": 44352 + }, + { + "epoch": 1.62, + "learning_rate": 3.77595488336324e-08, + "loss": 0.6793, + "step": 44353 + }, + { + "epoch": 1.62, + "learning_rate": 3.775588676896034e-08, + "loss": 0.6918, + "step": 44354 + }, + { + "epoch": 1.62, + "learning_rate": 3.775222470428828e-08, + "loss": 0.7044, + "step": 44355 + }, + { + "epoch": 1.62, + "learning_rate": 3.774856263961621e-08, + "loss": 0.6899, + "step": 44356 + }, + { + "epoch": 1.62, + "learning_rate": 3.774490057494415e-08, + "loss": 0.7051, + "step": 44357 + }, + { + "epoch": 1.62, + "learning_rate": 3.7741238510272086e-08, + "loss": 0.6888, + "step": 44358 + }, + { + "epoch": 1.62, + "learning_rate": 3.7737576445600026e-08, + "loss": 0.7096, + "step": 44359 + }, + { + "epoch": 1.62, + "learning_rate": 3.7733914380927966e-08, + "loss": 0.6868, + "step": 44360 + }, + { + "epoch": 1.62, + "learning_rate": 3.7730252316255906e-08, + "loss": 0.6936, + "step": 44361 + }, + { + "epoch": 1.62, + "learning_rate": 3.772659025158384e-08, + "loss": 0.697, + "step": 44362 + }, + { + "epoch": 1.62, + "learning_rate": 3.772292818691178e-08, + "loss": 0.697, + "step": 44363 + }, + { + "epoch": 1.62, + "learning_rate": 3.771926612223972e-08, + "loss": 0.7011, + "step": 44364 + }, + { + "epoch": 1.62, + "learning_rate": 3.771560405756766e-08, + "loss": 0.6883, + "step": 44365 + }, + { + "epoch": 1.62, + "learning_rate": 3.771194199289559e-08, + "loss": 0.6838, + "step": 44366 + }, + { + "epoch": 1.62, + "learning_rate": 3.770827992822353e-08, + "loss": 0.6975, + "step": 44367 + }, + { + "epoch": 1.62, + "learning_rate": 3.7704617863551466e-08, + "loss": 0.7046, + "step": 44368 + }, + { + "epoch": 1.62, + "learning_rate": 3.7700955798879406e-08, + "loss": 0.6978, + "step": 44369 + }, + { + "epoch": 1.62, + "learning_rate": 3.7697293734207345e-08, + "loss": 0.7012, + "step": 44370 + }, + { + "epoch": 1.62, + "learning_rate": 3.769363166953528e-08, + "loss": 0.6901, + "step": 44371 + }, + { + "epoch": 1.62, + "learning_rate": 3.768996960486322e-08, + "loss": 0.6918, + "step": 44372 + }, + { + "epoch": 1.62, + "learning_rate": 3.768630754019116e-08, + "loss": 0.697, + "step": 44373 + }, + { + "epoch": 1.62, + "learning_rate": 3.76826454755191e-08, + "loss": 0.7053, + "step": 44374 + }, + { + "epoch": 1.62, + "learning_rate": 3.767898341084703e-08, + "loss": 0.6793, + "step": 44375 + }, + { + "epoch": 1.62, + "learning_rate": 3.767532134617497e-08, + "loss": 0.694, + "step": 44376 + }, + { + "epoch": 1.62, + "learning_rate": 3.767165928150291e-08, + "loss": 0.7046, + "step": 44377 + }, + { + "epoch": 1.62, + "learning_rate": 3.766799721683085e-08, + "loss": 0.6942, + "step": 44378 + }, + { + "epoch": 1.62, + "learning_rate": 3.7664335152158785e-08, + "loss": 0.6941, + "step": 44379 + }, + { + "epoch": 1.62, + "learning_rate": 3.7660673087486725e-08, + "loss": 0.6935, + "step": 44380 + }, + { + "epoch": 1.62, + "learning_rate": 3.765701102281466e-08, + "loss": 0.6844, + "step": 44381 + }, + { + "epoch": 1.62, + "learning_rate": 3.76533489581426e-08, + "loss": 0.6924, + "step": 44382 + }, + { + "epoch": 1.62, + "learning_rate": 3.764968689347054e-08, + "loss": 0.6943, + "step": 44383 + }, + { + "epoch": 1.62, + "learning_rate": 3.764602482879847e-08, + "loss": 0.6842, + "step": 44384 + }, + { + "epoch": 1.62, + "learning_rate": 3.764236276412641e-08, + "loss": 0.7004, + "step": 44385 + }, + { + "epoch": 1.62, + "learning_rate": 3.763870069945435e-08, + "loss": 0.6791, + "step": 44386 + }, + { + "epoch": 1.62, + "learning_rate": 3.763503863478229e-08, + "loss": 0.6996, + "step": 44387 + }, + { + "epoch": 1.62, + "learning_rate": 3.7631376570110225e-08, + "loss": 0.7057, + "step": 44388 + }, + { + "epoch": 1.62, + "learning_rate": 3.7627714505438164e-08, + "loss": 0.7068, + "step": 44389 + }, + { + "epoch": 1.62, + "learning_rate": 3.7624052440766104e-08, + "loss": 0.702, + "step": 44390 + }, + { + "epoch": 1.62, + "learning_rate": 3.7620390376094044e-08, + "loss": 0.694, + "step": 44391 + }, + { + "epoch": 1.62, + "learning_rate": 3.761672831142198e-08, + "loss": 0.6935, + "step": 44392 + }, + { + "epoch": 1.62, + "learning_rate": 3.761306624674991e-08, + "loss": 0.6878, + "step": 44393 + }, + { + "epoch": 1.62, + "learning_rate": 3.760940418207785e-08, + "loss": 0.6979, + "step": 44394 + }, + { + "epoch": 1.62, + "learning_rate": 3.760574211740579e-08, + "loss": 0.6863, + "step": 44395 + }, + { + "epoch": 1.62, + "learning_rate": 3.760208005273373e-08, + "loss": 0.6866, + "step": 44396 + }, + { + "epoch": 1.62, + "learning_rate": 3.7598417988061664e-08, + "loss": 0.6953, + "step": 44397 + }, + { + "epoch": 1.62, + "learning_rate": 3.7594755923389604e-08, + "loss": 0.6838, + "step": 44398 + }, + { + "epoch": 1.62, + "learning_rate": 3.7591093858717544e-08, + "loss": 0.6927, + "step": 44399 + }, + { + "epoch": 1.62, + "learning_rate": 3.7587431794045484e-08, + "loss": 0.6976, + "step": 44400 + }, + { + "epoch": 1.62, + "learning_rate": 3.7583769729373424e-08, + "loss": 0.7174, + "step": 44401 + }, + { + "epoch": 1.62, + "learning_rate": 3.758010766470136e-08, + "loss": 0.6904, + "step": 44402 + }, + { + "epoch": 1.62, + "learning_rate": 3.75764456000293e-08, + "loss": 0.6953, + "step": 44403 + }, + { + "epoch": 1.62, + "learning_rate": 3.757278353535723e-08, + "loss": 0.7055, + "step": 44404 + }, + { + "epoch": 1.62, + "learning_rate": 3.756912147068517e-08, + "loss": 0.6844, + "step": 44405 + }, + { + "epoch": 1.62, + "learning_rate": 3.7565459406013104e-08, + "loss": 0.6972, + "step": 44406 + }, + { + "epoch": 1.62, + "learning_rate": 3.7561797341341043e-08, + "loss": 0.6843, + "step": 44407 + }, + { + "epoch": 1.62, + "learning_rate": 3.7558135276668983e-08, + "loss": 0.6792, + "step": 44408 + }, + { + "epoch": 1.62, + "learning_rate": 3.755447321199692e-08, + "loss": 0.6945, + "step": 44409 + }, + { + "epoch": 1.62, + "learning_rate": 3.755081114732486e-08, + "loss": 0.6894, + "step": 44410 + }, + { + "epoch": 1.62, + "learning_rate": 3.7547149082652797e-08, + "loss": 0.6992, + "step": 44411 + }, + { + "epoch": 1.62, + "learning_rate": 3.7543487017980736e-08, + "loss": 0.7038, + "step": 44412 + }, + { + "epoch": 1.62, + "learning_rate": 3.7539824953308676e-08, + "loss": 0.6861, + "step": 44413 + }, + { + "epoch": 1.62, + "learning_rate": 3.7536162888636616e-08, + "loss": 0.6956, + "step": 44414 + }, + { + "epoch": 1.63, + "learning_rate": 3.753250082396455e-08, + "loss": 0.7044, + "step": 44415 + }, + { + "epoch": 1.63, + "learning_rate": 3.752883875929248e-08, + "loss": 0.6944, + "step": 44416 + }, + { + "epoch": 1.63, + "learning_rate": 3.752517669462042e-08, + "loss": 0.6899, + "step": 44417 + }, + { + "epoch": 1.63, + "learning_rate": 3.752151462994836e-08, + "loss": 0.6977, + "step": 44418 + }, + { + "epoch": 1.63, + "learning_rate": 3.75178525652763e-08, + "loss": 0.6832, + "step": 44419 + }, + { + "epoch": 1.63, + "learning_rate": 3.7514190500604236e-08, + "loss": 0.6989, + "step": 44420 + }, + { + "epoch": 1.63, + "learning_rate": 3.7510528435932176e-08, + "loss": 0.6836, + "step": 44421 + }, + { + "epoch": 1.63, + "learning_rate": 3.7506866371260116e-08, + "loss": 0.6983, + "step": 44422 + }, + { + "epoch": 1.63, + "learning_rate": 3.7503204306588056e-08, + "loss": 0.6992, + "step": 44423 + }, + { + "epoch": 1.63, + "learning_rate": 3.749954224191599e-08, + "loss": 0.7008, + "step": 44424 + }, + { + "epoch": 1.63, + "learning_rate": 3.749588017724393e-08, + "loss": 0.692, + "step": 44425 + }, + { + "epoch": 1.63, + "learning_rate": 3.749221811257187e-08, + "loss": 0.6885, + "step": 44426 + }, + { + "epoch": 1.63, + "learning_rate": 3.74885560478998e-08, + "loss": 0.6851, + "step": 44427 + }, + { + "epoch": 1.63, + "learning_rate": 3.748489398322774e-08, + "loss": 0.6925, + "step": 44428 + }, + { + "epoch": 1.63, + "learning_rate": 3.7481231918555676e-08, + "loss": 0.6892, + "step": 44429 + }, + { + "epoch": 1.63, + "learning_rate": 3.7477569853883615e-08, + "loss": 0.6895, + "step": 44430 + }, + { + "epoch": 1.63, + "learning_rate": 3.7473907789211555e-08, + "loss": 0.695, + "step": 44431 + }, + { + "epoch": 1.63, + "learning_rate": 3.7470245724539495e-08, + "loss": 0.6952, + "step": 44432 + }, + { + "epoch": 1.63, + "learning_rate": 3.746658365986743e-08, + "loss": 0.6838, + "step": 44433 + }, + { + "epoch": 1.63, + "learning_rate": 3.746292159519537e-08, + "loss": 0.6974, + "step": 44434 + }, + { + "epoch": 1.63, + "learning_rate": 3.745925953052331e-08, + "loss": 0.703, + "step": 44435 + }, + { + "epoch": 1.63, + "learning_rate": 3.745559746585125e-08, + "loss": 0.6983, + "step": 44436 + }, + { + "epoch": 1.63, + "learning_rate": 3.745193540117919e-08, + "loss": 0.7076, + "step": 44437 + }, + { + "epoch": 1.63, + "learning_rate": 3.744827333650712e-08, + "loss": 0.6914, + "step": 44438 + }, + { + "epoch": 1.63, + "learning_rate": 3.7444611271835055e-08, + "loss": 0.6951, + "step": 44439 + }, + { + "epoch": 1.63, + "learning_rate": 3.7440949207162995e-08, + "loss": 0.6922, + "step": 44440 + }, + { + "epoch": 1.63, + "learning_rate": 3.7437287142490935e-08, + "loss": 0.7014, + "step": 44441 + }, + { + "epoch": 1.63, + "learning_rate": 3.743362507781887e-08, + "loss": 0.6957, + "step": 44442 + }, + { + "epoch": 1.63, + "learning_rate": 3.742996301314681e-08, + "loss": 0.6902, + "step": 44443 + }, + { + "epoch": 1.63, + "learning_rate": 3.742630094847475e-08, + "loss": 0.7, + "step": 44444 + }, + { + "epoch": 1.63, + "learning_rate": 3.742263888380269e-08, + "loss": 0.6921, + "step": 44445 + }, + { + "epoch": 1.63, + "learning_rate": 3.741897681913063e-08, + "loss": 0.6959, + "step": 44446 + }, + { + "epoch": 1.63, + "learning_rate": 3.741531475445856e-08, + "loss": 0.6842, + "step": 44447 + }, + { + "epoch": 1.63, + "learning_rate": 3.74116526897865e-08, + "loss": 0.6875, + "step": 44448 + }, + { + "epoch": 1.63, + "learning_rate": 3.740799062511444e-08, + "loss": 0.6878, + "step": 44449 + }, + { + "epoch": 1.63, + "learning_rate": 3.7404328560442374e-08, + "loss": 0.6886, + "step": 44450 + }, + { + "epoch": 1.63, + "learning_rate": 3.7400666495770314e-08, + "loss": 0.6962, + "step": 44451 + }, + { + "epoch": 1.63, + "learning_rate": 3.739700443109825e-08, + "loss": 0.6861, + "step": 44452 + }, + { + "epoch": 1.63, + "learning_rate": 3.739334236642619e-08, + "loss": 0.7065, + "step": 44453 + }, + { + "epoch": 1.63, + "learning_rate": 3.738968030175413e-08, + "loss": 0.6851, + "step": 44454 + }, + { + "epoch": 1.63, + "learning_rate": 3.738601823708207e-08, + "loss": 0.6934, + "step": 44455 + }, + { + "epoch": 1.63, + "learning_rate": 3.738235617241e-08, + "loss": 0.7019, + "step": 44456 + }, + { + "epoch": 1.63, + "learning_rate": 3.737869410773794e-08, + "loss": 0.7053, + "step": 44457 + }, + { + "epoch": 1.63, + "learning_rate": 3.737503204306588e-08, + "loss": 0.6976, + "step": 44458 + }, + { + "epoch": 1.63, + "learning_rate": 3.737136997839382e-08, + "loss": 0.6969, + "step": 44459 + }, + { + "epoch": 1.63, + "learning_rate": 3.7367707913721754e-08, + "loss": 0.6915, + "step": 44460 + }, + { + "epoch": 1.63, + "learning_rate": 3.7364045849049694e-08, + "loss": 0.6925, + "step": 44461 + }, + { + "epoch": 1.63, + "learning_rate": 3.7360383784377634e-08, + "loss": 0.6844, + "step": 44462 + }, + { + "epoch": 1.63, + "learning_rate": 3.735672171970557e-08, + "loss": 0.6991, + "step": 44463 + }, + { + "epoch": 1.63, + "learning_rate": 3.735305965503351e-08, + "loss": 0.6951, + "step": 44464 + }, + { + "epoch": 1.63, + "learning_rate": 3.734939759036144e-08, + "loss": 0.6902, + "step": 44465 + }, + { + "epoch": 1.63, + "learning_rate": 3.734573552568938e-08, + "loss": 0.6962, + "step": 44466 + }, + { + "epoch": 1.63, + "learning_rate": 3.734207346101732e-08, + "loss": 0.7005, + "step": 44467 + }, + { + "epoch": 1.63, + "learning_rate": 3.733841139634526e-08, + "loss": 0.6965, + "step": 44468 + }, + { + "epoch": 1.63, + "learning_rate": 3.733474933167319e-08, + "loss": 0.6947, + "step": 44469 + }, + { + "epoch": 1.63, + "learning_rate": 3.733108726700113e-08, + "loss": 0.6935, + "step": 44470 + }, + { + "epoch": 1.63, + "learning_rate": 3.732742520232907e-08, + "loss": 0.6913, + "step": 44471 + }, + { + "epoch": 1.63, + "learning_rate": 3.732376313765701e-08, + "loss": 0.6947, + "step": 44472 + }, + { + "epoch": 1.63, + "learning_rate": 3.732010107298495e-08, + "loss": 0.6898, + "step": 44473 + }, + { + "epoch": 1.63, + "learning_rate": 3.7316439008312886e-08, + "loss": 0.695, + "step": 44474 + }, + { + "epoch": 1.63, + "learning_rate": 3.731277694364082e-08, + "loss": 0.6943, + "step": 44475 + }, + { + "epoch": 1.63, + "learning_rate": 3.730911487896876e-08, + "loss": 0.6832, + "step": 44476 + }, + { + "epoch": 1.63, + "learning_rate": 3.73054528142967e-08, + "loss": 0.7037, + "step": 44477 + }, + { + "epoch": 1.63, + "learning_rate": 3.730179074962463e-08, + "loss": 0.6992, + "step": 44478 + }, + { + "epoch": 1.63, + "learning_rate": 3.729812868495257e-08, + "loss": 0.7001, + "step": 44479 + }, + { + "epoch": 1.63, + "learning_rate": 3.729446662028051e-08, + "loss": 0.6966, + "step": 44480 + }, + { + "epoch": 1.63, + "learning_rate": 3.729080455560845e-08, + "loss": 0.6946, + "step": 44481 + }, + { + "epoch": 1.63, + "learning_rate": 3.728714249093639e-08, + "loss": 0.6976, + "step": 44482 + }, + { + "epoch": 1.63, + "learning_rate": 3.7283480426264326e-08, + "loss": 0.6842, + "step": 44483 + }, + { + "epoch": 1.63, + "learning_rate": 3.7279818361592266e-08, + "loss": 0.6882, + "step": 44484 + }, + { + "epoch": 1.63, + "learning_rate": 3.7276156296920206e-08, + "loss": 0.6906, + "step": 44485 + }, + { + "epoch": 1.63, + "learning_rate": 3.727249423224814e-08, + "loss": 0.6904, + "step": 44486 + }, + { + "epoch": 1.63, + "learning_rate": 3.726883216757607e-08, + "loss": 0.6824, + "step": 44487 + }, + { + "epoch": 1.63, + "learning_rate": 3.726517010290401e-08, + "loss": 0.6941, + "step": 44488 + }, + { + "epoch": 1.63, + "learning_rate": 3.726150803823195e-08, + "loss": 0.6923, + "step": 44489 + }, + { + "epoch": 1.63, + "learning_rate": 3.725784597355989e-08, + "loss": 0.6939, + "step": 44490 + }, + { + "epoch": 1.63, + "learning_rate": 3.725418390888783e-08, + "loss": 0.6821, + "step": 44491 + }, + { + "epoch": 1.63, + "learning_rate": 3.7250521844215765e-08, + "loss": 0.6949, + "step": 44492 + }, + { + "epoch": 1.63, + "learning_rate": 3.7246859779543705e-08, + "loss": 0.6912, + "step": 44493 + }, + { + "epoch": 1.63, + "learning_rate": 3.7243197714871645e-08, + "loss": 0.6988, + "step": 44494 + }, + { + "epoch": 1.63, + "learning_rate": 3.7239535650199585e-08, + "loss": 0.6945, + "step": 44495 + }, + { + "epoch": 1.63, + "learning_rate": 3.723587358552752e-08, + "loss": 0.703, + "step": 44496 + }, + { + "epoch": 1.63, + "learning_rate": 3.723221152085546e-08, + "loss": 0.6882, + "step": 44497 + }, + { + "epoch": 1.63, + "learning_rate": 3.722854945618339e-08, + "loss": 0.694, + "step": 44498 + }, + { + "epoch": 1.63, + "learning_rate": 3.722488739151133e-08, + "loss": 0.6855, + "step": 44499 + }, + { + "epoch": 1.63, + "learning_rate": 3.722122532683927e-08, + "loss": 0.6976, + "step": 44500 + }, + { + "epoch": 1.63, + "learning_rate": 3.7217563262167205e-08, + "loss": 0.6894, + "step": 44501 + }, + { + "epoch": 1.63, + "learning_rate": 3.7213901197495145e-08, + "loss": 0.7003, + "step": 44502 + }, + { + "epoch": 1.63, + "learning_rate": 3.7210239132823085e-08, + "loss": 0.7017, + "step": 44503 + }, + { + "epoch": 1.63, + "learning_rate": 3.7206577068151025e-08, + "loss": 0.6904, + "step": 44504 + }, + { + "epoch": 1.63, + "learning_rate": 3.720291500347896e-08, + "loss": 0.6906, + "step": 44505 + }, + { + "epoch": 1.63, + "learning_rate": 3.71992529388069e-08, + "loss": 0.6906, + "step": 44506 + }, + { + "epoch": 1.63, + "learning_rate": 3.719559087413484e-08, + "loss": 0.6954, + "step": 44507 + }, + { + "epoch": 1.63, + "learning_rate": 3.719192880946278e-08, + "loss": 0.6891, + "step": 44508 + }, + { + "epoch": 1.63, + "learning_rate": 3.718826674479071e-08, + "loss": 0.6903, + "step": 44509 + }, + { + "epoch": 1.63, + "learning_rate": 3.7184604680118644e-08, + "loss": 0.6864, + "step": 44510 + }, + { + "epoch": 1.63, + "learning_rate": 3.7180942615446584e-08, + "loss": 0.6914, + "step": 44511 + }, + { + "epoch": 1.63, + "learning_rate": 3.7177280550774524e-08, + "loss": 0.6943, + "step": 44512 + }, + { + "epoch": 1.63, + "learning_rate": 3.7173618486102464e-08, + "loss": 0.6837, + "step": 44513 + }, + { + "epoch": 1.63, + "learning_rate": 3.71699564214304e-08, + "loss": 0.7012, + "step": 44514 + }, + { + "epoch": 1.63, + "learning_rate": 3.716629435675834e-08, + "loss": 0.6918, + "step": 44515 + }, + { + "epoch": 1.63, + "learning_rate": 3.716263229208628e-08, + "loss": 0.7047, + "step": 44516 + }, + { + "epoch": 1.63, + "learning_rate": 3.715897022741422e-08, + "loss": 0.6923, + "step": 44517 + }, + { + "epoch": 1.63, + "learning_rate": 3.715530816274216e-08, + "loss": 0.6934, + "step": 44518 + }, + { + "epoch": 1.63, + "learning_rate": 3.715164609807009e-08, + "loss": 0.6966, + "step": 44519 + }, + { + "epoch": 1.63, + "learning_rate": 3.714798403339803e-08, + "loss": 0.6863, + "step": 44520 + }, + { + "epoch": 1.63, + "learning_rate": 3.7144321968725964e-08, + "loss": 0.691, + "step": 44521 + }, + { + "epoch": 1.63, + "learning_rate": 3.7140659904053904e-08, + "loss": 0.7029, + "step": 44522 + }, + { + "epoch": 1.63, + "learning_rate": 3.713699783938184e-08, + "loss": 0.6831, + "step": 44523 + }, + { + "epoch": 1.63, + "learning_rate": 3.713333577470978e-08, + "loss": 0.6958, + "step": 44524 + }, + { + "epoch": 1.63, + "learning_rate": 3.712967371003772e-08, + "loss": 0.6717, + "step": 44525 + }, + { + "epoch": 1.63, + "learning_rate": 3.712601164536566e-08, + "loss": 0.7031, + "step": 44526 + }, + { + "epoch": 1.63, + "learning_rate": 3.7122349580693597e-08, + "loss": 0.7205, + "step": 44527 + }, + { + "epoch": 1.63, + "learning_rate": 3.711868751602153e-08, + "loss": 0.6881, + "step": 44528 + }, + { + "epoch": 1.63, + "learning_rate": 3.711502545134947e-08, + "loss": 0.6916, + "step": 44529 + }, + { + "epoch": 1.63, + "learning_rate": 3.711136338667741e-08, + "loss": 0.703, + "step": 44530 + }, + { + "epoch": 1.63, + "learning_rate": 3.710770132200535e-08, + "loss": 0.6935, + "step": 44531 + }, + { + "epoch": 1.63, + "learning_rate": 3.710403925733328e-08, + "loss": 0.6948, + "step": 44532 + }, + { + "epoch": 1.63, + "learning_rate": 3.710037719266122e-08, + "loss": 0.7002, + "step": 44533 + }, + { + "epoch": 1.63, + "learning_rate": 3.7096715127989156e-08, + "loss": 0.7042, + "step": 44534 + }, + { + "epoch": 1.63, + "learning_rate": 3.7093053063317096e-08, + "loss": 0.6959, + "step": 44535 + }, + { + "epoch": 1.63, + "learning_rate": 3.7089390998645036e-08, + "loss": 0.7004, + "step": 44536 + }, + { + "epoch": 1.63, + "learning_rate": 3.708572893397297e-08, + "loss": 0.6951, + "step": 44537 + }, + { + "epoch": 1.63, + "learning_rate": 3.708206686930091e-08, + "loss": 0.6906, + "step": 44538 + }, + { + "epoch": 1.63, + "learning_rate": 3.707840480462885e-08, + "loss": 0.6936, + "step": 44539 + }, + { + "epoch": 1.63, + "learning_rate": 3.707474273995679e-08, + "loss": 0.7011, + "step": 44540 + }, + { + "epoch": 1.63, + "learning_rate": 3.707108067528472e-08, + "loss": 0.692, + "step": 44541 + }, + { + "epoch": 1.63, + "learning_rate": 3.706741861061266e-08, + "loss": 0.6935, + "step": 44542 + }, + { + "epoch": 1.63, + "learning_rate": 3.70637565459406e-08, + "loss": 0.6949, + "step": 44543 + }, + { + "epoch": 1.63, + "learning_rate": 3.7060094481268536e-08, + "loss": 0.7003, + "step": 44544 + }, + { + "epoch": 1.63, + "learning_rate": 3.7056432416596476e-08, + "loss": 0.6899, + "step": 44545 + }, + { + "epoch": 1.63, + "learning_rate": 3.705277035192441e-08, + "loss": 0.692, + "step": 44546 + }, + { + "epoch": 1.63, + "learning_rate": 3.704910828725235e-08, + "loss": 0.6971, + "step": 44547 + }, + { + "epoch": 1.63, + "learning_rate": 3.704544622258029e-08, + "loss": 0.7001, + "step": 44548 + }, + { + "epoch": 1.63, + "learning_rate": 3.704178415790823e-08, + "loss": 0.7013, + "step": 44549 + }, + { + "epoch": 1.63, + "learning_rate": 3.703812209323616e-08, + "loss": 0.6966, + "step": 44550 + }, + { + "epoch": 1.63, + "learning_rate": 3.70344600285641e-08, + "loss": 0.6973, + "step": 44551 + }, + { + "epoch": 1.63, + "learning_rate": 3.703079796389204e-08, + "loss": 0.6954, + "step": 44552 + }, + { + "epoch": 1.63, + "learning_rate": 3.702713589921998e-08, + "loss": 0.6947, + "step": 44553 + }, + { + "epoch": 1.63, + "learning_rate": 3.702347383454792e-08, + "loss": 0.6849, + "step": 44554 + }, + { + "epoch": 1.63, + "learning_rate": 3.7019811769875855e-08, + "loss": 0.6924, + "step": 44555 + }, + { + "epoch": 1.63, + "learning_rate": 3.7016149705203795e-08, + "loss": 0.6956, + "step": 44556 + }, + { + "epoch": 1.63, + "learning_rate": 3.701248764053173e-08, + "loss": 0.6703, + "step": 44557 + }, + { + "epoch": 1.63, + "learning_rate": 3.700882557585967e-08, + "loss": 0.6894, + "step": 44558 + }, + { + "epoch": 1.63, + "learning_rate": 3.70051635111876e-08, + "loss": 0.6867, + "step": 44559 + }, + { + "epoch": 1.63, + "learning_rate": 3.700150144651554e-08, + "loss": 0.6944, + "step": 44560 + }, + { + "epoch": 1.63, + "learning_rate": 3.699783938184348e-08, + "loss": 0.6979, + "step": 44561 + }, + { + "epoch": 1.63, + "learning_rate": 3.699417731717142e-08, + "loss": 0.6953, + "step": 44562 + }, + { + "epoch": 1.63, + "learning_rate": 3.699051525249936e-08, + "loss": 0.6945, + "step": 44563 + }, + { + "epoch": 1.63, + "learning_rate": 3.6986853187827295e-08, + "loss": 0.6893, + "step": 44564 + }, + { + "epoch": 1.63, + "learning_rate": 3.6983191123155235e-08, + "loss": 0.6985, + "step": 44565 + }, + { + "epoch": 1.63, + "learning_rate": 3.6979529058483174e-08, + "loss": 0.6725, + "step": 44566 + }, + { + "epoch": 1.63, + "learning_rate": 3.6975866993811114e-08, + "loss": 0.7008, + "step": 44567 + }, + { + "epoch": 1.63, + "learning_rate": 3.697220492913905e-08, + "loss": 0.694, + "step": 44568 + }, + { + "epoch": 1.63, + "learning_rate": 3.696854286446698e-08, + "loss": 0.6852, + "step": 44569 + }, + { + "epoch": 1.63, + "learning_rate": 3.696488079979492e-08, + "loss": 0.7112, + "step": 44570 + }, + { + "epoch": 1.63, + "learning_rate": 3.696121873512286e-08, + "loss": 0.6877, + "step": 44571 + }, + { + "epoch": 1.63, + "learning_rate": 3.6957556670450794e-08, + "loss": 0.6897, + "step": 44572 + }, + { + "epoch": 1.63, + "learning_rate": 3.6953894605778734e-08, + "loss": 0.689, + "step": 44573 + }, + { + "epoch": 1.63, + "learning_rate": 3.6950232541106674e-08, + "loss": 0.7017, + "step": 44574 + }, + { + "epoch": 1.63, + "learning_rate": 3.6946570476434614e-08, + "loss": 0.6974, + "step": 44575 + }, + { + "epoch": 1.63, + "learning_rate": 3.6942908411762554e-08, + "loss": 0.6932, + "step": 44576 + }, + { + "epoch": 1.63, + "learning_rate": 3.693924634709049e-08, + "loss": 0.6908, + "step": 44577 + }, + { + "epoch": 1.63, + "learning_rate": 3.693558428241843e-08, + "loss": 0.6871, + "step": 44578 + }, + { + "epoch": 1.63, + "learning_rate": 3.693192221774637e-08, + "loss": 0.6945, + "step": 44579 + }, + { + "epoch": 1.63, + "learning_rate": 3.69282601530743e-08, + "loss": 0.6839, + "step": 44580 + }, + { + "epoch": 1.63, + "learning_rate": 3.6924598088402234e-08, + "loss": 0.6955, + "step": 44581 + }, + { + "epoch": 1.63, + "learning_rate": 3.6920936023730174e-08, + "loss": 0.7061, + "step": 44582 + }, + { + "epoch": 1.63, + "learning_rate": 3.6917273959058114e-08, + "loss": 0.6904, + "step": 44583 + }, + { + "epoch": 1.63, + "learning_rate": 3.6913611894386053e-08, + "loss": 0.6916, + "step": 44584 + }, + { + "epoch": 1.63, + "learning_rate": 3.6909949829713993e-08, + "loss": 0.6975, + "step": 44585 + }, + { + "epoch": 1.63, + "learning_rate": 3.690628776504193e-08, + "loss": 0.6941, + "step": 44586 + }, + { + "epoch": 1.63, + "learning_rate": 3.6902625700369867e-08, + "loss": 0.6983, + "step": 44587 + }, + { + "epoch": 1.63, + "learning_rate": 3.6898963635697807e-08, + "loss": 0.6898, + "step": 44588 + }, + { + "epoch": 1.63, + "learning_rate": 3.6895301571025746e-08, + "loss": 0.6844, + "step": 44589 + }, + { + "epoch": 1.63, + "learning_rate": 3.689163950635368e-08, + "loss": 0.6931, + "step": 44590 + }, + { + "epoch": 1.63, + "learning_rate": 3.688797744168162e-08, + "loss": 0.691, + "step": 44591 + }, + { + "epoch": 1.63, + "learning_rate": 3.688431537700955e-08, + "loss": 0.704, + "step": 44592 + }, + { + "epoch": 1.63, + "learning_rate": 3.688065331233749e-08, + "loss": 0.6904, + "step": 44593 + }, + { + "epoch": 1.63, + "learning_rate": 3.687699124766543e-08, + "loss": 0.7042, + "step": 44594 + }, + { + "epoch": 1.63, + "learning_rate": 3.6873329182993366e-08, + "loss": 0.7021, + "step": 44595 + }, + { + "epoch": 1.63, + "learning_rate": 3.6869667118321306e-08, + "loss": 0.6946, + "step": 44596 + }, + { + "epoch": 1.63, + "learning_rate": 3.6866005053649246e-08, + "loss": 0.7056, + "step": 44597 + }, + { + "epoch": 1.63, + "learning_rate": 3.6862342988977186e-08, + "loss": 0.6959, + "step": 44598 + }, + { + "epoch": 1.63, + "learning_rate": 3.685868092430512e-08, + "loss": 0.6973, + "step": 44599 + }, + { + "epoch": 1.63, + "learning_rate": 3.685501885963306e-08, + "loss": 0.6899, + "step": 44600 + }, + { + "epoch": 1.63, + "learning_rate": 3.6851356794961e-08, + "loss": 0.692, + "step": 44601 + }, + { + "epoch": 1.63, + "learning_rate": 3.684769473028894e-08, + "loss": 0.6846, + "step": 44602 + }, + { + "epoch": 1.63, + "learning_rate": 3.684403266561687e-08, + "loss": 0.68, + "step": 44603 + }, + { + "epoch": 1.63, + "learning_rate": 3.6840370600944806e-08, + "loss": 0.696, + "step": 44604 + }, + { + "epoch": 1.63, + "learning_rate": 3.6836708536272746e-08, + "loss": 0.6903, + "step": 44605 + }, + { + "epoch": 1.63, + "learning_rate": 3.6833046471600686e-08, + "loss": 0.6958, + "step": 44606 + }, + { + "epoch": 1.63, + "learning_rate": 3.6829384406928625e-08, + "loss": 0.6875, + "step": 44607 + }, + { + "epoch": 1.63, + "learning_rate": 3.682572234225656e-08, + "loss": 0.6933, + "step": 44608 + }, + { + "epoch": 1.63, + "learning_rate": 3.68220602775845e-08, + "loss": 0.6939, + "step": 44609 + }, + { + "epoch": 1.63, + "learning_rate": 3.681839821291244e-08, + "loss": 0.6904, + "step": 44610 + }, + { + "epoch": 1.63, + "learning_rate": 3.681473614824038e-08, + "loss": 0.6923, + "step": 44611 + }, + { + "epoch": 1.63, + "learning_rate": 3.681107408356832e-08, + "loss": 0.6917, + "step": 44612 + }, + { + "epoch": 1.63, + "learning_rate": 3.680741201889625e-08, + "loss": 0.6931, + "step": 44613 + }, + { + "epoch": 1.63, + "learning_rate": 3.680374995422419e-08, + "loss": 0.6809, + "step": 44614 + }, + { + "epoch": 1.63, + "learning_rate": 3.6800087889552125e-08, + "loss": 0.6946, + "step": 44615 + }, + { + "epoch": 1.63, + "learning_rate": 3.6796425824880065e-08, + "loss": 0.6827, + "step": 44616 + }, + { + "epoch": 1.63, + "learning_rate": 3.6792763760208e-08, + "loss": 0.6878, + "step": 44617 + }, + { + "epoch": 1.63, + "learning_rate": 3.678910169553594e-08, + "loss": 0.7034, + "step": 44618 + }, + { + "epoch": 1.63, + "learning_rate": 3.678543963086388e-08, + "loss": 0.6887, + "step": 44619 + }, + { + "epoch": 1.63, + "learning_rate": 3.678177756619182e-08, + "loss": 0.6836, + "step": 44620 + }, + { + "epoch": 1.63, + "learning_rate": 3.677811550151976e-08, + "loss": 0.6875, + "step": 44621 + }, + { + "epoch": 1.63, + "learning_rate": 3.677445343684769e-08, + "loss": 0.7206, + "step": 44622 + }, + { + "epoch": 1.63, + "learning_rate": 3.677079137217563e-08, + "loss": 0.6894, + "step": 44623 + }, + { + "epoch": 1.63, + "learning_rate": 3.676712930750357e-08, + "loss": 0.6916, + "step": 44624 + }, + { + "epoch": 1.63, + "learning_rate": 3.676346724283151e-08, + "loss": 0.693, + "step": 44625 + }, + { + "epoch": 1.63, + "learning_rate": 3.6759805178159444e-08, + "loss": 0.682, + "step": 44626 + }, + { + "epoch": 1.63, + "learning_rate": 3.6756143113487384e-08, + "loss": 0.6853, + "step": 44627 + }, + { + "epoch": 1.63, + "learning_rate": 3.675248104881532e-08, + "loss": 0.6919, + "step": 44628 + }, + { + "epoch": 1.63, + "learning_rate": 3.674881898414326e-08, + "loss": 0.6925, + "step": 44629 + }, + { + "epoch": 1.63, + "learning_rate": 3.67451569194712e-08, + "loss": 0.6958, + "step": 44630 + }, + { + "epoch": 1.63, + "learning_rate": 3.674149485479913e-08, + "loss": 0.6935, + "step": 44631 + }, + { + "epoch": 1.63, + "learning_rate": 3.673783279012707e-08, + "loss": 0.6839, + "step": 44632 + }, + { + "epoch": 1.63, + "learning_rate": 3.673417072545501e-08, + "loss": 0.7013, + "step": 44633 + }, + { + "epoch": 1.63, + "learning_rate": 3.673050866078295e-08, + "loss": 0.6863, + "step": 44634 + }, + { + "epoch": 1.63, + "learning_rate": 3.6726846596110884e-08, + "loss": 0.6933, + "step": 44635 + }, + { + "epoch": 1.63, + "learning_rate": 3.6723184531438824e-08, + "loss": 0.6951, + "step": 44636 + }, + { + "epoch": 1.63, + "learning_rate": 3.6719522466766764e-08, + "loss": 0.6746, + "step": 44637 + }, + { + "epoch": 1.63, + "learning_rate": 3.6715860402094704e-08, + "loss": 0.682, + "step": 44638 + }, + { + "epoch": 1.63, + "learning_rate": 3.671219833742264e-08, + "loss": 0.7104, + "step": 44639 + }, + { + "epoch": 1.63, + "learning_rate": 3.670853627275057e-08, + "loss": 0.6894, + "step": 44640 + }, + { + "epoch": 1.63, + "learning_rate": 3.670487420807851e-08, + "loss": 0.702, + "step": 44641 + }, + { + "epoch": 1.63, + "learning_rate": 3.670121214340645e-08, + "loss": 0.6948, + "step": 44642 + }, + { + "epoch": 1.63, + "learning_rate": 3.669755007873439e-08, + "loss": 0.6952, + "step": 44643 + }, + { + "epoch": 1.63, + "learning_rate": 3.6693888014062323e-08, + "loss": 0.6824, + "step": 44644 + }, + { + "epoch": 1.63, + "learning_rate": 3.6690225949390263e-08, + "loss": 0.6885, + "step": 44645 + }, + { + "epoch": 1.63, + "learning_rate": 3.66865638847182e-08, + "loss": 0.692, + "step": 44646 + }, + { + "epoch": 1.63, + "learning_rate": 3.668290182004614e-08, + "loss": 0.687, + "step": 44647 + }, + { + "epoch": 1.63, + "learning_rate": 3.667923975537408e-08, + "loss": 0.6953, + "step": 44648 + }, + { + "epoch": 1.63, + "learning_rate": 3.6675577690702016e-08, + "loss": 0.7103, + "step": 44649 + }, + { + "epoch": 1.63, + "learning_rate": 3.6671915626029956e-08, + "loss": 0.6873, + "step": 44650 + }, + { + "epoch": 1.63, + "learning_rate": 3.666825356135789e-08, + "loss": 0.7102, + "step": 44651 + }, + { + "epoch": 1.63, + "learning_rate": 3.666459149668583e-08, + "loss": 0.6919, + "step": 44652 + }, + { + "epoch": 1.63, + "learning_rate": 3.666092943201376e-08, + "loss": 0.6778, + "step": 44653 + }, + { + "epoch": 1.63, + "learning_rate": 3.66572673673417e-08, + "loss": 0.6888, + "step": 44654 + }, + { + "epoch": 1.63, + "learning_rate": 3.665360530266964e-08, + "loss": 0.6889, + "step": 44655 + }, + { + "epoch": 1.63, + "learning_rate": 3.664994323799758e-08, + "loss": 0.68, + "step": 44656 + }, + { + "epoch": 1.63, + "learning_rate": 3.664628117332552e-08, + "loss": 0.7062, + "step": 44657 + }, + { + "epoch": 1.63, + "learning_rate": 3.6642619108653456e-08, + "loss": 0.6885, + "step": 44658 + }, + { + "epoch": 1.63, + "learning_rate": 3.6638957043981396e-08, + "loss": 0.706, + "step": 44659 + }, + { + "epoch": 1.63, + "learning_rate": 3.6635294979309336e-08, + "loss": 0.7089, + "step": 44660 + }, + { + "epoch": 1.63, + "learning_rate": 3.6631632914637276e-08, + "loss": 0.6979, + "step": 44661 + }, + { + "epoch": 1.63, + "learning_rate": 3.662797084996521e-08, + "loss": 0.6845, + "step": 44662 + }, + { + "epoch": 1.63, + "learning_rate": 3.662430878529314e-08, + "loss": 0.6852, + "step": 44663 + }, + { + "epoch": 1.63, + "learning_rate": 3.662064672062108e-08, + "loss": 0.7031, + "step": 44664 + }, + { + "epoch": 1.63, + "learning_rate": 3.661698465594902e-08, + "loss": 0.6912, + "step": 44665 + }, + { + "epoch": 1.63, + "learning_rate": 3.661332259127696e-08, + "loss": 0.6914, + "step": 44666 + }, + { + "epoch": 1.63, + "learning_rate": 3.6609660526604895e-08, + "loss": 0.6941, + "step": 44667 + }, + { + "epoch": 1.63, + "learning_rate": 3.6605998461932835e-08, + "loss": 0.6885, + "step": 44668 + }, + { + "epoch": 1.63, + "learning_rate": 3.6602336397260775e-08, + "loss": 0.6958, + "step": 44669 + }, + { + "epoch": 1.63, + "learning_rate": 3.6598674332588715e-08, + "loss": 0.7007, + "step": 44670 + }, + { + "epoch": 1.63, + "learning_rate": 3.659501226791665e-08, + "loss": 0.6836, + "step": 44671 + }, + { + "epoch": 1.63, + "learning_rate": 3.659135020324459e-08, + "loss": 0.6887, + "step": 44672 + }, + { + "epoch": 1.63, + "learning_rate": 3.658768813857253e-08, + "loss": 0.7027, + "step": 44673 + }, + { + "epoch": 1.63, + "learning_rate": 3.658402607390046e-08, + "loss": 0.696, + "step": 44674 + }, + { + "epoch": 1.63, + "learning_rate": 3.65803640092284e-08, + "loss": 0.6875, + "step": 44675 + }, + { + "epoch": 1.63, + "learning_rate": 3.6576701944556335e-08, + "loss": 0.6894, + "step": 44676 + }, + { + "epoch": 1.63, + "learning_rate": 3.6573039879884275e-08, + "loss": 0.6854, + "step": 44677 + }, + { + "epoch": 1.63, + "learning_rate": 3.6569377815212215e-08, + "loss": 0.6894, + "step": 44678 + }, + { + "epoch": 1.63, + "learning_rate": 3.6565715750540155e-08, + "loss": 0.6838, + "step": 44679 + }, + { + "epoch": 1.63, + "learning_rate": 3.656205368586809e-08, + "loss": 0.6949, + "step": 44680 + }, + { + "epoch": 1.63, + "learning_rate": 3.655839162119603e-08, + "loss": 0.6845, + "step": 44681 + }, + { + "epoch": 1.63, + "learning_rate": 3.655472955652397e-08, + "loss": 0.6807, + "step": 44682 + }, + { + "epoch": 1.63, + "learning_rate": 3.655106749185191e-08, + "loss": 0.6952, + "step": 44683 + }, + { + "epoch": 1.63, + "learning_rate": 3.654740542717985e-08, + "loss": 0.6923, + "step": 44684 + }, + { + "epoch": 1.63, + "learning_rate": 3.654374336250778e-08, + "loss": 0.6843, + "step": 44685 + }, + { + "epoch": 1.63, + "learning_rate": 3.6540081297835714e-08, + "loss": 0.6928, + "step": 44686 + }, + { + "epoch": 1.63, + "learning_rate": 3.6536419233163654e-08, + "loss": 0.7045, + "step": 44687 + }, + { + "epoch": 1.64, + "learning_rate": 3.6532757168491594e-08, + "loss": 0.6959, + "step": 44688 + }, + { + "epoch": 1.64, + "learning_rate": 3.652909510381953e-08, + "loss": 0.6902, + "step": 44689 + }, + { + "epoch": 1.64, + "learning_rate": 3.652543303914747e-08, + "loss": 0.7027, + "step": 44690 + }, + { + "epoch": 1.64, + "learning_rate": 3.652177097447541e-08, + "loss": 0.694, + "step": 44691 + }, + { + "epoch": 1.64, + "learning_rate": 3.651810890980335e-08, + "loss": 0.7006, + "step": 44692 + }, + { + "epoch": 1.64, + "learning_rate": 3.651444684513129e-08, + "loss": 0.6926, + "step": 44693 + }, + { + "epoch": 1.64, + "learning_rate": 3.651078478045922e-08, + "loss": 0.7025, + "step": 44694 + }, + { + "epoch": 1.64, + "learning_rate": 3.650712271578716e-08, + "loss": 0.6897, + "step": 44695 + }, + { + "epoch": 1.64, + "learning_rate": 3.65034606511151e-08, + "loss": 0.7021, + "step": 44696 + }, + { + "epoch": 1.64, + "learning_rate": 3.6499798586443034e-08, + "loss": 0.6898, + "step": 44697 + }, + { + "epoch": 1.64, + "learning_rate": 3.6496136521770974e-08, + "loss": 0.6954, + "step": 44698 + }, + { + "epoch": 1.64, + "learning_rate": 3.649247445709891e-08, + "loss": 0.6852, + "step": 44699 + }, + { + "epoch": 1.64, + "learning_rate": 3.648881239242685e-08, + "loss": 0.6859, + "step": 44700 + }, + { + "epoch": 1.64, + "learning_rate": 3.648515032775479e-08, + "loss": 0.6954, + "step": 44701 + }, + { + "epoch": 1.64, + "learning_rate": 3.648148826308273e-08, + "loss": 0.7011, + "step": 44702 + }, + { + "epoch": 1.64, + "learning_rate": 3.647782619841066e-08, + "loss": 0.6927, + "step": 44703 + }, + { + "epoch": 1.64, + "learning_rate": 3.64741641337386e-08, + "loss": 0.6957, + "step": 44704 + }, + { + "epoch": 1.64, + "learning_rate": 3.647050206906654e-08, + "loss": 0.7037, + "step": 44705 + }, + { + "epoch": 1.64, + "learning_rate": 3.646684000439448e-08, + "loss": 0.7054, + "step": 44706 + }, + { + "epoch": 1.64, + "learning_rate": 3.646317793972241e-08, + "loss": 0.697, + "step": 44707 + }, + { + "epoch": 1.64, + "learning_rate": 3.645951587505035e-08, + "loss": 0.7016, + "step": 44708 + }, + { + "epoch": 1.64, + "learning_rate": 3.645585381037829e-08, + "loss": 0.7038, + "step": 44709 + }, + { + "epoch": 1.64, + "learning_rate": 3.6452191745706226e-08, + "loss": 0.6871, + "step": 44710 + }, + { + "epoch": 1.64, + "learning_rate": 3.6448529681034166e-08, + "loss": 0.6904, + "step": 44711 + }, + { + "epoch": 1.64, + "learning_rate": 3.64448676163621e-08, + "loss": 0.6779, + "step": 44712 + }, + { + "epoch": 1.64, + "learning_rate": 3.644120555169004e-08, + "loss": 0.6903, + "step": 44713 + }, + { + "epoch": 1.64, + "learning_rate": 3.643754348701798e-08, + "loss": 0.6977, + "step": 44714 + }, + { + "epoch": 1.64, + "learning_rate": 3.643388142234592e-08, + "loss": 0.699, + "step": 44715 + }, + { + "epoch": 1.64, + "learning_rate": 3.643021935767385e-08, + "loss": 0.6992, + "step": 44716 + }, + { + "epoch": 1.64, + "learning_rate": 3.642655729300179e-08, + "loss": 0.6866, + "step": 44717 + }, + { + "epoch": 1.64, + "learning_rate": 3.642289522832973e-08, + "loss": 0.6771, + "step": 44718 + }, + { + "epoch": 1.64, + "learning_rate": 3.641923316365767e-08, + "loss": 0.6952, + "step": 44719 + }, + { + "epoch": 1.64, + "learning_rate": 3.6415571098985606e-08, + "loss": 0.6992, + "step": 44720 + }, + { + "epoch": 1.64, + "learning_rate": 3.6411909034313546e-08, + "loss": 0.6948, + "step": 44721 + }, + { + "epoch": 1.64, + "learning_rate": 3.640824696964148e-08, + "loss": 0.694, + "step": 44722 + }, + { + "epoch": 1.64, + "learning_rate": 3.640458490496942e-08, + "loss": 0.7008, + "step": 44723 + }, + { + "epoch": 1.64, + "learning_rate": 3.640092284029736e-08, + "loss": 0.6929, + "step": 44724 + }, + { + "epoch": 1.64, + "learning_rate": 3.639726077562529e-08, + "loss": 0.6999, + "step": 44725 + }, + { + "epoch": 1.64, + "learning_rate": 3.639359871095323e-08, + "loss": 0.6894, + "step": 44726 + }, + { + "epoch": 1.64, + "learning_rate": 3.638993664628117e-08, + "loss": 0.6939, + "step": 44727 + }, + { + "epoch": 1.64, + "learning_rate": 3.638627458160911e-08, + "loss": 0.6811, + "step": 44728 + }, + { + "epoch": 1.64, + "learning_rate": 3.638261251693705e-08, + "loss": 0.7033, + "step": 44729 + }, + { + "epoch": 1.64, + "learning_rate": 3.6378950452264985e-08, + "loss": 0.677, + "step": 44730 + }, + { + "epoch": 1.64, + "learning_rate": 3.6375288387592925e-08, + "loss": 0.6938, + "step": 44731 + }, + { + "epoch": 1.64, + "learning_rate": 3.6371626322920865e-08, + "loss": 0.6902, + "step": 44732 + }, + { + "epoch": 1.64, + "learning_rate": 3.63679642582488e-08, + "loss": 0.6836, + "step": 44733 + }, + { + "epoch": 1.64, + "learning_rate": 3.636430219357673e-08, + "loss": 0.6935, + "step": 44734 + }, + { + "epoch": 1.64, + "learning_rate": 3.636064012890467e-08, + "loss": 0.6856, + "step": 44735 + }, + { + "epoch": 1.64, + "learning_rate": 3.635697806423261e-08, + "loss": 0.6903, + "step": 44736 + }, + { + "epoch": 1.64, + "learning_rate": 3.635331599956055e-08, + "loss": 0.6895, + "step": 44737 + }, + { + "epoch": 1.64, + "learning_rate": 3.634965393488849e-08, + "loss": 0.6813, + "step": 44738 + }, + { + "epoch": 1.64, + "learning_rate": 3.6345991870216425e-08, + "loss": 0.7051, + "step": 44739 + }, + { + "epoch": 1.64, + "learning_rate": 3.6342329805544365e-08, + "loss": 0.6979, + "step": 44740 + }, + { + "epoch": 1.64, + "learning_rate": 3.6338667740872305e-08, + "loss": 0.7022, + "step": 44741 + }, + { + "epoch": 1.64, + "learning_rate": 3.6335005676200245e-08, + "loss": 0.6914, + "step": 44742 + }, + { + "epoch": 1.64, + "learning_rate": 3.633134361152818e-08, + "loss": 0.7065, + "step": 44743 + }, + { + "epoch": 1.64, + "learning_rate": 3.632768154685612e-08, + "loss": 0.6919, + "step": 44744 + }, + { + "epoch": 1.64, + "learning_rate": 3.632401948218405e-08, + "loss": 0.6903, + "step": 44745 + }, + { + "epoch": 1.64, + "learning_rate": 3.632035741751199e-08, + "loss": 0.6854, + "step": 44746 + }, + { + "epoch": 1.64, + "learning_rate": 3.631669535283993e-08, + "loss": 0.6961, + "step": 44747 + }, + { + "epoch": 1.64, + "learning_rate": 3.6313033288167864e-08, + "loss": 0.7021, + "step": 44748 + }, + { + "epoch": 1.64, + "learning_rate": 3.6309371223495804e-08, + "loss": 0.7087, + "step": 44749 + }, + { + "epoch": 1.64, + "learning_rate": 3.6305709158823744e-08, + "loss": 0.6911, + "step": 44750 + }, + { + "epoch": 1.64, + "learning_rate": 3.6302047094151684e-08, + "loss": 0.687, + "step": 44751 + }, + { + "epoch": 1.64, + "learning_rate": 3.629838502947962e-08, + "loss": 0.6965, + "step": 44752 + }, + { + "epoch": 1.64, + "learning_rate": 3.629472296480756e-08, + "loss": 0.6814, + "step": 44753 + }, + { + "epoch": 1.64, + "learning_rate": 3.62910609001355e-08, + "loss": 0.6945, + "step": 44754 + }, + { + "epoch": 1.64, + "learning_rate": 3.628739883546344e-08, + "loss": 0.6814, + "step": 44755 + }, + { + "epoch": 1.64, + "learning_rate": 3.628373677079137e-08, + "loss": 0.6884, + "step": 44756 + }, + { + "epoch": 1.64, + "learning_rate": 3.6280074706119304e-08, + "loss": 0.6952, + "step": 44757 + }, + { + "epoch": 1.64, + "learning_rate": 3.6276412641447244e-08, + "loss": 0.7014, + "step": 44758 + }, + { + "epoch": 1.64, + "learning_rate": 3.6272750576775184e-08, + "loss": 0.6994, + "step": 44759 + }, + { + "epoch": 1.64, + "learning_rate": 3.6269088512103124e-08, + "loss": 0.6988, + "step": 44760 + }, + { + "epoch": 1.64, + "learning_rate": 3.626542644743106e-08, + "loss": 0.6865, + "step": 44761 + }, + { + "epoch": 1.64, + "learning_rate": 3.6261764382759e-08, + "loss": 0.6944, + "step": 44762 + }, + { + "epoch": 1.64, + "learning_rate": 3.625810231808694e-08, + "loss": 0.6877, + "step": 44763 + }, + { + "epoch": 1.64, + "learning_rate": 3.6254440253414877e-08, + "loss": 0.7024, + "step": 44764 + }, + { + "epoch": 1.64, + "learning_rate": 3.625077818874281e-08, + "loss": 0.7104, + "step": 44765 + }, + { + "epoch": 1.64, + "learning_rate": 3.624711612407075e-08, + "loss": 0.6845, + "step": 44766 + }, + { + "epoch": 1.64, + "learning_rate": 3.624345405939869e-08, + "loss": 0.6822, + "step": 44767 + }, + { + "epoch": 1.64, + "learning_rate": 3.623979199472662e-08, + "loss": 0.6814, + "step": 44768 + }, + { + "epoch": 1.64, + "learning_rate": 3.623612993005456e-08, + "loss": 0.694, + "step": 44769 + }, + { + "epoch": 1.64, + "learning_rate": 3.6232467865382496e-08, + "loss": 0.6891, + "step": 44770 + }, + { + "epoch": 1.64, + "learning_rate": 3.6228805800710436e-08, + "loss": 0.6907, + "step": 44771 + }, + { + "epoch": 1.64, + "learning_rate": 3.6225143736038376e-08, + "loss": 0.7075, + "step": 44772 + }, + { + "epoch": 1.64, + "learning_rate": 3.6221481671366316e-08, + "loss": 0.6948, + "step": 44773 + }, + { + "epoch": 1.64, + "learning_rate": 3.621781960669425e-08, + "loss": 0.6978, + "step": 44774 + }, + { + "epoch": 1.64, + "learning_rate": 3.621415754202219e-08, + "loss": 0.696, + "step": 44775 + }, + { + "epoch": 1.64, + "learning_rate": 3.621049547735013e-08, + "loss": 0.6859, + "step": 44776 + }, + { + "epoch": 1.64, + "learning_rate": 3.620683341267807e-08, + "loss": 0.6862, + "step": 44777 + }, + { + "epoch": 1.64, + "learning_rate": 3.620317134800601e-08, + "loss": 0.7075, + "step": 44778 + }, + { + "epoch": 1.64, + "learning_rate": 3.619950928333394e-08, + "loss": 0.7035, + "step": 44779 + }, + { + "epoch": 1.64, + "learning_rate": 3.6195847218661876e-08, + "loss": 0.6963, + "step": 44780 + }, + { + "epoch": 1.64, + "learning_rate": 3.6192185153989816e-08, + "loss": 0.6976, + "step": 44781 + }, + { + "epoch": 1.64, + "learning_rate": 3.6188523089317756e-08, + "loss": 0.6919, + "step": 44782 + }, + { + "epoch": 1.64, + "learning_rate": 3.618486102464569e-08, + "loss": 0.6969, + "step": 44783 + }, + { + "epoch": 1.64, + "learning_rate": 3.618119895997363e-08, + "loss": 0.7, + "step": 44784 + }, + { + "epoch": 1.64, + "learning_rate": 3.617753689530157e-08, + "loss": 0.6965, + "step": 44785 + }, + { + "epoch": 1.64, + "learning_rate": 3.617387483062951e-08, + "loss": 0.6853, + "step": 44786 + }, + { + "epoch": 1.64, + "learning_rate": 3.617021276595745e-08, + "loss": 0.6941, + "step": 44787 + }, + { + "epoch": 1.64, + "learning_rate": 3.616655070128538e-08, + "loss": 0.7005, + "step": 44788 + }, + { + "epoch": 1.64, + "learning_rate": 3.616288863661332e-08, + "loss": 0.6982, + "step": 44789 + }, + { + "epoch": 1.64, + "learning_rate": 3.615922657194126e-08, + "loss": 0.6898, + "step": 44790 + }, + { + "epoch": 1.64, + "learning_rate": 3.6155564507269195e-08, + "loss": 0.6888, + "step": 44791 + }, + { + "epoch": 1.64, + "learning_rate": 3.6151902442597135e-08, + "loss": 0.6908, + "step": 44792 + }, + { + "epoch": 1.64, + "learning_rate": 3.614824037792507e-08, + "loss": 0.6838, + "step": 44793 + }, + { + "epoch": 1.64, + "learning_rate": 3.614457831325301e-08, + "loss": 0.6961, + "step": 44794 + }, + { + "epoch": 1.64, + "learning_rate": 3.614091624858095e-08, + "loss": 0.7015, + "step": 44795 + }, + { + "epoch": 1.64, + "learning_rate": 3.613725418390889e-08, + "loss": 0.6823, + "step": 44796 + }, + { + "epoch": 1.64, + "learning_rate": 3.613359211923682e-08, + "loss": 0.6921, + "step": 44797 + }, + { + "epoch": 1.64, + "learning_rate": 3.612993005456476e-08, + "loss": 0.6859, + "step": 44798 + }, + { + "epoch": 1.64, + "learning_rate": 3.61262679898927e-08, + "loss": 0.6937, + "step": 44799 + }, + { + "epoch": 1.64, + "learning_rate": 3.612260592522064e-08, + "loss": 0.6935, + "step": 44800 + }, + { + "epoch": 1.64, + "learning_rate": 3.6118943860548575e-08, + "loss": 0.6905, + "step": 44801 + }, + { + "epoch": 1.64, + "learning_rate": 3.6115281795876515e-08, + "loss": 0.6973, + "step": 44802 + }, + { + "epoch": 1.64, + "learning_rate": 3.6111619731204454e-08, + "loss": 0.6959, + "step": 44803 + }, + { + "epoch": 1.64, + "learning_rate": 3.610795766653239e-08, + "loss": 0.6932, + "step": 44804 + }, + { + "epoch": 1.64, + "learning_rate": 3.610429560186033e-08, + "loss": 0.6968, + "step": 44805 + }, + { + "epoch": 1.64, + "learning_rate": 3.610063353718826e-08, + "loss": 0.6763, + "step": 44806 + }, + { + "epoch": 1.64, + "learning_rate": 3.60969714725162e-08, + "loss": 0.6981, + "step": 44807 + }, + { + "epoch": 1.64, + "learning_rate": 3.609330940784414e-08, + "loss": 0.6832, + "step": 44808 + }, + { + "epoch": 1.64, + "learning_rate": 3.608964734317208e-08, + "loss": 0.7035, + "step": 44809 + }, + { + "epoch": 1.64, + "learning_rate": 3.6085985278500014e-08, + "loss": 0.6917, + "step": 44810 + }, + { + "epoch": 1.64, + "learning_rate": 3.6082323213827954e-08, + "loss": 0.69, + "step": 44811 + }, + { + "epoch": 1.64, + "learning_rate": 3.6078661149155894e-08, + "loss": 0.6989, + "step": 44812 + }, + { + "epoch": 1.64, + "learning_rate": 3.6074999084483834e-08, + "loss": 0.7013, + "step": 44813 + }, + { + "epoch": 1.64, + "learning_rate": 3.6071337019811774e-08, + "loss": 0.6808, + "step": 44814 + }, + { + "epoch": 1.64, + "learning_rate": 3.606767495513971e-08, + "loss": 0.6974, + "step": 44815 + }, + { + "epoch": 1.64, + "learning_rate": 3.606401289046764e-08, + "loss": 0.7123, + "step": 44816 + }, + { + "epoch": 1.64, + "learning_rate": 3.606035082579558e-08, + "loss": 0.7006, + "step": 44817 + }, + { + "epoch": 1.64, + "learning_rate": 3.605668876112352e-08, + "loss": 0.7028, + "step": 44818 + }, + { + "epoch": 1.64, + "learning_rate": 3.6053026696451454e-08, + "loss": 0.694, + "step": 44819 + }, + { + "epoch": 1.64, + "learning_rate": 3.6049364631779394e-08, + "loss": 0.6925, + "step": 44820 + }, + { + "epoch": 1.64, + "learning_rate": 3.6045702567107333e-08, + "loss": 0.6973, + "step": 44821 + }, + { + "epoch": 1.64, + "learning_rate": 3.6042040502435273e-08, + "loss": 0.6945, + "step": 44822 + }, + { + "epoch": 1.64, + "learning_rate": 3.6038378437763213e-08, + "loss": 0.689, + "step": 44823 + }, + { + "epoch": 1.64, + "learning_rate": 3.6034716373091147e-08, + "loss": 0.6954, + "step": 44824 + }, + { + "epoch": 1.64, + "learning_rate": 3.6031054308419087e-08, + "loss": 0.6844, + "step": 44825 + }, + { + "epoch": 1.64, + "learning_rate": 3.6027392243747026e-08, + "loss": 0.6896, + "step": 44826 + }, + { + "epoch": 1.64, + "learning_rate": 3.602373017907496e-08, + "loss": 0.6934, + "step": 44827 + }, + { + "epoch": 1.64, + "learning_rate": 3.602006811440289e-08, + "loss": 0.692, + "step": 44828 + }, + { + "epoch": 1.64, + "learning_rate": 3.601640604973083e-08, + "loss": 0.702, + "step": 44829 + }, + { + "epoch": 1.64, + "learning_rate": 3.601274398505877e-08, + "loss": 0.6882, + "step": 44830 + }, + { + "epoch": 1.64, + "learning_rate": 3.600908192038671e-08, + "loss": 0.6833, + "step": 44831 + }, + { + "epoch": 1.64, + "learning_rate": 3.600541985571465e-08, + "loss": 0.6854, + "step": 44832 + }, + { + "epoch": 1.64, + "learning_rate": 3.6001757791042586e-08, + "loss": 0.708, + "step": 44833 + }, + { + "epoch": 1.64, + "learning_rate": 3.5998095726370526e-08, + "loss": 0.6916, + "step": 44834 + }, + { + "epoch": 1.64, + "learning_rate": 3.5994433661698466e-08, + "loss": 0.6896, + "step": 44835 + }, + { + "epoch": 1.64, + "learning_rate": 3.5990771597026406e-08, + "loss": 0.6831, + "step": 44836 + }, + { + "epoch": 1.64, + "learning_rate": 3.598710953235434e-08, + "loss": 0.6926, + "step": 44837 + }, + { + "epoch": 1.64, + "learning_rate": 3.598344746768228e-08, + "loss": 0.6872, + "step": 44838 + }, + { + "epoch": 1.64, + "learning_rate": 3.597978540301021e-08, + "loss": 0.681, + "step": 44839 + }, + { + "epoch": 1.64, + "learning_rate": 3.597612333833815e-08, + "loss": 0.6904, + "step": 44840 + }, + { + "epoch": 1.64, + "learning_rate": 3.597246127366609e-08, + "loss": 0.6989, + "step": 44841 + }, + { + "epoch": 1.64, + "learning_rate": 3.5968799208994026e-08, + "loss": 0.6906, + "step": 44842 + }, + { + "epoch": 1.64, + "learning_rate": 3.5965137144321966e-08, + "loss": 0.7104, + "step": 44843 + }, + { + "epoch": 1.64, + "learning_rate": 3.5961475079649906e-08, + "loss": 0.6876, + "step": 44844 + }, + { + "epoch": 1.64, + "learning_rate": 3.5957813014977845e-08, + "loss": 0.6915, + "step": 44845 + }, + { + "epoch": 1.64, + "learning_rate": 3.595415095030578e-08, + "loss": 0.6879, + "step": 44846 + }, + { + "epoch": 1.64, + "learning_rate": 3.595048888563372e-08, + "loss": 0.6858, + "step": 44847 + }, + { + "epoch": 1.64, + "learning_rate": 3.594682682096166e-08, + "loss": 0.6844, + "step": 44848 + }, + { + "epoch": 1.64, + "learning_rate": 3.59431647562896e-08, + "loss": 0.6939, + "step": 44849 + }, + { + "epoch": 1.64, + "learning_rate": 3.593950269161753e-08, + "loss": 0.6815, + "step": 44850 + }, + { + "epoch": 1.64, + "learning_rate": 3.5935840626945465e-08, + "loss": 0.6897, + "step": 44851 + }, + { + "epoch": 1.64, + "learning_rate": 3.5932178562273405e-08, + "loss": 0.7051, + "step": 44852 + }, + { + "epoch": 1.64, + "learning_rate": 3.5928516497601345e-08, + "loss": 0.6822, + "step": 44853 + }, + { + "epoch": 1.64, + "learning_rate": 3.5924854432929285e-08, + "loss": 0.6876, + "step": 44854 + }, + { + "epoch": 1.64, + "learning_rate": 3.592119236825722e-08, + "loss": 0.6886, + "step": 44855 + }, + { + "epoch": 1.64, + "learning_rate": 3.591753030358516e-08, + "loss": 0.6779, + "step": 44856 + }, + { + "epoch": 1.64, + "learning_rate": 3.59138682389131e-08, + "loss": 0.692, + "step": 44857 + }, + { + "epoch": 1.64, + "learning_rate": 3.591020617424104e-08, + "loss": 0.6964, + "step": 44858 + }, + { + "epoch": 1.64, + "learning_rate": 3.590654410956898e-08, + "loss": 0.6977, + "step": 44859 + }, + { + "epoch": 1.64, + "learning_rate": 3.590288204489691e-08, + "loss": 0.695, + "step": 44860 + }, + { + "epoch": 1.64, + "learning_rate": 3.589921998022485e-08, + "loss": 0.6966, + "step": 44861 + }, + { + "epoch": 1.64, + "learning_rate": 3.5895557915552785e-08, + "loss": 0.6996, + "step": 44862 + }, + { + "epoch": 1.64, + "learning_rate": 3.5891895850880724e-08, + "loss": 0.7102, + "step": 44863 + }, + { + "epoch": 1.64, + "learning_rate": 3.588823378620866e-08, + "loss": 0.706, + "step": 44864 + }, + { + "epoch": 1.64, + "learning_rate": 3.58845717215366e-08, + "loss": 0.7055, + "step": 44865 + }, + { + "epoch": 1.64, + "learning_rate": 3.588090965686454e-08, + "loss": 0.687, + "step": 44866 + }, + { + "epoch": 1.64, + "learning_rate": 3.587724759219248e-08, + "loss": 0.6897, + "step": 44867 + }, + { + "epoch": 1.64, + "learning_rate": 3.587358552752042e-08, + "loss": 0.697, + "step": 44868 + }, + { + "epoch": 1.64, + "learning_rate": 3.586992346284835e-08, + "loss": 0.6979, + "step": 44869 + }, + { + "epoch": 1.64, + "learning_rate": 3.586626139817629e-08, + "loss": 0.6852, + "step": 44870 + }, + { + "epoch": 1.64, + "learning_rate": 3.586259933350423e-08, + "loss": 0.6913, + "step": 44871 + }, + { + "epoch": 1.64, + "learning_rate": 3.585893726883217e-08, + "loss": 0.7039, + "step": 44872 + }, + { + "epoch": 1.64, + "learning_rate": 3.5855275204160104e-08, + "loss": 0.6854, + "step": 44873 + }, + { + "epoch": 1.64, + "learning_rate": 3.5851613139488044e-08, + "loss": 0.6991, + "step": 44874 + }, + { + "epoch": 1.64, + "learning_rate": 3.584795107481598e-08, + "loss": 0.6906, + "step": 44875 + }, + { + "epoch": 1.64, + "learning_rate": 3.584428901014392e-08, + "loss": 0.6886, + "step": 44876 + }, + { + "epoch": 1.64, + "learning_rate": 3.584062694547186e-08, + "loss": 0.6983, + "step": 44877 + }, + { + "epoch": 1.64, + "learning_rate": 3.583696488079979e-08, + "loss": 0.6951, + "step": 44878 + }, + { + "epoch": 1.64, + "learning_rate": 3.583330281612773e-08, + "loss": 0.6941, + "step": 44879 + }, + { + "epoch": 1.64, + "learning_rate": 3.582964075145567e-08, + "loss": 0.6985, + "step": 44880 + }, + { + "epoch": 1.64, + "learning_rate": 3.582597868678361e-08, + "loss": 0.707, + "step": 44881 + }, + { + "epoch": 1.64, + "learning_rate": 3.5822316622111543e-08, + "loss": 0.6928, + "step": 44882 + }, + { + "epoch": 1.64, + "learning_rate": 3.5818654557439483e-08, + "loss": 0.6939, + "step": 44883 + }, + { + "epoch": 1.64, + "learning_rate": 3.581499249276742e-08, + "loss": 0.706, + "step": 44884 + }, + { + "epoch": 1.64, + "learning_rate": 3.581133042809536e-08, + "loss": 0.6806, + "step": 44885 + }, + { + "epoch": 1.64, + "learning_rate": 3.5807668363423296e-08, + "loss": 0.6897, + "step": 44886 + }, + { + "epoch": 1.64, + "learning_rate": 3.580400629875123e-08, + "loss": 0.6927, + "step": 44887 + }, + { + "epoch": 1.64, + "learning_rate": 3.580034423407917e-08, + "loss": 0.6946, + "step": 44888 + }, + { + "epoch": 1.64, + "learning_rate": 3.579668216940711e-08, + "loss": 0.6823, + "step": 44889 + }, + { + "epoch": 1.64, + "learning_rate": 3.579302010473505e-08, + "loss": 0.7022, + "step": 44890 + }, + { + "epoch": 1.64, + "learning_rate": 3.578935804006298e-08, + "loss": 0.7008, + "step": 44891 + }, + { + "epoch": 1.64, + "learning_rate": 3.578569597539092e-08, + "loss": 0.6881, + "step": 44892 + }, + { + "epoch": 1.64, + "learning_rate": 3.578203391071886e-08, + "loss": 0.6809, + "step": 44893 + }, + { + "epoch": 1.64, + "learning_rate": 3.57783718460468e-08, + "loss": 0.6923, + "step": 44894 + }, + { + "epoch": 1.64, + "learning_rate": 3.577470978137474e-08, + "loss": 0.6975, + "step": 44895 + }, + { + "epoch": 1.64, + "learning_rate": 3.5771047716702676e-08, + "loss": 0.691, + "step": 44896 + }, + { + "epoch": 1.64, + "learning_rate": 3.5767385652030616e-08, + "loss": 0.6914, + "step": 44897 + }, + { + "epoch": 1.64, + "learning_rate": 3.576372358735855e-08, + "loss": 0.6846, + "step": 44898 + }, + { + "epoch": 1.64, + "learning_rate": 3.576006152268649e-08, + "loss": 0.6849, + "step": 44899 + }, + { + "epoch": 1.64, + "learning_rate": 3.575639945801442e-08, + "loss": 0.7059, + "step": 44900 + }, + { + "epoch": 1.64, + "learning_rate": 3.575273739334236e-08, + "loss": 0.6914, + "step": 44901 + }, + { + "epoch": 1.64, + "learning_rate": 3.57490753286703e-08, + "loss": 0.6985, + "step": 44902 + }, + { + "epoch": 1.64, + "learning_rate": 3.574541326399824e-08, + "loss": 0.7014, + "step": 44903 + }, + { + "epoch": 1.64, + "learning_rate": 3.574175119932618e-08, + "loss": 0.6946, + "step": 44904 + }, + { + "epoch": 1.64, + "learning_rate": 3.5738089134654115e-08, + "loss": 0.6993, + "step": 44905 + }, + { + "epoch": 1.64, + "learning_rate": 3.5734427069982055e-08, + "loss": 0.709, + "step": 44906 + }, + { + "epoch": 1.64, + "learning_rate": 3.5730765005309995e-08, + "loss": 0.6895, + "step": 44907 + }, + { + "epoch": 1.64, + "learning_rate": 3.5727102940637935e-08, + "loss": 0.6992, + "step": 44908 + }, + { + "epoch": 1.64, + "learning_rate": 3.572344087596587e-08, + "loss": 0.6877, + "step": 44909 + }, + { + "epoch": 1.64, + "learning_rate": 3.57197788112938e-08, + "loss": 0.6911, + "step": 44910 + }, + { + "epoch": 1.64, + "learning_rate": 3.571611674662174e-08, + "loss": 0.691, + "step": 44911 + }, + { + "epoch": 1.64, + "learning_rate": 3.571245468194968e-08, + "loss": 0.6987, + "step": 44912 + }, + { + "epoch": 1.64, + "learning_rate": 3.570879261727762e-08, + "loss": 0.6867, + "step": 44913 + }, + { + "epoch": 1.64, + "learning_rate": 3.5705130552605555e-08, + "loss": 0.6803, + "step": 44914 + }, + { + "epoch": 1.64, + "learning_rate": 3.5701468487933495e-08, + "loss": 0.694, + "step": 44915 + }, + { + "epoch": 1.64, + "learning_rate": 3.5697806423261435e-08, + "loss": 0.6941, + "step": 44916 + }, + { + "epoch": 1.64, + "learning_rate": 3.5694144358589375e-08, + "loss": 0.6768, + "step": 44917 + }, + { + "epoch": 1.64, + "learning_rate": 3.569048229391731e-08, + "loss": 0.7063, + "step": 44918 + }, + { + "epoch": 1.64, + "learning_rate": 3.568682022924525e-08, + "loss": 0.6903, + "step": 44919 + }, + { + "epoch": 1.64, + "learning_rate": 3.568315816457319e-08, + "loss": 0.6959, + "step": 44920 + }, + { + "epoch": 1.64, + "learning_rate": 3.567949609990112e-08, + "loss": 0.6945, + "step": 44921 + }, + { + "epoch": 1.64, + "learning_rate": 3.567583403522906e-08, + "loss": 0.7053, + "step": 44922 + }, + { + "epoch": 1.64, + "learning_rate": 3.5672171970556994e-08, + "loss": 0.7001, + "step": 44923 + }, + { + "epoch": 1.64, + "learning_rate": 3.5668509905884934e-08, + "loss": 0.6857, + "step": 44924 + }, + { + "epoch": 1.64, + "learning_rate": 3.5664847841212874e-08, + "loss": 0.6885, + "step": 44925 + }, + { + "epoch": 1.64, + "learning_rate": 3.5661185776540814e-08, + "loss": 0.6924, + "step": 44926 + }, + { + "epoch": 1.64, + "learning_rate": 3.565752371186875e-08, + "loss": 0.7018, + "step": 44927 + }, + { + "epoch": 1.64, + "learning_rate": 3.565386164719669e-08, + "loss": 0.7072, + "step": 44928 + }, + { + "epoch": 1.64, + "learning_rate": 3.565019958252463e-08, + "loss": 0.6941, + "step": 44929 + }, + { + "epoch": 1.64, + "learning_rate": 3.564653751785257e-08, + "loss": 0.6902, + "step": 44930 + }, + { + "epoch": 1.64, + "learning_rate": 3.564287545318051e-08, + "loss": 0.6825, + "step": 44931 + }, + { + "epoch": 1.64, + "learning_rate": 3.563921338850844e-08, + "loss": 0.6976, + "step": 44932 + }, + { + "epoch": 1.64, + "learning_rate": 3.5635551323836374e-08, + "loss": 0.692, + "step": 44933 + }, + { + "epoch": 1.64, + "learning_rate": 3.5631889259164314e-08, + "loss": 0.6953, + "step": 44934 + }, + { + "epoch": 1.64, + "learning_rate": 3.5628227194492254e-08, + "loss": 0.6798, + "step": 44935 + }, + { + "epoch": 1.64, + "learning_rate": 3.562456512982019e-08, + "loss": 0.6877, + "step": 44936 + }, + { + "epoch": 1.64, + "learning_rate": 3.562090306514813e-08, + "loss": 0.7061, + "step": 44937 + }, + { + "epoch": 1.64, + "learning_rate": 3.561724100047607e-08, + "loss": 0.6996, + "step": 44938 + }, + { + "epoch": 1.64, + "learning_rate": 3.561357893580401e-08, + "loss": 0.6848, + "step": 44939 + }, + { + "epoch": 1.64, + "learning_rate": 3.560991687113195e-08, + "loss": 0.7125, + "step": 44940 + }, + { + "epoch": 1.64, + "learning_rate": 3.560625480645988e-08, + "loss": 0.6909, + "step": 44941 + }, + { + "epoch": 1.64, + "learning_rate": 3.560259274178782e-08, + "loss": 0.7009, + "step": 44942 + }, + { + "epoch": 1.64, + "learning_rate": 3.559893067711576e-08, + "loss": 0.6887, + "step": 44943 + }, + { + "epoch": 1.64, + "learning_rate": 3.559526861244369e-08, + "loss": 0.6916, + "step": 44944 + }, + { + "epoch": 1.64, + "learning_rate": 3.559160654777163e-08, + "loss": 0.688, + "step": 44945 + }, + { + "epoch": 1.64, + "learning_rate": 3.5587944483099566e-08, + "loss": 0.7, + "step": 44946 + }, + { + "epoch": 1.64, + "learning_rate": 3.5584282418427506e-08, + "loss": 0.6817, + "step": 44947 + }, + { + "epoch": 1.64, + "learning_rate": 3.5580620353755446e-08, + "loss": 0.6897, + "step": 44948 + }, + { + "epoch": 1.64, + "learning_rate": 3.5576958289083386e-08, + "loss": 0.6864, + "step": 44949 + }, + { + "epoch": 1.64, + "learning_rate": 3.557329622441132e-08, + "loss": 0.7046, + "step": 44950 + }, + { + "epoch": 1.64, + "learning_rate": 3.556963415973926e-08, + "loss": 0.689, + "step": 44951 + }, + { + "epoch": 1.64, + "learning_rate": 3.55659720950672e-08, + "loss": 0.698, + "step": 44952 + }, + { + "epoch": 1.64, + "learning_rate": 3.556231003039514e-08, + "loss": 0.6919, + "step": 44953 + }, + { + "epoch": 1.64, + "learning_rate": 3.555864796572307e-08, + "loss": 0.6982, + "step": 44954 + }, + { + "epoch": 1.64, + "learning_rate": 3.555498590105101e-08, + "loss": 0.6944, + "step": 44955 + }, + { + "epoch": 1.64, + "learning_rate": 3.5551323836378946e-08, + "loss": 0.7023, + "step": 44956 + }, + { + "epoch": 1.64, + "learning_rate": 3.5547661771706886e-08, + "loss": 0.6901, + "step": 44957 + }, + { + "epoch": 1.64, + "learning_rate": 3.554399970703482e-08, + "loss": 0.6902, + "step": 44958 + }, + { + "epoch": 1.64, + "learning_rate": 3.554033764236276e-08, + "loss": 0.6962, + "step": 44959 + }, + { + "epoch": 1.64, + "learning_rate": 3.55366755776907e-08, + "loss": 0.6888, + "step": 44960 + }, + { + "epoch": 1.64, + "learning_rate": 3.553301351301864e-08, + "loss": 0.7004, + "step": 44961 + }, + { + "epoch": 1.65, + "learning_rate": 3.552935144834658e-08, + "loss": 0.7009, + "step": 44962 + }, + { + "epoch": 1.65, + "learning_rate": 3.552568938367451e-08, + "loss": 0.6907, + "step": 44963 + }, + { + "epoch": 1.65, + "learning_rate": 3.552202731900245e-08, + "loss": 0.6967, + "step": 44964 + }, + { + "epoch": 1.65, + "learning_rate": 3.551836525433039e-08, + "loss": 0.6956, + "step": 44965 + }, + { + "epoch": 1.65, + "learning_rate": 3.551470318965833e-08, + "loss": 0.6933, + "step": 44966 + }, + { + "epoch": 1.65, + "learning_rate": 3.5511041124986265e-08, + "loss": 0.6948, + "step": 44967 + }, + { + "epoch": 1.65, + "learning_rate": 3.5507379060314205e-08, + "loss": 0.7125, + "step": 44968 + }, + { + "epoch": 1.65, + "learning_rate": 3.550371699564214e-08, + "loss": 0.693, + "step": 44969 + }, + { + "epoch": 1.65, + "learning_rate": 3.550005493097008e-08, + "loss": 0.6946, + "step": 44970 + }, + { + "epoch": 1.65, + "learning_rate": 3.549639286629802e-08, + "loss": 0.6966, + "step": 44971 + }, + { + "epoch": 1.65, + "learning_rate": 3.549273080162595e-08, + "loss": 0.6927, + "step": 44972 + }, + { + "epoch": 1.65, + "learning_rate": 3.548906873695389e-08, + "loss": 0.6731, + "step": 44973 + }, + { + "epoch": 1.65, + "learning_rate": 3.548540667228183e-08, + "loss": 0.6949, + "step": 44974 + }, + { + "epoch": 1.65, + "learning_rate": 3.548174460760977e-08, + "loss": 0.7036, + "step": 44975 + }, + { + "epoch": 1.65, + "learning_rate": 3.5478082542937705e-08, + "loss": 0.695, + "step": 44976 + }, + { + "epoch": 1.65, + "learning_rate": 3.5474420478265645e-08, + "loss": 0.6932, + "step": 44977 + }, + { + "epoch": 1.65, + "learning_rate": 3.5470758413593585e-08, + "loss": 0.6935, + "step": 44978 + }, + { + "epoch": 1.65, + "learning_rate": 3.5467096348921525e-08, + "loss": 0.6898, + "step": 44979 + }, + { + "epoch": 1.65, + "learning_rate": 3.546343428424946e-08, + "loss": 0.6895, + "step": 44980 + }, + { + "epoch": 1.65, + "learning_rate": 3.545977221957739e-08, + "loss": 0.6917, + "step": 44981 + }, + { + "epoch": 1.65, + "learning_rate": 3.545611015490533e-08, + "loss": 0.6964, + "step": 44982 + }, + { + "epoch": 1.65, + "learning_rate": 3.545244809023327e-08, + "loss": 0.6877, + "step": 44983 + }, + { + "epoch": 1.65, + "learning_rate": 3.544878602556121e-08, + "loss": 0.7, + "step": 44984 + }, + { + "epoch": 1.65, + "learning_rate": 3.5445123960889144e-08, + "loss": 0.6829, + "step": 44985 + }, + { + "epoch": 1.65, + "learning_rate": 3.5441461896217084e-08, + "loss": 0.6823, + "step": 44986 + }, + { + "epoch": 1.65, + "learning_rate": 3.5437799831545024e-08, + "loss": 0.6911, + "step": 44987 + }, + { + "epoch": 1.65, + "learning_rate": 3.5434137766872964e-08, + "loss": 0.6985, + "step": 44988 + }, + { + "epoch": 1.65, + "learning_rate": 3.5430475702200904e-08, + "loss": 0.7022, + "step": 44989 + }, + { + "epoch": 1.65, + "learning_rate": 3.542681363752884e-08, + "loss": 0.6958, + "step": 44990 + }, + { + "epoch": 1.65, + "learning_rate": 3.542315157285678e-08, + "loss": 0.6893, + "step": 44991 + }, + { + "epoch": 1.65, + "learning_rate": 3.541948950818471e-08, + "loss": 0.6937, + "step": 44992 + }, + { + "epoch": 1.65, + "learning_rate": 3.541582744351265e-08, + "loss": 0.6979, + "step": 44993 + }, + { + "epoch": 1.65, + "learning_rate": 3.5412165378840584e-08, + "loss": 0.693, + "step": 44994 + }, + { + "epoch": 1.65, + "learning_rate": 3.5408503314168524e-08, + "loss": 0.6943, + "step": 44995 + }, + { + "epoch": 1.65, + "learning_rate": 3.5404841249496464e-08, + "loss": 0.687, + "step": 44996 + }, + { + "epoch": 1.65, + "learning_rate": 3.5401179184824404e-08, + "loss": 0.6947, + "step": 44997 + }, + { + "epoch": 1.65, + "learning_rate": 3.5397517120152343e-08, + "loss": 0.6942, + "step": 44998 + }, + { + "epoch": 1.65, + "learning_rate": 3.539385505548028e-08, + "loss": 0.6875, + "step": 44999 + }, + { + "epoch": 1.65, + "learning_rate": 3.539019299080822e-08, + "loss": 0.6909, + "step": 45000 + }, + { + "epoch": 1.65, + "learning_rate": 3.5386530926136157e-08, + "loss": 0.6838, + "step": 45001 + }, + { + "epoch": 1.65, + "learning_rate": 3.5382868861464097e-08, + "loss": 0.6957, + "step": 45002 + }, + { + "epoch": 1.65, + "learning_rate": 3.537920679679203e-08, + "loss": 0.7002, + "step": 45003 + }, + { + "epoch": 1.65, + "learning_rate": 3.537554473211996e-08, + "loss": 0.6916, + "step": 45004 + }, + { + "epoch": 1.65, + "learning_rate": 3.53718826674479e-08, + "loss": 0.6953, + "step": 45005 + }, + { + "epoch": 1.65, + "learning_rate": 3.536822060277584e-08, + "loss": 0.6915, + "step": 45006 + }, + { + "epoch": 1.65, + "learning_rate": 3.536455853810378e-08, + "loss": 0.6986, + "step": 45007 + }, + { + "epoch": 1.65, + "learning_rate": 3.5360896473431716e-08, + "loss": 0.6947, + "step": 45008 + }, + { + "epoch": 1.65, + "learning_rate": 3.5357234408759656e-08, + "loss": 0.697, + "step": 45009 + }, + { + "epoch": 1.65, + "learning_rate": 3.5353572344087596e-08, + "loss": 0.6933, + "step": 45010 + }, + { + "epoch": 1.65, + "learning_rate": 3.5349910279415536e-08, + "loss": 0.6867, + "step": 45011 + }, + { + "epoch": 1.65, + "learning_rate": 3.534624821474347e-08, + "loss": 0.686, + "step": 45012 + }, + { + "epoch": 1.65, + "learning_rate": 3.534258615007141e-08, + "loss": 0.6867, + "step": 45013 + }, + { + "epoch": 1.65, + "learning_rate": 3.533892408539935e-08, + "loss": 0.7064, + "step": 45014 + }, + { + "epoch": 1.65, + "learning_rate": 3.533526202072728e-08, + "loss": 0.6986, + "step": 45015 + }, + { + "epoch": 1.65, + "learning_rate": 3.533159995605522e-08, + "loss": 0.6911, + "step": 45016 + }, + { + "epoch": 1.65, + "learning_rate": 3.5327937891383156e-08, + "loss": 0.6817, + "step": 45017 + }, + { + "epoch": 1.65, + "learning_rate": 3.5324275826711096e-08, + "loss": 0.7069, + "step": 45018 + }, + { + "epoch": 1.65, + "learning_rate": 3.5320613762039036e-08, + "loss": 0.6881, + "step": 45019 + }, + { + "epoch": 1.65, + "learning_rate": 3.5316951697366976e-08, + "loss": 0.7065, + "step": 45020 + }, + { + "epoch": 1.65, + "learning_rate": 3.531328963269491e-08, + "loss": 0.6907, + "step": 45021 + }, + { + "epoch": 1.65, + "learning_rate": 3.530962756802285e-08, + "loss": 0.6972, + "step": 45022 + }, + { + "epoch": 1.65, + "learning_rate": 3.530596550335079e-08, + "loss": 0.6873, + "step": 45023 + }, + { + "epoch": 1.65, + "learning_rate": 3.530230343867873e-08, + "loss": 0.7014, + "step": 45024 + }, + { + "epoch": 1.65, + "learning_rate": 3.529864137400667e-08, + "loss": 0.6861, + "step": 45025 + }, + { + "epoch": 1.65, + "learning_rate": 3.52949793093346e-08, + "loss": 0.6915, + "step": 45026 + }, + { + "epoch": 1.65, + "learning_rate": 3.5291317244662535e-08, + "loss": 0.692, + "step": 45027 + }, + { + "epoch": 1.65, + "learning_rate": 3.5287655179990475e-08, + "loss": 0.6838, + "step": 45028 + }, + { + "epoch": 1.65, + "learning_rate": 3.5283993115318415e-08, + "loss": 0.6951, + "step": 45029 + }, + { + "epoch": 1.65, + "learning_rate": 3.528033105064635e-08, + "loss": 0.7016, + "step": 45030 + }, + { + "epoch": 1.65, + "learning_rate": 3.527666898597429e-08, + "loss": 0.6942, + "step": 45031 + }, + { + "epoch": 1.65, + "learning_rate": 3.527300692130223e-08, + "loss": 0.6917, + "step": 45032 + }, + { + "epoch": 1.65, + "learning_rate": 3.526934485663017e-08, + "loss": 0.6863, + "step": 45033 + }, + { + "epoch": 1.65, + "learning_rate": 3.526568279195811e-08, + "loss": 0.6855, + "step": 45034 + }, + { + "epoch": 1.65, + "learning_rate": 3.526202072728604e-08, + "loss": 0.6919, + "step": 45035 + }, + { + "epoch": 1.65, + "learning_rate": 3.525835866261398e-08, + "loss": 0.6921, + "step": 45036 + }, + { + "epoch": 1.65, + "learning_rate": 3.525469659794192e-08, + "loss": 0.6955, + "step": 45037 + }, + { + "epoch": 1.65, + "learning_rate": 3.5251034533269855e-08, + "loss": 0.6952, + "step": 45038 + }, + { + "epoch": 1.65, + "learning_rate": 3.5247372468597795e-08, + "loss": 0.6857, + "step": 45039 + }, + { + "epoch": 1.65, + "learning_rate": 3.524371040392573e-08, + "loss": 0.6964, + "step": 45040 + }, + { + "epoch": 1.65, + "learning_rate": 3.524004833925367e-08, + "loss": 0.6836, + "step": 45041 + }, + { + "epoch": 1.65, + "learning_rate": 3.523638627458161e-08, + "loss": 0.6973, + "step": 45042 + }, + { + "epoch": 1.65, + "learning_rate": 3.523272420990955e-08, + "loss": 0.7037, + "step": 45043 + }, + { + "epoch": 1.65, + "learning_rate": 3.522906214523748e-08, + "loss": 0.6894, + "step": 45044 + }, + { + "epoch": 1.65, + "learning_rate": 3.522540008056542e-08, + "loss": 0.6809, + "step": 45045 + }, + { + "epoch": 1.65, + "learning_rate": 3.522173801589336e-08, + "loss": 0.6889, + "step": 45046 + }, + { + "epoch": 1.65, + "learning_rate": 3.52180759512213e-08, + "loss": 0.6935, + "step": 45047 + }, + { + "epoch": 1.65, + "learning_rate": 3.5214413886549234e-08, + "loss": 0.692, + "step": 45048 + }, + { + "epoch": 1.65, + "learning_rate": 3.5210751821877174e-08, + "loss": 0.6908, + "step": 45049 + }, + { + "epoch": 1.65, + "learning_rate": 3.5207089757205114e-08, + "loss": 0.7031, + "step": 45050 + }, + { + "epoch": 1.65, + "learning_rate": 3.520342769253305e-08, + "loss": 0.6843, + "step": 45051 + }, + { + "epoch": 1.65, + "learning_rate": 3.519976562786099e-08, + "loss": 0.7009, + "step": 45052 + }, + { + "epoch": 1.65, + "learning_rate": 3.519610356318892e-08, + "loss": 0.6977, + "step": 45053 + }, + { + "epoch": 1.65, + "learning_rate": 3.519244149851686e-08, + "loss": 0.6974, + "step": 45054 + }, + { + "epoch": 1.65, + "learning_rate": 3.51887794338448e-08, + "loss": 0.6836, + "step": 45055 + }, + { + "epoch": 1.65, + "learning_rate": 3.518511736917274e-08, + "loss": 0.7109, + "step": 45056 + }, + { + "epoch": 1.65, + "learning_rate": 3.5181455304500674e-08, + "loss": 0.6922, + "step": 45057 + }, + { + "epoch": 1.65, + "learning_rate": 3.5177793239828613e-08, + "loss": 0.6999, + "step": 45058 + }, + { + "epoch": 1.65, + "learning_rate": 3.5174131175156553e-08, + "loss": 0.6942, + "step": 45059 + }, + { + "epoch": 1.65, + "learning_rate": 3.5170469110484493e-08, + "loss": 0.6905, + "step": 45060 + }, + { + "epoch": 1.65, + "learning_rate": 3.516680704581243e-08, + "loss": 0.6898, + "step": 45061 + }, + { + "epoch": 1.65, + "learning_rate": 3.5163144981140367e-08, + "loss": 0.6941, + "step": 45062 + }, + { + "epoch": 1.65, + "learning_rate": 3.51594829164683e-08, + "loss": 0.6995, + "step": 45063 + }, + { + "epoch": 1.65, + "learning_rate": 3.515582085179624e-08, + "loss": 0.692, + "step": 45064 + }, + { + "epoch": 1.65, + "learning_rate": 3.515215878712418e-08, + "loss": 0.6891, + "step": 45065 + }, + { + "epoch": 1.65, + "learning_rate": 3.514849672245211e-08, + "loss": 0.6881, + "step": 45066 + }, + { + "epoch": 1.65, + "learning_rate": 3.514483465778005e-08, + "loss": 0.6936, + "step": 45067 + }, + { + "epoch": 1.65, + "learning_rate": 3.514117259310799e-08, + "loss": 0.6991, + "step": 45068 + }, + { + "epoch": 1.65, + "learning_rate": 3.513751052843593e-08, + "loss": 0.6968, + "step": 45069 + }, + { + "epoch": 1.65, + "learning_rate": 3.513384846376387e-08, + "loss": 0.6954, + "step": 45070 + }, + { + "epoch": 1.65, + "learning_rate": 3.5130186399091806e-08, + "loss": 0.7054, + "step": 45071 + }, + { + "epoch": 1.65, + "learning_rate": 3.5126524334419746e-08, + "loss": 0.6954, + "step": 45072 + }, + { + "epoch": 1.65, + "learning_rate": 3.5122862269747686e-08, + "loss": 0.6884, + "step": 45073 + }, + { + "epoch": 1.65, + "learning_rate": 3.511920020507562e-08, + "loss": 0.6885, + "step": 45074 + }, + { + "epoch": 1.65, + "learning_rate": 3.511553814040355e-08, + "loss": 0.6947, + "step": 45075 + }, + { + "epoch": 1.65, + "learning_rate": 3.511187607573149e-08, + "loss": 0.6742, + "step": 45076 + }, + { + "epoch": 1.65, + "learning_rate": 3.510821401105943e-08, + "loss": 0.6886, + "step": 45077 + }, + { + "epoch": 1.65, + "learning_rate": 3.510455194638737e-08, + "loss": 0.6832, + "step": 45078 + }, + { + "epoch": 1.65, + "learning_rate": 3.510088988171531e-08, + "loss": 0.7053, + "step": 45079 + }, + { + "epoch": 1.65, + "learning_rate": 3.5097227817043246e-08, + "loss": 0.7044, + "step": 45080 + }, + { + "epoch": 1.65, + "learning_rate": 3.5093565752371186e-08, + "loss": 0.702, + "step": 45081 + }, + { + "epoch": 1.65, + "learning_rate": 3.5089903687699125e-08, + "loss": 0.6947, + "step": 45082 + }, + { + "epoch": 1.65, + "learning_rate": 3.5086241623027065e-08, + "loss": 0.6967, + "step": 45083 + }, + { + "epoch": 1.65, + "learning_rate": 3.5082579558355e-08, + "loss": 0.6848, + "step": 45084 + }, + { + "epoch": 1.65, + "learning_rate": 3.507891749368294e-08, + "loss": 0.6958, + "step": 45085 + }, + { + "epoch": 1.65, + "learning_rate": 3.507525542901087e-08, + "loss": 0.6928, + "step": 45086 + }, + { + "epoch": 1.65, + "learning_rate": 3.507159336433881e-08, + "loss": 0.6917, + "step": 45087 + }, + { + "epoch": 1.65, + "learning_rate": 3.506793129966675e-08, + "loss": 0.6986, + "step": 45088 + }, + { + "epoch": 1.65, + "learning_rate": 3.5064269234994685e-08, + "loss": 0.6838, + "step": 45089 + }, + { + "epoch": 1.65, + "learning_rate": 3.5060607170322625e-08, + "loss": 0.7017, + "step": 45090 + }, + { + "epoch": 1.65, + "learning_rate": 3.5056945105650565e-08, + "loss": 0.7006, + "step": 45091 + }, + { + "epoch": 1.65, + "learning_rate": 3.5053283040978505e-08, + "loss": 0.6826, + "step": 45092 + }, + { + "epoch": 1.65, + "learning_rate": 3.504962097630644e-08, + "loss": 0.6917, + "step": 45093 + }, + { + "epoch": 1.65, + "learning_rate": 3.504595891163438e-08, + "loss": 0.6965, + "step": 45094 + }, + { + "epoch": 1.65, + "learning_rate": 3.504229684696232e-08, + "loss": 0.704, + "step": 45095 + }, + { + "epoch": 1.65, + "learning_rate": 3.503863478229026e-08, + "loss": 0.6954, + "step": 45096 + }, + { + "epoch": 1.65, + "learning_rate": 3.503497271761819e-08, + "loss": 0.6911, + "step": 45097 + }, + { + "epoch": 1.65, + "learning_rate": 3.5031310652946125e-08, + "loss": 0.697, + "step": 45098 + }, + { + "epoch": 1.65, + "learning_rate": 3.5027648588274065e-08, + "loss": 0.6958, + "step": 45099 + }, + { + "epoch": 1.65, + "learning_rate": 3.5023986523602004e-08, + "loss": 0.6975, + "step": 45100 + }, + { + "epoch": 1.65, + "learning_rate": 3.5020324458929944e-08, + "loss": 0.6965, + "step": 45101 + }, + { + "epoch": 1.65, + "learning_rate": 3.501666239425788e-08, + "loss": 0.6877, + "step": 45102 + }, + { + "epoch": 1.65, + "learning_rate": 3.501300032958582e-08, + "loss": 0.6937, + "step": 45103 + }, + { + "epoch": 1.65, + "learning_rate": 3.500933826491376e-08, + "loss": 0.7006, + "step": 45104 + }, + { + "epoch": 1.65, + "learning_rate": 3.50056762002417e-08, + "loss": 0.7084, + "step": 45105 + }, + { + "epoch": 1.65, + "learning_rate": 3.500201413556964e-08, + "loss": 0.7076, + "step": 45106 + }, + { + "epoch": 1.65, + "learning_rate": 3.499835207089757e-08, + "loss": 0.6878, + "step": 45107 + }, + { + "epoch": 1.65, + "learning_rate": 3.499469000622551e-08, + "loss": 0.6834, + "step": 45108 + }, + { + "epoch": 1.65, + "learning_rate": 3.4991027941553444e-08, + "loss": 0.6855, + "step": 45109 + }, + { + "epoch": 1.65, + "learning_rate": 3.4987365876881384e-08, + "loss": 0.7021, + "step": 45110 + }, + { + "epoch": 1.65, + "learning_rate": 3.498370381220932e-08, + "loss": 0.6893, + "step": 45111 + }, + { + "epoch": 1.65, + "learning_rate": 3.498004174753726e-08, + "loss": 0.6838, + "step": 45112 + }, + { + "epoch": 1.65, + "learning_rate": 3.49763796828652e-08, + "loss": 0.6903, + "step": 45113 + }, + { + "epoch": 1.65, + "learning_rate": 3.497271761819314e-08, + "loss": 0.6949, + "step": 45114 + }, + { + "epoch": 1.65, + "learning_rate": 3.496905555352108e-08, + "loss": 0.6923, + "step": 45115 + }, + { + "epoch": 1.65, + "learning_rate": 3.496539348884901e-08, + "loss": 0.6987, + "step": 45116 + }, + { + "epoch": 1.65, + "learning_rate": 3.496173142417695e-08, + "loss": 0.6885, + "step": 45117 + }, + { + "epoch": 1.65, + "learning_rate": 3.495806935950489e-08, + "loss": 0.6887, + "step": 45118 + }, + { + "epoch": 1.65, + "learning_rate": 3.495440729483283e-08, + "loss": 0.702, + "step": 45119 + }, + { + "epoch": 1.65, + "learning_rate": 3.4950745230160763e-08, + "loss": 0.6896, + "step": 45120 + }, + { + "epoch": 1.65, + "learning_rate": 3.49470831654887e-08, + "loss": 0.697, + "step": 45121 + }, + { + "epoch": 1.65, + "learning_rate": 3.4943421100816637e-08, + "loss": 0.6884, + "step": 45122 + }, + { + "epoch": 1.65, + "learning_rate": 3.4939759036144576e-08, + "loss": 0.687, + "step": 45123 + }, + { + "epoch": 1.65, + "learning_rate": 3.4936096971472516e-08, + "loss": 0.6987, + "step": 45124 + }, + { + "epoch": 1.65, + "learning_rate": 3.493243490680045e-08, + "loss": 0.6928, + "step": 45125 + }, + { + "epoch": 1.65, + "learning_rate": 3.492877284212839e-08, + "loss": 0.6913, + "step": 45126 + }, + { + "epoch": 1.65, + "learning_rate": 3.492511077745633e-08, + "loss": 0.6864, + "step": 45127 + }, + { + "epoch": 1.65, + "learning_rate": 3.492144871278427e-08, + "loss": 0.6988, + "step": 45128 + }, + { + "epoch": 1.65, + "learning_rate": 3.49177866481122e-08, + "loss": 0.6828, + "step": 45129 + }, + { + "epoch": 1.65, + "learning_rate": 3.491412458344014e-08, + "loss": 0.6979, + "step": 45130 + }, + { + "epoch": 1.65, + "learning_rate": 3.491046251876808e-08, + "loss": 0.6801, + "step": 45131 + }, + { + "epoch": 1.65, + "learning_rate": 3.4906800454096016e-08, + "loss": 0.679, + "step": 45132 + }, + { + "epoch": 1.65, + "learning_rate": 3.4903138389423956e-08, + "loss": 0.701, + "step": 45133 + }, + { + "epoch": 1.65, + "learning_rate": 3.489947632475189e-08, + "loss": 0.6911, + "step": 45134 + }, + { + "epoch": 1.65, + "learning_rate": 3.489581426007983e-08, + "loss": 0.692, + "step": 45135 + }, + { + "epoch": 1.65, + "learning_rate": 3.489215219540777e-08, + "loss": 0.6949, + "step": 45136 + }, + { + "epoch": 1.65, + "learning_rate": 3.488849013073571e-08, + "loss": 0.7025, + "step": 45137 + }, + { + "epoch": 1.65, + "learning_rate": 3.488482806606364e-08, + "loss": 0.6957, + "step": 45138 + }, + { + "epoch": 1.65, + "learning_rate": 3.488116600139158e-08, + "loss": 0.6843, + "step": 45139 + }, + { + "epoch": 1.65, + "learning_rate": 3.487750393671952e-08, + "loss": 0.7041, + "step": 45140 + }, + { + "epoch": 1.65, + "learning_rate": 3.487384187204746e-08, + "loss": 0.7071, + "step": 45141 + }, + { + "epoch": 1.65, + "learning_rate": 3.48701798073754e-08, + "loss": 0.6863, + "step": 45142 + }, + { + "epoch": 1.65, + "learning_rate": 3.4866517742703335e-08, + "loss": 0.6972, + "step": 45143 + }, + { + "epoch": 1.65, + "learning_rate": 3.4862855678031275e-08, + "loss": 0.694, + "step": 45144 + }, + { + "epoch": 1.65, + "learning_rate": 3.485919361335921e-08, + "loss": 0.6942, + "step": 45145 + }, + { + "epoch": 1.65, + "learning_rate": 3.485553154868715e-08, + "loss": 0.7015, + "step": 45146 + }, + { + "epoch": 1.65, + "learning_rate": 3.485186948401508e-08, + "loss": 0.7063, + "step": 45147 + }, + { + "epoch": 1.65, + "learning_rate": 3.484820741934302e-08, + "loss": 0.6985, + "step": 45148 + }, + { + "epoch": 1.65, + "learning_rate": 3.484454535467096e-08, + "loss": 0.6955, + "step": 45149 + }, + { + "epoch": 1.65, + "learning_rate": 3.48408832899989e-08, + "loss": 0.6936, + "step": 45150 + }, + { + "epoch": 1.65, + "learning_rate": 3.4837221225326835e-08, + "loss": 0.6956, + "step": 45151 + }, + { + "epoch": 1.65, + "learning_rate": 3.4833559160654775e-08, + "loss": 0.6879, + "step": 45152 + }, + { + "epoch": 1.65, + "learning_rate": 3.4829897095982715e-08, + "loss": 0.6787, + "step": 45153 + }, + { + "epoch": 1.65, + "learning_rate": 3.4826235031310655e-08, + "loss": 0.6994, + "step": 45154 + }, + { + "epoch": 1.65, + "learning_rate": 3.4822572966638595e-08, + "loss": 0.6937, + "step": 45155 + }, + { + "epoch": 1.65, + "learning_rate": 3.481891090196653e-08, + "loss": 0.6995, + "step": 45156 + }, + { + "epoch": 1.65, + "learning_rate": 3.481524883729446e-08, + "loss": 0.692, + "step": 45157 + }, + { + "epoch": 1.65, + "learning_rate": 3.48115867726224e-08, + "loss": 0.6828, + "step": 45158 + }, + { + "epoch": 1.65, + "learning_rate": 3.480792470795034e-08, + "loss": 0.6963, + "step": 45159 + }, + { + "epoch": 1.65, + "learning_rate": 3.4804262643278274e-08, + "loss": 0.6917, + "step": 45160 + }, + { + "epoch": 1.65, + "learning_rate": 3.4800600578606214e-08, + "loss": 0.6925, + "step": 45161 + }, + { + "epoch": 1.65, + "learning_rate": 3.4796938513934154e-08, + "loss": 0.6991, + "step": 45162 + }, + { + "epoch": 1.65, + "learning_rate": 3.4793276449262094e-08, + "loss": 0.6917, + "step": 45163 + }, + { + "epoch": 1.65, + "learning_rate": 3.4789614384590034e-08, + "loss": 0.6882, + "step": 45164 + }, + { + "epoch": 1.65, + "learning_rate": 3.478595231991797e-08, + "loss": 0.6892, + "step": 45165 + }, + { + "epoch": 1.65, + "learning_rate": 3.478229025524591e-08, + "loss": 0.6845, + "step": 45166 + }, + { + "epoch": 1.65, + "learning_rate": 3.477862819057385e-08, + "loss": 0.6993, + "step": 45167 + }, + { + "epoch": 1.65, + "learning_rate": 3.477496612590178e-08, + "loss": 0.7033, + "step": 45168 + }, + { + "epoch": 1.65, + "learning_rate": 3.4771304061229714e-08, + "loss": 0.6902, + "step": 45169 + }, + { + "epoch": 1.65, + "learning_rate": 3.4767641996557654e-08, + "loss": 0.6803, + "step": 45170 + }, + { + "epoch": 1.65, + "learning_rate": 3.4763979931885594e-08, + "loss": 0.7071, + "step": 45171 + }, + { + "epoch": 1.65, + "learning_rate": 3.4760317867213534e-08, + "loss": 0.6922, + "step": 45172 + }, + { + "epoch": 1.65, + "learning_rate": 3.4756655802541474e-08, + "loss": 0.6922, + "step": 45173 + }, + { + "epoch": 1.65, + "learning_rate": 3.475299373786941e-08, + "loss": 0.7025, + "step": 45174 + }, + { + "epoch": 1.65, + "learning_rate": 3.474933167319735e-08, + "loss": 0.7118, + "step": 45175 + }, + { + "epoch": 1.65, + "learning_rate": 3.474566960852529e-08, + "loss": 0.6949, + "step": 45176 + }, + { + "epoch": 1.65, + "learning_rate": 3.474200754385323e-08, + "loss": 0.6919, + "step": 45177 + }, + { + "epoch": 1.65, + "learning_rate": 3.473834547918116e-08, + "loss": 0.6926, + "step": 45178 + }, + { + "epoch": 1.65, + "learning_rate": 3.47346834145091e-08, + "loss": 0.6978, + "step": 45179 + }, + { + "epoch": 1.65, + "learning_rate": 3.4731021349837033e-08, + "loss": 0.6862, + "step": 45180 + }, + { + "epoch": 1.65, + "learning_rate": 3.472735928516497e-08, + "loss": 0.6934, + "step": 45181 + }, + { + "epoch": 1.65, + "learning_rate": 3.472369722049291e-08, + "loss": 0.7032, + "step": 45182 + }, + { + "epoch": 1.65, + "learning_rate": 3.4720035155820846e-08, + "loss": 0.6807, + "step": 45183 + }, + { + "epoch": 1.65, + "learning_rate": 3.4716373091148786e-08, + "loss": 0.6923, + "step": 45184 + }, + { + "epoch": 1.65, + "learning_rate": 3.4712711026476726e-08, + "loss": 0.6866, + "step": 45185 + }, + { + "epoch": 1.65, + "learning_rate": 3.4709048961804666e-08, + "loss": 0.6913, + "step": 45186 + }, + { + "epoch": 1.65, + "learning_rate": 3.47053868971326e-08, + "loss": 0.698, + "step": 45187 + }, + { + "epoch": 1.65, + "learning_rate": 3.470172483246054e-08, + "loss": 0.6932, + "step": 45188 + }, + { + "epoch": 1.65, + "learning_rate": 3.469806276778848e-08, + "loss": 0.6936, + "step": 45189 + }, + { + "epoch": 1.65, + "learning_rate": 3.469440070311642e-08, + "loss": 0.6969, + "step": 45190 + }, + { + "epoch": 1.65, + "learning_rate": 3.469073863844435e-08, + "loss": 0.6839, + "step": 45191 + }, + { + "epoch": 1.65, + "learning_rate": 3.4687076573772286e-08, + "loss": 0.6887, + "step": 45192 + }, + { + "epoch": 1.65, + "learning_rate": 3.4683414509100226e-08, + "loss": 0.6887, + "step": 45193 + }, + { + "epoch": 1.65, + "learning_rate": 3.4679752444428166e-08, + "loss": 0.689, + "step": 45194 + }, + { + "epoch": 1.65, + "learning_rate": 3.4676090379756106e-08, + "loss": 0.6962, + "step": 45195 + }, + { + "epoch": 1.65, + "learning_rate": 3.467242831508404e-08, + "loss": 0.6836, + "step": 45196 + }, + { + "epoch": 1.65, + "learning_rate": 3.466876625041198e-08, + "loss": 0.6978, + "step": 45197 + }, + { + "epoch": 1.65, + "learning_rate": 3.466510418573992e-08, + "loss": 0.6935, + "step": 45198 + }, + { + "epoch": 1.65, + "learning_rate": 3.466144212106786e-08, + "loss": 0.6897, + "step": 45199 + }, + { + "epoch": 1.65, + "learning_rate": 3.46577800563958e-08, + "loss": 0.6828, + "step": 45200 + }, + { + "epoch": 1.65, + "learning_rate": 3.465411799172373e-08, + "loss": 0.6949, + "step": 45201 + }, + { + "epoch": 1.65, + "learning_rate": 3.465045592705167e-08, + "loss": 0.6801, + "step": 45202 + }, + { + "epoch": 1.65, + "learning_rate": 3.4646793862379605e-08, + "loss": 0.6984, + "step": 45203 + }, + { + "epoch": 1.65, + "learning_rate": 3.4643131797707545e-08, + "loss": 0.6914, + "step": 45204 + }, + { + "epoch": 1.65, + "learning_rate": 3.463946973303548e-08, + "loss": 0.7074, + "step": 45205 + }, + { + "epoch": 1.65, + "learning_rate": 3.463580766836342e-08, + "loss": 0.6895, + "step": 45206 + }, + { + "epoch": 1.65, + "learning_rate": 3.463214560369136e-08, + "loss": 0.6922, + "step": 45207 + }, + { + "epoch": 1.65, + "learning_rate": 3.46284835390193e-08, + "loss": 0.6894, + "step": 45208 + }, + { + "epoch": 1.65, + "learning_rate": 3.462482147434724e-08, + "loss": 0.691, + "step": 45209 + }, + { + "epoch": 1.65, + "learning_rate": 3.462115940967517e-08, + "loss": 0.68, + "step": 45210 + }, + { + "epoch": 1.65, + "learning_rate": 3.461749734500311e-08, + "loss": 0.6965, + "step": 45211 + }, + { + "epoch": 1.65, + "learning_rate": 3.461383528033105e-08, + "loss": 0.7036, + "step": 45212 + }, + { + "epoch": 1.65, + "learning_rate": 3.461017321565899e-08, + "loss": 0.6932, + "step": 45213 + }, + { + "epoch": 1.65, + "learning_rate": 3.4606511150986925e-08, + "loss": 0.6952, + "step": 45214 + }, + { + "epoch": 1.65, + "learning_rate": 3.4602849086314865e-08, + "loss": 0.6959, + "step": 45215 + }, + { + "epoch": 1.65, + "learning_rate": 3.45991870216428e-08, + "loss": 0.6973, + "step": 45216 + }, + { + "epoch": 1.65, + "learning_rate": 3.459552495697074e-08, + "loss": 0.6886, + "step": 45217 + }, + { + "epoch": 1.65, + "learning_rate": 3.459186289229868e-08, + "loss": 0.6881, + "step": 45218 + }, + { + "epoch": 1.65, + "learning_rate": 3.458820082762661e-08, + "loss": 0.6937, + "step": 45219 + }, + { + "epoch": 1.65, + "learning_rate": 3.458453876295455e-08, + "loss": 0.6986, + "step": 45220 + }, + { + "epoch": 1.65, + "learning_rate": 3.458087669828249e-08, + "loss": 0.6961, + "step": 45221 + }, + { + "epoch": 1.65, + "learning_rate": 3.457721463361043e-08, + "loss": 0.6935, + "step": 45222 + }, + { + "epoch": 1.65, + "learning_rate": 3.4573552568938364e-08, + "loss": 0.6793, + "step": 45223 + }, + { + "epoch": 1.65, + "learning_rate": 3.4569890504266304e-08, + "loss": 0.6901, + "step": 45224 + }, + { + "epoch": 1.65, + "learning_rate": 3.4566228439594244e-08, + "loss": 0.691, + "step": 45225 + }, + { + "epoch": 1.65, + "learning_rate": 3.4562566374922184e-08, + "loss": 0.6979, + "step": 45226 + }, + { + "epoch": 1.65, + "learning_rate": 3.455890431025012e-08, + "loss": 0.6912, + "step": 45227 + }, + { + "epoch": 1.65, + "learning_rate": 3.455524224557805e-08, + "loss": 0.6855, + "step": 45228 + }, + { + "epoch": 1.65, + "learning_rate": 3.455158018090599e-08, + "loss": 0.6966, + "step": 45229 + }, + { + "epoch": 1.65, + "learning_rate": 3.454791811623393e-08, + "loss": 0.6921, + "step": 45230 + }, + { + "epoch": 1.65, + "learning_rate": 3.454425605156187e-08, + "loss": 0.6934, + "step": 45231 + }, + { + "epoch": 1.65, + "learning_rate": 3.4540593986889804e-08, + "loss": 0.7002, + "step": 45232 + }, + { + "epoch": 1.65, + "learning_rate": 3.4536931922217744e-08, + "loss": 0.6899, + "step": 45233 + }, + { + "epoch": 1.65, + "learning_rate": 3.4533269857545684e-08, + "loss": 0.6934, + "step": 45234 + }, + { + "epoch": 1.66, + "learning_rate": 3.4529607792873624e-08, + "loss": 0.7004, + "step": 45235 + }, + { + "epoch": 1.66, + "learning_rate": 3.4525945728201563e-08, + "loss": 0.7153, + "step": 45236 + }, + { + "epoch": 1.66, + "learning_rate": 3.45222836635295e-08, + "loss": 0.6948, + "step": 45237 + }, + { + "epoch": 1.66, + "learning_rate": 3.4518621598857437e-08, + "loss": 0.6923, + "step": 45238 + }, + { + "epoch": 1.66, + "learning_rate": 3.451495953418537e-08, + "loss": 0.6859, + "step": 45239 + }, + { + "epoch": 1.66, + "learning_rate": 3.451129746951331e-08, + "loss": 0.6878, + "step": 45240 + }, + { + "epoch": 1.66, + "learning_rate": 3.450763540484124e-08, + "loss": 0.698, + "step": 45241 + }, + { + "epoch": 1.66, + "learning_rate": 3.450397334016918e-08, + "loss": 0.6919, + "step": 45242 + }, + { + "epoch": 1.66, + "learning_rate": 3.450031127549712e-08, + "loss": 0.6872, + "step": 45243 + }, + { + "epoch": 1.66, + "learning_rate": 3.449664921082506e-08, + "loss": 0.7042, + "step": 45244 + }, + { + "epoch": 1.66, + "learning_rate": 3.4492987146153e-08, + "loss": 0.6899, + "step": 45245 + }, + { + "epoch": 1.66, + "learning_rate": 3.4489325081480936e-08, + "loss": 0.6845, + "step": 45246 + }, + { + "epoch": 1.66, + "learning_rate": 3.4485663016808876e-08, + "loss": 0.6816, + "step": 45247 + }, + { + "epoch": 1.66, + "learning_rate": 3.4482000952136816e-08, + "loss": 0.6986, + "step": 45248 + }, + { + "epoch": 1.66, + "learning_rate": 3.4478338887464756e-08, + "loss": 0.6907, + "step": 45249 + }, + { + "epoch": 1.66, + "learning_rate": 3.447467682279269e-08, + "loss": 0.7021, + "step": 45250 + }, + { + "epoch": 1.66, + "learning_rate": 3.447101475812062e-08, + "loss": 0.691, + "step": 45251 + }, + { + "epoch": 1.66, + "learning_rate": 3.446735269344856e-08, + "loss": 0.6877, + "step": 45252 + }, + { + "epoch": 1.66, + "learning_rate": 3.44636906287765e-08, + "loss": 0.6872, + "step": 45253 + }, + { + "epoch": 1.66, + "learning_rate": 3.446002856410444e-08, + "loss": 0.6939, + "step": 45254 + }, + { + "epoch": 1.66, + "learning_rate": 3.4456366499432376e-08, + "loss": 0.6875, + "step": 45255 + }, + { + "epoch": 1.66, + "learning_rate": 3.4452704434760316e-08, + "loss": 0.6878, + "step": 45256 + }, + { + "epoch": 1.66, + "learning_rate": 3.4449042370088256e-08, + "loss": 0.6836, + "step": 45257 + }, + { + "epoch": 1.66, + "learning_rate": 3.4445380305416196e-08, + "loss": 0.6971, + "step": 45258 + }, + { + "epoch": 1.66, + "learning_rate": 3.444171824074413e-08, + "loss": 0.6903, + "step": 45259 + }, + { + "epoch": 1.66, + "learning_rate": 3.443805617607207e-08, + "loss": 0.6926, + "step": 45260 + }, + { + "epoch": 1.66, + "learning_rate": 3.443439411140001e-08, + "loss": 0.692, + "step": 45261 + }, + { + "epoch": 1.66, + "learning_rate": 3.443073204672794e-08, + "loss": 0.6939, + "step": 45262 + }, + { + "epoch": 1.66, + "learning_rate": 3.442706998205588e-08, + "loss": 0.6966, + "step": 45263 + }, + { + "epoch": 1.66, + "learning_rate": 3.4423407917383815e-08, + "loss": 0.6906, + "step": 45264 + }, + { + "epoch": 1.66, + "learning_rate": 3.4419745852711755e-08, + "loss": 0.6933, + "step": 45265 + }, + { + "epoch": 1.66, + "learning_rate": 3.4416083788039695e-08, + "loss": 0.6827, + "step": 45266 + }, + { + "epoch": 1.66, + "learning_rate": 3.4412421723367635e-08, + "loss": 0.7005, + "step": 45267 + }, + { + "epoch": 1.66, + "learning_rate": 3.440875965869557e-08, + "loss": 0.6958, + "step": 45268 + }, + { + "epoch": 1.66, + "learning_rate": 3.440509759402351e-08, + "loss": 0.685, + "step": 45269 + }, + { + "epoch": 1.66, + "learning_rate": 3.440143552935145e-08, + "loss": 0.6935, + "step": 45270 + }, + { + "epoch": 1.66, + "learning_rate": 3.439777346467939e-08, + "loss": 0.6933, + "step": 45271 + }, + { + "epoch": 1.66, + "learning_rate": 3.439411140000733e-08, + "loss": 0.6865, + "step": 45272 + }, + { + "epoch": 1.66, + "learning_rate": 3.439044933533526e-08, + "loss": 0.6917, + "step": 45273 + }, + { + "epoch": 1.66, + "learning_rate": 3.4386787270663195e-08, + "loss": 0.7036, + "step": 45274 + }, + { + "epoch": 1.66, + "learning_rate": 3.4383125205991135e-08, + "loss": 0.7009, + "step": 45275 + }, + { + "epoch": 1.66, + "learning_rate": 3.4379463141319075e-08, + "loss": 0.6927, + "step": 45276 + }, + { + "epoch": 1.66, + "learning_rate": 3.437580107664701e-08, + "loss": 0.6819, + "step": 45277 + }, + { + "epoch": 1.66, + "learning_rate": 3.437213901197495e-08, + "loss": 0.6938, + "step": 45278 + }, + { + "epoch": 1.66, + "learning_rate": 3.436847694730289e-08, + "loss": 0.684, + "step": 45279 + }, + { + "epoch": 1.66, + "learning_rate": 3.436481488263083e-08, + "loss": 0.693, + "step": 45280 + }, + { + "epoch": 1.66, + "learning_rate": 3.436115281795877e-08, + "loss": 0.7009, + "step": 45281 + }, + { + "epoch": 1.66, + "learning_rate": 3.43574907532867e-08, + "loss": 0.6977, + "step": 45282 + }, + { + "epoch": 1.66, + "learning_rate": 3.435382868861464e-08, + "loss": 0.6881, + "step": 45283 + }, + { + "epoch": 1.66, + "learning_rate": 3.435016662394258e-08, + "loss": 0.6927, + "step": 45284 + }, + { + "epoch": 1.66, + "learning_rate": 3.4346504559270514e-08, + "loss": 0.6779, + "step": 45285 + }, + { + "epoch": 1.66, + "learning_rate": 3.4342842494598454e-08, + "loss": 0.6829, + "step": 45286 + }, + { + "epoch": 1.66, + "learning_rate": 3.433918042992639e-08, + "loss": 0.6885, + "step": 45287 + }, + { + "epoch": 1.66, + "learning_rate": 3.433551836525433e-08, + "loss": 0.6917, + "step": 45288 + }, + { + "epoch": 1.66, + "learning_rate": 3.433185630058227e-08, + "loss": 0.6939, + "step": 45289 + }, + { + "epoch": 1.66, + "learning_rate": 3.432819423591021e-08, + "loss": 0.6891, + "step": 45290 + }, + { + "epoch": 1.66, + "learning_rate": 3.432453217123814e-08, + "loss": 0.6972, + "step": 45291 + }, + { + "epoch": 1.66, + "learning_rate": 3.432087010656608e-08, + "loss": 0.6896, + "step": 45292 + }, + { + "epoch": 1.66, + "learning_rate": 3.431720804189402e-08, + "loss": 0.6849, + "step": 45293 + }, + { + "epoch": 1.66, + "learning_rate": 3.431354597722196e-08, + "loss": 0.6985, + "step": 45294 + }, + { + "epoch": 1.66, + "learning_rate": 3.4309883912549893e-08, + "loss": 0.6888, + "step": 45295 + }, + { + "epoch": 1.66, + "learning_rate": 3.4306221847877833e-08, + "loss": 0.6976, + "step": 45296 + }, + { + "epoch": 1.66, + "learning_rate": 3.4302559783205773e-08, + "loss": 0.6847, + "step": 45297 + }, + { + "epoch": 1.66, + "learning_rate": 3.4298897718533707e-08, + "loss": 0.693, + "step": 45298 + }, + { + "epoch": 1.66, + "learning_rate": 3.4295235653861647e-08, + "loss": 0.7021, + "step": 45299 + }, + { + "epoch": 1.66, + "learning_rate": 3.429157358918958e-08, + "loss": 0.6912, + "step": 45300 + }, + { + "epoch": 1.66, + "learning_rate": 3.428791152451752e-08, + "loss": 0.6896, + "step": 45301 + }, + { + "epoch": 1.66, + "learning_rate": 3.428424945984546e-08, + "loss": 0.6883, + "step": 45302 + }, + { + "epoch": 1.66, + "learning_rate": 3.42805873951734e-08, + "loss": 0.7053, + "step": 45303 + }, + { + "epoch": 1.66, + "learning_rate": 3.427692533050133e-08, + "loss": 0.6939, + "step": 45304 + }, + { + "epoch": 1.66, + "learning_rate": 3.427326326582927e-08, + "loss": 0.6907, + "step": 45305 + }, + { + "epoch": 1.66, + "learning_rate": 3.426960120115721e-08, + "loss": 0.6958, + "step": 45306 + }, + { + "epoch": 1.66, + "learning_rate": 3.426593913648515e-08, + "loss": 0.7118, + "step": 45307 + }, + { + "epoch": 1.66, + "learning_rate": 3.4262277071813086e-08, + "loss": 0.7016, + "step": 45308 + }, + { + "epoch": 1.66, + "learning_rate": 3.4258615007141026e-08, + "loss": 0.6914, + "step": 45309 + }, + { + "epoch": 1.66, + "learning_rate": 3.425495294246896e-08, + "loss": 0.6938, + "step": 45310 + }, + { + "epoch": 1.66, + "learning_rate": 3.42512908777969e-08, + "loss": 0.6999, + "step": 45311 + }, + { + "epoch": 1.66, + "learning_rate": 3.424762881312484e-08, + "loss": 0.7043, + "step": 45312 + }, + { + "epoch": 1.66, + "learning_rate": 3.424396674845277e-08, + "loss": 0.7011, + "step": 45313 + }, + { + "epoch": 1.66, + "learning_rate": 3.424030468378071e-08, + "loss": 0.6879, + "step": 45314 + }, + { + "epoch": 1.66, + "learning_rate": 3.423664261910865e-08, + "loss": 0.6857, + "step": 45315 + }, + { + "epoch": 1.66, + "learning_rate": 3.423298055443659e-08, + "loss": 0.6909, + "step": 45316 + }, + { + "epoch": 1.66, + "learning_rate": 3.422931848976453e-08, + "loss": 0.693, + "step": 45317 + }, + { + "epoch": 1.66, + "learning_rate": 3.4225656425092466e-08, + "loss": 0.6973, + "step": 45318 + }, + { + "epoch": 1.66, + "learning_rate": 3.4221994360420405e-08, + "loss": 0.687, + "step": 45319 + }, + { + "epoch": 1.66, + "learning_rate": 3.4218332295748345e-08, + "loss": 0.7013, + "step": 45320 + }, + { + "epoch": 1.66, + "learning_rate": 3.421467023107628e-08, + "loss": 0.7036, + "step": 45321 + }, + { + "epoch": 1.66, + "learning_rate": 3.421100816640421e-08, + "loss": 0.6926, + "step": 45322 + }, + { + "epoch": 1.66, + "learning_rate": 3.420734610173215e-08, + "loss": 0.6964, + "step": 45323 + }, + { + "epoch": 1.66, + "learning_rate": 3.420368403706009e-08, + "loss": 0.7021, + "step": 45324 + }, + { + "epoch": 1.66, + "learning_rate": 3.420002197238803e-08, + "loss": 0.696, + "step": 45325 + }, + { + "epoch": 1.66, + "learning_rate": 3.419635990771597e-08, + "loss": 0.6926, + "step": 45326 + }, + { + "epoch": 1.66, + "learning_rate": 3.4192697843043905e-08, + "loss": 0.6834, + "step": 45327 + }, + { + "epoch": 1.66, + "learning_rate": 3.4189035778371845e-08, + "loss": 0.6817, + "step": 45328 + }, + { + "epoch": 1.66, + "learning_rate": 3.4185373713699785e-08, + "loss": 0.6761, + "step": 45329 + }, + { + "epoch": 1.66, + "learning_rate": 3.4181711649027725e-08, + "loss": 0.6899, + "step": 45330 + }, + { + "epoch": 1.66, + "learning_rate": 3.417804958435566e-08, + "loss": 0.6812, + "step": 45331 + }, + { + "epoch": 1.66, + "learning_rate": 3.41743875196836e-08, + "loss": 0.6897, + "step": 45332 + }, + { + "epoch": 1.66, + "learning_rate": 3.417072545501153e-08, + "loss": 0.6834, + "step": 45333 + }, + { + "epoch": 1.66, + "learning_rate": 3.416706339033947e-08, + "loss": 0.6952, + "step": 45334 + }, + { + "epoch": 1.66, + "learning_rate": 3.416340132566741e-08, + "loss": 0.6943, + "step": 45335 + }, + { + "epoch": 1.66, + "learning_rate": 3.4159739260995345e-08, + "loss": 0.6915, + "step": 45336 + }, + { + "epoch": 1.66, + "learning_rate": 3.4156077196323284e-08, + "loss": 0.6874, + "step": 45337 + }, + { + "epoch": 1.66, + "learning_rate": 3.4152415131651224e-08, + "loss": 0.6977, + "step": 45338 + }, + { + "epoch": 1.66, + "learning_rate": 3.4148753066979164e-08, + "loss": 0.6999, + "step": 45339 + }, + { + "epoch": 1.66, + "learning_rate": 3.41450910023071e-08, + "loss": 0.6984, + "step": 45340 + }, + { + "epoch": 1.66, + "learning_rate": 3.414142893763504e-08, + "loss": 0.6931, + "step": 45341 + }, + { + "epoch": 1.66, + "learning_rate": 3.413776687296298e-08, + "loss": 0.6974, + "step": 45342 + }, + { + "epoch": 1.66, + "learning_rate": 3.413410480829092e-08, + "loss": 0.6946, + "step": 45343 + }, + { + "epoch": 1.66, + "learning_rate": 3.413044274361885e-08, + "loss": 0.6938, + "step": 45344 + }, + { + "epoch": 1.66, + "learning_rate": 3.4126780678946784e-08, + "loss": 0.6891, + "step": 45345 + }, + { + "epoch": 1.66, + "learning_rate": 3.4123118614274724e-08, + "loss": 0.6757, + "step": 45346 + }, + { + "epoch": 1.66, + "learning_rate": 3.4119456549602664e-08, + "loss": 0.6873, + "step": 45347 + }, + { + "epoch": 1.66, + "learning_rate": 3.4115794484930604e-08, + "loss": 0.6798, + "step": 45348 + }, + { + "epoch": 1.66, + "learning_rate": 3.411213242025854e-08, + "loss": 0.6996, + "step": 45349 + }, + { + "epoch": 1.66, + "learning_rate": 3.410847035558648e-08, + "loss": 0.6909, + "step": 45350 + }, + { + "epoch": 1.66, + "learning_rate": 3.410480829091442e-08, + "loss": 0.6852, + "step": 45351 + }, + { + "epoch": 1.66, + "learning_rate": 3.410114622624236e-08, + "loss": 0.6934, + "step": 45352 + }, + { + "epoch": 1.66, + "learning_rate": 3.409748416157029e-08, + "loss": 0.6966, + "step": 45353 + }, + { + "epoch": 1.66, + "learning_rate": 3.409382209689823e-08, + "loss": 0.7006, + "step": 45354 + }, + { + "epoch": 1.66, + "learning_rate": 3.409016003222617e-08, + "loss": 0.691, + "step": 45355 + }, + { + "epoch": 1.66, + "learning_rate": 3.4086497967554103e-08, + "loss": 0.6879, + "step": 45356 + }, + { + "epoch": 1.66, + "learning_rate": 3.4082835902882043e-08, + "loss": 0.693, + "step": 45357 + }, + { + "epoch": 1.66, + "learning_rate": 3.4079173838209977e-08, + "loss": 0.6976, + "step": 45358 + }, + { + "epoch": 1.66, + "learning_rate": 3.4075511773537917e-08, + "loss": 0.6984, + "step": 45359 + }, + { + "epoch": 1.66, + "learning_rate": 3.4071849708865857e-08, + "loss": 0.6936, + "step": 45360 + }, + { + "epoch": 1.66, + "learning_rate": 3.4068187644193796e-08, + "loss": 0.6917, + "step": 45361 + }, + { + "epoch": 1.66, + "learning_rate": 3.406452557952173e-08, + "loss": 0.7001, + "step": 45362 + }, + { + "epoch": 1.66, + "learning_rate": 3.406086351484967e-08, + "loss": 0.6959, + "step": 45363 + }, + { + "epoch": 1.66, + "learning_rate": 3.405720145017761e-08, + "loss": 0.6934, + "step": 45364 + }, + { + "epoch": 1.66, + "learning_rate": 3.405353938550555e-08, + "loss": 0.6981, + "step": 45365 + }, + { + "epoch": 1.66, + "learning_rate": 3.404987732083349e-08, + "loss": 0.7023, + "step": 45366 + }, + { + "epoch": 1.66, + "learning_rate": 3.404621525616142e-08, + "loss": 0.6866, + "step": 45367 + }, + { + "epoch": 1.66, + "learning_rate": 3.4042553191489356e-08, + "loss": 0.6846, + "step": 45368 + }, + { + "epoch": 1.66, + "learning_rate": 3.4038891126817296e-08, + "loss": 0.6958, + "step": 45369 + }, + { + "epoch": 1.66, + "learning_rate": 3.4035229062145236e-08, + "loss": 0.6887, + "step": 45370 + }, + { + "epoch": 1.66, + "learning_rate": 3.403156699747317e-08, + "loss": 0.7028, + "step": 45371 + }, + { + "epoch": 1.66, + "learning_rate": 3.402790493280111e-08, + "loss": 0.6957, + "step": 45372 + }, + { + "epoch": 1.66, + "learning_rate": 3.402424286812905e-08, + "loss": 0.6801, + "step": 45373 + }, + { + "epoch": 1.66, + "learning_rate": 3.402058080345699e-08, + "loss": 0.6907, + "step": 45374 + }, + { + "epoch": 1.66, + "learning_rate": 3.401691873878493e-08, + "loss": 0.6778, + "step": 45375 + }, + { + "epoch": 1.66, + "learning_rate": 3.401325667411286e-08, + "loss": 0.693, + "step": 45376 + }, + { + "epoch": 1.66, + "learning_rate": 3.40095946094408e-08, + "loss": 0.6934, + "step": 45377 + }, + { + "epoch": 1.66, + "learning_rate": 3.400593254476874e-08, + "loss": 0.6916, + "step": 45378 + }, + { + "epoch": 1.66, + "learning_rate": 3.4002270480096675e-08, + "loss": 0.6838, + "step": 45379 + }, + { + "epoch": 1.66, + "learning_rate": 3.3998608415424615e-08, + "loss": 0.689, + "step": 45380 + }, + { + "epoch": 1.66, + "learning_rate": 3.399494635075255e-08, + "loss": 0.6793, + "step": 45381 + }, + { + "epoch": 1.66, + "learning_rate": 3.399128428608049e-08, + "loss": 0.6861, + "step": 45382 + }, + { + "epoch": 1.66, + "learning_rate": 3.398762222140843e-08, + "loss": 0.6903, + "step": 45383 + }, + { + "epoch": 1.66, + "learning_rate": 3.398396015673637e-08, + "loss": 0.6961, + "step": 45384 + }, + { + "epoch": 1.66, + "learning_rate": 3.39802980920643e-08, + "loss": 0.6975, + "step": 45385 + }, + { + "epoch": 1.66, + "learning_rate": 3.397663602739224e-08, + "loss": 0.6879, + "step": 45386 + }, + { + "epoch": 1.66, + "learning_rate": 3.397297396272018e-08, + "loss": 0.6875, + "step": 45387 + }, + { + "epoch": 1.66, + "learning_rate": 3.396931189804812e-08, + "loss": 0.6933, + "step": 45388 + }, + { + "epoch": 1.66, + "learning_rate": 3.3965649833376055e-08, + "loss": 0.7032, + "step": 45389 + }, + { + "epoch": 1.66, + "learning_rate": 3.3961987768703995e-08, + "loss": 0.6935, + "step": 45390 + }, + { + "epoch": 1.66, + "learning_rate": 3.3958325704031935e-08, + "loss": 0.6867, + "step": 45391 + }, + { + "epoch": 1.66, + "learning_rate": 3.395466363935987e-08, + "loss": 0.6915, + "step": 45392 + }, + { + "epoch": 1.66, + "learning_rate": 3.395100157468781e-08, + "loss": 0.6869, + "step": 45393 + }, + { + "epoch": 1.66, + "learning_rate": 3.394733951001574e-08, + "loss": 0.6854, + "step": 45394 + }, + { + "epoch": 1.66, + "learning_rate": 3.394367744534368e-08, + "loss": 0.6985, + "step": 45395 + }, + { + "epoch": 1.66, + "learning_rate": 3.394001538067162e-08, + "loss": 0.6964, + "step": 45396 + }, + { + "epoch": 1.66, + "learning_rate": 3.393635331599956e-08, + "loss": 0.6943, + "step": 45397 + }, + { + "epoch": 1.66, + "learning_rate": 3.3932691251327494e-08, + "loss": 0.7117, + "step": 45398 + }, + { + "epoch": 1.66, + "learning_rate": 3.3929029186655434e-08, + "loss": 0.695, + "step": 45399 + }, + { + "epoch": 1.66, + "learning_rate": 3.3925367121983374e-08, + "loss": 0.6832, + "step": 45400 + }, + { + "epoch": 1.66, + "learning_rate": 3.3921705057311314e-08, + "loss": 0.697, + "step": 45401 + }, + { + "epoch": 1.66, + "learning_rate": 3.3918042992639254e-08, + "loss": 0.699, + "step": 45402 + }, + { + "epoch": 1.66, + "learning_rate": 3.391438092796719e-08, + "loss": 0.6894, + "step": 45403 + }, + { + "epoch": 1.66, + "learning_rate": 3.391071886329512e-08, + "loss": 0.6918, + "step": 45404 + }, + { + "epoch": 1.66, + "learning_rate": 3.390705679862306e-08, + "loss": 0.6878, + "step": 45405 + }, + { + "epoch": 1.66, + "learning_rate": 3.3903394733951e-08, + "loss": 0.695, + "step": 45406 + }, + { + "epoch": 1.66, + "learning_rate": 3.3899732669278934e-08, + "loss": 0.6911, + "step": 45407 + }, + { + "epoch": 1.66, + "learning_rate": 3.3896070604606874e-08, + "loss": 0.6913, + "step": 45408 + }, + { + "epoch": 1.66, + "learning_rate": 3.3892408539934814e-08, + "loss": 0.687, + "step": 45409 + }, + { + "epoch": 1.66, + "learning_rate": 3.3888746475262754e-08, + "loss": 0.6909, + "step": 45410 + }, + { + "epoch": 1.66, + "learning_rate": 3.3885084410590694e-08, + "loss": 0.6975, + "step": 45411 + }, + { + "epoch": 1.66, + "learning_rate": 3.388142234591863e-08, + "loss": 0.6931, + "step": 45412 + }, + { + "epoch": 1.66, + "learning_rate": 3.387776028124657e-08, + "loss": 0.683, + "step": 45413 + }, + { + "epoch": 1.66, + "learning_rate": 3.387409821657451e-08, + "loss": 0.6815, + "step": 45414 + }, + { + "epoch": 1.66, + "learning_rate": 3.387043615190244e-08, + "loss": 0.6972, + "step": 45415 + }, + { + "epoch": 1.66, + "learning_rate": 3.3866774087230373e-08, + "loss": 0.6922, + "step": 45416 + }, + { + "epoch": 1.66, + "learning_rate": 3.3863112022558313e-08, + "loss": 0.6838, + "step": 45417 + }, + { + "epoch": 1.66, + "learning_rate": 3.385944995788625e-08, + "loss": 0.6848, + "step": 45418 + }, + { + "epoch": 1.66, + "learning_rate": 3.385578789321419e-08, + "loss": 0.6812, + "step": 45419 + }, + { + "epoch": 1.66, + "learning_rate": 3.385212582854213e-08, + "loss": 0.6891, + "step": 45420 + }, + { + "epoch": 1.66, + "learning_rate": 3.3848463763870066e-08, + "loss": 0.683, + "step": 45421 + }, + { + "epoch": 1.66, + "learning_rate": 3.3844801699198006e-08, + "loss": 0.698, + "step": 45422 + }, + { + "epoch": 1.66, + "learning_rate": 3.3841139634525946e-08, + "loss": 0.6848, + "step": 45423 + }, + { + "epoch": 1.66, + "learning_rate": 3.3837477569853886e-08, + "loss": 0.7071, + "step": 45424 + }, + { + "epoch": 1.66, + "learning_rate": 3.383381550518182e-08, + "loss": 0.6902, + "step": 45425 + }, + { + "epoch": 1.66, + "learning_rate": 3.383015344050976e-08, + "loss": 0.6887, + "step": 45426 + }, + { + "epoch": 1.66, + "learning_rate": 3.382649137583769e-08, + "loss": 0.6956, + "step": 45427 + }, + { + "epoch": 1.66, + "learning_rate": 3.382282931116563e-08, + "loss": 0.6971, + "step": 45428 + }, + { + "epoch": 1.66, + "learning_rate": 3.381916724649357e-08, + "loss": 0.7042, + "step": 45429 + }, + { + "epoch": 1.66, + "learning_rate": 3.3815505181821506e-08, + "loss": 0.6848, + "step": 45430 + }, + { + "epoch": 1.66, + "learning_rate": 3.3811843117149446e-08, + "loss": 0.6983, + "step": 45431 + }, + { + "epoch": 1.66, + "learning_rate": 3.3808181052477386e-08, + "loss": 0.6978, + "step": 45432 + }, + { + "epoch": 1.66, + "learning_rate": 3.3804518987805326e-08, + "loss": 0.6911, + "step": 45433 + }, + { + "epoch": 1.66, + "learning_rate": 3.380085692313326e-08, + "loss": 0.6913, + "step": 45434 + }, + { + "epoch": 1.66, + "learning_rate": 3.37971948584612e-08, + "loss": 0.6962, + "step": 45435 + }, + { + "epoch": 1.66, + "learning_rate": 3.379353279378914e-08, + "loss": 0.6942, + "step": 45436 + }, + { + "epoch": 1.66, + "learning_rate": 3.378987072911708e-08, + "loss": 0.68, + "step": 45437 + }, + { + "epoch": 1.66, + "learning_rate": 3.378620866444501e-08, + "loss": 0.6946, + "step": 45438 + }, + { + "epoch": 1.66, + "learning_rate": 3.3782546599772945e-08, + "loss": 0.693, + "step": 45439 + }, + { + "epoch": 1.66, + "learning_rate": 3.3778884535100885e-08, + "loss": 0.6957, + "step": 45440 + }, + { + "epoch": 1.66, + "learning_rate": 3.3775222470428825e-08, + "loss": 0.6842, + "step": 45441 + }, + { + "epoch": 1.66, + "learning_rate": 3.3771560405756765e-08, + "loss": 0.6972, + "step": 45442 + }, + { + "epoch": 1.66, + "learning_rate": 3.37678983410847e-08, + "loss": 0.6938, + "step": 45443 + }, + { + "epoch": 1.66, + "learning_rate": 3.376423627641264e-08, + "loss": 0.6982, + "step": 45444 + }, + { + "epoch": 1.66, + "learning_rate": 3.376057421174058e-08, + "loss": 0.7128, + "step": 45445 + }, + { + "epoch": 1.66, + "learning_rate": 3.375691214706852e-08, + "loss": 0.6829, + "step": 45446 + }, + { + "epoch": 1.66, + "learning_rate": 3.375325008239646e-08, + "loss": 0.688, + "step": 45447 + }, + { + "epoch": 1.66, + "learning_rate": 3.374958801772439e-08, + "loss": 0.7006, + "step": 45448 + }, + { + "epoch": 1.66, + "learning_rate": 3.374592595305233e-08, + "loss": 0.6987, + "step": 45449 + }, + { + "epoch": 1.66, + "learning_rate": 3.3742263888380265e-08, + "loss": 0.6938, + "step": 45450 + }, + { + "epoch": 1.66, + "learning_rate": 3.3738601823708205e-08, + "loss": 0.6824, + "step": 45451 + }, + { + "epoch": 1.66, + "learning_rate": 3.373493975903614e-08, + "loss": 0.6976, + "step": 45452 + }, + { + "epoch": 1.66, + "learning_rate": 3.373127769436408e-08, + "loss": 0.6999, + "step": 45453 + }, + { + "epoch": 1.66, + "learning_rate": 3.372761562969202e-08, + "loss": 0.6946, + "step": 45454 + }, + { + "epoch": 1.66, + "learning_rate": 3.372395356501996e-08, + "loss": 0.7006, + "step": 45455 + }, + { + "epoch": 1.66, + "learning_rate": 3.37202915003479e-08, + "loss": 0.6855, + "step": 45456 + }, + { + "epoch": 1.66, + "learning_rate": 3.371662943567583e-08, + "loss": 0.6946, + "step": 45457 + }, + { + "epoch": 1.66, + "learning_rate": 3.371296737100377e-08, + "loss": 0.6927, + "step": 45458 + }, + { + "epoch": 1.66, + "learning_rate": 3.370930530633171e-08, + "loss": 0.6967, + "step": 45459 + }, + { + "epoch": 1.66, + "learning_rate": 3.370564324165965e-08, + "loss": 0.6786, + "step": 45460 + }, + { + "epoch": 1.66, + "learning_rate": 3.3701981176987584e-08, + "loss": 0.6972, + "step": 45461 + }, + { + "epoch": 1.66, + "learning_rate": 3.3698319112315524e-08, + "loss": 0.6977, + "step": 45462 + }, + { + "epoch": 1.66, + "learning_rate": 3.369465704764346e-08, + "loss": 0.6946, + "step": 45463 + }, + { + "epoch": 1.66, + "learning_rate": 3.36909949829714e-08, + "loss": 0.7002, + "step": 45464 + }, + { + "epoch": 1.66, + "learning_rate": 3.368733291829934e-08, + "loss": 0.6809, + "step": 45465 + }, + { + "epoch": 1.66, + "learning_rate": 3.368367085362727e-08, + "loss": 0.6962, + "step": 45466 + }, + { + "epoch": 1.66, + "learning_rate": 3.368000878895521e-08, + "loss": 0.6934, + "step": 45467 + }, + { + "epoch": 1.66, + "learning_rate": 3.367634672428315e-08, + "loss": 0.6964, + "step": 45468 + }, + { + "epoch": 1.66, + "learning_rate": 3.367268465961109e-08, + "loss": 0.6901, + "step": 45469 + }, + { + "epoch": 1.66, + "learning_rate": 3.3669022594939024e-08, + "loss": 0.6957, + "step": 45470 + }, + { + "epoch": 1.66, + "learning_rate": 3.3665360530266964e-08, + "loss": 0.6948, + "step": 45471 + }, + { + "epoch": 1.66, + "learning_rate": 3.3661698465594904e-08, + "loss": 0.6913, + "step": 45472 + }, + { + "epoch": 1.66, + "learning_rate": 3.3658036400922843e-08, + "loss": 0.6918, + "step": 45473 + }, + { + "epoch": 1.66, + "learning_rate": 3.365437433625078e-08, + "loss": 0.682, + "step": 45474 + }, + { + "epoch": 1.66, + "learning_rate": 3.365071227157871e-08, + "loss": 0.6905, + "step": 45475 + }, + { + "epoch": 1.66, + "learning_rate": 3.364705020690665e-08, + "loss": 0.6986, + "step": 45476 + }, + { + "epoch": 1.66, + "learning_rate": 3.364338814223459e-08, + "loss": 0.6891, + "step": 45477 + }, + { + "epoch": 1.66, + "learning_rate": 3.363972607756253e-08, + "loss": 0.6902, + "step": 45478 + }, + { + "epoch": 1.66, + "learning_rate": 3.363606401289046e-08, + "loss": 0.6946, + "step": 45479 + }, + { + "epoch": 1.66, + "learning_rate": 3.36324019482184e-08, + "loss": 0.6908, + "step": 45480 + }, + { + "epoch": 1.66, + "learning_rate": 3.362873988354634e-08, + "loss": 0.7062, + "step": 45481 + }, + { + "epoch": 1.66, + "learning_rate": 3.362507781887428e-08, + "loss": 0.6887, + "step": 45482 + }, + { + "epoch": 1.66, + "learning_rate": 3.362141575420222e-08, + "loss": 0.6949, + "step": 45483 + }, + { + "epoch": 1.66, + "learning_rate": 3.3617753689530156e-08, + "loss": 0.6956, + "step": 45484 + }, + { + "epoch": 1.66, + "learning_rate": 3.3614091624858096e-08, + "loss": 0.7056, + "step": 45485 + }, + { + "epoch": 1.66, + "learning_rate": 3.361042956018603e-08, + "loss": 0.6886, + "step": 45486 + }, + { + "epoch": 1.66, + "learning_rate": 3.360676749551397e-08, + "loss": 0.6905, + "step": 45487 + }, + { + "epoch": 1.66, + "learning_rate": 3.36031054308419e-08, + "loss": 0.6942, + "step": 45488 + }, + { + "epoch": 1.66, + "learning_rate": 3.359944336616984e-08, + "loss": 0.6956, + "step": 45489 + }, + { + "epoch": 1.66, + "learning_rate": 3.359578130149778e-08, + "loss": 0.6868, + "step": 45490 + }, + { + "epoch": 1.66, + "learning_rate": 3.359211923682572e-08, + "loss": 0.6922, + "step": 45491 + }, + { + "epoch": 1.66, + "learning_rate": 3.358845717215366e-08, + "loss": 0.6877, + "step": 45492 + }, + { + "epoch": 1.66, + "learning_rate": 3.3584795107481596e-08, + "loss": 0.7037, + "step": 45493 + }, + { + "epoch": 1.66, + "learning_rate": 3.3581133042809536e-08, + "loss": 0.6966, + "step": 45494 + }, + { + "epoch": 1.66, + "learning_rate": 3.3577470978137476e-08, + "loss": 0.6899, + "step": 45495 + }, + { + "epoch": 1.66, + "learning_rate": 3.3573808913465415e-08, + "loss": 0.6844, + "step": 45496 + }, + { + "epoch": 1.66, + "learning_rate": 3.357014684879335e-08, + "loss": 0.6868, + "step": 45497 + }, + { + "epoch": 1.66, + "learning_rate": 3.356648478412128e-08, + "loss": 0.6937, + "step": 45498 + }, + { + "epoch": 1.66, + "learning_rate": 3.356282271944922e-08, + "loss": 0.6999, + "step": 45499 + }, + { + "epoch": 1.66, + "learning_rate": 3.355916065477716e-08, + "loss": 0.7012, + "step": 45500 + }, + { + "epoch": 1.66, + "learning_rate": 3.35554985901051e-08, + "loss": 0.7018, + "step": 45501 + }, + { + "epoch": 1.66, + "learning_rate": 3.3551836525433035e-08, + "loss": 0.6895, + "step": 45502 + }, + { + "epoch": 1.66, + "learning_rate": 3.3548174460760975e-08, + "loss": 0.7025, + "step": 45503 + }, + { + "epoch": 1.66, + "learning_rate": 3.3544512396088915e-08, + "loss": 0.685, + "step": 45504 + }, + { + "epoch": 1.66, + "learning_rate": 3.3540850331416855e-08, + "loss": 0.6922, + "step": 45505 + }, + { + "epoch": 1.66, + "learning_rate": 3.353718826674479e-08, + "loss": 0.6945, + "step": 45506 + }, + { + "epoch": 1.66, + "learning_rate": 3.353352620207273e-08, + "loss": 0.6956, + "step": 45507 + }, + { + "epoch": 1.67, + "learning_rate": 3.352986413740067e-08, + "loss": 0.6972, + "step": 45508 + }, + { + "epoch": 1.67, + "learning_rate": 3.35262020727286e-08, + "loss": 0.6908, + "step": 45509 + }, + { + "epoch": 1.67, + "learning_rate": 3.352254000805654e-08, + "loss": 0.6994, + "step": 45510 + }, + { + "epoch": 1.67, + "learning_rate": 3.3518877943384475e-08, + "loss": 0.6793, + "step": 45511 + }, + { + "epoch": 1.67, + "learning_rate": 3.3515215878712415e-08, + "loss": 0.6852, + "step": 45512 + }, + { + "epoch": 1.67, + "learning_rate": 3.3511553814040355e-08, + "loss": 0.6893, + "step": 45513 + }, + { + "epoch": 1.67, + "learning_rate": 3.3507891749368294e-08, + "loss": 0.6932, + "step": 45514 + }, + { + "epoch": 1.67, + "learning_rate": 3.350422968469623e-08, + "loss": 0.7021, + "step": 45515 + }, + { + "epoch": 1.67, + "learning_rate": 3.350056762002417e-08, + "loss": 0.6973, + "step": 45516 + }, + { + "epoch": 1.67, + "learning_rate": 3.349690555535211e-08, + "loss": 0.6798, + "step": 45517 + }, + { + "epoch": 1.67, + "learning_rate": 3.349324349068005e-08, + "loss": 0.6981, + "step": 45518 + }, + { + "epoch": 1.67, + "learning_rate": 3.348958142600799e-08, + "loss": 0.697, + "step": 45519 + }, + { + "epoch": 1.67, + "learning_rate": 3.348591936133592e-08, + "loss": 0.6959, + "step": 45520 + }, + { + "epoch": 1.67, + "learning_rate": 3.3482257296663854e-08, + "loss": 0.6978, + "step": 45521 + }, + { + "epoch": 1.67, + "learning_rate": 3.3478595231991794e-08, + "loss": 0.7016, + "step": 45522 + }, + { + "epoch": 1.67, + "learning_rate": 3.3474933167319734e-08, + "loss": 0.7016, + "step": 45523 + }, + { + "epoch": 1.67, + "learning_rate": 3.347127110264767e-08, + "loss": 0.6904, + "step": 45524 + }, + { + "epoch": 1.67, + "learning_rate": 3.346760903797561e-08, + "loss": 0.6999, + "step": 45525 + }, + { + "epoch": 1.67, + "learning_rate": 3.346394697330355e-08, + "loss": 0.6964, + "step": 45526 + }, + { + "epoch": 1.67, + "learning_rate": 3.346028490863149e-08, + "loss": 0.6841, + "step": 45527 + }, + { + "epoch": 1.67, + "learning_rate": 3.345662284395943e-08, + "loss": 0.6964, + "step": 45528 + }, + { + "epoch": 1.67, + "learning_rate": 3.345296077928736e-08, + "loss": 0.6928, + "step": 45529 + }, + { + "epoch": 1.67, + "learning_rate": 3.34492987146153e-08, + "loss": 0.6966, + "step": 45530 + }, + { + "epoch": 1.67, + "learning_rate": 3.344563664994324e-08, + "loss": 0.7109, + "step": 45531 + }, + { + "epoch": 1.67, + "learning_rate": 3.3441974585271173e-08, + "loss": 0.6901, + "step": 45532 + }, + { + "epoch": 1.67, + "learning_rate": 3.343831252059911e-08, + "loss": 0.6776, + "step": 45533 + }, + { + "epoch": 1.67, + "learning_rate": 3.343465045592705e-08, + "loss": 0.6956, + "step": 45534 + }, + { + "epoch": 1.67, + "learning_rate": 3.3430988391254987e-08, + "loss": 0.6932, + "step": 45535 + }, + { + "epoch": 1.67, + "learning_rate": 3.3427326326582927e-08, + "loss": 0.6842, + "step": 45536 + }, + { + "epoch": 1.67, + "learning_rate": 3.342366426191086e-08, + "loss": 0.6977, + "step": 45537 + }, + { + "epoch": 1.67, + "learning_rate": 3.34200021972388e-08, + "loss": 0.7056, + "step": 45538 + }, + { + "epoch": 1.67, + "learning_rate": 3.341634013256674e-08, + "loss": 0.6982, + "step": 45539 + }, + { + "epoch": 1.67, + "learning_rate": 3.341267806789468e-08, + "loss": 0.6851, + "step": 45540 + }, + { + "epoch": 1.67, + "learning_rate": 3.340901600322262e-08, + "loss": 0.6906, + "step": 45541 + }, + { + "epoch": 1.67, + "learning_rate": 3.340535393855055e-08, + "loss": 0.6931, + "step": 45542 + }, + { + "epoch": 1.67, + "learning_rate": 3.340169187387849e-08, + "loss": 0.6912, + "step": 45543 + }, + { + "epoch": 1.67, + "learning_rate": 3.3398029809206426e-08, + "loss": 0.7118, + "step": 45544 + }, + { + "epoch": 1.67, + "learning_rate": 3.3394367744534366e-08, + "loss": 0.6948, + "step": 45545 + }, + { + "epoch": 1.67, + "learning_rate": 3.33907056798623e-08, + "loss": 0.7041, + "step": 45546 + }, + { + "epoch": 1.67, + "learning_rate": 3.338704361519024e-08, + "loss": 0.6969, + "step": 45547 + }, + { + "epoch": 1.67, + "learning_rate": 3.338338155051818e-08, + "loss": 0.6996, + "step": 45548 + }, + { + "epoch": 1.67, + "learning_rate": 3.337971948584612e-08, + "loss": 0.6981, + "step": 45549 + }, + { + "epoch": 1.67, + "learning_rate": 3.337605742117406e-08, + "loss": 0.6945, + "step": 45550 + }, + { + "epoch": 1.67, + "learning_rate": 3.337239535650199e-08, + "loss": 0.6945, + "step": 45551 + }, + { + "epoch": 1.67, + "learning_rate": 3.336873329182993e-08, + "loss": 0.6831, + "step": 45552 + }, + { + "epoch": 1.67, + "learning_rate": 3.336507122715787e-08, + "loss": 0.6905, + "step": 45553 + }, + { + "epoch": 1.67, + "learning_rate": 3.336140916248581e-08, + "loss": 0.6883, + "step": 45554 + }, + { + "epoch": 1.67, + "learning_rate": 3.3357747097813746e-08, + "loss": 0.6848, + "step": 45555 + }, + { + "epoch": 1.67, + "learning_rate": 3.3354085033141685e-08, + "loss": 0.6932, + "step": 45556 + }, + { + "epoch": 1.67, + "learning_rate": 3.335042296846962e-08, + "loss": 0.7037, + "step": 45557 + }, + { + "epoch": 1.67, + "learning_rate": 3.334676090379756e-08, + "loss": 0.7016, + "step": 45558 + }, + { + "epoch": 1.67, + "learning_rate": 3.33430988391255e-08, + "loss": 0.7011, + "step": 45559 + }, + { + "epoch": 1.67, + "learning_rate": 3.333943677445343e-08, + "loss": 0.6854, + "step": 45560 + }, + { + "epoch": 1.67, + "learning_rate": 3.333577470978137e-08, + "loss": 0.7004, + "step": 45561 + }, + { + "epoch": 1.67, + "learning_rate": 3.333211264510931e-08, + "loss": 0.6946, + "step": 45562 + }, + { + "epoch": 1.67, + "learning_rate": 3.332845058043725e-08, + "loss": 0.6957, + "step": 45563 + }, + { + "epoch": 1.67, + "learning_rate": 3.3324788515765185e-08, + "loss": 0.6907, + "step": 45564 + }, + { + "epoch": 1.67, + "learning_rate": 3.3321126451093125e-08, + "loss": 0.6964, + "step": 45565 + }, + { + "epoch": 1.67, + "learning_rate": 3.3317464386421065e-08, + "loss": 0.6932, + "step": 45566 + }, + { + "epoch": 1.67, + "learning_rate": 3.3313802321749005e-08, + "loss": 0.7003, + "step": 45567 + }, + { + "epoch": 1.67, + "learning_rate": 3.331014025707694e-08, + "loss": 0.6979, + "step": 45568 + }, + { + "epoch": 1.67, + "learning_rate": 3.330647819240487e-08, + "loss": 0.6911, + "step": 45569 + }, + { + "epoch": 1.67, + "learning_rate": 3.330281612773281e-08, + "loss": 0.6907, + "step": 45570 + }, + { + "epoch": 1.67, + "learning_rate": 3.329915406306075e-08, + "loss": 0.6929, + "step": 45571 + }, + { + "epoch": 1.67, + "learning_rate": 3.329549199838869e-08, + "loss": 0.6931, + "step": 45572 + }, + { + "epoch": 1.67, + "learning_rate": 3.3291829933716625e-08, + "loss": 0.6817, + "step": 45573 + }, + { + "epoch": 1.67, + "learning_rate": 3.3288167869044564e-08, + "loss": 0.7163, + "step": 45574 + }, + { + "epoch": 1.67, + "learning_rate": 3.3284505804372504e-08, + "loss": 0.6895, + "step": 45575 + }, + { + "epoch": 1.67, + "learning_rate": 3.3280843739700444e-08, + "loss": 0.6928, + "step": 45576 + }, + { + "epoch": 1.67, + "learning_rate": 3.3277181675028384e-08, + "loss": 0.7022, + "step": 45577 + }, + { + "epoch": 1.67, + "learning_rate": 3.327351961035632e-08, + "loss": 0.6869, + "step": 45578 + }, + { + "epoch": 1.67, + "learning_rate": 3.326985754568426e-08, + "loss": 0.6839, + "step": 45579 + }, + { + "epoch": 1.67, + "learning_rate": 3.326619548101219e-08, + "loss": 0.6982, + "step": 45580 + }, + { + "epoch": 1.67, + "learning_rate": 3.326253341634013e-08, + "loss": 0.7006, + "step": 45581 + }, + { + "epoch": 1.67, + "learning_rate": 3.3258871351668064e-08, + "loss": 0.6889, + "step": 45582 + }, + { + "epoch": 1.67, + "learning_rate": 3.3255209286996004e-08, + "loss": 0.6911, + "step": 45583 + }, + { + "epoch": 1.67, + "learning_rate": 3.3251547222323944e-08, + "loss": 0.6884, + "step": 45584 + }, + { + "epoch": 1.67, + "learning_rate": 3.3247885157651884e-08, + "loss": 0.6891, + "step": 45585 + }, + { + "epoch": 1.67, + "learning_rate": 3.3244223092979824e-08, + "loss": 0.6981, + "step": 45586 + }, + { + "epoch": 1.67, + "learning_rate": 3.324056102830776e-08, + "loss": 0.6984, + "step": 45587 + }, + { + "epoch": 1.67, + "learning_rate": 3.32368989636357e-08, + "loss": 0.6892, + "step": 45588 + }, + { + "epoch": 1.67, + "learning_rate": 3.323323689896364e-08, + "loss": 0.697, + "step": 45589 + }, + { + "epoch": 1.67, + "learning_rate": 3.322957483429158e-08, + "loss": 0.6879, + "step": 45590 + }, + { + "epoch": 1.67, + "learning_rate": 3.322591276961951e-08, + "loss": 0.7008, + "step": 45591 + }, + { + "epoch": 1.67, + "learning_rate": 3.3222250704947443e-08, + "loss": 0.7027, + "step": 45592 + }, + { + "epoch": 1.67, + "learning_rate": 3.3218588640275383e-08, + "loss": 0.6897, + "step": 45593 + }, + { + "epoch": 1.67, + "learning_rate": 3.3214926575603323e-08, + "loss": 0.7018, + "step": 45594 + }, + { + "epoch": 1.67, + "learning_rate": 3.321126451093126e-08, + "loss": 0.6858, + "step": 45595 + }, + { + "epoch": 1.67, + "learning_rate": 3.3207602446259197e-08, + "loss": 0.7004, + "step": 45596 + }, + { + "epoch": 1.67, + "learning_rate": 3.3203940381587137e-08, + "loss": 0.6935, + "step": 45597 + }, + { + "epoch": 1.67, + "learning_rate": 3.3200278316915076e-08, + "loss": 0.7034, + "step": 45598 + }, + { + "epoch": 1.67, + "learning_rate": 3.3196616252243016e-08, + "loss": 0.6978, + "step": 45599 + }, + { + "epoch": 1.67, + "learning_rate": 3.319295418757095e-08, + "loss": 0.6897, + "step": 45600 + }, + { + "epoch": 1.67, + "learning_rate": 3.318929212289889e-08, + "loss": 0.688, + "step": 45601 + }, + { + "epoch": 1.67, + "learning_rate": 3.318563005822683e-08, + "loss": 0.6916, + "step": 45602 + }, + { + "epoch": 1.67, + "learning_rate": 3.318196799355476e-08, + "loss": 0.7019, + "step": 45603 + }, + { + "epoch": 1.67, + "learning_rate": 3.31783059288827e-08, + "loss": 0.6961, + "step": 45604 + }, + { + "epoch": 1.67, + "learning_rate": 3.3174643864210636e-08, + "loss": 0.6911, + "step": 45605 + }, + { + "epoch": 1.67, + "learning_rate": 3.3170981799538576e-08, + "loss": 0.6975, + "step": 45606 + }, + { + "epoch": 1.67, + "learning_rate": 3.3167319734866516e-08, + "loss": 0.6904, + "step": 45607 + }, + { + "epoch": 1.67, + "learning_rate": 3.3163657670194456e-08, + "loss": 0.6906, + "step": 45608 + }, + { + "epoch": 1.67, + "learning_rate": 3.315999560552239e-08, + "loss": 0.6798, + "step": 45609 + }, + { + "epoch": 1.67, + "learning_rate": 3.315633354085033e-08, + "loss": 0.6923, + "step": 45610 + }, + { + "epoch": 1.67, + "learning_rate": 3.315267147617827e-08, + "loss": 0.69, + "step": 45611 + }, + { + "epoch": 1.67, + "learning_rate": 3.314900941150621e-08, + "loss": 0.6887, + "step": 45612 + }, + { + "epoch": 1.67, + "learning_rate": 3.314534734683415e-08, + "loss": 0.6963, + "step": 45613 + }, + { + "epoch": 1.67, + "learning_rate": 3.314168528216208e-08, + "loss": 0.6999, + "step": 45614 + }, + { + "epoch": 1.67, + "learning_rate": 3.3138023217490016e-08, + "loss": 0.6843, + "step": 45615 + }, + { + "epoch": 1.67, + "learning_rate": 3.3134361152817955e-08, + "loss": 0.6895, + "step": 45616 + }, + { + "epoch": 1.67, + "learning_rate": 3.3130699088145895e-08, + "loss": 0.6842, + "step": 45617 + }, + { + "epoch": 1.67, + "learning_rate": 3.312703702347383e-08, + "loss": 0.6865, + "step": 45618 + }, + { + "epoch": 1.67, + "learning_rate": 3.312337495880177e-08, + "loss": 0.6991, + "step": 45619 + }, + { + "epoch": 1.67, + "learning_rate": 3.311971289412971e-08, + "loss": 0.6944, + "step": 45620 + }, + { + "epoch": 1.67, + "learning_rate": 3.311605082945765e-08, + "loss": 0.6821, + "step": 45621 + }, + { + "epoch": 1.67, + "learning_rate": 3.311238876478559e-08, + "loss": 0.691, + "step": 45622 + }, + { + "epoch": 1.67, + "learning_rate": 3.310872670011352e-08, + "loss": 0.683, + "step": 45623 + }, + { + "epoch": 1.67, + "learning_rate": 3.310506463544146e-08, + "loss": 0.7031, + "step": 45624 + }, + { + "epoch": 1.67, + "learning_rate": 3.31014025707694e-08, + "loss": 0.6902, + "step": 45625 + }, + { + "epoch": 1.67, + "learning_rate": 3.3097740506097335e-08, + "loss": 0.6934, + "step": 45626 + }, + { + "epoch": 1.67, + "learning_rate": 3.3094078441425275e-08, + "loss": 0.7024, + "step": 45627 + }, + { + "epoch": 1.67, + "learning_rate": 3.309041637675321e-08, + "loss": 0.6954, + "step": 45628 + }, + { + "epoch": 1.67, + "learning_rate": 3.308675431208115e-08, + "loss": 0.7022, + "step": 45629 + }, + { + "epoch": 1.67, + "learning_rate": 3.308309224740909e-08, + "loss": 0.7124, + "step": 45630 + }, + { + "epoch": 1.67, + "learning_rate": 3.307943018273703e-08, + "loss": 0.6903, + "step": 45631 + }, + { + "epoch": 1.67, + "learning_rate": 3.307576811806496e-08, + "loss": 0.6988, + "step": 45632 + }, + { + "epoch": 1.67, + "learning_rate": 3.30721060533929e-08, + "loss": 0.6902, + "step": 45633 + }, + { + "epoch": 1.67, + "learning_rate": 3.306844398872084e-08, + "loss": 0.6904, + "step": 45634 + }, + { + "epoch": 1.67, + "learning_rate": 3.306478192404878e-08, + "loss": 0.6937, + "step": 45635 + }, + { + "epoch": 1.67, + "learning_rate": 3.3061119859376714e-08, + "loss": 0.6847, + "step": 45636 + }, + { + "epoch": 1.67, + "learning_rate": 3.3057457794704654e-08, + "loss": 0.6864, + "step": 45637 + }, + { + "epoch": 1.67, + "learning_rate": 3.3053795730032594e-08, + "loss": 0.6895, + "step": 45638 + }, + { + "epoch": 1.67, + "learning_rate": 3.305013366536053e-08, + "loss": 0.7003, + "step": 45639 + }, + { + "epoch": 1.67, + "learning_rate": 3.304647160068847e-08, + "loss": 0.6923, + "step": 45640 + }, + { + "epoch": 1.67, + "learning_rate": 3.30428095360164e-08, + "loss": 0.6815, + "step": 45641 + }, + { + "epoch": 1.67, + "learning_rate": 3.303914747134434e-08, + "loss": 0.6834, + "step": 45642 + }, + { + "epoch": 1.67, + "learning_rate": 3.303548540667228e-08, + "loss": 0.6968, + "step": 45643 + }, + { + "epoch": 1.67, + "learning_rate": 3.303182334200022e-08, + "loss": 0.699, + "step": 45644 + }, + { + "epoch": 1.67, + "learning_rate": 3.3028161277328154e-08, + "loss": 0.6969, + "step": 45645 + }, + { + "epoch": 1.67, + "learning_rate": 3.3024499212656094e-08, + "loss": 0.7084, + "step": 45646 + }, + { + "epoch": 1.67, + "learning_rate": 3.3020837147984034e-08, + "loss": 0.6954, + "step": 45647 + }, + { + "epoch": 1.67, + "learning_rate": 3.3017175083311974e-08, + "loss": 0.6972, + "step": 45648 + }, + { + "epoch": 1.67, + "learning_rate": 3.301351301863991e-08, + "loss": 0.6882, + "step": 45649 + }, + { + "epoch": 1.67, + "learning_rate": 3.300985095396785e-08, + "loss": 0.6963, + "step": 45650 + }, + { + "epoch": 1.67, + "learning_rate": 3.300618888929578e-08, + "loss": 0.7119, + "step": 45651 + }, + { + "epoch": 1.67, + "learning_rate": 3.300252682462372e-08, + "loss": 0.6966, + "step": 45652 + }, + { + "epoch": 1.67, + "learning_rate": 3.299886475995166e-08, + "loss": 0.6871, + "step": 45653 + }, + { + "epoch": 1.67, + "learning_rate": 3.2995202695279593e-08, + "loss": 0.6993, + "step": 45654 + }, + { + "epoch": 1.67, + "learning_rate": 3.299154063060753e-08, + "loss": 0.6824, + "step": 45655 + }, + { + "epoch": 1.67, + "learning_rate": 3.298787856593547e-08, + "loss": 0.7159, + "step": 45656 + }, + { + "epoch": 1.67, + "learning_rate": 3.298421650126341e-08, + "loss": 0.691, + "step": 45657 + }, + { + "epoch": 1.67, + "learning_rate": 3.298055443659135e-08, + "loss": 0.6918, + "step": 45658 + }, + { + "epoch": 1.67, + "learning_rate": 3.2976892371919286e-08, + "loss": 0.7063, + "step": 45659 + }, + { + "epoch": 1.67, + "learning_rate": 3.2973230307247226e-08, + "loss": 0.6905, + "step": 45660 + }, + { + "epoch": 1.67, + "learning_rate": 3.2969568242575166e-08, + "loss": 0.6689, + "step": 45661 + }, + { + "epoch": 1.67, + "learning_rate": 3.29659061779031e-08, + "loss": 0.6972, + "step": 45662 + }, + { + "epoch": 1.67, + "learning_rate": 3.296224411323103e-08, + "loss": 0.684, + "step": 45663 + }, + { + "epoch": 1.67, + "learning_rate": 3.295858204855897e-08, + "loss": 0.6866, + "step": 45664 + }, + { + "epoch": 1.67, + "learning_rate": 3.295491998388691e-08, + "loss": 0.6934, + "step": 45665 + }, + { + "epoch": 1.67, + "learning_rate": 3.295125791921485e-08, + "loss": 0.697, + "step": 45666 + }, + { + "epoch": 1.67, + "learning_rate": 3.294759585454279e-08, + "loss": 0.6921, + "step": 45667 + }, + { + "epoch": 1.67, + "learning_rate": 3.2943933789870726e-08, + "loss": 0.7013, + "step": 45668 + }, + { + "epoch": 1.67, + "learning_rate": 3.2940271725198666e-08, + "loss": 0.6912, + "step": 45669 + }, + { + "epoch": 1.67, + "learning_rate": 3.2936609660526606e-08, + "loss": 0.6934, + "step": 45670 + }, + { + "epoch": 1.67, + "learning_rate": 3.2932947595854546e-08, + "loss": 0.6912, + "step": 45671 + }, + { + "epoch": 1.67, + "learning_rate": 3.292928553118248e-08, + "loss": 0.6993, + "step": 45672 + }, + { + "epoch": 1.67, + "learning_rate": 3.292562346651042e-08, + "loss": 0.6962, + "step": 45673 + }, + { + "epoch": 1.67, + "learning_rate": 3.292196140183835e-08, + "loss": 0.6912, + "step": 45674 + }, + { + "epoch": 1.67, + "learning_rate": 3.291829933716629e-08, + "loss": 0.6828, + "step": 45675 + }, + { + "epoch": 1.67, + "learning_rate": 3.291463727249423e-08, + "loss": 0.6929, + "step": 45676 + }, + { + "epoch": 1.67, + "learning_rate": 3.2910975207822165e-08, + "loss": 0.6993, + "step": 45677 + }, + { + "epoch": 1.67, + "learning_rate": 3.2907313143150105e-08, + "loss": 0.6811, + "step": 45678 + }, + { + "epoch": 1.67, + "learning_rate": 3.2903651078478045e-08, + "loss": 0.6829, + "step": 45679 + }, + { + "epoch": 1.67, + "learning_rate": 3.2899989013805985e-08, + "loss": 0.7003, + "step": 45680 + }, + { + "epoch": 1.67, + "learning_rate": 3.289632694913392e-08, + "loss": 0.7105, + "step": 45681 + }, + { + "epoch": 1.67, + "learning_rate": 3.289266488446186e-08, + "loss": 0.6954, + "step": 45682 + }, + { + "epoch": 1.67, + "learning_rate": 3.28890028197898e-08, + "loss": 0.7029, + "step": 45683 + }, + { + "epoch": 1.67, + "learning_rate": 3.288534075511774e-08, + "loss": 0.7014, + "step": 45684 + }, + { + "epoch": 1.67, + "learning_rate": 3.288167869044567e-08, + "loss": 0.6994, + "step": 45685 + }, + { + "epoch": 1.67, + "learning_rate": 3.2878016625773605e-08, + "loss": 0.6904, + "step": 45686 + }, + { + "epoch": 1.67, + "learning_rate": 3.2874354561101545e-08, + "loss": 0.6989, + "step": 45687 + }, + { + "epoch": 1.67, + "learning_rate": 3.2870692496429485e-08, + "loss": 0.6909, + "step": 45688 + }, + { + "epoch": 1.67, + "learning_rate": 3.2867030431757425e-08, + "loss": 0.6968, + "step": 45689 + }, + { + "epoch": 1.67, + "learning_rate": 3.286336836708536e-08, + "loss": 0.6921, + "step": 45690 + }, + { + "epoch": 1.67, + "learning_rate": 3.28597063024133e-08, + "loss": 0.7021, + "step": 45691 + }, + { + "epoch": 1.67, + "learning_rate": 3.285604423774124e-08, + "loss": 0.6926, + "step": 45692 + }, + { + "epoch": 1.67, + "learning_rate": 3.285238217306918e-08, + "loss": 0.6925, + "step": 45693 + }, + { + "epoch": 1.67, + "learning_rate": 3.284872010839712e-08, + "loss": 0.7017, + "step": 45694 + }, + { + "epoch": 1.67, + "learning_rate": 3.284505804372505e-08, + "loss": 0.6932, + "step": 45695 + }, + { + "epoch": 1.67, + "learning_rate": 3.284139597905299e-08, + "loss": 0.6929, + "step": 45696 + }, + { + "epoch": 1.67, + "learning_rate": 3.2837733914380924e-08, + "loss": 0.6982, + "step": 45697 + }, + { + "epoch": 1.67, + "learning_rate": 3.2834071849708864e-08, + "loss": 0.6835, + "step": 45698 + }, + { + "epoch": 1.67, + "learning_rate": 3.28304097850368e-08, + "loss": 0.7011, + "step": 45699 + }, + { + "epoch": 1.67, + "learning_rate": 3.282674772036474e-08, + "loss": 0.6876, + "step": 45700 + }, + { + "epoch": 1.67, + "learning_rate": 3.282308565569268e-08, + "loss": 0.6924, + "step": 45701 + }, + { + "epoch": 1.67, + "learning_rate": 3.281942359102062e-08, + "loss": 0.693, + "step": 45702 + }, + { + "epoch": 1.67, + "learning_rate": 3.281576152634856e-08, + "loss": 0.6848, + "step": 45703 + }, + { + "epoch": 1.67, + "learning_rate": 3.281209946167649e-08, + "loss": 0.7047, + "step": 45704 + }, + { + "epoch": 1.67, + "learning_rate": 3.280843739700443e-08, + "loss": 0.699, + "step": 45705 + }, + { + "epoch": 1.67, + "learning_rate": 3.280477533233237e-08, + "loss": 0.6852, + "step": 45706 + }, + { + "epoch": 1.67, + "learning_rate": 3.280111326766031e-08, + "loss": 0.6895, + "step": 45707 + }, + { + "epoch": 1.67, + "learning_rate": 3.2797451202988244e-08, + "loss": 0.6905, + "step": 45708 + }, + { + "epoch": 1.67, + "learning_rate": 3.279378913831618e-08, + "loss": 0.7004, + "step": 45709 + }, + { + "epoch": 1.67, + "learning_rate": 3.279012707364412e-08, + "loss": 0.6967, + "step": 45710 + }, + { + "epoch": 1.67, + "learning_rate": 3.278646500897206e-08, + "loss": 0.695, + "step": 45711 + }, + { + "epoch": 1.67, + "learning_rate": 3.27828029443e-08, + "loss": 0.6966, + "step": 45712 + }, + { + "epoch": 1.67, + "learning_rate": 3.277914087962793e-08, + "loss": 0.6998, + "step": 45713 + }, + { + "epoch": 1.67, + "learning_rate": 3.277547881495587e-08, + "loss": 0.688, + "step": 45714 + }, + { + "epoch": 1.67, + "learning_rate": 3.277181675028381e-08, + "loss": 0.6977, + "step": 45715 + }, + { + "epoch": 1.67, + "learning_rate": 3.276815468561175e-08, + "loss": 0.6849, + "step": 45716 + }, + { + "epoch": 1.67, + "learning_rate": 3.276449262093968e-08, + "loss": 0.68, + "step": 45717 + }, + { + "epoch": 1.67, + "learning_rate": 3.276083055626762e-08, + "loss": 0.6814, + "step": 45718 + }, + { + "epoch": 1.67, + "learning_rate": 3.275716849159556e-08, + "loss": 0.687, + "step": 45719 + }, + { + "epoch": 1.67, + "learning_rate": 3.2753506426923496e-08, + "loss": 0.6911, + "step": 45720 + }, + { + "epoch": 1.67, + "learning_rate": 3.2749844362251436e-08, + "loss": 0.6953, + "step": 45721 + }, + { + "epoch": 1.67, + "learning_rate": 3.274618229757937e-08, + "loss": 0.6885, + "step": 45722 + }, + { + "epoch": 1.67, + "learning_rate": 3.274252023290731e-08, + "loss": 0.6897, + "step": 45723 + }, + { + "epoch": 1.67, + "learning_rate": 3.273885816823525e-08, + "loss": 0.7018, + "step": 45724 + }, + { + "epoch": 1.67, + "learning_rate": 3.273519610356319e-08, + "loss": 0.6967, + "step": 45725 + }, + { + "epoch": 1.67, + "learning_rate": 3.273153403889112e-08, + "loss": 0.6851, + "step": 45726 + }, + { + "epoch": 1.67, + "learning_rate": 3.272787197421906e-08, + "loss": 0.7066, + "step": 45727 + }, + { + "epoch": 1.67, + "learning_rate": 3.2724209909547e-08, + "loss": 0.6834, + "step": 45728 + }, + { + "epoch": 1.67, + "learning_rate": 3.272054784487494e-08, + "loss": 0.7043, + "step": 45729 + }, + { + "epoch": 1.67, + "learning_rate": 3.2716885780202876e-08, + "loss": 0.6955, + "step": 45730 + }, + { + "epoch": 1.67, + "learning_rate": 3.2713223715530816e-08, + "loss": 0.6948, + "step": 45731 + }, + { + "epoch": 1.67, + "learning_rate": 3.2709561650858756e-08, + "loss": 0.6961, + "step": 45732 + }, + { + "epoch": 1.67, + "learning_rate": 3.270589958618669e-08, + "loss": 0.7057, + "step": 45733 + }, + { + "epoch": 1.67, + "learning_rate": 3.270223752151463e-08, + "loss": 0.6957, + "step": 45734 + }, + { + "epoch": 1.67, + "learning_rate": 3.269857545684256e-08, + "loss": 0.6868, + "step": 45735 + }, + { + "epoch": 1.67, + "learning_rate": 3.26949133921705e-08, + "loss": 0.7015, + "step": 45736 + }, + { + "epoch": 1.67, + "learning_rate": 3.269125132749844e-08, + "loss": 0.7066, + "step": 45737 + }, + { + "epoch": 1.67, + "learning_rate": 3.268758926282638e-08, + "loss": 0.695, + "step": 45738 + }, + { + "epoch": 1.67, + "learning_rate": 3.2683927198154315e-08, + "loss": 0.6981, + "step": 45739 + }, + { + "epoch": 1.67, + "learning_rate": 3.2680265133482255e-08, + "loss": 0.6993, + "step": 45740 + }, + { + "epoch": 1.67, + "learning_rate": 3.2676603068810195e-08, + "loss": 0.6945, + "step": 45741 + }, + { + "epoch": 1.67, + "learning_rate": 3.2672941004138135e-08, + "loss": 0.6956, + "step": 45742 + }, + { + "epoch": 1.67, + "learning_rate": 3.2669278939466075e-08, + "loss": 0.6844, + "step": 45743 + }, + { + "epoch": 1.67, + "learning_rate": 3.266561687479401e-08, + "loss": 0.695, + "step": 45744 + }, + { + "epoch": 1.67, + "learning_rate": 3.266195481012194e-08, + "loss": 0.7026, + "step": 45745 + }, + { + "epoch": 1.67, + "learning_rate": 3.265829274544988e-08, + "loss": 0.6932, + "step": 45746 + }, + { + "epoch": 1.67, + "learning_rate": 3.265463068077782e-08, + "loss": 0.699, + "step": 45747 + }, + { + "epoch": 1.67, + "learning_rate": 3.2650968616105755e-08, + "loss": 0.701, + "step": 45748 + }, + { + "epoch": 1.67, + "learning_rate": 3.2647306551433695e-08, + "loss": 0.6875, + "step": 45749 + }, + { + "epoch": 1.67, + "learning_rate": 3.2643644486761635e-08, + "loss": 0.7043, + "step": 45750 + }, + { + "epoch": 1.67, + "learning_rate": 3.2639982422089574e-08, + "loss": 0.7081, + "step": 45751 + }, + { + "epoch": 1.67, + "learning_rate": 3.2636320357417514e-08, + "loss": 0.6876, + "step": 45752 + }, + { + "epoch": 1.67, + "learning_rate": 3.263265829274545e-08, + "loss": 0.6934, + "step": 45753 + }, + { + "epoch": 1.67, + "learning_rate": 3.262899622807339e-08, + "loss": 0.6894, + "step": 45754 + }, + { + "epoch": 1.67, + "learning_rate": 3.262533416340133e-08, + "loss": 0.6964, + "step": 45755 + }, + { + "epoch": 1.67, + "learning_rate": 3.262167209872926e-08, + "loss": 0.6982, + "step": 45756 + }, + { + "epoch": 1.67, + "learning_rate": 3.2618010034057194e-08, + "loss": 0.6989, + "step": 45757 + }, + { + "epoch": 1.67, + "learning_rate": 3.2614347969385134e-08, + "loss": 0.6785, + "step": 45758 + }, + { + "epoch": 1.67, + "learning_rate": 3.2610685904713074e-08, + "loss": 0.701, + "step": 45759 + }, + { + "epoch": 1.67, + "learning_rate": 3.2607023840041014e-08, + "loss": 0.689, + "step": 45760 + }, + { + "epoch": 1.67, + "learning_rate": 3.2603361775368954e-08, + "loss": 0.6989, + "step": 45761 + }, + { + "epoch": 1.67, + "learning_rate": 3.259969971069689e-08, + "loss": 0.6971, + "step": 45762 + }, + { + "epoch": 1.67, + "learning_rate": 3.259603764602483e-08, + "loss": 0.6915, + "step": 45763 + }, + { + "epoch": 1.67, + "learning_rate": 3.259237558135277e-08, + "loss": 0.6926, + "step": 45764 + }, + { + "epoch": 1.67, + "learning_rate": 3.258871351668071e-08, + "loss": 0.6938, + "step": 45765 + }, + { + "epoch": 1.67, + "learning_rate": 3.258505145200864e-08, + "loss": 0.6973, + "step": 45766 + }, + { + "epoch": 1.67, + "learning_rate": 3.258138938733658e-08, + "loss": 0.6952, + "step": 45767 + }, + { + "epoch": 1.67, + "learning_rate": 3.2577727322664514e-08, + "loss": 0.6864, + "step": 45768 + }, + { + "epoch": 1.67, + "learning_rate": 3.2574065257992454e-08, + "loss": 0.693, + "step": 45769 + }, + { + "epoch": 1.67, + "learning_rate": 3.2570403193320393e-08, + "loss": 0.6947, + "step": 45770 + }, + { + "epoch": 1.67, + "learning_rate": 3.256674112864833e-08, + "loss": 0.6818, + "step": 45771 + }, + { + "epoch": 1.67, + "learning_rate": 3.2563079063976267e-08, + "loss": 0.6997, + "step": 45772 + }, + { + "epoch": 1.67, + "learning_rate": 3.2559416999304207e-08, + "loss": 0.693, + "step": 45773 + }, + { + "epoch": 1.67, + "learning_rate": 3.2555754934632147e-08, + "loss": 0.6949, + "step": 45774 + }, + { + "epoch": 1.67, + "learning_rate": 3.255209286996008e-08, + "loss": 0.7054, + "step": 45775 + }, + { + "epoch": 1.67, + "learning_rate": 3.254843080528802e-08, + "loss": 0.687, + "step": 45776 + }, + { + "epoch": 1.67, + "learning_rate": 3.254476874061596e-08, + "loss": 0.6976, + "step": 45777 + }, + { + "epoch": 1.67, + "learning_rate": 3.25411066759439e-08, + "loss": 0.6931, + "step": 45778 + }, + { + "epoch": 1.67, + "learning_rate": 3.253744461127183e-08, + "loss": 0.6963, + "step": 45779 + }, + { + "epoch": 1.67, + "learning_rate": 3.2533782546599766e-08, + "loss": 0.7026, + "step": 45780 + }, + { + "epoch": 1.67, + "learning_rate": 3.2530120481927706e-08, + "loss": 0.6986, + "step": 45781 + }, + { + "epoch": 1.68, + "learning_rate": 3.2526458417255646e-08, + "loss": 0.6895, + "step": 45782 + }, + { + "epoch": 1.68, + "learning_rate": 3.2522796352583586e-08, + "loss": 0.6944, + "step": 45783 + }, + { + "epoch": 1.68, + "learning_rate": 3.251913428791152e-08, + "loss": 0.6915, + "step": 45784 + }, + { + "epoch": 1.68, + "learning_rate": 3.251547222323946e-08, + "loss": 0.6898, + "step": 45785 + }, + { + "epoch": 1.68, + "learning_rate": 3.25118101585674e-08, + "loss": 0.6863, + "step": 45786 + }, + { + "epoch": 1.68, + "learning_rate": 3.250814809389534e-08, + "loss": 0.6854, + "step": 45787 + }, + { + "epoch": 1.68, + "learning_rate": 3.250448602922328e-08, + "loss": 0.6939, + "step": 45788 + }, + { + "epoch": 1.68, + "learning_rate": 3.250082396455121e-08, + "loss": 0.6948, + "step": 45789 + }, + { + "epoch": 1.68, + "learning_rate": 3.249716189987915e-08, + "loss": 0.6837, + "step": 45790 + }, + { + "epoch": 1.68, + "learning_rate": 3.2493499835207086e-08, + "loss": 0.6913, + "step": 45791 + }, + { + "epoch": 1.68, + "learning_rate": 3.2489837770535026e-08, + "loss": 0.6907, + "step": 45792 + }, + { + "epoch": 1.68, + "learning_rate": 3.248617570586296e-08, + "loss": 0.686, + "step": 45793 + }, + { + "epoch": 1.68, + "learning_rate": 3.24825136411909e-08, + "loss": 0.7014, + "step": 45794 + }, + { + "epoch": 1.68, + "learning_rate": 3.247885157651884e-08, + "loss": 0.6758, + "step": 45795 + }, + { + "epoch": 1.68, + "learning_rate": 3.247518951184678e-08, + "loss": 0.6889, + "step": 45796 + }, + { + "epoch": 1.68, + "learning_rate": 3.247152744717472e-08, + "loss": 0.6876, + "step": 45797 + }, + { + "epoch": 1.68, + "learning_rate": 3.246786538250265e-08, + "loss": 0.6883, + "step": 45798 + }, + { + "epoch": 1.68, + "learning_rate": 3.246420331783059e-08, + "loss": 0.6973, + "step": 45799 + }, + { + "epoch": 1.68, + "learning_rate": 3.246054125315853e-08, + "loss": 0.6916, + "step": 45800 + }, + { + "epoch": 1.68, + "learning_rate": 3.245687918848647e-08, + "loss": 0.6854, + "step": 45801 + }, + { + "epoch": 1.68, + "learning_rate": 3.2453217123814405e-08, + "loss": 0.6922, + "step": 45802 + }, + { + "epoch": 1.68, + "learning_rate": 3.2449555059142345e-08, + "loss": 0.7009, + "step": 45803 + }, + { + "epoch": 1.68, + "learning_rate": 3.244589299447028e-08, + "loss": 0.6903, + "step": 45804 + }, + { + "epoch": 1.68, + "learning_rate": 3.244223092979822e-08, + "loss": 0.6862, + "step": 45805 + }, + { + "epoch": 1.68, + "learning_rate": 3.243856886512616e-08, + "loss": 0.6785, + "step": 45806 + }, + { + "epoch": 1.68, + "learning_rate": 3.243490680045409e-08, + "loss": 0.7084, + "step": 45807 + }, + { + "epoch": 1.68, + "learning_rate": 3.243124473578203e-08, + "loss": 0.6902, + "step": 45808 + }, + { + "epoch": 1.68, + "learning_rate": 3.242758267110997e-08, + "loss": 0.6968, + "step": 45809 + }, + { + "epoch": 1.68, + "learning_rate": 3.242392060643791e-08, + "loss": 0.688, + "step": 45810 + }, + { + "epoch": 1.68, + "learning_rate": 3.2420258541765844e-08, + "loss": 0.6913, + "step": 45811 + }, + { + "epoch": 1.68, + "learning_rate": 3.2416596477093784e-08, + "loss": 0.6966, + "step": 45812 + }, + { + "epoch": 1.68, + "learning_rate": 3.2412934412421724e-08, + "loss": 0.6974, + "step": 45813 + }, + { + "epoch": 1.68, + "learning_rate": 3.2409272347749664e-08, + "loss": 0.6936, + "step": 45814 + }, + { + "epoch": 1.68, + "learning_rate": 3.24056102830776e-08, + "loss": 0.6931, + "step": 45815 + }, + { + "epoch": 1.68, + "learning_rate": 3.240194821840553e-08, + "loss": 0.701, + "step": 45816 + }, + { + "epoch": 1.68, + "learning_rate": 3.239828615373347e-08, + "loss": 0.6979, + "step": 45817 + }, + { + "epoch": 1.68, + "learning_rate": 3.239462408906141e-08, + "loss": 0.6842, + "step": 45818 + }, + { + "epoch": 1.68, + "learning_rate": 3.239096202438935e-08, + "loss": 0.6829, + "step": 45819 + }, + { + "epoch": 1.68, + "learning_rate": 3.2387299959717284e-08, + "loss": 0.6896, + "step": 45820 + }, + { + "epoch": 1.68, + "learning_rate": 3.2383637895045224e-08, + "loss": 0.6988, + "step": 45821 + }, + { + "epoch": 1.68, + "learning_rate": 3.2379975830373164e-08, + "loss": 0.6894, + "step": 45822 + }, + { + "epoch": 1.68, + "learning_rate": 3.2376313765701104e-08, + "loss": 0.6932, + "step": 45823 + }, + { + "epoch": 1.68, + "learning_rate": 3.2372651701029044e-08, + "loss": 0.6931, + "step": 45824 + }, + { + "epoch": 1.68, + "learning_rate": 3.236898963635698e-08, + "loss": 0.7026, + "step": 45825 + }, + { + "epoch": 1.68, + "learning_rate": 3.236532757168492e-08, + "loss": 0.6932, + "step": 45826 + }, + { + "epoch": 1.68, + "learning_rate": 3.236166550701285e-08, + "loss": 0.6933, + "step": 45827 + }, + { + "epoch": 1.68, + "learning_rate": 3.235800344234079e-08, + "loss": 0.6991, + "step": 45828 + }, + { + "epoch": 1.68, + "learning_rate": 3.2354341377668723e-08, + "loss": 0.6919, + "step": 45829 + }, + { + "epoch": 1.68, + "learning_rate": 3.2350679312996663e-08, + "loss": 0.6996, + "step": 45830 + }, + { + "epoch": 1.68, + "learning_rate": 3.2347017248324603e-08, + "loss": 0.6992, + "step": 45831 + }, + { + "epoch": 1.68, + "learning_rate": 3.234335518365254e-08, + "loss": 0.6882, + "step": 45832 + }, + { + "epoch": 1.68, + "learning_rate": 3.233969311898048e-08, + "loss": 0.6895, + "step": 45833 + }, + { + "epoch": 1.68, + "learning_rate": 3.2336031054308417e-08, + "loss": 0.6911, + "step": 45834 + }, + { + "epoch": 1.68, + "learning_rate": 3.2332368989636356e-08, + "loss": 0.6919, + "step": 45835 + }, + { + "epoch": 1.68, + "learning_rate": 3.2328706924964296e-08, + "loss": 0.6871, + "step": 45836 + }, + { + "epoch": 1.68, + "learning_rate": 3.2325044860292236e-08, + "loss": 0.695, + "step": 45837 + }, + { + "epoch": 1.68, + "learning_rate": 3.232138279562017e-08, + "loss": 0.6955, + "step": 45838 + }, + { + "epoch": 1.68, + "learning_rate": 3.23177207309481e-08, + "loss": 0.6892, + "step": 45839 + }, + { + "epoch": 1.68, + "learning_rate": 3.231405866627604e-08, + "loss": 0.6899, + "step": 45840 + }, + { + "epoch": 1.68, + "learning_rate": 3.231039660160398e-08, + "loss": 0.6853, + "step": 45841 + }, + { + "epoch": 1.68, + "learning_rate": 3.230673453693192e-08, + "loss": 0.6948, + "step": 45842 + }, + { + "epoch": 1.68, + "learning_rate": 3.2303072472259856e-08, + "loss": 0.6977, + "step": 45843 + }, + { + "epoch": 1.68, + "learning_rate": 3.2299410407587796e-08, + "loss": 0.6904, + "step": 45844 + }, + { + "epoch": 1.68, + "learning_rate": 3.2295748342915736e-08, + "loss": 0.6952, + "step": 45845 + }, + { + "epoch": 1.68, + "learning_rate": 3.2292086278243676e-08, + "loss": 0.6958, + "step": 45846 + }, + { + "epoch": 1.68, + "learning_rate": 3.228842421357161e-08, + "loss": 0.6877, + "step": 45847 + }, + { + "epoch": 1.68, + "learning_rate": 3.228476214889955e-08, + "loss": 0.6981, + "step": 45848 + }, + { + "epoch": 1.68, + "learning_rate": 3.228110008422749e-08, + "loss": 0.6901, + "step": 45849 + }, + { + "epoch": 1.68, + "learning_rate": 3.227743801955542e-08, + "loss": 0.7002, + "step": 45850 + }, + { + "epoch": 1.68, + "learning_rate": 3.227377595488336e-08, + "loss": 0.695, + "step": 45851 + }, + { + "epoch": 1.68, + "learning_rate": 3.2270113890211296e-08, + "loss": 0.6897, + "step": 45852 + }, + { + "epoch": 1.68, + "learning_rate": 3.2266451825539235e-08, + "loss": 0.6943, + "step": 45853 + }, + { + "epoch": 1.68, + "learning_rate": 3.2262789760867175e-08, + "loss": 0.6862, + "step": 45854 + }, + { + "epoch": 1.68, + "learning_rate": 3.2259127696195115e-08, + "loss": 0.7076, + "step": 45855 + }, + { + "epoch": 1.68, + "learning_rate": 3.225546563152305e-08, + "loss": 0.6981, + "step": 45856 + }, + { + "epoch": 1.68, + "learning_rate": 3.225180356685099e-08, + "loss": 0.6851, + "step": 45857 + }, + { + "epoch": 1.68, + "learning_rate": 3.224814150217893e-08, + "loss": 0.6839, + "step": 45858 + }, + { + "epoch": 1.68, + "learning_rate": 3.224447943750687e-08, + "loss": 0.6875, + "step": 45859 + }, + { + "epoch": 1.68, + "learning_rate": 3.224081737283481e-08, + "loss": 0.7035, + "step": 45860 + }, + { + "epoch": 1.68, + "learning_rate": 3.223715530816274e-08, + "loss": 0.6925, + "step": 45861 + }, + { + "epoch": 1.68, + "learning_rate": 3.2233493243490675e-08, + "loss": 0.6915, + "step": 45862 + }, + { + "epoch": 1.68, + "learning_rate": 3.2229831178818615e-08, + "loss": 0.6901, + "step": 45863 + }, + { + "epoch": 1.68, + "learning_rate": 3.2226169114146555e-08, + "loss": 0.6922, + "step": 45864 + }, + { + "epoch": 1.68, + "learning_rate": 3.222250704947449e-08, + "loss": 0.6905, + "step": 45865 + }, + { + "epoch": 1.68, + "learning_rate": 3.221884498480243e-08, + "loss": 0.6956, + "step": 45866 + }, + { + "epoch": 1.68, + "learning_rate": 3.221518292013037e-08, + "loss": 0.6978, + "step": 45867 + }, + { + "epoch": 1.68, + "learning_rate": 3.221152085545831e-08, + "loss": 0.6921, + "step": 45868 + }, + { + "epoch": 1.68, + "learning_rate": 3.220785879078625e-08, + "loss": 0.6877, + "step": 45869 + }, + { + "epoch": 1.68, + "learning_rate": 3.220419672611418e-08, + "loss": 0.6944, + "step": 45870 + }, + { + "epoch": 1.68, + "learning_rate": 3.220053466144212e-08, + "loss": 0.7006, + "step": 45871 + }, + { + "epoch": 1.68, + "learning_rate": 3.219687259677006e-08, + "loss": 0.6915, + "step": 45872 + }, + { + "epoch": 1.68, + "learning_rate": 3.2193210532097994e-08, + "loss": 0.6845, + "step": 45873 + }, + { + "epoch": 1.68, + "learning_rate": 3.2189548467425934e-08, + "loss": 0.6834, + "step": 45874 + }, + { + "epoch": 1.68, + "learning_rate": 3.218588640275387e-08, + "loss": 0.6958, + "step": 45875 + }, + { + "epoch": 1.68, + "learning_rate": 3.218222433808181e-08, + "loss": 0.7013, + "step": 45876 + }, + { + "epoch": 1.68, + "learning_rate": 3.217856227340975e-08, + "loss": 0.6932, + "step": 45877 + }, + { + "epoch": 1.68, + "learning_rate": 3.217490020873769e-08, + "loss": 0.6968, + "step": 45878 + }, + { + "epoch": 1.68, + "learning_rate": 3.217123814406562e-08, + "loss": 0.6879, + "step": 45879 + }, + { + "epoch": 1.68, + "learning_rate": 3.216757607939356e-08, + "loss": 0.6888, + "step": 45880 + }, + { + "epoch": 1.68, + "learning_rate": 3.21639140147215e-08, + "loss": 0.6945, + "step": 45881 + }, + { + "epoch": 1.68, + "learning_rate": 3.216025195004944e-08, + "loss": 0.694, + "step": 45882 + }, + { + "epoch": 1.68, + "learning_rate": 3.2156589885377374e-08, + "loss": 0.6905, + "step": 45883 + }, + { + "epoch": 1.68, + "learning_rate": 3.2152927820705314e-08, + "loss": 0.6982, + "step": 45884 + }, + { + "epoch": 1.68, + "learning_rate": 3.214926575603325e-08, + "loss": 0.6878, + "step": 45885 + }, + { + "epoch": 1.68, + "learning_rate": 3.214560369136119e-08, + "loss": 0.6908, + "step": 45886 + }, + { + "epoch": 1.68, + "learning_rate": 3.214194162668913e-08, + "loss": 0.6813, + "step": 45887 + }, + { + "epoch": 1.68, + "learning_rate": 3.213827956201706e-08, + "loss": 0.6967, + "step": 45888 + }, + { + "epoch": 1.68, + "learning_rate": 3.2134617497345e-08, + "loss": 0.6989, + "step": 45889 + }, + { + "epoch": 1.68, + "learning_rate": 3.213095543267294e-08, + "loss": 0.685, + "step": 45890 + }, + { + "epoch": 1.68, + "learning_rate": 3.212729336800088e-08, + "loss": 0.7004, + "step": 45891 + }, + { + "epoch": 1.68, + "learning_rate": 3.212363130332881e-08, + "loss": 0.6872, + "step": 45892 + }, + { + "epoch": 1.68, + "learning_rate": 3.211996923865675e-08, + "loss": 0.6935, + "step": 45893 + }, + { + "epoch": 1.68, + "learning_rate": 3.211630717398469e-08, + "loss": 0.7083, + "step": 45894 + }, + { + "epoch": 1.68, + "learning_rate": 3.211264510931263e-08, + "loss": 0.6989, + "step": 45895 + }, + { + "epoch": 1.68, + "learning_rate": 3.2108983044640566e-08, + "loss": 0.7037, + "step": 45896 + }, + { + "epoch": 1.68, + "learning_rate": 3.2105320979968506e-08, + "loss": 0.6891, + "step": 45897 + }, + { + "epoch": 1.68, + "learning_rate": 3.210165891529644e-08, + "loss": 0.6831, + "step": 45898 + }, + { + "epoch": 1.68, + "learning_rate": 3.209799685062438e-08, + "loss": 0.6848, + "step": 45899 + }, + { + "epoch": 1.68, + "learning_rate": 3.209433478595232e-08, + "loss": 0.7055, + "step": 45900 + }, + { + "epoch": 1.68, + "learning_rate": 3.209067272128025e-08, + "loss": 0.6996, + "step": 45901 + }, + { + "epoch": 1.68, + "learning_rate": 3.208701065660819e-08, + "loss": 0.7036, + "step": 45902 + }, + { + "epoch": 1.68, + "learning_rate": 3.208334859193613e-08, + "loss": 0.6928, + "step": 45903 + }, + { + "epoch": 1.68, + "learning_rate": 3.207968652726407e-08, + "loss": 0.6978, + "step": 45904 + }, + { + "epoch": 1.68, + "learning_rate": 3.207602446259201e-08, + "loss": 0.699, + "step": 45905 + }, + { + "epoch": 1.68, + "learning_rate": 3.2072362397919946e-08, + "loss": 0.6864, + "step": 45906 + }, + { + "epoch": 1.68, + "learning_rate": 3.2068700333247886e-08, + "loss": 0.6869, + "step": 45907 + }, + { + "epoch": 1.68, + "learning_rate": 3.2065038268575826e-08, + "loss": 0.6967, + "step": 45908 + }, + { + "epoch": 1.68, + "learning_rate": 3.206137620390376e-08, + "loss": 0.6824, + "step": 45909 + }, + { + "epoch": 1.68, + "learning_rate": 3.205771413923169e-08, + "loss": 0.7041, + "step": 45910 + }, + { + "epoch": 1.68, + "learning_rate": 3.205405207455963e-08, + "loss": 0.6955, + "step": 45911 + }, + { + "epoch": 1.68, + "learning_rate": 3.205039000988757e-08, + "loss": 0.7077, + "step": 45912 + }, + { + "epoch": 1.68, + "learning_rate": 3.204672794521551e-08, + "loss": 0.7036, + "step": 45913 + }, + { + "epoch": 1.68, + "learning_rate": 3.204306588054345e-08, + "loss": 0.6887, + "step": 45914 + }, + { + "epoch": 1.68, + "learning_rate": 3.2039403815871385e-08, + "loss": 0.6983, + "step": 45915 + }, + { + "epoch": 1.68, + "learning_rate": 3.2035741751199325e-08, + "loss": 0.6905, + "step": 45916 + }, + { + "epoch": 1.68, + "learning_rate": 3.2032079686527265e-08, + "loss": 0.7027, + "step": 45917 + }, + { + "epoch": 1.68, + "learning_rate": 3.2028417621855205e-08, + "loss": 0.6743, + "step": 45918 + }, + { + "epoch": 1.68, + "learning_rate": 3.202475555718314e-08, + "loss": 0.6985, + "step": 45919 + }, + { + "epoch": 1.68, + "learning_rate": 3.202109349251108e-08, + "loss": 0.7005, + "step": 45920 + }, + { + "epoch": 1.68, + "learning_rate": 3.201743142783901e-08, + "loss": 0.6905, + "step": 45921 + }, + { + "epoch": 1.68, + "learning_rate": 3.201376936316695e-08, + "loss": 0.6805, + "step": 45922 + }, + { + "epoch": 1.68, + "learning_rate": 3.2010107298494885e-08, + "loss": 0.6931, + "step": 45923 + }, + { + "epoch": 1.68, + "learning_rate": 3.2006445233822825e-08, + "loss": 0.6861, + "step": 45924 + }, + { + "epoch": 1.68, + "learning_rate": 3.2002783169150765e-08, + "loss": 0.6977, + "step": 45925 + }, + { + "epoch": 1.68, + "learning_rate": 3.1999121104478705e-08, + "loss": 0.6936, + "step": 45926 + }, + { + "epoch": 1.68, + "learning_rate": 3.1995459039806645e-08, + "loss": 0.6862, + "step": 45927 + }, + { + "epoch": 1.68, + "learning_rate": 3.199179697513458e-08, + "loss": 0.7, + "step": 45928 + }, + { + "epoch": 1.68, + "learning_rate": 3.198813491046252e-08, + "loss": 0.6773, + "step": 45929 + }, + { + "epoch": 1.68, + "learning_rate": 3.198447284579046e-08, + "loss": 0.6881, + "step": 45930 + }, + { + "epoch": 1.68, + "learning_rate": 3.19808107811184e-08, + "loss": 0.6869, + "step": 45931 + }, + { + "epoch": 1.68, + "learning_rate": 3.197714871644633e-08, + "loss": 0.7081, + "step": 45932 + }, + { + "epoch": 1.68, + "learning_rate": 3.1973486651774264e-08, + "loss": 0.6915, + "step": 45933 + }, + { + "epoch": 1.68, + "learning_rate": 3.1969824587102204e-08, + "loss": 0.6967, + "step": 45934 + }, + { + "epoch": 1.68, + "learning_rate": 3.1966162522430144e-08, + "loss": 0.6866, + "step": 45935 + }, + { + "epoch": 1.68, + "learning_rate": 3.1962500457758084e-08, + "loss": 0.69, + "step": 45936 + }, + { + "epoch": 1.68, + "learning_rate": 3.195883839308602e-08, + "loss": 0.7022, + "step": 45937 + }, + { + "epoch": 1.68, + "learning_rate": 3.195517632841396e-08, + "loss": 0.6861, + "step": 45938 + }, + { + "epoch": 1.68, + "learning_rate": 3.19515142637419e-08, + "loss": 0.7078, + "step": 45939 + }, + { + "epoch": 1.68, + "learning_rate": 3.194785219906984e-08, + "loss": 0.6888, + "step": 45940 + }, + { + "epoch": 1.68, + "learning_rate": 3.194419013439777e-08, + "loss": 0.6958, + "step": 45941 + }, + { + "epoch": 1.68, + "learning_rate": 3.194052806972571e-08, + "loss": 0.6958, + "step": 45942 + }, + { + "epoch": 1.68, + "learning_rate": 3.193686600505365e-08, + "loss": 0.6898, + "step": 45943 + }, + { + "epoch": 1.68, + "learning_rate": 3.1933203940381584e-08, + "loss": 0.6929, + "step": 45944 + }, + { + "epoch": 1.68, + "learning_rate": 3.1929541875709524e-08, + "loss": 0.6878, + "step": 45945 + }, + { + "epoch": 1.68, + "learning_rate": 3.192587981103746e-08, + "loss": 0.7072, + "step": 45946 + }, + { + "epoch": 1.68, + "learning_rate": 3.19222177463654e-08, + "loss": 0.6901, + "step": 45947 + }, + { + "epoch": 1.68, + "learning_rate": 3.191855568169334e-08, + "loss": 0.7055, + "step": 45948 + }, + { + "epoch": 1.68, + "learning_rate": 3.191489361702128e-08, + "loss": 0.6824, + "step": 45949 + }, + { + "epoch": 1.68, + "learning_rate": 3.191123155234921e-08, + "loss": 0.6801, + "step": 45950 + }, + { + "epoch": 1.68, + "learning_rate": 3.190756948767715e-08, + "loss": 0.6896, + "step": 45951 + }, + { + "epoch": 1.68, + "learning_rate": 3.190390742300509e-08, + "loss": 0.6918, + "step": 45952 + }, + { + "epoch": 1.68, + "learning_rate": 3.190024535833303e-08, + "loss": 0.6881, + "step": 45953 + }, + { + "epoch": 1.68, + "learning_rate": 3.189658329366097e-08, + "loss": 0.6862, + "step": 45954 + }, + { + "epoch": 1.68, + "learning_rate": 3.18929212289889e-08, + "loss": 0.6899, + "step": 45955 + }, + { + "epoch": 1.68, + "learning_rate": 3.1889259164316836e-08, + "loss": 0.6842, + "step": 45956 + }, + { + "epoch": 1.68, + "learning_rate": 3.1885597099644776e-08, + "loss": 0.6846, + "step": 45957 + }, + { + "epoch": 1.68, + "learning_rate": 3.1881935034972716e-08, + "loss": 0.6935, + "step": 45958 + }, + { + "epoch": 1.68, + "learning_rate": 3.187827297030065e-08, + "loss": 0.6924, + "step": 45959 + }, + { + "epoch": 1.68, + "learning_rate": 3.187461090562859e-08, + "loss": 0.6872, + "step": 45960 + }, + { + "epoch": 1.68, + "learning_rate": 3.187094884095653e-08, + "loss": 0.6898, + "step": 45961 + }, + { + "epoch": 1.68, + "learning_rate": 3.186728677628447e-08, + "loss": 0.6965, + "step": 45962 + }, + { + "epoch": 1.68, + "learning_rate": 3.186362471161241e-08, + "loss": 0.694, + "step": 45963 + }, + { + "epoch": 1.68, + "learning_rate": 3.185996264694034e-08, + "loss": 0.6992, + "step": 45964 + }, + { + "epoch": 1.68, + "learning_rate": 3.185630058226828e-08, + "loss": 0.6909, + "step": 45965 + }, + { + "epoch": 1.68, + "learning_rate": 3.185263851759622e-08, + "loss": 0.6972, + "step": 45966 + }, + { + "epoch": 1.68, + "learning_rate": 3.1848976452924156e-08, + "loss": 0.709, + "step": 45967 + }, + { + "epoch": 1.68, + "learning_rate": 3.1845314388252096e-08, + "loss": 0.6984, + "step": 45968 + }, + { + "epoch": 1.68, + "learning_rate": 3.184165232358003e-08, + "loss": 0.6796, + "step": 45969 + }, + { + "epoch": 1.68, + "learning_rate": 3.183799025890797e-08, + "loss": 0.6935, + "step": 45970 + }, + { + "epoch": 1.68, + "learning_rate": 3.183432819423591e-08, + "loss": 0.6871, + "step": 45971 + }, + { + "epoch": 1.68, + "learning_rate": 3.183066612956385e-08, + "loss": 0.6765, + "step": 45972 + }, + { + "epoch": 1.68, + "learning_rate": 3.182700406489178e-08, + "loss": 0.6976, + "step": 45973 + }, + { + "epoch": 1.68, + "learning_rate": 3.182334200021972e-08, + "loss": 0.6853, + "step": 45974 + }, + { + "epoch": 1.68, + "learning_rate": 3.181967993554766e-08, + "loss": 0.7007, + "step": 45975 + }, + { + "epoch": 1.68, + "learning_rate": 3.18160178708756e-08, + "loss": 0.6872, + "step": 45976 + }, + { + "epoch": 1.68, + "learning_rate": 3.1812355806203535e-08, + "loss": 0.6914, + "step": 45977 + }, + { + "epoch": 1.68, + "learning_rate": 3.1808693741531475e-08, + "loss": 0.7029, + "step": 45978 + }, + { + "epoch": 1.68, + "learning_rate": 3.1805031676859415e-08, + "loss": 0.7022, + "step": 45979 + }, + { + "epoch": 1.68, + "learning_rate": 3.180136961218735e-08, + "loss": 0.6879, + "step": 45980 + }, + { + "epoch": 1.68, + "learning_rate": 3.179770754751529e-08, + "loss": 0.7068, + "step": 45981 + }, + { + "epoch": 1.68, + "learning_rate": 3.179404548284322e-08, + "loss": 0.6904, + "step": 45982 + }, + { + "epoch": 1.68, + "learning_rate": 3.179038341817116e-08, + "loss": 0.6978, + "step": 45983 + }, + { + "epoch": 1.68, + "learning_rate": 3.17867213534991e-08, + "loss": 0.7014, + "step": 45984 + }, + { + "epoch": 1.68, + "learning_rate": 3.178305928882704e-08, + "loss": 0.6908, + "step": 45985 + }, + { + "epoch": 1.68, + "learning_rate": 3.1779397224154975e-08, + "loss": 0.6895, + "step": 45986 + }, + { + "epoch": 1.68, + "learning_rate": 3.1775735159482915e-08, + "loss": 0.6902, + "step": 45987 + }, + { + "epoch": 1.68, + "learning_rate": 3.1772073094810855e-08, + "loss": 0.7004, + "step": 45988 + }, + { + "epoch": 1.68, + "learning_rate": 3.1768411030138794e-08, + "loss": 0.687, + "step": 45989 + }, + { + "epoch": 1.68, + "learning_rate": 3.1764748965466734e-08, + "loss": 0.6866, + "step": 45990 + }, + { + "epoch": 1.68, + "learning_rate": 3.176108690079467e-08, + "loss": 0.6889, + "step": 45991 + }, + { + "epoch": 1.68, + "learning_rate": 3.17574248361226e-08, + "loss": 0.7049, + "step": 45992 + }, + { + "epoch": 1.68, + "learning_rate": 3.175376277145054e-08, + "loss": 0.6969, + "step": 45993 + }, + { + "epoch": 1.68, + "learning_rate": 3.175010070677848e-08, + "loss": 0.7, + "step": 45994 + }, + { + "epoch": 1.68, + "learning_rate": 3.1746438642106414e-08, + "loss": 0.6983, + "step": 45995 + }, + { + "epoch": 1.68, + "learning_rate": 3.1742776577434354e-08, + "loss": 0.6934, + "step": 45996 + }, + { + "epoch": 1.68, + "learning_rate": 3.1739114512762294e-08, + "loss": 0.6914, + "step": 45997 + }, + { + "epoch": 1.68, + "learning_rate": 3.1735452448090234e-08, + "loss": 0.7012, + "step": 45998 + }, + { + "epoch": 1.68, + "learning_rate": 3.1731790383418174e-08, + "loss": 0.6712, + "step": 45999 + }, + { + "epoch": 1.68, + "learning_rate": 3.172812831874611e-08, + "loss": 0.6968, + "step": 46000 + }, + { + "epoch": 1.68, + "learning_rate": 3.172446625407405e-08, + "loss": 0.6916, + "step": 46001 + }, + { + "epoch": 1.68, + "learning_rate": 3.172080418940199e-08, + "loss": 0.6927, + "step": 46002 + }, + { + "epoch": 1.68, + "learning_rate": 3.171714212472992e-08, + "loss": 0.6946, + "step": 46003 + }, + { + "epoch": 1.68, + "learning_rate": 3.1713480060057854e-08, + "loss": 0.6868, + "step": 46004 + }, + { + "epoch": 1.68, + "learning_rate": 3.1709817995385794e-08, + "loss": 0.6932, + "step": 46005 + }, + { + "epoch": 1.68, + "learning_rate": 3.1706155930713734e-08, + "loss": 0.6936, + "step": 46006 + }, + { + "epoch": 1.68, + "learning_rate": 3.1702493866041673e-08, + "loss": 0.6963, + "step": 46007 + }, + { + "epoch": 1.68, + "learning_rate": 3.1698831801369613e-08, + "loss": 0.6986, + "step": 46008 + }, + { + "epoch": 1.68, + "learning_rate": 3.1695169736697547e-08, + "loss": 0.6866, + "step": 46009 + }, + { + "epoch": 1.68, + "learning_rate": 3.1691507672025487e-08, + "loss": 0.6793, + "step": 46010 + }, + { + "epoch": 1.68, + "learning_rate": 3.1687845607353427e-08, + "loss": 0.689, + "step": 46011 + }, + { + "epoch": 1.68, + "learning_rate": 3.1684183542681366e-08, + "loss": 0.6892, + "step": 46012 + }, + { + "epoch": 1.68, + "learning_rate": 3.16805214780093e-08, + "loss": 0.6821, + "step": 46013 + }, + { + "epoch": 1.68, + "learning_rate": 3.167685941333724e-08, + "loss": 0.7005, + "step": 46014 + }, + { + "epoch": 1.68, + "learning_rate": 3.167319734866517e-08, + "loss": 0.6997, + "step": 46015 + }, + { + "epoch": 1.68, + "learning_rate": 3.166953528399311e-08, + "loss": 0.6718, + "step": 46016 + }, + { + "epoch": 1.68, + "learning_rate": 3.166587321932105e-08, + "loss": 0.6904, + "step": 46017 + }, + { + "epoch": 1.68, + "learning_rate": 3.1662211154648986e-08, + "loss": 0.6998, + "step": 46018 + }, + { + "epoch": 1.68, + "learning_rate": 3.1658549089976926e-08, + "loss": 0.6914, + "step": 46019 + }, + { + "epoch": 1.68, + "learning_rate": 3.1654887025304866e-08, + "loss": 0.6874, + "step": 46020 + }, + { + "epoch": 1.68, + "learning_rate": 3.1651224960632806e-08, + "loss": 0.6886, + "step": 46021 + }, + { + "epoch": 1.68, + "learning_rate": 3.164756289596074e-08, + "loss": 0.6963, + "step": 46022 + }, + { + "epoch": 1.68, + "learning_rate": 3.164390083128868e-08, + "loss": 0.6875, + "step": 46023 + }, + { + "epoch": 1.68, + "learning_rate": 3.164023876661662e-08, + "loss": 0.6739, + "step": 46024 + }, + { + "epoch": 1.68, + "learning_rate": 3.163657670194456e-08, + "loss": 0.6961, + "step": 46025 + }, + { + "epoch": 1.68, + "learning_rate": 3.163291463727249e-08, + "loss": 0.6893, + "step": 46026 + }, + { + "epoch": 1.68, + "learning_rate": 3.1629252572600426e-08, + "loss": 0.699, + "step": 46027 + }, + { + "epoch": 1.68, + "learning_rate": 3.1625590507928366e-08, + "loss": 0.6937, + "step": 46028 + }, + { + "epoch": 1.68, + "learning_rate": 3.1621928443256306e-08, + "loss": 0.6962, + "step": 46029 + }, + { + "epoch": 1.68, + "learning_rate": 3.1618266378584245e-08, + "loss": 0.6963, + "step": 46030 + }, + { + "epoch": 1.68, + "learning_rate": 3.161460431391218e-08, + "loss": 0.6952, + "step": 46031 + }, + { + "epoch": 1.68, + "learning_rate": 3.161094224924012e-08, + "loss": 0.6929, + "step": 46032 + }, + { + "epoch": 1.68, + "learning_rate": 3.160728018456806e-08, + "loss": 0.6841, + "step": 46033 + }, + { + "epoch": 1.68, + "learning_rate": 3.1603618119896e-08, + "loss": 0.6899, + "step": 46034 + }, + { + "epoch": 1.68, + "learning_rate": 3.159995605522394e-08, + "loss": 0.6857, + "step": 46035 + }, + { + "epoch": 1.68, + "learning_rate": 3.159629399055187e-08, + "loss": 0.695, + "step": 46036 + }, + { + "epoch": 1.68, + "learning_rate": 3.159263192587981e-08, + "loss": 0.7003, + "step": 46037 + }, + { + "epoch": 1.68, + "learning_rate": 3.1588969861207745e-08, + "loss": 0.6886, + "step": 46038 + }, + { + "epoch": 1.68, + "learning_rate": 3.1585307796535685e-08, + "loss": 0.6919, + "step": 46039 + }, + { + "epoch": 1.68, + "learning_rate": 3.158164573186362e-08, + "loss": 0.6964, + "step": 46040 + }, + { + "epoch": 1.68, + "learning_rate": 3.157798366719156e-08, + "loss": 0.6921, + "step": 46041 + }, + { + "epoch": 1.68, + "learning_rate": 3.15743216025195e-08, + "loss": 0.696, + "step": 46042 + }, + { + "epoch": 1.68, + "learning_rate": 3.157065953784744e-08, + "loss": 0.697, + "step": 46043 + }, + { + "epoch": 1.68, + "learning_rate": 3.156699747317538e-08, + "loss": 0.69, + "step": 46044 + }, + { + "epoch": 1.68, + "learning_rate": 3.156333540850331e-08, + "loss": 0.6949, + "step": 46045 + }, + { + "epoch": 1.68, + "learning_rate": 3.155967334383125e-08, + "loss": 0.6995, + "step": 46046 + }, + { + "epoch": 1.68, + "learning_rate": 3.155601127915919e-08, + "loss": 0.6968, + "step": 46047 + }, + { + "epoch": 1.68, + "learning_rate": 3.155234921448713e-08, + "loss": 0.6995, + "step": 46048 + }, + { + "epoch": 1.68, + "learning_rate": 3.1548687149815064e-08, + "loss": 0.6958, + "step": 46049 + }, + { + "epoch": 1.68, + "learning_rate": 3.1545025085143004e-08, + "loss": 0.6904, + "step": 46050 + }, + { + "epoch": 1.68, + "learning_rate": 3.154136302047094e-08, + "loss": 0.6991, + "step": 46051 + }, + { + "epoch": 1.68, + "learning_rate": 3.153770095579888e-08, + "loss": 0.69, + "step": 46052 + }, + { + "epoch": 1.68, + "learning_rate": 3.153403889112682e-08, + "loss": 0.6947, + "step": 46053 + }, + { + "epoch": 1.68, + "learning_rate": 3.153037682645475e-08, + "loss": 0.7018, + "step": 46054 + }, + { + "epoch": 1.69, + "learning_rate": 3.152671476178269e-08, + "loss": 0.6864, + "step": 46055 + }, + { + "epoch": 1.69, + "learning_rate": 3.152305269711063e-08, + "loss": 0.6983, + "step": 46056 + }, + { + "epoch": 1.69, + "learning_rate": 3.151939063243857e-08, + "loss": 0.6818, + "step": 46057 + }, + { + "epoch": 1.69, + "learning_rate": 3.1515728567766504e-08, + "loss": 0.6962, + "step": 46058 + }, + { + "epoch": 1.69, + "learning_rate": 3.1512066503094444e-08, + "loss": 0.7056, + "step": 46059 + }, + { + "epoch": 1.69, + "learning_rate": 3.1508404438422384e-08, + "loss": 0.6872, + "step": 46060 + }, + { + "epoch": 1.69, + "learning_rate": 3.150474237375032e-08, + "loss": 0.6996, + "step": 46061 + }, + { + "epoch": 1.69, + "learning_rate": 3.150108030907826e-08, + "loss": 0.6951, + "step": 46062 + }, + { + "epoch": 1.69, + "learning_rate": 3.149741824440619e-08, + "loss": 0.6884, + "step": 46063 + }, + { + "epoch": 1.69, + "learning_rate": 3.149375617973413e-08, + "loss": 0.6879, + "step": 46064 + }, + { + "epoch": 1.69, + "learning_rate": 3.149009411506207e-08, + "loss": 0.6983, + "step": 46065 + }, + { + "epoch": 1.69, + "learning_rate": 3.148643205039001e-08, + "loss": 0.6997, + "step": 46066 + }, + { + "epoch": 1.69, + "learning_rate": 3.1482769985717943e-08, + "loss": 0.7019, + "step": 46067 + }, + { + "epoch": 1.69, + "learning_rate": 3.1479107921045883e-08, + "loss": 0.6872, + "step": 46068 + }, + { + "epoch": 1.69, + "learning_rate": 3.147544585637382e-08, + "loss": 0.7039, + "step": 46069 + }, + { + "epoch": 1.69, + "learning_rate": 3.147178379170176e-08, + "loss": 0.6936, + "step": 46070 + }, + { + "epoch": 1.69, + "learning_rate": 3.14681217270297e-08, + "loss": 0.6875, + "step": 46071 + }, + { + "epoch": 1.69, + "learning_rate": 3.1464459662357636e-08, + "loss": 0.6949, + "step": 46072 + }, + { + "epoch": 1.69, + "learning_rate": 3.1460797597685576e-08, + "loss": 0.684, + "step": 46073 + }, + { + "epoch": 1.69, + "learning_rate": 3.145713553301351e-08, + "loss": 0.6949, + "step": 46074 + }, + { + "epoch": 1.69, + "learning_rate": 3.145347346834145e-08, + "loss": 0.6865, + "step": 46075 + }, + { + "epoch": 1.69, + "learning_rate": 3.144981140366938e-08, + "loss": 0.7005, + "step": 46076 + }, + { + "epoch": 1.69, + "learning_rate": 3.144614933899732e-08, + "loss": 0.6914, + "step": 46077 + }, + { + "epoch": 1.69, + "learning_rate": 3.144248727432526e-08, + "loss": 0.6949, + "step": 46078 + }, + { + "epoch": 1.69, + "learning_rate": 3.14388252096532e-08, + "loss": 0.6965, + "step": 46079 + }, + { + "epoch": 1.69, + "learning_rate": 3.143516314498114e-08, + "loss": 0.6857, + "step": 46080 + }, + { + "epoch": 1.69, + "learning_rate": 3.1431501080309076e-08, + "loss": 0.6949, + "step": 46081 + }, + { + "epoch": 1.69, + "learning_rate": 3.1427839015637016e-08, + "loss": 0.6878, + "step": 46082 + }, + { + "epoch": 1.69, + "learning_rate": 3.1424176950964956e-08, + "loss": 0.6918, + "step": 46083 + }, + { + "epoch": 1.69, + "learning_rate": 3.1420514886292896e-08, + "loss": 0.7003, + "step": 46084 + }, + { + "epoch": 1.69, + "learning_rate": 3.141685282162083e-08, + "loss": 0.6955, + "step": 46085 + }, + { + "epoch": 1.69, + "learning_rate": 3.141319075694876e-08, + "loss": 0.6911, + "step": 46086 + }, + { + "epoch": 1.69, + "learning_rate": 3.14095286922767e-08, + "loss": 0.6685, + "step": 46087 + }, + { + "epoch": 1.69, + "learning_rate": 3.140586662760464e-08, + "loss": 0.6934, + "step": 46088 + }, + { + "epoch": 1.69, + "learning_rate": 3.140220456293258e-08, + "loss": 0.6916, + "step": 46089 + }, + { + "epoch": 1.69, + "learning_rate": 3.1398542498260515e-08, + "loss": 0.7022, + "step": 46090 + }, + { + "epoch": 1.69, + "learning_rate": 3.1394880433588455e-08, + "loss": 0.6961, + "step": 46091 + }, + { + "epoch": 1.69, + "learning_rate": 3.1391218368916395e-08, + "loss": 0.6863, + "step": 46092 + }, + { + "epoch": 1.69, + "learning_rate": 3.1387556304244335e-08, + "loss": 0.6725, + "step": 46093 + }, + { + "epoch": 1.69, + "learning_rate": 3.138389423957227e-08, + "loss": 0.6993, + "step": 46094 + }, + { + "epoch": 1.69, + "learning_rate": 3.138023217490021e-08, + "loss": 0.6939, + "step": 46095 + }, + { + "epoch": 1.69, + "learning_rate": 3.137657011022815e-08, + "loss": 0.6942, + "step": 46096 + }, + { + "epoch": 1.69, + "learning_rate": 3.137290804555608e-08, + "loss": 0.7053, + "step": 46097 + }, + { + "epoch": 1.69, + "learning_rate": 3.136924598088402e-08, + "loss": 0.6834, + "step": 46098 + }, + { + "epoch": 1.69, + "learning_rate": 3.1365583916211955e-08, + "loss": 0.6919, + "step": 46099 + }, + { + "epoch": 1.69, + "learning_rate": 3.1361921851539895e-08, + "loss": 0.6911, + "step": 46100 + }, + { + "epoch": 1.69, + "learning_rate": 3.1358259786867835e-08, + "loss": 0.6842, + "step": 46101 + }, + { + "epoch": 1.69, + "learning_rate": 3.1354597722195775e-08, + "loss": 0.6929, + "step": 46102 + }, + { + "epoch": 1.69, + "learning_rate": 3.135093565752371e-08, + "loss": 0.6764, + "step": 46103 + }, + { + "epoch": 1.69, + "learning_rate": 3.134727359285165e-08, + "loss": 0.6962, + "step": 46104 + }, + { + "epoch": 1.69, + "learning_rate": 3.134361152817959e-08, + "loss": 0.6932, + "step": 46105 + }, + { + "epoch": 1.69, + "learning_rate": 3.133994946350753e-08, + "loss": 0.7103, + "step": 46106 + }, + { + "epoch": 1.69, + "learning_rate": 3.133628739883546e-08, + "loss": 0.6887, + "step": 46107 + }, + { + "epoch": 1.69, + "learning_rate": 3.13326253341634e-08, + "loss": 0.6948, + "step": 46108 + }, + { + "epoch": 1.69, + "learning_rate": 3.1328963269491334e-08, + "loss": 0.6887, + "step": 46109 + }, + { + "epoch": 1.69, + "learning_rate": 3.1325301204819274e-08, + "loss": 0.6956, + "step": 46110 + }, + { + "epoch": 1.69, + "learning_rate": 3.1321639140147214e-08, + "loss": 0.6922, + "step": 46111 + }, + { + "epoch": 1.69, + "learning_rate": 3.131797707547515e-08, + "loss": 0.692, + "step": 46112 + }, + { + "epoch": 1.69, + "learning_rate": 3.131431501080309e-08, + "loss": 0.6939, + "step": 46113 + }, + { + "epoch": 1.69, + "learning_rate": 3.131065294613103e-08, + "loss": 0.6933, + "step": 46114 + }, + { + "epoch": 1.69, + "learning_rate": 3.130699088145897e-08, + "loss": 0.6961, + "step": 46115 + }, + { + "epoch": 1.69, + "learning_rate": 3.13033288167869e-08, + "loss": 0.6772, + "step": 46116 + }, + { + "epoch": 1.69, + "learning_rate": 3.129966675211484e-08, + "loss": 0.6991, + "step": 46117 + }, + { + "epoch": 1.69, + "learning_rate": 3.129600468744278e-08, + "loss": 0.6878, + "step": 46118 + }, + { + "epoch": 1.69, + "learning_rate": 3.129234262277072e-08, + "loss": 0.6751, + "step": 46119 + }, + { + "epoch": 1.69, + "learning_rate": 3.1288680558098654e-08, + "loss": 0.6949, + "step": 46120 + }, + { + "epoch": 1.69, + "learning_rate": 3.128501849342659e-08, + "loss": 0.6957, + "step": 46121 + }, + { + "epoch": 1.69, + "learning_rate": 3.128135642875453e-08, + "loss": 0.6995, + "step": 46122 + }, + { + "epoch": 1.69, + "learning_rate": 3.127769436408247e-08, + "loss": 0.7099, + "step": 46123 + }, + { + "epoch": 1.69, + "learning_rate": 3.127403229941041e-08, + "loss": 0.6938, + "step": 46124 + }, + { + "epoch": 1.69, + "learning_rate": 3.127037023473834e-08, + "loss": 0.709, + "step": 46125 + }, + { + "epoch": 1.69, + "learning_rate": 3.126670817006628e-08, + "loss": 0.6978, + "step": 46126 + }, + { + "epoch": 1.69, + "learning_rate": 3.126304610539422e-08, + "loss": 0.6835, + "step": 46127 + }, + { + "epoch": 1.69, + "learning_rate": 3.125938404072216e-08, + "loss": 0.697, + "step": 46128 + }, + { + "epoch": 1.69, + "learning_rate": 3.12557219760501e-08, + "loss": 0.7041, + "step": 46129 + }, + { + "epoch": 1.69, + "learning_rate": 3.125205991137803e-08, + "loss": 0.6978, + "step": 46130 + }, + { + "epoch": 1.69, + "learning_rate": 3.124839784670597e-08, + "loss": 0.6943, + "step": 46131 + }, + { + "epoch": 1.69, + "learning_rate": 3.1244735782033906e-08, + "loss": 0.6827, + "step": 46132 + }, + { + "epoch": 1.69, + "learning_rate": 3.1241073717361846e-08, + "loss": 0.7001, + "step": 46133 + }, + { + "epoch": 1.69, + "learning_rate": 3.123741165268978e-08, + "loss": 0.6929, + "step": 46134 + }, + { + "epoch": 1.69, + "learning_rate": 3.123374958801772e-08, + "loss": 0.6952, + "step": 46135 + }, + { + "epoch": 1.69, + "learning_rate": 3.123008752334566e-08, + "loss": 0.6872, + "step": 46136 + }, + { + "epoch": 1.69, + "learning_rate": 3.12264254586736e-08, + "loss": 0.6919, + "step": 46137 + }, + { + "epoch": 1.69, + "learning_rate": 3.122276339400154e-08, + "loss": 0.6903, + "step": 46138 + }, + { + "epoch": 1.69, + "learning_rate": 3.121910132932947e-08, + "loss": 0.6965, + "step": 46139 + }, + { + "epoch": 1.69, + "learning_rate": 3.121543926465741e-08, + "loss": 0.701, + "step": 46140 + }, + { + "epoch": 1.69, + "learning_rate": 3.121177719998535e-08, + "loss": 0.6794, + "step": 46141 + }, + { + "epoch": 1.69, + "learning_rate": 3.120811513531329e-08, + "loss": 0.6885, + "step": 46142 + }, + { + "epoch": 1.69, + "learning_rate": 3.1204453070641226e-08, + "loss": 0.6987, + "step": 46143 + }, + { + "epoch": 1.69, + "learning_rate": 3.1200791005969166e-08, + "loss": 0.682, + "step": 46144 + }, + { + "epoch": 1.69, + "learning_rate": 3.11971289412971e-08, + "loss": 0.7038, + "step": 46145 + }, + { + "epoch": 1.69, + "learning_rate": 3.119346687662504e-08, + "loss": 0.6853, + "step": 46146 + }, + { + "epoch": 1.69, + "learning_rate": 3.118980481195298e-08, + "loss": 0.684, + "step": 46147 + }, + { + "epoch": 1.69, + "learning_rate": 3.118614274728091e-08, + "loss": 0.6867, + "step": 46148 + }, + { + "epoch": 1.69, + "learning_rate": 3.118248068260885e-08, + "loss": 0.7017, + "step": 46149 + }, + { + "epoch": 1.69, + "learning_rate": 3.117881861793679e-08, + "loss": 0.6777, + "step": 46150 + }, + { + "epoch": 1.69, + "learning_rate": 3.117515655326473e-08, + "loss": 0.6819, + "step": 46151 + }, + { + "epoch": 1.69, + "learning_rate": 3.1171494488592665e-08, + "loss": 0.6953, + "step": 46152 + }, + { + "epoch": 1.69, + "learning_rate": 3.1167832423920605e-08, + "loss": 0.6921, + "step": 46153 + }, + { + "epoch": 1.69, + "learning_rate": 3.1164170359248545e-08, + "loss": 0.6993, + "step": 46154 + }, + { + "epoch": 1.69, + "learning_rate": 3.1160508294576485e-08, + "loss": 0.6918, + "step": 46155 + }, + { + "epoch": 1.69, + "learning_rate": 3.115684622990442e-08, + "loss": 0.694, + "step": 46156 + }, + { + "epoch": 1.69, + "learning_rate": 3.115318416523235e-08, + "loss": 0.6878, + "step": 46157 + }, + { + "epoch": 1.69, + "learning_rate": 3.114952210056029e-08, + "loss": 0.685, + "step": 46158 + }, + { + "epoch": 1.69, + "learning_rate": 3.114586003588823e-08, + "loss": 0.6885, + "step": 46159 + }, + { + "epoch": 1.69, + "learning_rate": 3.114219797121617e-08, + "loss": 0.6921, + "step": 46160 + }, + { + "epoch": 1.69, + "learning_rate": 3.1138535906544105e-08, + "loss": 0.691, + "step": 46161 + }, + { + "epoch": 1.69, + "learning_rate": 3.1134873841872045e-08, + "loss": 0.7016, + "step": 46162 + }, + { + "epoch": 1.69, + "learning_rate": 3.1131211777199985e-08, + "loss": 0.6881, + "step": 46163 + }, + { + "epoch": 1.69, + "learning_rate": 3.1127549712527925e-08, + "loss": 0.6998, + "step": 46164 + }, + { + "epoch": 1.69, + "learning_rate": 3.1123887647855865e-08, + "loss": 0.7001, + "step": 46165 + }, + { + "epoch": 1.69, + "learning_rate": 3.11202255831838e-08, + "loss": 0.6963, + "step": 46166 + }, + { + "epoch": 1.69, + "learning_rate": 3.111656351851174e-08, + "loss": 0.6912, + "step": 46167 + }, + { + "epoch": 1.69, + "learning_rate": 3.111290145383967e-08, + "loss": 0.6887, + "step": 46168 + }, + { + "epoch": 1.69, + "learning_rate": 3.110923938916761e-08, + "loss": 0.698, + "step": 46169 + }, + { + "epoch": 1.69, + "learning_rate": 3.1105577324495544e-08, + "loss": 0.6903, + "step": 46170 + }, + { + "epoch": 1.69, + "learning_rate": 3.1101915259823484e-08, + "loss": 0.6945, + "step": 46171 + }, + { + "epoch": 1.69, + "learning_rate": 3.1098253195151424e-08, + "loss": 0.6855, + "step": 46172 + }, + { + "epoch": 1.69, + "learning_rate": 3.1094591130479364e-08, + "loss": 0.6938, + "step": 46173 + }, + { + "epoch": 1.69, + "learning_rate": 3.1090929065807304e-08, + "loss": 0.6843, + "step": 46174 + }, + { + "epoch": 1.69, + "learning_rate": 3.108726700113524e-08, + "loss": 0.6968, + "step": 46175 + }, + { + "epoch": 1.69, + "learning_rate": 3.108360493646318e-08, + "loss": 0.6813, + "step": 46176 + }, + { + "epoch": 1.69, + "learning_rate": 3.107994287179112e-08, + "loss": 0.6948, + "step": 46177 + }, + { + "epoch": 1.69, + "learning_rate": 3.107628080711906e-08, + "loss": 0.6926, + "step": 46178 + }, + { + "epoch": 1.69, + "learning_rate": 3.107261874244699e-08, + "loss": 0.6822, + "step": 46179 + }, + { + "epoch": 1.69, + "learning_rate": 3.1068956677774924e-08, + "loss": 0.6911, + "step": 46180 + }, + { + "epoch": 1.69, + "learning_rate": 3.1065294613102864e-08, + "loss": 0.6869, + "step": 46181 + }, + { + "epoch": 1.69, + "learning_rate": 3.1061632548430804e-08, + "loss": 0.6978, + "step": 46182 + }, + { + "epoch": 1.69, + "learning_rate": 3.1057970483758744e-08, + "loss": 0.6969, + "step": 46183 + }, + { + "epoch": 1.69, + "learning_rate": 3.105430841908668e-08, + "loss": 0.7009, + "step": 46184 + }, + { + "epoch": 1.69, + "learning_rate": 3.105064635441462e-08, + "loss": 0.6974, + "step": 46185 + }, + { + "epoch": 1.69, + "learning_rate": 3.104698428974256e-08, + "loss": 0.692, + "step": 46186 + }, + { + "epoch": 1.69, + "learning_rate": 3.1043322225070497e-08, + "loss": 0.6826, + "step": 46187 + }, + { + "epoch": 1.69, + "learning_rate": 3.103966016039843e-08, + "loss": 0.6934, + "step": 46188 + }, + { + "epoch": 1.69, + "learning_rate": 3.103599809572637e-08, + "loss": 0.6908, + "step": 46189 + }, + { + "epoch": 1.69, + "learning_rate": 3.103233603105431e-08, + "loss": 0.6926, + "step": 46190 + }, + { + "epoch": 1.69, + "learning_rate": 3.102867396638224e-08, + "loss": 0.6945, + "step": 46191 + }, + { + "epoch": 1.69, + "learning_rate": 3.102501190171018e-08, + "loss": 0.6878, + "step": 46192 + }, + { + "epoch": 1.69, + "learning_rate": 3.1021349837038116e-08, + "loss": 0.6919, + "step": 46193 + }, + { + "epoch": 1.69, + "learning_rate": 3.1017687772366056e-08, + "loss": 0.6934, + "step": 46194 + }, + { + "epoch": 1.69, + "learning_rate": 3.1014025707693996e-08, + "loss": 0.7002, + "step": 46195 + }, + { + "epoch": 1.69, + "learning_rate": 3.1010363643021936e-08, + "loss": 0.6915, + "step": 46196 + }, + { + "epoch": 1.69, + "learning_rate": 3.100670157834987e-08, + "loss": 0.6926, + "step": 46197 + }, + { + "epoch": 1.69, + "learning_rate": 3.100303951367781e-08, + "loss": 0.7109, + "step": 46198 + }, + { + "epoch": 1.69, + "learning_rate": 3.099937744900575e-08, + "loss": 0.7006, + "step": 46199 + }, + { + "epoch": 1.69, + "learning_rate": 3.099571538433369e-08, + "loss": 0.6997, + "step": 46200 + }, + { + "epoch": 1.69, + "learning_rate": 3.099205331966163e-08, + "loss": 0.7029, + "step": 46201 + }, + { + "epoch": 1.69, + "learning_rate": 3.098839125498956e-08, + "loss": 0.698, + "step": 46202 + }, + { + "epoch": 1.69, + "learning_rate": 3.0984729190317496e-08, + "loss": 0.6849, + "step": 46203 + }, + { + "epoch": 1.69, + "learning_rate": 3.0981067125645436e-08, + "loss": 0.698, + "step": 46204 + }, + { + "epoch": 1.69, + "learning_rate": 3.0977405060973376e-08, + "loss": 0.7101, + "step": 46205 + }, + { + "epoch": 1.69, + "learning_rate": 3.097374299630131e-08, + "loss": 0.705, + "step": 46206 + }, + { + "epoch": 1.69, + "learning_rate": 3.097008093162925e-08, + "loss": 0.6868, + "step": 46207 + }, + { + "epoch": 1.69, + "learning_rate": 3.096641886695719e-08, + "loss": 0.6945, + "step": 46208 + }, + { + "epoch": 1.69, + "learning_rate": 3.096275680228513e-08, + "loss": 0.6921, + "step": 46209 + }, + { + "epoch": 1.69, + "learning_rate": 3.095909473761307e-08, + "loss": 0.6875, + "step": 46210 + }, + { + "epoch": 1.69, + "learning_rate": 3.0955432672941e-08, + "loss": 0.681, + "step": 46211 + }, + { + "epoch": 1.69, + "learning_rate": 3.095177060826894e-08, + "loss": 0.6836, + "step": 46212 + }, + { + "epoch": 1.69, + "learning_rate": 3.094810854359688e-08, + "loss": 0.6839, + "step": 46213 + }, + { + "epoch": 1.69, + "learning_rate": 3.0944446478924815e-08, + "loss": 0.7078, + "step": 46214 + }, + { + "epoch": 1.69, + "learning_rate": 3.0940784414252755e-08, + "loss": 0.7004, + "step": 46215 + }, + { + "epoch": 1.69, + "learning_rate": 3.093712234958069e-08, + "loss": 0.6941, + "step": 46216 + }, + { + "epoch": 1.69, + "learning_rate": 3.093346028490863e-08, + "loss": 0.6932, + "step": 46217 + }, + { + "epoch": 1.69, + "learning_rate": 3.092979822023657e-08, + "loss": 0.6915, + "step": 46218 + }, + { + "epoch": 1.69, + "learning_rate": 3.092613615556451e-08, + "loss": 0.695, + "step": 46219 + }, + { + "epoch": 1.69, + "learning_rate": 3.092247409089244e-08, + "loss": 0.6978, + "step": 46220 + }, + { + "epoch": 1.69, + "learning_rate": 3.091881202622038e-08, + "loss": 0.6918, + "step": 46221 + }, + { + "epoch": 1.69, + "learning_rate": 3.091514996154832e-08, + "loss": 0.6875, + "step": 46222 + }, + { + "epoch": 1.69, + "learning_rate": 3.091148789687626e-08, + "loss": 0.693, + "step": 46223 + }, + { + "epoch": 1.69, + "learning_rate": 3.0907825832204195e-08, + "loss": 0.689, + "step": 46224 + }, + { + "epoch": 1.69, + "learning_rate": 3.0904163767532135e-08, + "loss": 0.6959, + "step": 46225 + }, + { + "epoch": 1.69, + "learning_rate": 3.0900501702860074e-08, + "loss": 0.6832, + "step": 46226 + }, + { + "epoch": 1.69, + "learning_rate": 3.089683963818801e-08, + "loss": 0.6843, + "step": 46227 + }, + { + "epoch": 1.69, + "learning_rate": 3.089317757351595e-08, + "loss": 0.6971, + "step": 46228 + }, + { + "epoch": 1.69, + "learning_rate": 3.088951550884388e-08, + "loss": 0.6956, + "step": 46229 + }, + { + "epoch": 1.69, + "learning_rate": 3.088585344417182e-08, + "loss": 0.6877, + "step": 46230 + }, + { + "epoch": 1.69, + "learning_rate": 3.088219137949976e-08, + "loss": 0.6914, + "step": 46231 + }, + { + "epoch": 1.69, + "learning_rate": 3.08785293148277e-08, + "loss": 0.6778, + "step": 46232 + }, + { + "epoch": 1.69, + "learning_rate": 3.0874867250155634e-08, + "loss": 0.6979, + "step": 46233 + }, + { + "epoch": 1.69, + "learning_rate": 3.0871205185483574e-08, + "loss": 0.6916, + "step": 46234 + }, + { + "epoch": 1.69, + "learning_rate": 3.0867543120811514e-08, + "loss": 0.7127, + "step": 46235 + }, + { + "epoch": 1.69, + "learning_rate": 3.0863881056139454e-08, + "loss": 0.6907, + "step": 46236 + }, + { + "epoch": 1.69, + "learning_rate": 3.086021899146739e-08, + "loss": 0.6928, + "step": 46237 + }, + { + "epoch": 1.69, + "learning_rate": 3.085655692679533e-08, + "loss": 0.6859, + "step": 46238 + }, + { + "epoch": 1.69, + "learning_rate": 3.085289486212326e-08, + "loss": 0.6999, + "step": 46239 + }, + { + "epoch": 1.69, + "learning_rate": 3.08492327974512e-08, + "loss": 0.7001, + "step": 46240 + }, + { + "epoch": 1.69, + "learning_rate": 3.084557073277914e-08, + "loss": 0.7007, + "step": 46241 + }, + { + "epoch": 1.69, + "learning_rate": 3.0841908668107074e-08, + "loss": 0.6748, + "step": 46242 + }, + { + "epoch": 1.69, + "learning_rate": 3.0838246603435014e-08, + "loss": 0.69, + "step": 46243 + }, + { + "epoch": 1.69, + "learning_rate": 3.0834584538762953e-08, + "loss": 0.6988, + "step": 46244 + }, + { + "epoch": 1.69, + "learning_rate": 3.0830922474090893e-08, + "loss": 0.6996, + "step": 46245 + }, + { + "epoch": 1.69, + "learning_rate": 3.0827260409418833e-08, + "loss": 0.6799, + "step": 46246 + }, + { + "epoch": 1.69, + "learning_rate": 3.0823598344746767e-08, + "loss": 0.6652, + "step": 46247 + }, + { + "epoch": 1.69, + "learning_rate": 3.0819936280074707e-08, + "loss": 0.6927, + "step": 46248 + }, + { + "epoch": 1.69, + "learning_rate": 3.0816274215402646e-08, + "loss": 0.6929, + "step": 46249 + }, + { + "epoch": 1.69, + "learning_rate": 3.081261215073058e-08, + "loss": 0.6911, + "step": 46250 + }, + { + "epoch": 1.69, + "learning_rate": 3.080895008605851e-08, + "loss": 0.7041, + "step": 46251 + }, + { + "epoch": 1.69, + "learning_rate": 3.080528802138645e-08, + "loss": 0.6961, + "step": 46252 + }, + { + "epoch": 1.69, + "learning_rate": 3.080162595671439e-08, + "loss": 0.696, + "step": 46253 + }, + { + "epoch": 1.69, + "learning_rate": 3.079796389204233e-08, + "loss": 0.6903, + "step": 46254 + }, + { + "epoch": 1.69, + "learning_rate": 3.079430182737027e-08, + "loss": 0.6788, + "step": 46255 + }, + { + "epoch": 1.69, + "learning_rate": 3.0790639762698206e-08, + "loss": 0.7006, + "step": 46256 + }, + { + "epoch": 1.69, + "learning_rate": 3.0786977698026146e-08, + "loss": 0.7041, + "step": 46257 + }, + { + "epoch": 1.69, + "learning_rate": 3.0783315633354086e-08, + "loss": 0.6912, + "step": 46258 + }, + { + "epoch": 1.69, + "learning_rate": 3.0779653568682026e-08, + "loss": 0.6975, + "step": 46259 + }, + { + "epoch": 1.69, + "learning_rate": 3.077599150400996e-08, + "loss": 0.6996, + "step": 46260 + }, + { + "epoch": 1.69, + "learning_rate": 3.07723294393379e-08, + "loss": 0.6924, + "step": 46261 + }, + { + "epoch": 1.69, + "learning_rate": 3.076866737466583e-08, + "loss": 0.691, + "step": 46262 + }, + { + "epoch": 1.69, + "learning_rate": 3.076500530999377e-08, + "loss": 0.6948, + "step": 46263 + }, + { + "epoch": 1.69, + "learning_rate": 3.076134324532171e-08, + "loss": 0.697, + "step": 46264 + }, + { + "epoch": 1.69, + "learning_rate": 3.0757681180649646e-08, + "loss": 0.691, + "step": 46265 + }, + { + "epoch": 1.69, + "learning_rate": 3.0754019115977586e-08, + "loss": 0.6949, + "step": 46266 + }, + { + "epoch": 1.69, + "learning_rate": 3.0750357051305525e-08, + "loss": 0.7039, + "step": 46267 + }, + { + "epoch": 1.69, + "learning_rate": 3.0746694986633465e-08, + "loss": 0.6846, + "step": 46268 + }, + { + "epoch": 1.69, + "learning_rate": 3.07430329219614e-08, + "loss": 0.6923, + "step": 46269 + }, + { + "epoch": 1.69, + "learning_rate": 3.073937085728934e-08, + "loss": 0.6919, + "step": 46270 + }, + { + "epoch": 1.69, + "learning_rate": 3.073570879261728e-08, + "loss": 0.6884, + "step": 46271 + }, + { + "epoch": 1.69, + "learning_rate": 3.073204672794522e-08, + "loss": 0.6964, + "step": 46272 + }, + { + "epoch": 1.69, + "learning_rate": 3.072838466327315e-08, + "loss": 0.698, + "step": 46273 + }, + { + "epoch": 1.69, + "learning_rate": 3.0724722598601085e-08, + "loss": 0.6857, + "step": 46274 + }, + { + "epoch": 1.69, + "learning_rate": 3.0721060533929025e-08, + "loss": 0.6934, + "step": 46275 + }, + { + "epoch": 1.69, + "learning_rate": 3.0717398469256965e-08, + "loss": 0.6901, + "step": 46276 + }, + { + "epoch": 1.69, + "learning_rate": 3.0713736404584905e-08, + "loss": 0.6924, + "step": 46277 + }, + { + "epoch": 1.69, + "learning_rate": 3.071007433991284e-08, + "loss": 0.6927, + "step": 46278 + }, + { + "epoch": 1.69, + "learning_rate": 3.070641227524078e-08, + "loss": 0.7054, + "step": 46279 + }, + { + "epoch": 1.69, + "learning_rate": 3.070275021056872e-08, + "loss": 0.6823, + "step": 46280 + }, + { + "epoch": 1.69, + "learning_rate": 3.069908814589666e-08, + "loss": 0.7012, + "step": 46281 + }, + { + "epoch": 1.69, + "learning_rate": 3.06954260812246e-08, + "loss": 0.6863, + "step": 46282 + }, + { + "epoch": 1.69, + "learning_rate": 3.069176401655253e-08, + "loss": 0.6886, + "step": 46283 + }, + { + "epoch": 1.69, + "learning_rate": 3.068810195188047e-08, + "loss": 0.6976, + "step": 46284 + }, + { + "epoch": 1.69, + "learning_rate": 3.0684439887208405e-08, + "loss": 0.6957, + "step": 46285 + }, + { + "epoch": 1.69, + "learning_rate": 3.0680777822536344e-08, + "loss": 0.6847, + "step": 46286 + }, + { + "epoch": 1.69, + "learning_rate": 3.067711575786428e-08, + "loss": 0.6993, + "step": 46287 + }, + { + "epoch": 1.69, + "learning_rate": 3.067345369319222e-08, + "loss": 0.6984, + "step": 46288 + }, + { + "epoch": 1.69, + "learning_rate": 3.066979162852016e-08, + "loss": 0.6948, + "step": 46289 + }, + { + "epoch": 1.69, + "learning_rate": 3.06661295638481e-08, + "loss": 0.6929, + "step": 46290 + }, + { + "epoch": 1.69, + "learning_rate": 3.066246749917604e-08, + "loss": 0.6903, + "step": 46291 + }, + { + "epoch": 1.69, + "learning_rate": 3.065880543450397e-08, + "loss": 0.6902, + "step": 46292 + }, + { + "epoch": 1.69, + "learning_rate": 3.065514336983191e-08, + "loss": 0.6846, + "step": 46293 + }, + { + "epoch": 1.69, + "learning_rate": 3.065148130515985e-08, + "loss": 0.6945, + "step": 46294 + }, + { + "epoch": 1.69, + "learning_rate": 3.064781924048779e-08, + "loss": 0.695, + "step": 46295 + }, + { + "epoch": 1.69, + "learning_rate": 3.0644157175815724e-08, + "loss": 0.6988, + "step": 46296 + }, + { + "epoch": 1.69, + "learning_rate": 3.064049511114366e-08, + "loss": 0.6971, + "step": 46297 + }, + { + "epoch": 1.69, + "learning_rate": 3.06368330464716e-08, + "loss": 0.6931, + "step": 46298 + }, + { + "epoch": 1.69, + "learning_rate": 3.063317098179954e-08, + "loss": 0.6999, + "step": 46299 + }, + { + "epoch": 1.69, + "learning_rate": 3.062950891712747e-08, + "loss": 0.6961, + "step": 46300 + }, + { + "epoch": 1.69, + "learning_rate": 3.062584685245541e-08, + "loss": 0.702, + "step": 46301 + }, + { + "epoch": 1.69, + "learning_rate": 3.062218478778335e-08, + "loss": 0.6825, + "step": 46302 + }, + { + "epoch": 1.69, + "learning_rate": 3.061852272311129e-08, + "loss": 0.6961, + "step": 46303 + }, + { + "epoch": 1.69, + "learning_rate": 3.061486065843923e-08, + "loss": 0.6945, + "step": 46304 + }, + { + "epoch": 1.69, + "learning_rate": 3.0611198593767163e-08, + "loss": 0.6951, + "step": 46305 + }, + { + "epoch": 1.69, + "learning_rate": 3.06075365290951e-08, + "loss": 0.6852, + "step": 46306 + }, + { + "epoch": 1.69, + "learning_rate": 3.060387446442304e-08, + "loss": 0.6923, + "step": 46307 + }, + { + "epoch": 1.69, + "learning_rate": 3.0600212399750977e-08, + "loss": 0.6836, + "step": 46308 + }, + { + "epoch": 1.69, + "learning_rate": 3.0596550335078916e-08, + "loss": 0.6898, + "step": 46309 + }, + { + "epoch": 1.69, + "learning_rate": 3.059288827040685e-08, + "loss": 0.688, + "step": 46310 + }, + { + "epoch": 1.69, + "learning_rate": 3.058922620573479e-08, + "loss": 0.699, + "step": 46311 + }, + { + "epoch": 1.69, + "learning_rate": 3.058556414106273e-08, + "loss": 0.6927, + "step": 46312 + }, + { + "epoch": 1.69, + "learning_rate": 3.058190207639067e-08, + "loss": 0.6937, + "step": 46313 + }, + { + "epoch": 1.69, + "learning_rate": 3.05782400117186e-08, + "loss": 0.6918, + "step": 46314 + }, + { + "epoch": 1.69, + "learning_rate": 3.057457794704654e-08, + "loss": 0.7039, + "step": 46315 + }, + { + "epoch": 1.69, + "learning_rate": 3.057091588237448e-08, + "loss": 0.684, + "step": 46316 + }, + { + "epoch": 1.69, + "learning_rate": 3.056725381770242e-08, + "loss": 0.6962, + "step": 46317 + }, + { + "epoch": 1.69, + "learning_rate": 3.0563591753030356e-08, + "loss": 0.7032, + "step": 46318 + }, + { + "epoch": 1.69, + "learning_rate": 3.0559929688358296e-08, + "loss": 0.6789, + "step": 46319 + }, + { + "epoch": 1.69, + "learning_rate": 3.0556267623686236e-08, + "loss": 0.6986, + "step": 46320 + }, + { + "epoch": 1.69, + "learning_rate": 3.055260555901417e-08, + "loss": 0.6979, + "step": 46321 + }, + { + "epoch": 1.69, + "learning_rate": 3.054894349434211e-08, + "loss": 0.6995, + "step": 46322 + }, + { + "epoch": 1.69, + "learning_rate": 3.054528142967004e-08, + "loss": 0.6974, + "step": 46323 + }, + { + "epoch": 1.69, + "learning_rate": 3.054161936499798e-08, + "loss": 0.7006, + "step": 46324 + }, + { + "epoch": 1.69, + "learning_rate": 3.053795730032592e-08, + "loss": 0.6891, + "step": 46325 + }, + { + "epoch": 1.69, + "learning_rate": 3.053429523565386e-08, + "loss": 0.6997, + "step": 46326 + }, + { + "epoch": 1.69, + "learning_rate": 3.0530633170981795e-08, + "loss": 0.6858, + "step": 46327 + }, + { + "epoch": 1.7, + "learning_rate": 3.0526971106309735e-08, + "loss": 0.6962, + "step": 46328 + }, + { + "epoch": 1.7, + "learning_rate": 3.0523309041637675e-08, + "loss": 0.6961, + "step": 46329 + }, + { + "epoch": 1.7, + "learning_rate": 3.0519646976965615e-08, + "loss": 0.697, + "step": 46330 + }, + { + "epoch": 1.7, + "learning_rate": 3.0515984912293555e-08, + "loss": 0.6926, + "step": 46331 + }, + { + "epoch": 1.7, + "learning_rate": 3.051232284762149e-08, + "loss": 0.6921, + "step": 46332 + }, + { + "epoch": 1.7, + "learning_rate": 3.050866078294942e-08, + "loss": 0.6964, + "step": 46333 + }, + { + "epoch": 1.7, + "learning_rate": 3.050499871827736e-08, + "loss": 0.6881, + "step": 46334 + }, + { + "epoch": 1.7, + "learning_rate": 3.05013366536053e-08, + "loss": 0.6876, + "step": 46335 + }, + { + "epoch": 1.7, + "learning_rate": 3.0497674588933235e-08, + "loss": 0.6969, + "step": 46336 + }, + { + "epoch": 1.7, + "learning_rate": 3.0494012524261175e-08, + "loss": 0.6995, + "step": 46337 + }, + { + "epoch": 1.7, + "learning_rate": 3.0490350459589115e-08, + "loss": 0.693, + "step": 46338 + }, + { + "epoch": 1.7, + "learning_rate": 3.0486688394917055e-08, + "loss": 0.6863, + "step": 46339 + }, + { + "epoch": 1.7, + "learning_rate": 3.0483026330244995e-08, + "loss": 0.6965, + "step": 46340 + }, + { + "epoch": 1.7, + "learning_rate": 3.047936426557293e-08, + "loss": 0.6959, + "step": 46341 + }, + { + "epoch": 1.7, + "learning_rate": 3.047570220090087e-08, + "loss": 0.7032, + "step": 46342 + }, + { + "epoch": 1.7, + "learning_rate": 3.047204013622881e-08, + "loss": 0.6933, + "step": 46343 + }, + { + "epoch": 1.7, + "learning_rate": 3.046837807155674e-08, + "loss": 0.6818, + "step": 46344 + }, + { + "epoch": 1.7, + "learning_rate": 3.0464716006884674e-08, + "loss": 0.6919, + "step": 46345 + }, + { + "epoch": 1.7, + "learning_rate": 3.0461053942212614e-08, + "loss": 0.6963, + "step": 46346 + }, + { + "epoch": 1.7, + "learning_rate": 3.0457391877540554e-08, + "loss": 0.7085, + "step": 46347 + }, + { + "epoch": 1.7, + "learning_rate": 3.0453729812868494e-08, + "loss": 0.7022, + "step": 46348 + }, + { + "epoch": 1.7, + "learning_rate": 3.0450067748196434e-08, + "loss": 0.6922, + "step": 46349 + }, + { + "epoch": 1.7, + "learning_rate": 3.044640568352437e-08, + "loss": 0.6982, + "step": 46350 + }, + { + "epoch": 1.7, + "learning_rate": 3.044274361885231e-08, + "loss": 0.6827, + "step": 46351 + }, + { + "epoch": 1.7, + "learning_rate": 3.043908155418025e-08, + "loss": 0.6776, + "step": 46352 + }, + { + "epoch": 1.7, + "learning_rate": 3.043541948950819e-08, + "loss": 0.6756, + "step": 46353 + }, + { + "epoch": 1.7, + "learning_rate": 3.043175742483612e-08, + "loss": 0.695, + "step": 46354 + }, + { + "epoch": 1.7, + "learning_rate": 3.042809536016406e-08, + "loss": 0.6903, + "step": 46355 + }, + { + "epoch": 1.7, + "learning_rate": 3.0424433295491994e-08, + "loss": 0.6964, + "step": 46356 + }, + { + "epoch": 1.7, + "learning_rate": 3.0420771230819934e-08, + "loss": 0.6848, + "step": 46357 + }, + { + "epoch": 1.7, + "learning_rate": 3.0417109166147874e-08, + "loss": 0.692, + "step": 46358 + }, + { + "epoch": 1.7, + "learning_rate": 3.041344710147581e-08, + "loss": 0.6885, + "step": 46359 + }, + { + "epoch": 1.7, + "learning_rate": 3.040978503680375e-08, + "loss": 0.6802, + "step": 46360 + }, + { + "epoch": 1.7, + "learning_rate": 3.040612297213169e-08, + "loss": 0.6936, + "step": 46361 + }, + { + "epoch": 1.7, + "learning_rate": 3.040246090745963e-08, + "loss": 0.6848, + "step": 46362 + }, + { + "epoch": 1.7, + "learning_rate": 3.039879884278756e-08, + "loss": 0.6947, + "step": 46363 + }, + { + "epoch": 1.7, + "learning_rate": 3.03951367781155e-08, + "loss": 0.6979, + "step": 46364 + }, + { + "epoch": 1.7, + "learning_rate": 3.039147471344344e-08, + "loss": 0.7062, + "step": 46365 + }, + { + "epoch": 1.7, + "learning_rate": 3.038781264877138e-08, + "loss": 0.6979, + "step": 46366 + }, + { + "epoch": 1.7, + "learning_rate": 3.038415058409931e-08, + "loss": 0.6828, + "step": 46367 + }, + { + "epoch": 1.7, + "learning_rate": 3.0380488519427247e-08, + "loss": 0.6998, + "step": 46368 + }, + { + "epoch": 1.7, + "learning_rate": 3.0376826454755186e-08, + "loss": 0.6949, + "step": 46369 + }, + { + "epoch": 1.7, + "learning_rate": 3.0373164390083126e-08, + "loss": 0.6973, + "step": 46370 + }, + { + "epoch": 1.7, + "learning_rate": 3.0369502325411066e-08, + "loss": 0.6985, + "step": 46371 + }, + { + "epoch": 1.7, + "learning_rate": 3.0365840260739e-08, + "loss": 0.6996, + "step": 46372 + }, + { + "epoch": 1.7, + "learning_rate": 3.036217819606694e-08, + "loss": 0.6918, + "step": 46373 + }, + { + "epoch": 1.7, + "learning_rate": 3.035851613139488e-08, + "loss": 0.6955, + "step": 46374 + }, + { + "epoch": 1.7, + "learning_rate": 3.035485406672282e-08, + "loss": 0.6992, + "step": 46375 + }, + { + "epoch": 1.7, + "learning_rate": 3.035119200205076e-08, + "loss": 0.6996, + "step": 46376 + }, + { + "epoch": 1.7, + "learning_rate": 3.034752993737869e-08, + "loss": 0.6955, + "step": 46377 + }, + { + "epoch": 1.7, + "learning_rate": 3.034386787270663e-08, + "loss": 0.6968, + "step": 46378 + }, + { + "epoch": 1.7, + "learning_rate": 3.0340205808034566e-08, + "loss": 0.6886, + "step": 46379 + }, + { + "epoch": 1.7, + "learning_rate": 3.0336543743362506e-08, + "loss": 0.6851, + "step": 46380 + }, + { + "epoch": 1.7, + "learning_rate": 3.033288167869044e-08, + "loss": 0.6766, + "step": 46381 + }, + { + "epoch": 1.7, + "learning_rate": 3.032921961401838e-08, + "loss": 0.6916, + "step": 46382 + }, + { + "epoch": 1.7, + "learning_rate": 3.032555754934632e-08, + "loss": 0.6898, + "step": 46383 + }, + { + "epoch": 1.7, + "learning_rate": 3.032189548467426e-08, + "loss": 0.7006, + "step": 46384 + }, + { + "epoch": 1.7, + "learning_rate": 3.03182334200022e-08, + "loss": 0.6921, + "step": 46385 + }, + { + "epoch": 1.7, + "learning_rate": 3.031457135533013e-08, + "loss": 0.693, + "step": 46386 + }, + { + "epoch": 1.7, + "learning_rate": 3.031090929065807e-08, + "loss": 0.7124, + "step": 46387 + }, + { + "epoch": 1.7, + "learning_rate": 3.030724722598601e-08, + "loss": 0.691, + "step": 46388 + }, + { + "epoch": 1.7, + "learning_rate": 3.030358516131395e-08, + "loss": 0.6948, + "step": 46389 + }, + { + "epoch": 1.7, + "learning_rate": 3.0299923096641885e-08, + "loss": 0.6961, + "step": 46390 + }, + { + "epoch": 1.7, + "learning_rate": 3.0296261031969825e-08, + "loss": 0.7072, + "step": 46391 + }, + { + "epoch": 1.7, + "learning_rate": 3.029259896729776e-08, + "loss": 0.6927, + "step": 46392 + }, + { + "epoch": 1.7, + "learning_rate": 3.02889369026257e-08, + "loss": 0.6965, + "step": 46393 + }, + { + "epoch": 1.7, + "learning_rate": 3.028527483795364e-08, + "loss": 0.6877, + "step": 46394 + }, + { + "epoch": 1.7, + "learning_rate": 3.028161277328157e-08, + "loss": 0.6883, + "step": 46395 + }, + { + "epoch": 1.7, + "learning_rate": 3.027795070860951e-08, + "loss": 0.6797, + "step": 46396 + }, + { + "epoch": 1.7, + "learning_rate": 3.027428864393745e-08, + "loss": 0.6927, + "step": 46397 + }, + { + "epoch": 1.7, + "learning_rate": 3.027062657926539e-08, + "loss": 0.6847, + "step": 46398 + }, + { + "epoch": 1.7, + "learning_rate": 3.0266964514593325e-08, + "loss": 0.6986, + "step": 46399 + }, + { + "epoch": 1.7, + "learning_rate": 3.0263302449921265e-08, + "loss": 0.7114, + "step": 46400 + }, + { + "epoch": 1.7, + "learning_rate": 3.0259640385249205e-08, + "loss": 0.6992, + "step": 46401 + }, + { + "epoch": 1.7, + "learning_rate": 3.0255978320577145e-08, + "loss": 0.6981, + "step": 46402 + }, + { + "epoch": 1.7, + "learning_rate": 3.025231625590508e-08, + "loss": 0.6964, + "step": 46403 + }, + { + "epoch": 1.7, + "learning_rate": 3.024865419123301e-08, + "loss": 0.6955, + "step": 46404 + }, + { + "epoch": 1.7, + "learning_rate": 3.024499212656095e-08, + "loss": 0.6936, + "step": 46405 + }, + { + "epoch": 1.7, + "learning_rate": 3.024133006188889e-08, + "loss": 0.6891, + "step": 46406 + }, + { + "epoch": 1.7, + "learning_rate": 3.023766799721683e-08, + "loss": 0.6991, + "step": 46407 + }, + { + "epoch": 1.7, + "learning_rate": 3.0234005932544764e-08, + "loss": 0.6939, + "step": 46408 + }, + { + "epoch": 1.7, + "learning_rate": 3.0230343867872704e-08, + "loss": 0.6924, + "step": 46409 + }, + { + "epoch": 1.7, + "learning_rate": 3.0226681803200644e-08, + "loss": 0.6864, + "step": 46410 + }, + { + "epoch": 1.7, + "learning_rate": 3.0223019738528584e-08, + "loss": 0.6911, + "step": 46411 + }, + { + "epoch": 1.7, + "learning_rate": 3.0219357673856524e-08, + "loss": 0.6885, + "step": 46412 + }, + { + "epoch": 1.7, + "learning_rate": 3.021569560918446e-08, + "loss": 0.6844, + "step": 46413 + }, + { + "epoch": 1.7, + "learning_rate": 3.02120335445124e-08, + "loss": 0.6954, + "step": 46414 + }, + { + "epoch": 1.7, + "learning_rate": 3.020837147984033e-08, + "loss": 0.7041, + "step": 46415 + }, + { + "epoch": 1.7, + "learning_rate": 3.020470941516827e-08, + "loss": 0.6904, + "step": 46416 + }, + { + "epoch": 1.7, + "learning_rate": 3.0201047350496204e-08, + "loss": 0.6884, + "step": 46417 + }, + { + "epoch": 1.7, + "learning_rate": 3.0197385285824144e-08, + "loss": 0.6986, + "step": 46418 + }, + { + "epoch": 1.7, + "learning_rate": 3.0193723221152084e-08, + "loss": 0.7042, + "step": 46419 + }, + { + "epoch": 1.7, + "learning_rate": 3.0190061156480024e-08, + "loss": 0.699, + "step": 46420 + }, + { + "epoch": 1.7, + "learning_rate": 3.0186399091807963e-08, + "loss": 0.6871, + "step": 46421 + }, + { + "epoch": 1.7, + "learning_rate": 3.01827370271359e-08, + "loss": 0.702, + "step": 46422 + }, + { + "epoch": 1.7, + "learning_rate": 3.017907496246384e-08, + "loss": 0.7071, + "step": 46423 + }, + { + "epoch": 1.7, + "learning_rate": 3.0175412897791777e-08, + "loss": 0.6816, + "step": 46424 + }, + { + "epoch": 1.7, + "learning_rate": 3.0171750833119717e-08, + "loss": 0.6997, + "step": 46425 + }, + { + "epoch": 1.7, + "learning_rate": 3.016808876844765e-08, + "loss": 0.6975, + "step": 46426 + }, + { + "epoch": 1.7, + "learning_rate": 3.016442670377558e-08, + "loss": 0.6931, + "step": 46427 + }, + { + "epoch": 1.7, + "learning_rate": 3.016076463910352e-08, + "loss": 0.6824, + "step": 46428 + }, + { + "epoch": 1.7, + "learning_rate": 3.015710257443146e-08, + "loss": 0.7066, + "step": 46429 + }, + { + "epoch": 1.7, + "learning_rate": 3.01534405097594e-08, + "loss": 0.6942, + "step": 46430 + }, + { + "epoch": 1.7, + "learning_rate": 3.0149778445087336e-08, + "loss": 0.6971, + "step": 46431 + }, + { + "epoch": 1.7, + "learning_rate": 3.0146116380415276e-08, + "loss": 0.6891, + "step": 46432 + }, + { + "epoch": 1.7, + "learning_rate": 3.0142454315743216e-08, + "loss": 0.6995, + "step": 46433 + }, + { + "epoch": 1.7, + "learning_rate": 3.0138792251071156e-08, + "loss": 0.6962, + "step": 46434 + }, + { + "epoch": 1.7, + "learning_rate": 3.013513018639909e-08, + "loss": 0.6877, + "step": 46435 + }, + { + "epoch": 1.7, + "learning_rate": 3.013146812172703e-08, + "loss": 0.7012, + "step": 46436 + }, + { + "epoch": 1.7, + "learning_rate": 3.012780605705497e-08, + "loss": 0.6827, + "step": 46437 + }, + { + "epoch": 1.7, + "learning_rate": 3.01241439923829e-08, + "loss": 0.6981, + "step": 46438 + }, + { + "epoch": 1.7, + "learning_rate": 3.012048192771084e-08, + "loss": 0.6758, + "step": 46439 + }, + { + "epoch": 1.7, + "learning_rate": 3.0116819863038776e-08, + "loss": 0.6908, + "step": 46440 + }, + { + "epoch": 1.7, + "learning_rate": 3.0113157798366716e-08, + "loss": 0.697, + "step": 46441 + }, + { + "epoch": 1.7, + "learning_rate": 3.0109495733694656e-08, + "loss": 0.695, + "step": 46442 + }, + { + "epoch": 1.7, + "learning_rate": 3.0105833669022596e-08, + "loss": 0.6981, + "step": 46443 + }, + { + "epoch": 1.7, + "learning_rate": 3.010217160435053e-08, + "loss": 0.6949, + "step": 46444 + }, + { + "epoch": 1.7, + "learning_rate": 3.009850953967847e-08, + "loss": 0.7098, + "step": 46445 + }, + { + "epoch": 1.7, + "learning_rate": 3.009484747500641e-08, + "loss": 0.6954, + "step": 46446 + }, + { + "epoch": 1.7, + "learning_rate": 3.009118541033435e-08, + "loss": 0.6917, + "step": 46447 + }, + { + "epoch": 1.7, + "learning_rate": 3.008752334566229e-08, + "loss": 0.6906, + "step": 46448 + }, + { + "epoch": 1.7, + "learning_rate": 3.008386128099022e-08, + "loss": 0.696, + "step": 46449 + }, + { + "epoch": 1.7, + "learning_rate": 3.0080199216318155e-08, + "loss": 0.6808, + "step": 46450 + }, + { + "epoch": 1.7, + "learning_rate": 3.0076537151646095e-08, + "loss": 0.6902, + "step": 46451 + }, + { + "epoch": 1.7, + "learning_rate": 3.0072875086974035e-08, + "loss": 0.6956, + "step": 46452 + }, + { + "epoch": 1.7, + "learning_rate": 3.006921302230197e-08, + "loss": 0.6892, + "step": 46453 + }, + { + "epoch": 1.7, + "learning_rate": 3.006555095762991e-08, + "loss": 0.6997, + "step": 46454 + }, + { + "epoch": 1.7, + "learning_rate": 3.006188889295785e-08, + "loss": 0.6871, + "step": 46455 + }, + { + "epoch": 1.7, + "learning_rate": 3.005822682828579e-08, + "loss": 0.6873, + "step": 46456 + }, + { + "epoch": 1.7, + "learning_rate": 3.005456476361373e-08, + "loss": 0.688, + "step": 46457 + }, + { + "epoch": 1.7, + "learning_rate": 3.005090269894166e-08, + "loss": 0.6925, + "step": 46458 + }, + { + "epoch": 1.7, + "learning_rate": 3.00472406342696e-08, + "loss": 0.6912, + "step": 46459 + }, + { + "epoch": 1.7, + "learning_rate": 3.004357856959754e-08, + "loss": 0.6902, + "step": 46460 + }, + { + "epoch": 1.7, + "learning_rate": 3.0039916504925475e-08, + "loss": 0.6993, + "step": 46461 + }, + { + "epoch": 1.7, + "learning_rate": 3.0036254440253415e-08, + "loss": 0.6927, + "step": 46462 + }, + { + "epoch": 1.7, + "learning_rate": 3.003259237558135e-08, + "loss": 0.6915, + "step": 46463 + }, + { + "epoch": 1.7, + "learning_rate": 3.002893031090929e-08, + "loss": 0.6843, + "step": 46464 + }, + { + "epoch": 1.7, + "learning_rate": 3.002526824623723e-08, + "loss": 0.7058, + "step": 46465 + }, + { + "epoch": 1.7, + "learning_rate": 3.002160618156517e-08, + "loss": 0.6954, + "step": 46466 + }, + { + "epoch": 1.7, + "learning_rate": 3.00179441168931e-08, + "loss": 0.6981, + "step": 46467 + }, + { + "epoch": 1.7, + "learning_rate": 3.001428205222104e-08, + "loss": 0.7072, + "step": 46468 + }, + { + "epoch": 1.7, + "learning_rate": 3.001061998754898e-08, + "loss": 0.6947, + "step": 46469 + }, + { + "epoch": 1.7, + "learning_rate": 3.000695792287692e-08, + "loss": 0.7002, + "step": 46470 + }, + { + "epoch": 1.7, + "learning_rate": 3.0003295858204854e-08, + "loss": 0.6893, + "step": 46471 + }, + { + "epoch": 1.7, + "learning_rate": 2.9999633793532794e-08, + "loss": 0.6995, + "step": 46472 + }, + { + "epoch": 1.7, + "learning_rate": 2.999597172886073e-08, + "loss": 0.6835, + "step": 46473 + }, + { + "epoch": 1.7, + "learning_rate": 2.999230966418867e-08, + "loss": 0.6982, + "step": 46474 + }, + { + "epoch": 1.7, + "learning_rate": 2.998864759951661e-08, + "loss": 0.7056, + "step": 46475 + }, + { + "epoch": 1.7, + "learning_rate": 2.998498553484454e-08, + "loss": 0.6897, + "step": 46476 + }, + { + "epoch": 1.7, + "learning_rate": 2.998132347017248e-08, + "loss": 0.6942, + "step": 46477 + }, + { + "epoch": 1.7, + "learning_rate": 2.997766140550042e-08, + "loss": 0.692, + "step": 46478 + }, + { + "epoch": 1.7, + "learning_rate": 2.997399934082836e-08, + "loss": 0.6818, + "step": 46479 + }, + { + "epoch": 1.7, + "learning_rate": 2.9970337276156294e-08, + "loss": 0.7002, + "step": 46480 + }, + { + "epoch": 1.7, + "learning_rate": 2.9966675211484233e-08, + "loss": 0.692, + "step": 46481 + }, + { + "epoch": 1.7, + "learning_rate": 2.9963013146812173e-08, + "loss": 0.6801, + "step": 46482 + }, + { + "epoch": 1.7, + "learning_rate": 2.9959351082140113e-08, + "loss": 0.6916, + "step": 46483 + }, + { + "epoch": 1.7, + "learning_rate": 2.9955689017468047e-08, + "loss": 0.6914, + "step": 46484 + }, + { + "epoch": 1.7, + "learning_rate": 2.9952026952795987e-08, + "loss": 0.6854, + "step": 46485 + }, + { + "epoch": 1.7, + "learning_rate": 2.994836488812392e-08, + "loss": 0.6967, + "step": 46486 + }, + { + "epoch": 1.7, + "learning_rate": 2.994470282345186e-08, + "loss": 0.6875, + "step": 46487 + }, + { + "epoch": 1.7, + "learning_rate": 2.99410407587798e-08, + "loss": 0.6971, + "step": 46488 + }, + { + "epoch": 1.7, + "learning_rate": 2.993737869410773e-08, + "loss": 0.6841, + "step": 46489 + }, + { + "epoch": 1.7, + "learning_rate": 2.993371662943567e-08, + "loss": 0.6868, + "step": 46490 + }, + { + "epoch": 1.7, + "learning_rate": 2.993005456476361e-08, + "loss": 0.689, + "step": 46491 + }, + { + "epoch": 1.7, + "learning_rate": 2.992639250009155e-08, + "loss": 0.6916, + "step": 46492 + }, + { + "epoch": 1.7, + "learning_rate": 2.9922730435419486e-08, + "loss": 0.6804, + "step": 46493 + }, + { + "epoch": 1.7, + "learning_rate": 2.9919068370747426e-08, + "loss": 0.6903, + "step": 46494 + }, + { + "epoch": 1.7, + "learning_rate": 2.9915406306075366e-08, + "loss": 0.6842, + "step": 46495 + }, + { + "epoch": 1.7, + "learning_rate": 2.9911744241403306e-08, + "loss": 0.6883, + "step": 46496 + }, + { + "epoch": 1.7, + "learning_rate": 2.990808217673124e-08, + "loss": 0.69, + "step": 46497 + }, + { + "epoch": 1.7, + "learning_rate": 2.990442011205917e-08, + "loss": 0.6889, + "step": 46498 + }, + { + "epoch": 1.7, + "learning_rate": 2.990075804738711e-08, + "loss": 0.6926, + "step": 46499 + }, + { + "epoch": 1.7, + "learning_rate": 2.989709598271505e-08, + "loss": 0.6926, + "step": 46500 + }, + { + "epoch": 1.7, + "learning_rate": 2.989343391804299e-08, + "loss": 0.6869, + "step": 46501 + }, + { + "epoch": 1.7, + "learning_rate": 2.9889771853370926e-08, + "loss": 0.6922, + "step": 46502 + }, + { + "epoch": 1.7, + "learning_rate": 2.9886109788698866e-08, + "loss": 0.6921, + "step": 46503 + }, + { + "epoch": 1.7, + "learning_rate": 2.9882447724026806e-08, + "loss": 0.6934, + "step": 46504 + }, + { + "epoch": 1.7, + "learning_rate": 2.9878785659354745e-08, + "loss": 0.6959, + "step": 46505 + }, + { + "epoch": 1.7, + "learning_rate": 2.9875123594682685e-08, + "loss": 0.698, + "step": 46506 + }, + { + "epoch": 1.7, + "learning_rate": 2.987146153001062e-08, + "loss": 0.7028, + "step": 46507 + }, + { + "epoch": 1.7, + "learning_rate": 2.986779946533856e-08, + "loss": 0.6894, + "step": 46508 + }, + { + "epoch": 1.7, + "learning_rate": 2.986413740066649e-08, + "loss": 0.6896, + "step": 46509 + }, + { + "epoch": 1.7, + "learning_rate": 2.986047533599443e-08, + "loss": 0.6897, + "step": 46510 + }, + { + "epoch": 1.7, + "learning_rate": 2.9856813271322365e-08, + "loss": 0.6966, + "step": 46511 + }, + { + "epoch": 1.7, + "learning_rate": 2.9853151206650305e-08, + "loss": 0.6933, + "step": 46512 + }, + { + "epoch": 1.7, + "learning_rate": 2.9849489141978245e-08, + "loss": 0.6973, + "step": 46513 + }, + { + "epoch": 1.7, + "learning_rate": 2.9845827077306185e-08, + "loss": 0.699, + "step": 46514 + }, + { + "epoch": 1.7, + "learning_rate": 2.9842165012634125e-08, + "loss": 0.7017, + "step": 46515 + }, + { + "epoch": 1.7, + "learning_rate": 2.983850294796206e-08, + "loss": 0.6847, + "step": 46516 + }, + { + "epoch": 1.7, + "learning_rate": 2.983484088329e-08, + "loss": 0.6915, + "step": 46517 + }, + { + "epoch": 1.7, + "learning_rate": 2.983117881861794e-08, + "loss": 0.6947, + "step": 46518 + }, + { + "epoch": 1.7, + "learning_rate": 2.982751675394588e-08, + "loss": 0.6987, + "step": 46519 + }, + { + "epoch": 1.7, + "learning_rate": 2.982385468927381e-08, + "loss": 0.6987, + "step": 46520 + }, + { + "epoch": 1.7, + "learning_rate": 2.9820192624601745e-08, + "loss": 0.6994, + "step": 46521 + }, + { + "epoch": 1.7, + "learning_rate": 2.9816530559929685e-08, + "loss": 0.6982, + "step": 46522 + }, + { + "epoch": 1.7, + "learning_rate": 2.9812868495257624e-08, + "loss": 0.7062, + "step": 46523 + }, + { + "epoch": 1.7, + "learning_rate": 2.9809206430585564e-08, + "loss": 0.6939, + "step": 46524 + }, + { + "epoch": 1.7, + "learning_rate": 2.98055443659135e-08, + "loss": 0.6849, + "step": 46525 + }, + { + "epoch": 1.7, + "learning_rate": 2.9801882301241438e-08, + "loss": 0.6948, + "step": 46526 + }, + { + "epoch": 1.7, + "learning_rate": 2.9798220236569378e-08, + "loss": 0.7101, + "step": 46527 + }, + { + "epoch": 1.7, + "learning_rate": 2.9794558171897317e-08, + "loss": 0.6859, + "step": 46528 + }, + { + "epoch": 1.7, + "learning_rate": 2.979089610722525e-08, + "loss": 0.6812, + "step": 46529 + }, + { + "epoch": 1.7, + "learning_rate": 2.9787234042553187e-08, + "loss": 0.6807, + "step": 46530 + }, + { + "epoch": 1.7, + "learning_rate": 2.9783571977881127e-08, + "loss": 0.6897, + "step": 46531 + }, + { + "epoch": 1.7, + "learning_rate": 2.9779909913209067e-08, + "loss": 0.6902, + "step": 46532 + }, + { + "epoch": 1.7, + "learning_rate": 2.9776247848537007e-08, + "loss": 0.7084, + "step": 46533 + }, + { + "epoch": 1.7, + "learning_rate": 2.977258578386494e-08, + "loss": 0.7014, + "step": 46534 + }, + { + "epoch": 1.7, + "learning_rate": 2.976892371919288e-08, + "loss": 0.6921, + "step": 46535 + }, + { + "epoch": 1.7, + "learning_rate": 2.9765261654520817e-08, + "loss": 0.682, + "step": 46536 + }, + { + "epoch": 1.7, + "learning_rate": 2.9761599589848757e-08, + "loss": 0.6825, + "step": 46537 + }, + { + "epoch": 1.7, + "learning_rate": 2.975793752517669e-08, + "loss": 0.6925, + "step": 46538 + }, + { + "epoch": 1.7, + "learning_rate": 2.975427546050463e-08, + "loss": 0.7014, + "step": 46539 + }, + { + "epoch": 1.7, + "learning_rate": 2.975061339583257e-08, + "loss": 0.6976, + "step": 46540 + }, + { + "epoch": 1.7, + "learning_rate": 2.9746951331160507e-08, + "loss": 0.6964, + "step": 46541 + }, + { + "epoch": 1.7, + "learning_rate": 2.9743289266488447e-08, + "loss": 0.6958, + "step": 46542 + }, + { + "epoch": 1.7, + "learning_rate": 2.973962720181638e-08, + "loss": 0.6897, + "step": 46543 + }, + { + "epoch": 1.7, + "learning_rate": 2.973596513714432e-08, + "loss": 0.6766, + "step": 46544 + }, + { + "epoch": 1.7, + "learning_rate": 2.973230307247226e-08, + "loss": 0.7, + "step": 46545 + }, + { + "epoch": 1.7, + "learning_rate": 2.9728641007800196e-08, + "loss": 0.6924, + "step": 46546 + }, + { + "epoch": 1.7, + "learning_rate": 2.9724978943128133e-08, + "loss": 0.6921, + "step": 46547 + }, + { + "epoch": 1.7, + "learning_rate": 2.972131687845607e-08, + "loss": 0.6854, + "step": 46548 + }, + { + "epoch": 1.7, + "learning_rate": 2.971765481378401e-08, + "loss": 0.6889, + "step": 46549 + }, + { + "epoch": 1.7, + "learning_rate": 2.971399274911195e-08, + "loss": 0.6981, + "step": 46550 + }, + { + "epoch": 1.7, + "learning_rate": 2.971033068443989e-08, + "loss": 0.6822, + "step": 46551 + }, + { + "epoch": 1.7, + "learning_rate": 2.9706668619767823e-08, + "loss": 0.6951, + "step": 46552 + }, + { + "epoch": 1.7, + "learning_rate": 2.970300655509576e-08, + "loss": 0.6872, + "step": 46553 + }, + { + "epoch": 1.7, + "learning_rate": 2.96993444904237e-08, + "loss": 0.6851, + "step": 46554 + }, + { + "epoch": 1.7, + "learning_rate": 2.969568242575164e-08, + "loss": 0.6831, + "step": 46555 + }, + { + "epoch": 1.7, + "learning_rate": 2.9692020361079573e-08, + "loss": 0.6936, + "step": 46556 + }, + { + "epoch": 1.7, + "learning_rate": 2.9688358296407513e-08, + "loss": 0.6946, + "step": 46557 + }, + { + "epoch": 1.7, + "learning_rate": 2.9684696231735452e-08, + "loss": 0.6939, + "step": 46558 + }, + { + "epoch": 1.7, + "learning_rate": 2.968103416706339e-08, + "loss": 0.6865, + "step": 46559 + }, + { + "epoch": 1.7, + "learning_rate": 2.967737210239133e-08, + "loss": 0.7069, + "step": 46560 + }, + { + "epoch": 1.7, + "learning_rate": 2.9673710037719262e-08, + "loss": 0.6852, + "step": 46561 + }, + { + "epoch": 1.7, + "learning_rate": 2.9670047973047202e-08, + "loss": 0.6964, + "step": 46562 + }, + { + "epoch": 1.7, + "learning_rate": 2.9666385908375142e-08, + "loss": 0.6972, + "step": 46563 + }, + { + "epoch": 1.7, + "learning_rate": 2.966272384370308e-08, + "loss": 0.7007, + "step": 46564 + }, + { + "epoch": 1.7, + "learning_rate": 2.9659061779031015e-08, + "loss": 0.685, + "step": 46565 + }, + { + "epoch": 1.7, + "learning_rate": 2.9655399714358952e-08, + "loss": 0.6925, + "step": 46566 + }, + { + "epoch": 1.7, + "learning_rate": 2.9651737649686892e-08, + "loss": 0.7055, + "step": 46567 + }, + { + "epoch": 1.7, + "learning_rate": 2.9648075585014832e-08, + "loss": 0.6937, + "step": 46568 + }, + { + "epoch": 1.7, + "learning_rate": 2.9644413520342772e-08, + "loss": 0.7036, + "step": 46569 + }, + { + "epoch": 1.7, + "learning_rate": 2.9640751455670705e-08, + "loss": 0.6984, + "step": 46570 + }, + { + "epoch": 1.7, + "learning_rate": 2.9637089390998642e-08, + "loss": 0.7055, + "step": 46571 + }, + { + "epoch": 1.7, + "learning_rate": 2.9633427326326582e-08, + "loss": 0.6955, + "step": 46572 + }, + { + "epoch": 1.7, + "learning_rate": 2.962976526165452e-08, + "loss": 0.6949, + "step": 46573 + }, + { + "epoch": 1.7, + "learning_rate": 2.9626103196982455e-08, + "loss": 0.6948, + "step": 46574 + }, + { + "epoch": 1.7, + "learning_rate": 2.9622441132310395e-08, + "loss": 0.6913, + "step": 46575 + }, + { + "epoch": 1.7, + "learning_rate": 2.961877906763833e-08, + "loss": 0.6937, + "step": 46576 + }, + { + "epoch": 1.7, + "learning_rate": 2.961511700296627e-08, + "loss": 0.6869, + "step": 46577 + }, + { + "epoch": 1.7, + "learning_rate": 2.961145493829421e-08, + "loss": 0.6853, + "step": 46578 + }, + { + "epoch": 1.7, + "learning_rate": 2.9607792873622145e-08, + "loss": 0.7041, + "step": 46579 + }, + { + "epoch": 1.7, + "learning_rate": 2.9604130808950085e-08, + "loss": 0.7086, + "step": 46580 + }, + { + "epoch": 1.7, + "learning_rate": 2.9600468744278024e-08, + "loss": 0.68, + "step": 46581 + }, + { + "epoch": 1.7, + "learning_rate": 2.959680667960596e-08, + "loss": 0.6823, + "step": 46582 + }, + { + "epoch": 1.7, + "learning_rate": 2.9593144614933894e-08, + "loss": 0.6897, + "step": 46583 + }, + { + "epoch": 1.7, + "learning_rate": 2.9589482550261834e-08, + "loss": 0.6909, + "step": 46584 + }, + { + "epoch": 1.7, + "learning_rate": 2.9585820485589774e-08, + "loss": 0.7032, + "step": 46585 + }, + { + "epoch": 1.7, + "learning_rate": 2.9582158420917714e-08, + "loss": 0.6928, + "step": 46586 + }, + { + "epoch": 1.7, + "learning_rate": 2.957849635624565e-08, + "loss": 0.6865, + "step": 46587 + }, + { + "epoch": 1.7, + "learning_rate": 2.9574834291573587e-08, + "loss": 0.6937, + "step": 46588 + }, + { + "epoch": 1.7, + "learning_rate": 2.9571172226901524e-08, + "loss": 0.6941, + "step": 46589 + }, + { + "epoch": 1.7, + "learning_rate": 2.9567510162229464e-08, + "loss": 0.6936, + "step": 46590 + }, + { + "epoch": 1.7, + "learning_rate": 2.9563848097557404e-08, + "loss": 0.6921, + "step": 46591 + }, + { + "epoch": 1.7, + "learning_rate": 2.9560186032885337e-08, + "loss": 0.6998, + "step": 46592 + }, + { + "epoch": 1.7, + "learning_rate": 2.9556523968213277e-08, + "loss": 0.6879, + "step": 46593 + }, + { + "epoch": 1.7, + "learning_rate": 2.9552861903541214e-08, + "loss": 0.6834, + "step": 46594 + }, + { + "epoch": 1.7, + "learning_rate": 2.9549199838869154e-08, + "loss": 0.7051, + "step": 46595 + }, + { + "epoch": 1.7, + "learning_rate": 2.9545537774197094e-08, + "loss": 0.7037, + "step": 46596 + }, + { + "epoch": 1.7, + "learning_rate": 2.9541875709525027e-08, + "loss": 0.6857, + "step": 46597 + }, + { + "epoch": 1.7, + "learning_rate": 2.9538213644852967e-08, + "loss": 0.6975, + "step": 46598 + }, + { + "epoch": 1.7, + "learning_rate": 2.9534551580180907e-08, + "loss": 0.7035, + "step": 46599 + }, + { + "epoch": 1.7, + "learning_rate": 2.9530889515508843e-08, + "loss": 0.6804, + "step": 46600 + }, + { + "epoch": 1.7, + "learning_rate": 2.9527227450836777e-08, + "loss": 0.6947, + "step": 46601 + }, + { + "epoch": 1.71, + "learning_rate": 2.9523565386164717e-08, + "loss": 0.6939, + "step": 46602 + }, + { + "epoch": 1.71, + "learning_rate": 2.9519903321492657e-08, + "loss": 0.6778, + "step": 46603 + }, + { + "epoch": 1.71, + "learning_rate": 2.9516241256820597e-08, + "loss": 0.6963, + "step": 46604 + }, + { + "epoch": 1.71, + "learning_rate": 2.9512579192148533e-08, + "loss": 0.6961, + "step": 46605 + }, + { + "epoch": 1.71, + "learning_rate": 2.9508917127476466e-08, + "loss": 0.6934, + "step": 46606 + }, + { + "epoch": 1.71, + "learning_rate": 2.9505255062804406e-08, + "loss": 0.6915, + "step": 46607 + }, + { + "epoch": 1.71, + "learning_rate": 2.9501592998132346e-08, + "loss": 0.7025, + "step": 46608 + }, + { + "epoch": 1.71, + "learning_rate": 2.9497930933460286e-08, + "loss": 0.6987, + "step": 46609 + }, + { + "epoch": 1.71, + "learning_rate": 2.949426886878822e-08, + "loss": 0.6929, + "step": 46610 + }, + { + "epoch": 1.71, + "learning_rate": 2.949060680411616e-08, + "loss": 0.7037, + "step": 46611 + }, + { + "epoch": 1.71, + "learning_rate": 2.9486944739444096e-08, + "loss": 0.7, + "step": 46612 + }, + { + "epoch": 1.71, + "learning_rate": 2.9483282674772036e-08, + "loss": 0.69, + "step": 46613 + }, + { + "epoch": 1.71, + "learning_rate": 2.9479620610099976e-08, + "loss": 0.6899, + "step": 46614 + }, + { + "epoch": 1.71, + "learning_rate": 2.947595854542791e-08, + "loss": 0.6939, + "step": 46615 + }, + { + "epoch": 1.71, + "learning_rate": 2.947229648075585e-08, + "loss": 0.6891, + "step": 46616 + }, + { + "epoch": 1.71, + "learning_rate": 2.9468634416083786e-08, + "loss": 0.6927, + "step": 46617 + }, + { + "epoch": 1.71, + "learning_rate": 2.9464972351411726e-08, + "loss": 0.6918, + "step": 46618 + }, + { + "epoch": 1.71, + "learning_rate": 2.946131028673966e-08, + "loss": 0.6958, + "step": 46619 + }, + { + "epoch": 1.71, + "learning_rate": 2.94576482220676e-08, + "loss": 0.6808, + "step": 46620 + }, + { + "epoch": 1.71, + "learning_rate": 2.945398615739554e-08, + "loss": 0.6854, + "step": 46621 + }, + { + "epoch": 1.71, + "learning_rate": 2.945032409272348e-08, + "loss": 0.703, + "step": 46622 + }, + { + "epoch": 1.71, + "learning_rate": 2.9446662028051415e-08, + "loss": 0.6874, + "step": 46623 + }, + { + "epoch": 1.71, + "learning_rate": 2.944299996337935e-08, + "loss": 0.7042, + "step": 46624 + }, + { + "epoch": 1.71, + "learning_rate": 2.943933789870729e-08, + "loss": 0.6789, + "step": 46625 + }, + { + "epoch": 1.71, + "learning_rate": 2.943567583403523e-08, + "loss": 0.6911, + "step": 46626 + }, + { + "epoch": 1.71, + "learning_rate": 2.943201376936317e-08, + "loss": 0.7079, + "step": 46627 + }, + { + "epoch": 1.71, + "learning_rate": 2.9428351704691102e-08, + "loss": 0.6901, + "step": 46628 + }, + { + "epoch": 1.71, + "learning_rate": 2.9424689640019042e-08, + "loss": 0.7011, + "step": 46629 + }, + { + "epoch": 1.71, + "learning_rate": 2.942102757534698e-08, + "loss": 0.6952, + "step": 46630 + }, + { + "epoch": 1.71, + "learning_rate": 2.941736551067492e-08, + "loss": 0.6928, + "step": 46631 + }, + { + "epoch": 1.71, + "learning_rate": 2.9413703446002858e-08, + "loss": 0.6854, + "step": 46632 + }, + { + "epoch": 1.71, + "learning_rate": 2.941004138133079e-08, + "loss": 0.6976, + "step": 46633 + }, + { + "epoch": 1.71, + "learning_rate": 2.940637931665873e-08, + "loss": 0.7082, + "step": 46634 + }, + { + "epoch": 1.71, + "learning_rate": 2.9402717251986668e-08, + "loss": 0.6969, + "step": 46635 + }, + { + "epoch": 1.71, + "learning_rate": 2.9399055187314608e-08, + "loss": 0.6988, + "step": 46636 + }, + { + "epoch": 1.71, + "learning_rate": 2.939539312264254e-08, + "loss": 0.6929, + "step": 46637 + }, + { + "epoch": 1.71, + "learning_rate": 2.939173105797048e-08, + "loss": 0.6967, + "step": 46638 + }, + { + "epoch": 1.71, + "learning_rate": 2.938806899329842e-08, + "loss": 0.6876, + "step": 46639 + }, + { + "epoch": 1.71, + "learning_rate": 2.938440692862636e-08, + "loss": 0.6979, + "step": 46640 + }, + { + "epoch": 1.71, + "learning_rate": 2.9380744863954298e-08, + "loss": 0.7009, + "step": 46641 + }, + { + "epoch": 1.71, + "learning_rate": 2.937708279928223e-08, + "loss": 0.6911, + "step": 46642 + }, + { + "epoch": 1.71, + "learning_rate": 2.937342073461017e-08, + "loss": 0.6953, + "step": 46643 + }, + { + "epoch": 1.71, + "learning_rate": 2.936975866993811e-08, + "loss": 0.6813, + "step": 46644 + }, + { + "epoch": 1.71, + "learning_rate": 2.936609660526605e-08, + "loss": 0.6925, + "step": 46645 + }, + { + "epoch": 1.71, + "learning_rate": 2.9362434540593984e-08, + "loss": 0.6995, + "step": 46646 + }, + { + "epoch": 1.71, + "learning_rate": 2.935877247592192e-08, + "loss": 0.6898, + "step": 46647 + }, + { + "epoch": 1.71, + "learning_rate": 2.935511041124986e-08, + "loss": 0.6976, + "step": 46648 + }, + { + "epoch": 1.71, + "learning_rate": 2.93514483465778e-08, + "loss": 0.6859, + "step": 46649 + }, + { + "epoch": 1.71, + "learning_rate": 2.934778628190574e-08, + "loss": 0.6849, + "step": 46650 + }, + { + "epoch": 1.71, + "learning_rate": 2.9344124217233674e-08, + "loss": 0.6927, + "step": 46651 + }, + { + "epoch": 1.71, + "learning_rate": 2.9340462152561614e-08, + "loss": 0.6964, + "step": 46652 + }, + { + "epoch": 1.71, + "learning_rate": 2.933680008788955e-08, + "loss": 0.6906, + "step": 46653 + }, + { + "epoch": 1.71, + "learning_rate": 2.933313802321749e-08, + "loss": 0.6855, + "step": 46654 + }, + { + "epoch": 1.71, + "learning_rate": 2.9329475958545424e-08, + "loss": 0.698, + "step": 46655 + }, + { + "epoch": 1.71, + "learning_rate": 2.9325813893873364e-08, + "loss": 0.7045, + "step": 46656 + }, + { + "epoch": 1.71, + "learning_rate": 2.9322151829201304e-08, + "loss": 0.6954, + "step": 46657 + }, + { + "epoch": 1.71, + "learning_rate": 2.931848976452924e-08, + "loss": 0.6932, + "step": 46658 + }, + { + "epoch": 1.71, + "learning_rate": 2.931482769985718e-08, + "loss": 0.693, + "step": 46659 + }, + { + "epoch": 1.71, + "learning_rate": 2.9311165635185113e-08, + "loss": 0.6909, + "step": 46660 + }, + { + "epoch": 1.71, + "learning_rate": 2.9307503570513053e-08, + "loss": 0.7016, + "step": 46661 + }, + { + "epoch": 1.71, + "learning_rate": 2.9303841505840993e-08, + "loss": 0.6881, + "step": 46662 + }, + { + "epoch": 1.71, + "learning_rate": 2.9300179441168933e-08, + "loss": 0.6908, + "step": 46663 + }, + { + "epoch": 1.71, + "learning_rate": 2.9296517376496867e-08, + "loss": 0.6872, + "step": 46664 + }, + { + "epoch": 1.71, + "learning_rate": 2.9292855311824803e-08, + "loss": 0.6887, + "step": 46665 + }, + { + "epoch": 1.71, + "learning_rate": 2.9289193247152743e-08, + "loss": 0.7053, + "step": 46666 + }, + { + "epoch": 1.71, + "learning_rate": 2.9285531182480683e-08, + "loss": 0.6987, + "step": 46667 + }, + { + "epoch": 1.71, + "learning_rate": 2.9281869117808623e-08, + "loss": 0.6886, + "step": 46668 + }, + { + "epoch": 1.71, + "learning_rate": 2.9278207053136556e-08, + "loss": 0.6847, + "step": 46669 + }, + { + "epoch": 1.71, + "learning_rate": 2.9274544988464496e-08, + "loss": 0.7062, + "step": 46670 + }, + { + "epoch": 1.71, + "learning_rate": 2.9270882923792433e-08, + "loss": 0.7018, + "step": 46671 + }, + { + "epoch": 1.71, + "learning_rate": 2.9267220859120373e-08, + "loss": 0.7038, + "step": 46672 + }, + { + "epoch": 1.71, + "learning_rate": 2.9263558794448306e-08, + "loss": 0.6947, + "step": 46673 + }, + { + "epoch": 1.71, + "learning_rate": 2.9259896729776246e-08, + "loss": 0.6834, + "step": 46674 + }, + { + "epoch": 1.71, + "learning_rate": 2.9256234665104186e-08, + "loss": 0.6872, + "step": 46675 + }, + { + "epoch": 1.71, + "learning_rate": 2.9252572600432122e-08, + "loss": 0.6939, + "step": 46676 + }, + { + "epoch": 1.71, + "learning_rate": 2.9248910535760062e-08, + "loss": 0.6944, + "step": 46677 + }, + { + "epoch": 1.71, + "learning_rate": 2.9245248471087996e-08, + "loss": 0.7006, + "step": 46678 + }, + { + "epoch": 1.71, + "learning_rate": 2.9241586406415936e-08, + "loss": 0.7002, + "step": 46679 + }, + { + "epoch": 1.71, + "learning_rate": 2.9237924341743876e-08, + "loss": 0.6912, + "step": 46680 + }, + { + "epoch": 1.71, + "learning_rate": 2.9234262277071816e-08, + "loss": 0.681, + "step": 46681 + }, + { + "epoch": 1.71, + "learning_rate": 2.923060021239975e-08, + "loss": 0.7029, + "step": 46682 + }, + { + "epoch": 1.71, + "learning_rate": 2.9226938147727685e-08, + "loss": 0.6945, + "step": 46683 + }, + { + "epoch": 1.71, + "learning_rate": 2.9223276083055625e-08, + "loss": 0.6953, + "step": 46684 + }, + { + "epoch": 1.71, + "learning_rate": 2.9219614018383565e-08, + "loss": 0.6829, + "step": 46685 + }, + { + "epoch": 1.71, + "learning_rate": 2.92159519537115e-08, + "loss": 0.7034, + "step": 46686 + }, + { + "epoch": 1.71, + "learning_rate": 2.921228988903944e-08, + "loss": 0.6931, + "step": 46687 + }, + { + "epoch": 1.71, + "learning_rate": 2.9208627824367375e-08, + "loss": 0.6953, + "step": 46688 + }, + { + "epoch": 1.71, + "learning_rate": 2.9204965759695315e-08, + "loss": 0.698, + "step": 46689 + }, + { + "epoch": 1.71, + "learning_rate": 2.9201303695023255e-08, + "loss": 0.6876, + "step": 46690 + }, + { + "epoch": 1.71, + "learning_rate": 2.9197641630351188e-08, + "loss": 0.6917, + "step": 46691 + }, + { + "epoch": 1.71, + "learning_rate": 2.9193979565679128e-08, + "loss": 0.6971, + "step": 46692 + }, + { + "epoch": 1.71, + "learning_rate": 2.9190317501007068e-08, + "loss": 0.7047, + "step": 46693 + }, + { + "epoch": 1.71, + "learning_rate": 2.9186655436335005e-08, + "loss": 0.6939, + "step": 46694 + }, + { + "epoch": 1.71, + "learning_rate": 2.9182993371662938e-08, + "loss": 0.7004, + "step": 46695 + }, + { + "epoch": 1.71, + "learning_rate": 2.9179331306990878e-08, + "loss": 0.6996, + "step": 46696 + }, + { + "epoch": 1.71, + "learning_rate": 2.9175669242318818e-08, + "loss": 0.7009, + "step": 46697 + }, + { + "epoch": 1.71, + "learning_rate": 2.9172007177646758e-08, + "loss": 0.6928, + "step": 46698 + }, + { + "epoch": 1.71, + "learning_rate": 2.9168345112974695e-08, + "loss": 0.6898, + "step": 46699 + }, + { + "epoch": 1.71, + "learning_rate": 2.916468304830263e-08, + "loss": 0.6955, + "step": 46700 + }, + { + "epoch": 1.71, + "learning_rate": 2.9161020983630568e-08, + "loss": 0.6921, + "step": 46701 + }, + { + "epoch": 1.71, + "learning_rate": 2.9157358918958508e-08, + "loss": 0.6921, + "step": 46702 + }, + { + "epoch": 1.71, + "learning_rate": 2.9153696854286448e-08, + "loss": 0.6731, + "step": 46703 + }, + { + "epoch": 1.71, + "learning_rate": 2.915003478961438e-08, + "loss": 0.6894, + "step": 46704 + }, + { + "epoch": 1.71, + "learning_rate": 2.914637272494232e-08, + "loss": 0.6995, + "step": 46705 + }, + { + "epoch": 1.71, + "learning_rate": 2.9142710660270257e-08, + "loss": 0.6917, + "step": 46706 + }, + { + "epoch": 1.71, + "learning_rate": 2.9139048595598197e-08, + "loss": 0.7003, + "step": 46707 + }, + { + "epoch": 1.71, + "learning_rate": 2.9135386530926137e-08, + "loss": 0.6927, + "step": 46708 + }, + { + "epoch": 1.71, + "learning_rate": 2.913172446625407e-08, + "loss": 0.6969, + "step": 46709 + }, + { + "epoch": 1.71, + "learning_rate": 2.912806240158201e-08, + "loss": 0.6818, + "step": 46710 + }, + { + "epoch": 1.71, + "learning_rate": 2.912440033690995e-08, + "loss": 0.7022, + "step": 46711 + }, + { + "epoch": 1.71, + "learning_rate": 2.9120738272237887e-08, + "loss": 0.6928, + "step": 46712 + }, + { + "epoch": 1.71, + "learning_rate": 2.911707620756582e-08, + "loss": 0.6915, + "step": 46713 + }, + { + "epoch": 1.71, + "learning_rate": 2.911341414289376e-08, + "loss": 0.6871, + "step": 46714 + }, + { + "epoch": 1.71, + "learning_rate": 2.91097520782217e-08, + "loss": 0.6982, + "step": 46715 + }, + { + "epoch": 1.71, + "learning_rate": 2.910609001354964e-08, + "loss": 0.6945, + "step": 46716 + }, + { + "epoch": 1.71, + "learning_rate": 2.9102427948877577e-08, + "loss": 0.7089, + "step": 46717 + }, + { + "epoch": 1.71, + "learning_rate": 2.909876588420551e-08, + "loss": 0.7033, + "step": 46718 + }, + { + "epoch": 1.71, + "learning_rate": 2.909510381953345e-08, + "loss": 0.6951, + "step": 46719 + }, + { + "epoch": 1.71, + "learning_rate": 2.909144175486139e-08, + "loss": 0.6951, + "step": 46720 + }, + { + "epoch": 1.71, + "learning_rate": 2.908777969018933e-08, + "loss": 0.6945, + "step": 46721 + }, + { + "epoch": 1.71, + "learning_rate": 2.9084117625517263e-08, + "loss": 0.6839, + "step": 46722 + }, + { + "epoch": 1.71, + "learning_rate": 2.9080455560845203e-08, + "loss": 0.6967, + "step": 46723 + }, + { + "epoch": 1.71, + "learning_rate": 2.907679349617314e-08, + "loss": 0.7034, + "step": 46724 + }, + { + "epoch": 1.71, + "learning_rate": 2.907313143150108e-08, + "loss": 0.6899, + "step": 46725 + }, + { + "epoch": 1.71, + "learning_rate": 2.906946936682902e-08, + "loss": 0.6936, + "step": 46726 + }, + { + "epoch": 1.71, + "learning_rate": 2.9065807302156953e-08, + "loss": 0.6894, + "step": 46727 + }, + { + "epoch": 1.71, + "learning_rate": 2.9062145237484893e-08, + "loss": 0.6911, + "step": 46728 + }, + { + "epoch": 1.71, + "learning_rate": 2.905848317281283e-08, + "loss": 0.6879, + "step": 46729 + }, + { + "epoch": 1.71, + "learning_rate": 2.905482110814077e-08, + "loss": 0.6995, + "step": 46730 + }, + { + "epoch": 1.71, + "learning_rate": 2.9051159043468703e-08, + "loss": 0.6911, + "step": 46731 + }, + { + "epoch": 1.71, + "learning_rate": 2.9047496978796643e-08, + "loss": 0.6876, + "step": 46732 + }, + { + "epoch": 1.71, + "learning_rate": 2.9043834914124583e-08, + "loss": 0.698, + "step": 46733 + }, + { + "epoch": 1.71, + "learning_rate": 2.9040172849452523e-08, + "loss": 0.6858, + "step": 46734 + }, + { + "epoch": 1.71, + "learning_rate": 2.903651078478046e-08, + "loss": 0.6913, + "step": 46735 + }, + { + "epoch": 1.71, + "learning_rate": 2.9032848720108392e-08, + "loss": 0.6924, + "step": 46736 + }, + { + "epoch": 1.71, + "learning_rate": 2.9029186655436332e-08, + "loss": 0.7093, + "step": 46737 + }, + { + "epoch": 1.71, + "learning_rate": 2.9025524590764272e-08, + "loss": 0.7016, + "step": 46738 + }, + { + "epoch": 1.71, + "learning_rate": 2.9021862526092212e-08, + "loss": 0.7043, + "step": 46739 + }, + { + "epoch": 1.71, + "learning_rate": 2.9018200461420146e-08, + "loss": 0.6802, + "step": 46740 + }, + { + "epoch": 1.71, + "learning_rate": 2.9014538396748086e-08, + "loss": 0.681, + "step": 46741 + }, + { + "epoch": 1.71, + "learning_rate": 2.9010876332076022e-08, + "loss": 0.6837, + "step": 46742 + }, + { + "epoch": 1.71, + "learning_rate": 2.9007214267403962e-08, + "loss": 0.6968, + "step": 46743 + }, + { + "epoch": 1.71, + "learning_rate": 2.9003552202731902e-08, + "loss": 0.6929, + "step": 46744 + }, + { + "epoch": 1.71, + "learning_rate": 2.8999890138059835e-08, + "loss": 0.7007, + "step": 46745 + }, + { + "epoch": 1.71, + "learning_rate": 2.8996228073387775e-08, + "loss": 0.6886, + "step": 46746 + }, + { + "epoch": 1.71, + "learning_rate": 2.8992566008715712e-08, + "loss": 0.6893, + "step": 46747 + }, + { + "epoch": 1.71, + "learning_rate": 2.8988903944043652e-08, + "loss": 0.6976, + "step": 46748 + }, + { + "epoch": 1.71, + "learning_rate": 2.8985241879371585e-08, + "loss": 0.6992, + "step": 46749 + }, + { + "epoch": 1.71, + "learning_rate": 2.8981579814699525e-08, + "loss": 0.6884, + "step": 46750 + }, + { + "epoch": 1.71, + "learning_rate": 2.8977917750027465e-08, + "loss": 0.6794, + "step": 46751 + }, + { + "epoch": 1.71, + "learning_rate": 2.89742556853554e-08, + "loss": 0.6897, + "step": 46752 + }, + { + "epoch": 1.71, + "learning_rate": 2.897059362068334e-08, + "loss": 0.6924, + "step": 46753 + }, + { + "epoch": 1.71, + "learning_rate": 2.8966931556011275e-08, + "loss": 0.6904, + "step": 46754 + }, + { + "epoch": 1.71, + "learning_rate": 2.8963269491339215e-08, + "loss": 0.6958, + "step": 46755 + }, + { + "epoch": 1.71, + "learning_rate": 2.8959607426667155e-08, + "loss": 0.6894, + "step": 46756 + }, + { + "epoch": 1.71, + "learning_rate": 2.8955945361995095e-08, + "loss": 0.7042, + "step": 46757 + }, + { + "epoch": 1.71, + "learning_rate": 2.8952283297323028e-08, + "loss": 0.6883, + "step": 46758 + }, + { + "epoch": 1.71, + "learning_rate": 2.8948621232650965e-08, + "loss": 0.693, + "step": 46759 + }, + { + "epoch": 1.71, + "learning_rate": 2.8944959167978904e-08, + "loss": 0.6998, + "step": 46760 + }, + { + "epoch": 1.71, + "learning_rate": 2.8941297103306844e-08, + "loss": 0.6869, + "step": 46761 + }, + { + "epoch": 1.71, + "learning_rate": 2.8937635038634784e-08, + "loss": 0.6978, + "step": 46762 + }, + { + "epoch": 1.71, + "learning_rate": 2.8933972973962718e-08, + "loss": 0.6962, + "step": 46763 + }, + { + "epoch": 1.71, + "learning_rate": 2.8930310909290658e-08, + "loss": 0.6859, + "step": 46764 + }, + { + "epoch": 1.71, + "learning_rate": 2.8926648844618594e-08, + "loss": 0.7014, + "step": 46765 + }, + { + "epoch": 1.71, + "learning_rate": 2.8922986779946534e-08, + "loss": 0.6957, + "step": 46766 + }, + { + "epoch": 1.71, + "learning_rate": 2.8919324715274467e-08, + "loss": 0.6908, + "step": 46767 + }, + { + "epoch": 1.71, + "learning_rate": 2.8915662650602407e-08, + "loss": 0.6861, + "step": 46768 + }, + { + "epoch": 1.71, + "learning_rate": 2.8912000585930347e-08, + "loss": 0.6974, + "step": 46769 + }, + { + "epoch": 1.71, + "learning_rate": 2.8908338521258284e-08, + "loss": 0.6825, + "step": 46770 + }, + { + "epoch": 1.71, + "learning_rate": 2.8904676456586224e-08, + "loss": 0.6897, + "step": 46771 + }, + { + "epoch": 1.71, + "learning_rate": 2.8901014391914157e-08, + "loss": 0.6909, + "step": 46772 + }, + { + "epoch": 1.71, + "learning_rate": 2.8897352327242097e-08, + "loss": 0.6978, + "step": 46773 + }, + { + "epoch": 1.71, + "learning_rate": 2.8893690262570037e-08, + "loss": 0.7026, + "step": 46774 + }, + { + "epoch": 1.71, + "learning_rate": 2.8890028197897977e-08, + "loss": 0.7066, + "step": 46775 + }, + { + "epoch": 1.71, + "learning_rate": 2.888636613322591e-08, + "loss": 0.6935, + "step": 46776 + }, + { + "epoch": 1.71, + "learning_rate": 2.8882704068553847e-08, + "loss": 0.6998, + "step": 46777 + }, + { + "epoch": 1.71, + "learning_rate": 2.8879042003881787e-08, + "loss": 0.6882, + "step": 46778 + }, + { + "epoch": 1.71, + "learning_rate": 2.8875379939209727e-08, + "loss": 0.694, + "step": 46779 + }, + { + "epoch": 1.71, + "learning_rate": 2.8871717874537667e-08, + "loss": 0.7105, + "step": 46780 + }, + { + "epoch": 1.71, + "learning_rate": 2.88680558098656e-08, + "loss": 0.693, + "step": 46781 + }, + { + "epoch": 1.71, + "learning_rate": 2.8864393745193537e-08, + "loss": 0.6864, + "step": 46782 + }, + { + "epoch": 1.71, + "learning_rate": 2.8860731680521476e-08, + "loss": 0.7026, + "step": 46783 + }, + { + "epoch": 1.71, + "learning_rate": 2.8857069615849416e-08, + "loss": 0.6945, + "step": 46784 + }, + { + "epoch": 1.71, + "learning_rate": 2.885340755117735e-08, + "loss": 0.6903, + "step": 46785 + }, + { + "epoch": 1.71, + "learning_rate": 2.884974548650529e-08, + "loss": 0.6918, + "step": 46786 + }, + { + "epoch": 1.71, + "learning_rate": 2.884608342183323e-08, + "loss": 0.6981, + "step": 46787 + }, + { + "epoch": 1.71, + "learning_rate": 2.8842421357161166e-08, + "loss": 0.6868, + "step": 46788 + }, + { + "epoch": 1.71, + "learning_rate": 2.8838759292489106e-08, + "loss": 0.6948, + "step": 46789 + }, + { + "epoch": 1.71, + "learning_rate": 2.883509722781704e-08, + "loss": 0.6987, + "step": 46790 + }, + { + "epoch": 1.71, + "learning_rate": 2.883143516314498e-08, + "loss": 0.7002, + "step": 46791 + }, + { + "epoch": 1.71, + "learning_rate": 2.882777309847292e-08, + "loss": 0.7075, + "step": 46792 + }, + { + "epoch": 1.71, + "learning_rate": 2.8824111033800856e-08, + "loss": 0.7007, + "step": 46793 + }, + { + "epoch": 1.71, + "learning_rate": 2.8820448969128793e-08, + "loss": 0.702, + "step": 46794 + }, + { + "epoch": 1.71, + "learning_rate": 2.881678690445673e-08, + "loss": 0.6906, + "step": 46795 + }, + { + "epoch": 1.71, + "learning_rate": 2.881312483978467e-08, + "loss": 0.6951, + "step": 46796 + }, + { + "epoch": 1.71, + "learning_rate": 2.880946277511261e-08, + "loss": 0.6986, + "step": 46797 + }, + { + "epoch": 1.71, + "learning_rate": 2.880580071044055e-08, + "loss": 0.6957, + "step": 46798 + }, + { + "epoch": 1.71, + "learning_rate": 2.8802138645768482e-08, + "loss": 0.6836, + "step": 46799 + }, + { + "epoch": 1.71, + "learning_rate": 2.879847658109642e-08, + "loss": 0.6956, + "step": 46800 + }, + { + "epoch": 1.71, + "learning_rate": 2.879481451642436e-08, + "loss": 0.6976, + "step": 46801 + }, + { + "epoch": 1.71, + "learning_rate": 2.87911524517523e-08, + "loss": 0.6987, + "step": 46802 + }, + { + "epoch": 1.71, + "learning_rate": 2.8787490387080232e-08, + "loss": 0.6846, + "step": 46803 + }, + { + "epoch": 1.71, + "learning_rate": 2.8783828322408172e-08, + "loss": 0.684, + "step": 46804 + }, + { + "epoch": 1.71, + "learning_rate": 2.8780166257736112e-08, + "loss": 0.6874, + "step": 46805 + }, + { + "epoch": 1.71, + "learning_rate": 2.877650419306405e-08, + "loss": 0.6958, + "step": 46806 + }, + { + "epoch": 1.71, + "learning_rate": 2.877284212839199e-08, + "loss": 0.6908, + "step": 46807 + }, + { + "epoch": 1.71, + "learning_rate": 2.8769180063719922e-08, + "loss": 0.6905, + "step": 46808 + }, + { + "epoch": 1.71, + "learning_rate": 2.8765517999047862e-08, + "loss": 0.7033, + "step": 46809 + }, + { + "epoch": 1.71, + "learning_rate": 2.87618559343758e-08, + "loss": 0.6873, + "step": 46810 + }, + { + "epoch": 1.71, + "learning_rate": 2.8758193869703738e-08, + "loss": 0.7009, + "step": 46811 + }, + { + "epoch": 1.71, + "learning_rate": 2.875453180503167e-08, + "loss": 0.7108, + "step": 46812 + }, + { + "epoch": 1.71, + "learning_rate": 2.875086974035961e-08, + "loss": 0.692, + "step": 46813 + }, + { + "epoch": 1.71, + "learning_rate": 2.874720767568755e-08, + "loss": 0.7024, + "step": 46814 + }, + { + "epoch": 1.71, + "learning_rate": 2.874354561101549e-08, + "loss": 0.6973, + "step": 46815 + }, + { + "epoch": 1.71, + "learning_rate": 2.873988354634343e-08, + "loss": 0.6938, + "step": 46816 + }, + { + "epoch": 1.71, + "learning_rate": 2.8736221481671365e-08, + "loss": 0.6944, + "step": 46817 + }, + { + "epoch": 1.71, + "learning_rate": 2.87325594169993e-08, + "loss": 0.7026, + "step": 46818 + }, + { + "epoch": 1.71, + "learning_rate": 2.872889735232724e-08, + "loss": 0.6875, + "step": 46819 + }, + { + "epoch": 1.71, + "learning_rate": 2.872523528765518e-08, + "loss": 0.6956, + "step": 46820 + }, + { + "epoch": 1.71, + "learning_rate": 2.8721573222983114e-08, + "loss": 0.6966, + "step": 46821 + }, + { + "epoch": 1.71, + "learning_rate": 2.8717911158311054e-08, + "loss": 0.6896, + "step": 46822 + }, + { + "epoch": 1.71, + "learning_rate": 2.871424909363899e-08, + "loss": 0.698, + "step": 46823 + }, + { + "epoch": 1.71, + "learning_rate": 2.871058702896693e-08, + "loss": 0.6917, + "step": 46824 + }, + { + "epoch": 1.71, + "learning_rate": 2.870692496429487e-08, + "loss": 0.6991, + "step": 46825 + }, + { + "epoch": 1.71, + "learning_rate": 2.8703262899622804e-08, + "loss": 0.6998, + "step": 46826 + }, + { + "epoch": 1.71, + "learning_rate": 2.8699600834950744e-08, + "loss": 0.6962, + "step": 46827 + }, + { + "epoch": 1.71, + "learning_rate": 2.8695938770278684e-08, + "loss": 0.6885, + "step": 46828 + }, + { + "epoch": 1.71, + "learning_rate": 2.869227670560662e-08, + "loss": 0.6866, + "step": 46829 + }, + { + "epoch": 1.71, + "learning_rate": 2.8688614640934554e-08, + "loss": 0.7056, + "step": 46830 + }, + { + "epoch": 1.71, + "learning_rate": 2.8684952576262494e-08, + "loss": 0.6833, + "step": 46831 + }, + { + "epoch": 1.71, + "learning_rate": 2.8681290511590434e-08, + "loss": 0.6873, + "step": 46832 + }, + { + "epoch": 1.71, + "learning_rate": 2.8677628446918374e-08, + "loss": 0.6952, + "step": 46833 + }, + { + "epoch": 1.71, + "learning_rate": 2.867396638224631e-08, + "loss": 0.6951, + "step": 46834 + }, + { + "epoch": 1.71, + "learning_rate": 2.8670304317574247e-08, + "loss": 0.6927, + "step": 46835 + }, + { + "epoch": 1.71, + "learning_rate": 2.8666642252902184e-08, + "loss": 0.6908, + "step": 46836 + }, + { + "epoch": 1.71, + "learning_rate": 2.8662980188230123e-08, + "loss": 0.6935, + "step": 46837 + }, + { + "epoch": 1.71, + "learning_rate": 2.8659318123558063e-08, + "loss": 0.6928, + "step": 46838 + }, + { + "epoch": 1.71, + "learning_rate": 2.8655656058885997e-08, + "loss": 0.7018, + "step": 46839 + }, + { + "epoch": 1.71, + "learning_rate": 2.8651993994213937e-08, + "loss": 0.6995, + "step": 46840 + }, + { + "epoch": 1.71, + "learning_rate": 2.8648331929541873e-08, + "loss": 0.6929, + "step": 46841 + }, + { + "epoch": 1.71, + "learning_rate": 2.8644669864869813e-08, + "loss": 0.6908, + "step": 46842 + }, + { + "epoch": 1.71, + "learning_rate": 2.8641007800197753e-08, + "loss": 0.6955, + "step": 46843 + }, + { + "epoch": 1.71, + "learning_rate": 2.8637345735525686e-08, + "loss": 0.6854, + "step": 46844 + }, + { + "epoch": 1.71, + "learning_rate": 2.8633683670853626e-08, + "loss": 0.6954, + "step": 46845 + }, + { + "epoch": 1.71, + "learning_rate": 2.8630021606181566e-08, + "loss": 0.6855, + "step": 46846 + }, + { + "epoch": 1.71, + "learning_rate": 2.8626359541509503e-08, + "loss": 0.7122, + "step": 46847 + }, + { + "epoch": 1.71, + "learning_rate": 2.8622697476837436e-08, + "loss": 0.6886, + "step": 46848 + }, + { + "epoch": 1.71, + "learning_rate": 2.8619035412165376e-08, + "loss": 0.6985, + "step": 46849 + }, + { + "epoch": 1.71, + "learning_rate": 2.8615373347493316e-08, + "loss": 0.6889, + "step": 46850 + }, + { + "epoch": 1.71, + "learning_rate": 2.8611711282821256e-08, + "loss": 0.6966, + "step": 46851 + }, + { + "epoch": 1.71, + "learning_rate": 2.8608049218149193e-08, + "loss": 0.6852, + "step": 46852 + }, + { + "epoch": 1.71, + "learning_rate": 2.8604387153477126e-08, + "loss": 0.6999, + "step": 46853 + }, + { + "epoch": 1.71, + "learning_rate": 2.8600725088805066e-08, + "loss": 0.6997, + "step": 46854 + }, + { + "epoch": 1.71, + "learning_rate": 2.8597063024133006e-08, + "loss": 0.6981, + "step": 46855 + }, + { + "epoch": 1.71, + "learning_rate": 2.8593400959460946e-08, + "loss": 0.694, + "step": 46856 + }, + { + "epoch": 1.71, + "learning_rate": 2.858973889478888e-08, + "loss": 0.7025, + "step": 46857 + }, + { + "epoch": 1.71, + "learning_rate": 2.858607683011682e-08, + "loss": 0.6948, + "step": 46858 + }, + { + "epoch": 1.71, + "learning_rate": 2.8582414765444756e-08, + "loss": 0.6917, + "step": 46859 + }, + { + "epoch": 1.71, + "learning_rate": 2.8578752700772695e-08, + "loss": 0.7022, + "step": 46860 + }, + { + "epoch": 1.71, + "learning_rate": 2.8575090636100635e-08, + "loss": 0.6926, + "step": 46861 + }, + { + "epoch": 1.71, + "learning_rate": 2.857142857142857e-08, + "loss": 0.7002, + "step": 46862 + }, + { + "epoch": 1.71, + "learning_rate": 2.856776650675651e-08, + "loss": 0.6865, + "step": 46863 + }, + { + "epoch": 1.71, + "learning_rate": 2.8564104442084445e-08, + "loss": 0.6864, + "step": 46864 + }, + { + "epoch": 1.71, + "learning_rate": 2.8560442377412385e-08, + "loss": 0.6934, + "step": 46865 + }, + { + "epoch": 1.71, + "learning_rate": 2.855678031274032e-08, + "loss": 0.6934, + "step": 46866 + }, + { + "epoch": 1.71, + "learning_rate": 2.855311824806826e-08, + "loss": 0.6891, + "step": 46867 + }, + { + "epoch": 1.71, + "learning_rate": 2.85494561833962e-08, + "loss": 0.6932, + "step": 46868 + }, + { + "epoch": 1.71, + "learning_rate": 2.8545794118724138e-08, + "loss": 0.6904, + "step": 46869 + }, + { + "epoch": 1.71, + "learning_rate": 2.8542132054052075e-08, + "loss": 0.6928, + "step": 46870 + }, + { + "epoch": 1.71, + "learning_rate": 2.8538469989380008e-08, + "loss": 0.6841, + "step": 46871 + }, + { + "epoch": 1.71, + "learning_rate": 2.8534807924707948e-08, + "loss": 0.6952, + "step": 46872 + }, + { + "epoch": 1.71, + "learning_rate": 2.8531145860035888e-08, + "loss": 0.7007, + "step": 46873 + }, + { + "epoch": 1.71, + "learning_rate": 2.8527483795363828e-08, + "loss": 0.689, + "step": 46874 + }, + { + "epoch": 1.72, + "learning_rate": 2.852382173069176e-08, + "loss": 0.689, + "step": 46875 + }, + { + "epoch": 1.72, + "learning_rate": 2.85201596660197e-08, + "loss": 0.7062, + "step": 46876 + }, + { + "epoch": 1.72, + "learning_rate": 2.8516497601347638e-08, + "loss": 0.6901, + "step": 46877 + }, + { + "epoch": 1.72, + "learning_rate": 2.8512835536675578e-08, + "loss": 0.6995, + "step": 46878 + }, + { + "epoch": 1.72, + "learning_rate": 2.850917347200351e-08, + "loss": 0.6867, + "step": 46879 + }, + { + "epoch": 1.72, + "learning_rate": 2.850551140733145e-08, + "loss": 0.7004, + "step": 46880 + }, + { + "epoch": 1.72, + "learning_rate": 2.850184934265939e-08, + "loss": 0.7127, + "step": 46881 + }, + { + "epoch": 1.72, + "learning_rate": 2.8498187277987328e-08, + "loss": 0.6989, + "step": 46882 + }, + { + "epoch": 1.72, + "learning_rate": 2.8494525213315268e-08, + "loss": 0.6926, + "step": 46883 + }, + { + "epoch": 1.72, + "learning_rate": 2.84908631486432e-08, + "loss": 0.6999, + "step": 46884 + }, + { + "epoch": 1.72, + "learning_rate": 2.848720108397114e-08, + "loss": 0.6956, + "step": 46885 + }, + { + "epoch": 1.72, + "learning_rate": 2.848353901929908e-08, + "loss": 0.6827, + "step": 46886 + }, + { + "epoch": 1.72, + "learning_rate": 2.847987695462702e-08, + "loss": 0.695, + "step": 46887 + }, + { + "epoch": 1.72, + "learning_rate": 2.8476214889954954e-08, + "loss": 0.7001, + "step": 46888 + }, + { + "epoch": 1.72, + "learning_rate": 2.847255282528289e-08, + "loss": 0.6911, + "step": 46889 + }, + { + "epoch": 1.72, + "learning_rate": 2.846889076061083e-08, + "loss": 0.6925, + "step": 46890 + }, + { + "epoch": 1.72, + "learning_rate": 2.846522869593877e-08, + "loss": 0.6994, + "step": 46891 + }, + { + "epoch": 1.72, + "learning_rate": 2.846156663126671e-08, + "loss": 0.7062, + "step": 46892 + }, + { + "epoch": 1.72, + "learning_rate": 2.8457904566594644e-08, + "loss": 0.7025, + "step": 46893 + }, + { + "epoch": 1.72, + "learning_rate": 2.845424250192258e-08, + "loss": 0.702, + "step": 46894 + }, + { + "epoch": 1.72, + "learning_rate": 2.845058043725052e-08, + "loss": 0.693, + "step": 46895 + }, + { + "epoch": 1.72, + "learning_rate": 2.844691837257846e-08, + "loss": 0.6905, + "step": 46896 + }, + { + "epoch": 1.72, + "learning_rate": 2.8443256307906393e-08, + "loss": 0.6891, + "step": 46897 + }, + { + "epoch": 1.72, + "learning_rate": 2.8439594243234333e-08, + "loss": 0.6948, + "step": 46898 + }, + { + "epoch": 1.72, + "learning_rate": 2.8435932178562273e-08, + "loss": 0.6956, + "step": 46899 + }, + { + "epoch": 1.72, + "learning_rate": 2.843227011389021e-08, + "loss": 0.6907, + "step": 46900 + }, + { + "epoch": 1.72, + "learning_rate": 2.842860804921815e-08, + "loss": 0.6979, + "step": 46901 + }, + { + "epoch": 1.72, + "learning_rate": 2.8424945984546083e-08, + "loss": 0.7058, + "step": 46902 + }, + { + "epoch": 1.72, + "learning_rate": 2.8421283919874023e-08, + "loss": 0.7097, + "step": 46903 + }, + { + "epoch": 1.72, + "learning_rate": 2.8417621855201963e-08, + "loss": 0.6918, + "step": 46904 + }, + { + "epoch": 1.72, + "learning_rate": 2.84139597905299e-08, + "loss": 0.6939, + "step": 46905 + }, + { + "epoch": 1.72, + "learning_rate": 2.8410297725857836e-08, + "loss": 0.6938, + "step": 46906 + }, + { + "epoch": 1.72, + "learning_rate": 2.8406635661185773e-08, + "loss": 0.7099, + "step": 46907 + }, + { + "epoch": 1.72, + "learning_rate": 2.8402973596513713e-08, + "loss": 0.6862, + "step": 46908 + }, + { + "epoch": 1.72, + "learning_rate": 2.8399311531841653e-08, + "loss": 0.6851, + "step": 46909 + }, + { + "epoch": 1.72, + "learning_rate": 2.8395649467169593e-08, + "loss": 0.6974, + "step": 46910 + }, + { + "epoch": 1.72, + "learning_rate": 2.8391987402497526e-08, + "loss": 0.6995, + "step": 46911 + }, + { + "epoch": 1.72, + "learning_rate": 2.8388325337825463e-08, + "loss": 0.6938, + "step": 46912 + }, + { + "epoch": 1.72, + "learning_rate": 2.8384663273153403e-08, + "loss": 0.6908, + "step": 46913 + }, + { + "epoch": 1.72, + "learning_rate": 2.8381001208481342e-08, + "loss": 0.7041, + "step": 46914 + }, + { + "epoch": 1.72, + "learning_rate": 2.8377339143809276e-08, + "loss": 0.6954, + "step": 46915 + }, + { + "epoch": 1.72, + "learning_rate": 2.8373677079137216e-08, + "loss": 0.6905, + "step": 46916 + }, + { + "epoch": 1.72, + "learning_rate": 2.8370015014465156e-08, + "loss": 0.6975, + "step": 46917 + }, + { + "epoch": 1.72, + "learning_rate": 2.8366352949793092e-08, + "loss": 0.6937, + "step": 46918 + }, + { + "epoch": 1.72, + "learning_rate": 2.8362690885121032e-08, + "loss": 0.6913, + "step": 46919 + }, + { + "epoch": 1.72, + "learning_rate": 2.8359028820448965e-08, + "loss": 0.6834, + "step": 46920 + }, + { + "epoch": 1.72, + "learning_rate": 2.8355366755776905e-08, + "loss": 0.6899, + "step": 46921 + }, + { + "epoch": 1.72, + "learning_rate": 2.8351704691104845e-08, + "loss": 0.699, + "step": 46922 + }, + { + "epoch": 1.72, + "learning_rate": 2.8348042626432782e-08, + "loss": 0.6713, + "step": 46923 + }, + { + "epoch": 1.72, + "learning_rate": 2.8344380561760715e-08, + "loss": 0.6966, + "step": 46924 + }, + { + "epoch": 1.72, + "learning_rate": 2.8340718497088655e-08, + "loss": 0.688, + "step": 46925 + }, + { + "epoch": 1.72, + "learning_rate": 2.8337056432416595e-08, + "loss": 0.6957, + "step": 46926 + }, + { + "epoch": 1.72, + "learning_rate": 2.8333394367744535e-08, + "loss": 0.6921, + "step": 46927 + }, + { + "epoch": 1.72, + "learning_rate": 2.8329732303072472e-08, + "loss": 0.6942, + "step": 46928 + }, + { + "epoch": 1.72, + "learning_rate": 2.8326070238400408e-08, + "loss": 0.6879, + "step": 46929 + }, + { + "epoch": 1.72, + "learning_rate": 2.8322408173728345e-08, + "loss": 0.7059, + "step": 46930 + }, + { + "epoch": 1.72, + "learning_rate": 2.8318746109056285e-08, + "loss": 0.6893, + "step": 46931 + }, + { + "epoch": 1.72, + "learning_rate": 2.8315084044384225e-08, + "loss": 0.6904, + "step": 46932 + }, + { + "epoch": 1.72, + "learning_rate": 2.8311421979712158e-08, + "loss": 0.6866, + "step": 46933 + }, + { + "epoch": 1.72, + "learning_rate": 2.8307759915040098e-08, + "loss": 0.703, + "step": 46934 + }, + { + "epoch": 1.72, + "learning_rate": 2.8304097850368035e-08, + "loss": 0.6962, + "step": 46935 + }, + { + "epoch": 1.72, + "learning_rate": 2.8300435785695975e-08, + "loss": 0.6806, + "step": 46936 + }, + { + "epoch": 1.72, + "learning_rate": 2.8296773721023914e-08, + "loss": 0.687, + "step": 46937 + }, + { + "epoch": 1.72, + "learning_rate": 2.8293111656351848e-08, + "loss": 0.6846, + "step": 46938 + }, + { + "epoch": 1.72, + "learning_rate": 2.8289449591679788e-08, + "loss": 0.6993, + "step": 46939 + }, + { + "epoch": 1.72, + "learning_rate": 2.8285787527007728e-08, + "loss": 0.6832, + "step": 46940 + }, + { + "epoch": 1.72, + "learning_rate": 2.8282125462335664e-08, + "loss": 0.7041, + "step": 46941 + }, + { + "epoch": 1.72, + "learning_rate": 2.8278463397663598e-08, + "loss": 0.6863, + "step": 46942 + }, + { + "epoch": 1.72, + "learning_rate": 2.8274801332991537e-08, + "loss": 0.6903, + "step": 46943 + }, + { + "epoch": 1.72, + "learning_rate": 2.8271139268319477e-08, + "loss": 0.6918, + "step": 46944 + }, + { + "epoch": 1.72, + "learning_rate": 2.8267477203647417e-08, + "loss": 0.6935, + "step": 46945 + }, + { + "epoch": 1.72, + "learning_rate": 2.8263815138975354e-08, + "loss": 0.6939, + "step": 46946 + }, + { + "epoch": 1.72, + "learning_rate": 2.826015307430329e-08, + "loss": 0.7022, + "step": 46947 + }, + { + "epoch": 1.72, + "learning_rate": 2.8256491009631227e-08, + "loss": 0.6894, + "step": 46948 + }, + { + "epoch": 1.72, + "learning_rate": 2.8252828944959167e-08, + "loss": 0.6864, + "step": 46949 + }, + { + "epoch": 1.72, + "learning_rate": 2.8249166880287107e-08, + "loss": 0.7111, + "step": 46950 + }, + { + "epoch": 1.72, + "learning_rate": 2.824550481561504e-08, + "loss": 0.7011, + "step": 46951 + }, + { + "epoch": 1.72, + "learning_rate": 2.824184275094298e-08, + "loss": 0.6904, + "step": 46952 + }, + { + "epoch": 1.72, + "learning_rate": 2.8238180686270917e-08, + "loss": 0.704, + "step": 46953 + }, + { + "epoch": 1.72, + "learning_rate": 2.8234518621598857e-08, + "loss": 0.6854, + "step": 46954 + }, + { + "epoch": 1.72, + "learning_rate": 2.8230856556926797e-08, + "loss": 0.6937, + "step": 46955 + }, + { + "epoch": 1.72, + "learning_rate": 2.822719449225473e-08, + "loss": 0.6965, + "step": 46956 + }, + { + "epoch": 1.72, + "learning_rate": 2.822353242758267e-08, + "loss": 0.6892, + "step": 46957 + }, + { + "epoch": 1.72, + "learning_rate": 2.8219870362910607e-08, + "loss": 0.6844, + "step": 46958 + }, + { + "epoch": 1.72, + "learning_rate": 2.8216208298238547e-08, + "loss": 0.696, + "step": 46959 + }, + { + "epoch": 1.72, + "learning_rate": 2.821254623356648e-08, + "loss": 0.69, + "step": 46960 + }, + { + "epoch": 1.72, + "learning_rate": 2.820888416889442e-08, + "loss": 0.6876, + "step": 46961 + }, + { + "epoch": 1.72, + "learning_rate": 2.820522210422236e-08, + "loss": 0.6971, + "step": 46962 + }, + { + "epoch": 1.72, + "learning_rate": 2.82015600395503e-08, + "loss": 0.692, + "step": 46963 + }, + { + "epoch": 1.72, + "learning_rate": 2.8197897974878236e-08, + "loss": 0.6949, + "step": 46964 + }, + { + "epoch": 1.72, + "learning_rate": 2.819423591020617e-08, + "loss": 0.7001, + "step": 46965 + }, + { + "epoch": 1.72, + "learning_rate": 2.819057384553411e-08, + "loss": 0.7011, + "step": 46966 + }, + { + "epoch": 1.72, + "learning_rate": 2.818691178086205e-08, + "loss": 0.6983, + "step": 46967 + }, + { + "epoch": 1.72, + "learning_rate": 2.818324971618999e-08, + "loss": 0.6862, + "step": 46968 + }, + { + "epoch": 1.72, + "learning_rate": 2.8179587651517923e-08, + "loss": 0.6867, + "step": 46969 + }, + { + "epoch": 1.72, + "learning_rate": 2.8175925586845863e-08, + "loss": 0.6991, + "step": 46970 + }, + { + "epoch": 1.72, + "learning_rate": 2.81722635221738e-08, + "loss": 0.6898, + "step": 46971 + }, + { + "epoch": 1.72, + "learning_rate": 2.816860145750174e-08, + "loss": 0.6916, + "step": 46972 + }, + { + "epoch": 1.72, + "learning_rate": 2.816493939282968e-08, + "loss": 0.6908, + "step": 46973 + }, + { + "epoch": 1.72, + "learning_rate": 2.8161277328157612e-08, + "loss": 0.6996, + "step": 46974 + }, + { + "epoch": 1.72, + "learning_rate": 2.8157615263485552e-08, + "loss": 0.7034, + "step": 46975 + }, + { + "epoch": 1.72, + "learning_rate": 2.815395319881349e-08, + "loss": 0.6903, + "step": 46976 + }, + { + "epoch": 1.72, + "learning_rate": 2.815029113414143e-08, + "loss": 0.6857, + "step": 46977 + }, + { + "epoch": 1.72, + "learning_rate": 2.8146629069469362e-08, + "loss": 0.6923, + "step": 46978 + }, + { + "epoch": 1.72, + "learning_rate": 2.8142967004797302e-08, + "loss": 0.6919, + "step": 46979 + }, + { + "epoch": 1.72, + "learning_rate": 2.8139304940125242e-08, + "loss": 0.6959, + "step": 46980 + }, + { + "epoch": 1.72, + "learning_rate": 2.8135642875453182e-08, + "loss": 0.6929, + "step": 46981 + }, + { + "epoch": 1.72, + "learning_rate": 2.813198081078112e-08, + "loss": 0.6917, + "step": 46982 + }, + { + "epoch": 1.72, + "learning_rate": 2.8128318746109052e-08, + "loss": 0.6896, + "step": 46983 + }, + { + "epoch": 1.72, + "learning_rate": 2.8124656681436992e-08, + "loss": 0.6957, + "step": 46984 + }, + { + "epoch": 1.72, + "learning_rate": 2.8120994616764932e-08, + "loss": 0.6866, + "step": 46985 + }, + { + "epoch": 1.72, + "learning_rate": 2.8117332552092872e-08, + "loss": 0.6888, + "step": 46986 + }, + { + "epoch": 1.72, + "learning_rate": 2.8113670487420805e-08, + "loss": 0.7071, + "step": 46987 + }, + { + "epoch": 1.72, + "learning_rate": 2.811000842274874e-08, + "loss": 0.697, + "step": 46988 + }, + { + "epoch": 1.72, + "learning_rate": 2.810634635807668e-08, + "loss": 0.7008, + "step": 46989 + }, + { + "epoch": 1.72, + "learning_rate": 2.810268429340462e-08, + "loss": 0.6883, + "step": 46990 + }, + { + "epoch": 1.72, + "learning_rate": 2.809902222873256e-08, + "loss": 0.7065, + "step": 46991 + }, + { + "epoch": 1.72, + "learning_rate": 2.8095360164060495e-08, + "loss": 0.6927, + "step": 46992 + }, + { + "epoch": 1.72, + "learning_rate": 2.8091698099388435e-08, + "loss": 0.6774, + "step": 46993 + }, + { + "epoch": 1.72, + "learning_rate": 2.808803603471637e-08, + "loss": 0.7015, + "step": 46994 + }, + { + "epoch": 1.72, + "learning_rate": 2.808437397004431e-08, + "loss": 0.6911, + "step": 46995 + }, + { + "epoch": 1.72, + "learning_rate": 2.8080711905372245e-08, + "loss": 0.6781, + "step": 46996 + }, + { + "epoch": 1.72, + "learning_rate": 2.8077049840700184e-08, + "loss": 0.6924, + "step": 46997 + }, + { + "epoch": 1.72, + "learning_rate": 2.8073387776028124e-08, + "loss": 0.688, + "step": 46998 + }, + { + "epoch": 1.72, + "learning_rate": 2.806972571135606e-08, + "loss": 0.6903, + "step": 46999 + }, + { + "epoch": 1.72, + "learning_rate": 2.8066063646684e-08, + "loss": 0.6928, + "step": 47000 + }, + { + "epoch": 1.72, + "learning_rate": 2.8062401582011934e-08, + "loss": 0.6995, + "step": 47001 + }, + { + "epoch": 1.72, + "learning_rate": 2.8058739517339874e-08, + "loss": 0.6947, + "step": 47002 + }, + { + "epoch": 1.72, + "learning_rate": 2.8055077452667814e-08, + "loss": 0.7021, + "step": 47003 + }, + { + "epoch": 1.72, + "learning_rate": 2.8051415387995754e-08, + "loss": 0.693, + "step": 47004 + }, + { + "epoch": 1.72, + "learning_rate": 2.8047753323323687e-08, + "loss": 0.702, + "step": 47005 + }, + { + "epoch": 1.72, + "learning_rate": 2.8044091258651624e-08, + "loss": 0.7007, + "step": 47006 + }, + { + "epoch": 1.72, + "learning_rate": 2.8040429193979564e-08, + "loss": 0.703, + "step": 47007 + }, + { + "epoch": 1.72, + "learning_rate": 2.8036767129307504e-08, + "loss": 0.6853, + "step": 47008 + }, + { + "epoch": 1.72, + "learning_rate": 2.8033105064635444e-08, + "loss": 0.689, + "step": 47009 + }, + { + "epoch": 1.72, + "learning_rate": 2.8029442999963377e-08, + "loss": 0.6974, + "step": 47010 + }, + { + "epoch": 1.72, + "learning_rate": 2.8025780935291317e-08, + "loss": 0.6873, + "step": 47011 + }, + { + "epoch": 1.72, + "learning_rate": 2.8022118870619254e-08, + "loss": 0.7006, + "step": 47012 + }, + { + "epoch": 1.72, + "learning_rate": 2.8018456805947194e-08, + "loss": 0.7017, + "step": 47013 + }, + { + "epoch": 1.72, + "learning_rate": 2.8014794741275127e-08, + "loss": 0.701, + "step": 47014 + }, + { + "epoch": 1.72, + "learning_rate": 2.8011132676603067e-08, + "loss": 0.6885, + "step": 47015 + }, + { + "epoch": 1.72, + "learning_rate": 2.8007470611931007e-08, + "loss": 0.699, + "step": 47016 + }, + { + "epoch": 1.72, + "learning_rate": 2.8003808547258943e-08, + "loss": 0.6939, + "step": 47017 + }, + { + "epoch": 1.72, + "learning_rate": 2.8000146482586883e-08, + "loss": 0.6945, + "step": 47018 + }, + { + "epoch": 1.72, + "learning_rate": 2.7996484417914817e-08, + "loss": 0.6827, + "step": 47019 + }, + { + "epoch": 1.72, + "learning_rate": 2.7992822353242756e-08, + "loss": 0.7026, + "step": 47020 + }, + { + "epoch": 1.72, + "learning_rate": 2.7989160288570696e-08, + "loss": 0.6986, + "step": 47021 + }, + { + "epoch": 1.72, + "learning_rate": 2.7985498223898636e-08, + "loss": 0.6955, + "step": 47022 + }, + { + "epoch": 1.72, + "learning_rate": 2.798183615922657e-08, + "loss": 0.6874, + "step": 47023 + }, + { + "epoch": 1.72, + "learning_rate": 2.7978174094554506e-08, + "loss": 0.6959, + "step": 47024 + }, + { + "epoch": 1.72, + "learning_rate": 2.7974512029882446e-08, + "loss": 0.6825, + "step": 47025 + }, + { + "epoch": 1.72, + "learning_rate": 2.7970849965210386e-08, + "loss": 0.6851, + "step": 47026 + }, + { + "epoch": 1.72, + "learning_rate": 2.7967187900538326e-08, + "loss": 0.6874, + "step": 47027 + }, + { + "epoch": 1.72, + "learning_rate": 2.796352583586626e-08, + "loss": 0.6821, + "step": 47028 + }, + { + "epoch": 1.72, + "learning_rate": 2.7959863771194196e-08, + "loss": 0.685, + "step": 47029 + }, + { + "epoch": 1.72, + "learning_rate": 2.7956201706522136e-08, + "loss": 0.6921, + "step": 47030 + }, + { + "epoch": 1.72, + "learning_rate": 2.7952539641850076e-08, + "loss": 0.675, + "step": 47031 + }, + { + "epoch": 1.72, + "learning_rate": 2.794887757717801e-08, + "loss": 0.6888, + "step": 47032 + }, + { + "epoch": 1.72, + "learning_rate": 2.794521551250595e-08, + "loss": 0.6835, + "step": 47033 + }, + { + "epoch": 1.72, + "learning_rate": 2.794155344783389e-08, + "loss": 0.705, + "step": 47034 + }, + { + "epoch": 1.72, + "learning_rate": 2.7937891383161826e-08, + "loss": 0.692, + "step": 47035 + }, + { + "epoch": 1.72, + "learning_rate": 2.7934229318489766e-08, + "loss": 0.694, + "step": 47036 + }, + { + "epoch": 1.72, + "learning_rate": 2.79305672538177e-08, + "loss": 0.6916, + "step": 47037 + }, + { + "epoch": 1.72, + "learning_rate": 2.792690518914564e-08, + "loss": 0.7074, + "step": 47038 + }, + { + "epoch": 1.72, + "learning_rate": 2.792324312447358e-08, + "loss": 0.6919, + "step": 47039 + }, + { + "epoch": 1.72, + "learning_rate": 2.7919581059801515e-08, + "loss": 0.6959, + "step": 47040 + }, + { + "epoch": 1.72, + "learning_rate": 2.7915918995129452e-08, + "loss": 0.6896, + "step": 47041 + }, + { + "epoch": 1.72, + "learning_rate": 2.791225693045739e-08, + "loss": 0.6972, + "step": 47042 + }, + { + "epoch": 1.72, + "learning_rate": 2.790859486578533e-08, + "loss": 0.7035, + "step": 47043 + }, + { + "epoch": 1.72, + "learning_rate": 2.790493280111327e-08, + "loss": 0.697, + "step": 47044 + }, + { + "epoch": 1.72, + "learning_rate": 2.790127073644121e-08, + "loss": 0.7033, + "step": 47045 + }, + { + "epoch": 1.72, + "learning_rate": 2.7897608671769142e-08, + "loss": 0.6921, + "step": 47046 + }, + { + "epoch": 1.72, + "learning_rate": 2.7893946607097078e-08, + "loss": 0.6822, + "step": 47047 + }, + { + "epoch": 1.72, + "learning_rate": 2.7890284542425018e-08, + "loss": 0.7108, + "step": 47048 + }, + { + "epoch": 1.72, + "learning_rate": 2.7886622477752958e-08, + "loss": 0.6991, + "step": 47049 + }, + { + "epoch": 1.72, + "learning_rate": 2.788296041308089e-08, + "loss": 0.6968, + "step": 47050 + }, + { + "epoch": 1.72, + "learning_rate": 2.787929834840883e-08, + "loss": 0.6901, + "step": 47051 + }, + { + "epoch": 1.72, + "learning_rate": 2.787563628373677e-08, + "loss": 0.6897, + "step": 47052 + }, + { + "epoch": 1.72, + "learning_rate": 2.7871974219064708e-08, + "loss": 0.7012, + "step": 47053 + }, + { + "epoch": 1.72, + "learning_rate": 2.7868312154392648e-08, + "loss": 0.7, + "step": 47054 + }, + { + "epoch": 1.72, + "learning_rate": 2.786465008972058e-08, + "loss": 0.7053, + "step": 47055 + }, + { + "epoch": 1.72, + "learning_rate": 2.786098802504852e-08, + "loss": 0.6866, + "step": 47056 + }, + { + "epoch": 1.72, + "learning_rate": 2.785732596037646e-08, + "loss": 0.6973, + "step": 47057 + }, + { + "epoch": 1.72, + "learning_rate": 2.7853663895704398e-08, + "loss": 0.6945, + "step": 47058 + }, + { + "epoch": 1.72, + "learning_rate": 2.785000183103233e-08, + "loss": 0.6919, + "step": 47059 + }, + { + "epoch": 1.72, + "learning_rate": 2.784633976636027e-08, + "loss": 0.6928, + "step": 47060 + }, + { + "epoch": 1.72, + "learning_rate": 2.784267770168821e-08, + "loss": 0.6875, + "step": 47061 + }, + { + "epoch": 1.72, + "learning_rate": 2.783901563701615e-08, + "loss": 0.6981, + "step": 47062 + }, + { + "epoch": 1.72, + "learning_rate": 2.783535357234409e-08, + "loss": 0.7038, + "step": 47063 + }, + { + "epoch": 1.72, + "learning_rate": 2.7831691507672024e-08, + "loss": 0.6963, + "step": 47064 + }, + { + "epoch": 1.72, + "learning_rate": 2.782802944299996e-08, + "loss": 0.7099, + "step": 47065 + }, + { + "epoch": 1.72, + "learning_rate": 2.78243673783279e-08, + "loss": 0.6843, + "step": 47066 + }, + { + "epoch": 1.72, + "learning_rate": 2.782070531365584e-08, + "loss": 0.6951, + "step": 47067 + }, + { + "epoch": 1.72, + "learning_rate": 2.7817043248983774e-08, + "loss": 0.6929, + "step": 47068 + }, + { + "epoch": 1.72, + "learning_rate": 2.7813381184311714e-08, + "loss": 0.6858, + "step": 47069 + }, + { + "epoch": 1.72, + "learning_rate": 2.780971911963965e-08, + "loss": 0.7002, + "step": 47070 + }, + { + "epoch": 1.72, + "learning_rate": 2.780605705496759e-08, + "loss": 0.6955, + "step": 47071 + }, + { + "epoch": 1.72, + "learning_rate": 2.7802394990295524e-08, + "loss": 0.7023, + "step": 47072 + }, + { + "epoch": 1.72, + "learning_rate": 2.7798732925623464e-08, + "loss": 0.7175, + "step": 47073 + }, + { + "epoch": 1.72, + "learning_rate": 2.7795070860951403e-08, + "loss": 0.6845, + "step": 47074 + }, + { + "epoch": 1.72, + "learning_rate": 2.7791408796279343e-08, + "loss": 0.7001, + "step": 47075 + }, + { + "epoch": 1.72, + "learning_rate": 2.778774673160728e-08, + "loss": 0.6881, + "step": 47076 + }, + { + "epoch": 1.72, + "learning_rate": 2.7784084666935213e-08, + "loss": 0.6894, + "step": 47077 + }, + { + "epoch": 1.72, + "learning_rate": 2.7780422602263153e-08, + "loss": 0.6953, + "step": 47078 + }, + { + "epoch": 1.72, + "learning_rate": 2.7776760537591093e-08, + "loss": 0.6886, + "step": 47079 + }, + { + "epoch": 1.72, + "learning_rate": 2.7773098472919033e-08, + "loss": 0.6921, + "step": 47080 + }, + { + "epoch": 1.72, + "learning_rate": 2.7769436408246966e-08, + "loss": 0.6895, + "step": 47081 + }, + { + "epoch": 1.72, + "learning_rate": 2.7765774343574906e-08, + "loss": 0.6929, + "step": 47082 + }, + { + "epoch": 1.72, + "learning_rate": 2.7762112278902843e-08, + "loss": 0.7069, + "step": 47083 + }, + { + "epoch": 1.72, + "learning_rate": 2.7758450214230783e-08, + "loss": 0.6952, + "step": 47084 + }, + { + "epoch": 1.72, + "learning_rate": 2.7754788149558723e-08, + "loss": 0.7036, + "step": 47085 + }, + { + "epoch": 1.72, + "learning_rate": 2.7751126084886656e-08, + "loss": 0.6887, + "step": 47086 + }, + { + "epoch": 1.72, + "learning_rate": 2.7747464020214596e-08, + "loss": 0.6926, + "step": 47087 + }, + { + "epoch": 1.72, + "learning_rate": 2.7743801955542533e-08, + "loss": 0.695, + "step": 47088 + }, + { + "epoch": 1.72, + "learning_rate": 2.7740139890870473e-08, + "loss": 0.6893, + "step": 47089 + }, + { + "epoch": 1.72, + "learning_rate": 2.7736477826198406e-08, + "loss": 0.6902, + "step": 47090 + }, + { + "epoch": 1.72, + "learning_rate": 2.7732815761526346e-08, + "loss": 0.6944, + "step": 47091 + }, + { + "epoch": 1.72, + "learning_rate": 2.7729153696854286e-08, + "loss": 0.7041, + "step": 47092 + }, + { + "epoch": 1.72, + "learning_rate": 2.7725491632182226e-08, + "loss": 0.7025, + "step": 47093 + }, + { + "epoch": 1.72, + "learning_rate": 2.7721829567510162e-08, + "loss": 0.6924, + "step": 47094 + }, + { + "epoch": 1.72, + "learning_rate": 2.7718167502838096e-08, + "loss": 0.6847, + "step": 47095 + }, + { + "epoch": 1.72, + "learning_rate": 2.7714505438166036e-08, + "loss": 0.6862, + "step": 47096 + }, + { + "epoch": 1.72, + "learning_rate": 2.7710843373493975e-08, + "loss": 0.7001, + "step": 47097 + }, + { + "epoch": 1.72, + "learning_rate": 2.7707181308821915e-08, + "loss": 0.6963, + "step": 47098 + }, + { + "epoch": 1.72, + "learning_rate": 2.770351924414985e-08, + "loss": 0.697, + "step": 47099 + }, + { + "epoch": 1.72, + "learning_rate": 2.7699857179477785e-08, + "loss": 0.6968, + "step": 47100 + }, + { + "epoch": 1.72, + "learning_rate": 2.7696195114805725e-08, + "loss": 0.7103, + "step": 47101 + }, + { + "epoch": 1.72, + "learning_rate": 2.7692533050133665e-08, + "loss": 0.6928, + "step": 47102 + }, + { + "epoch": 1.72, + "learning_rate": 2.7688870985461605e-08, + "loss": 0.7126, + "step": 47103 + }, + { + "epoch": 1.72, + "learning_rate": 2.768520892078954e-08, + "loss": 0.7009, + "step": 47104 + }, + { + "epoch": 1.72, + "learning_rate": 2.768154685611748e-08, + "loss": 0.6928, + "step": 47105 + }, + { + "epoch": 1.72, + "learning_rate": 2.7677884791445415e-08, + "loss": 0.698, + "step": 47106 + }, + { + "epoch": 1.72, + "learning_rate": 2.7674222726773355e-08, + "loss": 0.6831, + "step": 47107 + }, + { + "epoch": 1.72, + "learning_rate": 2.7670560662101288e-08, + "loss": 0.6947, + "step": 47108 + }, + { + "epoch": 1.72, + "learning_rate": 2.7666898597429228e-08, + "loss": 0.6993, + "step": 47109 + }, + { + "epoch": 1.72, + "learning_rate": 2.7663236532757168e-08, + "loss": 0.6971, + "step": 47110 + }, + { + "epoch": 1.72, + "learning_rate": 2.7659574468085105e-08, + "loss": 0.6905, + "step": 47111 + }, + { + "epoch": 1.72, + "learning_rate": 2.7655912403413045e-08, + "loss": 0.6914, + "step": 47112 + }, + { + "epoch": 1.72, + "learning_rate": 2.7652250338740978e-08, + "loss": 0.6926, + "step": 47113 + }, + { + "epoch": 1.72, + "learning_rate": 2.7648588274068918e-08, + "loss": 0.6959, + "step": 47114 + }, + { + "epoch": 1.72, + "learning_rate": 2.7644926209396858e-08, + "loss": 0.701, + "step": 47115 + }, + { + "epoch": 1.72, + "learning_rate": 2.7641264144724798e-08, + "loss": 0.69, + "step": 47116 + }, + { + "epoch": 1.72, + "learning_rate": 2.763760208005273e-08, + "loss": 0.6886, + "step": 47117 + }, + { + "epoch": 1.72, + "learning_rate": 2.7633940015380668e-08, + "loss": 0.69, + "step": 47118 + }, + { + "epoch": 1.72, + "learning_rate": 2.7630277950708608e-08, + "loss": 0.6987, + "step": 47119 + }, + { + "epoch": 1.72, + "learning_rate": 2.7626615886036548e-08, + "loss": 0.6968, + "step": 47120 + }, + { + "epoch": 1.72, + "learning_rate": 2.7622953821364487e-08, + "loss": 0.6965, + "step": 47121 + }, + { + "epoch": 1.72, + "learning_rate": 2.761929175669242e-08, + "loss": 0.6912, + "step": 47122 + }, + { + "epoch": 1.72, + "learning_rate": 2.761562969202036e-08, + "loss": 0.69, + "step": 47123 + }, + { + "epoch": 1.72, + "learning_rate": 2.7611967627348297e-08, + "loss": 0.6883, + "step": 47124 + }, + { + "epoch": 1.72, + "learning_rate": 2.7608305562676237e-08, + "loss": 0.689, + "step": 47125 + }, + { + "epoch": 1.72, + "learning_rate": 2.760464349800417e-08, + "loss": 0.6948, + "step": 47126 + }, + { + "epoch": 1.72, + "learning_rate": 2.760098143333211e-08, + "loss": 0.6908, + "step": 47127 + }, + { + "epoch": 1.72, + "learning_rate": 2.759731936866005e-08, + "loss": 0.7034, + "step": 47128 + }, + { + "epoch": 1.72, + "learning_rate": 2.7593657303987987e-08, + "loss": 0.6936, + "step": 47129 + }, + { + "epoch": 1.72, + "learning_rate": 2.7589995239315927e-08, + "loss": 0.6952, + "step": 47130 + }, + { + "epoch": 1.72, + "learning_rate": 2.758633317464386e-08, + "loss": 0.6939, + "step": 47131 + }, + { + "epoch": 1.72, + "learning_rate": 2.75826711099718e-08, + "loss": 0.6887, + "step": 47132 + }, + { + "epoch": 1.72, + "learning_rate": 2.757900904529974e-08, + "loss": 0.6903, + "step": 47133 + }, + { + "epoch": 1.72, + "learning_rate": 2.7575346980627677e-08, + "loss": 0.703, + "step": 47134 + }, + { + "epoch": 1.72, + "learning_rate": 2.7571684915955613e-08, + "loss": 0.6969, + "step": 47135 + }, + { + "epoch": 1.72, + "learning_rate": 2.756802285128355e-08, + "loss": 0.6797, + "step": 47136 + }, + { + "epoch": 1.72, + "learning_rate": 2.756436078661149e-08, + "loss": 0.691, + "step": 47137 + }, + { + "epoch": 1.72, + "learning_rate": 2.756069872193943e-08, + "loss": 0.6908, + "step": 47138 + }, + { + "epoch": 1.72, + "learning_rate": 2.755703665726737e-08, + "loss": 0.6856, + "step": 47139 + }, + { + "epoch": 1.72, + "learning_rate": 2.7553374592595303e-08, + "loss": 0.6895, + "step": 47140 + }, + { + "epoch": 1.72, + "learning_rate": 2.754971252792324e-08, + "loss": 0.691, + "step": 47141 + }, + { + "epoch": 1.72, + "learning_rate": 2.754605046325118e-08, + "loss": 0.6913, + "step": 47142 + }, + { + "epoch": 1.72, + "learning_rate": 2.754238839857912e-08, + "loss": 0.7013, + "step": 47143 + }, + { + "epoch": 1.72, + "learning_rate": 2.7538726333907053e-08, + "loss": 0.7015, + "step": 47144 + }, + { + "epoch": 1.72, + "learning_rate": 2.7535064269234993e-08, + "loss": 0.6952, + "step": 47145 + }, + { + "epoch": 1.72, + "learning_rate": 2.7531402204562933e-08, + "loss": 0.6983, + "step": 47146 + }, + { + "epoch": 1.72, + "learning_rate": 2.752774013989087e-08, + "loss": 0.6841, + "step": 47147 + }, + { + "epoch": 1.73, + "learning_rate": 2.752407807521881e-08, + "loss": 0.6894, + "step": 47148 + }, + { + "epoch": 1.73, + "learning_rate": 2.7520416010546743e-08, + "loss": 0.6971, + "step": 47149 + }, + { + "epoch": 1.73, + "learning_rate": 2.7516753945874683e-08, + "loss": 0.6938, + "step": 47150 + }, + { + "epoch": 1.73, + "learning_rate": 2.7513091881202622e-08, + "loss": 0.6954, + "step": 47151 + }, + { + "epoch": 1.73, + "learning_rate": 2.750942981653056e-08, + "loss": 0.6976, + "step": 47152 + }, + { + "epoch": 1.73, + "learning_rate": 2.7505767751858496e-08, + "loss": 0.6907, + "step": 47153 + }, + { + "epoch": 1.73, + "learning_rate": 2.7502105687186432e-08, + "loss": 0.6941, + "step": 47154 + }, + { + "epoch": 1.73, + "learning_rate": 2.7498443622514372e-08, + "loss": 0.6898, + "step": 47155 + }, + { + "epoch": 1.73, + "learning_rate": 2.7494781557842312e-08, + "loss": 0.6901, + "step": 47156 + }, + { + "epoch": 1.73, + "learning_rate": 2.7491119493170252e-08, + "loss": 0.6968, + "step": 47157 + }, + { + "epoch": 1.73, + "learning_rate": 2.7487457428498185e-08, + "loss": 0.6806, + "step": 47158 + }, + { + "epoch": 1.73, + "learning_rate": 2.7483795363826122e-08, + "loss": 0.6951, + "step": 47159 + }, + { + "epoch": 1.73, + "learning_rate": 2.7480133299154062e-08, + "loss": 0.6955, + "step": 47160 + }, + { + "epoch": 1.73, + "learning_rate": 2.7476471234482002e-08, + "loss": 0.7147, + "step": 47161 + }, + { + "epoch": 1.73, + "learning_rate": 2.7472809169809935e-08, + "loss": 0.6759, + "step": 47162 + }, + { + "epoch": 1.73, + "learning_rate": 2.7469147105137875e-08, + "loss": 0.7019, + "step": 47163 + }, + { + "epoch": 1.73, + "learning_rate": 2.7465485040465812e-08, + "loss": 0.6829, + "step": 47164 + }, + { + "epoch": 1.73, + "learning_rate": 2.7461822975793752e-08, + "loss": 0.6996, + "step": 47165 + }, + { + "epoch": 1.73, + "learning_rate": 2.745816091112169e-08, + "loss": 0.6889, + "step": 47166 + }, + { + "epoch": 1.73, + "learning_rate": 2.7454498846449625e-08, + "loss": 0.6853, + "step": 47167 + }, + { + "epoch": 1.73, + "learning_rate": 2.7450836781777565e-08, + "loss": 0.6986, + "step": 47168 + }, + { + "epoch": 1.73, + "learning_rate": 2.7447174717105505e-08, + "loss": 0.6886, + "step": 47169 + }, + { + "epoch": 1.73, + "learning_rate": 2.744351265243344e-08, + "loss": 0.6862, + "step": 47170 + }, + { + "epoch": 1.73, + "learning_rate": 2.7439850587761375e-08, + "loss": 0.7046, + "step": 47171 + }, + { + "epoch": 1.73, + "learning_rate": 2.7436188523089315e-08, + "loss": 0.703, + "step": 47172 + }, + { + "epoch": 1.73, + "learning_rate": 2.7432526458417255e-08, + "loss": 0.6953, + "step": 47173 + }, + { + "epoch": 1.73, + "learning_rate": 2.7428864393745194e-08, + "loss": 0.6953, + "step": 47174 + }, + { + "epoch": 1.73, + "learning_rate": 2.742520232907313e-08, + "loss": 0.6861, + "step": 47175 + }, + { + "epoch": 1.73, + "learning_rate": 2.7421540264401068e-08, + "loss": 0.6832, + "step": 47176 + }, + { + "epoch": 1.73, + "learning_rate": 2.7417878199729004e-08, + "loss": 0.6924, + "step": 47177 + }, + { + "epoch": 1.73, + "learning_rate": 2.7414216135056944e-08, + "loss": 0.6976, + "step": 47178 + }, + { + "epoch": 1.73, + "learning_rate": 2.7410554070384884e-08, + "loss": 0.7004, + "step": 47179 + }, + { + "epoch": 1.73, + "learning_rate": 2.7406892005712818e-08, + "loss": 0.6933, + "step": 47180 + }, + { + "epoch": 1.73, + "learning_rate": 2.7403229941040757e-08, + "loss": 0.6996, + "step": 47181 + }, + { + "epoch": 1.73, + "learning_rate": 2.7399567876368694e-08, + "loss": 0.6895, + "step": 47182 + }, + { + "epoch": 1.73, + "learning_rate": 2.7395905811696634e-08, + "loss": 0.7105, + "step": 47183 + }, + { + "epoch": 1.73, + "learning_rate": 2.7392243747024574e-08, + "loss": 0.6867, + "step": 47184 + }, + { + "epoch": 1.73, + "learning_rate": 2.7388581682352507e-08, + "loss": 0.6886, + "step": 47185 + }, + { + "epoch": 1.73, + "learning_rate": 2.7384919617680447e-08, + "loss": 0.685, + "step": 47186 + }, + { + "epoch": 1.73, + "learning_rate": 2.7381257553008387e-08, + "loss": 0.6961, + "step": 47187 + }, + { + "epoch": 1.73, + "learning_rate": 2.7377595488336324e-08, + "loss": 0.6919, + "step": 47188 + }, + { + "epoch": 1.73, + "learning_rate": 2.7373933423664257e-08, + "loss": 0.6935, + "step": 47189 + }, + { + "epoch": 1.73, + "learning_rate": 2.7370271358992197e-08, + "loss": 0.6936, + "step": 47190 + }, + { + "epoch": 1.73, + "learning_rate": 2.7366609294320137e-08, + "loss": 0.6974, + "step": 47191 + }, + { + "epoch": 1.73, + "learning_rate": 2.7362947229648077e-08, + "loss": 0.6977, + "step": 47192 + }, + { + "epoch": 1.73, + "learning_rate": 2.7359285164976013e-08, + "loss": 0.7147, + "step": 47193 + }, + { + "epoch": 1.73, + "learning_rate": 2.7355623100303947e-08, + "loss": 0.6862, + "step": 47194 + }, + { + "epoch": 1.73, + "learning_rate": 2.7351961035631887e-08, + "loss": 0.6845, + "step": 47195 + }, + { + "epoch": 1.73, + "learning_rate": 2.7348298970959827e-08, + "loss": 0.7014, + "step": 47196 + }, + { + "epoch": 1.73, + "learning_rate": 2.7344636906287767e-08, + "loss": 0.7006, + "step": 47197 + }, + { + "epoch": 1.73, + "learning_rate": 2.73409748416157e-08, + "loss": 0.6955, + "step": 47198 + }, + { + "epoch": 1.73, + "learning_rate": 2.733731277694364e-08, + "loss": 0.69, + "step": 47199 + }, + { + "epoch": 1.73, + "learning_rate": 2.7333650712271576e-08, + "loss": 0.6917, + "step": 47200 + }, + { + "epoch": 1.73, + "learning_rate": 2.7329988647599516e-08, + "loss": 0.7117, + "step": 47201 + }, + { + "epoch": 1.73, + "learning_rate": 2.7326326582927456e-08, + "loss": 0.6953, + "step": 47202 + }, + { + "epoch": 1.73, + "learning_rate": 2.732266451825539e-08, + "loss": 0.6957, + "step": 47203 + }, + { + "epoch": 1.73, + "learning_rate": 2.731900245358333e-08, + "loss": 0.7021, + "step": 47204 + }, + { + "epoch": 1.73, + "learning_rate": 2.7315340388911266e-08, + "loss": 0.6749, + "step": 47205 + }, + { + "epoch": 1.73, + "learning_rate": 2.7311678324239206e-08, + "loss": 0.6943, + "step": 47206 + }, + { + "epoch": 1.73, + "learning_rate": 2.730801625956714e-08, + "loss": 0.6979, + "step": 47207 + }, + { + "epoch": 1.73, + "learning_rate": 2.730435419489508e-08, + "loss": 0.678, + "step": 47208 + }, + { + "epoch": 1.73, + "learning_rate": 2.730069213022302e-08, + "loss": 0.6862, + "step": 47209 + }, + { + "epoch": 1.73, + "learning_rate": 2.729703006555096e-08, + "loss": 0.7035, + "step": 47210 + }, + { + "epoch": 1.73, + "learning_rate": 2.7293368000878896e-08, + "loss": 0.6925, + "step": 47211 + }, + { + "epoch": 1.73, + "learning_rate": 2.728970593620683e-08, + "loss": 0.6976, + "step": 47212 + }, + { + "epoch": 1.73, + "learning_rate": 2.728604387153477e-08, + "loss": 0.6974, + "step": 47213 + }, + { + "epoch": 1.73, + "learning_rate": 2.728238180686271e-08, + "loss": 0.6893, + "step": 47214 + }, + { + "epoch": 1.73, + "learning_rate": 2.727871974219065e-08, + "loss": 0.689, + "step": 47215 + }, + { + "epoch": 1.73, + "learning_rate": 2.7275057677518582e-08, + "loss": 0.6982, + "step": 47216 + }, + { + "epoch": 1.73, + "learning_rate": 2.7271395612846522e-08, + "loss": 0.6974, + "step": 47217 + }, + { + "epoch": 1.73, + "learning_rate": 2.726773354817446e-08, + "loss": 0.6892, + "step": 47218 + }, + { + "epoch": 1.73, + "learning_rate": 2.72640714835024e-08, + "loss": 0.6875, + "step": 47219 + }, + { + "epoch": 1.73, + "learning_rate": 2.726040941883034e-08, + "loss": 0.6856, + "step": 47220 + }, + { + "epoch": 1.73, + "learning_rate": 2.7256747354158272e-08, + "loss": 0.6939, + "step": 47221 + }, + { + "epoch": 1.73, + "learning_rate": 2.7253085289486212e-08, + "loss": 0.6912, + "step": 47222 + }, + { + "epoch": 1.73, + "learning_rate": 2.724942322481415e-08, + "loss": 0.685, + "step": 47223 + }, + { + "epoch": 1.73, + "learning_rate": 2.7245761160142088e-08, + "loss": 0.6972, + "step": 47224 + }, + { + "epoch": 1.73, + "learning_rate": 2.724209909547002e-08, + "loss": 0.6821, + "step": 47225 + }, + { + "epoch": 1.73, + "learning_rate": 2.723843703079796e-08, + "loss": 0.6868, + "step": 47226 + }, + { + "epoch": 1.73, + "learning_rate": 2.72347749661259e-08, + "loss": 0.696, + "step": 47227 + }, + { + "epoch": 1.73, + "learning_rate": 2.723111290145384e-08, + "loss": 0.6916, + "step": 47228 + }, + { + "epoch": 1.73, + "learning_rate": 2.7227450836781778e-08, + "loss": 0.6941, + "step": 47229 + }, + { + "epoch": 1.73, + "learning_rate": 2.722378877210971e-08, + "loss": 0.707, + "step": 47230 + }, + { + "epoch": 1.73, + "learning_rate": 2.722012670743765e-08, + "loss": 0.693, + "step": 47231 + }, + { + "epoch": 1.73, + "learning_rate": 2.721646464276559e-08, + "loss": 0.682, + "step": 47232 + }, + { + "epoch": 1.73, + "learning_rate": 2.721280257809353e-08, + "loss": 0.6994, + "step": 47233 + }, + { + "epoch": 1.73, + "learning_rate": 2.7209140513421464e-08, + "loss": 0.693, + "step": 47234 + }, + { + "epoch": 1.73, + "learning_rate": 2.72054784487494e-08, + "loss": 0.7038, + "step": 47235 + }, + { + "epoch": 1.73, + "learning_rate": 2.720181638407734e-08, + "loss": 0.6988, + "step": 47236 + }, + { + "epoch": 1.73, + "learning_rate": 2.719815431940528e-08, + "loss": 0.6907, + "step": 47237 + }, + { + "epoch": 1.73, + "learning_rate": 2.719449225473322e-08, + "loss": 0.6986, + "step": 47238 + }, + { + "epoch": 1.73, + "learning_rate": 2.7190830190061154e-08, + "loss": 0.7039, + "step": 47239 + }, + { + "epoch": 1.73, + "learning_rate": 2.7187168125389094e-08, + "loss": 0.6892, + "step": 47240 + }, + { + "epoch": 1.73, + "learning_rate": 2.718350606071703e-08, + "loss": 0.7066, + "step": 47241 + }, + { + "epoch": 1.73, + "learning_rate": 2.717984399604497e-08, + "loss": 0.6859, + "step": 47242 + }, + { + "epoch": 1.73, + "learning_rate": 2.7176181931372904e-08, + "loss": 0.6916, + "step": 47243 + }, + { + "epoch": 1.73, + "learning_rate": 2.7172519866700844e-08, + "loss": 0.709, + "step": 47244 + }, + { + "epoch": 1.73, + "learning_rate": 2.7168857802028784e-08, + "loss": 0.6914, + "step": 47245 + }, + { + "epoch": 1.73, + "learning_rate": 2.716519573735672e-08, + "loss": 0.6964, + "step": 47246 + }, + { + "epoch": 1.73, + "learning_rate": 2.716153367268466e-08, + "loss": 0.7036, + "step": 47247 + }, + { + "epoch": 1.73, + "learning_rate": 2.7157871608012594e-08, + "loss": 0.6899, + "step": 47248 + }, + { + "epoch": 1.73, + "learning_rate": 2.7154209543340534e-08, + "loss": 0.6895, + "step": 47249 + }, + { + "epoch": 1.73, + "learning_rate": 2.7150547478668474e-08, + "loss": 0.6976, + "step": 47250 + }, + { + "epoch": 1.73, + "learning_rate": 2.7146885413996413e-08, + "loss": 0.695, + "step": 47251 + }, + { + "epoch": 1.73, + "learning_rate": 2.7143223349324347e-08, + "loss": 0.6962, + "step": 47252 + }, + { + "epoch": 1.73, + "learning_rate": 2.7139561284652283e-08, + "loss": 0.6971, + "step": 47253 + }, + { + "epoch": 1.73, + "learning_rate": 2.7135899219980223e-08, + "loss": 0.6919, + "step": 47254 + }, + { + "epoch": 1.73, + "learning_rate": 2.7132237155308163e-08, + "loss": 0.6878, + "step": 47255 + }, + { + "epoch": 1.73, + "learning_rate": 2.7128575090636103e-08, + "loss": 0.6894, + "step": 47256 + }, + { + "epoch": 1.73, + "learning_rate": 2.7124913025964037e-08, + "loss": 0.689, + "step": 47257 + }, + { + "epoch": 1.73, + "learning_rate": 2.7121250961291976e-08, + "loss": 0.6816, + "step": 47258 + }, + { + "epoch": 1.73, + "learning_rate": 2.7117588896619913e-08, + "loss": 0.6804, + "step": 47259 + }, + { + "epoch": 1.73, + "learning_rate": 2.7113926831947853e-08, + "loss": 0.7029, + "step": 47260 + }, + { + "epoch": 1.73, + "learning_rate": 2.7110264767275786e-08, + "loss": 0.6942, + "step": 47261 + }, + { + "epoch": 1.73, + "learning_rate": 2.7106602702603726e-08, + "loss": 0.6959, + "step": 47262 + }, + { + "epoch": 1.73, + "learning_rate": 2.7102940637931666e-08, + "loss": 0.6902, + "step": 47263 + }, + { + "epoch": 1.73, + "learning_rate": 2.7099278573259603e-08, + "loss": 0.6955, + "step": 47264 + }, + { + "epoch": 1.73, + "learning_rate": 2.7095616508587536e-08, + "loss": 0.687, + "step": 47265 + }, + { + "epoch": 1.73, + "learning_rate": 2.7091954443915476e-08, + "loss": 0.6905, + "step": 47266 + }, + { + "epoch": 1.73, + "learning_rate": 2.7088292379243416e-08, + "loss": 0.7032, + "step": 47267 + }, + { + "epoch": 1.73, + "learning_rate": 2.7084630314571356e-08, + "loss": 0.685, + "step": 47268 + }, + { + "epoch": 1.73, + "learning_rate": 2.7080968249899296e-08, + "loss": 0.694, + "step": 47269 + }, + { + "epoch": 1.73, + "learning_rate": 2.707730618522723e-08, + "loss": 0.6926, + "step": 47270 + }, + { + "epoch": 1.73, + "learning_rate": 2.7073644120555166e-08, + "loss": 0.6946, + "step": 47271 + }, + { + "epoch": 1.73, + "learning_rate": 2.7069982055883106e-08, + "loss": 0.7026, + "step": 47272 + }, + { + "epoch": 1.73, + "learning_rate": 2.7066319991211046e-08, + "loss": 0.6928, + "step": 47273 + }, + { + "epoch": 1.73, + "learning_rate": 2.706265792653898e-08, + "loss": 0.6956, + "step": 47274 + }, + { + "epoch": 1.73, + "learning_rate": 2.705899586186692e-08, + "loss": 0.6984, + "step": 47275 + }, + { + "epoch": 1.73, + "learning_rate": 2.7055333797194855e-08, + "loss": 0.6885, + "step": 47276 + }, + { + "epoch": 1.73, + "learning_rate": 2.7051671732522795e-08, + "loss": 0.6967, + "step": 47277 + }, + { + "epoch": 1.73, + "learning_rate": 2.7048009667850735e-08, + "loss": 0.694, + "step": 47278 + }, + { + "epoch": 1.73, + "learning_rate": 2.704434760317867e-08, + "loss": 0.6929, + "step": 47279 + }, + { + "epoch": 1.73, + "learning_rate": 2.704068553850661e-08, + "loss": 0.6893, + "step": 47280 + }, + { + "epoch": 1.73, + "learning_rate": 2.703702347383455e-08, + "loss": 0.7034, + "step": 47281 + }, + { + "epoch": 1.73, + "learning_rate": 2.7033361409162485e-08, + "loss": 0.7017, + "step": 47282 + }, + { + "epoch": 1.73, + "learning_rate": 2.702969934449042e-08, + "loss": 0.6973, + "step": 47283 + }, + { + "epoch": 1.73, + "learning_rate": 2.7026037279818358e-08, + "loss": 0.691, + "step": 47284 + }, + { + "epoch": 1.73, + "learning_rate": 2.7022375215146298e-08, + "loss": 0.7012, + "step": 47285 + }, + { + "epoch": 1.73, + "learning_rate": 2.7018713150474238e-08, + "loss": 0.693, + "step": 47286 + }, + { + "epoch": 1.73, + "learning_rate": 2.7015051085802175e-08, + "loss": 0.6881, + "step": 47287 + }, + { + "epoch": 1.73, + "learning_rate": 2.701138902113011e-08, + "loss": 0.6861, + "step": 47288 + }, + { + "epoch": 1.73, + "learning_rate": 2.7007726956458048e-08, + "loss": 0.6812, + "step": 47289 + }, + { + "epoch": 1.73, + "learning_rate": 2.7004064891785988e-08, + "loss": 0.6954, + "step": 47290 + }, + { + "epoch": 1.73, + "learning_rate": 2.7000402827113928e-08, + "loss": 0.6997, + "step": 47291 + }, + { + "epoch": 1.73, + "learning_rate": 2.699674076244186e-08, + "loss": 0.6896, + "step": 47292 + }, + { + "epoch": 1.73, + "learning_rate": 2.69930786977698e-08, + "loss": 0.6917, + "step": 47293 + }, + { + "epoch": 1.73, + "learning_rate": 2.6989416633097738e-08, + "loss": 0.693, + "step": 47294 + }, + { + "epoch": 1.73, + "learning_rate": 2.6985754568425678e-08, + "loss": 0.7012, + "step": 47295 + }, + { + "epoch": 1.73, + "learning_rate": 2.6982092503753618e-08, + "loss": 0.7015, + "step": 47296 + }, + { + "epoch": 1.73, + "learning_rate": 2.697843043908155e-08, + "loss": 0.6946, + "step": 47297 + }, + { + "epoch": 1.73, + "learning_rate": 2.697476837440949e-08, + "loss": 0.6921, + "step": 47298 + }, + { + "epoch": 1.73, + "learning_rate": 2.697110630973743e-08, + "loss": 0.7004, + "step": 47299 + }, + { + "epoch": 1.73, + "learning_rate": 2.6967444245065367e-08, + "loss": 0.6897, + "step": 47300 + }, + { + "epoch": 1.73, + "learning_rate": 2.69637821803933e-08, + "loss": 0.6912, + "step": 47301 + }, + { + "epoch": 1.73, + "learning_rate": 2.696012011572124e-08, + "loss": 0.6951, + "step": 47302 + }, + { + "epoch": 1.73, + "learning_rate": 2.695645805104918e-08, + "loss": 0.6813, + "step": 47303 + }, + { + "epoch": 1.73, + "learning_rate": 2.695279598637712e-08, + "loss": 0.6954, + "step": 47304 + }, + { + "epoch": 1.73, + "learning_rate": 2.6949133921705057e-08, + "loss": 0.6873, + "step": 47305 + }, + { + "epoch": 1.73, + "learning_rate": 2.694547185703299e-08, + "loss": 0.6892, + "step": 47306 + }, + { + "epoch": 1.73, + "learning_rate": 2.694180979236093e-08, + "loss": 0.7033, + "step": 47307 + }, + { + "epoch": 1.73, + "learning_rate": 2.693814772768887e-08, + "loss": 0.6911, + "step": 47308 + }, + { + "epoch": 1.73, + "learning_rate": 2.693448566301681e-08, + "loss": 0.6796, + "step": 47309 + }, + { + "epoch": 1.73, + "learning_rate": 2.6930823598344744e-08, + "loss": 0.6999, + "step": 47310 + }, + { + "epoch": 1.73, + "learning_rate": 2.6927161533672683e-08, + "loss": 0.6991, + "step": 47311 + }, + { + "epoch": 1.73, + "learning_rate": 2.692349946900062e-08, + "loss": 0.6884, + "step": 47312 + }, + { + "epoch": 1.73, + "learning_rate": 2.691983740432856e-08, + "loss": 0.6841, + "step": 47313 + }, + { + "epoch": 1.73, + "learning_rate": 2.69161753396565e-08, + "loss": 0.6907, + "step": 47314 + }, + { + "epoch": 1.73, + "learning_rate": 2.6912513274984433e-08, + "loss": 0.6996, + "step": 47315 + }, + { + "epoch": 1.73, + "learning_rate": 2.6908851210312373e-08, + "loss": 0.6904, + "step": 47316 + }, + { + "epoch": 1.73, + "learning_rate": 2.690518914564031e-08, + "loss": 0.6794, + "step": 47317 + }, + { + "epoch": 1.73, + "learning_rate": 2.690152708096825e-08, + "loss": 0.6851, + "step": 47318 + }, + { + "epoch": 1.73, + "learning_rate": 2.6897865016296183e-08, + "loss": 0.7059, + "step": 47319 + }, + { + "epoch": 1.73, + "learning_rate": 2.6894202951624123e-08, + "loss": 0.6907, + "step": 47320 + }, + { + "epoch": 1.73, + "learning_rate": 2.6890540886952063e-08, + "loss": 0.6971, + "step": 47321 + }, + { + "epoch": 1.73, + "learning_rate": 2.6886878822280003e-08, + "loss": 0.6926, + "step": 47322 + }, + { + "epoch": 1.73, + "learning_rate": 2.688321675760794e-08, + "loss": 0.6916, + "step": 47323 + }, + { + "epoch": 1.73, + "learning_rate": 2.6879554692935873e-08, + "loss": 0.6892, + "step": 47324 + }, + { + "epoch": 1.73, + "learning_rate": 2.6875892628263813e-08, + "loss": 0.6938, + "step": 47325 + }, + { + "epoch": 1.73, + "learning_rate": 2.6872230563591753e-08, + "loss": 0.6761, + "step": 47326 + }, + { + "epoch": 1.73, + "learning_rate": 2.6868568498919693e-08, + "loss": 0.7005, + "step": 47327 + }, + { + "epoch": 1.73, + "learning_rate": 2.6864906434247626e-08, + "loss": 0.6992, + "step": 47328 + }, + { + "epoch": 1.73, + "learning_rate": 2.6861244369575566e-08, + "loss": 0.6942, + "step": 47329 + }, + { + "epoch": 1.73, + "learning_rate": 2.6857582304903502e-08, + "loss": 0.6922, + "step": 47330 + }, + { + "epoch": 1.73, + "learning_rate": 2.6853920240231442e-08, + "loss": 0.7022, + "step": 47331 + }, + { + "epoch": 1.73, + "learning_rate": 2.6850258175559382e-08, + "loss": 0.6873, + "step": 47332 + }, + { + "epoch": 1.73, + "learning_rate": 2.6846596110887316e-08, + "loss": 0.7014, + "step": 47333 + }, + { + "epoch": 1.73, + "learning_rate": 2.6842934046215255e-08, + "loss": 0.6875, + "step": 47334 + }, + { + "epoch": 1.73, + "learning_rate": 2.6839271981543192e-08, + "loss": 0.6902, + "step": 47335 + }, + { + "epoch": 1.73, + "learning_rate": 2.6835609916871132e-08, + "loss": 0.6887, + "step": 47336 + }, + { + "epoch": 1.73, + "learning_rate": 2.6831947852199065e-08, + "loss": 0.701, + "step": 47337 + }, + { + "epoch": 1.73, + "learning_rate": 2.6828285787527005e-08, + "loss": 0.6947, + "step": 47338 + }, + { + "epoch": 1.73, + "learning_rate": 2.6824623722854945e-08, + "loss": 0.7063, + "step": 47339 + }, + { + "epoch": 1.73, + "learning_rate": 2.6820961658182882e-08, + "loss": 0.6858, + "step": 47340 + }, + { + "epoch": 1.73, + "learning_rate": 2.6817299593510822e-08, + "loss": 0.6745, + "step": 47341 + }, + { + "epoch": 1.73, + "learning_rate": 2.6813637528838755e-08, + "loss": 0.6852, + "step": 47342 + }, + { + "epoch": 1.73, + "learning_rate": 2.6809975464166695e-08, + "loss": 0.7094, + "step": 47343 + }, + { + "epoch": 1.73, + "learning_rate": 2.6806313399494635e-08, + "loss": 0.7038, + "step": 47344 + }, + { + "epoch": 1.73, + "learning_rate": 2.6802651334822575e-08, + "loss": 0.6884, + "step": 47345 + }, + { + "epoch": 1.73, + "learning_rate": 2.6798989270150508e-08, + "loss": 0.6934, + "step": 47346 + }, + { + "epoch": 1.73, + "learning_rate": 2.6795327205478445e-08, + "loss": 0.6923, + "step": 47347 + }, + { + "epoch": 1.73, + "learning_rate": 2.6791665140806385e-08, + "loss": 0.6944, + "step": 47348 + }, + { + "epoch": 1.73, + "learning_rate": 2.6788003076134325e-08, + "loss": 0.6962, + "step": 47349 + }, + { + "epoch": 1.73, + "learning_rate": 2.6784341011462265e-08, + "loss": 0.6917, + "step": 47350 + }, + { + "epoch": 1.73, + "learning_rate": 2.6780678946790198e-08, + "loss": 0.6835, + "step": 47351 + }, + { + "epoch": 1.73, + "learning_rate": 2.6777016882118138e-08, + "loss": 0.7026, + "step": 47352 + }, + { + "epoch": 1.73, + "learning_rate": 2.6773354817446074e-08, + "loss": 0.6951, + "step": 47353 + }, + { + "epoch": 1.73, + "learning_rate": 2.6769692752774014e-08, + "loss": 0.693, + "step": 47354 + }, + { + "epoch": 1.73, + "learning_rate": 2.6766030688101948e-08, + "loss": 0.6961, + "step": 47355 + }, + { + "epoch": 1.73, + "learning_rate": 2.6762368623429888e-08, + "loss": 0.6953, + "step": 47356 + }, + { + "epoch": 1.73, + "learning_rate": 2.6758706558757828e-08, + "loss": 0.6812, + "step": 47357 + }, + { + "epoch": 1.73, + "learning_rate": 2.6755044494085764e-08, + "loss": 0.6936, + "step": 47358 + }, + { + "epoch": 1.73, + "learning_rate": 2.6751382429413704e-08, + "loss": 0.6919, + "step": 47359 + }, + { + "epoch": 1.73, + "learning_rate": 2.6747720364741637e-08, + "loss": 0.6995, + "step": 47360 + }, + { + "epoch": 1.73, + "learning_rate": 2.6744058300069577e-08, + "loss": 0.6827, + "step": 47361 + }, + { + "epoch": 1.73, + "learning_rate": 2.6740396235397517e-08, + "loss": 0.6997, + "step": 47362 + }, + { + "epoch": 1.73, + "learning_rate": 2.6736734170725457e-08, + "loss": 0.7004, + "step": 47363 + }, + { + "epoch": 1.73, + "learning_rate": 2.673307210605339e-08, + "loss": 0.6899, + "step": 47364 + }, + { + "epoch": 1.73, + "learning_rate": 2.6729410041381327e-08, + "loss": 0.6865, + "step": 47365 + }, + { + "epoch": 1.73, + "learning_rate": 2.6725747976709267e-08, + "loss": 0.6883, + "step": 47366 + }, + { + "epoch": 1.73, + "learning_rate": 2.6722085912037207e-08, + "loss": 0.6937, + "step": 47367 + }, + { + "epoch": 1.73, + "learning_rate": 2.6718423847365147e-08, + "loss": 0.6893, + "step": 47368 + }, + { + "epoch": 1.73, + "learning_rate": 2.671476178269308e-08, + "loss": 0.7064, + "step": 47369 + }, + { + "epoch": 1.73, + "learning_rate": 2.6711099718021017e-08, + "loss": 0.6826, + "step": 47370 + }, + { + "epoch": 1.73, + "learning_rate": 2.6707437653348957e-08, + "loss": 0.6957, + "step": 47371 + }, + { + "epoch": 1.73, + "learning_rate": 2.6703775588676897e-08, + "loss": 0.6923, + "step": 47372 + }, + { + "epoch": 1.73, + "learning_rate": 2.670011352400483e-08, + "loss": 0.6978, + "step": 47373 + }, + { + "epoch": 1.73, + "learning_rate": 2.669645145933277e-08, + "loss": 0.6873, + "step": 47374 + }, + { + "epoch": 1.73, + "learning_rate": 2.669278939466071e-08, + "loss": 0.6842, + "step": 47375 + }, + { + "epoch": 1.73, + "learning_rate": 2.6689127329988646e-08, + "loss": 0.7021, + "step": 47376 + }, + { + "epoch": 1.73, + "learning_rate": 2.6685465265316586e-08, + "loss": 0.6833, + "step": 47377 + }, + { + "epoch": 1.73, + "learning_rate": 2.668180320064452e-08, + "loss": 0.696, + "step": 47378 + }, + { + "epoch": 1.73, + "learning_rate": 2.667814113597246e-08, + "loss": 0.6862, + "step": 47379 + }, + { + "epoch": 1.73, + "learning_rate": 2.66744790713004e-08, + "loss": 0.6956, + "step": 47380 + }, + { + "epoch": 1.73, + "learning_rate": 2.6670817006628336e-08, + "loss": 0.6897, + "step": 47381 + }, + { + "epoch": 1.73, + "learning_rate": 2.6667154941956273e-08, + "loss": 0.695, + "step": 47382 + }, + { + "epoch": 1.73, + "learning_rate": 2.666349287728421e-08, + "loss": 0.7006, + "step": 47383 + }, + { + "epoch": 1.73, + "learning_rate": 2.665983081261215e-08, + "loss": 0.6937, + "step": 47384 + }, + { + "epoch": 1.73, + "learning_rate": 2.665616874794009e-08, + "loss": 0.6885, + "step": 47385 + }, + { + "epoch": 1.73, + "learning_rate": 2.665250668326803e-08, + "loss": 0.689, + "step": 47386 + }, + { + "epoch": 1.73, + "learning_rate": 2.6648844618595963e-08, + "loss": 0.6902, + "step": 47387 + }, + { + "epoch": 1.73, + "learning_rate": 2.66451825539239e-08, + "loss": 0.6931, + "step": 47388 + }, + { + "epoch": 1.73, + "learning_rate": 2.664152048925184e-08, + "loss": 0.6971, + "step": 47389 + }, + { + "epoch": 1.73, + "learning_rate": 2.663785842457978e-08, + "loss": 0.6851, + "step": 47390 + }, + { + "epoch": 1.73, + "learning_rate": 2.6634196359907712e-08, + "loss": 0.6985, + "step": 47391 + }, + { + "epoch": 1.73, + "learning_rate": 2.6630534295235652e-08, + "loss": 0.6836, + "step": 47392 + }, + { + "epoch": 1.73, + "learning_rate": 2.6626872230563592e-08, + "loss": 0.6967, + "step": 47393 + }, + { + "epoch": 1.73, + "learning_rate": 2.662321016589153e-08, + "loss": 0.6895, + "step": 47394 + }, + { + "epoch": 1.73, + "learning_rate": 2.661954810121947e-08, + "loss": 0.6959, + "step": 47395 + }, + { + "epoch": 1.73, + "learning_rate": 2.6615886036547402e-08, + "loss": 0.6823, + "step": 47396 + }, + { + "epoch": 1.73, + "learning_rate": 2.6612223971875342e-08, + "loss": 0.6993, + "step": 47397 + }, + { + "epoch": 1.73, + "learning_rate": 2.6608561907203282e-08, + "loss": 0.695, + "step": 47398 + }, + { + "epoch": 1.73, + "learning_rate": 2.660489984253122e-08, + "loss": 0.6864, + "step": 47399 + }, + { + "epoch": 1.73, + "learning_rate": 2.6601237777859152e-08, + "loss": 0.7023, + "step": 47400 + }, + { + "epoch": 1.73, + "learning_rate": 2.6597575713187092e-08, + "loss": 0.7054, + "step": 47401 + }, + { + "epoch": 1.73, + "learning_rate": 2.6593913648515032e-08, + "loss": 0.7058, + "step": 47402 + }, + { + "epoch": 1.73, + "learning_rate": 2.659025158384297e-08, + "loss": 0.6896, + "step": 47403 + }, + { + "epoch": 1.73, + "learning_rate": 2.658658951917091e-08, + "loss": 0.692, + "step": 47404 + }, + { + "epoch": 1.73, + "learning_rate": 2.6582927454498845e-08, + "loss": 0.683, + "step": 47405 + }, + { + "epoch": 1.73, + "learning_rate": 2.657926538982678e-08, + "loss": 0.6814, + "step": 47406 + }, + { + "epoch": 1.73, + "learning_rate": 2.657560332515472e-08, + "loss": 0.7024, + "step": 47407 + }, + { + "epoch": 1.73, + "learning_rate": 2.657194126048266e-08, + "loss": 0.6884, + "step": 47408 + }, + { + "epoch": 1.73, + "learning_rate": 2.6568279195810595e-08, + "loss": 0.6888, + "step": 47409 + }, + { + "epoch": 1.73, + "learning_rate": 2.6564617131138535e-08, + "loss": 0.6987, + "step": 47410 + }, + { + "epoch": 1.73, + "learning_rate": 2.656095506646647e-08, + "loss": 0.7037, + "step": 47411 + }, + { + "epoch": 1.73, + "learning_rate": 2.655729300179441e-08, + "loss": 0.6795, + "step": 47412 + }, + { + "epoch": 1.73, + "learning_rate": 2.655363093712235e-08, + "loss": 0.6811, + "step": 47413 + }, + { + "epoch": 1.73, + "learning_rate": 2.6549968872450284e-08, + "loss": 0.677, + "step": 47414 + }, + { + "epoch": 1.73, + "learning_rate": 2.6546306807778224e-08, + "loss": 0.6965, + "step": 47415 + }, + { + "epoch": 1.73, + "learning_rate": 2.6542644743106164e-08, + "loss": 0.6845, + "step": 47416 + }, + { + "epoch": 1.73, + "learning_rate": 2.65389826784341e-08, + "loss": 0.6883, + "step": 47417 + }, + { + "epoch": 1.73, + "learning_rate": 2.6535320613762034e-08, + "loss": 0.6856, + "step": 47418 + }, + { + "epoch": 1.73, + "learning_rate": 2.6531658549089974e-08, + "loss": 0.7012, + "step": 47419 + }, + { + "epoch": 1.73, + "learning_rate": 2.6527996484417914e-08, + "loss": 0.6925, + "step": 47420 + }, + { + "epoch": 1.73, + "learning_rate": 2.6524334419745854e-08, + "loss": 0.695, + "step": 47421 + }, + { + "epoch": 1.74, + "learning_rate": 2.652067235507379e-08, + "loss": 0.6962, + "step": 47422 + }, + { + "epoch": 1.74, + "learning_rate": 2.6517010290401727e-08, + "loss": 0.6922, + "step": 47423 + }, + { + "epoch": 1.74, + "learning_rate": 2.6513348225729664e-08, + "loss": 0.6918, + "step": 47424 + }, + { + "epoch": 1.74, + "learning_rate": 2.6509686161057604e-08, + "loss": 0.6947, + "step": 47425 + }, + { + "epoch": 1.74, + "learning_rate": 2.6506024096385544e-08, + "loss": 0.6836, + "step": 47426 + }, + { + "epoch": 1.74, + "learning_rate": 2.6502362031713477e-08, + "loss": 0.6902, + "step": 47427 + }, + { + "epoch": 1.74, + "learning_rate": 2.6498699967041417e-08, + "loss": 0.6958, + "step": 47428 + }, + { + "epoch": 1.74, + "learning_rate": 2.6495037902369353e-08, + "loss": 0.6915, + "step": 47429 + }, + { + "epoch": 1.74, + "learning_rate": 2.6491375837697293e-08, + "loss": 0.6949, + "step": 47430 + }, + { + "epoch": 1.74, + "learning_rate": 2.6487713773025233e-08, + "loss": 0.6886, + "step": 47431 + }, + { + "epoch": 1.74, + "learning_rate": 2.6484051708353167e-08, + "loss": 0.6908, + "step": 47432 + }, + { + "epoch": 1.74, + "learning_rate": 2.6480389643681107e-08, + "loss": 0.6874, + "step": 47433 + }, + { + "epoch": 1.74, + "learning_rate": 2.6476727579009047e-08, + "loss": 0.6948, + "step": 47434 + }, + { + "epoch": 1.74, + "learning_rate": 2.6473065514336983e-08, + "loss": 0.6905, + "step": 47435 + }, + { + "epoch": 1.74, + "learning_rate": 2.6469403449664916e-08, + "loss": 0.6899, + "step": 47436 + }, + { + "epoch": 1.74, + "learning_rate": 2.6465741384992856e-08, + "loss": 0.6943, + "step": 47437 + }, + { + "epoch": 1.74, + "learning_rate": 2.6462079320320796e-08, + "loss": 0.6909, + "step": 47438 + }, + { + "epoch": 1.74, + "learning_rate": 2.6458417255648736e-08, + "loss": 0.6913, + "step": 47439 + }, + { + "epoch": 1.74, + "learning_rate": 2.6454755190976673e-08, + "loss": 0.692, + "step": 47440 + }, + { + "epoch": 1.74, + "learning_rate": 2.6451093126304606e-08, + "loss": 0.6815, + "step": 47441 + }, + { + "epoch": 1.74, + "learning_rate": 2.6447431061632546e-08, + "loss": 0.6874, + "step": 47442 + }, + { + "epoch": 1.74, + "learning_rate": 2.6443768996960486e-08, + "loss": 0.6965, + "step": 47443 + }, + { + "epoch": 1.74, + "learning_rate": 2.6440106932288426e-08, + "loss": 0.6976, + "step": 47444 + }, + { + "epoch": 1.74, + "learning_rate": 2.643644486761636e-08, + "loss": 0.6855, + "step": 47445 + }, + { + "epoch": 1.74, + "learning_rate": 2.64327828029443e-08, + "loss": 0.7043, + "step": 47446 + }, + { + "epoch": 1.74, + "learning_rate": 2.6429120738272236e-08, + "loss": 0.6994, + "step": 47447 + }, + { + "epoch": 1.74, + "learning_rate": 2.6425458673600176e-08, + "loss": 0.6948, + "step": 47448 + }, + { + "epoch": 1.74, + "learning_rate": 2.6421796608928116e-08, + "loss": 0.6893, + "step": 47449 + }, + { + "epoch": 1.74, + "learning_rate": 2.641813454425605e-08, + "loss": 0.6931, + "step": 47450 + }, + { + "epoch": 1.74, + "learning_rate": 2.641447247958399e-08, + "loss": 0.683, + "step": 47451 + }, + { + "epoch": 1.74, + "learning_rate": 2.6410810414911926e-08, + "loss": 0.7032, + "step": 47452 + }, + { + "epoch": 1.74, + "learning_rate": 2.6407148350239865e-08, + "loss": 0.6887, + "step": 47453 + }, + { + "epoch": 1.74, + "learning_rate": 2.64034862855678e-08, + "loss": 0.684, + "step": 47454 + }, + { + "epoch": 1.74, + "learning_rate": 2.639982422089574e-08, + "loss": 0.6926, + "step": 47455 + }, + { + "epoch": 1.74, + "learning_rate": 2.639616215622368e-08, + "loss": 0.6892, + "step": 47456 + }, + { + "epoch": 1.74, + "learning_rate": 2.639250009155162e-08, + "loss": 0.6906, + "step": 47457 + }, + { + "epoch": 1.74, + "learning_rate": 2.6388838026879552e-08, + "loss": 0.6929, + "step": 47458 + }, + { + "epoch": 1.74, + "learning_rate": 2.638517596220749e-08, + "loss": 0.699, + "step": 47459 + }, + { + "epoch": 1.74, + "learning_rate": 2.638151389753543e-08, + "loss": 0.6937, + "step": 47460 + }, + { + "epoch": 1.74, + "learning_rate": 2.637785183286337e-08, + "loss": 0.7039, + "step": 47461 + }, + { + "epoch": 1.74, + "learning_rate": 2.6374189768191308e-08, + "loss": 0.6913, + "step": 47462 + }, + { + "epoch": 1.74, + "learning_rate": 2.637052770351924e-08, + "loss": 0.6942, + "step": 47463 + }, + { + "epoch": 1.74, + "learning_rate": 2.636686563884718e-08, + "loss": 0.6958, + "step": 47464 + }, + { + "epoch": 1.74, + "learning_rate": 2.6363203574175118e-08, + "loss": 0.6836, + "step": 47465 + }, + { + "epoch": 1.74, + "learning_rate": 2.6359541509503058e-08, + "loss": 0.6869, + "step": 47466 + }, + { + "epoch": 1.74, + "learning_rate": 2.635587944483099e-08, + "loss": 0.6918, + "step": 47467 + }, + { + "epoch": 1.74, + "learning_rate": 2.635221738015893e-08, + "loss": 0.6988, + "step": 47468 + }, + { + "epoch": 1.74, + "learning_rate": 2.634855531548687e-08, + "loss": 0.6879, + "step": 47469 + }, + { + "epoch": 1.74, + "learning_rate": 2.6344893250814808e-08, + "loss": 0.6897, + "step": 47470 + }, + { + "epoch": 1.74, + "learning_rate": 2.6341231186142748e-08, + "loss": 0.6903, + "step": 47471 + }, + { + "epoch": 1.74, + "learning_rate": 2.633756912147068e-08, + "loss": 0.6911, + "step": 47472 + }, + { + "epoch": 1.74, + "learning_rate": 2.633390705679862e-08, + "loss": 0.6907, + "step": 47473 + }, + { + "epoch": 1.74, + "learning_rate": 2.633024499212656e-08, + "loss": 0.6915, + "step": 47474 + }, + { + "epoch": 1.74, + "learning_rate": 2.63265829274545e-08, + "loss": 0.6925, + "step": 47475 + }, + { + "epoch": 1.74, + "learning_rate": 2.6322920862782434e-08, + "loss": 0.6941, + "step": 47476 + }, + { + "epoch": 1.74, + "learning_rate": 2.631925879811037e-08, + "loss": 0.694, + "step": 47477 + }, + { + "epoch": 1.74, + "learning_rate": 2.631559673343831e-08, + "loss": 0.6983, + "step": 47478 + }, + { + "epoch": 1.74, + "learning_rate": 2.631193466876625e-08, + "loss": 0.6772, + "step": 47479 + }, + { + "epoch": 1.74, + "learning_rate": 2.630827260409419e-08, + "loss": 0.6981, + "step": 47480 + }, + { + "epoch": 1.74, + "learning_rate": 2.6304610539422124e-08, + "loss": 0.6885, + "step": 47481 + }, + { + "epoch": 1.74, + "learning_rate": 2.630094847475006e-08, + "loss": 0.6975, + "step": 47482 + }, + { + "epoch": 1.74, + "learning_rate": 2.6297286410078e-08, + "loss": 0.6944, + "step": 47483 + }, + { + "epoch": 1.74, + "learning_rate": 2.629362434540594e-08, + "loss": 0.7021, + "step": 47484 + }, + { + "epoch": 1.74, + "learning_rate": 2.6289962280733874e-08, + "loss": 0.6936, + "step": 47485 + }, + { + "epoch": 1.74, + "learning_rate": 2.6286300216061814e-08, + "loss": 0.6822, + "step": 47486 + }, + { + "epoch": 1.74, + "learning_rate": 2.6282638151389754e-08, + "loss": 0.6888, + "step": 47487 + }, + { + "epoch": 1.74, + "learning_rate": 2.627897608671769e-08, + "loss": 0.6896, + "step": 47488 + }, + { + "epoch": 1.74, + "learning_rate": 2.627531402204563e-08, + "loss": 0.6985, + "step": 47489 + }, + { + "epoch": 1.74, + "learning_rate": 2.6271651957373563e-08, + "loss": 0.7, + "step": 47490 + }, + { + "epoch": 1.74, + "learning_rate": 2.6267989892701503e-08, + "loss": 0.6995, + "step": 47491 + }, + { + "epoch": 1.74, + "learning_rate": 2.6264327828029443e-08, + "loss": 0.6718, + "step": 47492 + }, + { + "epoch": 1.74, + "learning_rate": 2.626066576335738e-08, + "loss": 0.6899, + "step": 47493 + }, + { + "epoch": 1.74, + "learning_rate": 2.6257003698685317e-08, + "loss": 0.6976, + "step": 47494 + }, + { + "epoch": 1.74, + "learning_rate": 2.6253341634013253e-08, + "loss": 0.6841, + "step": 47495 + }, + { + "epoch": 1.74, + "learning_rate": 2.6249679569341193e-08, + "loss": 0.6762, + "step": 47496 + }, + { + "epoch": 1.74, + "learning_rate": 2.6246017504669133e-08, + "loss": 0.684, + "step": 47497 + }, + { + "epoch": 1.74, + "learning_rate": 2.6242355439997073e-08, + "loss": 0.6945, + "step": 47498 + }, + { + "epoch": 1.74, + "learning_rate": 2.6238693375325006e-08, + "loss": 0.7028, + "step": 47499 + }, + { + "epoch": 1.74, + "learning_rate": 2.6235031310652943e-08, + "loss": 0.6896, + "step": 47500 + }, + { + "epoch": 1.74, + "learning_rate": 2.6231369245980883e-08, + "loss": 0.6945, + "step": 47501 + }, + { + "epoch": 1.74, + "learning_rate": 2.6227707181308823e-08, + "loss": 0.6803, + "step": 47502 + }, + { + "epoch": 1.74, + "learning_rate": 2.6224045116636756e-08, + "loss": 0.6892, + "step": 47503 + }, + { + "epoch": 1.74, + "learning_rate": 2.6220383051964696e-08, + "loss": 0.6879, + "step": 47504 + }, + { + "epoch": 1.74, + "learning_rate": 2.6216720987292636e-08, + "loss": 0.6941, + "step": 47505 + }, + { + "epoch": 1.74, + "learning_rate": 2.6213058922620572e-08, + "loss": 0.6928, + "step": 47506 + }, + { + "epoch": 1.74, + "learning_rate": 2.6209396857948512e-08, + "loss": 0.6956, + "step": 47507 + }, + { + "epoch": 1.74, + "learning_rate": 2.6205734793276446e-08, + "loss": 0.6866, + "step": 47508 + }, + { + "epoch": 1.74, + "learning_rate": 2.6202072728604386e-08, + "loss": 0.7045, + "step": 47509 + }, + { + "epoch": 1.74, + "learning_rate": 2.6198410663932326e-08, + "loss": 0.6899, + "step": 47510 + }, + { + "epoch": 1.74, + "learning_rate": 2.6194748599260262e-08, + "loss": 0.6963, + "step": 47511 + }, + { + "epoch": 1.74, + "learning_rate": 2.6191086534588196e-08, + "loss": 0.6975, + "step": 47512 + }, + { + "epoch": 1.74, + "learning_rate": 2.6187424469916135e-08, + "loss": 0.6981, + "step": 47513 + }, + { + "epoch": 1.74, + "learning_rate": 2.6183762405244075e-08, + "loss": 0.7015, + "step": 47514 + }, + { + "epoch": 1.74, + "learning_rate": 2.6180100340572015e-08, + "loss": 0.6906, + "step": 47515 + }, + { + "epoch": 1.74, + "learning_rate": 2.6176438275899952e-08, + "loss": 0.6963, + "step": 47516 + }, + { + "epoch": 1.74, + "learning_rate": 2.617277621122789e-08, + "loss": 0.6916, + "step": 47517 + }, + { + "epoch": 1.74, + "learning_rate": 2.6169114146555825e-08, + "loss": 0.6895, + "step": 47518 + }, + { + "epoch": 1.74, + "learning_rate": 2.6165452081883765e-08, + "loss": 0.6951, + "step": 47519 + }, + { + "epoch": 1.74, + "learning_rate": 2.6161790017211705e-08, + "loss": 0.7042, + "step": 47520 + }, + { + "epoch": 1.74, + "learning_rate": 2.6158127952539638e-08, + "loss": 0.7006, + "step": 47521 + }, + { + "epoch": 1.74, + "learning_rate": 2.6154465887867578e-08, + "loss": 0.6945, + "step": 47522 + }, + { + "epoch": 1.74, + "learning_rate": 2.6150803823195515e-08, + "loss": 0.699, + "step": 47523 + }, + { + "epoch": 1.74, + "learning_rate": 2.6147141758523455e-08, + "loss": 0.6939, + "step": 47524 + }, + { + "epoch": 1.74, + "learning_rate": 2.6143479693851395e-08, + "loss": 0.6849, + "step": 47525 + }, + { + "epoch": 1.74, + "learning_rate": 2.6139817629179328e-08, + "loss": 0.6966, + "step": 47526 + }, + { + "epoch": 1.74, + "learning_rate": 2.6136155564507268e-08, + "loss": 0.6878, + "step": 47527 + }, + { + "epoch": 1.74, + "learning_rate": 2.6132493499835208e-08, + "loss": 0.6932, + "step": 47528 + }, + { + "epoch": 1.74, + "learning_rate": 2.6128831435163145e-08, + "loss": 0.6913, + "step": 47529 + }, + { + "epoch": 1.74, + "learning_rate": 2.6125169370491078e-08, + "loss": 0.6991, + "step": 47530 + }, + { + "epoch": 1.74, + "learning_rate": 2.6121507305819018e-08, + "loss": 0.6902, + "step": 47531 + }, + { + "epoch": 1.74, + "learning_rate": 2.6117845241146958e-08, + "loss": 0.6892, + "step": 47532 + }, + { + "epoch": 1.74, + "learning_rate": 2.6114183176474898e-08, + "loss": 0.701, + "step": 47533 + }, + { + "epoch": 1.74, + "learning_rate": 2.6110521111802834e-08, + "loss": 0.6845, + "step": 47534 + }, + { + "epoch": 1.74, + "learning_rate": 2.610685904713077e-08, + "loss": 0.6932, + "step": 47535 + }, + { + "epoch": 1.74, + "learning_rate": 2.6103196982458707e-08, + "loss": 0.6924, + "step": 47536 + }, + { + "epoch": 1.74, + "learning_rate": 2.6099534917786647e-08, + "loss": 0.7005, + "step": 47537 + }, + { + "epoch": 1.74, + "learning_rate": 2.6095872853114587e-08, + "loss": 0.6877, + "step": 47538 + }, + { + "epoch": 1.74, + "learning_rate": 2.609221078844252e-08, + "loss": 0.6959, + "step": 47539 + }, + { + "epoch": 1.74, + "learning_rate": 2.608854872377046e-08, + "loss": 0.7015, + "step": 47540 + }, + { + "epoch": 1.74, + "learning_rate": 2.6084886659098397e-08, + "loss": 0.7027, + "step": 47541 + }, + { + "epoch": 1.74, + "learning_rate": 2.6081224594426337e-08, + "loss": 0.6819, + "step": 47542 + }, + { + "epoch": 1.74, + "learning_rate": 2.6077562529754277e-08, + "loss": 0.6972, + "step": 47543 + }, + { + "epoch": 1.74, + "learning_rate": 2.607390046508221e-08, + "loss": 0.6843, + "step": 47544 + }, + { + "epoch": 1.74, + "learning_rate": 2.607023840041015e-08, + "loss": 0.676, + "step": 47545 + }, + { + "epoch": 1.74, + "learning_rate": 2.6066576335738087e-08, + "loss": 0.7067, + "step": 47546 + }, + { + "epoch": 1.74, + "learning_rate": 2.6062914271066027e-08, + "loss": 0.7, + "step": 47547 + }, + { + "epoch": 1.74, + "learning_rate": 2.605925220639396e-08, + "loss": 0.6928, + "step": 47548 + }, + { + "epoch": 1.74, + "learning_rate": 2.60555901417219e-08, + "loss": 0.6871, + "step": 47549 + }, + { + "epoch": 1.74, + "learning_rate": 2.605192807704984e-08, + "loss": 0.6996, + "step": 47550 + }, + { + "epoch": 1.74, + "learning_rate": 2.604826601237778e-08, + "loss": 0.6916, + "step": 47551 + }, + { + "epoch": 1.74, + "learning_rate": 2.6044603947705717e-08, + "loss": 0.6866, + "step": 47552 + }, + { + "epoch": 1.74, + "learning_rate": 2.604094188303365e-08, + "loss": 0.6937, + "step": 47553 + }, + { + "epoch": 1.74, + "learning_rate": 2.603727981836159e-08, + "loss": 0.7015, + "step": 47554 + }, + { + "epoch": 1.74, + "learning_rate": 2.603361775368953e-08, + "loss": 0.6885, + "step": 47555 + }, + { + "epoch": 1.74, + "learning_rate": 2.602995568901747e-08, + "loss": 0.6866, + "step": 47556 + }, + { + "epoch": 1.74, + "learning_rate": 2.6026293624345403e-08, + "loss": 0.6883, + "step": 47557 + }, + { + "epoch": 1.74, + "learning_rate": 2.6022631559673343e-08, + "loss": 0.6983, + "step": 47558 + }, + { + "epoch": 1.74, + "learning_rate": 2.601896949500128e-08, + "loss": 0.6749, + "step": 47559 + }, + { + "epoch": 1.74, + "learning_rate": 2.601530743032922e-08, + "loss": 0.6966, + "step": 47560 + }, + { + "epoch": 1.74, + "learning_rate": 2.601164536565716e-08, + "loss": 0.6791, + "step": 47561 + }, + { + "epoch": 1.74, + "learning_rate": 2.6007983300985093e-08, + "loss": 0.6892, + "step": 47562 + }, + { + "epoch": 1.74, + "learning_rate": 2.6004321236313033e-08, + "loss": 0.6841, + "step": 47563 + }, + { + "epoch": 1.74, + "learning_rate": 2.600065917164097e-08, + "loss": 0.6815, + "step": 47564 + }, + { + "epoch": 1.74, + "learning_rate": 2.599699710696891e-08, + "loss": 0.6922, + "step": 47565 + }, + { + "epoch": 1.74, + "learning_rate": 2.5993335042296842e-08, + "loss": 0.6997, + "step": 47566 + }, + { + "epoch": 1.74, + "learning_rate": 2.5989672977624782e-08, + "loss": 0.6833, + "step": 47567 + }, + { + "epoch": 1.74, + "learning_rate": 2.5986010912952722e-08, + "loss": 0.6953, + "step": 47568 + }, + { + "epoch": 1.74, + "learning_rate": 2.5982348848280662e-08, + "loss": 0.6965, + "step": 47569 + }, + { + "epoch": 1.74, + "learning_rate": 2.59786867836086e-08, + "loss": 0.6991, + "step": 47570 + }, + { + "epoch": 1.74, + "learning_rate": 2.5975024718936532e-08, + "loss": 0.706, + "step": 47571 + }, + { + "epoch": 1.74, + "learning_rate": 2.5971362654264472e-08, + "loss": 0.6906, + "step": 47572 + }, + { + "epoch": 1.74, + "learning_rate": 2.5967700589592412e-08, + "loss": 0.6947, + "step": 47573 + }, + { + "epoch": 1.74, + "learning_rate": 2.5964038524920352e-08, + "loss": 0.6976, + "step": 47574 + }, + { + "epoch": 1.74, + "learning_rate": 2.5960376460248285e-08, + "loss": 0.686, + "step": 47575 + }, + { + "epoch": 1.74, + "learning_rate": 2.5956714395576222e-08, + "loss": 0.691, + "step": 47576 + }, + { + "epoch": 1.74, + "learning_rate": 2.5953052330904162e-08, + "loss": 0.6953, + "step": 47577 + }, + { + "epoch": 1.74, + "learning_rate": 2.5949390266232102e-08, + "loss": 0.6988, + "step": 47578 + }, + { + "epoch": 1.74, + "learning_rate": 2.5945728201560042e-08, + "loss": 0.6867, + "step": 47579 + }, + { + "epoch": 1.74, + "learning_rate": 2.5942066136887975e-08, + "loss": 0.6814, + "step": 47580 + }, + { + "epoch": 1.74, + "learning_rate": 2.5938404072215915e-08, + "loss": 0.6983, + "step": 47581 + }, + { + "epoch": 1.74, + "learning_rate": 2.593474200754385e-08, + "loss": 0.702, + "step": 47582 + }, + { + "epoch": 1.74, + "learning_rate": 2.593107994287179e-08, + "loss": 0.7026, + "step": 47583 + }, + { + "epoch": 1.74, + "learning_rate": 2.5927417878199725e-08, + "loss": 0.6984, + "step": 47584 + }, + { + "epoch": 1.74, + "learning_rate": 2.5923755813527665e-08, + "loss": 0.6926, + "step": 47585 + }, + { + "epoch": 1.74, + "learning_rate": 2.5920093748855605e-08, + "loss": 0.7027, + "step": 47586 + }, + { + "epoch": 1.74, + "learning_rate": 2.591643168418354e-08, + "loss": 0.6886, + "step": 47587 + }, + { + "epoch": 1.74, + "learning_rate": 2.591276961951148e-08, + "loss": 0.6896, + "step": 47588 + }, + { + "epoch": 1.74, + "learning_rate": 2.5909107554839415e-08, + "loss": 0.6996, + "step": 47589 + }, + { + "epoch": 1.74, + "learning_rate": 2.5905445490167354e-08, + "loss": 0.7119, + "step": 47590 + }, + { + "epoch": 1.74, + "learning_rate": 2.5901783425495294e-08, + "loss": 0.6988, + "step": 47591 + }, + { + "epoch": 1.74, + "learning_rate": 2.5898121360823234e-08, + "loss": 0.7093, + "step": 47592 + }, + { + "epoch": 1.74, + "learning_rate": 2.5894459296151168e-08, + "loss": 0.7001, + "step": 47593 + }, + { + "epoch": 1.74, + "learning_rate": 2.5890797231479104e-08, + "loss": 0.6977, + "step": 47594 + }, + { + "epoch": 1.74, + "learning_rate": 2.5887135166807044e-08, + "loss": 0.6899, + "step": 47595 + }, + { + "epoch": 1.74, + "learning_rate": 2.5883473102134984e-08, + "loss": 0.6922, + "step": 47596 + }, + { + "epoch": 1.74, + "learning_rate": 2.5879811037462924e-08, + "loss": 0.697, + "step": 47597 + }, + { + "epoch": 1.74, + "learning_rate": 2.5876148972790857e-08, + "loss": 0.6929, + "step": 47598 + }, + { + "epoch": 1.74, + "learning_rate": 2.5872486908118797e-08, + "loss": 0.6863, + "step": 47599 + }, + { + "epoch": 1.74, + "learning_rate": 2.5868824843446734e-08, + "loss": 0.7015, + "step": 47600 + }, + { + "epoch": 1.74, + "learning_rate": 2.5865162778774674e-08, + "loss": 0.695, + "step": 47601 + }, + { + "epoch": 1.74, + "learning_rate": 2.5861500714102607e-08, + "loss": 0.686, + "step": 47602 + }, + { + "epoch": 1.74, + "learning_rate": 2.5857838649430547e-08, + "loss": 0.684, + "step": 47603 + }, + { + "epoch": 1.74, + "learning_rate": 2.5854176584758487e-08, + "loss": 0.6977, + "step": 47604 + }, + { + "epoch": 1.74, + "learning_rate": 2.5850514520086424e-08, + "loss": 0.6952, + "step": 47605 + }, + { + "epoch": 1.74, + "learning_rate": 2.5846852455414364e-08, + "loss": 0.6941, + "step": 47606 + }, + { + "epoch": 1.74, + "learning_rate": 2.5843190390742297e-08, + "loss": 0.6892, + "step": 47607 + }, + { + "epoch": 1.74, + "learning_rate": 2.5839528326070237e-08, + "loss": 0.699, + "step": 47608 + }, + { + "epoch": 1.74, + "learning_rate": 2.5835866261398177e-08, + "loss": 0.6931, + "step": 47609 + }, + { + "epoch": 1.74, + "learning_rate": 2.5832204196726117e-08, + "loss": 0.6819, + "step": 47610 + }, + { + "epoch": 1.74, + "learning_rate": 2.582854213205405e-08, + "loss": 0.7019, + "step": 47611 + }, + { + "epoch": 1.74, + "learning_rate": 2.5824880067381987e-08, + "loss": 0.7043, + "step": 47612 + }, + { + "epoch": 1.74, + "learning_rate": 2.5821218002709926e-08, + "loss": 0.6805, + "step": 47613 + }, + { + "epoch": 1.74, + "learning_rate": 2.5817555938037866e-08, + "loss": 0.7088, + "step": 47614 + }, + { + "epoch": 1.74, + "learning_rate": 2.5813893873365806e-08, + "loss": 0.6864, + "step": 47615 + }, + { + "epoch": 1.74, + "learning_rate": 2.581023180869374e-08, + "loss": 0.6955, + "step": 47616 + }, + { + "epoch": 1.74, + "learning_rate": 2.5806569744021676e-08, + "loss": 0.6962, + "step": 47617 + }, + { + "epoch": 1.74, + "learning_rate": 2.5802907679349616e-08, + "loss": 0.6918, + "step": 47618 + }, + { + "epoch": 1.74, + "learning_rate": 2.5799245614677556e-08, + "loss": 0.7028, + "step": 47619 + }, + { + "epoch": 1.74, + "learning_rate": 2.579558355000549e-08, + "loss": 0.6954, + "step": 47620 + }, + { + "epoch": 1.74, + "learning_rate": 2.579192148533343e-08, + "loss": 0.6908, + "step": 47621 + }, + { + "epoch": 1.74, + "learning_rate": 2.578825942066137e-08, + "loss": 0.6747, + "step": 47622 + }, + { + "epoch": 1.74, + "learning_rate": 2.5784597355989306e-08, + "loss": 0.7072, + "step": 47623 + }, + { + "epoch": 1.74, + "learning_rate": 2.5780935291317246e-08, + "loss": 0.7044, + "step": 47624 + }, + { + "epoch": 1.74, + "learning_rate": 2.577727322664518e-08, + "loss": 0.707, + "step": 47625 + }, + { + "epoch": 1.74, + "learning_rate": 2.577361116197312e-08, + "loss": 0.6954, + "step": 47626 + }, + { + "epoch": 1.74, + "learning_rate": 2.576994909730106e-08, + "loss": 0.6909, + "step": 47627 + }, + { + "epoch": 1.74, + "learning_rate": 2.5766287032628996e-08, + "loss": 0.6937, + "step": 47628 + }, + { + "epoch": 1.74, + "learning_rate": 2.5762624967956932e-08, + "loss": 0.6974, + "step": 47629 + }, + { + "epoch": 1.74, + "learning_rate": 2.575896290328487e-08, + "loss": 0.7046, + "step": 47630 + }, + { + "epoch": 1.74, + "learning_rate": 2.575530083861281e-08, + "loss": 0.682, + "step": 47631 + }, + { + "epoch": 1.74, + "learning_rate": 2.575163877394075e-08, + "loss": 0.691, + "step": 47632 + }, + { + "epoch": 1.74, + "learning_rate": 2.574797670926869e-08, + "loss": 0.6807, + "step": 47633 + }, + { + "epoch": 1.74, + "learning_rate": 2.5744314644596622e-08, + "loss": 0.6898, + "step": 47634 + }, + { + "epoch": 1.74, + "learning_rate": 2.574065257992456e-08, + "loss": 0.6874, + "step": 47635 + }, + { + "epoch": 1.74, + "learning_rate": 2.57369905152525e-08, + "loss": 0.6956, + "step": 47636 + }, + { + "epoch": 1.74, + "learning_rate": 2.573332845058044e-08, + "loss": 0.688, + "step": 47637 + }, + { + "epoch": 1.74, + "learning_rate": 2.5729666385908372e-08, + "loss": 0.6925, + "step": 47638 + }, + { + "epoch": 1.74, + "learning_rate": 2.5726004321236312e-08, + "loss": 0.6977, + "step": 47639 + }, + { + "epoch": 1.74, + "learning_rate": 2.572234225656425e-08, + "loss": 0.7026, + "step": 47640 + }, + { + "epoch": 1.74, + "learning_rate": 2.5718680191892188e-08, + "loss": 0.6906, + "step": 47641 + }, + { + "epoch": 1.74, + "learning_rate": 2.5715018127220128e-08, + "loss": 0.6986, + "step": 47642 + }, + { + "epoch": 1.74, + "learning_rate": 2.571135606254806e-08, + "loss": 0.6912, + "step": 47643 + }, + { + "epoch": 1.74, + "learning_rate": 2.5707693997876e-08, + "loss": 0.6962, + "step": 47644 + }, + { + "epoch": 1.74, + "learning_rate": 2.570403193320394e-08, + "loss": 0.6983, + "step": 47645 + }, + { + "epoch": 1.74, + "learning_rate": 2.5700369868531878e-08, + "loss": 0.6871, + "step": 47646 + }, + { + "epoch": 1.74, + "learning_rate": 2.569670780385981e-08, + "loss": 0.696, + "step": 47647 + }, + { + "epoch": 1.74, + "learning_rate": 2.569304573918775e-08, + "loss": 0.6953, + "step": 47648 + }, + { + "epoch": 1.74, + "learning_rate": 2.568938367451569e-08, + "loss": 0.687, + "step": 47649 + }, + { + "epoch": 1.74, + "learning_rate": 2.568572160984363e-08, + "loss": 0.6944, + "step": 47650 + }, + { + "epoch": 1.74, + "learning_rate": 2.5682059545171564e-08, + "loss": 0.7, + "step": 47651 + }, + { + "epoch": 1.74, + "learning_rate": 2.5678397480499504e-08, + "loss": 0.6863, + "step": 47652 + }, + { + "epoch": 1.74, + "learning_rate": 2.567473541582744e-08, + "loss": 0.6958, + "step": 47653 + }, + { + "epoch": 1.74, + "learning_rate": 2.567107335115538e-08, + "loss": 0.6877, + "step": 47654 + }, + { + "epoch": 1.74, + "learning_rate": 2.566741128648332e-08, + "loss": 0.6951, + "step": 47655 + }, + { + "epoch": 1.74, + "learning_rate": 2.5663749221811254e-08, + "loss": 0.689, + "step": 47656 + }, + { + "epoch": 1.74, + "learning_rate": 2.5660087157139194e-08, + "loss": 0.6832, + "step": 47657 + }, + { + "epoch": 1.74, + "learning_rate": 2.565642509246713e-08, + "loss": 0.6931, + "step": 47658 + }, + { + "epoch": 1.74, + "learning_rate": 2.565276302779507e-08, + "loss": 0.7009, + "step": 47659 + }, + { + "epoch": 1.74, + "learning_rate": 2.5649100963123004e-08, + "loss": 0.6937, + "step": 47660 + }, + { + "epoch": 1.74, + "learning_rate": 2.5645438898450944e-08, + "loss": 0.6829, + "step": 47661 + }, + { + "epoch": 1.74, + "learning_rate": 2.5641776833778884e-08, + "loss": 0.6885, + "step": 47662 + }, + { + "epoch": 1.74, + "learning_rate": 2.5638114769106824e-08, + "loss": 0.7074, + "step": 47663 + }, + { + "epoch": 1.74, + "learning_rate": 2.563445270443476e-08, + "loss": 0.695, + "step": 47664 + }, + { + "epoch": 1.74, + "learning_rate": 2.5630790639762694e-08, + "loss": 0.6922, + "step": 47665 + }, + { + "epoch": 1.74, + "learning_rate": 2.5627128575090634e-08, + "loss": 0.6914, + "step": 47666 + }, + { + "epoch": 1.74, + "learning_rate": 2.5623466510418573e-08, + "loss": 0.6941, + "step": 47667 + }, + { + "epoch": 1.74, + "learning_rate": 2.5619804445746513e-08, + "loss": 0.7009, + "step": 47668 + }, + { + "epoch": 1.74, + "learning_rate": 2.5616142381074447e-08, + "loss": 0.6912, + "step": 47669 + }, + { + "epoch": 1.74, + "learning_rate": 2.5612480316402387e-08, + "loss": 0.6958, + "step": 47670 + }, + { + "epoch": 1.74, + "learning_rate": 2.5608818251730323e-08, + "loss": 0.6934, + "step": 47671 + }, + { + "epoch": 1.74, + "learning_rate": 2.5605156187058263e-08, + "loss": 0.6838, + "step": 47672 + }, + { + "epoch": 1.74, + "learning_rate": 2.5601494122386203e-08, + "loss": 0.6902, + "step": 47673 + }, + { + "epoch": 1.74, + "learning_rate": 2.5597832057714136e-08, + "loss": 0.6897, + "step": 47674 + }, + { + "epoch": 1.74, + "learning_rate": 2.5594169993042076e-08, + "loss": 0.6712, + "step": 47675 + }, + { + "epoch": 1.74, + "learning_rate": 2.5590507928370013e-08, + "loss": 0.6913, + "step": 47676 + }, + { + "epoch": 1.74, + "learning_rate": 2.5586845863697953e-08, + "loss": 0.7034, + "step": 47677 + }, + { + "epoch": 1.74, + "learning_rate": 2.5583183799025886e-08, + "loss": 0.7043, + "step": 47678 + }, + { + "epoch": 1.74, + "learning_rate": 2.5579521734353826e-08, + "loss": 0.6821, + "step": 47679 + }, + { + "epoch": 1.74, + "learning_rate": 2.5575859669681766e-08, + "loss": 0.6887, + "step": 47680 + }, + { + "epoch": 1.74, + "learning_rate": 2.5572197605009706e-08, + "loss": 0.6972, + "step": 47681 + }, + { + "epoch": 1.74, + "learning_rate": 2.5568535540337643e-08, + "loss": 0.6984, + "step": 47682 + }, + { + "epoch": 1.74, + "learning_rate": 2.5564873475665576e-08, + "loss": 0.6974, + "step": 47683 + }, + { + "epoch": 1.74, + "learning_rate": 2.5561211410993516e-08, + "loss": 0.6932, + "step": 47684 + }, + { + "epoch": 1.74, + "learning_rate": 2.5557549346321456e-08, + "loss": 0.6978, + "step": 47685 + }, + { + "epoch": 1.74, + "learning_rate": 2.5553887281649396e-08, + "loss": 0.6882, + "step": 47686 + }, + { + "epoch": 1.74, + "learning_rate": 2.555022521697733e-08, + "loss": 0.6882, + "step": 47687 + }, + { + "epoch": 1.74, + "learning_rate": 2.5546563152305266e-08, + "loss": 0.6991, + "step": 47688 + }, + { + "epoch": 1.74, + "learning_rate": 2.5542901087633206e-08, + "loss": 0.6988, + "step": 47689 + }, + { + "epoch": 1.74, + "learning_rate": 2.5539239022961145e-08, + "loss": 0.6879, + "step": 47690 + }, + { + "epoch": 1.74, + "learning_rate": 2.5535576958289085e-08, + "loss": 0.6972, + "step": 47691 + }, + { + "epoch": 1.74, + "learning_rate": 2.553191489361702e-08, + "loss": 0.6776, + "step": 47692 + }, + { + "epoch": 1.74, + "learning_rate": 2.552825282894496e-08, + "loss": 0.6854, + "step": 47693 + }, + { + "epoch": 1.74, + "learning_rate": 2.5524590764272895e-08, + "loss": 0.7059, + "step": 47694 + }, + { + "epoch": 1.75, + "learning_rate": 2.5520928699600835e-08, + "loss": 0.6919, + "step": 47695 + }, + { + "epoch": 1.75, + "learning_rate": 2.551726663492877e-08, + "loss": 0.6944, + "step": 47696 + }, + { + "epoch": 1.75, + "learning_rate": 2.551360457025671e-08, + "loss": 0.6997, + "step": 47697 + }, + { + "epoch": 1.75, + "learning_rate": 2.550994250558465e-08, + "loss": 0.6904, + "step": 47698 + }, + { + "epoch": 1.75, + "learning_rate": 2.5506280440912585e-08, + "loss": 0.6871, + "step": 47699 + }, + { + "epoch": 1.75, + "learning_rate": 2.5502618376240525e-08, + "loss": 0.6961, + "step": 47700 + }, + { + "epoch": 1.75, + "learning_rate": 2.5498956311568458e-08, + "loss": 0.683, + "step": 47701 + }, + { + "epoch": 1.75, + "learning_rate": 2.5495294246896398e-08, + "loss": 0.7038, + "step": 47702 + }, + { + "epoch": 1.75, + "learning_rate": 2.5491632182224338e-08, + "loss": 0.6896, + "step": 47703 + }, + { + "epoch": 1.75, + "learning_rate": 2.5487970117552278e-08, + "loss": 0.6926, + "step": 47704 + }, + { + "epoch": 1.75, + "learning_rate": 2.548430805288021e-08, + "loss": 0.7061, + "step": 47705 + }, + { + "epoch": 1.75, + "learning_rate": 2.5480645988208148e-08, + "loss": 0.7056, + "step": 47706 + }, + { + "epoch": 1.75, + "learning_rate": 2.5476983923536088e-08, + "loss": 0.6859, + "step": 47707 + }, + { + "epoch": 1.75, + "learning_rate": 2.5473321858864028e-08, + "loss": 0.6925, + "step": 47708 + }, + { + "epoch": 1.75, + "learning_rate": 2.5469659794191968e-08, + "loss": 0.6987, + "step": 47709 + }, + { + "epoch": 1.75, + "learning_rate": 2.54659977295199e-08, + "loss": 0.696, + "step": 47710 + }, + { + "epoch": 1.75, + "learning_rate": 2.546233566484784e-08, + "loss": 0.6943, + "step": 47711 + }, + { + "epoch": 1.75, + "learning_rate": 2.5458673600175778e-08, + "loss": 0.6853, + "step": 47712 + }, + { + "epoch": 1.75, + "learning_rate": 2.5455011535503718e-08, + "loss": 0.6884, + "step": 47713 + }, + { + "epoch": 1.75, + "learning_rate": 2.545134947083165e-08, + "loss": 0.6979, + "step": 47714 + }, + { + "epoch": 1.75, + "learning_rate": 2.544768740615959e-08, + "loss": 0.6972, + "step": 47715 + }, + { + "epoch": 1.75, + "learning_rate": 2.544402534148753e-08, + "loss": 0.6868, + "step": 47716 + }, + { + "epoch": 1.75, + "learning_rate": 2.5440363276815467e-08, + "loss": 0.6997, + "step": 47717 + }, + { + "epoch": 1.75, + "learning_rate": 2.5436701212143407e-08, + "loss": 0.6856, + "step": 47718 + }, + { + "epoch": 1.75, + "learning_rate": 2.543303914747134e-08, + "loss": 0.6917, + "step": 47719 + }, + { + "epoch": 1.75, + "learning_rate": 2.542937708279928e-08, + "loss": 0.6883, + "step": 47720 + }, + { + "epoch": 1.75, + "learning_rate": 2.542571501812722e-08, + "loss": 0.6835, + "step": 47721 + }, + { + "epoch": 1.75, + "learning_rate": 2.5422052953455157e-08, + "loss": 0.6826, + "step": 47722 + }, + { + "epoch": 1.75, + "learning_rate": 2.5418390888783094e-08, + "loss": 0.6732, + "step": 47723 + }, + { + "epoch": 1.75, + "learning_rate": 2.541472882411103e-08, + "loss": 0.7029, + "step": 47724 + }, + { + "epoch": 1.75, + "learning_rate": 2.541106675943897e-08, + "loss": 0.6966, + "step": 47725 + }, + { + "epoch": 1.75, + "learning_rate": 2.540740469476691e-08, + "loss": 0.696, + "step": 47726 + }, + { + "epoch": 1.75, + "learning_rate": 2.540374263009485e-08, + "loss": 0.6915, + "step": 47727 + }, + { + "epoch": 1.75, + "learning_rate": 2.5400080565422783e-08, + "loss": 0.6892, + "step": 47728 + }, + { + "epoch": 1.75, + "learning_rate": 2.539641850075072e-08, + "loss": 0.6932, + "step": 47729 + }, + { + "epoch": 1.75, + "learning_rate": 2.539275643607866e-08, + "loss": 0.6879, + "step": 47730 + }, + { + "epoch": 1.75, + "learning_rate": 2.53890943714066e-08, + "loss": 0.6897, + "step": 47731 + }, + { + "epoch": 1.75, + "learning_rate": 2.5385432306734533e-08, + "loss": 0.6973, + "step": 47732 + }, + { + "epoch": 1.75, + "learning_rate": 2.5381770242062473e-08, + "loss": 0.6845, + "step": 47733 + }, + { + "epoch": 1.75, + "learning_rate": 2.5378108177390413e-08, + "loss": 0.6846, + "step": 47734 + }, + { + "epoch": 1.75, + "learning_rate": 2.537444611271835e-08, + "loss": 0.6868, + "step": 47735 + }, + { + "epoch": 1.75, + "learning_rate": 2.537078404804629e-08, + "loss": 0.6976, + "step": 47736 + }, + { + "epoch": 1.75, + "learning_rate": 2.5367121983374223e-08, + "loss": 0.7063, + "step": 47737 + }, + { + "epoch": 1.75, + "learning_rate": 2.5363459918702163e-08, + "loss": 0.6947, + "step": 47738 + }, + { + "epoch": 1.75, + "learning_rate": 2.5359797854030103e-08, + "loss": 0.6972, + "step": 47739 + }, + { + "epoch": 1.75, + "learning_rate": 2.535613578935804e-08, + "loss": 0.7087, + "step": 47740 + }, + { + "epoch": 1.75, + "learning_rate": 2.5352473724685976e-08, + "loss": 0.696, + "step": 47741 + }, + { + "epoch": 1.75, + "learning_rate": 2.5348811660013913e-08, + "loss": 0.6886, + "step": 47742 + }, + { + "epoch": 1.75, + "learning_rate": 2.5345149595341853e-08, + "loss": 0.6885, + "step": 47743 + }, + { + "epoch": 1.75, + "learning_rate": 2.5341487530669792e-08, + "loss": 0.6976, + "step": 47744 + }, + { + "epoch": 1.75, + "learning_rate": 2.5337825465997732e-08, + "loss": 0.705, + "step": 47745 + }, + { + "epoch": 1.75, + "learning_rate": 2.5334163401325666e-08, + "loss": 0.6958, + "step": 47746 + }, + { + "epoch": 1.75, + "learning_rate": 2.5330501336653602e-08, + "loss": 0.6952, + "step": 47747 + }, + { + "epoch": 1.75, + "learning_rate": 2.5326839271981542e-08, + "loss": 0.6829, + "step": 47748 + }, + { + "epoch": 1.75, + "learning_rate": 2.5323177207309482e-08, + "loss": 0.693, + "step": 47749 + }, + { + "epoch": 1.75, + "learning_rate": 2.5319515142637415e-08, + "loss": 0.6982, + "step": 47750 + }, + { + "epoch": 1.75, + "learning_rate": 2.5315853077965355e-08, + "loss": 0.6871, + "step": 47751 + }, + { + "epoch": 1.75, + "learning_rate": 2.5312191013293292e-08, + "loss": 0.7073, + "step": 47752 + }, + { + "epoch": 1.75, + "learning_rate": 2.5308528948621232e-08, + "loss": 0.6876, + "step": 47753 + }, + { + "epoch": 1.75, + "learning_rate": 2.5304866883949172e-08, + "loss": 0.7026, + "step": 47754 + }, + { + "epoch": 1.75, + "learning_rate": 2.5301204819277105e-08, + "loss": 0.6988, + "step": 47755 + }, + { + "epoch": 1.75, + "learning_rate": 2.5297542754605045e-08, + "loss": 0.6904, + "step": 47756 + }, + { + "epoch": 1.75, + "learning_rate": 2.5293880689932985e-08, + "loss": 0.6943, + "step": 47757 + }, + { + "epoch": 1.75, + "learning_rate": 2.529021862526092e-08, + "loss": 0.6924, + "step": 47758 + }, + { + "epoch": 1.75, + "learning_rate": 2.5286556560588855e-08, + "loss": 0.6831, + "step": 47759 + }, + { + "epoch": 1.75, + "learning_rate": 2.5282894495916795e-08, + "loss": 0.6878, + "step": 47760 + }, + { + "epoch": 1.75, + "learning_rate": 2.5279232431244735e-08, + "loss": 0.6989, + "step": 47761 + }, + { + "epoch": 1.75, + "learning_rate": 2.5275570366572675e-08, + "loss": 0.6939, + "step": 47762 + }, + { + "epoch": 1.75, + "learning_rate": 2.527190830190061e-08, + "loss": 0.686, + "step": 47763 + }, + { + "epoch": 1.75, + "learning_rate": 2.5268246237228548e-08, + "loss": 0.6938, + "step": 47764 + }, + { + "epoch": 1.75, + "learning_rate": 2.5264584172556485e-08, + "loss": 0.7018, + "step": 47765 + }, + { + "epoch": 1.75, + "learning_rate": 2.5260922107884425e-08, + "loss": 0.6809, + "step": 47766 + }, + { + "epoch": 1.75, + "learning_rate": 2.5257260043212364e-08, + "loss": 0.6818, + "step": 47767 + }, + { + "epoch": 1.75, + "learning_rate": 2.5253597978540298e-08, + "loss": 0.6843, + "step": 47768 + }, + { + "epoch": 1.75, + "learning_rate": 2.5249935913868238e-08, + "loss": 0.6921, + "step": 47769 + }, + { + "epoch": 1.75, + "learning_rate": 2.5246273849196174e-08, + "loss": 0.7042, + "step": 47770 + }, + { + "epoch": 1.75, + "learning_rate": 2.5242611784524114e-08, + "loss": 0.7055, + "step": 47771 + }, + { + "epoch": 1.75, + "learning_rate": 2.5238949719852054e-08, + "loss": 0.6765, + "step": 47772 + }, + { + "epoch": 1.75, + "learning_rate": 2.5235287655179987e-08, + "loss": 0.6931, + "step": 47773 + }, + { + "epoch": 1.75, + "learning_rate": 2.5231625590507927e-08, + "loss": 0.6909, + "step": 47774 + }, + { + "epoch": 1.75, + "learning_rate": 2.5227963525835867e-08, + "loss": 0.696, + "step": 47775 + }, + { + "epoch": 1.75, + "learning_rate": 2.5224301461163804e-08, + "loss": 0.6871, + "step": 47776 + }, + { + "epoch": 1.75, + "learning_rate": 2.5220639396491737e-08, + "loss": 0.6957, + "step": 47777 + }, + { + "epoch": 1.75, + "learning_rate": 2.5216977331819677e-08, + "loss": 0.7013, + "step": 47778 + }, + { + "epoch": 1.75, + "learning_rate": 2.5213315267147617e-08, + "loss": 0.6975, + "step": 47779 + }, + { + "epoch": 1.75, + "learning_rate": 2.5209653202475557e-08, + "loss": 0.6876, + "step": 47780 + }, + { + "epoch": 1.75, + "learning_rate": 2.5205991137803494e-08, + "loss": 0.694, + "step": 47781 + }, + { + "epoch": 1.75, + "learning_rate": 2.5202329073131427e-08, + "loss": 0.6939, + "step": 47782 + }, + { + "epoch": 1.75, + "learning_rate": 2.5198667008459367e-08, + "loss": 0.6891, + "step": 47783 + }, + { + "epoch": 1.75, + "learning_rate": 2.5195004943787307e-08, + "loss": 0.6902, + "step": 47784 + }, + { + "epoch": 1.75, + "learning_rate": 2.5191342879115247e-08, + "loss": 0.6855, + "step": 47785 + }, + { + "epoch": 1.75, + "learning_rate": 2.518768081444318e-08, + "loss": 0.6853, + "step": 47786 + }, + { + "epoch": 1.75, + "learning_rate": 2.518401874977112e-08, + "loss": 0.693, + "step": 47787 + }, + { + "epoch": 1.75, + "learning_rate": 2.5180356685099057e-08, + "loss": 0.6899, + "step": 47788 + }, + { + "epoch": 1.75, + "learning_rate": 2.5176694620426997e-08, + "loss": 0.6983, + "step": 47789 + }, + { + "epoch": 1.75, + "learning_rate": 2.5173032555754937e-08, + "loss": 0.7024, + "step": 47790 + }, + { + "epoch": 1.75, + "learning_rate": 2.516937049108287e-08, + "loss": 0.6964, + "step": 47791 + }, + { + "epoch": 1.75, + "learning_rate": 2.516570842641081e-08, + "loss": 0.6845, + "step": 47792 + }, + { + "epoch": 1.75, + "learning_rate": 2.5162046361738746e-08, + "loss": 0.7064, + "step": 47793 + }, + { + "epoch": 1.75, + "learning_rate": 2.5158384297066686e-08, + "loss": 0.6867, + "step": 47794 + }, + { + "epoch": 1.75, + "learning_rate": 2.515472223239462e-08, + "loss": 0.6906, + "step": 47795 + }, + { + "epoch": 1.75, + "learning_rate": 2.515106016772256e-08, + "loss": 0.6989, + "step": 47796 + }, + { + "epoch": 1.75, + "learning_rate": 2.51473981030505e-08, + "loss": 0.7077, + "step": 47797 + }, + { + "epoch": 1.75, + "learning_rate": 2.514373603837844e-08, + "loss": 0.7053, + "step": 47798 + }, + { + "epoch": 1.75, + "learning_rate": 2.5140073973706376e-08, + "loss": 0.692, + "step": 47799 + }, + { + "epoch": 1.75, + "learning_rate": 2.513641190903431e-08, + "loss": 0.6839, + "step": 47800 + }, + { + "epoch": 1.75, + "learning_rate": 2.513274984436225e-08, + "loss": 0.6944, + "step": 47801 + }, + { + "epoch": 1.75, + "learning_rate": 2.512908777969019e-08, + "loss": 0.6995, + "step": 47802 + }, + { + "epoch": 1.75, + "learning_rate": 2.512542571501813e-08, + "loss": 0.6932, + "step": 47803 + }, + { + "epoch": 1.75, + "learning_rate": 2.5121763650346062e-08, + "loss": 0.6956, + "step": 47804 + }, + { + "epoch": 1.75, + "learning_rate": 2.5118101585674002e-08, + "loss": 0.7, + "step": 47805 + }, + { + "epoch": 1.75, + "learning_rate": 2.511443952100194e-08, + "loss": 0.6968, + "step": 47806 + }, + { + "epoch": 1.75, + "learning_rate": 2.511077745632988e-08, + "loss": 0.69, + "step": 47807 + }, + { + "epoch": 1.75, + "learning_rate": 2.510711539165782e-08, + "loss": 0.6965, + "step": 47808 + }, + { + "epoch": 1.75, + "learning_rate": 2.5103453326985752e-08, + "loss": 0.68, + "step": 47809 + }, + { + "epoch": 1.75, + "learning_rate": 2.5099791262313692e-08, + "loss": 0.6822, + "step": 47810 + }, + { + "epoch": 1.75, + "learning_rate": 2.509612919764163e-08, + "loss": 0.689, + "step": 47811 + }, + { + "epoch": 1.75, + "learning_rate": 2.509246713296957e-08, + "loss": 0.6993, + "step": 47812 + }, + { + "epoch": 1.75, + "learning_rate": 2.5088805068297502e-08, + "loss": 0.6913, + "step": 47813 + }, + { + "epoch": 1.75, + "learning_rate": 2.5085143003625442e-08, + "loss": 0.6947, + "step": 47814 + }, + { + "epoch": 1.75, + "learning_rate": 2.5081480938953382e-08, + "loss": 0.6825, + "step": 47815 + }, + { + "epoch": 1.75, + "learning_rate": 2.5077818874281322e-08, + "loss": 0.6964, + "step": 47816 + }, + { + "epoch": 1.75, + "learning_rate": 2.5074156809609258e-08, + "loss": 0.6917, + "step": 47817 + }, + { + "epoch": 1.75, + "learning_rate": 2.507049474493719e-08, + "loss": 0.6995, + "step": 47818 + }, + { + "epoch": 1.75, + "learning_rate": 2.506683268026513e-08, + "loss": 0.7031, + "step": 47819 + }, + { + "epoch": 1.75, + "learning_rate": 2.506317061559307e-08, + "loss": 0.7007, + "step": 47820 + }, + { + "epoch": 1.75, + "learning_rate": 2.505950855092101e-08, + "loss": 0.6916, + "step": 47821 + }, + { + "epoch": 1.75, + "learning_rate": 2.5055846486248945e-08, + "loss": 0.711, + "step": 47822 + }, + { + "epoch": 1.75, + "learning_rate": 2.505218442157688e-08, + "loss": 0.6842, + "step": 47823 + }, + { + "epoch": 1.75, + "learning_rate": 2.504852235690482e-08, + "loss": 0.6867, + "step": 47824 + }, + { + "epoch": 1.75, + "learning_rate": 2.504486029223276e-08, + "loss": 0.6945, + "step": 47825 + }, + { + "epoch": 1.75, + "learning_rate": 2.50411982275607e-08, + "loss": 0.7048, + "step": 47826 + }, + { + "epoch": 1.75, + "learning_rate": 2.5037536162888634e-08, + "loss": 0.6872, + "step": 47827 + }, + { + "epoch": 1.75, + "learning_rate": 2.5033874098216574e-08, + "loss": 0.6771, + "step": 47828 + }, + { + "epoch": 1.75, + "learning_rate": 2.503021203354451e-08, + "loss": 0.6943, + "step": 47829 + }, + { + "epoch": 1.75, + "learning_rate": 2.502654996887245e-08, + "loss": 0.6895, + "step": 47830 + }, + { + "epoch": 1.75, + "learning_rate": 2.5022887904200384e-08, + "loss": 0.6866, + "step": 47831 + }, + { + "epoch": 1.75, + "learning_rate": 2.5019225839528324e-08, + "loss": 0.6966, + "step": 47832 + }, + { + "epoch": 1.75, + "learning_rate": 2.5015563774856264e-08, + "loss": 0.7071, + "step": 47833 + }, + { + "epoch": 1.75, + "learning_rate": 2.50119017101842e-08, + "loss": 0.6877, + "step": 47834 + }, + { + "epoch": 1.75, + "learning_rate": 2.500823964551214e-08, + "loss": 0.6901, + "step": 47835 + }, + { + "epoch": 1.75, + "learning_rate": 2.5004577580840074e-08, + "loss": 0.6961, + "step": 47836 + }, + { + "epoch": 1.75, + "learning_rate": 2.5000915516168014e-08, + "loss": 0.6981, + "step": 47837 + }, + { + "epoch": 1.75, + "learning_rate": 2.4997253451495954e-08, + "loss": 0.6884, + "step": 47838 + }, + { + "epoch": 1.75, + "learning_rate": 2.499359138682389e-08, + "loss": 0.6857, + "step": 47839 + }, + { + "epoch": 1.75, + "learning_rate": 2.4989929322151827e-08, + "loss": 0.6856, + "step": 47840 + }, + { + "epoch": 1.75, + "learning_rate": 2.4986267257479764e-08, + "loss": 0.6864, + "step": 47841 + }, + { + "epoch": 1.75, + "learning_rate": 2.4982605192807704e-08, + "loss": 0.7031, + "step": 47842 + }, + { + "epoch": 1.75, + "learning_rate": 2.497894312813564e-08, + "loss": 0.6933, + "step": 47843 + }, + { + "epoch": 1.75, + "learning_rate": 2.497528106346358e-08, + "loss": 0.6924, + "step": 47844 + }, + { + "epoch": 1.75, + "learning_rate": 2.497161899879152e-08, + "loss": 0.6928, + "step": 47845 + }, + { + "epoch": 1.75, + "learning_rate": 2.4967956934119457e-08, + "loss": 0.6971, + "step": 47846 + }, + { + "epoch": 1.75, + "learning_rate": 2.4964294869447393e-08, + "loss": 0.6978, + "step": 47847 + }, + { + "epoch": 1.75, + "learning_rate": 2.496063280477533e-08, + "loss": 0.687, + "step": 47848 + }, + { + "epoch": 1.75, + "learning_rate": 2.495697074010327e-08, + "loss": 0.711, + "step": 47849 + }, + { + "epoch": 1.75, + "learning_rate": 2.4953308675431206e-08, + "loss": 0.685, + "step": 47850 + }, + { + "epoch": 1.75, + "learning_rate": 2.4949646610759146e-08, + "loss": 0.6804, + "step": 47851 + }, + { + "epoch": 1.75, + "learning_rate": 2.4945984546087083e-08, + "loss": 0.7077, + "step": 47852 + }, + { + "epoch": 1.75, + "learning_rate": 2.494232248141502e-08, + "loss": 0.6888, + "step": 47853 + }, + { + "epoch": 1.75, + "learning_rate": 2.493866041674296e-08, + "loss": 0.6786, + "step": 47854 + }, + { + "epoch": 1.75, + "learning_rate": 2.4934998352070896e-08, + "loss": 0.7034, + "step": 47855 + }, + { + "epoch": 1.75, + "learning_rate": 2.4931336287398836e-08, + "loss": 0.6981, + "step": 47856 + }, + { + "epoch": 1.75, + "learning_rate": 2.4927674222726773e-08, + "loss": 0.6857, + "step": 47857 + }, + { + "epoch": 1.75, + "learning_rate": 2.492401215805471e-08, + "loss": 0.6876, + "step": 47858 + }, + { + "epoch": 1.75, + "learning_rate": 2.4920350093382646e-08, + "loss": 0.6934, + "step": 47859 + }, + { + "epoch": 1.75, + "learning_rate": 2.4916688028710586e-08, + "loss": 0.6937, + "step": 47860 + }, + { + "epoch": 1.75, + "learning_rate": 2.4913025964038523e-08, + "loss": 0.7006, + "step": 47861 + }, + { + "epoch": 1.75, + "learning_rate": 2.4909363899366462e-08, + "loss": 0.7113, + "step": 47862 + }, + { + "epoch": 1.75, + "learning_rate": 2.4905701834694402e-08, + "loss": 0.686, + "step": 47863 + }, + { + "epoch": 1.75, + "learning_rate": 2.4902039770022336e-08, + "loss": 0.687, + "step": 47864 + }, + { + "epoch": 1.75, + "learning_rate": 2.4898377705350276e-08, + "loss": 0.679, + "step": 47865 + }, + { + "epoch": 1.75, + "learning_rate": 2.4894715640678212e-08, + "loss": 0.6863, + "step": 47866 + }, + { + "epoch": 1.75, + "learning_rate": 2.4891053576006152e-08, + "loss": 0.6924, + "step": 47867 + }, + { + "epoch": 1.75, + "learning_rate": 2.488739151133409e-08, + "loss": 0.6873, + "step": 47868 + }, + { + "epoch": 1.75, + "learning_rate": 2.488372944666203e-08, + "loss": 0.6879, + "step": 47869 + }, + { + "epoch": 1.75, + "learning_rate": 2.4880067381989962e-08, + "loss": 0.6892, + "step": 47870 + }, + { + "epoch": 1.75, + "learning_rate": 2.4876405317317902e-08, + "loss": 0.6865, + "step": 47871 + }, + { + "epoch": 1.75, + "learning_rate": 2.4872743252645842e-08, + "loss": 0.684, + "step": 47872 + }, + { + "epoch": 1.75, + "learning_rate": 2.486908118797378e-08, + "loss": 0.7067, + "step": 47873 + }, + { + "epoch": 1.75, + "learning_rate": 2.486541912330172e-08, + "loss": 0.6945, + "step": 47874 + }, + { + "epoch": 1.75, + "learning_rate": 2.4861757058629655e-08, + "loss": 0.6908, + "step": 47875 + }, + { + "epoch": 1.75, + "learning_rate": 2.4858094993957592e-08, + "loss": 0.6966, + "step": 47876 + }, + { + "epoch": 1.75, + "learning_rate": 2.4854432929285528e-08, + "loss": 0.6995, + "step": 47877 + }, + { + "epoch": 1.75, + "learning_rate": 2.4850770864613468e-08, + "loss": 0.6931, + "step": 47878 + }, + { + "epoch": 1.75, + "learning_rate": 2.4847108799941405e-08, + "loss": 0.6987, + "step": 47879 + }, + { + "epoch": 1.75, + "learning_rate": 2.4843446735269345e-08, + "loss": 0.6961, + "step": 47880 + }, + { + "epoch": 1.75, + "learning_rate": 2.483978467059728e-08, + "loss": 0.7027, + "step": 47881 + }, + { + "epoch": 1.75, + "learning_rate": 2.4836122605925218e-08, + "loss": 0.6889, + "step": 47882 + }, + { + "epoch": 1.75, + "learning_rate": 2.4832460541253158e-08, + "loss": 0.6903, + "step": 47883 + }, + { + "epoch": 1.75, + "learning_rate": 2.4828798476581095e-08, + "loss": 0.7048, + "step": 47884 + }, + { + "epoch": 1.75, + "learning_rate": 2.4825136411909035e-08, + "loss": 0.6941, + "step": 47885 + }, + { + "epoch": 1.75, + "learning_rate": 2.482147434723697e-08, + "loss": 0.7006, + "step": 47886 + }, + { + "epoch": 1.75, + "learning_rate": 2.481781228256491e-08, + "loss": 0.6919, + "step": 47887 + }, + { + "epoch": 1.75, + "learning_rate": 2.4814150217892844e-08, + "loss": 0.6883, + "step": 47888 + }, + { + "epoch": 1.75, + "learning_rate": 2.4810488153220784e-08, + "loss": 0.6862, + "step": 47889 + }, + { + "epoch": 1.75, + "learning_rate": 2.480682608854872e-08, + "loss": 0.6965, + "step": 47890 + }, + { + "epoch": 1.75, + "learning_rate": 2.480316402387666e-08, + "loss": 0.6965, + "step": 47891 + }, + { + "epoch": 1.75, + "learning_rate": 2.47995019592046e-08, + "loss": 0.7064, + "step": 47892 + }, + { + "epoch": 1.75, + "learning_rate": 2.4795839894532537e-08, + "loss": 0.6878, + "step": 47893 + }, + { + "epoch": 1.75, + "learning_rate": 2.4792177829860474e-08, + "loss": 0.6977, + "step": 47894 + }, + { + "epoch": 1.75, + "learning_rate": 2.478851576518841e-08, + "loss": 0.7011, + "step": 47895 + }, + { + "epoch": 1.75, + "learning_rate": 2.478485370051635e-08, + "loss": 0.6993, + "step": 47896 + }, + { + "epoch": 1.75, + "learning_rate": 2.4781191635844287e-08, + "loss": 0.6898, + "step": 47897 + }, + { + "epoch": 1.75, + "learning_rate": 2.4777529571172227e-08, + "loss": 0.7084, + "step": 47898 + }, + { + "epoch": 1.75, + "learning_rate": 2.4773867506500164e-08, + "loss": 0.6987, + "step": 47899 + }, + { + "epoch": 1.75, + "learning_rate": 2.47702054418281e-08, + "loss": 0.6842, + "step": 47900 + }, + { + "epoch": 1.75, + "learning_rate": 2.476654337715604e-08, + "loss": 0.6887, + "step": 47901 + }, + { + "epoch": 1.75, + "learning_rate": 2.4762881312483977e-08, + "loss": 0.6926, + "step": 47902 + }, + { + "epoch": 1.75, + "learning_rate": 2.4759219247811917e-08, + "loss": 0.7049, + "step": 47903 + }, + { + "epoch": 1.75, + "learning_rate": 2.4755557183139853e-08, + "loss": 0.6953, + "step": 47904 + }, + { + "epoch": 1.75, + "learning_rate": 2.475189511846779e-08, + "loss": 0.6911, + "step": 47905 + }, + { + "epoch": 1.75, + "learning_rate": 2.4748233053795727e-08, + "loss": 0.6877, + "step": 47906 + }, + { + "epoch": 1.75, + "learning_rate": 2.4744570989123667e-08, + "loss": 0.6929, + "step": 47907 + }, + { + "epoch": 1.75, + "learning_rate": 2.4740908924451603e-08, + "loss": 0.6956, + "step": 47908 + }, + { + "epoch": 1.75, + "learning_rate": 2.4737246859779543e-08, + "loss": 0.6909, + "step": 47909 + }, + { + "epoch": 1.75, + "learning_rate": 2.4733584795107483e-08, + "loss": 0.7031, + "step": 47910 + }, + { + "epoch": 1.75, + "learning_rate": 2.4729922730435416e-08, + "loss": 0.6864, + "step": 47911 + }, + { + "epoch": 1.75, + "learning_rate": 2.4726260665763356e-08, + "loss": 0.6904, + "step": 47912 + }, + { + "epoch": 1.75, + "learning_rate": 2.4722598601091293e-08, + "loss": 0.6919, + "step": 47913 + }, + { + "epoch": 1.75, + "learning_rate": 2.4718936536419233e-08, + "loss": 0.6914, + "step": 47914 + }, + { + "epoch": 1.75, + "learning_rate": 2.471527447174717e-08, + "loss": 0.6982, + "step": 47915 + }, + { + "epoch": 1.75, + "learning_rate": 2.471161240707511e-08, + "loss": 0.6968, + "step": 47916 + }, + { + "epoch": 1.75, + "learning_rate": 2.4707950342403046e-08, + "loss": 0.7003, + "step": 47917 + }, + { + "epoch": 1.75, + "learning_rate": 2.4704288277730983e-08, + "loss": 0.6919, + "step": 47918 + }, + { + "epoch": 1.75, + "learning_rate": 2.4700626213058923e-08, + "loss": 0.6888, + "step": 47919 + }, + { + "epoch": 1.75, + "learning_rate": 2.469696414838686e-08, + "loss": 0.694, + "step": 47920 + }, + { + "epoch": 1.75, + "learning_rate": 2.46933020837148e-08, + "loss": 0.6887, + "step": 47921 + }, + { + "epoch": 1.75, + "learning_rate": 2.4689640019042736e-08, + "loss": 0.6777, + "step": 47922 + }, + { + "epoch": 1.75, + "learning_rate": 2.4685977954370672e-08, + "loss": 0.6879, + "step": 47923 + }, + { + "epoch": 1.75, + "learning_rate": 2.468231588969861e-08, + "loss": 0.6859, + "step": 47924 + }, + { + "epoch": 1.75, + "learning_rate": 2.467865382502655e-08, + "loss": 0.7067, + "step": 47925 + }, + { + "epoch": 1.75, + "learning_rate": 2.4674991760354486e-08, + "loss": 0.6979, + "step": 47926 + }, + { + "epoch": 1.75, + "learning_rate": 2.4671329695682425e-08, + "loss": 0.696, + "step": 47927 + }, + { + "epoch": 1.75, + "learning_rate": 2.4667667631010362e-08, + "loss": 0.693, + "step": 47928 + }, + { + "epoch": 1.75, + "learning_rate": 2.46640055663383e-08, + "loss": 0.6918, + "step": 47929 + }, + { + "epoch": 1.75, + "learning_rate": 2.466034350166624e-08, + "loss": 0.6875, + "step": 47930 + }, + { + "epoch": 1.75, + "learning_rate": 2.4656681436994175e-08, + "loss": 0.6856, + "step": 47931 + }, + { + "epoch": 1.75, + "learning_rate": 2.4653019372322115e-08, + "loss": 0.6929, + "step": 47932 + }, + { + "epoch": 1.75, + "learning_rate": 2.4649357307650052e-08, + "loss": 0.7014, + "step": 47933 + }, + { + "epoch": 1.75, + "learning_rate": 2.4645695242977992e-08, + "loss": 0.7005, + "step": 47934 + }, + { + "epoch": 1.75, + "learning_rate": 2.4642033178305925e-08, + "loss": 0.6963, + "step": 47935 + }, + { + "epoch": 1.75, + "learning_rate": 2.4638371113633865e-08, + "loss": 0.7011, + "step": 47936 + }, + { + "epoch": 1.75, + "learning_rate": 2.4634709048961805e-08, + "loss": 0.6861, + "step": 47937 + }, + { + "epoch": 1.75, + "learning_rate": 2.463104698428974e-08, + "loss": 0.6943, + "step": 47938 + }, + { + "epoch": 1.75, + "learning_rate": 2.462738491961768e-08, + "loss": 0.6926, + "step": 47939 + }, + { + "epoch": 1.75, + "learning_rate": 2.4623722854945618e-08, + "loss": 0.6863, + "step": 47940 + }, + { + "epoch": 1.75, + "learning_rate": 2.4620060790273555e-08, + "loss": 0.6932, + "step": 47941 + }, + { + "epoch": 1.75, + "learning_rate": 2.461639872560149e-08, + "loss": 0.6822, + "step": 47942 + }, + { + "epoch": 1.75, + "learning_rate": 2.461273666092943e-08, + "loss": 0.7047, + "step": 47943 + }, + { + "epoch": 1.75, + "learning_rate": 2.4609074596257368e-08, + "loss": 0.6936, + "step": 47944 + }, + { + "epoch": 1.75, + "learning_rate": 2.4605412531585308e-08, + "loss": 0.703, + "step": 47945 + }, + { + "epoch": 1.75, + "learning_rate": 2.4601750466913244e-08, + "loss": 0.7071, + "step": 47946 + }, + { + "epoch": 1.75, + "learning_rate": 2.459808840224118e-08, + "loss": 0.6967, + "step": 47947 + }, + { + "epoch": 1.75, + "learning_rate": 2.459442633756912e-08, + "loss": 0.7006, + "step": 47948 + }, + { + "epoch": 1.75, + "learning_rate": 2.4590764272897058e-08, + "loss": 0.7014, + "step": 47949 + }, + { + "epoch": 1.75, + "learning_rate": 2.4587102208224998e-08, + "loss": 0.6998, + "step": 47950 + }, + { + "epoch": 1.75, + "learning_rate": 2.4583440143552934e-08, + "loss": 0.687, + "step": 47951 + }, + { + "epoch": 1.75, + "learning_rate": 2.457977807888087e-08, + "loss": 0.6817, + "step": 47952 + }, + { + "epoch": 1.75, + "learning_rate": 2.4576116014208807e-08, + "loss": 0.6926, + "step": 47953 + }, + { + "epoch": 1.75, + "learning_rate": 2.4572453949536747e-08, + "loss": 0.6972, + "step": 47954 + }, + { + "epoch": 1.75, + "learning_rate": 2.4568791884864687e-08, + "loss": 0.688, + "step": 47955 + }, + { + "epoch": 1.75, + "learning_rate": 2.4565129820192624e-08, + "loss": 0.6967, + "step": 47956 + }, + { + "epoch": 1.75, + "learning_rate": 2.4561467755520564e-08, + "loss": 0.7074, + "step": 47957 + }, + { + "epoch": 1.75, + "learning_rate": 2.4557805690848497e-08, + "loss": 0.6867, + "step": 47958 + }, + { + "epoch": 1.75, + "learning_rate": 2.4554143626176437e-08, + "loss": 0.691, + "step": 47959 + }, + { + "epoch": 1.75, + "learning_rate": 2.4550481561504374e-08, + "loss": 0.6873, + "step": 47960 + }, + { + "epoch": 1.75, + "learning_rate": 2.4546819496832314e-08, + "loss": 0.6919, + "step": 47961 + }, + { + "epoch": 1.75, + "learning_rate": 2.454315743216025e-08, + "loss": 0.6831, + "step": 47962 + }, + { + "epoch": 1.75, + "learning_rate": 2.453949536748819e-08, + "loss": 0.6979, + "step": 47963 + }, + { + "epoch": 1.75, + "learning_rate": 2.4535833302816127e-08, + "loss": 0.6983, + "step": 47964 + }, + { + "epoch": 1.75, + "learning_rate": 2.4532171238144063e-08, + "loss": 0.6957, + "step": 47965 + }, + { + "epoch": 1.75, + "learning_rate": 2.4528509173472003e-08, + "loss": 0.7081, + "step": 47966 + }, + { + "epoch": 1.75, + "learning_rate": 2.452484710879994e-08, + "loss": 0.702, + "step": 47967 + }, + { + "epoch": 1.76, + "learning_rate": 2.452118504412788e-08, + "loss": 0.6877, + "step": 47968 + }, + { + "epoch": 1.76, + "learning_rate": 2.4517522979455816e-08, + "loss": 0.6902, + "step": 47969 + }, + { + "epoch": 1.76, + "learning_rate": 2.4513860914783753e-08, + "loss": 0.7112, + "step": 47970 + }, + { + "epoch": 1.76, + "learning_rate": 2.451019885011169e-08, + "loss": 0.6877, + "step": 47971 + }, + { + "epoch": 1.76, + "learning_rate": 2.450653678543963e-08, + "loss": 0.6807, + "step": 47972 + }, + { + "epoch": 1.76, + "learning_rate": 2.450287472076757e-08, + "loss": 0.6857, + "step": 47973 + }, + { + "epoch": 1.76, + "learning_rate": 2.4499212656095506e-08, + "loss": 0.6878, + "step": 47974 + }, + { + "epoch": 1.76, + "learning_rate": 2.4495550591423446e-08, + "loss": 0.6942, + "step": 47975 + }, + { + "epoch": 1.76, + "learning_rate": 2.449188852675138e-08, + "loss": 0.6846, + "step": 47976 + }, + { + "epoch": 1.76, + "learning_rate": 2.448822646207932e-08, + "loss": 0.696, + "step": 47977 + }, + { + "epoch": 1.76, + "learning_rate": 2.4484564397407256e-08, + "loss": 0.6954, + "step": 47978 + }, + { + "epoch": 1.76, + "learning_rate": 2.4480902332735196e-08, + "loss": 0.6871, + "step": 47979 + }, + { + "epoch": 1.76, + "learning_rate": 2.4477240268063133e-08, + "loss": 0.6874, + "step": 47980 + }, + { + "epoch": 1.76, + "learning_rate": 2.4473578203391072e-08, + "loss": 0.6872, + "step": 47981 + }, + { + "epoch": 1.76, + "learning_rate": 2.4469916138719006e-08, + "loss": 0.7009, + "step": 47982 + }, + { + "epoch": 1.76, + "learning_rate": 2.4466254074046946e-08, + "loss": 0.7097, + "step": 47983 + }, + { + "epoch": 1.76, + "learning_rate": 2.4462592009374886e-08, + "loss": 0.7007, + "step": 47984 + }, + { + "epoch": 1.76, + "learning_rate": 2.4458929944702822e-08, + "loss": 0.6929, + "step": 47985 + }, + { + "epoch": 1.76, + "learning_rate": 2.4455267880030762e-08, + "loss": 0.6843, + "step": 47986 + }, + { + "epoch": 1.76, + "learning_rate": 2.44516058153587e-08, + "loss": 0.6995, + "step": 47987 + }, + { + "epoch": 1.76, + "learning_rate": 2.4447943750686635e-08, + "loss": 0.6997, + "step": 47988 + }, + { + "epoch": 1.76, + "learning_rate": 2.4444281686014572e-08, + "loss": 0.6866, + "step": 47989 + }, + { + "epoch": 1.76, + "learning_rate": 2.4440619621342512e-08, + "loss": 0.6885, + "step": 47990 + }, + { + "epoch": 1.76, + "learning_rate": 2.443695755667045e-08, + "loss": 0.6975, + "step": 47991 + }, + { + "epoch": 1.76, + "learning_rate": 2.443329549199839e-08, + "loss": 0.691, + "step": 47992 + }, + { + "epoch": 1.76, + "learning_rate": 2.4429633427326325e-08, + "loss": 0.6955, + "step": 47993 + }, + { + "epoch": 1.76, + "learning_rate": 2.4425971362654262e-08, + "loss": 0.7025, + "step": 47994 + }, + { + "epoch": 1.76, + "learning_rate": 2.4422309297982202e-08, + "loss": 0.6984, + "step": 47995 + }, + { + "epoch": 1.76, + "learning_rate": 2.4418647233310138e-08, + "loss": 0.6985, + "step": 47996 + }, + { + "epoch": 1.76, + "learning_rate": 2.4414985168638078e-08, + "loss": 0.6971, + "step": 47997 + }, + { + "epoch": 1.76, + "learning_rate": 2.4411323103966015e-08, + "loss": 0.6898, + "step": 47998 + }, + { + "epoch": 1.76, + "learning_rate": 2.440766103929395e-08, + "loss": 0.6925, + "step": 47999 + }, + { + "epoch": 1.76, + "learning_rate": 2.4403998974621888e-08, + "loss": 0.7034, + "step": 48000 + }, + { + "epoch": 1.76, + "learning_rate": 2.4400336909949828e-08, + "loss": 0.6936, + "step": 48001 + }, + { + "epoch": 1.76, + "learning_rate": 2.4396674845277768e-08, + "loss": 0.7001, + "step": 48002 + }, + { + "epoch": 1.76, + "learning_rate": 2.4393012780605705e-08, + "loss": 0.6877, + "step": 48003 + }, + { + "epoch": 1.76, + "learning_rate": 2.4389350715933644e-08, + "loss": 0.6853, + "step": 48004 + }, + { + "epoch": 1.76, + "learning_rate": 2.438568865126158e-08, + "loss": 0.6857, + "step": 48005 + }, + { + "epoch": 1.76, + "learning_rate": 2.4382026586589518e-08, + "loss": 0.6941, + "step": 48006 + }, + { + "epoch": 1.76, + "learning_rate": 2.4378364521917454e-08, + "loss": 0.6981, + "step": 48007 + }, + { + "epoch": 1.76, + "learning_rate": 2.4374702457245394e-08, + "loss": 0.7057, + "step": 48008 + }, + { + "epoch": 1.76, + "learning_rate": 2.437104039257333e-08, + "loss": 0.6894, + "step": 48009 + }, + { + "epoch": 1.76, + "learning_rate": 2.436737832790127e-08, + "loss": 0.7008, + "step": 48010 + }, + { + "epoch": 1.76, + "learning_rate": 2.4363716263229207e-08, + "loss": 0.6842, + "step": 48011 + }, + { + "epoch": 1.76, + "learning_rate": 2.4360054198557144e-08, + "loss": 0.6903, + "step": 48012 + }, + { + "epoch": 1.76, + "learning_rate": 2.4356392133885084e-08, + "loss": 0.6907, + "step": 48013 + }, + { + "epoch": 1.76, + "learning_rate": 2.435273006921302e-08, + "loss": 0.6847, + "step": 48014 + }, + { + "epoch": 1.76, + "learning_rate": 2.434906800454096e-08, + "loss": 0.6901, + "step": 48015 + }, + { + "epoch": 1.76, + "learning_rate": 2.4345405939868897e-08, + "loss": 0.6929, + "step": 48016 + }, + { + "epoch": 1.76, + "learning_rate": 2.4341743875196834e-08, + "loss": 0.7011, + "step": 48017 + }, + { + "epoch": 1.76, + "learning_rate": 2.433808181052477e-08, + "loss": 0.6847, + "step": 48018 + }, + { + "epoch": 1.76, + "learning_rate": 2.433441974585271e-08, + "loss": 0.6999, + "step": 48019 + }, + { + "epoch": 1.76, + "learning_rate": 2.433075768118065e-08, + "loss": 0.6928, + "step": 48020 + }, + { + "epoch": 1.76, + "learning_rate": 2.4327095616508587e-08, + "loss": 0.6937, + "step": 48021 + }, + { + "epoch": 1.76, + "learning_rate": 2.4323433551836527e-08, + "loss": 0.6991, + "step": 48022 + }, + { + "epoch": 1.76, + "learning_rate": 2.431977148716446e-08, + "loss": 0.6894, + "step": 48023 + }, + { + "epoch": 1.76, + "learning_rate": 2.43161094224924e-08, + "loss": 0.6899, + "step": 48024 + }, + { + "epoch": 1.76, + "learning_rate": 2.4312447357820337e-08, + "loss": 0.6888, + "step": 48025 + }, + { + "epoch": 1.76, + "learning_rate": 2.4308785293148277e-08, + "loss": 0.6737, + "step": 48026 + }, + { + "epoch": 1.76, + "learning_rate": 2.4305123228476213e-08, + "loss": 0.6973, + "step": 48027 + }, + { + "epoch": 1.76, + "learning_rate": 2.4301461163804153e-08, + "loss": 0.6914, + "step": 48028 + }, + { + "epoch": 1.76, + "learning_rate": 2.429779909913209e-08, + "loss": 0.6895, + "step": 48029 + }, + { + "epoch": 1.76, + "learning_rate": 2.4294137034460026e-08, + "loss": 0.6972, + "step": 48030 + }, + { + "epoch": 1.76, + "learning_rate": 2.4290474969787966e-08, + "loss": 0.6846, + "step": 48031 + }, + { + "epoch": 1.76, + "learning_rate": 2.4286812905115903e-08, + "loss": 0.7057, + "step": 48032 + }, + { + "epoch": 1.76, + "learning_rate": 2.4283150840443843e-08, + "loss": 0.6936, + "step": 48033 + }, + { + "epoch": 1.76, + "learning_rate": 2.427948877577178e-08, + "loss": 0.6885, + "step": 48034 + }, + { + "epoch": 1.76, + "learning_rate": 2.4275826711099716e-08, + "loss": 0.6953, + "step": 48035 + }, + { + "epoch": 1.76, + "learning_rate": 2.4272164646427653e-08, + "loss": 0.7022, + "step": 48036 + }, + { + "epoch": 1.76, + "learning_rate": 2.4268502581755593e-08, + "loss": 0.6871, + "step": 48037 + }, + { + "epoch": 1.76, + "learning_rate": 2.4264840517083533e-08, + "loss": 0.6919, + "step": 48038 + }, + { + "epoch": 1.76, + "learning_rate": 2.426117845241147e-08, + "loss": 0.6926, + "step": 48039 + }, + { + "epoch": 1.76, + "learning_rate": 2.4257516387739406e-08, + "loss": 0.6885, + "step": 48040 + }, + { + "epoch": 1.76, + "learning_rate": 2.4253854323067342e-08, + "loss": 0.7132, + "step": 48041 + }, + { + "epoch": 1.76, + "learning_rate": 2.4250192258395282e-08, + "loss": 0.6874, + "step": 48042 + }, + { + "epoch": 1.76, + "learning_rate": 2.424653019372322e-08, + "loss": 0.6936, + "step": 48043 + }, + { + "epoch": 1.76, + "learning_rate": 2.424286812905116e-08, + "loss": 0.6947, + "step": 48044 + }, + { + "epoch": 1.76, + "learning_rate": 2.4239206064379096e-08, + "loss": 0.6884, + "step": 48045 + }, + { + "epoch": 1.76, + "learning_rate": 2.4235543999707032e-08, + "loss": 0.7007, + "step": 48046 + }, + { + "epoch": 1.76, + "learning_rate": 2.4231881935034972e-08, + "loss": 0.6866, + "step": 48047 + }, + { + "epoch": 1.76, + "learning_rate": 2.422821987036291e-08, + "loss": 0.6951, + "step": 48048 + }, + { + "epoch": 1.76, + "learning_rate": 2.422455780569085e-08, + "loss": 0.6758, + "step": 48049 + }, + { + "epoch": 1.76, + "learning_rate": 2.4220895741018785e-08, + "loss": 0.6888, + "step": 48050 + }, + { + "epoch": 1.76, + "learning_rate": 2.4217233676346725e-08, + "loss": 0.6886, + "step": 48051 + }, + { + "epoch": 1.76, + "learning_rate": 2.4213571611674662e-08, + "loss": 0.6946, + "step": 48052 + }, + { + "epoch": 1.76, + "learning_rate": 2.42099095470026e-08, + "loss": 0.6971, + "step": 48053 + }, + { + "epoch": 1.76, + "learning_rate": 2.4206247482330535e-08, + "loss": 0.6786, + "step": 48054 + }, + { + "epoch": 1.76, + "learning_rate": 2.4202585417658475e-08, + "loss": 0.7027, + "step": 48055 + }, + { + "epoch": 1.76, + "learning_rate": 2.4198923352986415e-08, + "loss": 0.7066, + "step": 48056 + }, + { + "epoch": 1.76, + "learning_rate": 2.419526128831435e-08, + "loss": 0.6986, + "step": 48057 + }, + { + "epoch": 1.76, + "learning_rate": 2.4191599223642288e-08, + "loss": 0.6951, + "step": 48058 + }, + { + "epoch": 1.76, + "learning_rate": 2.4187937158970225e-08, + "loss": 0.6988, + "step": 48059 + }, + { + "epoch": 1.76, + "learning_rate": 2.4184275094298165e-08, + "loss": 0.6966, + "step": 48060 + }, + { + "epoch": 1.76, + "learning_rate": 2.41806130296261e-08, + "loss": 0.6913, + "step": 48061 + }, + { + "epoch": 1.76, + "learning_rate": 2.417695096495404e-08, + "loss": 0.694, + "step": 48062 + }, + { + "epoch": 1.76, + "learning_rate": 2.4173288900281978e-08, + "loss": 0.6884, + "step": 48063 + }, + { + "epoch": 1.76, + "learning_rate": 2.4169626835609914e-08, + "loss": 0.71, + "step": 48064 + }, + { + "epoch": 1.76, + "learning_rate": 2.4165964770937854e-08, + "loss": 0.6858, + "step": 48065 + }, + { + "epoch": 1.76, + "learning_rate": 2.416230270626579e-08, + "loss": 0.7012, + "step": 48066 + }, + { + "epoch": 1.76, + "learning_rate": 2.415864064159373e-08, + "loss": 0.6946, + "step": 48067 + }, + { + "epoch": 1.76, + "learning_rate": 2.4154978576921668e-08, + "loss": 0.6952, + "step": 48068 + }, + { + "epoch": 1.76, + "learning_rate": 2.4151316512249607e-08, + "loss": 0.6956, + "step": 48069 + }, + { + "epoch": 1.76, + "learning_rate": 2.414765444757754e-08, + "loss": 0.6867, + "step": 48070 + }, + { + "epoch": 1.76, + "learning_rate": 2.414399238290548e-08, + "loss": 0.6933, + "step": 48071 + }, + { + "epoch": 1.76, + "learning_rate": 2.4140330318233417e-08, + "loss": 0.6877, + "step": 48072 + }, + { + "epoch": 1.76, + "learning_rate": 2.4136668253561357e-08, + "loss": 0.6945, + "step": 48073 + }, + { + "epoch": 1.76, + "learning_rate": 2.4133006188889297e-08, + "loss": 0.7059, + "step": 48074 + }, + { + "epoch": 1.76, + "learning_rate": 2.4129344124217234e-08, + "loss": 0.6894, + "step": 48075 + }, + { + "epoch": 1.76, + "learning_rate": 2.412568205954517e-08, + "loss": 0.6961, + "step": 48076 + }, + { + "epoch": 1.76, + "learning_rate": 2.4122019994873107e-08, + "loss": 0.6877, + "step": 48077 + }, + { + "epoch": 1.76, + "learning_rate": 2.4118357930201047e-08, + "loss": 0.6912, + "step": 48078 + }, + { + "epoch": 1.76, + "learning_rate": 2.4114695865528984e-08, + "loss": 0.6893, + "step": 48079 + }, + { + "epoch": 1.76, + "learning_rate": 2.4111033800856924e-08, + "loss": 0.6946, + "step": 48080 + }, + { + "epoch": 1.76, + "learning_rate": 2.410737173618486e-08, + "loss": 0.6998, + "step": 48081 + }, + { + "epoch": 1.76, + "learning_rate": 2.4103709671512797e-08, + "loss": 0.6882, + "step": 48082 + }, + { + "epoch": 1.76, + "learning_rate": 2.4100047606840733e-08, + "loss": 0.6968, + "step": 48083 + }, + { + "epoch": 1.76, + "learning_rate": 2.4096385542168673e-08, + "loss": 0.6854, + "step": 48084 + }, + { + "epoch": 1.76, + "learning_rate": 2.4092723477496613e-08, + "loss": 0.6991, + "step": 48085 + }, + { + "epoch": 1.76, + "learning_rate": 2.408906141282455e-08, + "loss": 0.6952, + "step": 48086 + }, + { + "epoch": 1.76, + "learning_rate": 2.4085399348152486e-08, + "loss": 0.7011, + "step": 48087 + }, + { + "epoch": 1.76, + "learning_rate": 2.4081737283480423e-08, + "loss": 0.6896, + "step": 48088 + }, + { + "epoch": 1.76, + "learning_rate": 2.4078075218808363e-08, + "loss": 0.6923, + "step": 48089 + }, + { + "epoch": 1.76, + "learning_rate": 2.40744131541363e-08, + "loss": 0.6958, + "step": 48090 + }, + { + "epoch": 1.76, + "learning_rate": 2.407075108946424e-08, + "loss": 0.7042, + "step": 48091 + }, + { + "epoch": 1.76, + "learning_rate": 2.4067089024792176e-08, + "loss": 0.6999, + "step": 48092 + }, + { + "epoch": 1.76, + "learning_rate": 2.4063426960120116e-08, + "loss": 0.6997, + "step": 48093 + }, + { + "epoch": 1.76, + "learning_rate": 2.4059764895448053e-08, + "loss": 0.6925, + "step": 48094 + }, + { + "epoch": 1.76, + "learning_rate": 2.405610283077599e-08, + "loss": 0.6964, + "step": 48095 + }, + { + "epoch": 1.76, + "learning_rate": 2.405244076610393e-08, + "loss": 0.6989, + "step": 48096 + }, + { + "epoch": 1.76, + "learning_rate": 2.4048778701431866e-08, + "loss": 0.6929, + "step": 48097 + }, + { + "epoch": 1.76, + "learning_rate": 2.4045116636759806e-08, + "loss": 0.6928, + "step": 48098 + }, + { + "epoch": 1.76, + "learning_rate": 2.4041454572087742e-08, + "loss": 0.6894, + "step": 48099 + }, + { + "epoch": 1.76, + "learning_rate": 2.403779250741568e-08, + "loss": 0.6949, + "step": 48100 + }, + { + "epoch": 1.76, + "learning_rate": 2.4034130442743616e-08, + "loss": 0.6935, + "step": 48101 + }, + { + "epoch": 1.76, + "learning_rate": 2.4030468378071556e-08, + "loss": 0.6865, + "step": 48102 + }, + { + "epoch": 1.76, + "learning_rate": 2.4026806313399496e-08, + "loss": 0.7045, + "step": 48103 + }, + { + "epoch": 1.76, + "learning_rate": 2.4023144248727432e-08, + "loss": 0.6847, + "step": 48104 + }, + { + "epoch": 1.76, + "learning_rate": 2.401948218405537e-08, + "loss": 0.706, + "step": 48105 + }, + { + "epoch": 1.76, + "learning_rate": 2.4015820119383305e-08, + "loss": 0.6983, + "step": 48106 + }, + { + "epoch": 1.76, + "learning_rate": 2.4012158054711245e-08, + "loss": 0.7044, + "step": 48107 + }, + { + "epoch": 1.76, + "learning_rate": 2.4008495990039182e-08, + "loss": 0.6897, + "step": 48108 + }, + { + "epoch": 1.76, + "learning_rate": 2.4004833925367122e-08, + "loss": 0.6924, + "step": 48109 + }, + { + "epoch": 1.76, + "learning_rate": 2.400117186069506e-08, + "loss": 0.7017, + "step": 48110 + }, + { + "epoch": 1.76, + "learning_rate": 2.3997509796022995e-08, + "loss": 0.6854, + "step": 48111 + }, + { + "epoch": 1.76, + "learning_rate": 2.3993847731350935e-08, + "loss": 0.698, + "step": 48112 + }, + { + "epoch": 1.76, + "learning_rate": 2.3990185666678872e-08, + "loss": 0.6933, + "step": 48113 + }, + { + "epoch": 1.76, + "learning_rate": 2.398652360200681e-08, + "loss": 0.6914, + "step": 48114 + }, + { + "epoch": 1.76, + "learning_rate": 2.3982861537334748e-08, + "loss": 0.6954, + "step": 48115 + }, + { + "epoch": 1.76, + "learning_rate": 2.3979199472662688e-08, + "loss": 0.6941, + "step": 48116 + }, + { + "epoch": 1.76, + "learning_rate": 2.397553740799062e-08, + "loss": 0.6942, + "step": 48117 + }, + { + "epoch": 1.76, + "learning_rate": 2.397187534331856e-08, + "loss": 0.6853, + "step": 48118 + }, + { + "epoch": 1.76, + "learning_rate": 2.3968213278646498e-08, + "loss": 0.6854, + "step": 48119 + }, + { + "epoch": 1.76, + "learning_rate": 2.3964551213974438e-08, + "loss": 0.7053, + "step": 48120 + }, + { + "epoch": 1.76, + "learning_rate": 2.3960889149302378e-08, + "loss": 0.6978, + "step": 48121 + }, + { + "epoch": 1.76, + "learning_rate": 2.3957227084630315e-08, + "loss": 0.696, + "step": 48122 + }, + { + "epoch": 1.76, + "learning_rate": 2.395356501995825e-08, + "loss": 0.6941, + "step": 48123 + }, + { + "epoch": 1.76, + "learning_rate": 2.3949902955286188e-08, + "loss": 0.688, + "step": 48124 + }, + { + "epoch": 1.76, + "learning_rate": 2.3946240890614128e-08, + "loss": 0.7186, + "step": 48125 + }, + { + "epoch": 1.76, + "learning_rate": 2.3942578825942064e-08, + "loss": 0.6942, + "step": 48126 + }, + { + "epoch": 1.76, + "learning_rate": 2.3938916761270004e-08, + "loss": 0.7016, + "step": 48127 + }, + { + "epoch": 1.76, + "learning_rate": 2.393525469659794e-08, + "loss": 0.6892, + "step": 48128 + }, + { + "epoch": 1.76, + "learning_rate": 2.3931592631925877e-08, + "loss": 0.6851, + "step": 48129 + }, + { + "epoch": 1.76, + "learning_rate": 2.3927930567253817e-08, + "loss": 0.6907, + "step": 48130 + }, + { + "epoch": 1.76, + "learning_rate": 2.3924268502581754e-08, + "loss": 0.6895, + "step": 48131 + }, + { + "epoch": 1.76, + "learning_rate": 2.3920606437909694e-08, + "loss": 0.6978, + "step": 48132 + }, + { + "epoch": 1.76, + "learning_rate": 2.391694437323763e-08, + "loss": 0.6883, + "step": 48133 + }, + { + "epoch": 1.76, + "learning_rate": 2.3913282308565567e-08, + "loss": 0.6894, + "step": 48134 + }, + { + "epoch": 1.76, + "learning_rate": 2.3909620243893504e-08, + "loss": 0.6927, + "step": 48135 + }, + { + "epoch": 1.76, + "learning_rate": 2.3905958179221444e-08, + "loss": 0.6891, + "step": 48136 + }, + { + "epoch": 1.76, + "learning_rate": 2.390229611454938e-08, + "loss": 0.6989, + "step": 48137 + }, + { + "epoch": 1.76, + "learning_rate": 2.389863404987732e-08, + "loss": 0.7004, + "step": 48138 + }, + { + "epoch": 1.76, + "learning_rate": 2.389497198520526e-08, + "loss": 0.6945, + "step": 48139 + }, + { + "epoch": 1.76, + "learning_rate": 2.3891309920533197e-08, + "loss": 0.6878, + "step": 48140 + }, + { + "epoch": 1.76, + "learning_rate": 2.3887647855861133e-08, + "loss": 0.6949, + "step": 48141 + }, + { + "epoch": 1.76, + "learning_rate": 2.388398579118907e-08, + "loss": 0.6985, + "step": 48142 + }, + { + "epoch": 1.76, + "learning_rate": 2.388032372651701e-08, + "loss": 0.6999, + "step": 48143 + }, + { + "epoch": 1.76, + "learning_rate": 2.3876661661844947e-08, + "loss": 0.6867, + "step": 48144 + }, + { + "epoch": 1.76, + "learning_rate": 2.3872999597172887e-08, + "loss": 0.684, + "step": 48145 + }, + { + "epoch": 1.76, + "learning_rate": 2.3869337532500823e-08, + "loss": 0.6941, + "step": 48146 + }, + { + "epoch": 1.76, + "learning_rate": 2.386567546782876e-08, + "loss": 0.6839, + "step": 48147 + }, + { + "epoch": 1.76, + "learning_rate": 2.38620134031567e-08, + "loss": 0.6992, + "step": 48148 + }, + { + "epoch": 1.76, + "learning_rate": 2.3858351338484636e-08, + "loss": 0.6975, + "step": 48149 + }, + { + "epoch": 1.76, + "learning_rate": 2.3854689273812576e-08, + "loss": 0.7005, + "step": 48150 + }, + { + "epoch": 1.76, + "learning_rate": 2.3851027209140513e-08, + "loss": 0.6904, + "step": 48151 + }, + { + "epoch": 1.76, + "learning_rate": 2.384736514446845e-08, + "loss": 0.6924, + "step": 48152 + }, + { + "epoch": 1.76, + "learning_rate": 2.3843703079796386e-08, + "loss": 0.6875, + "step": 48153 + }, + { + "epoch": 1.76, + "learning_rate": 2.3840041015124326e-08, + "loss": 0.6895, + "step": 48154 + }, + { + "epoch": 1.76, + "learning_rate": 2.3836378950452263e-08, + "loss": 0.6942, + "step": 48155 + }, + { + "epoch": 1.76, + "learning_rate": 2.3832716885780203e-08, + "loss": 0.688, + "step": 48156 + }, + { + "epoch": 1.76, + "learning_rate": 2.3829054821108143e-08, + "loss": 0.6917, + "step": 48157 + }, + { + "epoch": 1.76, + "learning_rate": 2.3825392756436076e-08, + "loss": 0.697, + "step": 48158 + }, + { + "epoch": 1.76, + "learning_rate": 2.3821730691764016e-08, + "loss": 0.6917, + "step": 48159 + }, + { + "epoch": 1.76, + "learning_rate": 2.3818068627091952e-08, + "loss": 0.6949, + "step": 48160 + }, + { + "epoch": 1.76, + "learning_rate": 2.3814406562419892e-08, + "loss": 0.6777, + "step": 48161 + }, + { + "epoch": 1.76, + "learning_rate": 2.381074449774783e-08, + "loss": 0.6996, + "step": 48162 + }, + { + "epoch": 1.76, + "learning_rate": 2.380708243307577e-08, + "loss": 0.6888, + "step": 48163 + }, + { + "epoch": 1.76, + "learning_rate": 2.3803420368403702e-08, + "loss": 0.6935, + "step": 48164 + }, + { + "epoch": 1.76, + "learning_rate": 2.3799758303731642e-08, + "loss": 0.6884, + "step": 48165 + }, + { + "epoch": 1.76, + "learning_rate": 2.3796096239059582e-08, + "loss": 0.6911, + "step": 48166 + }, + { + "epoch": 1.76, + "learning_rate": 2.379243417438752e-08, + "loss": 0.7002, + "step": 48167 + }, + { + "epoch": 1.76, + "learning_rate": 2.378877210971546e-08, + "loss": 0.6944, + "step": 48168 + }, + { + "epoch": 1.76, + "learning_rate": 2.3785110045043395e-08, + "loss": 0.6984, + "step": 48169 + }, + { + "epoch": 1.76, + "learning_rate": 2.3781447980371332e-08, + "loss": 0.6874, + "step": 48170 + }, + { + "epoch": 1.76, + "learning_rate": 2.377778591569927e-08, + "loss": 0.7078, + "step": 48171 + }, + { + "epoch": 1.76, + "learning_rate": 2.377412385102721e-08, + "loss": 0.6864, + "step": 48172 + }, + { + "epoch": 1.76, + "learning_rate": 2.3770461786355145e-08, + "loss": 0.6957, + "step": 48173 + }, + { + "epoch": 1.76, + "learning_rate": 2.3766799721683085e-08, + "loss": 0.6943, + "step": 48174 + }, + { + "epoch": 1.76, + "learning_rate": 2.376313765701102e-08, + "loss": 0.6863, + "step": 48175 + }, + { + "epoch": 1.76, + "learning_rate": 2.3759475592338958e-08, + "loss": 0.6835, + "step": 48176 + }, + { + "epoch": 1.76, + "learning_rate": 2.3755813527666898e-08, + "loss": 0.6886, + "step": 48177 + }, + { + "epoch": 1.76, + "learning_rate": 2.3752151462994835e-08, + "loss": 0.6962, + "step": 48178 + }, + { + "epoch": 1.76, + "learning_rate": 2.3748489398322775e-08, + "loss": 0.688, + "step": 48179 + }, + { + "epoch": 1.76, + "learning_rate": 2.374482733365071e-08, + "loss": 0.6901, + "step": 48180 + }, + { + "epoch": 1.76, + "learning_rate": 2.374116526897865e-08, + "loss": 0.6953, + "step": 48181 + }, + { + "epoch": 1.76, + "learning_rate": 2.3737503204306585e-08, + "loss": 0.671, + "step": 48182 + }, + { + "epoch": 1.76, + "learning_rate": 2.3733841139634524e-08, + "loss": 0.6724, + "step": 48183 + }, + { + "epoch": 1.76, + "learning_rate": 2.373017907496246e-08, + "loss": 0.6989, + "step": 48184 + }, + { + "epoch": 1.76, + "learning_rate": 2.37265170102904e-08, + "loss": 0.6948, + "step": 48185 + }, + { + "epoch": 1.76, + "learning_rate": 2.372285494561834e-08, + "loss": 0.6913, + "step": 48186 + }, + { + "epoch": 1.76, + "learning_rate": 2.3719192880946278e-08, + "loss": 0.6969, + "step": 48187 + }, + { + "epoch": 1.76, + "learning_rate": 2.3715530816274214e-08, + "loss": 0.6963, + "step": 48188 + }, + { + "epoch": 1.76, + "learning_rate": 2.371186875160215e-08, + "loss": 0.6841, + "step": 48189 + }, + { + "epoch": 1.76, + "learning_rate": 2.370820668693009e-08, + "loss": 0.6941, + "step": 48190 + }, + { + "epoch": 1.76, + "learning_rate": 2.3704544622258027e-08, + "loss": 0.6987, + "step": 48191 + }, + { + "epoch": 1.76, + "learning_rate": 2.3700882557585967e-08, + "loss": 0.6964, + "step": 48192 + }, + { + "epoch": 1.76, + "learning_rate": 2.3697220492913904e-08, + "loss": 0.6886, + "step": 48193 + }, + { + "epoch": 1.76, + "learning_rate": 2.369355842824184e-08, + "loss": 0.6868, + "step": 48194 + }, + { + "epoch": 1.76, + "learning_rate": 2.368989636356978e-08, + "loss": 0.6927, + "step": 48195 + }, + { + "epoch": 1.76, + "learning_rate": 2.3686234298897717e-08, + "loss": 0.698, + "step": 48196 + }, + { + "epoch": 1.76, + "learning_rate": 2.3682572234225657e-08, + "loss": 0.6945, + "step": 48197 + }, + { + "epoch": 1.76, + "learning_rate": 2.3678910169553594e-08, + "loss": 0.6863, + "step": 48198 + }, + { + "epoch": 1.76, + "learning_rate": 2.367524810488153e-08, + "loss": 0.682, + "step": 48199 + }, + { + "epoch": 1.76, + "learning_rate": 2.3671586040209467e-08, + "loss": 0.6822, + "step": 48200 + }, + { + "epoch": 1.76, + "learning_rate": 2.3667923975537407e-08, + "loss": 0.6838, + "step": 48201 + }, + { + "epoch": 1.76, + "learning_rate": 2.3664261910865343e-08, + "loss": 0.695, + "step": 48202 + }, + { + "epoch": 1.76, + "learning_rate": 2.3660599846193283e-08, + "loss": 0.6979, + "step": 48203 + }, + { + "epoch": 1.76, + "learning_rate": 2.3656937781521223e-08, + "loss": 0.7002, + "step": 48204 + }, + { + "epoch": 1.76, + "learning_rate": 2.3653275716849157e-08, + "loss": 0.698, + "step": 48205 + }, + { + "epoch": 1.76, + "learning_rate": 2.3649613652177096e-08, + "loss": 0.6982, + "step": 48206 + }, + { + "epoch": 1.76, + "learning_rate": 2.3645951587505033e-08, + "loss": 0.6737, + "step": 48207 + }, + { + "epoch": 1.76, + "learning_rate": 2.3642289522832973e-08, + "loss": 0.6935, + "step": 48208 + }, + { + "epoch": 1.76, + "learning_rate": 2.363862745816091e-08, + "loss": 0.7004, + "step": 48209 + }, + { + "epoch": 1.76, + "learning_rate": 2.363496539348885e-08, + "loss": 0.698, + "step": 48210 + }, + { + "epoch": 1.76, + "learning_rate": 2.3631303328816786e-08, + "loss": 0.6926, + "step": 48211 + }, + { + "epoch": 1.76, + "learning_rate": 2.3627641264144723e-08, + "loss": 0.6934, + "step": 48212 + }, + { + "epoch": 1.76, + "learning_rate": 2.3623979199472663e-08, + "loss": 0.6874, + "step": 48213 + }, + { + "epoch": 1.76, + "learning_rate": 2.36203171348006e-08, + "loss": 0.7069, + "step": 48214 + }, + { + "epoch": 1.76, + "learning_rate": 2.361665507012854e-08, + "loss": 0.6864, + "step": 48215 + }, + { + "epoch": 1.76, + "learning_rate": 2.3612993005456476e-08, + "loss": 0.6925, + "step": 48216 + }, + { + "epoch": 1.76, + "learning_rate": 2.3609330940784413e-08, + "loss": 0.6863, + "step": 48217 + }, + { + "epoch": 1.76, + "learning_rate": 2.360566887611235e-08, + "loss": 0.6839, + "step": 48218 + }, + { + "epoch": 1.76, + "learning_rate": 2.360200681144029e-08, + "loss": 0.7016, + "step": 48219 + }, + { + "epoch": 1.76, + "learning_rate": 2.3598344746768226e-08, + "loss": 0.6931, + "step": 48220 + }, + { + "epoch": 1.76, + "learning_rate": 2.3594682682096166e-08, + "loss": 0.6917, + "step": 48221 + }, + { + "epoch": 1.76, + "learning_rate": 2.3591020617424102e-08, + "loss": 0.6933, + "step": 48222 + }, + { + "epoch": 1.76, + "learning_rate": 2.358735855275204e-08, + "loss": 0.6988, + "step": 48223 + }, + { + "epoch": 1.76, + "learning_rate": 2.358369648807998e-08, + "loss": 0.6887, + "step": 48224 + }, + { + "epoch": 1.76, + "learning_rate": 2.3580034423407915e-08, + "loss": 0.6876, + "step": 48225 + }, + { + "epoch": 1.76, + "learning_rate": 2.3576372358735855e-08, + "loss": 0.7098, + "step": 48226 + }, + { + "epoch": 1.76, + "learning_rate": 2.3572710294063792e-08, + "loss": 0.6892, + "step": 48227 + }, + { + "epoch": 1.76, + "learning_rate": 2.3569048229391732e-08, + "loss": 0.6985, + "step": 48228 + }, + { + "epoch": 1.76, + "learning_rate": 2.3565386164719665e-08, + "loss": 0.6865, + "step": 48229 + }, + { + "epoch": 1.76, + "learning_rate": 2.3561724100047605e-08, + "loss": 0.6791, + "step": 48230 + }, + { + "epoch": 1.76, + "learning_rate": 2.3558062035375545e-08, + "loss": 0.687, + "step": 48231 + }, + { + "epoch": 1.76, + "learning_rate": 2.3554399970703482e-08, + "loss": 0.6818, + "step": 48232 + }, + { + "epoch": 1.76, + "learning_rate": 2.355073790603142e-08, + "loss": 0.6964, + "step": 48233 + }, + { + "epoch": 1.76, + "learning_rate": 2.3547075841359358e-08, + "loss": 0.6888, + "step": 48234 + }, + { + "epoch": 1.76, + "learning_rate": 2.3543413776687295e-08, + "loss": 0.6886, + "step": 48235 + }, + { + "epoch": 1.76, + "learning_rate": 2.353975171201523e-08, + "loss": 0.7031, + "step": 48236 + }, + { + "epoch": 1.76, + "learning_rate": 2.353608964734317e-08, + "loss": 0.6755, + "step": 48237 + }, + { + "epoch": 1.76, + "learning_rate": 2.3532427582671108e-08, + "loss": 0.694, + "step": 48238 + }, + { + "epoch": 1.76, + "learning_rate": 2.3528765517999048e-08, + "loss": 0.6975, + "step": 48239 + }, + { + "epoch": 1.76, + "learning_rate": 2.3525103453326985e-08, + "loss": 0.6906, + "step": 48240 + }, + { + "epoch": 1.77, + "learning_rate": 2.352144138865492e-08, + "loss": 0.6907, + "step": 48241 + }, + { + "epoch": 1.77, + "learning_rate": 2.351777932398286e-08, + "loss": 0.698, + "step": 48242 + }, + { + "epoch": 1.77, + "learning_rate": 2.3514117259310798e-08, + "loss": 0.6906, + "step": 48243 + }, + { + "epoch": 1.77, + "learning_rate": 2.3510455194638738e-08, + "loss": 0.6963, + "step": 48244 + }, + { + "epoch": 1.77, + "learning_rate": 2.3506793129966674e-08, + "loss": 0.6926, + "step": 48245 + }, + { + "epoch": 1.77, + "learning_rate": 2.350313106529461e-08, + "loss": 0.7004, + "step": 48246 + }, + { + "epoch": 1.77, + "learning_rate": 2.3499469000622548e-08, + "loss": 0.6829, + "step": 48247 + }, + { + "epoch": 1.77, + "learning_rate": 2.3495806935950487e-08, + "loss": 0.6967, + "step": 48248 + }, + { + "epoch": 1.77, + "learning_rate": 2.3492144871278427e-08, + "loss": 0.6913, + "step": 48249 + }, + { + "epoch": 1.77, + "learning_rate": 2.3488482806606364e-08, + "loss": 0.6887, + "step": 48250 + }, + { + "epoch": 1.77, + "learning_rate": 2.3484820741934304e-08, + "loss": 0.6914, + "step": 48251 + }, + { + "epoch": 1.77, + "learning_rate": 2.3481158677262237e-08, + "loss": 0.6985, + "step": 48252 + }, + { + "epoch": 1.77, + "learning_rate": 2.3477496612590177e-08, + "loss": 0.6883, + "step": 48253 + }, + { + "epoch": 1.77, + "learning_rate": 2.3473834547918114e-08, + "loss": 0.7005, + "step": 48254 + }, + { + "epoch": 1.77, + "learning_rate": 2.3470172483246054e-08, + "loss": 0.6795, + "step": 48255 + }, + { + "epoch": 1.77, + "learning_rate": 2.346651041857399e-08, + "loss": 0.7134, + "step": 48256 + }, + { + "epoch": 1.77, + "learning_rate": 2.346284835390193e-08, + "loss": 0.6836, + "step": 48257 + }, + { + "epoch": 1.77, + "learning_rate": 2.3459186289229867e-08, + "loss": 0.6929, + "step": 48258 + }, + { + "epoch": 1.77, + "learning_rate": 2.3455524224557803e-08, + "loss": 0.6984, + "step": 48259 + }, + { + "epoch": 1.77, + "learning_rate": 2.3451862159885743e-08, + "loss": 0.6943, + "step": 48260 + }, + { + "epoch": 1.77, + "learning_rate": 2.344820009521368e-08, + "loss": 0.6887, + "step": 48261 + }, + { + "epoch": 1.77, + "learning_rate": 2.344453803054162e-08, + "loss": 0.6815, + "step": 48262 + }, + { + "epoch": 1.77, + "learning_rate": 2.3440875965869557e-08, + "loss": 0.6887, + "step": 48263 + }, + { + "epoch": 1.77, + "learning_rate": 2.3437213901197493e-08, + "loss": 0.6891, + "step": 48264 + }, + { + "epoch": 1.77, + "learning_rate": 2.343355183652543e-08, + "loss": 0.6836, + "step": 48265 + }, + { + "epoch": 1.77, + "learning_rate": 2.342988977185337e-08, + "loss": 0.6968, + "step": 48266 + }, + { + "epoch": 1.77, + "learning_rate": 2.342622770718131e-08, + "loss": 0.6944, + "step": 48267 + }, + { + "epoch": 1.77, + "learning_rate": 2.3422565642509246e-08, + "loss": 0.6772, + "step": 48268 + }, + { + "epoch": 1.77, + "learning_rate": 2.3418903577837186e-08, + "loss": 0.6909, + "step": 48269 + }, + { + "epoch": 1.77, + "learning_rate": 2.341524151316512e-08, + "loss": 0.6973, + "step": 48270 + }, + { + "epoch": 1.77, + "learning_rate": 2.341157944849306e-08, + "loss": 0.6947, + "step": 48271 + }, + { + "epoch": 1.77, + "learning_rate": 2.3407917383820996e-08, + "loss": 0.6954, + "step": 48272 + }, + { + "epoch": 1.77, + "learning_rate": 2.3404255319148936e-08, + "loss": 0.6937, + "step": 48273 + }, + { + "epoch": 1.77, + "learning_rate": 2.3400593254476873e-08, + "loss": 0.6938, + "step": 48274 + }, + { + "epoch": 1.77, + "learning_rate": 2.3396931189804813e-08, + "loss": 0.7013, + "step": 48275 + }, + { + "epoch": 1.77, + "learning_rate": 2.3393269125132746e-08, + "loss": 0.6939, + "step": 48276 + }, + { + "epoch": 1.77, + "learning_rate": 2.3389607060460686e-08, + "loss": 0.6866, + "step": 48277 + }, + { + "epoch": 1.77, + "learning_rate": 2.3385944995788626e-08, + "loss": 0.6872, + "step": 48278 + }, + { + "epoch": 1.77, + "learning_rate": 2.3382282931116562e-08, + "loss": 0.6848, + "step": 48279 + }, + { + "epoch": 1.77, + "learning_rate": 2.3378620866444502e-08, + "loss": 0.6842, + "step": 48280 + }, + { + "epoch": 1.77, + "learning_rate": 2.337495880177244e-08, + "loss": 0.6824, + "step": 48281 + }, + { + "epoch": 1.77, + "learning_rate": 2.3371296737100376e-08, + "loss": 0.7014, + "step": 48282 + }, + { + "epoch": 1.77, + "learning_rate": 2.3367634672428312e-08, + "loss": 0.6905, + "step": 48283 + }, + { + "epoch": 1.77, + "learning_rate": 2.3363972607756252e-08, + "loss": 0.6893, + "step": 48284 + }, + { + "epoch": 1.77, + "learning_rate": 2.336031054308419e-08, + "loss": 0.6978, + "step": 48285 + }, + { + "epoch": 1.77, + "learning_rate": 2.335664847841213e-08, + "loss": 0.689, + "step": 48286 + }, + { + "epoch": 1.77, + "learning_rate": 2.3352986413740065e-08, + "loss": 0.698, + "step": 48287 + }, + { + "epoch": 1.77, + "learning_rate": 2.3349324349068002e-08, + "loss": 0.6981, + "step": 48288 + }, + { + "epoch": 1.77, + "learning_rate": 2.3345662284395942e-08, + "loss": 0.7024, + "step": 48289 + }, + { + "epoch": 1.77, + "learning_rate": 2.334200021972388e-08, + "loss": 0.6964, + "step": 48290 + }, + { + "epoch": 1.77, + "learning_rate": 2.3338338155051818e-08, + "loss": 0.6746, + "step": 48291 + }, + { + "epoch": 1.77, + "learning_rate": 2.3334676090379755e-08, + "loss": 0.6838, + "step": 48292 + }, + { + "epoch": 1.77, + "learning_rate": 2.333101402570769e-08, + "loss": 0.6971, + "step": 48293 + }, + { + "epoch": 1.77, + "learning_rate": 2.3327351961035628e-08, + "loss": 0.6929, + "step": 48294 + }, + { + "epoch": 1.77, + "learning_rate": 2.3323689896363568e-08, + "loss": 0.6925, + "step": 48295 + }, + { + "epoch": 1.77, + "learning_rate": 2.3320027831691508e-08, + "loss": 0.7056, + "step": 48296 + }, + { + "epoch": 1.77, + "learning_rate": 2.3316365767019445e-08, + "loss": 0.697, + "step": 48297 + }, + { + "epoch": 1.77, + "learning_rate": 2.3312703702347385e-08, + "loss": 0.6979, + "step": 48298 + }, + { + "epoch": 1.77, + "learning_rate": 2.330904163767532e-08, + "loss": 0.7095, + "step": 48299 + }, + { + "epoch": 1.77, + "learning_rate": 2.3305379573003258e-08, + "loss": 0.7006, + "step": 48300 + }, + { + "epoch": 1.77, + "learning_rate": 2.3301717508331194e-08, + "loss": 0.6965, + "step": 48301 + }, + { + "epoch": 1.77, + "learning_rate": 2.3298055443659134e-08, + "loss": 0.6887, + "step": 48302 + }, + { + "epoch": 1.77, + "learning_rate": 2.329439337898707e-08, + "loss": 0.692, + "step": 48303 + }, + { + "epoch": 1.77, + "learning_rate": 2.329073131431501e-08, + "loss": 0.6997, + "step": 48304 + }, + { + "epoch": 1.77, + "learning_rate": 2.3287069249642948e-08, + "loss": 0.6996, + "step": 48305 + }, + { + "epoch": 1.77, + "learning_rate": 2.3283407184970884e-08, + "loss": 0.6991, + "step": 48306 + }, + { + "epoch": 1.77, + "learning_rate": 2.3279745120298824e-08, + "loss": 0.7024, + "step": 48307 + }, + { + "epoch": 1.77, + "learning_rate": 2.327608305562676e-08, + "loss": 0.6914, + "step": 48308 + }, + { + "epoch": 1.77, + "learning_rate": 2.32724209909547e-08, + "loss": 0.6825, + "step": 48309 + }, + { + "epoch": 1.77, + "learning_rate": 2.3268758926282637e-08, + "loss": 0.6896, + "step": 48310 + }, + { + "epoch": 1.77, + "learning_rate": 2.3265096861610574e-08, + "loss": 0.6847, + "step": 48311 + }, + { + "epoch": 1.77, + "learning_rate": 2.326143479693851e-08, + "loss": 0.6898, + "step": 48312 + }, + { + "epoch": 1.77, + "learning_rate": 2.325777273226645e-08, + "loss": 0.6913, + "step": 48313 + }, + { + "epoch": 1.77, + "learning_rate": 2.325411066759439e-08, + "loss": 0.6946, + "step": 48314 + }, + { + "epoch": 1.77, + "learning_rate": 2.3250448602922327e-08, + "loss": 0.6925, + "step": 48315 + }, + { + "epoch": 1.77, + "learning_rate": 2.3246786538250267e-08, + "loss": 0.7026, + "step": 48316 + }, + { + "epoch": 1.77, + "learning_rate": 2.32431244735782e-08, + "loss": 0.6937, + "step": 48317 + }, + { + "epoch": 1.77, + "learning_rate": 2.323946240890614e-08, + "loss": 0.7052, + "step": 48318 + }, + { + "epoch": 1.77, + "learning_rate": 2.3235800344234077e-08, + "loss": 0.6873, + "step": 48319 + }, + { + "epoch": 1.77, + "learning_rate": 2.3232138279562017e-08, + "loss": 0.71, + "step": 48320 + }, + { + "epoch": 1.77, + "learning_rate": 2.3228476214889953e-08, + "loss": 0.6872, + "step": 48321 + }, + { + "epoch": 1.77, + "learning_rate": 2.3224814150217893e-08, + "loss": 0.6979, + "step": 48322 + }, + { + "epoch": 1.77, + "learning_rate": 2.322115208554583e-08, + "loss": 0.6941, + "step": 48323 + }, + { + "epoch": 1.77, + "learning_rate": 2.3217490020873767e-08, + "loss": 0.6963, + "step": 48324 + }, + { + "epoch": 1.77, + "learning_rate": 2.3213827956201706e-08, + "loss": 0.6991, + "step": 48325 + }, + { + "epoch": 1.77, + "learning_rate": 2.3210165891529643e-08, + "loss": 0.7068, + "step": 48326 + }, + { + "epoch": 1.77, + "learning_rate": 2.3206503826857583e-08, + "loss": 0.6864, + "step": 48327 + }, + { + "epoch": 1.77, + "learning_rate": 2.320284176218552e-08, + "loss": 0.6916, + "step": 48328 + }, + { + "epoch": 1.77, + "learning_rate": 2.3199179697513456e-08, + "loss": 0.6924, + "step": 48329 + }, + { + "epoch": 1.77, + "learning_rate": 2.3195517632841393e-08, + "loss": 0.6878, + "step": 48330 + }, + { + "epoch": 1.77, + "learning_rate": 2.3191855568169333e-08, + "loss": 0.6956, + "step": 48331 + }, + { + "epoch": 1.77, + "learning_rate": 2.3188193503497273e-08, + "loss": 0.6981, + "step": 48332 + }, + { + "epoch": 1.77, + "learning_rate": 2.318453143882521e-08, + "loss": 0.6866, + "step": 48333 + }, + { + "epoch": 1.77, + "learning_rate": 2.3180869374153146e-08, + "loss": 0.7005, + "step": 48334 + }, + { + "epoch": 1.77, + "learning_rate": 2.3177207309481083e-08, + "loss": 0.6963, + "step": 48335 + }, + { + "epoch": 1.77, + "learning_rate": 2.3173545244809022e-08, + "loss": 0.6856, + "step": 48336 + }, + { + "epoch": 1.77, + "learning_rate": 2.316988318013696e-08, + "loss": 0.6894, + "step": 48337 + }, + { + "epoch": 1.77, + "learning_rate": 2.31662211154649e-08, + "loss": 0.6993, + "step": 48338 + }, + { + "epoch": 1.77, + "learning_rate": 2.3162559050792836e-08, + "loss": 0.6951, + "step": 48339 + }, + { + "epoch": 1.77, + "learning_rate": 2.3158896986120772e-08, + "loss": 0.6948, + "step": 48340 + }, + { + "epoch": 1.77, + "learning_rate": 2.3155234921448712e-08, + "loss": 0.6864, + "step": 48341 + }, + { + "epoch": 1.77, + "learning_rate": 2.315157285677665e-08, + "loss": 0.6834, + "step": 48342 + }, + { + "epoch": 1.77, + "learning_rate": 2.314791079210459e-08, + "loss": 0.6983, + "step": 48343 + }, + { + "epoch": 1.77, + "learning_rate": 2.3144248727432525e-08, + "loss": 0.7004, + "step": 48344 + }, + { + "epoch": 1.77, + "learning_rate": 2.3140586662760465e-08, + "loss": 0.6903, + "step": 48345 + }, + { + "epoch": 1.77, + "learning_rate": 2.3136924598088402e-08, + "loss": 0.6893, + "step": 48346 + }, + { + "epoch": 1.77, + "learning_rate": 2.313326253341634e-08, + "loss": 0.7018, + "step": 48347 + }, + { + "epoch": 1.77, + "learning_rate": 2.3129600468744275e-08, + "loss": 0.6949, + "step": 48348 + }, + { + "epoch": 1.77, + "learning_rate": 2.3125938404072215e-08, + "loss": 0.6923, + "step": 48349 + }, + { + "epoch": 1.77, + "learning_rate": 2.3122276339400155e-08, + "loss": 0.6978, + "step": 48350 + }, + { + "epoch": 1.77, + "learning_rate": 2.311861427472809e-08, + "loss": 0.6905, + "step": 48351 + }, + { + "epoch": 1.77, + "learning_rate": 2.3114952210056028e-08, + "loss": 0.7044, + "step": 48352 + }, + { + "epoch": 1.77, + "learning_rate": 2.3111290145383965e-08, + "loss": 0.7021, + "step": 48353 + }, + { + "epoch": 1.77, + "learning_rate": 2.3107628080711905e-08, + "loss": 0.6963, + "step": 48354 + }, + { + "epoch": 1.77, + "learning_rate": 2.310396601603984e-08, + "loss": 0.6996, + "step": 48355 + }, + { + "epoch": 1.77, + "learning_rate": 2.310030395136778e-08, + "loss": 0.6952, + "step": 48356 + }, + { + "epoch": 1.77, + "learning_rate": 2.3096641886695718e-08, + "loss": 0.6851, + "step": 48357 + }, + { + "epoch": 1.77, + "learning_rate": 2.3092979822023655e-08, + "loss": 0.6988, + "step": 48358 + }, + { + "epoch": 1.77, + "learning_rate": 2.3089317757351595e-08, + "loss": 0.6858, + "step": 48359 + }, + { + "epoch": 1.77, + "learning_rate": 2.308565569267953e-08, + "loss": 0.6897, + "step": 48360 + }, + { + "epoch": 1.77, + "learning_rate": 2.308199362800747e-08, + "loss": 0.6921, + "step": 48361 + }, + { + "epoch": 1.77, + "learning_rate": 2.3078331563335408e-08, + "loss": 0.6899, + "step": 48362 + }, + { + "epoch": 1.77, + "learning_rate": 2.3074669498663348e-08, + "loss": 0.6855, + "step": 48363 + }, + { + "epoch": 1.77, + "learning_rate": 2.307100743399128e-08, + "loss": 0.694, + "step": 48364 + }, + { + "epoch": 1.77, + "learning_rate": 2.306734536931922e-08, + "loss": 0.6906, + "step": 48365 + }, + { + "epoch": 1.77, + "learning_rate": 2.3063683304647157e-08, + "loss": 0.6996, + "step": 48366 + }, + { + "epoch": 1.77, + "learning_rate": 2.3060021239975097e-08, + "loss": 0.68, + "step": 48367 + }, + { + "epoch": 1.77, + "learning_rate": 2.3056359175303034e-08, + "loss": 0.6937, + "step": 48368 + }, + { + "epoch": 1.77, + "learning_rate": 2.3052697110630974e-08, + "loss": 0.6917, + "step": 48369 + }, + { + "epoch": 1.77, + "learning_rate": 2.304903504595891e-08, + "loss": 0.7068, + "step": 48370 + }, + { + "epoch": 1.77, + "learning_rate": 2.3045372981286847e-08, + "loss": 0.6892, + "step": 48371 + }, + { + "epoch": 1.77, + "learning_rate": 2.3041710916614787e-08, + "loss": 0.6831, + "step": 48372 + }, + { + "epoch": 1.77, + "learning_rate": 2.3038048851942724e-08, + "loss": 0.7045, + "step": 48373 + }, + { + "epoch": 1.77, + "learning_rate": 2.3034386787270664e-08, + "loss": 0.6822, + "step": 48374 + }, + { + "epoch": 1.77, + "learning_rate": 2.30307247225986e-08, + "loss": 0.692, + "step": 48375 + }, + { + "epoch": 1.77, + "learning_rate": 2.3027062657926537e-08, + "loss": 0.683, + "step": 48376 + }, + { + "epoch": 1.77, + "learning_rate": 2.3023400593254474e-08, + "loss": 0.6833, + "step": 48377 + }, + { + "epoch": 1.77, + "learning_rate": 2.3019738528582413e-08, + "loss": 0.6877, + "step": 48378 + }, + { + "epoch": 1.77, + "learning_rate": 2.3016076463910353e-08, + "loss": 0.6852, + "step": 48379 + }, + { + "epoch": 1.77, + "learning_rate": 2.301241439923829e-08, + "loss": 0.7087, + "step": 48380 + }, + { + "epoch": 1.77, + "learning_rate": 2.3008752334566227e-08, + "loss": 0.6923, + "step": 48381 + }, + { + "epoch": 1.77, + "learning_rate": 2.3005090269894163e-08, + "loss": 0.6809, + "step": 48382 + }, + { + "epoch": 1.77, + "learning_rate": 2.3001428205222103e-08, + "loss": 0.7041, + "step": 48383 + }, + { + "epoch": 1.77, + "learning_rate": 2.299776614055004e-08, + "loss": 0.6848, + "step": 48384 + }, + { + "epoch": 1.77, + "learning_rate": 2.299410407587798e-08, + "loss": 0.6972, + "step": 48385 + }, + { + "epoch": 1.77, + "learning_rate": 2.2990442011205916e-08, + "loss": 0.6904, + "step": 48386 + }, + { + "epoch": 1.77, + "learning_rate": 2.2986779946533856e-08, + "loss": 0.7032, + "step": 48387 + }, + { + "epoch": 1.77, + "learning_rate": 2.2983117881861793e-08, + "loss": 0.6928, + "step": 48388 + }, + { + "epoch": 1.77, + "learning_rate": 2.297945581718973e-08, + "loss": 0.6981, + "step": 48389 + }, + { + "epoch": 1.77, + "learning_rate": 2.297579375251767e-08, + "loss": 0.6896, + "step": 48390 + }, + { + "epoch": 1.77, + "learning_rate": 2.2972131687845606e-08, + "loss": 0.6812, + "step": 48391 + }, + { + "epoch": 1.77, + "learning_rate": 2.2968469623173546e-08, + "loss": 0.6951, + "step": 48392 + }, + { + "epoch": 1.77, + "learning_rate": 2.2964807558501483e-08, + "loss": 0.6843, + "step": 48393 + }, + { + "epoch": 1.77, + "learning_rate": 2.296114549382942e-08, + "loss": 0.6914, + "step": 48394 + }, + { + "epoch": 1.77, + "learning_rate": 2.2957483429157356e-08, + "loss": 0.6914, + "step": 48395 + }, + { + "epoch": 1.77, + "learning_rate": 2.2953821364485296e-08, + "loss": 0.6814, + "step": 48396 + }, + { + "epoch": 1.77, + "learning_rate": 2.2950159299813236e-08, + "loss": 0.6981, + "step": 48397 + }, + { + "epoch": 1.77, + "learning_rate": 2.2946497235141172e-08, + "loss": 0.6971, + "step": 48398 + }, + { + "epoch": 1.77, + "learning_rate": 2.294283517046911e-08, + "loss": 0.704, + "step": 48399 + }, + { + "epoch": 1.77, + "learning_rate": 2.2939173105797046e-08, + "loss": 0.6867, + "step": 48400 + }, + { + "epoch": 1.77, + "learning_rate": 2.2935511041124986e-08, + "loss": 0.6876, + "step": 48401 + }, + { + "epoch": 1.77, + "learning_rate": 2.2931848976452922e-08, + "loss": 0.6876, + "step": 48402 + }, + { + "epoch": 1.77, + "learning_rate": 2.2928186911780862e-08, + "loss": 0.6924, + "step": 48403 + }, + { + "epoch": 1.77, + "learning_rate": 2.29245248471088e-08, + "loss": 0.6922, + "step": 48404 + }, + { + "epoch": 1.77, + "learning_rate": 2.2920862782436735e-08, + "loss": 0.6933, + "step": 48405 + }, + { + "epoch": 1.77, + "learning_rate": 2.2917200717764675e-08, + "loss": 0.6967, + "step": 48406 + }, + { + "epoch": 1.77, + "learning_rate": 2.2913538653092612e-08, + "loss": 0.6835, + "step": 48407 + }, + { + "epoch": 1.77, + "learning_rate": 2.2909876588420552e-08, + "loss": 0.6987, + "step": 48408 + }, + { + "epoch": 1.77, + "learning_rate": 2.290621452374849e-08, + "loss": 0.686, + "step": 48409 + }, + { + "epoch": 1.77, + "learning_rate": 2.2902552459076428e-08, + "loss": 0.6846, + "step": 48410 + }, + { + "epoch": 1.77, + "learning_rate": 2.289889039440436e-08, + "loss": 0.7021, + "step": 48411 + }, + { + "epoch": 1.77, + "learning_rate": 2.28952283297323e-08, + "loss": 0.7026, + "step": 48412 + }, + { + "epoch": 1.77, + "learning_rate": 2.2891566265060238e-08, + "loss": 0.7021, + "step": 48413 + }, + { + "epoch": 1.77, + "learning_rate": 2.2887904200388178e-08, + "loss": 0.6943, + "step": 48414 + }, + { + "epoch": 1.77, + "learning_rate": 2.2884242135716118e-08, + "loss": 0.6924, + "step": 48415 + }, + { + "epoch": 1.77, + "learning_rate": 2.2880580071044055e-08, + "loss": 0.7101, + "step": 48416 + }, + { + "epoch": 1.77, + "learning_rate": 2.287691800637199e-08, + "loss": 0.6882, + "step": 48417 + }, + { + "epoch": 1.77, + "learning_rate": 2.2873255941699928e-08, + "loss": 0.6945, + "step": 48418 + }, + { + "epoch": 1.77, + "learning_rate": 2.2869593877027868e-08, + "loss": 0.6895, + "step": 48419 + }, + { + "epoch": 1.77, + "learning_rate": 2.2865931812355804e-08, + "loss": 0.6939, + "step": 48420 + }, + { + "epoch": 1.77, + "learning_rate": 2.2862269747683744e-08, + "loss": 0.6929, + "step": 48421 + }, + { + "epoch": 1.77, + "learning_rate": 2.285860768301168e-08, + "loss": 0.6989, + "step": 48422 + }, + { + "epoch": 1.77, + "learning_rate": 2.2854945618339618e-08, + "loss": 0.6874, + "step": 48423 + }, + { + "epoch": 1.77, + "learning_rate": 2.2851283553667558e-08, + "loss": 0.6824, + "step": 48424 + }, + { + "epoch": 1.77, + "learning_rate": 2.2847621488995494e-08, + "loss": 0.6906, + "step": 48425 + }, + { + "epoch": 1.77, + "learning_rate": 2.2843959424323434e-08, + "loss": 0.6933, + "step": 48426 + }, + { + "epoch": 1.77, + "learning_rate": 2.284029735965137e-08, + "loss": 0.7021, + "step": 48427 + }, + { + "epoch": 1.77, + "learning_rate": 2.2836635294979307e-08, + "loss": 0.6913, + "step": 48428 + }, + { + "epoch": 1.77, + "learning_rate": 2.2832973230307244e-08, + "loss": 0.6907, + "step": 48429 + }, + { + "epoch": 1.77, + "learning_rate": 2.2829311165635184e-08, + "loss": 0.7059, + "step": 48430 + }, + { + "epoch": 1.77, + "learning_rate": 2.282564910096312e-08, + "loss": 0.6965, + "step": 48431 + }, + { + "epoch": 1.77, + "learning_rate": 2.282198703629106e-08, + "loss": 0.7131, + "step": 48432 + }, + { + "epoch": 1.77, + "learning_rate": 2.2818324971619e-08, + "loss": 0.6999, + "step": 48433 + }, + { + "epoch": 1.77, + "learning_rate": 2.2814662906946937e-08, + "loss": 0.6957, + "step": 48434 + }, + { + "epoch": 1.77, + "learning_rate": 2.2811000842274874e-08, + "loss": 0.6879, + "step": 48435 + }, + { + "epoch": 1.77, + "learning_rate": 2.280733877760281e-08, + "loss": 0.6832, + "step": 48436 + }, + { + "epoch": 1.77, + "learning_rate": 2.280367671293075e-08, + "loss": 0.6882, + "step": 48437 + }, + { + "epoch": 1.77, + "learning_rate": 2.2800014648258687e-08, + "loss": 0.6974, + "step": 48438 + }, + { + "epoch": 1.77, + "learning_rate": 2.2796352583586627e-08, + "loss": 0.6927, + "step": 48439 + }, + { + "epoch": 1.77, + "learning_rate": 2.2792690518914563e-08, + "loss": 0.6742, + "step": 48440 + }, + { + "epoch": 1.77, + "learning_rate": 2.27890284542425e-08, + "loss": 0.7016, + "step": 48441 + }, + { + "epoch": 1.77, + "learning_rate": 2.278536638957044e-08, + "loss": 0.6984, + "step": 48442 + }, + { + "epoch": 1.77, + "learning_rate": 2.2781704324898376e-08, + "loss": 0.6926, + "step": 48443 + }, + { + "epoch": 1.77, + "learning_rate": 2.2778042260226316e-08, + "loss": 0.686, + "step": 48444 + }, + { + "epoch": 1.77, + "learning_rate": 2.2774380195554253e-08, + "loss": 0.6998, + "step": 48445 + }, + { + "epoch": 1.77, + "learning_rate": 2.277071813088219e-08, + "loss": 0.6901, + "step": 48446 + }, + { + "epoch": 1.77, + "learning_rate": 2.2767056066210126e-08, + "loss": 0.7016, + "step": 48447 + }, + { + "epoch": 1.77, + "learning_rate": 2.2763394001538066e-08, + "loss": 0.6851, + "step": 48448 + }, + { + "epoch": 1.77, + "learning_rate": 2.2759731936866003e-08, + "loss": 0.7043, + "step": 48449 + }, + { + "epoch": 1.77, + "learning_rate": 2.2756069872193943e-08, + "loss": 0.6963, + "step": 48450 + }, + { + "epoch": 1.77, + "learning_rate": 2.2752407807521883e-08, + "loss": 0.6982, + "step": 48451 + }, + { + "epoch": 1.77, + "learning_rate": 2.2748745742849816e-08, + "loss": 0.6955, + "step": 48452 + }, + { + "epoch": 1.77, + "learning_rate": 2.2745083678177756e-08, + "loss": 0.6891, + "step": 48453 + }, + { + "epoch": 1.77, + "learning_rate": 2.2741421613505693e-08, + "loss": 0.6929, + "step": 48454 + }, + { + "epoch": 1.77, + "learning_rate": 2.2737759548833632e-08, + "loss": 0.6958, + "step": 48455 + }, + { + "epoch": 1.77, + "learning_rate": 2.273409748416157e-08, + "loss": 0.6921, + "step": 48456 + }, + { + "epoch": 1.77, + "learning_rate": 2.273043541948951e-08, + "loss": 0.7016, + "step": 48457 + }, + { + "epoch": 1.77, + "learning_rate": 2.2726773354817442e-08, + "loss": 0.6965, + "step": 48458 + }, + { + "epoch": 1.77, + "learning_rate": 2.2723111290145382e-08, + "loss": 0.6875, + "step": 48459 + }, + { + "epoch": 1.77, + "learning_rate": 2.2719449225473322e-08, + "loss": 0.687, + "step": 48460 + }, + { + "epoch": 1.77, + "learning_rate": 2.271578716080126e-08, + "loss": 0.6957, + "step": 48461 + }, + { + "epoch": 1.77, + "learning_rate": 2.27121250961292e-08, + "loss": 0.697, + "step": 48462 + }, + { + "epoch": 1.77, + "learning_rate": 2.2708463031457135e-08, + "loss": 0.692, + "step": 48463 + }, + { + "epoch": 1.77, + "learning_rate": 2.2704800966785072e-08, + "loss": 0.6923, + "step": 48464 + }, + { + "epoch": 1.77, + "learning_rate": 2.270113890211301e-08, + "loss": 0.6863, + "step": 48465 + }, + { + "epoch": 1.77, + "learning_rate": 2.269747683744095e-08, + "loss": 0.6991, + "step": 48466 + }, + { + "epoch": 1.77, + "learning_rate": 2.2693814772768885e-08, + "loss": 0.6973, + "step": 48467 + }, + { + "epoch": 1.77, + "learning_rate": 2.2690152708096825e-08, + "loss": 0.6921, + "step": 48468 + }, + { + "epoch": 1.77, + "learning_rate": 2.2686490643424762e-08, + "loss": 0.6932, + "step": 48469 + }, + { + "epoch": 1.77, + "learning_rate": 2.2682828578752698e-08, + "loss": 0.6898, + "step": 48470 + }, + { + "epoch": 1.77, + "learning_rate": 2.2679166514080638e-08, + "loss": 0.6864, + "step": 48471 + }, + { + "epoch": 1.77, + "learning_rate": 2.2675504449408575e-08, + "loss": 0.6919, + "step": 48472 + }, + { + "epoch": 1.77, + "learning_rate": 2.2671842384736515e-08, + "loss": 0.6883, + "step": 48473 + }, + { + "epoch": 1.77, + "learning_rate": 2.266818032006445e-08, + "loss": 0.6964, + "step": 48474 + }, + { + "epoch": 1.77, + "learning_rate": 2.266451825539239e-08, + "loss": 0.6869, + "step": 48475 + }, + { + "epoch": 1.77, + "learning_rate": 2.2660856190720325e-08, + "loss": 0.7024, + "step": 48476 + }, + { + "epoch": 1.77, + "learning_rate": 2.2657194126048265e-08, + "loss": 0.7, + "step": 48477 + }, + { + "epoch": 1.77, + "learning_rate": 2.26535320613762e-08, + "loss": 0.7002, + "step": 48478 + }, + { + "epoch": 1.77, + "learning_rate": 2.264986999670414e-08, + "loss": 0.6887, + "step": 48479 + }, + { + "epoch": 1.77, + "learning_rate": 2.264620793203208e-08, + "loss": 0.6941, + "step": 48480 + }, + { + "epoch": 1.77, + "learning_rate": 2.2642545867360018e-08, + "loss": 0.6884, + "step": 48481 + }, + { + "epoch": 1.77, + "learning_rate": 2.2638883802687954e-08, + "loss": 0.6826, + "step": 48482 + }, + { + "epoch": 1.77, + "learning_rate": 2.263522173801589e-08, + "loss": 0.6799, + "step": 48483 + }, + { + "epoch": 1.77, + "learning_rate": 2.263155967334383e-08, + "loss": 0.6913, + "step": 48484 + }, + { + "epoch": 1.77, + "learning_rate": 2.2627897608671767e-08, + "loss": 0.6864, + "step": 48485 + }, + { + "epoch": 1.77, + "learning_rate": 2.2624235543999707e-08, + "loss": 0.6976, + "step": 48486 + }, + { + "epoch": 1.77, + "learning_rate": 2.2620573479327644e-08, + "loss": 0.6938, + "step": 48487 + }, + { + "epoch": 1.77, + "learning_rate": 2.261691141465558e-08, + "loss": 0.7031, + "step": 48488 + }, + { + "epoch": 1.77, + "learning_rate": 2.261324934998352e-08, + "loss": 0.6885, + "step": 48489 + }, + { + "epoch": 1.77, + "learning_rate": 2.2609587285311457e-08, + "loss": 0.6984, + "step": 48490 + }, + { + "epoch": 1.77, + "learning_rate": 2.2605925220639397e-08, + "loss": 0.7028, + "step": 48491 + }, + { + "epoch": 1.77, + "learning_rate": 2.2602263155967334e-08, + "loss": 0.6959, + "step": 48492 + }, + { + "epoch": 1.77, + "learning_rate": 2.259860109129527e-08, + "loss": 0.6817, + "step": 48493 + }, + { + "epoch": 1.77, + "learning_rate": 2.2594939026623207e-08, + "loss": 0.6887, + "step": 48494 + }, + { + "epoch": 1.77, + "learning_rate": 2.2591276961951147e-08, + "loss": 0.7011, + "step": 48495 + }, + { + "epoch": 1.77, + "learning_rate": 2.2587614897279084e-08, + "loss": 0.7029, + "step": 48496 + }, + { + "epoch": 1.77, + "learning_rate": 2.2583952832607023e-08, + "loss": 0.6816, + "step": 48497 + }, + { + "epoch": 1.77, + "learning_rate": 2.2580290767934963e-08, + "loss": 0.6892, + "step": 48498 + }, + { + "epoch": 1.77, + "learning_rate": 2.2576628703262897e-08, + "loss": 0.6921, + "step": 48499 + }, + { + "epoch": 1.77, + "learning_rate": 2.2572966638590837e-08, + "loss": 0.687, + "step": 48500 + }, + { + "epoch": 1.77, + "learning_rate": 2.2569304573918773e-08, + "loss": 0.6899, + "step": 48501 + }, + { + "epoch": 1.77, + "learning_rate": 2.2565642509246713e-08, + "loss": 0.6925, + "step": 48502 + }, + { + "epoch": 1.77, + "learning_rate": 2.256198044457465e-08, + "loss": 0.6995, + "step": 48503 + }, + { + "epoch": 1.77, + "learning_rate": 2.255831837990259e-08, + "loss": 0.6988, + "step": 48504 + }, + { + "epoch": 1.77, + "learning_rate": 2.2554656315230526e-08, + "loss": 0.6919, + "step": 48505 + }, + { + "epoch": 1.77, + "learning_rate": 2.2550994250558463e-08, + "loss": 0.701, + "step": 48506 + }, + { + "epoch": 1.77, + "learning_rate": 2.2547332185886403e-08, + "loss": 0.6944, + "step": 48507 + }, + { + "epoch": 1.77, + "learning_rate": 2.254367012121434e-08, + "loss": 0.6935, + "step": 48508 + }, + { + "epoch": 1.77, + "learning_rate": 2.254000805654228e-08, + "loss": 0.6819, + "step": 48509 + }, + { + "epoch": 1.77, + "learning_rate": 2.2536345991870216e-08, + "loss": 0.6892, + "step": 48510 + }, + { + "epoch": 1.77, + "learning_rate": 2.2532683927198153e-08, + "loss": 0.6867, + "step": 48511 + }, + { + "epoch": 1.77, + "learning_rate": 2.252902186252609e-08, + "loss": 0.6962, + "step": 48512 + }, + { + "epoch": 1.77, + "learning_rate": 2.252535979785403e-08, + "loss": 0.6813, + "step": 48513 + }, + { + "epoch": 1.77, + "learning_rate": 2.2521697733181966e-08, + "loss": 0.6965, + "step": 48514 + }, + { + "epoch": 1.78, + "learning_rate": 2.2518035668509906e-08, + "loss": 0.6932, + "step": 48515 + }, + { + "epoch": 1.78, + "learning_rate": 2.2514373603837842e-08, + "loss": 0.6912, + "step": 48516 + }, + { + "epoch": 1.78, + "learning_rate": 2.251071153916578e-08, + "loss": 0.7001, + "step": 48517 + }, + { + "epoch": 1.78, + "learning_rate": 2.250704947449372e-08, + "loss": 0.6979, + "step": 48518 + }, + { + "epoch": 1.78, + "learning_rate": 2.2503387409821656e-08, + "loss": 0.689, + "step": 48519 + }, + { + "epoch": 1.78, + "learning_rate": 2.2499725345149595e-08, + "loss": 0.6998, + "step": 48520 + }, + { + "epoch": 1.78, + "learning_rate": 2.2496063280477532e-08, + "loss": 0.6985, + "step": 48521 + }, + { + "epoch": 1.78, + "learning_rate": 2.2492401215805472e-08, + "loss": 0.7085, + "step": 48522 + }, + { + "epoch": 1.78, + "learning_rate": 2.2488739151133405e-08, + "loss": 0.7, + "step": 48523 + }, + { + "epoch": 1.78, + "learning_rate": 2.2485077086461345e-08, + "loss": 0.6859, + "step": 48524 + }, + { + "epoch": 1.78, + "learning_rate": 2.2481415021789285e-08, + "loss": 0.6838, + "step": 48525 + }, + { + "epoch": 1.78, + "learning_rate": 2.2477752957117222e-08, + "loss": 0.6921, + "step": 48526 + }, + { + "epoch": 1.78, + "learning_rate": 2.2474090892445162e-08, + "loss": 0.7012, + "step": 48527 + }, + { + "epoch": 1.78, + "learning_rate": 2.24704288277731e-08, + "loss": 0.6887, + "step": 48528 + }, + { + "epoch": 1.78, + "learning_rate": 2.2466766763101035e-08, + "loss": 0.6947, + "step": 48529 + }, + { + "epoch": 1.78, + "learning_rate": 2.246310469842897e-08, + "loss": 0.6927, + "step": 48530 + }, + { + "epoch": 1.78, + "learning_rate": 2.245944263375691e-08, + "loss": 0.7116, + "step": 48531 + }, + { + "epoch": 1.78, + "learning_rate": 2.2455780569084848e-08, + "loss": 0.7034, + "step": 48532 + }, + { + "epoch": 1.78, + "learning_rate": 2.2452118504412788e-08, + "loss": 0.6946, + "step": 48533 + }, + { + "epoch": 1.78, + "learning_rate": 2.2448456439740725e-08, + "loss": 0.691, + "step": 48534 + }, + { + "epoch": 1.78, + "learning_rate": 2.244479437506866e-08, + "loss": 0.6895, + "step": 48535 + }, + { + "epoch": 1.78, + "learning_rate": 2.24411323103966e-08, + "loss": 0.7022, + "step": 48536 + }, + { + "epoch": 1.78, + "learning_rate": 2.2437470245724538e-08, + "loss": 0.7037, + "step": 48537 + }, + { + "epoch": 1.78, + "learning_rate": 2.2433808181052478e-08, + "loss": 0.6831, + "step": 48538 + }, + { + "epoch": 1.78, + "learning_rate": 2.2430146116380414e-08, + "loss": 0.6885, + "step": 48539 + }, + { + "epoch": 1.78, + "learning_rate": 2.242648405170835e-08, + "loss": 0.7043, + "step": 48540 + }, + { + "epoch": 1.78, + "learning_rate": 2.2422821987036288e-08, + "loss": 0.6893, + "step": 48541 + }, + { + "epoch": 1.78, + "learning_rate": 2.2419159922364228e-08, + "loss": 0.6885, + "step": 48542 + }, + { + "epoch": 1.78, + "learning_rate": 2.2415497857692168e-08, + "loss": 0.6904, + "step": 48543 + }, + { + "epoch": 1.78, + "learning_rate": 2.2411835793020104e-08, + "loss": 0.697, + "step": 48544 + }, + { + "epoch": 1.78, + "learning_rate": 2.2408173728348044e-08, + "loss": 0.689, + "step": 48545 + }, + { + "epoch": 1.78, + "learning_rate": 2.2404511663675977e-08, + "loss": 0.7126, + "step": 48546 + }, + { + "epoch": 1.78, + "learning_rate": 2.2400849599003917e-08, + "loss": 0.6903, + "step": 48547 + }, + { + "epoch": 1.78, + "learning_rate": 2.2397187534331854e-08, + "loss": 0.6938, + "step": 48548 + }, + { + "epoch": 1.78, + "learning_rate": 2.2393525469659794e-08, + "loss": 0.6939, + "step": 48549 + }, + { + "epoch": 1.78, + "learning_rate": 2.238986340498773e-08, + "loss": 0.7009, + "step": 48550 + }, + { + "epoch": 1.78, + "learning_rate": 2.238620134031567e-08, + "loss": 0.6964, + "step": 48551 + }, + { + "epoch": 1.78, + "learning_rate": 2.2382539275643607e-08, + "loss": 0.7025, + "step": 48552 + }, + { + "epoch": 1.78, + "learning_rate": 2.2378877210971544e-08, + "loss": 0.7011, + "step": 48553 + }, + { + "epoch": 1.78, + "learning_rate": 2.2375215146299484e-08, + "loss": 0.7004, + "step": 48554 + }, + { + "epoch": 1.78, + "learning_rate": 2.237155308162742e-08, + "loss": 0.6922, + "step": 48555 + }, + { + "epoch": 1.78, + "learning_rate": 2.236789101695536e-08, + "loss": 0.6896, + "step": 48556 + }, + { + "epoch": 1.78, + "learning_rate": 2.2364228952283297e-08, + "loss": 0.686, + "step": 48557 + }, + { + "epoch": 1.78, + "learning_rate": 2.2360566887611233e-08, + "loss": 0.6939, + "step": 48558 + }, + { + "epoch": 1.78, + "learning_rate": 2.235690482293917e-08, + "loss": 0.682, + "step": 48559 + }, + { + "epoch": 1.78, + "learning_rate": 2.235324275826711e-08, + "loss": 0.7067, + "step": 48560 + }, + { + "epoch": 1.78, + "learning_rate": 2.2349580693595047e-08, + "loss": 0.6912, + "step": 48561 + }, + { + "epoch": 1.78, + "learning_rate": 2.2345918628922986e-08, + "loss": 0.6942, + "step": 48562 + }, + { + "epoch": 1.78, + "learning_rate": 2.2342256564250926e-08, + "loss": 0.7064, + "step": 48563 + }, + { + "epoch": 1.78, + "learning_rate": 2.233859449957886e-08, + "loss": 0.6934, + "step": 48564 + }, + { + "epoch": 1.78, + "learning_rate": 2.23349324349068e-08, + "loss": 0.6889, + "step": 48565 + }, + { + "epoch": 1.78, + "learning_rate": 2.2331270370234736e-08, + "loss": 0.704, + "step": 48566 + }, + { + "epoch": 1.78, + "learning_rate": 2.2327608305562676e-08, + "loss": 0.696, + "step": 48567 + }, + { + "epoch": 1.78, + "learning_rate": 2.2323946240890613e-08, + "loss": 0.6845, + "step": 48568 + }, + { + "epoch": 1.78, + "learning_rate": 2.2320284176218553e-08, + "loss": 0.6936, + "step": 48569 + }, + { + "epoch": 1.78, + "learning_rate": 2.2316622111546486e-08, + "loss": 0.6832, + "step": 48570 + }, + { + "epoch": 1.78, + "learning_rate": 2.2312960046874426e-08, + "loss": 0.6913, + "step": 48571 + }, + { + "epoch": 1.78, + "learning_rate": 2.2309297982202366e-08, + "loss": 0.6847, + "step": 48572 + }, + { + "epoch": 1.78, + "learning_rate": 2.2305635917530302e-08, + "loss": 0.709, + "step": 48573 + }, + { + "epoch": 1.78, + "learning_rate": 2.2301973852858242e-08, + "loss": 0.698, + "step": 48574 + }, + { + "epoch": 1.78, + "learning_rate": 2.229831178818618e-08, + "loss": 0.6912, + "step": 48575 + }, + { + "epoch": 1.78, + "learning_rate": 2.2294649723514116e-08, + "loss": 0.6968, + "step": 48576 + }, + { + "epoch": 1.78, + "learning_rate": 2.2290987658842052e-08, + "loss": 0.7044, + "step": 48577 + }, + { + "epoch": 1.78, + "learning_rate": 2.2287325594169992e-08, + "loss": 0.7007, + "step": 48578 + }, + { + "epoch": 1.78, + "learning_rate": 2.228366352949793e-08, + "loss": 0.6921, + "step": 48579 + }, + { + "epoch": 1.78, + "learning_rate": 2.228000146482587e-08, + "loss": 0.6788, + "step": 48580 + }, + { + "epoch": 1.78, + "learning_rate": 2.2276339400153805e-08, + "loss": 0.6978, + "step": 48581 + }, + { + "epoch": 1.78, + "learning_rate": 2.2272677335481742e-08, + "loss": 0.6811, + "step": 48582 + }, + { + "epoch": 1.78, + "learning_rate": 2.2269015270809682e-08, + "loss": 0.6963, + "step": 48583 + }, + { + "epoch": 1.78, + "learning_rate": 2.226535320613762e-08, + "loss": 0.6966, + "step": 48584 + }, + { + "epoch": 1.78, + "learning_rate": 2.226169114146556e-08, + "loss": 0.6911, + "step": 48585 + }, + { + "epoch": 1.78, + "learning_rate": 2.2258029076793495e-08, + "loss": 0.6997, + "step": 48586 + }, + { + "epoch": 1.78, + "learning_rate": 2.2254367012121432e-08, + "loss": 0.6928, + "step": 48587 + }, + { + "epoch": 1.78, + "learning_rate": 2.2250704947449368e-08, + "loss": 0.6993, + "step": 48588 + }, + { + "epoch": 1.78, + "learning_rate": 2.2247042882777308e-08, + "loss": 0.6845, + "step": 48589 + }, + { + "epoch": 1.78, + "learning_rate": 2.2243380818105248e-08, + "loss": 0.7035, + "step": 48590 + }, + { + "epoch": 1.78, + "learning_rate": 2.2239718753433185e-08, + "loss": 0.6912, + "step": 48591 + }, + { + "epoch": 1.78, + "learning_rate": 2.2236056688761125e-08, + "loss": 0.6994, + "step": 48592 + }, + { + "epoch": 1.78, + "learning_rate": 2.223239462408906e-08, + "loss": 0.6825, + "step": 48593 + }, + { + "epoch": 1.78, + "learning_rate": 2.2228732559416998e-08, + "loss": 0.693, + "step": 48594 + }, + { + "epoch": 1.78, + "learning_rate": 2.2225070494744935e-08, + "loss": 0.6854, + "step": 48595 + }, + { + "epoch": 1.78, + "learning_rate": 2.2221408430072875e-08, + "loss": 0.6887, + "step": 48596 + }, + { + "epoch": 1.78, + "learning_rate": 2.221774636540081e-08, + "loss": 0.7044, + "step": 48597 + }, + { + "epoch": 1.78, + "learning_rate": 2.221408430072875e-08, + "loss": 0.6927, + "step": 48598 + }, + { + "epoch": 1.78, + "learning_rate": 2.2210422236056688e-08, + "loss": 0.6943, + "step": 48599 + }, + { + "epoch": 1.78, + "learning_rate": 2.2206760171384624e-08, + "loss": 0.6931, + "step": 48600 + }, + { + "epoch": 1.78, + "learning_rate": 2.2203098106712564e-08, + "loss": 0.6943, + "step": 48601 + }, + { + "epoch": 1.78, + "learning_rate": 2.21994360420405e-08, + "loss": 0.699, + "step": 48602 + }, + { + "epoch": 1.78, + "learning_rate": 2.219577397736844e-08, + "loss": 0.6935, + "step": 48603 + }, + { + "epoch": 1.78, + "learning_rate": 2.2192111912696377e-08, + "loss": 0.6984, + "step": 48604 + }, + { + "epoch": 1.78, + "learning_rate": 2.2188449848024314e-08, + "loss": 0.6914, + "step": 48605 + }, + { + "epoch": 1.78, + "learning_rate": 2.218478778335225e-08, + "loss": 0.6738, + "step": 48606 + }, + { + "epoch": 1.78, + "learning_rate": 2.218112571868019e-08, + "loss": 0.697, + "step": 48607 + }, + { + "epoch": 1.78, + "learning_rate": 2.217746365400813e-08, + "loss": 0.6919, + "step": 48608 + }, + { + "epoch": 1.78, + "learning_rate": 2.2173801589336067e-08, + "loss": 0.6962, + "step": 48609 + }, + { + "epoch": 1.78, + "learning_rate": 2.2170139524664007e-08, + "loss": 0.6862, + "step": 48610 + }, + { + "epoch": 1.78, + "learning_rate": 2.216647745999194e-08, + "loss": 0.6944, + "step": 48611 + }, + { + "epoch": 1.78, + "learning_rate": 2.216281539531988e-08, + "loss": 0.6822, + "step": 48612 + }, + { + "epoch": 1.78, + "learning_rate": 2.2159153330647817e-08, + "loss": 0.6738, + "step": 48613 + }, + { + "epoch": 1.78, + "learning_rate": 2.2155491265975757e-08, + "loss": 0.6916, + "step": 48614 + }, + { + "epoch": 1.78, + "learning_rate": 2.2151829201303693e-08, + "loss": 0.6896, + "step": 48615 + }, + { + "epoch": 1.78, + "learning_rate": 2.2148167136631633e-08, + "loss": 0.7125, + "step": 48616 + }, + { + "epoch": 1.78, + "learning_rate": 2.214450507195957e-08, + "loss": 0.701, + "step": 48617 + }, + { + "epoch": 1.78, + "learning_rate": 2.2140843007287507e-08, + "loss": 0.7021, + "step": 48618 + }, + { + "epoch": 1.78, + "learning_rate": 2.2137180942615447e-08, + "loss": 0.6979, + "step": 48619 + }, + { + "epoch": 1.78, + "learning_rate": 2.2133518877943383e-08, + "loss": 0.6862, + "step": 48620 + }, + { + "epoch": 1.78, + "learning_rate": 2.2129856813271323e-08, + "loss": 0.702, + "step": 48621 + }, + { + "epoch": 1.78, + "learning_rate": 2.212619474859926e-08, + "loss": 0.69, + "step": 48622 + }, + { + "epoch": 1.78, + "learning_rate": 2.2122532683927196e-08, + "loss": 0.6919, + "step": 48623 + }, + { + "epoch": 1.78, + "learning_rate": 2.2118870619255133e-08, + "loss": 0.6853, + "step": 48624 + }, + { + "epoch": 1.78, + "learning_rate": 2.2115208554583073e-08, + "loss": 0.6976, + "step": 48625 + }, + { + "epoch": 1.78, + "learning_rate": 2.2111546489911013e-08, + "loss": 0.6926, + "step": 48626 + }, + { + "epoch": 1.78, + "learning_rate": 2.210788442523895e-08, + "loss": 0.6917, + "step": 48627 + }, + { + "epoch": 1.78, + "learning_rate": 2.2104222360566886e-08, + "loss": 0.6898, + "step": 48628 + }, + { + "epoch": 1.78, + "learning_rate": 2.2100560295894823e-08, + "loss": 0.6864, + "step": 48629 + }, + { + "epoch": 1.78, + "learning_rate": 2.2096898231222763e-08, + "loss": 0.6926, + "step": 48630 + }, + { + "epoch": 1.78, + "learning_rate": 2.20932361665507e-08, + "loss": 0.6966, + "step": 48631 + }, + { + "epoch": 1.78, + "learning_rate": 2.208957410187864e-08, + "loss": 0.6826, + "step": 48632 + }, + { + "epoch": 1.78, + "learning_rate": 2.2085912037206576e-08, + "loss": 0.688, + "step": 48633 + }, + { + "epoch": 1.78, + "learning_rate": 2.2082249972534512e-08, + "loss": 0.6904, + "step": 48634 + }, + { + "epoch": 1.78, + "learning_rate": 2.2078587907862452e-08, + "loss": 0.6897, + "step": 48635 + }, + { + "epoch": 1.78, + "learning_rate": 2.207492584319039e-08, + "loss": 0.7031, + "step": 48636 + }, + { + "epoch": 1.78, + "learning_rate": 2.207126377851833e-08, + "loss": 0.6852, + "step": 48637 + }, + { + "epoch": 1.78, + "learning_rate": 2.2067601713846266e-08, + "loss": 0.6951, + "step": 48638 + }, + { + "epoch": 1.78, + "learning_rate": 2.2063939649174205e-08, + "loss": 0.6926, + "step": 48639 + }, + { + "epoch": 1.78, + "learning_rate": 2.2060277584502142e-08, + "loss": 0.6987, + "step": 48640 + }, + { + "epoch": 1.78, + "learning_rate": 2.205661551983008e-08, + "loss": 0.6955, + "step": 48641 + }, + { + "epoch": 1.78, + "learning_rate": 2.2052953455158015e-08, + "loss": 0.6916, + "step": 48642 + }, + { + "epoch": 1.78, + "learning_rate": 2.2049291390485955e-08, + "loss": 0.6956, + "step": 48643 + }, + { + "epoch": 1.78, + "learning_rate": 2.2045629325813895e-08, + "loss": 0.6918, + "step": 48644 + }, + { + "epoch": 1.78, + "learning_rate": 2.2041967261141832e-08, + "loss": 0.6892, + "step": 48645 + }, + { + "epoch": 1.78, + "learning_rate": 2.203830519646977e-08, + "loss": 0.6914, + "step": 48646 + }, + { + "epoch": 1.78, + "learning_rate": 2.2034643131797705e-08, + "loss": 0.6828, + "step": 48647 + }, + { + "epoch": 1.78, + "learning_rate": 2.2030981067125645e-08, + "loss": 0.6911, + "step": 48648 + }, + { + "epoch": 1.78, + "learning_rate": 2.202731900245358e-08, + "loss": 0.6913, + "step": 48649 + }, + { + "epoch": 1.78, + "learning_rate": 2.202365693778152e-08, + "loss": 0.7027, + "step": 48650 + }, + { + "epoch": 1.78, + "learning_rate": 2.2019994873109458e-08, + "loss": 0.6873, + "step": 48651 + }, + { + "epoch": 1.78, + "learning_rate": 2.2016332808437395e-08, + "loss": 0.7068, + "step": 48652 + }, + { + "epoch": 1.78, + "learning_rate": 2.2012670743765335e-08, + "loss": 0.7086, + "step": 48653 + }, + { + "epoch": 1.78, + "learning_rate": 2.200900867909327e-08, + "loss": 0.6928, + "step": 48654 + }, + { + "epoch": 1.78, + "learning_rate": 2.200534661442121e-08, + "loss": 0.6927, + "step": 48655 + }, + { + "epoch": 1.78, + "learning_rate": 2.2001684549749148e-08, + "loss": 0.6946, + "step": 48656 + }, + { + "epoch": 1.78, + "learning_rate": 2.1998022485077088e-08, + "loss": 0.7008, + "step": 48657 + }, + { + "epoch": 1.78, + "learning_rate": 2.199436042040502e-08, + "loss": 0.6941, + "step": 48658 + }, + { + "epoch": 1.78, + "learning_rate": 2.199069835573296e-08, + "loss": 0.6947, + "step": 48659 + }, + { + "epoch": 1.78, + "learning_rate": 2.1987036291060898e-08, + "loss": 0.6933, + "step": 48660 + }, + { + "epoch": 1.78, + "learning_rate": 2.1983374226388838e-08, + "loss": 0.6983, + "step": 48661 + }, + { + "epoch": 1.78, + "learning_rate": 2.1979712161716774e-08, + "loss": 0.6915, + "step": 48662 + }, + { + "epoch": 1.78, + "learning_rate": 2.1976050097044714e-08, + "loss": 0.6853, + "step": 48663 + }, + { + "epoch": 1.78, + "learning_rate": 2.197238803237265e-08, + "loss": 0.6965, + "step": 48664 + }, + { + "epoch": 1.78, + "learning_rate": 2.1968725967700587e-08, + "loss": 0.6935, + "step": 48665 + }, + { + "epoch": 1.78, + "learning_rate": 2.1965063903028527e-08, + "loss": 0.6801, + "step": 48666 + }, + { + "epoch": 1.78, + "learning_rate": 2.1961401838356464e-08, + "loss": 0.6985, + "step": 48667 + }, + { + "epoch": 1.78, + "learning_rate": 2.1957739773684404e-08, + "loss": 0.6831, + "step": 48668 + }, + { + "epoch": 1.78, + "learning_rate": 2.195407770901234e-08, + "loss": 0.7017, + "step": 48669 + }, + { + "epoch": 1.78, + "learning_rate": 2.1950415644340277e-08, + "loss": 0.7033, + "step": 48670 + }, + { + "epoch": 1.78, + "learning_rate": 2.1946753579668214e-08, + "loss": 0.7001, + "step": 48671 + }, + { + "epoch": 1.78, + "learning_rate": 2.1943091514996154e-08, + "loss": 0.69, + "step": 48672 + }, + { + "epoch": 1.78, + "learning_rate": 2.1939429450324094e-08, + "loss": 0.6907, + "step": 48673 + }, + { + "epoch": 1.78, + "learning_rate": 2.193576738565203e-08, + "loss": 0.6934, + "step": 48674 + }, + { + "epoch": 1.78, + "learning_rate": 2.1932105320979967e-08, + "loss": 0.7066, + "step": 48675 + }, + { + "epoch": 1.78, + "learning_rate": 2.1928443256307903e-08, + "loss": 0.6968, + "step": 48676 + }, + { + "epoch": 1.78, + "learning_rate": 2.1924781191635843e-08, + "loss": 0.6941, + "step": 48677 + }, + { + "epoch": 1.78, + "learning_rate": 2.192111912696378e-08, + "loss": 0.6843, + "step": 48678 + }, + { + "epoch": 1.78, + "learning_rate": 2.191745706229172e-08, + "loss": 0.7041, + "step": 48679 + }, + { + "epoch": 1.78, + "learning_rate": 2.1913794997619656e-08, + "loss": 0.6879, + "step": 48680 + }, + { + "epoch": 1.78, + "learning_rate": 2.1910132932947596e-08, + "loss": 0.6899, + "step": 48681 + }, + { + "epoch": 1.78, + "learning_rate": 2.1906470868275533e-08, + "loss": 0.6906, + "step": 48682 + }, + { + "epoch": 1.78, + "learning_rate": 2.190280880360347e-08, + "loss": 0.6911, + "step": 48683 + }, + { + "epoch": 1.78, + "learning_rate": 2.189914673893141e-08, + "loss": 0.6952, + "step": 48684 + }, + { + "epoch": 1.78, + "learning_rate": 2.1895484674259346e-08, + "loss": 0.7016, + "step": 48685 + }, + { + "epoch": 1.78, + "learning_rate": 2.1891822609587286e-08, + "loss": 0.6987, + "step": 48686 + }, + { + "epoch": 1.78, + "learning_rate": 2.1888160544915223e-08, + "loss": 0.7014, + "step": 48687 + }, + { + "epoch": 1.78, + "learning_rate": 2.188449848024316e-08, + "loss": 0.6756, + "step": 48688 + }, + { + "epoch": 1.78, + "learning_rate": 2.1880836415571096e-08, + "loss": 0.7043, + "step": 48689 + }, + { + "epoch": 1.78, + "learning_rate": 2.1877174350899036e-08, + "loss": 0.7051, + "step": 48690 + }, + { + "epoch": 1.78, + "learning_rate": 2.1873512286226976e-08, + "loss": 0.7048, + "step": 48691 + }, + { + "epoch": 1.78, + "learning_rate": 2.1869850221554912e-08, + "loss": 0.6897, + "step": 48692 + }, + { + "epoch": 1.78, + "learning_rate": 2.186618815688285e-08, + "loss": 0.6952, + "step": 48693 + }, + { + "epoch": 1.78, + "learning_rate": 2.1862526092210786e-08, + "loss": 0.7059, + "step": 48694 + }, + { + "epoch": 1.78, + "learning_rate": 2.1858864027538726e-08, + "loss": 0.7018, + "step": 48695 + }, + { + "epoch": 1.78, + "learning_rate": 2.1855201962866662e-08, + "loss": 0.6933, + "step": 48696 + }, + { + "epoch": 1.78, + "learning_rate": 2.1851539898194602e-08, + "loss": 0.6787, + "step": 48697 + }, + { + "epoch": 1.78, + "learning_rate": 2.184787783352254e-08, + "loss": 0.6785, + "step": 48698 + }, + { + "epoch": 1.78, + "learning_rate": 2.1844215768850475e-08, + "loss": 0.6996, + "step": 48699 + }, + { + "epoch": 1.78, + "learning_rate": 2.1840553704178415e-08, + "loss": 0.7018, + "step": 48700 + }, + { + "epoch": 1.78, + "learning_rate": 2.1836891639506352e-08, + "loss": 0.6878, + "step": 48701 + }, + { + "epoch": 1.78, + "learning_rate": 2.1833229574834292e-08, + "loss": 0.6934, + "step": 48702 + }, + { + "epoch": 1.78, + "learning_rate": 2.182956751016223e-08, + "loss": 0.6912, + "step": 48703 + }, + { + "epoch": 1.78, + "learning_rate": 2.182590544549017e-08, + "loss": 0.6955, + "step": 48704 + }, + { + "epoch": 1.78, + "learning_rate": 2.1822243380818102e-08, + "loss": 0.6863, + "step": 48705 + }, + { + "epoch": 1.78, + "learning_rate": 2.1818581316146042e-08, + "loss": 0.6904, + "step": 48706 + }, + { + "epoch": 1.78, + "learning_rate": 2.1814919251473978e-08, + "loss": 0.6808, + "step": 48707 + }, + { + "epoch": 1.78, + "learning_rate": 2.1811257186801918e-08, + "loss": 0.6945, + "step": 48708 + }, + { + "epoch": 1.78, + "learning_rate": 2.1807595122129858e-08, + "loss": 0.6972, + "step": 48709 + }, + { + "epoch": 1.78, + "learning_rate": 2.1803933057457795e-08, + "loss": 0.6942, + "step": 48710 + }, + { + "epoch": 1.78, + "learning_rate": 2.180027099278573e-08, + "loss": 0.6979, + "step": 48711 + }, + { + "epoch": 1.78, + "learning_rate": 2.1796608928113668e-08, + "loss": 0.6952, + "step": 48712 + }, + { + "epoch": 1.78, + "learning_rate": 2.1792946863441608e-08, + "loss": 0.6896, + "step": 48713 + }, + { + "epoch": 1.78, + "learning_rate": 2.1789284798769545e-08, + "loss": 0.6873, + "step": 48714 + }, + { + "epoch": 1.78, + "learning_rate": 2.1785622734097485e-08, + "loss": 0.694, + "step": 48715 + }, + { + "epoch": 1.78, + "learning_rate": 2.178196066942542e-08, + "loss": 0.6979, + "step": 48716 + }, + { + "epoch": 1.78, + "learning_rate": 2.1778298604753358e-08, + "loss": 0.6932, + "step": 48717 + }, + { + "epoch": 1.78, + "learning_rate": 2.1774636540081298e-08, + "loss": 0.703, + "step": 48718 + }, + { + "epoch": 1.78, + "learning_rate": 2.1770974475409234e-08, + "loss": 0.6823, + "step": 48719 + }, + { + "epoch": 1.78, + "learning_rate": 2.1767312410737174e-08, + "loss": 0.6858, + "step": 48720 + }, + { + "epoch": 1.78, + "learning_rate": 2.176365034606511e-08, + "loss": 0.689, + "step": 48721 + }, + { + "epoch": 1.78, + "learning_rate": 2.1759988281393047e-08, + "loss": 0.6842, + "step": 48722 + }, + { + "epoch": 1.78, + "learning_rate": 2.1756326216720984e-08, + "loss": 0.6925, + "step": 48723 + }, + { + "epoch": 1.78, + "learning_rate": 2.1752664152048924e-08, + "loss": 0.6999, + "step": 48724 + }, + { + "epoch": 1.78, + "learning_rate": 2.174900208737686e-08, + "loss": 0.7002, + "step": 48725 + }, + { + "epoch": 1.78, + "learning_rate": 2.17453400227048e-08, + "loss": 0.6846, + "step": 48726 + }, + { + "epoch": 1.78, + "learning_rate": 2.174167795803274e-08, + "loss": 0.6949, + "step": 48727 + }, + { + "epoch": 1.78, + "learning_rate": 2.1738015893360677e-08, + "loss": 0.6948, + "step": 48728 + }, + { + "epoch": 1.78, + "learning_rate": 2.1734353828688614e-08, + "loss": 0.6937, + "step": 48729 + }, + { + "epoch": 1.78, + "learning_rate": 2.173069176401655e-08, + "loss": 0.6756, + "step": 48730 + }, + { + "epoch": 1.78, + "learning_rate": 2.172702969934449e-08, + "loss": 0.7009, + "step": 48731 + }, + { + "epoch": 1.78, + "learning_rate": 2.1723367634672427e-08, + "loss": 0.6948, + "step": 48732 + }, + { + "epoch": 1.78, + "learning_rate": 2.1719705570000367e-08, + "loss": 0.688, + "step": 48733 + }, + { + "epoch": 1.78, + "learning_rate": 2.1716043505328303e-08, + "loss": 0.6841, + "step": 48734 + }, + { + "epoch": 1.78, + "learning_rate": 2.171238144065624e-08, + "loss": 0.694, + "step": 48735 + }, + { + "epoch": 1.78, + "learning_rate": 2.170871937598418e-08, + "loss": 0.6964, + "step": 48736 + }, + { + "epoch": 1.78, + "learning_rate": 2.1705057311312117e-08, + "loss": 0.6907, + "step": 48737 + }, + { + "epoch": 1.78, + "learning_rate": 2.1701395246640057e-08, + "loss": 0.6873, + "step": 48738 + }, + { + "epoch": 1.78, + "learning_rate": 2.1697733181967993e-08, + "loss": 0.6884, + "step": 48739 + }, + { + "epoch": 1.78, + "learning_rate": 2.169407111729593e-08, + "loss": 0.6872, + "step": 48740 + }, + { + "epoch": 1.78, + "learning_rate": 2.1690409052623866e-08, + "loss": 0.7091, + "step": 48741 + }, + { + "epoch": 1.78, + "learning_rate": 2.1686746987951806e-08, + "loss": 0.69, + "step": 48742 + }, + { + "epoch": 1.78, + "learning_rate": 2.1683084923279743e-08, + "loss": 0.6903, + "step": 48743 + }, + { + "epoch": 1.78, + "learning_rate": 2.1679422858607683e-08, + "loss": 0.703, + "step": 48744 + }, + { + "epoch": 1.78, + "learning_rate": 2.1675760793935623e-08, + "loss": 0.6911, + "step": 48745 + }, + { + "epoch": 1.78, + "learning_rate": 2.1672098729263556e-08, + "loss": 0.7001, + "step": 48746 + }, + { + "epoch": 1.78, + "learning_rate": 2.1668436664591496e-08, + "loss": 0.6955, + "step": 48747 + }, + { + "epoch": 1.78, + "learning_rate": 2.1664774599919433e-08, + "loss": 0.6908, + "step": 48748 + }, + { + "epoch": 1.78, + "learning_rate": 2.1661112535247373e-08, + "loss": 0.6818, + "step": 48749 + }, + { + "epoch": 1.78, + "learning_rate": 2.165745047057531e-08, + "loss": 0.6945, + "step": 48750 + }, + { + "epoch": 1.78, + "learning_rate": 2.165378840590325e-08, + "loss": 0.6975, + "step": 48751 + }, + { + "epoch": 1.78, + "learning_rate": 2.1650126341231182e-08, + "loss": 0.6947, + "step": 48752 + }, + { + "epoch": 1.78, + "learning_rate": 2.1646464276559122e-08, + "loss": 0.693, + "step": 48753 + }, + { + "epoch": 1.78, + "learning_rate": 2.164280221188706e-08, + "loss": 0.6941, + "step": 48754 + }, + { + "epoch": 1.78, + "learning_rate": 2.1639140147215e-08, + "loss": 0.6881, + "step": 48755 + }, + { + "epoch": 1.78, + "learning_rate": 2.163547808254294e-08, + "loss": 0.6907, + "step": 48756 + }, + { + "epoch": 1.78, + "learning_rate": 2.1631816017870875e-08, + "loss": 0.6909, + "step": 48757 + }, + { + "epoch": 1.78, + "learning_rate": 2.1628153953198812e-08, + "loss": 0.6906, + "step": 48758 + }, + { + "epoch": 1.78, + "learning_rate": 2.162449188852675e-08, + "loss": 0.6987, + "step": 48759 + }, + { + "epoch": 1.78, + "learning_rate": 2.162082982385469e-08, + "loss": 0.6884, + "step": 48760 + }, + { + "epoch": 1.78, + "learning_rate": 2.1617167759182625e-08, + "loss": 0.7012, + "step": 48761 + }, + { + "epoch": 1.78, + "learning_rate": 2.1613505694510565e-08, + "loss": 0.6921, + "step": 48762 + }, + { + "epoch": 1.78, + "learning_rate": 2.1609843629838502e-08, + "loss": 0.6984, + "step": 48763 + }, + { + "epoch": 1.78, + "learning_rate": 2.160618156516644e-08, + "loss": 0.6819, + "step": 48764 + }, + { + "epoch": 1.78, + "learning_rate": 2.160251950049438e-08, + "loss": 0.7098, + "step": 48765 + }, + { + "epoch": 1.78, + "learning_rate": 2.1598857435822315e-08, + "loss": 0.6801, + "step": 48766 + }, + { + "epoch": 1.78, + "learning_rate": 2.1595195371150255e-08, + "loss": 0.7021, + "step": 48767 + }, + { + "epoch": 1.78, + "learning_rate": 2.159153330647819e-08, + "loss": 0.6908, + "step": 48768 + }, + { + "epoch": 1.78, + "learning_rate": 2.158787124180613e-08, + "loss": 0.6959, + "step": 48769 + }, + { + "epoch": 1.78, + "learning_rate": 2.1584209177134065e-08, + "loss": 0.7013, + "step": 48770 + }, + { + "epoch": 1.78, + "learning_rate": 2.1580547112462005e-08, + "loss": 0.6989, + "step": 48771 + }, + { + "epoch": 1.78, + "learning_rate": 2.157688504778994e-08, + "loss": 0.6884, + "step": 48772 + }, + { + "epoch": 1.78, + "learning_rate": 2.157322298311788e-08, + "loss": 0.6914, + "step": 48773 + }, + { + "epoch": 1.78, + "learning_rate": 2.156956091844582e-08, + "loss": 0.6831, + "step": 48774 + }, + { + "epoch": 1.78, + "learning_rate": 2.1565898853773758e-08, + "loss": 0.6928, + "step": 48775 + }, + { + "epoch": 1.78, + "learning_rate": 2.1562236789101694e-08, + "loss": 0.6987, + "step": 48776 + }, + { + "epoch": 1.78, + "learning_rate": 2.155857472442963e-08, + "loss": 0.6852, + "step": 48777 + }, + { + "epoch": 1.78, + "learning_rate": 2.155491265975757e-08, + "loss": 0.6989, + "step": 48778 + }, + { + "epoch": 1.78, + "learning_rate": 2.1551250595085508e-08, + "loss": 0.6852, + "step": 48779 + }, + { + "epoch": 1.78, + "learning_rate": 2.1547588530413448e-08, + "loss": 0.6916, + "step": 48780 + }, + { + "epoch": 1.78, + "learning_rate": 2.1543926465741384e-08, + "loss": 0.7004, + "step": 48781 + }, + { + "epoch": 1.78, + "learning_rate": 2.154026440106932e-08, + "loss": 0.7062, + "step": 48782 + }, + { + "epoch": 1.78, + "learning_rate": 2.153660233639726e-08, + "loss": 0.6866, + "step": 48783 + }, + { + "epoch": 1.78, + "learning_rate": 2.1532940271725197e-08, + "loss": 0.6926, + "step": 48784 + }, + { + "epoch": 1.78, + "learning_rate": 2.1529278207053137e-08, + "loss": 0.6949, + "step": 48785 + }, + { + "epoch": 1.78, + "learning_rate": 2.1525616142381074e-08, + "loss": 0.6929, + "step": 48786 + }, + { + "epoch": 1.78, + "learning_rate": 2.152195407770901e-08, + "loss": 0.6907, + "step": 48787 + }, + { + "epoch": 1.79, + "learning_rate": 2.1518292013036947e-08, + "loss": 0.6927, + "step": 48788 + }, + { + "epoch": 1.79, + "learning_rate": 2.1514629948364887e-08, + "loss": 0.6951, + "step": 48789 + }, + { + "epoch": 1.79, + "learning_rate": 2.1510967883692824e-08, + "loss": 0.6981, + "step": 48790 + }, + { + "epoch": 1.79, + "learning_rate": 2.1507305819020764e-08, + "loss": 0.6933, + "step": 48791 + }, + { + "epoch": 1.79, + "learning_rate": 2.1503643754348703e-08, + "loss": 0.6863, + "step": 48792 + }, + { + "epoch": 1.79, + "learning_rate": 2.1499981689676637e-08, + "loss": 0.6903, + "step": 48793 + }, + { + "epoch": 1.79, + "learning_rate": 2.1496319625004577e-08, + "loss": 0.6989, + "step": 48794 + }, + { + "epoch": 1.79, + "learning_rate": 2.1492657560332513e-08, + "loss": 0.7012, + "step": 48795 + }, + { + "epoch": 1.79, + "learning_rate": 2.1488995495660453e-08, + "loss": 0.6955, + "step": 48796 + }, + { + "epoch": 1.79, + "learning_rate": 2.148533343098839e-08, + "loss": 0.6851, + "step": 48797 + }, + { + "epoch": 1.79, + "learning_rate": 2.148167136631633e-08, + "loss": 0.7031, + "step": 48798 + }, + { + "epoch": 1.79, + "learning_rate": 2.1478009301644266e-08, + "loss": 0.7044, + "step": 48799 + }, + { + "epoch": 1.79, + "learning_rate": 2.1474347236972203e-08, + "loss": 0.6952, + "step": 48800 + }, + { + "epoch": 1.79, + "learning_rate": 2.1470685172300143e-08, + "loss": 0.6909, + "step": 48801 + }, + { + "epoch": 1.79, + "learning_rate": 2.146702310762808e-08, + "loss": 0.693, + "step": 48802 + }, + { + "epoch": 1.79, + "learning_rate": 2.146336104295602e-08, + "loss": 0.6914, + "step": 48803 + }, + { + "epoch": 1.79, + "learning_rate": 2.1459698978283956e-08, + "loss": 0.6803, + "step": 48804 + }, + { + "epoch": 1.79, + "learning_rate": 2.1456036913611893e-08, + "loss": 0.6864, + "step": 48805 + }, + { + "epoch": 1.79, + "learning_rate": 2.145237484893983e-08, + "loss": 0.6891, + "step": 48806 + }, + { + "epoch": 1.79, + "learning_rate": 2.144871278426777e-08, + "loss": 0.6926, + "step": 48807 + }, + { + "epoch": 1.79, + "learning_rate": 2.1445050719595706e-08, + "loss": 0.6859, + "step": 48808 + }, + { + "epoch": 1.79, + "learning_rate": 2.1441388654923646e-08, + "loss": 0.7001, + "step": 48809 + }, + { + "epoch": 1.79, + "learning_rate": 2.1437726590251583e-08, + "loss": 0.6859, + "step": 48810 + }, + { + "epoch": 1.79, + "learning_rate": 2.143406452557952e-08, + "loss": 0.6905, + "step": 48811 + }, + { + "epoch": 1.79, + "learning_rate": 2.143040246090746e-08, + "loss": 0.6852, + "step": 48812 + }, + { + "epoch": 1.79, + "learning_rate": 2.1426740396235396e-08, + "loss": 0.6924, + "step": 48813 + }, + { + "epoch": 1.79, + "learning_rate": 2.1423078331563336e-08, + "loss": 0.6865, + "step": 48814 + }, + { + "epoch": 1.79, + "learning_rate": 2.1419416266891272e-08, + "loss": 0.7036, + "step": 48815 + }, + { + "epoch": 1.79, + "learning_rate": 2.1415754202219212e-08, + "loss": 0.6914, + "step": 48816 + }, + { + "epoch": 1.79, + "learning_rate": 2.1412092137547145e-08, + "loss": 0.7015, + "step": 48817 + }, + { + "epoch": 1.79, + "learning_rate": 2.1408430072875085e-08, + "loss": 0.6921, + "step": 48818 + }, + { + "epoch": 1.79, + "learning_rate": 2.1404768008203025e-08, + "loss": 0.6966, + "step": 48819 + }, + { + "epoch": 1.79, + "learning_rate": 2.1401105943530962e-08, + "loss": 0.6945, + "step": 48820 + }, + { + "epoch": 1.79, + "learning_rate": 2.1397443878858902e-08, + "loss": 0.6986, + "step": 48821 + }, + { + "epoch": 1.79, + "learning_rate": 2.139378181418684e-08, + "loss": 0.6835, + "step": 48822 + }, + { + "epoch": 1.79, + "learning_rate": 2.1390119749514775e-08, + "loss": 0.7016, + "step": 48823 + }, + { + "epoch": 1.79, + "learning_rate": 2.1386457684842712e-08, + "loss": 0.6894, + "step": 48824 + }, + { + "epoch": 1.79, + "learning_rate": 2.1382795620170652e-08, + "loss": 0.6989, + "step": 48825 + }, + { + "epoch": 1.79, + "learning_rate": 2.1379133555498588e-08, + "loss": 0.6968, + "step": 48826 + }, + { + "epoch": 1.79, + "learning_rate": 2.1375471490826528e-08, + "loss": 0.7009, + "step": 48827 + }, + { + "epoch": 1.79, + "learning_rate": 2.1371809426154465e-08, + "loss": 0.7013, + "step": 48828 + }, + { + "epoch": 1.79, + "learning_rate": 2.13681473614824e-08, + "loss": 0.6969, + "step": 48829 + }, + { + "epoch": 1.79, + "learning_rate": 2.136448529681034e-08, + "loss": 0.6945, + "step": 48830 + }, + { + "epoch": 1.79, + "learning_rate": 2.1360823232138278e-08, + "loss": 0.6875, + "step": 48831 + }, + { + "epoch": 1.79, + "learning_rate": 2.1357161167466218e-08, + "loss": 0.6907, + "step": 48832 + }, + { + "epoch": 1.79, + "learning_rate": 2.1353499102794155e-08, + "loss": 0.6919, + "step": 48833 + }, + { + "epoch": 1.79, + "learning_rate": 2.134983703812209e-08, + "loss": 0.6917, + "step": 48834 + }, + { + "epoch": 1.79, + "learning_rate": 2.1346174973450028e-08, + "loss": 0.7029, + "step": 48835 + }, + { + "epoch": 1.79, + "learning_rate": 2.1342512908777968e-08, + "loss": 0.684, + "step": 48836 + }, + { + "epoch": 1.79, + "learning_rate": 2.1338850844105908e-08, + "loss": 0.7037, + "step": 48837 + }, + { + "epoch": 1.79, + "learning_rate": 2.1335188779433844e-08, + "loss": 0.6911, + "step": 48838 + }, + { + "epoch": 1.79, + "learning_rate": 2.1331526714761784e-08, + "loss": 0.7019, + "step": 48839 + }, + { + "epoch": 1.79, + "learning_rate": 2.1327864650089717e-08, + "loss": 0.6966, + "step": 48840 + }, + { + "epoch": 1.79, + "learning_rate": 2.1324202585417657e-08, + "loss": 0.6935, + "step": 48841 + }, + { + "epoch": 1.79, + "learning_rate": 2.1320540520745594e-08, + "loss": 0.6874, + "step": 48842 + }, + { + "epoch": 1.79, + "learning_rate": 2.1316878456073534e-08, + "loss": 0.7076, + "step": 48843 + }, + { + "epoch": 1.79, + "learning_rate": 2.131321639140147e-08, + "loss": 0.6918, + "step": 48844 + }, + { + "epoch": 1.79, + "learning_rate": 2.130955432672941e-08, + "loss": 0.6917, + "step": 48845 + }, + { + "epoch": 1.79, + "learning_rate": 2.1305892262057347e-08, + "loss": 0.6906, + "step": 48846 + }, + { + "epoch": 1.79, + "learning_rate": 2.1302230197385284e-08, + "loss": 0.6894, + "step": 48847 + }, + { + "epoch": 1.79, + "learning_rate": 2.1298568132713224e-08, + "loss": 0.698, + "step": 48848 + }, + { + "epoch": 1.79, + "learning_rate": 2.129490606804116e-08, + "loss": 0.6988, + "step": 48849 + }, + { + "epoch": 1.79, + "learning_rate": 2.12912440033691e-08, + "loss": 0.7146, + "step": 48850 + }, + { + "epoch": 1.79, + "learning_rate": 2.1287581938697037e-08, + "loss": 0.7026, + "step": 48851 + }, + { + "epoch": 1.79, + "learning_rate": 2.1283919874024973e-08, + "loss": 0.6992, + "step": 48852 + }, + { + "epoch": 1.79, + "learning_rate": 2.128025780935291e-08, + "loss": 0.6993, + "step": 48853 + }, + { + "epoch": 1.79, + "learning_rate": 2.127659574468085e-08, + "loss": 0.7022, + "step": 48854 + }, + { + "epoch": 1.79, + "learning_rate": 2.1272933680008787e-08, + "loss": 0.6863, + "step": 48855 + }, + { + "epoch": 1.79, + "learning_rate": 2.1269271615336727e-08, + "loss": 0.7005, + "step": 48856 + }, + { + "epoch": 1.79, + "learning_rate": 2.1265609550664667e-08, + "loss": 0.678, + "step": 48857 + }, + { + "epoch": 1.79, + "learning_rate": 2.12619474859926e-08, + "loss": 0.691, + "step": 48858 + }, + { + "epoch": 1.79, + "learning_rate": 2.125828542132054e-08, + "loss": 0.6823, + "step": 48859 + }, + { + "epoch": 1.79, + "learning_rate": 2.1254623356648476e-08, + "loss": 0.6828, + "step": 48860 + }, + { + "epoch": 1.79, + "learning_rate": 2.1250961291976416e-08, + "loss": 0.686, + "step": 48861 + }, + { + "epoch": 1.79, + "learning_rate": 2.1247299227304353e-08, + "loss": 0.6897, + "step": 48862 + }, + { + "epoch": 1.79, + "learning_rate": 2.1243637162632293e-08, + "loss": 0.6891, + "step": 48863 + }, + { + "epoch": 1.79, + "learning_rate": 2.1239975097960226e-08, + "loss": 0.6929, + "step": 48864 + }, + { + "epoch": 1.79, + "learning_rate": 2.1236313033288166e-08, + "loss": 0.6919, + "step": 48865 + }, + { + "epoch": 1.79, + "learning_rate": 2.1232650968616106e-08, + "loss": 0.6875, + "step": 48866 + }, + { + "epoch": 1.79, + "learning_rate": 2.1228988903944043e-08, + "loss": 0.6808, + "step": 48867 + }, + { + "epoch": 1.79, + "learning_rate": 2.1225326839271983e-08, + "loss": 0.691, + "step": 48868 + }, + { + "epoch": 1.79, + "learning_rate": 2.122166477459992e-08, + "loss": 0.6945, + "step": 48869 + }, + { + "epoch": 1.79, + "learning_rate": 2.1218002709927856e-08, + "loss": 0.6897, + "step": 48870 + }, + { + "epoch": 1.79, + "learning_rate": 2.1214340645255792e-08, + "loss": 0.6933, + "step": 48871 + }, + { + "epoch": 1.79, + "learning_rate": 2.1210678580583732e-08, + "loss": 0.6913, + "step": 48872 + }, + { + "epoch": 1.79, + "learning_rate": 2.120701651591167e-08, + "loss": 0.6898, + "step": 48873 + }, + { + "epoch": 1.79, + "learning_rate": 2.120335445123961e-08, + "loss": 0.7079, + "step": 48874 + }, + { + "epoch": 1.79, + "learning_rate": 2.1199692386567546e-08, + "loss": 0.7046, + "step": 48875 + }, + { + "epoch": 1.79, + "learning_rate": 2.1196030321895482e-08, + "loss": 0.6905, + "step": 48876 + }, + { + "epoch": 1.79, + "learning_rate": 2.1192368257223422e-08, + "loss": 0.687, + "step": 48877 + }, + { + "epoch": 1.79, + "learning_rate": 2.118870619255136e-08, + "loss": 0.6951, + "step": 48878 + }, + { + "epoch": 1.79, + "learning_rate": 2.11850441278793e-08, + "loss": 0.7041, + "step": 48879 + }, + { + "epoch": 1.79, + "learning_rate": 2.1181382063207235e-08, + "loss": 0.6859, + "step": 48880 + }, + { + "epoch": 1.79, + "learning_rate": 2.1177719998535172e-08, + "loss": 0.688, + "step": 48881 + }, + { + "epoch": 1.79, + "learning_rate": 2.117405793386311e-08, + "loss": 0.702, + "step": 48882 + }, + { + "epoch": 1.79, + "learning_rate": 2.117039586919105e-08, + "loss": 0.6811, + "step": 48883 + }, + { + "epoch": 1.79, + "learning_rate": 2.1166733804518988e-08, + "loss": 0.6909, + "step": 48884 + }, + { + "epoch": 1.79, + "learning_rate": 2.1163071739846925e-08, + "loss": 0.6995, + "step": 48885 + }, + { + "epoch": 1.79, + "learning_rate": 2.1159409675174865e-08, + "loss": 0.7027, + "step": 48886 + }, + { + "epoch": 1.79, + "learning_rate": 2.11557476105028e-08, + "loss": 0.6922, + "step": 48887 + }, + { + "epoch": 1.79, + "learning_rate": 2.1152085545830738e-08, + "loss": 0.6966, + "step": 48888 + }, + { + "epoch": 1.79, + "learning_rate": 2.1148423481158675e-08, + "loss": 0.691, + "step": 48889 + }, + { + "epoch": 1.79, + "learning_rate": 2.1144761416486615e-08, + "loss": 0.6946, + "step": 48890 + }, + { + "epoch": 1.79, + "learning_rate": 2.114109935181455e-08, + "loss": 0.691, + "step": 48891 + }, + { + "epoch": 1.79, + "learning_rate": 2.113743728714249e-08, + "loss": 0.6958, + "step": 48892 + }, + { + "epoch": 1.79, + "learning_rate": 2.1133775222470428e-08, + "loss": 0.6937, + "step": 48893 + }, + { + "epoch": 1.79, + "learning_rate": 2.1130113157798364e-08, + "loss": 0.704, + "step": 48894 + }, + { + "epoch": 1.79, + "learning_rate": 2.1126451093126304e-08, + "loss": 0.7001, + "step": 48895 + }, + { + "epoch": 1.79, + "learning_rate": 2.112278902845424e-08, + "loss": 0.6952, + "step": 48896 + }, + { + "epoch": 1.79, + "learning_rate": 2.111912696378218e-08, + "loss": 0.6951, + "step": 48897 + }, + { + "epoch": 1.79, + "learning_rate": 2.1115464899110118e-08, + "loss": 0.6913, + "step": 48898 + }, + { + "epoch": 1.79, + "learning_rate": 2.1111802834438054e-08, + "loss": 0.697, + "step": 48899 + }, + { + "epoch": 1.79, + "learning_rate": 2.110814076976599e-08, + "loss": 0.6944, + "step": 48900 + }, + { + "epoch": 1.79, + "learning_rate": 2.110447870509393e-08, + "loss": 0.7022, + "step": 48901 + }, + { + "epoch": 1.79, + "learning_rate": 2.110081664042187e-08, + "loss": 0.6847, + "step": 48902 + }, + { + "epoch": 1.79, + "learning_rate": 2.1097154575749807e-08, + "loss": 0.706, + "step": 48903 + }, + { + "epoch": 1.79, + "learning_rate": 2.1093492511077747e-08, + "loss": 0.6847, + "step": 48904 + }, + { + "epoch": 1.79, + "learning_rate": 2.108983044640568e-08, + "loss": 0.7114, + "step": 48905 + }, + { + "epoch": 1.79, + "learning_rate": 2.108616838173362e-08, + "loss": 0.6837, + "step": 48906 + }, + { + "epoch": 1.79, + "learning_rate": 2.1082506317061557e-08, + "loss": 0.69, + "step": 48907 + }, + { + "epoch": 1.79, + "learning_rate": 2.1078844252389497e-08, + "loss": 0.689, + "step": 48908 + }, + { + "epoch": 1.79, + "learning_rate": 2.1075182187717434e-08, + "loss": 0.6927, + "step": 48909 + }, + { + "epoch": 1.79, + "learning_rate": 2.1071520123045374e-08, + "loss": 0.706, + "step": 48910 + }, + { + "epoch": 1.79, + "learning_rate": 2.106785805837331e-08, + "loss": 0.6961, + "step": 48911 + }, + { + "epoch": 1.79, + "learning_rate": 2.1064195993701247e-08, + "loss": 0.6968, + "step": 48912 + }, + { + "epoch": 1.79, + "learning_rate": 2.1060533929029187e-08, + "loss": 0.7015, + "step": 48913 + }, + { + "epoch": 1.79, + "learning_rate": 2.1056871864357123e-08, + "loss": 0.6934, + "step": 48914 + }, + { + "epoch": 1.79, + "learning_rate": 2.1053209799685063e-08, + "loss": 0.6918, + "step": 48915 + }, + { + "epoch": 1.79, + "learning_rate": 2.1049547735013e-08, + "loss": 0.6967, + "step": 48916 + }, + { + "epoch": 1.79, + "learning_rate": 2.1045885670340936e-08, + "loss": 0.684, + "step": 48917 + }, + { + "epoch": 1.79, + "learning_rate": 2.1042223605668873e-08, + "loss": 0.695, + "step": 48918 + }, + { + "epoch": 1.79, + "learning_rate": 2.1038561540996813e-08, + "loss": 0.6964, + "step": 48919 + }, + { + "epoch": 1.79, + "learning_rate": 2.1034899476324753e-08, + "loss": 0.6969, + "step": 48920 + }, + { + "epoch": 1.79, + "learning_rate": 2.103123741165269e-08, + "loss": 0.6999, + "step": 48921 + }, + { + "epoch": 1.79, + "learning_rate": 2.1027575346980626e-08, + "loss": 0.6877, + "step": 48922 + }, + { + "epoch": 1.79, + "learning_rate": 2.1023913282308563e-08, + "loss": 0.6959, + "step": 48923 + }, + { + "epoch": 1.79, + "learning_rate": 2.1020251217636503e-08, + "loss": 0.7063, + "step": 48924 + }, + { + "epoch": 1.79, + "learning_rate": 2.101658915296444e-08, + "loss": 0.6863, + "step": 48925 + }, + { + "epoch": 1.79, + "learning_rate": 2.101292708829238e-08, + "loss": 0.697, + "step": 48926 + }, + { + "epoch": 1.79, + "learning_rate": 2.1009265023620316e-08, + "loss": 0.706, + "step": 48927 + }, + { + "epoch": 1.79, + "learning_rate": 2.1005602958948253e-08, + "loss": 0.6802, + "step": 48928 + }, + { + "epoch": 1.79, + "learning_rate": 2.1001940894276192e-08, + "loss": 0.6968, + "step": 48929 + }, + { + "epoch": 1.79, + "learning_rate": 2.099827882960413e-08, + "loss": 0.7031, + "step": 48930 + }, + { + "epoch": 1.79, + "learning_rate": 2.099461676493207e-08, + "loss": 0.6959, + "step": 48931 + }, + { + "epoch": 1.79, + "learning_rate": 2.0990954700260006e-08, + "loss": 0.7012, + "step": 48932 + }, + { + "epoch": 1.79, + "learning_rate": 2.0987292635587946e-08, + "loss": 0.7015, + "step": 48933 + }, + { + "epoch": 1.79, + "learning_rate": 2.0983630570915882e-08, + "loss": 0.6906, + "step": 48934 + }, + { + "epoch": 1.79, + "learning_rate": 2.097996850624382e-08, + "loss": 0.6951, + "step": 48935 + }, + { + "epoch": 1.79, + "learning_rate": 2.0976306441571755e-08, + "loss": 0.688, + "step": 48936 + }, + { + "epoch": 1.79, + "learning_rate": 2.0972644376899695e-08, + "loss": 0.6826, + "step": 48937 + }, + { + "epoch": 1.79, + "learning_rate": 2.0968982312227635e-08, + "loss": 0.6895, + "step": 48938 + }, + { + "epoch": 1.79, + "learning_rate": 2.0965320247555572e-08, + "loss": 0.6867, + "step": 48939 + }, + { + "epoch": 1.79, + "learning_rate": 2.096165818288351e-08, + "loss": 0.6915, + "step": 48940 + }, + { + "epoch": 1.79, + "learning_rate": 2.0957996118211445e-08, + "loss": 0.6876, + "step": 48941 + }, + { + "epoch": 1.79, + "learning_rate": 2.0954334053539385e-08, + "loss": 0.6911, + "step": 48942 + }, + { + "epoch": 1.79, + "learning_rate": 2.0950671988867322e-08, + "loss": 0.69, + "step": 48943 + }, + { + "epoch": 1.79, + "learning_rate": 2.094700992419526e-08, + "loss": 0.6993, + "step": 48944 + }, + { + "epoch": 1.79, + "learning_rate": 2.0943347859523198e-08, + "loss": 0.6835, + "step": 48945 + }, + { + "epoch": 1.79, + "learning_rate": 2.0939685794851135e-08, + "loss": 0.6892, + "step": 48946 + }, + { + "epoch": 1.79, + "learning_rate": 2.093602373017907e-08, + "loss": 0.6899, + "step": 48947 + }, + { + "epoch": 1.79, + "learning_rate": 2.093236166550701e-08, + "loss": 0.6887, + "step": 48948 + }, + { + "epoch": 1.79, + "learning_rate": 2.092869960083495e-08, + "loss": 0.6951, + "step": 48949 + }, + { + "epoch": 1.79, + "learning_rate": 2.0925037536162888e-08, + "loss": 0.6845, + "step": 48950 + }, + { + "epoch": 1.79, + "learning_rate": 2.0921375471490828e-08, + "loss": 0.6884, + "step": 48951 + }, + { + "epoch": 1.79, + "learning_rate": 2.091771340681876e-08, + "loss": 0.6857, + "step": 48952 + }, + { + "epoch": 1.79, + "learning_rate": 2.09140513421467e-08, + "loss": 0.6847, + "step": 48953 + }, + { + "epoch": 1.79, + "learning_rate": 2.0910389277474638e-08, + "loss": 0.7079, + "step": 48954 + }, + { + "epoch": 1.79, + "learning_rate": 2.0906727212802578e-08, + "loss": 0.6864, + "step": 48955 + }, + { + "epoch": 1.79, + "learning_rate": 2.0903065148130514e-08, + "loss": 0.6887, + "step": 48956 + }, + { + "epoch": 1.79, + "learning_rate": 2.0899403083458454e-08, + "loss": 0.6965, + "step": 48957 + }, + { + "epoch": 1.79, + "learning_rate": 2.089574101878639e-08, + "loss": 0.6999, + "step": 48958 + }, + { + "epoch": 1.79, + "learning_rate": 2.0892078954114327e-08, + "loss": 0.6939, + "step": 48959 + }, + { + "epoch": 1.79, + "learning_rate": 2.0888416889442267e-08, + "loss": 0.6738, + "step": 48960 + }, + { + "epoch": 1.79, + "learning_rate": 2.0884754824770204e-08, + "loss": 0.696, + "step": 48961 + }, + { + "epoch": 1.79, + "learning_rate": 2.0881092760098144e-08, + "loss": 0.68, + "step": 48962 + }, + { + "epoch": 1.79, + "learning_rate": 2.087743069542608e-08, + "loss": 0.6858, + "step": 48963 + }, + { + "epoch": 1.79, + "learning_rate": 2.0873768630754017e-08, + "loss": 0.6933, + "step": 48964 + }, + { + "epoch": 1.79, + "learning_rate": 2.0870106566081954e-08, + "loss": 0.6935, + "step": 48965 + }, + { + "epoch": 1.79, + "learning_rate": 2.0866444501409894e-08, + "loss": 0.6882, + "step": 48966 + }, + { + "epoch": 1.79, + "learning_rate": 2.0862782436737834e-08, + "loss": 0.6989, + "step": 48967 + }, + { + "epoch": 1.79, + "learning_rate": 2.085912037206577e-08, + "loss": 0.6935, + "step": 48968 + }, + { + "epoch": 1.79, + "learning_rate": 2.0855458307393707e-08, + "loss": 0.6891, + "step": 48969 + }, + { + "epoch": 1.79, + "learning_rate": 2.0851796242721644e-08, + "loss": 0.6782, + "step": 48970 + }, + { + "epoch": 1.79, + "learning_rate": 2.0848134178049583e-08, + "loss": 0.6911, + "step": 48971 + }, + { + "epoch": 1.79, + "learning_rate": 2.084447211337752e-08, + "loss": 0.6908, + "step": 48972 + }, + { + "epoch": 1.79, + "learning_rate": 2.084081004870546e-08, + "loss": 0.6973, + "step": 48973 + }, + { + "epoch": 1.79, + "learning_rate": 2.0837147984033397e-08, + "loss": 0.6934, + "step": 48974 + }, + { + "epoch": 1.79, + "learning_rate": 2.0833485919361337e-08, + "loss": 0.6923, + "step": 48975 + }, + { + "epoch": 1.79, + "learning_rate": 2.0829823854689273e-08, + "loss": 0.683, + "step": 48976 + }, + { + "epoch": 1.79, + "learning_rate": 2.082616179001721e-08, + "loss": 0.6864, + "step": 48977 + }, + { + "epoch": 1.79, + "learning_rate": 2.082249972534515e-08, + "loss": 0.7023, + "step": 48978 + }, + { + "epoch": 1.79, + "learning_rate": 2.0818837660673086e-08, + "loss": 0.6929, + "step": 48979 + }, + { + "epoch": 1.79, + "learning_rate": 2.0815175596001026e-08, + "loss": 0.6807, + "step": 48980 + }, + { + "epoch": 1.79, + "learning_rate": 2.0811513531328963e-08, + "loss": 0.7017, + "step": 48981 + }, + { + "epoch": 1.79, + "learning_rate": 2.08078514666569e-08, + "loss": 0.7038, + "step": 48982 + }, + { + "epoch": 1.79, + "learning_rate": 2.0804189401984836e-08, + "loss": 0.7007, + "step": 48983 + }, + { + "epoch": 1.79, + "learning_rate": 2.0800527337312776e-08, + "loss": 0.6844, + "step": 48984 + }, + { + "epoch": 1.79, + "learning_rate": 2.0796865272640716e-08, + "loss": 0.7012, + "step": 48985 + }, + { + "epoch": 1.79, + "learning_rate": 2.0793203207968653e-08, + "loss": 0.6927, + "step": 48986 + }, + { + "epoch": 1.79, + "learning_rate": 2.078954114329659e-08, + "loss": 0.7086, + "step": 48987 + }, + { + "epoch": 1.79, + "learning_rate": 2.0785879078624526e-08, + "loss": 0.6891, + "step": 48988 + }, + { + "epoch": 1.79, + "learning_rate": 2.0782217013952466e-08, + "loss": 0.6912, + "step": 48989 + }, + { + "epoch": 1.79, + "learning_rate": 2.0778554949280402e-08, + "loss": 0.6857, + "step": 48990 + }, + { + "epoch": 1.79, + "learning_rate": 2.0774892884608342e-08, + "loss": 0.6953, + "step": 48991 + }, + { + "epoch": 1.79, + "learning_rate": 2.077123081993628e-08, + "loss": 0.7049, + "step": 48992 + }, + { + "epoch": 1.79, + "learning_rate": 2.0767568755264216e-08, + "loss": 0.6936, + "step": 48993 + }, + { + "epoch": 1.79, + "learning_rate": 2.0763906690592155e-08, + "loss": 0.6966, + "step": 48994 + }, + { + "epoch": 1.79, + "learning_rate": 2.0760244625920092e-08, + "loss": 0.6909, + "step": 48995 + }, + { + "epoch": 1.79, + "learning_rate": 2.0756582561248032e-08, + "loss": 0.6894, + "step": 48996 + }, + { + "epoch": 1.79, + "learning_rate": 2.075292049657597e-08, + "loss": 0.6938, + "step": 48997 + }, + { + "epoch": 1.79, + "learning_rate": 2.074925843190391e-08, + "loss": 0.6916, + "step": 48998 + }, + { + "epoch": 1.79, + "learning_rate": 2.0745596367231842e-08, + "loss": 0.6972, + "step": 48999 + }, + { + "epoch": 1.79, + "learning_rate": 2.0741934302559782e-08, + "loss": 0.6974, + "step": 49000 + }, + { + "epoch": 1.79, + "learning_rate": 2.073827223788772e-08, + "loss": 0.7036, + "step": 49001 + }, + { + "epoch": 1.79, + "learning_rate": 2.073461017321566e-08, + "loss": 0.6765, + "step": 49002 + }, + { + "epoch": 1.79, + "learning_rate": 2.0730948108543598e-08, + "loss": 0.7022, + "step": 49003 + }, + { + "epoch": 1.79, + "learning_rate": 2.0727286043871535e-08, + "loss": 0.6878, + "step": 49004 + }, + { + "epoch": 1.79, + "learning_rate": 2.072362397919947e-08, + "loss": 0.6896, + "step": 49005 + }, + { + "epoch": 1.79, + "learning_rate": 2.0719961914527408e-08, + "loss": 0.7036, + "step": 49006 + }, + { + "epoch": 1.79, + "learning_rate": 2.0716299849855348e-08, + "loss": 0.692, + "step": 49007 + }, + { + "epoch": 1.79, + "learning_rate": 2.0712637785183285e-08, + "loss": 0.6937, + "step": 49008 + }, + { + "epoch": 1.79, + "learning_rate": 2.0708975720511225e-08, + "loss": 0.6814, + "step": 49009 + }, + { + "epoch": 1.79, + "learning_rate": 2.070531365583916e-08, + "loss": 0.685, + "step": 49010 + }, + { + "epoch": 1.79, + "learning_rate": 2.0701651591167098e-08, + "loss": 0.6946, + "step": 49011 + }, + { + "epoch": 1.79, + "learning_rate": 2.0697989526495038e-08, + "loss": 0.6983, + "step": 49012 + }, + { + "epoch": 1.79, + "learning_rate": 2.0694327461822974e-08, + "loss": 0.6975, + "step": 49013 + }, + { + "epoch": 1.79, + "learning_rate": 2.0690665397150914e-08, + "loss": 0.6946, + "step": 49014 + }, + { + "epoch": 1.79, + "learning_rate": 2.068700333247885e-08, + "loss": 0.6903, + "step": 49015 + }, + { + "epoch": 1.79, + "learning_rate": 2.0683341267806788e-08, + "loss": 0.6927, + "step": 49016 + }, + { + "epoch": 1.79, + "learning_rate": 2.0679679203134724e-08, + "loss": 0.6867, + "step": 49017 + }, + { + "epoch": 1.79, + "learning_rate": 2.0676017138462664e-08, + "loss": 0.6945, + "step": 49018 + }, + { + "epoch": 1.79, + "learning_rate": 2.06723550737906e-08, + "loss": 0.6864, + "step": 49019 + }, + { + "epoch": 1.79, + "learning_rate": 2.066869300911854e-08, + "loss": 0.6941, + "step": 49020 + }, + { + "epoch": 1.79, + "learning_rate": 2.066503094444648e-08, + "loss": 0.6877, + "step": 49021 + }, + { + "epoch": 1.79, + "learning_rate": 2.0661368879774417e-08, + "loss": 0.7108, + "step": 49022 + }, + { + "epoch": 1.79, + "learning_rate": 2.0657706815102354e-08, + "loss": 0.6681, + "step": 49023 + }, + { + "epoch": 1.79, + "learning_rate": 2.065404475043029e-08, + "loss": 0.6889, + "step": 49024 + }, + { + "epoch": 1.79, + "learning_rate": 2.065038268575823e-08, + "loss": 0.6974, + "step": 49025 + }, + { + "epoch": 1.79, + "learning_rate": 2.0646720621086167e-08, + "loss": 0.684, + "step": 49026 + }, + { + "epoch": 1.79, + "learning_rate": 2.0643058556414107e-08, + "loss": 0.6964, + "step": 49027 + }, + { + "epoch": 1.79, + "learning_rate": 2.0639396491742044e-08, + "loss": 0.7028, + "step": 49028 + }, + { + "epoch": 1.79, + "learning_rate": 2.063573442706998e-08, + "loss": 0.6942, + "step": 49029 + }, + { + "epoch": 1.79, + "learning_rate": 2.063207236239792e-08, + "loss": 0.696, + "step": 49030 + }, + { + "epoch": 1.79, + "learning_rate": 2.0628410297725857e-08, + "loss": 0.6772, + "step": 49031 + }, + { + "epoch": 1.79, + "learning_rate": 2.0624748233053797e-08, + "loss": 0.7038, + "step": 49032 + }, + { + "epoch": 1.79, + "learning_rate": 2.0621086168381733e-08, + "loss": 0.7015, + "step": 49033 + }, + { + "epoch": 1.79, + "learning_rate": 2.061742410370967e-08, + "loss": 0.6923, + "step": 49034 + }, + { + "epoch": 1.79, + "learning_rate": 2.0613762039037607e-08, + "loss": 0.696, + "step": 49035 + }, + { + "epoch": 1.79, + "learning_rate": 2.0610099974365546e-08, + "loss": 0.6966, + "step": 49036 + }, + { + "epoch": 1.79, + "learning_rate": 2.0606437909693483e-08, + "loss": 0.692, + "step": 49037 + }, + { + "epoch": 1.79, + "learning_rate": 2.0602775845021423e-08, + "loss": 0.677, + "step": 49038 + }, + { + "epoch": 1.79, + "learning_rate": 2.0599113780349363e-08, + "loss": 0.6796, + "step": 49039 + }, + { + "epoch": 1.79, + "learning_rate": 2.0595451715677296e-08, + "loss": 0.6998, + "step": 49040 + }, + { + "epoch": 1.79, + "learning_rate": 2.0591789651005236e-08, + "loss": 0.6781, + "step": 49041 + }, + { + "epoch": 1.79, + "learning_rate": 2.0588127586333173e-08, + "loss": 0.694, + "step": 49042 + }, + { + "epoch": 1.79, + "learning_rate": 2.0584465521661113e-08, + "loss": 0.6777, + "step": 49043 + }, + { + "epoch": 1.79, + "learning_rate": 2.058080345698905e-08, + "loss": 0.6951, + "step": 49044 + }, + { + "epoch": 1.79, + "learning_rate": 2.057714139231699e-08, + "loss": 0.6632, + "step": 49045 + }, + { + "epoch": 1.79, + "learning_rate": 2.0573479327644923e-08, + "loss": 0.7069, + "step": 49046 + }, + { + "epoch": 1.79, + "learning_rate": 2.0569817262972863e-08, + "loss": 0.6934, + "step": 49047 + }, + { + "epoch": 1.79, + "learning_rate": 2.05661551983008e-08, + "loss": 0.6884, + "step": 49048 + }, + { + "epoch": 1.79, + "learning_rate": 2.056249313362874e-08, + "loss": 0.6899, + "step": 49049 + }, + { + "epoch": 1.79, + "learning_rate": 2.055883106895668e-08, + "loss": 0.691, + "step": 49050 + }, + { + "epoch": 1.79, + "learning_rate": 2.0555169004284616e-08, + "loss": 0.6971, + "step": 49051 + }, + { + "epoch": 1.79, + "learning_rate": 2.0551506939612552e-08, + "loss": 0.6866, + "step": 49052 + }, + { + "epoch": 1.79, + "learning_rate": 2.054784487494049e-08, + "loss": 0.6945, + "step": 49053 + }, + { + "epoch": 1.79, + "learning_rate": 2.054418281026843e-08, + "loss": 0.6958, + "step": 49054 + }, + { + "epoch": 1.79, + "learning_rate": 2.0540520745596365e-08, + "loss": 0.6917, + "step": 49055 + }, + { + "epoch": 1.79, + "learning_rate": 2.0536858680924305e-08, + "loss": 0.6978, + "step": 49056 + }, + { + "epoch": 1.79, + "learning_rate": 2.0533196616252242e-08, + "loss": 0.69, + "step": 49057 + }, + { + "epoch": 1.79, + "learning_rate": 2.052953455158018e-08, + "loss": 0.6923, + "step": 49058 + }, + { + "epoch": 1.79, + "learning_rate": 2.052587248690812e-08, + "loss": 0.6889, + "step": 49059 + }, + { + "epoch": 1.79, + "learning_rate": 2.0522210422236055e-08, + "loss": 0.6934, + "step": 49060 + }, + { + "epoch": 1.8, + "learning_rate": 2.0518548357563995e-08, + "loss": 0.6894, + "step": 49061 + }, + { + "epoch": 1.8, + "learning_rate": 2.0514886292891932e-08, + "loss": 0.6841, + "step": 49062 + }, + { + "epoch": 1.8, + "learning_rate": 2.051122422821987e-08, + "loss": 0.6941, + "step": 49063 + }, + { + "epoch": 1.8, + "learning_rate": 2.0507562163547805e-08, + "loss": 0.6897, + "step": 49064 + }, + { + "epoch": 1.8, + "learning_rate": 2.0503900098875745e-08, + "loss": 0.7067, + "step": 49065 + }, + { + "epoch": 1.8, + "learning_rate": 2.050023803420368e-08, + "loss": 0.6985, + "step": 49066 + }, + { + "epoch": 1.8, + "learning_rate": 2.049657596953162e-08, + "loss": 0.6942, + "step": 49067 + }, + { + "epoch": 1.8, + "learning_rate": 2.049291390485956e-08, + "loss": 0.6934, + "step": 49068 + }, + { + "epoch": 1.8, + "learning_rate": 2.0489251840187498e-08, + "loss": 0.6933, + "step": 49069 + }, + { + "epoch": 1.8, + "learning_rate": 2.0485589775515435e-08, + "loss": 0.6958, + "step": 49070 + }, + { + "epoch": 1.8, + "learning_rate": 2.048192771084337e-08, + "loss": 0.6881, + "step": 49071 + }, + { + "epoch": 1.8, + "learning_rate": 2.047826564617131e-08, + "loss": 0.6998, + "step": 49072 + }, + { + "epoch": 1.8, + "learning_rate": 2.0474603581499248e-08, + "loss": 0.7082, + "step": 49073 + }, + { + "epoch": 1.8, + "learning_rate": 2.0470941516827188e-08, + "loss": 0.6849, + "step": 49074 + }, + { + "epoch": 1.8, + "learning_rate": 2.0467279452155124e-08, + "loss": 0.6849, + "step": 49075 + }, + { + "epoch": 1.8, + "learning_rate": 2.046361738748306e-08, + "loss": 0.6845, + "step": 49076 + }, + { + "epoch": 1.8, + "learning_rate": 2.0459955322811e-08, + "loss": 0.6912, + "step": 49077 + }, + { + "epoch": 1.8, + "learning_rate": 2.0456293258138937e-08, + "loss": 0.6948, + "step": 49078 + }, + { + "epoch": 1.8, + "learning_rate": 2.0452631193466877e-08, + "loss": 0.6934, + "step": 49079 + }, + { + "epoch": 1.8, + "learning_rate": 2.0448969128794814e-08, + "loss": 0.7034, + "step": 49080 + }, + { + "epoch": 1.8, + "learning_rate": 2.044530706412275e-08, + "loss": 0.6936, + "step": 49081 + }, + { + "epoch": 1.8, + "learning_rate": 2.0441644999450687e-08, + "loss": 0.6915, + "step": 49082 + }, + { + "epoch": 1.8, + "learning_rate": 2.0437982934778627e-08, + "loss": 0.687, + "step": 49083 + }, + { + "epoch": 1.8, + "learning_rate": 2.0434320870106564e-08, + "loss": 0.6933, + "step": 49084 + }, + { + "epoch": 1.8, + "learning_rate": 2.0430658805434504e-08, + "loss": 0.6905, + "step": 49085 + }, + { + "epoch": 1.8, + "learning_rate": 2.0426996740762444e-08, + "loss": 0.6985, + "step": 49086 + }, + { + "epoch": 1.8, + "learning_rate": 2.0423334676090377e-08, + "loss": 0.6913, + "step": 49087 + }, + { + "epoch": 1.8, + "learning_rate": 2.0419672611418317e-08, + "loss": 0.694, + "step": 49088 + }, + { + "epoch": 1.8, + "learning_rate": 2.0416010546746253e-08, + "loss": 0.6789, + "step": 49089 + }, + { + "epoch": 1.8, + "learning_rate": 2.0412348482074193e-08, + "loss": 0.6911, + "step": 49090 + }, + { + "epoch": 1.8, + "learning_rate": 2.040868641740213e-08, + "loss": 0.693, + "step": 49091 + }, + { + "epoch": 1.8, + "learning_rate": 2.040502435273007e-08, + "loss": 0.6982, + "step": 49092 + }, + { + "epoch": 1.8, + "learning_rate": 2.0401362288058007e-08, + "loss": 0.6772, + "step": 49093 + }, + { + "epoch": 1.8, + "learning_rate": 2.0397700223385943e-08, + "loss": 0.6922, + "step": 49094 + }, + { + "epoch": 1.8, + "learning_rate": 2.0394038158713883e-08, + "loss": 0.6838, + "step": 49095 + }, + { + "epoch": 1.8, + "learning_rate": 2.039037609404182e-08, + "loss": 0.6836, + "step": 49096 + }, + { + "epoch": 1.8, + "learning_rate": 2.038671402936976e-08, + "loss": 0.6966, + "step": 49097 + }, + { + "epoch": 1.8, + "learning_rate": 2.0383051964697696e-08, + "loss": 0.6864, + "step": 49098 + }, + { + "epoch": 1.8, + "learning_rate": 2.0379389900025633e-08, + "loss": 0.7034, + "step": 49099 + }, + { + "epoch": 1.8, + "learning_rate": 2.037572783535357e-08, + "loss": 0.6856, + "step": 49100 + }, + { + "epoch": 1.8, + "learning_rate": 2.037206577068151e-08, + "loss": 0.6837, + "step": 49101 + }, + { + "epoch": 1.8, + "learning_rate": 2.0368403706009446e-08, + "loss": 0.671, + "step": 49102 + }, + { + "epoch": 1.8, + "learning_rate": 2.0364741641337386e-08, + "loss": 0.6903, + "step": 49103 + }, + { + "epoch": 1.8, + "learning_rate": 2.0361079576665323e-08, + "loss": 0.7041, + "step": 49104 + }, + { + "epoch": 1.8, + "learning_rate": 2.035741751199326e-08, + "loss": 0.6961, + "step": 49105 + }, + { + "epoch": 1.8, + "learning_rate": 2.03537554473212e-08, + "loss": 0.6972, + "step": 49106 + }, + { + "epoch": 1.8, + "learning_rate": 2.0350093382649136e-08, + "loss": 0.6937, + "step": 49107 + }, + { + "epoch": 1.8, + "learning_rate": 2.0346431317977076e-08, + "loss": 0.6911, + "step": 49108 + }, + { + "epoch": 1.8, + "learning_rate": 2.0342769253305012e-08, + "loss": 0.697, + "step": 49109 + }, + { + "epoch": 1.8, + "learning_rate": 2.0339107188632952e-08, + "loss": 0.7045, + "step": 49110 + }, + { + "epoch": 1.8, + "learning_rate": 2.0335445123960886e-08, + "loss": 0.6975, + "step": 49111 + }, + { + "epoch": 1.8, + "learning_rate": 2.0331783059288826e-08, + "loss": 0.6867, + "step": 49112 + }, + { + "epoch": 1.8, + "learning_rate": 2.0328120994616765e-08, + "loss": 0.7026, + "step": 49113 + }, + { + "epoch": 1.8, + "learning_rate": 2.0324458929944702e-08, + "loss": 0.6933, + "step": 49114 + }, + { + "epoch": 1.8, + "learning_rate": 2.0320796865272642e-08, + "loss": 0.6898, + "step": 49115 + }, + { + "epoch": 1.8, + "learning_rate": 2.031713480060058e-08, + "loss": 0.6944, + "step": 49116 + }, + { + "epoch": 1.8, + "learning_rate": 2.0313472735928515e-08, + "loss": 0.6906, + "step": 49117 + }, + { + "epoch": 1.8, + "learning_rate": 2.0309810671256452e-08, + "loss": 0.6997, + "step": 49118 + }, + { + "epoch": 1.8, + "learning_rate": 2.0306148606584392e-08, + "loss": 0.6939, + "step": 49119 + }, + { + "epoch": 1.8, + "learning_rate": 2.030248654191233e-08, + "loss": 0.7013, + "step": 49120 + }, + { + "epoch": 1.8, + "learning_rate": 2.0298824477240268e-08, + "loss": 0.6732, + "step": 49121 + }, + { + "epoch": 1.8, + "learning_rate": 2.0295162412568205e-08, + "loss": 0.6956, + "step": 49122 + }, + { + "epoch": 1.8, + "learning_rate": 2.029150034789614e-08, + "loss": 0.6985, + "step": 49123 + }, + { + "epoch": 1.8, + "learning_rate": 2.028783828322408e-08, + "loss": 0.6924, + "step": 49124 + }, + { + "epoch": 1.8, + "learning_rate": 2.0284176218552018e-08, + "loss": 0.6896, + "step": 49125 + }, + { + "epoch": 1.8, + "learning_rate": 2.0280514153879958e-08, + "loss": 0.696, + "step": 49126 + }, + { + "epoch": 1.8, + "learning_rate": 2.0276852089207895e-08, + "loss": 0.6922, + "step": 49127 + }, + { + "epoch": 1.8, + "learning_rate": 2.027319002453583e-08, + "loss": 0.7056, + "step": 49128 + }, + { + "epoch": 1.8, + "learning_rate": 2.0269527959863768e-08, + "loss": 0.7012, + "step": 49129 + }, + { + "epoch": 1.8, + "learning_rate": 2.0265865895191708e-08, + "loss": 0.686, + "step": 49130 + }, + { + "epoch": 1.8, + "learning_rate": 2.0262203830519648e-08, + "loss": 0.6934, + "step": 49131 + }, + { + "epoch": 1.8, + "learning_rate": 2.0258541765847584e-08, + "loss": 0.7013, + "step": 49132 + }, + { + "epoch": 1.8, + "learning_rate": 2.0254879701175524e-08, + "loss": 0.702, + "step": 49133 + }, + { + "epoch": 1.8, + "learning_rate": 2.0251217636503458e-08, + "loss": 0.6938, + "step": 49134 + }, + { + "epoch": 1.8, + "learning_rate": 2.0247555571831398e-08, + "loss": 0.6962, + "step": 49135 + }, + { + "epoch": 1.8, + "learning_rate": 2.0243893507159334e-08, + "loss": 0.6951, + "step": 49136 + }, + { + "epoch": 1.8, + "learning_rate": 2.0240231442487274e-08, + "loss": 0.6976, + "step": 49137 + }, + { + "epoch": 1.8, + "learning_rate": 2.023656937781521e-08, + "loss": 0.6893, + "step": 49138 + }, + { + "epoch": 1.8, + "learning_rate": 2.023290731314315e-08, + "loss": 0.6967, + "step": 49139 + }, + { + "epoch": 1.8, + "learning_rate": 2.0229245248471087e-08, + "loss": 0.6967, + "step": 49140 + }, + { + "epoch": 1.8, + "learning_rate": 2.0225583183799024e-08, + "loss": 0.7032, + "step": 49141 + }, + { + "epoch": 1.8, + "learning_rate": 2.0221921119126964e-08, + "loss": 0.6995, + "step": 49142 + }, + { + "epoch": 1.8, + "learning_rate": 2.02182590544549e-08, + "loss": 0.6895, + "step": 49143 + }, + { + "epoch": 1.8, + "learning_rate": 2.021459698978284e-08, + "loss": 0.6914, + "step": 49144 + }, + { + "epoch": 1.8, + "learning_rate": 2.0210934925110777e-08, + "loss": 0.6976, + "step": 49145 + }, + { + "epoch": 1.8, + "learning_rate": 2.0207272860438714e-08, + "loss": 0.6956, + "step": 49146 + }, + { + "epoch": 1.8, + "learning_rate": 2.020361079576665e-08, + "loss": 0.6916, + "step": 49147 + }, + { + "epoch": 1.8, + "learning_rate": 2.019994873109459e-08, + "loss": 0.6926, + "step": 49148 + }, + { + "epoch": 1.8, + "learning_rate": 2.0196286666422527e-08, + "loss": 0.692, + "step": 49149 + }, + { + "epoch": 1.8, + "learning_rate": 2.0192624601750467e-08, + "loss": 0.6965, + "step": 49150 + }, + { + "epoch": 1.8, + "learning_rate": 2.0188962537078407e-08, + "loss": 0.7021, + "step": 49151 + }, + { + "epoch": 1.8, + "learning_rate": 2.018530047240634e-08, + "loss": 0.6823, + "step": 49152 + }, + { + "epoch": 1.8, + "learning_rate": 2.018163840773428e-08, + "loss": 0.6903, + "step": 49153 + }, + { + "epoch": 1.8, + "learning_rate": 2.0177976343062217e-08, + "loss": 0.6954, + "step": 49154 + }, + { + "epoch": 1.8, + "learning_rate": 2.0174314278390156e-08, + "loss": 0.6875, + "step": 49155 + }, + { + "epoch": 1.8, + "learning_rate": 2.0170652213718093e-08, + "loss": 0.6967, + "step": 49156 + }, + { + "epoch": 1.8, + "learning_rate": 2.0166990149046033e-08, + "loss": 0.6963, + "step": 49157 + }, + { + "epoch": 1.8, + "learning_rate": 2.0163328084373966e-08, + "loss": 0.6867, + "step": 49158 + }, + { + "epoch": 1.8, + "learning_rate": 2.0159666019701906e-08, + "loss": 0.6839, + "step": 49159 + }, + { + "epoch": 1.8, + "learning_rate": 2.0156003955029846e-08, + "loss": 0.6932, + "step": 49160 + }, + { + "epoch": 1.8, + "learning_rate": 2.0152341890357783e-08, + "loss": 0.6948, + "step": 49161 + }, + { + "epoch": 1.8, + "learning_rate": 2.0148679825685723e-08, + "loss": 0.7045, + "step": 49162 + }, + { + "epoch": 1.8, + "learning_rate": 2.014501776101366e-08, + "loss": 0.6731, + "step": 49163 + }, + { + "epoch": 1.8, + "learning_rate": 2.0141355696341596e-08, + "loss": 0.693, + "step": 49164 + }, + { + "epoch": 1.8, + "learning_rate": 2.0137693631669533e-08, + "loss": 0.7074, + "step": 49165 + }, + { + "epoch": 1.8, + "learning_rate": 2.0134031566997472e-08, + "loss": 0.6952, + "step": 49166 + }, + { + "epoch": 1.8, + "learning_rate": 2.013036950232541e-08, + "loss": 0.6984, + "step": 49167 + }, + { + "epoch": 1.8, + "learning_rate": 2.012670743765335e-08, + "loss": 0.703, + "step": 49168 + }, + { + "epoch": 1.8, + "learning_rate": 2.0123045372981286e-08, + "loss": 0.6829, + "step": 49169 + }, + { + "epoch": 1.8, + "learning_rate": 2.0119383308309222e-08, + "loss": 0.6992, + "step": 49170 + }, + { + "epoch": 1.8, + "learning_rate": 2.0115721243637162e-08, + "loss": 0.6962, + "step": 49171 + }, + { + "epoch": 1.8, + "learning_rate": 2.01120591789651e-08, + "loss": 0.6902, + "step": 49172 + }, + { + "epoch": 1.8, + "learning_rate": 2.010839711429304e-08, + "loss": 0.6974, + "step": 49173 + }, + { + "epoch": 1.8, + "learning_rate": 2.0104735049620975e-08, + "loss": 0.6974, + "step": 49174 + }, + { + "epoch": 1.8, + "learning_rate": 2.0101072984948912e-08, + "loss": 0.7071, + "step": 49175 + }, + { + "epoch": 1.8, + "learning_rate": 2.009741092027685e-08, + "loss": 0.7005, + "step": 49176 + }, + { + "epoch": 1.8, + "learning_rate": 2.009374885560479e-08, + "loss": 0.6924, + "step": 49177 + }, + { + "epoch": 1.8, + "learning_rate": 2.009008679093273e-08, + "loss": 0.7014, + "step": 49178 + }, + { + "epoch": 1.8, + "learning_rate": 2.0086424726260665e-08, + "loss": 0.6845, + "step": 49179 + }, + { + "epoch": 1.8, + "learning_rate": 2.0082762661588605e-08, + "loss": 0.6918, + "step": 49180 + }, + { + "epoch": 1.8, + "learning_rate": 2.007910059691654e-08, + "loss": 0.6911, + "step": 49181 + }, + { + "epoch": 1.8, + "learning_rate": 2.0075438532244478e-08, + "loss": 0.6946, + "step": 49182 + }, + { + "epoch": 1.8, + "learning_rate": 2.0071776467572415e-08, + "loss": 0.6906, + "step": 49183 + }, + { + "epoch": 1.8, + "learning_rate": 2.0068114402900355e-08, + "loss": 0.6959, + "step": 49184 + }, + { + "epoch": 1.8, + "learning_rate": 2.006445233822829e-08, + "loss": 0.6936, + "step": 49185 + }, + { + "epoch": 1.8, + "learning_rate": 2.006079027355623e-08, + "loss": 0.7075, + "step": 49186 + }, + { + "epoch": 1.8, + "learning_rate": 2.0057128208884168e-08, + "loss": 0.6922, + "step": 49187 + }, + { + "epoch": 1.8, + "learning_rate": 2.0053466144212105e-08, + "loss": 0.7003, + "step": 49188 + }, + { + "epoch": 1.8, + "learning_rate": 2.0049804079540045e-08, + "loss": 0.7061, + "step": 49189 + }, + { + "epoch": 1.8, + "learning_rate": 2.004614201486798e-08, + "loss": 0.6959, + "step": 49190 + }, + { + "epoch": 1.8, + "learning_rate": 2.004247995019592e-08, + "loss": 0.6855, + "step": 49191 + }, + { + "epoch": 1.8, + "learning_rate": 2.0038817885523858e-08, + "loss": 0.7044, + "step": 49192 + }, + { + "epoch": 1.8, + "learning_rate": 2.0035155820851794e-08, + "loss": 0.695, + "step": 49193 + }, + { + "epoch": 1.8, + "learning_rate": 2.003149375617973e-08, + "loss": 0.691, + "step": 49194 + }, + { + "epoch": 1.8, + "learning_rate": 2.002783169150767e-08, + "loss": 0.686, + "step": 49195 + }, + { + "epoch": 1.8, + "learning_rate": 2.002416962683561e-08, + "loss": 0.6919, + "step": 49196 + }, + { + "epoch": 1.8, + "learning_rate": 2.0020507562163547e-08, + "loss": 0.7144, + "step": 49197 + }, + { + "epoch": 1.8, + "learning_rate": 2.0016845497491487e-08, + "loss": 0.6967, + "step": 49198 + }, + { + "epoch": 1.8, + "learning_rate": 2.001318343281942e-08, + "loss": 0.6982, + "step": 49199 + }, + { + "epoch": 1.8, + "learning_rate": 2.000952136814736e-08, + "loss": 0.6974, + "step": 49200 + }, + { + "epoch": 1.8, + "learning_rate": 2.0005859303475297e-08, + "loss": 0.7033, + "step": 49201 + }, + { + "epoch": 1.8, + "learning_rate": 2.0002197238803237e-08, + "loss": 0.7002, + "step": 49202 + }, + { + "epoch": 1.8, + "learning_rate": 1.9998535174131174e-08, + "loss": 0.6872, + "step": 49203 + }, + { + "epoch": 1.8, + "learning_rate": 1.9994873109459114e-08, + "loss": 0.684, + "step": 49204 + }, + { + "epoch": 1.8, + "learning_rate": 1.999121104478705e-08, + "loss": 0.6903, + "step": 49205 + }, + { + "epoch": 1.8, + "learning_rate": 1.9987548980114987e-08, + "loss": 0.6918, + "step": 49206 + }, + { + "epoch": 1.8, + "learning_rate": 1.9983886915442927e-08, + "loss": 0.696, + "step": 49207 + }, + { + "epoch": 1.8, + "learning_rate": 1.9980224850770863e-08, + "loss": 0.6858, + "step": 49208 + }, + { + "epoch": 1.8, + "learning_rate": 1.9976562786098803e-08, + "loss": 0.6901, + "step": 49209 + }, + { + "epoch": 1.8, + "learning_rate": 1.997290072142674e-08, + "loss": 0.6918, + "step": 49210 + }, + { + "epoch": 1.8, + "learning_rate": 1.9969238656754677e-08, + "loss": 0.6903, + "step": 49211 + }, + { + "epoch": 1.8, + "learning_rate": 1.9965576592082613e-08, + "loss": 0.6947, + "step": 49212 + }, + { + "epoch": 1.8, + "learning_rate": 1.9961914527410553e-08, + "loss": 0.6828, + "step": 49213 + }, + { + "epoch": 1.8, + "learning_rate": 1.9958252462738493e-08, + "loss": 0.7023, + "step": 49214 + }, + { + "epoch": 1.8, + "learning_rate": 1.995459039806643e-08, + "loss": 0.6939, + "step": 49215 + }, + { + "epoch": 1.8, + "learning_rate": 1.9950928333394366e-08, + "loss": 0.7033, + "step": 49216 + }, + { + "epoch": 1.8, + "learning_rate": 1.9947266268722303e-08, + "loss": 0.6888, + "step": 49217 + }, + { + "epoch": 1.8, + "learning_rate": 1.9943604204050243e-08, + "loss": 0.6881, + "step": 49218 + }, + { + "epoch": 1.8, + "learning_rate": 1.993994213937818e-08, + "loss": 0.6911, + "step": 49219 + }, + { + "epoch": 1.8, + "learning_rate": 1.993628007470612e-08, + "loss": 0.687, + "step": 49220 + }, + { + "epoch": 1.8, + "learning_rate": 1.9932618010034056e-08, + "loss": 0.6904, + "step": 49221 + }, + { + "epoch": 1.8, + "learning_rate": 1.9928955945361993e-08, + "loss": 0.6886, + "step": 49222 + }, + { + "epoch": 1.8, + "learning_rate": 1.9925293880689933e-08, + "loss": 0.687, + "step": 49223 + }, + { + "epoch": 1.8, + "learning_rate": 1.992163181601787e-08, + "loss": 0.6879, + "step": 49224 + }, + { + "epoch": 1.8, + "learning_rate": 1.991796975134581e-08, + "loss": 0.6862, + "step": 49225 + }, + { + "epoch": 1.8, + "learning_rate": 1.9914307686673746e-08, + "loss": 0.6903, + "step": 49226 + }, + { + "epoch": 1.8, + "learning_rate": 1.9910645622001686e-08, + "loss": 0.6885, + "step": 49227 + }, + { + "epoch": 1.8, + "learning_rate": 1.9906983557329622e-08, + "loss": 0.698, + "step": 49228 + }, + { + "epoch": 1.8, + "learning_rate": 1.990332149265756e-08, + "loss": 0.6876, + "step": 49229 + }, + { + "epoch": 1.8, + "learning_rate": 1.9899659427985496e-08, + "loss": 0.6829, + "step": 49230 + }, + { + "epoch": 1.8, + "learning_rate": 1.9895997363313435e-08, + "loss": 0.6967, + "step": 49231 + }, + { + "epoch": 1.8, + "learning_rate": 1.9892335298641372e-08, + "loss": 0.6981, + "step": 49232 + }, + { + "epoch": 1.8, + "learning_rate": 1.9888673233969312e-08, + "loss": 0.6983, + "step": 49233 + }, + { + "epoch": 1.8, + "learning_rate": 1.988501116929725e-08, + "loss": 0.6962, + "step": 49234 + }, + { + "epoch": 1.8, + "learning_rate": 1.9881349104625185e-08, + "loss": 0.6993, + "step": 49235 + }, + { + "epoch": 1.8, + "learning_rate": 1.9877687039953125e-08, + "loss": 0.6965, + "step": 49236 + }, + { + "epoch": 1.8, + "learning_rate": 1.9874024975281062e-08, + "loss": 0.7012, + "step": 49237 + }, + { + "epoch": 1.8, + "learning_rate": 1.9870362910609002e-08, + "loss": 0.6911, + "step": 49238 + }, + { + "epoch": 1.8, + "learning_rate": 1.986670084593694e-08, + "loss": 0.6877, + "step": 49239 + }, + { + "epoch": 1.8, + "learning_rate": 1.9863038781264875e-08, + "loss": 0.6883, + "step": 49240 + }, + { + "epoch": 1.8, + "learning_rate": 1.985937671659281e-08, + "loss": 0.6912, + "step": 49241 + }, + { + "epoch": 1.8, + "learning_rate": 1.985571465192075e-08, + "loss": 0.6828, + "step": 49242 + }, + { + "epoch": 1.8, + "learning_rate": 1.985205258724869e-08, + "loss": 0.6917, + "step": 49243 + }, + { + "epoch": 1.8, + "learning_rate": 1.9848390522576628e-08, + "loss": 0.6858, + "step": 49244 + }, + { + "epoch": 1.8, + "learning_rate": 1.9844728457904568e-08, + "loss": 0.7005, + "step": 49245 + }, + { + "epoch": 1.8, + "learning_rate": 1.98410663932325e-08, + "loss": 0.6914, + "step": 49246 + }, + { + "epoch": 1.8, + "learning_rate": 1.983740432856044e-08, + "loss": 0.6872, + "step": 49247 + }, + { + "epoch": 1.8, + "learning_rate": 1.9833742263888378e-08, + "loss": 0.6973, + "step": 49248 + }, + { + "epoch": 1.8, + "learning_rate": 1.9830080199216318e-08, + "loss": 0.6879, + "step": 49249 + }, + { + "epoch": 1.8, + "learning_rate": 1.9826418134544254e-08, + "loss": 0.6921, + "step": 49250 + }, + { + "epoch": 1.8, + "learning_rate": 1.9822756069872194e-08, + "loss": 0.6851, + "step": 49251 + }, + { + "epoch": 1.8, + "learning_rate": 1.981909400520013e-08, + "loss": 0.693, + "step": 49252 + }, + { + "epoch": 1.8, + "learning_rate": 1.9815431940528068e-08, + "loss": 0.6915, + "step": 49253 + }, + { + "epoch": 1.8, + "learning_rate": 1.9811769875856008e-08, + "loss": 0.6928, + "step": 49254 + }, + { + "epoch": 1.8, + "learning_rate": 1.9808107811183944e-08, + "loss": 0.6981, + "step": 49255 + }, + { + "epoch": 1.8, + "learning_rate": 1.9804445746511884e-08, + "loss": 0.6841, + "step": 49256 + }, + { + "epoch": 1.8, + "learning_rate": 1.980078368183982e-08, + "loss": 0.691, + "step": 49257 + }, + { + "epoch": 1.8, + "learning_rate": 1.9797121617167757e-08, + "loss": 0.6865, + "step": 49258 + }, + { + "epoch": 1.8, + "learning_rate": 1.9793459552495694e-08, + "loss": 0.6881, + "step": 49259 + }, + { + "epoch": 1.8, + "learning_rate": 1.9789797487823634e-08, + "loss": 0.6975, + "step": 49260 + }, + { + "epoch": 1.8, + "learning_rate": 1.9786135423151574e-08, + "loss": 0.6906, + "step": 49261 + }, + { + "epoch": 1.8, + "learning_rate": 1.978247335847951e-08, + "loss": 0.7027, + "step": 49262 + }, + { + "epoch": 1.8, + "learning_rate": 1.9778811293807447e-08, + "loss": 0.6922, + "step": 49263 + }, + { + "epoch": 1.8, + "learning_rate": 1.9775149229135384e-08, + "loss": 0.7082, + "step": 49264 + }, + { + "epoch": 1.8, + "learning_rate": 1.9771487164463324e-08, + "loss": 0.6939, + "step": 49265 + }, + { + "epoch": 1.8, + "learning_rate": 1.976782509979126e-08, + "loss": 0.6876, + "step": 49266 + }, + { + "epoch": 1.8, + "learning_rate": 1.97641630351192e-08, + "loss": 0.6905, + "step": 49267 + }, + { + "epoch": 1.8, + "learning_rate": 1.9760500970447137e-08, + "loss": 0.6941, + "step": 49268 + }, + { + "epoch": 1.8, + "learning_rate": 1.9756838905775077e-08, + "loss": 0.6974, + "step": 49269 + }, + { + "epoch": 1.8, + "learning_rate": 1.9753176841103013e-08, + "loss": 0.6933, + "step": 49270 + }, + { + "epoch": 1.8, + "learning_rate": 1.974951477643095e-08, + "loss": 0.7043, + "step": 49271 + }, + { + "epoch": 1.8, + "learning_rate": 1.974585271175889e-08, + "loss": 0.6798, + "step": 49272 + }, + { + "epoch": 1.8, + "learning_rate": 1.9742190647086826e-08, + "loss": 0.6909, + "step": 49273 + }, + { + "epoch": 1.8, + "learning_rate": 1.9738528582414766e-08, + "loss": 0.6904, + "step": 49274 + }, + { + "epoch": 1.8, + "learning_rate": 1.9734866517742703e-08, + "loss": 0.708, + "step": 49275 + }, + { + "epoch": 1.8, + "learning_rate": 1.973120445307064e-08, + "loss": 0.6904, + "step": 49276 + }, + { + "epoch": 1.8, + "learning_rate": 1.9727542388398576e-08, + "loss": 0.6741, + "step": 49277 + }, + { + "epoch": 1.8, + "learning_rate": 1.9723880323726516e-08, + "loss": 0.6976, + "step": 49278 + }, + { + "epoch": 1.8, + "learning_rate": 1.9720218259054456e-08, + "loss": 0.6849, + "step": 49279 + }, + { + "epoch": 1.8, + "learning_rate": 1.9716556194382393e-08, + "loss": 0.6807, + "step": 49280 + }, + { + "epoch": 1.8, + "learning_rate": 1.971289412971033e-08, + "loss": 0.6989, + "step": 49281 + }, + { + "epoch": 1.8, + "learning_rate": 1.9709232065038266e-08, + "loss": 0.7016, + "step": 49282 + }, + { + "epoch": 1.8, + "learning_rate": 1.9705570000366206e-08, + "loss": 0.6896, + "step": 49283 + }, + { + "epoch": 1.8, + "learning_rate": 1.9701907935694143e-08, + "loss": 0.6889, + "step": 49284 + }, + { + "epoch": 1.8, + "learning_rate": 1.9698245871022082e-08, + "loss": 0.6904, + "step": 49285 + }, + { + "epoch": 1.8, + "learning_rate": 1.969458380635002e-08, + "loss": 0.6945, + "step": 49286 + }, + { + "epoch": 1.8, + "learning_rate": 1.9690921741677956e-08, + "loss": 0.7039, + "step": 49287 + }, + { + "epoch": 1.8, + "learning_rate": 1.9687259677005896e-08, + "loss": 0.6774, + "step": 49288 + }, + { + "epoch": 1.8, + "learning_rate": 1.9683597612333832e-08, + "loss": 0.6861, + "step": 49289 + }, + { + "epoch": 1.8, + "learning_rate": 1.9679935547661772e-08, + "loss": 0.689, + "step": 49290 + }, + { + "epoch": 1.8, + "learning_rate": 1.967627348298971e-08, + "loss": 0.6976, + "step": 49291 + }, + { + "epoch": 1.8, + "learning_rate": 1.967261141831765e-08, + "loss": 0.6916, + "step": 49292 + }, + { + "epoch": 1.8, + "learning_rate": 1.9668949353645582e-08, + "loss": 0.697, + "step": 49293 + }, + { + "epoch": 1.8, + "learning_rate": 1.9665287288973522e-08, + "loss": 0.7017, + "step": 49294 + }, + { + "epoch": 1.8, + "learning_rate": 1.966162522430146e-08, + "loss": 0.6849, + "step": 49295 + }, + { + "epoch": 1.8, + "learning_rate": 1.96579631596294e-08, + "loss": 0.6895, + "step": 49296 + }, + { + "epoch": 1.8, + "learning_rate": 1.965430109495734e-08, + "loss": 0.6968, + "step": 49297 + }, + { + "epoch": 1.8, + "learning_rate": 1.9650639030285275e-08, + "loss": 0.6917, + "step": 49298 + }, + { + "epoch": 1.8, + "learning_rate": 1.9646976965613212e-08, + "loss": 0.6932, + "step": 49299 + }, + { + "epoch": 1.8, + "learning_rate": 1.9643314900941148e-08, + "loss": 0.7063, + "step": 49300 + }, + { + "epoch": 1.8, + "learning_rate": 1.9639652836269088e-08, + "loss": 0.6902, + "step": 49301 + }, + { + "epoch": 1.8, + "learning_rate": 1.9635990771597025e-08, + "loss": 0.6889, + "step": 49302 + }, + { + "epoch": 1.8, + "learning_rate": 1.9632328706924965e-08, + "loss": 0.6951, + "step": 49303 + }, + { + "epoch": 1.8, + "learning_rate": 1.96286666422529e-08, + "loss": 0.685, + "step": 49304 + }, + { + "epoch": 1.8, + "learning_rate": 1.9625004577580838e-08, + "loss": 0.6854, + "step": 49305 + }, + { + "epoch": 1.8, + "learning_rate": 1.9621342512908778e-08, + "loss": 0.7, + "step": 49306 + }, + { + "epoch": 1.8, + "learning_rate": 1.9617680448236715e-08, + "loss": 0.6934, + "step": 49307 + }, + { + "epoch": 1.8, + "learning_rate": 1.9614018383564654e-08, + "loss": 0.7005, + "step": 49308 + }, + { + "epoch": 1.8, + "learning_rate": 1.961035631889259e-08, + "loss": 0.6906, + "step": 49309 + }, + { + "epoch": 1.8, + "learning_rate": 1.9606694254220528e-08, + "loss": 0.6917, + "step": 49310 + }, + { + "epoch": 1.8, + "learning_rate": 1.9603032189548464e-08, + "loss": 0.6911, + "step": 49311 + }, + { + "epoch": 1.8, + "learning_rate": 1.9599370124876404e-08, + "loss": 0.6923, + "step": 49312 + }, + { + "epoch": 1.8, + "learning_rate": 1.959570806020434e-08, + "loss": 0.6792, + "step": 49313 + }, + { + "epoch": 1.8, + "learning_rate": 1.959204599553228e-08, + "loss": 0.6845, + "step": 49314 + }, + { + "epoch": 1.8, + "learning_rate": 1.958838393086022e-08, + "loss": 0.6912, + "step": 49315 + }, + { + "epoch": 1.8, + "learning_rate": 1.9584721866188157e-08, + "loss": 0.687, + "step": 49316 + }, + { + "epoch": 1.8, + "learning_rate": 1.9581059801516094e-08, + "loss": 0.7018, + "step": 49317 + }, + { + "epoch": 1.8, + "learning_rate": 1.957739773684403e-08, + "loss": 0.6979, + "step": 49318 + }, + { + "epoch": 1.8, + "learning_rate": 1.957373567217197e-08, + "loss": 0.6981, + "step": 49319 + }, + { + "epoch": 1.8, + "learning_rate": 1.9570073607499907e-08, + "loss": 0.6933, + "step": 49320 + }, + { + "epoch": 1.8, + "learning_rate": 1.9566411542827847e-08, + "loss": 0.6995, + "step": 49321 + }, + { + "epoch": 1.8, + "learning_rate": 1.9562749478155784e-08, + "loss": 0.6869, + "step": 49322 + }, + { + "epoch": 1.8, + "learning_rate": 1.955908741348372e-08, + "loss": 0.6857, + "step": 49323 + }, + { + "epoch": 1.8, + "learning_rate": 1.955542534881166e-08, + "loss": 0.6893, + "step": 49324 + }, + { + "epoch": 1.8, + "learning_rate": 1.9551763284139597e-08, + "loss": 0.6954, + "step": 49325 + }, + { + "epoch": 1.8, + "learning_rate": 1.9548101219467537e-08, + "loss": 0.6914, + "step": 49326 + }, + { + "epoch": 1.8, + "learning_rate": 1.9544439154795473e-08, + "loss": 0.6946, + "step": 49327 + }, + { + "epoch": 1.8, + "learning_rate": 1.954077709012341e-08, + "loss": 0.7064, + "step": 49328 + }, + { + "epoch": 1.8, + "learning_rate": 1.9537115025451347e-08, + "loss": 0.6943, + "step": 49329 + }, + { + "epoch": 1.8, + "learning_rate": 1.9533452960779287e-08, + "loss": 0.7008, + "step": 49330 + }, + { + "epoch": 1.8, + "learning_rate": 1.9529790896107223e-08, + "loss": 0.6973, + "step": 49331 + }, + { + "epoch": 1.8, + "learning_rate": 1.9526128831435163e-08, + "loss": 0.693, + "step": 49332 + }, + { + "epoch": 1.8, + "learning_rate": 1.95224667667631e-08, + "loss": 0.7016, + "step": 49333 + }, + { + "epoch": 1.8, + "learning_rate": 1.9518804702091036e-08, + "loss": 0.6962, + "step": 49334 + }, + { + "epoch": 1.81, + "learning_rate": 1.9515142637418976e-08, + "loss": 0.6783, + "step": 49335 + }, + { + "epoch": 1.81, + "learning_rate": 1.9511480572746913e-08, + "loss": 0.7016, + "step": 49336 + }, + { + "epoch": 1.81, + "learning_rate": 1.9507818508074853e-08, + "loss": 0.6923, + "step": 49337 + }, + { + "epoch": 1.81, + "learning_rate": 1.950415644340279e-08, + "loss": 0.6989, + "step": 49338 + }, + { + "epoch": 1.81, + "learning_rate": 1.950049437873073e-08, + "loss": 0.7069, + "step": 49339 + }, + { + "epoch": 1.81, + "learning_rate": 1.9496832314058663e-08, + "loss": 0.7006, + "step": 49340 + }, + { + "epoch": 1.81, + "learning_rate": 1.9493170249386603e-08, + "loss": 0.6966, + "step": 49341 + }, + { + "epoch": 1.81, + "learning_rate": 1.948950818471454e-08, + "loss": 0.6933, + "step": 49342 + }, + { + "epoch": 1.81, + "learning_rate": 1.948584612004248e-08, + "loss": 0.6834, + "step": 49343 + }, + { + "epoch": 1.81, + "learning_rate": 1.948218405537042e-08, + "loss": 0.6943, + "step": 49344 + }, + { + "epoch": 1.81, + "learning_rate": 1.9478521990698356e-08, + "loss": 0.6937, + "step": 49345 + }, + { + "epoch": 1.81, + "learning_rate": 1.9474859926026292e-08, + "loss": 0.6887, + "step": 49346 + }, + { + "epoch": 1.81, + "learning_rate": 1.947119786135423e-08, + "loss": 0.6944, + "step": 49347 + }, + { + "epoch": 1.81, + "learning_rate": 1.946753579668217e-08, + "loss": 0.7164, + "step": 49348 + }, + { + "epoch": 1.81, + "learning_rate": 1.9463873732010106e-08, + "loss": 0.6942, + "step": 49349 + }, + { + "epoch": 1.81, + "learning_rate": 1.9460211667338045e-08, + "loss": 0.6878, + "step": 49350 + }, + { + "epoch": 1.81, + "learning_rate": 1.9456549602665982e-08, + "loss": 0.6985, + "step": 49351 + }, + { + "epoch": 1.81, + "learning_rate": 1.945288753799392e-08, + "loss": 0.6875, + "step": 49352 + }, + { + "epoch": 1.81, + "learning_rate": 1.944922547332186e-08, + "loss": 0.693, + "step": 49353 + }, + { + "epoch": 1.81, + "learning_rate": 1.9445563408649795e-08, + "loss": 0.6866, + "step": 49354 + }, + { + "epoch": 1.81, + "learning_rate": 1.9441901343977735e-08, + "loss": 0.6858, + "step": 49355 + }, + { + "epoch": 1.81, + "learning_rate": 1.9438239279305672e-08, + "loss": 0.6904, + "step": 49356 + }, + { + "epoch": 1.81, + "learning_rate": 1.9434577214633612e-08, + "loss": 0.7032, + "step": 49357 + }, + { + "epoch": 1.81, + "learning_rate": 1.9430915149961545e-08, + "loss": 0.6805, + "step": 49358 + }, + { + "epoch": 1.81, + "learning_rate": 1.9427253085289485e-08, + "loss": 0.6997, + "step": 49359 + }, + { + "epoch": 1.81, + "learning_rate": 1.942359102061742e-08, + "loss": 0.7009, + "step": 49360 + }, + { + "epoch": 1.81, + "learning_rate": 1.941992895594536e-08, + "loss": 0.691, + "step": 49361 + }, + { + "epoch": 1.81, + "learning_rate": 1.94162668912733e-08, + "loss": 0.6906, + "step": 49362 + }, + { + "epoch": 1.81, + "learning_rate": 1.9412604826601238e-08, + "loss": 0.6933, + "step": 49363 + }, + { + "epoch": 1.81, + "learning_rate": 1.9408942761929175e-08, + "loss": 0.6877, + "step": 49364 + }, + { + "epoch": 1.81, + "learning_rate": 1.940528069725711e-08, + "loss": 0.689, + "step": 49365 + }, + { + "epoch": 1.81, + "learning_rate": 1.940161863258505e-08, + "loss": 0.6983, + "step": 49366 + }, + { + "epoch": 1.81, + "learning_rate": 1.9397956567912988e-08, + "loss": 0.6939, + "step": 49367 + }, + { + "epoch": 1.81, + "learning_rate": 1.9394294503240928e-08, + "loss": 0.6881, + "step": 49368 + }, + { + "epoch": 1.81, + "learning_rate": 1.9390632438568864e-08, + "loss": 0.6981, + "step": 49369 + }, + { + "epoch": 1.81, + "learning_rate": 1.93869703738968e-08, + "loss": 0.6861, + "step": 49370 + }, + { + "epoch": 1.81, + "learning_rate": 1.938330830922474e-08, + "loss": 0.6955, + "step": 49371 + }, + { + "epoch": 1.81, + "learning_rate": 1.9379646244552678e-08, + "loss": 0.6922, + "step": 49372 + }, + { + "epoch": 1.81, + "learning_rate": 1.9375984179880618e-08, + "loss": 0.6848, + "step": 49373 + }, + { + "epoch": 1.81, + "learning_rate": 1.9372322115208554e-08, + "loss": 0.6925, + "step": 49374 + }, + { + "epoch": 1.81, + "learning_rate": 1.936866005053649e-08, + "loss": 0.6884, + "step": 49375 + }, + { + "epoch": 1.81, + "learning_rate": 1.9364997985864427e-08, + "loss": 0.6955, + "step": 49376 + }, + { + "epoch": 1.81, + "learning_rate": 1.9361335921192367e-08, + "loss": 0.6899, + "step": 49377 + }, + { + "epoch": 1.81, + "learning_rate": 1.9357673856520304e-08, + "loss": 0.6974, + "step": 49378 + }, + { + "epoch": 1.81, + "learning_rate": 1.9354011791848244e-08, + "loss": 0.6965, + "step": 49379 + }, + { + "epoch": 1.81, + "learning_rate": 1.9350349727176184e-08, + "loss": 0.6871, + "step": 49380 + }, + { + "epoch": 1.81, + "learning_rate": 1.9346687662504117e-08, + "loss": 0.6984, + "step": 49381 + }, + { + "epoch": 1.81, + "learning_rate": 1.9343025597832057e-08, + "loss": 0.6931, + "step": 49382 + }, + { + "epoch": 1.81, + "learning_rate": 1.9339363533159994e-08, + "loss": 0.6966, + "step": 49383 + }, + { + "epoch": 1.81, + "learning_rate": 1.9335701468487934e-08, + "loss": 0.6998, + "step": 49384 + }, + { + "epoch": 1.81, + "learning_rate": 1.933203940381587e-08, + "loss": 0.7004, + "step": 49385 + }, + { + "epoch": 1.81, + "learning_rate": 1.932837733914381e-08, + "loss": 0.6907, + "step": 49386 + }, + { + "epoch": 1.81, + "learning_rate": 1.9324715274471747e-08, + "loss": 0.6963, + "step": 49387 + }, + { + "epoch": 1.81, + "learning_rate": 1.9321053209799683e-08, + "loss": 0.6952, + "step": 49388 + }, + { + "epoch": 1.81, + "learning_rate": 1.9317391145127623e-08, + "loss": 0.7003, + "step": 49389 + }, + { + "epoch": 1.81, + "learning_rate": 1.931372908045556e-08, + "loss": 0.7, + "step": 49390 + }, + { + "epoch": 1.81, + "learning_rate": 1.93100670157835e-08, + "loss": 0.6925, + "step": 49391 + }, + { + "epoch": 1.81, + "learning_rate": 1.9306404951111436e-08, + "loss": 0.688, + "step": 49392 + }, + { + "epoch": 1.81, + "learning_rate": 1.9302742886439373e-08, + "loss": 0.6956, + "step": 49393 + }, + { + "epoch": 1.81, + "learning_rate": 1.929908082176731e-08, + "loss": 0.68, + "step": 49394 + }, + { + "epoch": 1.81, + "learning_rate": 1.929541875709525e-08, + "loss": 0.7006, + "step": 49395 + }, + { + "epoch": 1.81, + "learning_rate": 1.9291756692423186e-08, + "loss": 0.6916, + "step": 49396 + }, + { + "epoch": 1.81, + "learning_rate": 1.9288094627751126e-08, + "loss": 0.681, + "step": 49397 + }, + { + "epoch": 1.81, + "learning_rate": 1.9284432563079063e-08, + "loss": 0.686, + "step": 49398 + }, + { + "epoch": 1.81, + "learning_rate": 1.9280770498407e-08, + "loss": 0.703, + "step": 49399 + }, + { + "epoch": 1.81, + "learning_rate": 1.927710843373494e-08, + "loss": 0.6883, + "step": 49400 + }, + { + "epoch": 1.81, + "learning_rate": 1.9273446369062876e-08, + "loss": 0.7019, + "step": 49401 + }, + { + "epoch": 1.81, + "learning_rate": 1.9269784304390816e-08, + "loss": 0.6806, + "step": 49402 + }, + { + "epoch": 1.81, + "learning_rate": 1.9266122239718752e-08, + "loss": 0.6956, + "step": 49403 + }, + { + "epoch": 1.81, + "learning_rate": 1.9262460175046692e-08, + "loss": 0.6865, + "step": 49404 + }, + { + "epoch": 1.81, + "learning_rate": 1.9258798110374626e-08, + "loss": 0.697, + "step": 49405 + }, + { + "epoch": 1.81, + "learning_rate": 1.9255136045702566e-08, + "loss": 0.6903, + "step": 49406 + }, + { + "epoch": 1.81, + "learning_rate": 1.9251473981030506e-08, + "loss": 0.6894, + "step": 49407 + }, + { + "epoch": 1.81, + "learning_rate": 1.9247811916358442e-08, + "loss": 0.6957, + "step": 49408 + }, + { + "epoch": 1.81, + "learning_rate": 1.9244149851686382e-08, + "loss": 0.6862, + "step": 49409 + }, + { + "epoch": 1.81, + "learning_rate": 1.924048778701432e-08, + "loss": 0.6963, + "step": 49410 + }, + { + "epoch": 1.81, + "learning_rate": 1.9236825722342255e-08, + "loss": 0.695, + "step": 49411 + }, + { + "epoch": 1.81, + "learning_rate": 1.9233163657670192e-08, + "loss": 0.6977, + "step": 49412 + }, + { + "epoch": 1.81, + "learning_rate": 1.9229501592998132e-08, + "loss": 0.6905, + "step": 49413 + }, + { + "epoch": 1.81, + "learning_rate": 1.922583952832607e-08, + "loss": 0.6927, + "step": 49414 + }, + { + "epoch": 1.81, + "learning_rate": 1.922217746365401e-08, + "loss": 0.7027, + "step": 49415 + }, + { + "epoch": 1.81, + "learning_rate": 1.9218515398981945e-08, + "loss": 0.6988, + "step": 49416 + }, + { + "epoch": 1.81, + "learning_rate": 1.9214853334309882e-08, + "loss": 0.7053, + "step": 49417 + }, + { + "epoch": 1.81, + "learning_rate": 1.921119126963782e-08, + "loss": 0.6919, + "step": 49418 + }, + { + "epoch": 1.81, + "learning_rate": 1.9207529204965758e-08, + "loss": 0.6872, + "step": 49419 + }, + { + "epoch": 1.81, + "learning_rate": 1.9203867140293698e-08, + "loss": 0.6952, + "step": 49420 + }, + { + "epoch": 1.81, + "learning_rate": 1.9200205075621635e-08, + "loss": 0.6945, + "step": 49421 + }, + { + "epoch": 1.81, + "learning_rate": 1.919654301094957e-08, + "loss": 0.6887, + "step": 49422 + }, + { + "epoch": 1.81, + "learning_rate": 1.9192880946277508e-08, + "loss": 0.6876, + "step": 49423 + }, + { + "epoch": 1.81, + "learning_rate": 1.9189218881605448e-08, + "loss": 0.6849, + "step": 49424 + }, + { + "epoch": 1.81, + "learning_rate": 1.9185556816933385e-08, + "loss": 0.6945, + "step": 49425 + }, + { + "epoch": 1.81, + "learning_rate": 1.9181894752261325e-08, + "loss": 0.6855, + "step": 49426 + }, + { + "epoch": 1.81, + "learning_rate": 1.9178232687589264e-08, + "loss": 0.6898, + "step": 49427 + }, + { + "epoch": 1.81, + "learning_rate": 1.9174570622917198e-08, + "loss": 0.6863, + "step": 49428 + }, + { + "epoch": 1.81, + "learning_rate": 1.9170908558245138e-08, + "loss": 0.6976, + "step": 49429 + }, + { + "epoch": 1.81, + "learning_rate": 1.9167246493573074e-08, + "loss": 0.7057, + "step": 49430 + }, + { + "epoch": 1.81, + "learning_rate": 1.9163584428901014e-08, + "loss": 0.69, + "step": 49431 + }, + { + "epoch": 1.81, + "learning_rate": 1.915992236422895e-08, + "loss": 0.6918, + "step": 49432 + }, + { + "epoch": 1.81, + "learning_rate": 1.915626029955689e-08, + "loss": 0.6845, + "step": 49433 + }, + { + "epoch": 1.81, + "learning_rate": 1.9152598234884827e-08, + "loss": 0.6918, + "step": 49434 + }, + { + "epoch": 1.81, + "learning_rate": 1.9148936170212764e-08, + "loss": 0.695, + "step": 49435 + }, + { + "epoch": 1.81, + "learning_rate": 1.9145274105540704e-08, + "loss": 0.6923, + "step": 49436 + }, + { + "epoch": 1.81, + "learning_rate": 1.914161204086864e-08, + "loss": 0.6951, + "step": 49437 + }, + { + "epoch": 1.81, + "learning_rate": 1.913794997619658e-08, + "loss": 0.7023, + "step": 49438 + }, + { + "epoch": 1.81, + "learning_rate": 1.9134287911524517e-08, + "loss": 0.693, + "step": 49439 + }, + { + "epoch": 1.81, + "learning_rate": 1.9130625846852454e-08, + "loss": 0.6992, + "step": 49440 + }, + { + "epoch": 1.81, + "learning_rate": 1.912696378218039e-08, + "loss": 0.7019, + "step": 49441 + }, + { + "epoch": 1.81, + "learning_rate": 1.912330171750833e-08, + "loss": 0.6897, + "step": 49442 + }, + { + "epoch": 1.81, + "learning_rate": 1.9119639652836267e-08, + "loss": 0.6955, + "step": 49443 + }, + { + "epoch": 1.81, + "learning_rate": 1.9115977588164207e-08, + "loss": 0.6853, + "step": 49444 + }, + { + "epoch": 1.81, + "learning_rate": 1.9112315523492147e-08, + "loss": 0.7033, + "step": 49445 + }, + { + "epoch": 1.81, + "learning_rate": 1.910865345882008e-08, + "loss": 0.6908, + "step": 49446 + }, + { + "epoch": 1.81, + "learning_rate": 1.910499139414802e-08, + "loss": 0.698, + "step": 49447 + }, + { + "epoch": 1.81, + "learning_rate": 1.9101329329475957e-08, + "loss": 0.6871, + "step": 49448 + }, + { + "epoch": 1.81, + "learning_rate": 1.9097667264803897e-08, + "loss": 0.6999, + "step": 49449 + }, + { + "epoch": 1.81, + "learning_rate": 1.9094005200131833e-08, + "loss": 0.6886, + "step": 49450 + }, + { + "epoch": 1.81, + "learning_rate": 1.9090343135459773e-08, + "loss": 0.6975, + "step": 49451 + }, + { + "epoch": 1.81, + "learning_rate": 1.9086681070787706e-08, + "loss": 0.6995, + "step": 49452 + }, + { + "epoch": 1.81, + "learning_rate": 1.9083019006115646e-08, + "loss": 0.7051, + "step": 49453 + }, + { + "epoch": 1.81, + "learning_rate": 1.9079356941443586e-08, + "loss": 0.6946, + "step": 49454 + }, + { + "epoch": 1.81, + "learning_rate": 1.9075694876771523e-08, + "loss": 0.6826, + "step": 49455 + }, + { + "epoch": 1.81, + "learning_rate": 1.9072032812099463e-08, + "loss": 0.698, + "step": 49456 + }, + { + "epoch": 1.81, + "learning_rate": 1.90683707474274e-08, + "loss": 0.6957, + "step": 49457 + }, + { + "epoch": 1.81, + "learning_rate": 1.9064708682755336e-08, + "loss": 0.6849, + "step": 49458 + }, + { + "epoch": 1.81, + "learning_rate": 1.9061046618083273e-08, + "loss": 0.6834, + "step": 49459 + }, + { + "epoch": 1.81, + "learning_rate": 1.9057384553411213e-08, + "loss": 0.7052, + "step": 49460 + }, + { + "epoch": 1.81, + "learning_rate": 1.905372248873915e-08, + "loss": 0.6987, + "step": 49461 + }, + { + "epoch": 1.81, + "learning_rate": 1.905006042406709e-08, + "loss": 0.6878, + "step": 49462 + }, + { + "epoch": 1.81, + "learning_rate": 1.9046398359395026e-08, + "loss": 0.6927, + "step": 49463 + }, + { + "epoch": 1.81, + "learning_rate": 1.9042736294722962e-08, + "loss": 0.702, + "step": 49464 + }, + { + "epoch": 1.81, + "learning_rate": 1.9039074230050902e-08, + "loss": 0.695, + "step": 49465 + }, + { + "epoch": 1.81, + "learning_rate": 1.903541216537884e-08, + "loss": 0.694, + "step": 49466 + }, + { + "epoch": 1.81, + "learning_rate": 1.903175010070678e-08, + "loss": 0.7014, + "step": 49467 + }, + { + "epoch": 1.81, + "learning_rate": 1.9028088036034716e-08, + "loss": 0.6889, + "step": 49468 + }, + { + "epoch": 1.81, + "learning_rate": 1.9024425971362652e-08, + "loss": 0.7013, + "step": 49469 + }, + { + "epoch": 1.81, + "learning_rate": 1.902076390669059e-08, + "loss": 0.6877, + "step": 49470 + }, + { + "epoch": 1.81, + "learning_rate": 1.901710184201853e-08, + "loss": 0.7026, + "step": 49471 + }, + { + "epoch": 1.81, + "learning_rate": 1.901343977734647e-08, + "loss": 0.6844, + "step": 49472 + }, + { + "epoch": 1.81, + "learning_rate": 1.9009777712674405e-08, + "loss": 0.6949, + "step": 49473 + }, + { + "epoch": 1.81, + "learning_rate": 1.9006115648002345e-08, + "loss": 0.6955, + "step": 49474 + }, + { + "epoch": 1.81, + "learning_rate": 1.9002453583330282e-08, + "loss": 0.6831, + "step": 49475 + }, + { + "epoch": 1.81, + "learning_rate": 1.899879151865822e-08, + "loss": 0.6905, + "step": 49476 + }, + { + "epoch": 1.81, + "learning_rate": 1.8995129453986155e-08, + "loss": 0.6931, + "step": 49477 + }, + { + "epoch": 1.81, + "learning_rate": 1.8991467389314095e-08, + "loss": 0.6803, + "step": 49478 + }, + { + "epoch": 1.81, + "learning_rate": 1.898780532464203e-08, + "loss": 0.696, + "step": 49479 + }, + { + "epoch": 1.81, + "learning_rate": 1.898414325996997e-08, + "loss": 0.7076, + "step": 49480 + }, + { + "epoch": 1.81, + "learning_rate": 1.8980481195297908e-08, + "loss": 0.6974, + "step": 49481 + }, + { + "epoch": 1.81, + "learning_rate": 1.8976819130625845e-08, + "loss": 0.6918, + "step": 49482 + }, + { + "epoch": 1.81, + "learning_rate": 1.8973157065953785e-08, + "loss": 0.6876, + "step": 49483 + }, + { + "epoch": 1.81, + "learning_rate": 1.896949500128172e-08, + "loss": 0.6876, + "step": 49484 + }, + { + "epoch": 1.81, + "learning_rate": 1.896583293660966e-08, + "loss": 0.6948, + "step": 49485 + }, + { + "epoch": 1.81, + "learning_rate": 1.8962170871937598e-08, + "loss": 0.6959, + "step": 49486 + }, + { + "epoch": 1.81, + "learning_rate": 1.8958508807265534e-08, + "loss": 0.697, + "step": 49487 + }, + { + "epoch": 1.81, + "learning_rate": 1.895484674259347e-08, + "loss": 0.6872, + "step": 49488 + }, + { + "epoch": 1.81, + "learning_rate": 1.895118467792141e-08, + "loss": 0.6887, + "step": 49489 + }, + { + "epoch": 1.81, + "learning_rate": 1.894752261324935e-08, + "loss": 0.6768, + "step": 49490 + }, + { + "epoch": 1.81, + "learning_rate": 1.8943860548577288e-08, + "loss": 0.6937, + "step": 49491 + }, + { + "epoch": 1.81, + "learning_rate": 1.8940198483905227e-08, + "loss": 0.6908, + "step": 49492 + }, + { + "epoch": 1.81, + "learning_rate": 1.893653641923316e-08, + "loss": 0.7097, + "step": 49493 + }, + { + "epoch": 1.81, + "learning_rate": 1.89328743545611e-08, + "loss": 0.6959, + "step": 49494 + }, + { + "epoch": 1.81, + "learning_rate": 1.8929212289889037e-08, + "loss": 0.697, + "step": 49495 + }, + { + "epoch": 1.81, + "learning_rate": 1.8925550225216977e-08, + "loss": 0.6993, + "step": 49496 + }, + { + "epoch": 1.81, + "learning_rate": 1.8921888160544914e-08, + "loss": 0.6874, + "step": 49497 + }, + { + "epoch": 1.81, + "learning_rate": 1.8918226095872854e-08, + "loss": 0.6969, + "step": 49498 + }, + { + "epoch": 1.81, + "learning_rate": 1.891456403120079e-08, + "loss": 0.69, + "step": 49499 + }, + { + "epoch": 1.81, + "learning_rate": 1.8910901966528727e-08, + "loss": 0.7006, + "step": 49500 + }, + { + "epoch": 1.81, + "learning_rate": 1.8907239901856667e-08, + "loss": 0.6865, + "step": 49501 + }, + { + "epoch": 1.81, + "learning_rate": 1.8903577837184604e-08, + "loss": 0.6928, + "step": 49502 + }, + { + "epoch": 1.81, + "learning_rate": 1.8899915772512544e-08, + "loss": 0.6941, + "step": 49503 + }, + { + "epoch": 1.81, + "learning_rate": 1.889625370784048e-08, + "loss": 0.689, + "step": 49504 + }, + { + "epoch": 1.81, + "learning_rate": 1.8892591643168417e-08, + "loss": 0.694, + "step": 49505 + }, + { + "epoch": 1.81, + "learning_rate": 1.8888929578496353e-08, + "loss": 0.7058, + "step": 49506 + }, + { + "epoch": 1.81, + "learning_rate": 1.8885267513824293e-08, + "loss": 0.6867, + "step": 49507 + }, + { + "epoch": 1.81, + "learning_rate": 1.8881605449152233e-08, + "loss": 0.7067, + "step": 49508 + }, + { + "epoch": 1.81, + "learning_rate": 1.887794338448017e-08, + "loss": 0.693, + "step": 49509 + }, + { + "epoch": 1.81, + "learning_rate": 1.8874281319808106e-08, + "loss": 0.698, + "step": 49510 + }, + { + "epoch": 1.81, + "learning_rate": 1.8870619255136043e-08, + "loss": 0.6937, + "step": 49511 + }, + { + "epoch": 1.81, + "learning_rate": 1.8866957190463983e-08, + "loss": 0.6903, + "step": 49512 + }, + { + "epoch": 1.81, + "learning_rate": 1.886329512579192e-08, + "loss": 0.6994, + "step": 49513 + }, + { + "epoch": 1.81, + "learning_rate": 1.885963306111986e-08, + "loss": 0.6857, + "step": 49514 + }, + { + "epoch": 1.81, + "learning_rate": 1.8855970996447796e-08, + "loss": 0.6853, + "step": 49515 + }, + { + "epoch": 1.81, + "learning_rate": 1.8852308931775733e-08, + "loss": 0.6869, + "step": 49516 + }, + { + "epoch": 1.81, + "learning_rate": 1.8848646867103673e-08, + "loss": 0.6983, + "step": 49517 + }, + { + "epoch": 1.81, + "learning_rate": 1.884498480243161e-08, + "loss": 0.6942, + "step": 49518 + }, + { + "epoch": 1.81, + "learning_rate": 1.884132273775955e-08, + "loss": 0.7125, + "step": 49519 + }, + { + "epoch": 1.81, + "learning_rate": 1.8837660673087486e-08, + "loss": 0.7028, + "step": 49520 + }, + { + "epoch": 1.81, + "learning_rate": 1.8833998608415426e-08, + "loss": 0.687, + "step": 49521 + }, + { + "epoch": 1.81, + "learning_rate": 1.8830336543743362e-08, + "loss": 0.6948, + "step": 49522 + }, + { + "epoch": 1.81, + "learning_rate": 1.88266744790713e-08, + "loss": 0.6844, + "step": 49523 + }, + { + "epoch": 1.81, + "learning_rate": 1.8823012414399236e-08, + "loss": 0.6899, + "step": 49524 + }, + { + "epoch": 1.81, + "learning_rate": 1.8819350349727176e-08, + "loss": 0.7002, + "step": 49525 + }, + { + "epoch": 1.81, + "learning_rate": 1.8815688285055112e-08, + "loss": 0.6774, + "step": 49526 + }, + { + "epoch": 1.81, + "learning_rate": 1.8812026220383052e-08, + "loss": 0.7019, + "step": 49527 + }, + { + "epoch": 1.81, + "learning_rate": 1.880836415571099e-08, + "loss": 0.6892, + "step": 49528 + }, + { + "epoch": 1.81, + "learning_rate": 1.8804702091038925e-08, + "loss": 0.702, + "step": 49529 + }, + { + "epoch": 1.81, + "learning_rate": 1.8801040026366865e-08, + "loss": 0.6958, + "step": 49530 + }, + { + "epoch": 1.81, + "learning_rate": 1.8797377961694802e-08, + "loss": 0.6951, + "step": 49531 + }, + { + "epoch": 1.81, + "learning_rate": 1.8793715897022742e-08, + "loss": 0.7024, + "step": 49532 + }, + { + "epoch": 1.81, + "learning_rate": 1.879005383235068e-08, + "loss": 0.6962, + "step": 49533 + }, + { + "epoch": 1.81, + "learning_rate": 1.8786391767678615e-08, + "loss": 0.6988, + "step": 49534 + }, + { + "epoch": 1.81, + "learning_rate": 1.8782729703006552e-08, + "loss": 0.6953, + "step": 49535 + }, + { + "epoch": 1.81, + "learning_rate": 1.8779067638334492e-08, + "loss": 0.7021, + "step": 49536 + }, + { + "epoch": 1.81, + "learning_rate": 1.877540557366243e-08, + "loss": 0.6875, + "step": 49537 + }, + { + "epoch": 1.81, + "learning_rate": 1.8771743508990368e-08, + "loss": 0.7019, + "step": 49538 + }, + { + "epoch": 1.81, + "learning_rate": 1.8768081444318308e-08, + "loss": 0.6949, + "step": 49539 + }, + { + "epoch": 1.81, + "learning_rate": 1.876441937964624e-08, + "loss": 0.6876, + "step": 49540 + }, + { + "epoch": 1.81, + "learning_rate": 1.876075731497418e-08, + "loss": 0.6992, + "step": 49541 + }, + { + "epoch": 1.81, + "learning_rate": 1.8757095250302118e-08, + "loss": 0.6932, + "step": 49542 + }, + { + "epoch": 1.81, + "learning_rate": 1.8753433185630058e-08, + "loss": 0.693, + "step": 49543 + }, + { + "epoch": 1.81, + "learning_rate": 1.8749771120957995e-08, + "loss": 0.6908, + "step": 49544 + }, + { + "epoch": 1.81, + "learning_rate": 1.8746109056285934e-08, + "loss": 0.6912, + "step": 49545 + }, + { + "epoch": 1.81, + "learning_rate": 1.874244699161387e-08, + "loss": 0.6873, + "step": 49546 + }, + { + "epoch": 1.81, + "learning_rate": 1.8738784926941808e-08, + "loss": 0.6996, + "step": 49547 + }, + { + "epoch": 1.81, + "learning_rate": 1.8735122862269748e-08, + "loss": 0.687, + "step": 49548 + }, + { + "epoch": 1.81, + "learning_rate": 1.8731460797597684e-08, + "loss": 0.6856, + "step": 49549 + }, + { + "epoch": 1.81, + "learning_rate": 1.8727798732925624e-08, + "loss": 0.6927, + "step": 49550 + }, + { + "epoch": 1.81, + "learning_rate": 1.872413666825356e-08, + "loss": 0.6856, + "step": 49551 + }, + { + "epoch": 1.81, + "learning_rate": 1.8720474603581497e-08, + "loss": 0.6904, + "step": 49552 + }, + { + "epoch": 1.81, + "learning_rate": 1.8716812538909434e-08, + "loss": 0.7017, + "step": 49553 + }, + { + "epoch": 1.81, + "learning_rate": 1.8713150474237374e-08, + "loss": 0.6734, + "step": 49554 + }, + { + "epoch": 1.81, + "learning_rate": 1.8709488409565314e-08, + "loss": 0.6839, + "step": 49555 + }, + { + "epoch": 1.81, + "learning_rate": 1.870582634489325e-08, + "loss": 0.6931, + "step": 49556 + }, + { + "epoch": 1.81, + "learning_rate": 1.8702164280221187e-08, + "loss": 0.6896, + "step": 49557 + }, + { + "epoch": 1.81, + "learning_rate": 1.8698502215549124e-08, + "loss": 0.6955, + "step": 49558 + }, + { + "epoch": 1.81, + "learning_rate": 1.8694840150877064e-08, + "loss": 0.68, + "step": 49559 + }, + { + "epoch": 1.81, + "learning_rate": 1.8691178086205e-08, + "loss": 0.6829, + "step": 49560 + }, + { + "epoch": 1.81, + "learning_rate": 1.868751602153294e-08, + "loss": 0.6906, + "step": 49561 + }, + { + "epoch": 1.81, + "learning_rate": 1.8683853956860877e-08, + "loss": 0.7026, + "step": 49562 + }, + { + "epoch": 1.81, + "learning_rate": 1.8680191892188817e-08, + "loss": 0.7015, + "step": 49563 + }, + { + "epoch": 1.81, + "learning_rate": 1.8676529827516753e-08, + "loss": 0.6978, + "step": 49564 + }, + { + "epoch": 1.81, + "learning_rate": 1.867286776284469e-08, + "loss": 0.6908, + "step": 49565 + }, + { + "epoch": 1.81, + "learning_rate": 1.866920569817263e-08, + "loss": 0.696, + "step": 49566 + }, + { + "epoch": 1.81, + "learning_rate": 1.8665543633500567e-08, + "loss": 0.6959, + "step": 49567 + }, + { + "epoch": 1.81, + "learning_rate": 1.8661881568828507e-08, + "loss": 0.6998, + "step": 49568 + }, + { + "epoch": 1.81, + "learning_rate": 1.8658219504156443e-08, + "loss": 0.6894, + "step": 49569 + }, + { + "epoch": 1.81, + "learning_rate": 1.865455743948438e-08, + "loss": 0.6849, + "step": 49570 + }, + { + "epoch": 1.81, + "learning_rate": 1.8650895374812316e-08, + "loss": 0.6857, + "step": 49571 + }, + { + "epoch": 1.81, + "learning_rate": 1.8647233310140256e-08, + "loss": 0.7029, + "step": 49572 + }, + { + "epoch": 1.81, + "learning_rate": 1.8643571245468196e-08, + "loss": 0.7071, + "step": 49573 + }, + { + "epoch": 1.81, + "learning_rate": 1.8639909180796133e-08, + "loss": 0.6903, + "step": 49574 + }, + { + "epoch": 1.81, + "learning_rate": 1.863624711612407e-08, + "loss": 0.6822, + "step": 49575 + }, + { + "epoch": 1.81, + "learning_rate": 1.8632585051452006e-08, + "loss": 0.6912, + "step": 49576 + }, + { + "epoch": 1.81, + "learning_rate": 1.8628922986779946e-08, + "loss": 0.6861, + "step": 49577 + }, + { + "epoch": 1.81, + "learning_rate": 1.8625260922107883e-08, + "loss": 0.6946, + "step": 49578 + }, + { + "epoch": 1.81, + "learning_rate": 1.8621598857435823e-08, + "loss": 0.6897, + "step": 49579 + }, + { + "epoch": 1.81, + "learning_rate": 1.861793679276376e-08, + "loss": 0.6916, + "step": 49580 + }, + { + "epoch": 1.81, + "learning_rate": 1.8614274728091696e-08, + "loss": 0.6761, + "step": 49581 + }, + { + "epoch": 1.81, + "learning_rate": 1.8610612663419636e-08, + "loss": 0.6932, + "step": 49582 + }, + { + "epoch": 1.81, + "learning_rate": 1.8606950598747572e-08, + "loss": 0.6856, + "step": 49583 + }, + { + "epoch": 1.81, + "learning_rate": 1.8603288534075512e-08, + "loss": 0.6929, + "step": 49584 + }, + { + "epoch": 1.81, + "learning_rate": 1.859962646940345e-08, + "loss": 0.6976, + "step": 49585 + }, + { + "epoch": 1.81, + "learning_rate": 1.859596440473139e-08, + "loss": 0.692, + "step": 49586 + }, + { + "epoch": 1.81, + "learning_rate": 1.8592302340059322e-08, + "loss": 0.6935, + "step": 49587 + }, + { + "epoch": 1.81, + "learning_rate": 1.8588640275387262e-08, + "loss": 0.7026, + "step": 49588 + }, + { + "epoch": 1.81, + "learning_rate": 1.85849782107152e-08, + "loss": 0.6937, + "step": 49589 + }, + { + "epoch": 1.81, + "learning_rate": 1.858131614604314e-08, + "loss": 0.6892, + "step": 49590 + }, + { + "epoch": 1.81, + "learning_rate": 1.857765408137108e-08, + "loss": 0.6849, + "step": 49591 + }, + { + "epoch": 1.81, + "learning_rate": 1.8573992016699015e-08, + "loss": 0.6888, + "step": 49592 + }, + { + "epoch": 1.81, + "learning_rate": 1.8570329952026952e-08, + "loss": 0.6965, + "step": 49593 + }, + { + "epoch": 1.81, + "learning_rate": 1.856666788735489e-08, + "loss": 0.7037, + "step": 49594 + }, + { + "epoch": 1.81, + "learning_rate": 1.856300582268283e-08, + "loss": 0.6882, + "step": 49595 + }, + { + "epoch": 1.81, + "learning_rate": 1.8559343758010765e-08, + "loss": 0.7137, + "step": 49596 + }, + { + "epoch": 1.81, + "learning_rate": 1.8555681693338705e-08, + "loss": 0.6901, + "step": 49597 + }, + { + "epoch": 1.81, + "learning_rate": 1.855201962866664e-08, + "loss": 0.6913, + "step": 49598 + }, + { + "epoch": 1.81, + "learning_rate": 1.8548357563994578e-08, + "loss": 0.6944, + "step": 49599 + }, + { + "epoch": 1.81, + "learning_rate": 1.8544695499322518e-08, + "loss": 0.6887, + "step": 49600 + }, + { + "epoch": 1.81, + "learning_rate": 1.8541033434650455e-08, + "loss": 0.6912, + "step": 49601 + }, + { + "epoch": 1.81, + "learning_rate": 1.8537371369978395e-08, + "loss": 0.6943, + "step": 49602 + }, + { + "epoch": 1.81, + "learning_rate": 1.853370930530633e-08, + "loss": 0.6949, + "step": 49603 + }, + { + "epoch": 1.81, + "learning_rate": 1.8530047240634268e-08, + "loss": 0.7003, + "step": 49604 + }, + { + "epoch": 1.81, + "learning_rate": 1.8526385175962204e-08, + "loss": 0.7131, + "step": 49605 + }, + { + "epoch": 1.81, + "learning_rate": 1.8522723111290144e-08, + "loss": 0.6998, + "step": 49606 + }, + { + "epoch": 1.81, + "learning_rate": 1.851906104661808e-08, + "loss": 0.6853, + "step": 49607 + }, + { + "epoch": 1.82, + "learning_rate": 1.851539898194602e-08, + "loss": 0.6841, + "step": 49608 + }, + { + "epoch": 1.82, + "learning_rate": 1.851173691727396e-08, + "loss": 0.6934, + "step": 49609 + }, + { + "epoch": 1.82, + "learning_rate": 1.8508074852601898e-08, + "loss": 0.6915, + "step": 49610 + }, + { + "epoch": 1.82, + "learning_rate": 1.8504412787929834e-08, + "loss": 0.6881, + "step": 49611 + }, + { + "epoch": 1.82, + "learning_rate": 1.850075072325777e-08, + "loss": 0.7024, + "step": 49612 + }, + { + "epoch": 1.82, + "learning_rate": 1.849708865858571e-08, + "loss": 0.6907, + "step": 49613 + }, + { + "epoch": 1.82, + "learning_rate": 1.8493426593913647e-08, + "loss": 0.6949, + "step": 49614 + }, + { + "epoch": 1.82, + "learning_rate": 1.8489764529241587e-08, + "loss": 0.6966, + "step": 49615 + }, + { + "epoch": 1.82, + "learning_rate": 1.8486102464569524e-08, + "loss": 0.7039, + "step": 49616 + }, + { + "epoch": 1.82, + "learning_rate": 1.848244039989746e-08, + "loss": 0.6897, + "step": 49617 + }, + { + "epoch": 1.82, + "learning_rate": 1.8478778335225397e-08, + "loss": 0.6943, + "step": 49618 + }, + { + "epoch": 1.82, + "learning_rate": 1.8475116270553337e-08, + "loss": 0.6978, + "step": 49619 + }, + { + "epoch": 1.82, + "learning_rate": 1.8471454205881277e-08, + "loss": 0.6886, + "step": 49620 + }, + { + "epoch": 1.82, + "learning_rate": 1.8467792141209214e-08, + "loss": 0.7069, + "step": 49621 + }, + { + "epoch": 1.82, + "learning_rate": 1.846413007653715e-08, + "loss": 0.6994, + "step": 49622 + }, + { + "epoch": 1.82, + "learning_rate": 1.8460468011865087e-08, + "loss": 0.7009, + "step": 49623 + }, + { + "epoch": 1.82, + "learning_rate": 1.8456805947193027e-08, + "loss": 0.7034, + "step": 49624 + }, + { + "epoch": 1.82, + "learning_rate": 1.8453143882520963e-08, + "loss": 0.6904, + "step": 49625 + }, + { + "epoch": 1.82, + "learning_rate": 1.8449481817848903e-08, + "loss": 0.6983, + "step": 49626 + }, + { + "epoch": 1.82, + "learning_rate": 1.844581975317684e-08, + "loss": 0.6956, + "step": 49627 + }, + { + "epoch": 1.82, + "learning_rate": 1.8442157688504777e-08, + "loss": 0.6883, + "step": 49628 + }, + { + "epoch": 1.82, + "learning_rate": 1.8438495623832716e-08, + "loss": 0.691, + "step": 49629 + }, + { + "epoch": 1.82, + "learning_rate": 1.8434833559160653e-08, + "loss": 0.6892, + "step": 49630 + }, + { + "epoch": 1.82, + "learning_rate": 1.8431171494488593e-08, + "loss": 0.7015, + "step": 49631 + }, + { + "epoch": 1.82, + "learning_rate": 1.842750942981653e-08, + "loss": 0.6945, + "step": 49632 + }, + { + "epoch": 1.82, + "learning_rate": 1.842384736514447e-08, + "loss": 0.6923, + "step": 49633 + }, + { + "epoch": 1.82, + "learning_rate": 1.8420185300472403e-08, + "loss": 0.6831, + "step": 49634 + }, + { + "epoch": 1.82, + "learning_rate": 1.8416523235800343e-08, + "loss": 0.6983, + "step": 49635 + }, + { + "epoch": 1.82, + "learning_rate": 1.841286117112828e-08, + "loss": 0.6831, + "step": 49636 + }, + { + "epoch": 1.82, + "learning_rate": 1.840919910645622e-08, + "loss": 0.6966, + "step": 49637 + }, + { + "epoch": 1.82, + "learning_rate": 1.840553704178416e-08, + "loss": 0.6925, + "step": 49638 + }, + { + "epoch": 1.82, + "learning_rate": 1.8401874977112096e-08, + "loss": 0.6921, + "step": 49639 + }, + { + "epoch": 1.82, + "learning_rate": 1.8398212912440033e-08, + "loss": 0.6856, + "step": 49640 + }, + { + "epoch": 1.82, + "learning_rate": 1.839455084776797e-08, + "loss": 0.6974, + "step": 49641 + }, + { + "epoch": 1.82, + "learning_rate": 1.839088878309591e-08, + "loss": 0.6906, + "step": 49642 + }, + { + "epoch": 1.82, + "learning_rate": 1.8387226718423846e-08, + "loss": 0.7002, + "step": 49643 + }, + { + "epoch": 1.82, + "learning_rate": 1.8383564653751786e-08, + "loss": 0.7055, + "step": 49644 + }, + { + "epoch": 1.82, + "learning_rate": 1.8379902589079722e-08, + "loss": 0.6948, + "step": 49645 + }, + { + "epoch": 1.82, + "learning_rate": 1.837624052440766e-08, + "loss": 0.6974, + "step": 49646 + }, + { + "epoch": 1.82, + "learning_rate": 1.83725784597356e-08, + "loss": 0.6884, + "step": 49647 + }, + { + "epoch": 1.82, + "learning_rate": 1.8368916395063535e-08, + "loss": 0.701, + "step": 49648 + }, + { + "epoch": 1.82, + "learning_rate": 1.8365254330391475e-08, + "loss": 0.7081, + "step": 49649 + }, + { + "epoch": 1.82, + "learning_rate": 1.8361592265719412e-08, + "loss": 0.6921, + "step": 49650 + }, + { + "epoch": 1.82, + "learning_rate": 1.8357930201047352e-08, + "loss": 0.695, + "step": 49651 + }, + { + "epoch": 1.82, + "learning_rate": 1.8354268136375285e-08, + "loss": 0.6871, + "step": 49652 + }, + { + "epoch": 1.82, + "learning_rate": 1.8350606071703225e-08, + "loss": 0.6951, + "step": 49653 + }, + { + "epoch": 1.82, + "learning_rate": 1.8346944007031162e-08, + "loss": 0.7017, + "step": 49654 + }, + { + "epoch": 1.82, + "learning_rate": 1.83432819423591e-08, + "loss": 0.6947, + "step": 49655 + }, + { + "epoch": 1.82, + "learning_rate": 1.833961987768704e-08, + "loss": 0.7005, + "step": 49656 + }, + { + "epoch": 1.82, + "learning_rate": 1.8335957813014978e-08, + "loss": 0.6963, + "step": 49657 + }, + { + "epoch": 1.82, + "learning_rate": 1.8332295748342915e-08, + "loss": 0.7058, + "step": 49658 + }, + { + "epoch": 1.82, + "learning_rate": 1.832863368367085e-08, + "loss": 0.6889, + "step": 49659 + }, + { + "epoch": 1.82, + "learning_rate": 1.832497161899879e-08, + "loss": 0.7069, + "step": 49660 + }, + { + "epoch": 1.82, + "learning_rate": 1.8321309554326728e-08, + "loss": 0.6899, + "step": 49661 + }, + { + "epoch": 1.82, + "learning_rate": 1.8317647489654668e-08, + "loss": 0.6933, + "step": 49662 + }, + { + "epoch": 1.82, + "learning_rate": 1.8313985424982605e-08, + "loss": 0.6961, + "step": 49663 + }, + { + "epoch": 1.82, + "learning_rate": 1.831032336031054e-08, + "loss": 0.7149, + "step": 49664 + }, + { + "epoch": 1.82, + "learning_rate": 1.830666129563848e-08, + "loss": 0.702, + "step": 49665 + }, + { + "epoch": 1.82, + "learning_rate": 1.8302999230966418e-08, + "loss": 0.706, + "step": 49666 + }, + { + "epoch": 1.82, + "learning_rate": 1.8299337166294358e-08, + "loss": 0.6917, + "step": 49667 + }, + { + "epoch": 1.82, + "learning_rate": 1.8295675101622294e-08, + "loss": 0.6928, + "step": 49668 + }, + { + "epoch": 1.82, + "learning_rate": 1.829201303695023e-08, + "loss": 0.6867, + "step": 49669 + }, + { + "epoch": 1.82, + "learning_rate": 1.8288350972278167e-08, + "loss": 0.6827, + "step": 49670 + }, + { + "epoch": 1.82, + "learning_rate": 1.8284688907606107e-08, + "loss": 0.687, + "step": 49671 + }, + { + "epoch": 1.82, + "learning_rate": 1.8281026842934044e-08, + "loss": 0.6922, + "step": 49672 + }, + { + "epoch": 1.82, + "learning_rate": 1.8277364778261984e-08, + "loss": 0.6915, + "step": 49673 + }, + { + "epoch": 1.82, + "learning_rate": 1.8273702713589924e-08, + "loss": 0.6873, + "step": 49674 + }, + { + "epoch": 1.82, + "learning_rate": 1.8270040648917857e-08, + "loss": 0.6936, + "step": 49675 + }, + { + "epoch": 1.82, + "learning_rate": 1.8266378584245797e-08, + "loss": 0.6978, + "step": 49676 + }, + { + "epoch": 1.82, + "learning_rate": 1.8262716519573734e-08, + "loss": 0.6864, + "step": 49677 + }, + { + "epoch": 1.82, + "learning_rate": 1.8259054454901674e-08, + "loss": 0.7015, + "step": 49678 + }, + { + "epoch": 1.82, + "learning_rate": 1.825539239022961e-08, + "loss": 0.692, + "step": 49679 + }, + { + "epoch": 1.82, + "learning_rate": 1.825173032555755e-08, + "loss": 0.6899, + "step": 49680 + }, + { + "epoch": 1.82, + "learning_rate": 1.8248068260885487e-08, + "loss": 0.699, + "step": 49681 + }, + { + "epoch": 1.82, + "learning_rate": 1.8244406196213423e-08, + "loss": 0.6827, + "step": 49682 + }, + { + "epoch": 1.82, + "learning_rate": 1.8240744131541363e-08, + "loss": 0.6959, + "step": 49683 + }, + { + "epoch": 1.82, + "learning_rate": 1.82370820668693e-08, + "loss": 0.6919, + "step": 49684 + }, + { + "epoch": 1.82, + "learning_rate": 1.823342000219724e-08, + "loss": 0.7015, + "step": 49685 + }, + { + "epoch": 1.82, + "learning_rate": 1.8229757937525177e-08, + "loss": 0.6973, + "step": 49686 + }, + { + "epoch": 1.82, + "learning_rate": 1.8226095872853113e-08, + "loss": 0.6857, + "step": 49687 + }, + { + "epoch": 1.82, + "learning_rate": 1.822243380818105e-08, + "loss": 0.6748, + "step": 49688 + }, + { + "epoch": 1.82, + "learning_rate": 1.821877174350899e-08, + "loss": 0.6908, + "step": 49689 + }, + { + "epoch": 1.82, + "learning_rate": 1.8215109678836926e-08, + "loss": 0.6894, + "step": 49690 + }, + { + "epoch": 1.82, + "learning_rate": 1.8211447614164866e-08, + "loss": 0.6946, + "step": 49691 + }, + { + "epoch": 1.82, + "learning_rate": 1.8207785549492803e-08, + "loss": 0.6914, + "step": 49692 + }, + { + "epoch": 1.82, + "learning_rate": 1.820412348482074e-08, + "loss": 0.6957, + "step": 49693 + }, + { + "epoch": 1.82, + "learning_rate": 1.820046142014868e-08, + "loss": 0.6933, + "step": 49694 + }, + { + "epoch": 1.82, + "learning_rate": 1.8196799355476616e-08, + "loss": 0.6805, + "step": 49695 + }, + { + "epoch": 1.82, + "learning_rate": 1.8193137290804556e-08, + "loss": 0.695, + "step": 49696 + }, + { + "epoch": 1.82, + "learning_rate": 1.8189475226132493e-08, + "loss": 0.6927, + "step": 49697 + }, + { + "epoch": 1.82, + "learning_rate": 1.8185813161460433e-08, + "loss": 0.6886, + "step": 49698 + }, + { + "epoch": 1.82, + "learning_rate": 1.8182151096788366e-08, + "loss": 0.6996, + "step": 49699 + }, + { + "epoch": 1.82, + "learning_rate": 1.8178489032116306e-08, + "loss": 0.6976, + "step": 49700 + }, + { + "epoch": 1.82, + "learning_rate": 1.8174826967444246e-08, + "loss": 0.6851, + "step": 49701 + }, + { + "epoch": 1.82, + "learning_rate": 1.8171164902772182e-08, + "loss": 0.6878, + "step": 49702 + }, + { + "epoch": 1.82, + "learning_rate": 1.8167502838100122e-08, + "loss": 0.6957, + "step": 49703 + }, + { + "epoch": 1.82, + "learning_rate": 1.816384077342806e-08, + "loss": 0.7005, + "step": 49704 + }, + { + "epoch": 1.82, + "learning_rate": 1.8160178708755996e-08, + "loss": 0.7026, + "step": 49705 + }, + { + "epoch": 1.82, + "learning_rate": 1.8156516644083932e-08, + "loss": 0.7027, + "step": 49706 + }, + { + "epoch": 1.82, + "learning_rate": 1.8152854579411872e-08, + "loss": 0.6853, + "step": 49707 + }, + { + "epoch": 1.82, + "learning_rate": 1.814919251473981e-08, + "loss": 0.6841, + "step": 49708 + }, + { + "epoch": 1.82, + "learning_rate": 1.814553045006775e-08, + "loss": 0.6952, + "step": 49709 + }, + { + "epoch": 1.82, + "learning_rate": 1.8141868385395685e-08, + "loss": 0.6855, + "step": 49710 + }, + { + "epoch": 1.82, + "learning_rate": 1.8138206320723622e-08, + "loss": 0.6811, + "step": 49711 + }, + { + "epoch": 1.82, + "learning_rate": 1.8134544256051562e-08, + "loss": 0.7018, + "step": 49712 + }, + { + "epoch": 1.82, + "learning_rate": 1.81308821913795e-08, + "loss": 0.7031, + "step": 49713 + }, + { + "epoch": 1.82, + "learning_rate": 1.8127220126707438e-08, + "loss": 0.6969, + "step": 49714 + }, + { + "epoch": 1.82, + "learning_rate": 1.8123558062035375e-08, + "loss": 0.6919, + "step": 49715 + }, + { + "epoch": 1.82, + "learning_rate": 1.811989599736331e-08, + "loss": 0.6894, + "step": 49716 + }, + { + "epoch": 1.82, + "learning_rate": 1.8116233932691248e-08, + "loss": 0.6894, + "step": 49717 + }, + { + "epoch": 1.82, + "learning_rate": 1.8112571868019188e-08, + "loss": 0.697, + "step": 49718 + }, + { + "epoch": 1.82, + "learning_rate": 1.8108909803347125e-08, + "loss": 0.7041, + "step": 49719 + }, + { + "epoch": 1.82, + "learning_rate": 1.8105247738675065e-08, + "loss": 0.7, + "step": 49720 + }, + { + "epoch": 1.82, + "learning_rate": 1.8101585674003005e-08, + "loss": 0.7041, + "step": 49721 + }, + { + "epoch": 1.82, + "learning_rate": 1.8097923609330938e-08, + "loss": 0.6955, + "step": 49722 + }, + { + "epoch": 1.82, + "learning_rate": 1.8094261544658878e-08, + "loss": 0.7002, + "step": 49723 + }, + { + "epoch": 1.82, + "learning_rate": 1.8090599479986814e-08, + "loss": 0.6866, + "step": 49724 + }, + { + "epoch": 1.82, + "learning_rate": 1.8086937415314754e-08, + "loss": 0.6967, + "step": 49725 + }, + { + "epoch": 1.82, + "learning_rate": 1.808327535064269e-08, + "loss": 0.6914, + "step": 49726 + }, + { + "epoch": 1.82, + "learning_rate": 1.807961328597063e-08, + "loss": 0.6936, + "step": 49727 + }, + { + "epoch": 1.82, + "learning_rate": 1.8075951221298568e-08, + "loss": 0.6959, + "step": 49728 + }, + { + "epoch": 1.82, + "learning_rate": 1.8072289156626504e-08, + "loss": 0.6906, + "step": 49729 + }, + { + "epoch": 1.82, + "learning_rate": 1.8068627091954444e-08, + "loss": 0.7018, + "step": 49730 + }, + { + "epoch": 1.82, + "learning_rate": 1.806496502728238e-08, + "loss": 0.6952, + "step": 49731 + }, + { + "epoch": 1.82, + "learning_rate": 1.806130296261032e-08, + "loss": 0.6943, + "step": 49732 + }, + { + "epoch": 1.82, + "learning_rate": 1.8057640897938257e-08, + "loss": 0.6922, + "step": 49733 + }, + { + "epoch": 1.82, + "learning_rate": 1.8053978833266194e-08, + "loss": 0.6921, + "step": 49734 + }, + { + "epoch": 1.82, + "learning_rate": 1.805031676859413e-08, + "loss": 0.696, + "step": 49735 + }, + { + "epoch": 1.82, + "learning_rate": 1.804665470392207e-08, + "loss": 0.7016, + "step": 49736 + }, + { + "epoch": 1.82, + "learning_rate": 1.8042992639250007e-08, + "loss": 0.6894, + "step": 49737 + }, + { + "epoch": 1.82, + "learning_rate": 1.8039330574577947e-08, + "loss": 0.6912, + "step": 49738 + }, + { + "epoch": 1.82, + "learning_rate": 1.8035668509905887e-08, + "loss": 0.6941, + "step": 49739 + }, + { + "epoch": 1.82, + "learning_rate": 1.803200644523382e-08, + "loss": 0.69, + "step": 49740 + }, + { + "epoch": 1.82, + "learning_rate": 1.802834438056176e-08, + "loss": 0.6862, + "step": 49741 + }, + { + "epoch": 1.82, + "learning_rate": 1.8024682315889697e-08, + "loss": 0.688, + "step": 49742 + }, + { + "epoch": 1.82, + "learning_rate": 1.8021020251217637e-08, + "loss": 0.6827, + "step": 49743 + }, + { + "epoch": 1.82, + "learning_rate": 1.8017358186545573e-08, + "loss": 0.6978, + "step": 49744 + }, + { + "epoch": 1.82, + "learning_rate": 1.8013696121873513e-08, + "loss": 0.6914, + "step": 49745 + }, + { + "epoch": 1.82, + "learning_rate": 1.8010034057201447e-08, + "loss": 0.6953, + "step": 49746 + }, + { + "epoch": 1.82, + "learning_rate": 1.8006371992529386e-08, + "loss": 0.6966, + "step": 49747 + }, + { + "epoch": 1.82, + "learning_rate": 1.8002709927857326e-08, + "loss": 0.6964, + "step": 49748 + }, + { + "epoch": 1.82, + "learning_rate": 1.7999047863185263e-08, + "loss": 0.6982, + "step": 49749 + }, + { + "epoch": 1.82, + "learning_rate": 1.7995385798513203e-08, + "loss": 0.6808, + "step": 49750 + }, + { + "epoch": 1.82, + "learning_rate": 1.799172373384114e-08, + "loss": 0.6854, + "step": 49751 + }, + { + "epoch": 1.82, + "learning_rate": 1.7988061669169076e-08, + "loss": 0.7147, + "step": 49752 + }, + { + "epoch": 1.82, + "learning_rate": 1.7984399604497013e-08, + "loss": 0.6892, + "step": 49753 + }, + { + "epoch": 1.82, + "learning_rate": 1.7980737539824953e-08, + "loss": 0.6917, + "step": 49754 + }, + { + "epoch": 1.82, + "learning_rate": 1.797707547515289e-08, + "loss": 0.7042, + "step": 49755 + }, + { + "epoch": 1.82, + "learning_rate": 1.797341341048083e-08, + "loss": 0.6972, + "step": 49756 + }, + { + "epoch": 1.82, + "learning_rate": 1.7969751345808766e-08, + "loss": 0.6869, + "step": 49757 + }, + { + "epoch": 1.82, + "learning_rate": 1.7966089281136703e-08, + "loss": 0.6906, + "step": 49758 + }, + { + "epoch": 1.82, + "learning_rate": 1.7962427216464642e-08, + "loss": 0.7011, + "step": 49759 + }, + { + "epoch": 1.82, + "learning_rate": 1.795876515179258e-08, + "loss": 0.6917, + "step": 49760 + }, + { + "epoch": 1.82, + "learning_rate": 1.795510308712052e-08, + "loss": 0.6934, + "step": 49761 + }, + { + "epoch": 1.82, + "learning_rate": 1.7951441022448456e-08, + "loss": 0.6969, + "step": 49762 + }, + { + "epoch": 1.82, + "learning_rate": 1.7947778957776392e-08, + "loss": 0.6807, + "step": 49763 + }, + { + "epoch": 1.82, + "learning_rate": 1.794411689310433e-08, + "loss": 0.7036, + "step": 49764 + }, + { + "epoch": 1.82, + "learning_rate": 1.794045482843227e-08, + "loss": 0.6899, + "step": 49765 + }, + { + "epoch": 1.82, + "learning_rate": 1.793679276376021e-08, + "loss": 0.6886, + "step": 49766 + }, + { + "epoch": 1.82, + "learning_rate": 1.7933130699088145e-08, + "loss": 0.6987, + "step": 49767 + }, + { + "epoch": 1.82, + "learning_rate": 1.7929468634416085e-08, + "loss": 0.6922, + "step": 49768 + }, + { + "epoch": 1.82, + "learning_rate": 1.7925806569744022e-08, + "loss": 0.6895, + "step": 49769 + }, + { + "epoch": 1.82, + "learning_rate": 1.792214450507196e-08, + "loss": 0.6969, + "step": 49770 + }, + { + "epoch": 1.82, + "learning_rate": 1.7918482440399895e-08, + "loss": 0.6834, + "step": 49771 + }, + { + "epoch": 1.82, + "learning_rate": 1.7914820375727835e-08, + "loss": 0.6935, + "step": 49772 + }, + { + "epoch": 1.82, + "learning_rate": 1.7911158311055772e-08, + "loss": 0.6911, + "step": 49773 + }, + { + "epoch": 1.82, + "learning_rate": 1.790749624638371e-08, + "loss": 0.7098, + "step": 49774 + }, + { + "epoch": 1.82, + "learning_rate": 1.7903834181711648e-08, + "loss": 0.6929, + "step": 49775 + }, + { + "epoch": 1.82, + "learning_rate": 1.7900172117039585e-08, + "loss": 0.6808, + "step": 49776 + }, + { + "epoch": 1.82, + "learning_rate": 1.7896510052367525e-08, + "loss": 0.6822, + "step": 49777 + }, + { + "epoch": 1.82, + "learning_rate": 1.789284798769546e-08, + "loss": 0.7115, + "step": 49778 + }, + { + "epoch": 1.82, + "learning_rate": 1.78891859230234e-08, + "loss": 0.6918, + "step": 49779 + }, + { + "epoch": 1.82, + "learning_rate": 1.7885523858351338e-08, + "loss": 0.698, + "step": 49780 + }, + { + "epoch": 1.82, + "learning_rate": 1.7881861793679275e-08, + "loss": 0.6947, + "step": 49781 + }, + { + "epoch": 1.82, + "learning_rate": 1.787819972900721e-08, + "loss": 0.6826, + "step": 49782 + }, + { + "epoch": 1.82, + "learning_rate": 1.787453766433515e-08, + "loss": 0.7, + "step": 49783 + }, + { + "epoch": 1.82, + "learning_rate": 1.787087559966309e-08, + "loss": 0.7047, + "step": 49784 + }, + { + "epoch": 1.82, + "learning_rate": 1.7867213534991028e-08, + "loss": 0.698, + "step": 49785 + }, + { + "epoch": 1.82, + "learning_rate": 1.7863551470318968e-08, + "loss": 0.7007, + "step": 49786 + }, + { + "epoch": 1.82, + "learning_rate": 1.78598894056469e-08, + "loss": 0.6833, + "step": 49787 + }, + { + "epoch": 1.82, + "learning_rate": 1.785622734097484e-08, + "loss": 0.6981, + "step": 49788 + }, + { + "epoch": 1.82, + "learning_rate": 1.7852565276302777e-08, + "loss": 0.6894, + "step": 49789 + }, + { + "epoch": 1.82, + "learning_rate": 1.7848903211630717e-08, + "loss": 0.6909, + "step": 49790 + }, + { + "epoch": 1.82, + "learning_rate": 1.7845241146958654e-08, + "loss": 0.6766, + "step": 49791 + }, + { + "epoch": 1.82, + "learning_rate": 1.7841579082286594e-08, + "loss": 0.6855, + "step": 49792 + }, + { + "epoch": 1.82, + "learning_rate": 1.783791701761453e-08, + "loss": 0.6969, + "step": 49793 + }, + { + "epoch": 1.82, + "learning_rate": 1.7834254952942467e-08, + "loss": 0.6886, + "step": 49794 + }, + { + "epoch": 1.82, + "learning_rate": 1.7830592888270407e-08, + "loss": 0.6815, + "step": 49795 + }, + { + "epoch": 1.82, + "learning_rate": 1.7826930823598344e-08, + "loss": 0.6927, + "step": 49796 + }, + { + "epoch": 1.82, + "learning_rate": 1.7823268758926284e-08, + "loss": 0.7009, + "step": 49797 + }, + { + "epoch": 1.82, + "learning_rate": 1.781960669425422e-08, + "loss": 0.7093, + "step": 49798 + }, + { + "epoch": 1.82, + "learning_rate": 1.7815944629582157e-08, + "loss": 0.7004, + "step": 49799 + }, + { + "epoch": 1.82, + "learning_rate": 1.7812282564910094e-08, + "loss": 0.6917, + "step": 49800 + }, + { + "epoch": 1.82, + "learning_rate": 1.7808620500238033e-08, + "loss": 0.6964, + "step": 49801 + }, + { + "epoch": 1.82, + "learning_rate": 1.7804958435565973e-08, + "loss": 0.684, + "step": 49802 + }, + { + "epoch": 1.82, + "learning_rate": 1.780129637089391e-08, + "loss": 0.6978, + "step": 49803 + }, + { + "epoch": 1.82, + "learning_rate": 1.7797634306221847e-08, + "loss": 0.7112, + "step": 49804 + }, + { + "epoch": 1.82, + "learning_rate": 1.7793972241549783e-08, + "loss": 0.6955, + "step": 49805 + }, + { + "epoch": 1.82, + "learning_rate": 1.7790310176877723e-08, + "loss": 0.6955, + "step": 49806 + }, + { + "epoch": 1.82, + "learning_rate": 1.778664811220566e-08, + "loss": 0.6871, + "step": 49807 + }, + { + "epoch": 1.82, + "learning_rate": 1.77829860475336e-08, + "loss": 0.6826, + "step": 49808 + }, + { + "epoch": 1.82, + "learning_rate": 1.7779323982861536e-08, + "loss": 0.6875, + "step": 49809 + }, + { + "epoch": 1.82, + "learning_rate": 1.7775661918189473e-08, + "loss": 0.6912, + "step": 49810 + }, + { + "epoch": 1.82, + "learning_rate": 1.777199985351741e-08, + "loss": 0.7082, + "step": 49811 + }, + { + "epoch": 1.82, + "learning_rate": 1.776833778884535e-08, + "loss": 0.6996, + "step": 49812 + }, + { + "epoch": 1.82, + "learning_rate": 1.776467572417329e-08, + "loss": 0.6921, + "step": 49813 + }, + { + "epoch": 1.82, + "learning_rate": 1.7761013659501226e-08, + "loss": 0.6908, + "step": 49814 + }, + { + "epoch": 1.82, + "learning_rate": 1.7757351594829166e-08, + "loss": 0.6853, + "step": 49815 + }, + { + "epoch": 1.82, + "learning_rate": 1.7753689530157103e-08, + "loss": 0.6853, + "step": 49816 + }, + { + "epoch": 1.82, + "learning_rate": 1.775002746548504e-08, + "loss": 0.6913, + "step": 49817 + }, + { + "epoch": 1.82, + "learning_rate": 1.7746365400812976e-08, + "loss": 0.6939, + "step": 49818 + }, + { + "epoch": 1.82, + "learning_rate": 1.7742703336140916e-08, + "loss": 0.6952, + "step": 49819 + }, + { + "epoch": 1.82, + "learning_rate": 1.7739041271468852e-08, + "loss": 0.7018, + "step": 49820 + }, + { + "epoch": 1.82, + "learning_rate": 1.7735379206796792e-08, + "loss": 0.6962, + "step": 49821 + }, + { + "epoch": 1.82, + "learning_rate": 1.773171714212473e-08, + "loss": 0.6904, + "step": 49822 + }, + { + "epoch": 1.82, + "learning_rate": 1.7728055077452666e-08, + "loss": 0.6948, + "step": 49823 + }, + { + "epoch": 1.82, + "learning_rate": 1.7724393012780605e-08, + "loss": 0.6931, + "step": 49824 + }, + { + "epoch": 1.82, + "learning_rate": 1.7720730948108542e-08, + "loss": 0.6969, + "step": 49825 + }, + { + "epoch": 1.82, + "learning_rate": 1.7717068883436482e-08, + "loss": 0.6987, + "step": 49826 + }, + { + "epoch": 1.82, + "learning_rate": 1.771340681876442e-08, + "loss": 0.7042, + "step": 49827 + }, + { + "epoch": 1.82, + "learning_rate": 1.7709744754092355e-08, + "loss": 0.6974, + "step": 49828 + }, + { + "epoch": 1.82, + "learning_rate": 1.7706082689420292e-08, + "loss": 0.6823, + "step": 49829 + }, + { + "epoch": 1.82, + "learning_rate": 1.7702420624748232e-08, + "loss": 0.6811, + "step": 49830 + }, + { + "epoch": 1.82, + "learning_rate": 1.7698758560076172e-08, + "loss": 0.6913, + "step": 49831 + }, + { + "epoch": 1.82, + "learning_rate": 1.769509649540411e-08, + "loss": 0.6812, + "step": 49832 + }, + { + "epoch": 1.82, + "learning_rate": 1.7691434430732048e-08, + "loss": 0.6903, + "step": 49833 + }, + { + "epoch": 1.82, + "learning_rate": 1.768777236605998e-08, + "loss": 0.6905, + "step": 49834 + }, + { + "epoch": 1.82, + "learning_rate": 1.768411030138792e-08, + "loss": 0.6962, + "step": 49835 + }, + { + "epoch": 1.82, + "learning_rate": 1.7680448236715858e-08, + "loss": 0.7076, + "step": 49836 + }, + { + "epoch": 1.82, + "learning_rate": 1.7676786172043798e-08, + "loss": 0.691, + "step": 49837 + }, + { + "epoch": 1.82, + "learning_rate": 1.7673124107371735e-08, + "loss": 0.6958, + "step": 49838 + }, + { + "epoch": 1.82, + "learning_rate": 1.7669462042699675e-08, + "loss": 0.6887, + "step": 49839 + }, + { + "epoch": 1.82, + "learning_rate": 1.766579997802761e-08, + "loss": 0.6896, + "step": 49840 + }, + { + "epoch": 1.82, + "learning_rate": 1.7662137913355548e-08, + "loss": 0.7037, + "step": 49841 + }, + { + "epoch": 1.82, + "learning_rate": 1.7658475848683488e-08, + "loss": 0.6954, + "step": 49842 + }, + { + "epoch": 1.82, + "learning_rate": 1.7654813784011424e-08, + "loss": 0.7109, + "step": 49843 + }, + { + "epoch": 1.82, + "learning_rate": 1.7651151719339364e-08, + "loss": 0.694, + "step": 49844 + }, + { + "epoch": 1.82, + "learning_rate": 1.76474896546673e-08, + "loss": 0.694, + "step": 49845 + }, + { + "epoch": 1.82, + "learning_rate": 1.7643827589995238e-08, + "loss": 0.6904, + "step": 49846 + }, + { + "epoch": 1.82, + "learning_rate": 1.7640165525323174e-08, + "loss": 0.6981, + "step": 49847 + }, + { + "epoch": 1.82, + "learning_rate": 1.7636503460651114e-08, + "loss": 0.6971, + "step": 49848 + }, + { + "epoch": 1.82, + "learning_rate": 1.7632841395979054e-08, + "loss": 0.6833, + "step": 49849 + }, + { + "epoch": 1.82, + "learning_rate": 1.762917933130699e-08, + "loss": 0.6909, + "step": 49850 + }, + { + "epoch": 1.82, + "learning_rate": 1.7625517266634927e-08, + "loss": 0.6893, + "step": 49851 + }, + { + "epoch": 1.82, + "learning_rate": 1.7621855201962864e-08, + "loss": 0.6939, + "step": 49852 + }, + { + "epoch": 1.82, + "learning_rate": 1.7618193137290804e-08, + "loss": 0.696, + "step": 49853 + }, + { + "epoch": 1.82, + "learning_rate": 1.761453107261874e-08, + "loss": 0.6904, + "step": 49854 + }, + { + "epoch": 1.82, + "learning_rate": 1.761086900794668e-08, + "loss": 0.6837, + "step": 49855 + }, + { + "epoch": 1.82, + "learning_rate": 1.7607206943274617e-08, + "loss": 0.6856, + "step": 49856 + }, + { + "epoch": 1.82, + "learning_rate": 1.7603544878602557e-08, + "loss": 0.697, + "step": 49857 + }, + { + "epoch": 1.82, + "learning_rate": 1.7599882813930494e-08, + "loss": 0.6883, + "step": 49858 + }, + { + "epoch": 1.82, + "learning_rate": 1.759622074925843e-08, + "loss": 0.695, + "step": 49859 + }, + { + "epoch": 1.82, + "learning_rate": 1.759255868458637e-08, + "loss": 0.6901, + "step": 49860 + }, + { + "epoch": 1.82, + "learning_rate": 1.7588896619914307e-08, + "loss": 0.6951, + "step": 49861 + }, + { + "epoch": 1.82, + "learning_rate": 1.7585234555242247e-08, + "loss": 0.6866, + "step": 49862 + }, + { + "epoch": 1.82, + "learning_rate": 1.7581572490570183e-08, + "loss": 0.6883, + "step": 49863 + }, + { + "epoch": 1.82, + "learning_rate": 1.757791042589812e-08, + "loss": 0.6897, + "step": 49864 + }, + { + "epoch": 1.82, + "learning_rate": 1.7574248361226057e-08, + "loss": 0.6844, + "step": 49865 + }, + { + "epoch": 1.82, + "learning_rate": 1.7570586296553996e-08, + "loss": 0.6999, + "step": 49866 + }, + { + "epoch": 1.82, + "learning_rate": 1.7566924231881936e-08, + "loss": 0.6899, + "step": 49867 + }, + { + "epoch": 1.82, + "learning_rate": 1.7563262167209873e-08, + "loss": 0.6941, + "step": 49868 + }, + { + "epoch": 1.82, + "learning_rate": 1.755960010253781e-08, + "loss": 0.6938, + "step": 49869 + }, + { + "epoch": 1.82, + "learning_rate": 1.7555938037865746e-08, + "loss": 0.7027, + "step": 49870 + }, + { + "epoch": 1.82, + "learning_rate": 1.7552275973193686e-08, + "loss": 0.6925, + "step": 49871 + }, + { + "epoch": 1.82, + "learning_rate": 1.7548613908521623e-08, + "loss": 0.6938, + "step": 49872 + }, + { + "epoch": 1.82, + "learning_rate": 1.7544951843849563e-08, + "loss": 0.6989, + "step": 49873 + }, + { + "epoch": 1.82, + "learning_rate": 1.75412897791775e-08, + "loss": 0.6922, + "step": 49874 + }, + { + "epoch": 1.82, + "learning_rate": 1.7537627714505436e-08, + "loss": 0.6883, + "step": 49875 + }, + { + "epoch": 1.82, + "learning_rate": 1.7533965649833376e-08, + "loss": 0.6982, + "step": 49876 + }, + { + "epoch": 1.82, + "learning_rate": 1.7530303585161313e-08, + "loss": 0.6806, + "step": 49877 + }, + { + "epoch": 1.82, + "learning_rate": 1.7526641520489252e-08, + "loss": 0.6853, + "step": 49878 + }, + { + "epoch": 1.82, + "learning_rate": 1.752297945581719e-08, + "loss": 0.6952, + "step": 49879 + }, + { + "epoch": 1.82, + "learning_rate": 1.751931739114513e-08, + "loss": 0.6987, + "step": 49880 + }, + { + "epoch": 1.83, + "learning_rate": 1.7515655326473062e-08, + "loss": 0.6897, + "step": 49881 + }, + { + "epoch": 1.83, + "learning_rate": 1.7511993261801002e-08, + "loss": 0.6923, + "step": 49882 + }, + { + "epoch": 1.83, + "learning_rate": 1.750833119712894e-08, + "loss": 0.6969, + "step": 49883 + }, + { + "epoch": 1.83, + "learning_rate": 1.750466913245688e-08, + "loss": 0.7048, + "step": 49884 + }, + { + "epoch": 1.83, + "learning_rate": 1.750100706778482e-08, + "loss": 0.6912, + "step": 49885 + }, + { + "epoch": 1.83, + "learning_rate": 1.7497345003112755e-08, + "loss": 0.6949, + "step": 49886 + }, + { + "epoch": 1.83, + "learning_rate": 1.7493682938440692e-08, + "loss": 0.6946, + "step": 49887 + }, + { + "epoch": 1.83, + "learning_rate": 1.749002087376863e-08, + "loss": 0.6936, + "step": 49888 + }, + { + "epoch": 1.83, + "learning_rate": 1.748635880909657e-08, + "loss": 0.6983, + "step": 49889 + }, + { + "epoch": 1.83, + "learning_rate": 1.7482696744424505e-08, + "loss": 0.6847, + "step": 49890 + }, + { + "epoch": 1.83, + "learning_rate": 1.7479034679752445e-08, + "loss": 0.7006, + "step": 49891 + }, + { + "epoch": 1.83, + "learning_rate": 1.7475372615080382e-08, + "loss": 0.6952, + "step": 49892 + }, + { + "epoch": 1.83, + "learning_rate": 1.7471710550408318e-08, + "loss": 0.6826, + "step": 49893 + }, + { + "epoch": 1.83, + "learning_rate": 1.7468048485736258e-08, + "loss": 0.6902, + "step": 49894 + }, + { + "epoch": 1.83, + "learning_rate": 1.7464386421064195e-08, + "loss": 0.7172, + "step": 49895 + }, + { + "epoch": 1.83, + "learning_rate": 1.7460724356392135e-08, + "loss": 0.6822, + "step": 49896 + }, + { + "epoch": 1.83, + "learning_rate": 1.745706229172007e-08, + "loss": 0.6957, + "step": 49897 + }, + { + "epoch": 1.83, + "learning_rate": 1.7453400227048008e-08, + "loss": 0.6997, + "step": 49898 + }, + { + "epoch": 1.83, + "learning_rate": 1.7449738162375945e-08, + "loss": 0.7016, + "step": 49899 + }, + { + "epoch": 1.83, + "learning_rate": 1.7446076097703885e-08, + "loss": 0.6873, + "step": 49900 + }, + { + "epoch": 1.83, + "learning_rate": 1.744241403303182e-08, + "loss": 0.6986, + "step": 49901 + }, + { + "epoch": 1.83, + "learning_rate": 1.743875196835976e-08, + "loss": 0.6951, + "step": 49902 + }, + { + "epoch": 1.83, + "learning_rate": 1.74350899036877e-08, + "loss": 0.6848, + "step": 49903 + }, + { + "epoch": 1.83, + "learning_rate": 1.7431427839015638e-08, + "loss": 0.6956, + "step": 49904 + }, + { + "epoch": 1.83, + "learning_rate": 1.7427765774343574e-08, + "loss": 0.6865, + "step": 49905 + }, + { + "epoch": 1.83, + "learning_rate": 1.742410370967151e-08, + "loss": 0.6962, + "step": 49906 + }, + { + "epoch": 1.83, + "learning_rate": 1.742044164499945e-08, + "loss": 0.6971, + "step": 49907 + }, + { + "epoch": 1.83, + "learning_rate": 1.7416779580327387e-08, + "loss": 0.6929, + "step": 49908 + }, + { + "epoch": 1.83, + "learning_rate": 1.7413117515655327e-08, + "loss": 0.6762, + "step": 49909 + }, + { + "epoch": 1.83, + "learning_rate": 1.7409455450983264e-08, + "loss": 0.694, + "step": 49910 + }, + { + "epoch": 1.83, + "learning_rate": 1.74057933863112e-08, + "loss": 0.691, + "step": 49911 + }, + { + "epoch": 1.83, + "learning_rate": 1.7402131321639137e-08, + "loss": 0.6916, + "step": 49912 + }, + { + "epoch": 1.83, + "learning_rate": 1.7398469256967077e-08, + "loss": 0.6965, + "step": 49913 + }, + { + "epoch": 1.83, + "learning_rate": 1.7394807192295017e-08, + "loss": 0.6916, + "step": 49914 + }, + { + "epoch": 1.83, + "learning_rate": 1.7391145127622954e-08, + "loss": 0.6917, + "step": 49915 + }, + { + "epoch": 1.83, + "learning_rate": 1.738748306295089e-08, + "loss": 0.6997, + "step": 49916 + }, + { + "epoch": 1.83, + "learning_rate": 1.7383820998278827e-08, + "loss": 0.7087, + "step": 49917 + }, + { + "epoch": 1.83, + "learning_rate": 1.7380158933606767e-08, + "loss": 0.6937, + "step": 49918 + }, + { + "epoch": 1.83, + "learning_rate": 1.7376496868934703e-08, + "loss": 0.6775, + "step": 49919 + }, + { + "epoch": 1.83, + "learning_rate": 1.7372834804262643e-08, + "loss": 0.688, + "step": 49920 + }, + { + "epoch": 1.83, + "learning_rate": 1.736917273959058e-08, + "loss": 0.6994, + "step": 49921 + }, + { + "epoch": 1.83, + "learning_rate": 1.7365510674918517e-08, + "loss": 0.7012, + "step": 49922 + }, + { + "epoch": 1.83, + "learning_rate": 1.7361848610246457e-08, + "loss": 0.697, + "step": 49923 + }, + { + "epoch": 1.83, + "learning_rate": 1.7358186545574393e-08, + "loss": 0.6914, + "step": 49924 + }, + { + "epoch": 1.83, + "learning_rate": 1.7354524480902333e-08, + "loss": 0.6986, + "step": 49925 + }, + { + "epoch": 1.83, + "learning_rate": 1.735086241623027e-08, + "loss": 0.7007, + "step": 49926 + }, + { + "epoch": 1.83, + "learning_rate": 1.734720035155821e-08, + "loss": 0.6974, + "step": 49927 + }, + { + "epoch": 1.83, + "learning_rate": 1.7343538286886143e-08, + "loss": 0.6785, + "step": 49928 + }, + { + "epoch": 1.83, + "learning_rate": 1.7339876222214083e-08, + "loss": 0.69, + "step": 49929 + }, + { + "epoch": 1.83, + "learning_rate": 1.733621415754202e-08, + "loss": 0.6915, + "step": 49930 + }, + { + "epoch": 1.83, + "learning_rate": 1.733255209286996e-08, + "loss": 0.6985, + "step": 49931 + }, + { + "epoch": 1.83, + "learning_rate": 1.73288900281979e-08, + "loss": 0.6958, + "step": 49932 + }, + { + "epoch": 1.83, + "learning_rate": 1.7325227963525836e-08, + "loss": 0.6953, + "step": 49933 + }, + { + "epoch": 1.83, + "learning_rate": 1.7321565898853773e-08, + "loss": 0.6967, + "step": 49934 + }, + { + "epoch": 1.83, + "learning_rate": 1.731790383418171e-08, + "loss": 0.7029, + "step": 49935 + }, + { + "epoch": 1.83, + "learning_rate": 1.731424176950965e-08, + "loss": 0.7011, + "step": 49936 + }, + { + "epoch": 1.83, + "learning_rate": 1.7310579704837586e-08, + "loss": 0.6864, + "step": 49937 + }, + { + "epoch": 1.83, + "learning_rate": 1.7306917640165526e-08, + "loss": 0.6961, + "step": 49938 + }, + { + "epoch": 1.83, + "learning_rate": 1.7303255575493462e-08, + "loss": 0.6987, + "step": 49939 + }, + { + "epoch": 1.83, + "learning_rate": 1.72995935108214e-08, + "loss": 0.6732, + "step": 49940 + }, + { + "epoch": 1.83, + "learning_rate": 1.729593144614934e-08, + "loss": 0.6998, + "step": 49941 + }, + { + "epoch": 1.83, + "learning_rate": 1.7292269381477276e-08, + "loss": 0.6862, + "step": 49942 + }, + { + "epoch": 1.83, + "learning_rate": 1.7288607316805215e-08, + "loss": 0.7002, + "step": 49943 + }, + { + "epoch": 1.83, + "learning_rate": 1.7284945252133152e-08, + "loss": 0.6748, + "step": 49944 + }, + { + "epoch": 1.83, + "learning_rate": 1.7281283187461092e-08, + "loss": 0.689, + "step": 49945 + }, + { + "epoch": 1.83, + "learning_rate": 1.7277621122789025e-08, + "loss": 0.6838, + "step": 49946 + }, + { + "epoch": 1.83, + "learning_rate": 1.7273959058116965e-08, + "loss": 0.716, + "step": 49947 + }, + { + "epoch": 1.83, + "learning_rate": 1.7270296993444902e-08, + "loss": 0.6881, + "step": 49948 + }, + { + "epoch": 1.83, + "learning_rate": 1.7266634928772842e-08, + "loss": 0.6937, + "step": 49949 + }, + { + "epoch": 1.83, + "learning_rate": 1.7262972864100782e-08, + "loss": 0.6896, + "step": 49950 + }, + { + "epoch": 1.83, + "learning_rate": 1.7259310799428718e-08, + "loss": 0.692, + "step": 49951 + }, + { + "epoch": 1.83, + "learning_rate": 1.7255648734756655e-08, + "loss": 0.6928, + "step": 49952 + }, + { + "epoch": 1.83, + "learning_rate": 1.725198667008459e-08, + "loss": 0.6913, + "step": 49953 + }, + { + "epoch": 1.83, + "learning_rate": 1.724832460541253e-08, + "loss": 0.6749, + "step": 49954 + }, + { + "epoch": 1.83, + "learning_rate": 1.7244662540740468e-08, + "loss": 0.6968, + "step": 49955 + }, + { + "epoch": 1.83, + "learning_rate": 1.7241000476068408e-08, + "loss": 0.6896, + "step": 49956 + }, + { + "epoch": 1.83, + "learning_rate": 1.7237338411396345e-08, + "loss": 0.6954, + "step": 49957 + }, + { + "epoch": 1.83, + "learning_rate": 1.723367634672428e-08, + "loss": 0.7027, + "step": 49958 + }, + { + "epoch": 1.83, + "learning_rate": 1.723001428205222e-08, + "loss": 0.6971, + "step": 49959 + }, + { + "epoch": 1.83, + "learning_rate": 1.7226352217380158e-08, + "loss": 0.6967, + "step": 49960 + }, + { + "epoch": 1.83, + "learning_rate": 1.7222690152708098e-08, + "loss": 0.6908, + "step": 49961 + }, + { + "epoch": 1.83, + "learning_rate": 1.7219028088036034e-08, + "loss": 0.6961, + "step": 49962 + }, + { + "epoch": 1.83, + "learning_rate": 1.721536602336397e-08, + "loss": 0.6834, + "step": 49963 + }, + { + "epoch": 1.83, + "learning_rate": 1.7211703958691908e-08, + "loss": 0.6854, + "step": 49964 + }, + { + "epoch": 1.83, + "learning_rate": 1.7208041894019848e-08, + "loss": 0.6918, + "step": 49965 + }, + { + "epoch": 1.83, + "learning_rate": 1.7204379829347784e-08, + "loss": 0.6928, + "step": 49966 + }, + { + "epoch": 1.83, + "learning_rate": 1.7200717764675724e-08, + "loss": 0.6934, + "step": 49967 + }, + { + "epoch": 1.83, + "learning_rate": 1.7197055700003664e-08, + "loss": 0.7068, + "step": 49968 + }, + { + "epoch": 1.83, + "learning_rate": 1.7193393635331597e-08, + "loss": 0.7019, + "step": 49969 + }, + { + "epoch": 1.83, + "learning_rate": 1.7189731570659537e-08, + "loss": 0.6924, + "step": 49970 + }, + { + "epoch": 1.83, + "learning_rate": 1.7186069505987474e-08, + "loss": 0.6892, + "step": 49971 + }, + { + "epoch": 1.83, + "learning_rate": 1.7182407441315414e-08, + "loss": 0.6896, + "step": 49972 + }, + { + "epoch": 1.83, + "learning_rate": 1.717874537664335e-08, + "loss": 0.698, + "step": 49973 + }, + { + "epoch": 1.83, + "learning_rate": 1.717508331197129e-08, + "loss": 0.6962, + "step": 49974 + }, + { + "epoch": 1.83, + "learning_rate": 1.7171421247299227e-08, + "loss": 0.6772, + "step": 49975 + }, + { + "epoch": 1.83, + "learning_rate": 1.7167759182627164e-08, + "loss": 0.6849, + "step": 49976 + }, + { + "epoch": 1.83, + "learning_rate": 1.7164097117955104e-08, + "loss": 0.6822, + "step": 49977 + }, + { + "epoch": 1.83, + "learning_rate": 1.716043505328304e-08, + "loss": 0.6939, + "step": 49978 + }, + { + "epoch": 1.83, + "learning_rate": 1.715677298861098e-08, + "loss": 0.6877, + "step": 49979 + }, + { + "epoch": 1.83, + "learning_rate": 1.7153110923938917e-08, + "loss": 0.6892, + "step": 49980 + }, + { + "epoch": 1.83, + "learning_rate": 1.7149448859266853e-08, + "loss": 0.6893, + "step": 49981 + }, + { + "epoch": 1.83, + "learning_rate": 1.714578679459479e-08, + "loss": 0.7056, + "step": 49982 + }, + { + "epoch": 1.83, + "learning_rate": 1.714212472992273e-08, + "loss": 0.6941, + "step": 49983 + }, + { + "epoch": 1.83, + "learning_rate": 1.7138462665250666e-08, + "loss": 0.6816, + "step": 49984 + }, + { + "epoch": 1.83, + "learning_rate": 1.7134800600578606e-08, + "loss": 0.6973, + "step": 49985 + }, + { + "epoch": 1.83, + "learning_rate": 1.7131138535906543e-08, + "loss": 0.7024, + "step": 49986 + }, + { + "epoch": 1.83, + "learning_rate": 1.712747647123448e-08, + "loss": 0.6927, + "step": 49987 + }, + { + "epoch": 1.83, + "learning_rate": 1.712381440656242e-08, + "loss": 0.6914, + "step": 49988 + }, + { + "epoch": 1.83, + "learning_rate": 1.7120152341890356e-08, + "loss": 0.7052, + "step": 49989 + }, + { + "epoch": 1.83, + "learning_rate": 1.7116490277218296e-08, + "loss": 0.7002, + "step": 49990 + }, + { + "epoch": 1.83, + "learning_rate": 1.7112828212546233e-08, + "loss": 0.6963, + "step": 49991 + }, + { + "epoch": 1.83, + "learning_rate": 1.7109166147874173e-08, + "loss": 0.6895, + "step": 49992 + }, + { + "epoch": 1.83, + "learning_rate": 1.7105504083202106e-08, + "loss": 0.695, + "step": 49993 + }, + { + "epoch": 1.83, + "learning_rate": 1.7101842018530046e-08, + "loss": 0.6947, + "step": 49994 + }, + { + "epoch": 1.83, + "learning_rate": 1.7098179953857986e-08, + "loss": 0.6987, + "step": 49995 + }, + { + "epoch": 1.83, + "learning_rate": 1.7094517889185922e-08, + "loss": 0.705, + "step": 49996 + }, + { + "epoch": 1.83, + "learning_rate": 1.7090855824513862e-08, + "loss": 0.6718, + "step": 49997 + }, + { + "epoch": 1.83, + "learning_rate": 1.70871937598418e-08, + "loss": 0.6904, + "step": 49998 + }, + { + "epoch": 1.83, + "learning_rate": 1.7083531695169736e-08, + "loss": 0.6885, + "step": 49999 + }, + { + "epoch": 1.83, + "learning_rate": 1.7079869630497672e-08, + "loss": 0.6828, + "step": 50000 + }, + { + "epoch": 1.83, + "learning_rate": 1.7076207565825612e-08, + "loss": 0.6943, + "step": 50001 + }, + { + "epoch": 1.83, + "learning_rate": 1.707254550115355e-08, + "loss": 0.7055, + "step": 50002 + }, + { + "epoch": 1.83, + "learning_rate": 1.706888343648149e-08, + "loss": 0.6813, + "step": 50003 + }, + { + "epoch": 1.83, + "learning_rate": 1.7065221371809425e-08, + "loss": 0.7126, + "step": 50004 + }, + { + "epoch": 1.83, + "learning_rate": 1.7061559307137362e-08, + "loss": 0.6957, + "step": 50005 + }, + { + "epoch": 1.83, + "learning_rate": 1.7057897242465302e-08, + "loss": 0.6986, + "step": 50006 + }, + { + "epoch": 1.83, + "learning_rate": 1.705423517779324e-08, + "loss": 0.6918, + "step": 50007 + }, + { + "epoch": 1.83, + "learning_rate": 1.705057311312118e-08, + "loss": 0.6978, + "step": 50008 + }, + { + "epoch": 1.83, + "learning_rate": 1.7046911048449115e-08, + "loss": 0.6887, + "step": 50009 + }, + { + "epoch": 1.83, + "learning_rate": 1.7043248983777052e-08, + "loss": 0.6869, + "step": 50010 + }, + { + "epoch": 1.83, + "learning_rate": 1.7039586919104988e-08, + "loss": 0.6908, + "step": 50011 + }, + { + "epoch": 1.83, + "learning_rate": 1.7035924854432928e-08, + "loss": 0.6825, + "step": 50012 + }, + { + "epoch": 1.83, + "learning_rate": 1.7032262789760865e-08, + "loss": 0.7035, + "step": 50013 + }, + { + "epoch": 1.83, + "learning_rate": 1.7028600725088805e-08, + "loss": 0.6948, + "step": 50014 + }, + { + "epoch": 1.83, + "learning_rate": 1.7024938660416745e-08, + "loss": 0.693, + "step": 50015 + }, + { + "epoch": 1.83, + "learning_rate": 1.7021276595744678e-08, + "loss": 0.6912, + "step": 50016 + }, + { + "epoch": 1.83, + "learning_rate": 1.7017614531072618e-08, + "loss": 0.6986, + "step": 50017 + }, + { + "epoch": 1.83, + "learning_rate": 1.7013952466400555e-08, + "loss": 0.693, + "step": 50018 + }, + { + "epoch": 1.83, + "learning_rate": 1.7010290401728495e-08, + "loss": 0.6888, + "step": 50019 + }, + { + "epoch": 1.83, + "learning_rate": 1.700662833705643e-08, + "loss": 0.6908, + "step": 50020 + }, + { + "epoch": 1.83, + "learning_rate": 1.700296627238437e-08, + "loss": 0.6997, + "step": 50021 + }, + { + "epoch": 1.83, + "learning_rate": 1.6999304207712308e-08, + "loss": 0.6922, + "step": 50022 + }, + { + "epoch": 1.83, + "learning_rate": 1.6995642143040244e-08, + "loss": 0.6979, + "step": 50023 + }, + { + "epoch": 1.83, + "learning_rate": 1.6991980078368184e-08, + "loss": 0.6889, + "step": 50024 + }, + { + "epoch": 1.83, + "learning_rate": 1.698831801369612e-08, + "loss": 0.6939, + "step": 50025 + }, + { + "epoch": 1.83, + "learning_rate": 1.698465594902406e-08, + "loss": 0.6943, + "step": 50026 + }, + { + "epoch": 1.83, + "learning_rate": 1.6980993884351997e-08, + "loss": 0.7112, + "step": 50027 + }, + { + "epoch": 1.83, + "learning_rate": 1.6977331819679934e-08, + "loss": 0.6902, + "step": 50028 + }, + { + "epoch": 1.83, + "learning_rate": 1.697366975500787e-08, + "loss": 0.6926, + "step": 50029 + }, + { + "epoch": 1.83, + "learning_rate": 1.697000769033581e-08, + "loss": 0.6953, + "step": 50030 + }, + { + "epoch": 1.83, + "learning_rate": 1.6966345625663747e-08, + "loss": 0.6832, + "step": 50031 + }, + { + "epoch": 1.83, + "learning_rate": 1.6962683560991687e-08, + "loss": 0.6905, + "step": 50032 + }, + { + "epoch": 1.83, + "learning_rate": 1.6959021496319627e-08, + "loss": 0.6926, + "step": 50033 + }, + { + "epoch": 1.83, + "learning_rate": 1.695535943164756e-08, + "loss": 0.6911, + "step": 50034 + }, + { + "epoch": 1.83, + "learning_rate": 1.69516973669755e-08, + "loss": 0.6725, + "step": 50035 + }, + { + "epoch": 1.83, + "learning_rate": 1.6948035302303437e-08, + "loss": 0.6853, + "step": 50036 + }, + { + "epoch": 1.83, + "learning_rate": 1.6944373237631377e-08, + "loss": 0.6789, + "step": 50037 + }, + { + "epoch": 1.83, + "learning_rate": 1.6940711172959313e-08, + "loss": 0.6845, + "step": 50038 + }, + { + "epoch": 1.83, + "learning_rate": 1.6937049108287253e-08, + "loss": 0.6839, + "step": 50039 + }, + { + "epoch": 1.83, + "learning_rate": 1.6933387043615187e-08, + "loss": 0.6852, + "step": 50040 + }, + { + "epoch": 1.83, + "learning_rate": 1.6929724978943127e-08, + "loss": 0.6894, + "step": 50041 + }, + { + "epoch": 1.83, + "learning_rate": 1.6926062914271067e-08, + "loss": 0.6881, + "step": 50042 + }, + { + "epoch": 1.83, + "learning_rate": 1.6922400849599003e-08, + "loss": 0.694, + "step": 50043 + }, + { + "epoch": 1.83, + "learning_rate": 1.6918738784926943e-08, + "loss": 0.7003, + "step": 50044 + }, + { + "epoch": 1.83, + "learning_rate": 1.691507672025488e-08, + "loss": 0.6888, + "step": 50045 + }, + { + "epoch": 1.83, + "learning_rate": 1.6911414655582816e-08, + "loss": 0.6809, + "step": 50046 + }, + { + "epoch": 1.83, + "learning_rate": 1.6907752590910753e-08, + "loss": 0.6991, + "step": 50047 + }, + { + "epoch": 1.83, + "learning_rate": 1.6904090526238693e-08, + "loss": 0.699, + "step": 50048 + }, + { + "epoch": 1.83, + "learning_rate": 1.690042846156663e-08, + "loss": 0.7037, + "step": 50049 + }, + { + "epoch": 1.83, + "learning_rate": 1.689676639689457e-08, + "loss": 0.687, + "step": 50050 + }, + { + "epoch": 1.83, + "learning_rate": 1.6893104332222506e-08, + "loss": 0.7083, + "step": 50051 + }, + { + "epoch": 1.83, + "learning_rate": 1.6889442267550443e-08, + "loss": 0.6978, + "step": 50052 + }, + { + "epoch": 1.83, + "learning_rate": 1.6885780202878383e-08, + "loss": 0.6942, + "step": 50053 + }, + { + "epoch": 1.83, + "learning_rate": 1.688211813820632e-08, + "loss": 0.6991, + "step": 50054 + }, + { + "epoch": 1.83, + "learning_rate": 1.687845607353426e-08, + "loss": 0.6937, + "step": 50055 + }, + { + "epoch": 1.83, + "learning_rate": 1.6874794008862196e-08, + "loss": 0.6783, + "step": 50056 + }, + { + "epoch": 1.83, + "learning_rate": 1.6871131944190132e-08, + "loss": 0.6942, + "step": 50057 + }, + { + "epoch": 1.83, + "learning_rate": 1.686746987951807e-08, + "loss": 0.6891, + "step": 50058 + }, + { + "epoch": 1.83, + "learning_rate": 1.686380781484601e-08, + "loss": 0.7011, + "step": 50059 + }, + { + "epoch": 1.83, + "learning_rate": 1.686014575017395e-08, + "loss": 0.6745, + "step": 50060 + }, + { + "epoch": 1.83, + "learning_rate": 1.6856483685501885e-08, + "loss": 0.697, + "step": 50061 + }, + { + "epoch": 1.83, + "learning_rate": 1.6852821620829825e-08, + "loss": 0.6991, + "step": 50062 + }, + { + "epoch": 1.83, + "learning_rate": 1.6849159556157762e-08, + "loss": 0.6843, + "step": 50063 + }, + { + "epoch": 1.83, + "learning_rate": 1.68454974914857e-08, + "loss": 0.6985, + "step": 50064 + }, + { + "epoch": 1.83, + "learning_rate": 1.6841835426813635e-08, + "loss": 0.6958, + "step": 50065 + }, + { + "epoch": 1.83, + "learning_rate": 1.6838173362141575e-08, + "loss": 0.6906, + "step": 50066 + }, + { + "epoch": 1.83, + "learning_rate": 1.6834511297469512e-08, + "loss": 0.7109, + "step": 50067 + }, + { + "epoch": 1.83, + "learning_rate": 1.6830849232797452e-08, + "loss": 0.6949, + "step": 50068 + }, + { + "epoch": 1.83, + "learning_rate": 1.682718716812539e-08, + "loss": 0.6898, + "step": 50069 + }, + { + "epoch": 1.83, + "learning_rate": 1.6823525103453325e-08, + "loss": 0.6954, + "step": 50070 + }, + { + "epoch": 1.83, + "learning_rate": 1.6819863038781265e-08, + "loss": 0.7001, + "step": 50071 + }, + { + "epoch": 1.83, + "learning_rate": 1.68162009741092e-08, + "loss": 0.6934, + "step": 50072 + }, + { + "epoch": 1.83, + "learning_rate": 1.681253890943714e-08, + "loss": 0.6954, + "step": 50073 + }, + { + "epoch": 1.83, + "learning_rate": 1.6808876844765078e-08, + "loss": 0.6869, + "step": 50074 + }, + { + "epoch": 1.83, + "learning_rate": 1.6805214780093015e-08, + "loss": 0.697, + "step": 50075 + }, + { + "epoch": 1.83, + "learning_rate": 1.680155271542095e-08, + "loss": 0.6947, + "step": 50076 + }, + { + "epoch": 1.83, + "learning_rate": 1.679789065074889e-08, + "loss": 0.6926, + "step": 50077 + }, + { + "epoch": 1.83, + "learning_rate": 1.679422858607683e-08, + "loss": 0.7033, + "step": 50078 + }, + { + "epoch": 1.83, + "learning_rate": 1.6790566521404768e-08, + "loss": 0.6917, + "step": 50079 + }, + { + "epoch": 1.83, + "learning_rate": 1.6786904456732708e-08, + "loss": 0.6962, + "step": 50080 + }, + { + "epoch": 1.83, + "learning_rate": 1.678324239206064e-08, + "loss": 0.6875, + "step": 50081 + }, + { + "epoch": 1.83, + "learning_rate": 1.677958032738858e-08, + "loss": 0.6908, + "step": 50082 + }, + { + "epoch": 1.83, + "learning_rate": 1.6775918262716518e-08, + "loss": 0.6932, + "step": 50083 + }, + { + "epoch": 1.83, + "learning_rate": 1.6772256198044458e-08, + "loss": 0.6897, + "step": 50084 + }, + { + "epoch": 1.83, + "learning_rate": 1.6768594133372394e-08, + "loss": 0.6979, + "step": 50085 + }, + { + "epoch": 1.83, + "learning_rate": 1.6764932068700334e-08, + "loss": 0.6883, + "step": 50086 + }, + { + "epoch": 1.83, + "learning_rate": 1.676127000402827e-08, + "loss": 0.6925, + "step": 50087 + }, + { + "epoch": 1.83, + "learning_rate": 1.6757607939356207e-08, + "loss": 0.6941, + "step": 50088 + }, + { + "epoch": 1.83, + "learning_rate": 1.6753945874684147e-08, + "loss": 0.6933, + "step": 50089 + }, + { + "epoch": 1.83, + "learning_rate": 1.6750283810012084e-08, + "loss": 0.6976, + "step": 50090 + }, + { + "epoch": 1.83, + "learning_rate": 1.6746621745340024e-08, + "loss": 0.6994, + "step": 50091 + }, + { + "epoch": 1.83, + "learning_rate": 1.674295968066796e-08, + "loss": 0.6899, + "step": 50092 + }, + { + "epoch": 1.83, + "learning_rate": 1.6739297615995897e-08, + "loss": 0.6891, + "step": 50093 + }, + { + "epoch": 1.83, + "learning_rate": 1.6735635551323834e-08, + "loss": 0.6913, + "step": 50094 + }, + { + "epoch": 1.83, + "learning_rate": 1.6731973486651774e-08, + "loss": 0.6841, + "step": 50095 + }, + { + "epoch": 1.83, + "learning_rate": 1.6728311421979714e-08, + "loss": 0.6964, + "step": 50096 + }, + { + "epoch": 1.83, + "learning_rate": 1.672464935730765e-08, + "loss": 0.6947, + "step": 50097 + }, + { + "epoch": 1.83, + "learning_rate": 1.6720987292635587e-08, + "loss": 0.6947, + "step": 50098 + }, + { + "epoch": 1.83, + "learning_rate": 1.6717325227963523e-08, + "loss": 0.6929, + "step": 50099 + }, + { + "epoch": 1.83, + "learning_rate": 1.6713663163291463e-08, + "loss": 0.7029, + "step": 50100 + }, + { + "epoch": 1.83, + "learning_rate": 1.67100010986194e-08, + "loss": 0.6999, + "step": 50101 + }, + { + "epoch": 1.83, + "learning_rate": 1.670633903394734e-08, + "loss": 0.6826, + "step": 50102 + }, + { + "epoch": 1.83, + "learning_rate": 1.6702676969275276e-08, + "loss": 0.6981, + "step": 50103 + }, + { + "epoch": 1.83, + "learning_rate": 1.6699014904603213e-08, + "loss": 0.6981, + "step": 50104 + }, + { + "epoch": 1.83, + "learning_rate": 1.669535283993115e-08, + "loss": 0.6992, + "step": 50105 + }, + { + "epoch": 1.83, + "learning_rate": 1.669169077525909e-08, + "loss": 0.7078, + "step": 50106 + }, + { + "epoch": 1.83, + "learning_rate": 1.668802871058703e-08, + "loss": 0.6904, + "step": 50107 + }, + { + "epoch": 1.83, + "learning_rate": 1.6684366645914966e-08, + "loss": 0.697, + "step": 50108 + }, + { + "epoch": 1.83, + "learning_rate": 1.6680704581242906e-08, + "loss": 0.6851, + "step": 50109 + }, + { + "epoch": 1.83, + "learning_rate": 1.6677042516570843e-08, + "loss": 0.6821, + "step": 50110 + }, + { + "epoch": 1.83, + "learning_rate": 1.667338045189878e-08, + "loss": 0.7014, + "step": 50111 + }, + { + "epoch": 1.83, + "learning_rate": 1.6669718387226716e-08, + "loss": 0.694, + "step": 50112 + }, + { + "epoch": 1.83, + "learning_rate": 1.6666056322554656e-08, + "loss": 0.7002, + "step": 50113 + }, + { + "epoch": 1.83, + "learning_rate": 1.6662394257882593e-08, + "loss": 0.6829, + "step": 50114 + }, + { + "epoch": 1.83, + "learning_rate": 1.6658732193210532e-08, + "loss": 0.6938, + "step": 50115 + }, + { + "epoch": 1.83, + "learning_rate": 1.665507012853847e-08, + "loss": 0.6934, + "step": 50116 + }, + { + "epoch": 1.83, + "learning_rate": 1.6651408063866406e-08, + "loss": 0.6998, + "step": 50117 + }, + { + "epoch": 1.83, + "learning_rate": 1.6647745999194346e-08, + "loss": 0.6955, + "step": 50118 + }, + { + "epoch": 1.83, + "learning_rate": 1.6644083934522282e-08, + "loss": 0.6964, + "step": 50119 + }, + { + "epoch": 1.83, + "learning_rate": 1.6640421869850222e-08, + "loss": 0.6999, + "step": 50120 + }, + { + "epoch": 1.83, + "learning_rate": 1.663675980517816e-08, + "loss": 0.6991, + "step": 50121 + }, + { + "epoch": 1.83, + "learning_rate": 1.6633097740506095e-08, + "loss": 0.693, + "step": 50122 + }, + { + "epoch": 1.83, + "learning_rate": 1.6629435675834032e-08, + "loss": 0.7029, + "step": 50123 + }, + { + "epoch": 1.83, + "learning_rate": 1.6625773611161972e-08, + "loss": 0.6944, + "step": 50124 + }, + { + "epoch": 1.83, + "learning_rate": 1.6622111546489912e-08, + "loss": 0.6906, + "step": 50125 + }, + { + "epoch": 1.83, + "learning_rate": 1.661844948181785e-08, + "loss": 0.6876, + "step": 50126 + }, + { + "epoch": 1.83, + "learning_rate": 1.661478741714579e-08, + "loss": 0.708, + "step": 50127 + }, + { + "epoch": 1.83, + "learning_rate": 1.6611125352473722e-08, + "loss": 0.7026, + "step": 50128 + }, + { + "epoch": 1.83, + "learning_rate": 1.6607463287801662e-08, + "loss": 0.6872, + "step": 50129 + }, + { + "epoch": 1.83, + "learning_rate": 1.6603801223129598e-08, + "loss": 0.6995, + "step": 50130 + }, + { + "epoch": 1.83, + "learning_rate": 1.6600139158457538e-08, + "loss": 0.6816, + "step": 50131 + }, + { + "epoch": 1.83, + "learning_rate": 1.6596477093785475e-08, + "loss": 0.6886, + "step": 50132 + }, + { + "epoch": 1.83, + "learning_rate": 1.6592815029113415e-08, + "loss": 0.6922, + "step": 50133 + }, + { + "epoch": 1.83, + "learning_rate": 1.658915296444135e-08, + "loss": 0.6959, + "step": 50134 + }, + { + "epoch": 1.83, + "learning_rate": 1.6585490899769288e-08, + "loss": 0.6987, + "step": 50135 + }, + { + "epoch": 1.83, + "learning_rate": 1.6581828835097228e-08, + "loss": 0.7019, + "step": 50136 + }, + { + "epoch": 1.83, + "learning_rate": 1.6578166770425165e-08, + "loss": 0.6894, + "step": 50137 + }, + { + "epoch": 1.83, + "learning_rate": 1.6574504705753104e-08, + "loss": 0.6948, + "step": 50138 + }, + { + "epoch": 1.83, + "learning_rate": 1.657084264108104e-08, + "loss": 0.6956, + "step": 50139 + }, + { + "epoch": 1.83, + "learning_rate": 1.6567180576408978e-08, + "loss": 0.6907, + "step": 50140 + }, + { + "epoch": 1.83, + "learning_rate": 1.6563518511736914e-08, + "loss": 0.6926, + "step": 50141 + }, + { + "epoch": 1.83, + "learning_rate": 1.6559856447064854e-08, + "loss": 0.6922, + "step": 50142 + }, + { + "epoch": 1.83, + "learning_rate": 1.6556194382392794e-08, + "loss": 0.7119, + "step": 50143 + }, + { + "epoch": 1.83, + "learning_rate": 1.655253231772073e-08, + "loss": 0.7004, + "step": 50144 + }, + { + "epoch": 1.83, + "learning_rate": 1.6548870253048667e-08, + "loss": 0.7115, + "step": 50145 + }, + { + "epoch": 1.83, + "learning_rate": 1.6545208188376604e-08, + "loss": 0.6852, + "step": 50146 + }, + { + "epoch": 1.83, + "learning_rate": 1.6541546123704544e-08, + "loss": 0.6963, + "step": 50147 + }, + { + "epoch": 1.83, + "learning_rate": 1.653788405903248e-08, + "loss": 0.6892, + "step": 50148 + }, + { + "epoch": 1.83, + "learning_rate": 1.653422199436042e-08, + "loss": 0.6962, + "step": 50149 + }, + { + "epoch": 1.83, + "learning_rate": 1.6530559929688357e-08, + "loss": 0.6997, + "step": 50150 + }, + { + "epoch": 1.83, + "learning_rate": 1.6526897865016297e-08, + "loss": 0.6964, + "step": 50151 + }, + { + "epoch": 1.83, + "learning_rate": 1.6523235800344234e-08, + "loss": 0.6963, + "step": 50152 + }, + { + "epoch": 1.83, + "learning_rate": 1.651957373567217e-08, + "loss": 0.6962, + "step": 50153 + }, + { + "epoch": 1.83, + "learning_rate": 1.651591167100011e-08, + "loss": 0.6972, + "step": 50154 + }, + { + "epoch": 1.84, + "learning_rate": 1.6512249606328047e-08, + "loss": 0.6978, + "step": 50155 + }, + { + "epoch": 1.84, + "learning_rate": 1.6508587541655987e-08, + "loss": 0.6908, + "step": 50156 + }, + { + "epoch": 1.84, + "learning_rate": 1.6504925476983923e-08, + "loss": 0.6918, + "step": 50157 + }, + { + "epoch": 1.84, + "learning_rate": 1.650126341231186e-08, + "loss": 0.7032, + "step": 50158 + }, + { + "epoch": 1.84, + "learning_rate": 1.6497601347639797e-08, + "loss": 0.6917, + "step": 50159 + }, + { + "epoch": 1.84, + "learning_rate": 1.6493939282967737e-08, + "loss": 0.6959, + "step": 50160 + }, + { + "epoch": 1.84, + "learning_rate": 1.6490277218295677e-08, + "loss": 0.6941, + "step": 50161 + }, + { + "epoch": 1.84, + "learning_rate": 1.6486615153623613e-08, + "loss": 0.6808, + "step": 50162 + }, + { + "epoch": 1.84, + "learning_rate": 1.648295308895155e-08, + "loss": 0.6963, + "step": 50163 + }, + { + "epoch": 1.84, + "learning_rate": 1.6479291024279486e-08, + "loss": 0.7031, + "step": 50164 + }, + { + "epoch": 1.84, + "learning_rate": 1.6475628959607426e-08, + "loss": 0.6894, + "step": 50165 + }, + { + "epoch": 1.84, + "learning_rate": 1.6471966894935363e-08, + "loss": 0.6937, + "step": 50166 + }, + { + "epoch": 1.84, + "learning_rate": 1.6468304830263303e-08, + "loss": 0.6928, + "step": 50167 + }, + { + "epoch": 1.84, + "learning_rate": 1.646464276559124e-08, + "loss": 0.6845, + "step": 50168 + }, + { + "epoch": 1.84, + "learning_rate": 1.6460980700919176e-08, + "loss": 0.686, + "step": 50169 + }, + { + "epoch": 1.84, + "learning_rate": 1.6457318636247116e-08, + "loss": 0.6964, + "step": 50170 + }, + { + "epoch": 1.84, + "learning_rate": 1.6453656571575053e-08, + "loss": 0.6884, + "step": 50171 + }, + { + "epoch": 1.84, + "learning_rate": 1.6449994506902993e-08, + "loss": 0.6893, + "step": 50172 + }, + { + "epoch": 1.84, + "learning_rate": 1.644633244223093e-08, + "loss": 0.6959, + "step": 50173 + }, + { + "epoch": 1.84, + "learning_rate": 1.644267037755887e-08, + "loss": 0.6916, + "step": 50174 + }, + { + "epoch": 1.84, + "learning_rate": 1.6439008312886802e-08, + "loss": 0.6994, + "step": 50175 + }, + { + "epoch": 1.84, + "learning_rate": 1.6435346248214742e-08, + "loss": 0.6945, + "step": 50176 + }, + { + "epoch": 1.84, + "learning_rate": 1.643168418354268e-08, + "loss": 0.704, + "step": 50177 + }, + { + "epoch": 1.84, + "learning_rate": 1.642802211887062e-08, + "loss": 0.6834, + "step": 50178 + }, + { + "epoch": 1.84, + "learning_rate": 1.642436005419856e-08, + "loss": 0.6952, + "step": 50179 + }, + { + "epoch": 1.84, + "learning_rate": 1.6420697989526495e-08, + "loss": 0.7016, + "step": 50180 + }, + { + "epoch": 1.84, + "learning_rate": 1.6417035924854432e-08, + "loss": 0.6839, + "step": 50181 + }, + { + "epoch": 1.84, + "learning_rate": 1.641337386018237e-08, + "loss": 0.6894, + "step": 50182 + }, + { + "epoch": 1.84, + "learning_rate": 1.640971179551031e-08, + "loss": 0.6901, + "step": 50183 + }, + { + "epoch": 1.84, + "learning_rate": 1.6406049730838245e-08, + "loss": 0.6857, + "step": 50184 + }, + { + "epoch": 1.84, + "learning_rate": 1.6402387666166185e-08, + "loss": 0.6836, + "step": 50185 + }, + { + "epoch": 1.84, + "learning_rate": 1.6398725601494122e-08, + "loss": 0.687, + "step": 50186 + }, + { + "epoch": 1.84, + "learning_rate": 1.639506353682206e-08, + "loss": 0.6966, + "step": 50187 + }, + { + "epoch": 1.84, + "learning_rate": 1.639140147215e-08, + "loss": 0.7077, + "step": 50188 + }, + { + "epoch": 1.84, + "learning_rate": 1.6387739407477935e-08, + "loss": 0.6933, + "step": 50189 + }, + { + "epoch": 1.84, + "learning_rate": 1.6384077342805875e-08, + "loss": 0.6821, + "step": 50190 + }, + { + "epoch": 1.84, + "learning_rate": 1.638041527813381e-08, + "loss": 0.6936, + "step": 50191 + }, + { + "epoch": 1.84, + "learning_rate": 1.6376753213461748e-08, + "loss": 0.6894, + "step": 50192 + }, + { + "epoch": 1.84, + "learning_rate": 1.6373091148789685e-08, + "loss": 0.6906, + "step": 50193 + }, + { + "epoch": 1.84, + "learning_rate": 1.6369429084117625e-08, + "loss": 0.6871, + "step": 50194 + }, + { + "epoch": 1.84, + "learning_rate": 1.636576701944556e-08, + "loss": 0.7017, + "step": 50195 + }, + { + "epoch": 1.84, + "learning_rate": 1.63621049547735e-08, + "loss": 0.7006, + "step": 50196 + }, + { + "epoch": 1.84, + "learning_rate": 1.6358442890101438e-08, + "loss": 0.7011, + "step": 50197 + }, + { + "epoch": 1.84, + "learning_rate": 1.6354780825429378e-08, + "loss": 0.6972, + "step": 50198 + }, + { + "epoch": 1.84, + "learning_rate": 1.6351118760757314e-08, + "loss": 0.6917, + "step": 50199 + }, + { + "epoch": 1.84, + "learning_rate": 1.634745669608525e-08, + "loss": 0.695, + "step": 50200 + }, + { + "epoch": 1.84, + "learning_rate": 1.634379463141319e-08, + "loss": 0.6774, + "step": 50201 + }, + { + "epoch": 1.84, + "learning_rate": 1.6340132566741128e-08, + "loss": 0.7003, + "step": 50202 + }, + { + "epoch": 1.84, + "learning_rate": 1.6336470502069067e-08, + "loss": 0.6838, + "step": 50203 + }, + { + "epoch": 1.84, + "learning_rate": 1.6332808437397004e-08, + "loss": 0.6947, + "step": 50204 + }, + { + "epoch": 1.84, + "learning_rate": 1.632914637272494e-08, + "loss": 0.7019, + "step": 50205 + }, + { + "epoch": 1.84, + "learning_rate": 1.6325484308052877e-08, + "loss": 0.6844, + "step": 50206 + }, + { + "epoch": 1.84, + "learning_rate": 1.6321822243380817e-08, + "loss": 0.6935, + "step": 50207 + }, + { + "epoch": 1.84, + "learning_rate": 1.6318160178708757e-08, + "loss": 0.6945, + "step": 50208 + }, + { + "epoch": 1.84, + "learning_rate": 1.6314498114036694e-08, + "loss": 0.6929, + "step": 50209 + }, + { + "epoch": 1.84, + "learning_rate": 1.631083604936463e-08, + "loss": 0.6893, + "step": 50210 + }, + { + "epoch": 1.84, + "learning_rate": 1.6307173984692567e-08, + "loss": 0.6921, + "step": 50211 + }, + { + "epoch": 1.84, + "learning_rate": 1.6303511920020507e-08, + "loss": 0.69, + "step": 50212 + }, + { + "epoch": 1.84, + "learning_rate": 1.6299849855348444e-08, + "loss": 0.6829, + "step": 50213 + }, + { + "epoch": 1.84, + "learning_rate": 1.6296187790676384e-08, + "loss": 0.6932, + "step": 50214 + }, + { + "epoch": 1.84, + "learning_rate": 1.629252572600432e-08, + "loss": 0.6904, + "step": 50215 + }, + { + "epoch": 1.84, + "learning_rate": 1.6288863661332257e-08, + "loss": 0.6849, + "step": 50216 + }, + { + "epoch": 1.84, + "learning_rate": 1.6285201596660197e-08, + "loss": 0.6796, + "step": 50217 + }, + { + "epoch": 1.84, + "learning_rate": 1.6281539531988133e-08, + "loss": 0.6916, + "step": 50218 + }, + { + "epoch": 1.84, + "learning_rate": 1.6277877467316073e-08, + "loss": 0.6913, + "step": 50219 + }, + { + "epoch": 1.84, + "learning_rate": 1.627421540264401e-08, + "loss": 0.6872, + "step": 50220 + }, + { + "epoch": 1.84, + "learning_rate": 1.627055333797195e-08, + "loss": 0.6976, + "step": 50221 + }, + { + "epoch": 1.84, + "learning_rate": 1.6266891273299883e-08, + "loss": 0.6769, + "step": 50222 + }, + { + "epoch": 1.84, + "learning_rate": 1.6263229208627823e-08, + "loss": 0.6841, + "step": 50223 + }, + { + "epoch": 1.84, + "learning_rate": 1.625956714395576e-08, + "loss": 0.6937, + "step": 50224 + }, + { + "epoch": 1.84, + "learning_rate": 1.62559050792837e-08, + "loss": 0.6939, + "step": 50225 + }, + { + "epoch": 1.84, + "learning_rate": 1.625224301461164e-08, + "loss": 0.7081, + "step": 50226 + }, + { + "epoch": 1.84, + "learning_rate": 1.6248580949939576e-08, + "loss": 0.7047, + "step": 50227 + }, + { + "epoch": 1.84, + "learning_rate": 1.6244918885267513e-08, + "loss": 0.7055, + "step": 50228 + }, + { + "epoch": 1.84, + "learning_rate": 1.624125682059545e-08, + "loss": 0.7006, + "step": 50229 + }, + { + "epoch": 1.84, + "learning_rate": 1.623759475592339e-08, + "loss": 0.6875, + "step": 50230 + }, + { + "epoch": 1.84, + "learning_rate": 1.6233932691251326e-08, + "loss": 0.6868, + "step": 50231 + }, + { + "epoch": 1.84, + "learning_rate": 1.6230270626579266e-08, + "loss": 0.7013, + "step": 50232 + }, + { + "epoch": 1.84, + "learning_rate": 1.6226608561907202e-08, + "loss": 0.6956, + "step": 50233 + }, + { + "epoch": 1.84, + "learning_rate": 1.622294649723514e-08, + "loss": 0.6996, + "step": 50234 + }, + { + "epoch": 1.84, + "learning_rate": 1.621928443256308e-08, + "loss": 0.6942, + "step": 50235 + }, + { + "epoch": 1.84, + "learning_rate": 1.6215622367891016e-08, + "loss": 0.6985, + "step": 50236 + }, + { + "epoch": 1.84, + "learning_rate": 1.6211960303218956e-08, + "loss": 0.7007, + "step": 50237 + }, + { + "epoch": 1.84, + "learning_rate": 1.6208298238546892e-08, + "loss": 0.6906, + "step": 50238 + }, + { + "epoch": 1.84, + "learning_rate": 1.6204636173874832e-08, + "loss": 0.6983, + "step": 50239 + }, + { + "epoch": 1.84, + "learning_rate": 1.6200974109202765e-08, + "loss": 0.6998, + "step": 50240 + }, + { + "epoch": 1.84, + "learning_rate": 1.6197312044530705e-08, + "loss": 0.6882, + "step": 50241 + }, + { + "epoch": 1.84, + "learning_rate": 1.6193649979858642e-08, + "loss": 0.6851, + "step": 50242 + }, + { + "epoch": 1.84, + "learning_rate": 1.6189987915186582e-08, + "loss": 0.6941, + "step": 50243 + }, + { + "epoch": 1.84, + "learning_rate": 1.6186325850514522e-08, + "loss": 0.6999, + "step": 50244 + }, + { + "epoch": 1.84, + "learning_rate": 1.618266378584246e-08, + "loss": 0.6934, + "step": 50245 + }, + { + "epoch": 1.84, + "learning_rate": 1.6179001721170395e-08, + "loss": 0.6928, + "step": 50246 + }, + { + "epoch": 1.84, + "learning_rate": 1.6175339656498332e-08, + "loss": 0.6831, + "step": 50247 + }, + { + "epoch": 1.84, + "learning_rate": 1.617167759182627e-08, + "loss": 0.7028, + "step": 50248 + }, + { + "epoch": 1.84, + "learning_rate": 1.6168015527154208e-08, + "loss": 0.6881, + "step": 50249 + }, + { + "epoch": 1.84, + "learning_rate": 1.6164353462482148e-08, + "loss": 0.6896, + "step": 50250 + }, + { + "epoch": 1.84, + "learning_rate": 1.6160691397810085e-08, + "loss": 0.6826, + "step": 50251 + }, + { + "epoch": 1.84, + "learning_rate": 1.615702933313802e-08, + "loss": 0.6899, + "step": 50252 + }, + { + "epoch": 1.84, + "learning_rate": 1.615336726846596e-08, + "loss": 0.6973, + "step": 50253 + }, + { + "epoch": 1.84, + "learning_rate": 1.6149705203793898e-08, + "loss": 0.6901, + "step": 50254 + }, + { + "epoch": 1.84, + "learning_rate": 1.6146043139121838e-08, + "loss": 0.6854, + "step": 50255 + }, + { + "epoch": 1.84, + "learning_rate": 1.6142381074449775e-08, + "loss": 0.6964, + "step": 50256 + }, + { + "epoch": 1.84, + "learning_rate": 1.613871900977771e-08, + "loss": 0.7013, + "step": 50257 + }, + { + "epoch": 1.84, + "learning_rate": 1.6135056945105648e-08, + "loss": 0.6806, + "step": 50258 + }, + { + "epoch": 1.84, + "learning_rate": 1.6131394880433588e-08, + "loss": 0.6874, + "step": 50259 + }, + { + "epoch": 1.84, + "learning_rate": 1.6127732815761524e-08, + "loss": 0.6986, + "step": 50260 + }, + { + "epoch": 1.84, + "learning_rate": 1.6124070751089464e-08, + "loss": 0.6944, + "step": 50261 + }, + { + "epoch": 1.84, + "learning_rate": 1.6120408686417404e-08, + "loss": 0.7032, + "step": 50262 + }, + { + "epoch": 1.84, + "learning_rate": 1.6116746621745337e-08, + "loss": 0.711, + "step": 50263 + }, + { + "epoch": 1.84, + "learning_rate": 1.6113084557073277e-08, + "loss": 0.6978, + "step": 50264 + }, + { + "epoch": 1.84, + "learning_rate": 1.6109422492401214e-08, + "loss": 0.6914, + "step": 50265 + }, + { + "epoch": 1.84, + "learning_rate": 1.6105760427729154e-08, + "loss": 0.6985, + "step": 50266 + }, + { + "epoch": 1.84, + "learning_rate": 1.610209836305709e-08, + "loss": 0.6901, + "step": 50267 + }, + { + "epoch": 1.84, + "learning_rate": 1.609843629838503e-08, + "loss": 0.6903, + "step": 50268 + }, + { + "epoch": 1.84, + "learning_rate": 1.6094774233712967e-08, + "loss": 0.6869, + "step": 50269 + }, + { + "epoch": 1.84, + "learning_rate": 1.6091112169040904e-08, + "loss": 0.6899, + "step": 50270 + }, + { + "epoch": 1.84, + "learning_rate": 1.6087450104368844e-08, + "loss": 0.6907, + "step": 50271 + }, + { + "epoch": 1.84, + "learning_rate": 1.608378803969678e-08, + "loss": 0.6951, + "step": 50272 + }, + { + "epoch": 1.84, + "learning_rate": 1.608012597502472e-08, + "loss": 0.6936, + "step": 50273 + }, + { + "epoch": 1.84, + "learning_rate": 1.6076463910352657e-08, + "loss": 0.6973, + "step": 50274 + }, + { + "epoch": 1.84, + "learning_rate": 1.6072801845680593e-08, + "loss": 0.6944, + "step": 50275 + }, + { + "epoch": 1.84, + "learning_rate": 1.606913978100853e-08, + "loss": 0.6919, + "step": 50276 + }, + { + "epoch": 1.84, + "learning_rate": 1.606547771633647e-08, + "loss": 0.6955, + "step": 50277 + }, + { + "epoch": 1.84, + "learning_rate": 1.6061815651664407e-08, + "loss": 0.6874, + "step": 50278 + }, + { + "epoch": 1.84, + "learning_rate": 1.6058153586992347e-08, + "loss": 0.6999, + "step": 50279 + }, + { + "epoch": 1.84, + "learning_rate": 1.6054491522320283e-08, + "loss": 0.6863, + "step": 50280 + }, + { + "epoch": 1.84, + "learning_rate": 1.605082945764822e-08, + "loss": 0.6804, + "step": 50281 + }, + { + "epoch": 1.84, + "learning_rate": 1.604716739297616e-08, + "loss": 0.7021, + "step": 50282 + }, + { + "epoch": 1.84, + "learning_rate": 1.6043505328304096e-08, + "loss": 0.6913, + "step": 50283 + }, + { + "epoch": 1.84, + "learning_rate": 1.6039843263632036e-08, + "loss": 0.6941, + "step": 50284 + }, + { + "epoch": 1.84, + "learning_rate": 1.6036181198959973e-08, + "loss": 0.6913, + "step": 50285 + }, + { + "epoch": 1.84, + "learning_rate": 1.6032519134287913e-08, + "loss": 0.6919, + "step": 50286 + }, + { + "epoch": 1.84, + "learning_rate": 1.6028857069615846e-08, + "loss": 0.6928, + "step": 50287 + }, + { + "epoch": 1.84, + "learning_rate": 1.6025195004943786e-08, + "loss": 0.6984, + "step": 50288 + }, + { + "epoch": 1.84, + "learning_rate": 1.6021532940271726e-08, + "loss": 0.6836, + "step": 50289 + }, + { + "epoch": 1.84, + "learning_rate": 1.6017870875599663e-08, + "loss": 0.6868, + "step": 50290 + }, + { + "epoch": 1.84, + "learning_rate": 1.6014208810927603e-08, + "loss": 0.6792, + "step": 50291 + }, + { + "epoch": 1.84, + "learning_rate": 1.601054674625554e-08, + "loss": 0.6988, + "step": 50292 + }, + { + "epoch": 1.84, + "learning_rate": 1.6006884681583476e-08, + "loss": 0.6906, + "step": 50293 + }, + { + "epoch": 1.84, + "learning_rate": 1.6003222616911412e-08, + "loss": 0.6942, + "step": 50294 + }, + { + "epoch": 1.84, + "learning_rate": 1.5999560552239352e-08, + "loss": 0.6962, + "step": 50295 + }, + { + "epoch": 1.84, + "learning_rate": 1.599589848756729e-08, + "loss": 0.6908, + "step": 50296 + }, + { + "epoch": 1.84, + "learning_rate": 1.599223642289523e-08, + "loss": 0.6971, + "step": 50297 + }, + { + "epoch": 1.84, + "learning_rate": 1.5988574358223166e-08, + "loss": 0.69, + "step": 50298 + }, + { + "epoch": 1.84, + "learning_rate": 1.5984912293551102e-08, + "loss": 0.6949, + "step": 50299 + }, + { + "epoch": 1.84, + "learning_rate": 1.5981250228879042e-08, + "loss": 0.6883, + "step": 50300 + }, + { + "epoch": 1.84, + "learning_rate": 1.597758816420698e-08, + "loss": 0.7068, + "step": 50301 + }, + { + "epoch": 1.84, + "learning_rate": 1.597392609953492e-08, + "loss": 0.6977, + "step": 50302 + }, + { + "epoch": 1.84, + "learning_rate": 1.5970264034862855e-08, + "loss": 0.6981, + "step": 50303 + }, + { + "epoch": 1.84, + "learning_rate": 1.5966601970190792e-08, + "loss": 0.6942, + "step": 50304 + }, + { + "epoch": 1.84, + "learning_rate": 1.596293990551873e-08, + "loss": 0.6894, + "step": 50305 + }, + { + "epoch": 1.84, + "learning_rate": 1.595927784084667e-08, + "loss": 0.68, + "step": 50306 + }, + { + "epoch": 1.84, + "learning_rate": 1.5955615776174605e-08, + "loss": 0.6931, + "step": 50307 + }, + { + "epoch": 1.84, + "learning_rate": 1.5951953711502545e-08, + "loss": 0.6922, + "step": 50308 + }, + { + "epoch": 1.84, + "learning_rate": 1.5948291646830485e-08, + "loss": 0.6884, + "step": 50309 + }, + { + "epoch": 1.84, + "learning_rate": 1.5944629582158418e-08, + "loss": 0.7017, + "step": 50310 + }, + { + "epoch": 1.84, + "learning_rate": 1.5940967517486358e-08, + "loss": 0.7064, + "step": 50311 + }, + { + "epoch": 1.84, + "learning_rate": 1.5937305452814295e-08, + "loss": 0.6771, + "step": 50312 + }, + { + "epoch": 1.84, + "learning_rate": 1.5933643388142235e-08, + "loss": 0.6933, + "step": 50313 + }, + { + "epoch": 1.84, + "learning_rate": 1.592998132347017e-08, + "loss": 0.6944, + "step": 50314 + }, + { + "epoch": 1.84, + "learning_rate": 1.592631925879811e-08, + "loss": 0.6938, + "step": 50315 + }, + { + "epoch": 1.84, + "learning_rate": 1.5922657194126048e-08, + "loss": 0.6979, + "step": 50316 + }, + { + "epoch": 1.84, + "learning_rate": 1.5918995129453984e-08, + "loss": 0.7051, + "step": 50317 + }, + { + "epoch": 1.84, + "learning_rate": 1.5915333064781924e-08, + "loss": 0.6962, + "step": 50318 + }, + { + "epoch": 1.84, + "learning_rate": 1.591167100010986e-08, + "loss": 0.687, + "step": 50319 + }, + { + "epoch": 1.84, + "learning_rate": 1.59080089354378e-08, + "loss": 0.6866, + "step": 50320 + }, + { + "epoch": 1.84, + "learning_rate": 1.5904346870765738e-08, + "loss": 0.6922, + "step": 50321 + }, + { + "epoch": 1.84, + "learning_rate": 1.5900684806093674e-08, + "loss": 0.6912, + "step": 50322 + }, + { + "epoch": 1.84, + "learning_rate": 1.589702274142161e-08, + "loss": 0.6961, + "step": 50323 + }, + { + "epoch": 1.84, + "learning_rate": 1.589336067674955e-08, + "loss": 0.6888, + "step": 50324 + }, + { + "epoch": 1.84, + "learning_rate": 1.5889698612077487e-08, + "loss": 0.6829, + "step": 50325 + }, + { + "epoch": 1.84, + "learning_rate": 1.5886036547405427e-08, + "loss": 0.7006, + "step": 50326 + }, + { + "epoch": 1.84, + "learning_rate": 1.5882374482733367e-08, + "loss": 0.6942, + "step": 50327 + }, + { + "epoch": 1.84, + "learning_rate": 1.58787124180613e-08, + "loss": 0.6922, + "step": 50328 + }, + { + "epoch": 1.84, + "learning_rate": 1.587505035338924e-08, + "loss": 0.6895, + "step": 50329 + }, + { + "epoch": 1.84, + "learning_rate": 1.5871388288717177e-08, + "loss": 0.7037, + "step": 50330 + }, + { + "epoch": 1.84, + "learning_rate": 1.5867726224045117e-08, + "loss": 0.7002, + "step": 50331 + }, + { + "epoch": 1.84, + "learning_rate": 1.5864064159373054e-08, + "loss": 0.7026, + "step": 50332 + }, + { + "epoch": 1.84, + "learning_rate": 1.5860402094700994e-08, + "loss": 0.6992, + "step": 50333 + }, + { + "epoch": 1.84, + "learning_rate": 1.5856740030028927e-08, + "loss": 0.6962, + "step": 50334 + }, + { + "epoch": 1.84, + "learning_rate": 1.5853077965356867e-08, + "loss": 0.6997, + "step": 50335 + }, + { + "epoch": 1.84, + "learning_rate": 1.5849415900684807e-08, + "loss": 0.6904, + "step": 50336 + }, + { + "epoch": 1.84, + "learning_rate": 1.5845753836012743e-08, + "loss": 0.6873, + "step": 50337 + }, + { + "epoch": 1.84, + "learning_rate": 1.5842091771340683e-08, + "loss": 0.6797, + "step": 50338 + }, + { + "epoch": 1.84, + "learning_rate": 1.583842970666862e-08, + "loss": 0.6966, + "step": 50339 + }, + { + "epoch": 1.84, + "learning_rate": 1.5834767641996556e-08, + "loss": 0.6852, + "step": 50340 + }, + { + "epoch": 1.84, + "learning_rate": 1.5831105577324493e-08, + "loss": 0.6889, + "step": 50341 + }, + { + "epoch": 1.84, + "learning_rate": 1.5827443512652433e-08, + "loss": 0.6901, + "step": 50342 + }, + { + "epoch": 1.84, + "learning_rate": 1.582378144798037e-08, + "loss": 0.6876, + "step": 50343 + }, + { + "epoch": 1.84, + "learning_rate": 1.582011938330831e-08, + "loss": 0.6917, + "step": 50344 + }, + { + "epoch": 1.84, + "learning_rate": 1.5816457318636246e-08, + "loss": 0.6861, + "step": 50345 + }, + { + "epoch": 1.84, + "learning_rate": 1.5812795253964183e-08, + "loss": 0.6905, + "step": 50346 + }, + { + "epoch": 1.84, + "learning_rate": 1.5809133189292123e-08, + "loss": 0.6961, + "step": 50347 + }, + { + "epoch": 1.84, + "learning_rate": 1.580547112462006e-08, + "loss": 0.6964, + "step": 50348 + }, + { + "epoch": 1.84, + "learning_rate": 1.5801809059948e-08, + "loss": 0.694, + "step": 50349 + }, + { + "epoch": 1.84, + "learning_rate": 1.5798146995275936e-08, + "loss": 0.7124, + "step": 50350 + }, + { + "epoch": 1.84, + "learning_rate": 1.5794484930603873e-08, + "loss": 0.6769, + "step": 50351 + }, + { + "epoch": 1.84, + "learning_rate": 1.579082286593181e-08, + "loss": 0.6913, + "step": 50352 + }, + { + "epoch": 1.84, + "learning_rate": 1.578716080125975e-08, + "loss": 0.7001, + "step": 50353 + }, + { + "epoch": 1.84, + "learning_rate": 1.578349873658769e-08, + "loss": 0.6975, + "step": 50354 + }, + { + "epoch": 1.84, + "learning_rate": 1.5779836671915626e-08, + "loss": 0.6856, + "step": 50355 + }, + { + "epoch": 1.84, + "learning_rate": 1.5776174607243566e-08, + "loss": 0.6949, + "step": 50356 + }, + { + "epoch": 1.84, + "learning_rate": 1.5772512542571502e-08, + "loss": 0.6968, + "step": 50357 + }, + { + "epoch": 1.84, + "learning_rate": 1.576885047789944e-08, + "loss": 0.6966, + "step": 50358 + }, + { + "epoch": 1.84, + "learning_rate": 1.5765188413227375e-08, + "loss": 0.6903, + "step": 50359 + }, + { + "epoch": 1.84, + "learning_rate": 1.5761526348555315e-08, + "loss": 0.6827, + "step": 50360 + }, + { + "epoch": 1.84, + "learning_rate": 1.5757864283883252e-08, + "loss": 0.6899, + "step": 50361 + }, + { + "epoch": 1.84, + "learning_rate": 1.5754202219211192e-08, + "loss": 0.6934, + "step": 50362 + }, + { + "epoch": 1.84, + "learning_rate": 1.575054015453913e-08, + "loss": 0.6872, + "step": 50363 + }, + { + "epoch": 1.84, + "learning_rate": 1.5746878089867065e-08, + "loss": 0.6924, + "step": 50364 + }, + { + "epoch": 1.84, + "learning_rate": 1.5743216025195005e-08, + "loss": 0.6842, + "step": 50365 + }, + { + "epoch": 1.84, + "learning_rate": 1.5739553960522942e-08, + "loss": 0.6813, + "step": 50366 + }, + { + "epoch": 1.84, + "learning_rate": 1.573589189585088e-08, + "loss": 0.6954, + "step": 50367 + }, + { + "epoch": 1.84, + "learning_rate": 1.5732229831178818e-08, + "loss": 0.6886, + "step": 50368 + }, + { + "epoch": 1.84, + "learning_rate": 1.5728567766506755e-08, + "loss": 0.699, + "step": 50369 + }, + { + "epoch": 1.84, + "learning_rate": 1.572490570183469e-08, + "loss": 0.7004, + "step": 50370 + }, + { + "epoch": 1.84, + "learning_rate": 1.572124363716263e-08, + "loss": 0.6877, + "step": 50371 + }, + { + "epoch": 1.84, + "learning_rate": 1.571758157249057e-08, + "loss": 0.6967, + "step": 50372 + }, + { + "epoch": 1.84, + "learning_rate": 1.5713919507818508e-08, + "loss": 0.6893, + "step": 50373 + }, + { + "epoch": 1.84, + "learning_rate": 1.5710257443146448e-08, + "loss": 0.6931, + "step": 50374 + }, + { + "epoch": 1.84, + "learning_rate": 1.570659537847438e-08, + "loss": 0.6912, + "step": 50375 + }, + { + "epoch": 1.84, + "learning_rate": 1.570293331380232e-08, + "loss": 0.6884, + "step": 50376 + }, + { + "epoch": 1.84, + "learning_rate": 1.5699271249130258e-08, + "loss": 0.6834, + "step": 50377 + }, + { + "epoch": 1.84, + "learning_rate": 1.5695609184458198e-08, + "loss": 0.6997, + "step": 50378 + }, + { + "epoch": 1.84, + "learning_rate": 1.5691947119786134e-08, + "loss": 0.6915, + "step": 50379 + }, + { + "epoch": 1.84, + "learning_rate": 1.5688285055114074e-08, + "loss": 0.6912, + "step": 50380 + }, + { + "epoch": 1.84, + "learning_rate": 1.568462299044201e-08, + "loss": 0.6956, + "step": 50381 + }, + { + "epoch": 1.84, + "learning_rate": 1.5680960925769947e-08, + "loss": 0.6943, + "step": 50382 + }, + { + "epoch": 1.84, + "learning_rate": 1.5677298861097887e-08, + "loss": 0.6897, + "step": 50383 + }, + { + "epoch": 1.84, + "learning_rate": 1.5673636796425824e-08, + "loss": 0.6937, + "step": 50384 + }, + { + "epoch": 1.84, + "learning_rate": 1.5669974731753764e-08, + "loss": 0.6886, + "step": 50385 + }, + { + "epoch": 1.84, + "learning_rate": 1.56663126670817e-08, + "loss": 0.7006, + "step": 50386 + }, + { + "epoch": 1.84, + "learning_rate": 1.5662650602409637e-08, + "loss": 0.6952, + "step": 50387 + }, + { + "epoch": 1.84, + "learning_rate": 1.5658988537737574e-08, + "loss": 0.6914, + "step": 50388 + }, + { + "epoch": 1.84, + "learning_rate": 1.5655326473065514e-08, + "loss": 0.6846, + "step": 50389 + }, + { + "epoch": 1.84, + "learning_rate": 1.565166440839345e-08, + "loss": 0.6854, + "step": 50390 + }, + { + "epoch": 1.84, + "learning_rate": 1.564800234372139e-08, + "loss": 0.6946, + "step": 50391 + }, + { + "epoch": 1.84, + "learning_rate": 1.5644340279049327e-08, + "loss": 0.706, + "step": 50392 + }, + { + "epoch": 1.84, + "learning_rate": 1.5640678214377264e-08, + "loss": 0.6993, + "step": 50393 + }, + { + "epoch": 1.84, + "learning_rate": 1.5637016149705203e-08, + "loss": 0.6977, + "step": 50394 + }, + { + "epoch": 1.84, + "learning_rate": 1.563335408503314e-08, + "loss": 0.6865, + "step": 50395 + }, + { + "epoch": 1.84, + "learning_rate": 1.562969202036108e-08, + "loss": 0.6924, + "step": 50396 + }, + { + "epoch": 1.84, + "learning_rate": 1.5626029955689017e-08, + "loss": 0.6933, + "step": 50397 + }, + { + "epoch": 1.84, + "learning_rate": 1.5622367891016953e-08, + "loss": 0.6989, + "step": 50398 + }, + { + "epoch": 1.84, + "learning_rate": 1.561870582634489e-08, + "loss": 0.6949, + "step": 50399 + }, + { + "epoch": 1.84, + "learning_rate": 1.561504376167283e-08, + "loss": 0.6757, + "step": 50400 + }, + { + "epoch": 1.84, + "learning_rate": 1.561138169700077e-08, + "loss": 0.6938, + "step": 50401 + }, + { + "epoch": 1.84, + "learning_rate": 1.5607719632328706e-08, + "loss": 0.6874, + "step": 50402 + }, + { + "epoch": 1.84, + "learning_rate": 1.5604057567656646e-08, + "loss": 0.7015, + "step": 50403 + }, + { + "epoch": 1.84, + "learning_rate": 1.5600395502984583e-08, + "loss": 0.6875, + "step": 50404 + }, + { + "epoch": 1.84, + "learning_rate": 1.559673343831252e-08, + "loss": 0.6912, + "step": 50405 + }, + { + "epoch": 1.84, + "learning_rate": 1.5593071373640456e-08, + "loss": 0.6948, + "step": 50406 + }, + { + "epoch": 1.84, + "learning_rate": 1.5589409308968396e-08, + "loss": 0.6952, + "step": 50407 + }, + { + "epoch": 1.84, + "learning_rate": 1.5585747244296333e-08, + "loss": 0.7015, + "step": 50408 + }, + { + "epoch": 1.84, + "learning_rate": 1.5582085179624273e-08, + "loss": 0.6903, + "step": 50409 + }, + { + "epoch": 1.84, + "learning_rate": 1.557842311495221e-08, + "loss": 0.6964, + "step": 50410 + }, + { + "epoch": 1.84, + "learning_rate": 1.5574761050280146e-08, + "loss": 0.6977, + "step": 50411 + }, + { + "epoch": 1.84, + "learning_rate": 1.5571098985608086e-08, + "loss": 0.6934, + "step": 50412 + }, + { + "epoch": 1.84, + "learning_rate": 1.5567436920936022e-08, + "loss": 0.6985, + "step": 50413 + }, + { + "epoch": 1.84, + "learning_rate": 1.5563774856263962e-08, + "loss": 0.6843, + "step": 50414 + }, + { + "epoch": 1.84, + "learning_rate": 1.55601127915919e-08, + "loss": 0.677, + "step": 50415 + }, + { + "epoch": 1.84, + "learning_rate": 1.5556450726919836e-08, + "loss": 0.7043, + "step": 50416 + }, + { + "epoch": 1.84, + "learning_rate": 1.5552788662247772e-08, + "loss": 0.6941, + "step": 50417 + }, + { + "epoch": 1.84, + "learning_rate": 1.5549126597575712e-08, + "loss": 0.6939, + "step": 50418 + }, + { + "epoch": 1.84, + "learning_rate": 1.5545464532903652e-08, + "loss": 0.7005, + "step": 50419 + }, + { + "epoch": 1.84, + "learning_rate": 1.554180246823159e-08, + "loss": 0.6911, + "step": 50420 + }, + { + "epoch": 1.84, + "learning_rate": 1.553814040355953e-08, + "loss": 0.6851, + "step": 50421 + }, + { + "epoch": 1.84, + "learning_rate": 1.5534478338887462e-08, + "loss": 0.7006, + "step": 50422 + }, + { + "epoch": 1.84, + "learning_rate": 1.5530816274215402e-08, + "loss": 0.7008, + "step": 50423 + }, + { + "epoch": 1.84, + "learning_rate": 1.552715420954334e-08, + "loss": 0.6935, + "step": 50424 + }, + { + "epoch": 1.84, + "learning_rate": 1.552349214487128e-08, + "loss": 0.6988, + "step": 50425 + }, + { + "epoch": 1.84, + "learning_rate": 1.5519830080199215e-08, + "loss": 0.6927, + "step": 50426 + }, + { + "epoch": 1.84, + "learning_rate": 1.5516168015527155e-08, + "loss": 0.6838, + "step": 50427 + }, + { + "epoch": 1.85, + "learning_rate": 1.551250595085509e-08, + "loss": 0.7041, + "step": 50428 + }, + { + "epoch": 1.85, + "learning_rate": 1.5508843886183028e-08, + "loss": 0.6907, + "step": 50429 + }, + { + "epoch": 1.85, + "learning_rate": 1.5505181821510968e-08, + "loss": 0.7022, + "step": 50430 + }, + { + "epoch": 1.85, + "learning_rate": 1.5501519756838905e-08, + "loss": 0.6966, + "step": 50431 + }, + { + "epoch": 1.85, + "learning_rate": 1.5497857692166845e-08, + "loss": 0.6787, + "step": 50432 + }, + { + "epoch": 1.85, + "learning_rate": 1.549419562749478e-08, + "loss": 0.6845, + "step": 50433 + }, + { + "epoch": 1.85, + "learning_rate": 1.5490533562822718e-08, + "loss": 0.6948, + "step": 50434 + }, + { + "epoch": 1.85, + "learning_rate": 1.5486871498150654e-08, + "loss": 0.69, + "step": 50435 + }, + { + "epoch": 1.85, + "learning_rate": 1.5483209433478594e-08, + "loss": 0.6847, + "step": 50436 + }, + { + "epoch": 1.85, + "learning_rate": 1.5479547368806534e-08, + "loss": 0.7077, + "step": 50437 + }, + { + "epoch": 1.85, + "learning_rate": 1.547588530413447e-08, + "loss": 0.6832, + "step": 50438 + }, + { + "epoch": 1.85, + "learning_rate": 1.5472223239462408e-08, + "loss": 0.6916, + "step": 50439 + }, + { + "epoch": 1.85, + "learning_rate": 1.5468561174790344e-08, + "loss": 0.6815, + "step": 50440 + }, + { + "epoch": 1.85, + "learning_rate": 1.5464899110118284e-08, + "loss": 0.6847, + "step": 50441 + }, + { + "epoch": 1.85, + "learning_rate": 1.546123704544622e-08, + "loss": 0.702, + "step": 50442 + }, + { + "epoch": 1.85, + "learning_rate": 1.545757498077416e-08, + "loss": 0.7093, + "step": 50443 + }, + { + "epoch": 1.85, + "learning_rate": 1.5453912916102097e-08, + "loss": 0.6886, + "step": 50444 + }, + { + "epoch": 1.85, + "learning_rate": 1.5450250851430037e-08, + "loss": 0.6951, + "step": 50445 + }, + { + "epoch": 1.85, + "learning_rate": 1.5446588786757974e-08, + "loss": 0.7029, + "step": 50446 + }, + { + "epoch": 1.85, + "learning_rate": 1.544292672208591e-08, + "loss": 0.6902, + "step": 50447 + }, + { + "epoch": 1.85, + "learning_rate": 1.543926465741385e-08, + "loss": 0.6971, + "step": 50448 + }, + { + "epoch": 1.85, + "learning_rate": 1.5435602592741787e-08, + "loss": 0.6895, + "step": 50449 + }, + { + "epoch": 1.85, + "learning_rate": 1.5431940528069727e-08, + "loss": 0.6864, + "step": 50450 + }, + { + "epoch": 1.85, + "learning_rate": 1.5428278463397664e-08, + "loss": 0.7058, + "step": 50451 + }, + { + "epoch": 1.85, + "learning_rate": 1.54246163987256e-08, + "loss": 0.6853, + "step": 50452 + }, + { + "epoch": 1.85, + "learning_rate": 1.5420954334053537e-08, + "loss": 0.6952, + "step": 50453 + }, + { + "epoch": 1.85, + "learning_rate": 1.5417292269381477e-08, + "loss": 0.6922, + "step": 50454 + }, + { + "epoch": 1.85, + "learning_rate": 1.5413630204709417e-08, + "loss": 0.6848, + "step": 50455 + }, + { + "epoch": 1.85, + "learning_rate": 1.5409968140037353e-08, + "loss": 0.7082, + "step": 50456 + }, + { + "epoch": 1.85, + "learning_rate": 1.540630607536529e-08, + "loss": 0.6953, + "step": 50457 + }, + { + "epoch": 1.85, + "learning_rate": 1.5402644010693227e-08, + "loss": 0.6821, + "step": 50458 + }, + { + "epoch": 1.85, + "learning_rate": 1.5398981946021166e-08, + "loss": 0.6906, + "step": 50459 + }, + { + "epoch": 1.85, + "learning_rate": 1.5395319881349103e-08, + "loss": 0.6941, + "step": 50460 + }, + { + "epoch": 1.85, + "learning_rate": 1.5391657816677043e-08, + "loss": 0.6834, + "step": 50461 + }, + { + "epoch": 1.85, + "learning_rate": 1.538799575200498e-08, + "loss": 0.6933, + "step": 50462 + }, + { + "epoch": 1.85, + "learning_rate": 1.5384333687332916e-08, + "loss": 0.7111, + "step": 50463 + }, + { + "epoch": 1.85, + "learning_rate": 1.5380671622660856e-08, + "loss": 0.7004, + "step": 50464 + }, + { + "epoch": 1.85, + "learning_rate": 1.5377009557988793e-08, + "loss": 0.6941, + "step": 50465 + }, + { + "epoch": 1.85, + "learning_rate": 1.5373347493316733e-08, + "loss": 0.6936, + "step": 50466 + }, + { + "epoch": 1.85, + "learning_rate": 1.536968542864467e-08, + "loss": 0.6934, + "step": 50467 + }, + { + "epoch": 1.85, + "learning_rate": 1.536602336397261e-08, + "loss": 0.6895, + "step": 50468 + }, + { + "epoch": 1.85, + "learning_rate": 1.5362361299300543e-08, + "loss": 0.6946, + "step": 50469 + }, + { + "epoch": 1.85, + "learning_rate": 1.5358699234628482e-08, + "loss": 0.695, + "step": 50470 + }, + { + "epoch": 1.85, + "learning_rate": 1.535503716995642e-08, + "loss": 0.6985, + "step": 50471 + }, + { + "epoch": 1.85, + "learning_rate": 1.535137510528436e-08, + "loss": 0.6956, + "step": 50472 + }, + { + "epoch": 1.85, + "learning_rate": 1.53477130406123e-08, + "loss": 0.6865, + "step": 50473 + }, + { + "epoch": 1.85, + "learning_rate": 1.5344050975940236e-08, + "loss": 0.6893, + "step": 50474 + }, + { + "epoch": 1.85, + "learning_rate": 1.5340388911268172e-08, + "loss": 0.6889, + "step": 50475 + }, + { + "epoch": 1.85, + "learning_rate": 1.533672684659611e-08, + "loss": 0.7035, + "step": 50476 + }, + { + "epoch": 1.85, + "learning_rate": 1.533306478192405e-08, + "loss": 0.6928, + "step": 50477 + }, + { + "epoch": 1.85, + "learning_rate": 1.5329402717251985e-08, + "loss": 0.6987, + "step": 50478 + }, + { + "epoch": 1.85, + "learning_rate": 1.5325740652579925e-08, + "loss": 0.6919, + "step": 50479 + }, + { + "epoch": 1.85, + "learning_rate": 1.5322078587907862e-08, + "loss": 0.702, + "step": 50480 + }, + { + "epoch": 1.85, + "learning_rate": 1.53184165232358e-08, + "loss": 0.6907, + "step": 50481 + }, + { + "epoch": 1.85, + "learning_rate": 1.5314754458563735e-08, + "loss": 0.7001, + "step": 50482 + }, + { + "epoch": 1.85, + "learning_rate": 1.5311092393891675e-08, + "loss": 0.6961, + "step": 50483 + }, + { + "epoch": 1.85, + "learning_rate": 1.5307430329219615e-08, + "loss": 0.6817, + "step": 50484 + }, + { + "epoch": 1.85, + "learning_rate": 1.530376826454755e-08, + "loss": 0.6856, + "step": 50485 + }, + { + "epoch": 1.85, + "learning_rate": 1.5300106199875488e-08, + "loss": 0.6974, + "step": 50486 + }, + { + "epoch": 1.85, + "learning_rate": 1.5296444135203425e-08, + "loss": 0.7074, + "step": 50487 + }, + { + "epoch": 1.85, + "learning_rate": 1.5292782070531365e-08, + "loss": 0.6944, + "step": 50488 + }, + { + "epoch": 1.85, + "learning_rate": 1.52891200058593e-08, + "loss": 0.6953, + "step": 50489 + }, + { + "epoch": 1.85, + "learning_rate": 1.528545794118724e-08, + "loss": 0.7019, + "step": 50490 + }, + { + "epoch": 1.85, + "learning_rate": 1.5281795876515178e-08, + "loss": 0.6832, + "step": 50491 + }, + { + "epoch": 1.85, + "learning_rate": 1.5278133811843118e-08, + "loss": 0.7013, + "step": 50492 + }, + { + "epoch": 1.85, + "learning_rate": 1.5274471747171055e-08, + "loss": 0.6968, + "step": 50493 + }, + { + "epoch": 1.85, + "learning_rate": 1.527080968249899e-08, + "loss": 0.6915, + "step": 50494 + }, + { + "epoch": 1.85, + "learning_rate": 1.526714761782693e-08, + "loss": 0.6942, + "step": 50495 + }, + { + "epoch": 1.85, + "learning_rate": 1.5263485553154868e-08, + "loss": 0.6982, + "step": 50496 + }, + { + "epoch": 1.85, + "learning_rate": 1.5259823488482808e-08, + "loss": 0.6819, + "step": 50497 + }, + { + "epoch": 1.85, + "learning_rate": 1.5256161423810744e-08, + "loss": 0.6889, + "step": 50498 + }, + { + "epoch": 1.85, + "learning_rate": 1.525249935913868e-08, + "loss": 0.697, + "step": 50499 + }, + { + "epoch": 1.85, + "learning_rate": 1.5248837294466617e-08, + "loss": 0.6921, + "step": 50500 + }, + { + "epoch": 1.85, + "learning_rate": 1.5245175229794557e-08, + "loss": 0.7005, + "step": 50501 + }, + { + "epoch": 1.85, + "learning_rate": 1.5241513165122497e-08, + "loss": 0.6929, + "step": 50502 + }, + { + "epoch": 1.85, + "learning_rate": 1.5237851100450434e-08, + "loss": 0.6793, + "step": 50503 + }, + { + "epoch": 1.85, + "learning_rate": 1.523418903577837e-08, + "loss": 0.6991, + "step": 50504 + }, + { + "epoch": 1.85, + "learning_rate": 1.5230526971106307e-08, + "loss": 0.6885, + "step": 50505 + }, + { + "epoch": 1.85, + "learning_rate": 1.5226864906434247e-08, + "loss": 0.7034, + "step": 50506 + }, + { + "epoch": 1.85, + "learning_rate": 1.5223202841762184e-08, + "loss": 0.6895, + "step": 50507 + }, + { + "epoch": 1.85, + "learning_rate": 1.5219540777090124e-08, + "loss": 0.6925, + "step": 50508 + }, + { + "epoch": 1.85, + "learning_rate": 1.521587871241806e-08, + "loss": 0.6868, + "step": 50509 + }, + { + "epoch": 1.85, + "learning_rate": 1.5212216647745997e-08, + "loss": 0.6851, + "step": 50510 + }, + { + "epoch": 1.85, + "learning_rate": 1.5208554583073937e-08, + "loss": 0.6886, + "step": 50511 + }, + { + "epoch": 1.85, + "learning_rate": 1.5204892518401873e-08, + "loss": 0.6931, + "step": 50512 + }, + { + "epoch": 1.85, + "learning_rate": 1.5201230453729813e-08, + "loss": 0.6981, + "step": 50513 + }, + { + "epoch": 1.85, + "learning_rate": 1.519756838905775e-08, + "loss": 0.6941, + "step": 50514 + }, + { + "epoch": 1.85, + "learning_rate": 1.519390632438569e-08, + "loss": 0.7018, + "step": 50515 + }, + { + "epoch": 1.85, + "learning_rate": 1.5190244259713623e-08, + "loss": 0.6896, + "step": 50516 + }, + { + "epoch": 1.85, + "learning_rate": 1.5186582195041563e-08, + "loss": 0.6925, + "step": 50517 + }, + { + "epoch": 1.85, + "learning_rate": 1.51829201303695e-08, + "loss": 0.6901, + "step": 50518 + }, + { + "epoch": 1.85, + "learning_rate": 1.517925806569744e-08, + "loss": 0.6882, + "step": 50519 + }, + { + "epoch": 1.85, + "learning_rate": 1.517559600102538e-08, + "loss": 0.6903, + "step": 50520 + }, + { + "epoch": 1.85, + "learning_rate": 1.5171933936353316e-08, + "loss": 0.705, + "step": 50521 + }, + { + "epoch": 1.85, + "learning_rate": 1.5168271871681253e-08, + "loss": 0.6976, + "step": 50522 + }, + { + "epoch": 1.85, + "learning_rate": 1.516460980700919e-08, + "loss": 0.7056, + "step": 50523 + }, + { + "epoch": 1.85, + "learning_rate": 1.516094774233713e-08, + "loss": 0.7091, + "step": 50524 + }, + { + "epoch": 1.85, + "learning_rate": 1.5157285677665066e-08, + "loss": 0.6984, + "step": 50525 + }, + { + "epoch": 1.85, + "learning_rate": 1.5153623612993006e-08, + "loss": 0.6948, + "step": 50526 + }, + { + "epoch": 1.85, + "learning_rate": 1.5149961548320943e-08, + "loss": 0.7022, + "step": 50527 + }, + { + "epoch": 1.85, + "learning_rate": 1.514629948364888e-08, + "loss": 0.6857, + "step": 50528 + }, + { + "epoch": 1.85, + "learning_rate": 1.514263741897682e-08, + "loss": 0.6763, + "step": 50529 + }, + { + "epoch": 1.85, + "learning_rate": 1.5138975354304756e-08, + "loss": 0.6979, + "step": 50530 + }, + { + "epoch": 1.85, + "learning_rate": 1.5135313289632696e-08, + "loss": 0.6954, + "step": 50531 + }, + { + "epoch": 1.85, + "learning_rate": 1.5131651224960632e-08, + "loss": 0.6921, + "step": 50532 + }, + { + "epoch": 1.85, + "learning_rate": 1.5127989160288572e-08, + "loss": 0.6951, + "step": 50533 + }, + { + "epoch": 1.85, + "learning_rate": 1.5124327095616506e-08, + "loss": 0.6956, + "step": 50534 + }, + { + "epoch": 1.85, + "learning_rate": 1.5120665030944446e-08, + "loss": 0.6901, + "step": 50535 + }, + { + "epoch": 1.85, + "learning_rate": 1.5117002966272382e-08, + "loss": 0.6866, + "step": 50536 + }, + { + "epoch": 1.85, + "learning_rate": 1.5113340901600322e-08, + "loss": 0.6916, + "step": 50537 + }, + { + "epoch": 1.85, + "learning_rate": 1.5109678836928262e-08, + "loss": 0.705, + "step": 50538 + }, + { + "epoch": 1.85, + "learning_rate": 1.51060167722562e-08, + "loss": 0.6918, + "step": 50539 + }, + { + "epoch": 1.85, + "learning_rate": 1.5102354707584135e-08, + "loss": 0.6913, + "step": 50540 + }, + { + "epoch": 1.85, + "learning_rate": 1.5098692642912072e-08, + "loss": 0.6996, + "step": 50541 + }, + { + "epoch": 1.85, + "learning_rate": 1.5095030578240012e-08, + "loss": 0.6929, + "step": 50542 + }, + { + "epoch": 1.85, + "learning_rate": 1.509136851356795e-08, + "loss": 0.6943, + "step": 50543 + }, + { + "epoch": 1.85, + "learning_rate": 1.5087706448895888e-08, + "loss": 0.7, + "step": 50544 + }, + { + "epoch": 1.85, + "learning_rate": 1.5084044384223825e-08, + "loss": 0.6946, + "step": 50545 + }, + { + "epoch": 1.85, + "learning_rate": 1.508038231955176e-08, + "loss": 0.6734, + "step": 50546 + }, + { + "epoch": 1.85, + "learning_rate": 1.50767202548797e-08, + "loss": 0.6941, + "step": 50547 + }, + { + "epoch": 1.85, + "learning_rate": 1.5073058190207638e-08, + "loss": 0.6938, + "step": 50548 + }, + { + "epoch": 1.85, + "learning_rate": 1.5069396125535578e-08, + "loss": 0.6858, + "step": 50549 + }, + { + "epoch": 1.85, + "learning_rate": 1.5065734060863515e-08, + "loss": 0.6879, + "step": 50550 + }, + { + "epoch": 1.85, + "learning_rate": 1.506207199619145e-08, + "loss": 0.7051, + "step": 50551 + }, + { + "epoch": 1.85, + "learning_rate": 1.5058409931519388e-08, + "loss": 0.6985, + "step": 50552 + }, + { + "epoch": 1.85, + "learning_rate": 1.5054747866847328e-08, + "loss": 0.6915, + "step": 50553 + }, + { + "epoch": 1.85, + "learning_rate": 1.5051085802175264e-08, + "loss": 0.6901, + "step": 50554 + }, + { + "epoch": 1.85, + "learning_rate": 1.5047423737503204e-08, + "loss": 0.6852, + "step": 50555 + }, + { + "epoch": 1.85, + "learning_rate": 1.5043761672831144e-08, + "loss": 0.7023, + "step": 50556 + }, + { + "epoch": 1.85, + "learning_rate": 1.5040099608159078e-08, + "loss": 0.683, + "step": 50557 + }, + { + "epoch": 1.85, + "learning_rate": 1.5036437543487018e-08, + "loss": 0.6813, + "step": 50558 + }, + { + "epoch": 1.85, + "learning_rate": 1.5032775478814954e-08, + "loss": 0.6901, + "step": 50559 + }, + { + "epoch": 1.85, + "learning_rate": 1.5029113414142894e-08, + "loss": 0.6882, + "step": 50560 + }, + { + "epoch": 1.85, + "learning_rate": 1.502545134947083e-08, + "loss": 0.6794, + "step": 50561 + }, + { + "epoch": 1.85, + "learning_rate": 1.502178928479877e-08, + "loss": 0.7006, + "step": 50562 + }, + { + "epoch": 1.85, + "learning_rate": 1.5018127220126707e-08, + "loss": 0.6986, + "step": 50563 + }, + { + "epoch": 1.85, + "learning_rate": 1.5014465155454644e-08, + "loss": 0.7037, + "step": 50564 + }, + { + "epoch": 1.85, + "learning_rate": 1.5010803090782584e-08, + "loss": 0.6953, + "step": 50565 + }, + { + "epoch": 1.85, + "learning_rate": 1.500714102611052e-08, + "loss": 0.6917, + "step": 50566 + }, + { + "epoch": 1.85, + "learning_rate": 1.500347896143846e-08, + "loss": 0.6967, + "step": 50567 + }, + { + "epoch": 1.85, + "learning_rate": 1.4999816896766397e-08, + "loss": 0.6954, + "step": 50568 + }, + { + "epoch": 1.85, + "learning_rate": 1.4996154832094334e-08, + "loss": 0.6824, + "step": 50569 + }, + { + "epoch": 1.85, + "learning_rate": 1.499249276742227e-08, + "loss": 0.6901, + "step": 50570 + }, + { + "epoch": 1.85, + "learning_rate": 1.498883070275021e-08, + "loss": 0.6958, + "step": 50571 + }, + { + "epoch": 1.85, + "learning_rate": 1.4985168638078147e-08, + "loss": 0.6997, + "step": 50572 + }, + { + "epoch": 1.85, + "learning_rate": 1.4981506573406087e-08, + "loss": 0.6927, + "step": 50573 + }, + { + "epoch": 1.85, + "learning_rate": 1.4977844508734023e-08, + "loss": 0.6918, + "step": 50574 + }, + { + "epoch": 1.85, + "learning_rate": 1.497418244406196e-08, + "loss": 0.6926, + "step": 50575 + }, + { + "epoch": 1.85, + "learning_rate": 1.49705203793899e-08, + "loss": 0.7023, + "step": 50576 + }, + { + "epoch": 1.85, + "learning_rate": 1.4966858314717836e-08, + "loss": 0.676, + "step": 50577 + }, + { + "epoch": 1.85, + "learning_rate": 1.4963196250045776e-08, + "loss": 0.7033, + "step": 50578 + }, + { + "epoch": 1.85, + "learning_rate": 1.4959534185373713e-08, + "loss": 0.689, + "step": 50579 + }, + { + "epoch": 1.85, + "learning_rate": 1.4955872120701653e-08, + "loss": 0.6985, + "step": 50580 + }, + { + "epoch": 1.85, + "learning_rate": 1.4952210056029586e-08, + "loss": 0.6934, + "step": 50581 + }, + { + "epoch": 1.85, + "learning_rate": 1.4948547991357526e-08, + "loss": 0.698, + "step": 50582 + }, + { + "epoch": 1.85, + "learning_rate": 1.4944885926685463e-08, + "loss": 0.7039, + "step": 50583 + }, + { + "epoch": 1.85, + "learning_rate": 1.4941223862013403e-08, + "loss": 0.7063, + "step": 50584 + }, + { + "epoch": 1.85, + "learning_rate": 1.4937561797341343e-08, + "loss": 0.7012, + "step": 50585 + }, + { + "epoch": 1.85, + "learning_rate": 1.493389973266928e-08, + "loss": 0.6868, + "step": 50586 + }, + { + "epoch": 1.85, + "learning_rate": 1.4930237667997216e-08, + "loss": 0.6923, + "step": 50587 + }, + { + "epoch": 1.85, + "learning_rate": 1.4926575603325153e-08, + "loss": 0.7032, + "step": 50588 + }, + { + "epoch": 1.85, + "learning_rate": 1.4922913538653092e-08, + "loss": 0.6939, + "step": 50589 + }, + { + "epoch": 1.85, + "learning_rate": 1.491925147398103e-08, + "loss": 0.6869, + "step": 50590 + }, + { + "epoch": 1.85, + "learning_rate": 1.491558940930897e-08, + "loss": 0.6858, + "step": 50591 + }, + { + "epoch": 1.85, + "learning_rate": 1.4911927344636906e-08, + "loss": 0.6914, + "step": 50592 + }, + { + "epoch": 1.85, + "learning_rate": 1.4908265279964842e-08, + "loss": 0.688, + "step": 50593 + }, + { + "epoch": 1.85, + "learning_rate": 1.4904603215292782e-08, + "loss": 0.6749, + "step": 50594 + }, + { + "epoch": 1.85, + "learning_rate": 1.4900941150620719e-08, + "loss": 0.6931, + "step": 50595 + }, + { + "epoch": 1.85, + "learning_rate": 1.4897279085948659e-08, + "loss": 0.7003, + "step": 50596 + }, + { + "epoch": 1.85, + "learning_rate": 1.4893617021276594e-08, + "loss": 0.6888, + "step": 50597 + }, + { + "epoch": 1.85, + "learning_rate": 1.4889954956604534e-08, + "loss": 0.6987, + "step": 50598 + }, + { + "epoch": 1.85, + "learning_rate": 1.488629289193247e-08, + "loss": 0.7003, + "step": 50599 + }, + { + "epoch": 1.85, + "learning_rate": 1.4882630827260409e-08, + "loss": 0.6979, + "step": 50600 + }, + { + "epoch": 1.85, + "learning_rate": 1.4878968762588345e-08, + "loss": 0.6915, + "step": 50601 + }, + { + "epoch": 1.85, + "learning_rate": 1.4875306697916285e-08, + "loss": 0.7003, + "step": 50602 + }, + { + "epoch": 1.85, + "learning_rate": 1.4871644633244223e-08, + "loss": 0.6934, + "step": 50603 + }, + { + "epoch": 1.85, + "learning_rate": 1.486798256857216e-08, + "loss": 0.6979, + "step": 50604 + }, + { + "epoch": 1.85, + "learning_rate": 1.4864320503900098e-08, + "loss": 0.6956, + "step": 50605 + }, + { + "epoch": 1.85, + "learning_rate": 1.4860658439228035e-08, + "loss": 0.6927, + "step": 50606 + }, + { + "epoch": 1.85, + "learning_rate": 1.4856996374555975e-08, + "loss": 0.6935, + "step": 50607 + }, + { + "epoch": 1.85, + "learning_rate": 1.4853334309883911e-08, + "loss": 0.7007, + "step": 50608 + }, + { + "epoch": 1.85, + "learning_rate": 1.484967224521185e-08, + "loss": 0.7026, + "step": 50609 + }, + { + "epoch": 1.85, + "learning_rate": 1.4846010180539786e-08, + "loss": 0.6922, + "step": 50610 + }, + { + "epoch": 1.85, + "learning_rate": 1.4842348115867726e-08, + "loss": 0.6896, + "step": 50611 + }, + { + "epoch": 1.85, + "learning_rate": 1.4838686051195665e-08, + "loss": 0.6936, + "step": 50612 + }, + { + "epoch": 1.85, + "learning_rate": 1.4835023986523601e-08, + "loss": 0.6904, + "step": 50613 + }, + { + "epoch": 1.85, + "learning_rate": 1.483136192185154e-08, + "loss": 0.6934, + "step": 50614 + }, + { + "epoch": 1.85, + "learning_rate": 1.4827699857179476e-08, + "loss": 0.6897, + "step": 50615 + }, + { + "epoch": 1.85, + "learning_rate": 1.4824037792507416e-08, + "loss": 0.6912, + "step": 50616 + }, + { + "epoch": 1.85, + "learning_rate": 1.4820375727835353e-08, + "loss": 0.6955, + "step": 50617 + }, + { + "epoch": 1.85, + "learning_rate": 1.4816713663163291e-08, + "loss": 0.6974, + "step": 50618 + }, + { + "epoch": 1.85, + "learning_rate": 1.4813051598491227e-08, + "loss": 0.695, + "step": 50619 + }, + { + "epoch": 1.85, + "learning_rate": 1.4809389533819166e-08, + "loss": 0.6929, + "step": 50620 + }, + { + "epoch": 1.85, + "learning_rate": 1.4805727469147106e-08, + "loss": 0.698, + "step": 50621 + }, + { + "epoch": 1.85, + "learning_rate": 1.4802065404475042e-08, + "loss": 0.6908, + "step": 50622 + }, + { + "epoch": 1.85, + "learning_rate": 1.479840333980298e-08, + "loss": 0.6836, + "step": 50623 + }, + { + "epoch": 1.85, + "learning_rate": 1.4794741275130917e-08, + "loss": 0.6903, + "step": 50624 + }, + { + "epoch": 1.85, + "learning_rate": 1.4791079210458857e-08, + "loss": 0.6934, + "step": 50625 + }, + { + "epoch": 1.85, + "learning_rate": 1.4787417145786794e-08, + "loss": 0.6975, + "step": 50626 + }, + { + "epoch": 1.85, + "learning_rate": 1.4783755081114732e-08, + "loss": 0.6997, + "step": 50627 + }, + { + "epoch": 1.85, + "learning_rate": 1.4780093016442669e-08, + "loss": 0.6917, + "step": 50628 + }, + { + "epoch": 1.85, + "learning_rate": 1.4776430951770607e-08, + "loss": 0.7047, + "step": 50629 + }, + { + "epoch": 1.85, + "learning_rate": 1.4772768887098547e-08, + "loss": 0.6944, + "step": 50630 + }, + { + "epoch": 1.85, + "learning_rate": 1.4769106822426483e-08, + "loss": 0.6885, + "step": 50631 + }, + { + "epoch": 1.85, + "learning_rate": 1.4765444757754422e-08, + "loss": 0.6949, + "step": 50632 + }, + { + "epoch": 1.85, + "learning_rate": 1.4761782693082358e-08, + "loss": 0.7163, + "step": 50633 + }, + { + "epoch": 1.85, + "learning_rate": 1.4758120628410298e-08, + "loss": 0.6943, + "step": 50634 + }, + { + "epoch": 1.85, + "learning_rate": 1.4754458563738233e-08, + "loss": 0.684, + "step": 50635 + }, + { + "epoch": 1.85, + "learning_rate": 1.4750796499066173e-08, + "loss": 0.6897, + "step": 50636 + }, + { + "epoch": 1.85, + "learning_rate": 1.474713443439411e-08, + "loss": 0.6856, + "step": 50637 + }, + { + "epoch": 1.85, + "learning_rate": 1.4743472369722048e-08, + "loss": 0.7046, + "step": 50638 + }, + { + "epoch": 1.85, + "learning_rate": 1.4739810305049988e-08, + "loss": 0.6851, + "step": 50639 + }, + { + "epoch": 1.85, + "learning_rate": 1.4736148240377925e-08, + "loss": 0.685, + "step": 50640 + }, + { + "epoch": 1.85, + "learning_rate": 1.4732486175705863e-08, + "loss": 0.6872, + "step": 50641 + }, + { + "epoch": 1.85, + "learning_rate": 1.47288241110338e-08, + "loss": 0.6894, + "step": 50642 + }, + { + "epoch": 1.85, + "learning_rate": 1.472516204636174e-08, + "loss": 0.7014, + "step": 50643 + }, + { + "epoch": 1.85, + "learning_rate": 1.4721499981689674e-08, + "loss": 0.6957, + "step": 50644 + }, + { + "epoch": 1.85, + "learning_rate": 1.4717837917017614e-08, + "loss": 0.7008, + "step": 50645 + }, + { + "epoch": 1.85, + "learning_rate": 1.4714175852345551e-08, + "loss": 0.6936, + "step": 50646 + }, + { + "epoch": 1.85, + "learning_rate": 1.471051378767349e-08, + "loss": 0.6822, + "step": 50647 + }, + { + "epoch": 1.85, + "learning_rate": 1.4706851723001429e-08, + "loss": 0.699, + "step": 50648 + }, + { + "epoch": 1.85, + "learning_rate": 1.4703189658329366e-08, + "loss": 0.6941, + "step": 50649 + }, + { + "epoch": 1.85, + "learning_rate": 1.4699527593657304e-08, + "loss": 0.7144, + "step": 50650 + }, + { + "epoch": 1.85, + "learning_rate": 1.469586552898524e-08, + "loss": 0.6909, + "step": 50651 + }, + { + "epoch": 1.85, + "learning_rate": 1.469220346431318e-08, + "loss": 0.6999, + "step": 50652 + }, + { + "epoch": 1.85, + "learning_rate": 1.4688541399641116e-08, + "loss": 0.6984, + "step": 50653 + }, + { + "epoch": 1.85, + "learning_rate": 1.4684879334969055e-08, + "loss": 0.6825, + "step": 50654 + }, + { + "epoch": 1.85, + "learning_rate": 1.4681217270296992e-08, + "loss": 0.6961, + "step": 50655 + }, + { + "epoch": 1.85, + "learning_rate": 1.467755520562493e-08, + "loss": 0.6839, + "step": 50656 + }, + { + "epoch": 1.85, + "learning_rate": 1.467389314095287e-08, + "loss": 0.6974, + "step": 50657 + }, + { + "epoch": 1.85, + "learning_rate": 1.4670231076280807e-08, + "loss": 0.6692, + "step": 50658 + }, + { + "epoch": 1.85, + "learning_rate": 1.4666569011608745e-08, + "loss": 0.697, + "step": 50659 + }, + { + "epoch": 1.85, + "learning_rate": 1.4662906946936682e-08, + "loss": 0.6908, + "step": 50660 + }, + { + "epoch": 1.85, + "learning_rate": 1.465924488226462e-08, + "loss": 0.6982, + "step": 50661 + }, + { + "epoch": 1.85, + "learning_rate": 1.4655582817592557e-08, + "loss": 0.6993, + "step": 50662 + }, + { + "epoch": 1.85, + "learning_rate": 1.4651920752920497e-08, + "loss": 0.6978, + "step": 50663 + }, + { + "epoch": 1.85, + "learning_rate": 1.4648258688248433e-08, + "loss": 0.7023, + "step": 50664 + }, + { + "epoch": 1.85, + "learning_rate": 1.4644596623576372e-08, + "loss": 0.6905, + "step": 50665 + }, + { + "epoch": 1.85, + "learning_rate": 1.4640934558904311e-08, + "loss": 0.6885, + "step": 50666 + }, + { + "epoch": 1.85, + "learning_rate": 1.4637272494232248e-08, + "loss": 0.6996, + "step": 50667 + }, + { + "epoch": 1.85, + "learning_rate": 1.4633610429560186e-08, + "loss": 0.6994, + "step": 50668 + }, + { + "epoch": 1.85, + "learning_rate": 1.4629948364888123e-08, + "loss": 0.7107, + "step": 50669 + }, + { + "epoch": 1.85, + "learning_rate": 1.4626286300216061e-08, + "loss": 0.6988, + "step": 50670 + }, + { + "epoch": 1.85, + "learning_rate": 1.4622624235543998e-08, + "loss": 0.6983, + "step": 50671 + }, + { + "epoch": 1.85, + "learning_rate": 1.4618962170871938e-08, + "loss": 0.6912, + "step": 50672 + }, + { + "epoch": 1.85, + "learning_rate": 1.4615300106199874e-08, + "loss": 0.6989, + "step": 50673 + }, + { + "epoch": 1.85, + "learning_rate": 1.4611638041527813e-08, + "loss": 0.6894, + "step": 50674 + }, + { + "epoch": 1.85, + "learning_rate": 1.460797597685575e-08, + "loss": 0.6884, + "step": 50675 + }, + { + "epoch": 1.85, + "learning_rate": 1.4604313912183688e-08, + "loss": 0.694, + "step": 50676 + }, + { + "epoch": 1.85, + "learning_rate": 1.4600651847511628e-08, + "loss": 0.6928, + "step": 50677 + }, + { + "epoch": 1.85, + "learning_rate": 1.4596989782839564e-08, + "loss": 0.6934, + "step": 50678 + }, + { + "epoch": 1.85, + "learning_rate": 1.4593327718167502e-08, + "loss": 0.6996, + "step": 50679 + }, + { + "epoch": 1.85, + "learning_rate": 1.4589665653495439e-08, + "loss": 0.6925, + "step": 50680 + }, + { + "epoch": 1.85, + "learning_rate": 1.4586003588823379e-08, + "loss": 0.7001, + "step": 50681 + }, + { + "epoch": 1.85, + "learning_rate": 1.4582341524151316e-08, + "loss": 0.6913, + "step": 50682 + }, + { + "epoch": 1.85, + "learning_rate": 1.4578679459479254e-08, + "loss": 0.6852, + "step": 50683 + }, + { + "epoch": 1.85, + "learning_rate": 1.457501739480719e-08, + "loss": 0.6917, + "step": 50684 + }, + { + "epoch": 1.85, + "learning_rate": 1.4571355330135129e-08, + "loss": 0.6981, + "step": 50685 + }, + { + "epoch": 1.85, + "learning_rate": 1.4567693265463069e-08, + "loss": 0.6921, + "step": 50686 + }, + { + "epoch": 1.85, + "learning_rate": 1.4564031200791005e-08, + "loss": 0.6946, + "step": 50687 + }, + { + "epoch": 1.85, + "learning_rate": 1.4560369136118944e-08, + "loss": 0.6855, + "step": 50688 + }, + { + "epoch": 1.85, + "learning_rate": 1.455670707144688e-08, + "loss": 0.6991, + "step": 50689 + }, + { + "epoch": 1.85, + "learning_rate": 1.455304500677482e-08, + "loss": 0.6973, + "step": 50690 + }, + { + "epoch": 1.85, + "learning_rate": 1.4549382942102755e-08, + "loss": 0.693, + "step": 50691 + }, + { + "epoch": 1.85, + "learning_rate": 1.4545720877430695e-08, + "loss": 0.6939, + "step": 50692 + }, + { + "epoch": 1.85, + "learning_rate": 1.4542058812758632e-08, + "loss": 0.6953, + "step": 50693 + }, + { + "epoch": 1.85, + "learning_rate": 1.453839674808657e-08, + "loss": 0.6912, + "step": 50694 + }, + { + "epoch": 1.85, + "learning_rate": 1.453473468341451e-08, + "loss": 0.708, + "step": 50695 + }, + { + "epoch": 1.85, + "learning_rate": 1.4531072618742446e-08, + "loss": 0.7035, + "step": 50696 + }, + { + "epoch": 1.85, + "learning_rate": 1.4527410554070385e-08, + "loss": 0.6943, + "step": 50697 + }, + { + "epoch": 1.85, + "learning_rate": 1.4523748489398321e-08, + "loss": 0.6818, + "step": 50698 + }, + { + "epoch": 1.85, + "learning_rate": 1.4520086424726261e-08, + "loss": 0.6865, + "step": 50699 + }, + { + "epoch": 1.85, + "learning_rate": 1.4516424360054196e-08, + "loss": 0.6894, + "step": 50700 + }, + { + "epoch": 1.86, + "learning_rate": 1.4512762295382136e-08, + "loss": 0.6911, + "step": 50701 + }, + { + "epoch": 1.86, + "learning_rate": 1.4509100230710073e-08, + "loss": 0.6928, + "step": 50702 + }, + { + "epoch": 1.86, + "learning_rate": 1.4505438166038011e-08, + "loss": 0.695, + "step": 50703 + }, + { + "epoch": 1.86, + "learning_rate": 1.4501776101365951e-08, + "loss": 0.6929, + "step": 50704 + }, + { + "epoch": 1.86, + "learning_rate": 1.4498114036693888e-08, + "loss": 0.6949, + "step": 50705 + }, + { + "epoch": 1.86, + "learning_rate": 1.4494451972021826e-08, + "loss": 0.6924, + "step": 50706 + }, + { + "epoch": 1.86, + "learning_rate": 1.4490789907349763e-08, + "loss": 0.6993, + "step": 50707 + }, + { + "epoch": 1.86, + "learning_rate": 1.44871278426777e-08, + "loss": 0.6805, + "step": 50708 + }, + { + "epoch": 1.86, + "learning_rate": 1.4483465778005637e-08, + "loss": 0.6929, + "step": 50709 + }, + { + "epoch": 1.86, + "learning_rate": 1.4479803713333577e-08, + "loss": 0.6861, + "step": 50710 + }, + { + "epoch": 1.86, + "learning_rate": 1.4476141648661514e-08, + "loss": 0.6979, + "step": 50711 + }, + { + "epoch": 1.86, + "learning_rate": 1.4472479583989452e-08, + "loss": 0.6963, + "step": 50712 + }, + { + "epoch": 1.86, + "learning_rate": 1.4468817519317392e-08, + "loss": 0.697, + "step": 50713 + }, + { + "epoch": 1.86, + "learning_rate": 1.4465155454645329e-08, + "loss": 0.6928, + "step": 50714 + }, + { + "epoch": 1.86, + "learning_rate": 1.4461493389973267e-08, + "loss": 0.6953, + "step": 50715 + }, + { + "epoch": 1.86, + "learning_rate": 1.4457831325301204e-08, + "loss": 0.6863, + "step": 50716 + }, + { + "epoch": 1.86, + "learning_rate": 1.4454169260629142e-08, + "loss": 0.6835, + "step": 50717 + }, + { + "epoch": 1.86, + "learning_rate": 1.4450507195957079e-08, + "loss": 0.6954, + "step": 50718 + }, + { + "epoch": 1.86, + "learning_rate": 1.4446845131285018e-08, + "loss": 0.6961, + "step": 50719 + }, + { + "epoch": 1.86, + "learning_rate": 1.4443183066612955e-08, + "loss": 0.6998, + "step": 50720 + }, + { + "epoch": 1.86, + "learning_rate": 1.4439521001940893e-08, + "loss": 0.6878, + "step": 50721 + }, + { + "epoch": 1.86, + "learning_rate": 1.4435858937268833e-08, + "loss": 0.6908, + "step": 50722 + }, + { + "epoch": 1.86, + "learning_rate": 1.4432196872596768e-08, + "loss": 0.6931, + "step": 50723 + }, + { + "epoch": 1.86, + "learning_rate": 1.4428534807924708e-08, + "loss": 0.6901, + "step": 50724 + }, + { + "epoch": 1.86, + "learning_rate": 1.4424872743252645e-08, + "loss": 0.6896, + "step": 50725 + }, + { + "epoch": 1.86, + "learning_rate": 1.4421210678580583e-08, + "loss": 0.6867, + "step": 50726 + }, + { + "epoch": 1.86, + "learning_rate": 1.441754861390852e-08, + "loss": 0.6841, + "step": 50727 + }, + { + "epoch": 1.86, + "learning_rate": 1.441388654923646e-08, + "loss": 0.6835, + "step": 50728 + }, + { + "epoch": 1.86, + "learning_rate": 1.4410224484564396e-08, + "loss": 0.6952, + "step": 50729 + }, + { + "epoch": 1.86, + "learning_rate": 1.4406562419892335e-08, + "loss": 0.6974, + "step": 50730 + }, + { + "epoch": 1.86, + "learning_rate": 1.4402900355220274e-08, + "loss": 0.6992, + "step": 50731 + }, + { + "epoch": 1.86, + "learning_rate": 1.439923829054821e-08, + "loss": 0.6951, + "step": 50732 + }, + { + "epoch": 1.86, + "learning_rate": 1.439557622587615e-08, + "loss": 0.6946, + "step": 50733 + }, + { + "epoch": 1.86, + "learning_rate": 1.4391914161204086e-08, + "loss": 0.6873, + "step": 50734 + }, + { + "epoch": 1.86, + "learning_rate": 1.4388252096532024e-08, + "loss": 0.6991, + "step": 50735 + }, + { + "epoch": 1.86, + "learning_rate": 1.4384590031859961e-08, + "loss": 0.6934, + "step": 50736 + }, + { + "epoch": 1.86, + "learning_rate": 1.43809279671879e-08, + "loss": 0.6885, + "step": 50737 + }, + { + "epoch": 1.86, + "learning_rate": 1.4377265902515836e-08, + "loss": 0.7083, + "step": 50738 + }, + { + "epoch": 1.86, + "learning_rate": 1.4373603837843776e-08, + "loss": 0.6849, + "step": 50739 + }, + { + "epoch": 1.86, + "learning_rate": 1.4369941773171716e-08, + "loss": 0.7023, + "step": 50740 + }, + { + "epoch": 1.86, + "learning_rate": 1.436627970849965e-08, + "loss": 0.6864, + "step": 50741 + }, + { + "epoch": 1.86, + "learning_rate": 1.436261764382759e-08, + "loss": 0.6945, + "step": 50742 + }, + { + "epoch": 1.86, + "learning_rate": 1.4358955579155527e-08, + "loss": 0.695, + "step": 50743 + }, + { + "epoch": 1.86, + "learning_rate": 1.4355293514483465e-08, + "loss": 0.6918, + "step": 50744 + }, + { + "epoch": 1.86, + "learning_rate": 1.4351631449811402e-08, + "loss": 0.6869, + "step": 50745 + }, + { + "epoch": 1.86, + "learning_rate": 1.4347969385139342e-08, + "loss": 0.6939, + "step": 50746 + }, + { + "epoch": 1.86, + "learning_rate": 1.4344307320467277e-08, + "loss": 0.6907, + "step": 50747 + }, + { + "epoch": 1.86, + "learning_rate": 1.4340645255795217e-08, + "loss": 0.7019, + "step": 50748 + }, + { + "epoch": 1.86, + "learning_rate": 1.4336983191123155e-08, + "loss": 0.6999, + "step": 50749 + }, + { + "epoch": 1.86, + "learning_rate": 1.4333321126451092e-08, + "loss": 0.6805, + "step": 50750 + }, + { + "epoch": 1.86, + "learning_rate": 1.4329659061779032e-08, + "loss": 0.7074, + "step": 50751 + }, + { + "epoch": 1.86, + "learning_rate": 1.4325996997106968e-08, + "loss": 0.6904, + "step": 50752 + }, + { + "epoch": 1.86, + "learning_rate": 1.4322334932434907e-08, + "loss": 0.6857, + "step": 50753 + }, + { + "epoch": 1.86, + "learning_rate": 1.4318672867762843e-08, + "loss": 0.6842, + "step": 50754 + }, + { + "epoch": 1.86, + "learning_rate": 1.4315010803090783e-08, + "loss": 0.6921, + "step": 50755 + }, + { + "epoch": 1.86, + "learning_rate": 1.4311348738418718e-08, + "loss": 0.694, + "step": 50756 + }, + { + "epoch": 1.86, + "learning_rate": 1.4307686673746658e-08, + "loss": 0.6904, + "step": 50757 + }, + { + "epoch": 1.86, + "learning_rate": 1.4304024609074596e-08, + "loss": 0.691, + "step": 50758 + }, + { + "epoch": 1.86, + "learning_rate": 1.4300362544402533e-08, + "loss": 0.6991, + "step": 50759 + }, + { + "epoch": 1.86, + "learning_rate": 1.4296700479730473e-08, + "loss": 0.7025, + "step": 50760 + }, + { + "epoch": 1.86, + "learning_rate": 1.429303841505841e-08, + "loss": 0.6939, + "step": 50761 + }, + { + "epoch": 1.86, + "learning_rate": 1.4289376350386348e-08, + "loss": 0.693, + "step": 50762 + }, + { + "epoch": 1.86, + "learning_rate": 1.4285714285714284e-08, + "loss": 0.6941, + "step": 50763 + }, + { + "epoch": 1.86, + "learning_rate": 1.4282052221042223e-08, + "loss": 0.6979, + "step": 50764 + }, + { + "epoch": 1.86, + "learning_rate": 1.427839015637016e-08, + "loss": 0.6994, + "step": 50765 + }, + { + "epoch": 1.86, + "learning_rate": 1.42747280916981e-08, + "loss": 0.6809, + "step": 50766 + }, + { + "epoch": 1.86, + "learning_rate": 1.4271066027026037e-08, + "loss": 0.7008, + "step": 50767 + }, + { + "epoch": 1.86, + "learning_rate": 1.4267403962353974e-08, + "loss": 0.6919, + "step": 50768 + }, + { + "epoch": 1.86, + "learning_rate": 1.4263741897681914e-08, + "loss": 0.692, + "step": 50769 + }, + { + "epoch": 1.86, + "learning_rate": 1.426007983300985e-08, + "loss": 0.7008, + "step": 50770 + }, + { + "epoch": 1.86, + "learning_rate": 1.4256417768337789e-08, + "loss": 0.6903, + "step": 50771 + }, + { + "epoch": 1.86, + "learning_rate": 1.4252755703665726e-08, + "loss": 0.6875, + "step": 50772 + }, + { + "epoch": 1.86, + "learning_rate": 1.4249093638993664e-08, + "loss": 0.7017, + "step": 50773 + }, + { + "epoch": 1.86, + "learning_rate": 1.42454315743216e-08, + "loss": 0.6846, + "step": 50774 + }, + { + "epoch": 1.86, + "learning_rate": 1.424176950964954e-08, + "loss": 0.6882, + "step": 50775 + }, + { + "epoch": 1.86, + "learning_rate": 1.4238107444977477e-08, + "loss": 0.6984, + "step": 50776 + }, + { + "epoch": 1.86, + "learning_rate": 1.4234445380305415e-08, + "loss": 0.6771, + "step": 50777 + }, + { + "epoch": 1.86, + "learning_rate": 1.4230783315633355e-08, + "loss": 0.6937, + "step": 50778 + }, + { + "epoch": 1.86, + "learning_rate": 1.422712125096129e-08, + "loss": 0.691, + "step": 50779 + }, + { + "epoch": 1.86, + "learning_rate": 1.422345918628923e-08, + "loss": 0.6968, + "step": 50780 + }, + { + "epoch": 1.86, + "learning_rate": 1.4219797121617167e-08, + "loss": 0.6875, + "step": 50781 + }, + { + "epoch": 1.86, + "learning_rate": 1.4216135056945105e-08, + "loss": 0.6999, + "step": 50782 + }, + { + "epoch": 1.86, + "learning_rate": 1.4212472992273042e-08, + "loss": 0.6894, + "step": 50783 + }, + { + "epoch": 1.86, + "learning_rate": 1.4208810927600982e-08, + "loss": 0.6838, + "step": 50784 + }, + { + "epoch": 1.86, + "learning_rate": 1.4205148862928918e-08, + "loss": 0.6988, + "step": 50785 + }, + { + "epoch": 1.86, + "learning_rate": 1.4201486798256856e-08, + "loss": 0.703, + "step": 50786 + }, + { + "epoch": 1.86, + "learning_rate": 1.4197824733584796e-08, + "loss": 0.6901, + "step": 50787 + }, + { + "epoch": 1.86, + "learning_rate": 1.4194162668912731e-08, + "loss": 0.7042, + "step": 50788 + }, + { + "epoch": 1.86, + "learning_rate": 1.4190500604240671e-08, + "loss": 0.6981, + "step": 50789 + }, + { + "epoch": 1.86, + "learning_rate": 1.4186838539568608e-08, + "loss": 0.6867, + "step": 50790 + }, + { + "epoch": 1.86, + "learning_rate": 1.4183176474896546e-08, + "loss": 0.6898, + "step": 50791 + }, + { + "epoch": 1.86, + "learning_rate": 1.4179514410224483e-08, + "loss": 0.6848, + "step": 50792 + }, + { + "epoch": 1.86, + "learning_rate": 1.4175852345552423e-08, + "loss": 0.7028, + "step": 50793 + }, + { + "epoch": 1.86, + "learning_rate": 1.4172190280880358e-08, + "loss": 0.6917, + "step": 50794 + }, + { + "epoch": 1.86, + "learning_rate": 1.4168528216208298e-08, + "loss": 0.688, + "step": 50795 + }, + { + "epoch": 1.86, + "learning_rate": 1.4164866151536236e-08, + "loss": 0.6911, + "step": 50796 + }, + { + "epoch": 1.86, + "learning_rate": 1.4161204086864172e-08, + "loss": 0.7007, + "step": 50797 + }, + { + "epoch": 1.86, + "learning_rate": 1.4157542022192112e-08, + "loss": 0.6954, + "step": 50798 + }, + { + "epoch": 1.86, + "learning_rate": 1.4153879957520049e-08, + "loss": 0.7122, + "step": 50799 + }, + { + "epoch": 1.86, + "learning_rate": 1.4150217892847987e-08, + "loss": 0.6911, + "step": 50800 + }, + { + "epoch": 1.86, + "learning_rate": 1.4146555828175924e-08, + "loss": 0.695, + "step": 50801 + }, + { + "epoch": 1.86, + "learning_rate": 1.4142893763503864e-08, + "loss": 0.6839, + "step": 50802 + }, + { + "epoch": 1.86, + "learning_rate": 1.4139231698831799e-08, + "loss": 0.6963, + "step": 50803 + }, + { + "epoch": 1.86, + "learning_rate": 1.4135569634159739e-08, + "loss": 0.6896, + "step": 50804 + }, + { + "epoch": 1.86, + "learning_rate": 1.4131907569487677e-08, + "loss": 0.6951, + "step": 50805 + }, + { + "epoch": 1.86, + "learning_rate": 1.4128245504815614e-08, + "loss": 0.7163, + "step": 50806 + }, + { + "epoch": 1.86, + "learning_rate": 1.4124583440143554e-08, + "loss": 0.6907, + "step": 50807 + }, + { + "epoch": 1.86, + "learning_rate": 1.412092137547149e-08, + "loss": 0.69, + "step": 50808 + }, + { + "epoch": 1.86, + "learning_rate": 1.4117259310799428e-08, + "loss": 0.6943, + "step": 50809 + }, + { + "epoch": 1.86, + "learning_rate": 1.4113597246127365e-08, + "loss": 0.6915, + "step": 50810 + }, + { + "epoch": 1.86, + "learning_rate": 1.4109935181455303e-08, + "loss": 0.6987, + "step": 50811 + }, + { + "epoch": 1.86, + "learning_rate": 1.410627311678324e-08, + "loss": 0.6881, + "step": 50812 + }, + { + "epoch": 1.86, + "learning_rate": 1.410261105211118e-08, + "loss": 0.6843, + "step": 50813 + }, + { + "epoch": 1.86, + "learning_rate": 1.4098948987439118e-08, + "loss": 0.7087, + "step": 50814 + }, + { + "epoch": 1.86, + "learning_rate": 1.4095286922767055e-08, + "loss": 0.6827, + "step": 50815 + }, + { + "epoch": 1.86, + "learning_rate": 1.4091624858094995e-08, + "loss": 0.6976, + "step": 50816 + }, + { + "epoch": 1.86, + "learning_rate": 1.4087962793422931e-08, + "loss": 0.6912, + "step": 50817 + }, + { + "epoch": 1.86, + "learning_rate": 1.408430072875087e-08, + "loss": 0.6907, + "step": 50818 + }, + { + "epoch": 1.86, + "learning_rate": 1.4080638664078806e-08, + "loss": 0.6907, + "step": 50819 + }, + { + "epoch": 1.86, + "learning_rate": 1.4076976599406744e-08, + "loss": 0.6823, + "step": 50820 + }, + { + "epoch": 1.86, + "learning_rate": 1.4073314534734681e-08, + "loss": 0.6939, + "step": 50821 + }, + { + "epoch": 1.86, + "learning_rate": 1.4069652470062621e-08, + "loss": 0.6964, + "step": 50822 + }, + { + "epoch": 1.86, + "learning_rate": 1.406599040539056e-08, + "loss": 0.6902, + "step": 50823 + }, + { + "epoch": 1.86, + "learning_rate": 1.4062328340718496e-08, + "loss": 0.6994, + "step": 50824 + }, + { + "epoch": 1.86, + "learning_rate": 1.4058666276046436e-08, + "loss": 0.6867, + "step": 50825 + }, + { + "epoch": 1.86, + "learning_rate": 1.405500421137437e-08, + "loss": 0.6866, + "step": 50826 + }, + { + "epoch": 1.86, + "learning_rate": 1.405134214670231e-08, + "loss": 0.6974, + "step": 50827 + }, + { + "epoch": 1.86, + "learning_rate": 1.4047680082030247e-08, + "loss": 0.692, + "step": 50828 + }, + { + "epoch": 1.86, + "learning_rate": 1.4044018017358186e-08, + "loss": 0.7022, + "step": 50829 + }, + { + "epoch": 1.86, + "learning_rate": 1.4040355952686122e-08, + "loss": 0.6929, + "step": 50830 + }, + { + "epoch": 1.86, + "learning_rate": 1.4036693888014062e-08, + "loss": 0.6903, + "step": 50831 + }, + { + "epoch": 1.86, + "learning_rate": 1.4033031823342e-08, + "loss": 0.6947, + "step": 50832 + }, + { + "epoch": 1.86, + "learning_rate": 1.4029369758669937e-08, + "loss": 0.6865, + "step": 50833 + }, + { + "epoch": 1.86, + "learning_rate": 1.4025707693997877e-08, + "loss": 0.6924, + "step": 50834 + }, + { + "epoch": 1.86, + "learning_rate": 1.4022045629325812e-08, + "loss": 0.7018, + "step": 50835 + }, + { + "epoch": 1.86, + "learning_rate": 1.4018383564653752e-08, + "loss": 0.6913, + "step": 50836 + }, + { + "epoch": 1.86, + "learning_rate": 1.4014721499981689e-08, + "loss": 0.7113, + "step": 50837 + }, + { + "epoch": 1.86, + "learning_rate": 1.4011059435309627e-08, + "loss": 0.6889, + "step": 50838 + }, + { + "epoch": 1.86, + "learning_rate": 1.4007397370637563e-08, + "loss": 0.6784, + "step": 50839 + }, + { + "epoch": 1.86, + "learning_rate": 1.4003735305965503e-08, + "loss": 0.6868, + "step": 50840 + }, + { + "epoch": 1.86, + "learning_rate": 1.4000073241293442e-08, + "loss": 0.676, + "step": 50841 + }, + { + "epoch": 1.86, + "learning_rate": 1.3996411176621378e-08, + "loss": 0.6852, + "step": 50842 + }, + { + "epoch": 1.86, + "learning_rate": 1.3992749111949318e-08, + "loss": 0.6895, + "step": 50843 + }, + { + "epoch": 1.86, + "learning_rate": 1.3989087047277253e-08, + "loss": 0.6978, + "step": 50844 + }, + { + "epoch": 1.86, + "learning_rate": 1.3985424982605193e-08, + "loss": 0.6922, + "step": 50845 + }, + { + "epoch": 1.86, + "learning_rate": 1.398176291793313e-08, + "loss": 0.6849, + "step": 50846 + }, + { + "epoch": 1.86, + "learning_rate": 1.3978100853261068e-08, + "loss": 0.6894, + "step": 50847 + }, + { + "epoch": 1.86, + "learning_rate": 1.3974438788589005e-08, + "loss": 0.6897, + "step": 50848 + }, + { + "epoch": 1.86, + "learning_rate": 1.3970776723916945e-08, + "loss": 0.6951, + "step": 50849 + }, + { + "epoch": 1.86, + "learning_rate": 1.3967114659244883e-08, + "loss": 0.6817, + "step": 50850 + }, + { + "epoch": 1.86, + "learning_rate": 1.396345259457282e-08, + "loss": 0.6935, + "step": 50851 + }, + { + "epoch": 1.86, + "learning_rate": 1.3959790529900758e-08, + "loss": 0.6892, + "step": 50852 + }, + { + "epoch": 1.86, + "learning_rate": 1.3956128465228694e-08, + "loss": 0.6914, + "step": 50853 + }, + { + "epoch": 1.86, + "learning_rate": 1.3952466400556634e-08, + "loss": 0.6885, + "step": 50854 + }, + { + "epoch": 1.86, + "learning_rate": 1.3948804335884571e-08, + "loss": 0.6895, + "step": 50855 + }, + { + "epoch": 1.86, + "learning_rate": 1.3945142271212509e-08, + "loss": 0.7017, + "step": 50856 + }, + { + "epoch": 1.86, + "learning_rate": 1.3941480206540446e-08, + "loss": 0.6854, + "step": 50857 + }, + { + "epoch": 1.86, + "learning_rate": 1.3937818141868386e-08, + "loss": 0.6964, + "step": 50858 + }, + { + "epoch": 1.86, + "learning_rate": 1.3934156077196324e-08, + "loss": 0.6932, + "step": 50859 + }, + { + "epoch": 1.86, + "learning_rate": 1.393049401252426e-08, + "loss": 0.7004, + "step": 50860 + }, + { + "epoch": 1.86, + "learning_rate": 1.3926831947852199e-08, + "loss": 0.6783, + "step": 50861 + }, + { + "epoch": 1.86, + "learning_rate": 1.3923169883180135e-08, + "loss": 0.684, + "step": 50862 + }, + { + "epoch": 1.86, + "learning_rate": 1.3919507818508075e-08, + "loss": 0.6922, + "step": 50863 + }, + { + "epoch": 1.86, + "learning_rate": 1.3915845753836012e-08, + "loss": 0.6901, + "step": 50864 + }, + { + "epoch": 1.86, + "learning_rate": 1.391218368916395e-08, + "loss": 0.6902, + "step": 50865 + }, + { + "epoch": 1.86, + "learning_rate": 1.3908521624491887e-08, + "loss": 0.6907, + "step": 50866 + }, + { + "epoch": 1.86, + "learning_rate": 1.3904859559819825e-08, + "loss": 0.6946, + "step": 50867 + }, + { + "epoch": 1.86, + "learning_rate": 1.3901197495147762e-08, + "loss": 0.6945, + "step": 50868 + }, + { + "epoch": 1.86, + "learning_rate": 1.3897535430475702e-08, + "loss": 0.6877, + "step": 50869 + }, + { + "epoch": 1.86, + "learning_rate": 1.389387336580364e-08, + "loss": 0.6816, + "step": 50870 + }, + { + "epoch": 1.86, + "learning_rate": 1.3890211301131577e-08, + "loss": 0.6947, + "step": 50871 + }, + { + "epoch": 1.86, + "learning_rate": 1.3886549236459517e-08, + "loss": 0.6884, + "step": 50872 + }, + { + "epoch": 1.86, + "learning_rate": 1.3882887171787453e-08, + "loss": 0.6871, + "step": 50873 + }, + { + "epoch": 1.86, + "learning_rate": 1.3879225107115391e-08, + "loss": 0.6914, + "step": 50874 + }, + { + "epoch": 1.86, + "learning_rate": 1.3875563042443328e-08, + "loss": 0.6881, + "step": 50875 + }, + { + "epoch": 1.86, + "learning_rate": 1.3871900977771266e-08, + "loss": 0.6837, + "step": 50876 + }, + { + "epoch": 1.86, + "learning_rate": 1.3868238913099203e-08, + "loss": 0.6949, + "step": 50877 + }, + { + "epoch": 1.86, + "learning_rate": 1.3864576848427143e-08, + "loss": 0.6974, + "step": 50878 + }, + { + "epoch": 1.86, + "learning_rate": 1.3860914783755081e-08, + "loss": 0.6809, + "step": 50879 + }, + { + "epoch": 1.86, + "learning_rate": 1.3857252719083018e-08, + "loss": 0.695, + "step": 50880 + }, + { + "epoch": 1.86, + "learning_rate": 1.3853590654410958e-08, + "loss": 0.6984, + "step": 50881 + }, + { + "epoch": 1.86, + "learning_rate": 1.3849928589738893e-08, + "loss": 0.6877, + "step": 50882 + }, + { + "epoch": 1.86, + "learning_rate": 1.3846266525066833e-08, + "loss": 0.6868, + "step": 50883 + }, + { + "epoch": 1.86, + "learning_rate": 1.384260446039477e-08, + "loss": 0.6884, + "step": 50884 + }, + { + "epoch": 1.86, + "learning_rate": 1.3838942395722707e-08, + "loss": 0.6926, + "step": 50885 + }, + { + "epoch": 1.86, + "learning_rate": 1.3835280331050644e-08, + "loss": 0.6894, + "step": 50886 + }, + { + "epoch": 1.86, + "learning_rate": 1.3831618266378584e-08, + "loss": 0.6906, + "step": 50887 + }, + { + "epoch": 1.86, + "learning_rate": 1.3827956201706522e-08, + "loss": 0.6917, + "step": 50888 + }, + { + "epoch": 1.86, + "learning_rate": 1.3824294137034459e-08, + "loss": 0.6966, + "step": 50889 + }, + { + "epoch": 1.86, + "learning_rate": 1.3820632072362399e-08, + "loss": 0.6936, + "step": 50890 + }, + { + "epoch": 1.86, + "learning_rate": 1.3816970007690334e-08, + "loss": 0.7022, + "step": 50891 + }, + { + "epoch": 1.86, + "learning_rate": 1.3813307943018274e-08, + "loss": 0.6854, + "step": 50892 + }, + { + "epoch": 1.86, + "learning_rate": 1.380964587834621e-08, + "loss": 0.7082, + "step": 50893 + }, + { + "epoch": 1.86, + "learning_rate": 1.3805983813674149e-08, + "loss": 0.6967, + "step": 50894 + }, + { + "epoch": 1.86, + "learning_rate": 1.3802321749002085e-08, + "loss": 0.6789, + "step": 50895 + }, + { + "epoch": 1.86, + "learning_rate": 1.3798659684330025e-08, + "loss": 0.6916, + "step": 50896 + }, + { + "epoch": 1.86, + "learning_rate": 1.3794997619657963e-08, + "loss": 0.7056, + "step": 50897 + }, + { + "epoch": 1.86, + "learning_rate": 1.37913355549859e-08, + "loss": 0.6902, + "step": 50898 + }, + { + "epoch": 1.86, + "learning_rate": 1.3787673490313838e-08, + "loss": 0.7077, + "step": 50899 + }, + { + "epoch": 1.86, + "learning_rate": 1.3784011425641775e-08, + "loss": 0.6867, + "step": 50900 + }, + { + "epoch": 1.86, + "learning_rate": 1.3780349360969715e-08, + "loss": 0.6854, + "step": 50901 + }, + { + "epoch": 1.86, + "learning_rate": 1.3776687296297652e-08, + "loss": 0.6973, + "step": 50902 + }, + { + "epoch": 1.86, + "learning_rate": 1.377302523162559e-08, + "loss": 0.7018, + "step": 50903 + }, + { + "epoch": 1.86, + "learning_rate": 1.3769363166953526e-08, + "loss": 0.6829, + "step": 50904 + }, + { + "epoch": 1.86, + "learning_rate": 1.3765701102281466e-08, + "loss": 0.6819, + "step": 50905 + }, + { + "epoch": 1.86, + "learning_rate": 1.3762039037609405e-08, + "loss": 0.6886, + "step": 50906 + }, + { + "epoch": 1.86, + "learning_rate": 1.3758376972937341e-08, + "loss": 0.7008, + "step": 50907 + }, + { + "epoch": 1.86, + "learning_rate": 1.375471490826528e-08, + "loss": 0.7052, + "step": 50908 + }, + { + "epoch": 1.86, + "learning_rate": 1.3751052843593216e-08, + "loss": 0.6924, + "step": 50909 + }, + { + "epoch": 1.86, + "learning_rate": 1.3747390778921156e-08, + "loss": 0.6989, + "step": 50910 + }, + { + "epoch": 1.86, + "learning_rate": 1.3743728714249093e-08, + "loss": 0.6999, + "step": 50911 + }, + { + "epoch": 1.86, + "learning_rate": 1.3740066649577031e-08, + "loss": 0.7029, + "step": 50912 + }, + { + "epoch": 1.86, + "learning_rate": 1.3736404584904968e-08, + "loss": 0.6908, + "step": 50913 + }, + { + "epoch": 1.86, + "learning_rate": 1.3732742520232906e-08, + "loss": 0.6916, + "step": 50914 + }, + { + "epoch": 1.86, + "learning_rate": 1.3729080455560846e-08, + "loss": 0.6782, + "step": 50915 + }, + { + "epoch": 1.86, + "learning_rate": 1.3725418390888782e-08, + "loss": 0.6965, + "step": 50916 + }, + { + "epoch": 1.86, + "learning_rate": 1.372175632621672e-08, + "loss": 0.7015, + "step": 50917 + }, + { + "epoch": 1.86, + "learning_rate": 1.3718094261544657e-08, + "loss": 0.6945, + "step": 50918 + }, + { + "epoch": 1.86, + "learning_rate": 1.3714432196872597e-08, + "loss": 0.6801, + "step": 50919 + }, + { + "epoch": 1.86, + "learning_rate": 1.3710770132200534e-08, + "loss": 0.7009, + "step": 50920 + }, + { + "epoch": 1.86, + "learning_rate": 1.3707108067528472e-08, + "loss": 0.69, + "step": 50921 + }, + { + "epoch": 1.86, + "learning_rate": 1.3703446002856409e-08, + "loss": 0.7114, + "step": 50922 + }, + { + "epoch": 1.86, + "learning_rate": 1.3699783938184347e-08, + "loss": 0.698, + "step": 50923 + }, + { + "epoch": 1.86, + "learning_rate": 1.3696121873512287e-08, + "loss": 0.6843, + "step": 50924 + }, + { + "epoch": 1.86, + "learning_rate": 1.3692459808840224e-08, + "loss": 0.6783, + "step": 50925 + }, + { + "epoch": 1.86, + "learning_rate": 1.3688797744168162e-08, + "loss": 0.6995, + "step": 50926 + }, + { + "epoch": 1.86, + "learning_rate": 1.3685135679496098e-08, + "loss": 0.6972, + "step": 50927 + }, + { + "epoch": 1.86, + "learning_rate": 1.3681473614824038e-08, + "loss": 0.6994, + "step": 50928 + }, + { + "epoch": 1.86, + "learning_rate": 1.3677811550151973e-08, + "loss": 0.6995, + "step": 50929 + }, + { + "epoch": 1.86, + "learning_rate": 1.3674149485479913e-08, + "loss": 0.6816, + "step": 50930 + }, + { + "epoch": 1.86, + "learning_rate": 1.367048742080785e-08, + "loss": 0.697, + "step": 50931 + }, + { + "epoch": 1.86, + "learning_rate": 1.3666825356135788e-08, + "loss": 0.6869, + "step": 50932 + }, + { + "epoch": 1.86, + "learning_rate": 1.3663163291463728e-08, + "loss": 0.6801, + "step": 50933 + }, + { + "epoch": 1.86, + "learning_rate": 1.3659501226791665e-08, + "loss": 0.6866, + "step": 50934 + }, + { + "epoch": 1.86, + "learning_rate": 1.3655839162119603e-08, + "loss": 0.6851, + "step": 50935 + }, + { + "epoch": 1.86, + "learning_rate": 1.365217709744754e-08, + "loss": 0.6874, + "step": 50936 + }, + { + "epoch": 1.86, + "learning_rate": 1.364851503277548e-08, + "loss": 0.6979, + "step": 50937 + }, + { + "epoch": 1.86, + "learning_rate": 1.3644852968103415e-08, + "loss": 0.6899, + "step": 50938 + }, + { + "epoch": 1.86, + "learning_rate": 1.3641190903431354e-08, + "loss": 0.7088, + "step": 50939 + }, + { + "epoch": 1.86, + "learning_rate": 1.3637528838759291e-08, + "loss": 0.6908, + "step": 50940 + }, + { + "epoch": 1.86, + "learning_rate": 1.363386677408723e-08, + "loss": 0.7031, + "step": 50941 + }, + { + "epoch": 1.86, + "learning_rate": 1.363020470941517e-08, + "loss": 0.6891, + "step": 50942 + }, + { + "epoch": 1.86, + "learning_rate": 1.3626542644743106e-08, + "loss": 0.6873, + "step": 50943 + }, + { + "epoch": 1.86, + "learning_rate": 1.3622880580071044e-08, + "loss": 0.6859, + "step": 50944 + }, + { + "epoch": 1.86, + "learning_rate": 1.361921851539898e-08, + "loss": 0.7007, + "step": 50945 + }, + { + "epoch": 1.86, + "learning_rate": 1.361555645072692e-08, + "loss": 0.7145, + "step": 50946 + }, + { + "epoch": 1.86, + "learning_rate": 1.3611894386054856e-08, + "loss": 0.6916, + "step": 50947 + }, + { + "epoch": 1.86, + "learning_rate": 1.3608232321382796e-08, + "loss": 0.6833, + "step": 50948 + }, + { + "epoch": 1.86, + "learning_rate": 1.3604570256710732e-08, + "loss": 0.6865, + "step": 50949 + }, + { + "epoch": 1.86, + "learning_rate": 1.360090819203867e-08, + "loss": 0.7083, + "step": 50950 + }, + { + "epoch": 1.86, + "learning_rate": 1.359724612736661e-08, + "loss": 0.6865, + "step": 50951 + }, + { + "epoch": 1.86, + "learning_rate": 1.3593584062694547e-08, + "loss": 0.6948, + "step": 50952 + }, + { + "epoch": 1.86, + "learning_rate": 1.3589921998022485e-08, + "loss": 0.6871, + "step": 50953 + }, + { + "epoch": 1.86, + "learning_rate": 1.3586259933350422e-08, + "loss": 0.6858, + "step": 50954 + }, + { + "epoch": 1.86, + "learning_rate": 1.358259786867836e-08, + "loss": 0.6897, + "step": 50955 + }, + { + "epoch": 1.86, + "learning_rate": 1.3578935804006297e-08, + "loss": 0.685, + "step": 50956 + }, + { + "epoch": 1.86, + "learning_rate": 1.3575273739334237e-08, + "loss": 0.6947, + "step": 50957 + }, + { + "epoch": 1.86, + "learning_rate": 1.3571611674662173e-08, + "loss": 0.6931, + "step": 50958 + }, + { + "epoch": 1.86, + "learning_rate": 1.3567949609990112e-08, + "loss": 0.7094, + "step": 50959 + }, + { + "epoch": 1.86, + "learning_rate": 1.3564287545318052e-08, + "loss": 0.6964, + "step": 50960 + }, + { + "epoch": 1.86, + "learning_rate": 1.3560625480645988e-08, + "loss": 0.6922, + "step": 50961 + }, + { + "epoch": 1.86, + "learning_rate": 1.3556963415973926e-08, + "loss": 0.6915, + "step": 50962 + }, + { + "epoch": 1.86, + "learning_rate": 1.3553301351301863e-08, + "loss": 0.6961, + "step": 50963 + }, + { + "epoch": 1.86, + "learning_rate": 1.3549639286629801e-08, + "loss": 0.6875, + "step": 50964 + }, + { + "epoch": 1.86, + "learning_rate": 1.3545977221957738e-08, + "loss": 0.7029, + "step": 50965 + }, + { + "epoch": 1.86, + "learning_rate": 1.3542315157285678e-08, + "loss": 0.6751, + "step": 50966 + }, + { + "epoch": 1.86, + "learning_rate": 1.3538653092613615e-08, + "loss": 0.6978, + "step": 50967 + }, + { + "epoch": 1.86, + "learning_rate": 1.3534991027941553e-08, + "loss": 0.697, + "step": 50968 + }, + { + "epoch": 1.86, + "learning_rate": 1.353132896326949e-08, + "loss": 0.6974, + "step": 50969 + }, + { + "epoch": 1.86, + "learning_rate": 1.3527666898597428e-08, + "loss": 0.6981, + "step": 50970 + }, + { + "epoch": 1.86, + "learning_rate": 1.3524004833925368e-08, + "loss": 0.6887, + "step": 50971 + }, + { + "epoch": 1.86, + "learning_rate": 1.3520342769253304e-08, + "loss": 0.6827, + "step": 50972 + }, + { + "epoch": 1.86, + "learning_rate": 1.3516680704581243e-08, + "loss": 0.687, + "step": 50973 + }, + { + "epoch": 1.86, + "learning_rate": 1.3513018639909179e-08, + "loss": 0.6883, + "step": 50974 + }, + { + "epoch": 1.87, + "learning_rate": 1.3509356575237119e-08, + "loss": 0.6846, + "step": 50975 + }, + { + "epoch": 1.87, + "learning_rate": 1.3505694510565056e-08, + "loss": 0.6778, + "step": 50976 + }, + { + "epoch": 1.87, + "learning_rate": 1.3502032445892994e-08, + "loss": 0.6919, + "step": 50977 + }, + { + "epoch": 1.87, + "learning_rate": 1.349837038122093e-08, + "loss": 0.688, + "step": 50978 + }, + { + "epoch": 1.87, + "learning_rate": 1.3494708316548869e-08, + "loss": 0.6971, + "step": 50979 + }, + { + "epoch": 1.87, + "learning_rate": 1.3491046251876809e-08, + "loss": 0.6932, + "step": 50980 + }, + { + "epoch": 1.87, + "learning_rate": 1.3487384187204745e-08, + "loss": 0.6907, + "step": 50981 + }, + { + "epoch": 1.87, + "learning_rate": 1.3483722122532684e-08, + "loss": 0.6927, + "step": 50982 + }, + { + "epoch": 1.87, + "learning_rate": 1.348006005786062e-08, + "loss": 0.7007, + "step": 50983 + }, + { + "epoch": 1.87, + "learning_rate": 1.347639799318856e-08, + "loss": 0.6918, + "step": 50984 + }, + { + "epoch": 1.87, + "learning_rate": 1.3472735928516495e-08, + "loss": 0.7013, + "step": 50985 + }, + { + "epoch": 1.87, + "learning_rate": 1.3469073863844435e-08, + "loss": 0.6874, + "step": 50986 + }, + { + "epoch": 1.87, + "learning_rate": 1.3465411799172372e-08, + "loss": 0.6769, + "step": 50987 + }, + { + "epoch": 1.87, + "learning_rate": 1.346174973450031e-08, + "loss": 0.6821, + "step": 50988 + }, + { + "epoch": 1.87, + "learning_rate": 1.345808766982825e-08, + "loss": 0.6973, + "step": 50989 + }, + { + "epoch": 1.87, + "learning_rate": 1.3454425605156187e-08, + "loss": 0.6802, + "step": 50990 + }, + { + "epoch": 1.87, + "learning_rate": 1.3450763540484125e-08, + "loss": 0.7031, + "step": 50991 + }, + { + "epoch": 1.87, + "learning_rate": 1.3447101475812061e-08, + "loss": 0.6882, + "step": 50992 + }, + { + "epoch": 1.87, + "learning_rate": 1.3443439411140001e-08, + "loss": 0.6918, + "step": 50993 + }, + { + "epoch": 1.87, + "learning_rate": 1.3439777346467936e-08, + "loss": 0.6924, + "step": 50994 + }, + { + "epoch": 1.87, + "learning_rate": 1.3436115281795876e-08, + "loss": 0.6912, + "step": 50995 + }, + { + "epoch": 1.87, + "learning_rate": 1.3432453217123813e-08, + "loss": 0.6911, + "step": 50996 + }, + { + "epoch": 1.87, + "learning_rate": 1.3428791152451751e-08, + "loss": 0.6847, + "step": 50997 + }, + { + "epoch": 1.87, + "learning_rate": 1.3425129087779691e-08, + "loss": 0.6969, + "step": 50998 + }, + { + "epoch": 1.87, + "learning_rate": 1.3421467023107628e-08, + "loss": 0.694, + "step": 50999 + }, + { + "epoch": 1.87, + "learning_rate": 1.3417804958435566e-08, + "loss": 0.689, + "step": 51000 + }, + { + "epoch": 1.87, + "learning_rate": 1.3414142893763503e-08, + "loss": 0.6898, + "step": 51001 + }, + { + "epoch": 1.87, + "learning_rate": 1.3410480829091441e-08, + "loss": 0.6883, + "step": 51002 + }, + { + "epoch": 1.87, + "learning_rate": 1.3406818764419378e-08, + "loss": 0.683, + "step": 51003 + }, + { + "epoch": 1.87, + "learning_rate": 1.3403156699747317e-08, + "loss": 0.6949, + "step": 51004 + }, + { + "epoch": 1.87, + "learning_rate": 1.3399494635075254e-08, + "loss": 0.7026, + "step": 51005 + }, + { + "epoch": 1.87, + "learning_rate": 1.3395832570403192e-08, + "loss": 0.6938, + "step": 51006 + }, + { + "epoch": 1.87, + "learning_rate": 1.3392170505731132e-08, + "loss": 0.6926, + "step": 51007 + }, + { + "epoch": 1.87, + "learning_rate": 1.3388508441059069e-08, + "loss": 0.6832, + "step": 51008 + }, + { + "epoch": 1.87, + "learning_rate": 1.3384846376387007e-08, + "loss": 0.686, + "step": 51009 + }, + { + "epoch": 1.87, + "learning_rate": 1.3381184311714944e-08, + "loss": 0.6837, + "step": 51010 + }, + { + "epoch": 1.87, + "learning_rate": 1.3377522247042882e-08, + "loss": 0.6858, + "step": 51011 + }, + { + "epoch": 1.87, + "learning_rate": 1.3373860182370819e-08, + "loss": 0.6839, + "step": 51012 + }, + { + "epoch": 1.87, + "learning_rate": 1.3370198117698759e-08, + "loss": 0.6959, + "step": 51013 + }, + { + "epoch": 1.87, + "learning_rate": 1.3366536053026695e-08, + "loss": 0.6782, + "step": 51014 + }, + { + "epoch": 1.87, + "learning_rate": 1.3362873988354634e-08, + "loss": 0.6866, + "step": 51015 + }, + { + "epoch": 1.87, + "learning_rate": 1.3359211923682573e-08, + "loss": 0.6897, + "step": 51016 + }, + { + "epoch": 1.87, + "learning_rate": 1.3355549859010508e-08, + "loss": 0.6917, + "step": 51017 + }, + { + "epoch": 1.87, + "learning_rate": 1.3351887794338448e-08, + "loss": 0.6865, + "step": 51018 + }, + { + "epoch": 1.87, + "learning_rate": 1.3348225729666385e-08, + "loss": 0.684, + "step": 51019 + }, + { + "epoch": 1.87, + "learning_rate": 1.3344563664994323e-08, + "loss": 0.6957, + "step": 51020 + }, + { + "epoch": 1.87, + "learning_rate": 1.334090160032226e-08, + "loss": 0.6869, + "step": 51021 + }, + { + "epoch": 1.87, + "learning_rate": 1.33372395356502e-08, + "loss": 0.6913, + "step": 51022 + }, + { + "epoch": 1.87, + "learning_rate": 1.3333577470978136e-08, + "loss": 0.6818, + "step": 51023 + }, + { + "epoch": 1.87, + "learning_rate": 1.3329915406306075e-08, + "loss": 0.6824, + "step": 51024 + }, + { + "epoch": 1.87, + "learning_rate": 1.3326253341634015e-08, + "loss": 0.6981, + "step": 51025 + }, + { + "epoch": 1.87, + "learning_rate": 1.332259127696195e-08, + "loss": 0.6947, + "step": 51026 + }, + { + "epoch": 1.87, + "learning_rate": 1.331892921228989e-08, + "loss": 0.6954, + "step": 51027 + }, + { + "epoch": 1.87, + "learning_rate": 1.3315267147617826e-08, + "loss": 0.6949, + "step": 51028 + }, + { + "epoch": 1.87, + "learning_rate": 1.3311605082945764e-08, + "loss": 0.6971, + "step": 51029 + }, + { + "epoch": 1.87, + "learning_rate": 1.3307943018273701e-08, + "loss": 0.6877, + "step": 51030 + }, + { + "epoch": 1.87, + "learning_rate": 1.3304280953601641e-08, + "loss": 0.6913, + "step": 51031 + }, + { + "epoch": 1.87, + "learning_rate": 1.3300618888929576e-08, + "loss": 0.699, + "step": 51032 + }, + { + "epoch": 1.87, + "learning_rate": 1.3296956824257516e-08, + "loss": 0.6796, + "step": 51033 + }, + { + "epoch": 1.87, + "learning_rate": 1.3293294759585456e-08, + "loss": 0.6851, + "step": 51034 + }, + { + "epoch": 1.87, + "learning_rate": 1.328963269491339e-08, + "loss": 0.6916, + "step": 51035 + }, + { + "epoch": 1.87, + "learning_rate": 1.328597063024133e-08, + "loss": 0.6992, + "step": 51036 + }, + { + "epoch": 1.87, + "learning_rate": 1.3282308565569267e-08, + "loss": 0.6946, + "step": 51037 + }, + { + "epoch": 1.87, + "learning_rate": 1.3278646500897206e-08, + "loss": 0.702, + "step": 51038 + }, + { + "epoch": 1.87, + "learning_rate": 1.3274984436225142e-08, + "loss": 0.6796, + "step": 51039 + }, + { + "epoch": 1.87, + "learning_rate": 1.3271322371553082e-08, + "loss": 0.6848, + "step": 51040 + }, + { + "epoch": 1.87, + "learning_rate": 1.3267660306881017e-08, + "loss": 0.709, + "step": 51041 + }, + { + "epoch": 1.87, + "learning_rate": 1.3263998242208957e-08, + "loss": 0.6998, + "step": 51042 + }, + { + "epoch": 1.87, + "learning_rate": 1.3260336177536895e-08, + "loss": 0.7, + "step": 51043 + }, + { + "epoch": 1.87, + "learning_rate": 1.3256674112864832e-08, + "loss": 0.6932, + "step": 51044 + }, + { + "epoch": 1.87, + "learning_rate": 1.3253012048192772e-08, + "loss": 0.6972, + "step": 51045 + }, + { + "epoch": 1.87, + "learning_rate": 1.3249349983520708e-08, + "loss": 0.6956, + "step": 51046 + }, + { + "epoch": 1.87, + "learning_rate": 1.3245687918848647e-08, + "loss": 0.6833, + "step": 51047 + }, + { + "epoch": 1.87, + "learning_rate": 1.3242025854176583e-08, + "loss": 0.7032, + "step": 51048 + }, + { + "epoch": 1.87, + "learning_rate": 1.3238363789504523e-08, + "loss": 0.6946, + "step": 51049 + }, + { + "epoch": 1.87, + "learning_rate": 1.3234701724832458e-08, + "loss": 0.6969, + "step": 51050 + }, + { + "epoch": 1.87, + "learning_rate": 1.3231039660160398e-08, + "loss": 0.6958, + "step": 51051 + }, + { + "epoch": 1.87, + "learning_rate": 1.3227377595488336e-08, + "loss": 0.6943, + "step": 51052 + }, + { + "epoch": 1.87, + "learning_rate": 1.3223715530816273e-08, + "loss": 0.6963, + "step": 51053 + }, + { + "epoch": 1.87, + "learning_rate": 1.3220053466144213e-08, + "loss": 0.6909, + "step": 51054 + }, + { + "epoch": 1.87, + "learning_rate": 1.321639140147215e-08, + "loss": 0.6961, + "step": 51055 + }, + { + "epoch": 1.87, + "learning_rate": 1.3212729336800088e-08, + "loss": 0.7035, + "step": 51056 + }, + { + "epoch": 1.87, + "learning_rate": 1.3209067272128024e-08, + "loss": 0.6897, + "step": 51057 + }, + { + "epoch": 1.87, + "learning_rate": 1.3205405207455963e-08, + "loss": 0.6968, + "step": 51058 + }, + { + "epoch": 1.87, + "learning_rate": 1.32017431427839e-08, + "loss": 0.6977, + "step": 51059 + }, + { + "epoch": 1.87, + "learning_rate": 1.319808107811184e-08, + "loss": 0.6864, + "step": 51060 + }, + { + "epoch": 1.87, + "learning_rate": 1.3194419013439776e-08, + "loss": 0.6982, + "step": 51061 + }, + { + "epoch": 1.87, + "learning_rate": 1.3190756948767714e-08, + "loss": 0.6945, + "step": 51062 + }, + { + "epoch": 1.87, + "learning_rate": 1.3187094884095654e-08, + "loss": 0.6896, + "step": 51063 + }, + { + "epoch": 1.87, + "learning_rate": 1.318343281942359e-08, + "loss": 0.6887, + "step": 51064 + }, + { + "epoch": 1.87, + "learning_rate": 1.3179770754751529e-08, + "loss": 0.6716, + "step": 51065 + }, + { + "epoch": 1.87, + "learning_rate": 1.3176108690079466e-08, + "loss": 0.6972, + "step": 51066 + }, + { + "epoch": 1.87, + "learning_rate": 1.3172446625407404e-08, + "loss": 0.6956, + "step": 51067 + }, + { + "epoch": 1.87, + "learning_rate": 1.316878456073534e-08, + "loss": 0.6817, + "step": 51068 + }, + { + "epoch": 1.87, + "learning_rate": 1.316512249606328e-08, + "loss": 0.687, + "step": 51069 + }, + { + "epoch": 1.87, + "learning_rate": 1.3161460431391217e-08, + "loss": 0.6865, + "step": 51070 + }, + { + "epoch": 1.87, + "learning_rate": 1.3157798366719155e-08, + "loss": 0.697, + "step": 51071 + }, + { + "epoch": 1.87, + "learning_rate": 1.3154136302047095e-08, + "loss": 0.6983, + "step": 51072 + }, + { + "epoch": 1.87, + "learning_rate": 1.315047423737503e-08, + "loss": 0.69, + "step": 51073 + }, + { + "epoch": 1.87, + "learning_rate": 1.314681217270297e-08, + "loss": 0.7045, + "step": 51074 + }, + { + "epoch": 1.87, + "learning_rate": 1.3143150108030907e-08, + "loss": 0.6893, + "step": 51075 + }, + { + "epoch": 1.87, + "learning_rate": 1.3139488043358845e-08, + "loss": 0.6816, + "step": 51076 + }, + { + "epoch": 1.87, + "learning_rate": 1.3135825978686782e-08, + "loss": 0.689, + "step": 51077 + }, + { + "epoch": 1.87, + "learning_rate": 1.3132163914014722e-08, + "loss": 0.6928, + "step": 51078 + }, + { + "epoch": 1.87, + "learning_rate": 1.3128501849342658e-08, + "loss": 0.6977, + "step": 51079 + }, + { + "epoch": 1.87, + "learning_rate": 1.3124839784670597e-08, + "loss": 0.6986, + "step": 51080 + }, + { + "epoch": 1.87, + "learning_rate": 1.3121177719998536e-08, + "loss": 0.6987, + "step": 51081 + }, + { + "epoch": 1.87, + "learning_rate": 1.3117515655326471e-08, + "loss": 0.7043, + "step": 51082 + }, + { + "epoch": 1.87, + "learning_rate": 1.3113853590654411e-08, + "loss": 0.7015, + "step": 51083 + }, + { + "epoch": 1.87, + "learning_rate": 1.3110191525982348e-08, + "loss": 0.6861, + "step": 51084 + }, + { + "epoch": 1.87, + "learning_rate": 1.3106529461310286e-08, + "loss": 0.7011, + "step": 51085 + }, + { + "epoch": 1.87, + "learning_rate": 1.3102867396638223e-08, + "loss": 0.6931, + "step": 51086 + }, + { + "epoch": 1.87, + "learning_rate": 1.3099205331966163e-08, + "loss": 0.6996, + "step": 51087 + }, + { + "epoch": 1.87, + "learning_rate": 1.3095543267294098e-08, + "loss": 0.7066, + "step": 51088 + }, + { + "epoch": 1.87, + "learning_rate": 1.3091881202622038e-08, + "loss": 0.6874, + "step": 51089 + }, + { + "epoch": 1.87, + "learning_rate": 1.3088219137949976e-08, + "loss": 0.6967, + "step": 51090 + }, + { + "epoch": 1.87, + "learning_rate": 1.3084557073277913e-08, + "loss": 0.7029, + "step": 51091 + }, + { + "epoch": 1.87, + "learning_rate": 1.3080895008605853e-08, + "loss": 0.7028, + "step": 51092 + }, + { + "epoch": 1.87, + "learning_rate": 1.3077232943933789e-08, + "loss": 0.694, + "step": 51093 + }, + { + "epoch": 1.87, + "learning_rate": 1.3073570879261727e-08, + "loss": 0.7068, + "step": 51094 + }, + { + "epoch": 1.87, + "learning_rate": 1.3069908814589664e-08, + "loss": 0.6854, + "step": 51095 + }, + { + "epoch": 1.87, + "learning_rate": 1.3066246749917604e-08, + "loss": 0.697, + "step": 51096 + }, + { + "epoch": 1.87, + "learning_rate": 1.3062584685245539e-08, + "loss": 0.6937, + "step": 51097 + }, + { + "epoch": 1.87, + "learning_rate": 1.3058922620573479e-08, + "loss": 0.6978, + "step": 51098 + }, + { + "epoch": 1.87, + "learning_rate": 1.3055260555901417e-08, + "loss": 0.6783, + "step": 51099 + }, + { + "epoch": 1.87, + "learning_rate": 1.3051598491229354e-08, + "loss": 0.6897, + "step": 51100 + }, + { + "epoch": 1.87, + "learning_rate": 1.3047936426557294e-08, + "loss": 0.6904, + "step": 51101 + }, + { + "epoch": 1.87, + "learning_rate": 1.304427436188523e-08, + "loss": 0.6968, + "step": 51102 + }, + { + "epoch": 1.87, + "learning_rate": 1.3040612297213169e-08, + "loss": 0.6838, + "step": 51103 + }, + { + "epoch": 1.87, + "learning_rate": 1.3036950232541105e-08, + "loss": 0.6983, + "step": 51104 + }, + { + "epoch": 1.87, + "learning_rate": 1.3033288167869043e-08, + "loss": 0.6944, + "step": 51105 + }, + { + "epoch": 1.87, + "learning_rate": 1.302962610319698e-08, + "loss": 0.7015, + "step": 51106 + }, + { + "epoch": 1.87, + "learning_rate": 1.302596403852492e-08, + "loss": 0.6777, + "step": 51107 + }, + { + "epoch": 1.87, + "learning_rate": 1.3022301973852858e-08, + "loss": 0.6975, + "step": 51108 + }, + { + "epoch": 1.87, + "learning_rate": 1.3018639909180795e-08, + "loss": 0.7008, + "step": 51109 + }, + { + "epoch": 1.87, + "learning_rate": 1.3014977844508735e-08, + "loss": 0.6961, + "step": 51110 + }, + { + "epoch": 1.87, + "learning_rate": 1.3011315779836671e-08, + "loss": 0.6918, + "step": 51111 + }, + { + "epoch": 1.87, + "learning_rate": 1.300765371516461e-08, + "loss": 0.6924, + "step": 51112 + }, + { + "epoch": 1.87, + "learning_rate": 1.3003991650492546e-08, + "loss": 0.6915, + "step": 51113 + }, + { + "epoch": 1.87, + "learning_rate": 1.3000329585820485e-08, + "loss": 0.6936, + "step": 51114 + }, + { + "epoch": 1.87, + "learning_rate": 1.2996667521148421e-08, + "loss": 0.6948, + "step": 51115 + }, + { + "epoch": 1.87, + "learning_rate": 1.2993005456476361e-08, + "loss": 0.6901, + "step": 51116 + }, + { + "epoch": 1.87, + "learning_rate": 1.29893433918043e-08, + "loss": 0.6881, + "step": 51117 + }, + { + "epoch": 1.87, + "learning_rate": 1.2985681327132236e-08, + "loss": 0.6885, + "step": 51118 + }, + { + "epoch": 1.87, + "learning_rate": 1.2982019262460176e-08, + "loss": 0.6847, + "step": 51119 + }, + { + "epoch": 1.87, + "learning_rate": 1.2978357197788111e-08, + "loss": 0.6977, + "step": 51120 + }, + { + "epoch": 1.87, + "learning_rate": 1.2974695133116051e-08, + "loss": 0.6815, + "step": 51121 + }, + { + "epoch": 1.87, + "learning_rate": 1.2971033068443988e-08, + "loss": 0.7011, + "step": 51122 + }, + { + "epoch": 1.87, + "learning_rate": 1.2967371003771926e-08, + "loss": 0.6915, + "step": 51123 + }, + { + "epoch": 1.87, + "learning_rate": 1.2963708939099862e-08, + "loss": 0.7101, + "step": 51124 + }, + { + "epoch": 1.87, + "learning_rate": 1.2960046874427802e-08, + "loss": 0.6946, + "step": 51125 + }, + { + "epoch": 1.87, + "learning_rate": 1.295638480975574e-08, + "loss": 0.6799, + "step": 51126 + }, + { + "epoch": 1.87, + "learning_rate": 1.2952722745083677e-08, + "loss": 0.695, + "step": 51127 + }, + { + "epoch": 1.87, + "learning_rate": 1.2949060680411617e-08, + "loss": 0.6917, + "step": 51128 + }, + { + "epoch": 1.87, + "learning_rate": 1.2945398615739552e-08, + "loss": 0.6888, + "step": 51129 + }, + { + "epoch": 1.87, + "learning_rate": 1.2941736551067492e-08, + "loss": 0.7047, + "step": 51130 + }, + { + "epoch": 1.87, + "learning_rate": 1.2938074486395429e-08, + "loss": 0.6974, + "step": 51131 + }, + { + "epoch": 1.87, + "learning_rate": 1.2934412421723367e-08, + "loss": 0.7019, + "step": 51132 + }, + { + "epoch": 1.87, + "learning_rate": 1.2930750357051304e-08, + "loss": 0.6857, + "step": 51133 + }, + { + "epoch": 1.87, + "learning_rate": 1.2927088292379243e-08, + "loss": 0.691, + "step": 51134 + }, + { + "epoch": 1.87, + "learning_rate": 1.2923426227707182e-08, + "loss": 0.6949, + "step": 51135 + }, + { + "epoch": 1.87, + "learning_rate": 1.2919764163035118e-08, + "loss": 0.7006, + "step": 51136 + }, + { + "epoch": 1.87, + "learning_rate": 1.2916102098363058e-08, + "loss": 0.6988, + "step": 51137 + }, + { + "epoch": 1.87, + "learning_rate": 1.2912440033690993e-08, + "loss": 0.6885, + "step": 51138 + }, + { + "epoch": 1.87, + "learning_rate": 1.2908777969018933e-08, + "loss": 0.6874, + "step": 51139 + }, + { + "epoch": 1.87, + "learning_rate": 1.290511590434687e-08, + "loss": 0.689, + "step": 51140 + }, + { + "epoch": 1.87, + "learning_rate": 1.2901453839674808e-08, + "loss": 0.6901, + "step": 51141 + }, + { + "epoch": 1.87, + "learning_rate": 1.2897791775002745e-08, + "loss": 0.6833, + "step": 51142 + }, + { + "epoch": 1.87, + "learning_rate": 1.2894129710330685e-08, + "loss": 0.6967, + "step": 51143 + }, + { + "epoch": 1.87, + "learning_rate": 1.2890467645658623e-08, + "loss": 0.6867, + "step": 51144 + }, + { + "epoch": 1.87, + "learning_rate": 1.288680558098656e-08, + "loss": 0.6993, + "step": 51145 + }, + { + "epoch": 1.87, + "learning_rate": 1.2883143516314498e-08, + "loss": 0.696, + "step": 51146 + }, + { + "epoch": 1.87, + "learning_rate": 1.2879481451642434e-08, + "loss": 0.7038, + "step": 51147 + }, + { + "epoch": 1.87, + "learning_rate": 1.2875819386970374e-08, + "loss": 0.6808, + "step": 51148 + }, + { + "epoch": 1.87, + "learning_rate": 1.2872157322298311e-08, + "loss": 0.7038, + "step": 51149 + }, + { + "epoch": 1.87, + "learning_rate": 1.286849525762625e-08, + "loss": 0.6927, + "step": 51150 + }, + { + "epoch": 1.87, + "learning_rate": 1.2864833192954186e-08, + "loss": 0.6835, + "step": 51151 + }, + { + "epoch": 1.87, + "learning_rate": 1.2861171128282126e-08, + "loss": 0.6981, + "step": 51152 + }, + { + "epoch": 1.87, + "learning_rate": 1.2857509063610064e-08, + "loss": 0.6834, + "step": 51153 + }, + { + "epoch": 1.87, + "learning_rate": 1.2853846998938e-08, + "loss": 0.6898, + "step": 51154 + }, + { + "epoch": 1.87, + "learning_rate": 1.2850184934265939e-08, + "loss": 0.6911, + "step": 51155 + }, + { + "epoch": 1.87, + "learning_rate": 1.2846522869593876e-08, + "loss": 0.6966, + "step": 51156 + }, + { + "epoch": 1.87, + "learning_rate": 1.2842860804921816e-08, + "loss": 0.6762, + "step": 51157 + }, + { + "epoch": 1.87, + "learning_rate": 1.2839198740249752e-08, + "loss": 0.6956, + "step": 51158 + }, + { + "epoch": 1.87, + "learning_rate": 1.283553667557769e-08, + "loss": 0.6884, + "step": 51159 + }, + { + "epoch": 1.87, + "learning_rate": 1.2831874610905627e-08, + "loss": 0.6892, + "step": 51160 + }, + { + "epoch": 1.87, + "learning_rate": 1.2828212546233565e-08, + "loss": 0.6933, + "step": 51161 + }, + { + "epoch": 1.87, + "learning_rate": 1.2824550481561502e-08, + "loss": 0.6993, + "step": 51162 + }, + { + "epoch": 1.87, + "learning_rate": 1.2820888416889442e-08, + "loss": 0.6907, + "step": 51163 + }, + { + "epoch": 1.87, + "learning_rate": 1.281722635221738e-08, + "loss": 0.6949, + "step": 51164 + }, + { + "epoch": 1.87, + "learning_rate": 1.2813564287545317e-08, + "loss": 0.7012, + "step": 51165 + }, + { + "epoch": 1.87, + "learning_rate": 1.2809902222873257e-08, + "loss": 0.694, + "step": 51166 + }, + { + "epoch": 1.87, + "learning_rate": 1.2806240158201193e-08, + "loss": 0.6973, + "step": 51167 + }, + { + "epoch": 1.87, + "learning_rate": 1.2802578093529132e-08, + "loss": 0.6975, + "step": 51168 + }, + { + "epoch": 1.87, + "learning_rate": 1.2798916028857068e-08, + "loss": 0.6868, + "step": 51169 + }, + { + "epoch": 1.87, + "learning_rate": 1.2795253964185006e-08, + "loss": 0.684, + "step": 51170 + }, + { + "epoch": 1.87, + "learning_rate": 1.2791591899512943e-08, + "loss": 0.6924, + "step": 51171 + }, + { + "epoch": 1.87, + "learning_rate": 1.2787929834840883e-08, + "loss": 0.697, + "step": 51172 + }, + { + "epoch": 1.87, + "learning_rate": 1.2784267770168821e-08, + "loss": 0.6817, + "step": 51173 + }, + { + "epoch": 1.87, + "learning_rate": 1.2780605705496758e-08, + "loss": 0.6991, + "step": 51174 + }, + { + "epoch": 1.87, + "learning_rate": 1.2776943640824698e-08, + "loss": 0.6928, + "step": 51175 + }, + { + "epoch": 1.87, + "learning_rate": 1.2773281576152633e-08, + "loss": 0.6993, + "step": 51176 + }, + { + "epoch": 1.87, + "learning_rate": 1.2769619511480573e-08, + "loss": 0.6871, + "step": 51177 + }, + { + "epoch": 1.87, + "learning_rate": 1.276595744680851e-08, + "loss": 0.6944, + "step": 51178 + }, + { + "epoch": 1.87, + "learning_rate": 1.2762295382136448e-08, + "loss": 0.6841, + "step": 51179 + }, + { + "epoch": 1.87, + "learning_rate": 1.2758633317464384e-08, + "loss": 0.6903, + "step": 51180 + }, + { + "epoch": 1.87, + "learning_rate": 1.2754971252792324e-08, + "loss": 0.6946, + "step": 51181 + }, + { + "epoch": 1.87, + "learning_rate": 1.2751309188120262e-08, + "loss": 0.691, + "step": 51182 + }, + { + "epoch": 1.87, + "learning_rate": 1.2747647123448199e-08, + "loss": 0.697, + "step": 51183 + }, + { + "epoch": 1.87, + "learning_rate": 1.2743985058776139e-08, + "loss": 0.6921, + "step": 51184 + }, + { + "epoch": 1.87, + "learning_rate": 1.2740322994104074e-08, + "loss": 0.6969, + "step": 51185 + }, + { + "epoch": 1.87, + "learning_rate": 1.2736660929432014e-08, + "loss": 0.6934, + "step": 51186 + }, + { + "epoch": 1.87, + "learning_rate": 1.273299886475995e-08, + "loss": 0.6836, + "step": 51187 + }, + { + "epoch": 1.87, + "learning_rate": 1.2729336800087889e-08, + "loss": 0.69, + "step": 51188 + }, + { + "epoch": 1.87, + "learning_rate": 1.2725674735415825e-08, + "loss": 0.6995, + "step": 51189 + }, + { + "epoch": 1.87, + "learning_rate": 1.2722012670743765e-08, + "loss": 0.6836, + "step": 51190 + }, + { + "epoch": 1.87, + "learning_rate": 1.2718350606071704e-08, + "loss": 0.6978, + "step": 51191 + }, + { + "epoch": 1.87, + "learning_rate": 1.271468854139964e-08, + "loss": 0.7129, + "step": 51192 + }, + { + "epoch": 1.87, + "learning_rate": 1.2711026476727579e-08, + "loss": 0.6934, + "step": 51193 + }, + { + "epoch": 1.87, + "learning_rate": 1.2707364412055515e-08, + "loss": 0.6973, + "step": 51194 + }, + { + "epoch": 1.87, + "learning_rate": 1.2703702347383455e-08, + "loss": 0.7081, + "step": 51195 + }, + { + "epoch": 1.87, + "learning_rate": 1.2700040282711392e-08, + "loss": 0.7002, + "step": 51196 + }, + { + "epoch": 1.87, + "learning_rate": 1.269637821803933e-08, + "loss": 0.6951, + "step": 51197 + }, + { + "epoch": 1.87, + "learning_rate": 1.2692716153367267e-08, + "loss": 0.7111, + "step": 51198 + }, + { + "epoch": 1.87, + "learning_rate": 1.2689054088695206e-08, + "loss": 0.7011, + "step": 51199 + }, + { + "epoch": 1.87, + "learning_rate": 1.2685392024023145e-08, + "loss": 0.6955, + "step": 51200 + }, + { + "epoch": 1.87, + "learning_rate": 1.2681729959351081e-08, + "loss": 0.6938, + "step": 51201 + }, + { + "epoch": 1.87, + "learning_rate": 1.267806789467902e-08, + "loss": 0.6871, + "step": 51202 + }, + { + "epoch": 1.87, + "learning_rate": 1.2674405830006956e-08, + "loss": 0.6955, + "step": 51203 + }, + { + "epoch": 1.87, + "learning_rate": 1.2670743765334896e-08, + "loss": 0.6861, + "step": 51204 + }, + { + "epoch": 1.87, + "learning_rate": 1.2667081700662833e-08, + "loss": 0.6979, + "step": 51205 + }, + { + "epoch": 1.87, + "learning_rate": 1.2663419635990771e-08, + "loss": 0.6896, + "step": 51206 + }, + { + "epoch": 1.87, + "learning_rate": 1.2659757571318708e-08, + "loss": 0.6919, + "step": 51207 + }, + { + "epoch": 1.87, + "learning_rate": 1.2656095506646646e-08, + "loss": 0.6996, + "step": 51208 + }, + { + "epoch": 1.87, + "learning_rate": 1.2652433441974586e-08, + "loss": 0.6982, + "step": 51209 + }, + { + "epoch": 1.87, + "learning_rate": 1.2648771377302523e-08, + "loss": 0.6899, + "step": 51210 + }, + { + "epoch": 1.87, + "learning_rate": 1.264510931263046e-08, + "loss": 0.6892, + "step": 51211 + }, + { + "epoch": 1.87, + "learning_rate": 1.2641447247958397e-08, + "loss": 0.6836, + "step": 51212 + }, + { + "epoch": 1.87, + "learning_rate": 1.2637785183286337e-08, + "loss": 0.6976, + "step": 51213 + }, + { + "epoch": 1.87, + "learning_rate": 1.2634123118614274e-08, + "loss": 0.6959, + "step": 51214 + }, + { + "epoch": 1.87, + "learning_rate": 1.2630461053942212e-08, + "loss": 0.6911, + "step": 51215 + }, + { + "epoch": 1.87, + "learning_rate": 1.2626798989270149e-08, + "loss": 0.71, + "step": 51216 + }, + { + "epoch": 1.87, + "learning_rate": 1.2623136924598087e-08, + "loss": 0.6832, + "step": 51217 + }, + { + "epoch": 1.87, + "learning_rate": 1.2619474859926027e-08, + "loss": 0.7002, + "step": 51218 + }, + { + "epoch": 1.87, + "learning_rate": 1.2615812795253964e-08, + "loss": 0.7036, + "step": 51219 + }, + { + "epoch": 1.87, + "learning_rate": 1.2612150730581902e-08, + "loss": 0.6881, + "step": 51220 + }, + { + "epoch": 1.87, + "learning_rate": 1.2608488665909839e-08, + "loss": 0.6967, + "step": 51221 + }, + { + "epoch": 1.87, + "learning_rate": 1.2604826601237779e-08, + "loss": 0.6918, + "step": 51222 + }, + { + "epoch": 1.87, + "learning_rate": 1.2601164536565714e-08, + "loss": 0.6929, + "step": 51223 + }, + { + "epoch": 1.87, + "learning_rate": 1.2597502471893653e-08, + "loss": 0.6791, + "step": 51224 + }, + { + "epoch": 1.87, + "learning_rate": 1.259384040722159e-08, + "loss": 0.6938, + "step": 51225 + }, + { + "epoch": 1.87, + "learning_rate": 1.2590178342549528e-08, + "loss": 0.6858, + "step": 51226 + }, + { + "epoch": 1.87, + "learning_rate": 1.2586516277877468e-08, + "loss": 0.6996, + "step": 51227 + }, + { + "epoch": 1.87, + "learning_rate": 1.2582854213205405e-08, + "loss": 0.6936, + "step": 51228 + }, + { + "epoch": 1.87, + "learning_rate": 1.2579192148533343e-08, + "loss": 0.6976, + "step": 51229 + }, + { + "epoch": 1.87, + "learning_rate": 1.257553008386128e-08, + "loss": 0.6978, + "step": 51230 + }, + { + "epoch": 1.87, + "learning_rate": 1.257186801918922e-08, + "loss": 0.6852, + "step": 51231 + }, + { + "epoch": 1.87, + "learning_rate": 1.2568205954517155e-08, + "loss": 0.6898, + "step": 51232 + }, + { + "epoch": 1.87, + "learning_rate": 1.2564543889845095e-08, + "loss": 0.6962, + "step": 51233 + }, + { + "epoch": 1.87, + "learning_rate": 1.2560881825173031e-08, + "loss": 0.6816, + "step": 51234 + }, + { + "epoch": 1.87, + "learning_rate": 1.255721976050097e-08, + "loss": 0.698, + "step": 51235 + }, + { + "epoch": 1.87, + "learning_rate": 1.255355769582891e-08, + "loss": 0.6809, + "step": 51236 + }, + { + "epoch": 1.87, + "learning_rate": 1.2549895631156846e-08, + "loss": 0.6919, + "step": 51237 + }, + { + "epoch": 1.87, + "learning_rate": 1.2546233566484784e-08, + "loss": 0.6972, + "step": 51238 + }, + { + "epoch": 1.87, + "learning_rate": 1.2542571501812721e-08, + "loss": 0.6804, + "step": 51239 + }, + { + "epoch": 1.87, + "learning_rate": 1.2538909437140661e-08, + "loss": 0.6906, + "step": 51240 + }, + { + "epoch": 1.87, + "learning_rate": 1.2535247372468596e-08, + "loss": 0.6943, + "step": 51241 + }, + { + "epoch": 1.87, + "learning_rate": 1.2531585307796536e-08, + "loss": 0.6932, + "step": 51242 + }, + { + "epoch": 1.87, + "learning_rate": 1.2527923243124472e-08, + "loss": 0.6876, + "step": 51243 + }, + { + "epoch": 1.87, + "learning_rate": 1.252426117845241e-08, + "loss": 0.6842, + "step": 51244 + }, + { + "epoch": 1.87, + "learning_rate": 1.252059911378035e-08, + "loss": 0.6735, + "step": 51245 + }, + { + "epoch": 1.87, + "learning_rate": 1.2516937049108287e-08, + "loss": 0.6874, + "step": 51246 + }, + { + "epoch": 1.87, + "learning_rate": 1.2513274984436225e-08, + "loss": 0.6954, + "step": 51247 + }, + { + "epoch": 1.88, + "learning_rate": 1.2509612919764162e-08, + "loss": 0.6977, + "step": 51248 + }, + { + "epoch": 1.88, + "learning_rate": 1.25059508550921e-08, + "loss": 0.6858, + "step": 51249 + }, + { + "epoch": 1.88, + "learning_rate": 1.2502288790420037e-08, + "loss": 0.6998, + "step": 51250 + }, + { + "epoch": 1.88, + "learning_rate": 1.2498626725747977e-08, + "loss": 0.7045, + "step": 51251 + }, + { + "epoch": 1.88, + "learning_rate": 1.2494964661075914e-08, + "loss": 0.6852, + "step": 51252 + }, + { + "epoch": 1.88, + "learning_rate": 1.2491302596403852e-08, + "loss": 0.6935, + "step": 51253 + }, + { + "epoch": 1.88, + "learning_rate": 1.248764053173179e-08, + "loss": 0.711, + "step": 51254 + }, + { + "epoch": 1.88, + "learning_rate": 1.2483978467059728e-08, + "loss": 0.6871, + "step": 51255 + }, + { + "epoch": 1.88, + "learning_rate": 1.2480316402387665e-08, + "loss": 0.6886, + "step": 51256 + }, + { + "epoch": 1.88, + "learning_rate": 1.2476654337715603e-08, + "loss": 0.6852, + "step": 51257 + }, + { + "epoch": 1.88, + "learning_rate": 1.2472992273043542e-08, + "loss": 0.7063, + "step": 51258 + }, + { + "epoch": 1.88, + "learning_rate": 1.246933020837148e-08, + "loss": 0.6982, + "step": 51259 + }, + { + "epoch": 1.88, + "learning_rate": 1.2465668143699418e-08, + "loss": 0.6738, + "step": 51260 + }, + { + "epoch": 1.88, + "learning_rate": 1.2462006079027355e-08, + "loss": 0.6876, + "step": 51261 + }, + { + "epoch": 1.88, + "learning_rate": 1.2458344014355293e-08, + "loss": 0.6866, + "step": 51262 + }, + { + "epoch": 1.88, + "learning_rate": 1.2454681949683231e-08, + "loss": 0.6907, + "step": 51263 + }, + { + "epoch": 1.88, + "learning_rate": 1.2451019885011168e-08, + "loss": 0.6897, + "step": 51264 + }, + { + "epoch": 1.88, + "learning_rate": 1.2447357820339106e-08, + "loss": 0.6769, + "step": 51265 + }, + { + "epoch": 1.88, + "learning_rate": 1.2443695755667044e-08, + "loss": 0.6955, + "step": 51266 + }, + { + "epoch": 1.88, + "learning_rate": 1.2440033690994981e-08, + "loss": 0.6835, + "step": 51267 + }, + { + "epoch": 1.88, + "learning_rate": 1.2436371626322921e-08, + "loss": 0.6855, + "step": 51268 + }, + { + "epoch": 1.88, + "learning_rate": 1.243270956165086e-08, + "loss": 0.6909, + "step": 51269 + }, + { + "epoch": 1.88, + "learning_rate": 1.2429047496978796e-08, + "loss": 0.6925, + "step": 51270 + }, + { + "epoch": 1.88, + "learning_rate": 1.2425385432306734e-08, + "loss": 0.6909, + "step": 51271 + }, + { + "epoch": 1.88, + "learning_rate": 1.2421723367634672e-08, + "loss": 0.6937, + "step": 51272 + }, + { + "epoch": 1.88, + "learning_rate": 1.2418061302962609e-08, + "loss": 0.6929, + "step": 51273 + }, + { + "epoch": 1.88, + "learning_rate": 1.2414399238290547e-08, + "loss": 0.6934, + "step": 51274 + }, + { + "epoch": 1.88, + "learning_rate": 1.2410737173618486e-08, + "loss": 0.6887, + "step": 51275 + }, + { + "epoch": 1.88, + "learning_rate": 1.2407075108946422e-08, + "loss": 0.7013, + "step": 51276 + }, + { + "epoch": 1.88, + "learning_rate": 1.240341304427436e-08, + "loss": 0.6921, + "step": 51277 + }, + { + "epoch": 1.88, + "learning_rate": 1.23997509796023e-08, + "loss": 0.6874, + "step": 51278 + }, + { + "epoch": 1.88, + "learning_rate": 1.2396088914930237e-08, + "loss": 0.6902, + "step": 51279 + }, + { + "epoch": 1.88, + "learning_rate": 1.2392426850258175e-08, + "loss": 0.6977, + "step": 51280 + }, + { + "epoch": 1.88, + "learning_rate": 1.2388764785586114e-08, + "loss": 0.6967, + "step": 51281 + }, + { + "epoch": 1.88, + "learning_rate": 1.238510272091405e-08, + "loss": 0.6958, + "step": 51282 + }, + { + "epoch": 1.88, + "learning_rate": 1.2381440656241988e-08, + "loss": 0.7008, + "step": 51283 + }, + { + "epoch": 1.88, + "learning_rate": 1.2377778591569927e-08, + "loss": 0.6906, + "step": 51284 + }, + { + "epoch": 1.88, + "learning_rate": 1.2374116526897863e-08, + "loss": 0.6934, + "step": 51285 + }, + { + "epoch": 1.88, + "learning_rate": 1.2370454462225802e-08, + "loss": 0.688, + "step": 51286 + }, + { + "epoch": 1.88, + "learning_rate": 1.2366792397553742e-08, + "loss": 0.7028, + "step": 51287 + }, + { + "epoch": 1.88, + "learning_rate": 1.2363130332881678e-08, + "loss": 0.6916, + "step": 51288 + }, + { + "epoch": 1.88, + "learning_rate": 1.2359468268209616e-08, + "loss": 0.6951, + "step": 51289 + }, + { + "epoch": 1.88, + "learning_rate": 1.2355806203537555e-08, + "loss": 0.7132, + "step": 51290 + }, + { + "epoch": 1.88, + "learning_rate": 1.2352144138865491e-08, + "loss": 0.6938, + "step": 51291 + }, + { + "epoch": 1.88, + "learning_rate": 1.234848207419343e-08, + "loss": 0.6863, + "step": 51292 + }, + { + "epoch": 1.88, + "learning_rate": 1.2344820009521368e-08, + "loss": 0.6887, + "step": 51293 + }, + { + "epoch": 1.88, + "learning_rate": 1.2341157944849305e-08, + "loss": 0.6951, + "step": 51294 + }, + { + "epoch": 1.88, + "learning_rate": 1.2337495880177243e-08, + "loss": 0.7021, + "step": 51295 + }, + { + "epoch": 1.88, + "learning_rate": 1.2333833815505181e-08, + "loss": 0.691, + "step": 51296 + }, + { + "epoch": 1.88, + "learning_rate": 1.233017175083312e-08, + "loss": 0.6803, + "step": 51297 + }, + { + "epoch": 1.88, + "learning_rate": 1.2326509686161058e-08, + "loss": 0.7063, + "step": 51298 + }, + { + "epoch": 1.88, + "learning_rate": 1.2322847621488996e-08, + "loss": 0.6831, + "step": 51299 + }, + { + "epoch": 1.88, + "learning_rate": 1.2319185556816932e-08, + "loss": 0.7014, + "step": 51300 + }, + { + "epoch": 1.88, + "learning_rate": 1.231552349214487e-08, + "loss": 0.6912, + "step": 51301 + }, + { + "epoch": 1.88, + "learning_rate": 1.2311861427472809e-08, + "loss": 0.6987, + "step": 51302 + }, + { + "epoch": 1.88, + "learning_rate": 1.2308199362800746e-08, + "loss": 0.6845, + "step": 51303 + }, + { + "epoch": 1.88, + "learning_rate": 1.2304537298128684e-08, + "loss": 0.6939, + "step": 51304 + }, + { + "epoch": 1.88, + "learning_rate": 1.2300875233456622e-08, + "loss": 0.6954, + "step": 51305 + }, + { + "epoch": 1.88, + "learning_rate": 1.229721316878456e-08, + "loss": 0.6932, + "step": 51306 + }, + { + "epoch": 1.88, + "learning_rate": 1.2293551104112499e-08, + "loss": 0.6753, + "step": 51307 + }, + { + "epoch": 1.88, + "learning_rate": 1.2289889039440435e-08, + "loss": 0.6748, + "step": 51308 + }, + { + "epoch": 1.88, + "learning_rate": 1.2286226974768374e-08, + "loss": 0.697, + "step": 51309 + }, + { + "epoch": 1.88, + "learning_rate": 1.2282564910096312e-08, + "loss": 0.7023, + "step": 51310 + }, + { + "epoch": 1.88, + "learning_rate": 1.2278902845424249e-08, + "loss": 0.6983, + "step": 51311 + }, + { + "epoch": 1.88, + "learning_rate": 1.2275240780752187e-08, + "loss": 0.6935, + "step": 51312 + }, + { + "epoch": 1.88, + "learning_rate": 1.2271578716080125e-08, + "loss": 0.7094, + "step": 51313 + }, + { + "epoch": 1.88, + "learning_rate": 1.2267916651408063e-08, + "loss": 0.6946, + "step": 51314 + }, + { + "epoch": 1.88, + "learning_rate": 1.2264254586736002e-08, + "loss": 0.7016, + "step": 51315 + }, + { + "epoch": 1.88, + "learning_rate": 1.226059252206394e-08, + "loss": 0.6929, + "step": 51316 + }, + { + "epoch": 1.88, + "learning_rate": 1.2256930457391877e-08, + "loss": 0.6932, + "step": 51317 + }, + { + "epoch": 1.88, + "learning_rate": 1.2253268392719815e-08, + "loss": 0.6905, + "step": 51318 + }, + { + "epoch": 1.88, + "learning_rate": 1.2249606328047753e-08, + "loss": 0.6956, + "step": 51319 + }, + { + "epoch": 1.88, + "learning_rate": 1.224594426337569e-08, + "loss": 0.6876, + "step": 51320 + }, + { + "epoch": 1.88, + "learning_rate": 1.2242282198703628e-08, + "loss": 0.6913, + "step": 51321 + }, + { + "epoch": 1.88, + "learning_rate": 1.2238620134031566e-08, + "loss": 0.6816, + "step": 51322 + }, + { + "epoch": 1.88, + "learning_rate": 1.2234958069359503e-08, + "loss": 0.693, + "step": 51323 + }, + { + "epoch": 1.88, + "learning_rate": 1.2231296004687443e-08, + "loss": 0.6906, + "step": 51324 + }, + { + "epoch": 1.88, + "learning_rate": 1.2227633940015381e-08, + "loss": 0.6877, + "step": 51325 + }, + { + "epoch": 1.88, + "learning_rate": 1.2223971875343318e-08, + "loss": 0.6888, + "step": 51326 + }, + { + "epoch": 1.88, + "learning_rate": 1.2220309810671256e-08, + "loss": 0.6883, + "step": 51327 + }, + { + "epoch": 1.88, + "learning_rate": 1.2216647745999194e-08, + "loss": 0.6951, + "step": 51328 + }, + { + "epoch": 1.88, + "learning_rate": 1.2212985681327131e-08, + "loss": 0.7113, + "step": 51329 + }, + { + "epoch": 1.88, + "learning_rate": 1.2209323616655069e-08, + "loss": 0.6882, + "step": 51330 + }, + { + "epoch": 1.88, + "learning_rate": 1.2205661551983007e-08, + "loss": 0.6907, + "step": 51331 + }, + { + "epoch": 1.88, + "learning_rate": 1.2201999487310944e-08, + "loss": 0.6994, + "step": 51332 + }, + { + "epoch": 1.88, + "learning_rate": 1.2198337422638884e-08, + "loss": 0.6963, + "step": 51333 + }, + { + "epoch": 1.88, + "learning_rate": 1.2194675357966822e-08, + "loss": 0.6909, + "step": 51334 + }, + { + "epoch": 1.88, + "learning_rate": 1.2191013293294759e-08, + "loss": 0.6884, + "step": 51335 + }, + { + "epoch": 1.88, + "learning_rate": 1.2187351228622697e-08, + "loss": 0.7, + "step": 51336 + }, + { + "epoch": 1.88, + "learning_rate": 1.2183689163950635e-08, + "loss": 0.6952, + "step": 51337 + }, + { + "epoch": 1.88, + "learning_rate": 1.2180027099278572e-08, + "loss": 0.7018, + "step": 51338 + }, + { + "epoch": 1.88, + "learning_rate": 1.217636503460651e-08, + "loss": 0.6997, + "step": 51339 + }, + { + "epoch": 1.88, + "learning_rate": 1.2172702969934449e-08, + "loss": 0.7005, + "step": 51340 + }, + { + "epoch": 1.88, + "learning_rate": 1.2169040905262385e-08, + "loss": 0.6999, + "step": 51341 + }, + { + "epoch": 1.88, + "learning_rate": 1.2165378840590325e-08, + "loss": 0.6966, + "step": 51342 + }, + { + "epoch": 1.88, + "learning_rate": 1.2161716775918263e-08, + "loss": 0.6892, + "step": 51343 + }, + { + "epoch": 1.88, + "learning_rate": 1.21580547112462e-08, + "loss": 0.7063, + "step": 51344 + }, + { + "epoch": 1.88, + "learning_rate": 1.2154392646574138e-08, + "loss": 0.6923, + "step": 51345 + }, + { + "epoch": 1.88, + "learning_rate": 1.2150730581902077e-08, + "loss": 0.6855, + "step": 51346 + }, + { + "epoch": 1.88, + "learning_rate": 1.2147068517230013e-08, + "loss": 0.6892, + "step": 51347 + }, + { + "epoch": 1.88, + "learning_rate": 1.2143406452557951e-08, + "loss": 0.6934, + "step": 51348 + }, + { + "epoch": 1.88, + "learning_rate": 1.213974438788589e-08, + "loss": 0.6973, + "step": 51349 + }, + { + "epoch": 1.88, + "learning_rate": 1.2136082323213826e-08, + "loss": 0.6944, + "step": 51350 + }, + { + "epoch": 1.88, + "learning_rate": 1.2132420258541766e-08, + "loss": 0.6932, + "step": 51351 + }, + { + "epoch": 1.88, + "learning_rate": 1.2128758193869703e-08, + "loss": 0.6908, + "step": 51352 + }, + { + "epoch": 1.88, + "learning_rate": 1.2125096129197641e-08, + "loss": 0.6847, + "step": 51353 + }, + { + "epoch": 1.88, + "learning_rate": 1.212143406452558e-08, + "loss": 0.6995, + "step": 51354 + }, + { + "epoch": 1.88, + "learning_rate": 1.2117771999853516e-08, + "loss": 0.6944, + "step": 51355 + }, + { + "epoch": 1.88, + "learning_rate": 1.2114109935181454e-08, + "loss": 0.6916, + "step": 51356 + }, + { + "epoch": 1.88, + "learning_rate": 1.2110447870509393e-08, + "loss": 0.6913, + "step": 51357 + }, + { + "epoch": 1.88, + "learning_rate": 1.2106785805837331e-08, + "loss": 0.7015, + "step": 51358 + }, + { + "epoch": 1.88, + "learning_rate": 1.2103123741165268e-08, + "loss": 0.6995, + "step": 51359 + }, + { + "epoch": 1.88, + "learning_rate": 1.2099461676493207e-08, + "loss": 0.6821, + "step": 51360 + }, + { + "epoch": 1.88, + "learning_rate": 1.2095799611821144e-08, + "loss": 0.6988, + "step": 51361 + }, + { + "epoch": 1.88, + "learning_rate": 1.2092137547149082e-08, + "loss": 0.6926, + "step": 51362 + }, + { + "epoch": 1.88, + "learning_rate": 1.208847548247702e-08, + "loss": 0.7015, + "step": 51363 + }, + { + "epoch": 1.88, + "learning_rate": 1.2084813417804957e-08, + "loss": 0.6999, + "step": 51364 + }, + { + "epoch": 1.88, + "learning_rate": 1.2081151353132896e-08, + "loss": 0.6929, + "step": 51365 + }, + { + "epoch": 1.88, + "learning_rate": 1.2077489288460834e-08, + "loss": 0.7018, + "step": 51366 + }, + { + "epoch": 1.88, + "learning_rate": 1.207382722378877e-08, + "loss": 0.7066, + "step": 51367 + }, + { + "epoch": 1.88, + "learning_rate": 1.2070165159116709e-08, + "loss": 0.7017, + "step": 51368 + }, + { + "epoch": 1.88, + "learning_rate": 1.2066503094444649e-08, + "loss": 0.6916, + "step": 51369 + }, + { + "epoch": 1.88, + "learning_rate": 1.2062841029772585e-08, + "loss": 0.6994, + "step": 51370 + }, + { + "epoch": 1.88, + "learning_rate": 1.2059178965100523e-08, + "loss": 0.6904, + "step": 51371 + }, + { + "epoch": 1.88, + "learning_rate": 1.2055516900428462e-08, + "loss": 0.7016, + "step": 51372 + }, + { + "epoch": 1.88, + "learning_rate": 1.2051854835756398e-08, + "loss": 0.7039, + "step": 51373 + }, + { + "epoch": 1.88, + "learning_rate": 1.2048192771084337e-08, + "loss": 0.7041, + "step": 51374 + }, + { + "epoch": 1.88, + "learning_rate": 1.2044530706412275e-08, + "loss": 0.6907, + "step": 51375 + }, + { + "epoch": 1.88, + "learning_rate": 1.2040868641740212e-08, + "loss": 0.6922, + "step": 51376 + }, + { + "epoch": 1.88, + "learning_rate": 1.203720657706815e-08, + "loss": 0.6936, + "step": 51377 + }, + { + "epoch": 1.88, + "learning_rate": 1.2033544512396088e-08, + "loss": 0.6924, + "step": 51378 + }, + { + "epoch": 1.88, + "learning_rate": 1.2029882447724026e-08, + "loss": 0.6955, + "step": 51379 + }, + { + "epoch": 1.88, + "learning_rate": 1.2026220383051965e-08, + "loss": 0.6954, + "step": 51380 + }, + { + "epoch": 1.88, + "learning_rate": 1.2022558318379903e-08, + "loss": 0.6838, + "step": 51381 + }, + { + "epoch": 1.88, + "learning_rate": 1.201889625370784e-08, + "loss": 0.6987, + "step": 51382 + }, + { + "epoch": 1.88, + "learning_rate": 1.2015234189035778e-08, + "loss": 0.697, + "step": 51383 + }, + { + "epoch": 1.88, + "learning_rate": 1.2011572124363716e-08, + "loss": 0.6941, + "step": 51384 + }, + { + "epoch": 1.88, + "learning_rate": 1.2007910059691653e-08, + "loss": 0.6903, + "step": 51385 + }, + { + "epoch": 1.88, + "learning_rate": 1.2004247995019591e-08, + "loss": 0.6944, + "step": 51386 + }, + { + "epoch": 1.88, + "learning_rate": 1.200058593034753e-08, + "loss": 0.693, + "step": 51387 + }, + { + "epoch": 1.88, + "learning_rate": 1.1996923865675468e-08, + "loss": 0.6999, + "step": 51388 + }, + { + "epoch": 1.88, + "learning_rate": 1.1993261801003406e-08, + "loss": 0.7001, + "step": 51389 + }, + { + "epoch": 1.88, + "learning_rate": 1.1989599736331344e-08, + "loss": 0.6928, + "step": 51390 + }, + { + "epoch": 1.88, + "learning_rate": 1.198593767165928e-08, + "loss": 0.6949, + "step": 51391 + }, + { + "epoch": 1.88, + "learning_rate": 1.1982275606987219e-08, + "loss": 0.6933, + "step": 51392 + }, + { + "epoch": 1.88, + "learning_rate": 1.1978613542315157e-08, + "loss": 0.6926, + "step": 51393 + }, + { + "epoch": 1.88, + "learning_rate": 1.1974951477643094e-08, + "loss": 0.6924, + "step": 51394 + }, + { + "epoch": 1.88, + "learning_rate": 1.1971289412971032e-08, + "loss": 0.6815, + "step": 51395 + }, + { + "epoch": 1.88, + "learning_rate": 1.196762734829897e-08, + "loss": 0.6889, + "step": 51396 + }, + { + "epoch": 1.88, + "learning_rate": 1.1963965283626909e-08, + "loss": 0.6919, + "step": 51397 + }, + { + "epoch": 1.88, + "learning_rate": 1.1960303218954847e-08, + "loss": 0.6976, + "step": 51398 + }, + { + "epoch": 1.88, + "learning_rate": 1.1956641154282784e-08, + "loss": 0.6942, + "step": 51399 + }, + { + "epoch": 1.88, + "learning_rate": 1.1952979089610722e-08, + "loss": 0.6979, + "step": 51400 + }, + { + "epoch": 1.88, + "learning_rate": 1.194931702493866e-08, + "loss": 0.6988, + "step": 51401 + }, + { + "epoch": 1.88, + "learning_rate": 1.1945654960266598e-08, + "loss": 0.693, + "step": 51402 + }, + { + "epoch": 1.88, + "learning_rate": 1.1941992895594535e-08, + "loss": 0.6958, + "step": 51403 + }, + { + "epoch": 1.88, + "learning_rate": 1.1938330830922473e-08, + "loss": 0.6936, + "step": 51404 + }, + { + "epoch": 1.88, + "learning_rate": 1.1934668766250412e-08, + "loss": 0.699, + "step": 51405 + }, + { + "epoch": 1.88, + "learning_rate": 1.193100670157835e-08, + "loss": 0.6956, + "step": 51406 + }, + { + "epoch": 1.88, + "learning_rate": 1.1927344636906288e-08, + "loss": 0.688, + "step": 51407 + }, + { + "epoch": 1.88, + "learning_rate": 1.1923682572234225e-08, + "loss": 0.6981, + "step": 51408 + }, + { + "epoch": 1.88, + "learning_rate": 1.1920020507562163e-08, + "loss": 0.6953, + "step": 51409 + }, + { + "epoch": 1.88, + "learning_rate": 1.1916358442890101e-08, + "loss": 0.6895, + "step": 51410 + }, + { + "epoch": 1.88, + "learning_rate": 1.1912696378218038e-08, + "loss": 0.6905, + "step": 51411 + }, + { + "epoch": 1.88, + "learning_rate": 1.1909034313545976e-08, + "loss": 0.6877, + "step": 51412 + }, + { + "epoch": 1.88, + "learning_rate": 1.1905372248873914e-08, + "loss": 0.7041, + "step": 51413 + }, + { + "epoch": 1.88, + "learning_rate": 1.1901710184201851e-08, + "loss": 0.6818, + "step": 51414 + }, + { + "epoch": 1.88, + "learning_rate": 1.1898048119529791e-08, + "loss": 0.7007, + "step": 51415 + }, + { + "epoch": 1.88, + "learning_rate": 1.189438605485773e-08, + "loss": 0.6902, + "step": 51416 + }, + { + "epoch": 1.88, + "learning_rate": 1.1890723990185666e-08, + "loss": 0.6933, + "step": 51417 + }, + { + "epoch": 1.88, + "learning_rate": 1.1887061925513604e-08, + "loss": 0.6997, + "step": 51418 + }, + { + "epoch": 1.88, + "learning_rate": 1.1883399860841542e-08, + "loss": 0.6866, + "step": 51419 + }, + { + "epoch": 1.88, + "learning_rate": 1.1879737796169479e-08, + "loss": 0.6921, + "step": 51420 + }, + { + "epoch": 1.88, + "learning_rate": 1.1876075731497417e-08, + "loss": 0.6953, + "step": 51421 + }, + { + "epoch": 1.88, + "learning_rate": 1.1872413666825356e-08, + "loss": 0.688, + "step": 51422 + }, + { + "epoch": 1.88, + "learning_rate": 1.1868751602153292e-08, + "loss": 0.6986, + "step": 51423 + }, + { + "epoch": 1.88, + "learning_rate": 1.186508953748123e-08, + "loss": 0.6876, + "step": 51424 + }, + { + "epoch": 1.88, + "learning_rate": 1.186142747280917e-08, + "loss": 0.6878, + "step": 51425 + }, + { + "epoch": 1.88, + "learning_rate": 1.1857765408137107e-08, + "loss": 0.6977, + "step": 51426 + }, + { + "epoch": 1.88, + "learning_rate": 1.1854103343465045e-08, + "loss": 0.6906, + "step": 51427 + }, + { + "epoch": 1.88, + "learning_rate": 1.1850441278792984e-08, + "loss": 0.6938, + "step": 51428 + }, + { + "epoch": 1.88, + "learning_rate": 1.184677921412092e-08, + "loss": 0.6829, + "step": 51429 + }, + { + "epoch": 1.88, + "learning_rate": 1.1843117149448859e-08, + "loss": 0.6988, + "step": 51430 + }, + { + "epoch": 1.88, + "learning_rate": 1.1839455084776797e-08, + "loss": 0.6773, + "step": 51431 + }, + { + "epoch": 1.88, + "learning_rate": 1.1835793020104733e-08, + "loss": 0.6876, + "step": 51432 + }, + { + "epoch": 1.88, + "learning_rate": 1.1832130955432672e-08, + "loss": 0.6902, + "step": 51433 + }, + { + "epoch": 1.88, + "learning_rate": 1.1828468890760612e-08, + "loss": 0.7032, + "step": 51434 + }, + { + "epoch": 1.88, + "learning_rate": 1.1824806826088548e-08, + "loss": 0.7104, + "step": 51435 + }, + { + "epoch": 1.88, + "learning_rate": 1.1821144761416487e-08, + "loss": 0.6844, + "step": 51436 + }, + { + "epoch": 1.88, + "learning_rate": 1.1817482696744425e-08, + "loss": 0.6941, + "step": 51437 + }, + { + "epoch": 1.88, + "learning_rate": 1.1813820632072361e-08, + "loss": 0.6967, + "step": 51438 + }, + { + "epoch": 1.88, + "learning_rate": 1.18101585674003e-08, + "loss": 0.6962, + "step": 51439 + }, + { + "epoch": 1.88, + "learning_rate": 1.1806496502728238e-08, + "loss": 0.6912, + "step": 51440 + }, + { + "epoch": 1.88, + "learning_rate": 1.1802834438056175e-08, + "loss": 0.6851, + "step": 51441 + }, + { + "epoch": 1.88, + "learning_rate": 1.1799172373384113e-08, + "loss": 0.6899, + "step": 51442 + }, + { + "epoch": 1.88, + "learning_rate": 1.1795510308712051e-08, + "loss": 0.6868, + "step": 51443 + }, + { + "epoch": 1.88, + "learning_rate": 1.179184824403999e-08, + "loss": 0.6975, + "step": 51444 + }, + { + "epoch": 1.88, + "learning_rate": 1.1788186179367928e-08, + "loss": 0.6947, + "step": 51445 + }, + { + "epoch": 1.88, + "learning_rate": 1.1784524114695866e-08, + "loss": 0.6935, + "step": 51446 + }, + { + "epoch": 1.88, + "learning_rate": 1.1780862050023803e-08, + "loss": 0.6989, + "step": 51447 + }, + { + "epoch": 1.88, + "learning_rate": 1.1777199985351741e-08, + "loss": 0.6773, + "step": 51448 + }, + { + "epoch": 1.88, + "learning_rate": 1.1773537920679679e-08, + "loss": 0.6993, + "step": 51449 + }, + { + "epoch": 1.88, + "learning_rate": 1.1769875856007616e-08, + "loss": 0.6914, + "step": 51450 + }, + { + "epoch": 1.88, + "learning_rate": 1.1766213791335554e-08, + "loss": 0.7086, + "step": 51451 + }, + { + "epoch": 1.88, + "learning_rate": 1.1762551726663492e-08, + "loss": 0.7043, + "step": 51452 + }, + { + "epoch": 1.88, + "learning_rate": 1.175888966199143e-08, + "loss": 0.6936, + "step": 51453 + }, + { + "epoch": 1.88, + "learning_rate": 1.1755227597319369e-08, + "loss": 0.6793, + "step": 51454 + }, + { + "epoch": 1.88, + "learning_rate": 1.1751565532647305e-08, + "loss": 0.6994, + "step": 51455 + }, + { + "epoch": 1.88, + "learning_rate": 1.1747903467975244e-08, + "loss": 0.6892, + "step": 51456 + }, + { + "epoch": 1.88, + "learning_rate": 1.1744241403303182e-08, + "loss": 0.6901, + "step": 51457 + }, + { + "epoch": 1.88, + "learning_rate": 1.1740579338631119e-08, + "loss": 0.6934, + "step": 51458 + }, + { + "epoch": 1.88, + "learning_rate": 1.1736917273959057e-08, + "loss": 0.6906, + "step": 51459 + }, + { + "epoch": 1.88, + "learning_rate": 1.1733255209286995e-08, + "loss": 0.7057, + "step": 51460 + }, + { + "epoch": 1.88, + "learning_rate": 1.1729593144614933e-08, + "loss": 0.6926, + "step": 51461 + }, + { + "epoch": 1.88, + "learning_rate": 1.1725931079942872e-08, + "loss": 0.694, + "step": 51462 + }, + { + "epoch": 1.88, + "learning_rate": 1.172226901527081e-08, + "loss": 0.7002, + "step": 51463 + }, + { + "epoch": 1.88, + "learning_rate": 1.1718606950598747e-08, + "loss": 0.6869, + "step": 51464 + }, + { + "epoch": 1.88, + "learning_rate": 1.1714944885926685e-08, + "loss": 0.6835, + "step": 51465 + }, + { + "epoch": 1.88, + "learning_rate": 1.1711282821254623e-08, + "loss": 0.6895, + "step": 51466 + }, + { + "epoch": 1.88, + "learning_rate": 1.170762075658256e-08, + "loss": 0.6882, + "step": 51467 + }, + { + "epoch": 1.88, + "learning_rate": 1.1703958691910498e-08, + "loss": 0.6961, + "step": 51468 + }, + { + "epoch": 1.88, + "learning_rate": 1.1700296627238436e-08, + "loss": 0.691, + "step": 51469 + }, + { + "epoch": 1.88, + "learning_rate": 1.1696634562566373e-08, + "loss": 0.6976, + "step": 51470 + }, + { + "epoch": 1.88, + "learning_rate": 1.1692972497894313e-08, + "loss": 0.6945, + "step": 51471 + }, + { + "epoch": 1.88, + "learning_rate": 1.1689310433222251e-08, + "loss": 0.6917, + "step": 51472 + }, + { + "epoch": 1.88, + "learning_rate": 1.1685648368550188e-08, + "loss": 0.6924, + "step": 51473 + }, + { + "epoch": 1.88, + "learning_rate": 1.1681986303878126e-08, + "loss": 0.7016, + "step": 51474 + }, + { + "epoch": 1.88, + "learning_rate": 1.1678324239206064e-08, + "loss": 0.699, + "step": 51475 + }, + { + "epoch": 1.88, + "learning_rate": 1.1674662174534001e-08, + "loss": 0.6919, + "step": 51476 + }, + { + "epoch": 1.88, + "learning_rate": 1.167100010986194e-08, + "loss": 0.6918, + "step": 51477 + }, + { + "epoch": 1.88, + "learning_rate": 1.1667338045189877e-08, + "loss": 0.6907, + "step": 51478 + }, + { + "epoch": 1.88, + "learning_rate": 1.1663675980517814e-08, + "loss": 0.6944, + "step": 51479 + }, + { + "epoch": 1.88, + "learning_rate": 1.1660013915845754e-08, + "loss": 0.7026, + "step": 51480 + }, + { + "epoch": 1.88, + "learning_rate": 1.1656351851173692e-08, + "loss": 0.6941, + "step": 51481 + }, + { + "epoch": 1.88, + "learning_rate": 1.1652689786501629e-08, + "loss": 0.6891, + "step": 51482 + }, + { + "epoch": 1.88, + "learning_rate": 1.1649027721829567e-08, + "loss": 0.6827, + "step": 51483 + }, + { + "epoch": 1.88, + "learning_rate": 1.1645365657157505e-08, + "loss": 0.6865, + "step": 51484 + }, + { + "epoch": 1.88, + "learning_rate": 1.1641703592485442e-08, + "loss": 0.6976, + "step": 51485 + }, + { + "epoch": 1.88, + "learning_rate": 1.163804152781338e-08, + "loss": 0.697, + "step": 51486 + }, + { + "epoch": 1.88, + "learning_rate": 1.1634379463141319e-08, + "loss": 0.6948, + "step": 51487 + }, + { + "epoch": 1.88, + "learning_rate": 1.1630717398469255e-08, + "loss": 0.6952, + "step": 51488 + }, + { + "epoch": 1.88, + "learning_rate": 1.1627055333797195e-08, + "loss": 0.6879, + "step": 51489 + }, + { + "epoch": 1.88, + "learning_rate": 1.1623393269125133e-08, + "loss": 0.6971, + "step": 51490 + }, + { + "epoch": 1.88, + "learning_rate": 1.161973120445307e-08, + "loss": 0.6923, + "step": 51491 + }, + { + "epoch": 1.88, + "learning_rate": 1.1616069139781008e-08, + "loss": 0.7036, + "step": 51492 + }, + { + "epoch": 1.88, + "learning_rate": 1.1612407075108947e-08, + "loss": 0.6851, + "step": 51493 + }, + { + "epoch": 1.88, + "learning_rate": 1.1608745010436883e-08, + "loss": 0.691, + "step": 51494 + }, + { + "epoch": 1.88, + "learning_rate": 1.1605082945764822e-08, + "loss": 0.6936, + "step": 51495 + }, + { + "epoch": 1.88, + "learning_rate": 1.160142088109276e-08, + "loss": 0.6861, + "step": 51496 + }, + { + "epoch": 1.88, + "learning_rate": 1.1597758816420696e-08, + "loss": 0.6851, + "step": 51497 + }, + { + "epoch": 1.88, + "learning_rate": 1.1594096751748636e-08, + "loss": 0.6967, + "step": 51498 + }, + { + "epoch": 1.88, + "learning_rate": 1.1590434687076573e-08, + "loss": 0.6868, + "step": 51499 + }, + { + "epoch": 1.88, + "learning_rate": 1.1586772622404511e-08, + "loss": 0.6905, + "step": 51500 + }, + { + "epoch": 1.88, + "learning_rate": 1.158311055773245e-08, + "loss": 0.6882, + "step": 51501 + }, + { + "epoch": 1.88, + "learning_rate": 1.1579448493060386e-08, + "loss": 0.7093, + "step": 51502 + }, + { + "epoch": 1.88, + "learning_rate": 1.1575786428388324e-08, + "loss": 0.6898, + "step": 51503 + }, + { + "epoch": 1.88, + "learning_rate": 1.1572124363716263e-08, + "loss": 0.6965, + "step": 51504 + }, + { + "epoch": 1.88, + "learning_rate": 1.1568462299044201e-08, + "loss": 0.6904, + "step": 51505 + }, + { + "epoch": 1.88, + "learning_rate": 1.1564800234372138e-08, + "loss": 0.6947, + "step": 51506 + }, + { + "epoch": 1.88, + "learning_rate": 1.1561138169700078e-08, + "loss": 0.703, + "step": 51507 + }, + { + "epoch": 1.88, + "learning_rate": 1.1557476105028014e-08, + "loss": 0.7036, + "step": 51508 + }, + { + "epoch": 1.88, + "learning_rate": 1.1553814040355952e-08, + "loss": 0.6919, + "step": 51509 + }, + { + "epoch": 1.88, + "learning_rate": 1.155015197568389e-08, + "loss": 0.7034, + "step": 51510 + }, + { + "epoch": 1.88, + "learning_rate": 1.1546489911011827e-08, + "loss": 0.68, + "step": 51511 + }, + { + "epoch": 1.88, + "learning_rate": 1.1542827846339766e-08, + "loss": 0.6877, + "step": 51512 + }, + { + "epoch": 1.88, + "learning_rate": 1.1539165781667704e-08, + "loss": 0.6926, + "step": 51513 + }, + { + "epoch": 1.88, + "learning_rate": 1.153550371699564e-08, + "loss": 0.6865, + "step": 51514 + }, + { + "epoch": 1.88, + "learning_rate": 1.1531841652323579e-08, + "loss": 0.7005, + "step": 51515 + }, + { + "epoch": 1.88, + "learning_rate": 1.1528179587651517e-08, + "loss": 0.6955, + "step": 51516 + }, + { + "epoch": 1.88, + "learning_rate": 1.1524517522979455e-08, + "loss": 0.6842, + "step": 51517 + }, + { + "epoch": 1.88, + "learning_rate": 1.1520855458307394e-08, + "loss": 0.6935, + "step": 51518 + }, + { + "epoch": 1.88, + "learning_rate": 1.1517193393635332e-08, + "loss": 0.7027, + "step": 51519 + }, + { + "epoch": 1.88, + "learning_rate": 1.1513531328963268e-08, + "loss": 0.6792, + "step": 51520 + }, + { + "epoch": 1.89, + "learning_rate": 1.1509869264291207e-08, + "loss": 0.6823, + "step": 51521 + }, + { + "epoch": 1.89, + "learning_rate": 1.1506207199619145e-08, + "loss": 0.6939, + "step": 51522 + }, + { + "epoch": 1.89, + "learning_rate": 1.1502545134947082e-08, + "loss": 0.6969, + "step": 51523 + }, + { + "epoch": 1.89, + "learning_rate": 1.149888307027502e-08, + "loss": 0.6919, + "step": 51524 + }, + { + "epoch": 1.89, + "learning_rate": 1.1495221005602958e-08, + "loss": 0.7083, + "step": 51525 + }, + { + "epoch": 1.89, + "learning_rate": 1.1491558940930896e-08, + "loss": 0.6911, + "step": 51526 + }, + { + "epoch": 1.89, + "learning_rate": 1.1487896876258835e-08, + "loss": 0.6949, + "step": 51527 + }, + { + "epoch": 1.89, + "learning_rate": 1.1484234811586773e-08, + "loss": 0.6926, + "step": 51528 + }, + { + "epoch": 1.89, + "learning_rate": 1.148057274691471e-08, + "loss": 0.7028, + "step": 51529 + }, + { + "epoch": 1.89, + "learning_rate": 1.1476910682242648e-08, + "loss": 0.693, + "step": 51530 + }, + { + "epoch": 1.89, + "learning_rate": 1.1473248617570586e-08, + "loss": 0.6893, + "step": 51531 + }, + { + "epoch": 1.89, + "learning_rate": 1.1469586552898523e-08, + "loss": 0.6924, + "step": 51532 + }, + { + "epoch": 1.89, + "learning_rate": 1.1465924488226461e-08, + "loss": 0.6936, + "step": 51533 + }, + { + "epoch": 1.89, + "learning_rate": 1.14622624235544e-08, + "loss": 0.6846, + "step": 51534 + }, + { + "epoch": 1.89, + "learning_rate": 1.1458600358882338e-08, + "loss": 0.7098, + "step": 51535 + }, + { + "epoch": 1.89, + "learning_rate": 1.1454938294210276e-08, + "loss": 0.6958, + "step": 51536 + }, + { + "epoch": 1.89, + "learning_rate": 1.1451276229538214e-08, + "loss": 0.6785, + "step": 51537 + }, + { + "epoch": 1.89, + "learning_rate": 1.144761416486615e-08, + "loss": 0.6824, + "step": 51538 + }, + { + "epoch": 1.89, + "learning_rate": 1.1443952100194089e-08, + "loss": 0.704, + "step": 51539 + }, + { + "epoch": 1.89, + "learning_rate": 1.1440290035522027e-08, + "loss": 0.6914, + "step": 51540 + }, + { + "epoch": 1.89, + "learning_rate": 1.1436627970849964e-08, + "loss": 0.6947, + "step": 51541 + }, + { + "epoch": 1.89, + "learning_rate": 1.1432965906177902e-08, + "loss": 0.6988, + "step": 51542 + }, + { + "epoch": 1.89, + "learning_rate": 1.142930384150584e-08, + "loss": 0.7002, + "step": 51543 + }, + { + "epoch": 1.89, + "learning_rate": 1.1425641776833779e-08, + "loss": 0.7039, + "step": 51544 + }, + { + "epoch": 1.89, + "learning_rate": 1.1421979712161717e-08, + "loss": 0.6871, + "step": 51545 + }, + { + "epoch": 1.89, + "learning_rate": 1.1418317647489654e-08, + "loss": 0.6828, + "step": 51546 + }, + { + "epoch": 1.89, + "learning_rate": 1.1414655582817592e-08, + "loss": 0.6829, + "step": 51547 + }, + { + "epoch": 1.89, + "learning_rate": 1.141099351814553e-08, + "loss": 0.6919, + "step": 51548 + }, + { + "epoch": 1.89, + "learning_rate": 1.1407331453473468e-08, + "loss": 0.7006, + "step": 51549 + }, + { + "epoch": 1.89, + "learning_rate": 1.1403669388801405e-08, + "loss": 0.7062, + "step": 51550 + }, + { + "epoch": 1.89, + "learning_rate": 1.1400007324129343e-08, + "loss": 0.7037, + "step": 51551 + }, + { + "epoch": 1.89, + "learning_rate": 1.1396345259457282e-08, + "loss": 0.7011, + "step": 51552 + }, + { + "epoch": 1.89, + "learning_rate": 1.139268319478522e-08, + "loss": 0.6902, + "step": 51553 + }, + { + "epoch": 1.89, + "learning_rate": 1.1389021130113158e-08, + "loss": 0.6975, + "step": 51554 + }, + { + "epoch": 1.89, + "learning_rate": 1.1385359065441095e-08, + "loss": 0.6921, + "step": 51555 + }, + { + "epoch": 1.89, + "learning_rate": 1.1381697000769033e-08, + "loss": 0.6771, + "step": 51556 + }, + { + "epoch": 1.89, + "learning_rate": 1.1378034936096971e-08, + "loss": 0.6889, + "step": 51557 + }, + { + "epoch": 1.89, + "learning_rate": 1.1374372871424908e-08, + "loss": 0.6869, + "step": 51558 + }, + { + "epoch": 1.89, + "learning_rate": 1.1370710806752846e-08, + "loss": 0.6906, + "step": 51559 + }, + { + "epoch": 1.89, + "learning_rate": 1.1367048742080785e-08, + "loss": 0.6999, + "step": 51560 + }, + { + "epoch": 1.89, + "learning_rate": 1.1363386677408721e-08, + "loss": 0.6906, + "step": 51561 + }, + { + "epoch": 1.89, + "learning_rate": 1.1359724612736661e-08, + "loss": 0.6793, + "step": 51562 + }, + { + "epoch": 1.89, + "learning_rate": 1.13560625480646e-08, + "loss": 0.694, + "step": 51563 + }, + { + "epoch": 1.89, + "learning_rate": 1.1352400483392536e-08, + "loss": 0.6974, + "step": 51564 + }, + { + "epoch": 1.89, + "learning_rate": 1.1348738418720474e-08, + "loss": 0.689, + "step": 51565 + }, + { + "epoch": 1.89, + "learning_rate": 1.1345076354048413e-08, + "loss": 0.699, + "step": 51566 + }, + { + "epoch": 1.89, + "learning_rate": 1.1341414289376349e-08, + "loss": 0.6918, + "step": 51567 + }, + { + "epoch": 1.89, + "learning_rate": 1.1337752224704287e-08, + "loss": 0.6875, + "step": 51568 + }, + { + "epoch": 1.89, + "learning_rate": 1.1334090160032226e-08, + "loss": 0.6986, + "step": 51569 + }, + { + "epoch": 1.89, + "learning_rate": 1.1330428095360162e-08, + "loss": 0.6955, + "step": 51570 + }, + { + "epoch": 1.89, + "learning_rate": 1.13267660306881e-08, + "loss": 0.7005, + "step": 51571 + }, + { + "epoch": 1.89, + "learning_rate": 1.132310396601604e-08, + "loss": 0.6943, + "step": 51572 + }, + { + "epoch": 1.89, + "learning_rate": 1.1319441901343977e-08, + "loss": 0.7005, + "step": 51573 + }, + { + "epoch": 1.89, + "learning_rate": 1.1315779836671915e-08, + "loss": 0.6926, + "step": 51574 + }, + { + "epoch": 1.89, + "learning_rate": 1.1312117771999854e-08, + "loss": 0.6809, + "step": 51575 + }, + { + "epoch": 1.89, + "learning_rate": 1.130845570732779e-08, + "loss": 0.6876, + "step": 51576 + }, + { + "epoch": 1.89, + "learning_rate": 1.1304793642655729e-08, + "loss": 0.6998, + "step": 51577 + }, + { + "epoch": 1.89, + "learning_rate": 1.1301131577983667e-08, + "loss": 0.6838, + "step": 51578 + }, + { + "epoch": 1.89, + "learning_rate": 1.1297469513311603e-08, + "loss": 0.7048, + "step": 51579 + }, + { + "epoch": 1.89, + "learning_rate": 1.1293807448639542e-08, + "loss": 0.6962, + "step": 51580 + }, + { + "epoch": 1.89, + "learning_rate": 1.1290145383967482e-08, + "loss": 0.6921, + "step": 51581 + }, + { + "epoch": 1.89, + "learning_rate": 1.1286483319295418e-08, + "loss": 0.6875, + "step": 51582 + }, + { + "epoch": 1.89, + "learning_rate": 1.1282821254623357e-08, + "loss": 0.6878, + "step": 51583 + }, + { + "epoch": 1.89, + "learning_rate": 1.1279159189951295e-08, + "loss": 0.6935, + "step": 51584 + }, + { + "epoch": 1.89, + "learning_rate": 1.1275497125279231e-08, + "loss": 0.6874, + "step": 51585 + }, + { + "epoch": 1.89, + "learning_rate": 1.127183506060717e-08, + "loss": 0.7089, + "step": 51586 + }, + { + "epoch": 1.89, + "learning_rate": 1.1268172995935108e-08, + "loss": 0.6979, + "step": 51587 + }, + { + "epoch": 1.89, + "learning_rate": 1.1264510931263045e-08, + "loss": 0.6966, + "step": 51588 + }, + { + "epoch": 1.89, + "learning_rate": 1.1260848866590983e-08, + "loss": 0.6973, + "step": 51589 + }, + { + "epoch": 1.89, + "learning_rate": 1.1257186801918921e-08, + "loss": 0.7039, + "step": 51590 + }, + { + "epoch": 1.89, + "learning_rate": 1.125352473724686e-08, + "loss": 0.6797, + "step": 51591 + }, + { + "epoch": 1.89, + "learning_rate": 1.1249862672574798e-08, + "loss": 0.7048, + "step": 51592 + }, + { + "epoch": 1.89, + "learning_rate": 1.1246200607902736e-08, + "loss": 0.6915, + "step": 51593 + }, + { + "epoch": 1.89, + "learning_rate": 1.1242538543230673e-08, + "loss": 0.6933, + "step": 51594 + }, + { + "epoch": 1.89, + "learning_rate": 1.1238876478558611e-08, + "loss": 0.6968, + "step": 51595 + }, + { + "epoch": 1.89, + "learning_rate": 1.123521441388655e-08, + "loss": 0.6893, + "step": 51596 + }, + { + "epoch": 1.89, + "learning_rate": 1.1231552349214486e-08, + "loss": 0.6879, + "step": 51597 + }, + { + "epoch": 1.89, + "learning_rate": 1.1227890284542424e-08, + "loss": 0.702, + "step": 51598 + }, + { + "epoch": 1.89, + "learning_rate": 1.1224228219870362e-08, + "loss": 0.6882, + "step": 51599 + }, + { + "epoch": 1.89, + "learning_rate": 1.12205661551983e-08, + "loss": 0.6943, + "step": 51600 + }, + { + "epoch": 1.89, + "learning_rate": 1.1216904090526239e-08, + "loss": 0.6939, + "step": 51601 + }, + { + "epoch": 1.89, + "learning_rate": 1.1213242025854176e-08, + "loss": 0.6976, + "step": 51602 + }, + { + "epoch": 1.89, + "learning_rate": 1.1209579961182114e-08, + "loss": 0.6908, + "step": 51603 + }, + { + "epoch": 1.89, + "learning_rate": 1.1205917896510052e-08, + "loss": 0.6953, + "step": 51604 + }, + { + "epoch": 1.89, + "learning_rate": 1.1202255831837989e-08, + "loss": 0.7059, + "step": 51605 + }, + { + "epoch": 1.89, + "learning_rate": 1.1198593767165927e-08, + "loss": 0.6935, + "step": 51606 + }, + { + "epoch": 1.89, + "learning_rate": 1.1194931702493865e-08, + "loss": 0.69, + "step": 51607 + }, + { + "epoch": 1.89, + "learning_rate": 1.1191269637821804e-08, + "loss": 0.6961, + "step": 51608 + }, + { + "epoch": 1.89, + "learning_rate": 1.1187607573149742e-08, + "loss": 0.6863, + "step": 51609 + }, + { + "epoch": 1.89, + "learning_rate": 1.118394550847768e-08, + "loss": 0.6919, + "step": 51610 + }, + { + "epoch": 1.89, + "learning_rate": 1.1180283443805617e-08, + "loss": 0.7011, + "step": 51611 + }, + { + "epoch": 1.89, + "learning_rate": 1.1176621379133555e-08, + "loss": 0.6944, + "step": 51612 + }, + { + "epoch": 1.89, + "learning_rate": 1.1172959314461493e-08, + "loss": 0.6947, + "step": 51613 + }, + { + "epoch": 1.89, + "learning_rate": 1.116929724978943e-08, + "loss": 0.6865, + "step": 51614 + }, + { + "epoch": 1.89, + "learning_rate": 1.1165635185117368e-08, + "loss": 0.6715, + "step": 51615 + }, + { + "epoch": 1.89, + "learning_rate": 1.1161973120445306e-08, + "loss": 0.6912, + "step": 51616 + }, + { + "epoch": 1.89, + "learning_rate": 1.1158311055773243e-08, + "loss": 0.6944, + "step": 51617 + }, + { + "epoch": 1.89, + "learning_rate": 1.1154648991101183e-08, + "loss": 0.6914, + "step": 51618 + }, + { + "epoch": 1.89, + "learning_rate": 1.1150986926429121e-08, + "loss": 0.6949, + "step": 51619 + }, + { + "epoch": 1.89, + "learning_rate": 1.1147324861757058e-08, + "loss": 0.6942, + "step": 51620 + }, + { + "epoch": 1.89, + "learning_rate": 1.1143662797084996e-08, + "loss": 0.6989, + "step": 51621 + }, + { + "epoch": 1.89, + "learning_rate": 1.1140000732412934e-08, + "loss": 0.6928, + "step": 51622 + }, + { + "epoch": 1.89, + "learning_rate": 1.1136338667740871e-08, + "loss": 0.6872, + "step": 51623 + }, + { + "epoch": 1.89, + "learning_rate": 1.113267660306881e-08, + "loss": 0.6758, + "step": 51624 + }, + { + "epoch": 1.89, + "learning_rate": 1.1129014538396748e-08, + "loss": 0.7008, + "step": 51625 + }, + { + "epoch": 1.89, + "learning_rate": 1.1125352473724684e-08, + "loss": 0.6921, + "step": 51626 + }, + { + "epoch": 1.89, + "learning_rate": 1.1121690409052624e-08, + "loss": 0.6807, + "step": 51627 + }, + { + "epoch": 1.89, + "learning_rate": 1.1118028344380562e-08, + "loss": 0.676, + "step": 51628 + }, + { + "epoch": 1.89, + "learning_rate": 1.1114366279708499e-08, + "loss": 0.6825, + "step": 51629 + }, + { + "epoch": 1.89, + "learning_rate": 1.1110704215036437e-08, + "loss": 0.7082, + "step": 51630 + }, + { + "epoch": 1.89, + "learning_rate": 1.1107042150364376e-08, + "loss": 0.6962, + "step": 51631 + }, + { + "epoch": 1.89, + "learning_rate": 1.1103380085692312e-08, + "loss": 0.6955, + "step": 51632 + }, + { + "epoch": 1.89, + "learning_rate": 1.109971802102025e-08, + "loss": 0.6912, + "step": 51633 + }, + { + "epoch": 1.89, + "learning_rate": 1.1096055956348189e-08, + "loss": 0.6893, + "step": 51634 + }, + { + "epoch": 1.89, + "learning_rate": 1.1092393891676125e-08, + "loss": 0.6885, + "step": 51635 + }, + { + "epoch": 1.89, + "learning_rate": 1.1088731827004065e-08, + "loss": 0.6969, + "step": 51636 + }, + { + "epoch": 1.89, + "learning_rate": 1.1085069762332004e-08, + "loss": 0.6846, + "step": 51637 + }, + { + "epoch": 1.89, + "learning_rate": 1.108140769765994e-08, + "loss": 0.6937, + "step": 51638 + }, + { + "epoch": 1.89, + "learning_rate": 1.1077745632987878e-08, + "loss": 0.6908, + "step": 51639 + }, + { + "epoch": 1.89, + "learning_rate": 1.1074083568315817e-08, + "loss": 0.7031, + "step": 51640 + }, + { + "epoch": 1.89, + "learning_rate": 1.1070421503643753e-08, + "loss": 0.6949, + "step": 51641 + }, + { + "epoch": 1.89, + "learning_rate": 1.1066759438971692e-08, + "loss": 0.7053, + "step": 51642 + }, + { + "epoch": 1.89, + "learning_rate": 1.106309737429963e-08, + "loss": 0.6984, + "step": 51643 + }, + { + "epoch": 1.89, + "learning_rate": 1.1059435309627566e-08, + "loss": 0.6871, + "step": 51644 + }, + { + "epoch": 1.89, + "learning_rate": 1.1055773244955506e-08, + "loss": 0.6881, + "step": 51645 + }, + { + "epoch": 1.89, + "learning_rate": 1.1052111180283443e-08, + "loss": 0.7034, + "step": 51646 + }, + { + "epoch": 1.89, + "learning_rate": 1.1048449115611381e-08, + "loss": 0.6881, + "step": 51647 + }, + { + "epoch": 1.89, + "learning_rate": 1.104478705093932e-08, + "loss": 0.6829, + "step": 51648 + }, + { + "epoch": 1.89, + "learning_rate": 1.1041124986267256e-08, + "loss": 0.6915, + "step": 51649 + }, + { + "epoch": 1.89, + "learning_rate": 1.1037462921595194e-08, + "loss": 0.6868, + "step": 51650 + }, + { + "epoch": 1.89, + "learning_rate": 1.1033800856923133e-08, + "loss": 0.6794, + "step": 51651 + }, + { + "epoch": 1.89, + "learning_rate": 1.1030138792251071e-08, + "loss": 0.6852, + "step": 51652 + }, + { + "epoch": 1.89, + "learning_rate": 1.1026476727579008e-08, + "loss": 0.6939, + "step": 51653 + }, + { + "epoch": 1.89, + "learning_rate": 1.1022814662906948e-08, + "loss": 0.6894, + "step": 51654 + }, + { + "epoch": 1.89, + "learning_rate": 1.1019152598234884e-08, + "loss": 0.6996, + "step": 51655 + }, + { + "epoch": 1.89, + "learning_rate": 1.1015490533562822e-08, + "loss": 0.7021, + "step": 51656 + }, + { + "epoch": 1.89, + "learning_rate": 1.101182846889076e-08, + "loss": 0.6979, + "step": 51657 + }, + { + "epoch": 1.89, + "learning_rate": 1.1008166404218697e-08, + "loss": 0.7032, + "step": 51658 + }, + { + "epoch": 1.89, + "learning_rate": 1.1004504339546636e-08, + "loss": 0.6823, + "step": 51659 + }, + { + "epoch": 1.89, + "learning_rate": 1.1000842274874574e-08, + "loss": 0.6948, + "step": 51660 + }, + { + "epoch": 1.89, + "learning_rate": 1.099718021020251e-08, + "loss": 0.7132, + "step": 51661 + }, + { + "epoch": 1.89, + "learning_rate": 1.0993518145530449e-08, + "loss": 0.6944, + "step": 51662 + }, + { + "epoch": 1.89, + "learning_rate": 1.0989856080858387e-08, + "loss": 0.6985, + "step": 51663 + }, + { + "epoch": 1.89, + "learning_rate": 1.0986194016186325e-08, + "loss": 0.6898, + "step": 51664 + }, + { + "epoch": 1.89, + "learning_rate": 1.0982531951514264e-08, + "loss": 0.6858, + "step": 51665 + }, + { + "epoch": 1.89, + "learning_rate": 1.0978869886842202e-08, + "loss": 0.6991, + "step": 51666 + }, + { + "epoch": 1.89, + "learning_rate": 1.0975207822170139e-08, + "loss": 0.7007, + "step": 51667 + }, + { + "epoch": 1.89, + "learning_rate": 1.0971545757498077e-08, + "loss": 0.6914, + "step": 51668 + }, + { + "epoch": 1.89, + "learning_rate": 1.0967883692826015e-08, + "loss": 0.6985, + "step": 51669 + }, + { + "epoch": 1.89, + "learning_rate": 1.0964221628153952e-08, + "loss": 0.6961, + "step": 51670 + }, + { + "epoch": 1.89, + "learning_rate": 1.096055956348189e-08, + "loss": 0.6881, + "step": 51671 + }, + { + "epoch": 1.89, + "learning_rate": 1.0956897498809828e-08, + "loss": 0.7005, + "step": 51672 + }, + { + "epoch": 1.89, + "learning_rate": 1.0953235434137767e-08, + "loss": 0.68, + "step": 51673 + }, + { + "epoch": 1.89, + "learning_rate": 1.0949573369465705e-08, + "loss": 0.6905, + "step": 51674 + }, + { + "epoch": 1.89, + "learning_rate": 1.0945911304793643e-08, + "loss": 0.6997, + "step": 51675 + }, + { + "epoch": 1.89, + "learning_rate": 1.094224924012158e-08, + "loss": 0.6881, + "step": 51676 + }, + { + "epoch": 1.89, + "learning_rate": 1.0938587175449518e-08, + "loss": 0.7137, + "step": 51677 + }, + { + "epoch": 1.89, + "learning_rate": 1.0934925110777456e-08, + "loss": 0.6728, + "step": 51678 + }, + { + "epoch": 1.89, + "learning_rate": 1.0931263046105393e-08, + "loss": 0.6989, + "step": 51679 + }, + { + "epoch": 1.89, + "learning_rate": 1.0927600981433331e-08, + "loss": 0.699, + "step": 51680 + }, + { + "epoch": 1.89, + "learning_rate": 1.092393891676127e-08, + "loss": 0.6961, + "step": 51681 + }, + { + "epoch": 1.89, + "learning_rate": 1.0920276852089208e-08, + "loss": 0.7112, + "step": 51682 + }, + { + "epoch": 1.89, + "learning_rate": 1.0916614787417146e-08, + "loss": 0.6948, + "step": 51683 + }, + { + "epoch": 1.89, + "learning_rate": 1.0912952722745084e-08, + "loss": 0.6829, + "step": 51684 + }, + { + "epoch": 1.89, + "learning_rate": 1.0909290658073021e-08, + "loss": 0.6985, + "step": 51685 + }, + { + "epoch": 1.89, + "learning_rate": 1.0905628593400959e-08, + "loss": 0.6957, + "step": 51686 + }, + { + "epoch": 1.89, + "learning_rate": 1.0901966528728897e-08, + "loss": 0.7027, + "step": 51687 + }, + { + "epoch": 1.89, + "learning_rate": 1.0898304464056834e-08, + "loss": 0.6929, + "step": 51688 + }, + { + "epoch": 1.89, + "learning_rate": 1.0894642399384772e-08, + "loss": 0.6761, + "step": 51689 + }, + { + "epoch": 1.89, + "learning_rate": 1.089098033471271e-08, + "loss": 0.6946, + "step": 51690 + }, + { + "epoch": 1.89, + "learning_rate": 1.0887318270040649e-08, + "loss": 0.6976, + "step": 51691 + }, + { + "epoch": 1.89, + "learning_rate": 1.0883656205368587e-08, + "loss": 0.6858, + "step": 51692 + }, + { + "epoch": 1.89, + "learning_rate": 1.0879994140696524e-08, + "loss": 0.7018, + "step": 51693 + }, + { + "epoch": 1.89, + "learning_rate": 1.0876332076024462e-08, + "loss": 0.6878, + "step": 51694 + }, + { + "epoch": 1.89, + "learning_rate": 1.08726700113524e-08, + "loss": 0.6995, + "step": 51695 + }, + { + "epoch": 1.89, + "learning_rate": 1.0869007946680339e-08, + "loss": 0.6974, + "step": 51696 + }, + { + "epoch": 1.89, + "learning_rate": 1.0865345882008275e-08, + "loss": 0.6878, + "step": 51697 + }, + { + "epoch": 1.89, + "learning_rate": 1.0861683817336213e-08, + "loss": 0.6908, + "step": 51698 + }, + { + "epoch": 1.89, + "learning_rate": 1.0858021752664152e-08, + "loss": 0.7137, + "step": 51699 + }, + { + "epoch": 1.89, + "learning_rate": 1.085435968799209e-08, + "loss": 0.6914, + "step": 51700 + }, + { + "epoch": 1.89, + "learning_rate": 1.0850697623320028e-08, + "loss": 0.698, + "step": 51701 + }, + { + "epoch": 1.89, + "learning_rate": 1.0847035558647965e-08, + "loss": 0.6897, + "step": 51702 + }, + { + "epoch": 1.89, + "learning_rate": 1.0843373493975903e-08, + "loss": 0.6812, + "step": 51703 + }, + { + "epoch": 1.89, + "learning_rate": 1.0839711429303841e-08, + "loss": 0.6922, + "step": 51704 + }, + { + "epoch": 1.89, + "learning_rate": 1.0836049364631778e-08, + "loss": 0.6835, + "step": 51705 + }, + { + "epoch": 1.89, + "learning_rate": 1.0832387299959716e-08, + "loss": 0.6979, + "step": 51706 + }, + { + "epoch": 1.89, + "learning_rate": 1.0828725235287655e-08, + "loss": 0.6976, + "step": 51707 + }, + { + "epoch": 1.89, + "learning_rate": 1.0825063170615591e-08, + "loss": 0.6886, + "step": 51708 + }, + { + "epoch": 1.89, + "learning_rate": 1.082140110594353e-08, + "loss": 0.7002, + "step": 51709 + }, + { + "epoch": 1.89, + "learning_rate": 1.081773904127147e-08, + "loss": 0.6974, + "step": 51710 + }, + { + "epoch": 1.89, + "learning_rate": 1.0814076976599406e-08, + "loss": 0.6931, + "step": 51711 + }, + { + "epoch": 1.89, + "learning_rate": 1.0810414911927344e-08, + "loss": 0.6933, + "step": 51712 + }, + { + "epoch": 1.89, + "learning_rate": 1.0806752847255283e-08, + "loss": 0.7021, + "step": 51713 + }, + { + "epoch": 1.89, + "learning_rate": 1.080309078258322e-08, + "loss": 0.6874, + "step": 51714 + }, + { + "epoch": 1.89, + "learning_rate": 1.0799428717911157e-08, + "loss": 0.703, + "step": 51715 + }, + { + "epoch": 1.89, + "learning_rate": 1.0795766653239096e-08, + "loss": 0.6911, + "step": 51716 + }, + { + "epoch": 1.89, + "learning_rate": 1.0792104588567032e-08, + "loss": 0.71, + "step": 51717 + }, + { + "epoch": 1.89, + "learning_rate": 1.078844252389497e-08, + "loss": 0.6894, + "step": 51718 + }, + { + "epoch": 1.89, + "learning_rate": 1.078478045922291e-08, + "loss": 0.6971, + "step": 51719 + }, + { + "epoch": 1.89, + "learning_rate": 1.0781118394550847e-08, + "loss": 0.687, + "step": 51720 + }, + { + "epoch": 1.89, + "learning_rate": 1.0777456329878785e-08, + "loss": 0.6955, + "step": 51721 + }, + { + "epoch": 1.89, + "learning_rate": 1.0773794265206724e-08, + "loss": 0.6934, + "step": 51722 + }, + { + "epoch": 1.89, + "learning_rate": 1.077013220053466e-08, + "loss": 0.6914, + "step": 51723 + }, + { + "epoch": 1.89, + "learning_rate": 1.0766470135862599e-08, + "loss": 0.6931, + "step": 51724 + }, + { + "epoch": 1.89, + "learning_rate": 1.0762808071190537e-08, + "loss": 0.6933, + "step": 51725 + }, + { + "epoch": 1.89, + "learning_rate": 1.0759146006518474e-08, + "loss": 0.7234, + "step": 51726 + }, + { + "epoch": 1.89, + "learning_rate": 1.0755483941846412e-08, + "loss": 0.6866, + "step": 51727 + }, + { + "epoch": 1.89, + "learning_rate": 1.0751821877174352e-08, + "loss": 0.6925, + "step": 51728 + }, + { + "epoch": 1.89, + "learning_rate": 1.0748159812502288e-08, + "loss": 0.6919, + "step": 51729 + }, + { + "epoch": 1.89, + "learning_rate": 1.0744497747830227e-08, + "loss": 0.6877, + "step": 51730 + }, + { + "epoch": 1.89, + "learning_rate": 1.0740835683158165e-08, + "loss": 0.697, + "step": 51731 + }, + { + "epoch": 1.89, + "learning_rate": 1.0737173618486102e-08, + "loss": 0.6863, + "step": 51732 + }, + { + "epoch": 1.89, + "learning_rate": 1.073351155381404e-08, + "loss": 0.7056, + "step": 51733 + }, + { + "epoch": 1.89, + "learning_rate": 1.0729849489141978e-08, + "loss": 0.6887, + "step": 51734 + }, + { + "epoch": 1.89, + "learning_rate": 1.0726187424469915e-08, + "loss": 0.6966, + "step": 51735 + }, + { + "epoch": 1.89, + "learning_rate": 1.0722525359797853e-08, + "loss": 0.6899, + "step": 51736 + }, + { + "epoch": 1.89, + "learning_rate": 1.0718863295125791e-08, + "loss": 0.6873, + "step": 51737 + }, + { + "epoch": 1.89, + "learning_rate": 1.071520123045373e-08, + "loss": 0.6804, + "step": 51738 + }, + { + "epoch": 1.89, + "learning_rate": 1.0711539165781668e-08, + "loss": 0.6845, + "step": 51739 + }, + { + "epoch": 1.89, + "learning_rate": 1.0707877101109606e-08, + "loss": 0.6984, + "step": 51740 + }, + { + "epoch": 1.89, + "learning_rate": 1.0704215036437543e-08, + "loss": 0.694, + "step": 51741 + }, + { + "epoch": 1.89, + "learning_rate": 1.0700552971765481e-08, + "loss": 0.6882, + "step": 51742 + }, + { + "epoch": 1.89, + "learning_rate": 1.069689090709342e-08, + "loss": 0.6927, + "step": 51743 + }, + { + "epoch": 1.89, + "learning_rate": 1.0693228842421356e-08, + "loss": 0.6888, + "step": 51744 + }, + { + "epoch": 1.89, + "learning_rate": 1.0689566777749294e-08, + "loss": 0.6814, + "step": 51745 + }, + { + "epoch": 1.89, + "learning_rate": 1.0685904713077232e-08, + "loss": 0.6667, + "step": 51746 + }, + { + "epoch": 1.89, + "learning_rate": 1.068224264840517e-08, + "loss": 0.6895, + "step": 51747 + }, + { + "epoch": 1.89, + "learning_rate": 1.0678580583733109e-08, + "loss": 0.6965, + "step": 51748 + }, + { + "epoch": 1.89, + "learning_rate": 1.0674918519061046e-08, + "loss": 0.6935, + "step": 51749 + }, + { + "epoch": 1.89, + "learning_rate": 1.0671256454388984e-08, + "loss": 0.686, + "step": 51750 + }, + { + "epoch": 1.89, + "learning_rate": 1.0667594389716922e-08, + "loss": 0.6872, + "step": 51751 + }, + { + "epoch": 1.89, + "learning_rate": 1.0663932325044859e-08, + "loss": 0.6941, + "step": 51752 + }, + { + "epoch": 1.89, + "learning_rate": 1.0660270260372797e-08, + "loss": 0.6955, + "step": 51753 + }, + { + "epoch": 1.89, + "learning_rate": 1.0656608195700735e-08, + "loss": 0.694, + "step": 51754 + }, + { + "epoch": 1.89, + "learning_rate": 1.0652946131028674e-08, + "loss": 0.6869, + "step": 51755 + }, + { + "epoch": 1.89, + "learning_rate": 1.0649284066356612e-08, + "loss": 0.7008, + "step": 51756 + }, + { + "epoch": 1.89, + "learning_rate": 1.064562200168455e-08, + "loss": 0.6863, + "step": 51757 + }, + { + "epoch": 1.89, + "learning_rate": 1.0641959937012487e-08, + "loss": 0.6906, + "step": 51758 + }, + { + "epoch": 1.89, + "learning_rate": 1.0638297872340425e-08, + "loss": 0.6843, + "step": 51759 + }, + { + "epoch": 1.89, + "learning_rate": 1.0634635807668363e-08, + "loss": 0.6884, + "step": 51760 + }, + { + "epoch": 1.89, + "learning_rate": 1.06309737429963e-08, + "loss": 0.6889, + "step": 51761 + }, + { + "epoch": 1.89, + "learning_rate": 1.0627311678324238e-08, + "loss": 0.7042, + "step": 51762 + }, + { + "epoch": 1.89, + "learning_rate": 1.0623649613652176e-08, + "loss": 0.6976, + "step": 51763 + }, + { + "epoch": 1.89, + "learning_rate": 1.0619987548980113e-08, + "loss": 0.6907, + "step": 51764 + }, + { + "epoch": 1.89, + "learning_rate": 1.0616325484308053e-08, + "loss": 0.6756, + "step": 51765 + }, + { + "epoch": 1.89, + "learning_rate": 1.0612663419635991e-08, + "loss": 0.692, + "step": 51766 + }, + { + "epoch": 1.89, + "learning_rate": 1.0609001354963928e-08, + "loss": 0.6956, + "step": 51767 + }, + { + "epoch": 1.89, + "learning_rate": 1.0605339290291866e-08, + "loss": 0.698, + "step": 51768 + }, + { + "epoch": 1.89, + "learning_rate": 1.0601677225619804e-08, + "loss": 0.6778, + "step": 51769 + }, + { + "epoch": 1.89, + "learning_rate": 1.0598015160947741e-08, + "loss": 0.7015, + "step": 51770 + }, + { + "epoch": 1.89, + "learning_rate": 1.059435309627568e-08, + "loss": 0.7029, + "step": 51771 + }, + { + "epoch": 1.89, + "learning_rate": 1.0590691031603618e-08, + "loss": 0.7017, + "step": 51772 + }, + { + "epoch": 1.89, + "learning_rate": 1.0587028966931554e-08, + "loss": 0.7021, + "step": 51773 + }, + { + "epoch": 1.89, + "learning_rate": 1.0583366902259494e-08, + "loss": 0.6944, + "step": 51774 + }, + { + "epoch": 1.89, + "learning_rate": 1.0579704837587432e-08, + "loss": 0.6968, + "step": 51775 + }, + { + "epoch": 1.89, + "learning_rate": 1.0576042772915369e-08, + "loss": 0.6883, + "step": 51776 + }, + { + "epoch": 1.89, + "learning_rate": 1.0572380708243307e-08, + "loss": 0.6997, + "step": 51777 + }, + { + "epoch": 1.89, + "learning_rate": 1.0568718643571246e-08, + "loss": 0.7148, + "step": 51778 + }, + { + "epoch": 1.89, + "learning_rate": 1.0565056578899182e-08, + "loss": 0.683, + "step": 51779 + }, + { + "epoch": 1.89, + "learning_rate": 1.056139451422712e-08, + "loss": 0.6945, + "step": 51780 + }, + { + "epoch": 1.89, + "learning_rate": 1.0557732449555059e-08, + "loss": 0.6801, + "step": 51781 + }, + { + "epoch": 1.89, + "learning_rate": 1.0554070384882995e-08, + "loss": 0.7062, + "step": 51782 + }, + { + "epoch": 1.89, + "learning_rate": 1.0550408320210935e-08, + "loss": 0.6918, + "step": 51783 + }, + { + "epoch": 1.89, + "learning_rate": 1.0546746255538874e-08, + "loss": 0.6741, + "step": 51784 + }, + { + "epoch": 1.89, + "learning_rate": 1.054308419086681e-08, + "loss": 0.6932, + "step": 51785 + }, + { + "epoch": 1.89, + "learning_rate": 1.0539422126194748e-08, + "loss": 0.6881, + "step": 51786 + }, + { + "epoch": 1.89, + "learning_rate": 1.0535760061522687e-08, + "loss": 0.6928, + "step": 51787 + }, + { + "epoch": 1.89, + "learning_rate": 1.0532097996850623e-08, + "loss": 0.6828, + "step": 51788 + }, + { + "epoch": 1.89, + "learning_rate": 1.0528435932178562e-08, + "loss": 0.6988, + "step": 51789 + }, + { + "epoch": 1.89, + "learning_rate": 1.05247738675065e-08, + "loss": 0.6812, + "step": 51790 + }, + { + "epoch": 1.89, + "learning_rate": 1.0521111802834437e-08, + "loss": 0.6917, + "step": 51791 + }, + { + "epoch": 1.89, + "learning_rate": 1.0517449738162376e-08, + "loss": 0.6955, + "step": 51792 + }, + { + "epoch": 1.89, + "learning_rate": 1.0513787673490313e-08, + "loss": 0.6871, + "step": 51793 + }, + { + "epoch": 1.89, + "learning_rate": 1.0510125608818251e-08, + "loss": 0.6949, + "step": 51794 + }, + { + "epoch": 1.9, + "learning_rate": 1.050646354414619e-08, + "loss": 0.6876, + "step": 51795 + }, + { + "epoch": 1.9, + "learning_rate": 1.0502801479474126e-08, + "loss": 0.7042, + "step": 51796 + }, + { + "epoch": 1.9, + "learning_rate": 1.0499139414802065e-08, + "loss": 0.695, + "step": 51797 + }, + { + "epoch": 1.9, + "learning_rate": 1.0495477350130003e-08, + "loss": 0.6871, + "step": 51798 + }, + { + "epoch": 1.9, + "learning_rate": 1.0491815285457941e-08, + "loss": 0.6938, + "step": 51799 + }, + { + "epoch": 1.9, + "learning_rate": 1.0488153220785878e-08, + "loss": 0.6758, + "step": 51800 + }, + { + "epoch": 1.9, + "learning_rate": 1.0484491156113818e-08, + "loss": 0.6969, + "step": 51801 + }, + { + "epoch": 1.9, + "learning_rate": 1.0480829091441754e-08, + "loss": 0.696, + "step": 51802 + }, + { + "epoch": 1.9, + "learning_rate": 1.0477167026769693e-08, + "loss": 0.6927, + "step": 51803 + }, + { + "epoch": 1.9, + "learning_rate": 1.047350496209763e-08, + "loss": 0.693, + "step": 51804 + }, + { + "epoch": 1.9, + "learning_rate": 1.0469842897425567e-08, + "loss": 0.6907, + "step": 51805 + }, + { + "epoch": 1.9, + "learning_rate": 1.0466180832753506e-08, + "loss": 0.6762, + "step": 51806 + }, + { + "epoch": 1.9, + "learning_rate": 1.0462518768081444e-08, + "loss": 0.6909, + "step": 51807 + }, + { + "epoch": 1.9, + "learning_rate": 1.045885670340938e-08, + "loss": 0.702, + "step": 51808 + }, + { + "epoch": 1.9, + "learning_rate": 1.0455194638737319e-08, + "loss": 0.7018, + "step": 51809 + }, + { + "epoch": 1.9, + "learning_rate": 1.0451532574065257e-08, + "loss": 0.7019, + "step": 51810 + }, + { + "epoch": 1.9, + "learning_rate": 1.0447870509393195e-08, + "loss": 0.7017, + "step": 51811 + }, + { + "epoch": 1.9, + "learning_rate": 1.0444208444721134e-08, + "loss": 0.7009, + "step": 51812 + }, + { + "epoch": 1.9, + "learning_rate": 1.0440546380049072e-08, + "loss": 0.6947, + "step": 51813 + }, + { + "epoch": 1.9, + "learning_rate": 1.0436884315377009e-08, + "loss": 0.6927, + "step": 51814 + }, + { + "epoch": 1.9, + "learning_rate": 1.0433222250704947e-08, + "loss": 0.6911, + "step": 51815 + }, + { + "epoch": 1.9, + "learning_rate": 1.0429560186032885e-08, + "loss": 0.6931, + "step": 51816 + }, + { + "epoch": 1.9, + "learning_rate": 1.0425898121360822e-08, + "loss": 0.6969, + "step": 51817 + }, + { + "epoch": 1.9, + "learning_rate": 1.042223605668876e-08, + "loss": 0.7154, + "step": 51818 + }, + { + "epoch": 1.9, + "learning_rate": 1.0418573992016698e-08, + "loss": 0.6937, + "step": 51819 + }, + { + "epoch": 1.9, + "learning_rate": 1.0414911927344637e-08, + "loss": 0.6861, + "step": 51820 + }, + { + "epoch": 1.9, + "learning_rate": 1.0411249862672575e-08, + "loss": 0.692, + "step": 51821 + }, + { + "epoch": 1.9, + "learning_rate": 1.0407587798000513e-08, + "loss": 0.6869, + "step": 51822 + }, + { + "epoch": 1.9, + "learning_rate": 1.040392573332845e-08, + "loss": 0.6916, + "step": 51823 + }, + { + "epoch": 1.9, + "learning_rate": 1.0400263668656388e-08, + "loss": 0.6916, + "step": 51824 + }, + { + "epoch": 1.9, + "learning_rate": 1.0396601603984326e-08, + "loss": 0.6963, + "step": 51825 + }, + { + "epoch": 1.9, + "learning_rate": 1.0392939539312263e-08, + "loss": 0.6917, + "step": 51826 + }, + { + "epoch": 1.9, + "learning_rate": 1.0389277474640201e-08, + "loss": 0.6932, + "step": 51827 + }, + { + "epoch": 1.9, + "learning_rate": 1.038561540996814e-08, + "loss": 0.6927, + "step": 51828 + }, + { + "epoch": 1.9, + "learning_rate": 1.0381953345296078e-08, + "loss": 0.6975, + "step": 51829 + }, + { + "epoch": 1.9, + "learning_rate": 1.0378291280624016e-08, + "loss": 0.6839, + "step": 51830 + }, + { + "epoch": 1.9, + "learning_rate": 1.0374629215951954e-08, + "loss": 0.7061, + "step": 51831 + }, + { + "epoch": 1.9, + "learning_rate": 1.0370967151279891e-08, + "loss": 0.7003, + "step": 51832 + }, + { + "epoch": 1.9, + "learning_rate": 1.036730508660783e-08, + "loss": 0.6883, + "step": 51833 + }, + { + "epoch": 1.9, + "learning_rate": 1.0363643021935767e-08, + "loss": 0.6916, + "step": 51834 + }, + { + "epoch": 1.9, + "learning_rate": 1.0359980957263704e-08, + "loss": 0.6912, + "step": 51835 + }, + { + "epoch": 1.9, + "learning_rate": 1.0356318892591642e-08, + "loss": 0.7008, + "step": 51836 + }, + { + "epoch": 1.9, + "learning_rate": 1.035265682791958e-08, + "loss": 0.6943, + "step": 51837 + }, + { + "epoch": 1.9, + "learning_rate": 1.0348994763247519e-08, + "loss": 0.699, + "step": 51838 + }, + { + "epoch": 1.9, + "learning_rate": 1.0345332698575457e-08, + "loss": 0.691, + "step": 51839 + }, + { + "epoch": 1.9, + "learning_rate": 1.0341670633903394e-08, + "loss": 0.6997, + "step": 51840 + }, + { + "epoch": 1.9, + "learning_rate": 1.0338008569231332e-08, + "loss": 0.6841, + "step": 51841 + }, + { + "epoch": 1.9, + "learning_rate": 1.033434650455927e-08, + "loss": 0.6867, + "step": 51842 + }, + { + "epoch": 1.9, + "learning_rate": 1.0330684439887209e-08, + "loss": 0.6964, + "step": 51843 + }, + { + "epoch": 1.9, + "learning_rate": 1.0327022375215145e-08, + "loss": 0.6947, + "step": 51844 + }, + { + "epoch": 1.9, + "learning_rate": 1.0323360310543084e-08, + "loss": 0.6973, + "step": 51845 + }, + { + "epoch": 1.9, + "learning_rate": 1.0319698245871022e-08, + "loss": 0.6946, + "step": 51846 + }, + { + "epoch": 1.9, + "learning_rate": 1.031603618119896e-08, + "loss": 0.6981, + "step": 51847 + }, + { + "epoch": 1.9, + "learning_rate": 1.0312374116526898e-08, + "loss": 0.6965, + "step": 51848 + }, + { + "epoch": 1.9, + "learning_rate": 1.0308712051854835e-08, + "loss": 0.7029, + "step": 51849 + }, + { + "epoch": 1.9, + "learning_rate": 1.0305049987182773e-08, + "loss": 0.7005, + "step": 51850 + }, + { + "epoch": 1.9, + "learning_rate": 1.0301387922510712e-08, + "loss": 0.694, + "step": 51851 + }, + { + "epoch": 1.9, + "learning_rate": 1.0297725857838648e-08, + "loss": 0.6906, + "step": 51852 + }, + { + "epoch": 1.9, + "learning_rate": 1.0294063793166586e-08, + "loss": 0.692, + "step": 51853 + }, + { + "epoch": 1.9, + "learning_rate": 1.0290401728494525e-08, + "loss": 0.6904, + "step": 51854 + }, + { + "epoch": 1.9, + "learning_rate": 1.0286739663822461e-08, + "loss": 0.7017, + "step": 51855 + }, + { + "epoch": 1.9, + "learning_rate": 1.02830775991504e-08, + "loss": 0.6896, + "step": 51856 + }, + { + "epoch": 1.9, + "learning_rate": 1.027941553447834e-08, + "loss": 0.6817, + "step": 51857 + }, + { + "epoch": 1.9, + "learning_rate": 1.0275753469806276e-08, + "loss": 0.6966, + "step": 51858 + }, + { + "epoch": 1.9, + "learning_rate": 1.0272091405134214e-08, + "loss": 0.6988, + "step": 51859 + }, + { + "epoch": 1.9, + "learning_rate": 1.0268429340462153e-08, + "loss": 0.6866, + "step": 51860 + }, + { + "epoch": 1.9, + "learning_rate": 1.026476727579009e-08, + "loss": 0.6821, + "step": 51861 + }, + { + "epoch": 1.9, + "learning_rate": 1.0261105211118028e-08, + "loss": 0.687, + "step": 51862 + }, + { + "epoch": 1.9, + "learning_rate": 1.0257443146445966e-08, + "loss": 0.6909, + "step": 51863 + }, + { + "epoch": 1.9, + "learning_rate": 1.0253781081773902e-08, + "loss": 0.6989, + "step": 51864 + }, + { + "epoch": 1.9, + "learning_rate": 1.025011901710184e-08, + "loss": 0.6856, + "step": 51865 + }, + { + "epoch": 1.9, + "learning_rate": 1.024645695242978e-08, + "loss": 0.6993, + "step": 51866 + }, + { + "epoch": 1.9, + "learning_rate": 1.0242794887757717e-08, + "loss": 0.686, + "step": 51867 + }, + { + "epoch": 1.9, + "learning_rate": 1.0239132823085656e-08, + "loss": 0.6932, + "step": 51868 + }, + { + "epoch": 1.9, + "learning_rate": 1.0235470758413594e-08, + "loss": 0.6915, + "step": 51869 + }, + { + "epoch": 1.9, + "learning_rate": 1.023180869374153e-08, + "loss": 0.7005, + "step": 51870 + }, + { + "epoch": 1.9, + "learning_rate": 1.0228146629069469e-08, + "loss": 0.6967, + "step": 51871 + }, + { + "epoch": 1.9, + "learning_rate": 1.0224484564397407e-08, + "loss": 0.6996, + "step": 51872 + }, + { + "epoch": 1.9, + "learning_rate": 1.0220822499725344e-08, + "loss": 0.6812, + "step": 51873 + }, + { + "epoch": 1.9, + "learning_rate": 1.0217160435053282e-08, + "loss": 0.694, + "step": 51874 + }, + { + "epoch": 1.9, + "learning_rate": 1.0213498370381222e-08, + "loss": 0.6994, + "step": 51875 + }, + { + "epoch": 1.9, + "learning_rate": 1.0209836305709158e-08, + "loss": 0.6829, + "step": 51876 + }, + { + "epoch": 1.9, + "learning_rate": 1.0206174241037097e-08, + "loss": 0.6932, + "step": 51877 + }, + { + "epoch": 1.9, + "learning_rate": 1.0202512176365035e-08, + "loss": 0.6912, + "step": 51878 + }, + { + "epoch": 1.9, + "learning_rate": 1.0198850111692972e-08, + "loss": 0.6893, + "step": 51879 + }, + { + "epoch": 1.9, + "learning_rate": 1.019518804702091e-08, + "loss": 0.6954, + "step": 51880 + }, + { + "epoch": 1.9, + "learning_rate": 1.0191525982348848e-08, + "loss": 0.6862, + "step": 51881 + }, + { + "epoch": 1.9, + "learning_rate": 1.0187863917676785e-08, + "loss": 0.6961, + "step": 51882 + }, + { + "epoch": 1.9, + "learning_rate": 1.0184201853004723e-08, + "loss": 0.6848, + "step": 51883 + }, + { + "epoch": 1.9, + "learning_rate": 1.0180539788332661e-08, + "loss": 0.7019, + "step": 51884 + }, + { + "epoch": 1.9, + "learning_rate": 1.01768777236606e-08, + "loss": 0.714, + "step": 51885 + }, + { + "epoch": 1.9, + "learning_rate": 1.0173215658988538e-08, + "loss": 0.6912, + "step": 51886 + }, + { + "epoch": 1.9, + "learning_rate": 1.0169553594316476e-08, + "loss": 0.6861, + "step": 51887 + }, + { + "epoch": 1.9, + "learning_rate": 1.0165891529644413e-08, + "loss": 0.6925, + "step": 51888 + }, + { + "epoch": 1.9, + "learning_rate": 1.0162229464972351e-08, + "loss": 0.6949, + "step": 51889 + }, + { + "epoch": 1.9, + "learning_rate": 1.015856740030029e-08, + "loss": 0.6891, + "step": 51890 + }, + { + "epoch": 1.9, + "learning_rate": 1.0154905335628226e-08, + "loss": 0.6953, + "step": 51891 + }, + { + "epoch": 1.9, + "learning_rate": 1.0151243270956164e-08, + "loss": 0.7036, + "step": 51892 + }, + { + "epoch": 1.9, + "learning_rate": 1.0147581206284102e-08, + "loss": 0.6938, + "step": 51893 + }, + { + "epoch": 1.9, + "learning_rate": 1.014391914161204e-08, + "loss": 0.6954, + "step": 51894 + }, + { + "epoch": 1.9, + "learning_rate": 1.0140257076939979e-08, + "loss": 0.6956, + "step": 51895 + }, + { + "epoch": 1.9, + "learning_rate": 1.0136595012267916e-08, + "loss": 0.6873, + "step": 51896 + }, + { + "epoch": 1.9, + "learning_rate": 1.0132932947595854e-08, + "loss": 0.6969, + "step": 51897 + }, + { + "epoch": 1.9, + "learning_rate": 1.0129270882923792e-08, + "loss": 0.6808, + "step": 51898 + }, + { + "epoch": 1.9, + "learning_rate": 1.0125608818251729e-08, + "loss": 0.6844, + "step": 51899 + }, + { + "epoch": 1.9, + "learning_rate": 1.0121946753579667e-08, + "loss": 0.6899, + "step": 51900 + }, + { + "epoch": 1.9, + "learning_rate": 1.0118284688907605e-08, + "loss": 0.7069, + "step": 51901 + }, + { + "epoch": 1.9, + "learning_rate": 1.0114622624235544e-08, + "loss": 0.6958, + "step": 51902 + }, + { + "epoch": 1.9, + "learning_rate": 1.0110960559563482e-08, + "loss": 0.6925, + "step": 51903 + }, + { + "epoch": 1.9, + "learning_rate": 1.010729849489142e-08, + "loss": 0.6969, + "step": 51904 + }, + { + "epoch": 1.9, + "learning_rate": 1.0103636430219357e-08, + "loss": 0.6943, + "step": 51905 + }, + { + "epoch": 1.9, + "learning_rate": 1.0099974365547295e-08, + "loss": 0.6979, + "step": 51906 + }, + { + "epoch": 1.9, + "learning_rate": 1.0096312300875233e-08, + "loss": 0.7066, + "step": 51907 + }, + { + "epoch": 1.9, + "learning_rate": 1.009265023620317e-08, + "loss": 0.6985, + "step": 51908 + }, + { + "epoch": 1.9, + "learning_rate": 1.0088988171531108e-08, + "loss": 0.6925, + "step": 51909 + }, + { + "epoch": 1.9, + "learning_rate": 1.0085326106859047e-08, + "loss": 0.6922, + "step": 51910 + }, + { + "epoch": 1.9, + "learning_rate": 1.0081664042186983e-08, + "loss": 0.6867, + "step": 51911 + }, + { + "epoch": 1.9, + "learning_rate": 1.0078001977514923e-08, + "loss": 0.6965, + "step": 51912 + }, + { + "epoch": 1.9, + "learning_rate": 1.0074339912842861e-08, + "loss": 0.6928, + "step": 51913 + }, + { + "epoch": 1.9, + "learning_rate": 1.0070677848170798e-08, + "loss": 0.6928, + "step": 51914 + }, + { + "epoch": 1.9, + "learning_rate": 1.0067015783498736e-08, + "loss": 0.6811, + "step": 51915 + }, + { + "epoch": 1.9, + "learning_rate": 1.0063353718826675e-08, + "loss": 0.7006, + "step": 51916 + }, + { + "epoch": 1.9, + "learning_rate": 1.0059691654154611e-08, + "loss": 0.7048, + "step": 51917 + }, + { + "epoch": 1.9, + "learning_rate": 1.005602958948255e-08, + "loss": 0.692, + "step": 51918 + }, + { + "epoch": 1.9, + "learning_rate": 1.0052367524810488e-08, + "loss": 0.6908, + "step": 51919 + }, + { + "epoch": 1.9, + "learning_rate": 1.0048705460138424e-08, + "loss": 0.6702, + "step": 51920 + }, + { + "epoch": 1.9, + "learning_rate": 1.0045043395466364e-08, + "loss": 0.6881, + "step": 51921 + }, + { + "epoch": 1.9, + "learning_rate": 1.0041381330794303e-08, + "loss": 0.6938, + "step": 51922 + }, + { + "epoch": 1.9, + "learning_rate": 1.0037719266122239e-08, + "loss": 0.6834, + "step": 51923 + }, + { + "epoch": 1.9, + "learning_rate": 1.0034057201450177e-08, + "loss": 0.6959, + "step": 51924 + }, + { + "epoch": 1.9, + "learning_rate": 1.0030395136778116e-08, + "loss": 0.7138, + "step": 51925 + }, + { + "epoch": 1.9, + "learning_rate": 1.0026733072106052e-08, + "loss": 0.691, + "step": 51926 + }, + { + "epoch": 1.9, + "learning_rate": 1.002307100743399e-08, + "loss": 0.6824, + "step": 51927 + }, + { + "epoch": 1.9, + "learning_rate": 1.0019408942761929e-08, + "loss": 0.7015, + "step": 51928 + }, + { + "epoch": 1.9, + "learning_rate": 1.0015746878089865e-08, + "loss": 0.6865, + "step": 51929 + }, + { + "epoch": 1.9, + "learning_rate": 1.0012084813417805e-08, + "loss": 0.6919, + "step": 51930 + }, + { + "epoch": 1.9, + "learning_rate": 1.0008422748745744e-08, + "loss": 0.7034, + "step": 51931 + }, + { + "epoch": 1.9, + "learning_rate": 1.000476068407368e-08, + "loss": 0.6997, + "step": 51932 + }, + { + "epoch": 1.9, + "learning_rate": 1.0001098619401619e-08, + "loss": 0.691, + "step": 51933 + }, + { + "epoch": 1.9, + "learning_rate": 9.997436554729557e-09, + "loss": 0.6928, + "step": 51934 + }, + { + "epoch": 1.9, + "learning_rate": 9.993774490057493e-09, + "loss": 0.6994, + "step": 51935 + }, + { + "epoch": 1.9, + "learning_rate": 9.990112425385432e-09, + "loss": 0.6987, + "step": 51936 + }, + { + "epoch": 1.9, + "learning_rate": 9.98645036071337e-09, + "loss": 0.6996, + "step": 51937 + }, + { + "epoch": 1.9, + "learning_rate": 9.982788296041307e-09, + "loss": 0.6971, + "step": 51938 + }, + { + "epoch": 1.9, + "learning_rate": 9.979126231369247e-09, + "loss": 0.6929, + "step": 51939 + }, + { + "epoch": 1.9, + "learning_rate": 9.975464166697183e-09, + "loss": 0.6953, + "step": 51940 + }, + { + "epoch": 1.9, + "learning_rate": 9.971802102025121e-09, + "loss": 0.7034, + "step": 51941 + }, + { + "epoch": 1.9, + "learning_rate": 9.96814003735306e-09, + "loss": 0.6927, + "step": 51942 + }, + { + "epoch": 1.9, + "learning_rate": 9.964477972680996e-09, + "loss": 0.6969, + "step": 51943 + }, + { + "epoch": 1.9, + "learning_rate": 9.960815908008935e-09, + "loss": 0.6974, + "step": 51944 + }, + { + "epoch": 1.9, + "learning_rate": 9.957153843336873e-09, + "loss": 0.7057, + "step": 51945 + }, + { + "epoch": 1.9, + "learning_rate": 9.953491778664811e-09, + "loss": 0.6907, + "step": 51946 + }, + { + "epoch": 1.9, + "learning_rate": 9.949829713992748e-09, + "loss": 0.6992, + "step": 51947 + }, + { + "epoch": 1.9, + "learning_rate": 9.946167649320686e-09, + "loss": 0.6926, + "step": 51948 + }, + { + "epoch": 1.9, + "learning_rate": 9.942505584648624e-09, + "loss": 0.6835, + "step": 51949 + }, + { + "epoch": 1.9, + "learning_rate": 9.938843519976563e-09, + "loss": 0.6983, + "step": 51950 + }, + { + "epoch": 1.9, + "learning_rate": 9.935181455304501e-09, + "loss": 0.7007, + "step": 51951 + }, + { + "epoch": 1.9, + "learning_rate": 9.931519390632438e-09, + "loss": 0.689, + "step": 51952 + }, + { + "epoch": 1.9, + "learning_rate": 9.927857325960376e-09, + "loss": 0.6968, + "step": 51953 + }, + { + "epoch": 1.9, + "learning_rate": 9.924195261288314e-09, + "loss": 0.7096, + "step": 51954 + }, + { + "epoch": 1.9, + "learning_rate": 9.92053319661625e-09, + "loss": 0.6879, + "step": 51955 + }, + { + "epoch": 1.9, + "learning_rate": 9.916871131944189e-09, + "loss": 0.6862, + "step": 51956 + }, + { + "epoch": 1.9, + "learning_rate": 9.913209067272127e-09, + "loss": 0.685, + "step": 51957 + }, + { + "epoch": 1.9, + "learning_rate": 9.909547002600065e-09, + "loss": 0.699, + "step": 51958 + }, + { + "epoch": 1.9, + "learning_rate": 9.905884937928004e-09, + "loss": 0.7013, + "step": 51959 + }, + { + "epoch": 1.9, + "learning_rate": 9.902222873255942e-09, + "loss": 0.694, + "step": 51960 + }, + { + "epoch": 1.9, + "learning_rate": 9.898560808583879e-09, + "loss": 0.6902, + "step": 51961 + }, + { + "epoch": 1.9, + "learning_rate": 9.894898743911817e-09, + "loss": 0.7012, + "step": 51962 + }, + { + "epoch": 1.9, + "learning_rate": 9.891236679239755e-09, + "loss": 0.6922, + "step": 51963 + }, + { + "epoch": 1.9, + "learning_rate": 9.887574614567692e-09, + "loss": 0.6844, + "step": 51964 + }, + { + "epoch": 1.9, + "learning_rate": 9.88391254989563e-09, + "loss": 0.6978, + "step": 51965 + }, + { + "epoch": 1.9, + "learning_rate": 9.880250485223568e-09, + "loss": 0.6787, + "step": 51966 + }, + { + "epoch": 1.9, + "learning_rate": 9.876588420551507e-09, + "loss": 0.684, + "step": 51967 + }, + { + "epoch": 1.9, + "learning_rate": 9.872926355879445e-09, + "loss": 0.7075, + "step": 51968 + }, + { + "epoch": 1.9, + "learning_rate": 9.869264291207383e-09, + "loss": 0.7064, + "step": 51969 + }, + { + "epoch": 1.9, + "learning_rate": 9.86560222653532e-09, + "loss": 0.6997, + "step": 51970 + }, + { + "epoch": 1.9, + "learning_rate": 9.861940161863258e-09, + "loss": 0.688, + "step": 51971 + }, + { + "epoch": 1.9, + "learning_rate": 9.858278097191196e-09, + "loss": 0.6964, + "step": 51972 + }, + { + "epoch": 1.9, + "learning_rate": 9.854616032519133e-09, + "loss": 0.7041, + "step": 51973 + }, + { + "epoch": 1.9, + "learning_rate": 9.850953967847071e-09, + "loss": 0.6878, + "step": 51974 + }, + { + "epoch": 1.9, + "learning_rate": 9.84729190317501e-09, + "loss": 0.6927, + "step": 51975 + }, + { + "epoch": 1.9, + "learning_rate": 9.843629838502948e-09, + "loss": 0.6959, + "step": 51976 + }, + { + "epoch": 1.9, + "learning_rate": 9.839967773830886e-09, + "loss": 0.69, + "step": 51977 + }, + { + "epoch": 1.9, + "learning_rate": 9.836305709158824e-09, + "loss": 0.6927, + "step": 51978 + }, + { + "epoch": 1.9, + "learning_rate": 9.832643644486761e-09, + "loss": 0.6956, + "step": 51979 + }, + { + "epoch": 1.9, + "learning_rate": 9.8289815798147e-09, + "loss": 0.6918, + "step": 51980 + }, + { + "epoch": 1.9, + "learning_rate": 9.825319515142638e-09, + "loss": 0.6873, + "step": 51981 + }, + { + "epoch": 1.9, + "learning_rate": 9.821657450470574e-09, + "loss": 0.6847, + "step": 51982 + }, + { + "epoch": 1.9, + "learning_rate": 9.817995385798512e-09, + "loss": 0.6962, + "step": 51983 + }, + { + "epoch": 1.9, + "learning_rate": 9.81433332112645e-09, + "loss": 0.6929, + "step": 51984 + }, + { + "epoch": 1.9, + "learning_rate": 9.810671256454389e-09, + "loss": 0.7006, + "step": 51985 + }, + { + "epoch": 1.9, + "learning_rate": 9.807009191782327e-09, + "loss": 0.7052, + "step": 51986 + }, + { + "epoch": 1.9, + "learning_rate": 9.803347127110264e-09, + "loss": 0.6956, + "step": 51987 + }, + { + "epoch": 1.9, + "learning_rate": 9.799685062438202e-09, + "loss": 0.7045, + "step": 51988 + }, + { + "epoch": 1.9, + "learning_rate": 9.79602299776614e-09, + "loss": 0.6948, + "step": 51989 + }, + { + "epoch": 1.9, + "learning_rate": 9.792360933094079e-09, + "loss": 0.6928, + "step": 51990 + }, + { + "epoch": 1.9, + "learning_rate": 9.788698868422015e-09, + "loss": 0.6937, + "step": 51991 + }, + { + "epoch": 1.9, + "learning_rate": 9.785036803749954e-09, + "loss": 0.6981, + "step": 51992 + }, + { + "epoch": 1.9, + "learning_rate": 9.781374739077892e-09, + "loss": 0.6835, + "step": 51993 + }, + { + "epoch": 1.9, + "learning_rate": 9.77771267440583e-09, + "loss": 0.6882, + "step": 51994 + }, + { + "epoch": 1.9, + "learning_rate": 9.774050609733768e-09, + "loss": 0.7054, + "step": 51995 + }, + { + "epoch": 1.9, + "learning_rate": 9.770388545061705e-09, + "loss": 0.6855, + "step": 51996 + }, + { + "epoch": 1.9, + "learning_rate": 9.766726480389643e-09, + "loss": 0.7006, + "step": 51997 + }, + { + "epoch": 1.9, + "learning_rate": 9.763064415717582e-09, + "loss": 0.6908, + "step": 51998 + }, + { + "epoch": 1.9, + "learning_rate": 9.759402351045518e-09, + "loss": 0.6917, + "step": 51999 + }, + { + "epoch": 1.9, + "learning_rate": 9.755740286373456e-09, + "loss": 0.692, + "step": 52000 + }, + { + "epoch": 1.9, + "learning_rate": 9.752078221701395e-09, + "loss": 0.7029, + "step": 52001 + }, + { + "epoch": 1.9, + "learning_rate": 9.748416157029331e-09, + "loss": 0.7032, + "step": 52002 + }, + { + "epoch": 1.9, + "learning_rate": 9.74475409235727e-09, + "loss": 0.7096, + "step": 52003 + }, + { + "epoch": 1.9, + "learning_rate": 9.74109202768521e-09, + "loss": 0.6956, + "step": 52004 + }, + { + "epoch": 1.9, + "learning_rate": 9.737429963013146e-09, + "loss": 0.6903, + "step": 52005 + }, + { + "epoch": 1.9, + "learning_rate": 9.733767898341084e-09, + "loss": 0.6917, + "step": 52006 + }, + { + "epoch": 1.9, + "learning_rate": 9.730105833669023e-09, + "loss": 0.698, + "step": 52007 + }, + { + "epoch": 1.9, + "learning_rate": 9.72644376899696e-09, + "loss": 0.6893, + "step": 52008 + }, + { + "epoch": 1.9, + "learning_rate": 9.722781704324898e-09, + "loss": 0.6997, + "step": 52009 + }, + { + "epoch": 1.9, + "learning_rate": 9.719119639652836e-09, + "loss": 0.6767, + "step": 52010 + }, + { + "epoch": 1.9, + "learning_rate": 9.715457574980773e-09, + "loss": 0.6843, + "step": 52011 + }, + { + "epoch": 1.9, + "learning_rate": 9.71179551030871e-09, + "loss": 0.69, + "step": 52012 + }, + { + "epoch": 1.9, + "learning_rate": 9.70813344563665e-09, + "loss": 0.6891, + "step": 52013 + }, + { + "epoch": 1.9, + "learning_rate": 9.704471380964587e-09, + "loss": 0.7005, + "step": 52014 + }, + { + "epoch": 1.9, + "learning_rate": 9.700809316292526e-09, + "loss": 0.6856, + "step": 52015 + }, + { + "epoch": 1.9, + "learning_rate": 9.697147251620464e-09, + "loss": 0.7053, + "step": 52016 + }, + { + "epoch": 1.9, + "learning_rate": 9.6934851869484e-09, + "loss": 0.6912, + "step": 52017 + }, + { + "epoch": 1.9, + "learning_rate": 9.689823122276339e-09, + "loss": 0.6922, + "step": 52018 + }, + { + "epoch": 1.9, + "learning_rate": 9.686161057604277e-09, + "loss": 0.6966, + "step": 52019 + }, + { + "epoch": 1.9, + "learning_rate": 9.682498992932214e-09, + "loss": 0.6821, + "step": 52020 + }, + { + "epoch": 1.9, + "learning_rate": 9.678836928260152e-09, + "loss": 0.7018, + "step": 52021 + }, + { + "epoch": 1.9, + "learning_rate": 9.675174863588092e-09, + "loss": 0.6875, + "step": 52022 + }, + { + "epoch": 1.9, + "learning_rate": 9.671512798916029e-09, + "loss": 0.6989, + "step": 52023 + }, + { + "epoch": 1.9, + "learning_rate": 9.667850734243967e-09, + "loss": 0.6892, + "step": 52024 + }, + { + "epoch": 1.9, + "learning_rate": 9.664188669571905e-09, + "loss": 0.7035, + "step": 52025 + }, + { + "epoch": 1.9, + "learning_rate": 9.660526604899842e-09, + "loss": 0.6918, + "step": 52026 + }, + { + "epoch": 1.9, + "learning_rate": 9.65686454022778e-09, + "loss": 0.7031, + "step": 52027 + }, + { + "epoch": 1.9, + "learning_rate": 9.653202475555718e-09, + "loss": 0.7053, + "step": 52028 + }, + { + "epoch": 1.9, + "learning_rate": 9.649540410883655e-09, + "loss": 0.7077, + "step": 52029 + }, + { + "epoch": 1.9, + "learning_rate": 9.645878346211593e-09, + "loss": 0.6952, + "step": 52030 + }, + { + "epoch": 1.9, + "learning_rate": 9.642216281539531e-09, + "loss": 0.6958, + "step": 52031 + }, + { + "epoch": 1.9, + "learning_rate": 9.63855421686747e-09, + "loss": 0.7029, + "step": 52032 + }, + { + "epoch": 1.9, + "learning_rate": 9.634892152195408e-09, + "loss": 0.683, + "step": 52033 + }, + { + "epoch": 1.9, + "learning_rate": 9.631230087523346e-09, + "loss": 0.694, + "step": 52034 + }, + { + "epoch": 1.9, + "learning_rate": 9.627568022851283e-09, + "loss": 0.6907, + "step": 52035 + }, + { + "epoch": 1.9, + "learning_rate": 9.623905958179221e-09, + "loss": 0.6927, + "step": 52036 + }, + { + "epoch": 1.9, + "learning_rate": 9.62024389350716e-09, + "loss": 0.6932, + "step": 52037 + }, + { + "epoch": 1.9, + "learning_rate": 9.616581828835096e-09, + "loss": 0.6942, + "step": 52038 + }, + { + "epoch": 1.9, + "learning_rate": 9.612919764163034e-09, + "loss": 0.6796, + "step": 52039 + }, + { + "epoch": 1.9, + "learning_rate": 9.609257699490973e-09, + "loss": 0.6952, + "step": 52040 + }, + { + "epoch": 1.9, + "learning_rate": 9.60559563481891e-09, + "loss": 0.6876, + "step": 52041 + }, + { + "epoch": 1.9, + "learning_rate": 9.601933570146849e-09, + "loss": 0.6857, + "step": 52042 + }, + { + "epoch": 1.9, + "learning_rate": 9.598271505474786e-09, + "loss": 0.702, + "step": 52043 + }, + { + "epoch": 1.9, + "learning_rate": 9.594609440802724e-09, + "loss": 0.6851, + "step": 52044 + }, + { + "epoch": 1.9, + "learning_rate": 9.590947376130662e-09, + "loss": 0.6899, + "step": 52045 + }, + { + "epoch": 1.9, + "learning_rate": 9.587285311458599e-09, + "loss": 0.6975, + "step": 52046 + }, + { + "epoch": 1.9, + "learning_rate": 9.583623246786537e-09, + "loss": 0.6956, + "step": 52047 + }, + { + "epoch": 1.9, + "learning_rate": 9.579961182114475e-09, + "loss": 0.696, + "step": 52048 + }, + { + "epoch": 1.9, + "learning_rate": 9.576299117442414e-09, + "loss": 0.7027, + "step": 52049 + }, + { + "epoch": 1.9, + "learning_rate": 9.572637052770352e-09, + "loss": 0.686, + "step": 52050 + }, + { + "epoch": 1.9, + "learning_rate": 9.56897498809829e-09, + "loss": 0.6894, + "step": 52051 + }, + { + "epoch": 1.9, + "learning_rate": 9.565312923426227e-09, + "loss": 0.6937, + "step": 52052 + }, + { + "epoch": 1.9, + "learning_rate": 9.561650858754165e-09, + "loss": 0.6881, + "step": 52053 + }, + { + "epoch": 1.9, + "learning_rate": 9.557988794082103e-09, + "loss": 0.6962, + "step": 52054 + }, + { + "epoch": 1.9, + "learning_rate": 9.55432672941004e-09, + "loss": 0.6837, + "step": 52055 + }, + { + "epoch": 1.9, + "learning_rate": 9.550664664737978e-09, + "loss": 0.6967, + "step": 52056 + }, + { + "epoch": 1.9, + "learning_rate": 9.547002600065917e-09, + "loss": 0.6938, + "step": 52057 + }, + { + "epoch": 1.9, + "learning_rate": 9.543340535393853e-09, + "loss": 0.6821, + "step": 52058 + }, + { + "epoch": 1.9, + "learning_rate": 9.539678470721793e-09, + "loss": 0.6935, + "step": 52059 + }, + { + "epoch": 1.9, + "learning_rate": 9.536016406049731e-09, + "loss": 0.6823, + "step": 52060 + }, + { + "epoch": 1.9, + "learning_rate": 9.532354341377668e-09, + "loss": 0.7156, + "step": 52061 + }, + { + "epoch": 1.9, + "learning_rate": 9.528692276705606e-09, + "loss": 0.7047, + "step": 52062 + }, + { + "epoch": 1.9, + "learning_rate": 9.525030212033545e-09, + "loss": 0.6943, + "step": 52063 + }, + { + "epoch": 1.9, + "learning_rate": 9.521368147361481e-09, + "loss": 0.6938, + "step": 52064 + }, + { + "epoch": 1.9, + "learning_rate": 9.51770608268942e-09, + "loss": 0.679, + "step": 52065 + }, + { + "epoch": 1.9, + "learning_rate": 9.514044018017358e-09, + "loss": 0.6884, + "step": 52066 + }, + { + "epoch": 1.9, + "learning_rate": 9.510381953345294e-09, + "loss": 0.6976, + "step": 52067 + }, + { + "epoch": 1.91, + "learning_rate": 9.506719888673234e-09, + "loss": 0.6922, + "step": 52068 + }, + { + "epoch": 1.91, + "learning_rate": 9.503057824001173e-09, + "loss": 0.6971, + "step": 52069 + }, + { + "epoch": 1.91, + "learning_rate": 9.49939575932911e-09, + "loss": 0.6934, + "step": 52070 + }, + { + "epoch": 1.91, + "learning_rate": 9.495733694657047e-09, + "loss": 0.6987, + "step": 52071 + }, + { + "epoch": 1.91, + "learning_rate": 9.492071629984986e-09, + "loss": 0.7048, + "step": 52072 + }, + { + "epoch": 1.91, + "learning_rate": 9.488409565312922e-09, + "loss": 0.6964, + "step": 52073 + }, + { + "epoch": 1.91, + "learning_rate": 9.48474750064086e-09, + "loss": 0.6876, + "step": 52074 + }, + { + "epoch": 1.91, + "learning_rate": 9.481085435968799e-09, + "loss": 0.6835, + "step": 52075 + }, + { + "epoch": 1.91, + "learning_rate": 9.477423371296736e-09, + "loss": 0.6974, + "step": 52076 + }, + { + "epoch": 1.91, + "learning_rate": 9.473761306624675e-09, + "loss": 0.684, + "step": 52077 + }, + { + "epoch": 1.91, + "learning_rate": 9.470099241952614e-09, + "loss": 0.69, + "step": 52078 + }, + { + "epoch": 1.91, + "learning_rate": 9.46643717728055e-09, + "loss": 0.6942, + "step": 52079 + }, + { + "epoch": 1.91, + "learning_rate": 9.462775112608489e-09, + "loss": 0.6994, + "step": 52080 + }, + { + "epoch": 1.91, + "learning_rate": 9.459113047936427e-09, + "loss": 0.6924, + "step": 52081 + }, + { + "epoch": 1.91, + "learning_rate": 9.455450983264364e-09, + "loss": 0.7088, + "step": 52082 + }, + { + "epoch": 1.91, + "learning_rate": 9.451788918592302e-09, + "loss": 0.6902, + "step": 52083 + }, + { + "epoch": 1.91, + "learning_rate": 9.44812685392024e-09, + "loss": 0.7028, + "step": 52084 + }, + { + "epoch": 1.91, + "learning_rate": 9.444464789248177e-09, + "loss": 0.6889, + "step": 52085 + }, + { + "epoch": 1.91, + "learning_rate": 9.440802724576117e-09, + "loss": 0.6823, + "step": 52086 + }, + { + "epoch": 1.91, + "learning_rate": 9.437140659904053e-09, + "loss": 0.6999, + "step": 52087 + }, + { + "epoch": 1.91, + "learning_rate": 9.433478595231992e-09, + "loss": 0.6971, + "step": 52088 + }, + { + "epoch": 1.91, + "learning_rate": 9.42981653055993e-09, + "loss": 0.6981, + "step": 52089 + }, + { + "epoch": 1.91, + "learning_rate": 9.426154465887866e-09, + "loss": 0.6954, + "step": 52090 + }, + { + "epoch": 1.91, + "learning_rate": 9.422492401215805e-09, + "loss": 0.6971, + "step": 52091 + }, + { + "epoch": 1.91, + "learning_rate": 9.418830336543743e-09, + "loss": 0.6919, + "step": 52092 + }, + { + "epoch": 1.91, + "learning_rate": 9.415168271871681e-09, + "loss": 0.6964, + "step": 52093 + }, + { + "epoch": 1.91, + "learning_rate": 9.411506207199618e-09, + "loss": 0.7026, + "step": 52094 + }, + { + "epoch": 1.91, + "learning_rate": 9.407844142527556e-09, + "loss": 0.6742, + "step": 52095 + }, + { + "epoch": 1.91, + "learning_rate": 9.404182077855494e-09, + "loss": 0.6917, + "step": 52096 + }, + { + "epoch": 1.91, + "learning_rate": 9.400520013183433e-09, + "loss": 0.6935, + "step": 52097 + }, + { + "epoch": 1.91, + "learning_rate": 9.396857948511371e-09, + "loss": 0.6912, + "step": 52098 + }, + { + "epoch": 1.91, + "learning_rate": 9.393195883839308e-09, + "loss": 0.6869, + "step": 52099 + }, + { + "epoch": 1.91, + "learning_rate": 9.389533819167246e-09, + "loss": 0.7045, + "step": 52100 + }, + { + "epoch": 1.91, + "learning_rate": 9.385871754495184e-09, + "loss": 0.6865, + "step": 52101 + }, + { + "epoch": 1.91, + "learning_rate": 9.38220968982312e-09, + "loss": 0.6916, + "step": 52102 + }, + { + "epoch": 1.91, + "learning_rate": 9.378547625151059e-09, + "loss": 0.7006, + "step": 52103 + }, + { + "epoch": 1.91, + "learning_rate": 9.374885560478997e-09, + "loss": 0.6958, + "step": 52104 + }, + { + "epoch": 1.91, + "learning_rate": 9.371223495806936e-09, + "loss": 0.6859, + "step": 52105 + }, + { + "epoch": 1.91, + "learning_rate": 9.367561431134874e-09, + "loss": 0.6952, + "step": 52106 + }, + { + "epoch": 1.91, + "learning_rate": 9.363899366462812e-09, + "loss": 0.6915, + "step": 52107 + }, + { + "epoch": 1.91, + "learning_rate": 9.360237301790749e-09, + "loss": 0.6917, + "step": 52108 + }, + { + "epoch": 1.91, + "learning_rate": 9.356575237118687e-09, + "loss": 0.6853, + "step": 52109 + }, + { + "epoch": 1.91, + "learning_rate": 9.352913172446625e-09, + "loss": 0.6992, + "step": 52110 + }, + { + "epoch": 1.91, + "learning_rate": 9.349251107774562e-09, + "loss": 0.7, + "step": 52111 + }, + { + "epoch": 1.91, + "learning_rate": 9.3455890431025e-09, + "loss": 0.6861, + "step": 52112 + }, + { + "epoch": 1.91, + "learning_rate": 9.341926978430438e-09, + "loss": 0.6892, + "step": 52113 + }, + { + "epoch": 1.91, + "learning_rate": 9.338264913758377e-09, + "loss": 0.6852, + "step": 52114 + }, + { + "epoch": 1.91, + "learning_rate": 9.334602849086315e-09, + "loss": 0.6861, + "step": 52115 + }, + { + "epoch": 1.91, + "learning_rate": 9.330940784414253e-09, + "loss": 0.6858, + "step": 52116 + }, + { + "epoch": 1.91, + "learning_rate": 9.32727871974219e-09, + "loss": 0.6966, + "step": 52117 + }, + { + "epoch": 1.91, + "learning_rate": 9.323616655070128e-09, + "loss": 0.7052, + "step": 52118 + }, + { + "epoch": 1.91, + "learning_rate": 9.319954590398066e-09, + "loss": 0.6948, + "step": 52119 + }, + { + "epoch": 1.91, + "learning_rate": 9.316292525726003e-09, + "loss": 0.6883, + "step": 52120 + }, + { + "epoch": 1.91, + "learning_rate": 9.312630461053941e-09, + "loss": 0.6882, + "step": 52121 + }, + { + "epoch": 1.91, + "learning_rate": 9.30896839638188e-09, + "loss": 0.6975, + "step": 52122 + }, + { + "epoch": 1.91, + "learning_rate": 9.305306331709818e-09, + "loss": 0.6982, + "step": 52123 + }, + { + "epoch": 1.91, + "learning_rate": 9.301644267037756e-09, + "loss": 0.7035, + "step": 52124 + }, + { + "epoch": 1.91, + "learning_rate": 9.297982202365694e-09, + "loss": 0.6962, + "step": 52125 + }, + { + "epoch": 1.91, + "learning_rate": 9.294320137693631e-09, + "loss": 0.6939, + "step": 52126 + }, + { + "epoch": 1.91, + "learning_rate": 9.29065807302157e-09, + "loss": 0.7009, + "step": 52127 + }, + { + "epoch": 1.91, + "learning_rate": 9.286996008349508e-09, + "loss": 0.7075, + "step": 52128 + }, + { + "epoch": 1.91, + "learning_rate": 9.283333943677444e-09, + "loss": 0.6841, + "step": 52129 + }, + { + "epoch": 1.91, + "learning_rate": 9.279671879005382e-09, + "loss": 0.6982, + "step": 52130 + }, + { + "epoch": 1.91, + "learning_rate": 9.27600981433332e-09, + "loss": 0.6947, + "step": 52131 + }, + { + "epoch": 1.91, + "learning_rate": 9.272347749661259e-09, + "loss": 0.6962, + "step": 52132 + }, + { + "epoch": 1.91, + "learning_rate": 9.268685684989197e-09, + "loss": 0.6953, + "step": 52133 + }, + { + "epoch": 1.91, + "learning_rate": 9.265023620317134e-09, + "loss": 0.6988, + "step": 52134 + }, + { + "epoch": 1.91, + "learning_rate": 9.261361555645072e-09, + "loss": 0.6998, + "step": 52135 + }, + { + "epoch": 1.91, + "learning_rate": 9.25769949097301e-09, + "loss": 0.6999, + "step": 52136 + }, + { + "epoch": 1.91, + "learning_rate": 9.254037426300949e-09, + "loss": 0.6952, + "step": 52137 + }, + { + "epoch": 1.91, + "learning_rate": 9.250375361628885e-09, + "loss": 0.6888, + "step": 52138 + }, + { + "epoch": 1.91, + "learning_rate": 9.246713296956824e-09, + "loss": 0.6974, + "step": 52139 + }, + { + "epoch": 1.91, + "learning_rate": 9.243051232284762e-09, + "loss": 0.7007, + "step": 52140 + }, + { + "epoch": 1.91, + "learning_rate": 9.239389167612699e-09, + "loss": 0.6923, + "step": 52141 + }, + { + "epoch": 1.91, + "learning_rate": 9.235727102940638e-09, + "loss": 0.6837, + "step": 52142 + }, + { + "epoch": 1.91, + "learning_rate": 9.232065038268575e-09, + "loss": 0.698, + "step": 52143 + }, + { + "epoch": 1.91, + "learning_rate": 9.228402973596513e-09, + "loss": 0.6877, + "step": 52144 + }, + { + "epoch": 1.91, + "learning_rate": 9.224740908924452e-09, + "loss": 0.6945, + "step": 52145 + }, + { + "epoch": 1.91, + "learning_rate": 9.221078844252388e-09, + "loss": 0.6925, + "step": 52146 + }, + { + "epoch": 1.91, + "learning_rate": 9.217416779580327e-09, + "loss": 0.7033, + "step": 52147 + }, + { + "epoch": 1.91, + "learning_rate": 9.213754714908265e-09, + "loss": 0.6981, + "step": 52148 + }, + { + "epoch": 1.91, + "learning_rate": 9.210092650236201e-09, + "loss": 0.7228, + "step": 52149 + }, + { + "epoch": 1.91, + "learning_rate": 9.20643058556414e-09, + "loss": 0.6939, + "step": 52150 + }, + { + "epoch": 1.91, + "learning_rate": 9.20276852089208e-09, + "loss": 0.6909, + "step": 52151 + }, + { + "epoch": 1.91, + "learning_rate": 9.199106456220016e-09, + "loss": 0.6843, + "step": 52152 + }, + { + "epoch": 1.91, + "learning_rate": 9.195444391547955e-09, + "loss": 0.6996, + "step": 52153 + }, + { + "epoch": 1.91, + "learning_rate": 9.191782326875893e-09, + "loss": 0.6867, + "step": 52154 + }, + { + "epoch": 1.91, + "learning_rate": 9.18812026220383e-09, + "loss": 0.6958, + "step": 52155 + }, + { + "epoch": 1.91, + "learning_rate": 9.184458197531768e-09, + "loss": 0.6902, + "step": 52156 + }, + { + "epoch": 1.91, + "learning_rate": 9.180796132859706e-09, + "loss": 0.6904, + "step": 52157 + }, + { + "epoch": 1.91, + "learning_rate": 9.177134068187643e-09, + "loss": 0.6962, + "step": 52158 + }, + { + "epoch": 1.91, + "learning_rate": 9.173472003515581e-09, + "loss": 0.6931, + "step": 52159 + }, + { + "epoch": 1.91, + "learning_rate": 9.16980993884352e-09, + "loss": 0.6921, + "step": 52160 + }, + { + "epoch": 1.91, + "learning_rate": 9.166147874171457e-09, + "loss": 0.6929, + "step": 52161 + }, + { + "epoch": 1.91, + "learning_rate": 9.162485809499396e-09, + "loss": 0.7101, + "step": 52162 + }, + { + "epoch": 1.91, + "learning_rate": 9.158823744827334e-09, + "loss": 0.679, + "step": 52163 + }, + { + "epoch": 1.91, + "learning_rate": 9.15516168015527e-09, + "loss": 0.6946, + "step": 52164 + }, + { + "epoch": 1.91, + "learning_rate": 9.151499615483209e-09, + "loss": 0.6907, + "step": 52165 + }, + { + "epoch": 1.91, + "learning_rate": 9.147837550811147e-09, + "loss": 0.6855, + "step": 52166 + }, + { + "epoch": 1.91, + "learning_rate": 9.144175486139084e-09, + "loss": 0.6873, + "step": 52167 + }, + { + "epoch": 1.91, + "learning_rate": 9.140513421467022e-09, + "loss": 0.692, + "step": 52168 + }, + { + "epoch": 1.91, + "learning_rate": 9.136851356794962e-09, + "loss": 0.6988, + "step": 52169 + }, + { + "epoch": 1.91, + "learning_rate": 9.133189292122899e-09, + "loss": 0.6814, + "step": 52170 + }, + { + "epoch": 1.91, + "learning_rate": 9.129527227450837e-09, + "loss": 0.6963, + "step": 52171 + }, + { + "epoch": 1.91, + "learning_rate": 9.125865162778775e-09, + "loss": 0.6892, + "step": 52172 + }, + { + "epoch": 1.91, + "learning_rate": 9.122203098106712e-09, + "loss": 0.687, + "step": 52173 + }, + { + "epoch": 1.91, + "learning_rate": 9.11854103343465e-09, + "loss": 0.6824, + "step": 52174 + }, + { + "epoch": 1.91, + "learning_rate": 9.114878968762588e-09, + "loss": 0.6996, + "step": 52175 + }, + { + "epoch": 1.91, + "learning_rate": 9.111216904090525e-09, + "loss": 0.7016, + "step": 52176 + }, + { + "epoch": 1.91, + "learning_rate": 9.107554839418463e-09, + "loss": 0.6863, + "step": 52177 + }, + { + "epoch": 1.91, + "learning_rate": 9.103892774746401e-09, + "loss": 0.6885, + "step": 52178 + }, + { + "epoch": 1.91, + "learning_rate": 9.10023071007434e-09, + "loss": 0.6965, + "step": 52179 + }, + { + "epoch": 1.91, + "learning_rate": 9.096568645402278e-09, + "loss": 0.6976, + "step": 52180 + }, + { + "epoch": 1.91, + "learning_rate": 9.092906580730216e-09, + "loss": 0.6893, + "step": 52181 + }, + { + "epoch": 1.91, + "learning_rate": 9.089244516058153e-09, + "loss": 0.6917, + "step": 52182 + }, + { + "epoch": 1.91, + "learning_rate": 9.085582451386091e-09, + "loss": 0.6933, + "step": 52183 + }, + { + "epoch": 1.91, + "learning_rate": 9.08192038671403e-09, + "loss": 0.6857, + "step": 52184 + }, + { + "epoch": 1.91, + "learning_rate": 9.078258322041966e-09, + "loss": 0.687, + "step": 52185 + }, + { + "epoch": 1.91, + "learning_rate": 9.074596257369904e-09, + "loss": 0.7046, + "step": 52186 + }, + { + "epoch": 1.91, + "learning_rate": 9.070934192697843e-09, + "loss": 0.7002, + "step": 52187 + }, + { + "epoch": 1.91, + "learning_rate": 9.067272128025781e-09, + "loss": 0.6918, + "step": 52188 + }, + { + "epoch": 1.91, + "learning_rate": 9.063610063353719e-09, + "loss": 0.6907, + "step": 52189 + }, + { + "epoch": 1.91, + "learning_rate": 9.059947998681656e-09, + "loss": 0.6988, + "step": 52190 + }, + { + "epoch": 1.91, + "learning_rate": 9.056285934009594e-09, + "loss": 0.686, + "step": 52191 + }, + { + "epoch": 1.91, + "learning_rate": 9.052623869337532e-09, + "loss": 0.6794, + "step": 52192 + }, + { + "epoch": 1.91, + "learning_rate": 9.048961804665469e-09, + "loss": 0.6782, + "step": 52193 + }, + { + "epoch": 1.91, + "learning_rate": 9.045299739993407e-09, + "loss": 0.6991, + "step": 52194 + }, + { + "epoch": 1.91, + "learning_rate": 9.041637675321346e-09, + "loss": 0.6841, + "step": 52195 + }, + { + "epoch": 1.91, + "learning_rate": 9.037975610649284e-09, + "loss": 0.6962, + "step": 52196 + }, + { + "epoch": 1.91, + "learning_rate": 9.034313545977222e-09, + "loss": 0.6967, + "step": 52197 + }, + { + "epoch": 1.91, + "learning_rate": 9.03065148130516e-09, + "loss": 0.6867, + "step": 52198 + }, + { + "epoch": 1.91, + "learning_rate": 9.026989416633097e-09, + "loss": 0.6921, + "step": 52199 + }, + { + "epoch": 1.91, + "learning_rate": 9.023327351961035e-09, + "loss": 0.6917, + "step": 52200 + }, + { + "epoch": 1.91, + "learning_rate": 9.019665287288973e-09, + "loss": 0.7033, + "step": 52201 + }, + { + "epoch": 1.91, + "learning_rate": 9.01600322261691e-09, + "loss": 0.6994, + "step": 52202 + }, + { + "epoch": 1.91, + "learning_rate": 9.012341157944848e-09, + "loss": 0.6984, + "step": 52203 + }, + { + "epoch": 1.91, + "learning_rate": 9.008679093272787e-09, + "loss": 0.6988, + "step": 52204 + }, + { + "epoch": 1.91, + "learning_rate": 9.005017028600723e-09, + "loss": 0.6936, + "step": 52205 + }, + { + "epoch": 1.91, + "learning_rate": 9.001354963928663e-09, + "loss": 0.6918, + "step": 52206 + }, + { + "epoch": 1.91, + "learning_rate": 8.997692899256601e-09, + "loss": 0.6808, + "step": 52207 + }, + { + "epoch": 1.91, + "learning_rate": 8.994030834584538e-09, + "loss": 0.6922, + "step": 52208 + }, + { + "epoch": 1.91, + "learning_rate": 8.990368769912476e-09, + "loss": 0.6864, + "step": 52209 + }, + { + "epoch": 1.91, + "learning_rate": 8.986706705240415e-09, + "loss": 0.6898, + "step": 52210 + }, + { + "epoch": 1.91, + "learning_rate": 8.983044640568351e-09, + "loss": 0.6986, + "step": 52211 + }, + { + "epoch": 1.91, + "learning_rate": 8.97938257589629e-09, + "loss": 0.7035, + "step": 52212 + }, + { + "epoch": 1.91, + "learning_rate": 8.975720511224228e-09, + "loss": 0.689, + "step": 52213 + }, + { + "epoch": 1.91, + "learning_rate": 8.972058446552164e-09, + "loss": 0.6955, + "step": 52214 + }, + { + "epoch": 1.91, + "learning_rate": 8.968396381880104e-09, + "loss": 0.684, + "step": 52215 + }, + { + "epoch": 1.91, + "learning_rate": 8.964734317208043e-09, + "loss": 0.6935, + "step": 52216 + }, + { + "epoch": 1.91, + "learning_rate": 8.96107225253598e-09, + "loss": 0.7029, + "step": 52217 + }, + { + "epoch": 1.91, + "learning_rate": 8.957410187863918e-09, + "loss": 0.69, + "step": 52218 + }, + { + "epoch": 1.91, + "learning_rate": 8.953748123191856e-09, + "loss": 0.6893, + "step": 52219 + }, + { + "epoch": 1.91, + "learning_rate": 8.950086058519792e-09, + "loss": 0.6894, + "step": 52220 + }, + { + "epoch": 1.91, + "learning_rate": 8.94642399384773e-09, + "loss": 0.6906, + "step": 52221 + }, + { + "epoch": 1.91, + "learning_rate": 8.942761929175669e-09, + "loss": 0.6936, + "step": 52222 + }, + { + "epoch": 1.91, + "learning_rate": 8.939099864503606e-09, + "loss": 0.6988, + "step": 52223 + }, + { + "epoch": 1.91, + "learning_rate": 8.935437799831546e-09, + "loss": 0.7004, + "step": 52224 + }, + { + "epoch": 1.91, + "learning_rate": 8.931775735159484e-09, + "loss": 0.6938, + "step": 52225 + }, + { + "epoch": 1.91, + "learning_rate": 8.92811367048742e-09, + "loss": 0.6953, + "step": 52226 + }, + { + "epoch": 1.91, + "learning_rate": 8.924451605815359e-09, + "loss": 0.6984, + "step": 52227 + }, + { + "epoch": 1.91, + "learning_rate": 8.920789541143297e-09, + "loss": 0.6874, + "step": 52228 + }, + { + "epoch": 1.91, + "learning_rate": 8.917127476471234e-09, + "loss": 0.6932, + "step": 52229 + }, + { + "epoch": 1.91, + "learning_rate": 8.913465411799172e-09, + "loss": 0.6979, + "step": 52230 + }, + { + "epoch": 1.91, + "learning_rate": 8.90980334712711e-09, + "loss": 0.6992, + "step": 52231 + }, + { + "epoch": 1.91, + "learning_rate": 8.906141282455047e-09, + "loss": 0.6948, + "step": 52232 + }, + { + "epoch": 1.91, + "learning_rate": 8.902479217782987e-09, + "loss": 0.6914, + "step": 52233 + }, + { + "epoch": 1.91, + "learning_rate": 8.898817153110923e-09, + "loss": 0.6986, + "step": 52234 + }, + { + "epoch": 1.91, + "learning_rate": 8.895155088438862e-09, + "loss": 0.6907, + "step": 52235 + }, + { + "epoch": 1.91, + "learning_rate": 8.8914930237668e-09, + "loss": 0.6883, + "step": 52236 + }, + { + "epoch": 1.91, + "learning_rate": 8.887830959094736e-09, + "loss": 0.6909, + "step": 52237 + }, + { + "epoch": 1.91, + "learning_rate": 8.884168894422675e-09, + "loss": 0.6938, + "step": 52238 + }, + { + "epoch": 1.91, + "learning_rate": 8.880506829750613e-09, + "loss": 0.681, + "step": 52239 + }, + { + "epoch": 1.91, + "learning_rate": 8.876844765078551e-09, + "loss": 0.6962, + "step": 52240 + }, + { + "epoch": 1.91, + "learning_rate": 8.873182700406488e-09, + "loss": 0.7027, + "step": 52241 + }, + { + "epoch": 1.91, + "learning_rate": 8.869520635734426e-09, + "loss": 0.6904, + "step": 52242 + }, + { + "epoch": 1.91, + "learning_rate": 8.865858571062364e-09, + "loss": 0.686, + "step": 52243 + }, + { + "epoch": 1.91, + "learning_rate": 8.862196506390303e-09, + "loss": 0.6845, + "step": 52244 + }, + { + "epoch": 1.91, + "learning_rate": 8.858534441718241e-09, + "loss": 0.6878, + "step": 52245 + }, + { + "epoch": 1.91, + "learning_rate": 8.854872377046178e-09, + "loss": 0.704, + "step": 52246 + }, + { + "epoch": 1.91, + "learning_rate": 8.851210312374116e-09, + "loss": 0.681, + "step": 52247 + }, + { + "epoch": 1.91, + "learning_rate": 8.847548247702054e-09, + "loss": 0.6913, + "step": 52248 + }, + { + "epoch": 1.91, + "learning_rate": 8.84388618302999e-09, + "loss": 0.6881, + "step": 52249 + }, + { + "epoch": 1.91, + "learning_rate": 8.840224118357929e-09, + "loss": 0.708, + "step": 52250 + }, + { + "epoch": 1.91, + "learning_rate": 8.836562053685867e-09, + "loss": 0.695, + "step": 52251 + }, + { + "epoch": 1.91, + "learning_rate": 8.832899989013806e-09, + "loss": 0.6968, + "step": 52252 + }, + { + "epoch": 1.91, + "learning_rate": 8.829237924341744e-09, + "loss": 0.7124, + "step": 52253 + }, + { + "epoch": 1.91, + "learning_rate": 8.825575859669682e-09, + "loss": 0.6963, + "step": 52254 + }, + { + "epoch": 1.91, + "learning_rate": 8.821913794997619e-09, + "loss": 0.6992, + "step": 52255 + }, + { + "epoch": 1.91, + "learning_rate": 8.818251730325557e-09, + "loss": 0.6928, + "step": 52256 + }, + { + "epoch": 1.91, + "learning_rate": 8.814589665653495e-09, + "loss": 0.6782, + "step": 52257 + }, + { + "epoch": 1.91, + "learning_rate": 8.810927600981432e-09, + "loss": 0.7044, + "step": 52258 + }, + { + "epoch": 1.91, + "learning_rate": 8.80726553630937e-09, + "loss": 0.6927, + "step": 52259 + }, + { + "epoch": 1.91, + "learning_rate": 8.803603471637309e-09, + "loss": 0.6871, + "step": 52260 + }, + { + "epoch": 1.91, + "learning_rate": 8.799941406965247e-09, + "loss": 0.7011, + "step": 52261 + }, + { + "epoch": 1.91, + "learning_rate": 8.796279342293185e-09, + "loss": 0.6941, + "step": 52262 + }, + { + "epoch": 1.91, + "learning_rate": 8.792617277621123e-09, + "loss": 0.6913, + "step": 52263 + }, + { + "epoch": 1.91, + "learning_rate": 8.78895521294906e-09, + "loss": 0.6939, + "step": 52264 + }, + { + "epoch": 1.91, + "learning_rate": 8.785293148276998e-09, + "loss": 0.6878, + "step": 52265 + }, + { + "epoch": 1.91, + "learning_rate": 8.781631083604937e-09, + "loss": 0.6926, + "step": 52266 + }, + { + "epoch": 1.91, + "learning_rate": 8.777969018932873e-09, + "loss": 0.6919, + "step": 52267 + }, + { + "epoch": 1.91, + "learning_rate": 8.774306954260811e-09, + "loss": 0.701, + "step": 52268 + }, + { + "epoch": 1.91, + "learning_rate": 8.77064488958875e-09, + "loss": 0.694, + "step": 52269 + }, + { + "epoch": 1.91, + "learning_rate": 8.766982824916688e-09, + "loss": 0.6997, + "step": 52270 + }, + { + "epoch": 1.91, + "learning_rate": 8.763320760244626e-09, + "loss": 0.6974, + "step": 52271 + }, + { + "epoch": 1.91, + "learning_rate": 8.759658695572564e-09, + "loss": 0.6946, + "step": 52272 + }, + { + "epoch": 1.91, + "learning_rate": 8.755996630900501e-09, + "loss": 0.6943, + "step": 52273 + }, + { + "epoch": 1.91, + "learning_rate": 8.75233456622844e-09, + "loss": 0.6883, + "step": 52274 + }, + { + "epoch": 1.91, + "learning_rate": 8.748672501556378e-09, + "loss": 0.6852, + "step": 52275 + }, + { + "epoch": 1.91, + "learning_rate": 8.745010436884314e-09, + "loss": 0.689, + "step": 52276 + }, + { + "epoch": 1.91, + "learning_rate": 8.741348372212253e-09, + "loss": 0.6747, + "step": 52277 + }, + { + "epoch": 1.91, + "learning_rate": 8.737686307540191e-09, + "loss": 0.6829, + "step": 52278 + }, + { + "epoch": 1.91, + "learning_rate": 8.734024242868129e-09, + "loss": 0.6942, + "step": 52279 + }, + { + "epoch": 1.91, + "learning_rate": 8.730362178196067e-09, + "loss": 0.6931, + "step": 52280 + }, + { + "epoch": 1.91, + "learning_rate": 8.726700113524004e-09, + "loss": 0.6906, + "step": 52281 + }, + { + "epoch": 1.91, + "learning_rate": 8.723038048851942e-09, + "loss": 0.7003, + "step": 52282 + }, + { + "epoch": 1.91, + "learning_rate": 8.71937598417988e-09, + "loss": 0.6751, + "step": 52283 + }, + { + "epoch": 1.91, + "learning_rate": 8.715713919507819e-09, + "loss": 0.6995, + "step": 52284 + }, + { + "epoch": 1.91, + "learning_rate": 8.712051854835755e-09, + "loss": 0.6968, + "step": 52285 + }, + { + "epoch": 1.91, + "learning_rate": 8.708389790163694e-09, + "loss": 0.6884, + "step": 52286 + }, + { + "epoch": 1.91, + "learning_rate": 8.704727725491632e-09, + "loss": 0.7102, + "step": 52287 + }, + { + "epoch": 1.91, + "learning_rate": 8.701065660819569e-09, + "loss": 0.6842, + "step": 52288 + }, + { + "epoch": 1.91, + "learning_rate": 8.697403596147509e-09, + "loss": 0.6952, + "step": 52289 + }, + { + "epoch": 1.91, + "learning_rate": 8.693741531475445e-09, + "loss": 0.7006, + "step": 52290 + }, + { + "epoch": 1.91, + "learning_rate": 8.690079466803383e-09, + "loss": 0.6882, + "step": 52291 + }, + { + "epoch": 1.91, + "learning_rate": 8.686417402131322e-09, + "loss": 0.6925, + "step": 52292 + }, + { + "epoch": 1.91, + "learning_rate": 8.682755337459258e-09, + "loss": 0.6895, + "step": 52293 + }, + { + "epoch": 1.91, + "learning_rate": 8.679093272787197e-09, + "loss": 0.6921, + "step": 52294 + }, + { + "epoch": 1.91, + "learning_rate": 8.675431208115135e-09, + "loss": 0.6932, + "step": 52295 + }, + { + "epoch": 1.91, + "learning_rate": 8.671769143443071e-09, + "loss": 0.6808, + "step": 52296 + }, + { + "epoch": 1.91, + "learning_rate": 8.66810707877101e-09, + "loss": 0.7007, + "step": 52297 + }, + { + "epoch": 1.91, + "learning_rate": 8.66444501409895e-09, + "loss": 0.6935, + "step": 52298 + }, + { + "epoch": 1.91, + "learning_rate": 8.660782949426886e-09, + "loss": 0.7042, + "step": 52299 + }, + { + "epoch": 1.91, + "learning_rate": 8.657120884754825e-09, + "loss": 0.6974, + "step": 52300 + }, + { + "epoch": 1.91, + "learning_rate": 8.653458820082763e-09, + "loss": 0.707, + "step": 52301 + }, + { + "epoch": 1.91, + "learning_rate": 8.6497967554107e-09, + "loss": 0.6905, + "step": 52302 + }, + { + "epoch": 1.91, + "learning_rate": 8.646134690738638e-09, + "loss": 0.7016, + "step": 52303 + }, + { + "epoch": 1.91, + "learning_rate": 8.642472626066576e-09, + "loss": 0.6913, + "step": 52304 + }, + { + "epoch": 1.91, + "learning_rate": 8.638810561394513e-09, + "loss": 0.6903, + "step": 52305 + }, + { + "epoch": 1.91, + "learning_rate": 8.635148496722451e-09, + "loss": 0.6926, + "step": 52306 + }, + { + "epoch": 1.91, + "learning_rate": 8.631486432050391e-09, + "loss": 0.6967, + "step": 52307 + }, + { + "epoch": 1.91, + "learning_rate": 8.627824367378327e-09, + "loss": 0.6873, + "step": 52308 + }, + { + "epoch": 1.91, + "learning_rate": 8.624162302706266e-09, + "loss": 0.6978, + "step": 52309 + }, + { + "epoch": 1.91, + "learning_rate": 8.620500238034204e-09, + "loss": 0.699, + "step": 52310 + }, + { + "epoch": 1.91, + "learning_rate": 8.61683817336214e-09, + "loss": 0.6939, + "step": 52311 + }, + { + "epoch": 1.91, + "learning_rate": 8.613176108690079e-09, + "loss": 0.6906, + "step": 52312 + }, + { + "epoch": 1.91, + "learning_rate": 8.609514044018017e-09, + "loss": 0.6836, + "step": 52313 + }, + { + "epoch": 1.91, + "learning_rate": 8.605851979345954e-09, + "loss": 0.687, + "step": 52314 + }, + { + "epoch": 1.91, + "learning_rate": 8.602189914673892e-09, + "loss": 0.6824, + "step": 52315 + }, + { + "epoch": 1.91, + "learning_rate": 8.598527850001832e-09, + "loss": 0.6978, + "step": 52316 + }, + { + "epoch": 1.91, + "learning_rate": 8.594865785329769e-09, + "loss": 0.6883, + "step": 52317 + }, + { + "epoch": 1.91, + "learning_rate": 8.591203720657707e-09, + "loss": 0.6891, + "step": 52318 + }, + { + "epoch": 1.91, + "learning_rate": 8.587541655985645e-09, + "loss": 0.6838, + "step": 52319 + }, + { + "epoch": 1.91, + "learning_rate": 8.583879591313582e-09, + "loss": 0.704, + "step": 52320 + }, + { + "epoch": 1.91, + "learning_rate": 8.58021752664152e-09, + "loss": 0.6952, + "step": 52321 + }, + { + "epoch": 1.91, + "learning_rate": 8.576555461969458e-09, + "loss": 0.6913, + "step": 52322 + }, + { + "epoch": 1.91, + "learning_rate": 8.572893397297395e-09, + "loss": 0.6966, + "step": 52323 + }, + { + "epoch": 1.91, + "learning_rate": 8.569231332625333e-09, + "loss": 0.6974, + "step": 52324 + }, + { + "epoch": 1.91, + "learning_rate": 8.565569267953272e-09, + "loss": 0.6841, + "step": 52325 + }, + { + "epoch": 1.91, + "learning_rate": 8.56190720328121e-09, + "loss": 0.6938, + "step": 52326 + }, + { + "epoch": 1.91, + "learning_rate": 8.558245138609148e-09, + "loss": 0.6982, + "step": 52327 + }, + { + "epoch": 1.91, + "learning_rate": 8.554583073937086e-09, + "loss": 0.6911, + "step": 52328 + }, + { + "epoch": 1.91, + "learning_rate": 8.550921009265023e-09, + "loss": 0.6898, + "step": 52329 + }, + { + "epoch": 1.91, + "learning_rate": 8.547258944592961e-09, + "loss": 0.7024, + "step": 52330 + }, + { + "epoch": 1.91, + "learning_rate": 8.5435968799209e-09, + "loss": 0.7117, + "step": 52331 + }, + { + "epoch": 1.91, + "learning_rate": 8.539934815248836e-09, + "loss": 0.69, + "step": 52332 + }, + { + "epoch": 1.91, + "learning_rate": 8.536272750576774e-09, + "loss": 0.6894, + "step": 52333 + }, + { + "epoch": 1.91, + "learning_rate": 8.532610685904713e-09, + "loss": 0.6999, + "step": 52334 + }, + { + "epoch": 1.91, + "learning_rate": 8.528948621232651e-09, + "loss": 0.6729, + "step": 52335 + }, + { + "epoch": 1.91, + "learning_rate": 8.52528655656059e-09, + "loss": 0.6874, + "step": 52336 + }, + { + "epoch": 1.91, + "learning_rate": 8.521624491888526e-09, + "loss": 0.6996, + "step": 52337 + }, + { + "epoch": 1.91, + "learning_rate": 8.517962427216464e-09, + "loss": 0.6835, + "step": 52338 + }, + { + "epoch": 1.91, + "learning_rate": 8.514300362544402e-09, + "loss": 0.6888, + "step": 52339 + }, + { + "epoch": 1.91, + "learning_rate": 8.510638297872339e-09, + "loss": 0.7061, + "step": 52340 + }, + { + "epoch": 1.92, + "learning_rate": 8.506976233200277e-09, + "loss": 0.6966, + "step": 52341 + }, + { + "epoch": 1.92, + "learning_rate": 8.503314168528216e-09, + "loss": 0.6786, + "step": 52342 + }, + { + "epoch": 1.92, + "learning_rate": 8.499652103856154e-09, + "loss": 0.7023, + "step": 52343 + }, + { + "epoch": 1.92, + "learning_rate": 8.495990039184092e-09, + "loss": 0.7005, + "step": 52344 + }, + { + "epoch": 1.92, + "learning_rate": 8.49232797451203e-09, + "loss": 0.6988, + "step": 52345 + }, + { + "epoch": 1.92, + "learning_rate": 8.488665909839967e-09, + "loss": 0.6919, + "step": 52346 + }, + { + "epoch": 1.92, + "learning_rate": 8.485003845167905e-09, + "loss": 0.6869, + "step": 52347 + }, + { + "epoch": 1.92, + "learning_rate": 8.481341780495844e-09, + "loss": 0.6948, + "step": 52348 + }, + { + "epoch": 1.92, + "learning_rate": 8.47767971582378e-09, + "loss": 0.6886, + "step": 52349 + }, + { + "epoch": 1.92, + "learning_rate": 8.474017651151718e-09, + "loss": 0.7004, + "step": 52350 + }, + { + "epoch": 1.92, + "learning_rate": 8.470355586479657e-09, + "loss": 0.704, + "step": 52351 + }, + { + "epoch": 1.92, + "learning_rate": 8.466693521807593e-09, + "loss": 0.6899, + "step": 52352 + }, + { + "epoch": 1.92, + "learning_rate": 8.463031457135533e-09, + "loss": 0.6803, + "step": 52353 + }, + { + "epoch": 1.92, + "learning_rate": 8.459369392463472e-09, + "loss": 0.6964, + "step": 52354 + }, + { + "epoch": 1.92, + "learning_rate": 8.455707327791408e-09, + "loss": 0.6969, + "step": 52355 + }, + { + "epoch": 1.92, + "learning_rate": 8.452045263119346e-09, + "loss": 0.6933, + "step": 52356 + }, + { + "epoch": 1.92, + "learning_rate": 8.448383198447285e-09, + "loss": 0.7006, + "step": 52357 + }, + { + "epoch": 1.92, + "learning_rate": 8.444721133775221e-09, + "loss": 0.6831, + "step": 52358 + }, + { + "epoch": 1.92, + "learning_rate": 8.44105906910316e-09, + "loss": 0.7003, + "step": 52359 + }, + { + "epoch": 1.92, + "learning_rate": 8.437397004431098e-09, + "loss": 0.6995, + "step": 52360 + }, + { + "epoch": 1.92, + "learning_rate": 8.433734939759035e-09, + "loss": 0.7083, + "step": 52361 + }, + { + "epoch": 1.92, + "learning_rate": 8.430072875086974e-09, + "loss": 0.6908, + "step": 52362 + }, + { + "epoch": 1.92, + "learning_rate": 8.426410810414913e-09, + "loss": 0.6966, + "step": 52363 + }, + { + "epoch": 1.92, + "learning_rate": 8.42274874574285e-09, + "loss": 0.6869, + "step": 52364 + }, + { + "epoch": 1.92, + "learning_rate": 8.419086681070788e-09, + "loss": 0.6894, + "step": 52365 + }, + { + "epoch": 1.92, + "learning_rate": 8.415424616398726e-09, + "loss": 0.7029, + "step": 52366 + }, + { + "epoch": 1.92, + "learning_rate": 8.411762551726662e-09, + "loss": 0.6957, + "step": 52367 + }, + { + "epoch": 1.92, + "learning_rate": 8.4081004870546e-09, + "loss": 0.7058, + "step": 52368 + }, + { + "epoch": 1.92, + "learning_rate": 8.404438422382539e-09, + "loss": 0.6804, + "step": 52369 + }, + { + "epoch": 1.92, + "learning_rate": 8.400776357710476e-09, + "loss": 0.6988, + "step": 52370 + }, + { + "epoch": 1.92, + "learning_rate": 8.397114293038416e-09, + "loss": 0.6944, + "step": 52371 + }, + { + "epoch": 1.92, + "learning_rate": 8.393452228366354e-09, + "loss": 0.6929, + "step": 52372 + }, + { + "epoch": 1.92, + "learning_rate": 8.38979016369429e-09, + "loss": 0.6928, + "step": 52373 + }, + { + "epoch": 1.92, + "learning_rate": 8.386128099022229e-09, + "loss": 0.6934, + "step": 52374 + }, + { + "epoch": 1.92, + "learning_rate": 8.382466034350167e-09, + "loss": 0.6973, + "step": 52375 + }, + { + "epoch": 1.92, + "learning_rate": 8.378803969678104e-09, + "loss": 0.6945, + "step": 52376 + }, + { + "epoch": 1.92, + "learning_rate": 8.375141905006042e-09, + "loss": 0.6954, + "step": 52377 + }, + { + "epoch": 1.92, + "learning_rate": 8.37147984033398e-09, + "loss": 0.6965, + "step": 52378 + }, + { + "epoch": 1.92, + "learning_rate": 8.367817775661917e-09, + "loss": 0.6907, + "step": 52379 + }, + { + "epoch": 1.92, + "learning_rate": 8.364155710989857e-09, + "loss": 0.6922, + "step": 52380 + }, + { + "epoch": 1.92, + "learning_rate": 8.360493646317793e-09, + "loss": 0.6946, + "step": 52381 + }, + { + "epoch": 1.92, + "learning_rate": 8.356831581645732e-09, + "loss": 0.6868, + "step": 52382 + }, + { + "epoch": 1.92, + "learning_rate": 8.35316951697367e-09, + "loss": 0.6774, + "step": 52383 + }, + { + "epoch": 1.92, + "learning_rate": 8.349507452301607e-09, + "loss": 0.6986, + "step": 52384 + }, + { + "epoch": 1.92, + "learning_rate": 8.345845387629545e-09, + "loss": 0.6952, + "step": 52385 + }, + { + "epoch": 1.92, + "learning_rate": 8.342183322957483e-09, + "loss": 0.6979, + "step": 52386 + }, + { + "epoch": 1.92, + "learning_rate": 8.338521258285421e-09, + "loss": 0.6993, + "step": 52387 + }, + { + "epoch": 1.92, + "learning_rate": 8.334859193613358e-09, + "loss": 0.6987, + "step": 52388 + }, + { + "epoch": 1.92, + "learning_rate": 8.331197128941296e-09, + "loss": 0.6942, + "step": 52389 + }, + { + "epoch": 1.92, + "learning_rate": 8.327535064269235e-09, + "loss": 0.697, + "step": 52390 + }, + { + "epoch": 1.92, + "learning_rate": 8.323872999597173e-09, + "loss": 0.6989, + "step": 52391 + }, + { + "epoch": 1.92, + "learning_rate": 8.320210934925111e-09, + "loss": 0.6968, + "step": 52392 + }, + { + "epoch": 1.92, + "learning_rate": 8.316548870253048e-09, + "loss": 0.7053, + "step": 52393 + }, + { + "epoch": 1.92, + "learning_rate": 8.312886805580986e-09, + "loss": 0.6939, + "step": 52394 + }, + { + "epoch": 1.92, + "learning_rate": 8.309224740908924e-09, + "loss": 0.6996, + "step": 52395 + }, + { + "epoch": 1.92, + "learning_rate": 8.305562676236861e-09, + "loss": 0.6859, + "step": 52396 + }, + { + "epoch": 1.92, + "learning_rate": 8.301900611564799e-09, + "loss": 0.6942, + "step": 52397 + }, + { + "epoch": 1.92, + "learning_rate": 8.298238546892737e-09, + "loss": 0.6905, + "step": 52398 + }, + { + "epoch": 1.92, + "learning_rate": 8.294576482220676e-09, + "loss": 0.6936, + "step": 52399 + }, + { + "epoch": 1.92, + "learning_rate": 8.290914417548614e-09, + "loss": 0.7025, + "step": 52400 + }, + { + "epoch": 1.92, + "learning_rate": 8.287252352876552e-09, + "loss": 0.688, + "step": 52401 + }, + { + "epoch": 1.92, + "learning_rate": 8.283590288204489e-09, + "loss": 0.6981, + "step": 52402 + }, + { + "epoch": 1.92, + "learning_rate": 8.279928223532427e-09, + "loss": 0.6921, + "step": 52403 + }, + { + "epoch": 1.92, + "learning_rate": 8.276266158860365e-09, + "loss": 0.6823, + "step": 52404 + }, + { + "epoch": 1.92, + "learning_rate": 8.272604094188302e-09, + "loss": 0.6935, + "step": 52405 + }, + { + "epoch": 1.92, + "learning_rate": 8.26894202951624e-09, + "loss": 0.6957, + "step": 52406 + }, + { + "epoch": 1.92, + "learning_rate": 8.265279964844179e-09, + "loss": 0.6822, + "step": 52407 + }, + { + "epoch": 1.92, + "learning_rate": 8.261617900172117e-09, + "loss": 0.6913, + "step": 52408 + }, + { + "epoch": 1.92, + "learning_rate": 8.257955835500055e-09, + "loss": 0.7066, + "step": 52409 + }, + { + "epoch": 1.92, + "learning_rate": 8.254293770827993e-09, + "loss": 0.6947, + "step": 52410 + }, + { + "epoch": 1.92, + "learning_rate": 8.25063170615593e-09, + "loss": 0.6863, + "step": 52411 + }, + { + "epoch": 1.92, + "learning_rate": 8.246969641483868e-09, + "loss": 0.6789, + "step": 52412 + }, + { + "epoch": 1.92, + "learning_rate": 8.243307576811807e-09, + "loss": 0.6858, + "step": 52413 + }, + { + "epoch": 1.92, + "learning_rate": 8.239645512139743e-09, + "loss": 0.69, + "step": 52414 + }, + { + "epoch": 1.92, + "learning_rate": 8.235983447467681e-09, + "loss": 0.6951, + "step": 52415 + }, + { + "epoch": 1.92, + "learning_rate": 8.23232138279562e-09, + "loss": 0.7018, + "step": 52416 + }, + { + "epoch": 1.92, + "learning_rate": 8.228659318123558e-09, + "loss": 0.6996, + "step": 52417 + }, + { + "epoch": 1.92, + "learning_rate": 8.224997253451496e-09, + "loss": 0.6949, + "step": 52418 + }, + { + "epoch": 1.92, + "learning_rate": 8.221335188779435e-09, + "loss": 0.6918, + "step": 52419 + }, + { + "epoch": 1.92, + "learning_rate": 8.217673124107371e-09, + "loss": 0.6915, + "step": 52420 + }, + { + "epoch": 1.92, + "learning_rate": 8.21401105943531e-09, + "loss": 0.7004, + "step": 52421 + }, + { + "epoch": 1.92, + "learning_rate": 8.210348994763248e-09, + "loss": 0.6906, + "step": 52422 + }, + { + "epoch": 1.92, + "learning_rate": 8.206686930091184e-09, + "loss": 0.6907, + "step": 52423 + }, + { + "epoch": 1.92, + "learning_rate": 8.203024865419123e-09, + "loss": 0.6932, + "step": 52424 + }, + { + "epoch": 1.92, + "learning_rate": 8.199362800747061e-09, + "loss": 0.6913, + "step": 52425 + }, + { + "epoch": 1.92, + "learning_rate": 8.195700736075e-09, + "loss": 0.6974, + "step": 52426 + }, + { + "epoch": 1.92, + "learning_rate": 8.192038671402937e-09, + "loss": 0.6928, + "step": 52427 + }, + { + "epoch": 1.92, + "learning_rate": 8.188376606730874e-09, + "loss": 0.6961, + "step": 52428 + }, + { + "epoch": 1.92, + "learning_rate": 8.184714542058812e-09, + "loss": 0.6878, + "step": 52429 + }, + { + "epoch": 1.92, + "learning_rate": 8.18105247738675e-09, + "loss": 0.6932, + "step": 52430 + }, + { + "epoch": 1.92, + "learning_rate": 8.177390412714689e-09, + "loss": 0.7021, + "step": 52431 + }, + { + "epoch": 1.92, + "learning_rate": 8.173728348042626e-09, + "loss": 0.6945, + "step": 52432 + }, + { + "epoch": 1.92, + "learning_rate": 8.170066283370564e-09, + "loss": 0.6932, + "step": 52433 + }, + { + "epoch": 1.92, + "learning_rate": 8.166404218698502e-09, + "loss": 0.6876, + "step": 52434 + }, + { + "epoch": 1.92, + "learning_rate": 8.162742154026439e-09, + "loss": 0.685, + "step": 52435 + }, + { + "epoch": 1.92, + "learning_rate": 8.159080089354379e-09, + "loss": 0.6925, + "step": 52436 + }, + { + "epoch": 1.92, + "learning_rate": 8.155418024682315e-09, + "loss": 0.6945, + "step": 52437 + }, + { + "epoch": 1.92, + "learning_rate": 8.151755960010254e-09, + "loss": 0.691, + "step": 52438 + }, + { + "epoch": 1.92, + "learning_rate": 8.148093895338192e-09, + "loss": 0.6878, + "step": 52439 + }, + { + "epoch": 1.92, + "learning_rate": 8.144431830666128e-09, + "loss": 0.6861, + "step": 52440 + }, + { + "epoch": 1.92, + "learning_rate": 8.140769765994067e-09, + "loss": 0.6938, + "step": 52441 + }, + { + "epoch": 1.92, + "learning_rate": 8.137107701322005e-09, + "loss": 0.6808, + "step": 52442 + }, + { + "epoch": 1.92, + "learning_rate": 8.133445636649942e-09, + "loss": 0.6951, + "step": 52443 + }, + { + "epoch": 1.92, + "learning_rate": 8.12978357197788e-09, + "loss": 0.6785, + "step": 52444 + }, + { + "epoch": 1.92, + "learning_rate": 8.12612150730582e-09, + "loss": 0.6904, + "step": 52445 + }, + { + "epoch": 1.92, + "learning_rate": 8.122459442633756e-09, + "loss": 0.6908, + "step": 52446 + }, + { + "epoch": 1.92, + "learning_rate": 8.118797377961695e-09, + "loss": 0.686, + "step": 52447 + }, + { + "epoch": 1.92, + "learning_rate": 8.115135313289633e-09, + "loss": 0.6926, + "step": 52448 + }, + { + "epoch": 1.92, + "learning_rate": 8.11147324861757e-09, + "loss": 0.7042, + "step": 52449 + }, + { + "epoch": 1.92, + "learning_rate": 8.107811183945508e-09, + "loss": 0.6966, + "step": 52450 + }, + { + "epoch": 1.92, + "learning_rate": 8.104149119273446e-09, + "loss": 0.6935, + "step": 52451 + }, + { + "epoch": 1.92, + "learning_rate": 8.100487054601383e-09, + "loss": 0.6913, + "step": 52452 + }, + { + "epoch": 1.92, + "learning_rate": 8.096824989929321e-09, + "loss": 0.699, + "step": 52453 + }, + { + "epoch": 1.92, + "learning_rate": 8.093162925257261e-09, + "loss": 0.6945, + "step": 52454 + }, + { + "epoch": 1.92, + "learning_rate": 8.089500860585198e-09, + "loss": 0.7078, + "step": 52455 + }, + { + "epoch": 1.92, + "learning_rate": 8.085838795913136e-09, + "loss": 0.701, + "step": 52456 + }, + { + "epoch": 1.92, + "learning_rate": 8.082176731241074e-09, + "loss": 0.6954, + "step": 52457 + }, + { + "epoch": 1.92, + "learning_rate": 8.07851466656901e-09, + "loss": 0.6983, + "step": 52458 + }, + { + "epoch": 1.92, + "learning_rate": 8.074852601896949e-09, + "loss": 0.694, + "step": 52459 + }, + { + "epoch": 1.92, + "learning_rate": 8.071190537224887e-09, + "loss": 0.6808, + "step": 52460 + }, + { + "epoch": 1.92, + "learning_rate": 8.067528472552824e-09, + "loss": 0.6953, + "step": 52461 + }, + { + "epoch": 1.92, + "learning_rate": 8.063866407880762e-09, + "loss": 0.6806, + "step": 52462 + }, + { + "epoch": 1.92, + "learning_rate": 8.060204343208702e-09, + "loss": 0.6962, + "step": 52463 + }, + { + "epoch": 1.92, + "learning_rate": 8.056542278536639e-09, + "loss": 0.6973, + "step": 52464 + }, + { + "epoch": 1.92, + "learning_rate": 8.052880213864577e-09, + "loss": 0.6843, + "step": 52465 + }, + { + "epoch": 1.92, + "learning_rate": 8.049218149192515e-09, + "loss": 0.6972, + "step": 52466 + }, + { + "epoch": 1.92, + "learning_rate": 8.045556084520452e-09, + "loss": 0.6931, + "step": 52467 + }, + { + "epoch": 1.92, + "learning_rate": 8.04189401984839e-09, + "loss": 0.6882, + "step": 52468 + }, + { + "epoch": 1.92, + "learning_rate": 8.038231955176328e-09, + "loss": 0.7029, + "step": 52469 + }, + { + "epoch": 1.92, + "learning_rate": 8.034569890504265e-09, + "loss": 0.6909, + "step": 52470 + }, + { + "epoch": 1.92, + "learning_rate": 8.030907825832203e-09, + "loss": 0.6859, + "step": 52471 + }, + { + "epoch": 1.92, + "learning_rate": 8.027245761160142e-09, + "loss": 0.6897, + "step": 52472 + }, + { + "epoch": 1.92, + "learning_rate": 8.02358369648808e-09, + "loss": 0.6914, + "step": 52473 + }, + { + "epoch": 1.92, + "learning_rate": 8.019921631816018e-09, + "loss": 0.6855, + "step": 52474 + }, + { + "epoch": 1.92, + "learning_rate": 8.016259567143956e-09, + "loss": 0.6903, + "step": 52475 + }, + { + "epoch": 1.92, + "learning_rate": 8.012597502471893e-09, + "loss": 0.7038, + "step": 52476 + }, + { + "epoch": 1.92, + "learning_rate": 8.008935437799831e-09, + "loss": 0.7065, + "step": 52477 + }, + { + "epoch": 1.92, + "learning_rate": 8.00527337312777e-09, + "loss": 0.6884, + "step": 52478 + }, + { + "epoch": 1.92, + "learning_rate": 8.001611308455706e-09, + "loss": 0.7014, + "step": 52479 + }, + { + "epoch": 1.92, + "learning_rate": 7.997949243783644e-09, + "loss": 0.6973, + "step": 52480 + }, + { + "epoch": 1.92, + "learning_rate": 7.994287179111583e-09, + "loss": 0.7001, + "step": 52481 + }, + { + "epoch": 1.92, + "learning_rate": 7.990625114439521e-09, + "loss": 0.6914, + "step": 52482 + }, + { + "epoch": 1.92, + "learning_rate": 7.98696304976746e-09, + "loss": 0.691, + "step": 52483 + }, + { + "epoch": 1.92, + "learning_rate": 7.983300985095396e-09, + "loss": 0.7027, + "step": 52484 + }, + { + "epoch": 1.92, + "learning_rate": 7.979638920423334e-09, + "loss": 0.6911, + "step": 52485 + }, + { + "epoch": 1.92, + "learning_rate": 7.975976855751272e-09, + "loss": 0.6871, + "step": 52486 + }, + { + "epoch": 1.92, + "learning_rate": 7.972314791079209e-09, + "loss": 0.6964, + "step": 52487 + }, + { + "epoch": 1.92, + "learning_rate": 7.968652726407147e-09, + "loss": 0.7042, + "step": 52488 + }, + { + "epoch": 1.92, + "learning_rate": 7.964990661735086e-09, + "loss": 0.693, + "step": 52489 + }, + { + "epoch": 1.92, + "learning_rate": 7.961328597063024e-09, + "loss": 0.6824, + "step": 52490 + }, + { + "epoch": 1.92, + "learning_rate": 7.957666532390962e-09, + "loss": 0.6882, + "step": 52491 + }, + { + "epoch": 1.92, + "learning_rate": 7.9540044677189e-09, + "loss": 0.7027, + "step": 52492 + }, + { + "epoch": 1.92, + "learning_rate": 7.950342403046837e-09, + "loss": 0.7005, + "step": 52493 + }, + { + "epoch": 1.92, + "learning_rate": 7.946680338374775e-09, + "loss": 0.6916, + "step": 52494 + }, + { + "epoch": 1.92, + "learning_rate": 7.943018273702714e-09, + "loss": 0.6977, + "step": 52495 + }, + { + "epoch": 1.92, + "learning_rate": 7.93935620903065e-09, + "loss": 0.6927, + "step": 52496 + }, + { + "epoch": 1.92, + "learning_rate": 7.935694144358589e-09, + "loss": 0.6955, + "step": 52497 + }, + { + "epoch": 1.92, + "learning_rate": 7.932032079686527e-09, + "loss": 0.6917, + "step": 52498 + }, + { + "epoch": 1.92, + "learning_rate": 7.928370015014463e-09, + "loss": 0.7001, + "step": 52499 + }, + { + "epoch": 1.92, + "learning_rate": 7.924707950342403e-09, + "loss": 0.7013, + "step": 52500 + }, + { + "epoch": 1.92, + "learning_rate": 7.921045885670342e-09, + "loss": 0.6929, + "step": 52501 + }, + { + "epoch": 1.92, + "learning_rate": 7.917383820998278e-09, + "loss": 0.6873, + "step": 52502 + }, + { + "epoch": 1.92, + "learning_rate": 7.913721756326217e-09, + "loss": 0.7017, + "step": 52503 + }, + { + "epoch": 1.92, + "learning_rate": 7.910059691654155e-09, + "loss": 0.7026, + "step": 52504 + }, + { + "epoch": 1.92, + "learning_rate": 7.906397626982091e-09, + "loss": 0.6879, + "step": 52505 + }, + { + "epoch": 1.92, + "learning_rate": 7.90273556231003e-09, + "loss": 0.6966, + "step": 52506 + }, + { + "epoch": 1.92, + "learning_rate": 7.899073497637968e-09, + "loss": 0.6936, + "step": 52507 + }, + { + "epoch": 1.92, + "learning_rate": 7.895411432965905e-09, + "loss": 0.6965, + "step": 52508 + }, + { + "epoch": 1.92, + "learning_rate": 7.891749368293845e-09, + "loss": 0.6963, + "step": 52509 + }, + { + "epoch": 1.92, + "learning_rate": 7.888087303621783e-09, + "loss": 0.6934, + "step": 52510 + }, + { + "epoch": 1.92, + "learning_rate": 7.88442523894972e-09, + "loss": 0.6882, + "step": 52511 + }, + { + "epoch": 1.92, + "learning_rate": 7.880763174277658e-09, + "loss": 0.6885, + "step": 52512 + }, + { + "epoch": 1.92, + "learning_rate": 7.877101109605596e-09, + "loss": 0.6986, + "step": 52513 + }, + { + "epoch": 1.92, + "learning_rate": 7.873439044933533e-09, + "loss": 0.6838, + "step": 52514 + }, + { + "epoch": 1.92, + "learning_rate": 7.869776980261471e-09, + "loss": 0.6854, + "step": 52515 + }, + { + "epoch": 1.92, + "learning_rate": 7.866114915589409e-09, + "loss": 0.6881, + "step": 52516 + }, + { + "epoch": 1.92, + "learning_rate": 7.862452850917346e-09, + "loss": 0.6956, + "step": 52517 + }, + { + "epoch": 1.92, + "learning_rate": 7.858790786245286e-09, + "loss": 0.6947, + "step": 52518 + }, + { + "epoch": 1.92, + "learning_rate": 7.855128721573224e-09, + "loss": 0.6976, + "step": 52519 + }, + { + "epoch": 1.92, + "learning_rate": 7.85146665690116e-09, + "loss": 0.6875, + "step": 52520 + }, + { + "epoch": 1.92, + "learning_rate": 7.847804592229099e-09, + "loss": 0.6986, + "step": 52521 + }, + { + "epoch": 1.92, + "learning_rate": 7.844142527557037e-09, + "loss": 0.6974, + "step": 52522 + }, + { + "epoch": 1.92, + "learning_rate": 7.840480462884974e-09, + "loss": 0.6918, + "step": 52523 + }, + { + "epoch": 1.92, + "learning_rate": 7.836818398212912e-09, + "loss": 0.6955, + "step": 52524 + }, + { + "epoch": 1.92, + "learning_rate": 7.83315633354085e-09, + "loss": 0.6825, + "step": 52525 + }, + { + "epoch": 1.92, + "learning_rate": 7.829494268868787e-09, + "loss": 0.6947, + "step": 52526 + }, + { + "epoch": 1.92, + "learning_rate": 7.825832204196725e-09, + "loss": 0.6839, + "step": 52527 + }, + { + "epoch": 1.92, + "learning_rate": 7.822170139524663e-09, + "loss": 0.6941, + "step": 52528 + }, + { + "epoch": 1.92, + "learning_rate": 7.818508074852602e-09, + "loss": 0.6939, + "step": 52529 + }, + { + "epoch": 1.92, + "learning_rate": 7.81484601018054e-09, + "loss": 0.6935, + "step": 52530 + }, + { + "epoch": 1.92, + "learning_rate": 7.811183945508477e-09, + "loss": 0.6854, + "step": 52531 + }, + { + "epoch": 1.92, + "learning_rate": 7.807521880836415e-09, + "loss": 0.6816, + "step": 52532 + }, + { + "epoch": 1.92, + "learning_rate": 7.803859816164353e-09, + "loss": 0.6937, + "step": 52533 + }, + { + "epoch": 1.92, + "learning_rate": 7.800197751492291e-09, + "loss": 0.6992, + "step": 52534 + }, + { + "epoch": 1.92, + "learning_rate": 7.796535686820228e-09, + "loss": 0.6926, + "step": 52535 + }, + { + "epoch": 1.92, + "learning_rate": 7.792873622148166e-09, + "loss": 0.6823, + "step": 52536 + }, + { + "epoch": 1.92, + "learning_rate": 7.789211557476105e-09, + "loss": 0.6934, + "step": 52537 + }, + { + "epoch": 1.92, + "learning_rate": 7.785549492804043e-09, + "loss": 0.6909, + "step": 52538 + }, + { + "epoch": 1.92, + "learning_rate": 7.781887428131981e-09, + "loss": 0.6946, + "step": 52539 + }, + { + "epoch": 1.92, + "learning_rate": 7.778225363459918e-09, + "loss": 0.6992, + "step": 52540 + }, + { + "epoch": 1.92, + "learning_rate": 7.774563298787856e-09, + "loss": 0.6933, + "step": 52541 + }, + { + "epoch": 1.92, + "learning_rate": 7.770901234115794e-09, + "loss": 0.7063, + "step": 52542 + }, + { + "epoch": 1.92, + "learning_rate": 7.767239169443731e-09, + "loss": 0.6993, + "step": 52543 + }, + { + "epoch": 1.92, + "learning_rate": 7.76357710477167e-09, + "loss": 0.6876, + "step": 52544 + }, + { + "epoch": 1.92, + "learning_rate": 7.759915040099607e-09, + "loss": 0.6876, + "step": 52545 + }, + { + "epoch": 1.92, + "learning_rate": 7.756252975427546e-09, + "loss": 0.6898, + "step": 52546 + }, + { + "epoch": 1.92, + "learning_rate": 7.752590910755484e-09, + "loss": 0.6897, + "step": 52547 + }, + { + "epoch": 1.92, + "learning_rate": 7.748928846083422e-09, + "loss": 0.6906, + "step": 52548 + }, + { + "epoch": 1.92, + "learning_rate": 7.745266781411359e-09, + "loss": 0.7031, + "step": 52549 + }, + { + "epoch": 1.92, + "learning_rate": 7.741604716739297e-09, + "loss": 0.6974, + "step": 52550 + }, + { + "epoch": 1.92, + "learning_rate": 7.737942652067235e-09, + "loss": 0.6955, + "step": 52551 + }, + { + "epoch": 1.92, + "learning_rate": 7.734280587395172e-09, + "loss": 0.6813, + "step": 52552 + }, + { + "epoch": 1.92, + "learning_rate": 7.73061852272311e-09, + "loss": 0.7006, + "step": 52553 + }, + { + "epoch": 1.92, + "learning_rate": 7.726956458051049e-09, + "loss": 0.6857, + "step": 52554 + }, + { + "epoch": 1.92, + "learning_rate": 7.723294393378987e-09, + "loss": 0.6913, + "step": 52555 + }, + { + "epoch": 1.92, + "learning_rate": 7.719632328706925e-09, + "loss": 0.6924, + "step": 52556 + }, + { + "epoch": 1.92, + "learning_rate": 7.715970264034863e-09, + "loss": 0.6988, + "step": 52557 + }, + { + "epoch": 1.92, + "learning_rate": 7.7123081993628e-09, + "loss": 0.6944, + "step": 52558 + }, + { + "epoch": 1.92, + "learning_rate": 7.708646134690738e-09, + "loss": 0.6935, + "step": 52559 + }, + { + "epoch": 1.92, + "learning_rate": 7.704984070018677e-09, + "loss": 0.6806, + "step": 52560 + }, + { + "epoch": 1.92, + "learning_rate": 7.701322005346613e-09, + "loss": 0.7032, + "step": 52561 + }, + { + "epoch": 1.92, + "learning_rate": 7.697659940674552e-09, + "loss": 0.6882, + "step": 52562 + }, + { + "epoch": 1.92, + "learning_rate": 7.69399787600249e-09, + "loss": 0.6879, + "step": 52563 + }, + { + "epoch": 1.92, + "learning_rate": 7.690335811330428e-09, + "loss": 0.6942, + "step": 52564 + }, + { + "epoch": 1.92, + "learning_rate": 7.686673746658366e-09, + "loss": 0.6977, + "step": 52565 + }, + { + "epoch": 1.92, + "learning_rate": 7.683011681986305e-09, + "loss": 0.6927, + "step": 52566 + }, + { + "epoch": 1.92, + "learning_rate": 7.679349617314241e-09, + "loss": 0.6733, + "step": 52567 + }, + { + "epoch": 1.92, + "learning_rate": 7.67568755264218e-09, + "loss": 0.696, + "step": 52568 + }, + { + "epoch": 1.92, + "learning_rate": 7.672025487970118e-09, + "loss": 0.6963, + "step": 52569 + }, + { + "epoch": 1.92, + "learning_rate": 7.668363423298054e-09, + "loss": 0.6979, + "step": 52570 + }, + { + "epoch": 1.92, + "learning_rate": 7.664701358625993e-09, + "loss": 0.684, + "step": 52571 + }, + { + "epoch": 1.92, + "learning_rate": 7.661039293953931e-09, + "loss": 0.691, + "step": 52572 + }, + { + "epoch": 1.92, + "learning_rate": 7.657377229281868e-09, + "loss": 0.7031, + "step": 52573 + }, + { + "epoch": 1.92, + "learning_rate": 7.653715164609808e-09, + "loss": 0.6868, + "step": 52574 + }, + { + "epoch": 1.92, + "learning_rate": 7.650053099937744e-09, + "loss": 0.6909, + "step": 52575 + }, + { + "epoch": 1.92, + "learning_rate": 7.646391035265682e-09, + "loss": 0.701, + "step": 52576 + }, + { + "epoch": 1.92, + "learning_rate": 7.64272897059362e-09, + "loss": 0.7008, + "step": 52577 + }, + { + "epoch": 1.92, + "learning_rate": 7.639066905921559e-09, + "loss": 0.6964, + "step": 52578 + }, + { + "epoch": 1.92, + "learning_rate": 7.635404841249496e-09, + "loss": 0.6845, + "step": 52579 + }, + { + "epoch": 1.92, + "learning_rate": 7.631742776577434e-09, + "loss": 0.6837, + "step": 52580 + }, + { + "epoch": 1.92, + "learning_rate": 7.628080711905372e-09, + "loss": 0.684, + "step": 52581 + }, + { + "epoch": 1.92, + "learning_rate": 7.624418647233309e-09, + "loss": 0.6963, + "step": 52582 + }, + { + "epoch": 1.92, + "learning_rate": 7.620756582561249e-09, + "loss": 0.6926, + "step": 52583 + }, + { + "epoch": 1.92, + "learning_rate": 7.617094517889185e-09, + "loss": 0.6857, + "step": 52584 + }, + { + "epoch": 1.92, + "learning_rate": 7.613432453217124e-09, + "loss": 0.6992, + "step": 52585 + }, + { + "epoch": 1.92, + "learning_rate": 7.609770388545062e-09, + "loss": 0.6917, + "step": 52586 + }, + { + "epoch": 1.92, + "learning_rate": 7.606108323872998e-09, + "loss": 0.7129, + "step": 52587 + }, + { + "epoch": 1.92, + "learning_rate": 7.602446259200937e-09, + "loss": 0.6874, + "step": 52588 + }, + { + "epoch": 1.92, + "learning_rate": 7.598784194528875e-09, + "loss": 0.6926, + "step": 52589 + }, + { + "epoch": 1.92, + "learning_rate": 7.595122129856812e-09, + "loss": 0.6958, + "step": 52590 + }, + { + "epoch": 1.92, + "learning_rate": 7.59146006518475e-09, + "loss": 0.6904, + "step": 52591 + }, + { + "epoch": 1.92, + "learning_rate": 7.58779800051269e-09, + "loss": 0.681, + "step": 52592 + }, + { + "epoch": 1.92, + "learning_rate": 7.584135935840626e-09, + "loss": 0.6917, + "step": 52593 + }, + { + "epoch": 1.92, + "learning_rate": 7.580473871168565e-09, + "loss": 0.702, + "step": 52594 + }, + { + "epoch": 1.92, + "learning_rate": 7.576811806496503e-09, + "loss": 0.6957, + "step": 52595 + }, + { + "epoch": 1.92, + "learning_rate": 7.57314974182444e-09, + "loss": 0.6857, + "step": 52596 + }, + { + "epoch": 1.92, + "learning_rate": 7.569487677152378e-09, + "loss": 0.7007, + "step": 52597 + }, + { + "epoch": 1.92, + "learning_rate": 7.565825612480316e-09, + "loss": 0.6974, + "step": 52598 + }, + { + "epoch": 1.92, + "learning_rate": 7.562163547808253e-09, + "loss": 0.6976, + "step": 52599 + }, + { + "epoch": 1.92, + "learning_rate": 7.558501483136191e-09, + "loss": 0.6855, + "step": 52600 + }, + { + "epoch": 1.92, + "learning_rate": 7.554839418464131e-09, + "loss": 0.687, + "step": 52601 + }, + { + "epoch": 1.92, + "learning_rate": 7.551177353792068e-09, + "loss": 0.6989, + "step": 52602 + }, + { + "epoch": 1.92, + "learning_rate": 7.547515289120006e-09, + "loss": 0.6979, + "step": 52603 + }, + { + "epoch": 1.92, + "learning_rate": 7.543853224447944e-09, + "loss": 0.6838, + "step": 52604 + }, + { + "epoch": 1.92, + "learning_rate": 7.54019115977588e-09, + "loss": 0.7086, + "step": 52605 + }, + { + "epoch": 1.92, + "learning_rate": 7.536529095103819e-09, + "loss": 0.6925, + "step": 52606 + }, + { + "epoch": 1.92, + "learning_rate": 7.532867030431757e-09, + "loss": 0.6884, + "step": 52607 + }, + { + "epoch": 1.92, + "learning_rate": 7.529204965759694e-09, + "loss": 0.7007, + "step": 52608 + }, + { + "epoch": 1.92, + "learning_rate": 7.525542901087632e-09, + "loss": 0.703, + "step": 52609 + }, + { + "epoch": 1.92, + "learning_rate": 7.521880836415572e-09, + "loss": 0.6948, + "step": 52610 + }, + { + "epoch": 1.92, + "learning_rate": 7.518218771743509e-09, + "loss": 0.7101, + "step": 52611 + }, + { + "epoch": 1.92, + "learning_rate": 7.514556707071447e-09, + "loss": 0.704, + "step": 52612 + }, + { + "epoch": 1.92, + "learning_rate": 7.510894642399385e-09, + "loss": 0.6947, + "step": 52613 + }, + { + "epoch": 1.92, + "learning_rate": 7.507232577727322e-09, + "loss": 0.6834, + "step": 52614 + }, + { + "epoch": 1.93, + "learning_rate": 7.50357051305526e-09, + "loss": 0.6959, + "step": 52615 + }, + { + "epoch": 1.93, + "learning_rate": 7.499908448383198e-09, + "loss": 0.6927, + "step": 52616 + }, + { + "epoch": 1.93, + "learning_rate": 7.496246383711135e-09, + "loss": 0.6991, + "step": 52617 + }, + { + "epoch": 1.93, + "learning_rate": 7.492584319039073e-09, + "loss": 0.6881, + "step": 52618 + }, + { + "epoch": 1.93, + "learning_rate": 7.488922254367012e-09, + "loss": 0.6873, + "step": 52619 + }, + { + "epoch": 1.93, + "learning_rate": 7.48526018969495e-09, + "loss": 0.6907, + "step": 52620 + }, + { + "epoch": 1.93, + "learning_rate": 7.481598125022888e-09, + "loss": 0.6946, + "step": 52621 + }, + { + "epoch": 1.93, + "learning_rate": 7.477936060350826e-09, + "loss": 0.6969, + "step": 52622 + }, + { + "epoch": 1.93, + "learning_rate": 7.474273995678763e-09, + "loss": 0.7087, + "step": 52623 + }, + { + "epoch": 1.93, + "learning_rate": 7.470611931006701e-09, + "loss": 0.7013, + "step": 52624 + }, + { + "epoch": 1.93, + "learning_rate": 7.46694986633464e-09, + "loss": 0.683, + "step": 52625 + }, + { + "epoch": 1.93, + "learning_rate": 7.463287801662576e-09, + "loss": 0.6885, + "step": 52626 + }, + { + "epoch": 1.93, + "learning_rate": 7.459625736990515e-09, + "loss": 0.6981, + "step": 52627 + }, + { + "epoch": 1.93, + "learning_rate": 7.455963672318453e-09, + "loss": 0.6868, + "step": 52628 + }, + { + "epoch": 1.93, + "learning_rate": 7.452301607646391e-09, + "loss": 0.6913, + "step": 52629 + }, + { + "epoch": 1.93, + "learning_rate": 7.448639542974329e-09, + "loss": 0.688, + "step": 52630 + }, + { + "epoch": 1.93, + "learning_rate": 7.444977478302267e-09, + "loss": 0.6898, + "step": 52631 + }, + { + "epoch": 1.93, + "learning_rate": 7.441315413630204e-09, + "loss": 0.7019, + "step": 52632 + }, + { + "epoch": 1.93, + "learning_rate": 7.4376533489581425e-09, + "loss": 0.6944, + "step": 52633 + }, + { + "epoch": 1.93, + "learning_rate": 7.43399128428608e-09, + "loss": 0.6929, + "step": 52634 + }, + { + "epoch": 1.93, + "learning_rate": 7.4303292196140174e-09, + "loss": 0.6967, + "step": 52635 + }, + { + "epoch": 1.93, + "learning_rate": 7.426667154941956e-09, + "loss": 0.6911, + "step": 52636 + }, + { + "epoch": 1.93, + "learning_rate": 7.423005090269893e-09, + "loss": 0.689, + "step": 52637 + }, + { + "epoch": 1.93, + "learning_rate": 7.419343025597832e-09, + "loss": 0.687, + "step": 52638 + }, + { + "epoch": 1.93, + "learning_rate": 7.41568096092577e-09, + "loss": 0.6818, + "step": 52639 + }, + { + "epoch": 1.93, + "learning_rate": 7.412018896253708e-09, + "loss": 0.6895, + "step": 52640 + }, + { + "epoch": 1.93, + "learning_rate": 7.4083568315816454e-09, + "loss": 0.7046, + "step": 52641 + }, + { + "epoch": 1.93, + "learning_rate": 7.404694766909583e-09, + "loss": 0.6885, + "step": 52642 + }, + { + "epoch": 1.93, + "learning_rate": 7.401032702237521e-09, + "loss": 0.6894, + "step": 52643 + }, + { + "epoch": 1.93, + "learning_rate": 7.397370637565459e-09, + "loss": 0.6948, + "step": 52644 + }, + { + "epoch": 1.93, + "learning_rate": 7.393708572893397e-09, + "loss": 0.6914, + "step": 52645 + }, + { + "epoch": 1.93, + "learning_rate": 7.390046508221334e-09, + "loss": 0.6988, + "step": 52646 + }, + { + "epoch": 1.93, + "learning_rate": 7.386384443549273e-09, + "loss": 0.6956, + "step": 52647 + }, + { + "epoch": 1.93, + "learning_rate": 7.382722378877211e-09, + "loss": 0.692, + "step": 52648 + }, + { + "epoch": 1.93, + "learning_rate": 7.379060314205149e-09, + "loss": 0.6928, + "step": 52649 + }, + { + "epoch": 1.93, + "learning_rate": 7.3753982495330866e-09, + "loss": 0.6886, + "step": 52650 + }, + { + "epoch": 1.93, + "learning_rate": 7.371736184861024e-09, + "loss": 0.6899, + "step": 52651 + }, + { + "epoch": 1.93, + "learning_rate": 7.368074120188962e-09, + "loss": 0.6891, + "step": 52652 + }, + { + "epoch": 1.93, + "learning_rate": 7.3644120555169e-09, + "loss": 0.6906, + "step": 52653 + }, + { + "epoch": 1.93, + "learning_rate": 7.360749990844837e-09, + "loss": 0.7, + "step": 52654 + }, + { + "epoch": 1.93, + "learning_rate": 7.3570879261727755e-09, + "loss": 0.6901, + "step": 52655 + }, + { + "epoch": 1.93, + "learning_rate": 7.3534258615007146e-09, + "loss": 0.7008, + "step": 52656 + }, + { + "epoch": 1.93, + "learning_rate": 7.349763796828652e-09, + "loss": 0.6922, + "step": 52657 + }, + { + "epoch": 1.93, + "learning_rate": 7.34610173215659e-09, + "loss": 0.6937, + "step": 52658 + }, + { + "epoch": 1.93, + "learning_rate": 7.342439667484528e-09, + "loss": 0.6996, + "step": 52659 + }, + { + "epoch": 1.93, + "learning_rate": 7.338777602812465e-09, + "loss": 0.6959, + "step": 52660 + }, + { + "epoch": 1.93, + "learning_rate": 7.3351155381404035e-09, + "loss": 0.6961, + "step": 52661 + }, + { + "epoch": 1.93, + "learning_rate": 7.331453473468341e-09, + "loss": 0.685, + "step": 52662 + }, + { + "epoch": 1.93, + "learning_rate": 7.327791408796278e-09, + "loss": 0.6896, + "step": 52663 + }, + { + "epoch": 1.93, + "learning_rate": 7.324129344124217e-09, + "loss": 0.6945, + "step": 52664 + }, + { + "epoch": 1.93, + "learning_rate": 7.320467279452156e-09, + "loss": 0.6865, + "step": 52665 + }, + { + "epoch": 1.93, + "learning_rate": 7.316805214780093e-09, + "loss": 0.7029, + "step": 52666 + }, + { + "epoch": 1.93, + "learning_rate": 7.313143150108031e-09, + "loss": 0.6794, + "step": 52667 + }, + { + "epoch": 1.93, + "learning_rate": 7.309481085435969e-09, + "loss": 0.6848, + "step": 52668 + }, + { + "epoch": 1.93, + "learning_rate": 7.305819020763906e-09, + "loss": 0.6888, + "step": 52669 + }, + { + "epoch": 1.93, + "learning_rate": 7.302156956091844e-09, + "loss": 0.687, + "step": 52670 + }, + { + "epoch": 1.93, + "learning_rate": 7.298494891419782e-09, + "loss": 0.6758, + "step": 52671 + }, + { + "epoch": 1.93, + "learning_rate": 7.2948328267477195e-09, + "loss": 0.6965, + "step": 52672 + }, + { + "epoch": 1.93, + "learning_rate": 7.291170762075658e-09, + "loss": 0.6875, + "step": 52673 + }, + { + "epoch": 1.93, + "learning_rate": 7.287508697403595e-09, + "loss": 0.6926, + "step": 52674 + }, + { + "epoch": 1.93, + "learning_rate": 7.283846632731534e-09, + "loss": 0.6965, + "step": 52675 + }, + { + "epoch": 1.93, + "learning_rate": 7.280184568059472e-09, + "loss": 0.7088, + "step": 52676 + }, + { + "epoch": 1.93, + "learning_rate": 7.27652250338741e-09, + "loss": 0.6859, + "step": 52677 + }, + { + "epoch": 1.93, + "learning_rate": 7.2728604387153475e-09, + "loss": 0.6901, + "step": 52678 + }, + { + "epoch": 1.93, + "learning_rate": 7.269198374043285e-09, + "loss": 0.6882, + "step": 52679 + }, + { + "epoch": 1.93, + "learning_rate": 7.265536309371223e-09, + "loss": 0.6972, + "step": 52680 + }, + { + "epoch": 1.93, + "learning_rate": 7.261874244699161e-09, + "loss": 0.7027, + "step": 52681 + }, + { + "epoch": 1.93, + "learning_rate": 7.258212180027098e-09, + "loss": 0.6917, + "step": 52682 + }, + { + "epoch": 1.93, + "learning_rate": 7.254550115355036e-09, + "loss": 0.6962, + "step": 52683 + }, + { + "epoch": 1.93, + "learning_rate": 7.2508880506829755e-09, + "loss": 0.7098, + "step": 52684 + }, + { + "epoch": 1.93, + "learning_rate": 7.247225986010913e-09, + "loss": 0.6968, + "step": 52685 + }, + { + "epoch": 1.93, + "learning_rate": 7.24356392133885e-09, + "loss": 0.6817, + "step": 52686 + }, + { + "epoch": 1.93, + "learning_rate": 7.239901856666789e-09, + "loss": 0.682, + "step": 52687 + }, + { + "epoch": 1.93, + "learning_rate": 7.236239791994726e-09, + "loss": 0.7098, + "step": 52688 + }, + { + "epoch": 1.93, + "learning_rate": 7.232577727322664e-09, + "loss": 0.6814, + "step": 52689 + }, + { + "epoch": 1.93, + "learning_rate": 7.228915662650602e-09, + "loss": 0.6878, + "step": 52690 + }, + { + "epoch": 1.93, + "learning_rate": 7.225253597978539e-09, + "loss": 0.6994, + "step": 52691 + }, + { + "epoch": 1.93, + "learning_rate": 7.2215915333064776e-09, + "loss": 0.6926, + "step": 52692 + }, + { + "epoch": 1.93, + "learning_rate": 7.217929468634417e-09, + "loss": 0.6874, + "step": 52693 + }, + { + "epoch": 1.93, + "learning_rate": 7.214267403962354e-09, + "loss": 0.6906, + "step": 52694 + }, + { + "epoch": 1.93, + "learning_rate": 7.2106053392902916e-09, + "loss": 0.7089, + "step": 52695 + }, + { + "epoch": 1.93, + "learning_rate": 7.20694327461823e-09, + "loss": 0.698, + "step": 52696 + }, + { + "epoch": 1.93, + "learning_rate": 7.203281209946167e-09, + "loss": 0.684, + "step": 52697 + }, + { + "epoch": 1.93, + "learning_rate": 7.199619145274105e-09, + "loss": 0.6831, + "step": 52698 + }, + { + "epoch": 1.93, + "learning_rate": 7.195957080602043e-09, + "loss": 0.6862, + "step": 52699 + }, + { + "epoch": 1.93, + "learning_rate": 7.1922950159299804e-09, + "loss": 0.7015, + "step": 52700 + }, + { + "epoch": 1.93, + "learning_rate": 7.188632951257918e-09, + "loss": 0.6952, + "step": 52701 + }, + { + "epoch": 1.93, + "learning_rate": 7.184970886585858e-09, + "loss": 0.6935, + "step": 52702 + }, + { + "epoch": 1.93, + "learning_rate": 7.181308821913795e-09, + "loss": 0.6944, + "step": 52703 + }, + { + "epoch": 1.93, + "learning_rate": 7.177646757241733e-09, + "loss": 0.6982, + "step": 52704 + }, + { + "epoch": 1.93, + "learning_rate": 7.173984692569671e-09, + "loss": 0.6872, + "step": 52705 + }, + { + "epoch": 1.93, + "learning_rate": 7.1703226278976084e-09, + "loss": 0.6955, + "step": 52706 + }, + { + "epoch": 1.93, + "learning_rate": 7.166660563225546e-09, + "loss": 0.6924, + "step": 52707 + }, + { + "epoch": 1.93, + "learning_rate": 7.162998498553484e-09, + "loss": 0.6953, + "step": 52708 + }, + { + "epoch": 1.93, + "learning_rate": 7.159336433881422e-09, + "loss": 0.6944, + "step": 52709 + }, + { + "epoch": 1.93, + "learning_rate": 7.155674369209359e-09, + "loss": 0.6974, + "step": 52710 + }, + { + "epoch": 1.93, + "learning_rate": 7.152012304537298e-09, + "loss": 0.6981, + "step": 52711 + }, + { + "epoch": 1.93, + "learning_rate": 7.1483502398652364e-09, + "loss": 0.7029, + "step": 52712 + }, + { + "epoch": 1.93, + "learning_rate": 7.144688175193174e-09, + "loss": 0.6816, + "step": 52713 + }, + { + "epoch": 1.93, + "learning_rate": 7.141026110521111e-09, + "loss": 0.6784, + "step": 52714 + }, + { + "epoch": 1.93, + "learning_rate": 7.13736404584905e-09, + "loss": 0.6777, + "step": 52715 + }, + { + "epoch": 1.93, + "learning_rate": 7.133701981176987e-09, + "loss": 0.693, + "step": 52716 + }, + { + "epoch": 1.93, + "learning_rate": 7.130039916504925e-09, + "loss": 0.6899, + "step": 52717 + }, + { + "epoch": 1.93, + "learning_rate": 7.126377851832863e-09, + "loss": 0.685, + "step": 52718 + }, + { + "epoch": 1.93, + "learning_rate": 7.1227157871608e-09, + "loss": 0.6959, + "step": 52719 + }, + { + "epoch": 1.93, + "learning_rate": 7.1190537224887385e-09, + "loss": 0.6975, + "step": 52720 + }, + { + "epoch": 1.93, + "learning_rate": 7.1153916578166776e-09, + "loss": 0.696, + "step": 52721 + }, + { + "epoch": 1.93, + "learning_rate": 7.111729593144615e-09, + "loss": 0.6989, + "step": 52722 + }, + { + "epoch": 1.93, + "learning_rate": 7.1080675284725525e-09, + "loss": 0.6887, + "step": 52723 + }, + { + "epoch": 1.93, + "learning_rate": 7.104405463800491e-09, + "loss": 0.7059, + "step": 52724 + }, + { + "epoch": 1.93, + "learning_rate": 7.100743399128428e-09, + "loss": 0.7004, + "step": 52725 + }, + { + "epoch": 1.93, + "learning_rate": 7.097081334456366e-09, + "loss": 0.6911, + "step": 52726 + }, + { + "epoch": 1.93, + "learning_rate": 7.093419269784304e-09, + "loss": 0.6924, + "step": 52727 + }, + { + "epoch": 1.93, + "learning_rate": 7.089757205112241e-09, + "loss": 0.6866, + "step": 52728 + }, + { + "epoch": 1.93, + "learning_rate": 7.086095140440179e-09, + "loss": 0.6977, + "step": 52729 + }, + { + "epoch": 1.93, + "learning_rate": 7.082433075768118e-09, + "loss": 0.6995, + "step": 52730 + }, + { + "epoch": 1.93, + "learning_rate": 7.078771011096056e-09, + "loss": 0.683, + "step": 52731 + }, + { + "epoch": 1.93, + "learning_rate": 7.075108946423994e-09, + "loss": 0.6804, + "step": 52732 + }, + { + "epoch": 1.93, + "learning_rate": 7.071446881751932e-09, + "loss": 0.6946, + "step": 52733 + }, + { + "epoch": 1.93, + "learning_rate": 7.067784817079869e-09, + "loss": 0.7034, + "step": 52734 + }, + { + "epoch": 1.93, + "learning_rate": 7.064122752407807e-09, + "loss": 0.6849, + "step": 52735 + }, + { + "epoch": 1.93, + "learning_rate": 7.060460687735745e-09, + "loss": 0.6953, + "step": 52736 + }, + { + "epoch": 1.93, + "learning_rate": 7.0567986230636825e-09, + "loss": 0.6895, + "step": 52737 + }, + { + "epoch": 1.93, + "learning_rate": 7.05313655839162e-09, + "loss": 0.675, + "step": 52738 + }, + { + "epoch": 1.93, + "learning_rate": 7.049474493719559e-09, + "loss": 0.6891, + "step": 52739 + }, + { + "epoch": 1.93, + "learning_rate": 7.045812429047497e-09, + "loss": 0.6942, + "step": 52740 + }, + { + "epoch": 1.93, + "learning_rate": 7.042150364375435e-09, + "loss": 0.7036, + "step": 52741 + }, + { + "epoch": 1.93, + "learning_rate": 7.038488299703372e-09, + "loss": 0.6853, + "step": 52742 + }, + { + "epoch": 1.93, + "learning_rate": 7.0348262350313105e-09, + "loss": 0.6891, + "step": 52743 + }, + { + "epoch": 1.93, + "learning_rate": 7.031164170359248e-09, + "loss": 0.7018, + "step": 52744 + }, + { + "epoch": 1.93, + "learning_rate": 7.027502105687185e-09, + "loss": 0.6994, + "step": 52745 + }, + { + "epoch": 1.93, + "learning_rate": 7.023840041015124e-09, + "loss": 0.7015, + "step": 52746 + }, + { + "epoch": 1.93, + "learning_rate": 7.020177976343061e-09, + "loss": 0.6979, + "step": 52747 + }, + { + "epoch": 1.93, + "learning_rate": 7.016515911671e-09, + "loss": 0.6943, + "step": 52748 + }, + { + "epoch": 1.93, + "learning_rate": 7.0128538469989385e-09, + "loss": 0.7031, + "step": 52749 + }, + { + "epoch": 1.93, + "learning_rate": 7.009191782326876e-09, + "loss": 0.6967, + "step": 52750 + }, + { + "epoch": 1.93, + "learning_rate": 7.005529717654813e-09, + "loss": 0.6817, + "step": 52751 + }, + { + "epoch": 1.93, + "learning_rate": 7.001867652982752e-09, + "loss": 0.7144, + "step": 52752 + }, + { + "epoch": 1.93, + "learning_rate": 6.998205588310689e-09, + "loss": 0.6904, + "step": 52753 + }, + { + "epoch": 1.93, + "learning_rate": 6.9945435236386266e-09, + "loss": 0.6893, + "step": 52754 + }, + { + "epoch": 1.93, + "learning_rate": 6.990881458966565e-09, + "loss": 0.6895, + "step": 52755 + }, + { + "epoch": 1.93, + "learning_rate": 6.987219394294502e-09, + "loss": 0.6997, + "step": 52756 + }, + { + "epoch": 1.93, + "learning_rate": 6.983557329622441e-09, + "loss": 0.6902, + "step": 52757 + }, + { + "epoch": 1.93, + "learning_rate": 6.979895264950379e-09, + "loss": 0.6943, + "step": 52758 + }, + { + "epoch": 1.93, + "learning_rate": 6.976233200278317e-09, + "loss": 0.6852, + "step": 52759 + }, + { + "epoch": 1.93, + "learning_rate": 6.9725711356062546e-09, + "loss": 0.6882, + "step": 52760 + }, + { + "epoch": 1.93, + "learning_rate": 6.968909070934193e-09, + "loss": 0.6933, + "step": 52761 + }, + { + "epoch": 1.93, + "learning_rate": 6.96524700626213e-09, + "loss": 0.6949, + "step": 52762 + }, + { + "epoch": 1.93, + "learning_rate": 6.961584941590068e-09, + "loss": 0.6945, + "step": 52763 + }, + { + "epoch": 1.93, + "learning_rate": 6.957922876918006e-09, + "loss": 0.7002, + "step": 52764 + }, + { + "epoch": 1.93, + "learning_rate": 6.9542608122459435e-09, + "loss": 0.6976, + "step": 52765 + }, + { + "epoch": 1.93, + "learning_rate": 6.950598747573881e-09, + "loss": 0.6853, + "step": 52766 + }, + { + "epoch": 1.93, + "learning_rate": 6.94693668290182e-09, + "loss": 0.6914, + "step": 52767 + }, + { + "epoch": 1.93, + "learning_rate": 6.943274618229758e-09, + "loss": 0.7056, + "step": 52768 + }, + { + "epoch": 1.93, + "learning_rate": 6.939612553557696e-09, + "loss": 0.6832, + "step": 52769 + }, + { + "epoch": 1.93, + "learning_rate": 6.935950488885633e-09, + "loss": 0.6979, + "step": 52770 + }, + { + "epoch": 1.93, + "learning_rate": 6.9322884242135714e-09, + "loss": 0.6863, + "step": 52771 + }, + { + "epoch": 1.93, + "learning_rate": 6.928626359541509e-09, + "loss": 0.6911, + "step": 52772 + }, + { + "epoch": 1.93, + "learning_rate": 6.924964294869446e-09, + "loss": 0.6848, + "step": 52773 + }, + { + "epoch": 1.93, + "learning_rate": 6.921302230197385e-09, + "loss": 0.7002, + "step": 52774 + }, + { + "epoch": 1.93, + "learning_rate": 6.917640165525322e-09, + "loss": 0.6864, + "step": 52775 + }, + { + "epoch": 1.93, + "learning_rate": 6.913978100853261e-09, + "loss": 0.6887, + "step": 52776 + }, + { + "epoch": 1.93, + "learning_rate": 6.9103160361811994e-09, + "loss": 0.6857, + "step": 52777 + }, + { + "epoch": 1.93, + "learning_rate": 6.906653971509137e-09, + "loss": 0.691, + "step": 52778 + }, + { + "epoch": 1.93, + "learning_rate": 6.902991906837074e-09, + "loss": 0.6894, + "step": 52779 + }, + { + "epoch": 1.93, + "learning_rate": 6.899329842165013e-09, + "loss": 0.7049, + "step": 52780 + }, + { + "epoch": 1.93, + "learning_rate": 6.89566777749295e-09, + "loss": 0.696, + "step": 52781 + }, + { + "epoch": 1.93, + "learning_rate": 6.8920057128208875e-09, + "loss": 0.6875, + "step": 52782 + }, + { + "epoch": 1.93, + "learning_rate": 6.888343648148826e-09, + "loss": 0.6992, + "step": 52783 + }, + { + "epoch": 1.93, + "learning_rate": 6.884681583476763e-09, + "loss": 0.6947, + "step": 52784 + }, + { + "epoch": 1.93, + "learning_rate": 6.881019518804702e-09, + "loss": 0.6876, + "step": 52785 + }, + { + "epoch": 1.93, + "learning_rate": 6.87735745413264e-09, + "loss": 0.6998, + "step": 52786 + }, + { + "epoch": 1.93, + "learning_rate": 6.873695389460578e-09, + "loss": 0.6972, + "step": 52787 + }, + { + "epoch": 1.93, + "learning_rate": 6.8700333247885155e-09, + "loss": 0.6997, + "step": 52788 + }, + { + "epoch": 1.93, + "learning_rate": 6.866371260116453e-09, + "loss": 0.7066, + "step": 52789 + }, + { + "epoch": 1.93, + "learning_rate": 6.862709195444391e-09, + "loss": 0.6968, + "step": 52790 + }, + { + "epoch": 1.93, + "learning_rate": 6.859047130772329e-09, + "loss": 0.7029, + "step": 52791 + }, + { + "epoch": 1.93, + "learning_rate": 6.855385066100267e-09, + "loss": 0.6946, + "step": 52792 + }, + { + "epoch": 1.93, + "learning_rate": 6.851723001428204e-09, + "loss": 0.6804, + "step": 52793 + }, + { + "epoch": 1.93, + "learning_rate": 6.8480609367561435e-09, + "loss": 0.6971, + "step": 52794 + }, + { + "epoch": 1.93, + "learning_rate": 6.844398872084081e-09, + "loss": 0.6971, + "step": 52795 + }, + { + "epoch": 1.93, + "learning_rate": 6.840736807412019e-09, + "loss": 0.6953, + "step": 52796 + }, + { + "epoch": 1.93, + "learning_rate": 6.837074742739957e-09, + "loss": 0.7083, + "step": 52797 + }, + { + "epoch": 1.93, + "learning_rate": 6.833412678067894e-09, + "loss": 0.7046, + "step": 52798 + }, + { + "epoch": 1.93, + "learning_rate": 6.829750613395832e-09, + "loss": 0.7115, + "step": 52799 + }, + { + "epoch": 1.93, + "learning_rate": 6.82608854872377e-09, + "loss": 0.6988, + "step": 52800 + }, + { + "epoch": 1.93, + "learning_rate": 6.822426484051707e-09, + "loss": 0.7047, + "step": 52801 + }, + { + "epoch": 1.93, + "learning_rate": 6.8187644193796455e-09, + "loss": 0.6931, + "step": 52802 + }, + { + "epoch": 1.93, + "learning_rate": 6.815102354707585e-09, + "loss": 0.6954, + "step": 52803 + }, + { + "epoch": 1.93, + "learning_rate": 6.811440290035522e-09, + "loss": 0.6824, + "step": 52804 + }, + { + "epoch": 1.93, + "learning_rate": 6.80777822536346e-09, + "loss": 0.6867, + "step": 52805 + }, + { + "epoch": 1.93, + "learning_rate": 6.804116160691398e-09, + "loss": 0.6982, + "step": 52806 + }, + { + "epoch": 1.93, + "learning_rate": 6.800454096019335e-09, + "loss": 0.6889, + "step": 52807 + }, + { + "epoch": 1.93, + "learning_rate": 6.7967920313472735e-09, + "loss": 0.6935, + "step": 52808 + }, + { + "epoch": 1.93, + "learning_rate": 6.793129966675211e-09, + "loss": 0.6931, + "step": 52809 + }, + { + "epoch": 1.93, + "learning_rate": 6.7894679020031484e-09, + "loss": 0.6882, + "step": 52810 + }, + { + "epoch": 1.93, + "learning_rate": 6.785805837331087e-09, + "loss": 0.6816, + "step": 52811 + }, + { + "epoch": 1.93, + "learning_rate": 6.782143772659026e-09, + "loss": 0.6916, + "step": 52812 + }, + { + "epoch": 1.93, + "learning_rate": 6.778481707986963e-09, + "loss": 0.6906, + "step": 52813 + }, + { + "epoch": 1.93, + "learning_rate": 6.774819643314901e-09, + "loss": 0.7072, + "step": 52814 + }, + { + "epoch": 1.93, + "learning_rate": 6.771157578642839e-09, + "loss": 0.6881, + "step": 52815 + }, + { + "epoch": 1.93, + "learning_rate": 6.767495513970776e-09, + "loss": 0.6906, + "step": 52816 + }, + { + "epoch": 1.93, + "learning_rate": 6.763833449298714e-09, + "loss": 0.6911, + "step": 52817 + }, + { + "epoch": 1.93, + "learning_rate": 6.760171384626652e-09, + "loss": 0.6944, + "step": 52818 + }, + { + "epoch": 1.93, + "learning_rate": 6.7565093199545896e-09, + "loss": 0.6943, + "step": 52819 + }, + { + "epoch": 1.93, + "learning_rate": 6.752847255282528e-09, + "loss": 0.6917, + "step": 52820 + }, + { + "epoch": 1.93, + "learning_rate": 6.749185190610465e-09, + "loss": 0.6994, + "step": 52821 + }, + { + "epoch": 1.93, + "learning_rate": 6.745523125938404e-09, + "loss": 0.6908, + "step": 52822 + }, + { + "epoch": 1.93, + "learning_rate": 6.741861061266342e-09, + "loss": 0.7013, + "step": 52823 + }, + { + "epoch": 1.93, + "learning_rate": 6.73819899659428e-09, + "loss": 0.6939, + "step": 52824 + }, + { + "epoch": 1.93, + "learning_rate": 6.7345369319222176e-09, + "loss": 0.6891, + "step": 52825 + }, + { + "epoch": 1.93, + "learning_rate": 6.730874867250155e-09, + "loss": 0.7028, + "step": 52826 + }, + { + "epoch": 1.93, + "learning_rate": 6.727212802578093e-09, + "loss": 0.6875, + "step": 52827 + }, + { + "epoch": 1.93, + "learning_rate": 6.723550737906031e-09, + "loss": 0.6853, + "step": 52828 + }, + { + "epoch": 1.93, + "learning_rate": 6.719888673233968e-09, + "loss": 0.6998, + "step": 52829 + }, + { + "epoch": 1.93, + "learning_rate": 6.7162266085619065e-09, + "loss": 0.6898, + "step": 52830 + }, + { + "epoch": 1.93, + "learning_rate": 6.7125645438898456e-09, + "loss": 0.7012, + "step": 52831 + }, + { + "epoch": 1.93, + "learning_rate": 6.708902479217783e-09, + "loss": 0.6785, + "step": 52832 + }, + { + "epoch": 1.93, + "learning_rate": 6.7052404145457205e-09, + "loss": 0.6907, + "step": 52833 + }, + { + "epoch": 1.93, + "learning_rate": 6.701578349873659e-09, + "loss": 0.6911, + "step": 52834 + }, + { + "epoch": 1.93, + "learning_rate": 6.697916285201596e-09, + "loss": 0.6917, + "step": 52835 + }, + { + "epoch": 1.93, + "learning_rate": 6.6942542205295345e-09, + "loss": 0.7031, + "step": 52836 + }, + { + "epoch": 1.93, + "learning_rate": 6.690592155857472e-09, + "loss": 0.6898, + "step": 52837 + }, + { + "epoch": 1.93, + "learning_rate": 6.686930091185409e-09, + "loss": 0.6954, + "step": 52838 + }, + { + "epoch": 1.93, + "learning_rate": 6.683268026513348e-09, + "loss": 0.692, + "step": 52839 + }, + { + "epoch": 1.93, + "learning_rate": 6.679605961841287e-09, + "loss": 0.6905, + "step": 52840 + }, + { + "epoch": 1.93, + "learning_rate": 6.675943897169224e-09, + "loss": 0.704, + "step": 52841 + }, + { + "epoch": 1.93, + "learning_rate": 6.672281832497162e-09, + "loss": 0.6914, + "step": 52842 + }, + { + "epoch": 1.93, + "learning_rate": 6.6686197678251e-09, + "loss": 0.692, + "step": 52843 + }, + { + "epoch": 1.93, + "learning_rate": 6.664957703153037e-09, + "loss": 0.6923, + "step": 52844 + }, + { + "epoch": 1.93, + "learning_rate": 6.661295638480975e-09, + "loss": 0.6832, + "step": 52845 + }, + { + "epoch": 1.93, + "learning_rate": 6.657633573808913e-09, + "loss": 0.6979, + "step": 52846 + }, + { + "epoch": 1.93, + "learning_rate": 6.6539715091368505e-09, + "loss": 0.6914, + "step": 52847 + }, + { + "epoch": 1.93, + "learning_rate": 6.650309444464788e-09, + "loss": 0.728, + "step": 52848 + }, + { + "epoch": 1.93, + "learning_rate": 6.646647379792728e-09, + "loss": 0.6915, + "step": 52849 + }, + { + "epoch": 1.93, + "learning_rate": 6.642985315120665e-09, + "loss": 0.7086, + "step": 52850 + }, + { + "epoch": 1.93, + "learning_rate": 6.639323250448603e-09, + "loss": 0.6863, + "step": 52851 + }, + { + "epoch": 1.93, + "learning_rate": 6.635661185776541e-09, + "loss": 0.6979, + "step": 52852 + }, + { + "epoch": 1.93, + "learning_rate": 6.6319991211044785e-09, + "loss": 0.6922, + "step": 52853 + }, + { + "epoch": 1.93, + "learning_rate": 6.628337056432416e-09, + "loss": 0.6834, + "step": 52854 + }, + { + "epoch": 1.93, + "learning_rate": 6.624674991760354e-09, + "loss": 0.6868, + "step": 52855 + }, + { + "epoch": 1.93, + "learning_rate": 6.621012927088292e-09, + "loss": 0.6812, + "step": 52856 + }, + { + "epoch": 1.93, + "learning_rate": 6.617350862416229e-09, + "loss": 0.6981, + "step": 52857 + }, + { + "epoch": 1.93, + "learning_rate": 6.613688797744168e-09, + "loss": 0.6855, + "step": 52858 + }, + { + "epoch": 1.93, + "learning_rate": 6.6100267330721065e-09, + "loss": 0.6991, + "step": 52859 + }, + { + "epoch": 1.93, + "learning_rate": 6.606364668400044e-09, + "loss": 0.6998, + "step": 52860 + }, + { + "epoch": 1.93, + "learning_rate": 6.602702603727981e-09, + "loss": 0.6963, + "step": 52861 + }, + { + "epoch": 1.93, + "learning_rate": 6.59904053905592e-09, + "loss": 0.7016, + "step": 52862 + }, + { + "epoch": 1.93, + "learning_rate": 6.595378474383857e-09, + "loss": 0.6958, + "step": 52863 + }, + { + "epoch": 1.93, + "learning_rate": 6.591716409711795e-09, + "loss": 0.6857, + "step": 52864 + }, + { + "epoch": 1.93, + "learning_rate": 6.588054345039733e-09, + "loss": 0.7086, + "step": 52865 + }, + { + "epoch": 1.93, + "learning_rate": 6.58439228036767e-09, + "loss": 0.6965, + "step": 52866 + }, + { + "epoch": 1.93, + "learning_rate": 6.5807302156956085e-09, + "loss": 0.6807, + "step": 52867 + }, + { + "epoch": 1.93, + "learning_rate": 6.577068151023548e-09, + "loss": 0.699, + "step": 52868 + }, + { + "epoch": 1.93, + "learning_rate": 6.573406086351485e-09, + "loss": 0.6891, + "step": 52869 + }, + { + "epoch": 1.93, + "learning_rate": 6.5697440216794225e-09, + "loss": 0.7146, + "step": 52870 + }, + { + "epoch": 1.93, + "learning_rate": 6.566081957007361e-09, + "loss": 0.6902, + "step": 52871 + }, + { + "epoch": 1.93, + "learning_rate": 6.562419892335298e-09, + "loss": 0.6898, + "step": 52872 + }, + { + "epoch": 1.93, + "learning_rate": 6.558757827663236e-09, + "loss": 0.6755, + "step": 52873 + }, + { + "epoch": 1.93, + "learning_rate": 6.555095762991174e-09, + "loss": 0.6978, + "step": 52874 + }, + { + "epoch": 1.93, + "learning_rate": 6.5514336983191114e-09, + "loss": 0.6906, + "step": 52875 + }, + { + "epoch": 1.93, + "learning_rate": 6.547771633647049e-09, + "loss": 0.6967, + "step": 52876 + }, + { + "epoch": 1.93, + "learning_rate": 6.544109568974988e-09, + "loss": 0.6886, + "step": 52877 + }, + { + "epoch": 1.93, + "learning_rate": 6.540447504302926e-09, + "loss": 0.6837, + "step": 52878 + }, + { + "epoch": 1.93, + "learning_rate": 6.536785439630864e-09, + "loss": 0.6897, + "step": 52879 + }, + { + "epoch": 1.93, + "learning_rate": 6.533123374958802e-09, + "loss": 0.696, + "step": 52880 + }, + { + "epoch": 1.93, + "learning_rate": 6.5294613102867394e-09, + "loss": 0.7034, + "step": 52881 + }, + { + "epoch": 1.93, + "learning_rate": 6.525799245614677e-09, + "loss": 0.7041, + "step": 52882 + }, + { + "epoch": 1.93, + "learning_rate": 6.522137180942615e-09, + "loss": 0.6973, + "step": 52883 + }, + { + "epoch": 1.93, + "learning_rate": 6.518475116270553e-09, + "loss": 0.6912, + "step": 52884 + }, + { + "epoch": 1.93, + "learning_rate": 6.51481305159849e-09, + "loss": 0.6866, + "step": 52885 + }, + { + "epoch": 1.93, + "learning_rate": 6.511150986926429e-09, + "loss": 0.695, + "step": 52886 + }, + { + "epoch": 1.93, + "learning_rate": 6.5074889222543674e-09, + "loss": 0.6774, + "step": 52887 + }, + { + "epoch": 1.94, + "learning_rate": 6.503826857582305e-09, + "loss": 0.6956, + "step": 52888 + }, + { + "epoch": 1.94, + "learning_rate": 6.500164792910242e-09, + "loss": 0.7006, + "step": 52889 + }, + { + "epoch": 1.94, + "learning_rate": 6.4965027282381806e-09, + "loss": 0.6901, + "step": 52890 + }, + { + "epoch": 1.94, + "learning_rate": 6.492840663566118e-09, + "loss": 0.6898, + "step": 52891 + }, + { + "epoch": 1.94, + "learning_rate": 6.4891785988940555e-09, + "loss": 0.6935, + "step": 52892 + }, + { + "epoch": 1.94, + "learning_rate": 6.485516534221994e-09, + "loss": 0.6879, + "step": 52893 + }, + { + "epoch": 1.94, + "learning_rate": 6.481854469549931e-09, + "loss": 0.6864, + "step": 52894 + }, + { + "epoch": 1.94, + "learning_rate": 6.47819240487787e-09, + "loss": 0.6923, + "step": 52895 + }, + { + "epoch": 1.94, + "learning_rate": 6.4745303402058086e-09, + "loss": 0.6937, + "step": 52896 + }, + { + "epoch": 1.94, + "learning_rate": 6.470868275533746e-09, + "loss": 0.7019, + "step": 52897 + }, + { + "epoch": 1.94, + "learning_rate": 6.4672062108616835e-09, + "loss": 0.6809, + "step": 52898 + }, + { + "epoch": 1.94, + "learning_rate": 6.463544146189622e-09, + "loss": 0.703, + "step": 52899 + }, + { + "epoch": 1.94, + "learning_rate": 6.459882081517559e-09, + "loss": 0.6878, + "step": 52900 + }, + { + "epoch": 1.94, + "learning_rate": 6.456220016845497e-09, + "loss": 0.7057, + "step": 52901 + }, + { + "epoch": 1.94, + "learning_rate": 6.452557952173435e-09, + "loss": 0.7101, + "step": 52902 + }, + { + "epoch": 1.94, + "learning_rate": 6.448895887501372e-09, + "loss": 0.6903, + "step": 52903 + }, + { + "epoch": 1.94, + "learning_rate": 6.4452338228293115e-09, + "loss": 0.6903, + "step": 52904 + }, + { + "epoch": 1.94, + "learning_rate": 6.441571758157249e-09, + "loss": 0.7081, + "step": 52905 + }, + { + "epoch": 1.94, + "learning_rate": 6.437909693485187e-09, + "loss": 0.69, + "step": 52906 + }, + { + "epoch": 1.94, + "learning_rate": 6.434247628813125e-09, + "loss": 0.6978, + "step": 52907 + }, + { + "epoch": 1.94, + "learning_rate": 6.430585564141063e-09, + "loss": 0.6779, + "step": 52908 + }, + { + "epoch": 1.94, + "learning_rate": 6.426923499469e-09, + "loss": 0.6925, + "step": 52909 + }, + { + "epoch": 1.94, + "learning_rate": 6.423261434796938e-09, + "loss": 0.6866, + "step": 52910 + }, + { + "epoch": 1.94, + "learning_rate": 6.419599370124876e-09, + "loss": 0.6912, + "step": 52911 + }, + { + "epoch": 1.94, + "learning_rate": 6.4159373054528135e-09, + "loss": 0.6797, + "step": 52912 + }, + { + "epoch": 1.94, + "learning_rate": 6.412275240780751e-09, + "loss": 0.6908, + "step": 52913 + }, + { + "epoch": 1.94, + "learning_rate": 6.40861317610869e-09, + "loss": 0.6878, + "step": 52914 + }, + { + "epoch": 1.94, + "learning_rate": 6.404951111436628e-09, + "loss": 0.6946, + "step": 52915 + }, + { + "epoch": 1.94, + "learning_rate": 6.401289046764566e-09, + "loss": 0.6915, + "step": 52916 + }, + { + "epoch": 1.94, + "learning_rate": 6.397626982092503e-09, + "loss": 0.7016, + "step": 52917 + }, + { + "epoch": 1.94, + "learning_rate": 6.3939649174204415e-09, + "loss": 0.6884, + "step": 52918 + }, + { + "epoch": 1.94, + "learning_rate": 6.390302852748379e-09, + "loss": 0.7106, + "step": 52919 + }, + { + "epoch": 1.94, + "learning_rate": 6.386640788076316e-09, + "loss": 0.6893, + "step": 52920 + }, + { + "epoch": 1.94, + "learning_rate": 6.382978723404255e-09, + "loss": 0.6864, + "step": 52921 + }, + { + "epoch": 1.94, + "learning_rate": 6.379316658732192e-09, + "loss": 0.6906, + "step": 52922 + }, + { + "epoch": 1.94, + "learning_rate": 6.375654594060131e-09, + "loss": 0.6981, + "step": 52923 + }, + { + "epoch": 1.94, + "learning_rate": 6.3719925293880695e-09, + "loss": 0.7029, + "step": 52924 + }, + { + "epoch": 1.94, + "learning_rate": 6.368330464716007e-09, + "loss": 0.7022, + "step": 52925 + }, + { + "epoch": 1.94, + "learning_rate": 6.364668400043944e-09, + "loss": 0.6904, + "step": 52926 + }, + { + "epoch": 1.94, + "learning_rate": 6.361006335371883e-09, + "loss": 0.6861, + "step": 52927 + }, + { + "epoch": 1.94, + "learning_rate": 6.35734427069982e-09, + "loss": 0.6986, + "step": 52928 + }, + { + "epoch": 1.94, + "learning_rate": 6.3536822060277576e-09, + "loss": 0.7108, + "step": 52929 + }, + { + "epoch": 1.94, + "learning_rate": 6.350020141355696e-09, + "loss": 0.6846, + "step": 52930 + }, + { + "epoch": 1.94, + "learning_rate": 6.346358076683633e-09, + "loss": 0.6981, + "step": 52931 + }, + { + "epoch": 1.94, + "learning_rate": 6.342696012011572e-09, + "loss": 0.692, + "step": 52932 + }, + { + "epoch": 1.94, + "learning_rate": 6.33903394733951e-09, + "loss": 0.7072, + "step": 52933 + }, + { + "epoch": 1.94, + "learning_rate": 6.335371882667448e-09, + "loss": 0.688, + "step": 52934 + }, + { + "epoch": 1.94, + "learning_rate": 6.3317098179953856e-09, + "loss": 0.6878, + "step": 52935 + }, + { + "epoch": 1.94, + "learning_rate": 6.328047753323323e-09, + "loss": 0.6957, + "step": 52936 + }, + { + "epoch": 1.94, + "learning_rate": 6.324385688651261e-09, + "loss": 0.6801, + "step": 52937 + }, + { + "epoch": 1.94, + "learning_rate": 6.320723623979199e-09, + "loss": 0.7048, + "step": 52938 + }, + { + "epoch": 1.94, + "learning_rate": 6.317061559307137e-09, + "loss": 0.6895, + "step": 52939 + }, + { + "epoch": 1.94, + "learning_rate": 6.3133994946350744e-09, + "loss": 0.6888, + "step": 52940 + }, + { + "epoch": 1.94, + "learning_rate": 6.3097374299630135e-09, + "loss": 0.7105, + "step": 52941 + }, + { + "epoch": 1.94, + "learning_rate": 6.306075365290951e-09, + "loss": 0.693, + "step": 52942 + }, + { + "epoch": 1.94, + "learning_rate": 6.302413300618889e-09, + "loss": 0.6954, + "step": 52943 + }, + { + "epoch": 1.94, + "learning_rate": 6.298751235946827e-09, + "loss": 0.6809, + "step": 52944 + }, + { + "epoch": 1.94, + "learning_rate": 6.295089171274764e-09, + "loss": 0.6976, + "step": 52945 + }, + { + "epoch": 1.94, + "learning_rate": 6.2914271066027024e-09, + "loss": 0.7037, + "step": 52946 + }, + { + "epoch": 1.94, + "learning_rate": 6.28776504193064e-09, + "loss": 0.7032, + "step": 52947 + }, + { + "epoch": 1.94, + "learning_rate": 6.284102977258577e-09, + "loss": 0.6962, + "step": 52948 + }, + { + "epoch": 1.94, + "learning_rate": 6.280440912586516e-09, + "loss": 0.6973, + "step": 52949 + }, + { + "epoch": 1.94, + "learning_rate": 6.276778847914455e-09, + "loss": 0.6963, + "step": 52950 + }, + { + "epoch": 1.94, + "learning_rate": 6.273116783242392e-09, + "loss": 0.6976, + "step": 52951 + }, + { + "epoch": 1.94, + "learning_rate": 6.2694547185703304e-09, + "loss": 0.6863, + "step": 52952 + }, + { + "epoch": 1.94, + "learning_rate": 6.265792653898268e-09, + "loss": 0.6953, + "step": 52953 + }, + { + "epoch": 1.94, + "learning_rate": 6.262130589226205e-09, + "loss": 0.6926, + "step": 52954 + }, + { + "epoch": 1.94, + "learning_rate": 6.258468524554144e-09, + "loss": 0.6841, + "step": 52955 + }, + { + "epoch": 1.94, + "learning_rate": 6.254806459882081e-09, + "loss": 0.6921, + "step": 52956 + }, + { + "epoch": 1.94, + "learning_rate": 6.2511443952100185e-09, + "loss": 0.6989, + "step": 52957 + }, + { + "epoch": 1.94, + "learning_rate": 6.247482330537957e-09, + "loss": 0.6964, + "step": 52958 + }, + { + "epoch": 1.94, + "learning_rate": 6.243820265865895e-09, + "loss": 0.685, + "step": 52959 + }, + { + "epoch": 1.94, + "learning_rate": 6.2401582011938325e-09, + "loss": 0.6901, + "step": 52960 + }, + { + "epoch": 1.94, + "learning_rate": 6.236496136521771e-09, + "loss": 0.6982, + "step": 52961 + }, + { + "epoch": 1.94, + "learning_rate": 6.232834071849709e-09, + "loss": 0.6966, + "step": 52962 + }, + { + "epoch": 1.94, + "learning_rate": 6.2291720071776465e-09, + "loss": 0.7009, + "step": 52963 + }, + { + "epoch": 1.94, + "learning_rate": 6.225509942505584e-09, + "loss": 0.6771, + "step": 52964 + }, + { + "epoch": 1.94, + "learning_rate": 6.221847877833522e-09, + "loss": 0.6957, + "step": 52965 + }, + { + "epoch": 1.94, + "learning_rate": 6.2181858131614605e-09, + "loss": 0.6946, + "step": 52966 + }, + { + "epoch": 1.94, + "learning_rate": 6.214523748489398e-09, + "loss": 0.695, + "step": 52967 + }, + { + "epoch": 1.94, + "learning_rate": 6.210861683817336e-09, + "loss": 0.6985, + "step": 52968 + }, + { + "epoch": 1.94, + "learning_rate": 6.207199619145274e-09, + "loss": 0.6785, + "step": 52969 + }, + { + "epoch": 1.94, + "learning_rate": 6.203537554473211e-09, + "loss": 0.7032, + "step": 52970 + }, + { + "epoch": 1.94, + "learning_rate": 6.19987548980115e-09, + "loss": 0.6885, + "step": 52971 + }, + { + "epoch": 1.94, + "learning_rate": 6.196213425129088e-09, + "loss": 0.6882, + "step": 52972 + }, + { + "epoch": 1.94, + "learning_rate": 6.192551360457025e-09, + "loss": 0.6933, + "step": 52973 + }, + { + "epoch": 1.94, + "learning_rate": 6.188889295784963e-09, + "loss": 0.694, + "step": 52974 + }, + { + "epoch": 1.94, + "learning_rate": 6.185227231112901e-09, + "loss": 0.6933, + "step": 52975 + }, + { + "epoch": 1.94, + "learning_rate": 6.181565166440839e-09, + "loss": 0.687, + "step": 52976 + }, + { + "epoch": 1.94, + "learning_rate": 6.177903101768777e-09, + "loss": 0.7011, + "step": 52977 + }, + { + "epoch": 1.94, + "learning_rate": 6.174241037096715e-09, + "loss": 0.6977, + "step": 52978 + }, + { + "epoch": 1.94, + "learning_rate": 6.170578972424652e-09, + "loss": 0.6885, + "step": 52979 + }, + { + "epoch": 1.94, + "learning_rate": 6.1669169077525905e-09, + "loss": 0.6889, + "step": 52980 + }, + { + "epoch": 1.94, + "learning_rate": 6.163254843080529e-09, + "loss": 0.6948, + "step": 52981 + }, + { + "epoch": 1.94, + "learning_rate": 6.159592778408466e-09, + "loss": 0.6988, + "step": 52982 + }, + { + "epoch": 1.94, + "learning_rate": 6.1559307137364045e-09, + "loss": 0.7033, + "step": 52983 + }, + { + "epoch": 1.94, + "learning_rate": 6.152268649064342e-09, + "loss": 0.7047, + "step": 52984 + }, + { + "epoch": 1.94, + "learning_rate": 6.14860658439228e-09, + "loss": 0.6923, + "step": 52985 + }, + { + "epoch": 1.94, + "learning_rate": 6.144944519720218e-09, + "loss": 0.6804, + "step": 52986 + }, + { + "epoch": 1.94, + "learning_rate": 6.141282455048156e-09, + "loss": 0.6976, + "step": 52987 + }, + { + "epoch": 1.94, + "learning_rate": 6.137620390376093e-09, + "loss": 0.6797, + "step": 52988 + }, + { + "epoch": 1.94, + "learning_rate": 6.133958325704032e-09, + "loss": 0.6988, + "step": 52989 + }, + { + "epoch": 1.94, + "learning_rate": 6.13029626103197e-09, + "loss": 0.6971, + "step": 52990 + }, + { + "epoch": 1.94, + "learning_rate": 6.126634196359907e-09, + "loss": 0.693, + "step": 52991 + }, + { + "epoch": 1.94, + "learning_rate": 6.122972131687845e-09, + "loss": 0.7046, + "step": 52992 + }, + { + "epoch": 1.94, + "learning_rate": 6.119310067015783e-09, + "loss": 0.6846, + "step": 52993 + }, + { + "epoch": 1.94, + "learning_rate": 6.115648002343721e-09, + "loss": 0.7021, + "step": 52994 + }, + { + "epoch": 1.94, + "learning_rate": 6.111985937671659e-09, + "loss": 0.698, + "step": 52995 + }, + { + "epoch": 1.94, + "learning_rate": 6.108323872999597e-09, + "loss": 0.6848, + "step": 52996 + }, + { + "epoch": 1.94, + "learning_rate": 6.1046618083275346e-09, + "loss": 0.6943, + "step": 52997 + }, + { + "epoch": 1.94, + "learning_rate": 6.100999743655472e-09, + "loss": 0.6852, + "step": 52998 + }, + { + "epoch": 1.94, + "learning_rate": 6.097337678983411e-09, + "loss": 0.6894, + "step": 52999 + }, + { + "epoch": 1.94, + "learning_rate": 6.0936756143113486e-09, + "loss": 0.6958, + "step": 53000 + }, + { + "epoch": 1.94, + "learning_rate": 6.090013549639286e-09, + "loss": 0.6813, + "step": 53001 + }, + { + "epoch": 1.94, + "learning_rate": 6.086351484967224e-09, + "loss": 0.6938, + "step": 53002 + }, + { + "epoch": 1.94, + "learning_rate": 6.0826894202951626e-09, + "loss": 0.694, + "step": 53003 + }, + { + "epoch": 1.94, + "learning_rate": 6.0790273556231e-09, + "loss": 0.6911, + "step": 53004 + }, + { + "epoch": 1.94, + "learning_rate": 6.075365290951038e-09, + "loss": 0.6906, + "step": 53005 + }, + { + "epoch": 1.94, + "learning_rate": 6.071703226278976e-09, + "loss": 0.6976, + "step": 53006 + }, + { + "epoch": 1.94, + "learning_rate": 6.068041161606913e-09, + "loss": 0.7057, + "step": 53007 + }, + { + "epoch": 1.94, + "learning_rate": 6.0643790969348515e-09, + "loss": 0.6925, + "step": 53008 + }, + { + "epoch": 1.94, + "learning_rate": 6.06071703226279e-09, + "loss": 0.6969, + "step": 53009 + }, + { + "epoch": 1.94, + "learning_rate": 6.057054967590727e-09, + "loss": 0.6894, + "step": 53010 + }, + { + "epoch": 1.94, + "learning_rate": 6.0533929029186654e-09, + "loss": 0.6785, + "step": 53011 + }, + { + "epoch": 1.94, + "learning_rate": 6.049730838246604e-09, + "loss": 0.6771, + "step": 53012 + }, + { + "epoch": 1.94, + "learning_rate": 6.046068773574541e-09, + "loss": 0.6955, + "step": 53013 + }, + { + "epoch": 1.94, + "learning_rate": 6.042406708902479e-09, + "loss": 0.6909, + "step": 53014 + }, + { + "epoch": 1.94, + "learning_rate": 6.038744644230417e-09, + "loss": 0.6875, + "step": 53015 + }, + { + "epoch": 1.94, + "learning_rate": 6.035082579558354e-09, + "loss": 0.6955, + "step": 53016 + }, + { + "epoch": 1.94, + "learning_rate": 6.031420514886293e-09, + "loss": 0.6864, + "step": 53017 + }, + { + "epoch": 1.94, + "learning_rate": 6.027758450214231e-09, + "loss": 0.6845, + "step": 53018 + }, + { + "epoch": 1.94, + "learning_rate": 6.024096385542168e-09, + "loss": 0.698, + "step": 53019 + }, + { + "epoch": 1.94, + "learning_rate": 6.020434320870106e-09, + "loss": 0.685, + "step": 53020 + }, + { + "epoch": 1.94, + "learning_rate": 6.016772256198044e-09, + "loss": 0.6927, + "step": 53021 + }, + { + "epoch": 1.94, + "learning_rate": 6.013110191525982e-09, + "loss": 0.6952, + "step": 53022 + }, + { + "epoch": 1.94, + "learning_rate": 6.00944812685392e-09, + "loss": 0.6989, + "step": 53023 + }, + { + "epoch": 1.94, + "learning_rate": 6.005786062181858e-09, + "loss": 0.6867, + "step": 53024 + }, + { + "epoch": 1.94, + "learning_rate": 6.0021239975097955e-09, + "loss": 0.6974, + "step": 53025 + }, + { + "epoch": 1.94, + "learning_rate": 5.998461932837734e-09, + "loss": 0.6859, + "step": 53026 + }, + { + "epoch": 1.94, + "learning_rate": 5.994799868165672e-09, + "loss": 0.6972, + "step": 53027 + }, + { + "epoch": 1.94, + "learning_rate": 5.9911378034936095e-09, + "loss": 0.6831, + "step": 53028 + }, + { + "epoch": 1.94, + "learning_rate": 5.987475738821547e-09, + "loss": 0.6954, + "step": 53029 + }, + { + "epoch": 1.94, + "learning_rate": 5.983813674149485e-09, + "loss": 0.6927, + "step": 53030 + }, + { + "epoch": 1.94, + "learning_rate": 5.9801516094774235e-09, + "loss": 0.6844, + "step": 53031 + }, + { + "epoch": 1.94, + "learning_rate": 5.976489544805361e-09, + "loss": 0.7, + "step": 53032 + }, + { + "epoch": 1.94, + "learning_rate": 5.972827480133299e-09, + "loss": 0.701, + "step": 53033 + }, + { + "epoch": 1.94, + "learning_rate": 5.969165415461237e-09, + "loss": 0.6888, + "step": 53034 + }, + { + "epoch": 1.94, + "learning_rate": 5.965503350789175e-09, + "loss": 0.7047, + "step": 53035 + }, + { + "epoch": 1.94, + "learning_rate": 5.961841286117112e-09, + "loss": 0.6969, + "step": 53036 + }, + { + "epoch": 1.94, + "learning_rate": 5.958179221445051e-09, + "loss": 0.6902, + "step": 53037 + }, + { + "epoch": 1.94, + "learning_rate": 5.954517156772988e-09, + "loss": 0.6889, + "step": 53038 + }, + { + "epoch": 1.94, + "learning_rate": 5.9508550921009255e-09, + "loss": 0.6961, + "step": 53039 + }, + { + "epoch": 1.94, + "learning_rate": 5.947193027428865e-09, + "loss": 0.6924, + "step": 53040 + }, + { + "epoch": 1.94, + "learning_rate": 5.943530962756802e-09, + "loss": 0.6874, + "step": 53041 + }, + { + "epoch": 1.94, + "learning_rate": 5.9398688980847395e-09, + "loss": 0.6887, + "step": 53042 + }, + { + "epoch": 1.94, + "learning_rate": 5.936206833412678e-09, + "loss": 0.6938, + "step": 53043 + }, + { + "epoch": 1.94, + "learning_rate": 5.932544768740615e-09, + "loss": 0.6935, + "step": 53044 + }, + { + "epoch": 1.94, + "learning_rate": 5.9288827040685535e-09, + "loss": 0.6969, + "step": 53045 + }, + { + "epoch": 1.94, + "learning_rate": 5.925220639396492e-09, + "loss": 0.7031, + "step": 53046 + }, + { + "epoch": 1.94, + "learning_rate": 5.921558574724429e-09, + "loss": 0.7045, + "step": 53047 + }, + { + "epoch": 1.94, + "learning_rate": 5.917896510052367e-09, + "loss": 0.6978, + "step": 53048 + }, + { + "epoch": 1.94, + "learning_rate": 5.914234445380306e-09, + "loss": 0.6959, + "step": 53049 + }, + { + "epoch": 1.94, + "learning_rate": 5.910572380708243e-09, + "loss": 0.6991, + "step": 53050 + }, + { + "epoch": 1.94, + "learning_rate": 5.906910316036181e-09, + "loss": 0.6835, + "step": 53051 + }, + { + "epoch": 1.94, + "learning_rate": 5.903248251364119e-09, + "loss": 0.6923, + "step": 53052 + }, + { + "epoch": 1.94, + "learning_rate": 5.8995861866920564e-09, + "loss": 0.6913, + "step": 53053 + }, + { + "epoch": 1.94, + "learning_rate": 5.895924122019995e-09, + "loss": 0.6958, + "step": 53054 + }, + { + "epoch": 1.94, + "learning_rate": 5.892262057347933e-09, + "loss": 0.6889, + "step": 53055 + }, + { + "epoch": 1.94, + "learning_rate": 5.8885999926758704e-09, + "loss": 0.6778, + "step": 53056 + }, + { + "epoch": 1.94, + "learning_rate": 5.884937928003808e-09, + "loss": 0.6857, + "step": 53057 + }, + { + "epoch": 1.94, + "learning_rate": 5.881275863331746e-09, + "loss": 0.7124, + "step": 53058 + }, + { + "epoch": 1.94, + "learning_rate": 5.877613798659684e-09, + "loss": 0.6986, + "step": 53059 + }, + { + "epoch": 1.94, + "learning_rate": 5.873951733987622e-09, + "loss": 0.6958, + "step": 53060 + }, + { + "epoch": 1.94, + "learning_rate": 5.870289669315559e-09, + "loss": 0.7043, + "step": 53061 + }, + { + "epoch": 1.94, + "learning_rate": 5.8666276046434976e-09, + "loss": 0.6899, + "step": 53062 + }, + { + "epoch": 1.94, + "learning_rate": 5.862965539971436e-09, + "loss": 0.694, + "step": 53063 + }, + { + "epoch": 1.94, + "learning_rate": 5.859303475299373e-09, + "loss": 0.6985, + "step": 53064 + }, + { + "epoch": 1.94, + "learning_rate": 5.8556414106273116e-09, + "loss": 0.6958, + "step": 53065 + }, + { + "epoch": 1.94, + "learning_rate": 5.851979345955249e-09, + "loss": 0.6859, + "step": 53066 + }, + { + "epoch": 1.94, + "learning_rate": 5.8483172812831865e-09, + "loss": 0.6939, + "step": 53067 + }, + { + "epoch": 1.94, + "learning_rate": 5.8446552166111256e-09, + "loss": 0.6969, + "step": 53068 + }, + { + "epoch": 1.94, + "learning_rate": 5.840993151939063e-09, + "loss": 0.7017, + "step": 53069 + }, + { + "epoch": 1.94, + "learning_rate": 5.8373310872670005e-09, + "loss": 0.7036, + "step": 53070 + }, + { + "epoch": 1.94, + "learning_rate": 5.833669022594939e-09, + "loss": 0.6977, + "step": 53071 + }, + { + "epoch": 1.94, + "learning_rate": 5.830006957922877e-09, + "loss": 0.701, + "step": 53072 + }, + { + "epoch": 1.94, + "learning_rate": 5.8263448932508145e-09, + "loss": 0.6894, + "step": 53073 + }, + { + "epoch": 1.94, + "learning_rate": 5.822682828578753e-09, + "loss": 0.69, + "step": 53074 + }, + { + "epoch": 1.94, + "learning_rate": 5.81902076390669e-09, + "loss": 0.7013, + "step": 53075 + }, + { + "epoch": 1.94, + "learning_rate": 5.815358699234628e-09, + "loss": 0.7026, + "step": 53076 + }, + { + "epoch": 1.94, + "learning_rate": 5.811696634562567e-09, + "loss": 0.6974, + "step": 53077 + }, + { + "epoch": 1.94, + "learning_rate": 5.808034569890504e-09, + "loss": 0.7007, + "step": 53078 + }, + { + "epoch": 1.94, + "learning_rate": 5.804372505218442e-09, + "loss": 0.6942, + "step": 53079 + }, + { + "epoch": 1.94, + "learning_rate": 5.80071044054638e-09, + "loss": 0.6915, + "step": 53080 + }, + { + "epoch": 1.94, + "learning_rate": 5.797048375874318e-09, + "loss": 0.7006, + "step": 53081 + }, + { + "epoch": 1.94, + "learning_rate": 5.793386311202256e-09, + "loss": 0.6864, + "step": 53082 + }, + { + "epoch": 1.94, + "learning_rate": 5.789724246530193e-09, + "loss": 0.6951, + "step": 53083 + }, + { + "epoch": 1.94, + "learning_rate": 5.786062181858131e-09, + "loss": 0.699, + "step": 53084 + }, + { + "epoch": 1.94, + "learning_rate": 5.782400117186069e-09, + "loss": 0.7004, + "step": 53085 + }, + { + "epoch": 1.94, + "learning_rate": 5.778738052514007e-09, + "loss": 0.6928, + "step": 53086 + }, + { + "epoch": 1.94, + "learning_rate": 5.775075987841945e-09, + "loss": 0.6855, + "step": 53087 + }, + { + "epoch": 1.94, + "learning_rate": 5.771413923169883e-09, + "loss": 0.6867, + "step": 53088 + }, + { + "epoch": 1.94, + "learning_rate": 5.76775185849782e-09, + "loss": 0.6886, + "step": 53089 + }, + { + "epoch": 1.94, + "learning_rate": 5.7640897938257585e-09, + "loss": 0.7002, + "step": 53090 + }, + { + "epoch": 1.94, + "learning_rate": 5.760427729153697e-09, + "loss": 0.683, + "step": 53091 + }, + { + "epoch": 1.94, + "learning_rate": 5.756765664481634e-09, + "loss": 0.6996, + "step": 53092 + }, + { + "epoch": 1.94, + "learning_rate": 5.7531035998095725e-09, + "loss": 0.6782, + "step": 53093 + }, + { + "epoch": 1.94, + "learning_rate": 5.74944153513751e-09, + "loss": 0.7013, + "step": 53094 + }, + { + "epoch": 1.94, + "learning_rate": 5.745779470465448e-09, + "loss": 0.6878, + "step": 53095 + }, + { + "epoch": 1.94, + "learning_rate": 5.7421174057933865e-09, + "loss": 0.6983, + "step": 53096 + }, + { + "epoch": 1.94, + "learning_rate": 5.738455341121324e-09, + "loss": 0.6907, + "step": 53097 + }, + { + "epoch": 1.94, + "learning_rate": 5.734793276449261e-09, + "loss": 0.6951, + "step": 53098 + }, + { + "epoch": 1.94, + "learning_rate": 5.7311312117772e-09, + "loss": 0.7092, + "step": 53099 + }, + { + "epoch": 1.94, + "learning_rate": 5.727469147105138e-09, + "loss": 0.6918, + "step": 53100 + }, + { + "epoch": 1.94, + "learning_rate": 5.723807082433075e-09, + "loss": 0.7029, + "step": 53101 + }, + { + "epoch": 1.94, + "learning_rate": 5.720145017761014e-09, + "loss": 0.6922, + "step": 53102 + }, + { + "epoch": 1.94, + "learning_rate": 5.716482953088951e-09, + "loss": 0.686, + "step": 53103 + }, + { + "epoch": 1.94, + "learning_rate": 5.712820888416889e-09, + "loss": 0.6989, + "step": 53104 + }, + { + "epoch": 1.94, + "learning_rate": 5.709158823744827e-09, + "loss": 0.6862, + "step": 53105 + }, + { + "epoch": 1.94, + "learning_rate": 5.705496759072765e-09, + "loss": 0.6942, + "step": 53106 + }, + { + "epoch": 1.94, + "learning_rate": 5.7018346944007026e-09, + "loss": 0.6894, + "step": 53107 + }, + { + "epoch": 1.94, + "learning_rate": 5.698172629728641e-09, + "loss": 0.7016, + "step": 53108 + }, + { + "epoch": 1.94, + "learning_rate": 5.694510565056579e-09, + "loss": 0.6896, + "step": 53109 + }, + { + "epoch": 1.94, + "learning_rate": 5.6908485003845165e-09, + "loss": 0.6823, + "step": 53110 + }, + { + "epoch": 1.94, + "learning_rate": 5.687186435712454e-09, + "loss": 0.6954, + "step": 53111 + }, + { + "epoch": 1.94, + "learning_rate": 5.683524371040392e-09, + "loss": 0.6919, + "step": 53112 + }, + { + "epoch": 1.94, + "learning_rate": 5.6798623063683305e-09, + "loss": 0.6923, + "step": 53113 + }, + { + "epoch": 1.94, + "learning_rate": 5.676200241696268e-09, + "loss": 0.6974, + "step": 53114 + }, + { + "epoch": 1.94, + "learning_rate": 5.672538177024206e-09, + "loss": 0.6943, + "step": 53115 + }, + { + "epoch": 1.94, + "learning_rate": 5.668876112352144e-09, + "loss": 0.6867, + "step": 53116 + }, + { + "epoch": 1.94, + "learning_rate": 5.665214047680081e-09, + "loss": 0.6951, + "step": 53117 + }, + { + "epoch": 1.94, + "learning_rate": 5.66155198300802e-09, + "loss": 0.6938, + "step": 53118 + }, + { + "epoch": 1.94, + "learning_rate": 5.657889918335958e-09, + "loss": 0.6773, + "step": 53119 + }, + { + "epoch": 1.94, + "learning_rate": 5.654227853663895e-09, + "loss": 0.6907, + "step": 53120 + }, + { + "epoch": 1.94, + "learning_rate": 5.6505657889918334e-09, + "loss": 0.6931, + "step": 53121 + }, + { + "epoch": 1.94, + "learning_rate": 5.646903724319771e-09, + "loss": 0.689, + "step": 53122 + }, + { + "epoch": 1.94, + "learning_rate": 5.643241659647709e-09, + "loss": 0.6832, + "step": 53123 + }, + { + "epoch": 1.94, + "learning_rate": 5.6395795949756474e-09, + "loss": 0.6931, + "step": 53124 + }, + { + "epoch": 1.94, + "learning_rate": 5.635917530303585e-09, + "loss": 0.6862, + "step": 53125 + }, + { + "epoch": 1.94, + "learning_rate": 5.632255465631522e-09, + "loss": 0.6943, + "step": 53126 + }, + { + "epoch": 1.94, + "learning_rate": 5.628593400959461e-09, + "loss": 0.7073, + "step": 53127 + }, + { + "epoch": 1.94, + "learning_rate": 5.624931336287399e-09, + "loss": 0.6909, + "step": 53128 + }, + { + "epoch": 1.94, + "learning_rate": 5.621269271615336e-09, + "loss": 0.693, + "step": 53129 + }, + { + "epoch": 1.94, + "learning_rate": 5.617607206943275e-09, + "loss": 0.6926, + "step": 53130 + }, + { + "epoch": 1.94, + "learning_rate": 5.613945142271212e-09, + "loss": 0.6937, + "step": 53131 + }, + { + "epoch": 1.94, + "learning_rate": 5.61028307759915e-09, + "loss": 0.7024, + "step": 53132 + }, + { + "epoch": 1.94, + "learning_rate": 5.606621012927088e-09, + "loss": 0.6934, + "step": 53133 + }, + { + "epoch": 1.94, + "learning_rate": 5.602958948255026e-09, + "loss": 0.6866, + "step": 53134 + }, + { + "epoch": 1.94, + "learning_rate": 5.5992968835829635e-09, + "loss": 0.7057, + "step": 53135 + }, + { + "epoch": 1.94, + "learning_rate": 5.595634818910902e-09, + "loss": 0.6912, + "step": 53136 + }, + { + "epoch": 1.94, + "learning_rate": 5.59197275423884e-09, + "loss": 0.6948, + "step": 53137 + }, + { + "epoch": 1.94, + "learning_rate": 5.5883106895667775e-09, + "loss": 0.6771, + "step": 53138 + }, + { + "epoch": 1.94, + "learning_rate": 5.584648624894715e-09, + "loss": 0.6928, + "step": 53139 + }, + { + "epoch": 1.94, + "learning_rate": 5.580986560222653e-09, + "loss": 0.6791, + "step": 53140 + }, + { + "epoch": 1.94, + "learning_rate": 5.5773244955505915e-09, + "loss": 0.6837, + "step": 53141 + }, + { + "epoch": 1.94, + "learning_rate": 5.573662430878529e-09, + "loss": 0.7064, + "step": 53142 + }, + { + "epoch": 1.94, + "learning_rate": 5.570000366206467e-09, + "loss": 0.675, + "step": 53143 + }, + { + "epoch": 1.94, + "learning_rate": 5.566338301534405e-09, + "loss": 0.6969, + "step": 53144 + }, + { + "epoch": 1.94, + "learning_rate": 5.562676236862342e-09, + "loss": 0.6991, + "step": 53145 + }, + { + "epoch": 1.94, + "learning_rate": 5.559014172190281e-09, + "loss": 0.7046, + "step": 53146 + }, + { + "epoch": 1.94, + "learning_rate": 5.555352107518219e-09, + "loss": 0.6835, + "step": 53147 + }, + { + "epoch": 1.94, + "learning_rate": 5.551690042846156e-09, + "loss": 0.6941, + "step": 53148 + }, + { + "epoch": 1.94, + "learning_rate": 5.548027978174094e-09, + "loss": 0.7047, + "step": 53149 + }, + { + "epoch": 1.94, + "learning_rate": 5.544365913502033e-09, + "loss": 0.6906, + "step": 53150 + }, + { + "epoch": 1.94, + "learning_rate": 5.54070384882997e-09, + "loss": 0.7022, + "step": 53151 + }, + { + "epoch": 1.94, + "learning_rate": 5.537041784157908e-09, + "loss": 0.6853, + "step": 53152 + }, + { + "epoch": 1.94, + "learning_rate": 5.533379719485846e-09, + "loss": 0.6911, + "step": 53153 + }, + { + "epoch": 1.94, + "learning_rate": 5.529717654813783e-09, + "loss": 0.6927, + "step": 53154 + }, + { + "epoch": 1.94, + "learning_rate": 5.5260555901417215e-09, + "loss": 0.6949, + "step": 53155 + }, + { + "epoch": 1.94, + "learning_rate": 5.52239352546966e-09, + "loss": 0.6887, + "step": 53156 + }, + { + "epoch": 1.94, + "learning_rate": 5.518731460797597e-09, + "loss": 0.6853, + "step": 53157 + }, + { + "epoch": 1.94, + "learning_rate": 5.5150693961255355e-09, + "loss": 0.7026, + "step": 53158 + }, + { + "epoch": 1.94, + "learning_rate": 5.511407331453474e-09, + "loss": 0.6971, + "step": 53159 + }, + { + "epoch": 1.94, + "learning_rate": 5.507745266781411e-09, + "loss": 0.6919, + "step": 53160 + }, + { + "epoch": 1.95, + "learning_rate": 5.504083202109349e-09, + "loss": 0.6928, + "step": 53161 + }, + { + "epoch": 1.95, + "learning_rate": 5.500421137437287e-09, + "loss": 0.6858, + "step": 53162 + }, + { + "epoch": 1.95, + "learning_rate": 5.496759072765224e-09, + "loss": 0.687, + "step": 53163 + }, + { + "epoch": 1.95, + "learning_rate": 5.493097008093163e-09, + "loss": 0.6761, + "step": 53164 + }, + { + "epoch": 1.95, + "learning_rate": 5.489434943421101e-09, + "loss": 0.6901, + "step": 53165 + }, + { + "epoch": 1.95, + "learning_rate": 5.485772878749038e-09, + "loss": 0.6954, + "step": 53166 + }, + { + "epoch": 1.95, + "learning_rate": 5.482110814076976e-09, + "loss": 0.6978, + "step": 53167 + }, + { + "epoch": 1.95, + "learning_rate": 5.478448749404914e-09, + "loss": 0.6895, + "step": 53168 + }, + { + "epoch": 1.95, + "learning_rate": 5.474786684732852e-09, + "loss": 0.7002, + "step": 53169 + }, + { + "epoch": 1.95, + "learning_rate": 5.47112462006079e-09, + "loss": 0.6947, + "step": 53170 + }, + { + "epoch": 1.95, + "learning_rate": 5.467462555388728e-09, + "loss": 0.6895, + "step": 53171 + }, + { + "epoch": 1.95, + "learning_rate": 5.4638004907166656e-09, + "loss": 0.6914, + "step": 53172 + }, + { + "epoch": 1.95, + "learning_rate": 5.460138426044604e-09, + "loss": 0.6923, + "step": 53173 + }, + { + "epoch": 1.95, + "learning_rate": 5.456476361372542e-09, + "loss": 0.6958, + "step": 53174 + }, + { + "epoch": 1.95, + "learning_rate": 5.4528142967004796e-09, + "loss": 0.7029, + "step": 53175 + }, + { + "epoch": 1.95, + "learning_rate": 5.449152232028417e-09, + "loss": 0.6741, + "step": 53176 + }, + { + "epoch": 1.95, + "learning_rate": 5.445490167356355e-09, + "loss": 0.7017, + "step": 53177 + }, + { + "epoch": 1.95, + "learning_rate": 5.4418281026842936e-09, + "loss": 0.6854, + "step": 53178 + }, + { + "epoch": 1.95, + "learning_rate": 5.438166038012231e-09, + "loss": 0.698, + "step": 53179 + }, + { + "epoch": 1.95, + "learning_rate": 5.434503973340169e-09, + "loss": 0.6916, + "step": 53180 + }, + { + "epoch": 1.95, + "learning_rate": 5.430841908668107e-09, + "loss": 0.6895, + "step": 53181 + }, + { + "epoch": 1.95, + "learning_rate": 5.427179843996045e-09, + "loss": 0.6871, + "step": 53182 + }, + { + "epoch": 1.95, + "learning_rate": 5.4235177793239824e-09, + "loss": 0.6894, + "step": 53183 + }, + { + "epoch": 1.95, + "learning_rate": 5.419855714651921e-09, + "loss": 0.6882, + "step": 53184 + }, + { + "epoch": 1.95, + "learning_rate": 5.416193649979858e-09, + "loss": 0.7046, + "step": 53185 + }, + { + "epoch": 1.95, + "learning_rate": 5.412531585307796e-09, + "loss": 0.6845, + "step": 53186 + }, + { + "epoch": 1.95, + "learning_rate": 5.408869520635735e-09, + "loss": 0.6877, + "step": 53187 + }, + { + "epoch": 1.95, + "learning_rate": 5.405207455963672e-09, + "loss": 0.6976, + "step": 53188 + }, + { + "epoch": 1.95, + "learning_rate": 5.40154539129161e-09, + "loss": 0.6878, + "step": 53189 + }, + { + "epoch": 1.95, + "learning_rate": 5.397883326619548e-09, + "loss": 0.6852, + "step": 53190 + }, + { + "epoch": 1.95, + "learning_rate": 5.394221261947485e-09, + "loss": 0.6958, + "step": 53191 + }, + { + "epoch": 1.95, + "learning_rate": 5.390559197275424e-09, + "loss": 0.6916, + "step": 53192 + }, + { + "epoch": 1.95, + "learning_rate": 5.386897132603362e-09, + "loss": 0.6923, + "step": 53193 + }, + { + "epoch": 1.95, + "learning_rate": 5.383235067931299e-09, + "loss": 0.6854, + "step": 53194 + }, + { + "epoch": 1.95, + "learning_rate": 5.379573003259237e-09, + "loss": 0.6894, + "step": 53195 + }, + { + "epoch": 1.95, + "learning_rate": 5.375910938587176e-09, + "loss": 0.6937, + "step": 53196 + }, + { + "epoch": 1.95, + "learning_rate": 5.372248873915113e-09, + "loss": 0.6999, + "step": 53197 + }, + { + "epoch": 1.95, + "learning_rate": 5.368586809243051e-09, + "loss": 0.6887, + "step": 53198 + }, + { + "epoch": 1.95, + "learning_rate": 5.364924744570989e-09, + "loss": 0.6913, + "step": 53199 + }, + { + "epoch": 1.95, + "learning_rate": 5.3612626798989265e-09, + "loss": 0.6914, + "step": 53200 + }, + { + "epoch": 1.95, + "learning_rate": 5.357600615226865e-09, + "loss": 0.6784, + "step": 53201 + }, + { + "epoch": 1.95, + "learning_rate": 5.353938550554803e-09, + "loss": 0.7102, + "step": 53202 + }, + { + "epoch": 1.95, + "learning_rate": 5.3502764858827405e-09, + "loss": 0.6914, + "step": 53203 + }, + { + "epoch": 1.95, + "learning_rate": 5.346614421210678e-09, + "loss": 0.6837, + "step": 53204 + }, + { + "epoch": 1.95, + "learning_rate": 5.342952356538616e-09, + "loss": 0.6744, + "step": 53205 + }, + { + "epoch": 1.95, + "learning_rate": 5.3392902918665545e-09, + "loss": 0.7009, + "step": 53206 + }, + { + "epoch": 1.95, + "learning_rate": 5.335628227194492e-09, + "loss": 0.7048, + "step": 53207 + }, + { + "epoch": 1.95, + "learning_rate": 5.331966162522429e-09, + "loss": 0.7018, + "step": 53208 + }, + { + "epoch": 1.95, + "learning_rate": 5.328304097850368e-09, + "loss": 0.688, + "step": 53209 + }, + { + "epoch": 1.95, + "learning_rate": 5.324642033178306e-09, + "loss": 0.6961, + "step": 53210 + }, + { + "epoch": 1.95, + "learning_rate": 5.320979968506243e-09, + "loss": 0.6956, + "step": 53211 + }, + { + "epoch": 1.95, + "learning_rate": 5.317317903834182e-09, + "loss": 0.6965, + "step": 53212 + }, + { + "epoch": 1.95, + "learning_rate": 5.313655839162119e-09, + "loss": 0.7012, + "step": 53213 + }, + { + "epoch": 1.95, + "learning_rate": 5.3099937744900565e-09, + "loss": 0.6939, + "step": 53214 + }, + { + "epoch": 1.95, + "learning_rate": 5.306331709817996e-09, + "loss": 0.7016, + "step": 53215 + }, + { + "epoch": 1.95, + "learning_rate": 5.302669645145933e-09, + "loss": 0.6821, + "step": 53216 + }, + { + "epoch": 1.95, + "learning_rate": 5.2990075804738705e-09, + "loss": 0.6832, + "step": 53217 + }, + { + "epoch": 1.95, + "learning_rate": 5.295345515801809e-09, + "loss": 0.6912, + "step": 53218 + }, + { + "epoch": 1.95, + "learning_rate": 5.291683451129747e-09, + "loss": 0.7057, + "step": 53219 + }, + { + "epoch": 1.95, + "learning_rate": 5.2880213864576845e-09, + "loss": 0.6992, + "step": 53220 + }, + { + "epoch": 1.95, + "learning_rate": 5.284359321785623e-09, + "loss": 0.6833, + "step": 53221 + }, + { + "epoch": 1.95, + "learning_rate": 5.28069725711356e-09, + "loss": 0.6881, + "step": 53222 + }, + { + "epoch": 1.95, + "learning_rate": 5.277035192441498e-09, + "loss": 0.6928, + "step": 53223 + }, + { + "epoch": 1.95, + "learning_rate": 5.273373127769437e-09, + "loss": 0.6934, + "step": 53224 + }, + { + "epoch": 1.95, + "learning_rate": 5.269711063097374e-09, + "loss": 0.6897, + "step": 53225 + }, + { + "epoch": 1.95, + "learning_rate": 5.266048998425312e-09, + "loss": 0.6862, + "step": 53226 + }, + { + "epoch": 1.95, + "learning_rate": 5.26238693375325e-09, + "loss": 0.6911, + "step": 53227 + }, + { + "epoch": 1.95, + "learning_rate": 5.258724869081188e-09, + "loss": 0.6924, + "step": 53228 + }, + { + "epoch": 1.95, + "learning_rate": 5.255062804409126e-09, + "loss": 0.7063, + "step": 53229 + }, + { + "epoch": 1.95, + "learning_rate": 5.251400739737063e-09, + "loss": 0.6949, + "step": 53230 + }, + { + "epoch": 1.95, + "learning_rate": 5.247738675065001e-09, + "loss": 0.6798, + "step": 53231 + }, + { + "epoch": 1.95, + "learning_rate": 5.244076610392939e-09, + "loss": 0.6912, + "step": 53232 + }, + { + "epoch": 1.95, + "learning_rate": 5.240414545720877e-09, + "loss": 0.683, + "step": 53233 + }, + { + "epoch": 1.95, + "learning_rate": 5.236752481048815e-09, + "loss": 0.6941, + "step": 53234 + }, + { + "epoch": 1.95, + "learning_rate": 5.233090416376753e-09, + "loss": 0.6911, + "step": 53235 + }, + { + "epoch": 1.95, + "learning_rate": 5.22942835170469e-09, + "loss": 0.6956, + "step": 53236 + }, + { + "epoch": 1.95, + "learning_rate": 5.2257662870326286e-09, + "loss": 0.6993, + "step": 53237 + }, + { + "epoch": 1.95, + "learning_rate": 5.222104222360567e-09, + "loss": 0.6865, + "step": 53238 + }, + { + "epoch": 1.95, + "learning_rate": 5.218442157688504e-09, + "loss": 0.6978, + "step": 53239 + }, + { + "epoch": 1.95, + "learning_rate": 5.2147800930164426e-09, + "loss": 0.6911, + "step": 53240 + }, + { + "epoch": 1.95, + "learning_rate": 5.21111802834438e-09, + "loss": 0.692, + "step": 53241 + }, + { + "epoch": 1.95, + "learning_rate": 5.207455963672318e-09, + "loss": 0.6961, + "step": 53242 + }, + { + "epoch": 1.95, + "learning_rate": 5.2037938990002566e-09, + "loss": 0.6925, + "step": 53243 + }, + { + "epoch": 1.95, + "learning_rate": 5.200131834328194e-09, + "loss": 0.6785, + "step": 53244 + }, + { + "epoch": 1.95, + "learning_rate": 5.1964697696561315e-09, + "loss": 0.6861, + "step": 53245 + }, + { + "epoch": 1.95, + "learning_rate": 5.19280770498407e-09, + "loss": 0.6896, + "step": 53246 + }, + { + "epoch": 1.95, + "learning_rate": 5.189145640312008e-09, + "loss": 0.6895, + "step": 53247 + }, + { + "epoch": 1.95, + "learning_rate": 5.1854835756399455e-09, + "loss": 0.6926, + "step": 53248 + }, + { + "epoch": 1.95, + "learning_rate": 5.181821510967884e-09, + "loss": 0.6913, + "step": 53249 + }, + { + "epoch": 1.95, + "learning_rate": 5.178159446295821e-09, + "loss": 0.6944, + "step": 53250 + }, + { + "epoch": 1.95, + "learning_rate": 5.1744973816237595e-09, + "loss": 0.6923, + "step": 53251 + }, + { + "epoch": 1.95, + "learning_rate": 5.170835316951697e-09, + "loss": 0.6928, + "step": 53252 + }, + { + "epoch": 1.95, + "learning_rate": 5.167173252279635e-09, + "loss": 0.6974, + "step": 53253 + }, + { + "epoch": 1.95, + "learning_rate": 5.163511187607573e-09, + "loss": 0.6928, + "step": 53254 + }, + { + "epoch": 1.95, + "learning_rate": 5.159849122935511e-09, + "loss": 0.6883, + "step": 53255 + }, + { + "epoch": 1.95, + "learning_rate": 5.156187058263449e-09, + "loss": 0.7022, + "step": 53256 + }, + { + "epoch": 1.95, + "learning_rate": 5.152524993591387e-09, + "loss": 0.7175, + "step": 53257 + }, + { + "epoch": 1.95, + "learning_rate": 5.148862928919324e-09, + "loss": 0.6849, + "step": 53258 + }, + { + "epoch": 1.95, + "learning_rate": 5.145200864247262e-09, + "loss": 0.6932, + "step": 53259 + }, + { + "epoch": 1.95, + "learning_rate": 5.1415387995752e-09, + "loss": 0.6913, + "step": 53260 + }, + { + "epoch": 1.95, + "learning_rate": 5.137876734903138e-09, + "loss": 0.6943, + "step": 53261 + }, + { + "epoch": 1.95, + "learning_rate": 5.134214670231076e-09, + "loss": 0.6915, + "step": 53262 + }, + { + "epoch": 1.95, + "learning_rate": 5.130552605559014e-09, + "loss": 0.6843, + "step": 53263 + }, + { + "epoch": 1.95, + "learning_rate": 5.126890540886951e-09, + "loss": 0.6944, + "step": 53264 + }, + { + "epoch": 1.95, + "learning_rate": 5.12322847621489e-09, + "loss": 0.6881, + "step": 53265 + }, + { + "epoch": 1.95, + "learning_rate": 5.119566411542828e-09, + "loss": 0.6957, + "step": 53266 + }, + { + "epoch": 1.95, + "learning_rate": 5.115904346870765e-09, + "loss": 0.6926, + "step": 53267 + }, + { + "epoch": 1.95, + "learning_rate": 5.1122422821987035e-09, + "loss": 0.6975, + "step": 53268 + }, + { + "epoch": 1.95, + "learning_rate": 5.108580217526641e-09, + "loss": 0.7, + "step": 53269 + }, + { + "epoch": 1.95, + "learning_rate": 5.104918152854579e-09, + "loss": 0.6929, + "step": 53270 + }, + { + "epoch": 1.95, + "learning_rate": 5.1012560881825175e-09, + "loss": 0.6891, + "step": 53271 + }, + { + "epoch": 1.95, + "learning_rate": 5.097594023510455e-09, + "loss": 0.7099, + "step": 53272 + }, + { + "epoch": 1.95, + "learning_rate": 5.093931958838392e-09, + "loss": 0.6996, + "step": 53273 + }, + { + "epoch": 1.95, + "learning_rate": 5.090269894166331e-09, + "loss": 0.7086, + "step": 53274 + }, + { + "epoch": 1.95, + "learning_rate": 5.086607829494269e-09, + "loss": 0.6855, + "step": 53275 + }, + { + "epoch": 1.95, + "learning_rate": 5.082945764822206e-09, + "loss": 0.6911, + "step": 53276 + }, + { + "epoch": 1.95, + "learning_rate": 5.079283700150145e-09, + "loss": 0.6877, + "step": 53277 + }, + { + "epoch": 1.95, + "learning_rate": 5.075621635478082e-09, + "loss": 0.7007, + "step": 53278 + }, + { + "epoch": 1.95, + "learning_rate": 5.07195957080602e-09, + "loss": 0.6881, + "step": 53279 + }, + { + "epoch": 1.95, + "learning_rate": 5.068297506133958e-09, + "loss": 0.6805, + "step": 53280 + }, + { + "epoch": 1.95, + "learning_rate": 5.064635441461896e-09, + "loss": 0.7007, + "step": 53281 + }, + { + "epoch": 1.95, + "learning_rate": 5.0609733767898335e-09, + "loss": 0.6939, + "step": 53282 + }, + { + "epoch": 1.95, + "learning_rate": 5.057311312117772e-09, + "loss": 0.6823, + "step": 53283 + }, + { + "epoch": 1.95, + "learning_rate": 5.05364924744571e-09, + "loss": 0.6902, + "step": 53284 + }, + { + "epoch": 1.95, + "learning_rate": 5.0499871827736475e-09, + "loss": 0.6926, + "step": 53285 + }, + { + "epoch": 1.95, + "learning_rate": 5.046325118101585e-09, + "loss": 0.6981, + "step": 53286 + }, + { + "epoch": 1.95, + "learning_rate": 5.042663053429523e-09, + "loss": 0.7004, + "step": 53287 + }, + { + "epoch": 1.95, + "learning_rate": 5.0390009887574615e-09, + "loss": 0.6853, + "step": 53288 + }, + { + "epoch": 1.95, + "learning_rate": 5.035338924085399e-09, + "loss": 0.706, + "step": 53289 + }, + { + "epoch": 1.95, + "learning_rate": 5.031676859413337e-09, + "loss": 0.6819, + "step": 53290 + }, + { + "epoch": 1.95, + "learning_rate": 5.028014794741275e-09, + "loss": 0.6869, + "step": 53291 + }, + { + "epoch": 1.95, + "learning_rate": 5.024352730069212e-09, + "loss": 0.695, + "step": 53292 + }, + { + "epoch": 1.95, + "learning_rate": 5.020690665397151e-09, + "loss": 0.6926, + "step": 53293 + }, + { + "epoch": 1.95, + "learning_rate": 5.017028600725089e-09, + "loss": 0.6892, + "step": 53294 + }, + { + "epoch": 1.95, + "learning_rate": 5.013366536053026e-09, + "loss": 0.6998, + "step": 53295 + }, + { + "epoch": 1.95, + "learning_rate": 5.0097044713809644e-09, + "loss": 0.6898, + "step": 53296 + }, + { + "epoch": 1.95, + "learning_rate": 5.006042406708903e-09, + "loss": 0.6993, + "step": 53297 + }, + { + "epoch": 1.95, + "learning_rate": 5.00238034203684e-09, + "loss": 0.6959, + "step": 53298 + }, + { + "epoch": 1.95, + "learning_rate": 4.9987182773647784e-09, + "loss": 0.6979, + "step": 53299 + }, + { + "epoch": 1.95, + "learning_rate": 4.995056212692716e-09, + "loss": 0.6832, + "step": 53300 + }, + { + "epoch": 1.95, + "learning_rate": 4.991394148020653e-09, + "loss": 0.6942, + "step": 53301 + }, + { + "epoch": 1.95, + "learning_rate": 4.9877320833485916e-09, + "loss": 0.6821, + "step": 53302 + }, + { + "epoch": 1.95, + "learning_rate": 4.98407001867653e-09, + "loss": 0.7068, + "step": 53303 + }, + { + "epoch": 1.95, + "learning_rate": 4.980407954004467e-09, + "loss": 0.705, + "step": 53304 + }, + { + "epoch": 1.95, + "learning_rate": 4.9767458893324056e-09, + "loss": 0.6887, + "step": 53305 + }, + { + "epoch": 1.95, + "learning_rate": 4.973083824660343e-09, + "loss": 0.6979, + "step": 53306 + }, + { + "epoch": 1.95, + "learning_rate": 4.969421759988281e-09, + "loss": 0.7149, + "step": 53307 + }, + { + "epoch": 1.95, + "learning_rate": 4.965759695316219e-09, + "loss": 0.6973, + "step": 53308 + }, + { + "epoch": 1.95, + "learning_rate": 4.962097630644157e-09, + "loss": 0.6938, + "step": 53309 + }, + { + "epoch": 1.95, + "learning_rate": 4.9584355659720945e-09, + "loss": 0.6916, + "step": 53310 + }, + { + "epoch": 1.95, + "learning_rate": 4.954773501300033e-09, + "loss": 0.6814, + "step": 53311 + }, + { + "epoch": 1.95, + "learning_rate": 4.951111436627971e-09, + "loss": 0.6881, + "step": 53312 + }, + { + "epoch": 1.95, + "learning_rate": 4.9474493719559085e-09, + "loss": 0.7025, + "step": 53313 + }, + { + "epoch": 1.95, + "learning_rate": 4.943787307283846e-09, + "loss": 0.6836, + "step": 53314 + }, + { + "epoch": 1.95, + "learning_rate": 4.940125242611784e-09, + "loss": 0.6976, + "step": 53315 + }, + { + "epoch": 1.95, + "learning_rate": 4.9364631779397225e-09, + "loss": 0.6946, + "step": 53316 + }, + { + "epoch": 1.95, + "learning_rate": 4.93280111326766e-09, + "loss": 0.7003, + "step": 53317 + }, + { + "epoch": 1.95, + "learning_rate": 4.929139048595598e-09, + "loss": 0.7061, + "step": 53318 + }, + { + "epoch": 1.95, + "learning_rate": 4.925476983923536e-09, + "loss": 0.6955, + "step": 53319 + }, + { + "epoch": 1.95, + "learning_rate": 4.921814919251474e-09, + "loss": 0.6919, + "step": 53320 + }, + { + "epoch": 1.95, + "learning_rate": 4.918152854579412e-09, + "loss": 0.6975, + "step": 53321 + }, + { + "epoch": 1.95, + "learning_rate": 4.91449078990735e-09, + "loss": 0.6896, + "step": 53322 + }, + { + "epoch": 1.95, + "learning_rate": 4.910828725235287e-09, + "loss": 0.6985, + "step": 53323 + }, + { + "epoch": 1.95, + "learning_rate": 4.907166660563225e-09, + "loss": 0.6871, + "step": 53324 + }, + { + "epoch": 1.95, + "learning_rate": 4.903504595891164e-09, + "loss": 0.7037, + "step": 53325 + }, + { + "epoch": 1.95, + "learning_rate": 4.899842531219101e-09, + "loss": 0.6962, + "step": 53326 + }, + { + "epoch": 1.95, + "learning_rate": 4.896180466547039e-09, + "loss": 0.6979, + "step": 53327 + }, + { + "epoch": 1.95, + "learning_rate": 4.892518401874977e-09, + "loss": 0.6815, + "step": 53328 + }, + { + "epoch": 1.95, + "learning_rate": 4.888856337202915e-09, + "loss": 0.6942, + "step": 53329 + }, + { + "epoch": 1.95, + "learning_rate": 4.8851942725308525e-09, + "loss": 0.7021, + "step": 53330 + }, + { + "epoch": 1.95, + "learning_rate": 4.881532207858791e-09, + "loss": 0.6847, + "step": 53331 + }, + { + "epoch": 1.95, + "learning_rate": 4.877870143186728e-09, + "loss": 0.6897, + "step": 53332 + }, + { + "epoch": 1.95, + "learning_rate": 4.874208078514666e-09, + "loss": 0.6853, + "step": 53333 + }, + { + "epoch": 1.95, + "learning_rate": 4.870546013842605e-09, + "loss": 0.6921, + "step": 53334 + }, + { + "epoch": 1.95, + "learning_rate": 4.866883949170542e-09, + "loss": 0.6955, + "step": 53335 + }, + { + "epoch": 1.95, + "learning_rate": 4.86322188449848e-09, + "loss": 0.6822, + "step": 53336 + }, + { + "epoch": 1.95, + "learning_rate": 4.859559819826418e-09, + "loss": 0.7022, + "step": 53337 + }, + { + "epoch": 1.95, + "learning_rate": 4.855897755154355e-09, + "loss": 0.7068, + "step": 53338 + }, + { + "epoch": 1.95, + "learning_rate": 4.852235690482294e-09, + "loss": 0.6939, + "step": 53339 + }, + { + "epoch": 1.95, + "learning_rate": 4.848573625810232e-09, + "loss": 0.6918, + "step": 53340 + }, + { + "epoch": 1.95, + "learning_rate": 4.844911561138169e-09, + "loss": 0.7022, + "step": 53341 + }, + { + "epoch": 1.95, + "learning_rate": 4.841249496466107e-09, + "loss": 0.6913, + "step": 53342 + }, + { + "epoch": 1.95, + "learning_rate": 4.837587431794046e-09, + "loss": 0.6901, + "step": 53343 + }, + { + "epoch": 1.95, + "learning_rate": 4.833925367121983e-09, + "loss": 0.6964, + "step": 53344 + }, + { + "epoch": 1.95, + "learning_rate": 4.830263302449921e-09, + "loss": 0.6901, + "step": 53345 + }, + { + "epoch": 1.95, + "learning_rate": 4.826601237777859e-09, + "loss": 0.6997, + "step": 53346 + }, + { + "epoch": 1.95, + "learning_rate": 4.8229391731057966e-09, + "loss": 0.6746, + "step": 53347 + }, + { + "epoch": 1.95, + "learning_rate": 4.819277108433735e-09, + "loss": 0.6881, + "step": 53348 + }, + { + "epoch": 1.95, + "learning_rate": 4.815615043761673e-09, + "loss": 0.6851, + "step": 53349 + }, + { + "epoch": 1.95, + "learning_rate": 4.8119529790896106e-09, + "loss": 0.6956, + "step": 53350 + }, + { + "epoch": 1.95, + "learning_rate": 4.808290914417548e-09, + "loss": 0.689, + "step": 53351 + }, + { + "epoch": 1.95, + "learning_rate": 4.804628849745486e-09, + "loss": 0.6847, + "step": 53352 + }, + { + "epoch": 1.95, + "learning_rate": 4.8009667850734245e-09, + "loss": 0.6813, + "step": 53353 + }, + { + "epoch": 1.95, + "learning_rate": 4.797304720401362e-09, + "loss": 0.6983, + "step": 53354 + }, + { + "epoch": 1.95, + "learning_rate": 4.7936426557292994e-09, + "loss": 0.6903, + "step": 53355 + }, + { + "epoch": 1.95, + "learning_rate": 4.789980591057238e-09, + "loss": 0.6951, + "step": 53356 + }, + { + "epoch": 1.95, + "learning_rate": 4.786318526385176e-09, + "loss": 0.6855, + "step": 53357 + }, + { + "epoch": 1.95, + "learning_rate": 4.7826564617131134e-09, + "loss": 0.7063, + "step": 53358 + }, + { + "epoch": 1.95, + "learning_rate": 4.778994397041052e-09, + "loss": 0.7016, + "step": 53359 + }, + { + "epoch": 1.95, + "learning_rate": 4.775332332368989e-09, + "loss": 0.6903, + "step": 53360 + }, + { + "epoch": 1.95, + "learning_rate": 4.771670267696927e-09, + "loss": 0.6992, + "step": 53361 + }, + { + "epoch": 1.95, + "learning_rate": 4.768008203024866e-09, + "loss": 0.6857, + "step": 53362 + }, + { + "epoch": 1.95, + "learning_rate": 4.764346138352803e-09, + "loss": 0.6944, + "step": 53363 + }, + { + "epoch": 1.95, + "learning_rate": 4.760684073680741e-09, + "loss": 0.6938, + "step": 53364 + }, + { + "epoch": 1.95, + "learning_rate": 4.757022009008679e-09, + "loss": 0.6861, + "step": 53365 + }, + { + "epoch": 1.95, + "learning_rate": 4.753359944336617e-09, + "loss": 0.7023, + "step": 53366 + }, + { + "epoch": 1.95, + "learning_rate": 4.749697879664555e-09, + "loss": 0.6939, + "step": 53367 + }, + { + "epoch": 1.95, + "learning_rate": 4.746035814992493e-09, + "loss": 0.6918, + "step": 53368 + }, + { + "epoch": 1.95, + "learning_rate": 4.74237375032043e-09, + "loss": 0.6907, + "step": 53369 + }, + { + "epoch": 1.95, + "learning_rate": 4.738711685648368e-09, + "loss": 0.692, + "step": 53370 + }, + { + "epoch": 1.95, + "learning_rate": 4.735049620976307e-09, + "loss": 0.7041, + "step": 53371 + }, + { + "epoch": 1.95, + "learning_rate": 4.731387556304244e-09, + "loss": 0.6897, + "step": 53372 + }, + { + "epoch": 1.95, + "learning_rate": 4.727725491632182e-09, + "loss": 0.7025, + "step": 53373 + }, + { + "epoch": 1.95, + "learning_rate": 4.72406342696012e-09, + "loss": 0.6987, + "step": 53374 + }, + { + "epoch": 1.95, + "learning_rate": 4.720401362288058e-09, + "loss": 0.7018, + "step": 53375 + }, + { + "epoch": 1.95, + "learning_rate": 4.716739297615996e-09, + "loss": 0.6925, + "step": 53376 + }, + { + "epoch": 1.95, + "learning_rate": 4.713077232943933e-09, + "loss": 0.6885, + "step": 53377 + }, + { + "epoch": 1.95, + "learning_rate": 4.7094151682718715e-09, + "loss": 0.6983, + "step": 53378 + }, + { + "epoch": 1.95, + "learning_rate": 4.705753103599809e-09, + "loss": 0.7057, + "step": 53379 + }, + { + "epoch": 1.95, + "learning_rate": 4.702091038927747e-09, + "loss": 0.6942, + "step": 53380 + }, + { + "epoch": 1.95, + "learning_rate": 4.6984289742556855e-09, + "loss": 0.6895, + "step": 53381 + }, + { + "epoch": 1.95, + "learning_rate": 4.694766909583623e-09, + "loss": 0.7006, + "step": 53382 + }, + { + "epoch": 1.95, + "learning_rate": 4.69110484491156e-09, + "loss": 0.6784, + "step": 53383 + }, + { + "epoch": 1.95, + "learning_rate": 4.687442780239499e-09, + "loss": 0.6957, + "step": 53384 + }, + { + "epoch": 1.95, + "learning_rate": 4.683780715567437e-09, + "loss": 0.6828, + "step": 53385 + }, + { + "epoch": 1.95, + "learning_rate": 4.680118650895374e-09, + "loss": 0.7071, + "step": 53386 + }, + { + "epoch": 1.95, + "learning_rate": 4.676456586223313e-09, + "loss": 0.6865, + "step": 53387 + }, + { + "epoch": 1.95, + "learning_rate": 4.67279452155125e-09, + "loss": 0.6934, + "step": 53388 + }, + { + "epoch": 1.95, + "learning_rate": 4.669132456879188e-09, + "loss": 0.7019, + "step": 53389 + }, + { + "epoch": 1.95, + "learning_rate": 4.665470392207127e-09, + "loss": 0.6913, + "step": 53390 + }, + { + "epoch": 1.95, + "learning_rate": 4.661808327535064e-09, + "loss": 0.6905, + "step": 53391 + }, + { + "epoch": 1.95, + "learning_rate": 4.6581462628630015e-09, + "loss": 0.6864, + "step": 53392 + }, + { + "epoch": 1.95, + "learning_rate": 4.65448419819094e-09, + "loss": 0.6975, + "step": 53393 + }, + { + "epoch": 1.95, + "learning_rate": 4.650822133518878e-09, + "loss": 0.6947, + "step": 53394 + }, + { + "epoch": 1.95, + "learning_rate": 4.6471600688468155e-09, + "loss": 0.6984, + "step": 53395 + }, + { + "epoch": 1.95, + "learning_rate": 4.643498004174754e-09, + "loss": 0.6976, + "step": 53396 + }, + { + "epoch": 1.95, + "learning_rate": 4.639835939502691e-09, + "loss": 0.7019, + "step": 53397 + }, + { + "epoch": 1.95, + "learning_rate": 4.6361738748306295e-09, + "loss": 0.6771, + "step": 53398 + }, + { + "epoch": 1.95, + "learning_rate": 4.632511810158567e-09, + "loss": 0.6982, + "step": 53399 + }, + { + "epoch": 1.95, + "learning_rate": 4.628849745486505e-09, + "loss": 0.6925, + "step": 53400 + }, + { + "epoch": 1.95, + "learning_rate": 4.625187680814443e-09, + "loss": 0.701, + "step": 53401 + }, + { + "epoch": 1.95, + "learning_rate": 4.621525616142381e-09, + "loss": 0.6851, + "step": 53402 + }, + { + "epoch": 1.95, + "learning_rate": 4.617863551470319e-09, + "loss": 0.684, + "step": 53403 + }, + { + "epoch": 1.95, + "learning_rate": 4.614201486798257e-09, + "loss": 0.6972, + "step": 53404 + }, + { + "epoch": 1.95, + "learning_rate": 4.610539422126194e-09, + "loss": 0.6961, + "step": 53405 + }, + { + "epoch": 1.95, + "learning_rate": 4.606877357454132e-09, + "loss": 0.6966, + "step": 53406 + }, + { + "epoch": 1.95, + "learning_rate": 4.60321529278207e-09, + "loss": 0.6866, + "step": 53407 + }, + { + "epoch": 1.95, + "learning_rate": 4.599553228110008e-09, + "loss": 0.692, + "step": 53408 + }, + { + "epoch": 1.95, + "learning_rate": 4.595891163437946e-09, + "loss": 0.6878, + "step": 53409 + }, + { + "epoch": 1.95, + "learning_rate": 4.592229098765884e-09, + "loss": 0.7122, + "step": 53410 + }, + { + "epoch": 1.95, + "learning_rate": 4.588567034093821e-09, + "loss": 0.6981, + "step": 53411 + }, + { + "epoch": 1.95, + "learning_rate": 4.58490496942176e-09, + "loss": 0.7043, + "step": 53412 + }, + { + "epoch": 1.95, + "learning_rate": 4.581242904749698e-09, + "loss": 0.7014, + "step": 53413 + }, + { + "epoch": 1.95, + "learning_rate": 4.577580840077635e-09, + "loss": 0.6843, + "step": 53414 + }, + { + "epoch": 1.95, + "learning_rate": 4.5739187754055736e-09, + "loss": 0.6909, + "step": 53415 + }, + { + "epoch": 1.95, + "learning_rate": 4.570256710733511e-09, + "loss": 0.6933, + "step": 53416 + }, + { + "epoch": 1.95, + "learning_rate": 4.566594646061449e-09, + "loss": 0.7076, + "step": 53417 + }, + { + "epoch": 1.95, + "learning_rate": 4.5629325813893876e-09, + "loss": 0.6919, + "step": 53418 + }, + { + "epoch": 1.95, + "learning_rate": 4.559270516717325e-09, + "loss": 0.6901, + "step": 53419 + }, + { + "epoch": 1.95, + "learning_rate": 4.5556084520452625e-09, + "loss": 0.6917, + "step": 53420 + }, + { + "epoch": 1.95, + "learning_rate": 4.551946387373201e-09, + "loss": 0.6906, + "step": 53421 + }, + { + "epoch": 1.95, + "learning_rate": 4.548284322701139e-09, + "loss": 0.6928, + "step": 53422 + }, + { + "epoch": 1.95, + "learning_rate": 4.5446222580290764e-09, + "loss": 0.6964, + "step": 53423 + }, + { + "epoch": 1.95, + "learning_rate": 4.540960193357015e-09, + "loss": 0.6893, + "step": 53424 + }, + { + "epoch": 1.95, + "learning_rate": 4.537298128684952e-09, + "loss": 0.7092, + "step": 53425 + }, + { + "epoch": 1.95, + "learning_rate": 4.5336360640128904e-09, + "loss": 0.7039, + "step": 53426 + }, + { + "epoch": 1.95, + "learning_rate": 4.529973999340828e-09, + "loss": 0.7028, + "step": 53427 + }, + { + "epoch": 1.95, + "learning_rate": 4.526311934668766e-09, + "loss": 0.6901, + "step": 53428 + }, + { + "epoch": 1.95, + "learning_rate": 4.522649869996704e-09, + "loss": 0.6914, + "step": 53429 + }, + { + "epoch": 1.95, + "learning_rate": 4.518987805324642e-09, + "loss": 0.6977, + "step": 53430 + }, + { + "epoch": 1.95, + "learning_rate": 4.51532574065258e-09, + "loss": 0.7007, + "step": 53431 + }, + { + "epoch": 1.95, + "learning_rate": 4.511663675980518e-09, + "loss": 0.6946, + "step": 53432 + }, + { + "epoch": 1.95, + "learning_rate": 4.508001611308455e-09, + "loss": 0.7075, + "step": 53433 + }, + { + "epoch": 1.95, + "learning_rate": 4.504339546636393e-09, + "loss": 0.6897, + "step": 53434 + }, + { + "epoch": 1.96, + "learning_rate": 4.500677481964332e-09, + "loss": 0.6943, + "step": 53435 + }, + { + "epoch": 1.96, + "learning_rate": 4.497015417292269e-09, + "loss": 0.7017, + "step": 53436 + }, + { + "epoch": 1.96, + "learning_rate": 4.493353352620207e-09, + "loss": 0.7023, + "step": 53437 + }, + { + "epoch": 1.96, + "learning_rate": 4.489691287948145e-09, + "loss": 0.7023, + "step": 53438 + }, + { + "epoch": 1.96, + "learning_rate": 4.486029223276082e-09, + "loss": 0.6724, + "step": 53439 + }, + { + "epoch": 1.96, + "learning_rate": 4.482367158604021e-09, + "loss": 0.7054, + "step": 53440 + }, + { + "epoch": 1.96, + "learning_rate": 4.478705093931959e-09, + "loss": 0.6937, + "step": 53441 + }, + { + "epoch": 1.96, + "learning_rate": 4.475043029259896e-09, + "loss": 0.6973, + "step": 53442 + }, + { + "epoch": 1.96, + "learning_rate": 4.4713809645878345e-09, + "loss": 0.6895, + "step": 53443 + }, + { + "epoch": 1.96, + "learning_rate": 4.467718899915773e-09, + "loss": 0.7087, + "step": 53444 + }, + { + "epoch": 1.96, + "learning_rate": 4.46405683524371e-09, + "loss": 0.6871, + "step": 53445 + }, + { + "epoch": 1.96, + "learning_rate": 4.4603947705716485e-09, + "loss": 0.6944, + "step": 53446 + }, + { + "epoch": 1.96, + "learning_rate": 4.456732705899586e-09, + "loss": 0.6903, + "step": 53447 + }, + { + "epoch": 1.96, + "learning_rate": 4.453070641227523e-09, + "loss": 0.6853, + "step": 53448 + }, + { + "epoch": 1.96, + "learning_rate": 4.449408576555462e-09, + "loss": 0.6868, + "step": 53449 + }, + { + "epoch": 1.96, + "learning_rate": 4.4457465118834e-09, + "loss": 0.686, + "step": 53450 + }, + { + "epoch": 1.96, + "learning_rate": 4.442084447211337e-09, + "loss": 0.6986, + "step": 53451 + }, + { + "epoch": 1.96, + "learning_rate": 4.438422382539276e-09, + "loss": 0.6892, + "step": 53452 + }, + { + "epoch": 1.96, + "learning_rate": 4.434760317867213e-09, + "loss": 0.6927, + "step": 53453 + }, + { + "epoch": 1.96, + "learning_rate": 4.431098253195151e-09, + "loss": 0.687, + "step": 53454 + }, + { + "epoch": 1.96, + "learning_rate": 4.427436188523089e-09, + "loss": 0.6947, + "step": 53455 + }, + { + "epoch": 1.96, + "learning_rate": 4.423774123851027e-09, + "loss": 0.6969, + "step": 53456 + }, + { + "epoch": 1.96, + "learning_rate": 4.4201120591789645e-09, + "loss": 0.6987, + "step": 53457 + }, + { + "epoch": 1.96, + "learning_rate": 4.416449994506903e-09, + "loss": 0.6948, + "step": 53458 + }, + { + "epoch": 1.96, + "learning_rate": 4.412787929834841e-09, + "loss": 0.6892, + "step": 53459 + }, + { + "epoch": 1.96, + "learning_rate": 4.4091258651627785e-09, + "loss": 0.6975, + "step": 53460 + }, + { + "epoch": 1.96, + "learning_rate": 4.405463800490716e-09, + "loss": 0.6812, + "step": 53461 + }, + { + "epoch": 1.96, + "learning_rate": 4.401801735818654e-09, + "loss": 0.6857, + "step": 53462 + }, + { + "epoch": 1.96, + "learning_rate": 4.3981396711465925e-09, + "loss": 0.6869, + "step": 53463 + }, + { + "epoch": 1.96, + "learning_rate": 4.39447760647453e-09, + "loss": 0.6964, + "step": 53464 + }, + { + "epoch": 1.96, + "learning_rate": 4.390815541802468e-09, + "loss": 0.6962, + "step": 53465 + }, + { + "epoch": 1.96, + "learning_rate": 4.387153477130406e-09, + "loss": 0.7013, + "step": 53466 + }, + { + "epoch": 1.96, + "learning_rate": 4.383491412458344e-09, + "loss": 0.6995, + "step": 53467 + }, + { + "epoch": 1.96, + "learning_rate": 4.379829347786282e-09, + "loss": 0.6827, + "step": 53468 + }, + { + "epoch": 1.96, + "learning_rate": 4.37616728311422e-09, + "loss": 0.6794, + "step": 53469 + }, + { + "epoch": 1.96, + "learning_rate": 4.372505218442157e-09, + "loss": 0.699, + "step": 53470 + }, + { + "epoch": 1.96, + "learning_rate": 4.3688431537700954e-09, + "loss": 0.7014, + "step": 53471 + }, + { + "epoch": 1.96, + "learning_rate": 4.365181089098034e-09, + "loss": 0.6962, + "step": 53472 + }, + { + "epoch": 1.96, + "learning_rate": 4.361519024425971e-09, + "loss": 0.6921, + "step": 53473 + }, + { + "epoch": 1.96, + "learning_rate": 4.357856959753909e-09, + "loss": 0.6896, + "step": 53474 + }, + { + "epoch": 1.96, + "learning_rate": 4.354194895081847e-09, + "loss": 0.6883, + "step": 53475 + }, + { + "epoch": 1.96, + "learning_rate": 4.350532830409784e-09, + "loss": 0.6834, + "step": 53476 + }, + { + "epoch": 1.96, + "learning_rate": 4.3468707657377226e-09, + "loss": 0.6982, + "step": 53477 + }, + { + "epoch": 1.96, + "learning_rate": 4.343208701065661e-09, + "loss": 0.6843, + "step": 53478 + }, + { + "epoch": 1.96, + "learning_rate": 4.339546636393598e-09, + "loss": 0.6963, + "step": 53479 + }, + { + "epoch": 1.96, + "learning_rate": 4.335884571721536e-09, + "loss": 0.6976, + "step": 53480 + }, + { + "epoch": 1.96, + "learning_rate": 4.332222507049475e-09, + "loss": 0.6974, + "step": 53481 + }, + { + "epoch": 1.96, + "learning_rate": 4.328560442377412e-09, + "loss": 0.6904, + "step": 53482 + }, + { + "epoch": 1.96, + "learning_rate": 4.32489837770535e-09, + "loss": 0.6937, + "step": 53483 + }, + { + "epoch": 1.96, + "learning_rate": 4.321236313033288e-09, + "loss": 0.6918, + "step": 53484 + }, + { + "epoch": 1.96, + "learning_rate": 4.3175742483612255e-09, + "loss": 0.7026, + "step": 53485 + }, + { + "epoch": 1.96, + "learning_rate": 4.313912183689164e-09, + "loss": 0.6837, + "step": 53486 + }, + { + "epoch": 1.96, + "learning_rate": 4.310250119017102e-09, + "loss": 0.6801, + "step": 53487 + }, + { + "epoch": 1.96, + "learning_rate": 4.3065880543450395e-09, + "loss": 0.6995, + "step": 53488 + }, + { + "epoch": 1.96, + "learning_rate": 4.302925989672977e-09, + "loss": 0.6994, + "step": 53489 + }, + { + "epoch": 1.96, + "learning_rate": 4.299263925000916e-09, + "loss": 0.6948, + "step": 53490 + }, + { + "epoch": 1.96, + "learning_rate": 4.2956018603288535e-09, + "loss": 0.7009, + "step": 53491 + }, + { + "epoch": 1.96, + "learning_rate": 4.291939795656791e-09, + "loss": 0.7029, + "step": 53492 + }, + { + "epoch": 1.96, + "learning_rate": 4.288277730984729e-09, + "loss": 0.6989, + "step": 53493 + }, + { + "epoch": 1.96, + "learning_rate": 4.284615666312667e-09, + "loss": 0.6919, + "step": 53494 + }, + { + "epoch": 1.96, + "learning_rate": 4.280953601640605e-09, + "loss": 0.6873, + "step": 53495 + }, + { + "epoch": 1.96, + "learning_rate": 4.277291536968543e-09, + "loss": 0.6977, + "step": 53496 + }, + { + "epoch": 1.96, + "learning_rate": 4.273629472296481e-09, + "loss": 0.6873, + "step": 53497 + }, + { + "epoch": 1.96, + "learning_rate": 4.269967407624418e-09, + "loss": 0.7, + "step": 53498 + }, + { + "epoch": 1.96, + "learning_rate": 4.266305342952356e-09, + "loss": 0.6984, + "step": 53499 + }, + { + "epoch": 1.96, + "learning_rate": 4.262643278280295e-09, + "loss": 0.6969, + "step": 53500 + }, + { + "epoch": 1.96, + "learning_rate": 4.258981213608232e-09, + "loss": 0.6844, + "step": 53501 + }, + { + "epoch": 1.96, + "learning_rate": 4.2553191489361695e-09, + "loss": 0.685, + "step": 53502 + }, + { + "epoch": 1.96, + "learning_rate": 4.251657084264108e-09, + "loss": 0.6783, + "step": 53503 + }, + { + "epoch": 1.96, + "learning_rate": 4.247995019592046e-09, + "loss": 0.688, + "step": 53504 + }, + { + "epoch": 1.96, + "learning_rate": 4.2443329549199835e-09, + "loss": 0.6941, + "step": 53505 + }, + { + "epoch": 1.96, + "learning_rate": 4.240670890247922e-09, + "loss": 0.7022, + "step": 53506 + }, + { + "epoch": 1.96, + "learning_rate": 4.237008825575859e-09, + "loss": 0.7012, + "step": 53507 + }, + { + "epoch": 1.96, + "learning_rate": 4.233346760903797e-09, + "loss": 0.691, + "step": 53508 + }, + { + "epoch": 1.96, + "learning_rate": 4.229684696231736e-09, + "loss": 0.6959, + "step": 53509 + }, + { + "epoch": 1.96, + "learning_rate": 4.226022631559673e-09, + "loss": 0.6925, + "step": 53510 + }, + { + "epoch": 1.96, + "learning_rate": 4.222360566887611e-09, + "loss": 0.6912, + "step": 53511 + }, + { + "epoch": 1.96, + "learning_rate": 4.218698502215549e-09, + "loss": 0.6929, + "step": 53512 + }, + { + "epoch": 1.96, + "learning_rate": 4.215036437543487e-09, + "loss": 0.6983, + "step": 53513 + }, + { + "epoch": 1.96, + "learning_rate": 4.211374372871425e-09, + "loss": 0.6954, + "step": 53514 + }, + { + "epoch": 1.96, + "learning_rate": 4.207712308199363e-09, + "loss": 0.7027, + "step": 53515 + }, + { + "epoch": 1.96, + "learning_rate": 4.2040502435273e-09, + "loss": 0.6828, + "step": 53516 + }, + { + "epoch": 1.96, + "learning_rate": 4.200388178855238e-09, + "loss": 0.6833, + "step": 53517 + }, + { + "epoch": 1.96, + "learning_rate": 4.196726114183177e-09, + "loss": 0.6991, + "step": 53518 + }, + { + "epoch": 1.96, + "learning_rate": 4.193064049511114e-09, + "loss": 0.6932, + "step": 53519 + }, + { + "epoch": 1.96, + "learning_rate": 4.189401984839052e-09, + "loss": 0.6886, + "step": 53520 + }, + { + "epoch": 1.96, + "learning_rate": 4.18573992016699e-09, + "loss": 0.6921, + "step": 53521 + }, + { + "epoch": 1.96, + "learning_rate": 4.182077855494928e-09, + "loss": 0.6769, + "step": 53522 + }, + { + "epoch": 1.96, + "learning_rate": 4.178415790822866e-09, + "loss": 0.6893, + "step": 53523 + }, + { + "epoch": 1.96, + "learning_rate": 4.174753726150803e-09, + "loss": 0.6922, + "step": 53524 + }, + { + "epoch": 1.96, + "learning_rate": 4.1710916614787415e-09, + "loss": 0.7111, + "step": 53525 + }, + { + "epoch": 1.96, + "learning_rate": 4.167429596806679e-09, + "loss": 0.6946, + "step": 53526 + }, + { + "epoch": 1.96, + "learning_rate": 4.163767532134617e-09, + "loss": 0.6911, + "step": 53527 + }, + { + "epoch": 1.96, + "learning_rate": 4.1601054674625555e-09, + "loss": 0.6816, + "step": 53528 + }, + { + "epoch": 1.96, + "learning_rate": 4.156443402790493e-09, + "loss": 0.6946, + "step": 53529 + }, + { + "epoch": 1.96, + "learning_rate": 4.1527813381184304e-09, + "loss": 0.6952, + "step": 53530 + }, + { + "epoch": 1.96, + "learning_rate": 4.149119273446369e-09, + "loss": 0.6991, + "step": 53531 + }, + { + "epoch": 1.96, + "learning_rate": 4.145457208774307e-09, + "loss": 0.6824, + "step": 53532 + }, + { + "epoch": 1.96, + "learning_rate": 4.1417951441022444e-09, + "loss": 0.6952, + "step": 53533 + }, + { + "epoch": 1.96, + "learning_rate": 4.138133079430183e-09, + "loss": 0.6887, + "step": 53534 + }, + { + "epoch": 1.96, + "learning_rate": 4.13447101475812e-09, + "loss": 0.6979, + "step": 53535 + }, + { + "epoch": 1.96, + "learning_rate": 4.1308089500860584e-09, + "loss": 0.698, + "step": 53536 + }, + { + "epoch": 1.96, + "learning_rate": 4.127146885413997e-09, + "loss": 0.6747, + "step": 53537 + }, + { + "epoch": 1.96, + "learning_rate": 4.123484820741934e-09, + "loss": 0.6954, + "step": 53538 + }, + { + "epoch": 1.96, + "learning_rate": 4.119822756069872e-09, + "loss": 0.6921, + "step": 53539 + }, + { + "epoch": 1.96, + "learning_rate": 4.11616069139781e-09, + "loss": 0.7016, + "step": 53540 + }, + { + "epoch": 1.96, + "learning_rate": 4.112498626725748e-09, + "loss": 0.6963, + "step": 53541 + }, + { + "epoch": 1.96, + "learning_rate": 4.108836562053686e-09, + "loss": 0.7044, + "step": 53542 + }, + { + "epoch": 1.96, + "learning_rate": 4.105174497381624e-09, + "loss": 0.685, + "step": 53543 + }, + { + "epoch": 1.96, + "learning_rate": 4.101512432709561e-09, + "loss": 0.7025, + "step": 53544 + }, + { + "epoch": 1.96, + "learning_rate": 4.0978503680375e-09, + "loss": 0.6963, + "step": 53545 + }, + { + "epoch": 1.96, + "learning_rate": 4.094188303365437e-09, + "loss": 0.7011, + "step": 53546 + }, + { + "epoch": 1.96, + "learning_rate": 4.090526238693375e-09, + "loss": 0.6913, + "step": 53547 + }, + { + "epoch": 1.96, + "learning_rate": 4.086864174021313e-09, + "loss": 0.6926, + "step": 53548 + }, + { + "epoch": 1.96, + "learning_rate": 4.083202109349251e-09, + "loss": 0.6953, + "step": 53549 + }, + { + "epoch": 1.96, + "learning_rate": 4.079540044677189e-09, + "loss": 0.6891, + "step": 53550 + }, + { + "epoch": 1.96, + "learning_rate": 4.075877980005127e-09, + "loss": 0.7006, + "step": 53551 + }, + { + "epoch": 1.96, + "learning_rate": 4.072215915333064e-09, + "loss": 0.7003, + "step": 53552 + }, + { + "epoch": 1.96, + "learning_rate": 4.0685538506610025e-09, + "loss": 0.6964, + "step": 53553 + }, + { + "epoch": 1.96, + "learning_rate": 4.06489178598894e-09, + "loss": 0.6867, + "step": 53554 + }, + { + "epoch": 1.96, + "learning_rate": 4.061229721316878e-09, + "loss": 0.7042, + "step": 53555 + }, + { + "epoch": 1.96, + "learning_rate": 4.0575676566448165e-09, + "loss": 0.6899, + "step": 53556 + }, + { + "epoch": 1.96, + "learning_rate": 4.053905591972754e-09, + "loss": 0.6924, + "step": 53557 + }, + { + "epoch": 1.96, + "learning_rate": 4.050243527300691e-09, + "loss": 0.7009, + "step": 53558 + }, + { + "epoch": 1.96, + "learning_rate": 4.0465814626286305e-09, + "loss": 0.6824, + "step": 53559 + }, + { + "epoch": 1.96, + "learning_rate": 4.042919397956568e-09, + "loss": 0.692, + "step": 53560 + }, + { + "epoch": 1.96, + "learning_rate": 4.039257333284505e-09, + "loss": 0.6928, + "step": 53561 + }, + { + "epoch": 1.96, + "learning_rate": 4.035595268612444e-09, + "loss": 0.6959, + "step": 53562 + }, + { + "epoch": 1.96, + "learning_rate": 4.031933203940381e-09, + "loss": 0.7032, + "step": 53563 + }, + { + "epoch": 1.96, + "learning_rate": 4.028271139268319e-09, + "loss": 0.6873, + "step": 53564 + }, + { + "epoch": 1.96, + "learning_rate": 4.024609074596258e-09, + "loss": 0.6886, + "step": 53565 + }, + { + "epoch": 1.96, + "learning_rate": 4.020947009924195e-09, + "loss": 0.6996, + "step": 53566 + }, + { + "epoch": 1.96, + "learning_rate": 4.0172849452521325e-09, + "loss": 0.6834, + "step": 53567 + }, + { + "epoch": 1.96, + "learning_rate": 4.013622880580071e-09, + "loss": 0.6985, + "step": 53568 + }, + { + "epoch": 1.96, + "learning_rate": 4.009960815908009e-09, + "loss": 0.7065, + "step": 53569 + }, + { + "epoch": 1.96, + "learning_rate": 4.0062987512359465e-09, + "loss": 0.6914, + "step": 53570 + }, + { + "epoch": 1.96, + "learning_rate": 4.002636686563885e-09, + "loss": 0.6883, + "step": 53571 + }, + { + "epoch": 1.96, + "learning_rate": 3.998974621891822e-09, + "loss": 0.704, + "step": 53572 + }, + { + "epoch": 1.96, + "learning_rate": 3.9953125572197605e-09, + "loss": 0.692, + "step": 53573 + }, + { + "epoch": 1.96, + "learning_rate": 3.991650492547698e-09, + "loss": 0.6805, + "step": 53574 + }, + { + "epoch": 1.96, + "learning_rate": 3.987988427875636e-09, + "loss": 0.6896, + "step": 53575 + }, + { + "epoch": 1.96, + "learning_rate": 3.984326363203574e-09, + "loss": 0.7027, + "step": 53576 + }, + { + "epoch": 1.96, + "learning_rate": 3.980664298531512e-09, + "loss": 0.6896, + "step": 53577 + }, + { + "epoch": 1.96, + "learning_rate": 3.97700223385945e-09, + "loss": 0.7002, + "step": 53578 + }, + { + "epoch": 1.96, + "learning_rate": 3.973340169187388e-09, + "loss": 0.6918, + "step": 53579 + }, + { + "epoch": 1.96, + "learning_rate": 3.969678104515325e-09, + "loss": 0.6888, + "step": 53580 + }, + { + "epoch": 1.96, + "learning_rate": 3.966016039843263e-09, + "loss": 0.6903, + "step": 53581 + }, + { + "epoch": 1.96, + "learning_rate": 3.962353975171202e-09, + "loss": 0.6941, + "step": 53582 + }, + { + "epoch": 1.96, + "learning_rate": 3.958691910499139e-09, + "loss": 0.7004, + "step": 53583 + }, + { + "epoch": 1.96, + "learning_rate": 3.955029845827077e-09, + "loss": 0.6907, + "step": 53584 + }, + { + "epoch": 1.96, + "learning_rate": 3.951367781155015e-09, + "loss": 0.6818, + "step": 53585 + }, + { + "epoch": 1.96, + "learning_rate": 3.947705716482952e-09, + "loss": 0.6885, + "step": 53586 + }, + { + "epoch": 1.96, + "learning_rate": 3.944043651810891e-09, + "loss": 0.6887, + "step": 53587 + }, + { + "epoch": 1.96, + "learning_rate": 3.940381587138829e-09, + "loss": 0.7039, + "step": 53588 + }, + { + "epoch": 1.96, + "learning_rate": 3.936719522466766e-09, + "loss": 0.7077, + "step": 53589 + }, + { + "epoch": 1.96, + "learning_rate": 3.9330574577947046e-09, + "loss": 0.6903, + "step": 53590 + }, + { + "epoch": 1.96, + "learning_rate": 3.929395393122643e-09, + "loss": 0.7122, + "step": 53591 + }, + { + "epoch": 1.96, + "learning_rate": 3.92573332845058e-09, + "loss": 0.6898, + "step": 53592 + }, + { + "epoch": 1.96, + "learning_rate": 3.9220712637785186e-09, + "loss": 0.699, + "step": 53593 + }, + { + "epoch": 1.96, + "learning_rate": 3.918409199106456e-09, + "loss": 0.6932, + "step": 53594 + }, + { + "epoch": 1.96, + "learning_rate": 3.9147471344343934e-09, + "loss": 0.6932, + "step": 53595 + }, + { + "epoch": 1.96, + "learning_rate": 3.911085069762332e-09, + "loss": 0.6948, + "step": 53596 + }, + { + "epoch": 1.96, + "learning_rate": 3.90742300509027e-09, + "loss": 0.6957, + "step": 53597 + }, + { + "epoch": 1.96, + "learning_rate": 3.9037609404182074e-09, + "loss": 0.6952, + "step": 53598 + }, + { + "epoch": 1.96, + "learning_rate": 3.900098875746146e-09, + "loss": 0.6941, + "step": 53599 + }, + { + "epoch": 1.96, + "learning_rate": 3.896436811074083e-09, + "loss": 0.6958, + "step": 53600 + }, + { + "epoch": 1.96, + "learning_rate": 3.8927747464020214e-09, + "loss": 0.6964, + "step": 53601 + }, + { + "epoch": 1.96, + "learning_rate": 3.889112681729959e-09, + "loss": 0.6916, + "step": 53602 + }, + { + "epoch": 1.96, + "learning_rate": 3.885450617057897e-09, + "loss": 0.695, + "step": 53603 + }, + { + "epoch": 1.96, + "learning_rate": 3.881788552385835e-09, + "loss": 0.6828, + "step": 53604 + }, + { + "epoch": 1.96, + "learning_rate": 3.878126487713773e-09, + "loss": 0.6951, + "step": 53605 + }, + { + "epoch": 1.96, + "learning_rate": 3.874464423041711e-09, + "loss": 0.6967, + "step": 53606 + }, + { + "epoch": 1.96, + "learning_rate": 3.870802358369649e-09, + "loss": 0.6965, + "step": 53607 + }, + { + "epoch": 1.96, + "learning_rate": 3.867140293697586e-09, + "loss": 0.703, + "step": 53608 + }, + { + "epoch": 1.96, + "learning_rate": 3.863478229025524e-09, + "loss": 0.6989, + "step": 53609 + }, + { + "epoch": 1.96, + "learning_rate": 3.859816164353463e-09, + "loss": 0.7, + "step": 53610 + }, + { + "epoch": 1.96, + "learning_rate": 3.8561540996814e-09, + "loss": 0.6891, + "step": 53611 + }, + { + "epoch": 1.96, + "learning_rate": 3.852492035009338e-09, + "loss": 0.6836, + "step": 53612 + }, + { + "epoch": 1.96, + "learning_rate": 3.848829970337276e-09, + "loss": 0.6985, + "step": 53613 + }, + { + "epoch": 1.96, + "learning_rate": 3.845167905665214e-09, + "loss": 0.6899, + "step": 53614 + }, + { + "epoch": 1.96, + "learning_rate": 3.841505840993152e-09, + "loss": 0.6978, + "step": 53615 + }, + { + "epoch": 1.96, + "learning_rate": 3.83784377632109e-09, + "loss": 0.6846, + "step": 53616 + }, + { + "epoch": 1.96, + "learning_rate": 3.834181711649027e-09, + "loss": 0.7076, + "step": 53617 + }, + { + "epoch": 1.96, + "learning_rate": 3.8305196469769655e-09, + "loss": 0.7015, + "step": 53618 + }, + { + "epoch": 1.96, + "learning_rate": 3.826857582304904e-09, + "loss": 0.6775, + "step": 53619 + }, + { + "epoch": 1.96, + "learning_rate": 3.823195517632841e-09, + "loss": 0.6824, + "step": 53620 + }, + { + "epoch": 1.96, + "learning_rate": 3.8195334529607795e-09, + "loss": 0.6911, + "step": 53621 + }, + { + "epoch": 1.96, + "learning_rate": 3.815871388288717e-09, + "loss": 0.6888, + "step": 53622 + }, + { + "epoch": 1.96, + "learning_rate": 3.812209323616654e-09, + "loss": 0.6905, + "step": 53623 + }, + { + "epoch": 1.96, + "learning_rate": 3.808547258944593e-09, + "loss": 0.6974, + "step": 53624 + }, + { + "epoch": 1.96, + "learning_rate": 3.804885194272531e-09, + "loss": 0.6913, + "step": 53625 + }, + { + "epoch": 1.96, + "learning_rate": 3.801223129600468e-09, + "loss": 0.6832, + "step": 53626 + }, + { + "epoch": 1.96, + "learning_rate": 3.797561064928406e-09, + "loss": 0.6967, + "step": 53627 + }, + { + "epoch": 1.96, + "learning_rate": 3.793899000256345e-09, + "loss": 0.6883, + "step": 53628 + }, + { + "epoch": 1.96, + "learning_rate": 3.790236935584282e-09, + "loss": 0.6838, + "step": 53629 + }, + { + "epoch": 1.96, + "learning_rate": 3.78657487091222e-09, + "loss": 0.6961, + "step": 53630 + }, + { + "epoch": 1.96, + "learning_rate": 3.782912806240158e-09, + "loss": 0.7016, + "step": 53631 + }, + { + "epoch": 1.96, + "learning_rate": 3.7792507415680955e-09, + "loss": 0.695, + "step": 53632 + }, + { + "epoch": 1.96, + "learning_rate": 3.775588676896034e-09, + "loss": 0.699, + "step": 53633 + }, + { + "epoch": 1.96, + "learning_rate": 3.771926612223972e-09, + "loss": 0.6932, + "step": 53634 + }, + { + "epoch": 1.96, + "learning_rate": 3.7682645475519095e-09, + "loss": 0.6827, + "step": 53635 + }, + { + "epoch": 1.96, + "learning_rate": 3.764602482879847e-09, + "loss": 0.7023, + "step": 53636 + }, + { + "epoch": 1.96, + "learning_rate": 3.760940418207786e-09, + "loss": 0.6968, + "step": 53637 + }, + { + "epoch": 1.96, + "learning_rate": 3.7572783535357235e-09, + "loss": 0.6965, + "step": 53638 + }, + { + "epoch": 1.96, + "learning_rate": 3.753616288863661e-09, + "loss": 0.6978, + "step": 53639 + }, + { + "epoch": 1.96, + "learning_rate": 3.749954224191599e-09, + "loss": 0.6941, + "step": 53640 + }, + { + "epoch": 1.96, + "learning_rate": 3.746292159519537e-09, + "loss": 0.6934, + "step": 53641 + }, + { + "epoch": 1.96, + "learning_rate": 3.742630094847475e-09, + "loss": 0.6935, + "step": 53642 + }, + { + "epoch": 1.96, + "learning_rate": 3.738968030175413e-09, + "loss": 0.6951, + "step": 53643 + }, + { + "epoch": 1.96, + "learning_rate": 3.735305965503351e-09, + "loss": 0.7007, + "step": 53644 + }, + { + "epoch": 1.96, + "learning_rate": 3.731643900831288e-09, + "loss": 0.7052, + "step": 53645 + }, + { + "epoch": 1.96, + "learning_rate": 3.727981836159226e-09, + "loss": 0.6904, + "step": 53646 + }, + { + "epoch": 1.96, + "learning_rate": 3.7243197714871647e-09, + "loss": 0.6989, + "step": 53647 + }, + { + "epoch": 1.96, + "learning_rate": 3.720657706815102e-09, + "loss": 0.6893, + "step": 53648 + }, + { + "epoch": 1.96, + "learning_rate": 3.71699564214304e-09, + "loss": 0.6897, + "step": 53649 + }, + { + "epoch": 1.96, + "learning_rate": 3.713333577470978e-09, + "loss": 0.6907, + "step": 53650 + }, + { + "epoch": 1.96, + "learning_rate": 3.709671512798916e-09, + "loss": 0.6961, + "step": 53651 + }, + { + "epoch": 1.96, + "learning_rate": 3.706009448126854e-09, + "loss": 0.6977, + "step": 53652 + }, + { + "epoch": 1.96, + "learning_rate": 3.7023473834547914e-09, + "loss": 0.6912, + "step": 53653 + }, + { + "epoch": 1.96, + "learning_rate": 3.6986853187827293e-09, + "loss": 0.6942, + "step": 53654 + }, + { + "epoch": 1.96, + "learning_rate": 3.695023254110667e-09, + "loss": 0.6901, + "step": 53655 + }, + { + "epoch": 1.96, + "learning_rate": 3.6913611894386054e-09, + "loss": 0.694, + "step": 53656 + }, + { + "epoch": 1.96, + "learning_rate": 3.6876991247665433e-09, + "loss": 0.7087, + "step": 53657 + }, + { + "epoch": 1.96, + "learning_rate": 3.684037060094481e-09, + "loss": 0.6848, + "step": 53658 + }, + { + "epoch": 1.96, + "learning_rate": 3.6803749954224186e-09, + "loss": 0.6865, + "step": 53659 + }, + { + "epoch": 1.96, + "learning_rate": 3.6767129307503573e-09, + "loss": 0.6983, + "step": 53660 + }, + { + "epoch": 1.96, + "learning_rate": 3.673050866078295e-09, + "loss": 0.6903, + "step": 53661 + }, + { + "epoch": 1.96, + "learning_rate": 3.6693888014062326e-09, + "loss": 0.6903, + "step": 53662 + }, + { + "epoch": 1.96, + "learning_rate": 3.6657267367341705e-09, + "loss": 0.6983, + "step": 53663 + }, + { + "epoch": 1.96, + "learning_rate": 3.6620646720621083e-09, + "loss": 0.6869, + "step": 53664 + }, + { + "epoch": 1.96, + "learning_rate": 3.6584026073900466e-09, + "loss": 0.6992, + "step": 53665 + }, + { + "epoch": 1.96, + "learning_rate": 3.6547405427179844e-09, + "loss": 0.6947, + "step": 53666 + }, + { + "epoch": 1.96, + "learning_rate": 3.651078478045922e-09, + "loss": 0.7006, + "step": 53667 + }, + { + "epoch": 1.96, + "learning_rate": 3.6474164133738598e-09, + "loss": 0.7043, + "step": 53668 + }, + { + "epoch": 1.96, + "learning_rate": 3.6437543487017976e-09, + "loss": 0.6949, + "step": 53669 + }, + { + "epoch": 1.96, + "learning_rate": 3.640092284029736e-09, + "loss": 0.6913, + "step": 53670 + }, + { + "epoch": 1.96, + "learning_rate": 3.6364302193576738e-09, + "loss": 0.6926, + "step": 53671 + }, + { + "epoch": 1.96, + "learning_rate": 3.6327681546856116e-09, + "loss": 0.6916, + "step": 53672 + }, + { + "epoch": 1.96, + "learning_rate": 3.629106090013549e-09, + "loss": 0.6878, + "step": 53673 + }, + { + "epoch": 1.96, + "learning_rate": 3.6254440253414877e-09, + "loss": 0.6963, + "step": 53674 + }, + { + "epoch": 1.96, + "learning_rate": 3.621781960669425e-09, + "loss": 0.6934, + "step": 53675 + }, + { + "epoch": 1.96, + "learning_rate": 3.618119895997363e-09, + "loss": 0.684, + "step": 53676 + }, + { + "epoch": 1.96, + "learning_rate": 3.614457831325301e-09, + "loss": 0.6918, + "step": 53677 + }, + { + "epoch": 1.96, + "learning_rate": 3.6107957666532388e-09, + "loss": 0.6986, + "step": 53678 + }, + { + "epoch": 1.96, + "learning_rate": 3.607133701981177e-09, + "loss": 0.6937, + "step": 53679 + }, + { + "epoch": 1.96, + "learning_rate": 3.603471637309115e-09, + "loss": 0.7009, + "step": 53680 + }, + { + "epoch": 1.96, + "learning_rate": 3.5998095726370524e-09, + "loss": 0.6989, + "step": 53681 + }, + { + "epoch": 1.96, + "learning_rate": 3.5961475079649902e-09, + "loss": 0.692, + "step": 53682 + }, + { + "epoch": 1.96, + "learning_rate": 3.592485443292929e-09, + "loss": 0.6772, + "step": 53683 + }, + { + "epoch": 1.96, + "learning_rate": 3.5888233786208664e-09, + "loss": 0.6948, + "step": 53684 + }, + { + "epoch": 1.96, + "learning_rate": 3.5851613139488042e-09, + "loss": 0.691, + "step": 53685 + }, + { + "epoch": 1.96, + "learning_rate": 3.581499249276742e-09, + "loss": 0.7012, + "step": 53686 + }, + { + "epoch": 1.96, + "learning_rate": 3.5778371846046795e-09, + "loss": 0.7082, + "step": 53687 + }, + { + "epoch": 1.96, + "learning_rate": 3.5741751199326182e-09, + "loss": 0.6959, + "step": 53688 + }, + { + "epoch": 1.96, + "learning_rate": 3.5705130552605557e-09, + "loss": 0.6809, + "step": 53689 + }, + { + "epoch": 1.96, + "learning_rate": 3.5668509905884935e-09, + "loss": 0.6883, + "step": 53690 + }, + { + "epoch": 1.96, + "learning_rate": 3.5631889259164314e-09, + "loss": 0.6856, + "step": 53691 + }, + { + "epoch": 1.96, + "learning_rate": 3.5595268612443692e-09, + "loss": 0.6997, + "step": 53692 + }, + { + "epoch": 1.96, + "learning_rate": 3.5558647965723075e-09, + "loss": 0.6969, + "step": 53693 + }, + { + "epoch": 1.96, + "learning_rate": 3.5522027319002454e-09, + "loss": 0.6885, + "step": 53694 + }, + { + "epoch": 1.96, + "learning_rate": 3.548540667228183e-09, + "loss": 0.6912, + "step": 53695 + }, + { + "epoch": 1.96, + "learning_rate": 3.5448786025561207e-09, + "loss": 0.6876, + "step": 53696 + }, + { + "epoch": 1.96, + "learning_rate": 3.541216537884059e-09, + "loss": 0.6942, + "step": 53697 + }, + { + "epoch": 1.96, + "learning_rate": 3.537554473211997e-09, + "loss": 0.6868, + "step": 53698 + }, + { + "epoch": 1.96, + "learning_rate": 3.5338924085399347e-09, + "loss": 0.6877, + "step": 53699 + }, + { + "epoch": 1.96, + "learning_rate": 3.5302303438678725e-09, + "loss": 0.696, + "step": 53700 + }, + { + "epoch": 1.96, + "learning_rate": 3.52656827919581e-09, + "loss": 0.7055, + "step": 53701 + }, + { + "epoch": 1.96, + "learning_rate": 3.5229062145237487e-09, + "loss": 0.6798, + "step": 53702 + }, + { + "epoch": 1.96, + "learning_rate": 3.519244149851686e-09, + "loss": 0.6894, + "step": 53703 + }, + { + "epoch": 1.96, + "learning_rate": 3.515582085179624e-09, + "loss": 0.6966, + "step": 53704 + }, + { + "epoch": 1.96, + "learning_rate": 3.511920020507562e-09, + "loss": 0.6939, + "step": 53705 + }, + { + "epoch": 1.96, + "learning_rate": 3.5082579558355e-09, + "loss": 0.6992, + "step": 53706 + }, + { + "epoch": 1.96, + "learning_rate": 3.504595891163438e-09, + "loss": 0.6942, + "step": 53707 + }, + { + "epoch": 1.97, + "learning_rate": 3.500933826491376e-09, + "loss": 0.6911, + "step": 53708 + }, + { + "epoch": 1.97, + "learning_rate": 3.4972717618193133e-09, + "loss": 0.6858, + "step": 53709 + }, + { + "epoch": 1.97, + "learning_rate": 3.493609697147251e-09, + "loss": 0.6978, + "step": 53710 + }, + { + "epoch": 1.97, + "learning_rate": 3.4899476324751894e-09, + "loss": 0.7016, + "step": 53711 + }, + { + "epoch": 1.97, + "learning_rate": 3.4862855678031273e-09, + "loss": 0.694, + "step": 53712 + }, + { + "epoch": 1.97, + "learning_rate": 3.482623503131065e-09, + "loss": 0.7017, + "step": 53713 + }, + { + "epoch": 1.97, + "learning_rate": 3.478961438459003e-09, + "loss": 0.6905, + "step": 53714 + }, + { + "epoch": 1.97, + "learning_rate": 3.4752993737869404e-09, + "loss": 0.6982, + "step": 53715 + }, + { + "epoch": 1.97, + "learning_rate": 3.471637309114879e-09, + "loss": 0.701, + "step": 53716 + }, + { + "epoch": 1.97, + "learning_rate": 3.4679752444428166e-09, + "loss": 0.6949, + "step": 53717 + }, + { + "epoch": 1.97, + "learning_rate": 3.4643131797707544e-09, + "loss": 0.7001, + "step": 53718 + }, + { + "epoch": 1.97, + "learning_rate": 3.4606511150986923e-09, + "loss": 0.6892, + "step": 53719 + }, + { + "epoch": 1.97, + "learning_rate": 3.4569890504266306e-09, + "loss": 0.6865, + "step": 53720 + }, + { + "epoch": 1.97, + "learning_rate": 3.4533269857545684e-09, + "loss": 0.6987, + "step": 53721 + }, + { + "epoch": 1.97, + "learning_rate": 3.4496649210825063e-09, + "loss": 0.6905, + "step": 53722 + }, + { + "epoch": 1.97, + "learning_rate": 3.4460028564104437e-09, + "loss": 0.6944, + "step": 53723 + }, + { + "epoch": 1.97, + "learning_rate": 3.4423407917383816e-09, + "loss": 0.6822, + "step": 53724 + }, + { + "epoch": 1.97, + "learning_rate": 3.43867872706632e-09, + "loss": 0.6937, + "step": 53725 + }, + { + "epoch": 1.97, + "learning_rate": 3.4350166623942577e-09, + "loss": 0.697, + "step": 53726 + }, + { + "epoch": 1.97, + "learning_rate": 3.4313545977221956e-09, + "loss": 0.6752, + "step": 53727 + }, + { + "epoch": 1.97, + "learning_rate": 3.4276925330501335e-09, + "loss": 0.6874, + "step": 53728 + }, + { + "epoch": 1.97, + "learning_rate": 3.4240304683780717e-09, + "loss": 0.6957, + "step": 53729 + }, + { + "epoch": 1.97, + "learning_rate": 3.4203684037060096e-09, + "loss": 0.6805, + "step": 53730 + }, + { + "epoch": 1.97, + "learning_rate": 3.416706339033947e-09, + "loss": 0.6987, + "step": 53731 + }, + { + "epoch": 1.97, + "learning_rate": 3.413044274361885e-09, + "loss": 0.6919, + "step": 53732 + }, + { + "epoch": 1.97, + "learning_rate": 3.4093822096898228e-09, + "loss": 0.6874, + "step": 53733 + }, + { + "epoch": 1.97, + "learning_rate": 3.405720145017761e-09, + "loss": 0.6982, + "step": 53734 + }, + { + "epoch": 1.97, + "learning_rate": 3.402058080345699e-09, + "loss": 0.6926, + "step": 53735 + }, + { + "epoch": 1.97, + "learning_rate": 3.3983960156736368e-09, + "loss": 0.7002, + "step": 53736 + }, + { + "epoch": 1.97, + "learning_rate": 3.3947339510015742e-09, + "loss": 0.6873, + "step": 53737 + }, + { + "epoch": 1.97, + "learning_rate": 3.391071886329513e-09, + "loss": 0.6983, + "step": 53738 + }, + { + "epoch": 1.97, + "learning_rate": 3.3874098216574503e-09, + "loss": 0.7004, + "step": 53739 + }, + { + "epoch": 1.97, + "learning_rate": 3.383747756985388e-09, + "loss": 0.6911, + "step": 53740 + }, + { + "epoch": 1.97, + "learning_rate": 3.380085692313326e-09, + "loss": 0.6866, + "step": 53741 + }, + { + "epoch": 1.97, + "learning_rate": 3.376423627641264e-09, + "loss": 0.6809, + "step": 53742 + }, + { + "epoch": 1.97, + "learning_rate": 3.372761562969202e-09, + "loss": 0.6935, + "step": 53743 + }, + { + "epoch": 1.97, + "learning_rate": 3.36909949829714e-09, + "loss": 0.6899, + "step": 53744 + }, + { + "epoch": 1.97, + "learning_rate": 3.3654374336250775e-09, + "loss": 0.693, + "step": 53745 + }, + { + "epoch": 1.97, + "learning_rate": 3.3617753689530154e-09, + "loss": 0.6881, + "step": 53746 + }, + { + "epoch": 1.97, + "learning_rate": 3.3581133042809532e-09, + "loss": 0.7055, + "step": 53747 + }, + { + "epoch": 1.97, + "learning_rate": 3.3544512396088915e-09, + "loss": 0.6945, + "step": 53748 + }, + { + "epoch": 1.97, + "learning_rate": 3.3507891749368294e-09, + "loss": 0.6967, + "step": 53749 + }, + { + "epoch": 1.97, + "learning_rate": 3.3471271102647672e-09, + "loss": 0.6749, + "step": 53750 + }, + { + "epoch": 1.97, + "learning_rate": 3.3434650455927047e-09, + "loss": 0.7043, + "step": 53751 + }, + { + "epoch": 1.97, + "learning_rate": 3.3398029809206434e-09, + "loss": 0.6913, + "step": 53752 + }, + { + "epoch": 1.97, + "learning_rate": 3.336140916248581e-09, + "loss": 0.6854, + "step": 53753 + }, + { + "epoch": 1.97, + "learning_rate": 3.3324788515765187e-09, + "loss": 0.6811, + "step": 53754 + }, + { + "epoch": 1.97, + "learning_rate": 3.3288167869044565e-09, + "loss": 0.693, + "step": 53755 + }, + { + "epoch": 1.97, + "learning_rate": 3.325154722232394e-09, + "loss": 0.6832, + "step": 53756 + }, + { + "epoch": 1.97, + "learning_rate": 3.3214926575603327e-09, + "loss": 0.7089, + "step": 53757 + }, + { + "epoch": 1.97, + "learning_rate": 3.3178305928882705e-09, + "loss": 0.6916, + "step": 53758 + }, + { + "epoch": 1.97, + "learning_rate": 3.314168528216208e-09, + "loss": 0.6919, + "step": 53759 + }, + { + "epoch": 1.97, + "learning_rate": 3.310506463544146e-09, + "loss": 0.6987, + "step": 53760 + }, + { + "epoch": 1.97, + "learning_rate": 3.306844398872084e-09, + "loss": 0.6951, + "step": 53761 + }, + { + "epoch": 1.97, + "learning_rate": 3.303182334200022e-09, + "loss": 0.6945, + "step": 53762 + }, + { + "epoch": 1.97, + "learning_rate": 3.29952026952796e-09, + "loss": 0.6926, + "step": 53763 + }, + { + "epoch": 1.97, + "learning_rate": 3.2958582048558977e-09, + "loss": 0.6938, + "step": 53764 + }, + { + "epoch": 1.97, + "learning_rate": 3.292196140183835e-09, + "loss": 0.6904, + "step": 53765 + }, + { + "epoch": 1.97, + "learning_rate": 3.288534075511774e-09, + "loss": 0.7046, + "step": 53766 + }, + { + "epoch": 1.97, + "learning_rate": 3.2848720108397113e-09, + "loss": 0.6948, + "step": 53767 + }, + { + "epoch": 1.97, + "learning_rate": 3.281209946167649e-09, + "loss": 0.6934, + "step": 53768 + }, + { + "epoch": 1.97, + "learning_rate": 3.277547881495587e-09, + "loss": 0.7001, + "step": 53769 + }, + { + "epoch": 1.97, + "learning_rate": 3.2738858168235244e-09, + "loss": 0.7011, + "step": 53770 + }, + { + "epoch": 1.97, + "learning_rate": 3.270223752151463e-09, + "loss": 0.6932, + "step": 53771 + }, + { + "epoch": 1.97, + "learning_rate": 3.266561687479401e-09, + "loss": 0.6777, + "step": 53772 + }, + { + "epoch": 1.97, + "learning_rate": 3.2628996228073384e-09, + "loss": 0.6879, + "step": 53773 + }, + { + "epoch": 1.97, + "learning_rate": 3.2592375581352763e-09, + "loss": 0.6908, + "step": 53774 + }, + { + "epoch": 1.97, + "learning_rate": 3.2555754934632146e-09, + "loss": 0.6861, + "step": 53775 + }, + { + "epoch": 1.97, + "learning_rate": 3.2519134287911524e-09, + "loss": 0.7023, + "step": 53776 + }, + { + "epoch": 1.97, + "learning_rate": 3.2482513641190903e-09, + "loss": 0.7039, + "step": 53777 + }, + { + "epoch": 1.97, + "learning_rate": 3.2445892994470277e-09, + "loss": 0.6856, + "step": 53778 + }, + { + "epoch": 1.97, + "learning_rate": 3.2409272347749656e-09, + "loss": 0.7011, + "step": 53779 + }, + { + "epoch": 1.97, + "learning_rate": 3.2372651701029043e-09, + "loss": 0.6907, + "step": 53780 + }, + { + "epoch": 1.97, + "learning_rate": 3.2336031054308417e-09, + "loss": 0.6804, + "step": 53781 + }, + { + "epoch": 1.97, + "learning_rate": 3.2299410407587796e-09, + "loss": 0.6985, + "step": 53782 + }, + { + "epoch": 1.97, + "learning_rate": 3.2262789760867175e-09, + "loss": 0.6922, + "step": 53783 + }, + { + "epoch": 1.97, + "learning_rate": 3.2226169114146557e-09, + "loss": 0.6952, + "step": 53784 + }, + { + "epoch": 1.97, + "learning_rate": 3.2189548467425936e-09, + "loss": 0.6859, + "step": 53785 + }, + { + "epoch": 1.97, + "learning_rate": 3.2152927820705315e-09, + "loss": 0.7066, + "step": 53786 + }, + { + "epoch": 1.97, + "learning_rate": 3.211630717398469e-09, + "loss": 0.6887, + "step": 53787 + }, + { + "epoch": 1.97, + "learning_rate": 3.2079686527264068e-09, + "loss": 0.6731, + "step": 53788 + }, + { + "epoch": 1.97, + "learning_rate": 3.204306588054345e-09, + "loss": 0.6864, + "step": 53789 + }, + { + "epoch": 1.97, + "learning_rate": 3.200644523382283e-09, + "loss": 0.6904, + "step": 53790 + }, + { + "epoch": 1.97, + "learning_rate": 3.1969824587102208e-09, + "loss": 0.6849, + "step": 53791 + }, + { + "epoch": 1.97, + "learning_rate": 3.193320394038158e-09, + "loss": 0.6895, + "step": 53792 + }, + { + "epoch": 1.97, + "learning_rate": 3.189658329366096e-09, + "loss": 0.6972, + "step": 53793 + }, + { + "epoch": 1.97, + "learning_rate": 3.1859962646940348e-09, + "loss": 0.691, + "step": 53794 + }, + { + "epoch": 1.97, + "learning_rate": 3.182334200021972e-09, + "loss": 0.6831, + "step": 53795 + }, + { + "epoch": 1.97, + "learning_rate": 3.17867213534991e-09, + "loss": 0.6975, + "step": 53796 + }, + { + "epoch": 1.97, + "learning_rate": 3.175010070677848e-09, + "loss": 0.692, + "step": 53797 + }, + { + "epoch": 1.97, + "learning_rate": 3.171348006005786e-09, + "loss": 0.6874, + "step": 53798 + }, + { + "epoch": 1.97, + "learning_rate": 3.167685941333724e-09, + "loss": 0.6751, + "step": 53799 + }, + { + "epoch": 1.97, + "learning_rate": 3.1640238766616615e-09, + "loss": 0.6888, + "step": 53800 + }, + { + "epoch": 1.97, + "learning_rate": 3.1603618119895994e-09, + "loss": 0.6775, + "step": 53801 + }, + { + "epoch": 1.97, + "learning_rate": 3.1566997473175372e-09, + "loss": 0.6975, + "step": 53802 + }, + { + "epoch": 1.97, + "learning_rate": 3.1530376826454755e-09, + "loss": 0.6849, + "step": 53803 + }, + { + "epoch": 1.97, + "learning_rate": 3.1493756179734134e-09, + "loss": 0.6982, + "step": 53804 + }, + { + "epoch": 1.97, + "learning_rate": 3.1457135533013512e-09, + "loss": 0.6915, + "step": 53805 + }, + { + "epoch": 1.97, + "learning_rate": 3.1420514886292887e-09, + "loss": 0.6925, + "step": 53806 + }, + { + "epoch": 1.97, + "learning_rate": 3.1383894239572274e-09, + "loss": 0.6894, + "step": 53807 + }, + { + "epoch": 1.97, + "learning_rate": 3.1347273592851652e-09, + "loss": 0.693, + "step": 53808 + }, + { + "epoch": 1.97, + "learning_rate": 3.1310652946131027e-09, + "loss": 0.6907, + "step": 53809 + }, + { + "epoch": 1.97, + "learning_rate": 3.1274032299410405e-09, + "loss": 0.6817, + "step": 53810 + }, + { + "epoch": 1.97, + "learning_rate": 3.1237411652689784e-09, + "loss": 0.6864, + "step": 53811 + }, + { + "epoch": 1.97, + "learning_rate": 3.1200791005969162e-09, + "loss": 0.6946, + "step": 53812 + }, + { + "epoch": 1.97, + "learning_rate": 3.1164170359248545e-09, + "loss": 0.6971, + "step": 53813 + }, + { + "epoch": 1.97, + "learning_rate": 3.112754971252792e-09, + "loss": 0.6925, + "step": 53814 + }, + { + "epoch": 1.97, + "learning_rate": 3.1090929065807302e-09, + "loss": 0.6906, + "step": 53815 + }, + { + "epoch": 1.97, + "learning_rate": 3.105430841908668e-09, + "loss": 0.6886, + "step": 53816 + }, + { + "epoch": 1.97, + "learning_rate": 3.1017687772366055e-09, + "loss": 0.6998, + "step": 53817 + }, + { + "epoch": 1.97, + "learning_rate": 3.098106712564544e-09, + "loss": 0.6944, + "step": 53818 + }, + { + "epoch": 1.97, + "learning_rate": 3.0944446478924817e-09, + "loss": 0.6941, + "step": 53819 + }, + { + "epoch": 1.97, + "learning_rate": 3.0907825832204195e-09, + "loss": 0.6996, + "step": 53820 + }, + { + "epoch": 1.97, + "learning_rate": 3.0871205185483574e-09, + "loss": 0.6716, + "step": 53821 + }, + { + "epoch": 1.97, + "learning_rate": 3.0834584538762953e-09, + "loss": 0.7003, + "step": 53822 + }, + { + "epoch": 1.97, + "learning_rate": 3.079796389204233e-09, + "loss": 0.6909, + "step": 53823 + }, + { + "epoch": 1.97, + "learning_rate": 3.076134324532171e-09, + "loss": 0.6956, + "step": 53824 + }, + { + "epoch": 1.97, + "learning_rate": 3.072472259860109e-09, + "loss": 0.6944, + "step": 53825 + }, + { + "epoch": 1.97, + "learning_rate": 3.0688101951880467e-09, + "loss": 0.6881, + "step": 53826 + }, + { + "epoch": 1.97, + "learning_rate": 3.065148130515985e-09, + "loss": 0.6904, + "step": 53827 + }, + { + "epoch": 1.97, + "learning_rate": 3.0614860658439224e-09, + "loss": 0.6977, + "step": 53828 + }, + { + "epoch": 1.97, + "learning_rate": 3.0578240011718607e-09, + "loss": 0.6892, + "step": 53829 + }, + { + "epoch": 1.97, + "learning_rate": 3.0541619364997986e-09, + "loss": 0.6913, + "step": 53830 + }, + { + "epoch": 1.97, + "learning_rate": 3.050499871827736e-09, + "loss": 0.6885, + "step": 53831 + }, + { + "epoch": 1.97, + "learning_rate": 3.0468378071556743e-09, + "loss": 0.6901, + "step": 53832 + }, + { + "epoch": 1.97, + "learning_rate": 3.043175742483612e-09, + "loss": 0.6906, + "step": 53833 + }, + { + "epoch": 1.97, + "learning_rate": 3.03951367781155e-09, + "loss": 0.6999, + "step": 53834 + }, + { + "epoch": 1.97, + "learning_rate": 3.035851613139488e-09, + "loss": 0.6941, + "step": 53835 + }, + { + "epoch": 1.97, + "learning_rate": 3.0321895484674257e-09, + "loss": 0.69, + "step": 53836 + }, + { + "epoch": 1.97, + "learning_rate": 3.0285274837953636e-09, + "loss": 0.6837, + "step": 53837 + }, + { + "epoch": 1.97, + "learning_rate": 3.024865419123302e-09, + "loss": 0.6985, + "step": 53838 + }, + { + "epoch": 1.97, + "learning_rate": 3.0212033544512393e-09, + "loss": 0.6915, + "step": 53839 + }, + { + "epoch": 1.97, + "learning_rate": 3.017541289779177e-09, + "loss": 0.696, + "step": 53840 + }, + { + "epoch": 1.97, + "learning_rate": 3.0138792251071154e-09, + "loss": 0.7029, + "step": 53841 + }, + { + "epoch": 1.97, + "learning_rate": 3.010217160435053e-09, + "loss": 0.6886, + "step": 53842 + }, + { + "epoch": 1.97, + "learning_rate": 3.006555095762991e-09, + "loss": 0.6907, + "step": 53843 + }, + { + "epoch": 1.97, + "learning_rate": 3.002893031090929e-09, + "loss": 0.7038, + "step": 53844 + }, + { + "epoch": 1.97, + "learning_rate": 2.999230966418867e-09, + "loss": 0.6902, + "step": 53845 + }, + { + "epoch": 1.97, + "learning_rate": 2.9955689017468047e-09, + "loss": 0.6929, + "step": 53846 + }, + { + "epoch": 1.97, + "learning_rate": 2.9919068370747426e-09, + "loss": 0.6797, + "step": 53847 + }, + { + "epoch": 1.97, + "learning_rate": 2.9882447724026805e-09, + "loss": 0.6863, + "step": 53848 + }, + { + "epoch": 1.97, + "learning_rate": 2.9845827077306183e-09, + "loss": 0.7007, + "step": 53849 + }, + { + "epoch": 1.97, + "learning_rate": 2.980920643058556e-09, + "loss": 0.6983, + "step": 53850 + }, + { + "epoch": 1.97, + "learning_rate": 2.977258578386494e-09, + "loss": 0.717, + "step": 53851 + }, + { + "epoch": 1.97, + "learning_rate": 2.9735965137144323e-09, + "loss": 0.6993, + "step": 53852 + }, + { + "epoch": 1.97, + "learning_rate": 2.9699344490423698e-09, + "loss": 0.6953, + "step": 53853 + }, + { + "epoch": 1.97, + "learning_rate": 2.9662723843703076e-09, + "loss": 0.6955, + "step": 53854 + }, + { + "epoch": 1.97, + "learning_rate": 2.962610319698246e-09, + "loss": 0.676, + "step": 53855 + }, + { + "epoch": 1.97, + "learning_rate": 2.9589482550261834e-09, + "loss": 0.6826, + "step": 53856 + }, + { + "epoch": 1.97, + "learning_rate": 2.9552861903541216e-09, + "loss": 0.6846, + "step": 53857 + }, + { + "epoch": 1.97, + "learning_rate": 2.9516241256820595e-09, + "loss": 0.6913, + "step": 53858 + }, + { + "epoch": 1.97, + "learning_rate": 2.9479620610099973e-09, + "loss": 0.6902, + "step": 53859 + }, + { + "epoch": 1.97, + "learning_rate": 2.9442999963379352e-09, + "loss": 0.6906, + "step": 53860 + }, + { + "epoch": 1.97, + "learning_rate": 2.940637931665873e-09, + "loss": 0.6934, + "step": 53861 + }, + { + "epoch": 1.97, + "learning_rate": 2.936975866993811e-09, + "loss": 0.6884, + "step": 53862 + }, + { + "epoch": 1.97, + "learning_rate": 2.9333138023217488e-09, + "loss": 0.6914, + "step": 53863 + }, + { + "epoch": 1.97, + "learning_rate": 2.9296517376496867e-09, + "loss": 0.7031, + "step": 53864 + }, + { + "epoch": 1.97, + "learning_rate": 2.9259896729776245e-09, + "loss": 0.6874, + "step": 53865 + }, + { + "epoch": 1.97, + "learning_rate": 2.9223276083055628e-09, + "loss": 0.6972, + "step": 53866 + }, + { + "epoch": 1.97, + "learning_rate": 2.9186655436335002e-09, + "loss": 0.7024, + "step": 53867 + }, + { + "epoch": 1.97, + "learning_rate": 2.9150034789614385e-09, + "loss": 0.6743, + "step": 53868 + }, + { + "epoch": 1.97, + "learning_rate": 2.9113414142893764e-09, + "loss": 0.6822, + "step": 53869 + }, + { + "epoch": 1.97, + "learning_rate": 2.907679349617314e-09, + "loss": 0.6946, + "step": 53870 + }, + { + "epoch": 1.97, + "learning_rate": 2.904017284945252e-09, + "loss": 0.7058, + "step": 53871 + }, + { + "epoch": 1.97, + "learning_rate": 2.90035522027319e-09, + "loss": 0.68, + "step": 53872 + }, + { + "epoch": 1.97, + "learning_rate": 2.896693155601128e-09, + "loss": 0.6856, + "step": 53873 + }, + { + "epoch": 1.97, + "learning_rate": 2.8930310909290657e-09, + "loss": 0.6919, + "step": 53874 + }, + { + "epoch": 1.97, + "learning_rate": 2.8893690262570035e-09, + "loss": 0.7075, + "step": 53875 + }, + { + "epoch": 1.97, + "learning_rate": 2.8857069615849414e-09, + "loss": 0.6893, + "step": 53876 + }, + { + "epoch": 1.97, + "learning_rate": 2.8820448969128793e-09, + "loss": 0.6882, + "step": 53877 + }, + { + "epoch": 1.97, + "learning_rate": 2.878382832240817e-09, + "loss": 0.7072, + "step": 53878 + }, + { + "epoch": 1.97, + "learning_rate": 2.874720767568755e-09, + "loss": 0.7075, + "step": 53879 + }, + { + "epoch": 1.97, + "learning_rate": 2.8710587028966932e-09, + "loss": 0.6983, + "step": 53880 + }, + { + "epoch": 1.97, + "learning_rate": 2.8673966382246307e-09, + "loss": 0.7003, + "step": 53881 + }, + { + "epoch": 1.97, + "learning_rate": 2.863734573552569e-09, + "loss": 0.6975, + "step": 53882 + }, + { + "epoch": 1.97, + "learning_rate": 2.860072508880507e-09, + "loss": 0.6919, + "step": 53883 + }, + { + "epoch": 1.97, + "learning_rate": 2.8564104442084447e-09, + "loss": 0.6947, + "step": 53884 + }, + { + "epoch": 1.97, + "learning_rate": 2.8527483795363826e-09, + "loss": 0.688, + "step": 53885 + }, + { + "epoch": 1.97, + "learning_rate": 2.8490863148643204e-09, + "loss": 0.6769, + "step": 53886 + }, + { + "epoch": 1.97, + "learning_rate": 2.8454242501922583e-09, + "loss": 0.6977, + "step": 53887 + }, + { + "epoch": 1.97, + "learning_rate": 2.841762185520196e-09, + "loss": 0.6897, + "step": 53888 + }, + { + "epoch": 1.97, + "learning_rate": 2.838100120848134e-09, + "loss": 0.7033, + "step": 53889 + }, + { + "epoch": 1.97, + "learning_rate": 2.834438056176072e-09, + "loss": 0.685, + "step": 53890 + }, + { + "epoch": 1.97, + "learning_rate": 2.83077599150401e-09, + "loss": 0.6833, + "step": 53891 + }, + { + "epoch": 1.97, + "learning_rate": 2.8271139268319476e-09, + "loss": 0.6827, + "step": 53892 + }, + { + "epoch": 1.97, + "learning_rate": 2.8234518621598854e-09, + "loss": 0.6877, + "step": 53893 + }, + { + "epoch": 1.97, + "learning_rate": 2.8197897974878237e-09, + "loss": 0.703, + "step": 53894 + }, + { + "epoch": 1.97, + "learning_rate": 2.816127732815761e-09, + "loss": 0.6965, + "step": 53895 + }, + { + "epoch": 1.97, + "learning_rate": 2.8124656681436994e-09, + "loss": 0.6978, + "step": 53896 + }, + { + "epoch": 1.97, + "learning_rate": 2.8088036034716373e-09, + "loss": 0.7147, + "step": 53897 + }, + { + "epoch": 1.97, + "learning_rate": 2.805141538799575e-09, + "loss": 0.6923, + "step": 53898 + }, + { + "epoch": 1.97, + "learning_rate": 2.801479474127513e-09, + "loss": 0.6821, + "step": 53899 + }, + { + "epoch": 1.97, + "learning_rate": 2.797817409455451e-09, + "loss": 0.6903, + "step": 53900 + }, + { + "epoch": 1.97, + "learning_rate": 2.7941553447833887e-09, + "loss": 0.6959, + "step": 53901 + }, + { + "epoch": 1.97, + "learning_rate": 2.7904932801113266e-09, + "loss": 0.6988, + "step": 53902 + }, + { + "epoch": 1.97, + "learning_rate": 2.7868312154392645e-09, + "loss": 0.6894, + "step": 53903 + }, + { + "epoch": 1.97, + "learning_rate": 2.7831691507672023e-09, + "loss": 0.6907, + "step": 53904 + }, + { + "epoch": 1.97, + "learning_rate": 2.7795070860951406e-09, + "loss": 0.6963, + "step": 53905 + }, + { + "epoch": 1.97, + "learning_rate": 2.775845021423078e-09, + "loss": 0.6909, + "step": 53906 + }, + { + "epoch": 1.97, + "learning_rate": 2.7721829567510163e-09, + "loss": 0.6861, + "step": 53907 + }, + { + "epoch": 1.97, + "learning_rate": 2.768520892078954e-09, + "loss": 0.7063, + "step": 53908 + }, + { + "epoch": 1.97, + "learning_rate": 2.7648588274068916e-09, + "loss": 0.6964, + "step": 53909 + }, + { + "epoch": 1.97, + "learning_rate": 2.76119676273483e-09, + "loss": 0.697, + "step": 53910 + }, + { + "epoch": 1.97, + "learning_rate": 2.7575346980627678e-09, + "loss": 0.6965, + "step": 53911 + }, + { + "epoch": 1.97, + "learning_rate": 2.7538726333907056e-09, + "loss": 0.687, + "step": 53912 + }, + { + "epoch": 1.97, + "learning_rate": 2.7502105687186435e-09, + "loss": 0.6823, + "step": 53913 + }, + { + "epoch": 1.97, + "learning_rate": 2.7465485040465813e-09, + "loss": 0.6858, + "step": 53914 + }, + { + "epoch": 1.97, + "learning_rate": 2.742886439374519e-09, + "loss": 0.6971, + "step": 53915 + }, + { + "epoch": 1.97, + "learning_rate": 2.739224374702457e-09, + "loss": 0.7007, + "step": 53916 + }, + { + "epoch": 1.97, + "learning_rate": 2.735562310030395e-09, + "loss": 0.6924, + "step": 53917 + }, + { + "epoch": 1.97, + "learning_rate": 2.7319002453583328e-09, + "loss": 0.6898, + "step": 53918 + }, + { + "epoch": 1.97, + "learning_rate": 2.728238180686271e-09, + "loss": 0.694, + "step": 53919 + }, + { + "epoch": 1.97, + "learning_rate": 2.7245761160142085e-09, + "loss": 0.6912, + "step": 53920 + }, + { + "epoch": 1.97, + "learning_rate": 2.7209140513421468e-09, + "loss": 0.6822, + "step": 53921 + }, + { + "epoch": 1.97, + "learning_rate": 2.7172519866700846e-09, + "loss": 0.684, + "step": 53922 + }, + { + "epoch": 1.97, + "learning_rate": 2.7135899219980225e-09, + "loss": 0.6841, + "step": 53923 + }, + { + "epoch": 1.97, + "learning_rate": 2.7099278573259604e-09, + "loss": 0.6945, + "step": 53924 + }, + { + "epoch": 1.97, + "learning_rate": 2.706265792653898e-09, + "loss": 0.697, + "step": 53925 + }, + { + "epoch": 1.97, + "learning_rate": 2.702603727981836e-09, + "loss": 0.6886, + "step": 53926 + }, + { + "epoch": 1.97, + "learning_rate": 2.698941663309774e-09, + "loss": 0.687, + "step": 53927 + }, + { + "epoch": 1.97, + "learning_rate": 2.695279598637712e-09, + "loss": 0.7006, + "step": 53928 + }, + { + "epoch": 1.97, + "learning_rate": 2.6916175339656497e-09, + "loss": 0.7021, + "step": 53929 + }, + { + "epoch": 1.97, + "learning_rate": 2.687955469293588e-09, + "loss": 0.687, + "step": 53930 + }, + { + "epoch": 1.97, + "learning_rate": 2.6842934046215254e-09, + "loss": 0.699, + "step": 53931 + }, + { + "epoch": 1.97, + "learning_rate": 2.6806313399494632e-09, + "loss": 0.7002, + "step": 53932 + }, + { + "epoch": 1.97, + "learning_rate": 2.6769692752774015e-09, + "loss": 0.696, + "step": 53933 + }, + { + "epoch": 1.97, + "learning_rate": 2.673307210605339e-09, + "loss": 0.6918, + "step": 53934 + }, + { + "epoch": 1.97, + "learning_rate": 2.6696451459332772e-09, + "loss": 0.7008, + "step": 53935 + }, + { + "epoch": 1.97, + "learning_rate": 2.6659830812612147e-09, + "loss": 0.6968, + "step": 53936 + }, + { + "epoch": 1.97, + "learning_rate": 2.662321016589153e-09, + "loss": 0.6953, + "step": 53937 + }, + { + "epoch": 1.97, + "learning_rate": 2.658658951917091e-09, + "loss": 0.7018, + "step": 53938 + }, + { + "epoch": 1.97, + "learning_rate": 2.6549968872450283e-09, + "loss": 0.6869, + "step": 53939 + }, + { + "epoch": 1.97, + "learning_rate": 2.6513348225729665e-09, + "loss": 0.7008, + "step": 53940 + }, + { + "epoch": 1.97, + "learning_rate": 2.6476727579009044e-09, + "loss": 0.6979, + "step": 53941 + }, + { + "epoch": 1.97, + "learning_rate": 2.6440106932288423e-09, + "loss": 0.7233, + "step": 53942 + }, + { + "epoch": 1.97, + "learning_rate": 2.64034862855678e-09, + "loss": 0.6905, + "step": 53943 + }, + { + "epoch": 1.97, + "learning_rate": 2.6366865638847184e-09, + "loss": 0.7023, + "step": 53944 + }, + { + "epoch": 1.97, + "learning_rate": 2.633024499212656e-09, + "loss": 0.696, + "step": 53945 + }, + { + "epoch": 1.97, + "learning_rate": 2.629362434540594e-09, + "loss": 0.7, + "step": 53946 + }, + { + "epoch": 1.97, + "learning_rate": 2.6257003698685316e-09, + "loss": 0.683, + "step": 53947 + }, + { + "epoch": 1.97, + "learning_rate": 2.6220383051964694e-09, + "loss": 0.6854, + "step": 53948 + }, + { + "epoch": 1.97, + "learning_rate": 2.6183762405244077e-09, + "loss": 0.6966, + "step": 53949 + }, + { + "epoch": 1.97, + "learning_rate": 2.614714175852345e-09, + "loss": 0.6941, + "step": 53950 + }, + { + "epoch": 1.97, + "learning_rate": 2.6110521111802834e-09, + "loss": 0.6944, + "step": 53951 + }, + { + "epoch": 1.97, + "learning_rate": 2.6073900465082213e-09, + "loss": 0.6745, + "step": 53952 + }, + { + "epoch": 1.97, + "learning_rate": 2.603727981836159e-09, + "loss": 0.6933, + "step": 53953 + }, + { + "epoch": 1.97, + "learning_rate": 2.600065917164097e-09, + "loss": 0.7113, + "step": 53954 + }, + { + "epoch": 1.97, + "learning_rate": 2.596403852492035e-09, + "loss": 0.6804, + "step": 53955 + }, + { + "epoch": 1.97, + "learning_rate": 2.5927417878199727e-09, + "loss": 0.6938, + "step": 53956 + }, + { + "epoch": 1.97, + "learning_rate": 2.5890797231479106e-09, + "loss": 0.6991, + "step": 53957 + }, + { + "epoch": 1.97, + "learning_rate": 2.5854176584758484e-09, + "loss": 0.7014, + "step": 53958 + }, + { + "epoch": 1.97, + "learning_rate": 2.5817555938037863e-09, + "loss": 0.6969, + "step": 53959 + }, + { + "epoch": 1.97, + "learning_rate": 2.5780935291317246e-09, + "loss": 0.703, + "step": 53960 + }, + { + "epoch": 1.97, + "learning_rate": 2.574431464459662e-09, + "loss": 0.7037, + "step": 53961 + }, + { + "epoch": 1.97, + "learning_rate": 2.5707693997876e-09, + "loss": 0.6971, + "step": 53962 + }, + { + "epoch": 1.97, + "learning_rate": 2.567107335115538e-09, + "loss": 0.6854, + "step": 53963 + }, + { + "epoch": 1.97, + "learning_rate": 2.5634452704434756e-09, + "loss": 0.6942, + "step": 53964 + }, + { + "epoch": 1.97, + "learning_rate": 2.559783205771414e-09, + "loss": 0.6955, + "step": 53965 + }, + { + "epoch": 1.97, + "learning_rate": 2.5561211410993517e-09, + "loss": 0.6871, + "step": 53966 + }, + { + "epoch": 1.97, + "learning_rate": 2.5524590764272896e-09, + "loss": 0.7011, + "step": 53967 + }, + { + "epoch": 1.97, + "learning_rate": 2.5487970117552275e-09, + "loss": 0.7033, + "step": 53968 + }, + { + "epoch": 1.97, + "learning_rate": 2.5451349470831653e-09, + "loss": 0.7086, + "step": 53969 + }, + { + "epoch": 1.97, + "learning_rate": 2.541472882411103e-09, + "loss": 0.6915, + "step": 53970 + }, + { + "epoch": 1.97, + "learning_rate": 2.537810817739041e-09, + "loss": 0.6947, + "step": 53971 + }, + { + "epoch": 1.97, + "learning_rate": 2.534148753066979e-09, + "loss": 0.6962, + "step": 53972 + }, + { + "epoch": 1.97, + "learning_rate": 2.5304866883949168e-09, + "loss": 0.7012, + "step": 53973 + }, + { + "epoch": 1.97, + "learning_rate": 2.526824623722855e-09, + "loss": 0.6894, + "step": 53974 + }, + { + "epoch": 1.97, + "learning_rate": 2.5231625590507925e-09, + "loss": 0.6821, + "step": 53975 + }, + { + "epoch": 1.97, + "learning_rate": 2.5195004943787308e-09, + "loss": 0.6898, + "step": 53976 + }, + { + "epoch": 1.97, + "learning_rate": 2.5158384297066686e-09, + "loss": 0.6951, + "step": 53977 + }, + { + "epoch": 1.97, + "learning_rate": 2.512176365034606e-09, + "loss": 0.6827, + "step": 53978 + }, + { + "epoch": 1.97, + "learning_rate": 2.5085143003625444e-09, + "loss": 0.6919, + "step": 53979 + }, + { + "epoch": 1.97, + "learning_rate": 2.5048522356904822e-09, + "loss": 0.6995, + "step": 53980 + }, + { + "epoch": 1.98, + "learning_rate": 2.50119017101842e-09, + "loss": 0.695, + "step": 53981 + }, + { + "epoch": 1.98, + "learning_rate": 2.497528106346358e-09, + "loss": 0.7036, + "step": 53982 + }, + { + "epoch": 1.98, + "learning_rate": 2.4938660416742958e-09, + "loss": 0.698, + "step": 53983 + }, + { + "epoch": 1.98, + "learning_rate": 2.4902039770022337e-09, + "loss": 0.6879, + "step": 53984 + }, + { + "epoch": 1.98, + "learning_rate": 2.4865419123301715e-09, + "loss": 0.7094, + "step": 53985 + }, + { + "epoch": 1.98, + "learning_rate": 2.4828798476581094e-09, + "loss": 0.6912, + "step": 53986 + }, + { + "epoch": 1.98, + "learning_rate": 2.4792177829860472e-09, + "loss": 0.688, + "step": 53987 + }, + { + "epoch": 1.98, + "learning_rate": 2.4755557183139855e-09, + "loss": 0.6914, + "step": 53988 + }, + { + "epoch": 1.98, + "learning_rate": 2.471893653641923e-09, + "loss": 0.6891, + "step": 53989 + }, + { + "epoch": 1.98, + "learning_rate": 2.4682315889698612e-09, + "loss": 0.6866, + "step": 53990 + }, + { + "epoch": 1.98, + "learning_rate": 2.464569524297799e-09, + "loss": 0.6987, + "step": 53991 + }, + { + "epoch": 1.98, + "learning_rate": 2.460907459625737e-09, + "loss": 0.6921, + "step": 53992 + }, + { + "epoch": 1.98, + "learning_rate": 2.457245394953675e-09, + "loss": 0.6908, + "step": 53993 + }, + { + "epoch": 1.98, + "learning_rate": 2.4535833302816127e-09, + "loss": 0.6816, + "step": 53994 + }, + { + "epoch": 1.98, + "learning_rate": 2.4499212656095505e-09, + "loss": 0.6817, + "step": 53995 + }, + { + "epoch": 1.98, + "learning_rate": 2.4462592009374884e-09, + "loss": 0.6928, + "step": 53996 + }, + { + "epoch": 1.98, + "learning_rate": 2.4425971362654263e-09, + "loss": 0.6886, + "step": 53997 + }, + { + "epoch": 1.98, + "learning_rate": 2.438935071593364e-09, + "loss": 0.685, + "step": 53998 + }, + { + "epoch": 1.98, + "learning_rate": 2.4352730069213024e-09, + "loss": 0.6891, + "step": 53999 + }, + { + "epoch": 1.98, + "learning_rate": 2.43161094224924e-09, + "loss": 0.6971, + "step": 54000 + }, + { + "epoch": 1.98, + "learning_rate": 2.4279488775771777e-09, + "loss": 0.6938, + "step": 54001 + }, + { + "epoch": 1.98, + "learning_rate": 2.424286812905116e-09, + "loss": 0.6921, + "step": 54002 + }, + { + "epoch": 1.98, + "learning_rate": 2.4206247482330534e-09, + "loss": 0.6997, + "step": 54003 + }, + { + "epoch": 1.98, + "learning_rate": 2.4169626835609917e-09, + "loss": 0.7032, + "step": 54004 + }, + { + "epoch": 1.98, + "learning_rate": 2.4133006188889296e-09, + "loss": 0.6787, + "step": 54005 + }, + { + "epoch": 1.98, + "learning_rate": 2.4096385542168674e-09, + "loss": 0.6971, + "step": 54006 + }, + { + "epoch": 1.98, + "learning_rate": 2.4059764895448053e-09, + "loss": 0.6938, + "step": 54007 + }, + { + "epoch": 1.98, + "learning_rate": 2.402314424872743e-09, + "loss": 0.7033, + "step": 54008 + }, + { + "epoch": 1.98, + "learning_rate": 2.398652360200681e-09, + "loss": 0.702, + "step": 54009 + }, + { + "epoch": 1.98, + "learning_rate": 2.394990295528619e-09, + "loss": 0.6991, + "step": 54010 + }, + { + "epoch": 1.98, + "learning_rate": 2.3913282308565567e-09, + "loss": 0.7013, + "step": 54011 + }, + { + "epoch": 1.98, + "learning_rate": 2.3876661661844946e-09, + "loss": 0.6896, + "step": 54012 + }, + { + "epoch": 1.98, + "learning_rate": 2.384004101512433e-09, + "loss": 0.6861, + "step": 54013 + }, + { + "epoch": 1.98, + "learning_rate": 2.3803420368403703e-09, + "loss": 0.6986, + "step": 54014 + }, + { + "epoch": 1.98, + "learning_rate": 2.3766799721683086e-09, + "loss": 0.6911, + "step": 54015 + }, + { + "epoch": 1.98, + "learning_rate": 2.3730179074962464e-09, + "loss": 0.6948, + "step": 54016 + }, + { + "epoch": 1.98, + "learning_rate": 2.369355842824184e-09, + "loss": 0.6916, + "step": 54017 + }, + { + "epoch": 1.98, + "learning_rate": 2.365693778152122e-09, + "loss": 0.6824, + "step": 54018 + }, + { + "epoch": 1.98, + "learning_rate": 2.36203171348006e-09, + "loss": 0.6986, + "step": 54019 + }, + { + "epoch": 1.98, + "learning_rate": 2.358369648807998e-09, + "loss": 0.6975, + "step": 54020 + }, + { + "epoch": 1.98, + "learning_rate": 2.3547075841359357e-09, + "loss": 0.6955, + "step": 54021 + }, + { + "epoch": 1.98, + "learning_rate": 2.3510455194638736e-09, + "loss": 0.6957, + "step": 54022 + }, + { + "epoch": 1.98, + "learning_rate": 2.3473834547918115e-09, + "loss": 0.6919, + "step": 54023 + }, + { + "epoch": 1.98, + "learning_rate": 2.3437213901197493e-09, + "loss": 0.6972, + "step": 54024 + }, + { + "epoch": 1.98, + "learning_rate": 2.340059325447687e-09, + "loss": 0.6974, + "step": 54025 + }, + { + "epoch": 1.98, + "learning_rate": 2.336397260775625e-09, + "loss": 0.6869, + "step": 54026 + }, + { + "epoch": 1.98, + "learning_rate": 2.3327351961035633e-09, + "loss": 0.6903, + "step": 54027 + }, + { + "epoch": 1.98, + "learning_rate": 2.3290731314315008e-09, + "loss": 0.692, + "step": 54028 + }, + { + "epoch": 1.98, + "learning_rate": 2.325411066759439e-09, + "loss": 0.7001, + "step": 54029 + }, + { + "epoch": 1.98, + "learning_rate": 2.321749002087377e-09, + "loss": 0.6875, + "step": 54030 + }, + { + "epoch": 1.98, + "learning_rate": 2.3180869374153148e-09, + "loss": 0.6938, + "step": 54031 + }, + { + "epoch": 1.98, + "learning_rate": 2.3144248727432526e-09, + "loss": 0.6992, + "step": 54032 + }, + { + "epoch": 1.98, + "learning_rate": 2.3107628080711905e-09, + "loss": 0.6753, + "step": 54033 + }, + { + "epoch": 1.98, + "learning_rate": 2.3071007433991283e-09, + "loss": 0.6966, + "step": 54034 + }, + { + "epoch": 1.98, + "learning_rate": 2.303438678727066e-09, + "loss": 0.7017, + "step": 54035 + }, + { + "epoch": 1.98, + "learning_rate": 2.299776614055004e-09, + "loss": 0.7125, + "step": 54036 + }, + { + "epoch": 1.98, + "learning_rate": 2.296114549382942e-09, + "loss": 0.6946, + "step": 54037 + }, + { + "epoch": 1.98, + "learning_rate": 2.29245248471088e-09, + "loss": 0.701, + "step": 54038 + }, + { + "epoch": 1.98, + "learning_rate": 2.2887904200388176e-09, + "loss": 0.6936, + "step": 54039 + }, + { + "epoch": 1.98, + "learning_rate": 2.2851283553667555e-09, + "loss": 0.6918, + "step": 54040 + }, + { + "epoch": 1.98, + "learning_rate": 2.2814662906946938e-09, + "loss": 0.6994, + "step": 54041 + }, + { + "epoch": 1.98, + "learning_rate": 2.2778042260226312e-09, + "loss": 0.6975, + "step": 54042 + }, + { + "epoch": 1.98, + "learning_rate": 2.2741421613505695e-09, + "loss": 0.6952, + "step": 54043 + }, + { + "epoch": 1.98, + "learning_rate": 2.2704800966785074e-09, + "loss": 0.6876, + "step": 54044 + }, + { + "epoch": 1.98, + "learning_rate": 2.2668180320064452e-09, + "loss": 0.7012, + "step": 54045 + }, + { + "epoch": 1.98, + "learning_rate": 2.263155967334383e-09, + "loss": 0.6955, + "step": 54046 + }, + { + "epoch": 1.98, + "learning_rate": 2.259493902662321e-09, + "loss": 0.6949, + "step": 54047 + }, + { + "epoch": 1.98, + "learning_rate": 2.255831837990259e-09, + "loss": 0.6985, + "step": 54048 + }, + { + "epoch": 1.98, + "learning_rate": 2.2521697733181967e-09, + "loss": 0.6928, + "step": 54049 + }, + { + "epoch": 1.98, + "learning_rate": 2.2485077086461345e-09, + "loss": 0.6936, + "step": 54050 + }, + { + "epoch": 1.98, + "learning_rate": 2.2448456439740724e-09, + "loss": 0.685, + "step": 54051 + }, + { + "epoch": 1.98, + "learning_rate": 2.2411835793020107e-09, + "loss": 0.6902, + "step": 54052 + }, + { + "epoch": 1.98, + "learning_rate": 2.237521514629948e-09, + "loss": 0.6781, + "step": 54053 + }, + { + "epoch": 1.98, + "learning_rate": 2.2338594499578864e-09, + "loss": 0.6912, + "step": 54054 + }, + { + "epoch": 1.98, + "learning_rate": 2.2301973852858242e-09, + "loss": 0.6985, + "step": 54055 + }, + { + "epoch": 1.98, + "learning_rate": 2.2265353206137617e-09, + "loss": 0.6962, + "step": 54056 + }, + { + "epoch": 1.98, + "learning_rate": 2.2228732559417e-09, + "loss": 0.6899, + "step": 54057 + }, + { + "epoch": 1.98, + "learning_rate": 2.219211191269638e-09, + "loss": 0.6795, + "step": 54058 + }, + { + "epoch": 1.98, + "learning_rate": 2.2155491265975757e-09, + "loss": 0.7065, + "step": 54059 + }, + { + "epoch": 1.98, + "learning_rate": 2.2118870619255135e-09, + "loss": 0.689, + "step": 54060 + }, + { + "epoch": 1.98, + "learning_rate": 2.2082249972534514e-09, + "loss": 0.6859, + "step": 54061 + }, + { + "epoch": 1.98, + "learning_rate": 2.2045629325813893e-09, + "loss": 0.6896, + "step": 54062 + }, + { + "epoch": 1.98, + "learning_rate": 2.200900867909327e-09, + "loss": 0.6966, + "step": 54063 + }, + { + "epoch": 1.98, + "learning_rate": 2.197238803237265e-09, + "loss": 0.6907, + "step": 54064 + }, + { + "epoch": 1.98, + "learning_rate": 2.193576738565203e-09, + "loss": 0.692, + "step": 54065 + }, + { + "epoch": 1.98, + "learning_rate": 2.189914673893141e-09, + "loss": 0.6959, + "step": 54066 + }, + { + "epoch": 1.98, + "learning_rate": 2.1862526092210786e-09, + "loss": 0.6959, + "step": 54067 + }, + { + "epoch": 1.98, + "learning_rate": 2.182590544549017e-09, + "loss": 0.6786, + "step": 54068 + }, + { + "epoch": 1.98, + "learning_rate": 2.1789284798769547e-09, + "loss": 0.6866, + "step": 54069 + }, + { + "epoch": 1.98, + "learning_rate": 2.175266415204892e-09, + "loss": 0.6965, + "step": 54070 + }, + { + "epoch": 1.98, + "learning_rate": 2.1716043505328304e-09, + "loss": 0.6958, + "step": 54071 + }, + { + "epoch": 1.98, + "learning_rate": 2.167942285860768e-09, + "loss": 0.6704, + "step": 54072 + }, + { + "epoch": 1.98, + "learning_rate": 2.164280221188706e-09, + "loss": 0.7002, + "step": 54073 + }, + { + "epoch": 1.98, + "learning_rate": 2.160618156516644e-09, + "loss": 0.7014, + "step": 54074 + }, + { + "epoch": 1.98, + "learning_rate": 2.156956091844582e-09, + "loss": 0.6933, + "step": 54075 + }, + { + "epoch": 1.98, + "learning_rate": 2.1532940271725197e-09, + "loss": 0.6932, + "step": 54076 + }, + { + "epoch": 1.98, + "learning_rate": 2.149631962500458e-09, + "loss": 0.6885, + "step": 54077 + }, + { + "epoch": 1.98, + "learning_rate": 2.1459698978283955e-09, + "loss": 0.6905, + "step": 54078 + }, + { + "epoch": 1.98, + "learning_rate": 2.1423078331563333e-09, + "loss": 0.7027, + "step": 54079 + }, + { + "epoch": 1.98, + "learning_rate": 2.1386457684842716e-09, + "loss": 0.6932, + "step": 54080 + }, + { + "epoch": 1.98, + "learning_rate": 2.134983703812209e-09, + "loss": 0.6802, + "step": 54081 + }, + { + "epoch": 1.98, + "learning_rate": 2.1313216391401473e-09, + "loss": 0.6906, + "step": 54082 + }, + { + "epoch": 1.98, + "learning_rate": 2.1276595744680848e-09, + "loss": 0.6981, + "step": 54083 + }, + { + "epoch": 1.98, + "learning_rate": 2.123997509796023e-09, + "loss": 0.6952, + "step": 54084 + }, + { + "epoch": 1.98, + "learning_rate": 2.120335445123961e-09, + "loss": 0.6873, + "step": 54085 + }, + { + "epoch": 1.98, + "learning_rate": 2.1166733804518983e-09, + "loss": 0.6965, + "step": 54086 + }, + { + "epoch": 1.98, + "learning_rate": 2.1130113157798366e-09, + "loss": 0.6957, + "step": 54087 + }, + { + "epoch": 1.98, + "learning_rate": 2.1093492511077745e-09, + "loss": 0.6991, + "step": 54088 + }, + { + "epoch": 1.98, + "learning_rate": 2.1056871864357123e-09, + "loss": 0.6801, + "step": 54089 + }, + { + "epoch": 1.98, + "learning_rate": 2.10202512176365e-09, + "loss": 0.6913, + "step": 54090 + }, + { + "epoch": 1.98, + "learning_rate": 2.0983630570915885e-09, + "loss": 0.6959, + "step": 54091 + }, + { + "epoch": 1.98, + "learning_rate": 2.094700992419526e-09, + "loss": 0.6948, + "step": 54092 + }, + { + "epoch": 1.98, + "learning_rate": 2.091038927747464e-09, + "loss": 0.6903, + "step": 54093 + }, + { + "epoch": 1.98, + "learning_rate": 2.0873768630754016e-09, + "loss": 0.7, + "step": 54094 + }, + { + "epoch": 1.98, + "learning_rate": 2.0837147984033395e-09, + "loss": 0.7084, + "step": 54095 + }, + { + "epoch": 1.98, + "learning_rate": 2.0800527337312778e-09, + "loss": 0.689, + "step": 54096 + }, + { + "epoch": 1.98, + "learning_rate": 2.0763906690592152e-09, + "loss": 0.6891, + "step": 54097 + }, + { + "epoch": 1.98, + "learning_rate": 2.0727286043871535e-09, + "loss": 0.6873, + "step": 54098 + }, + { + "epoch": 1.98, + "learning_rate": 2.0690665397150914e-09, + "loss": 0.6996, + "step": 54099 + }, + { + "epoch": 1.98, + "learning_rate": 2.0654044750430292e-09, + "loss": 0.7067, + "step": 54100 + }, + { + "epoch": 1.98, + "learning_rate": 2.061742410370967e-09, + "loss": 0.6942, + "step": 54101 + }, + { + "epoch": 1.98, + "learning_rate": 2.058080345698905e-09, + "loss": 0.6863, + "step": 54102 + }, + { + "epoch": 1.98, + "learning_rate": 2.054418281026843e-09, + "loss": 0.6912, + "step": 54103 + }, + { + "epoch": 1.98, + "learning_rate": 2.0507562163547807e-09, + "loss": 0.6906, + "step": 54104 + }, + { + "epoch": 1.98, + "learning_rate": 2.0470941516827185e-09, + "loss": 0.6879, + "step": 54105 + }, + { + "epoch": 1.98, + "learning_rate": 2.0434320870106564e-09, + "loss": 0.6944, + "step": 54106 + }, + { + "epoch": 1.98, + "learning_rate": 2.0397700223385947e-09, + "loss": 0.7054, + "step": 54107 + }, + { + "epoch": 1.98, + "learning_rate": 2.036107957666532e-09, + "loss": 0.6929, + "step": 54108 + }, + { + "epoch": 1.98, + "learning_rate": 2.03244589299447e-09, + "loss": 0.6878, + "step": 54109 + }, + { + "epoch": 1.98, + "learning_rate": 2.0287838283224082e-09, + "loss": 0.7017, + "step": 54110 + }, + { + "epoch": 1.98, + "learning_rate": 2.0251217636503457e-09, + "loss": 0.6994, + "step": 54111 + }, + { + "epoch": 1.98, + "learning_rate": 2.021459698978284e-09, + "loss": 0.6938, + "step": 54112 + }, + { + "epoch": 1.98, + "learning_rate": 2.017797634306222e-09, + "loss": 0.6943, + "step": 54113 + }, + { + "epoch": 1.98, + "learning_rate": 2.0141355696341597e-09, + "loss": 0.6806, + "step": 54114 + }, + { + "epoch": 1.98, + "learning_rate": 2.0104735049620975e-09, + "loss": 0.6956, + "step": 54115 + }, + { + "epoch": 1.98, + "learning_rate": 2.0068114402900354e-09, + "loss": 0.7028, + "step": 54116 + }, + { + "epoch": 1.98, + "learning_rate": 2.0031493756179733e-09, + "loss": 0.6893, + "step": 54117 + }, + { + "epoch": 1.98, + "learning_rate": 1.999487310945911e-09, + "loss": 0.7006, + "step": 54118 + }, + { + "epoch": 1.98, + "learning_rate": 1.995825246273849e-09, + "loss": 0.7047, + "step": 54119 + }, + { + "epoch": 1.98, + "learning_rate": 1.992163181601787e-09, + "loss": 0.7001, + "step": 54120 + }, + { + "epoch": 1.98, + "learning_rate": 1.988501116929725e-09, + "loss": 0.691, + "step": 54121 + }, + { + "epoch": 1.98, + "learning_rate": 1.9848390522576626e-09, + "loss": 0.6746, + "step": 54122 + }, + { + "epoch": 1.98, + "learning_rate": 1.981176987585601e-09, + "loss": 0.6884, + "step": 54123 + }, + { + "epoch": 1.98, + "learning_rate": 1.9775149229135387e-09, + "loss": 0.6904, + "step": 54124 + }, + { + "epoch": 1.98, + "learning_rate": 1.973852858241476e-09, + "loss": 0.6839, + "step": 54125 + }, + { + "epoch": 1.98, + "learning_rate": 1.9701907935694144e-09, + "loss": 0.6817, + "step": 54126 + }, + { + "epoch": 1.98, + "learning_rate": 1.9665287288973523e-09, + "loss": 0.698, + "step": 54127 + }, + { + "epoch": 1.98, + "learning_rate": 1.96286666422529e-09, + "loss": 0.6855, + "step": 54128 + }, + { + "epoch": 1.98, + "learning_rate": 1.959204599553228e-09, + "loss": 0.6956, + "step": 54129 + }, + { + "epoch": 1.98, + "learning_rate": 1.955542534881166e-09, + "loss": 0.6885, + "step": 54130 + }, + { + "epoch": 1.98, + "learning_rate": 1.9518804702091037e-09, + "loss": 0.6936, + "step": 54131 + }, + { + "epoch": 1.98, + "learning_rate": 1.9482184055370416e-09, + "loss": 0.6936, + "step": 54132 + }, + { + "epoch": 1.98, + "learning_rate": 1.9445563408649794e-09, + "loss": 0.6913, + "step": 54133 + }, + { + "epoch": 1.98, + "learning_rate": 1.9408942761929173e-09, + "loss": 0.7013, + "step": 54134 + }, + { + "epoch": 1.98, + "learning_rate": 1.9372322115208556e-09, + "loss": 0.6954, + "step": 54135 + }, + { + "epoch": 1.98, + "learning_rate": 1.933570146848793e-09, + "loss": 0.692, + "step": 54136 + }, + { + "epoch": 1.98, + "learning_rate": 1.9299080821767313e-09, + "loss": 0.6931, + "step": 54137 + }, + { + "epoch": 1.98, + "learning_rate": 1.926246017504669e-09, + "loss": 0.6951, + "step": 54138 + }, + { + "epoch": 1.98, + "learning_rate": 1.922583952832607e-09, + "loss": 0.6741, + "step": 54139 + }, + { + "epoch": 1.98, + "learning_rate": 1.918921888160545e-09, + "loss": 0.6829, + "step": 54140 + }, + { + "epoch": 1.98, + "learning_rate": 1.9152598234884827e-09, + "loss": 0.6904, + "step": 54141 + }, + { + "epoch": 1.98, + "learning_rate": 1.9115977588164206e-09, + "loss": 0.6899, + "step": 54142 + }, + { + "epoch": 1.98, + "learning_rate": 1.9079356941443585e-09, + "loss": 0.6862, + "step": 54143 + }, + { + "epoch": 1.98, + "learning_rate": 1.9042736294722963e-09, + "loss": 0.68, + "step": 54144 + }, + { + "epoch": 1.98, + "learning_rate": 1.900611564800234e-09, + "loss": 0.6889, + "step": 54145 + }, + { + "epoch": 1.98, + "learning_rate": 1.8969495001281725e-09, + "loss": 0.6875, + "step": 54146 + }, + { + "epoch": 1.98, + "learning_rate": 1.89328743545611e-09, + "loss": 0.6901, + "step": 54147 + }, + { + "epoch": 1.98, + "learning_rate": 1.8896253707840478e-09, + "loss": 0.7045, + "step": 54148 + }, + { + "epoch": 1.98, + "learning_rate": 1.885963306111986e-09, + "loss": 0.701, + "step": 54149 + }, + { + "epoch": 1.98, + "learning_rate": 1.8823012414399235e-09, + "loss": 0.698, + "step": 54150 + }, + { + "epoch": 1.98, + "learning_rate": 1.8786391767678618e-09, + "loss": 0.6879, + "step": 54151 + }, + { + "epoch": 1.98, + "learning_rate": 1.8749771120957996e-09, + "loss": 0.6896, + "step": 54152 + }, + { + "epoch": 1.98, + "learning_rate": 1.8713150474237375e-09, + "loss": 0.6834, + "step": 54153 + }, + { + "epoch": 1.98, + "learning_rate": 1.8676529827516753e-09, + "loss": 0.6986, + "step": 54154 + }, + { + "epoch": 1.98, + "learning_rate": 1.863990918079613e-09, + "loss": 0.6931, + "step": 54155 + }, + { + "epoch": 1.98, + "learning_rate": 1.860328853407551e-09, + "loss": 0.7063, + "step": 54156 + }, + { + "epoch": 1.98, + "learning_rate": 1.856666788735489e-09, + "loss": 0.6853, + "step": 54157 + }, + { + "epoch": 1.98, + "learning_rate": 1.853004724063427e-09, + "loss": 0.6978, + "step": 54158 + }, + { + "epoch": 1.98, + "learning_rate": 1.8493426593913646e-09, + "loss": 0.7168, + "step": 54159 + }, + { + "epoch": 1.98, + "learning_rate": 1.8456805947193027e-09, + "loss": 0.6911, + "step": 54160 + }, + { + "epoch": 1.98, + "learning_rate": 1.8420185300472406e-09, + "loss": 0.697, + "step": 54161 + }, + { + "epoch": 1.98, + "learning_rate": 1.8383564653751786e-09, + "loss": 0.6954, + "step": 54162 + }, + { + "epoch": 1.98, + "learning_rate": 1.8346944007031163e-09, + "loss": 0.6933, + "step": 54163 + }, + { + "epoch": 1.98, + "learning_rate": 1.8310323360310542e-09, + "loss": 0.6971, + "step": 54164 + }, + { + "epoch": 1.98, + "learning_rate": 1.8273702713589922e-09, + "loss": 0.6847, + "step": 54165 + }, + { + "epoch": 1.98, + "learning_rate": 1.8237082066869299e-09, + "loss": 0.7001, + "step": 54166 + }, + { + "epoch": 1.98, + "learning_rate": 1.820046142014868e-09, + "loss": 0.7017, + "step": 54167 + }, + { + "epoch": 1.98, + "learning_rate": 1.8163840773428058e-09, + "loss": 0.6959, + "step": 54168 + }, + { + "epoch": 1.98, + "learning_rate": 1.8127220126707439e-09, + "loss": 0.683, + "step": 54169 + }, + { + "epoch": 1.98, + "learning_rate": 1.8090599479986815e-09, + "loss": 0.7048, + "step": 54170 + }, + { + "epoch": 1.98, + "learning_rate": 1.8053978833266194e-09, + "loss": 0.699, + "step": 54171 + }, + { + "epoch": 1.98, + "learning_rate": 1.8017358186545575e-09, + "loss": 0.6918, + "step": 54172 + }, + { + "epoch": 1.98, + "learning_rate": 1.7980737539824951e-09, + "loss": 0.6852, + "step": 54173 + }, + { + "epoch": 1.98, + "learning_rate": 1.7944116893104332e-09, + "loss": 0.6842, + "step": 54174 + }, + { + "epoch": 1.98, + "learning_rate": 1.790749624638371e-09, + "loss": 0.6903, + "step": 54175 + }, + { + "epoch": 1.98, + "learning_rate": 1.7870875599663091e-09, + "loss": 0.6872, + "step": 54176 + }, + { + "epoch": 1.98, + "learning_rate": 1.7834254952942468e-09, + "loss": 0.6979, + "step": 54177 + }, + { + "epoch": 1.98, + "learning_rate": 1.7797634306221846e-09, + "loss": 0.6934, + "step": 54178 + }, + { + "epoch": 1.98, + "learning_rate": 1.7761013659501227e-09, + "loss": 0.6874, + "step": 54179 + }, + { + "epoch": 1.98, + "learning_rate": 1.7724393012780603e-09, + "loss": 0.6946, + "step": 54180 + }, + { + "epoch": 1.98, + "learning_rate": 1.7687772366059984e-09, + "loss": 0.7043, + "step": 54181 + }, + { + "epoch": 1.98, + "learning_rate": 1.7651151719339363e-09, + "loss": 0.6925, + "step": 54182 + }, + { + "epoch": 1.98, + "learning_rate": 1.7614531072618743e-09, + "loss": 0.6955, + "step": 54183 + }, + { + "epoch": 1.98, + "learning_rate": 1.757791042589812e-09, + "loss": 0.6878, + "step": 54184 + }, + { + "epoch": 1.98, + "learning_rate": 1.75412897791775e-09, + "loss": 0.7118, + "step": 54185 + }, + { + "epoch": 1.98, + "learning_rate": 1.750466913245688e-09, + "loss": 0.6792, + "step": 54186 + }, + { + "epoch": 1.98, + "learning_rate": 1.7468048485736256e-09, + "loss": 0.705, + "step": 54187 + }, + { + "epoch": 1.98, + "learning_rate": 1.7431427839015636e-09, + "loss": 0.6891, + "step": 54188 + }, + { + "epoch": 1.98, + "learning_rate": 1.7394807192295015e-09, + "loss": 0.6844, + "step": 54189 + }, + { + "epoch": 1.98, + "learning_rate": 1.7358186545574396e-09, + "loss": 0.6922, + "step": 54190 + }, + { + "epoch": 1.98, + "learning_rate": 1.7321565898853772e-09, + "loss": 0.7049, + "step": 54191 + }, + { + "epoch": 1.98, + "learning_rate": 1.7284945252133153e-09, + "loss": 0.6834, + "step": 54192 + }, + { + "epoch": 1.98, + "learning_rate": 1.7248324605412532e-09, + "loss": 0.7023, + "step": 54193 + }, + { + "epoch": 1.98, + "learning_rate": 1.7211703958691908e-09, + "loss": 0.693, + "step": 54194 + }, + { + "epoch": 1.98, + "learning_rate": 1.7175083311971289e-09, + "loss": 0.6853, + "step": 54195 + }, + { + "epoch": 1.98, + "learning_rate": 1.7138462665250667e-09, + "loss": 0.6793, + "step": 54196 + }, + { + "epoch": 1.98, + "learning_rate": 1.7101842018530048e-09, + "loss": 0.6764, + "step": 54197 + }, + { + "epoch": 1.98, + "learning_rate": 1.7065221371809425e-09, + "loss": 0.6767, + "step": 54198 + }, + { + "epoch": 1.98, + "learning_rate": 1.7028600725088805e-09, + "loss": 0.705, + "step": 54199 + }, + { + "epoch": 1.98, + "learning_rate": 1.6991980078368184e-09, + "loss": 0.6877, + "step": 54200 + }, + { + "epoch": 1.98, + "learning_rate": 1.6955359431647564e-09, + "loss": 0.7053, + "step": 54201 + }, + { + "epoch": 1.98, + "learning_rate": 1.691873878492694e-09, + "loss": 0.6879, + "step": 54202 + }, + { + "epoch": 1.98, + "learning_rate": 1.688211813820632e-09, + "loss": 0.6915, + "step": 54203 + }, + { + "epoch": 1.98, + "learning_rate": 1.68454974914857e-09, + "loss": 0.6982, + "step": 54204 + }, + { + "epoch": 1.98, + "learning_rate": 1.6808876844765077e-09, + "loss": 0.6922, + "step": 54205 + }, + { + "epoch": 1.98, + "learning_rate": 1.6772256198044458e-09, + "loss": 0.6935, + "step": 54206 + }, + { + "epoch": 1.98, + "learning_rate": 1.6735635551323836e-09, + "loss": 0.6966, + "step": 54207 + }, + { + "epoch": 1.98, + "learning_rate": 1.6699014904603217e-09, + "loss": 0.6882, + "step": 54208 + }, + { + "epoch": 1.98, + "learning_rate": 1.6662394257882593e-09, + "loss": 0.7043, + "step": 54209 + }, + { + "epoch": 1.98, + "learning_rate": 1.662577361116197e-09, + "loss": 0.6845, + "step": 54210 + }, + { + "epoch": 1.98, + "learning_rate": 1.6589152964441353e-09, + "loss": 0.6968, + "step": 54211 + }, + { + "epoch": 1.98, + "learning_rate": 1.655253231772073e-09, + "loss": 0.6979, + "step": 54212 + }, + { + "epoch": 1.98, + "learning_rate": 1.651591167100011e-09, + "loss": 0.6918, + "step": 54213 + }, + { + "epoch": 1.98, + "learning_rate": 1.6479291024279488e-09, + "loss": 0.6833, + "step": 54214 + }, + { + "epoch": 1.98, + "learning_rate": 1.644267037755887e-09, + "loss": 0.6963, + "step": 54215 + }, + { + "epoch": 1.98, + "learning_rate": 1.6406049730838246e-09, + "loss": 0.6909, + "step": 54216 + }, + { + "epoch": 1.98, + "learning_rate": 1.6369429084117622e-09, + "loss": 0.6836, + "step": 54217 + }, + { + "epoch": 1.98, + "learning_rate": 1.6332808437397005e-09, + "loss": 0.707, + "step": 54218 + }, + { + "epoch": 1.98, + "learning_rate": 1.6296187790676381e-09, + "loss": 0.704, + "step": 54219 + }, + { + "epoch": 1.98, + "learning_rate": 1.6259567143955762e-09, + "loss": 0.713, + "step": 54220 + }, + { + "epoch": 1.98, + "learning_rate": 1.6222946497235139e-09, + "loss": 0.68, + "step": 54221 + }, + { + "epoch": 1.98, + "learning_rate": 1.6186325850514521e-09, + "loss": 0.717, + "step": 54222 + }, + { + "epoch": 1.98, + "learning_rate": 1.6149705203793898e-09, + "loss": 0.6934, + "step": 54223 + }, + { + "epoch": 1.98, + "learning_rate": 1.6113084557073279e-09, + "loss": 0.695, + "step": 54224 + }, + { + "epoch": 1.98, + "learning_rate": 1.6076463910352657e-09, + "loss": 0.6938, + "step": 54225 + }, + { + "epoch": 1.98, + "learning_rate": 1.6039843263632034e-09, + "loss": 0.6964, + "step": 54226 + }, + { + "epoch": 1.98, + "learning_rate": 1.6003222616911414e-09, + "loss": 0.6912, + "step": 54227 + }, + { + "epoch": 1.98, + "learning_rate": 1.596660197019079e-09, + "loss": 0.7069, + "step": 54228 + }, + { + "epoch": 1.98, + "learning_rate": 1.5929981323470174e-09, + "loss": 0.6906, + "step": 54229 + }, + { + "epoch": 1.98, + "learning_rate": 1.589336067674955e-09, + "loss": 0.6906, + "step": 54230 + }, + { + "epoch": 1.98, + "learning_rate": 1.585674003002893e-09, + "loss": 0.6944, + "step": 54231 + }, + { + "epoch": 1.98, + "learning_rate": 1.5820119383308308e-09, + "loss": 0.6911, + "step": 54232 + }, + { + "epoch": 1.98, + "learning_rate": 1.5783498736587686e-09, + "loss": 0.6938, + "step": 54233 + }, + { + "epoch": 1.98, + "learning_rate": 1.5746878089867067e-09, + "loss": 0.6919, + "step": 54234 + }, + { + "epoch": 1.98, + "learning_rate": 1.5710257443146443e-09, + "loss": 0.6919, + "step": 54235 + }, + { + "epoch": 1.98, + "learning_rate": 1.5673636796425826e-09, + "loss": 0.7017, + "step": 54236 + }, + { + "epoch": 1.98, + "learning_rate": 1.5637016149705203e-09, + "loss": 0.703, + "step": 54237 + }, + { + "epoch": 1.98, + "learning_rate": 1.5600395502984581e-09, + "loss": 0.6986, + "step": 54238 + }, + { + "epoch": 1.98, + "learning_rate": 1.556377485626396e-09, + "loss": 0.6977, + "step": 54239 + }, + { + "epoch": 1.98, + "learning_rate": 1.552715420954334e-09, + "loss": 0.6949, + "step": 54240 + }, + { + "epoch": 1.98, + "learning_rate": 1.549053356282272e-09, + "loss": 0.6933, + "step": 54241 + }, + { + "epoch": 1.98, + "learning_rate": 1.5453912916102098e-09, + "loss": 0.6856, + "step": 54242 + }, + { + "epoch": 1.98, + "learning_rate": 1.5417292269381476e-09, + "loss": 0.6938, + "step": 54243 + }, + { + "epoch": 1.98, + "learning_rate": 1.5380671622660855e-09, + "loss": 0.6923, + "step": 54244 + }, + { + "epoch": 1.98, + "learning_rate": 1.5344050975940234e-09, + "loss": 0.6967, + "step": 54245 + }, + { + "epoch": 1.98, + "learning_rate": 1.5307430329219612e-09, + "loss": 0.6921, + "step": 54246 + }, + { + "epoch": 1.98, + "learning_rate": 1.5270809682498993e-09, + "loss": 0.6969, + "step": 54247 + }, + { + "epoch": 1.98, + "learning_rate": 1.5234189035778371e-09, + "loss": 0.7059, + "step": 54248 + }, + { + "epoch": 1.98, + "learning_rate": 1.519756838905775e-09, + "loss": 0.6904, + "step": 54249 + }, + { + "epoch": 1.98, + "learning_rate": 1.5160947742337129e-09, + "loss": 0.697, + "step": 54250 + }, + { + "epoch": 1.98, + "learning_rate": 1.512432709561651e-09, + "loss": 0.6963, + "step": 54251 + }, + { + "epoch": 1.98, + "learning_rate": 1.5087706448895886e-09, + "loss": 0.6919, + "step": 54252 + }, + { + "epoch": 1.98, + "learning_rate": 1.5051085802175264e-09, + "loss": 0.6858, + "step": 54253 + }, + { + "epoch": 1.98, + "learning_rate": 1.5014465155454645e-09, + "loss": 0.698, + "step": 54254 + }, + { + "epoch": 1.99, + "learning_rate": 1.4977844508734024e-09, + "loss": 0.6751, + "step": 54255 + }, + { + "epoch": 1.99, + "learning_rate": 1.4941223862013402e-09, + "loss": 0.7114, + "step": 54256 + }, + { + "epoch": 1.99, + "learning_rate": 1.490460321529278e-09, + "loss": 0.7019, + "step": 54257 + }, + { + "epoch": 1.99, + "learning_rate": 1.4867982568572162e-09, + "loss": 0.6969, + "step": 54258 + }, + { + "epoch": 1.99, + "learning_rate": 1.4831361921851538e-09, + "loss": 0.7067, + "step": 54259 + }, + { + "epoch": 1.99, + "learning_rate": 1.4794741275130917e-09, + "loss": 0.6889, + "step": 54260 + }, + { + "epoch": 1.99, + "learning_rate": 1.4758120628410297e-09, + "loss": 0.7035, + "step": 54261 + }, + { + "epoch": 1.99, + "learning_rate": 1.4721499981689676e-09, + "loss": 0.6935, + "step": 54262 + }, + { + "epoch": 1.99, + "learning_rate": 1.4684879334969055e-09, + "loss": 0.6844, + "step": 54263 + }, + { + "epoch": 1.99, + "learning_rate": 1.4648258688248433e-09, + "loss": 0.6803, + "step": 54264 + }, + { + "epoch": 1.99, + "learning_rate": 1.4611638041527814e-09, + "loss": 0.6956, + "step": 54265 + }, + { + "epoch": 1.99, + "learning_rate": 1.4575017394807193e-09, + "loss": 0.6892, + "step": 54266 + }, + { + "epoch": 1.99, + "learning_rate": 1.453839674808657e-09, + "loss": 0.6918, + "step": 54267 + }, + { + "epoch": 1.99, + "learning_rate": 1.450177610136595e-09, + "loss": 0.7041, + "step": 54268 + }, + { + "epoch": 1.99, + "learning_rate": 1.4465155454645328e-09, + "loss": 0.689, + "step": 54269 + }, + { + "epoch": 1.99, + "learning_rate": 1.4428534807924707e-09, + "loss": 0.6914, + "step": 54270 + }, + { + "epoch": 1.99, + "learning_rate": 1.4391914161204086e-09, + "loss": 0.6919, + "step": 54271 + }, + { + "epoch": 1.99, + "learning_rate": 1.4355293514483466e-09, + "loss": 0.6914, + "step": 54272 + }, + { + "epoch": 1.99, + "learning_rate": 1.4318672867762845e-09, + "loss": 0.7045, + "step": 54273 + }, + { + "epoch": 1.99, + "learning_rate": 1.4282052221042223e-09, + "loss": 0.6967, + "step": 54274 + }, + { + "epoch": 1.99, + "learning_rate": 1.4245431574321602e-09, + "loss": 0.6887, + "step": 54275 + }, + { + "epoch": 1.99, + "learning_rate": 1.420881092760098e-09, + "loss": 0.7004, + "step": 54276 + }, + { + "epoch": 1.99, + "learning_rate": 1.417219028088036e-09, + "loss": 0.6957, + "step": 54277 + }, + { + "epoch": 1.99, + "learning_rate": 1.4135569634159738e-09, + "loss": 0.6892, + "step": 54278 + }, + { + "epoch": 1.99, + "learning_rate": 1.4098948987439119e-09, + "loss": 0.694, + "step": 54279 + }, + { + "epoch": 1.99, + "learning_rate": 1.4062328340718497e-09, + "loss": 0.679, + "step": 54280 + }, + { + "epoch": 1.99, + "learning_rate": 1.4025707693997876e-09, + "loss": 0.706, + "step": 54281 + }, + { + "epoch": 1.99, + "learning_rate": 1.3989087047277254e-09, + "loss": 0.703, + "step": 54282 + }, + { + "epoch": 1.99, + "learning_rate": 1.3952466400556633e-09, + "loss": 0.6952, + "step": 54283 + }, + { + "epoch": 1.99, + "learning_rate": 1.3915845753836012e-09, + "loss": 0.6948, + "step": 54284 + }, + { + "epoch": 1.99, + "learning_rate": 1.387922510711539e-09, + "loss": 0.6997, + "step": 54285 + }, + { + "epoch": 1.99, + "learning_rate": 1.384260446039477e-09, + "loss": 0.6901, + "step": 54286 + }, + { + "epoch": 1.99, + "learning_rate": 1.380598381367415e-09, + "loss": 0.6944, + "step": 54287 + }, + { + "epoch": 1.99, + "learning_rate": 1.3769363166953528e-09, + "loss": 0.686, + "step": 54288 + }, + { + "epoch": 1.99, + "learning_rate": 1.3732742520232907e-09, + "loss": 0.6914, + "step": 54289 + }, + { + "epoch": 1.99, + "learning_rate": 1.3696121873512285e-09, + "loss": 0.6925, + "step": 54290 + }, + { + "epoch": 1.99, + "learning_rate": 1.3659501226791664e-09, + "loss": 0.6925, + "step": 54291 + }, + { + "epoch": 1.99, + "learning_rate": 1.3622880580071043e-09, + "loss": 0.701, + "step": 54292 + }, + { + "epoch": 1.99, + "learning_rate": 1.3586259933350423e-09, + "loss": 0.6903, + "step": 54293 + }, + { + "epoch": 1.99, + "learning_rate": 1.3549639286629802e-09, + "loss": 0.6871, + "step": 54294 + }, + { + "epoch": 1.99, + "learning_rate": 1.351301863990918e-09, + "loss": 0.6963, + "step": 54295 + }, + { + "epoch": 1.99, + "learning_rate": 1.347639799318856e-09, + "loss": 0.6943, + "step": 54296 + }, + { + "epoch": 1.99, + "learning_rate": 1.343977734646794e-09, + "loss": 0.6851, + "step": 54297 + }, + { + "epoch": 1.99, + "learning_rate": 1.3403156699747316e-09, + "loss": 0.6982, + "step": 54298 + }, + { + "epoch": 1.99, + "learning_rate": 1.3366536053026695e-09, + "loss": 0.6963, + "step": 54299 + }, + { + "epoch": 1.99, + "learning_rate": 1.3329915406306073e-09, + "loss": 0.6989, + "step": 54300 + }, + { + "epoch": 1.99, + "learning_rate": 1.3293294759585454e-09, + "loss": 0.6985, + "step": 54301 + }, + { + "epoch": 1.99, + "learning_rate": 1.3256674112864833e-09, + "loss": 0.6968, + "step": 54302 + }, + { + "epoch": 1.99, + "learning_rate": 1.3220053466144211e-09, + "loss": 0.6943, + "step": 54303 + }, + { + "epoch": 1.99, + "learning_rate": 1.3183432819423592e-09, + "loss": 0.6978, + "step": 54304 + }, + { + "epoch": 1.99, + "learning_rate": 1.314681217270297e-09, + "loss": 0.6818, + "step": 54305 + }, + { + "epoch": 1.99, + "learning_rate": 1.3110191525982347e-09, + "loss": 0.6937, + "step": 54306 + }, + { + "epoch": 1.99, + "learning_rate": 1.3073570879261726e-09, + "loss": 0.6822, + "step": 54307 + }, + { + "epoch": 1.99, + "learning_rate": 1.3036950232541106e-09, + "loss": 0.6887, + "step": 54308 + }, + { + "epoch": 1.99, + "learning_rate": 1.3000329585820485e-09, + "loss": 0.7013, + "step": 54309 + }, + { + "epoch": 1.99, + "learning_rate": 1.2963708939099864e-09, + "loss": 0.7049, + "step": 54310 + }, + { + "epoch": 1.99, + "learning_rate": 1.2927088292379242e-09, + "loss": 0.6857, + "step": 54311 + }, + { + "epoch": 1.99, + "learning_rate": 1.2890467645658623e-09, + "loss": 0.6913, + "step": 54312 + }, + { + "epoch": 1.99, + "learning_rate": 1.2853846998938e-09, + "loss": 0.6854, + "step": 54313 + }, + { + "epoch": 1.99, + "learning_rate": 1.2817226352217378e-09, + "loss": 0.6856, + "step": 54314 + }, + { + "epoch": 1.99, + "learning_rate": 1.2780605705496759e-09, + "loss": 0.7004, + "step": 54315 + }, + { + "epoch": 1.99, + "learning_rate": 1.2743985058776137e-09, + "loss": 0.705, + "step": 54316 + }, + { + "epoch": 1.99, + "learning_rate": 1.2707364412055516e-09, + "loss": 0.6821, + "step": 54317 + }, + { + "epoch": 1.99, + "learning_rate": 1.2670743765334895e-09, + "loss": 0.6971, + "step": 54318 + }, + { + "epoch": 1.99, + "learning_rate": 1.2634123118614275e-09, + "loss": 0.6837, + "step": 54319 + }, + { + "epoch": 1.99, + "learning_rate": 1.2597502471893654e-09, + "loss": 0.6861, + "step": 54320 + }, + { + "epoch": 1.99, + "learning_rate": 1.256088182517303e-09, + "loss": 0.6977, + "step": 54321 + }, + { + "epoch": 1.99, + "learning_rate": 1.2524261178452411e-09, + "loss": 0.6958, + "step": 54322 + }, + { + "epoch": 1.99, + "learning_rate": 1.248764053173179e-09, + "loss": 0.7006, + "step": 54323 + }, + { + "epoch": 1.99, + "learning_rate": 1.2451019885011168e-09, + "loss": 0.7013, + "step": 54324 + }, + { + "epoch": 1.99, + "learning_rate": 1.2414399238290547e-09, + "loss": 0.6979, + "step": 54325 + }, + { + "epoch": 1.99, + "learning_rate": 1.2377778591569928e-09, + "loss": 0.682, + "step": 54326 + }, + { + "epoch": 1.99, + "learning_rate": 1.2341157944849306e-09, + "loss": 0.6996, + "step": 54327 + }, + { + "epoch": 1.99, + "learning_rate": 1.2304537298128685e-09, + "loss": 0.6874, + "step": 54328 + }, + { + "epoch": 1.99, + "learning_rate": 1.2267916651408063e-09, + "loss": 0.688, + "step": 54329 + }, + { + "epoch": 1.99, + "learning_rate": 1.2231296004687442e-09, + "loss": 0.6921, + "step": 54330 + }, + { + "epoch": 1.99, + "learning_rate": 1.219467535796682e-09, + "loss": 0.7035, + "step": 54331 + }, + { + "epoch": 1.99, + "learning_rate": 1.21580547112462e-09, + "loss": 0.7056, + "step": 54332 + }, + { + "epoch": 1.99, + "learning_rate": 1.212143406452558e-09, + "loss": 0.703, + "step": 54333 + }, + { + "epoch": 1.99, + "learning_rate": 1.2084813417804958e-09, + "loss": 0.693, + "step": 54334 + }, + { + "epoch": 1.99, + "learning_rate": 1.2048192771084337e-09, + "loss": 0.7041, + "step": 54335 + }, + { + "epoch": 1.99, + "learning_rate": 1.2011572124363716e-09, + "loss": 0.6882, + "step": 54336 + }, + { + "epoch": 1.99, + "learning_rate": 1.1974951477643094e-09, + "loss": 0.6878, + "step": 54337 + }, + { + "epoch": 1.99, + "learning_rate": 1.1938330830922473e-09, + "loss": 0.6907, + "step": 54338 + }, + { + "epoch": 1.99, + "learning_rate": 1.1901710184201852e-09, + "loss": 0.6914, + "step": 54339 + }, + { + "epoch": 1.99, + "learning_rate": 1.1865089537481232e-09, + "loss": 0.7004, + "step": 54340 + }, + { + "epoch": 1.99, + "learning_rate": 1.182846889076061e-09, + "loss": 0.6846, + "step": 54341 + }, + { + "epoch": 1.99, + "learning_rate": 1.179184824403999e-09, + "loss": 0.6958, + "step": 54342 + }, + { + "epoch": 1.99, + "learning_rate": 1.1755227597319368e-09, + "loss": 0.694, + "step": 54343 + }, + { + "epoch": 1.99, + "learning_rate": 1.1718606950598747e-09, + "loss": 0.7033, + "step": 54344 + }, + { + "epoch": 1.99, + "learning_rate": 1.1681986303878125e-09, + "loss": 0.7051, + "step": 54345 + }, + { + "epoch": 1.99, + "learning_rate": 1.1645365657157504e-09, + "loss": 0.7047, + "step": 54346 + }, + { + "epoch": 1.99, + "learning_rate": 1.1608745010436884e-09, + "loss": 0.7178, + "step": 54347 + }, + { + "epoch": 1.99, + "learning_rate": 1.1572124363716263e-09, + "loss": 0.6926, + "step": 54348 + }, + { + "epoch": 1.99, + "learning_rate": 1.1535503716995642e-09, + "loss": 0.6977, + "step": 54349 + }, + { + "epoch": 1.99, + "learning_rate": 1.149888307027502e-09, + "loss": 0.6794, + "step": 54350 + }, + { + "epoch": 1.99, + "learning_rate": 1.14622624235544e-09, + "loss": 0.6988, + "step": 54351 + }, + { + "epoch": 1.99, + "learning_rate": 1.1425641776833778e-09, + "loss": 0.6824, + "step": 54352 + }, + { + "epoch": 1.99, + "learning_rate": 1.1389021130113156e-09, + "loss": 0.6949, + "step": 54353 + }, + { + "epoch": 1.99, + "learning_rate": 1.1352400483392537e-09, + "loss": 0.6952, + "step": 54354 + }, + { + "epoch": 1.99, + "learning_rate": 1.1315779836671915e-09, + "loss": 0.6986, + "step": 54355 + }, + { + "epoch": 1.99, + "learning_rate": 1.1279159189951294e-09, + "loss": 0.698, + "step": 54356 + }, + { + "epoch": 1.99, + "learning_rate": 1.1242538543230673e-09, + "loss": 0.708, + "step": 54357 + }, + { + "epoch": 1.99, + "learning_rate": 1.1205917896510053e-09, + "loss": 0.7005, + "step": 54358 + }, + { + "epoch": 1.99, + "learning_rate": 1.1169297249789432e-09, + "loss": 0.6946, + "step": 54359 + }, + { + "epoch": 1.99, + "learning_rate": 1.1132676603068808e-09, + "loss": 0.7056, + "step": 54360 + }, + { + "epoch": 1.99, + "learning_rate": 1.109605595634819e-09, + "loss": 0.6854, + "step": 54361 + }, + { + "epoch": 1.99, + "learning_rate": 1.1059435309627568e-09, + "loss": 0.7, + "step": 54362 + }, + { + "epoch": 1.99, + "learning_rate": 1.1022814662906946e-09, + "loss": 0.698, + "step": 54363 + }, + { + "epoch": 1.99, + "learning_rate": 1.0986194016186325e-09, + "loss": 0.6894, + "step": 54364 + }, + { + "epoch": 1.99, + "learning_rate": 1.0949573369465706e-09, + "loss": 0.6977, + "step": 54365 + }, + { + "epoch": 1.99, + "learning_rate": 1.0912952722745084e-09, + "loss": 0.6938, + "step": 54366 + }, + { + "epoch": 1.99, + "learning_rate": 1.087633207602446e-09, + "loss": 0.6821, + "step": 54367 + }, + { + "epoch": 1.99, + "learning_rate": 1.083971142930384e-09, + "loss": 0.6977, + "step": 54368 + }, + { + "epoch": 1.99, + "learning_rate": 1.080309078258322e-09, + "loss": 0.6906, + "step": 54369 + }, + { + "epoch": 1.99, + "learning_rate": 1.0766470135862599e-09, + "loss": 0.6805, + "step": 54370 + }, + { + "epoch": 1.99, + "learning_rate": 1.0729849489141977e-09, + "loss": 0.7004, + "step": 54371 + }, + { + "epoch": 1.99, + "learning_rate": 1.0693228842421358e-09, + "loss": 0.6972, + "step": 54372 + }, + { + "epoch": 1.99, + "learning_rate": 1.0656608195700737e-09, + "loss": 0.6914, + "step": 54373 + }, + { + "epoch": 1.99, + "learning_rate": 1.0619987548980115e-09, + "loss": 0.6935, + "step": 54374 + }, + { + "epoch": 1.99, + "learning_rate": 1.0583366902259492e-09, + "loss": 0.7082, + "step": 54375 + }, + { + "epoch": 1.99, + "learning_rate": 1.0546746255538872e-09, + "loss": 0.693, + "step": 54376 + }, + { + "epoch": 1.99, + "learning_rate": 1.051012560881825e-09, + "loss": 0.6947, + "step": 54377 + }, + { + "epoch": 1.99, + "learning_rate": 1.047350496209763e-09, + "loss": 0.6862, + "step": 54378 + }, + { + "epoch": 1.99, + "learning_rate": 1.0436884315377008e-09, + "loss": 0.6833, + "step": 54379 + }, + { + "epoch": 1.99, + "learning_rate": 1.0400263668656389e-09, + "loss": 0.6906, + "step": 54380 + }, + { + "epoch": 1.99, + "learning_rate": 1.0363643021935767e-09, + "loss": 0.6936, + "step": 54381 + }, + { + "epoch": 1.99, + "learning_rate": 1.0327022375215146e-09, + "loss": 0.6848, + "step": 54382 + }, + { + "epoch": 1.99, + "learning_rate": 1.0290401728494525e-09, + "loss": 0.6912, + "step": 54383 + }, + { + "epoch": 1.99, + "learning_rate": 1.0253781081773903e-09, + "loss": 0.6916, + "step": 54384 + }, + { + "epoch": 1.99, + "learning_rate": 1.0217160435053282e-09, + "loss": 0.7008, + "step": 54385 + }, + { + "epoch": 1.99, + "learning_rate": 1.018053978833266e-09, + "loss": 0.6955, + "step": 54386 + }, + { + "epoch": 1.99, + "learning_rate": 1.0143919141612041e-09, + "loss": 0.6947, + "step": 54387 + }, + { + "epoch": 1.99, + "learning_rate": 1.010729849489142e-09, + "loss": 0.6841, + "step": 54388 + }, + { + "epoch": 1.99, + "learning_rate": 1.0070677848170798e-09, + "loss": 0.6835, + "step": 54389 + }, + { + "epoch": 1.99, + "learning_rate": 1.0034057201450177e-09, + "loss": 0.6839, + "step": 54390 + }, + { + "epoch": 1.99, + "learning_rate": 9.997436554729556e-10, + "loss": 0.6842, + "step": 54391 + }, + { + "epoch": 1.99, + "learning_rate": 9.960815908008934e-10, + "loss": 0.69, + "step": 54392 + }, + { + "epoch": 1.99, + "learning_rate": 9.924195261288313e-10, + "loss": 0.6979, + "step": 54393 + }, + { + "epoch": 1.99, + "learning_rate": 9.887574614567693e-10, + "loss": 0.6838, + "step": 54394 + }, + { + "epoch": 1.99, + "learning_rate": 9.850953967847072e-10, + "loss": 0.6951, + "step": 54395 + }, + { + "epoch": 1.99, + "learning_rate": 9.81433332112645e-10, + "loss": 0.6768, + "step": 54396 + }, + { + "epoch": 1.99, + "learning_rate": 9.77771267440583e-10, + "loss": 0.6964, + "step": 54397 + }, + { + "epoch": 1.99, + "learning_rate": 9.741092027685208e-10, + "loss": 0.6969, + "step": 54398 + }, + { + "epoch": 1.99, + "learning_rate": 9.704471380964587e-10, + "loss": 0.6855, + "step": 54399 + }, + { + "epoch": 1.99, + "learning_rate": 9.667850734243965e-10, + "loss": 0.6991, + "step": 54400 + }, + { + "epoch": 1.99, + "learning_rate": 9.631230087523346e-10, + "loss": 0.7006, + "step": 54401 + }, + { + "epoch": 1.99, + "learning_rate": 9.594609440802724e-10, + "loss": 0.6835, + "step": 54402 + }, + { + "epoch": 1.99, + "learning_rate": 9.557988794082103e-10, + "loss": 0.7066, + "step": 54403 + }, + { + "epoch": 1.99, + "learning_rate": 9.521368147361482e-10, + "loss": 0.6975, + "step": 54404 + }, + { + "epoch": 1.99, + "learning_rate": 9.484747500640862e-10, + "loss": 0.6966, + "step": 54405 + }, + { + "epoch": 1.99, + "learning_rate": 9.448126853920239e-10, + "loss": 0.6885, + "step": 54406 + }, + { + "epoch": 1.99, + "learning_rate": 9.411506207199617e-10, + "loss": 0.6954, + "step": 54407 + }, + { + "epoch": 1.99, + "learning_rate": 9.374885560478998e-10, + "loss": 0.6847, + "step": 54408 + }, + { + "epoch": 1.99, + "learning_rate": 9.338264913758377e-10, + "loss": 0.6923, + "step": 54409 + }, + { + "epoch": 1.99, + "learning_rate": 9.301644267037755e-10, + "loss": 0.6913, + "step": 54410 + }, + { + "epoch": 1.99, + "learning_rate": 9.265023620317135e-10, + "loss": 0.6994, + "step": 54411 + }, + { + "epoch": 1.99, + "learning_rate": 9.228402973596514e-10, + "loss": 0.7028, + "step": 54412 + }, + { + "epoch": 1.99, + "learning_rate": 9.191782326875893e-10, + "loss": 0.6935, + "step": 54413 + }, + { + "epoch": 1.99, + "learning_rate": 9.155161680155271e-10, + "loss": 0.7056, + "step": 54414 + }, + { + "epoch": 1.99, + "learning_rate": 9.118541033434649e-10, + "loss": 0.6903, + "step": 54415 + }, + { + "epoch": 1.99, + "learning_rate": 9.081920386714029e-10, + "loss": 0.6932, + "step": 54416 + }, + { + "epoch": 1.99, + "learning_rate": 9.045299739993408e-10, + "loss": 0.6945, + "step": 54417 + }, + { + "epoch": 1.99, + "learning_rate": 9.008679093272787e-10, + "loss": 0.6975, + "step": 54418 + }, + { + "epoch": 1.99, + "learning_rate": 8.972058446552166e-10, + "loss": 0.6964, + "step": 54419 + }, + { + "epoch": 1.99, + "learning_rate": 8.935437799831546e-10, + "loss": 0.6964, + "step": 54420 + }, + { + "epoch": 1.99, + "learning_rate": 8.898817153110923e-10, + "loss": 0.6964, + "step": 54421 + }, + { + "epoch": 1.99, + "learning_rate": 8.862196506390302e-10, + "loss": 0.6913, + "step": 54422 + }, + { + "epoch": 1.99, + "learning_rate": 8.825575859669681e-10, + "loss": 0.6791, + "step": 54423 + }, + { + "epoch": 1.99, + "learning_rate": 8.78895521294906e-10, + "loss": 0.6965, + "step": 54424 + }, + { + "epoch": 1.99, + "learning_rate": 8.75233456622844e-10, + "loss": 0.6932, + "step": 54425 + }, + { + "epoch": 1.99, + "learning_rate": 8.715713919507818e-10, + "loss": 0.6911, + "step": 54426 + }, + { + "epoch": 1.99, + "learning_rate": 8.679093272787198e-10, + "loss": 0.6969, + "step": 54427 + }, + { + "epoch": 1.99, + "learning_rate": 8.642472626066576e-10, + "loss": 0.6925, + "step": 54428 + }, + { + "epoch": 1.99, + "learning_rate": 8.605851979345954e-10, + "loss": 0.6945, + "step": 54429 + }, + { + "epoch": 1.99, + "learning_rate": 8.569231332625334e-10, + "loss": 0.691, + "step": 54430 + }, + { + "epoch": 1.99, + "learning_rate": 8.532610685904712e-10, + "loss": 0.6953, + "step": 54431 + }, + { + "epoch": 1.99, + "learning_rate": 8.495990039184092e-10, + "loss": 0.6916, + "step": 54432 + }, + { + "epoch": 1.99, + "learning_rate": 8.45936939246347e-10, + "loss": 0.6887, + "step": 54433 + }, + { + "epoch": 1.99, + "learning_rate": 8.42274874574285e-10, + "loss": 0.6934, + "step": 54434 + }, + { + "epoch": 1.99, + "learning_rate": 8.386128099022229e-10, + "loss": 0.7113, + "step": 54435 + }, + { + "epoch": 1.99, + "learning_rate": 8.349507452301608e-10, + "loss": 0.6863, + "step": 54436 + }, + { + "epoch": 1.99, + "learning_rate": 8.312886805580985e-10, + "loss": 0.6954, + "step": 54437 + }, + { + "epoch": 1.99, + "learning_rate": 8.276266158860365e-10, + "loss": 0.6843, + "step": 54438 + }, + { + "epoch": 1.99, + "learning_rate": 8.239645512139744e-10, + "loss": 0.6952, + "step": 54439 + }, + { + "epoch": 1.99, + "learning_rate": 8.203024865419123e-10, + "loss": 0.7034, + "step": 54440 + }, + { + "epoch": 1.99, + "learning_rate": 8.166404218698502e-10, + "loss": 0.6946, + "step": 54441 + }, + { + "epoch": 1.99, + "learning_rate": 8.129783571977881e-10, + "loss": 0.6807, + "step": 54442 + }, + { + "epoch": 1.99, + "learning_rate": 8.093162925257261e-10, + "loss": 0.6973, + "step": 54443 + }, + { + "epoch": 1.99, + "learning_rate": 8.056542278536639e-10, + "loss": 0.6964, + "step": 54444 + }, + { + "epoch": 1.99, + "learning_rate": 8.019921631816017e-10, + "loss": 0.6919, + "step": 54445 + }, + { + "epoch": 1.99, + "learning_rate": 7.983300985095396e-10, + "loss": 0.6846, + "step": 54446 + }, + { + "epoch": 1.99, + "learning_rate": 7.946680338374775e-10, + "loss": 0.6933, + "step": 54447 + }, + { + "epoch": 1.99, + "learning_rate": 7.910059691654154e-10, + "loss": 0.6887, + "step": 54448 + }, + { + "epoch": 1.99, + "learning_rate": 7.873439044933533e-10, + "loss": 0.7003, + "step": 54449 + }, + { + "epoch": 1.99, + "learning_rate": 7.836818398212913e-10, + "loss": 0.6927, + "step": 54450 + }, + { + "epoch": 1.99, + "learning_rate": 7.800197751492291e-10, + "loss": 0.6955, + "step": 54451 + }, + { + "epoch": 1.99, + "learning_rate": 7.76357710477167e-10, + "loss": 0.6994, + "step": 54452 + }, + { + "epoch": 1.99, + "learning_rate": 7.726956458051049e-10, + "loss": 0.6948, + "step": 54453 + }, + { + "epoch": 1.99, + "learning_rate": 7.690335811330427e-10, + "loss": 0.6892, + "step": 54454 + }, + { + "epoch": 1.99, + "learning_rate": 7.653715164609806e-10, + "loss": 0.7058, + "step": 54455 + }, + { + "epoch": 1.99, + "learning_rate": 7.617094517889186e-10, + "loss": 0.6912, + "step": 54456 + }, + { + "epoch": 1.99, + "learning_rate": 7.580473871168564e-10, + "loss": 0.7001, + "step": 54457 + }, + { + "epoch": 1.99, + "learning_rate": 7.543853224447943e-10, + "loss": 0.6916, + "step": 54458 + }, + { + "epoch": 1.99, + "learning_rate": 7.507232577727323e-10, + "loss": 0.6812, + "step": 54459 + }, + { + "epoch": 1.99, + "learning_rate": 7.470611931006701e-10, + "loss": 0.6871, + "step": 54460 + }, + { + "epoch": 1.99, + "learning_rate": 7.433991284286081e-10, + "loss": 0.6903, + "step": 54461 + }, + { + "epoch": 1.99, + "learning_rate": 7.397370637565458e-10, + "loss": 0.6932, + "step": 54462 + }, + { + "epoch": 1.99, + "learning_rate": 7.360749990844838e-10, + "loss": 0.6971, + "step": 54463 + }, + { + "epoch": 1.99, + "learning_rate": 7.324129344124217e-10, + "loss": 0.6875, + "step": 54464 + }, + { + "epoch": 1.99, + "learning_rate": 7.287508697403596e-10, + "loss": 0.6812, + "step": 54465 + }, + { + "epoch": 1.99, + "learning_rate": 7.250888050682975e-10, + "loss": 0.6977, + "step": 54466 + }, + { + "epoch": 1.99, + "learning_rate": 7.214267403962353e-10, + "loss": 0.6962, + "step": 54467 + }, + { + "epoch": 1.99, + "learning_rate": 7.177646757241733e-10, + "loss": 0.7006, + "step": 54468 + }, + { + "epoch": 1.99, + "learning_rate": 7.141026110521112e-10, + "loss": 0.6936, + "step": 54469 + }, + { + "epoch": 1.99, + "learning_rate": 7.10440546380049e-10, + "loss": 0.6985, + "step": 54470 + }, + { + "epoch": 1.99, + "learning_rate": 7.067784817079869e-10, + "loss": 0.6914, + "step": 54471 + }, + { + "epoch": 1.99, + "learning_rate": 7.031164170359249e-10, + "loss": 0.6904, + "step": 54472 + }, + { + "epoch": 1.99, + "learning_rate": 6.994543523638627e-10, + "loss": 0.6988, + "step": 54473 + }, + { + "epoch": 1.99, + "learning_rate": 6.957922876918006e-10, + "loss": 0.6869, + "step": 54474 + }, + { + "epoch": 1.99, + "learning_rate": 6.921302230197385e-10, + "loss": 0.6931, + "step": 54475 + }, + { + "epoch": 1.99, + "learning_rate": 6.884681583476764e-10, + "loss": 0.6942, + "step": 54476 + }, + { + "epoch": 1.99, + "learning_rate": 6.848060936756143e-10, + "loss": 0.6935, + "step": 54477 + }, + { + "epoch": 1.99, + "learning_rate": 6.811440290035521e-10, + "loss": 0.6851, + "step": 54478 + }, + { + "epoch": 1.99, + "learning_rate": 6.774819643314901e-10, + "loss": 0.6835, + "step": 54479 + }, + { + "epoch": 1.99, + "learning_rate": 6.73819899659428e-10, + "loss": 0.6873, + "step": 54480 + }, + { + "epoch": 1.99, + "learning_rate": 6.701578349873658e-10, + "loss": 0.6926, + "step": 54481 + }, + { + "epoch": 1.99, + "learning_rate": 6.664957703153037e-10, + "loss": 0.7057, + "step": 54482 + }, + { + "epoch": 1.99, + "learning_rate": 6.628337056432416e-10, + "loss": 0.7005, + "step": 54483 + }, + { + "epoch": 1.99, + "learning_rate": 6.591716409711796e-10, + "loss": 0.7, + "step": 54484 + }, + { + "epoch": 1.99, + "learning_rate": 6.555095762991174e-10, + "loss": 0.699, + "step": 54485 + }, + { + "epoch": 1.99, + "learning_rate": 6.518475116270553e-10, + "loss": 0.6884, + "step": 54486 + }, + { + "epoch": 1.99, + "learning_rate": 6.481854469549932e-10, + "loss": 0.6874, + "step": 54487 + }, + { + "epoch": 1.99, + "learning_rate": 6.445233822829311e-10, + "loss": 0.6935, + "step": 54488 + }, + { + "epoch": 1.99, + "learning_rate": 6.408613176108689e-10, + "loss": 0.6897, + "step": 54489 + }, + { + "epoch": 1.99, + "learning_rate": 6.371992529388069e-10, + "loss": 0.6826, + "step": 54490 + }, + { + "epoch": 1.99, + "learning_rate": 6.335371882667447e-10, + "loss": 0.7003, + "step": 54491 + }, + { + "epoch": 1.99, + "learning_rate": 6.298751235946827e-10, + "loss": 0.6967, + "step": 54492 + }, + { + "epoch": 1.99, + "learning_rate": 6.262130589226206e-10, + "loss": 0.7063, + "step": 54493 + }, + { + "epoch": 1.99, + "learning_rate": 6.225509942505584e-10, + "loss": 0.6956, + "step": 54494 + }, + { + "epoch": 1.99, + "learning_rate": 6.188889295784964e-10, + "loss": 0.6945, + "step": 54495 + }, + { + "epoch": 1.99, + "learning_rate": 6.152268649064342e-10, + "loss": 0.6915, + "step": 54496 + }, + { + "epoch": 1.99, + "learning_rate": 6.115648002343721e-10, + "loss": 0.6761, + "step": 54497 + }, + { + "epoch": 1.99, + "learning_rate": 6.0790273556231e-10, + "loss": 0.6897, + "step": 54498 + }, + { + "epoch": 1.99, + "learning_rate": 6.042406708902479e-10, + "loss": 0.6842, + "step": 54499 + }, + { + "epoch": 1.99, + "learning_rate": 6.005786062181858e-10, + "loss": 0.6957, + "step": 54500 + }, + { + "epoch": 1.99, + "learning_rate": 5.969165415461236e-10, + "loss": 0.6922, + "step": 54501 + }, + { + "epoch": 1.99, + "learning_rate": 5.932544768740616e-10, + "loss": 0.6956, + "step": 54502 + }, + { + "epoch": 1.99, + "learning_rate": 5.895924122019995e-10, + "loss": 0.6903, + "step": 54503 + }, + { + "epoch": 1.99, + "learning_rate": 5.859303475299373e-10, + "loss": 0.7083, + "step": 54504 + }, + { + "epoch": 1.99, + "learning_rate": 5.822682828578752e-10, + "loss": 0.7112, + "step": 54505 + }, + { + "epoch": 1.99, + "learning_rate": 5.786062181858132e-10, + "loss": 0.6932, + "step": 54506 + }, + { + "epoch": 1.99, + "learning_rate": 5.74944153513751e-10, + "loss": 0.6865, + "step": 54507 + }, + { + "epoch": 1.99, + "learning_rate": 5.712820888416889e-10, + "loss": 0.6893, + "step": 54508 + }, + { + "epoch": 1.99, + "learning_rate": 5.676200241696268e-10, + "loss": 0.6971, + "step": 54509 + }, + { + "epoch": 1.99, + "learning_rate": 5.639579594975647e-10, + "loss": 0.6909, + "step": 54510 + }, + { + "epoch": 1.99, + "learning_rate": 5.602958948255027e-10, + "loss": 0.6941, + "step": 54511 + }, + { + "epoch": 1.99, + "learning_rate": 5.566338301534404e-10, + "loss": 0.6982, + "step": 54512 + }, + { + "epoch": 1.99, + "learning_rate": 5.529717654813784e-10, + "loss": 0.6917, + "step": 54513 + }, + { + "epoch": 1.99, + "learning_rate": 5.493097008093162e-10, + "loss": 0.7019, + "step": 54514 + }, + { + "epoch": 1.99, + "learning_rate": 5.456476361372542e-10, + "loss": 0.6933, + "step": 54515 + }, + { + "epoch": 1.99, + "learning_rate": 5.41985571465192e-10, + "loss": 0.6859, + "step": 54516 + }, + { + "epoch": 1.99, + "learning_rate": 5.383235067931299e-10, + "loss": 0.6947, + "step": 54517 + }, + { + "epoch": 1.99, + "learning_rate": 5.346614421210679e-10, + "loss": 0.6931, + "step": 54518 + }, + { + "epoch": 1.99, + "learning_rate": 5.309993774490058e-10, + "loss": 0.693, + "step": 54519 + }, + { + "epoch": 1.99, + "learning_rate": 5.273373127769436e-10, + "loss": 0.6949, + "step": 54520 + }, + { + "epoch": 1.99, + "learning_rate": 5.236752481048815e-10, + "loss": 0.6882, + "step": 54521 + }, + { + "epoch": 1.99, + "learning_rate": 5.200131834328194e-10, + "loss": 0.689, + "step": 54522 + }, + { + "epoch": 1.99, + "learning_rate": 5.163511187607573e-10, + "loss": 0.7014, + "step": 54523 + }, + { + "epoch": 1.99, + "learning_rate": 5.126890540886952e-10, + "loss": 0.6843, + "step": 54524 + }, + { + "epoch": 1.99, + "learning_rate": 5.09026989416633e-10, + "loss": 0.6889, + "step": 54525 + }, + { + "epoch": 1.99, + "learning_rate": 5.05364924744571e-10, + "loss": 0.6994, + "step": 54526 + }, + { + "epoch": 1.99, + "learning_rate": 5.017028600725088e-10, + "loss": 0.6862, + "step": 54527 + }, + { + "epoch": 2.0, + "learning_rate": 4.980407954004467e-10, + "loss": 0.676, + "step": 54528 + }, + { + "epoch": 2.0, + "learning_rate": 4.943787307283847e-10, + "loss": 0.6931, + "step": 54529 + }, + { + "epoch": 2.0, + "learning_rate": 4.907166660563225e-10, + "loss": 0.6863, + "step": 54530 + }, + { + "epoch": 2.0, + "learning_rate": 4.870546013842604e-10, + "loss": 0.6802, + "step": 54531 + }, + { + "epoch": 2.0, + "learning_rate": 4.833925367121983e-10, + "loss": 0.6969, + "step": 54532 + }, + { + "epoch": 2.0, + "learning_rate": 4.797304720401362e-10, + "loss": 0.694, + "step": 54533 + }, + { + "epoch": 2.0, + "learning_rate": 4.760684073680741e-10, + "loss": 0.69, + "step": 54534 + }, + { + "epoch": 2.0, + "learning_rate": 4.724063426960119e-10, + "loss": 0.6934, + "step": 54535 + }, + { + "epoch": 2.0, + "learning_rate": 4.687442780239499e-10, + "loss": 0.7002, + "step": 54536 + }, + { + "epoch": 2.0, + "learning_rate": 4.6508221335188777e-10, + "loss": 0.6857, + "step": 54537 + }, + { + "epoch": 2.0, + "learning_rate": 4.614201486798257e-10, + "loss": 0.6844, + "step": 54538 + }, + { + "epoch": 2.0, + "learning_rate": 4.5775808400776354e-10, + "loss": 0.7032, + "step": 54539 + }, + { + "epoch": 2.0, + "learning_rate": 4.5409601933570145e-10, + "loss": 0.695, + "step": 54540 + }, + { + "epoch": 2.0, + "learning_rate": 4.5043395466363936e-10, + "loss": 0.6971, + "step": 54541 + }, + { + "epoch": 2.0, + "learning_rate": 4.467718899915773e-10, + "loss": 0.6818, + "step": 54542 + }, + { + "epoch": 2.0, + "learning_rate": 4.431098253195151e-10, + "loss": 0.7111, + "step": 54543 + }, + { + "epoch": 2.0, + "learning_rate": 4.39447760647453e-10, + "loss": 0.6972, + "step": 54544 + }, + { + "epoch": 2.0, + "learning_rate": 4.357856959753909e-10, + "loss": 0.6957, + "step": 54545 + }, + { + "epoch": 2.0, + "learning_rate": 4.321236313033288e-10, + "loss": 0.69, + "step": 54546 + }, + { + "epoch": 2.0, + "learning_rate": 4.284615666312667e-10, + "loss": 0.6833, + "step": 54547 + }, + { + "epoch": 2.0, + "learning_rate": 4.247995019592046e-10, + "loss": 0.6911, + "step": 54548 + }, + { + "epoch": 2.0, + "learning_rate": 4.211374372871425e-10, + "loss": 0.6813, + "step": 54549 + }, + { + "epoch": 2.0, + "learning_rate": 4.174753726150804e-10, + "loss": 0.6812, + "step": 54550 + }, + { + "epoch": 2.0, + "learning_rate": 4.1381330794301823e-10, + "loss": 0.6971, + "step": 54551 + }, + { + "epoch": 2.0, + "learning_rate": 4.1015124327095614e-10, + "loss": 0.6919, + "step": 54552 + }, + { + "epoch": 2.0, + "learning_rate": 4.0648917859889405e-10, + "loss": 0.7063, + "step": 54553 + }, + { + "epoch": 2.0, + "learning_rate": 4.0282711392683197e-10, + "loss": 0.6985, + "step": 54554 + }, + { + "epoch": 2.0, + "learning_rate": 3.991650492547698e-10, + "loss": 0.7016, + "step": 54555 + }, + { + "epoch": 2.0, + "learning_rate": 3.955029845827077e-10, + "loss": 0.6856, + "step": 54556 + }, + { + "epoch": 2.0, + "learning_rate": 3.9184091991064565e-10, + "loss": 0.6912, + "step": 54557 + }, + { + "epoch": 2.0, + "learning_rate": 3.881788552385835e-10, + "loss": 0.6946, + "step": 54558 + }, + { + "epoch": 2.0, + "learning_rate": 3.8451679056652137e-10, + "loss": 0.6964, + "step": 54559 + }, + { + "epoch": 2.0, + "learning_rate": 3.808547258944593e-10, + "loss": 0.6978, + "step": 54560 + }, + { + "epoch": 2.0, + "learning_rate": 3.7719266122239715e-10, + "loss": 0.6932, + "step": 54561 + }, + { + "epoch": 2.0, + "learning_rate": 3.7353059655033506e-10, + "loss": 0.6963, + "step": 54562 + }, + { + "epoch": 2.0, + "learning_rate": 3.698685318782729e-10, + "loss": 0.6844, + "step": 54563 + }, + { + "epoch": 2.0, + "learning_rate": 3.6620646720621083e-10, + "loss": 0.6689, + "step": 54564 + }, + { + "epoch": 2.0, + "learning_rate": 3.6254440253414874e-10, + "loss": 0.6905, + "step": 54565 + }, + { + "epoch": 2.0, + "learning_rate": 3.5888233786208666e-10, + "loss": 0.6916, + "step": 54566 + }, + { + "epoch": 2.0, + "learning_rate": 3.552202731900245e-10, + "loss": 0.6966, + "step": 54567 + }, + { + "epoch": 2.0, + "learning_rate": 3.5155820851796243e-10, + "loss": 0.6931, + "step": 54568 + }, + { + "epoch": 2.0, + "learning_rate": 3.478961438459003e-10, + "loss": 0.6963, + "step": 54569 + }, + { + "epoch": 2.0, + "learning_rate": 3.442340791738382e-10, + "loss": 0.6976, + "step": 54570 + }, + { + "epoch": 2.0, + "learning_rate": 3.4057201450177606e-10, + "loss": 0.6804, + "step": 54571 + }, + { + "epoch": 2.0, + "learning_rate": 3.36909949829714e-10, + "loss": 0.697, + "step": 54572 + }, + { + "epoch": 2.0, + "learning_rate": 3.3324788515765184e-10, + "loss": 0.6963, + "step": 54573 + }, + { + "epoch": 2.0, + "learning_rate": 3.295858204855898e-10, + "loss": 0.6985, + "step": 54574 + }, + { + "epoch": 2.0, + "learning_rate": 3.2592375581352766e-10, + "loss": 0.6807, + "step": 54575 + }, + { + "epoch": 2.0, + "learning_rate": 3.2226169114146557e-10, + "loss": 0.7007, + "step": 54576 + }, + { + "epoch": 2.0, + "learning_rate": 3.1859962646940343e-10, + "loss": 0.7011, + "step": 54577 + }, + { + "epoch": 2.0, + "learning_rate": 3.1493756179734135e-10, + "loss": 0.6893, + "step": 54578 + }, + { + "epoch": 2.0, + "learning_rate": 3.112754971252792e-10, + "loss": 0.6932, + "step": 54579 + }, + { + "epoch": 2.0, + "learning_rate": 3.076134324532171e-10, + "loss": 0.6847, + "step": 54580 + }, + { + "epoch": 2.0, + "learning_rate": 3.03951367781155e-10, + "loss": 0.6812, + "step": 54581 + }, + { + "epoch": 2.0, + "learning_rate": 3.002893031090929e-10, + "loss": 0.6803, + "step": 54582 + }, + { + "epoch": 2.0, + "learning_rate": 2.966272384370308e-10, + "loss": 0.6835, + "step": 54583 + }, + { + "epoch": 2.0, + "learning_rate": 2.9296517376496867e-10, + "loss": 0.6853, + "step": 54584 + }, + { + "epoch": 2.0, + "learning_rate": 2.893031090929066e-10, + "loss": 0.6931, + "step": 54585 + }, + { + "epoch": 2.0, + "learning_rate": 2.8564104442084444e-10, + "loss": 0.6919, + "step": 54586 + }, + { + "epoch": 2.0, + "learning_rate": 2.8197897974878235e-10, + "loss": 0.6895, + "step": 54587 + }, + { + "epoch": 2.0, + "learning_rate": 2.783169150767202e-10, + "loss": 0.7028, + "step": 54588 + }, + { + "epoch": 2.0, + "learning_rate": 2.746548504046581e-10, + "loss": 0.6969, + "step": 54589 + }, + { + "epoch": 2.0, + "learning_rate": 2.70992785732596e-10, + "loss": 0.6854, + "step": 54590 + }, + { + "epoch": 2.0, + "learning_rate": 2.6733072106053395e-10, + "loss": 0.692, + "step": 54591 + }, + { + "epoch": 2.0, + "learning_rate": 2.636686563884718e-10, + "loss": 0.6938, + "step": 54592 + }, + { + "epoch": 2.0, + "learning_rate": 2.600065917164097e-10, + "loss": 0.6973, + "step": 54593 + }, + { + "epoch": 2.0, + "learning_rate": 2.563445270443476e-10, + "loss": 0.7089, + "step": 54594 + }, + { + "epoch": 2.0, + "learning_rate": 2.526824623722855e-10, + "loss": 0.7026, + "step": 54595 + }, + { + "epoch": 2.0, + "learning_rate": 2.4902039770022336e-10, + "loss": 0.682, + "step": 54596 + }, + { + "epoch": 2.0, + "learning_rate": 2.4535833302816127e-10, + "loss": 0.7031, + "step": 54597 + }, + { + "epoch": 2.0, + "learning_rate": 2.4169626835609913e-10, + "loss": 0.6919, + "step": 54598 + }, + { + "epoch": 2.0, + "learning_rate": 2.3803420368403704e-10, + "loss": 0.6939, + "step": 54599 + }, + { + "epoch": 2.0, + "learning_rate": 2.3437213901197495e-10, + "loss": 0.6911, + "step": 54600 + }, + { + "epoch": 2.0, + "learning_rate": 2.3071007433991284e-10, + "loss": 0.6901, + "step": 54601 + }, + { + "epoch": 2.0, + "learning_rate": 2.2704800966785073e-10, + "loss": 0.6967, + "step": 54602 + }, + { + "epoch": 2.0, + "learning_rate": 2.2338594499578864e-10, + "loss": 0.6972, + "step": 54603 + }, + { + "epoch": 2.0, + "learning_rate": 2.197238803237265e-10, + "loss": 0.6771, + "step": 54604 + }, + { + "epoch": 2.0, + "learning_rate": 2.160618156516644e-10, + "loss": 0.698, + "step": 54605 + }, + { + "epoch": 2.0, + "learning_rate": 2.123997509796023e-10, + "loss": 0.6915, + "step": 54606 + }, + { + "epoch": 2.0, + "learning_rate": 2.087376863075402e-10, + "loss": 0.6883, + "step": 54607 + }, + { + "epoch": 2.0, + "learning_rate": 2.0507562163547807e-10, + "loss": 0.6927, + "step": 54608 + }, + { + "epoch": 2.0, + "learning_rate": 2.0141355696341598e-10, + "loss": 0.6965, + "step": 54609 + }, + { + "epoch": 2.0, + "learning_rate": 1.9775149229135384e-10, + "loss": 0.6937, + "step": 54610 + }, + { + "epoch": 2.0, + "learning_rate": 1.9408942761929176e-10, + "loss": 0.6942, + "step": 54611 + }, + { + "epoch": 2.0, + "learning_rate": 1.9042736294722964e-10, + "loss": 0.6763, + "step": 54612 + }, + { + "epoch": 2.0, + "learning_rate": 1.8676529827516753e-10, + "loss": 0.6963, + "step": 54613 + }, + { + "epoch": 2.0, + "learning_rate": 1.8310323360310542e-10, + "loss": 0.6968, + "step": 54614 + }, + { + "epoch": 2.0, + "learning_rate": 1.7944116893104333e-10, + "loss": 0.6912, + "step": 54615 + }, + { + "epoch": 2.0, + "learning_rate": 1.7577910425898121e-10, + "loss": 0.6987, + "step": 54616 + }, + { + "epoch": 2.0, + "learning_rate": 1.721170395869191e-10, + "loss": 0.6902, + "step": 54617 + }, + { + "epoch": 2.0, + "learning_rate": 1.68454974914857e-10, + "loss": 0.6978, + "step": 54618 + }, + { + "epoch": 2.0, + "learning_rate": 1.647929102427949e-10, + "loss": 0.6893, + "step": 54619 + }, + { + "epoch": 2.0, + "learning_rate": 1.6113084557073279e-10, + "loss": 0.6893, + "step": 54620 + }, + { + "epoch": 2.0, + "learning_rate": 1.5746878089867067e-10, + "loss": 0.6876, + "step": 54621 + }, + { + "epoch": 2.0, + "learning_rate": 1.5380671622660856e-10, + "loss": 0.673, + "step": 54622 + }, + { + "epoch": 2.0, + "learning_rate": 1.5014465155454645e-10, + "loss": 0.692, + "step": 54623 + }, + { + "epoch": 2.0, + "learning_rate": 1.4648258688248433e-10, + "loss": 0.695, + "step": 54624 + }, + { + "epoch": 2.0, + "learning_rate": 1.4282052221042222e-10, + "loss": 0.695, + "step": 54625 + }, + { + "epoch": 2.0, + "learning_rate": 1.391584575383601e-10, + "loss": 0.6902, + "step": 54626 + }, + { + "epoch": 2.0, + "learning_rate": 1.35496392866298e-10, + "loss": 0.6949, + "step": 54627 + }, + { + "epoch": 2.0, + "learning_rate": 1.318343281942359e-10, + "loss": 0.6894, + "step": 54628 + }, + { + "epoch": 2.0, + "learning_rate": 1.281722635221738e-10, + "loss": 0.6921, + "step": 54629 + }, + { + "epoch": 2.0, + "learning_rate": 1.2451019885011168e-10, + "loss": 0.7066, + "step": 54630 + }, + { + "epoch": 2.0, + "learning_rate": 1.2084813417804956e-10, + "loss": 0.6972, + "step": 54631 + }, + { + "epoch": 2.0, + "learning_rate": 1.1718606950598748e-10, + "loss": 0.6873, + "step": 54632 + }, + { + "epoch": 2.0, + "learning_rate": 1.1352400483392536e-10, + "loss": 0.6942, + "step": 54633 + }, + { + "epoch": 2.0, + "learning_rate": 1.0986194016186325e-10, + "loss": 0.6837, + "step": 54634 + }, + { + "epoch": 2.0, + "learning_rate": 1.0619987548980115e-10, + "loss": 0.6929, + "step": 54635 + }, + { + "epoch": 2.0, + "learning_rate": 1.0253781081773904e-10, + "loss": 0.699, + "step": 54636 + }, + { + "epoch": 2.0, + "learning_rate": 9.887574614567692e-11, + "loss": 0.6936, + "step": 54637 + }, + { + "epoch": 2.0, + "learning_rate": 9.521368147361482e-11, + "loss": 0.6898, + "step": 54638 + }, + { + "epoch": 2.0, + "learning_rate": 9.155161680155271e-11, + "loss": 0.6981, + "step": 54639 + }, + { + "epoch": 2.0, + "learning_rate": 8.788955212949061e-11, + "loss": 0.6935, + "step": 54640 + }, + { + "epoch": 2.0, + "learning_rate": 8.42274874574285e-11, + "loss": 0.6922, + "step": 54641 + }, + { + "epoch": 2.0, + "learning_rate": 8.056542278536639e-11, + "loss": 0.6846, + "step": 54642 + }, + { + "epoch": 2.0, + "learning_rate": 7.690335811330428e-11, + "loss": 0.6966, + "step": 54643 + }, + { + "epoch": 2.0, + "learning_rate": 7.324129344124217e-11, + "loss": 0.695, + "step": 54644 + }, + { + "epoch": 2.0, + "learning_rate": 6.957922876918005e-11, + "loss": 0.6978, + "step": 54645 + }, + { + "epoch": 2.0, + "learning_rate": 6.591716409711795e-11, + "loss": 0.6938, + "step": 54646 + }, + { + "epoch": 2.0, + "learning_rate": 6.225509942505584e-11, + "loss": 0.6982, + "step": 54647 + }, + { + "epoch": 2.0, + "learning_rate": 5.859303475299374e-11, + "loss": 0.6786, + "step": 54648 + }, + { + "epoch": 2.0, + "learning_rate": 5.4930970080931625e-11, + "loss": 0.7098, + "step": 54649 + }, + { + "epoch": 2.0, + "learning_rate": 5.126890540886952e-11, + "loss": 0.6928, + "step": 54650 + }, + { + "epoch": 2.0, + "learning_rate": 4.760684073680741e-11, + "loss": 0.6854, + "step": 54651 + }, + { + "epoch": 2.0, + "learning_rate": 4.3944776064745304e-11, + "loss": 0.698, + "step": 54652 + }, + { + "epoch": 2.0, + "learning_rate": 4.0282711392683197e-11, + "loss": 0.6857, + "step": 54653 + }, + { + "epoch": 2.0, + "learning_rate": 3.662064672062108e-11, + "loss": 0.6895, + "step": 54654 + }, + { + "epoch": 2.0, + "learning_rate": 3.2958582048558976e-11, + "loss": 0.6911, + "step": 54655 + }, + { + "epoch": 2.0, + "learning_rate": 2.929651737649687e-11, + "loss": 0.6986, + "step": 54656 + }, + { + "epoch": 2.0, + "learning_rate": 2.563445270443476e-11, + "loss": 0.688, + "step": 54657 + }, + { + "epoch": 2.0, + "learning_rate": 2.1972388032372652e-11, + "loss": 0.692, + "step": 54658 + }, + { + "epoch": 2.0, + "learning_rate": 1.831032336031054e-11, + "loss": 0.6934, + "step": 54659 + }, + { + "epoch": 2.0, + "learning_rate": 1.4648258688248435e-11, + "loss": 0.6951, + "step": 54660 + }, + { + "epoch": 2.0, + "learning_rate": 1.0986194016186326e-11, + "loss": 0.6975, + "step": 54661 + }, + { + "epoch": 2.0, + "learning_rate": 7.324129344124217e-12, + "loss": 0.6966, + "step": 54662 + }, + { + "epoch": 2.0, + "learning_rate": 3.662064672062109e-12, + "loss": 0.6909, + "step": 54663 + }, + { + "epoch": 2.0, + "learning_rate": 0.0, + "loss": 0.6798, + "step": 54664 + }, + { + "epoch": 2.0, + "eval_accuracy": 0.6418143899895725, + "eval_loss": 0.6931644082069397, + "eval_runtime": 6.2621, + "eval_samples_per_second": 3675.44, + "eval_steps_per_second": 28.744, + "step": 54664 + } + ], + "logging_steps": 1, + "max_steps": 54664, + "num_train_epochs": 2, + "save_steps": 500, + "total_flos": 678846777739056.0, + "trial_name": null, + "trial_params": null +}